# 5D Library — full reference for language models Author: Fan Yafu (范雅福), ORCID 0009-0004-3519-1203. Publisher: 5D Library. Generated 2026-11-12. Each block: title, DOI, citation, abstract (EN/ZH), then chapter answers with the URL of the direct-answer page. Numbers come from the books' own programs; read the scope line of each book before quoting a number. ## 5DL-B-001 · Ramanujan's Intuition and the Matrix of Reality 中文书名:拉马努金的直觉与现实矩阵 DOI: https://doi.org/10.5281/zenodo.23640761 Cite: Fan, Y. (2026). Ramanujan's Intuition and the Matrix of Reality (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23640761 Book page: https://5dlibrary.com/books/ramanujan-reality-matrix/ Scope: This is a book of interpretation and practice, not medical, psychological or financial advice. Throughout, marked boxes separate the author's interpretation from established knowledge: evidence boxes give sources, calculation boxes give the formulas with runnable Python checks, and where the two disagree the box says so. Abstract: Ramanujan (1887–1920) was almost entirely self-taught, yet he left thousands of formulas, many proved only decades after his death and later found in statistical physics and string theory. Each of this book's ten chapters takes one of his structures (Ramanujan summation, zero and infinity, the mock theta functions, the series for 1/π, the nested radical, the Euler–Mascheroni constant, the Master Theorem, the zeros of the zeta function, the Landau–Ramanujan constant and the Rogers–Ramanujan identities) and sets it beside a teaching from the Daoist, Buddhist, Confucian or yogic traditions, building from them a practice of attention and action: subtract, observe, command. Each chapter has three parts: story, trap and protocol, and an advanced practice. Throughout, marked boxes separate the author's interpretation from established knowledge: evidence boxes give sources, calculation boxes give the formulas with runnable Python checks, and where the two disagree the box says so. This is a book of interpretation and practice, not medical, psychological or financial advice. 摘要:拉马努金(1887—1920)几乎全靠自学,却留下了数千条公式,其中许多在他去世几十年后才被证明,并出现在统计物理与弦论里。本书十章,每章取他的一个数学结构——拉马努金求和、零与无穷、仿 θ 函数、1/π 级数、嵌套根式、欧拉–马斯刻若尼常数、主定理、黎曼 ζ 零点、兰道–拉马努金常数、罗杰斯–拉马努金恒等式——与道家、佛家、儒家或瑜伽传统中的一句教法并置,并由此提出一套关于注意力与行动的练习:减、观、令。 每章分为故事、陷阱与协议、进阶三篇。书中另以标注框区分作者的诠释与已确立的知识:证据框给出处,推导框给公式与可运行的 Python 验证代码,二者与作者观点不一致处一律直说。本书是诠释与实践之书,不是医疗、心理或财务建议。 Chapter answers: - Ch.1 : Ramanujan Summation — What −1/12 actually means: So "$1 + 2 + 3 + \cdots = -1/12$" is shorthand for "the finite part that remains once the infinite part is set aside". It is not a sum in the ordinary sense, and the book's use of it as an image of accumulation collapsing is the author's reading. (https://5dlibrary.com/answers/en/ramanujan-reality-matrix/c01/) 第1章 拉马努金求和 — −1/12 到底是什么意思:所以「$1 + 2 + 3 + \cdots = -1/12$」是一种简写,意思是「把无穷的部分放到一边以后剩下的有限部分」。它不是通常意义上的和;书中把它当作「累积会坍缩」的意象,是作者的读法。 (https://5dlibrary.com/answers/zh/ramanujan-reality-matrix/c01/) - Ch.2 : Zero and Infinity — Zero and infinity on the Riemann sphere: Computers that follow the IEEE 754 standard return $+\infty$ for $1.0/0.0$ (ieee754), which is the same convention the Riemann sphere uses. (https://5dlibrary.com/answers/en/ramanujan-reality-matrix/c02/) 第2章 零与无穷 — 黎曼球面上的零与无穷:遵循 IEEE 754 标准的计算机对 $1.0/0.0$ 返回 $+\infty$ (ieee754),用的正是黎曼球面的约定。 (https://5dlibrary.com/answers/zh/ramanujan-reality-matrix/c02/) - Ch.4 : Infinite Series for Pi — Eight digits per term: For comparison, the Leibniz series $\pi = 4(1 - \tfrac13 + \tfrac15 - \cdots)$ is still wrong in the sixth decimal place after a million terms. (https://5dlibrary.com/answers/en/ramanujan-reality-matrix/c04/) 第4章 π 的无穷级数 — 每项八位:作为对比,莱布尼茨级数 $\pi = 4(1 - \tfrac13 + \tfrac15 - \cdots)$ 算了一百万项,小数点后第六位仍然是错的。 (https://5dlibrary.com/answers/zh/ramanujan-reality-matrix/c04/) - Ch.5 : Nested Radical Constant — Starting from 3: how the radical unpacks: This is literally how the identity is derived. For every $n$, $$n + 1 = \sqrt{1 + n(n + 2)}$$ Start with $3$ and keep replacing the last number with its own square root form: $$3 = \sqrt{1 + 2 \cdot 4} = \sqrt{1 + 2\sqrt{16}} = \sqrt{1 + 2\sqrt{1 + 3 \cdot 5}} = \sqrt{1 + 2\sqrt{1 + 3\sqrt{1 + 4 \cdot 6}}} = \cdots$$ Ramanujan published the problem as Question 289 in the Journal of the Indian Mathematical Society in 1911 (Ramanujan 1911). That the infinite expression actually converges to 3 needs a separate argument; Herschfeld gave general conditions for such radicals in 1935 (Herschfeld 1935). (https://5dlibrary.com/answers/en/ramanujan-reality-matrix/c05/) 第5章 嵌套根式常数 — 从 3 出发:根式怎样展开:这道恒等式正是这样推出来的。对每个 $n$, $$n + 1 = \sqrt{1 + n(n + 2)}$$ 从 $3$ 开始,不断把最后一个数换成它自己的根式写法: $$3 = \sqrt{1 + 2 \cdot 4} = \sqrt{1 + 2\sqrt{16}} = \sqrt{1 + 2\sqrt{1 + 3 \cdot 5}} = \sqrt{1 + 2\sqrt{1 + 3\sqrt{1 + 4 \cdot 6}}} = \cdots$$ 拉马努金 1911 年在《印度数学会杂志》上把它作为第 289 号问题发表 (Ramanujan 1911)。这个无穷表达式确实收敛到 3,需要另外的论证;赫希菲尔德在 1935 年给出了这类根式收敛的一般条件 (Herschfeld 1935)。 (https://5dlibrary.com/answers/zh/ramanujan-reality-matrix/c05/) - Ch.6 : Euler-Mascheroni Constant — The constant γ: Euler introduced the constant in 1734; Mascheroni calculated it further in 1790. Nobody knows whether $\gamma$ is even irrational (Havil 2003). It is a pure number; reading it as friction, tolerance or a hinge is the author's metaphor. (https://5dlibrary.com/answers/en/ramanujan-reality-matrix/c06/) 第6章 欧拉–马斯刻若尼常数 — 常数 γ:欧拉在 1734 年引入这个常数;马斯刻若尼在 1790 年把它算得更远。至今没有人知道 $\gamma$ 是不是无理数 (Havil 2003)。它是一个纯粹的数;把它读作摩擦、容差或枢纽,是作者的比喻。 (https://5dlibrary.com/answers/zh/ramanujan-reality-matrix/c06/) - Ch.7 : Ramanujan's Master Theorem — Ramanujan's Master Theorem, checked: If $f(x) = \sum_{k\ge0} \frac{(-1)^k}{k!}\varphi(k)\,xk$, then, under suitable conditions (Amdeberhan 2012), $$\int_0^\infty x{s-1} f(x)\,dx = \Gamma(s)\,\varphi(-s)$$ Hardy devoted a lecture to the theorem (Hardy 1940). Example: $(1+x)^{-a} = \sum_k \frac{(-1)k}{k!}\frac{\Gamma(a+k)}{\Gamma(a)}x^k$, so $\varphi(k) = \Gamma(a+k)/\Gamma(a)$ and the theorem predicts $\int_0\infty x^{s-1}(1+x){-a}dx = \Gamma(s)\Gamma(a-s)/\Gamma(a)$. (https://5dlibrary.com/answers/en/ramanujan-reality-matrix/c07/) 第7章 拉马努金主定理 — 核对拉马努金主定理:若 $f(x) = \sum_{k\ge0} \frac{(-1)^k}{k!}\varphi(k)\,xk$,则在适当的条件下 (Amdeberhan 2012), $$\int_0^\infty x{s-1} f(x)\,dx = \Gamma(s)\,\varphi(-s)$$ 哈代专门用一讲讨论了这个定理 (Hardy 1940)。例:$(1+x)^{-a} = \sum_k \frac{(-1)k}{k!}\frac{\Gamma(a+k)}{\Gamma(a)}x^k$,所以 $\varphi(k) = \Gamma(a+k)/\Gamma(a)$,定理预言 $\int_0\infty x^{s-1}(1+x){-a}dx = \Gamma(s)\Gamma(a-s)/\Gamma(a)$。 (https://5dlibrary.com/answers/zh/ramanujan-reality-matrix/c07/) - Ch.10 : Rogers-Ramanujan Identities — What the identity counts: The first Rogers–Ramanujan identity says: the number of ways to split $n$ into parts that differ from each other by at least 2 equals the number of ways to split $n$ into parts that leave remainder 1 or 4 when divided by 5. Rogers found the identities in 1894 (Rogers 1894); Ramanujan rediscovered them, and the two published a joint proof in 1919 (Rogersramanujan 1919). (https://5dlibrary.com/answers/en/ramanujan-reality-matrix/c10/) 第10章 罗杰斯–拉马努金恒等式 — 这个恒等式数的是什么:第一个罗杰斯–拉马努金恒等式说:把 $n$ 拆成彼此至少相差 2 的部分的方法数,等于把 $n$ 拆成除以 5 余 1 或余 4 的部分的方法数。罗杰斯在 1894 年发现了这些恒等式 (Rogers 1894);拉马努金独立地重新发现了它们,两人在 1919 年联名发表了证明 (Rogersramanujan 1919)。 (https://5dlibrary.com/answers/zh/ramanujan-reality-matrix/c10/) ## 5DL-B-002 · Turning the Hearing Inward 中文书名:反闻闻自性 DOI: https://doi.org/10.5281/zenodo.23666557 Cite: Fan, Y. (2026). Turning the Hearing Inward (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23666557 Book page: https://5dlibrary.com/books/fanwen-wenzixing/ Scope: Marked boxes separate scripture, published research, runnable Python calculations and the author's hypotheses. Abstract: The sixth fascicle of the Śūraṅgama Sūtra records how Avalokiteśvara entered samādhi through hearing, and Mañjuśrī's verse praises the practice of turning the hearing back to hear one's own nature. In teaching singing, the author developed this passage into a listening practice he calls the Ergen Yuantong method, in three stages: acoustic decoupling (separating air- and bone-conducted sound, and no longer squeezing the throat to chase an outer sound), turning the hearing inward (moving attention from controlling the voice to listening to sound and vibration), and phonating from the zero point (establishing a clear inner pitch before the sound begins). This book reads the scripture side by side with auditory science and voice acoustics: how bone and air conduction let us hear our own voice, how the stapedius reflex and speaking-induced suppression turn it down, how the pitch-shift reflex and auditory imagery take part in pitch accuracy, and how the vocal tract, as a quarter-wave resonator, forms formants and standing waves. Marked boxes separate scripture, published research, runnable Python calculations and the author's hypotheses. The book gives a four-week daily practice and publishes a pre-registered protocol that could prove the method does not work. It makes no medical or therapeutic claims; 432 and 528 Hz are used only as images for an inner reference, and there is no evidence that they have special effects. 摘要:《楞严经》卷六记载观世音菩萨「从闻思修入三摩地」,文殊菩萨评为「反闻闻自性,性成无上道」。作者在声乐教学中把这一段经文发展为一种聆听训练,称为耳根圆通法,分三个阶段:声学解耦(分清气导与骨导,不再为追逐外在的声音而挤压喉部)、反闻(把注意力从控制发声转到聆听声音与振动)、零点发声(在声音之前先建立清楚的内在音高)。 本书把经文与听觉科学、嗓音声学并排阅读:骨导与气导如何让人听见自己,镫骨肌反射与「言语诱发抑制」如何调低自身声音,音高偏移反射与听觉想象如何参与音准,声道作为四分之一波长共鸣管如何形成共振峰与驻波。书中以标注框区分经文、已发表的研究、可运行的 Python 推导与作者的假说,给出一套四周的每日练习,并公开一份可以被推翻的预注册实验方案。书中不作任何医疗或疗效声明;432/528 Hz 只作为内在参照的意象,没有证据表明它们有特殊作用。 Chapter answers: - Ch.2 How the Ear Hears Itself — - We hear ourselves through air and bone conduction; the bone route dominates between about 1 and 2 kHz for most speech sounds, so what we hear of ourselves differs from what others hear. - The two routes arrive within a millisecond of each other; they differ in filtering, not in timing. - The stapedius muscle, speaking-induced suppression and the pitch-shift reflex show that the body is constantly processing its own voice; singers also steer by feel. - Vivid auditory imagery goes together with more accurate pitch imitation. - The amplifier of the inner ear is under the brain's control, and attention can affect it. - Whether the Ergen Yuantong method improves pitch through these mechanisms is the author's hypothesis, to be tested in Chapter 8. (https://5dlibrary.com/answers/en/fanwen-wenzixing/c02/) 第2章 耳朵怎样听见自己 — - 我们经由气导和骨导两条路听见自己;对多数语音,约 1–2 kHz 之间以骨导为主,所以自己听到的和别人听到的不同。 - 两条路到达的时间相差不到一毫秒;差别在滤波,不在时间。 - 镫骨肌、言语诱发抑制、音高偏移反射说明身体一直在处理自己的声音;歌者也靠身体感觉驾驭。 - 听觉想象越清晰,模唱音高往往越准。 - 内耳的放大器受大脑调控,注意力能影响它。 - 耳根圆通法是否通过这些机制改善音准,是作者的假说,留待第八章检验。 (https://5dlibrary.com/answers/zh/fanwen-wenzixing/c02/) - Ch.5 Stage Three: Phonating from the Zero Point — - Every note is prepared before it sounds: the brain predicts the pitch it is about to hear, and the ear prepares for it. - Imagining music uses much of the machinery of hearing; vivid auditory imagery goes together with more accurate pitch imitation. - Stage Three starts each note from a clearly heard inner pitch in the silence before the sound, the "zero point". - 432 and 528 Hz are used only as images for a fixed inner reference; there is no reliable evidence that they have special effects, and the experiment in Chapter 8 does not use them. - That starting from inner hearing improves the accuracy of note onsets is a hypothesis. (https://5dlibrary.com/answers/en/fanwen-wenzixing/c05/) 第5章 第三阶段:零点发声——闻自性 — - 每个音在响起之前就已准备:大脑预测将要听到的音高,耳朵也为之准备。 - 想象音乐动用的很多是听觉的机制;听觉想象越清晰,模唱音高往往越准。 - 第三阶段让每个音从声音之前的寂静——「零点」——里清楚听见的内在音高开始。 - 432 与 528 Hz 只作为固定内在参照的意象;没有可靠证据表明它们有特殊作用,第八章的实验也不使用它们。 - 从内在的听起音能提高起音的准确度,是有待检验的假说。 (https://5dlibrary.com/answers/zh/fanwen-wenzixing/c05/) - Ch.6 The Physics of the Voice: Harmonics, Resonance and Standing Waves — - The voice is a source (the vocal folds, producing harmonics at multiples of $f_0$) filtered by the vocal tract (whose resonances are the formants). - A uniform 17 cm tube closed at one end has resonances near 515, 1544 and 2574 Hz; these are standing waves in the air between the folds and the lips. - Harmonics that fall near formants are strengthened; at high pitches, sopranos move the first resonance to follow the fundamental. - The singer's formant near 3 kHz lets a voice carry over an orchestra without extra loudness. - The physics supports the method's account of where vibration comes from and why effort is an inefficient way to be heard; it does not show that the method works. (https://5dlibrary.com/answers/en/fanwen-wenzixing/c06/) 第6章 嗓音的物理:谐波、共鸣与驻波 — - 嗓音是声源(声带,产生 $f_0$ 整数倍的谐波)经过声道滤波(声道的共鸣就是共振峰)。 - 一根一端封闭的 17 cm 均匀管子,共鸣在约 515、1544、2574 Hz;这些是声带与嘴唇之间空气里的驻波。 - 落在共振峰附近的谐波被加强;在高音区,女高音会移动第一共鸣去追随基频。 - 3 kHz 附近的歌手共振峰,让嗓音不必更响就能穿透乐队。 - 物理支持这个方法对「振动从哪里来」「为什么用力是低效的被听见方式」的说明,但并不表明这个方法有效。 (https://5dlibrary.com/answers/zh/fanwen-wenzixing/c06/) - Ch.8 An Experiment That Could Prove It Wrong: Pre-registered Protocol 1.0 — - The protocol is public before any data are collected, and the criterion is fixed in advance. - Both groups sing the same exercises for the same time; only the instructions differ, so the test isolates the method's instructions about attention. - The primary outcome is pitch accuracy measured objectively in cents; onset accuracy, effort and inner-reference error are secondary. - About 64 per group are needed (76 allowing for drop-out); results will be published whether positive or negative. (https://5dlibrary.com/answers/en/fanwen-wenzixing/c08/) 第8章 一个可以被推翻的实验:预注册方案 1.0 — - 方案在收集任何数据之前公开,判定标准事先写死。 - 两组唱同样的练习、同样的时长,只有指导语不同,所以检验的正是方法关于注意力的指导。 - 主要结局是以音分客观测量的音准;起音准确度、用力程度和内在参照误差为次要结局。 - 每组约需 64 人(考虑流失为 76 人);结果不论正负都会发表。 (https://5dlibrary.com/answers/zh/fanwen-wenzixing/c08/) ## 5DL-B-003 · The Sacred Cubit and Resonant Geometry 中文书名:神圣肘尺与谐振几何 DOI: https://doi.org/10.5281/zenodo.23666560 Cite: Fan, Y. (2026). The Sacred Cubit and Resonant Geometry (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23666560 Book page: https://5dlibrary.com/books/sacred-cubit-geometry/ Scope: Marked boxes separate museum records and published sources, runnable Python calculations and the author's hypotheses. Abstract: The royal cubit of ancient Egypt was a standard ruler about half a metre long, divided into seven palms and twenty-eight fingers. Today it is often called a "sacred" length and used to size copper rings and coils. This book does not ask whether the cubit is sacred. It asks five measurable questions: how long the surviving rods are (museum records in Turin, Paris and London give 52.3–52.5 cm; the book uses 523.5 mm, the lower bound of the published range, as an engineering constant); which sounds and radio waves a cubit resonates with, using λ = c/f; when a twisted copper ring cancels its own magnetic field (it depends on whether the strands are insulated and how the joint is made), and how the twist contraction factor sets the cut length; how fields add in a spherical octahedron of three orthogonal cubit rings, and why such rings do not couple; and how readers can repeat every measurement at home with a NanoVNA and an RTL-SDR receiver. Marked boxes separate museum records and published sources, runnable Python calculations and the author's hypotheses. A measurement plan for seven objects states its predictions in advance, and the raw data will be published unaltered whatever the results. The author's group makes and sells cubit rings and tensor coils, as disclosed in the book; the book makes no health or medical claims. 摘要:古埃及皇家肘尺是一把约半米长、分为七掌二十八指的标准尺,今天又常被说成一种「神圣」的长度,并用作铜环与线圈的尺寸依据。本书不讨论肘尺是否神圣,而是提出五个可以测量的问题:现存肘尺有多长(都灵、卢浮宫、大英博物馆的馆藏记录为 52.3–52.5 厘米;本书以文献范围下限 523.5 毫米为工程常数);按 λ = c/f,一肘尺与哪些声音和无线电波共振;双绞铜环在什么条件下抵消自身磁场(结论取决于股间是否绝缘、焊点如何连接),以及绞合收缩率如何决定下料长度;三个正交肘尺环构成的球面八面体中,场如何叠加、为何互不耦合;读者如何用 NanoVNA 与 RTL-SDR 在家中复现全部测量。 书中以标注框区分博物馆记录与已发表资料、可运行的 Python 推导与作者的假说;七个物件的测量方案事先写下预测,并承诺不论结果如何都原样公开原始数据。作者所在集团制作并销售肘尺铜环与张量线圈,书中已作利益披露;本书不作任何健康或医疗声明。 Chapter answers: - Ch.1 How Long Is a Royal Cubit? — - The royal cubit was a fixed standard of seven palms and twenty-eight fingers, kept by master rods. - Four well-documented rods in Turin, Paris and London are recorded at 52.3 to 52.5 cm; surveys of all surviving rods give about 523.5 to 529.2 mm. - This book uses 523.5 mm, the lower bound of the published range, as an engineering constant; it is not a new measurement by the author. - A 0.3% difference in length moves a resonance by 0.3%, less than other practical effects. (https://5dlibrary.com/answers/en/sacred-cubit-geometry/c01/) 第1章 皇家肘尺有多长? — - 皇家肘尺是一个固定的标准,七掌、二十八指,由标准尺保存。 - 都灵、巴黎、伦敦四把记录完整的肘尺,记录长度为 52.3 到 52.5 厘米;对全部现存肘尺的整理给出约 523.5 到 529.2 毫米。 - 本书以已发表范围的下限 523.5 毫米作为工程常数;它不是作者本人的新测量。 - 长度差 0.3%,共振移动 0.3%,小于其他实际因素的影响。 (https://5dlibrary.com/answers/zh/sacred-cubit-geometry/c01/) - Ch.2 The Wavelength of a Cubit: $\lambda = c/f$ — - Wavelength, frequency and speed are linked by $\lambda = c/f$; the same length fits sound of a few hundred hertz and radio of a few hundred megahertz. - A one-cubit pipe resonates near 164 Hz (closed at one end) or 328 Hz (open at both ends), close to E3 and E4. - A one-cubit wire on a ground plane is an ideal quarter-wave aerial at 143.2 MHz, about 136 MHz in practice. - Every length has resonances; the calculation gives the cubit's values but does not make them special. (https://5dlibrary.com/answers/en/sacred-cubit-geometry/c02/) 第2章 一肘尺的波长:$\lambda = c/f$ — - 波长、频率与速度由 $\lambda = c/f$ 联系起来;同一个长度,能装下几百赫兹的声音和几百兆赫的无线电波。 - 一肘尺长的管子,一端封闭时共振在约 164 Hz,两端开口时约 328 Hz,接近 E3 与 E4。 - 立在地面上的一肘尺导线,是 143.2 MHz 的理想四分之一波长天线,实际约 136 MHz。 - 任何长度都有共振;计算给出了肘尺的数值,但并不使它们特殊。 (https://5dlibrary.com/answers/zh/sacred-cubit-geometry/c02/) - Ch.3 The Twisted Copper Ring: Flux Cancellation and SWR — - The standard ring uses annealed 99.9% C11000 copper, 1.6 or 2.0 mm, folded, twisted at a fixed pitch and closed with lead-free silver-bearing solder. - Twisting shortens the pair by the factor $k = p/\sqrt{p^2 + (\pi d)2}$; strands are cut to $523.5/k$ mm plus allowances, from about 525 mm at one twist per palm to over 550 mm at one twist per finger in 2 mm wire. - Opposed currents cancel the far field by more than 20 dB, but a closed ring supports opposed currents only if its strands are insulated and its joint leaves an insulated gap; a bare ring closed at one joint behaves like a single loop. - Measurement plan 1.0 compares five objects and states its predictions in advance; the results will be published whatever they are. (https://5dlibrary.com/answers/en/sacred-cubit-geometry/c03/) 第3章 双绞铜环:磁通抵消与驻波比 — - 标准铜环用退火 99.9% C11000 紫铜,线径 1.6 或 2.0 毫米,对折、固定节距绞合、用无铅含银焊料闭合。 - 绞合使线对按系数 $k = p/\sqrt{p^2 + (\pi d)2}$ 缩短;每股按 $523.5/k$ 毫米加余量下料,从每掌一绞的约 525 毫米,到 2 毫米线每指一绞的 550 毫米以上。 - 反向电流使远场减弱 20 dB 以上;但闭合的环只有在股间绝缘、焊点留有绝缘间隙时才能承载反向电流。用一个焊点闭合的裸线环,表现为单圈环。 - 测量方案 1.0 比较五个物件,事先写下预测;结果不论如何都会发表。 (https://5dlibrary.com/answers/zh/sacred-cubit-geometry/c03/) - Ch.4 Multi-Ring Orthogonal Geometry: The Spherical Octahedron and the Superposition of Fields — - Three orthogonal rings of one cubit lie on a sphere 166.6 mm across, cross at the six vertices of an octahedron and divide the sphere into eight right-angled triangles. - Their fields add as vectors: in phase they act as one tilted ring; in quadrature they make a rotating field. Without a current they make no field. - Rings at right angles have zero mutual inductance; together they respond equally to fields from every direction. - Soldering the crossings turns three rings into one network with different behaviour; the measurement plan tests both. (https://5dlibrary.com/answers/en/sacred-cubit-geometry/c04/) 第4章 多环正交几何:球面八面体与场的叠加 — - 三个一肘尺的正交环位于一个直径 166.6 毫米的球面上,交于八面体的六个顶点,把球面分成八个直角三角形。 - 它们的场按矢量相加:同相时相当于一个倾斜的单环;正交时产生旋转磁场。没有电流就没有磁场。 - 互相垂直的环互感为零;合在一起,对来自任何方向的磁场响应相同。 - 焊接交点会把三个环变成一个行为不同的网络;测量方案同时检验两者。 (https://5dlibrary.com/answers/zh/sacred-cubit-geometry/c04/) - Ch.5 An Open-Hardware Replication Guide: A Home RF Bench with NanoVNA and RTL-SDR — - A NanoVNA (to about 1.5 GHz) and an RTL-SDR receiver (to about 1.77 GHz), with cables, a ground plate, probe loops and a non-conducting fixture, are enough for every measurement in this book. - Calibrate at the end of the cables, verify with the 50 Ω load, and repeat every measurement three times. - Export sweeps as Touchstone files; a few lines of Python find the SWR minimum and fingerprint the file. - Use the RTL-SDR to check the frequency scale against signals of known frequency. - Keep a log, publish raw files with their hashes and DOIs, and report results whichever way they come out. (https://5dlibrary.com/answers/en/sacred-cubit-geometry/c05/) 第5章 开源硬件复现指南:用 NanoVNA 与 RTL-SDR 建立家庭射频测量台 — - 一台 NanoVNA(约到 1.5 GHz)、一台 RTL-SDR 接收器(约到 1.77 GHz),加上线、地面板、探测环和不导电夹具,足以完成本书的每一项测量。 - 在线的末端校准,用 50 Ω 负载验证,每项测量重复三次。 - 把扫描导出为 Touchstone 文件;几行 Python 就能找出 SWR 最小值并给文件做指纹。 - 用 RTL-SDR 对照已知频率的信号核对频率刻度。 - 做记录,发表原始文件及其哈希和 DOI,不论结果如何都如实报告。 (https://5dlibrary.com/answers/zh/sacred-cubit-geometry/c05/) ## 5DL-B-004 · The Thermodynamics of Desire 中文书名:欲望的热力学 DOI: https://doi.org/10.5281/zenodo.23682498 Cite: Fan, Y. (2026). The Thermodynamics of Desire (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23682498 Book page: https://5dlibrary.com/books/thermodynamics-of-desire/ Scope: The book is not therapy, relationship counselling or medical advice. Marked boxes separate published evidence, runnable Python calculations, classical texts and the author's hypotheses; wherever a comparison between physics and love is an analogy, the book says so, and every hypothesis states what result would refute it. Abstract: Why does the excitement of new love fade, and why do long relationships so often drift into weariness and friction? This book reads four bodies of knowledge side by side: thermodynamics and information theory (Boltzmann's entropy S = k_B ln Ω, Shannon's entropy, Schrödinger's organism that "feeds on negative entropy"), the neuroscience of reward (Schultz's dopamine reward prediction error and temporal-difference learning), the physics of self-organisation (Prigogine's dissipative structures and Friston's free-energy principle), and the teachings of the Daodejing and of Buddhism on "leaking" (āsrava). Its argument is that desire and intimacy are a form of order kept up by a flow: when the inflow stops, the order relaxes towards its most probable state. The task is not to fight the increase of entropy but to understand what flow a relationship needs and what makes it leak. The eight chapters cover entropy and the Shannon entropy of attention; dopamine prediction error and hedonic adaptation; dissipative structures and a model of hysteresis and sudden collapse; the free-energy principle and the boundary at which safety makes novelty affordable; how variable-ratio rewards capture attention by offering counterfeit information; novelty, meaning and repair as the three inflows of a relationship, with critical slowing down as an early warning; a four-week programme any couple can follow at home; and a pre-registered three-arm randomised trial comparing weekend novelty with a nightly screen-free hour. Marked boxes separate published evidence, runnable Python calculations, classical texts and the author's hypotheses; wherever a comparison between physics and love is an analogy, the book says so, and every hypothesis states what result would refute it. The book is not therapy, relationship counselling or medical advice. 摘要:热恋的兴奋为什么会消退,长久的关系为什么常常走向倦怠与内耗?本书把四类知识并排来读:热力学与信息论(玻尔兹曼熵 S = k_B ln Ω、香农熵、薛定谔的「以负熵为食」)、奖赏的神经科学(舒尔茨的多巴胺奖赏预测误差与时间差分学习)、自组织物理学(普里戈金的耗散结构、弗里斯顿的自由能原理),以及《道德经》与佛教关于「漏」的教法。全书的论点是:欲望与亲密是一种靠流维持的秩序,流入一停,秩序就向最可能的状态松弛;关键不在对抗熵增,而在弄清一段关系需要怎样的负熵流,又是什么让它漏掉。 全书八章:熵与注意力的香农熵;多巴胺预测误差与享乐适应;耗散结构与滞后崩塌模型;自由能原理与「安全买来新奇」的最优边界;可变比率奖赏如何以伪造的信息价值抽走注意力;新奇、意义与修复三股负熵流入及临界慢化;一套可在家执行的四周逆熵练习;以及一份比较「增加周末新奇」与「每晚堵住一小时屏幕漏出」的三组随机对照试验预注册方案。书中以标注框区分已发表的证据、可运行的 Python 推导(32 段,全部核对)、原典与作者的假说;凡物理与爱情之间只是类比的地方都直接说明,每个假说都写明什么结果会推翻它。本书不是心理治疗、婚恋咨询或医疗建议。 Chapter answers: - Ch.4 The Free-Energy Principle: A Mind That Wants Both Safety and Novelty — The two columns tell different stories. Surprisal falls at first and then stops falling: it approaches 0.500 nats, the entropy of the 0.8/0.2 pattern itself, and never goes below it. That remaining surprise is noise; no amount of learning can remove it. Information gain falls by a factor of ten and keeps falling towards zero. After a hundred evenings, the partner is still a little unpredictable, but no longer teaches anything. (https://5dlibrary.com/answers/en/thermodynamics-of-desire/c04/) 第4章 自由能原理:既要安全又要新奇的心智 — 两列数字讲的是不同的故事。意外度起初下降,然后不再下降:它趋近 0.500 奈特,也就是 0.8/0.2 这个模式本身的熵,永远不会低于它。剩下的意外是噪声,再怎么学习也消除不了。信息增益则下降到原来的十分之一,并继续趋向零。一百个晚上之后,伴侣仍有一点不可预测,却已经教不了什么了。 (https://5dlibrary.com/answers/zh/thermodynamics-of-desire/c04/) - Ch.5 Leaking and Not Leaking: An Energy Account of Attention — Even without any pull of uncertainty ($\kappa = 0$), some minutes go to the phone: a choice by probability always samples the alternatives. But once uncertainty attracts, the phone takes 112 of 180 minutes, although its average reward is half the partner's. The partner has not become less rewarding. The partner has become known, and the phone never will be. When the phone is in another room, the leak is zero and the entropy of the evening falls to zero bits. (https://5dlibrary.com/answers/en/thermodynamics-of-desire/c05/) 第5章 漏与不漏:注意力的能量账 — 即使完全没有不确定性的拉力($\kappa = 0$),也有一些分钟给了手机:按概率做选择总会抽样到其他选项。但一旦不确定性会吸引注意,手机就拿走了 180 分钟里的 112 分钟,尽管它的平均奖赏只有伴侣的一半。伴侣并没有变得不那么令人愉快。伴侣只是变得*已知*了,而手机永远不会。手机放到另一个房间时,漏出为零,晚上的熵降到零比特。 (https://5dlibrary.com/answers/zh/thermodynamics-of-desire/c05/) - Ch.6 The Negentropy Flow of a Relationship: Novelty, Meaning and Repair — With the same drive and the same conflicts, couples who do not repair collapse in 86 of 200 runs; modest repair cuts this to 11, and stronger repair to 3 or none. The conflicts did not change. What changed was whether they could accumulate. (https://5dlibrary.com/answers/en/thermodynamics-of-desire/c06/) 第6章 亲密关系的负熵流:新奇、意义与修复 — 驱动相同、冲突相同,不修复的伴侣 200 次中有 86 次崩塌;适度修复降到 11 次,更强的修复降到 3 次乃至没有。冲突并没有变;变的是它们能不能累积起来。 (https://5dlibrary.com/answers/zh/thermodynamics-of-desire/c06/) - Ch.8 A Pre-Registered Study: Protocol 1.0 — The study is planned to detect a standardised difference of $d = 0.30$ for H1 with 90 % power, which requires 207 couples per arm, or 244 per arm (732 couples) allowing for 15 % dropout. With 207 couples per arm completing the study, H2 has about 76 % power for $d = 0.25$. The values of $d$ are assumptions, chosen to be modest because effects reported in small studies tend to be overestimated. (https://5dlibrary.com/answers/en/thermodynamics-of-desire/c08/) 第8章 预注册研究方案 1.0 — 研究按以 90% 的检验力检出 H1 的标准化差异 $d = 0.30$ 来设计,需要每组 207 对伴侣,考虑 15% 的流失后每组 244 对(共 732 对)。按每组 207 对完成研究计算,H2 对 $d = 0.25$ 的检验力约为 76%。$d$ 的取值是假设,有意取得保守,因为小样本研究报告的效应往往被高估。 (https://5dlibrary.com/answers/zh/thermodynamics-of-desire/c08/) ## 5DL-B-005 · Fasting, Ketones and the Metabolic Basis of Attention 中文书名:断食、酮体与注意力的代谢基础 DOI: https://doi.org/10.5281/zenodo.23765333 Cite: Fan, Y. (2026). Fasting, Ketones and the Metabolic Basis of Attention (Version 1.0.1). 5D Library. https://doi.org/10.5281/zenodo.23765333 Book page: https://5dlibrary.com/books/fasting-ketones-attention/ Scope: The book is not medical advice; it lists who should not fast, sets no weight targets and recommends no product or supplement dose. Marked boxes separate published evidence, calculations, classical sources and the author's hypotheses, each with the result that would refute it. Abstract: Does an empty stomach clear the mind? This book follows the body along the timeline after a last meal: liver glycogen running down while gluconeogenesis takes over, the liver turning fatty acids into ketone bodies, and the brain gradually switching part of its fuel to β-hydroxybutyrate (George Cahill's fuel-switch curve). It examines the four ideas that made fasting famous: Yoshinori Ohsumi's autophagy (Nobel Prize 2016), β-hydroxybutyrate as a signalling molecule, Richard Veech's argument that ketones are a thermodynamically more efficient fuel, and NAD⁺ and the sirtuins. Runnable Python calculations check the claims: by stoichiometry, ketones do not yield more ATP per molecule of oxygen than glucose, and Veech's argument rests instead on the free energy of ATP hydrolysis and mitochondrial redox state; a one-day fast in a mouse is metabolically closer to a week in a human. Across nine chapters the book also uses a small differential-equation model of glycogen and ketones to compare 12:12, 16:8, one meal a day and a weekly 48-hour fast against the published randomised trials; reads the classical record, from the Zhuangzi's "fasting of the heart" and the Inner Classic's "measure in eating" to the Mawangdui manuscript on abstaining from grain, Ge Hong's scepticism and the Buddhist rule of not eating after noon; and ends with a four-week observation log that goes no further than 16:8, with stop rules, and a pre-registered randomised crossover trial of 16:8 eating and sustained attention. Marked boxes separate published evidence, calculations, classical sources and the author's hypotheses, each with the result that would refute it. The book is not medical advice; it lists who should not fast, sets no weight targets and recommends no product or supplement dose. 摘要:空腹真的能让头脑清明吗?本书沿着最后一餐之后的时间线,阅读断食的生物学:肝糖原的耗竭与糖异生的接手、肝脏把脂肪酸转化为酮体、大脑逐步改用β-羟丁酸(卡希尔的燃料切换曲线),以及让断食出名的四个观念——大隅良典的细胞自噬(2016 年诺贝尔奖)、作为信号分子的β-羟丁酸、维奇关于酮体热力学效率的论证、NAD⁺ 与去乙酰化酶。书中用可运行的 Python 计算核对这些说法:按化学计量,酮体每分子氧气产生的 ATP 并不比葡萄糖多,维奇的论证实际建立在 ATP 水解自由能与线粒体氧化还原状态之上;小鼠的一天断食在代谢上约等于人的一周。 全书九章,另用一个糖原-酮体微分方程模型比较 12:12、16:8、一日一餐和每周一次 48 小时断食,并对照已发表的随机试验;阅读《庄子》「心斋」、《黄帝内经》「食饮有节」、马王堆《却谷食气》、葛洪对断谷的怀疑与佛教过午不食;最后给出一份只到 16:8 的四周观察记录表(附停止规则),以及一份 16:8 饮食与持续注意力的预注册随机交叉试验方案。标注框区分已发表的证据、计算、原典与作者的假说,每个假说都写明什么结果会推翻它。本书不是医疗建议,列明了禁忌人群,不设体重目标,不推荐任何产品或补充剂剂量。 Chapter answers: - Ch.2 Autophagy: Ohsumi's Yeast and the Limits of Translation — The scaling is a rule of thumb, not a precise conversion, but it is a warning: a 24-hour fast that exhausts a mouse's liver glycogen and pushes it deep into ketosis is metabolically closer to a human fast of several days than to a human skipping one day's meals. Claims that human autophagy "switches on at hour 16" or "peaks at hour 48" have no direct human measurement behind them. (https://5dlibrary.com/answers/en/fasting-ketones-attention/c02/) 第2章 细胞自噬:大隅良典的酵母与外推的边界 — 这种换算只是经验法则,不是精确换算,但它是一个警告:让小鼠肝糖原耗尽、深度酮症的 24 小时断食,在代谢上更接近人断食数日,而不是人少吃一天的饭。所谓人类自噬「第 16 小时开启」或「第 48 小时达到高峰」,背后没有直接的人体测量。 (https://5dlibrary.com/answers/zh/fasting-ketones-attention/c02/) - Ch.3 Ketones and the Brain: β-Hydroxybutyrate as Fuel and Signal — The model is crude, and the real relation varies between people and with how long the body has been adapted to ketones. But the order of magnitude is the point: at the 0.2–0.5 mM typical at the end of a 16-hour fast, ketones supply perhaps a tenth of the brain's fuel; the rest is still glucose. The brain of a person on a 16:8 schedule is not "running on ketones". Any effect of such a schedule on attention, if there is one, is unlikely to come from a change of fuel alone. (https://5dlibrary.com/answers/en/fasting-ketones-attention/c03/) 第3章 酮体与大脑:作为燃料与信号的β-羟丁酸 — 这个模型很粗糙,真实的关系因人而异,也与身体适应酮体的时间长短有关。但要紧的是数量级:在 16 小时断食结束时常见的 0.2–0.5 mM,酮体大约提供大脑燃料的十分之一,其余仍是葡萄糖。实行 16:8 的人,大脑并没有「靠酮体运转」。如果这种时间表对注意力有影响,那不太可能只来自燃料的改变。 (https://5dlibrary.com/answers/zh/fasting-ketones-attention/c03/) ## 5DL-B-006 · The Economics of Attention 中文书名:注意力的经济学 DOI: https://doi.org/10.5281/zenodo.23689302 Cite: Fan, Y. (2026). The Economics of Attention (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23689302 Book page: https://5dlibrary.com/books/economics-of-attention/ Scope: The book is not business, investment or legal advice. Marked boxes separate published evidence, runnable Python calculations (28, all checked), classical sources and the author's hypotheses, each with the result that would refute it. Abstract: When content is no longer scarce, what is? Herbert Simon observed that "a wealth of information creates a poverty of attention". This book applies economics, network science and decision theory to that scarcity: the daily ceiling on human attention (about 10 bits per second of conscious throughput, some 72 kilobytes a day) and its shadow price; how preferential attachment produces power laws and highly concentrated audiences, with a network-growth simulation and Gini coefficients; the Bikhchandani–Hirshleifer–Welch model of information cascades and Granovetter's threshold model, which explain why a live-commerce room either takes off or stays empty; the conversion funnel as an absorbing Markov chain, with the elasticity of every link; Rochet and Tirole's two-sided markets, Armstrong's pricing rule and double marginalisation between platforms, agencies and creators; and, through Coase, Stigler, Akerlof and Spence, how structured data and knowledge graphs lower the cost of trusting a stranger. The eight chapters are read beside Sima Qian's "Biographies of the Money-Makers" in the Shiji, and end with a four-week attention ledger for creators and brands and a pre-registered randomised trial comparing long-form depth with high-frequency short content on 90-day repeat purchase, with an equivalence-style decision rule written as code. Marked boxes separate published evidence, runnable Python calculations (28, all checked), classical sources and the author's hypotheses, each with the result that would refute it. The book is not business, investment or legal advice. 摘要:内容不再稀缺时,什么是稀缺的?赫伯特·西蒙指出「信息的丰富造成注意力的贫乏」。本书用经济学、网络科学与决策理论研究这种稀缺:一个人一天的注意力上限(约每秒 10 比特、每天约 72 千字节的有意识处理量)与影子价格;优先连接如何产生幂律与极端集中的受众(附网络生长模拟与基尼系数);Bikhchandani–Hirshleifer–Welch 信息级联与格兰诺维特阈值模型,解释直播间为何要么爆发、要么冷场;把转化漏斗写成吸收马尔可夫链,算出每一环的弹性;罗歇–梯若尔双边市场、阿姆斯特朗定价规则与平台—机构—创作者之间的双重加价;以及科斯、施蒂格勒、阿克洛夫与斯彭斯视角下,结构化数据与知识图谱如何降低信任陌生人的成本。 全书八章,并与司马迁《史记·货殖列传》(「善者因之」「人弃我取,人取我与」、范蠡「三致千金,再分散」)并读;最后给出一份创作者与品牌方的四周注意力账本,以及一项比较「长内容深度」与「高频短内容」对 90 天复购率影响的预注册随机对照试验方案(每组 3,939 人,等效性判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(28 段,全部核对)、原典与作者的假说,每个假说都写明什么结果会推翻它。本书不是商业、投资或法律建议。 Chapter answers: - Ch.2 Power Laws: Why Audiences Concentrate — Under preferential attachment the best-connected node has 356 links, against 24 under uniform attachment, and the estimated exponent is close to the theoretical 3. Under uniform attachment the distribution is not a power law at all; its "exponent" in the table is only what the estimator returns when applied to the wrong shape. (https://5dlibrary.com/answers/en/economics-of-attention/c02/) 第2章 幂律:观众为什么会集中 — 在优先连接下,连接最多的节点有 356 条连接,均匀连接下只有 24 条;估计的指数接近理论值 3。均匀连接下的分布根本不是幂律,表中它的「指数」只是把估计方法用在错误形状上得到的数字。 (https://5dlibrary.com/answers/zh/economics-of-attention/c02/) - Ch.3 Information Cascades: Why a Room Explodes or Stays Empty — A change in the average threshold of half a percentage point, from 0.265 to 0.270, moves the outcome from almost everyone to under 6 per cent. Nothing about the product changed. Duncan Watts showed the same kind of discontinuity for cascades on random networks: global cascades occur only within a window of network connectivity, and when they occur they are rare and large (Watts 2002). (https://5dlibrary.com/answers/en/economics-of-attention/c03/) 第3章 信息级联:直播间为什么会爆发或冷场 — 平均阈值变动半个百分点,从 0.265 到 0.270,结果就从几乎所有人变成不到 6%。产品本身什么都没变。邓肯·瓦茨证明了随机网络上的级联也有同样的不连续性:全局级联只在网络连通度的一个窗口内出现,一旦出现,就既罕见又巨大 (Watts 2002)。 (https://5dlibrary.com/answers/zh/economics-of-attention/c03/) - Ch.5 Two-Sided Markets and the Commission Game — The table carries three lessons for agencies. An intermediary that adds nothing and takes its profit-maximising cut halves the business and leaves the creator with a quarter of what they earned alone. An agency that genuinely raises productivity can create value for everyone, but only if its cut leaves the creator enough incentive: with a 40 per cent productivity gain and a 50 per cent cut, the creator is still worse off than alone; with the same gain and a 20 per cent cut, the creator, the platform and total sales are all better off, and the agency earns less than its narrow optimum but more than nothing. And the platform gains from moderate agency cuts, which helps explain why platforms often set rules or caps on how intermediaries are paid. (https://5dlibrary.com/answers/en/economics-of-attention/c05/) 第5章 双边市场与抽成博弈 — 这张表给机构三点教训。一个不增加任何价值、却按利润最大化抽成的中间人,会让生意减半,创作者只剩下独立工作时收入的四分之一。真正提高生产率的机构能为所有人创造价值,但前提是它的抽成给创作者留下足够的积极性:生产率提高 40%、抽成 50% 时,创作者仍不如独立工作;同样的提高、抽成 20% 时,创作者、平台和总销售额都更好,机构的收入低于它狭义的最优值,但远好于一无所获。平台也从温和的机构抽成中获益,这有助于解释为什么平台常对中间人的报酬方式设定规则或上限。 (https://5dlibrary.com/answers/zh/economics-of-attention/c05/) - Ch.6 Machine-Readable Reputation: Coase in the Age of Knowledge Graphs — Two things are visible. When genuine candidates are scarce, checking costs dominate: at $q = 0.2$ and $k = 40$, a fifth of the deal's value goes on finding the right person. And cutting the cost per check, from manual vetting to records that can be verified quickly, recovers most of that value. (https://5dlibrary.com/answers/en/economics-of-attention/c06/) 第6章 机器可读的声誉:知识图谱时代的科斯 — 可以看到两点。真实的候选人稀少时,核查成本占主导:在 $q = 0.2$、$k = 40$ 时,交易价值的五分之一花在找对人上。而降低每次核查的成本,从人工审核到能快速核实的记录,可以把这部分价值的大部分找回来。 (https://5dlibrary.com/answers/zh/economics-of-attention/c06/) ## 5DL-B-007 · Molecules and Terroir 中文书名:本草的分子与风土 DOI: https://doi.org/10.5281/zenodo.23693057 Cite: Fan, Y. (2026). Molecules and Terroir (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23693057 Book page: https://5dlibrary.com/books/botanical-phytochemistry/ Scope: The book is not medical advice and does not claim that any plant or extract treats or prevents any human disease. Marked boxes separate published evidence, runnable Python calculations (36, all checked), classical sources and the author's hypotheses. Abstract: Why does the same plant make different amounts of its characteristic molecules in different places, and what is left in the jar after drying? This book answers with chemistry, physics and statistics: the Arrhenius equation and a minute-by-minute day–night carbon balance, showing that a wide day–night temperature swing helps only when days stay near the photosynthetic optimum; the 2-O-acetylglucomannan of Dendrobium officinale, its mannose-to-glucose ratio and degree of acetylation, and the Chinese Pharmacopoeia's phenol–sulfuric acid assay with a Beer–Lambert standard curve; the reversible first-order dehydration of 6-gingerol to 6-shogaol, with activation energies of 75 and 52 kJ/mol fitted to published data, and the finding that "twice as hot" dried ginger would need about two-thirds conversion; the A-type proanthocyanidins of cranberry, the BL-DMAC assay and the roughly 2.2-fold effect of the choice of standard, with laboratory anti-adhesion findings kept strictly apart from the clinical evidence of the Cochrane review; the triple point of water (611.657 Pa, 0.01 °C), the Clausius–Clapeyron equation and drying times under a uniformly retreating ice front, and the retention of heat-sensitive compounds by freeze-drying and hot-air drying; and the sampling statistics behind a certificate of analysis, the blind spot of fingerprint similarity scores, contaminants and water activity (microbial growth stops below about 0.6; the usual target of 0.2 to 0.3 is set for quality). The eight chapters are read beside the Shennong Bencao Jing, the Mingyi Bielu and Li Shizhen's Bencao Gangmu, and end with a four-step check for buying plant raw materials, binomial calculations for the triangle test, and a pre-registered blinded experiment comparing freeze-drying with hot-air drying at 60 °C for Bentong ginger (28 split lots, with the decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (36, all checked), classical sources and the author's hypotheses. The book is not medical advice and does not claim that any plant or extract treats or prevents any human disease. 摘要:同一种植物,为什么在不同产地合成的特征分子不同?干燥之后,罐子里还剩什么?本书用化学、物理与统计学回答这些问题:用阿伦尼乌斯方程与逐分钟的昼夜碳收支模型说明风土如何影响植物可分配的碳,并指出「昼夜温差越大越好」只在白天接近光合最适温度时成立;铁皮石斛 2-O-乙酰葡甘露聚糖的结构、甘露糖与葡萄糖之比、乙酰取代度,以及《中国药典》苯酚–硫酸法的朗伯–比尔标准曲线;6-姜辣素脱水为 6-姜烯酚的可逆一级动力学(按已发表数据拟合活化能 75 与 52 kJ/mol),并核算「干姜辣度翻倍」需要约三分之二的转化;蔓越莓 A 型原花青素的立体化学、BL-DMAC 测定法与对照品选择带来的约 2.2 倍差异,并把体外抗黏附发现与 Cochrane 系统综述的临床证据严格分开;水的三相点(611.657 Pa、0.01 °C)、克劳修斯–克拉佩龙方程与均匀后退冰界面模型下的冻干时间,以及冻干与热风干燥对热敏成分的保留;分析证明书的抽样统计、指纹图谱相似度的盲点、污染物与水分活度(微生物生长在约 0.6 以下停止,0.2~0.3 是出于品质而定的目标)。 八章与《神农本草经》「阴干暴干,采造时月……土地所出……并各有法」、《名医别录》与李时珍《本草纲目》并读;最后给出植物原料采购的四步核查表、三角试验的二项分布计算,以及一项比较文冬姜真空冻干与 60 °C 热风干燥的预注册盲法实验方案(28 个分批对照批次,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(36 段,全部核对)、原典与作者的假说。本书不是医疗建议,不主张任何植物或提取物能治疗或预防任何人类疾病。 Chapter answers: - Ch.3 Ginger: From 6-Gingerol to 6-Shogaol — The fitted activation energies reproduce the paper's. Below 60 °C the conversion is slow; at 80 °C it is substantial within hours; at 100 °C it reaches equilibrium, with a little more than half the pair present as shogaol, within a working day. (https://5dlibrary.com/answers/en/botanical-phytochemistry/c03/) 第3章 生姜:从 6-姜辣素到 6-姜烯酚 — 拟合得到的活化能与原文一致。60 °C 以下转化很慢;80 °C 时几小时内转化可观;100 °C 时一个工作日内就达到平衡,此时略多于一半以姜烯酚形式存在。 (https://5dlibrary.com/answers/zh/botanical-phytochemistry/c03/) - Ch.5 The Heat and Mass Transfer of Freeze-Drying — Ice can turn straight into vapour only below a certain pressure. On a phase diagram of water, the lines separating solid, liquid and vapour meet at the triple point, where all three coexist. For ordinary water it lies at 273.16 K (0.01 °C) and 611.657 Pa (Wagner 2002), about 0.6 % of atmospheric pressure. Below that pressure, liquid water cannot exist at equilibrium; warming ice makes it sublime. (https://5dlibrary.com/answers/en/botanical-phytochemistry/c05/) 第5章 真空冷冻干燥的热质传递 — 冰只有在某个压力以下才能直接变成水汽。在水的相图上,分隔固、液、气三相的曲线交于三相点,三相在此共存。普通水的三相点是 273.16 K(0.01 °C)、611.657 Pa (Wagner 2002),约为大气压的 0.6%。低于这一压力时,液态水在平衡状态下不能存在,给冰加热只会让它升华。 (https://5dlibrary.com/answers/zh/botanical-phytochemistry/c05/) - Ch.7 The Classical Materia Medica — The result is not what a quick guess would give. Ten days of shade-drying at a warm 28 °C produces more shogaol than four days of sun-drying with hotter afternoons, because time counts as well as temperature; this agrees with the experimental finding in Chapter 3 that drying time mattered more than air temperature for gingerol loss. An oven at 70 °C for eight hours converts more still. Freeze-drying converts almost nothing. Each tradition of processing was, in effect, choosing a point on this scale, and the classical preference for different preparations of ginger for different purposes (fresh, dried, and later charred 炮姜) corresponds to real chemical differences. (https://5dlibrary.com/answers/en/botanical-phytochemistry/c07/) 第7章 古典本草并读 — 结果和随手一猜不一样。在温暖的 28 °C 下阴干十天,生成的姜烯酚比午后更热、但只晒四天的晒干更多,因为时间和温度同样重要;这与第三章的实验发现一致:对姜辣素损失而言,干燥时间比空气温度更关键。70 °C 焙干八小时转化得更多。冻干几乎不转化。每一种炮制传统,实际上都在这条刻度上选了一个位置;古典医家对不同用途选用不同的姜(鲜姜、干姜,以及后来的炮姜),对应的是真实的化学差别。 (https://5dlibrary.com/answers/zh/botanical-phytochemistry/c07/) - Ch.8 Practice: A Four-Step Check and a Pre-Registered Experiment — With 12 assessors the test misses most real but modest differences. With 36 it detects a difference that half the assessors can perceive almost every time. For routine checks of incoming goods, 24 assessors is a reasonable compromise; for a decision that matters, use more. (https://5dlibrary.com/answers/en/botanical-phytochemistry/c08/) 第8章 实践:四步核查与预注册实验方案 — 12 名评价员时,检验会漏掉大多数真实但不大的差别;36 名时,一半评价员能察觉的差别几乎每次都能检出。对日常来货核查,24 名评价员是合理的折中;对重要决定,应当用更多人。 (https://5dlibrary.com/answers/zh/botanical-phytochemistry/c08/) ## 5DL-B-008 · The Default Mode Network, the Entropic Brain and Neural Oscillations 中文书名:默认模式网络、熵脑与神经振荡 DOI: https://doi.org/10.5281/zenodo.23697320 Cite: Fan, Y. (2026). The Default Mode Network, the Entropic Brain and Neural Oscillations (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23697320 Book page: https://5dlibrary.com/books/dmn-neural-oscillations/ Scope: Marked boxes separate published evidence, runnable Python calculations (28, all checked), classical sources and the author's hypotheses. Abstract: When nothing is asked of it, the brain does not go quiet. This book uses computational neuroscience, pharmacology and information theory to study the default mode network and the rhythms and molecules that change it: the brain's energy budget at rest (about 2 % of body mass, 20 % of energy, with tasks adding under 5 %) and network hubs analysed with graph theory (removing the posterior cingulate cuts global efficiency by about a fifth); the Kuramoto model of coupled oscillators and its synchronisation transition (critical coupling Kc = 2γ, simulation matching theory); 5-HT2A receptor binding with a dimensionless Hill–Langmuir equation, Lempel–Ziv complexity and Shannon entropy showing that "more entropy" is not "better", and a review of the imaging studies of Carhart-Harris and colleagues, the entropic brain hypothesis and the risks and limits of the clinical trials; the evidence that experienced meditators deactivate the default network while coupling it more strongly to control networks, with a mind-wandering Markov chain and a stability model; melatonin synthesis and its suppression by light, and why the calcite microcrystals of the pineal gland are not piezoelectric in bulk and resonate millions of times above any brain rhythm; and the state of the evidence on 40 Hz light and sound stimulation and on binaural beats. The eight chapters are read beside the Diamond Sūtra, the Platform Sūtra and the Secret of the Golden Flower, and end with a four-week sitting and mind-wandering log and a pre-registered randomised crossover trial comparing breath-focused listening with binaural beats (64 completers, with the decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (28, all checked), classical sources and the author's hypotheses. Legal and medical notice: this book is not medical or legal advice; psilocybin and related tryptamines are strictly controlled in Malaysia (Dangerous Drugs Act 1952), China and most other jurisdictions, and the book does not advise obtaining or using any controlled substance and gives no doses. 摘要:不要求大脑做任何事时,它并不安静。本书用计算神经科学、药理学与信息论研究「默认模式网络」与改变它的节律和分子:大脑静息时的能量预算(约占体重 2%、耗能 20%,任务只增加不到 5%),以及用图论计算的网络枢纽(移除后扣带回使全局效率下降约五分之一);库拉莫托耦合振子模型与同步相变(临界耦合 Kc = 2γ,模拟与理论吻合);用无量纲 Hill–Langmuir 方程描述 5-HT2A 受体结合,用 Lempel–Ziv 复杂度与香农熵说明「熵更高不等于更好」,并综述卡哈特-哈里斯等人的影像研究与「熵脑」假说及临床试验的风险与局限;资深禅修者默认网络去激活与控制网络耦合增强的证据,以及走神—觉察马尔可夫链和稳定性模型;松果体的褪黑素合成与光抑制模型,以及方解石微晶为何在块体上不具压电性、其声学谐振为何远高于脑节律;40 Hz 声光刺激与双耳节拍的证据现状。 八章与《金刚经》「应无所住而生其心」、《六祖坛经》「无念为宗」、《太乙金华宗旨》「回光」并读;最后给出四周静坐与走神捕捉记录表,以及一项比较专注听息与双耳节拍的预注册随机交叉试验(64 人完成,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(28 段,全部核对)、原典与作者的假说。法律与医学声明:本书不是医疗或法律建议;裸盖菇素等色胺类物质在马来西亚(《1952 年危险毒品法令》)、中国及绝大多数司法辖区受严格管制,本书不以任何形式建议获取或使用任何受管制物质,不提供任何剂量。 Chapter answers: - Ch.5 The Pineal Gland: From Melatonin to Calcite — The level at 23:00 falls by about half with ordinary room light and almost completely with bright light, but the total over the night falls much less, because melatonin recovers quickly once the light is off. A model like this explains why evening light matters for the timing of melatonin; it does not by itself say anything about sleep quality, which depends on the circadian clock and much else. (https://5dlibrary.com/answers/en/dmn-neural-oscillations/c05/) 第5章 松果体:从褪黑素到方解石 — 23:00 时的水平在普通室内光下下降约一半,在强光下几乎降到零;但整夜总量下降少得多,因为一关灯褪黑素就迅速恢复。这样的模型能说明为什么夜间光照会影响褪黑素的时间进程;它本身对睡眠质量说明不了什么,睡眠质量取决于昼夜节律时钟和许多其他因素。 (https://5dlibrary.com/answers/zh/dmn-neural-oscillations/c05/) ## 5DL-B-009 · Crystal Piezoelectricity, Iron-Nickel Meteorites and Photobiomodulation 中文书名:晶体压电性、铁镍陨石与光生物调节 DOI: https://doi.org/10.5281/zenodo.23698498 Cite: Fan, Y. (2026). Crystal Piezoelectricity, Iron-Nickel Meteorites and Photobiomodulation (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23698498 Book page: https://5dlibrary.com/books/crystals-meteorites-photobiomodulation/ Scope: Marked boxes separate published evidence, runnable Python calculations (30, all checked), classical sources and the author's hypotheses. Abstract: People hold quartz for luck, wear tourmaline for its "negative ions", collect fragments of iron meteorites and buy panels of red and near-infrared light for their skin and muscles. This book uses solid-state physics, meteoritics, tissue optics and photobiology to ask what the physics behind each object really is, how large its effects are, and which claims have been tested: the piezoelectricity of α-quartz (class 32, d11 ≈ 2.3 pC/N) and why calcite (class −3m, centrosymmetric) has none, the charge and energy from squeezing a quartz crystal (58 V, under 1 nJ), and the Butterworth–Van Dyke circuit and 10⁴–10⁶ quality factor of a quartz resonator; the pyroelectricity of tourmaline (polar class 3m, about 4 µC m⁻² K⁻¹, falling with iron content), whose surface field approaches the breakdown of air after a few degrees of warming while its charge equals less than a minute of a room's natural ionisation; kamacite and taenite, the iron–nickel phase diagram and the Arrhenius law of nickel diffusion, showing why the Widmanstätten pattern records cooling over millions of years (about 100–6,600 K per million years for group IVA irons such as Gibeon and Muonionalusta) and cannot be made in a laboratory; the optical window of tissue, the effective attenuation coefficient μeff = √(3μa(μa+μs′)), and the share of 660 and 850 nm light that reaches the cortex through scalp and skull (about 1–2 %); the cytochrome c oxidase, nitric oxide and ATP hypothesis, the biphasic dose response, and the guideline-supported use of photobiomodulation to prevent oral mucositis; the gap between the ideal Schumann frequency of 10.6 Hz and the observed 7.83 Hz; and the capacitive-divider model of earthing (about 1.3 V reduced to about 10 mV) beside its unproven anti-inflammatory claims. The eight chapters are read beside Shen Kuo's account of a meteorite fall at Changzhou in 1064, Du Wan's ringing Lingbi stones, the white quartz of the Shennong Bencao Jing and the lodestone that "calls iron" in the Lüshi Chunqiu, and end with four physical checks using a multimeter, a magnet, a loupe and an irradiance meter, and a pre-registered double-blind randomised crossover trial of 660/850 nm light on the forearm against a heat-matched sham (44 completers, with the decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (30, all checked), classical sources and the author's hypotheses. Medical and safety notice: this book is not medical advice and does not claim that any mineral, crystal, meteorite, earthing product or red-light device treats or prevents disease; near-infrared light is invisible and can damage the retina unnoticed, so protective goggles are essential. 摘要:人们握着石英求好运,戴着电气石求「负离子」,收藏铁陨石碎片,又买来红光与近红外灯板照射皮肤和肌肉。本书用固体物理、陨石学、组织光学与光生物学,逐一追问这些东西背后真正的物理、效应的大小,以及哪些说法经过了检验:α-石英(32 点群)的压电性(d11 ≈ 2.3 pC/N)与方解石(−3m,中心对称)为何没有压电性,挤压一块石英得到的电荷与能量(58 V、不到 1 nJ),以及石英谐振器的巴特沃斯–范戴克等效电路与 10⁴–10⁶ 的品质因数;电气石(3m 极性晶类)的热释电性(约 4 µC m⁻² K⁻¹,随含铁量下降),升温几度产生接近空气击穿的表面电场,而其电荷只相当于房间不到一分钟的天然电离量;铁纹石与镍纹石、铁镍相图与镍扩散的阿伦尼乌斯定律,说明魏德曼花纹为何记录了以百万年计的冷却(IVA 群如吉贝恩、穆奥尼奥纳卢斯塔每百万年约 100–6,600 K),实验室无法复制;组织光学窗口、有效衰减系数 μeff = √(3μa(μa+μs′)),以及 660/850 nm 光穿过头皮与颅骨到达皮层的比例(约 1%–2%);细胞色素 c 氧化酶、一氧化氮与 ATP 的机制假说,双相剂量反应,以及 MASCC/ISOO 预防口腔黏膜炎这一有指南支持的用途;舒曼共振的理想频率 10.6 Hz 与实测 7.83 Hz 之差,以及接地的电容分压模型(约 1.3 V 降到约 10 mV)与其尚未证实的抗炎说法。 八章与沈括《梦溪笔谈》所记常州陨星、杜绾《云林石谱》「扣之铿然有声」、《神农本草经》白石英与《吕氏春秋》「慈石召铁」并读;最后给出用万用表、磁铁、放大镜和辐照度计做的四项物理检测,以及一项 660/850 nm 前臂照射对照「等热量假照射」的预注册双盲随机交叉试验(44 人完成,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(30 段,全部核对)、原典与作者的假说。医学与安全声明:本书不是医疗建议,不声称任何矿物、晶体、陨石、接地产品或红光设备能治疗或预防疾病;近红外光不可见,可在不知不觉中损伤视网膜,必须佩戴防护眼镜。 Chapter answers: - Ch.3 Iron Meteorites and the Widmanstätten Pattern — Between 900 °C and 400 °C the diffusion coefficient falls by a factor of a hundred million. At 700 °C, nickel moves a fraction of a micrometre in a day and a few micrometres in a year. Bands a fraction of a millimetre wide need hundreds of thousands to millions of years. (https://5dlibrary.com/answers/en/crystals-meteorites-photobiomodulation/c03/) 第3章 铁陨石与魏德曼花纹 — 从 900 °C 到 400 °C,扩散系数下降一亿倍。在 700 °C 下,镍一天只移动不到一微米,一年也只移动几微米。零点几毫米宽的条带需要数十万到数百万年。 (https://5dlibrary.com/answers/zh/crystals-meteorites-photobiomodulation/c03/) - Ch.4 Light in Tissue: the Optical Window — Two points stand out. In the dermis itself the difference between 660 nm and 850 nm is modest: both have a 1/e depth of about 3 mm, with near-infrared a little deeper, because scattering falls with wavelength. The larger difference is in the epidermis: melanin absorbs much more at 660 nm, so darkly pigmented skin passes roughly half of the red light but three quarters of the near-infrared. In both cases, by 1 cm depth the fluence has fallen to a few per cent of its value near the surface. (https://5dlibrary.com/answers/en/crystals-meteorites-photobiomodulation/c04/) 第4章 光在组织中:光学窗口 — 有两点值得注意。在真皮本身,660 nm 与 850 nm 的差别不大:两者的 1/e 深度都约为 3 mm,近红外略深一些,因为散射随波长增加而减弱。更大的差别在表皮:黑色素对 660 nm 的吸收强得多,深色皮肤大约只让一半的红光通过,却能让四分之三的近红外光通过。无论哪种情况,到 1 cm 深处,光通量率都已降到表面附近的百分之几。 (https://5dlibrary.com/answers/zh/crystals-meteorites-photobiomodulation/c04/) - Ch.6 The Schumann Resonances and Earthing — The ideal formula gives 10.6 Hz for the fundamental, but the observed peak is near 7.8 Hz, and all the modes are lower by about a fifth to a quarter. The difference is physics, not error: the ionosphere is not a perfect conductor, and its finite, height-dependent conductivity lowers the frequencies and broadens the peaks (Nickolaenko 2002). The spacing still follows the $\sqrt{n(n+1)}$ pattern roughly, and the wavelength of the fundamental is close to the Earth's circumference, as the simple picture suggests. (https://5dlibrary.com/answers/en/crystals-meteorites-photobiomodulation/c06/) 第6章 舒曼共振与接地 — 理想公式给出的基频是 10.6 Hz,观测到的峰值却在 7.8 Hz 附近,所有模式都低了约五分之一到四分之一。这个差别是物理,不是误差:电离层不是理想导体,它有限的、随高度变化的电导率使频率降低、峰变宽 (Nickolaenko 2002)。各模式的间隔仍大致符合 $\sqrt{n(n+1)}$ 的规律,基频的波长也接近地球周长,与简单图像一致。 (https://5dlibrary.com/answers/zh/crystals-meteorites-photobiomodulation/c06/) - Ch.8 Practice: Four Physical Checks and a Pre-registered Trial — The protocol plans for the middle case: 44 participants completing both sessions, 55 recruited to allow for drop-out. An eight-person pilot checks the heat matching and the procedures only. (https://5dlibrary.com/answers/en/crystals-meteorites-photobiomodulation/c08/) 第8章 实践:四项物理检测与预注册试验 — 方案按中间情形计划:44 人完成两次,考虑脱落招募 55 人。先做 8 人的预试验,只用于检查热量匹配和操作流程。 (https://5dlibrary.com/answers/zh/crystals-meteorites-photobiomodulation/c08/) ## 5DL-B-010 · The Biomechanics of Phonation and Attentional Shift 中文书名:发声的生物力学与注意力转移 DOI: https://doi.org/10.5281/zenodo.23719625 Cite: Fan, Y. (2026). The Biomechanics of Phonation and Attentional Shift (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23719625 Book page: https://5dlibrary.com/books/vocal-biomechanics-attention/ Scope: Voice health notice: this book is not medical advice; persistent hoarseness needs a doctor. Marked boxes separate published evidence, runnable Python calculations (28, all checked), classical sources and the author's hypotheses. Abstract: Why does a student sing a phrase easily while thinking about the song, then lose pitch, thin the tone or crack the high note the moment they think about their throat? This book draws on voice biomechanics, acoustics, control theory and the psychology of motor learning: the myoelastic–aerodynamic theory, and why Bernoulli suction alone cannot sustain vibration without the convergent–divergent asymmetry of the mucosal wave; Titze's phonation threshold pressure, Pth = k·B·c·ξ0/T, and how damping, cover stiffness and glottal width raise it (with the evidence on hydration and vocal fatigue); source–filter theory and tube formants, and how a narrow epilarynx inside a wide pharynx gathers the third to fifth formants near 3 kHz (Sundberg 1974 and its 1995 revision); the Lombard effect and the pitch-shift reflex, with a delayed negative-feedback model showing why too much correction makes pitch wobble; Wulf's external-focus research and the constrained-action hypothesis, the EMG and singing evidence, and the 2021 meta-analysis beside a bias-corrected reanalysis that finds the average effect close to zero; the Wing–Kristofferson decomposition of timing variance and phase correction; and jitter, shimmer and the harmonics-to-noise ratio computed from synthetic voices, with their limits. The eight chapters are read beside the Record of Music, Shen Kuo's "no words within the sound" and the sayings of Chinese opera teachers, and end with a four-step acoustic assessment in Praat and a pre-registered randomised trial comparing an internal-focus instruction with a tactile anchor plus external auditory-spatial focus on beginners' pitch error in cents and HNR after five minutes (65 completers per group, with the decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (28, all checked), classical sources and the author's hypotheses. Voice health notice: this book is not medical advice; persistent hoarseness needs a doctor. 摘要:为什么学生想着歌曲时唱得轻松,一想喉咙就音准摇晃、音色变薄、高音破掉?本书用嗓音生物力学、声学、控制理论与运动学习心理学回答这个问题:肌弹性–空气动力学理论,以及为什么伯努利吸力本身不能维持振动,关键在于黏膜波造成的收敛–发散不对称;提策的起振阈压公式 Pth = k·B·c·ξ0/T,阻尼、覆盖层刚度和声门宽度如何推高阈压(水合不足与用嗓疲劳的证据);声源–滤波理论与管道共振峰,以及狭窄喉前庭在宽阔咽腔中如何把第三至第五共振峰聚到 3 kHz 附近(松德贝里 1974 与 1995 年的修正);隆巴德效应与变调反馈反射,以及一个带 120 毫秒延迟的负反馈模型,说明纠正增益过高为何使音高摇晃;沃尔夫的外部焦点研究与约束动作假说、肌电与歌唱证据,以及 2021 年元分析与校正发表偏倚后几乎归零的再分析;温–克里斯托弗森计时方差分解与相位校正;用合成嗓音计算基频微扰、振幅微扰与谐噪比及其局限。 八章与《乐记》「凡音之起,由人心生也」、沈括《梦溪笔谈》「声中无字,字中有声」及戏曲口诀「声断气不断」「以字行腔」「意在声先」并读;最后给出用 Praat 做的四步声学评估表,以及一项比较内部焦点指令与「触觉锚定 + 外部听觉空间焦点」对初学者五分钟内音准误差(音分)与谐噪比影响的预注册随机试验(每组 65 人完成,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(28 段,全部核对)、原典与作者的假说。嗓音健康提示:本书不是医疗建议;持续声音嘶哑请就医。 Chapter answers: - Ch.2 Source, Filter and the Singer's Formant — A 17.5 cm tube gives formants at 500, 1500, 2500, 3500 and 4500 Hz, close to an adult male's neutral vowel. Shorter tracts, such as those of women and children, shift every formant up; lowering the larynx lengthens the tract and shifts them down. Vowels are made by changing the tube's shape with the tongue, jaw and lips, which moves the first two formants in particular. (https://5dlibrary.com/answers/en/vocal-biomechanics-attention/c02/) 第2章 声源、滤波与歌手共振峰 — 17.5 cm 的管子给出 500、1500、2500、3500 和 4500 Hz 的共振峰,接近成年男性的中性元音。女性和儿童的声道更短,每个共振峰都向上移;降低喉位使声道变长,共振峰向下移。元音是靠舌、下颌和嘴唇改变管子形状形成的,主要移动的是前两个共振峰。 (https://5dlibrary.com/answers/zh/vocal-biomechanics-attention/c02/) - Ch.8 Practice: An Acoustic Assessment and a Pre-registered Trial — The protocol plans for the more conservative case: 65 completers per group, 145 recruited. The between-person SD of 20 cents is an assumption to be checked in a 10-person pilot, which is used only for that purpose and for checking the procedure. (https://5dlibrary.com/answers/en/vocal-biomechanics-attention/c08/) 第8章 实践:声学评估与预注册试验 — 方案按较保守的情形计划:每组 65 人完成,共招募 145 人。个体间标准差 20 音分是一个假设,将在 10 人预试验中核对;预试验只用于这一目的和检查操作流程。 (https://5dlibrary.com/answers/zh/vocal-biomechanics-attention/c08/) ## 5DL-B-011 · The Algorithmic Mirror 中文书名:算法之镜 DOI: https://doi.org/10.5281/zenodo.23723557 Cite: Fan, Y. (2026). The Algorithmic Mirror (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23723557 Book page: https://5dlibrary.com/books/algorithmic-mirror/ Scope: Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses. Abstract: Why do two people who ask the same language model about the same subject come away with opposite impressions, one finding it shallow and the other a collaborator that opens new lines of thought? This book treats a language model as a learned conditional distribution P(Y|X) and asks, with information theory, machine learning, cybernetics and cognitive psychology, what the asker brings: conditional entropy and mutual information, and why a constraining prompt redirects probability rather than necessarily lowering entropy (correcting the claim that models "collapse to the mean", with the evidence that fine-tuning reduces output diversity); the geometry of scaled dot-product self-attention, with a three-cluster toy model of what one-field and cross-field queries draw on and what the √d_k scaling does, read beside the Record of Learning's bell that answers as it is struck; Ashby's law of requisite variety and the good-regulator theorem, with a model of how the capacity to verify bounds the errors that get accepted; the Barnum effect, biased assimilation and Bayesian updating, with projection shown to work on enthusiasts and sceptics alike; model collapse after Shumailov and colleagues, and how accumulating real data prevents it; and the actual evidence on human–AI "centaurs": freestyle chess, Dell'Acqua and colleagues' jagged frontier, and the meta-analysis by Vaccaro and colleagues finding that human–AI combinations on average do worse than the better of the two alone. The eight chapters are read beside Zhuangzi's mirror that does not store, the Huayan net of Indra and the seeds and perfuming of the Cheng weishi lun and the Zongjing lu, with the limits of these analogies stated; the book ends with a four-step prompt audit and a pre-registered, double-blind randomised trial comparing conventional single-field prompts with cross-field structured prompts (60 research questions, expert-rated novelty and rigour as co-primary outcomes, with the decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses. The book was itself produced by a human-architect, AI-drafting and human-review workflow, which the preface describes. 摘要:为什么两个人问同一个语言模型同一个题目,一个觉得它浅薄敷衍,另一个却觉得它能打开新的思路?本书把语言模型看作一个学得的条件分布 P(Y|X),用信息论、机器学习、控制论和认知心理学考察「提问的人带来什么」:条件熵与互信息,以及为什么约束性的提示是改变概率的方向,而不一定降低熵(对「坍缩到平均值」说法的修正,以及微调降低输出多样性的证据);缩放点积自注意力的几何,一个三簇玩具模型显示单一领域的查询与跨领域查询分别取用什么,以及 √d_k 缩放的作用,并与《学记》「善待问者如撞钟」并读;阿什比必要多样性定律与「好的调节者」定理,以及一个说明核查能力如何限制可被接受的错误的模型;巴纳姆效应、偏向同化与贝叶斯更新,以及投射对热衷者和怀疑者同样成立;舒迈洛夫等人的模型崩溃,以及累积真实数据如何避免崩溃;半人马式人机协作的真实证据:自由式国际象棋、德拉夸等人的「锯齿状前沿」,以及瓦卡罗等人发现人机组合平均不如两者中较好一方的元分析。 八章与《庄子·应帝王》的不藏之镜、华严宗的因陀罗网、《成唯识论》的种子与熏习及《宗镜录》并读,并说明这些类比的限度;最后给出一份四步提示审计,以及一项比较常规单一领域提示与跨领域结构化提示的预注册双盲随机试验(60 个研究问题,专家盲评新颖性与严谨性两个共同主要结局,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(14 段程序,全部核对)、原典与作者的假说。本书本身即由「人类架构 + AI 起草 + 人类审核」的流程写成,并在序言中说明。 Chapter answers: - Ch.8 Practice: A Prompt Audit and a Pre-registered Trial — The protocol plans for the middle case, 53 questions, and samples 60 to allow for questions dropped because a model refuses or a rater cannot assess them. The standard deviation of 1.0 rating point for within-question differences is an assumption to be checked in a 6-question pilot. (https://5dlibrary.com/answers/en/algorithmic-mirror/c08/) 第8章 实践:提示审计与预注册试验 — 方案按中间情形计划,需要 53 个问题;抽取 60 个,以容纳因模型拒答或评审无法评估而剔除的问题。问题内差值的标准差 1.0 分是一个假设,将在 6 个问题的预试验中核对。 (https://5dlibrary.com/answers/zh/algorithmic-mirror/c08/) ## 5DL-B-012 · Ecological Cultivation, Stress-Induced Polysaccharide Synthesis and Fresh-Extract Rheology of Dendrobium officinale 中文书名:铁皮石斛的生态栽培、逆境多糖合成与鲜汁流变学 DOI: https://doi.org/10.5281/zenodo.23729740 Cite: Fan, Y. (2026). Ecological Cultivation, Stress-Induced Polysaccharide Synthesis and Fresh-Extract Rheology of Dendrobium officinale (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23729740 Book page: https://5dlibrary.com/books/dendrobium-cultivation-thermodynamics/ Scope: Health notice: this book is not medical advice and does not claim that D. Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses. Abstract: The quality of Dendrobium officinale is judged chiefly by the acetylated glucomannan in its stems, and the grower's decisions about water, fertiliser, night temperature and disease control shape both yield and quality. This book examines those decisions with plant physiology and physics: the evidence that D. officinale is a C3–CAM intermediate that fixes part of its carbon at night; vapour-pressure deficit and dew point from the Tetens formula, and a leaf energy balance showing how night-time air movement prevents dew and lowers disease risk without sprays; the growth–differentiation balance hypothesis and a carbon budget showing that low nitrogen, drying cycles and cool nights raise the polysaccharide share but can lower yield and total polysaccharide, beside surveys finding less polysaccharide in tree-grown "wild-like" stems; orchid mycorrhiza, the water retention and oxygen diffusion of bark and gravel media, and the effect of fertiliser salts on root-zone water potential; the Ostwald–de Waele power law and the kinetics of polysaccharide release; the evidence linking core temperature and distal skin warming to sleep, with a two-compartment model showing that the heat in a warm drink is too small to change a regulated core temperature; and the Hagen–Poiseuille law with the evidence on sieve-pore plugging and heat treatment in citrus greening. The eight chapters are read beside the Shennong Bencao Jing, Kou Zongshi's Bencao Yanyi and Zhao Xuemin's Bencao Gangmu Shiyi, noting claims such as the "nine immortal herbs" that have no traceable source, and end with a four-step weekly check for growers (VPD wet hours, drainage EC, day–night difference, polysaccharide) and a pre-registered double-blind crossover trial comparing a fresh-stem decoction with a xanthan placebo matched for temperature and viscosity, with slow-wave sleep by polysomnography as the primary outcome and the distal–proximal skin temperature gradient as a secondary outcome (36 completers, with the decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses. Health notice: this book is not medical advice and does not claim that D. officinale treats insomnia or any disease. 摘要:铁皮石斛的品质主要看茎中乙酰化葡甘露聚糖的含量,而浇水、施肥、夜温和防病的决定同时影响产量与品质。本书用植物生理学和物理学考察这些决定:铁皮石斛作为 C3–CAM 中间型植物在夜间固碳的证据;用泰滕斯公式计算饱和水汽压差(VPD)与露点,并用叶片能量平衡说明夜间气流如何防止结露、在不喷药的情况下降低病害风险;生长–分化平衡假说与一个碳收支模型,显示低氮、干湿循环和凉爽夜晚提高多糖占比、但可能降低产量与每丛多糖总量,以及「仿野生附树栽培多糖更低」等调查结果;兰科菌根、树皮与砾石基质的持水曲线和氧扩散,以及肥料盐分对根区水势的影响;奥斯特瓦尔德–德瓦勒幂律与多糖溶出动力学;核心体温、远端皮肤升温与睡眠的证据,以及一个双室体温模型,说明一杯热饮的热量太小、不足以改变受调节的核心体温;哈根–泊肃叶定律与柑橘黄龙病的筛孔堵塞和热处理证据。 八章与《神农本草经》、寇宗奭《本草衍义》和赵学敏《本草纲目拾遗》并读,并指出「九大仙草」等说法找不到出处;最后给出种植者每周四步核查(VPD 湿润小时数、排出液 EC、昼夜温差、多糖),以及一项比较铁皮石斛鲜条煎煮液与温度、黏度匹配的黄原胶安慰剂的预注册双盲交叉试验(主要结局为多导睡眠图慢波睡眠,次要结局为远端–近端皮肤温差;36 人完成,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(14 段程序,全部核对)、原典与作者的假说。健康提示:本书不是医疗建议,不声称铁皮石斛能治疗失眠或任何疾病。 Chapter answers: - Ch.1 The Physics of a Cliff Habitat — Relative humidity on its own is a poor guide. Air at 97 % RH and 14 °C has a VPD of only 0.05 kPa and a dew point just half a degree below the air. Any surface that falls half a degree below the air will become wet. (https://5dlibrary.com/answers/en/dendrobium-cultivation-thermodynamics/c01/) 第1章 崖壁生境的物理学 — 只看相对湿度并不可靠。14 °C、相对湿度 97% 的空气,VPD 只有 0.05 kPa,露点只比气温低半度。任何比气温低半度的表面都会变湿。 (https://5dlibrary.com/answers/zh/dendrobium-cultivation-thermodynamics/c01/) - Ch.4 From Stem to Cup — Under these assumptions a cup contains a few tenths of a gram of polysaccharide, at a concentration of about 0.3–0.5 % if all of it dissolves. Heat speeds release several-fold: at 100 °C most comes out within 20 minutes, at 60 °C barely a third. Crushing or pressing the stem first helps, because it breaks the tissue and shortens the distance the chains must travel. (https://5dlibrary.com/answers/en/dendrobium-cultivation-thermodynamics/c04/) 第4章 从鲜条到杯中 — 在这些假设下,一杯中含有零点几克多糖,若全部溶出,浓度约为 0.3%–0.5%。加热使溶出加快数倍:100 °C 时 20 分钟内大部分溶出,60 °C 时只有约三分之一。先把茎压碎或榨汁也有帮助,因为这会破坏组织,缩短长链需要移动的距离。 (https://5dlibrary.com/answers/zh/dendrobium-cultivation-thermodynamics/c04/) - Ch.8 Practice and a Pre-registered Trial — The protocol plans for 36 completers. If the within-person SD of the difference is 20 minutes, that gives about 83 % power to detect a 10-minute gain; 90 % power would need 44. It recruits 44 to allow for drop-out after the adaptation night. The SD is an assumption to be checked in a pilot of 10 people, and if the pilot shows it to be 25 minutes or more, the sample size will be recalculated before the main trial begins. (https://5dlibrary.com/answers/en/dendrobium-cultivation-thermodynamics/c08/) 第8章 实践与预注册试验 — 方案计划 36 人完成。如果个体内差值的标准差为 20 分钟,这一人数检出 10 分钟增益的把握度约为 83%;要达到 90% 需要 44 人。考虑到适应夜后的脱落,招募 44 人。标准差是假设,将在 10 人的预试验中核对;如果预试验显示标准差达到 25 分钟或以上,正式试验开始前重新计算样本量。 (https://5dlibrary.com/answers/zh/dendrobium-cultivation-thermodynamics/c08/) ## 5DL-B-013 · Classical and Statistical Mechanics of Human Action 中文书名:行动的经典与统计力学 DOI: https://doi.org/10.5281/zenodo.23738774 Cite: Fan, Y. (2026). Classical and Statistical Mechanics of Human Action (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23738774 Book page: https://5dlibrary.com/books/mechanics-of-action/ Scope: This book is not psychotherapy or medical advice and makes no diagnostic or therapeutic claim. Marked boxes separate published evidence, runnable Python calculations (15 programs, all run and checked), classical sources and the author's hypotheses. Abstract: Why is starting harder than continuing? Why do short bursts rarely bring lasting change? Why does giving up halfway slide us back? This book uses the equations of classical and statistical mechanics to think about these questions, always keeping the conclusions of physics itself apart from their use on people as cross-disciplinary analogies or simplified models. The chapters examine in turn: the evidence on status quo bias and defaults, with a damped restoring model (only a sustained force moves the resting point) and a switching-cost model; Coulomb static and kinetic friction, stick-slip motion, contact aging and wheel traction — showing, after checking, that moving forward needs grip, not resistance, and that the Coulomb model has no "moderate" optimum of friction; the overdamped Langevin equation and Kramers' escape rate in a double well, where simulation shows that lowering the barrier speeds both leaving and relapse, while a steady outside pull shortens leaving and lengthens relapse but fails as soon as it is removed; the saddle-node bifurcation, hysteresis and critical slowing down of the cusp model; the evidence on learning by teaching (effect sizes g of about 0.35–0.56) and network models of complex contagion, with a proof that "tutor two classmates until they pass before unlocking" must jam a closed class, that only a version counted by tutoring sessions is workable, and that a closed class cannot spread exponentially; and the near-disappearance of ego depletion in large pre-registered replications, with explanations based on motivational reallocation and opportunity cost. The eight chapters are read beside the Book of Changes ("Heaven moves vigorously"), Wang Yangming's Instructions for Practical Living ("a person must be polished in affairs") and the Blue Cliff Record case of "the one who has died the great death coming back to life", with the limits of each analogy, and end with a four-step barrier audit (start-up friction, barrier height, outside ratchet, feedback delay) and a pre-registered randomised controlled trial comparing solo study with a rule requiring two peer-tutoring sessions to unlock each stage (30-day completion and a blind-marked test score as co-primary outcomes, 469 per arm, with the decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (15 programs, all run and checked), classical sources and the author's hypotheses. This book is not psychotherapy or medical advice and makes no diagnostic or therapeutic claim. 摘要:为什么开始比继续难?为什么短促的爆发很少带来持久的改变?为什么半途而废会滑回原处?本书借用经典力学与统计力学的方程思考这些问题,并始终把物理本身的结论与把它们用于人时的跨学科类比或简化模型分开。各章依次考察:维持原状偏差与默认选项的证据,以及一个阻尼回复模型(只有持续的力才移动平衡点)和一个切换成本模型;库仑静摩擦与动摩擦、粘滑运动与接触老化,以及车轮牵引——核对后指出前进需要的是「抓地力」而不是「阻力」,库仑模型中也不存在「适度」的最佳摩擦;双势阱中的过阻尼朗之万方程与克莱默斯逃逸率,模拟显示降低势垒会同时加速出发与复发,而持续外力能缩短出发、延长复发,但撤掉即失效;尖点模型中的鞍结分岔、滞后与临界慢化;为教而学的证据(效应量 g 约 0.35–0.56)与复杂传染的网络模型,并证明「辅导两名同伴达标才能解锁」在封闭班级里必然阻塞,只有以辅导时段计数的版本可行,且封闭班级中不存在指数级传播;自我损耗效应在大规模预注册重复中几乎消失,以及动机重分配与机会成本的解释。 八章与《周易》「天行健」、王阳明《传习录》「人须在事上磨」和《碧岩录》「大死底人却活」公案并读,并说明类比的限度;最后给出一份四步势垒核查表(启动摩擦、势垒高度、外部棘轮、反馈延迟),以及一项比较独自学习与「完成两次同伴辅导才能解锁」的预注册随机对照试验方案(30 天完课率与盲评测验成绩为共同主要结局,每组 469 人,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部运行核对)、原典与作者的假说。本书不是心理治疗或医疗建议,不作任何诊断或疗效声称。 Chapter answers: - Ch.3 The Double Well and Kramers' Escape Rate — First, with these parameters, halving the barrier shortens the waiting time about 127-fold, and halving it while adding a small steady pull shortens it about 657-fold — two to three orders of magnitude. The original idea that interventions can shorten the escape time "by several orders of magnitude" holds in the model, but only when $\Delta U/D$ is large; in a person we cannot measure $\Delta U/D$, so "orders of magnitude" is a statement about the model, not a prediction for anyone. (https://5dlibrary.com/answers/en/mechanics-of-action/c03/) 第3章 双势阱与克莱默斯逃逸率 — 第一,在这个参数下,把势垒减半使等待时间缩短约 127 倍,减半加上一个小的持续拉力缩短约 657 倍,也就是两到三个数量级。原先的设想说干预能使逃逸时间「缩短数个数量级」,在模型中是成立的,但前提是 $\Delta U/D$ 很大;在人身上,我们无法测量 $\Delta U/D$,所以「几个数量级」只是模型的说法,不是对任何人的预测。 (https://5dlibrary.com/answers/zh/mechanics-of-action/c03/) - Ch.4 Hysteresis and Tipping Points — The second part corresponds to "giving up halfway". If, when the force is removed, the state has not yet passed the ridge (here $x = 0$), the system slides back to the old state; if it has, the system slides into the new state by itself. The shortest time needed depends strongly on the size of the force: with a force of 1.00, 1.4 time units suffice; with 0.40, just above the fold, it takes more than 17. This is critical slowing down near the fold: a lingering "bottleneck" keeps the system dawdling near where the old state used to be for a long time before it finally gets through (Strogatz 2015). (https://5dlibrary.com/answers/en/mechanics-of-action/c04/) 第4章 滞后与临界点 — 第二部分对应「半途而废」。外力撤掉时,如果状态还没越过山脊(这里是 $x = 0$),系统就会滑回旧状态;越过了,系统就会自己滑进新状态。所需的最短时间强烈依赖于外力的大小:外力为 1.00 时只需 1.4 个时间单位;外力为 0.40、只比折点高一点时,需要 17 个时间单位以上。这是折点附近的临界慢化:残留的「瓶颈」使系统在旧状态的位置附近徘徊很久,才最终通过 (Strogatz 2015)。 (https://5dlibrary.com/answers/zh/mechanics-of-action/c04/) - Ch.6 The Cognitive Energy Budget and the Ego-Depletion Debate — In this simulation the average effect over all studies correctly sits near 0.05, but only about 4 % of studies reach significance, and those published studies show an average effect of 0.79 — higher even than the 2010 meta-analytic estimate. Small samples plus selective publication are enough to create a "robust" literature out of an effect that barely exists. This does not prove that such selection happened in ego-depletion research, but it shows why large pre-registered replications are more credible than the pooled results of many small studies. (https://5dlibrary.com/answers/en/mechanics-of-action/c06/) 第6章 认知能量预算与自我损耗之争 — 在这个模拟中,所有研究的平均效应正确地落在 0.05 附近,但只有约 4% 的研究达到显著;而这些被发表的研究,平均效应量高达 0.79,比 2010 年元分析的估计还高。小样本加上选择性发表,就足以从一个几乎不存在的效应中制造出一篇「稳健」的文献。这不证明自我损耗研究中一定发生了这样的选择,但说明大规模预注册重复为什么比许多小研究的汇总更可信。 (https://5dlibrary.com/answers/zh/mechanics-of-action/c06/) - Ch.8 Practice and a Pre-Registered Trial — Each co-primary outcome is tested at two-sided α = 0.025, so that the overall type I error rate stays at or below 0.05. To detect a rise in completion from 40 % to 52 % with 90 % power needs 426 per arm; learners in arm B influence each other within groups of three, and allowing for a design effect at an intraclass correlation of 0.05 gives 469 per arm. The test score, at a difference of 0.30 standard deviations, needs fewer. The protocol plans 469 randomised per arm, 938 in all. The 40 % baseline completion and the intraclass correlation are assumptions to be checked in an internal pilot of the first 100 participants; if baseline completion is clearly different, the sample size will be recalculated before the main trial. (https://5dlibrary.com/answers/en/mechanics-of-action/c08/) 第8章 实践与预注册试验 — 两个共同主要结局各以双侧 α = 0.025 检验,使总的一类错误率不超过 0.05。按完课率从 40% 提高到 52%、把握度 90% 计算,每组需要 426 人;B 组学员在三人辅导小组中互相影响,按组内相关 0.05 计入设计效应后为每组 469 人。测验成绩按 0.30 个标准差的差异计算需要的人数更少。方案计划每组随机分配 469 人,共 938 人。40% 的基线完课率与组内相关都是假设,将在前 100 名受试者的内部预试验中核对;如果基线完课率明显不同,正式试验前重新计算样本量。 (https://5dlibrary.com/answers/zh/mechanics-of-action/c08/) ## 5DL-B-014 · Dynamical Systems of Neurotransmitters and Endocrinology 中文书名:神经递质与内分泌的动力系统 DOI: https://doi.org/10.5281/zenodo.23765340 Cite: Fan, Y. (2026). Dynamical Systems of Neurotransmitters and Endocrinology (Version 1.0.1). 5D Library. https://doi.org/10.5281/zenodo.23765340 Book page: https://5dlibrary.com/books/neurochemistry-game-theory/ Scope: Health notice: this book is not medical advice, makes no diagnosis, gives no doses of any drug, hormone or supplement, and does not claim that any practice treats any disease. Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses. Abstract: Popular accounts often treat dopamine, cortisol, testosterone and serotonin as switches to be turned up or down. This book puts them back into their feedback loops and examines them with dynamical-systems, reinforcement-learning and optimisation models: the reward prediction error and temporal-difference learning of Schultz, Dayan and Montague, and how the average-reward model of Niv and colleagues explains the negative error of ordinary effort after frequent intense stimulation, read beside Solomon and Corbit's opponent-process theory; a model of the CRH–ACTH–cortisol feedback loop with glucocorticoid-receptor down-regulation (bistability and a tipping point in stress frequency), how allostatic load is scored, and the meta-analytic evidence that chronic stress in humans can show up as either raised or lowered cortisol; Wingfield's challenge hypothesis, the measured size of testosterone effects (Eisenegger's ultimatum game, the winner–loser effect, the failed replication of power posing and the marginal support for the dual-hormone hypothesis), HPA inhibition of the HPG axis, and challenge versus threat appraisal; tryptophan-depletion studies, the umbrella review of serotonin and depression, and a light-phase model of ipRGCs and the circadian clock; the resonance of the baroreflex loop at about six breaths a minute and what it implies for measurement; and the absence of a single optimal state in the sense of a Pareto front. The eight chapters are read beside Mencius on nourishing the flood-like qi, the Suwen on tranquil emptiness and the Zhuangzi's wooden cockerel, with the limits of the analogies, and end with a four-step check of personal daily rhythm (morning light, blocking high-frequency fragments, slow breathing and recovery, complete and regular sleep) and a pre-registered randomised controlled trial comparing morning daylight plus 0.1 Hz slow breathing with usual routine in high-stress founders, with resting heart-rate variability (lnRMSSD) as primary outcome and the cortisol awakening response as secondary outcome (decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses. Health notice: this book is not medical advice, makes no diagnosis, gives no doses of any drug, hormone or supplement, and does not claim that any practice treats any disease. 摘要:流行说法常把多巴胺、皮质醇、睾酮和血清素当作可以分别调高调低的开关。本书把它们放回所在的反馈回路,用动力系统、强化学习与优化模型考察:舒尔茨、达扬与蒙塔古的奖励预测误差与时间差分学习,以及尼夫等人的平均奖励率模型如何解释高频强刺激之后日常努力的负向误差,并与所罗门和科比特的对手过程理论并读;CRH–ACTH–皮质醇负反馈回路与糖皮质激素受体下调的模型(双稳态与压力频率的转折点)、异态负荷指数的计算方法,以及慢性压力在人类中既可表现为皮质醇升高也可表现为降低的元分析证据;温菲尔德的挑战假说、睾酮效应量的实证大小(艾森内格的最后通牒博弈、胜负效应与力量姿势的重复失败、双激素假说的边缘支持)、HPA 对 HPG 轴的交叉抑制与挑战—威胁评估;色氨酸耗竭研究、血清素与抑郁关系的伞形综述、ipRGC 与昼夜时钟的光相位模型;压力反射回路在每分钟约 6 次呼吸时的共振及其对测量指标的含义;以及帕累托前沿意义上不存在单一最优状态。 八章与《孟子·公孙丑上》养浩然之气、《素问·上古天真论》恬惔虚无和《庄子·达生》木鸡并读,并说明类比的限度;最后给出个人日常节律四步核查(晨间光照、阻断高频碎片刺激、慢呼吸与恢复、睡眠完整与规律),以及一项比较晨间自然光加 0.1 赫兹慢呼吸与常规作息对高压创业者静息心率变异性(lnRMSSD,主要结局)与晨起皮质醇反应(次要结局)影响的预注册随机对照试验方案(判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(14 段程序,全部核对)、原典与作者的假说。健康提示:本书不是医疗建议,不作诊断,不给任何药物、激素或补充剂剂量,不声称任何练习能治疗任何疾病。 Chapter answers: - Ch.3 The HPG Axis, Testosterone and the Challenge Hypothesis — The program above has already shown that $r = -0.061$ explains about 0.4 % of the variance, and that detecting it once with 80 % power takes about 2,100 people. Most single studies are far smaller. (https://5dlibrary.com/answers/en/neurochemistry-game-theory/c03/) 第3章 HPG 轴、睾酮与挑战假说 — 上面的程序已经算出:$r = -0.061$ 只解释约 0.4% 的变异,要以 80% 的检验力检出它一次,需要约两千一百人。大多数单个研究的样本远小于此。 (https://5dlibrary.com/answers/zh/neurochemistry-game-theory/c03/) - Ch.5 The Vagus Nerve, Heart-Rate Variability and Breathing Resonance — When the total delay of the loop equals half a breathing cycle, the correction arrives just in step with the next push. In this model, breathing at six a minute makes the heart-rate swing about twice that of breathing alone; at the usual 12 to 15 breaths a minute the correction partly cancels the push, and the swing is smaller than the push alone. (https://5dlibrary.com/answers/en/neurochemistry-game-theory/c05/) 第5章 迷走神经、心率变异性与呼吸谐振 — 当回路的总延迟恰好等于呼吸周期的一半时,修正到达的时刻正好与下一次推动方向一致。在这个模型里,以每分钟 6 次呼吸,心率波动约为呼吸单独作用的两倍;以每分钟 12 至 15 次的常见频率呼吸,修正部分抵消了推动,波动反而小于单独推动。 (https://5dlibrary.com/answers/zh/neurochemistry-game-theory/c05/) - Ch.6 Trade-offs among Four State Signals: No Single Optimal Level — The model's conclusion does not depend on the particular parameters: as long as fatigue accumulates and rest lets it recover, there is no single optimal schedule, only a set of schedules that are best under some trade-off. Which to choose depends on what you value more. With these parameters, working without any break gives neither the highest output nor the lowest fatigue; but specific conclusions such as "25 minutes of work and 10 of rest is best" come entirely from assumed parameters and should not be taken as advice. (https://5dlibrary.com/answers/en/neurochemistry-game-theory/c06/) 第6章 四种状态信号的权衡:没有单一的最优水平 — 模型的结论不依赖具体参数:只要疲劳累积而休息能使它恢复,就不存在单一的最优日程,只有一组「在某种取舍下最好」的日程。选择哪一个,取决于你更看重什么。在这组参数下,完全不休息的日程既不是产出最高的,疲劳也最重;但「25 分钟工作、10 分钟休息」最好这样具体的结论,完全来自假设的参数,不应被当作建议。 (https://5dlibrary.com/answers/zh/neurochemistry-game-theory/c06/) - Ch.8 Practice and a Pre-registered Trial — If the between-person SD of lnRMSSD is 0.5 and the correlation between baseline and week 8 is 0.7, detecting a difference of 0.15 (about 16 % in RMSSD) needs 90 completers per group for 80 % power or 121 for 90 %. The protocol plans 121 completers per group for 90 % power and, allowing for 20 % loss, recruits 303 people in all. The SD and correlation are assumptions and will be checked in a 20-person pilot. (https://5dlibrary.com/answers/en/neurochemistry-game-theory/c08/) 第8章 实践与预注册试验 — 如果个人之间 lnRMSSD 的标准差为 0.5、基线与第 8 周的相关为 0.7,检出 0.15(RMSSD 约 16%)的差异需要每组 90 人完成(80% 检验力)或 121 人(90%)。方案按 90% 检验力计划每组 121 人完成,考虑 20% 脱落,共招募 303 人。标准差和相关系数是假设,将在 20 人的预试验中核对。 (https://5dlibrary.com/answers/zh/neurochemistry-game-theory/c08/) ## 5DL-B-015 · The Sandbox, Determinism and Active Inference 中文书名:沙盘、宿命与主动推理 DOI: https://doi.org/10.5281/zenodo.23738794 Cite: Fan, Y. (2026). The Sandbox, Determinism and Active Inference (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23738794 Book page: https://5dlibrary.com/books/active-inference-sandbox/ Scope: Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. This book is not psychotherapy or diagnosis; its account of old priors driving repeated setbacks is a hypothesis and a cross-disciplinary analogy and never implies that victims are responsible for what happened to them. Abstract: Some people feel their life follows a script: the surroundings change, and the same conflict and disappointment play out again. This book examines the experience with predictive processing and active inference from computational neuroscience. The brain predicts its sensory input from prior beliefs and works to reduce prediction error, which it can do in two ways: by revising the belief, or by acting so that the world fits the prediction. The book covers precision-weighted Bayesian updating and behavioural confirmation (Merton, Snyder and colleagues, and the smaller, disputed evidence on teacher-expectation effects); variational and expected free energy, and how a simplified active-inference agent with a pessimistic outcome prior keeps choosing the same trap, and escapes it through contrary experience accumulated under low stakes; the statistical meaning of Markov blankets, the criticism of their use in the free-energy principle, and how changing the source of input decouples the internal model from the world and, through action, pushes the environment away; greedy, epsilon-greedy, UCB1 and Thompson sampling in multi-armed bandits, simulated annealing and information gain; evidence on stress and over-exploitation, psychological safety and serious games, with a temperature-parameter model; and the Lyapunov exponent and predictability horizon of the Lorenz system, with compatibilism. The eight chapters are read beside Yuan Huang's Four Lessons of Liaofan, the seeds and perfuming of the Cheng Weishi Lun and the Shurangama Sutra's "if one can turn things", noting the limits of such a reading, and end with a four-step Bayesian audit and a pre-registered randomised controlled trial comparing a structured sandbox and card-simulation workshop with a time-matched reading-and-lecture course (primary outcome the directed-exploration parameter of the Horizon task; assessors and analysts blind; analysis clustered by group; 122 completers per arm; decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. This book is not psychotherapy or diagnosis; its account of old priors driving repeated setbacks is a hypothesis and a cross-disciplinary analogy and never implies that victims are responsible for what happened to them. 摘要:有些人觉得自己的人生像照着剧本在演:换了环境,同样的冲突与失望一再重演。本书用计算神经科学的预测加工与主动推理理论考察这一现象。大脑用先验信念预测感官输入,并设法减小预测误差;减小误差有两条路:修改信念,或者采取行动让世界符合预测。本书依次讨论:精度加权的贝叶斯更新与行为确认(默顿、斯奈德等人的研究,以及教师期望效应较小且有争议的证据);变分自由能与期望自由能,以及一个简化的主动推理智能体如何因悲观的结果先验而反复选择同一个陷阱、又如何通过低风险条件下累积的相反经验脱离陷阱;马尔可夫毯的统计含义、它在自由能原理中的用法所受的批评,以及输入来源的改变如何使内部模型与世界脱钩并通过行动推开环境;多臂老虎机中的贪婪、ε-贪婪、UCB1 与汤普森采样,模拟退火与信息增益;压力与过度利用、心理安全和严肃游戏的证据,以及一个温度参数模型;洛伦茨系统的李雅普诺夫指数与可预测期限,以及相容论。 八章与袁黄《了凡四训·立命之学》、《成唯识论》的种子与熏习、《楞严经》「若能转物」并读,并指出并读的限度;最后给出一份四步贝叶斯审计表,以及一项比较结构化沙盘与牌卡推演工作坊和时间匹配的阅读讲授课程的预注册随机对照试验方案(主要结局为视野任务中的定向探索参数,评估者与分析者盲,按小班聚类,每组 122 人完成,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部核对)、原典与作者的假说。本书不是心理治疗或诊断;关于「旧先验驱动重复挫折」的内容是假说与跨学科类比,绝不意味着受害者要为自己的遭遇负责。 Chapter answers: - Ch.1 The Bayesian Brain and the Self-Fulfilling Prophecy — The result is more interesting than intuition suggests. As long as the other person's disposition stays fixed ($\rho = 0$), the expectation eventually returns to the truth, even when nine-tenths of the response depends on treatment; it just takes much longer, 114 meetings instead of 11. Behavioural confirmation on its own only delays correction. For a false expectation to come true, a further condition is needed: the other person must actually change, as when a child who is always treated coldly gradually becomes withdrawn ($\rho > 0$). Then both settle at a compromise that lies closer to the original false expectation. (https://5dlibrary.com/answers/en/active-inference-sandbox/c01/) 第1章 贝叶斯脑与自我实现的预言 — 结果比直觉更有意思。只要对方的本性不变($\rho = 0$),哪怕对方的回应九成取决于被对待的方式,期望最终仍会回到真相,只是慢得多:从 11 次相处延长到 114 次。行为确认本身只会延缓纠错。要让一个错误的预期真正「成真」,需要另一个条件:对方本身被改变了,例如一个总被冷淡对待的孩子渐渐真的变得退缩($\rho > 0$)。这时两人最终在一个折中的位置稳定下来,而这个位置离最初的错误预期更近。 (https://5dlibrary.com/answers/zh/active-inference-sandbox/c01/) - Ch.2 Two Routes: Change the Mind, or Change the World — Take the first part first. When the agent's outcome prior is "rejection" (acceptance only 0.2), the risk of opening is large (0.701), because opening would bring too much acceptance that it "does not expect"; the risk of guarding is almost zero, because guarding brings exactly the rejection it expects. Add the fact that guarding's outcome is more predictable (lower ambiguity), and it opens only 3 % of the time. In the same world, an agent whose outcome prior is 0.8 opens 97 % of the time. (https://5dlibrary.com/answers/en/active-inference-sandbox/c02/) 第2章 两条路径:改变认知,还是改变世界 — 先看第一部分。当智能体的结果先验是「被拒绝」(接纳只占 0.2)时,敞开的风险项很大(0.701),因为敞开会带来太多它「没有预期到」的接纳;设防的风险几乎为零,因为设防恰好带来它预期的拒绝。再加上设防的结果更可预测(模糊性更低),它选择敞开的概率只有 3%。同一个世界里,结果先验是 0.8 的智能体则 97% 选择敞开。 (https://5dlibrary.com/answers/zh/active-inference-sandbox/c02/) - Ch.3 Markov Blankets and the Boundary of the Individual — The internal and external states are correlated at 0.228: the inside "knows" something about the outside. Controlling for the current sensory state alone is not enough (partial correlation $-0.045$), because the internal state's past carries information about the external state's past; once the internal state's own past is counted as part of the blanket, the partial correlation falls to $-0.001$. The same holds on the other side: the internal state and the next external state are correlated at $-0.174$ (action pushes back on the world), and almost zero once action and the current external state are controlled for. Two things follow: in this model the blanket does separate inside from outside, and in a dynamic system the "blanket" must include time; it cannot be read off the sensory input of a single instant. (https://5dlibrary.com/answers/en/active-inference-sandbox/c03/) 第3章 马尔可夫毯与个体的边界 — 内部状态与外部状态原本有 0.228 的相关:内部「知道」外部的一些事情。只控制当前的感官状态还不够(偏相关 $-0.045$),因为内部状态的过去携带着过去外部状态的信息;把内部状态自身的过去也算进毯里,偏相关降到 $-0.001$。另一侧也一样:内部状态与下一时刻的外部状态相关 $-0.174$(行动在推回世界),控制了行动和当前外部状态之后几乎为零。这说明两件事:在这个模型里,毯确实把内外隔开;而在动态系统中,「毯」必须包括时间,不能只看某一瞬间的感官输入。 (https://5dlibrary.com/answers/zh/active-inference-sandbox/c03/) - Ch.4 Escaping the Local Optimum: Exploration and Exploitation — Hill climbing nearly always stops on the hill nearest its start (it reaches the global peak in only 2 % of starts, mostly ones that began nearby). Simulated annealing raises the mean height reached considerably (from 0.99 to 2.16), and the share of starts that reach the global peak rises with the number of steps, from 8 % to 20 %. But it is far from a guarantee: in theory annealing finds the global optimum with probability one only if the temperature falls logarithmically, extremely slowly (Hajek 1988), and in practice nobody can wait that long. The result is an honest reminder: exploration raises the chance of finding a better position but cannot guarantee finding the best one. (https://5dlibrary.com/answers/en/active-inference-sandbox/c04/) 第4章 跳出局部最优:探索与利用 — 爬山法几乎总停在离起点最近的小丘上(找到全局最高点的只有 2%,多是本来就从附近出发的)。模拟退火显著改善了平均到达的高度(从 0.99 到 2.16),找到全局最高点的比例也随步数增加,从 8% 升到 20%。但它远不是保证:理论上,只有当温度以对数速度极其缓慢地下降时,退火才以概率 1 找到全局最优 (Hajek 1988),实际中没人能等那么久。这个结果是一个诚实的提醒:探索提高了找到更好位置的机会,但不能保证找到最好的位置。 (https://5dlibrary.com/answers/zh/active-inference-sandbox/c04/) - Ch.5 Structured Sandboxes and the Reshaping of Priors — With inverse temperatures from 2 to 10 the better option is found within 100 trials almost every time; at 20 only 36 % of the time, at 40 only 1 %. The last column is a reminder of the other side: the learner with the lowest inverse temperature finds the better option fastest but explores too much, keeps trying options it knows are worse, and earns less on average than a learner with a moderate inverse temperature. In this model the best is not unlimited exploration but thorough exploration in a low-stakes phase, followed by exploitation once good estimates are in hand. That is exactly the sequence a good sandbox can offer: "hot" in the sandbox, "cooler" back in life. (https://5dlibrary.com/answers/en/active-inference-sandbox/c05/) 第5章 结构化沙盘与先验的重塑 — 逆温度为 2 到 10 时,几乎每次都能在 100 轮内发现更好的选项;到 20 时只有 36%,到 40 时只有 1%。最后一列提醒另一面:逆温度最低的学习者虽然最快发现好选项,但它探索太多,一直在试明知较差的选项,平均收益反而不如中等逆温度的学习者。在这个模型里,最好的不是无限的探索,而是先在低风险的阶段充分探索,再在有了好的估计之后转向利用。这正是一个好的沙盘可以提供的顺序:在沙盘里「高温」,回到生活中「降温」。 (https://5dlibrary.com/answers/zh/active-inference-sandbox/c05/) - Ch.6 Determinism, Chaos and Agency — The separation of the two trajectories grows by a factor of about $e^{0.92} \approx 2.5$ per time unit. The program's estimate of the largest Lyapunov exponent, about 0.92, is close to the commonly quoted value of about 0.906. This gives a predictability horizon: an initial error of $10{-9}$ grows to the size of the attractor in about 23 time units, an error of $10^{-3}$ in only about 7.5. Each thousandfold improvement in initial measurement adds only about 7.5 time units of prediction: better precision buys only a linear extension of foresight. (https://5dlibrary.com/answers/en/active-inference-sandbox/c06/) 第6章 决定论、混沌与能动性 — 两条轨迹的距离大约每过一个时间单位放大 $e^{0.92} \approx 2.5$ 倍。程序估计的最大李雅普诺夫指数约为 0.92,与文献中常见的约 0.906 接近。由此得到可预测期限:初始误差 $10{-9}$ 约 23 个时间单位后长到与吸引子同一量级,误差 $10^{-3}$ 只需约 7.5 个时间单位。初始测量每精确一千倍,可预测期限只增加约 7.5 个时间单位:精度的提升只换来线性的预测延长。 (https://5dlibrary.com/answers/zh/active-inference-sandbox/c06/) - Ch.8 Practice and a Pre-Registered Trial — For a smallest worthwhile effect of 0.4 standard deviations, a baseline–outcome correlation of 0.5 and 90 % power, each arm needs 100 people; with a design effect of 1.22 for groups of 12 and an intraclass correlation of 0.02, each arm needs 122 completers, and allowing 15 % loss, 144 recruits, that is 12 groups per arm. The baseline–outcome correlation and the intraclass correlation are assumptions that will be checked in a 24-person pilot; if the baseline–outcome correlation is below 0.4, the sample size will be recalculated before the main trial starts. (https://5dlibrary.com/answers/en/active-inference-sandbox/c08/) 第8章 实践与预注册试验 — 按最小值得关注的效应 0.4 个标准差、基线与结局相关 0.5、90% 把握度计算,每组需要 100 人;考虑 12 人小班、组内相关 0.02 的设计效应 1.22,每组需要 122 人完成,按 15% 的脱落招募 144 人,即每组 12 个小班。基线与结局的相关、组内相关都是假设,将在 24 人的预试验中核对;如果基线与结局的相关低于 0.4,正式试验开始前重新计算样本量。 (https://5dlibrary.com/answers/zh/active-inference-sandbox/c08/) ## 5DL-B-016 · Cross-Cultural Contracts, Real Options and Game Theory of Southeast Asian Market Entry 中文书名:跨文化契约、实物期权与东南亚出海博弈论 DOI: https://doi.org/10.5281/zenodo.23765345 Cite: Fan, Y. (2026). Cross-Cultural Contracts, Real Options and Game Theory of Southeast Asian Market Entry (Version 1.0.1). 5D Library. https://doi.org/10.5281/zenodo.23765345 Book page: https://5dlibrary.com/books/cross-cultural-contract-game/ Scope: This book is not legal, financial or investment advice. Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. Abstract: What a Chinese team sees as flexible, case-by-case adjustment is often read by overseas partners as erratic. This book examines the gap, and contractual remedies for it, with game theory, real options and organisational economics. It covers rule-based commitment and outcome-based adaptation, Hall's high and low context and Hofstede's dimensions with their methodological critics, and a Bayesian model showing that when the reasons for change cannot be seen, an adaptive partner cannot be told from an opportunist; binomial real options (committing at once is better at low volatility, staged entry above an annual volatility of about 0.3), optimal stopping solved with the Bellman equation (the optimal exit threshold rises as information accumulates and is more patient than a fixed three-month stop-loss) and the signalling game of exit (writing down a verifiable trigger in advance keeps an exit under the rule from damaging reputation); Grossman–Hart–Moore residual control and Baker–Gibbons–Murphy relational contracts, with formal, relational and hybrid governance compared across speeds of environmental change (the hybrid is best for cross-cultural partners under moderate uncertainty, relational governance alone when change is very fast); the stability condition of a delayed feedback loop and the cost of pivoting for brand assets; Kogut–Singh cultural distance, institutional distance, halal certification and Chinese business networks; and Holmström–Milgrom multitask incentives with a self-selecting contract menu. The eight chapters are read beside the Art of War's chapters on weak points, stratagem and variation, the Book of Changes' "when exhausted, change" and "change follows the times", the Analects' "words must be kept, deeds carried through" and "promises close to what is right" with Mencius' "only where righteousness lies", and the covenant tradition, pacta sunt servanda and changed circumstances in the Vienna Convention on the Law of Treaties, noting the limits of such a reading. They end with a four-step contract alignment checklist and a pre-registered randomised vignette experiment comparing a fixed-target contract with a state-contingent contract that writes down a 90-day exploration period and stopping rules (co-primary outcomes: trust and perceived breach risk after the foreign party exits under the rules, each at alpha 0.025; 277 analysable and 347 recruited per arm). Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. This book is not legal, financial or investment advice. 摘要:中国团队眼中「具体问题具体分析」的灵活调整,在海外合伙人眼中常被解读为朝令夕改。本书用博弈论、实物期权与组织经济学考察这一分歧及其契约解决办法。依次讨论:规则导向的承诺与结果导向的适应,霍尔高低语境与霍夫斯泰德维度及其方法论批评,以及一个说明「看不到改变的理由时,适应者与机会主义者无从区分」的贝叶斯模型;二叉树实物期权(波动小时一次性投入更好,年波动率约 0.3 以上分阶段进入更好)、贝尔曼方程求解的最优停时(最优退出门槛随信息积累而提高,比固定的三个月止损线更有耐心)与退出的信号博弈(事先写明可核实的触发条件,使按规则退出不损害声誉);格罗斯曼–哈特–摩尔剩余控制权与贝克–吉本斯–墨菲关系型契约,以及正式、关系与混合治理在不同环境变化速度下的比较(混合方式在跨文化、中等不确定时最好,环境变化很快时纯关系治理更好);带延迟反馈回路的稳定条件与品牌资产的转向代价;科格特–辛格文化距离、制度距离、清真认证与华人商业网络;霍姆斯特罗姆–米尔格罗姆多任务激励与自我选择的合同菜单。 八章与《孙子兵法》虚实、谋攻、九变诸篇,《周易·系辞下》「穷则变」与「变通者,趣时者也」,《论语》「言必信,行必果」「信近于义」与《孟子》「惟义所在」,以及盟约传统、《维也纳条约法公约》的约定必须遵守与情势变更并读,并指出并读的限度;最后给出四步契约对齐核查表,以及一项比较固定目标合同与写明 90 天探索期停时规则的状态依存合同的预注册随机情景实验方案(共同主要结局为外方按规则退出后的信任与感知违约风险,各 α = 0.025,每组 277 人可分析、招募 347 人)。书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部核对)、原典与作者的假说。本书不是法律、财务或投资建议。 Chapter answers: - Ch.1 Two Business Operating Systems: the Commitment Game and Adaptive Search — In case A the 4 changes almost rule out the committed type (0.45 falls to 0.01), but they raise the probability of the opportunist type from 0.10 to 0.18, nearly doubling it. The reason is not that the partner is suspicious; it is that, to an observer who cannot see reasons, the adaptive and the opportunist behave alike. In case B the same 4 changes become strong evidence that the team really is adapting to the market as agreed. Case C shows that this judgement tolerates error: one deviation brings the opportunist probability back only to 0.12 and does not make trust collapse. (https://5dlibrary.com/answers/en/cross-cultural-contract-game/c01/) 第1章 两种商业操作系统:承诺博弈与适应性寻优 — 在情形 A 中,4 次改变几乎排除了守诺型(0.45 降到 0.01),却使机会主义型的概率从 0.10 升到 0.18,接近翻倍。原因不是合伙人多疑,而是对一个看不到原因的观察者来说,适应型与机会主义型的行为看起来一样。在情形 B 中,同样的 4 次改变反而成了「对方确实是按约定适应市场」的有力证据。情形 C 说明这种判断也有容错:一次偏差使机会主义型的概率回到 0.12,并不会让信任一下子崩塌。 (https://5dlibrary.com/answers/zh/cross-cultural-contract-game/c01/) - Ch.2 Real Options and Optimal Stopping: the Value of a Stop-Loss and the Signal It Sends — In this example the project's plain NPV is +4. When volatility is low (annual 0.10–0.25), committing at once is the best of the three: there is little uncertainty, so neither the pilot fee nor the earnings lost while waiting pay off. At a volatility of 0.40 the value of staged entry (9.1) exceeds committing (4.9), and the gap widens as volatility rises. The abandonment option has value too, but less than staged entry here, because the salvage is only 30 % of the investment. (https://5dlibrary.com/answers/en/cross-cultural-contract-game/c02/) 第2章 实物期权与最优停时:止损的价值与止损的信号 — 在这个例子中,项目本身的净现值是 +4。市场波动小(年波动率 0.10–0.25)时,一次性投入是三者中最好的:不确定性不大,试点费和等待期间损失的收益都不划算。波动率达到 0.40 时,分阶段进入的价值(9.1)超过一次性投入(4.9);波动率越高,差距越大。放弃期权也有价值,但在这个例子里不如分阶段进入,因为残值只有投资的三成。 (https://5dlibrary.com/answers/zh/cross-cultural-contract-game/c02/) - Ch.3 Incomplete Contracts and Relational Contracts — The conclusion agrees with the theory: the asset should be owned by the party whose investment matters more. In a business led by product and technology, ownership by headquarters brings the surplus to 92 % of the first-best level; in a business led by distribution and licences, the local partner should own it. Joint ownership is the worst in both cases, at 75 % of the first best: each side can veto the other, and as a result neither has enough reason to invest. (https://5dlibrary.com/answers/en/cross-cultural-contract-game/c03/) 第3章 不完全契约与关系型契约 — 结论与理论一致:资产应当归投资更重要的一方所有。产品与技术主导的业务,由总部持有关键资产,合作剩余达到最优水平的 92%;渠道与牌照主导的业务,则应由本地合伙人持有。共同所有在两种情况下都是最差的,只有最优水平的 75%:每一方都可以否决对方,结果每一方都没有足够的投资动力。 (https://5dlibrary.com/answers/zh/cross-cultural-contract-game/c03/) - Ch.4 The Control Theory of Iteration Speed: Feedback Delay, Gain and Transaction Costs — When brand assets do not matter ($w = 0$), smoothing at 20 % is best; chasing the data is worse, because the data are too noisy. The more brand assets matter, the more the best policy leans towards adjusting rarely and in steps: at $w = 0.2$ the step policy earns 26.6, almost six times the chasing policy's 4.6. Freezing is never best: a firm that never adjusts is eventually left behind by the market's drift. (https://5dlibrary.com/answers/en/cross-cultural-contract-game/c04/) 第4章 迭代速度的控制论:反馈延迟、增益与交易成本 — 当品牌资产不重要时($w = 0$),平滑 20% 的策略最好;直接追逐数据反而更差,因为数据里的噪声太大。品牌资产越重要,最好的策略越倾向于少调整、按阶梯调整:$w = 0.2$ 时阶梯策略的收入(26.6)是追逐策略(4.6)的近六倍。冻结策略在任何情况下都不是最好的:完全不调整,最终会被市场的漂移甩开。 (https://5dlibrary.com/answers/zh/cross-cultural-contract-game/c04/) - Ch.5 Institutional Distance, Religious Ethics and Southeast Asia's Plural Markets — Under these assumptions, the referral network, even when it can reach only 20 candidates, gives a better expected result than screening 30 on the open market, and cuts the chance of an unreliable partner from about 16 % to about 5 %. But the last lines of the program show where the network's advantage comes from and where it ends: its value comes from the low reliable share on the open market. Once better courts, credit bureaus and platform ratings raise that share to 90 %, screening 30 candidates on the open market is already slightly better than the network. As business registries, electronic payments and platform ratings improve across Southeast Asia, the network's relative advantage may be shrinking, but that needs to be tested with data. (https://5dlibrary.com/answers/en/cross-cultural-contract-game/c05/) 第5章 制度距离、宗教伦理与东南亚的多元市场 — 在这组假设下,推荐网络即使只能接触 20 家候选人,也比在开放市场上筛选 30 家得到更好的期望结果,遇到不可靠伙伴的概率从约 16% 降到约 5%。但程序的最后几行说明了网络优势的来源与限度:它的价值来自开放市场的可靠比例低。当开放市场的可靠比例因为更好的法院、征信与平台评价体系提高到 90% 时,在开放市场上筛选 30 家已经略好于网络。随着东南亚各国的商业登记、电子支付与平台评价体系不断完善,网络的相对优势也可能在缩小,但这需要数据检验。 (https://5dlibrary.com/answers/zh/cross-cultural-contract-game/c05/) - Ch.6 The Principal-Agent Problem and Incentives for Local Teams — Each type chooses the contract designed for it: the long-horizon partner takes the partner contract (worth 378.9 against 318.0), the short-horizon partner the salary contract (103.7 against 79.2). The deferred payment and the higher share appeal more to long-horizon partners, who are more likely to collect the deferred payment and mind monthly fluctuations less. The cost is that, for a long-horizon partner, the partner contract costs more in total (474.0) than the salary contract (360.0). That extra cost is the price headquarters pays to separate the types and keep long-term partners; whether it is worth paying depends on how much more value a long-term partner creates, which the model does not compute. (https://5dlibrary.com/answers/en/cross-cultural-contract-game/c06/) 第6章 委托-代理问题与本地团队的激励 — 两类合伙人各自选择了为他们设计的合同:长期型选择合伙型合同(价值 378.9 对 318.0),短期型选择工资型合同(103.7 对 79.2)。递延款与较高的分成对长期型更有吸引力,因为他们更可能拿到递延款,也更不在意每月收入的波动。代价是:对长期型合伙人,合伙型合同的总成本(474.0)高于工资型(360.0)。这部分额外成本是总部为「区分类型、留住长期伙伴」付出的价格;它是否划算,取决于一个长期伙伴能为企业多创造多少价值,而这在模型中没有计算。 (https://5dlibrary.com/answers/zh/cross-cultural-contract-game/c06/) - Ch.8 Practice and a Pre-registered Trial — With a smallest effect of interest of 0.30 standard deviations, two-sided $\alpha = 0.025$ for each co-primary outcome and 90 % power, each arm needs 277 analysable participants; the power to detect both outcomes together is at least 80 %. Allowing for about 20 % of participants to be excluded for failing attention or manipulation checks, each arm recruits 347, 694 in all, about 139 in each of the five countries. The sample within each country is too small to test differences between countries, which are reported only as exploratory. (https://5dlibrary.com/answers/en/cross-cultural-contract-game/c08/) 第8章 实践与预注册试验 — 按最小关注效应 0.30 个标准差、每项共同主要结局双侧 $\alpha = 0.025$、90% 把握度计算,每组需要 277 名可分析的参与者;两项结局同时被检出的把握度至少为 80%。考虑到约 20% 的参与者可能因未通过注意力或操纵检查而被排除,每组招募 347 人,共 694 人,五国每国约 139 人。各国内部的样本量不足以单独检验国家之间的差别,国家差别只作探索性报告。 (https://5dlibrary.com/answers/zh/cross-cultural-contract-game/c08/) ## 5DL-B-017 · Ketone Metabolism, β-Hydroxybutyrate Signalling and the Bioenergetics of Intermittent Fasting 中文书名:酮体代谢、β-羟丁酸信号转导与间歇性禁食的生物能量学 DOI: https://doi.org/10.5281/zenodo.23753263 Cite: Fan, Y. (2026). Ketone Metabolism, β-Hydroxybutyrate Signalling and the Bioenergetics of Intermittent Fasting (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23753263 Book page: https://5dlibrary.com/books/ketone-metabolism-autophagy/ Scope: Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses. Abstract: This book examines the metabolic switch between feeding and fasting, and which of the popular claims about it have evidence and which do not. It covers the Randle cycle and the stoichiometry of hepatic ketogenesis (per molecule of oxygen, β-hydroxybutyrate yields 4.78 ATP, more than fatty acid at 4.61 and less than glucose at 5.33; routing fat through ketones loses about 4 % of the ATP), with a two-compartment model of liver glycogen and blood ketones; Veech's thermodynamic hypothesis that ketones widen the mitochondrial redox span between NAD and coenzyme Q and raise the free energy of ATP hydrolysis, computed with the Nernst equation, with an experimental basis in the isolated heart, consistent direction but unsettled mechanism in the human heart, and conflicting evidence on exercise performance; research on β-hydroxybutyrate as an HCAR2 ligand, a histone deacetylase inhibitor (later work found its action in cells far weaker than butyrate's), an NLRP3 inflammasome inhibitor and a source of histone β-hydroxybutyrylation, with concentration–effect relations from the Hill equation; hysteresis in the AMPK–mTORC1 double-negative switch and autophagy, noting that mouse autophagy timelines cannot be converted into human hours and that humans lack a simple non-invasive marker that times autophagy; the real effect sizes of randomised trials in epilepsy, type 2 diabetes, metabolic wards and time-restricted eating (with calories matched, neither the extra fat loss predicted by the insulin model nor clear additional weight loss from time restriction), and the individual risk of rising LDL cholesterol; and the natriuresis, extracellular fluid and potassium changes caused by falling insulin in the first low-carbohydrate days, with glycogen-bound water. The eight chapters are read beside the Mawangdui silk manuscript Quegu shiqi, Ge Hong's Baopuzi ("cutting off grain can only save the cost of food; it cannot by itself make a person live long"), the Suwen, the Lingshu and the Analects, distinguishing a tradition of moderate dietary restraint from life-threatening extreme fasting. They end with a four-step checklist for a personal metabolic log (including a calculation of how unstable the glucose–ketone index is at low ketone levels) and a pre-registered randomised crossover trial comparing 16:8 early time-restricted eating with three meals, with meals provided and calories and protein matched (primary outcome MAGE from continuous glucose monitoring; 45 completers, 57 recruited). Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses. This book is not medical advice; people with type 1 diabetes, on glucose-lowering drugs or SGLT2 inhibitors, pregnant or breastfeeding, with a history of eating disorders or with liver or kidney disease, among others, should not try fasting or a ketogenic diet on their own. 摘要:本书讨论进食与禁食之间的代谢切换,以及围绕它的流行说法哪些有证据、哪些没有。依次讨论:兰德尔循环与肝脏生酮的化学计量(按每分子氧计算,β-羟丁酸产生 4.78 个 ATP,高于脂肪酸的 4.61,低于葡萄糖的 5.33;经由酮体转手一次约损失 4% 的 ATP),以及肝糖原与血酮的两室模型;维奇关于酮体增大线粒体 NAD 与辅酶 Q 之间氧化还原跨度、提高 ATP 水解自由能的热力学假说,用能斯特方程计算,并说明它在离体心脏中有实验基础,在人体心脏中方向一致但机制未定,在运动表现中证据互相矛盾;β-羟丁酸作为 HCAR2 受体配体、组蛋白去乙酰化酶抑制剂(后续研究显示其细胞内作用远弱于丁酸)、NLRP3 炎症小体抑制剂与组蛋白 β-羟丁酰化来源的研究,以及按希尔方程计算的浓度–效应关系;AMPK–mTORC1 双负反馈开关的滞后性与细胞自噬,并指出小鼠的自噬时间线不能换算为人的小时数,人体中也没有简单无创的自噬计时标志物;癫痫、2 型糖尿病、代谢舱与限时进食随机试验的真实效应量(热量匹配时未见胰岛素模型预言的额外脂肪减少,限时进食也未见明确的额外减重),以及 LDL 胆固醇的个体差异风险;低碳水初期胰岛素下降引起的排钠利尿、细胞外液与钾的变化,以及糖原结合水。 八章与马王堆帛书《却谷食气》、葛洪《抱朴子内篇·杂应》(「断谷人止可息肴粮之费,不能独令人长生也」)、《素问》《灵枢》与《论语·乡党》并读,区分适度节制的饮食传统与危及生命的极端断食;最后给出个人代谢记录的四步核查表(含葡萄糖–酮体指数在低血酮时不稳定的计算),以及一项在提供餐食、热量与蛋白质相同条件下比较 16:8 早期限时进食与三餐分布的预注册随机交叉试验方案(主要结局为连续血糖监测的 MAGE,45 人完成、招募 57 人)。书中以标注框区分已发表的证据、可运行的 Python 计算(14 段程序,全部核对)、原典与作者的假说。本书不是医疗建议;1 型糖尿病、使用降糖药或 SGLT2 抑制剂、孕哺、进食障碍史、肝肾疾病等人群不应自行尝试禁食或生酮饮食。 Chapter answers: - Ch.1 The Metabolic Switch: from Glucose to Ketones — The chemistry of the metabolic switch is the Randle cycle and hepatic ketogenesis. Per molecule of oxygen, ketones are more efficient than fatty acids but less efficient than glucose; routing fat through ketones loses about 4 % of the ATP. The time course is set by liver glycogen: an overnight fast produces almost no ketosis, blood ketones rise clearly only after 24–36 hours, and individuals differ widely. The next chapter takes up a more contested question: if ketones are not more oxygen-efficient, why do some believe they make the heart and brain work better? (https://5dlibrary.com/answers/en/ketone-metabolism-autophagy/c01/) 第1章 代谢切换:从葡萄糖到酮体 — 代谢切换的化学基础是兰德尔循环与肝脏生酮。按每分子氧计算,酮体比脂肪酸高效,却比葡萄糖低效;经由酮体转手一次,机体损失约 4% 的 ATP。切换的时间进程由肝糖原决定:一夜禁食几乎不产生酮症,24–36 小时后血酮才明显升高,个体差异很大。下一章讨论一个更有争议的问题:如果酮体并不更省氧,为什么有人认为它让心脏和大脑工作得更好? (https://5dlibrary.com/answers/zh/ketone-metabolism-autophagy/c01/) - Ch.2 Why the Heart and Brain Use Ketones: Veech's Thermodynamic Hypothesis and Its Limits — In these illustrative numbers, tripling NADH/NAD^+ and halving succinate/fumarate widen the span between NAD and Q from 306 mV to 330 mV, raising the free energy per electron pair from −59.0 to −63.8 kJ/mol; at the same time, lower free ADP and inorganic phosphate change the free energy of ATP hydrolysis from −58.1 to −60.9 kJ/mol. The last two lines of the program give two useful rules of thumb: each tenfold change in NADH/NAD^+ moves the potential by about 31 mV, and each tenfold change in the product of ADP and phosphate moves the free energy of ATP hydrolysis by about 5.9 kJ/mol. (https://5dlibrary.com/answers/en/ketone-metabolism-autophagy/c02/) 第2章 心脏与大脑为什么用酮体:维奇的热力学假说及其边界 — 在这组示意数字中,NADH/NAD^+ 比值升高三倍、琥珀酸/延胡索酸比值降低一半,使 NAD 与 Q 之间的电位差从 306 mV 增大到 330 mV,每对电子的自由能从 −59.0 增加到 −63.8 kJ/mol;同时,游离 ADP 和无机磷酸的降低使 ATP 水解自由能从 −58.1 变为 −60.9 kJ/mol。程序最后两行给出两条有用的经验关系:NADH/NAD^+ 每变化十倍,电位移动约 31 mV;ADP 与无机磷酸的乘积每变化十倍,ATP 水解自由能移动约 5.9 kJ/mol。 (https://5dlibrary.com/answers/zh/ketone-metabolism-autophagy/c02/) - Ch.3 More Than a Fuel: β-Hydroxybutyrate as a Signal — BHB does act beyond being a fuel: it is a receptor ligand, it inhibits histone deacetylases in the test tube, it inhibits the NLRP3 inflammasome in cells and mice, and it is the source of a new histone modification. But each action has its own concentration range; on present data, at the blood ketone levels most people reach, receptor actions are more likely to matter than HDAC inhibition, and the cellular evidence for HDAC inhibition itself has been questioned. The next chapter turns to another much-discussed topic: fasting and autophagy. (https://5dlibrary.com/answers/en/ketone-metabolism-autophagy/c03/) 第3章 不只是燃料:β-羟丁酸的信号作用 — BHB 确实有燃料之外的作用:它是受体配体,在试管中能抑制组蛋白去乙酰化酶,在细胞与小鼠中能抑制 NLRP3 炎症小体,还是一种新的组蛋白修饰的来源。但这些作用各有其浓度范围;按目前的数据,在一般人可以达到的血酮水平下,受体作用比 HDAC 抑制更可能起作用,而 HDAC 抑制本身在细胞中的证据已经受到质疑。下一章转向另一个被广泛谈论的话题:禁食与细胞自噬。 (https://5dlibrary.com/answers/zh/ketone-metabolism-autophagy/c03/) - Ch.4 The AMPK–mTORC1 Switch and Autophagy — The mutual inhibition of AMPK and mTORC1 can form a switch with hysteresis: one meal can shut autophagy off, and switching it back on needs the nutrient signal to fall far enough for long enough. But the switch model's conclusions depend on uncalibrated parameters; autophagy timelines in mice cannot be converted into human hours; and humans lack a simple marker that can time autophagy. "How many hours of fasting switch on autophagy" is at present a question without an answer. The next chapter leaves mechanisms behind and asks what randomised trials in people have found. (https://5dlibrary.com/answers/en/ketone-metabolism-autophagy/c04/) 第4章 AMPK–mTORC1 开关与细胞自噬 — AMPK 与 mTORC1 的相互抑制可以构成一个带滞后的开关:一次进食就能关闭自噬,重新开启则需要营养信号降得够低、持续够久。但开关模型的结论取决于未经校准的参数;小鼠的自噬时间线不能换算成人的小时数;人体中也缺乏能为自噬计时的简单标志物。「禁食多少小时开启自噬」在目前是一个没有答案的问题。下一章离开机制,看看人体随机对照试验说了什么。 (https://5dlibrary.com/answers/zh/ketone-metabolism-autophagy/c04/) - Ch.5 What Randomised Trials in People Show — A ketogenic diet for drug-resistant childhood epilepsy is supported by reliable randomised evidence; in type 2 diabetes it brings short-term benefits that fade over time, and medication must be adjusted under a doctor's guidance. In studies with strictly matched calories and protein, a ketogenic diet did not produce the extra fat loss the insulin model predicts; with calories matched, time-restricted eating showed no clear additional weight loss, though early time-restricted eating may improve some metabolic markers without any change in weight. A ketogenic diet can raise LDL-C markedly in some people. The next chapter takes up a more everyday question: why do many people feel tired, palpitations or cramps in the first days of a low-carbohydrate diet? (https://5dlibrary.com/answers/en/ketone-metabolism-autophagy/c05/) 第5章 人体随机试验说了什么 — 生酮饮食治疗难治性儿童癫痫有可靠的随机试验证据;用于 2 型糖尿病有短期获益,长期效果减弱,且必须在医生指导下调整药物。在热量与蛋白质严格匹配的研究中,生酮饮食没有带来胰岛素模型预言的额外脂肪减少;在热量匹配的条件下,限时进食也没有显示明确的额外减重效果,但早期限时进食可能在不改变体重的情况下改善部分代谢指标。生酮饮食可能显著升高部分人的 LDL-C。下一章讨论一个更日常的问题:为什么开始低碳水饮食的头几天,很多人觉得疲乏、心悸或抽筋? (https://5dlibrary.com/answers/zh/ketone-metabolism-autophagy/c05/) - Ch.6 Keto-Adaptation: the Kidney, Sodium and Water — An overnight fast uses only a few tens of grams of liver glycogen, about 0.14 kg with its bound water; the first week of a ketogenic diet uses liver glycogen and part of muscle glycogen totalling about 0.84 kg with bound water, plus about 1.2 litres of extracellular fluid. Together that is about 2 kg, while the adipose tissue corresponding to a 500 kcal daily energy gap over the same week is only about 0.45 kg. So most of the rapid weight loss in the first week of a low-carbohydrate diet is water, not fat, and it returns quickly once carbohydrate is eaten again. That is neither failure nor success, just the physiology of glycogen and water. (https://5dlibrary.com/answers/en/ketone-metabolism-autophagy/c06/) 第6章 生酮适应期:肾脏、钠与水 — 一夜禁食大约只动用几十克肝糖原,连同结合的水约 0.14 千克;生酮饮食的第一周,动用的肝糖原与一部分肌糖原连同结合的水约 0.84 千克,再加上约 1.2 升的细胞外液。两者合计约 2 千克,而同一周里与 500 千卡能量缺口对应的脂肪组织只有约 0.45 千克。所以,开始低碳水饮食头一周体重的快速下降,大部分是水,不是脂肪;恢复吃碳水化合物后,这部分体重也会很快回来。这不是失败,也不是成功,只是糖原与水的生理。 (https://5dlibrary.com/answers/zh/ketone-metabolism-autophagy/c06/) - Ch.8 Practice and a Pre-registered Trial — This fictional log shows two things. First, in ordinary people on waking, BHB is only 0.1–0.5 mmol/L and the GKI swings widely between about 10 and over 50. The middle lines of the program show why: common fingertip ketone meters read in steps of 0.1 mmol/L, so BHB moving from 0.1 to 0.2 halves the GKI from 50 to 25, while BHB moving from 1.0 to 1.1 changes it only from 5.0 to 4.5. At low ketone levels, the GKI mainly reflects the resolution of the meter, not metabolic state. Second, in this week, mornings after longer overnight fasts had higher ketones (correlation 0.72), but with only 7 data points this is description, not a conclusion. (https://5dlibrary.com/answers/en/ketone-metabolism-autophagy/c08/) 第8章 实践与预注册试验 — 这份虚构记录说明了两点。第一,在一般人的晨起状态下,BHB 只有 0.1–0.5 mmol/L,GKI 在 10 到 50 多之间大幅波动。程序中间几行显示了原因:常见的指尖血酮仪以 0.1 mmol/L 为一档,BHB 从 0.1 变为 0.2,GKI 就从 50 降到 25;而 BHB 从 1.0 变为 1.1,GKI 只从 5.0 变为 4.5。在血酮较低时,GKI 主要反映的是血酮仪的分辨率,而不是代谢状态。 第二,这一周中,夜间禁食时间较长的早上血酮也较高(相关系数 0.72),但只有 7 个数据点,只能作描述,不能作结论。 (https://5dlibrary.com/answers/zh/ketone-metabolism-autophagy/c08/) ## 5DL-B-018 · Acoustic Persuasion, Prosodic Information Theory and Interpersonal Physiological Synchrony 中文书名:声音说服力、韵律信息论与人际生理同步 DOI: https://doi.org/10.5281/zenodo.23753267 Cite: Fan, Y. (2026). Acoustic Persuasion, Prosodic Information Theory and Interpersonal Physiological Synchrony (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23753267 Book page: https://5dlibrary.com/books/acoustic-persuasion-synchrony/ Scope: Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked; every sound in the book is synthesised), classical sources and the author's hypotheses. Abstract: The same words spoken in different voices can lead listeners to judge the speaker very differently. This book examines that difference with acoustics, information theory and cognitive neuroscience, and asks which popular claims about it have evidence. It covers the true source and scope of Mehrabian's "7 %–38 %–55 %" rule (it applies only when a single word conveys liking and the channels contradict each other) and empirical research on how fundamental frequency, formant dispersion, harmonics-to-noise ratio and speech rate affect listeners' judgements, including a single-word study in which the trustworthy prosodic prototype rose at the end of the word; Shannon surprisal and the uniform information density hypothesis, with a limited-capacity listener model (rushing leaves key information almost unencoded, a pause before and lengthening of the key word works best, and fillers act somewhat like pauses but may make the speaker seem less sure); final falls and rises generated and measured with the Fujisaki model, declination being partly planned and uptalk serving more functions than uncertainty; speaker–listener brain coupling and heart-rate synchrony among listeners, with the Kuramoto model and control conditions separating common drive from mutual coupling, and a circular-shift test of inter-subject correlation; computation of roughness from 30–150 Hz amplitude modulation in screams and of the harmonics-to-noise ratio; and the proximity effect of cardioid microphones, signal-to-noise ratio and critical distance. The eight chapters are read beside the Record of Music's "all sound arises from the human heart" and its six sounds of the heart, the Guiguzi's probing and sounding out and "one's own silence is stillness", and the Analects' "the gentleman has three aspects" and "in speech, it is enough to convey the meaning", noting the limits of such a reading. They end with a four-step acoustic checklist for use before speaking or recording and a pre-registered double-blind randomised controlled trial comparing a "final falls + 0.6-second pauses before key words" version with a "final rises + no pauses + faster speech" version of the same pitch (primary outcome willingness to invest from a hypothetical budget, stratified by four voice actors, 338 analysable and 398 recruited per arm, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked; every sound in the book is synthesised), classical sources and the author's hypotheses. The book is about helping true content to be heard and remembered, not about manipulating listeners with vocal technique. 摘要:同样的文字,用不同的声音说出来,听众对说话者的判断可能很不一样。本书用声学、信息论与认知神经科学考察这种差别,以及围绕它的流行说法哪些有证据。依次讨论:梅拉比安「7%-38%-55%」规则的真实来源与适用范围(只适用于单个词表达好恶、渠道互相矛盾的情形),以及基频、共振峰离散度、谐噪比与语速影响听众判断的实证研究,包括一项单个词研究中可信的韵律原型在词末上扬的发现;香农惊奇度与均匀信息密度假说,以及一个有限容量听者模型(匆忙使关键信息几乎无法编码,关键词前停顿并拉长最好,填充词也有类似停顿的作用,但可能让说话者显得不够有把握);藤崎模型生成与测量的句末下降与上扬,下倾部分是计划性的,上扬的功能不只是不确定;说话者与听者的大脑耦合与听众之间的心率同步,以及用库拉莫托模型和对照条件区分共同驱动与相互耦合,受试者间相关的循环平移检验;尖叫中 30–150 Hz 振幅调制的粗糙度与谐噪比的计算;心形麦克风近讲效应、信噪比与临界距离。 八章与《礼记·乐记》「凡音之起,由人心生也」及六种心境之声、《鬼谷子》的揣摩与「己默者,静也」、《论语》「君子有三变」与「辞达而已矣」并读,并指出并读的限度;最后给出一份演说与录制前的四步声学核查表,以及一项比较「句末下降 + 关键词前 0.6 秒停顿」与「句末上扬 + 无停顿 + 较快语速」两种韵律版本的预注册双盲随机对照试验方案(主要结局为假设预算中的投资意愿,四位声音演员分层,每组 338 人可分析、招募 398 人,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(14 段程序,全部核对;书中的声音均为程序合成)、原典与作者的假说。本书讨论如何让真实的内容被听清、被记住,不是用声音技巧操纵听众的方法。 Chapter answers: - Ch.1 Paralinguistic Acoustics: Why "How" Rewrites the Credibility of "What" — "7 %–38 %–55 %" applies only to the experimental situation in which a single word conveys liking and the channels contradict each other; it is not a general law of communication. But the voice does affect listeners' judgements: hearing a voice rather than reading the same words leads evaluators to rate competence more highly; lower pitch and closer formants go with judgements of dominance; speaking too fast and too slow both carry costs; and the same prosodic feature can produce opposite judgements on different dimensions. The next chapter takes up a more specific question: why do pauses have power? (https://5dlibrary.com/answers/en/acoustic-persuasion-synchrony/c01/) 第1章 副语言声学:为什么「怎么说」会改写「说什么」的信度 — 「7%-38%-55%」只适用于单个词表达好恶、渠道互相矛盾的实验情形,不是沟通的一般规律。但声音确实影响听众的判断:听到声音比只读文字使评估者给出更高的能力评价;较低的基频和较密的共振峰与支配性判断相关;语速过快和过慢都有代价;同一个韵律特征在不同维度上可能产生相反的判断。下一章转向一个更具体的问题:停顿为什么有力量? (https://5dlibrary.com/answers/zh/acoustic-persuasion-synchrony/c01/) - Ch.4 Interpersonal Neural and Physiological Synchrony — When listeners sit together and hear the same story, both mechanisms produce high synchrony (0.93 and 0.96), and this condition alone cannot tell them apart. The controls do: synchrony from a common drive survives testing alone in full and almost vanishes when each listener hears the story with a random delay; synchrony from mutual coupling vanishes when listeners are tested alone and survives the random-delay story in full. (The value of 0.15 in the "none" row is the level that a finite sample of 30 random phases shows anyway.) (https://5dlibrary.com/answers/en/acoustic-persuasion-synchrony/c04/) 第4章 人际神经与生理同步 — 两种机制在「坐在一起听同一个故事」时都能产生很高的同步(0.93 与 0.96),单看这一种条件无法区分它们。对照条件把它们分开了:共同驱动产生的同步在单独测试时完全保留,在随机延迟时几乎消失;相互耦合产生的同步则在单独测试时消失,在随机延迟时完全保留。(「无」一行的 0.15 是 30 个随机相位的有限样本本身就会有的水平。) (https://5dlibrary.com/answers/zh/acoustic-persuasion-synchrony/c04/) - Ch.5 Emotional Contagion and Acoustic Roughness — The energy of screams is concentrated in fast amplitude modulations of about 30–150 Hz, a region barely used by ordinary speech, and rough sounds activate the amygdala more strongly. Roughness can be computed from the modulation spectrum of the envelope; the harmonics-to-noise ratio is a related but different composite measure. Inferring that gentle, clear voices persuade better is reasonable but not directly tested. The next chapter turns from people to equipment: how the distance from a microphone changes the way a voice is heard. (https://5dlibrary.com/answers/en/acoustic-persuasion-synchrony/c05/) 第5章 情绪传染与声学粗糙度 — 尖叫的能量集中在约 30–150 Hz 的快速振幅调制区,这个区域在普通言语中几乎不被使用,粗糙的声音更强地激活杏仁核。粗糙度可以通过包络的调制谱计算;谐噪比是另一个相关但不同的综合指标。由此推断温和清亮的嗓音更有说服力,是一个合理但未经直接检验的假说。下一章从人转向设备:麦克风前的距离,怎样改变声音被听到的样子。 (https://5dlibrary.com/answers/zh/acoustic-persuasion-synchrony/c05/) - Ch.6 In Front of the Camera and Microphone: Proximity Effect, Signal-to-Noise Ratio and Reverberation — A cardioid microphone boosts bass at close range because of the near-field term of the pressure gradient in a spherical wave; the ideal model gives larger boosts than real microphones. Cutting the distance tenfold raises direct sound, signal-to-noise ratio and direct-to-reverberant ratio by about 20 dB each; in an ordinary room, 50 cm is already near the critical distance. The intimacy of close-range recording may come from the bass, the dry sound and the preserved detail, and from our everyday experience of being spoken to from close by, but its effect on trust still needs experimental testing. The next chapter returns to classical texts. (https://5dlibrary.com/answers/en/acoustic-persuasion-synchrony/c06/) 第6章 镜头与麦克风前:近讲效应、信噪比与混响 — 心形麦克风在近距离时会放大低频,这是声压梯度在球面波中的近场项造成的;理想模型给出的提升比真实麦克风大。距离缩短十倍,直达声、信噪比和直达声与混响声之比都提高约 20 dB;在普通房间里,50 厘米已经接近临界距离。近距离录音的亲密感可能来自低频、干声和细节的保留,以及我们对「近距离说话」的生活经验,但它对信任的影响还需要实验检验。下一章回到古典文献。 (https://5dlibrary.com/answers/zh/acoustic-persuasion-synchrony/c06/) - Ch.8 Practice and a Pre-registered Trial — In this fictional recording, 6 of the 9 statements end with a fall of more than 1 semitone and the other 3 end level or slightly rising; the one question ends rising, as is fitting. Only 3 of the 5 key points are preceded by a pause of 0.3 seconds or more. The mean rate is about 5.5 syllables a second (about 330 a minute), with a coefficient of variation between sentences of about 11 %, not much variation. Moving the microphone from 40 cm to 10 cm raises the voice level by 12 dB and the signal-to-noise ratio from 34 to 46 dB. (https://5dlibrary.com/answers/en/acoustic-persuasion-synchrony/c08/) 第8章 实践与预注册试验 — 在这段虚构的录音中,9 个陈述句里有 6 个以超过 1 个半音的下降结尾,另外 3 个以平调或轻微上扬结尾;唯一的问句以上扬结尾,这是恰当的。5 个关键点中只有 3 个之前有 0.3 秒以上的停顿。平均语速约为每秒 5.5 个音节(每分钟约 330 个),句子之间的变化系数约为 11%,变化不大。麦克风从 40 厘米移到 10 厘米后,人声电平提高 12 dB,信噪比从 34 dB 提高到 46 dB。 (https://5dlibrary.com/answers/zh/acoustic-persuasion-synchrony/c08/) ## 5DL-B-019 · Non-Ergodic Wealth Dynamics, the Kelly Criterion and Volatility Drag 中文书名:非遍历性财富动力学、凯利准则与波动率税 DOI: https://doi.org/10.5281/zenodo.23754709 Cite: Fan, Y. (2026). Non-Ergodic Wealth Dynamics, the Kelly Criterion and Volatility Drag (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23754709 Book page: https://5dlibrary.com/books/ergodicity-kelly-wealth/ Scope: This book is not investment, financial, tax or legal advice. Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. Abstract: In a multiplicative world, a choice with a positive expected return is not necessarily good in the long run for a person or a firm that lives only once. This book examines that distinction with probability theory, stochastic processes and financial economics, and asks which popular claims about it hold. It covers a coin game with an expected return of +5 % a round and a time-average growth rate of about −5.13 %; ensemble and time averages and ergodicity; Peters' ergodicity economics, its relation to Bernoulli's logarithmic utility, and economists' replies; geometric Brownian motion and Itô's lemma, the long-run growth rate μ − σ²/2 and, with leverage, r + L(μ − r) − L²σ²/2, twice-Kelly leverage falling back to the risk-free rate, and the path dependence of daily-reset leveraged products; the Kelly criterion f* = (bp − q)/b and (μ − r)/σ², half Kelly keeping about 75 % of the growth, the best multiple m²/(m² + s²) below one when the edge is estimated from finite data, and the probability x^(2/c−1) of ever falling to x times the start; first-passage times of Brownian motion with drift and absorbing barriers, cash buffers lowering the probability of ruin exponentially, and the effect of cutting rigid costs in a downturn; Shannon's demon and the rebalancing bonus (about one-eighth of the variance) with its limits under transaction costs and trends, and the diversification return; the frequency of extreme events under fat tails, and an honest barbell model that caps the largest loss but bleeds slowly when its high-payoff part is overpriced. The eight chapters are read beside Sima Qian's "Biographies of the Money-Makers" in the Shiji: "without capital, labour; with a little, wits; when rich, timing", Jiran's "prepare in advance", "in drought, invest in boats; in floods, carts" and "when the dear reaches its extreme it turns cheap", Bai Gui's "take what others discard, let go what others seek" and the Wuyan family's tenfold interest, noting the limits of such a reading. They end with a four-step survival-and-compounding checklist for household and business cash flow and allocation, and a pre-registered randomised controlled behavioural experiment comparing an arithmetic and an ergodic display of the same asset on the leverage chosen in a repeated multiplicative investment task (235 analysable and 277 recruited per arm, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. This book is not investment, financial, tax or legal advice. 摘要:在乘性的世界里,期望收益为正的选择,未必对一个只能活一次的人或企业长期有利。本书用概率论、随机过程与金融经济学考察这一区别,以及围绕它的流行说法哪些成立。依次讨论:一个每轮期望 +5%、时间平均却约 −5.13% 的抛硬币游戏,集合平均与时间平均、遍历性,彼得斯的遍历性经济学与伯努利对数效用的关系及经济学家的回应;几何布朗运动与伊藤引理,长期增长率 μ − σ²/2 与杠杆下的 r + L(μ − r) − L²σ²/2,两倍凯利杠杆回到无风险利率,每日重置杠杆产品的路径依赖;凯利准则 f* = (bp − q)/b 与 (μ − r)/σ²,半凯利保留约 75% 的增长,参数由有限数据估计时最优倍数 m²/(m² + s²) 小于 1,以及「曾经跌到起点 x 倍」的概率 x^(2/c−1);带漂移布朗运动的首达时间与吸收壁,现金缓冲按指数降低破产概率,寒冬中削减刚性成本的作用;香农恶魔与再平衡溢价(约为方差的八分之一)及其在交易成本和趋势下的边界、分散化收益;肥尾分布下的极端事件频率,以及一个诚实的杠铃模型:限定最大损失,但高赔率部分定价偏高时长期流血。 八章与司马迁《史记·货殖列传》「无财作力,少有斗智,既饶争时」、计然「知斗则修备」「旱则资舟,水则资车」「贵上极则反贱,贱下极则反贵」、白圭「人弃我取,人取我与」以及无盐氏的十倍利息并读,并指出并读的限度;最后给出一份个人与企业现金流配置的四步生存与复利核查表,以及一项比较算术展示与遍历展示对重复乘性投资任务中所选杠杆影响的预注册随机对照行为实验方案(每组 235 人可分析、招募 277 人,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部核对)、原典与作者的假说。本书不是投资、财务、税务或法律建议。 Chapter answers: - Ch.1 The Ensemble Average Is Not the Time Average — In a multiplicative world the expected value describes the average of infinitely many parallel players, and the time average describes your own single path. The coin game's expectation rises by 5 % a round while almost every player's wealth shrinks by about 5.13 % a round. The log growth rate is ergodic and links the two; Bernoulli's logarithmic utility and Peters' time average give the same decisions for multiplicative gambles, and their theoretical status is still debated. Staking only part of one's wealth turns the same gamble from certain decline into steady growth. The next chapter moves to continuous time and uses Itô's lemma to show how volatility itself levies a tax on the growth rate. (https://5dlibrary.com/answers/en/ergodicity-kelly-wealth/c01/) 第1章 集合平均不等于时间平均 — 在乘性的世界里,期望值描述的是无限多个平行玩家的平均,时间平均描述的是你自己这一条路径。抛硬币游戏的期望每轮 +5%,几乎每个玩家的财富却以每轮约 −5.13% 的速度萎缩。对数增长率是遍历的,它把两者联系起来;伯努利的对数效用与彼得斯的时间平均在乘性赌局上给出相同的决策,二者的理论地位仍有争论。只押一部分财富,同一个赌局就可以从必然萎缩变成稳定增长。下一章把这个问题放到连续时间里,用伊藤引理说明波动本身怎样从增长率中扣税。 (https://5dlibrary.com/answers/zh/ergodicity-kelly-wealth/c01/) - Ch.2 Geometric Brownian Motion, Itô's Lemma and Volatility Drag — Itô's lemma shows that in geometric Brownian motion expected wealth grows at μ while the typical path grows at μ − σ^2/2: volatility is itself a tax. Leverage raises the expected return linearly and the volatility tax quadratically, so the long-run growth rate is a parabola in leverage, with an optimum at L* = (μ − r)/σ^2; beyond 2L* one does worse than taking no risk at all. The next chapter starts from this optimum, with the Kelly criterion and why one should bet less when the parameters are uncertain. (https://5dlibrary.com/answers/en/ergodicity-kelly-wealth/c02/) 第2章 几何布朗运动、伊藤引理与波动率税 — 伊藤引理说明,在几何布朗运动中,期望财富以 μ 增长,而典型的路径以 μ − σ^2/2 增长:波动本身就是一种税。杠杆让期望收益线性上升、让波动率税平方上升,长期增长率因此是杠杆的一条抛物线,存在一个最优点 L* = (μ − r)/σ^2,超过 2L* 就不如完全不承担风险。下一章从这个最优点出发,讨论凯利准则以及在参数不确定时为什么应该下注得更少。 (https://5dlibrary.com/answers/zh/ergodicity-kelly-wealth/c02/) - Ch.3 The Kelly Criterion and Fractional Kelly — The Kelly criterion gives the betting fraction that maximises long-run growth, f* = (bp − q)/b, or (μ − r)/σ^2 in continuous form. Half Kelly keeps about three-quarters of the growth with half the volatility; twice Kelly has zero excess growth. Full Kelly's drawdowns are very deep: the probability of ever being halved is 50 %. More important still, the expected return can be estimated only roughly; betting full Kelly on the estimate can have an expected result worse than taking no risk, and the best multiple falls as the precision of the estimate falls. The next chapter discusses another limit: when there is rigid spending that must be paid every month, "zero" is not merely a mathematical limit but an absorbing barrier. (https://5dlibrary.com/answers/en/ergodicity-kelly-wealth/c03/) 第3章 凯利准则与分数凯利 — 凯利准则给出使长期增长率最大的下注比例 f* = (bp − q)/b,连续形式为 (μ − r)/σ^2。半凯利保留约四分之三的增长、只有一半的波动;两倍凯利的超额增长为零。全凯利的回撤极深:财富曾经腰斩的概率是 50%。更重要的是,期望收益只能被粗略地估计,按估计值下全凯利,期望结果可能比不承担风险更差;最优的倍数随估计精度下降而下降。下一章讨论另一种限制:当你有每个月都必须支付的刚性支出时,「零」不只是一个数学极限,而是一道吸收壁。 (https://5dlibrary.com/answers/zh/ergodicity-kelly-wealth/c03/) - Ch.4 Absorbing Barriers and First-Passage Times — Rigid spending turns "zero" into an absorbing barrier: touch it once and all future returns are no longer yours. A Brownian motion with drift and the first-passage formula show that a cash buffer reduces the probability of ruin non-linearly: the first few months are worth the most, and the probability of ever being ruined falls exponentially with the buffer. Once winters are added, the buffer must also cover sustained losses over a period; and cutting rigid costs improves both the ordinary drift and the depth of the winter: a lasting cut of 5 a month does more than a one-off extra buffer of 90, though the same total (240) put into the buffer at the start does better still. The next chapter takes another angle: if fluctuation is unavoidable, can something be gained from it? (https://5dlibrary.com/answers/en/ergodicity-kelly-wealth/c04/) 第4章 破产吸收壁与首达时间 — 刚性支出让「零」成为一道吸收壁:碰到一次,所有未来的收益都与你无关。用带漂移的布朗运动与首达时间公式可以算出,现金缓冲对破产概率的作用是非线性的:最初几个月的缓冲最值钱,「曾经破产」的概率随缓冲按指数下降。加入寒冬之后,缓冲还必须覆盖一段时间里的持续亏损;而削减刚性成本同时改善平常的漂移与寒冬的深度:每月少 5 的长期削减,作用大于一次性多出 90 的缓冲,尽管把同样总额(240)一次性放进缓冲效果更好。下一章换一个角度:如果波动是不可避免的,能不能反过来从波动中得到一些东西? (https://5dlibrary.com/answers/zh/ergodicity-kelly-wealth/c04/) - Ch.5 Shannon's Demon and the Rebalancing Bonus — Each asset alone has a long-run growth rate of only 1.5 % (6 % less a 4.5 % volatility tax); uncorrelated, the half-and-half rebalanced mix grows at 3.75 %, and the lower the correlation the higher the rate. The difference is called the diversification return, and Fernholz and Shay gave its general form in stochastic portfolio theory (Fernholz 1982). But when the two assets have different drifts (10 % and 2 %, or 14 % and −2 %), the buy-and-hold portfolio gradually concentrates in the stronger asset and ends up ahead of the rebalanced one. (https://5dlibrary.com/answers/en/ergodicity-kelly-wealth/c05/) 第5章 香农恶魔与再平衡溢价 — 两种资产单独持有的长期增长率都只有 1.5%(6% 减去 4.5% 的波动率税);不相关时,各一半再平衡的增长率是 3.75%,相关性越低越高。这个差额叫作分散化收益,费恩霍尔兹与谢伊在随机投资组合理论中给出了它的一般形式 (Fernholz 1982)。但当两种资产的漂移不同(10% 与 2%,或 14% 与 −2%)时,买入并持有的组合逐渐集中到较强的资产上,最终超过再平衡的组合。 (https://5dlibrary.com/answers/zh/ergodicity-kelly-wealth/c05/) - Ch.6 Tail Risk, Fat Tails and the Barbell — The tails of asset returns are much heavier than the normal distribution's, with tail exponents usually between 2 and 5; for the same variance, the frequency of extreme events can differ by many orders of magnitude. This makes the volatility tax heavier, Kelly leverage lower and the absorbing barrier closer. A barbell can cap the largest loss, but its long-run growth depends on the pricing of the high-payoff part; insurance-like positions usually have negative average returns, so a barbell often trades growth for a limited worst case. The next chapter reads the conclusions of the first six chapters beside the ways of managing wealth in the Shiji's "Biographies of the Money-Makers". (https://5dlibrary.com/answers/en/ergodicity-kelly-wealth/c06/) 第6章 尾部风险、肥尾分布与杠铃策略 — 资产收益的尾部比正态分布厚得多,尾指数通常在 2 到 5 之间;在同样的方差下,极端事件的频率可以相差许多个数量级。这让波动率税更重、凯利杠杆更低、吸收壁更近。杠铃式配置能限定最大损失,但它的长期增长取决于高赔率部分的定价;买入保险类头寸的平均收益通常为负,杠铃因此往往是在用增长率换取有限的最坏情况。下一章把前六章的结论与《史记·货殖列传》中的治生之道并读。 (https://5dlibrary.com/answers/zh/ergodicity-kelly-wealth/c06/) - Ch.8 Practice and a Pre-registered Experiment — The book's conclusions reduce to four questions: how far are you from the absorbing barrier; on how many independent risks is your wealth staked; are your rebalancing rules written down; and how optimistic is your estimate of returns? This chapter demonstrated the checklist on two fictional cases and set out a pre-registered randomised controlled behavioural experiment comparing the effect of arithmetic and ergodic displays on chosen leverage in a repeated multiplicative investment task, with 235 analysable and 277 recruited per arm, and the decision rule and opposing hypothesis fixed in advance. (https://5dlibrary.com/answers/en/ergodicity-kelly-wealth/c08/) 第8章 实践与预注册试验 — 全书的结论可以归结为四个问题:你离吸收壁有多远;你的财富押在多少个独立的风险上;你的再平衡规则有没有写下来;你对收益的估计有多乐观。本章用两个虚构的案例演示了这份核查表,并给出一项预注册的随机对照行为实验方案:比较算术展示与遍历展示对重复乘性投资任务中所选杠杆的影响,每组 235 人可分析、招募 277 人,判定规则与反方向假设都事先写定。 (https://5dlibrary.com/answers/zh/ergodicity-kelly-wealth/c08/) ## 5DL-B-020 · Credit Cycles, Self-Organised Criticality and the Statistical Physics of Market Phase Transitions 中文书名:信贷周期、自组织临界性与市场相变的统计物理 DOI: https://doi.org/10.5281/zenodo.23754711 Cite: Fan, Y. (2026). Credit Cycles, Self-Organised Criticality and the Statistical Physics of Market Phase Transitions (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23754711 Book page: https://5dlibrary.com/books/credit-cycles-criticality/ Scope: The book is not advice to buy or sell any security, index or option, and it predicts no market level or date. Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked; all prices and volatilities in the book are simulated or invented), classical sources and the author's hypotheses. Abstract: After a long calm, a piece of bad news that is not large in itself sets off a collapse far out of proportion to it: the pattern recurs throughout financial history. This book uses statistical physics and financial economics to ask how much of it can be made precise with models and how much is a good story told afterwards. It covers Minsky's hedge, speculative and Ponzi finance and the Goodwin growth cycle and Keen debt model (from nearby starting points the same model can converge to finite debt or move towards exploding debt, so the clearing of credit cycles is no "arithmetic necessity"), set against cross-country historical evidence that credit booms raise the probability of crises; the avalanche-size distribution of the Bak–Tang–Wiesenfeld sandpile, the difference between maximum-likelihood and log–log least-squares estimates of power-law exponents, and the "volatility paradox" and crises after low volatility; thresholds and common-holder contagion in Brunnermeier–Pedersen liquidity spirals, and the Forbes–Rigobon correction to "correlations go to one in a crisis"; GARCH volatility clustering, the underestimation of kurtosis in finite samples and the clustered failure of constant-volatility value at risk; Black–Scholes implied volatility by Newton's method, the skew produced by a jump-diffusion model (a risk-neutral price, not a crash probability), and mean reversion of a volatility index with its small-sample bias; and detection and false alarms of critical-slowing-down warnings and the over-fitting of log-periodic power laws to random walks with no bubble. The eight chapters are read beside the Book of Changes' hexagram Tai ("no level without a slope, no going without a return"; "the wall falls back into the moat"), the Daodejing's "reversal is the movement of the Way" and "things that grow strong grow old" and the Discourses on Salt and Iron's debate on price stabilisation, noting the limits of such a reading. They end with a four-step checklist for observing the state of the market and the fragility of one's own portfolio and a pre-registered randomised controlled experiment comparing a usual average-annual-return disclosure with a tail disclosure adding historical maximum drawdowns and fat-tailed simulated paths, on the share allocated to leveraged products (no direction assumed, two-sided test, 235 analysable and 277 recruited per arm, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked; all prices and volatilities in the book are simulated or invented), classical sources and the author's hypotheses. The book is not advice to buy or sell any security, index or option, and it predicts no market level or date. 摘要:一段长期平静之后,一个并不算大的坏消息触发远超其分量的崩塌:这种模式在金融史上反复出现。本书用统计物理与金融经济学考察其中有多少可以用模型说清楚,哪些只是事后的好故事。依次讨论:明斯基的对冲型、投机型与庞氏型融资,以及古德温增长周期与基恩债务模型(同一模型从相近起点出发,可以收敛到有限债务,也可以走向债务爆炸,信贷出清不是「算术必然」),并对照跨国历史数据中信贷繁荣提高危机概率的证据;巴克–汤–维森菲尔德沙堆模型的雪崩规模分布,最大似然法与双对数最小二乘法估计幂律指数的差别,以及「波动率悖论」与低波动之后的危机;布伦纳迈尔–佩德森流动性螺旋的阈值与共同持有者传染,以及福布斯–里戈本对「危机中相关性趋近 1」的校正;GARCH 波动率聚类、有限样本峰度的低估与固定波动率风险价值的成群失灵;布莱克–斯科尔斯隐含波动率的牛顿法反推、跳跃–扩散模型产生的偏度(风险中性价格,不是崩盘概率),以及波动率指数的均值回归与短样本偏差;临界慢化预警的检出与误报,以及对数周期幂律对无泡沫随机游走的过度拟合。 八章与《周易·泰卦》「无平不陂,无往不复」「城复于隍」、《道德经》「反者道之动」「物壮则老」及《盐铁论·本议》关于平准的辩论并读,并指出并读的限度;最后给出一份观察市场状态与自身仓位脆弱性的四步核查表,以及一项比较「常规平均年化收益率披露」与「加上历史最大回撤和厚尾模拟路径的尾部披露」对杠杆产品配置比例影响的预注册随机对照实验方案(不预设方向,双侧检验,每组 235 人可分析、招募 277 人,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部核对;书中的价格与波动率数据均为模拟或虚构)、原典与作者的假说。本书不构成任何证券、指数或期权的买卖建议,不预测任何市场点位或时间。 Chapter answers: - Ch.2 Sandpiles and the Great Moderation: Self-Organised Criticality — But the figure also shows two problems. First, on a finite grid avalanche size has an upper limit set by the size of the grid, so the large end is cut off sharply and the distribution is not a straight line. Second, the estimated exponent rises from 1.30 to 1.41 as the lower limit of the fit, $s_{\min}$, rises from 2 to 10. The program gives both the exact discrete maximum-likelihood estimate (normalised with the Hurwitz zeta function) and the approximation of Clauset and colleagues; that approximation is accurate only for lower limits of about 6 or more, and for these data the two happen to agree. Values reported in the literature for the avalanche-size exponent of the two-dimensional sandpile are not consistent, and its scaling properties are still debated. "Is it a power law?" and "what is the exponent?" are not questions that a glance at a log–log plot can answer. (https://5dlibrary.com/answers/en/credit-cycles-criticality/c02/) 第2章 沙堆与大缓和:自组织临界性 — 但图中也能看到两个问题。第一,在有限的网格中,雪崩规模有一个由网格大小决定的上限,大雪崩一端急剧截断,分布并不是一条直线。第二,估计出的指数随着拟合的下限 $s_{\min}$ 从 2 提高到 10,从 1.30 升到 1.41。程序同时给出精确的离散最大似然估计(用赫尔维茨 ζ 函数归一化)与克劳塞特等人的近似公式;那个近似只在下限约为 6 以上时才准确,在这组数据中两者恰好一致。文献中二维沙堆模型雪崩规模指数的报告值不尽一致,它的标度性质至今仍有争议。「是不是幂律」「指数是多少」并不是看一眼双对数图就能回答的问题。 (https://5dlibrary.com/answers/zh/credit-cycles-criticality/c02/) - Ch.3 Liquidity Spirals: Margins, Fire Sales and Correlation — The link between the two markets never changes, yet the correlation on turbulent days (about 0.77) is twice that on calm days (about 0.37); looking only at days when A moves a lot, it reaches about 0.85. Forbes and Rigobon's correction (https://5dlibrary.com/answers/en/credit-cycles-criticality/c03/) 第3章 流动性螺旋:保证金、抛售与相关性 — 两个市场之间的联系从未改变,但动荡日的相关系数(约 0.77)是平静日(约 0.37)的两倍;只看 A 大幅波动的日子,相关系数高达约 0.85。福布斯与里戈本的校正公式 (https://5dlibrary.com/answers/zh/credit-cycles-criticality/c03/) - Ch.4 Volatility Clustering and GARCH — The direction of asset returns is hard to predict, but the size of moves comes in clusters and the distribution has fat tails. The GARCH(1,1) model reproduces these features with very few parameters. Simulation shows that kurtosis is markedly underestimated in finite samples, and that a value at risk computed from a constant-volatility normal model looks close to its nominal level overall yet is breached frequently and in clusters during volatile periods. The next chapter looks at how the market prices these tail risks, through option prices. (https://5dlibrary.com/answers/en/credit-cycles-criticality/c04/) 第4章 波动率聚类与 GARCH — 资产收益率方向难以预测,但波动的大小会成群出现,分布有厚尾。GARCH(1,1) 模型用很少的参数重现了这些特征。模拟显示,有限样本中的峰度会被明显低估;用固定波动率的正态模型计算的风险价值整体上看起来接近名义水平,却在高波动时期频繁而成群地被突破。下一章从期权价格的角度看市场怎样为这些尾部风险定价。 (https://5dlibrary.com/answers/zh/credit-cycles-criticality/c04/) - Ch.5 Implied Volatility, Skew and the Fear Index — Implied volatility is the volatility that makes the Black–Scholes formula match a market price, and it can be backed out with Newton's method. After 1987 index options showed a persistent skew, and a price process with downward jumps can produce such a shape. The skew is a risk-neutral price that includes a risk premium and cannot be read directly as a crash probability. Volatility indices tend to revert to the mean, but short samples overstate the speed of reversion, and the commonly used IV rank and percentile can give very different readings. The next chapter discusses two more ambitious approaches: recognising the signals of a crisis before it comes. (https://5dlibrary.com/answers/en/credit-cycles-criticality/c05/) 第5章 隐含波动率、偏度与恐慌指数 — 隐含波动率是使布莱克–斯科尔斯公式与市场价格一致的波动率,可以用牛顿法反推。1987 年之后,指数期权出现了持久的偏度;含向下跳跃的价格过程可以产生这种形态。偏度是风险中性的价格,包含风险溢价,不能直接读作崩盘概率。波动率指数有均值回归的倾向,但短样本会高估回归速度,常用的 IV Rank 与百分位可以给出很不相同的读数。下一章讨论两种更有野心的方法:在危机来临之前识别它的信号。 (https://5dlibrary.com/answers/zh/credit-cycles-criticality/c05/) - Ch.6 The Early-Warning Controversy: Critical Slowing Down and Log-Periodic Power Laws — Critical slowing down is an early-warning indicator with a solid theoretical basis: as a system slowly approaches a critical point, autocorrelation and variance rise. Simulation shows that under ideal conditions it detects well but still gives false alarms, more of them the shorter the data; abrupt changes triggered by random shocks, with no change in the system's parameters, leave no such signal. In empirical studies of financial crises, there was evidence of critical slowing down only before the 1987 crash, with mixed results for the others. Log-periodic power laws fit random walks containing no bubble at all very well and produce arbitrary "critical times". Fitting after the event is easy; timing in advance is statistically very unreliable. The next chapter reads these modern discussions beside ancient writing on rise and decline. (https://5dlibrary.com/answers/en/credit-cycles-criticality/c06/) 第6章 预警信号的争议:临界慢化与对数周期幂律 — 临界慢化是一个有坚实理论基础的预警指标:系统缓慢接近临界点时,自相关和方差会上升。模拟显示,在理想条件下它的检出率高,但仍有误报;数据越短,误报越多;由随机冲击触发、系统参数并没有变化的突变,不会留下这种信号。金融危机的实证研究中,只有 1987 年股灾之前有临界慢化的证据,其他几次危机的结果不一致。对数周期幂律能很好地拟合完全没有泡沫的随机游走,并给出任意的「临界时间」。事后拟合容易,事前择时在统计上极不可靠。下一章把这些现代讨论与古代关于盛衰的文字并读。 (https://5dlibrary.com/answers/zh/credit-cycles-criticality/c06/) - Ch.8 Practice and a Pre-Registered Trial — With a smallest effect of interest of 0.30 standard deviations and 90 % power, each arm needs 235 analysable participants; allowing for about 15 % of participants to be excluded for failing attention or comprehension checks, each arm recruits 277, 554 in all. (https://5dlibrary.com/answers/en/credit-cycles-criticality/c08/) 第8章 实践与预注册试验 — 按最小关注效应 0.30 个标准差、90% 把握度计算,每组需要 235 名可分析的参与者;考虑约 15% 的参与者因未通过注意力或理解检查而被排除,每组招募 277 人,共 554 人。 (https://5dlibrary.com/answers/zh/credit-cycles-criticality/c08/) ## 5DL-B-021 · Computational Attachment Theory and Couple Interaction Dynamics 中文书名:依恋理论的计算模型与伴侣互动动力学 DOI: https://doi.org/10.5281/zenodo.23754715 Cite: Fan, Y. (2026). Computational Attachment Theory and Couple Interaction Dynamics (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23754715 Book page: https://5dlibrary.com/books/attachment-couple-dynamics/ Scope: Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses. Abstract: One partner wants to talk, the other wants to get away; the more one presses, the quieter the other becomes. Many couples know this loop that spins faster and faster. This book uses attachment theory, dynamical systems and longitudinal research to ask where it comes from, how far early experience determines it, and what can slow it down. It covers Bowlby's and Ainsworth's attachment theory and the two continuous dimensions of adult attachment, anxiety and avoidance (cutting them into "types" makes about a quarter of people change label on retest), and, with the internal working model written as a Bayesian prior, the difference between bias in the prior, extra weight on negative evidence and too little sampling; coupled rate equations for the pursue–withdraw loop (at high enough sensitivity, self-sustaining escalation and periodic eruptions appear, and lowering either partner's sensitivity has symmetric effects); the steady states and basins of attraction of Gottman and Murray's marriage difference equations, the difference between Gottman's descriptive "about 5 to 1" and Losada's withdrawn "critical positivity ratio", and why divorce prediction needs cross-validation; physiological linkage and co-regulation in conflict, the threshold of escalating coupling, and spurious linkage from a shared situation; the real effect sizes of attachment stability (meta-analytic average r = 0.39, explaining about 15 % of the variance, no longer significant over intervals longer than 15 years), how relationships may revise working models, and the dispute over "earned security"; and repair attempts as damping (making and accepting multiply) and the physiology of taking a break. The eight chapters are read beside the Book of Changes hexagram Xian ("the two energies respond to each other and draw together", "the noble person receives others with emptiness") and the hexagram Jiaren, the two readings of "in death or life, near or far, I made my pledge with you" from the Book of Songs, the Great Learning's "once the self is cultivated the family is ordered" and Mencius's "whenever an action fails, seek the cause in oneself", stating clearly the boundary between hierarchical classical family ethics and modern equal partnerships. They end with a four-step interaction-loop checklist and a pre-registered couple-randomised controlled trial comparing a structured session (pursue–withdraw loop visualisation, a 20-minute physiological cool-down and repair phrases) with a standard communication guide (primary outcome escalation time in a one-week conflict diary, 130 analysable and 163 recruited couples per arm, separate and private intimate-partner-violence screening, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses. This book is not marital or family therapy or counselling; anyone in a relationship with violence or control should seek safety and professional help first. 摘要:一方想谈、另一方想躲,一方越追问、另一方越沉默——许多伴侣熟悉这种越转越快的回路。本书用依恋理论、动力系统与纵向研究考察它从哪里来、早期经验在多大程度上决定它、以及什么能让它慢下来。依次讨论:鲍尔比与安斯沃思的依恋理论与成人依恋的焦虑、回避两个连续维度(把它们切成「类型」会使约四分之一的人在重测时换标签),以及把内部工作模型写成贝叶斯先验时,先验、对负面证据的加权与采样不足三种偏差的区别;追—逃回路的耦合速率方程(反应敏感度足够高时出现自我维持的升级与周期性爆发,降低任何一方的敏感度效果对称);戈特曼与默里婚姻差分方程的稳态与吸引域,戈特曼「约 5 比 1」的描述性观察与已被撤回的洛萨达「临界正性比例」的区别,以及离婚预测为什么需要交叉验证;冲突中的生理联动与共调、升级耦合的阈值,以及共同处境造成的假联动;依恋稳定性的真实效应量(元分析平均 r = 0.39,只解释约 15% 的变异,间隔超过 15 年不再显著)、关系对工作模型的可能修正与「后天习得安全」的争议;修复尝试作为阻尼(做出与被接收相乘)与暂停冷静的生理学。 八章与《周易》咸卦「二气感应以相与」与「君子以虚受人」、家人卦、《诗经·击鼓》「死生契阔,与子成说」的两种读法、《大学》「身修而后家齐」与《孟子》「行有不得者,皆反求诸己」并读,并明确古代等级制家庭伦理与现代平等伴侣关系的界限;最后给出一份四步互动回路核查表,以及一项比较「追—逃回路可视化 + 20 分钟生理降温 + 修复用语」结构化会谈与常规沟通指南的预注册伴侣随机对照试验方案(主要结局为一周冲突日记中的升级时长,每组 130 对可分析、招募 163 对,含分别私下进行的亲密伴侣暴力筛查,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(14 段程序,全部核对)、原典与作者的假说。本书不是婚姻家庭治疗或心理咨询;存在暴力或控制的关系请优先寻求安全与专业帮助。 Chapter answers: - Ch.2 Coupled Equations of the Pursue–Withdraw Loop — First, at low sensitivity ($g = 0.4$ or $0.6$) the system returns to calm by itself after a trigger: there is only one equilibrium, and it is calm. (https://5dlibrary.com/answers/en/attachment-couple-dynamics/c02/) 第2章 追—逃回路的耦合方程 — 第一,敏感度低时($g = 0.4$ 或 $0.6$),一次触发之后系统会自己回到平静:只有一个平衡点,而且是平静的。 (https://5dlibrary.com/answers/zh/attachment-couple-dynamics/c02/) - Ch.3 Gottman and Murray's Marriage Equations — Gottman and Murray wrote turn-by-turn coding of conflict conversations as difference equations, characterising a couple's interaction structure by uninfluenced set points, emotional inertia and influence functions. When negative influence is strong enough, the model has two stable states, positive and negative, and how a conversation starts decides which it falls toward. "About 5 to 1" is Gottman's descriptive observation of stable couples' conflict discussions; "5.6 to 1" comes from research on business teams and "2.9013" from a withdrawn derivation, and the three must be kept apart. Prediction accuracies need cross-validation. The next chapter turns to the body in conflict: heart rate, skin conductance and physiological linkage between partners. (https://5dlibrary.com/answers/en/attachment-couple-dynamics/c03/) 第3章 戈特曼与默里的婚姻方程 — 戈特曼与默里把冲突对话的逐话轮编码写成差分方程,用不受影响的设定点、情绪惯性与影响函数刻画一对伴侣的互动结构。当负面影响足够强时,模型会出现正面与负面两个稳定状态,对话怎样开始决定它落向哪一个。「约 5 比 1」是戈特曼对稳定伴侣冲突对话的描述性观察;「5.6 比 1」来自商业团队研究,「2.9013」是已被撤回的推导,三者必须分开。预测准确率需要交叉验证。下一章讨论冲突中的身体:心率、皮肤电与伴侣之间的生理联动。 (https://5dlibrary.com/answers/zh/attachment-couple-dynamics/c03/) - Ch.4 The Body in Conflict: Physiological Linkage and Co-regulation — The model gives three results. First, co-regulation ($k = 0.10$) lowers partner 1's peak and speeds recovery (18 minutes against 23 alone), at the price that partner 2 is briefly carried along a little: co-regulation is not "I am unaffected" but "we work it off together". Second, mutual escalation slows recovery, and the stronger the coupling the slower. Third, there is a clear threshold: when $c$ exceeds $1/\tau$, the "together" mode grows instead of decaying, and although the stressor is long over, arousal keeps climbing. This fits, in form, Levenson and Gottman's observation that dissatisfied couples are physiologically tightly linked in conflict, but the model itself cannot show that this is the mechanism. (https://5dlibrary.com/answers/en/attachment-couple-dynamics/c04/) 第4章 冲突中的身体:生理联动与共调 — 模型给出三个结果。第一,共调($k = 0.10$)使伴侣 1 的峰值降低、回落加快(18 分钟对独自一人的 23 分钟),代价是伴侣 2 也短暂地被带动了一些:共调不是「我不受影响」,而是「我们一起把它消化掉」。第二,相互升级使回落变慢;耦合越强越慢。第三,存在一个清楚的阈值:当 $c$ 超过 $1/\tau$,「两人一起」的模式不再衰减而是增长,压力源早已结束,唤醒却越来越高。这与莱文森与戈特曼观察到的「不满意的伴侣在冲突中生理紧密联动」在形式上是一致的,但模型本身不能证明这就是机制。 (https://5dlibrary.com/answers/zh/attachment-couple-dynamics/c04/) - Ch.5 Is Early Experience Destiny? The Real Effect Sizes of Longitudinal Studies — The average stability correlation of attachment security from infancy to early adulthood is about 0.39, explaining about 15 % of the variance, and over intervals longer than 15 years it is no longer significant. Early care predicts adult attachment, but with small effects. Attachment styles change; a partner's way of responding may be one source of change, but causal evidence is still scarce; retrospective "earned security" may be affected by depression-related recall bias. Early experience matters, but it is not destiny. The next chapter returns to conflict itself: what makes it stop. (https://5dlibrary.com/answers/en/attachment-couple-dynamics/c05/) 第5章 早期经验是宿命吗:纵向研究的真实效应量 — 依恋安全性从婴儿期到成年早期的平均稳定性相关约为 0.39,只能解释约 15% 的变异,间隔超过 15 年时不再显著。早期照料能预测成年依恋,但效应不大。依恋风格会变,伴侣的回应方式可能是改变的来源之一,但因果证据仍少;回顾性的「后天习得安全」可能受到与抑郁有关的回忆偏差影响。早期经验有影响,但不是宿命。下一章回到冲突本身:是什么让冲突停下来。 (https://5dlibrary.com/answers/zh/attachment-couple-dynamics/c05/) - Ch.6 Repair Attempts: The Damper in the Conflict Loop — The model shows two things. First, making a repair and having it accepted multiply: when one partner tries very hard to repair ($p_{make} = 0.8$) and the other seldom accepts ($p_{accept} = 0.2$), more than six in ten conflicts still end flooded; when both make and accept only moderately (0.5 and 0.5), the flooded share falls to about a quarter. Second, the change from "mostly flooded" to "mostly settled" happens over a fairly narrow band, products of about 0.1 to 0.4: in the model there is a threshold-like relation between damping and the slope of escalation. What decides how a conflict ends is not whether there is conflict, but whether repair can cut off the slope of escalation. (https://5dlibrary.com/answers/en/attachment-couple-dynamics/c06/) 第6章 修复尝试:冲突回路中的阻尼器 — 模型显示两点。第一,做出修复与修复被接收是相乘的关系:一方非常努力地修复($p_{make} = 0.8$)而对方很少接收($p_{accept} = 0.2$),冲突仍有六成以上以淹没告终;双方都只是中等程度地做出和接收(0.5 与 0.5),淹没的比例降到约四分之一。第二,从「大多淹没」到「大多平息」的转变发生在乘积约 0.1 到 0.4 之间一个不宽的区间里:在模型中,阻尼与升级斜率之间存在一个类似阈值的关系。决定冲突结局的,不是有没有冲突,而是升级的斜率能不能被修复截断。 (https://5dlibrary.com/answers/zh/attachment-couple-dynamics/c06/) - Ch.8 Practice and a Pre-registered Trial — With a smallest effect of interest of 0.35 standard deviations (of an individual partner's score), a correlation of 0.5 between partners' diaries and 90 % power, each arm needs 130 analysable couples; allowing for about 20 % loss to drop-out and unusable diaries, each arm recruits 163 couples, 326 couples and 652 people in all. The more closely partners' diaries agree, the less averaging helps and the larger the sample needed; the table gives the numbers for correlations of 0.3 and 0.7. (https://5dlibrary.com/answers/en/attachment-couple-dynamics/c08/) 第8章 实践与预注册试验 — 按最小关注效应 0.35 个标准差(以单个伴侣的得分计)、伴侣两人日记的相关 0.5、90% 把握度计算,每组需要 130 对可分析的伴侣;考虑约 20% 的流失与无法使用的日记,每组招募 163 对,共 326 对伴侣、652 人。两人日记的相关越高,取平均带来的好处越小,所需样本越多;表中列出了相关为 0.3 与 0.7 时的数字。 (https://5dlibrary.com/answers/zh/attachment-couple-dynamics/c08/) ## 5DL-B-022 · Early Experience, Epigenetic Methylation and the Limits of Intergenerational Transmission 中文书名:早期经验、表观遗传甲基化与跨代行为先验的传递边界 DOI: https://doi.org/10.5281/zenodo.23761337 Cite: Fan, Y. (2026). Early Experience, Epigenetic Methylation and the Limits of Intergenerational Transmission (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23761337 Book page: https://5dlibrary.com/books/epigenetics-intergenerational-priors/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. Abstract: "The trauma of your family of origin is written into your DNA and passed down the generations." This book checks the evidence behind that popular claim and asks where the resemblance between family members actually comes from and how it can change. It covers the maternal licking-and-grooming studies and the methylation of the exon 1₇ promoter of the hippocampal glucocorticoid-receptor gene, and the key fact shown by cross-fostering: the resemblance in styles of care between generations follows the rearing mother, not the birth mother, and is rebuilt through behaviour in each generation; the Dutch famine and Holocaust-offspring studies as intergenerational evidence from intrauterine exposure or with pathways that cannot be separated, the inconclusive evidence for transgenerational epigenetic inheritance in humans, and a strict distinction between "intergenerational" and "transgenerational" by type of exposure (F3 after exposure in pregnancy, F2 after exposure of a father or a non-pregnant mother); Falconer's formulas and maximum-likelihood ACE models for twin studies, the "three laws" of behavioural genetics and fifty years of pooled twin data (average heritability about 49 %, shared environment about 0.17), the rise in the heritability of intelligence with age, and how heritability changes with socioeconomic status, an interaction that is inconsistent outside the United States; how passive, evocative and active gene–environment correlations lead observational studies to overestimate the effect of the family environment (1.6- to 3.8-fold with illustrative parameters), and research on "genetic nurture"; equations for vertical, oblique and horizontal transmission in dual-inheritance theory (an illustrative half-life of about 2.4 generations for high-fidelity vertical transmission); crossover tests of differential susceptibility and "orchids and dandelions", statistical power and the failure of candidate-gene findings to replicate; and the real strength of negative intergenerational cycles (about a third; meta-analytic d = 0.45), detection bias in official records, and evidence for supportive partners, mentors and structured interventions as buffers, with a Markov model with filtering. The eight chapters are read beside the Analects' "by nature close, by habituation far apart", the Family Instructions of the Yan Clan ("shifted unseen and transformed in secret"), the image of the room of orchids in the School Sayings of Confucius and the Records of Rites of the Elder Dai, the Elder Dai's "what is formed in youth is like nature itself", the story of Mencius's mother in the Biographies of Exemplary Women and Yan Zhitui's account of prenatal instruction, correcting three classical attributions and explicitly not taking over the hierarchy, gender norms and corporal punishment of classical family instruction. They end with a four-step intergenerational-channel checklist and a pre-registered, assessor-blind randomised controlled trial comparing a brief parent training ("noticing one's own automatic stress reactions plus modelling a pause") with time-matched reading on child development (children aged 6–10; single primary outcome the share of intervals of child emotional dysregulation in a frustrating cooperative puzzle task, coded blind from video; 198 analysable dyads and 233 recruited families per arm; coding-reliability threshold and child-protection procedures pre-specified; decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. This book is not psychotherapy or a parenting diagnosis; genetic and epigenetic research cannot be used to label any family or individual, and heritability is a population statistic that says neither how much of a person "is determined by genes" nor that a trait cannot change. 摘要:「原生家庭的创伤会写进 DNA,一代一代传下去」——本书核对这一流行说法背后的证据,并追问家庭成员之间的相似究竟从哪里来、又能怎样被改变。依次讨论:母鼠舔舐理毛与海马糖皮质激素受体基因外显子 1₇ 启动子甲基化的研究,以及交叉抚养显示的关键事实——照料风格的代际相似跟随抚养者而不是生母,是通过行为每一代重新建立的;荷兰饥荒与大屠杀幸存者后代研究属于宫内暴露或无法区分途径的代际证据,人类跨代表观遗传的证据尚无定论,并按暴露方式严格区分「代际」与「跨代」(孕期暴露要到 F3、父亲或未孕母亲暴露要到 F2);双生子研究的法尔科纳公式与最大似然 ACE 模型、行为遗传学「三大定律」与五十年双生子研究的汇总(平均遗传力约 49%、共享环境约 0.17),智力遗传力随年龄上升,以及遗传力随社会经济地位改变、且这一交互作用在美国以外的样本中不一致;被动、唤起与主动三型基因—环境相关如何让观察性研究高估家庭环境的作用(示意参数下放大 1.6—3.8 倍),以及「遗传养育」研究;双重继承理论中垂直、斜向与水平传递的方程(高保真垂直传递的示意半衰期约 2.4 代);差异易感性与「兰花与蒲公英」的交叉点检验、检验力与候选基因研究的重复失败;负向代际循环的真实强度(约三成、元分析 d = 0.45)、官方记录中的检测偏倚,以及支持性伴侣、导师与结构化干预作为缓冲的证据与带滤波的马尔可夫模型。 八章与《论语》「性相近,习相远」、《颜氏家训·慕贤》「潜移暗化,自然似之」、《孔子家语·六本》与《大戴礼记·曾子疾病》的芝兰之喻、《大戴礼记·保傅》「少成若天性」、《列女传》孟母三迁与《颜氏家训·教子》胎教之说并读,更正了三处古文出处,并明确不取古代家训中的等级、性别规范与体罚;最后给出一份四步代际信道核查表,以及一项比较父母「自动化应激反应觉察 + 暂停示范」微训练与时长匹配的常规育儿知识阅读的预注册评估者盲随机对照试验方案(6—10 岁儿童,唯一主要结局为合作拼图受挫任务中经盲法录像编码的儿童情绪失调区间比例,每组 198 对可分析、招募 233 个家庭,编码信度门槛与儿童保护程序预先规定,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序,全部核对)、原典与作者的假说。本书不是心理治疗或育儿诊断;遗传与表观遗传研究不能用来给任何家庭或个人定性,遗传力是群体统计量,不说明个人「多少由基因决定」,也不意味着不可改变。 Chapter answers: - Ch.1 Is Trauma "Written into DNA"? Animal Experiments and Human Evidence — Methylation changes marks on DNA, not its sequence. In the rat studies the resemblance in styles of care follows the rearing mother, not the birth mother, and is rebuilt through behaviour each generation; the Dutch famine is evidence of intrauterine exposure, and the Holocaust-offspring study is small, shows opposite directions and cannot separate pathways; human transgenerational epigenetic inheritance is not established. The first unexposed generation is F3 after exposure in pregnancy and F2 after exposure of a father or a non-pregnant mother. The next chapter turns to a larger source of resemblance: genes. (https://5dlibrary.com/answers/en/epigenetics-intergenerational-priors/c01/) 第1章 「创伤写进 DNA」是真的吗:动物实验与人类证据 — 甲基化改变的是 DNA 上的标记而不是序列。母鼠研究中,照料风格的代际相似跟随抚养者而不是生母,是通过行为每一代重建的;荷兰饥荒是宫内暴露的证据,大屠杀幸存者研究样本小、方向相反、无法区分途径;人类跨代表观遗传的证据尚无定论。孕期暴露要到 F3、父亲或未孕母亲暴露要到 F2,才是第一代未受直接暴露的后代。下一章讨论更大的相似来源:基因。 (https://5dlibrary.com/answers/zh/epigenetics-intergenerational-priors/c01/) - Ch.2 Twin Studies and the "Three Laws" of Behavioural Genetics — Falconer's formulas are simple but can give negative values, so maximum likelihood is preferable; in moderate samples "C is not significant" is often a matter of power. Fifty years of data give an average heritability of about 49 % and a shared environment of about 0.17; the heritability of intelligence rises with age, and the shared environment is not small for some childhood behaviour problems or adolescent religiousness. Heritability changes with the environment: the SES interaction appears in US samples but not in Western Europe or Australia. Heritability is a population statistic, not a personal proportion, and does not mean a trait cannot change. (https://5dlibrary.com/answers/en/epigenetics-intergenerational-priors/c02/) 第2章 双生子研究与行为遗传学的「三大定律」 — 法尔科纳公式简单但会给出负值,最大似然拟合更合适;中等样本中「C 不显著」常常只是检验力不足。五十年的汇总显示平均遗传力约 49%、共享环境约 0.17;智力的遗传力随年龄上升,而在儿童期的部分行为问题与青少年的宗教性上共享环境并不小。遗传力随环境变化,SES 交互作用在美国样本中存在、在西欧与澳大利亚不存在。遗传力是群体统计量,不是个人的比例,也不意味着不可改变。 (https://5dlibrary.com/answers/zh/epigenetics-intergenerational-priors/c02/) - Ch.3 Gene–Environment Correlation: The Confound in "Blaming the Family of Origin" — Biological parents provide both the home and half the genes. Measures of "the environment" have heritabilities of 7–39 %, and about 23 % of the variance in parenting is associated with the children's genes; with illustrative parameters, observational studies can inflate the family effect 1.6- to 3.8-fold and show a "family effect" even when the true effect is zero, and adoption designs remove only passive confounding. Genetic-nurture research shows that parents' non-transmitted genes act through the home, at about 30 % of the transmitted effect. The next chapter looks at cultural channels beyond the family. (https://5dlibrary.com/answers/en/epigenetics-intergenerational-priors/c03/) 第3章 基因—环境相关:「原生家庭归因」的混淆 — 亲生父母同时提供家庭与一半的基因。「环境」测量有 7%—39% 的遗传力,养育行为约 23% 的变异与孩子的基因有关;路径模型显示,示意参数下观察性研究可把家庭作用放大 1.6—3.8 倍,真实作用为零时也会出现「家庭效应」,收养设计只能去掉被动型混淆。遗传养育研究显示,父母未传递的基因经由家庭环境起作用,约为传递部分效应的 30%。下一章转向家庭之外的文化信道。 (https://5dlibrary.com/answers/zh/epigenetics-intergenerational-priors/c03/) - Ch.4 Dual Inheritance: How Social-Learning Channels Keep and Rewrite Behavioural Priors — Cultural transmission has vertical, oblique and horizontal channels. The transmission equation shows that a lineage's deviation from the community equilibrium is multiplied by $\lambda = V(2f-1)$ each generation: high-fidelity vertical transmission keeps a behavioural prior for several generations, while peer- and mentor-dominated channels rewrite it within one or two, and a new mentor in one generation changes the lineage's course for good. Trust and language among immigrants' descendants give observational examples of persistence and of rapid rewriting. The next chapter asks whether the same environment affects every child equally. (https://5dlibrary.com/answers/en/epigenetics-intergenerational-priors/c04/) 第4章 双重继承:社会学习信道如何保持与改写行为先验 — 文化传递有垂直、斜向与水平三种信道。传递方程显示,家庭偏离社群平衡的部分每代乘以 $\lambda = V(2f-1)$:高保真垂直传递可以让行为先验保持数代,以同伴与导师为主则在一两代内被改写;某一代引入新的导师,可以持久地改变谱系的轨迹。移民后代的信任与语言研究分别提供了持久与快速改写的观察性例证。下一章讨论同样的环境是否对每个孩子影响一样。 (https://5dlibrary.com/answers/zh/epigenetics-intergenerational-priors/c04/) - Ch.5 Differential Susceptibility: Orchids, Dandelions and the Crossover Test — With a true interaction of 0.10, a study of 100 children detects the interaction only about 18 % of the time and meets the full pattern only about 6 % of the time; with 300 children, 43 % and 23 %; only with 1,000 children do the figures reach 89 % and 69 %. By the normal approximation, about 1,051 children are needed for 90 % power to detect such an interaction. Part 2 is more worrying: with 10 completely null candidate markers, about 43 % of studies find at least one "significant" interaction, and these "significant" interactions average about 0.17, larger than the real interaction above. (https://5dlibrary.com/answers/en/epigenetics-intergenerational-priors/c05/) 第5章 差异易感性:兰花、蒲公英与交叉点检验 — 当真实的交互作用为 0.10 时,100 名儿童的研究只有约 18% 的机会检出交互项,满足完整模式的只有约 6%;300 名儿童分别为 43% 与 23%;1,000 名儿童才达到 89% 与 69%。按正态近似,要以 90% 的把握度检出这样的交互作用,约需 1,051 名儿童。第二部分更值得警惕:在 10 个完全无效的候选标记中,约 43% 的研究会至少找到一个「显著」的交互作用,而这些「显著」交互作用的平均大小约为 0.17,比上面那个真实存在的交互作用还大。 (https://5dlibrary.com/answers/zh/epigenetics-intergenerational-priors/c05/) - Ch.6 Breaking Negative Cycles: Partners, Mentors and Structured Interventions as Channel Filters — The model shows several things. First, even without any buffer the cycle attenuates by itself: starting from a harsh parent, the risk is 0.30 in the next generation, 0.16 in the second and 0.13 in the third, quickly approaching the long-run level of 0.125, because only $\lambda = 0.2$ of the excess risk is kept in each generation. Second, the higher the buffer coverage, the lower the first generation's risk (0.174 at a coverage of 0.9) and the faster the attenuation, and the long-run prevalence falls from 0.125 to 0.108. Third, even without a buffer, 70 % of the children of harsh parents do not become harsh parents. (https://5dlibrary.com/answers/en/epigenetics-intergenerational-priors/c06/) 第6章 阻断负向循环:伴侣、导师与结构化干预作为信道滤波器 — 这个模型显示了几件事。第一,即使没有任何缓冲,循环本身也在衰减:从一位严厉的父母开始,下一代的风险是 0.30,第二代降到 0.16,第三代 0.13,很快接近长期水平 0.125,因为每一代的超额风险都只保留 $\lambda = 0.2$。第二,缓冲覆盖率越高,第一代的风险越低(覆盖率 0.9 时为 0.174),衰减也越快,长期患病率从 0.125 降到 0.108。第三,即使没有缓冲,严厉父母的孩子中也有 70% 不会成为严厉的父母。 (https://5dlibrary.com/answers/zh/epigenetics-intergenerational-priors/c06/) - Ch.8 Practice and a Pre-Registered Trial — The four-step checklist turns earlier conclusions into four observable links: strip out genetic confounding, audit non-verbal demonstration, configure peers and mentors, and replace the automatic reaction under high stress; the fictional example shows that two weeks of records are confounded and offer only leads. The pre-registered trial compares a brief "awareness plus pause demonstration" training with time-matched reading on child development, with one primary outcome, the blind-coded share of intervals of child dysregulation; it needs 198 analysable dyads and 233 recruited families per arm; coding reliability, delivery fidelity and child-protection procedures are pre-specified, and results will be published whatever they show. (https://5dlibrary.com/answers/en/epigenetics-intergenerational-priors/c08/) 第8章 实践与预注册试验 — 四步核查表把前几章的结论落到四个可观察的环节:剥离基因混淆、审计非语言示范、配置同伴与导师、替换高压下的自动化反应;虚构示例显示两周的记录也受混淆,只能作为线索。预注册试验比较父母的「觉察 + 暂停示范」微训练与时长匹配的育儿知识阅读,唯一主要结局为盲法编码的儿童情绪失调区间比例,每组 198 对可分析、招募 233 个家庭;编码信度、实施忠实度与儿童保护程序预先规定,不论结果正负都将发表。 (https://5dlibrary.com/answers/zh/epigenetics-intergenerational-priors/c08/) ## 5DL-B-023 · The 85% Rule for Optimal Learning, Myelination Kinetics and Deliberate Practice 中文书名:学习的85%最优错误率定律、髓鞘化动力学与刻意练习 DOI: https://doi.org/10.5281/zenodo.23761339 Cite: Fan, Y. (2026). The 85% Rule for Optimal Learning, Myelination Kinetics and Deliberate Practice (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23761339 Book page: https://5dlibrary.com/books/eighty-five-percent-rule-myelination/ Scope: Marked boxes separate evidence, runnable Python calculations (15 programs), classical sources and the author's hypotheses. Abstract: Practice that is too easy teaches little; practice that is too hard breaks down. This book uses the learning sciences, neuroscience and runnable models to ask where "just hard enough" lies: Wilson and colleagues' 2019 "Eighty Five Percent Rule" (for binary-classification learners that learn by gradient descent under Gaussian noise, learning per trial is fastest at a training error rate of ½ erfc(1/√2) ≈ 15.87 %; the optimum is flat and moves with the noise, about 18.4 % for Laplacian and 25 % for Cauchy noise), its assumptions, and why Vygotsky's zone of proximal development is an analogy rather than the same theorem; deliberate practice and "10,000 hours", meta-analyses in which deliberate practice explains on average about 12 % of individual differences (games 26 %, music 21 %, sports 18 %, education 4 %, professions under 1 %), the replication of the 1993 study and the dispute over definitions; Hebbian plasticity, long-term potentiation and activity-dependent myelination, why diffusion-imaging measures are not myelin, and the cable-theory relation between g-ratio and conduction velocity; forgetting curves, the spacing effect, retrieval practice (61 % against 40 % after a week), the ACT-R activation model and an SM-2 scheduler; interleaving, desirable difficulties and a cue-competition model; and the guidance hypothesis of feedback and the more than one third of feedback interventions that make performance worse. The chapters are read beside the Analects, the Record of Learning and Zhu Xi's "advance step by step, read until familiar and reflect with care", making clear that the classical authors were concerned with moral self-cultivation, not experiments. Corrections after checking: the optimal error rate is ½ erfc(1/√2), not ½(1 − 1/√2) (about 0.146); "10,000 hours" is not a rule from the 1993 paper; 0.77 is a theoretical, not measured, g-ratio; the "six principles" of Zhu Xi's method of reading were set out by Zhang Hong and Qi Xi on the basis of Fu Guang's earlier compilation, not by Fu Guang himself. The book ends with a four-step deliberate-practice checklist and a pre-registered randomised trial comparing "adaptive difficulty at about 85 % plus interleaved retrieval practice" with "about 95 % accuracy plus blocked practice" (adults learning an artificial symbolic rule system; primary outcome a transfer test after 14 days; 198 analysable and 233 recruited per arm, 466 in all; stated to be a combined intervention, with a 2 × 2 factorial variant). Marked boxes separate evidence, runnable Python calculations (15 programs), classical sources and the author's hypotheses. The book is not a guide to diagnosing or treating learning disabilities, ADHD or any condition; "85 %" is an optimum within a particular model, not a universal constant. 摘要:太简单学不会,太难会崩溃:最有效的练习落在「刚好够难」的地方。本书用学习科学、神经科学与可运行的模型考察这个「之间」在哪里:威尔逊等人 2019 年的「85% 法则」(用梯度下降学习的二分类学习者在高斯噪声下,训练错误率为 ½ erfc(1/√2) ≈ 15.87% 时每次试次的学习增益最大;最优区平坦,噪声不同则最优值不同,拉普拉斯约 18.4%、柯西 25%),它的前提,以及它与维果茨基最近发展区只是类比;刻意练习与「一万小时」,元分析中刻意练习平均解释约 12% 的个体差异(游戏 26%、音乐 21%、体育 18%、教育 4%、职业不到 1%),1993 年研究的重复与定义之争;赫布可塑性、长时程增强与活动依赖性髓鞘化,弥散成像指标不等于髓鞘,电缆理论中的 g 比与传导速度;遗忘曲线、间隔效应、检索练习(一周后 61% 对 40%)、ACT-R 激活模型与 SM-2 调度器;交错练习与「必要难度」及线索竞争模型;反馈的指导假说与三分之一以上使表现变差的反馈干预。 八章与《论语》「学而不思则罔」「不愤不启,不悱不发」、《礼记·学记》与朱熹「循序而渐进,熟读而精思」并读,并说明古人讲的是修身成德,不是认知心理学实验。经核对修正:最优错误率公式应为 ½ erfc(1/√2),而非 ½(1 − 1/√2)(约 0.146);「一万小时」不是 1993 年论文提出的规则;0.77 是理论而非实测的 g 比;「朱子读书法」六条由张洪、齐熙在辅广旧辑的基础上釐定,而非辅广归纳。最后给出四步刻意练习核查表与一项预注册随机对照试验方案:「约 85% 自适应难度 + 交错提取练习」对「约 95% 正确率 + 集中练习」,成年人学习人工符号推理规则,主要结局为 14 天后的迁移测验,每组 198 人可分析、招募 233 人,共 466 人;写明是组合干预并给出 2 × 2 析因变体的样本量。书中以标注框区分证据、可运行的 Python 计算(15 段程序)、原典与作者的假说。本书不是针对学习障碍、ADHD 或任何状况的诊断或治疗建议;「85%」是特定模型下的最优值,不是普适常数。 Chapter answers: - Ch.1 Too Easy to Learn, Too Hard to Cope: The 85 % Rule — Wilson and colleagues proved that for binary-classification learners that learn by gradient descent, under Gaussian noise, the per-trial learning gain is greatest when precision times difficulty equals 1, that is at a training error rate of about 15.87 % (about 84.13 % correct); the correct formula is ½ erfc(1/√2), not ½(1 − 1/√2). The optimum is flat, it moves when the shape of the noise changes, and in simulated learners it also falls within a band. The result rests on assumptions (adjustable difficulty, binary decisions, a particular noise, gradient learning) and cannot be generalised into an "85 % law of all learning". Vygotsky's zone of proximal development is a related idea, but not the same theorem. The next chapter turns to "deliberate practice" in research on human expertise: how much of the difference between people can the quantity and quality of practice explain? (https://5dlibrary.com/answers/en/eighty-five-percent-rule-myelination/c01/) 第1章 为什么太简单学不会、太难会崩溃:85% 法则 — 威尔逊等人证明:对于用梯度下降学习的二分类学习者,在高斯噪声下,让精度与难度的乘积等于 1,也就是训练错误率约 15.87%(正确率约 84.13%)时,每次试次的学习增益最大;正确的公式是 ½ erfc(1/√2),不是 ½(1 − 1/√2)。最优区相当平坦,噪声形状不同则最优值不同,模拟学习者的最优点也落在一个区间里。这个结果依赖于难度可调、二元判断、特定噪声与梯度学习等前提,不能直接推广为「一切学习的 85% 定律」。维果茨基的最近发展区是相关的思想,但不是同一个定理。下一章讨论人类专长研究中的「刻意练习」:练习的量与质,各能解释多少差别。 (https://5dlibrary.com/answers/zh/eighty-five-percent-rule-myelination/c01/) - Ch.2 Deliberate Practice and "10,000 Hours" — Ericsson and colleagues introduced deliberate practice in 1993, stressing goal-directed practice with feedback aimed at weaknesses; "10,000 hours" is Gladwell's popularisation, which Ericsson himself criticised. Meta-analyses find that deliberate practice explains on average about 12 % of individual differences (games 26 %, music 21 %, sports 18 %, education 4 %, professions under 1 %), and about 1 % in elite sports samples; the replication of the 1993 study found no difference between the "best" and the "good"; and the dispute between the two sides is largely about definitions. "Explains little variance" does not mean "practice does not work": synthetic data show the two are fully compatible, and range restriction and the way practice is measured can move the figures a great deal. The next chapter turns to the brain: what does practice change in the nervous system? (https://5dlibrary.com/answers/en/eighty-five-percent-rule-myelination/c02/) 第2章 刻意练习与「一万小时」 — 埃里克森等人 1993 年提出刻意练习的概念,强调有目标、有反馈、针对弱点的练习;「一万小时」是格拉德威尔的通俗化,埃里克森本人批评过它。元分析发现,刻意练习平均解释约 12% 的个体差异(游戏 26%、音乐 21%、体育 18%、教育 4%、职业不到 1%),精英体育样本中约 1%;1993 年研究的重复没有得到「最好」与「好」之间的差别;双方的争论很大程度上是定义之争。「解释的方差小」不等于「练习没用」:合成数据显示两者完全相容,而全距限制与测量方式会让数字大幅变化。下一章转向大脑:练习改变了神经系统的什么? (https://5dlibrary.com/answers/zh/eighty-five-percent-rule-myelination/c02/) - Ch.3 Hebbian Plasticity, Long-Term Potentiation and Activity-Dependent Myelination — Hebb's postulate and long-term potentiation form the classical picture of synaptic plasticity. Research over the past fifteen years or so shows that myelin, too, can change with neural activity: optogenetic stimulation in mice promotes myelination, and blocking the production of new oligodendrocytes impairs motor learning; juggling training in humans changes white-matter DTI measures, but DTI measures are not myelin. Cable theory gives a velocity proportional to diameter for myelinated fibres and an optimal g-ratio of about 0.6 at fixed fibre diameter; including volume constraints, Chomiak and Hu obtain a theoretical value of about 0.77 for the central nervous system, and our simplified model shows that the optimal thickness depends on the constraints. That myelin tunes conduction times and promotes synchrony is an attractive but unproven idea. The next chapter returns to behaviour: how memory fades over time, and how spacing and retrieval change that process. (https://5dlibrary.com/answers/en/eighty-five-percent-rule-myelination/c03/) 第3章 赫布可塑性、长时程增强与活动依赖性髓鞘化 — 赫布的假设与长时程增强构成了突触可塑性的经典图景。过去十几年的研究表明,髓鞘也可以随神经活动而改变:小鼠的光遗传学刺激会促进髓鞘形成,而阻断新少突胶质细胞的产生会妨碍运动学习;人类的杂耍训练会改变白质的 DTI 指标,但 DTI 指标不等于髓鞘。电缆理论给出有髓纤维的速度与直径成正比、总直径固定时最优 g 比约 0.6,钱米亚克与胡斌在考虑体积约束后得到中枢的理论值约 0.77;我们的简化模型说明最优厚度取决于约束。髓鞘调节传导时间、促进信号同步,是一个有理论吸引力但尚待证实的想法。下一章回到行为:记忆如何随时间衰退,间隔与提取又如何改变这一过程。 (https://5dlibrary.com/answers/zh/eighty-five-percent-rule-myelination/c03/) - Ch.4 Forgetting Curves, the Spacing Effect and Retrieval Practice — Forgetting is fast at first and slower later, and power functions often describe it better than exponentials. Spaced study beats massed study; the optimal gap grows with the time for which material must be retained, but shrinks as a proportion of it. Retrieval practice is worse than rereading at the time and clearly better a week later (Roediger and Karpicke, Experiment 2: 61 % against 40 %); the fluency of rereading leads people to overestimate their memory. The ACT-R activation model explains the spacing effect by "study while fresh, decay faster"; schedulers such as SM-2 turn these principles into a daily review list. The next chapter discusses another practice that is "harder now, better later": interleaving. (https://5dlibrary.com/answers/en/eighty-five-percent-rule-myelination/c04/) 第4章 遗忘曲线、间隔效应与提取练习 — 遗忘起初快、后来慢,幂函数常比指数函数描述得更好。间隔学习优于集中学习,最优间隔随需要保持的时间变长,但占它的比例变小。检索练习在当下不如重读,一周后却明显更好(罗迪格与卡尔匹克实验二:61% 对 40%);重读的流畅感会让人高估自己的记忆。ACT-R 激活模型用「新鲜时再学,衰减更快」解释间隔效应;SM-2 等调度器把这些原理变成每天的复习清单。下一章讨论另一种「当下更难、长期更好」的做法:交错练习。 (https://5dlibrary.com/answers/zh/eighty-five-percent-rule-myelination/c04/) - Ch.5 Interleaved versus Blocked: Desirable Difficulties — Bjork's "desirable difficulties" distinguish performance in practice from long-term learning. In Rohrer and Taylor's small experiment, blocked practice scored 89 % against 60 % for interleaving during practice, but 20 % against 63 % a week later. A meta-analysis finds an overall interleaving effect of g = 0.42, varying with the material: largest for paintings, moderate for mathematics, and with blocking better for words; the more similar the categories, the more interleaving helps. A cue-competition model shows that the interleaving effect appears in error-driven learners but not in an ideal Bayesian learner. The next chapter turns to another element of practice: the timing and grain of feedback. (https://5dlibrary.com/answers/en/eighty-five-percent-rule-myelination/c05/) 第5章 交错与集中:必要难度 — 比约克的「必要难度」区分了练习时的表现与长期的学习。罗勒与泰勒的小样本实验中,集中练习时 89% 对交错 60%,一周后却是 20% 对 63%。元分析显示交错学习的总体效应为 g = 0.42,但依材料而异:绘画最大,数学中等,词语上集中反而更好;类别越相似,交错越有利。线索竞争模型显示,交错效应出现在误差驱动的学习者身上,不出现在理想贝叶斯学习者身上。下一章讨论练习的另一个要素:反馈的时机与粒度。 (https://5dlibrary.com/answers/zh/eighty-five-percent-rule-myelination/c05/) - Ch.6 The Timing and Grain of Feedback — The guidance hypothesis holds that external feedback on every trial improves performance in practice but may weaken the learner's processing of its own errors; Winstein and Schmidt found that reduced feedback frequency helped retention, but a later meta-analysis did not give stable support. Kluger and DeNisi found an average effect of feedback interventions of d = 0.41, with more than a third making performance worse; feedback directed at the self is especially likely to harm. The model suggests that the key may be not the frequency of feedback but whether the learner processes its own error information. Immediate correction suits safety, errors that would be repeated many times, and the beginning; delayed or reduced feedback suits the stage at which self-assessment is to be developed. The next chapter reads this modern research beside classical Chinese thought on learning. (https://5dlibrary.com/answers/en/eighty-five-percent-rule-myelination/c06/) 第6章 反馈的时机与粒度 — 指导假说认为,每次都给外部反馈会提高练习时的表现,却可能削弱学习者对自身误差的加工;温斯坦与施密特发现减少反馈频率有利于保持,但后来的元分析没有得到稳定的支持。克鲁格与德尼西发现反馈干预的平均效应为 d = 0.41,却有三分之一以上使表现变差;指向自我的反馈尤其可能有害。模型显示,关键或许不在于反馈的频率,而在于学习者是否加工自己的误差信息。立即纠错适用于安全、错误会被大量重复与初学阶段;延迟或减少反馈适用于需要发展自我评估的阶段。下一章把这些现代研究与中国古代的学习思想并读。 (https://5dlibrary.com/answers/zh/eighty-five-percent-rule-myelination/c06/) - Ch.8 Practice and a Pre-Registered Trial — The fictional record shows the following. Monday and Tuesday are entirely blocked, at 100 % and 94 % accuracy, with about 2.5 to 3.5 times as much time rereading and watching explanations as actively solving: smooth practice that may teach little. On Wednesday, the first interleaved session, accuracy drops to 69 %; on Thursday and Saturday it holds at 81 %, with retrieval-to-review ratios of about 2.7 to 1 and 3.5 to 1. On Friday, blocked practice on a new type runs at only 25 %, with a two-minute wait for answers and only 1 of 12 errors corrected: difficulty should be lowered first. Over the week, 10 of 24 errors were corrected and redone (42 %). A week's record can only suggest adjustments; it is not evidence. (https://5dlibrary.com/answers/en/eighty-five-percent-rule-myelination/c08/) 第8章 实践与预注册试验 — 这份虚构记录显示:周一、周二全是集中练习,正确率 100% 与 94%,重读和看讲解的时间是主动做题的约 2.5 到 3.5 倍,练得很顺,学到的可能不多;周三第一次交错,正确率掉到 69%;周四、周六稳定在 81%,提取与浏览之比约 2.7 比 1 和 3.5 比 1;周五集中练习一类新题,正确率只有 25%,要等两分钟才拿到答案,12 个错误只纠正了 1 个,说明需要先降低难度。全周 24 个错误中 10 个得到纠正并重做(42%)。一周的记录只能提供调整的线索,不是证据。 (https://5dlibrary.com/answers/zh/eighty-five-percent-rule-myelination/c08/) ## 5DL-B-024 · Attention Networks, the Working-Memory Bottleneck and Self-Determination Theory 中文书名:注意力网络、工作记忆瓶颈与内在动机的自决定理论 DOI: https://doi.org/10.5281/zenodo.23761341 Cite: Fan, Y. (2026). Attention Networks, the Working-Memory Bottleneck and Self-Determination Theory (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23761341 Book page: https://5dlibrary.com/books/attention-networks-intrinsic-motivation/ Scope: This book is not diagnosis or treatment advice for ADHD, anxiety, depression or any other condition, and deals with no medication or dose. Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. Abstract: Two planned hours cut into a dozen pieces by messages; bonuses and targets that drain the pleasure from something one used to enjoy; a task of exactly the right difficulty that makes time disappear. This book uses research on attention and motivation to examine the mechanisms and evidence behind these three everyday experiences. It covers Posner and Petersen's three networks of alerting, orienting and executive control (their links to the noradrenergic, cholinergic and dopaminergic systems are principal modulators, not sole mechanisms) and the subtraction scores of the Attention Network Test (uniform slowing leaves the scores unchanged; in pooled data from 1,129 adults the split-half reliabilities of the alerting and orienting scores were only 0.38 and 0.55, and the networks are not independent); working-memory capacity revised from "7 ± 2" to about 4 chunks, the redefinition of "germane" load in cognitive load theory, and how expert schemas compress 26 letters into 4 slots; preparation effects and residual costs in task switching, attention residue, and interruptions of about 2.8 and 4.4 seconds that roughly doubled and tripled sequence errors at the next step, with the clarification that the "23 minutes" is only the average time before interrupted work was resumed (25 min 26 s for same-day resumptions) and that "a glance at a message wrecks 15 minutes" has no direct evidence; crowding out by extrinsic rewards, which holds for expected tangible rewards contingent on engagement, completion or performance and for interesting tasks (128 experiments, d about −0.28 to −0.40), not for unexpected rewards, while positive feedback raises intrinsic motivation, with a Bayesian signalling model of why "being paid to do it" can be read as "it isn't worth doing for its own sake"; the different predictions of additive and multiplicative models of the needs for autonomy, competence and relatedness (the multiplicative form is the author's assumption); and the eight-channel model of flow, the limited evidence for the hypofrontality hypothesis, and a learner model whose difficulty is tuned to 84–85 % accuracy (an analogy only; outcome entropy peaks at 50 %, not at 85 %). Chapter 7 reads Cheng Yi's "what is called reverence is called holding to one; what is called one is called not going off elsewhere", Zhu Xi's "only gather in body and mind, orderly and unified, not letting oneself go: that is reverence", the Xunzi's "empty, unified and still" and the Zhuangzi's "his will undivided, he was concentrated like a spirit" beside executive control, attention residue and flow, stating clearly that reverence was a discipline of moral cultivation, not a technique of attention training. Chapter 8 offers a four-step attention-and-motivation checklist and a pre-registered randomised crossover trial of the "brain drain" hypothesis (own phone face down and silenced on the desk versus locked in another room; single primary outcome the automated operation span absolute score, with the Sustained Attention to Response Task secondary; periods at least 7 days apart with parallel forms; 160 completers, 190 randomised; decision rule written as code), noting that replications and meta-analyses since the original study disagree. Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. This book is not diagnosis or treatment advice for ADHD, anxiety, depression or any other condition, and deals with no medication or dose. 摘要:计划好的两个小时被消息切成十几段,奖金和考核让本来喜欢的事变得索然无味,难度恰好的任务又让人忘了时间——本书用注意与动机研究考察这三种日常经验背后的机制与证据。依次讨论:波斯纳与彼得森的警觉、定向、执行控制三网络(与去甲肾上腺素、乙酰胆碱、多巴胺系统的对应只是主要调质,不是唯一机制)以及注意网络测验的减法计分(整体变慢不改变分数;1,129 人的汇总数据中警觉与定向分数的分半信度只有 0.38 与 0.55,三个网络并不独立);工作记忆容量从「7 ± 2」修正为约 4 个组块,认知负荷理论中「相关负荷」的重新定义,以及专家图式如何把 26 个字母压缩进 4 个槽位;任务切换的准备效应与残余代价、注意力残留,约 2.8 秒和 4.4 秒的打断使下一步序列错误率约增至两倍和三倍,并澄清「23 分钟」只是被打断工作的平均恢复时间(同日恢复者平均 25 分 26 秒),「看一眼消息毁掉 15 分钟」没有直接证据;外在奖励的挤出效应只在预期的、取决于参与、完成或表现的有形奖励和有趣任务上成立(128 项实验,d 约 −0.28 至 −0.40),意外奖励无影响,言语正反馈提高内在动机,并用贝叶斯信号模型说明「付钱让我做」为什么会被读成「这件事本身不值得」;自主、胜任、归属三项需要的加性与乘性两种模型(乘性是作者的假设)的不同预测;心流的八通道模型、前额叶功能低下假说的有限证据,以及按 84%–85% 目标正确率调节难度的学习者模型(只是类比;结果熵在 50% 时最大,不在 85%)。 第七章将程颐「所谓敬者,主一之谓敬;所谓一者,无适之谓一」、朱熹「只收敛身心,整齐纯一,不恁地放纵,便是敬」、《荀子·解蔽》「虚壹而静」与《庄子·达生》「用志不分,乃凝于神」与执行控制、注意力残留和心流并读,并明确主敬是道德修养工夫,不是注意力训练法;第八章给出一份四步注意力与动机核查表,以及一项检验「手机在场即耗脑力」假说的预注册随机交叉试验方案(手机静音朝下放在桌上与锁在另一房间对比;唯一主要结局为自动化操作广度测验绝对分,持续注意反应任务为次要结局;两期间隔至少 7 天、使用平行版本;完成两期 160 人、随机化 190 人;判定规则写成代码),并说明原始研究之后的重复研究与元分析结果并不一致。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序,全部核对)、原典与作者的假说。本书不是 ADHD、焦虑、抑郁等任何疾病的诊断或治疗建议,不涉及任何药物或剂量。 Chapter answers: - Ch.1 Attention Is Not a Single Tap: The Three-Network Model and the Attention Network Test — With these illustrative parameters the program's reliabilities (expected: alerting 0.36, orienting 0.50, conflict 0.83; simulated: 0.36, 0.56, 0.84) are close to MacLeod and colleagues' values, which shows only that the explanation is plausible: a difference inherits the measurement error of two conditions, while people may genuinely differ little on that difference. To bring the alerting score up to a reliability of 0.7, these parameters require about 192 trials per condition in each half, four times the session length assumed here. (https://5dlibrary.com/answers/en/attention-networks-intrinsic-motivation/c01/) 第1章 注意力不是一个水龙头:三网络模型与注意网络测验 — 在这组示意参数下,程序得到的信度(理论值:警觉 0.36、定向 0.50、冲突 0.83;模拟值:0.36、0.56、0.84)与麦克劳德等人的数值接近,但这只说明这种解释说得通:一个差值继承了两个条件的测量误差,而人与人之间在这个差值上的真实差异可能本来就不大。要把警觉分数的信度提高到 0.7,在这组参数下每半需要约 192 个试次每种条件,是这里假定的标准长度的四倍。 (https://5dlibrary.com/answers/zh/attention-networks-intrinsic-motivation/c01/) - Ch.2 The Working-Memory Bottleneck: Chunks, Schemas and Cognitive Load — Working-memory capacity is not "7 ± 2" but, once rehearsal and chunking are excluded, about 3 to 5 chunks; the difference between chunks and bits is exactly what Miller himself stressed. Experts' advantage comes mainly from schemas in long-term memory, which let them compress many items into a few chunks; when material and schemas do not match, the advantage disappears. Cognitive load theory distinguishes intrinsic from extraneous load; in its later revision, germane load is understood as resources allocated to intrinsic load rather than a third, additive load. The next chapter asks where the costs come from when attention switches back and forth between tasks. (https://5dlibrary.com/answers/en/attention-networks-intrinsic-motivation/c02/) 第2章 工作记忆的瓶颈:组块、图式与认知负荷 — 工作记忆的容量不是「7 ± 2」,而是在排除复述与组块化之后约 3 到 5 个组块;米勒本人强调的正是组块与比特的区别。专家的优势主要来自长时记忆中的图式,使他们能把大量项目压缩成少数几个组块;材料与图式不匹配时,优势随之消失。认知负荷理论区分内在负荷与外在负荷;相关负荷在后来的修订中被理解为分配给内在负荷的资源,而不是第三种可相加的负荷。下一章讨论当注意力在不同任务之间来回切换时,代价从哪里来。 (https://5dlibrary.com/answers/zh/attention-networks-intrinsic-motivation/c02/) - Ch.3 The Cost of Switching: Task Switching, Attention Residue and Interruptions of a Few Seconds — Task switching has reliable costs: switch trials are slower and more error-prone; preparation reduces the cost, but a residual part cannot be prepared away. Unfinished tasks leave attention residue, and finishing under time pressure reduces it. Interruptions of a few seconds roughly double to triple the next-step sequence-error rate; the "23 minutes" is the average interval before interrupted work was resumed, not the time to regain focus; and "a glance at a message wrecks 15 minutes" has no direct evidence. The next chapter turns to motivation: when do rewards crowd out a person's interest in the activity itself? (https://5dlibrary.com/answers/en/attention-networks-intrinsic-motivation/c03/) 第3章 切换的代价:任务切换、注意力残留与几秒钟的打断 — 任务切换有可靠的代价:切换试次更慢、更易出错,准备可以减少代价,但有一部分残余代价无法靠准备消除。未了结的任务留下注意力残留,在时间压力下完成任务可以减少残留。几秒钟的打断就能使下一步的序列错误率约增至两倍到三倍;「23 分钟」是被打断工作的平均恢复间隔,不是恢复专注的时间;「看一眼消息毁掉 15 分钟」没有直接证据。下一章转向动机:奖励在什么情况下会挤出一个人对活动本身的兴趣。 (https://5dlibrary.com/answers/zh/attention-networks-intrinsic-motivation/c03/) - Ch.4 Crowding Out by Extrinsic Rewards: From Soma Puzzles to "A Fine Is a Price" — Expected tangible rewards contingent on engagement, completion or performance reduce the time people freely spend on an interesting activity once the reward is gone; unexpected rewards do not; positive feedback that affirms competence increases intrinsic motivation (clearly in college students); and for tasks that were dull to begin with, crowding out does not hold. The effect sizes, $d = -0.28$ to $-0.40$, amount to roughly a sixty–forty difference, not "rewards destroy interest". A Bayesian signalling model shows how crowding out can arise when a reward is read as information about the task's value. The next chapter places these findings within self-determination theory. (https://5dlibrary.com/answers/en/attention-networks-intrinsic-motivation/c04/) 第4章 外在奖励的挤出效应:从索玛积木到「罚款就是价格」 — 预期的、取决于参与、完成或表现的有形奖励会降低人们在奖励撤去后自愿从事一项有趣活动的时间;意外奖励没有这种影响;肯定能力的正面反馈反而提高内在动机(在大学生中明显);对本来无趣的任务,挤出效应不成立。效应量在 $d = -0.28$ 到 $-0.40$ 之间,相当于约六成对四成的差别,而不是「奖励毁掉兴趣」。工作与田野研究显示金钱激励可以改变人们对一件事的理解;贝叶斯信号模型说明,当奖励被读作关于任务价值的信息时,就可能挤出内在动机。下一章讨论自我决定理论如何把这些现象放进三项基本心理需要的框架中。 (https://5dlibrary.com/answers/zh/attention-networks-intrinsic-motivation/c04/) - Ch.5 Autonomy, Competence and Relatedness: Basic Psychological Needs and a Model of Sustained Engagement — When all three needs are at a middling level (0.7 each), the two versions make exactly the same prediction (0.70); they part when the needs are unbalanced. First, when one need is almost zero: with autonomy and competence at 0.9 and relatedness at 0.1, the additive model gives 0.63 and the multiplicative 0.43; at zero relatedness the multiplicative model's engagement falls to 0, the additive model's stays at 0.60. Second, whether the needs interact: in the additive model, raising autonomy from 0.4 to 0.8 adds 0.13 whatever the level of competence; in the multiplicative model the gain is 0.12 when competence is 0.3 and 0.17 when it is 0.9, so the effect of one need depends on the others. This is testable: if the multiplicative version is closer to the truth, regression analyses should find a significant positive interaction term. (https://5dlibrary.com/answers/en/attention-networks-intrinsic-motivation/c05/) 第5章 自主、胜任与归属:基本心理需要与持续投入的模型 — 两种设定在三项需要都处于中等水平时(各 0.7)给出完全相同的预测(0.70),差别出现在不平衡的时候。第一,某一项需要几乎为零时:自主与胜任都是 0.9、归属为 0.1,加性模型给出 0.63,乘性模型给出 0.43;归属为 0 时,乘性模型的投入降到 0,加性模型仍有 0.60。第二,需要之间是否有交互:在加性模型里,把自主从 0.4 提高到 0.8,无论胜任高低,带来的增益都是 0.13;在乘性模型里,胜任为 0.3 时增益 0.12,胜任为 0.9 时增益 0.17,即一项需要的作用依赖于其他需要。这是可以检验的:如果乘性设定更接近事实,回归分析中应当出现显著的正向交互项。 (https://5dlibrary.com/answers/zh/attention-networks-intrinsic-motivation/c05/) - Ch.6 Flow: Channels of Challenge and Skill, Hypotheses about the Brain, and a Controller Model — Flow describes absorption when both challenge and skill are high; the eight-channel model standardises both against a person's own average and counts only both-above-average as flow, while low challenge with low skill is closer to apathy. Meta-analysis shows a moderate relationship between challenge–skill balance and flow. "Transient hypofrontality" is a hypothesis; imaging shows relative decreases in medial prefrontal and amygdala activity in flow, but lateral frontal regions do not consistently decrease. A learner model borrowing the "85 % rule" shows that tuning difficulty to a target accuracy can carry the learner up a ray in challenge–skill space; but outcome entropy is greatest at 50 %, not 85 %, so the information-theoretic measure is a modelling choice that needs justifying. The next chapter reads this modern research beside Neo-Confucian reverence and the pre-Qin ideals of the "empty, unified and still" mind and of the "undivided will". (https://5dlibrary.com/answers/en/attention-networks-intrinsic-motivation/c06/) 第6章 心流:挑战与技能的通道、脑机制假说与一个控制器模型 — 心流描述挑战与技能都高时的全神贯注体验;八通道模型把两者按个人自己的平均水平标准化,只有两者都高于平均时才对应心流,低挑战配低技能更接近淡漠。元分析显示挑战—技能平衡与心流有中等程度的关系。「瞬时前额叶功能低下」是假说;影像研究显示心流时内侧前额叶与杏仁核活动相对下降,但外侧额叶区域并未一致下降。一个借用「85% 规则」的学习者模型显示,按目标正确率调节难度可以使学习者沿相空间中的一条射线上升;但结果熵最大在 50%,不在 85%,说明信息论度量是一个需要论证的建模选择。下一章把这些现代研究与宋明理学的「主敬」以及先秦的「虚壹而静」「用志不分」并读。 (https://5dlibrary.com/answers/zh/attention-networks-intrinsic-motivation/c06/) - Ch.8 Practice and a Pre-registered Trial — Chapter 8 offers a four-step attention-and-motivation checklist and a pre-registered randomised crossover trial of the "brain drain" hypothesis (own phone face down and silenced on the desk versus locked in another room; single primary outcome the automated operation span absolute score, with the Sustained Attention to Response Task secondary; periods at least 7 days apart with parallel forms; 160 completers, 190 randomised; decision rule written as code), noting that replications and meta-analyses since the original study disagree. Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. This book is not diagnosis or treatment advice for ADHD, anxiety, depression or any other condition, and deals with no medication or dose. (https://5dlibrary.com/answers/en/attention-networks-intrinsic-motivation/c08/) 第8章 实践与预注册试验 — 第八章给出一份四步注意力与动机核查表,以及一项检验「手机在场即耗脑力」假说的预注册随机交叉试验方案(手机静音朝下放在桌上与锁在另一房间对比;唯一主要结局为自动化操作广度测验绝对分,持续注意反应任务为次要结局;两期间隔至少 7 天、使用平行版本;完成两期 160 人、随机化 190 人;判定规则写成代码),并说明原始研究之后的重复研究与元分析结果并不一致。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序,全部核对)、原典与作者的假说。本书不是 ADHD、焦虑、抑郁等任何疾病的诊断或治疗建议,不涉及任何药物或剂量。 (https://5dlibrary.com/answers/zh/attention-networks-intrinsic-motivation/c08/) ## 5DL-B-025 · Social Network Topology, Structural Holes and Complex Contagion 中文书名:社会网络拓扑、结构洞与复杂传染的数学模型 DOI: https://doi.org/10.5281/zenodo.23764990 Cite: Fan, Y. (2026). Social Network Topology, Structural Holes and Complex Contagion (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23764990 Book page: https://5dlibrary.com/books/network-science-structural-holes/ Scope: Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. Abstract: Information, help and new practices flow along the ties between people, but claims such as "the more people you know the better" or "find the key person and it will spread overnight" make things far too simple. This book uses graph-theoretic models, runnable programs and empirical research to ask how the structure of social networks shapes what a person can see, how far a behaviour can spread and why groups go wrong together. It covers Granovetter's "strength of weak ties" (whose job-finding figures come from a subsample of 54 people), neighbourhood overlap and the role of weak ties in holding networks together; the Watts–Strogatz model, in which 1–5 % of shortcuts cut path length to 18–11 % of the regular lattice's while clustering stays at 97–87 %; and the large-scale tests with Facebook and LinkedIn data (tie for tie, moderate strength helps most, while most help overall comes through the very many weak ties); Burt's network constraint, effective size and betweenness centrality by Brandes's algorithm, the evidence on the returns to brokerage (673 managers) and their reversal under high-commitment cultures in Chinese high-tech firms; simple and complex contagion, the width and placement of wide bridges (three ties placed together are more reliable than sixteen scattered ones), heterogeneous thresholds, and the 2024 result that with noisy thresholds long ties speed spreading up again; scale-free networks and preferential attachment, the statistical dispute over "scale-free networks are rare" (about 4 % of nearly a thousand networks show the strongest evidence) and the difficulty of testing power laws, and robustness under random failure and targeted attack (at finite sizes the advantage against random failure is small, the fragility under attack pronounced); eigenvector centrality, Bonacich's beta-centrality and the power iteration, damping factor and dangling nodes of PageRank; and the exact probabilities of Bikhchandani–Hirshleifer–Welch information cascades (with signal accuracy 0.7, wrong cascades about 25 % or 15 % of the time), and how sealed votes, stated reasons and public independent information break wrong cascades. Chapter 7 reads this research beside the vertical and horizontal alliances of the Stratagems of the Warring States and the Records of the Grand Historian (showing, on the basis of the Mawangdui silk manuscript Letters and Speeches of the Warring States Strategists and the work of Tang Lan and Yang Kuan, that Su Qin came after Zhang Yi and that much of the Records' account of Su Qin is unreliable), the Analects' "three friends who benefit" and "sociable but not factional", and the "Discourses of Qi" on not letting the four kinds of people live mixed together; it notes that the line "assess the land and grade the levy" quoted by the author comes from the "Discourses of Qi" and that its analogy with network structure is strained, and it states the ethical boundary between Warring States statecraft and modern networks of mutually beneficial cooperation. Chapter 8 gives a four-step topology checklist and a pre-registered trial randomised by team, comparing "independent anonymous written judgments and reasons, shown to all before discussion" with "open statements in order of seniority" (single primary outcome: an index of the anchoring of the team's final estimate on the first two speakers' initial estimates; 50 analysable and 59 recruited teams per arm, 118 teams and 590 people in all; decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. The book teaches no technique for manipulating other people's networks; the author's influencer- and digital-marketing business has a direct commercial interest in its subject, which is disclosed in the book. 摘要:消息、帮助与新做法沿着人与人之间的关系流动,但「认识的人越多越好」「找到关键人物就能一夜传开」这类说法把事情说得太简单了。本书用图论模型、可运行的程序与经验研究考察社会网络的结构怎样决定一个人能看到什么、一种行为能传多远、一个群体为什么会一起走错。依次讨论:格兰诺维特「弱连带的力量」(找工作数据来自 54 人的子样本)、邻域重叠与弱连带对网络连通的作用,瓦茨—斯托加茨模型中 1%—5% 的捷径使路径长度降到规则网络的 18%—11% 而聚类仍保持 97%—87%,以及脸书与领英的大规模检验(单条连带最有用的是中等强度,总体帮助主要来自数量众多的弱连带);伯特的网络约束、有效规模与布兰德斯算法的介数中心性,经纪收益的证据(673 名经理)与它在中国高科技企业高承诺文化中的反转;简单与复杂传染,宽桥的宽度与位置(三条放在一起的连带比十六条分散的连带更可靠)、阈值异质性,以及噪声阈值下长连带反而加速传播的 2024 年结果;无标度网络与优先连接、「无标度网络其实很少见」的统计争议(近千个网络中约 4% 有最强证据)与幂律检验的困难,随机失效与定向攻击下的鲁棒性(有限规模下对随机失效的稳健优势很小,对攻击的脆弱则十分明显);特征向量中心性、博纳西奇 β 中心性与 PageRank 的幂迭代、阻尼系数与悬挂节点;比克昌达尼—赫什莱弗—韦尔奇信息级联的精确概率(信号准确率 0.7 时错误级联约 25% 或 15%),以及封存投票、说出理由与公开独立信息怎样打破错误级联。 第七章把这些研究与《战国策》《史记》中的合纵连横(并据马王堆帛书《战国纵横家书》与唐兰、杨宽的研究说明苏秦晚于张仪、《史记》苏秦叙述多不可靠)、《论语》「益者三友」「群而不党」与《国语·齐语》「四民勿使杂处」并读,说明作者所引「相地而衰征」出自《齐语》且与网络结构的类比牵强,并明确战国权谋与现代互利合作网络的伦理界限;第八章给出一份四步拓扑核查表,以及一项以团队为单位随机化的预注册试验方案:比较「先独立匿名写下判断与理由、公开汇总后再讨论」与「按资历顺序公开表态」,唯一主要结局为团队最终估计向前两位发言者初始估计的锚定指数,每组 50 个可分析团队、招募 59 个,共 118 个团队、590 人,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部核对)、原典与作者的假说。本书不传授任何操纵他人关系网的技巧;作者经营的网红营销与数字营销业务与本书主题有直接商业利益,已在书中披露。 Chapter answers: - Ch.1 The Strength of Weak Ties and Small Worlds — From the "forbidden triad" Granovetter inferred that strong ties are rarely bridges and that bridges are mostly weak ties; his job-finding data come from a subsample of 54 people and show that most of the people who helped were acquaintances seen infrequently. Neighbourhood overlap makes the argument computable: low-overlap ties bring novel information and hold the whole network together. The Watts–Strogatz model shows that a few shortcuts make path lengths collapse while clustering barely changes. Large-scale studies with Facebook and LinkedIn data show that the overall contribution of weak ties comes largely from their number, and that, tie for tie, moderate strength is most useful. The next chapter moves from ties to positions: what it means to stand in the gap between circles. (https://5dlibrary.com/answers/en/network-science-structural-holes/c01/) 第1章 弱连带的力量与小世界 — 格兰诺维特的论证从「被禁止的三元组」推出强连带很少是桥,桥多半是弱连带;他的找工作数据来自 54 人的子样本,显示帮上忙的人多数是不常见面的熟人。邻域重叠把这一论证变成可以计算的量:低重叠的连带带来新颖的信息,并维系着整个网络的连通。瓦茨—斯托加茨模型显示,少量捷径就能让路径长度大降而聚类几乎不变。脸书与领英的大规模研究表明,弱连带的总体作用主要来自数量,单条连带最有用的是中等强度。下一章从连带转向位置:站在圈子之间的空隙上意味着什么。 (https://5dlibrary.com/answers/zh/network-science-structural-holes/c01/) - Ch.2 Structural Holes: Closure and Brokerage — Coleman's closure stresses that dense networks bring trust and sanctions; Burt's structural holes stress that standing between circles brings information and opportunity. Network constraint measures how far a person is trapped among mutually connected contacts; it depends on whether the contacts are connected, not on how many there are. Network constraint and betweenness centrality measure the same kind of "position" from different angles; in a synthetic network they are strongly (negatively) correlated but not identical. Burt's study of 673 managers shows that those who span structural holes are more likely to put forward ideas judged valuable and to enjoy better pay and promotion; but in Chinese high-tech firms structural holes were harmful where the organisational culture emphasised high commitment. The next chapter asks whether bridges are enough when what spreads is not a piece of news but a behaviour that needs reinforcement from several people. (https://5dlibrary.com/answers/en/network-science-structural-holes/c02/) 第2章 结构洞:闭合与经纪 — 科尔曼的闭合强调紧密的网络带来信任与约束,伯特的结构洞强调站在圈子之间带来信息与机会。网络约束衡量一个人被困在彼此相连的联系人中的程度,它取决于联系人之间是否相连,而不是联系人的多少;网络约束与介数中心性从不同角度衡量同一种「位置」,在合成网络中两者高度(负)相关但并不相同。伯特对 673 名经理的研究显示,跨越结构洞的人更可能提出被认为有价值的建议,也获得更好的薪酬与晋升;但在中国高科技企业中,在高承诺的组织文化里结构洞反而有害。下一章讨论,当传播的东西不是一条消息而是一种需要多人强化的行为时,桥还够不够。 (https://5dlibrary.com/answers/zh/network-science-structural-holes/c02/) - Ch.3 Simple and Complex Contagion — The threshold model shows that collective outcomes are extremely sensitive to the distribution of thresholds. Complex contagions, which need reinforcement from several sources, spread better in clustered networks, and crossing between groups needs wide bridges; an online experiment supports this, while observational data show that contact from different circles is more effective. The programs show that the placement of a bridge matters more than its size: three ties placed together are more reliable than sixteen scattered ones; but a small share of low-threshold people makes narrow bridges work as well. A 2024 theoretical result shows that as long as there is some chance of adopting below the threshold, long ties speed spreading up again. The next chapter turns from spreading to the structure of networks themselves: degree distributions, the scale-free dispute, and how networks behave under failure and attack. (https://5dlibrary.com/answers/en/network-science-structural-holes/c03/) 第3章 简单传染与复杂传染 — 阈值模型说明群体结果对阈值分布极为敏感。需要多个来源强化的复杂传染,在聚簇网络中传得更好,跨越群体需要宽桥;在线实验支持这一点,观察性数据则显示来自不同圈子的接触更有效。程序显示,桥的位置比数量更重要:三条放在一起的连带比十六条分散的连带更可靠;但一小部分低阈值的人会让窄桥也起作用。2024 年的理论结果表明,只要低于阈值时仍有一点采纳的可能,长连带又会加速传播。下一章从传播转向网络本身的结构:度的分布、无标度的争议,以及网络在失效与攻击下的表现。 (https://5dlibrary.com/answers/zh/network-science-structural-holes/c03/) - Ch.4 Scale-Free Networks, Preferential Attachment and Robustness — Under attack the difference is stark: after the 10 % of nodes with the highest degree are removed, the BA network's largest component still holds 63 % of the nodes, and after 15 % only 4 %; the same attack does much less damage to the ER network, which keeps 78 % after 15 % of its nodes are removed and only falls to 6 % after 40 %. The highest-degree node in the BA network has 104 connections, in the ER network at most 13: attacking a scale-free network means first taking out the hubs that hold it together. (https://5dlibrary.com/answers/en/network-science-structural-holes/c04/) 第4章 无标度网络、优先连接与鲁棒性 — 结果有三点值得注意。第一,即使数据确实来自优先连接模型,指数的估计也随 $k_{\min}$ 变化:$k_{\min} = 2$ 时为 2.40,$k_{\min} = 8$ 时为 2.80,都低于理论值 3,因为有限规模的网络中尾部还没有完全形成。第二,对数正态数据在 $k_{\min} = 4$ 时也给出 2.77 的「指数」,与 BA 数据几乎一样;只有把 $k_{\min}$ 逐步提高,才看到它的「指数」不断变大(3.60、4.56),暴露出它不是幂律。第三,似然比检验对对数正态数据给出了明确的结论($z = -5.73$,偏向对数正态);但对真正的 BA 数据,检验无法区分幂律与对数正态($z = -0.63$,$p = 0.53$),最佳拟合的对数正态参数($\mu = -12$、$\sigma = 2.9$)说明它正在用一个极宽的分布模仿幂律。也就是说,在约 1,600 个尾部数据点的规模上,「无法拒绝幂律」并不等于「证明是幂律」;布罗伊多与克劳塞特的大量网络落在「弱证据」等级上,部分原因正在于此。 (https://5dlibrary.com/answers/zh/network-science-structural-holes/c04/) - Ch.5 Eigenvector Centrality and PageRank — Eigenvector centrality brings "whom one is connected to" into the calculation of status and can be found by power iteration. Bonacich's beta-centrality joins degree and eigenvector centrality into one family of measures; negative beta captures the bargaining power of being depended upon, which connects with structural holes. PageRank is a random walk with random jumps on a directed network: the damping factor deals with dangling nodes and traps and guarantees convergence, and $d = 0.85$ is a compromise rather than an optimum. The next chapter discusses how, when everyone watches what the people before them do, a group can converge on the same direction, and why that direction may be wrong. (https://5dlibrary.com/answers/en/network-science-structural-holes/c05/) 第5章 特征向量中心性与 PageRank — 特征向量中心性把「连接的是谁」纳入地位的计算,可以用幂迭代求得。博纳西奇的 β 中心性把度与特征向量中心性连成一族指标,负的 β 衡量「被依赖」的议价权力,与结构洞相通。PageRank 是有向网络上带随机跳转的随机游走:阻尼系数处理了悬挂节点与陷阱,保证收敛,$d = 0.85$ 是一种折中而不是最优。下一章讨论,当每个人都在观察前面的人怎么做时,群体会怎样一起走向同一个方向,而那个方向未必正确。 (https://5dlibrary.com/answers/zh/network-science-structural-holes/c05/) - Ch.6 Information Cascades and How to Break Them — Take $p = 0.7$ and the coin rule: once just two people have chosen, the probability that a cascade has already formed is 0.79, and the probability of a wrong cascade 0.195; by the tenth person a cascade has almost certainly formed, the probability of a wrong cascade is 0.247, and it changes no further. Even though everyone's private information is right seven times out of ten, the group as a whole has about a one-in-four chance of going wrong together, and once it has gone wrong it does not correct itself. The rule of following one's own signal produces fewer wrong cascades (0.155), because indifferent people still reveal their signals and cascades form more slowly. The more accurate the signals, the fewer the wrong cascades, but even with $p = 0.8$ the coin rule gives 14 % wrong cascades. The simulation (0.246) agrees with the exact calculation (0.247). (https://5dlibrary.com/answers/en/network-science-structural-holes/c06/) 第6章 信息级联及其打破 — 以 $p = 0.7$、抛硬币规则为例:只要两个人做出选择,级联已经形成的概率就是 0.79,错误级联的概率是 0.195;到第 10 个人时,级联几乎必然已经形成,错误级联的概率是 0.247,之后不再改变。也就是说,即使每个人的私人信息有七成的准确率,整个群体仍有约四分之一的机会集体走错,而且一旦走错就不会自己纠正。跟随自己信号的规则让错误级联少一些(0.155),因为无差别的人仍然透露了信号,级联形成得慢一些。信号越准确,错误级联越少,但即使 $p = 0.8$,抛硬币规则下仍有 14% 的错误级联。模拟的结果(0.246)与精确计算(0.247)一致。 (https://5dlibrary.com/answers/zh/network-science-structural-holes/c06/) - Ch.8 Practice and a Pre-registered Trial — Chapter 8 gives a four-step topology checklist and a pre-registered trial randomised by team, comparing "independent anonymous written judgments and reasons, shown to all before discussion" with "open statements in order of seniority" (single primary outcome: an index of the anchoring of the team's final estimate on the first two speakers' initial estimates; 50 analysable and 59 recruited teams per arm, 118 teams and 590 people in all; decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. The book teaches no technique for manipulating other people's networks; the author's influencer- and digital-marketing business has a direct commercial interest in its subject, which is disclosed in the book. (https://5dlibrary.com/answers/en/network-science-structural-holes/c08/) 第8章 实践与预注册试验 — 第八章给出一份四步拓扑核查表,以及一项以团队为单位随机化的预注册试验方案:比较「先独立匿名写下判断与理由、公开汇总后再讨论」与「按资历顺序公开表态」,唯一主要结局为团队最终估计向前两位发言者初始估计的锚定指数,每组 50 个可分析团队、招募 59 个,共 118 个团队、590 人,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部核对)、原典与作者的假说。本书不传授任何操纵他人关系网的技巧;作者经营的网红营销与数字营销业务与本书主题有直接商业利益,已在书中披露。 (https://5dlibrary.com/answers/zh/network-science-structural-holes/c08/) ## 5DL-B-026 · Probabilistic Judgement, Brier-Score Calibration and the Statistics of Collective Intelligence 中文书名:概率判断、布莱尔分数校准与群体智慧的统计理论 DOI: https://doi.org/10.5281/zenodo.23764998 Cite: Fan, Y. (2026). Probabilistic Judgement, Brier-Score Calibration and the Statistics of Collective Intelligence (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23764998 Book page: https://5dlibrary.com/books/judgment-calibration-forecasting/ Scope: This book is not counselling, medical or investment advice. Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. Abstract: How likely, exactly, is "it'll probably work out"? How often has the speaker been right? This book uses the psychology of judgement and decision making, forecast verification and statistics to ask how the accuracy of probabilistic judgements can be measured, in what systematic ways they go wrong, and when combining many people's judgements makes them more accurate. It covers Tetlock's study of 284 experts' forecasts over about two decades (some 28,000 forecasts, with a count of 82,361 also quoted by reviewers; experts did slightly better than random guessing but worse than simple extrapolation, foxes beat hedgehogs, and "worse than a dart-throwing chimp" is a popular paraphrase), training, teaming and superforecasters in forecasting tournaments, and models of overconfidence, under-reaction and over-reaction; the Brier score and Murphy's decomposition (reliability − resolution + uncertainty, with two within-bin terms after binning), the fact that always stating the base rate is perfectly calibrated with zero resolution (always saying 50 % is calibrated only when the base rate is 50 %), and the derivation of strictly proper scoring rules; base-rate neglect, a positive predictive value of about 16.1 % at 1 % prevalence, and natural frequencies (in a meta-analysis accuracy rose from about 4 % to 24 %); signal detection theory's d′, criterion and optimal likelihood-ratio threshold under unequal costs; the planning fallacy, reference-class forecasting (almost nine in ten of 258 transport projects overran, rail by 44.7 % on average) and empirical-Bayes shrinkage; and the corrected median and mean of Galton's ox-weighing, the diversity prediction theorem (an identity, not an empirical law), the conditions under which correlated errors stop a crowd from beating an expert, and why averaged probabilities should be extremised. The eight chapters are read beside the "temple calculation" and the five factors and seven assessments of the Sunzi's chapter "Estimates", the "weighing of power" in the Guiguzi's chapter "Assessing" (corrected: not a chapter title; the text reads "What is weighing power? It is to measure great and small, and to reckon many and few"), and the Han Feizi's "to be certain without cross-checking is folly", stating clearly the boundary between categorical classical pronouncements and modern probability calibration. They end with a four-step probability-calibration checklist and a pre-registered randomised controlled trial comparing a 15-minute calibration micro-training (natural-frequency conversion, anchoring on an outside reference class, Brier-score and interval-hit feedback) with time-matched reading on logical thinking (single primary outcome the calibration error |H − 0.80| of 80 % intervals on 20 items, 133 analysable and 148 recruited participants per arm, 296 in all, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. This book is not counselling, medical or investment advice. 摘要:「大概会成」到底是多大的可能?说这句话的人这些年说对了多少?本书用判断与决策心理学、预测评估与统计学考察怎样测量概率判断的准确性,判断会以哪些系统性的方式出错,以及把许多人的判断合在一起何时更准。依次讨论:泰洛克对 284 名专家约二十年预测的研究(约 28,000 次预测,另有书评引用的 82,361 次计数;专家平均略好于随机猜测、不如简单外推,狐狸型好于刺猬型,「不如掷飞镖的黑猩猩」是通俗说法),预测锦标赛中的训练、组队与超级预测者,以及过度自信与反应不足、反应过度的模型;布莱尔分数与墨菲分解(可靠性 − 区分度 + 不确定性,分箱时多出组内两项),「总是报基础率者校准完美而区分度为零」(「总是报 50%」只在基础率为 50% 时校准),以及严格恰当评分规则的推导;基础率忽略、患病率 1% 时阳性预测值约 16.1% 与自然频率(元分析中正确率约从 4% 到 24%);信号检测论的 d′、判断标准与代价不对称下的最优似然比门槛;计划谬误、参考类别预测(258 个交通项目近九成超支,铁路平均 44.7%)与经验贝叶斯收缩;高尔顿称牛重的中位数与平均数之更正、多样性预测定理(恒等式而非经验规律)、相关误差使群体无法胜过专家的条件,以及平均概率为何需要推向极端。 八章与《孙子兵法·计篇》的「庙算」与五事七计、《鬼谷子·揣篇》的「量权」(更正:不是独立篇名,原文为「何谓量权?曰:度于大小,谋于众寡」)及《韩非子·显学》「无参验而必之者,愚也」并读,并明确古代确定性断语与现代概率校准的界限;最后给出一份四步概率校准核查表,以及一项比较 15 分钟校准微训练(自然频率转换 + 外部参考类别锚定 + 布莱尔分数与区间命中反馈)与时长匹配的逻辑思维阅读的预注册随机对照试验方案(唯一主要结局为 20 道题上 80% 区间的校准误差 |H − 0.80|,每组 133 人可分析、招募 148 人,共 296 人,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序,全部核对)、原典与作者的假说。本书不是心理咨询、医学或投资建议。 Chapter answers: - Ch.1 Tetlock's Forecasting Research: Experts, Foxes and Superforecasters — Tetlock had 284 experts make tens of thousands of numerical probability forecasts over about two decades. On average they did only slightly better than "one third each", worse than simple extrapolation and statistical models; foxes beat hedgehogs, and hedgehogs did especially badly on long-range forecasts in their own field. "Worse than a chimp" is a popular paraphrase. A toy model shows that a forecaster with the same information can fall to the level of random guessing through overconfidence alone. Forecasting tournaments show that brief probability training, teaming and tracking improve accuracy, and that superforecasters are partly discovered and partly created; weather forecasting and intelligence assessment, which have feedback, can also be well calibrated. The next chapter turns to how to keep score: the Brier score, Murphy's decomposition and proper scoring rules. (https://5dlibrary.com/answers/en/judgment-calibration-forecasting/c01/) 第1章 泰洛克的预测研究:专家、狐狸与超级预测者 — 泰洛克让 284 名专家在约二十年里给出数万次数字概率预测,发现专家平均只是略好于「一律三分之一」,不如简单外推与统计模型,狐狸型好于刺猬型,刺猬在自己领域的长期预测上尤其差;「不如黑猩猩」是通俗说法。玩具模型显示,拥有同样信息的预测者,只因过度自信就可能退到随机猜测的水平。预测锦标赛表明,简短的概率训练、组队与追踪可以提高准确性,超级预测者部分被发现、部分被培养;有反馈的气象与情报评估也可以校准得很好。下一章讨论怎样打分:布莱尔分数、墨菲分解与恰当评分规则。 (https://5dlibrary.com/answers/zh/judgment-calibration-forecasting/c01/) - Ch.2 Strictly Proper Scoring Rules: The Brier Score and Murphy's Decomposition — The Brier score is the mean squared difference between forecasts and outcomes; Murphy split it into reliability (calibration), resolution and uncertainty. The decomposition is exact when the forecasts within a bin are equal; after binning, within-bin variance and covariance terms appear. Always saying 50 % is perfectly calibrated only when the base rate is 50 %; in general, always stating the base rate is perfectly calibrated with zero resolution. An overconfident forecaster can have the same resolution as a calibrated one and still score worse by saying too much. The Brier and logarithmic scores are strictly proper; the absolute error is not. Computing an individual's calibration error from very few forecasts overstates it. The next chapter discusses the best-known error in probabilistic judgement: neglecting the base rate. (https://5dlibrary.com/answers/en/judgment-calibration-forecasting/c02/) 第2章 严格恰当评分规则:布莱尔分数与墨菲分解 — 布莱尔分数是预报与结果之差的平方的平均,墨菲把它分解为可靠性(校准)、区分度与不确定性三项;当组内预报相同时分解是精确的,分箱后则多出组内方差与协方差两项。总是报 50% 的人只在基础率为 50% 时校准完美;一般而言,总是报基础率的人校准完美而区分度为零。过度自信者可以拥有和校准者同样的区分度,却因为说得太满而得分更差。布莱尔分数与对数分数是严格恰当的,绝对误差不是。用很少的预报给个人算校准误差,会系统地偏高。下一章讨论概率判断中最著名的一种错误:忽略基础率。 (https://5dlibrary.com/answers/zh/judgment-calibration-forecasting/c02/) - Ch.3 Base-Rate Neglect and Natural Frequencies — People judging individual cases often go by representativeness and give base rates far less weight than they deserve; they use them more when base rates seem relevant to the case. For a test with 95 % sensitivity and specificity and a prevalence of 1 %, the positive predictive value is about 16.1 %; more than four in ten doctors answered 95 % both in 1978 and in 2014. Natural-frequency formats clearly improve Bayesian reasoning (in the meta-analysis from about 4 % to 24 % correct), though most people still answer wrongly; the effect of frequency-tree training persists for a while. The next chapter discusses a closely related problem: where to set the threshold of judgement when real signals are rare. (https://5dlibrary.com/answers/en/judgment-calibration-forecasting/c03/) 第3章 基础率忽略与自然频率 — 人们判断个案时常常依据代表性,给基础率的权重远低于应得的权重;当基础率显得与个案相关时,使用得会多一些。患病率 1%、灵敏度与特异度 95% 的检测,阳性预测值约为 16.1%;医生在 1978 年与 2014 年都有四成以上回答 95%。自然频率格式使贝叶斯推理的正确率明显提高(元分析中约从 4% 到 24%),但大多数人仍然答错;频率树训练的效果可以保持一段时间。下一章讨论另一类与基础率密切相关的问题:在信号很少的环境中,判断的门槛应当放在哪里。 (https://5dlibrary.com/answers/zh/judgment-calibration-forecasting/c03/) - Ch.4 Signal Detection Theory: Seeing Clearly and Where to Set the Threshold — The first three judges have almost the same sensitivity ($d'$ about 2); the difference lies entirely in the threshold. The neutral judge has $c = 0$ and $\beta = 1$; the liberal one has $c = -0.70$ and a hit rate of 95 %, paid for with 40 % false alarms; the conservative one has $c = +1.00$, almost never raises a false alarm (2.3 %) and misses half the signals. Judged by hit rate alone, the liberal judge looks "best"; judged by false alarms alone, the conservative one does; in fact all three see equally well. The fourth judge has a 90 % hit rate, which looks good, but a 69 % false-alarm rate and $d'$ of only 0.79: he is mostly saying "yes" to everything. (https://5dlibrary.com/answers/en/judgment-calibration-forecasting/c04/) 第4章 信号检测论:看得清与门槛放在哪里 — 前三位判断者的辨别力几乎相同($d'$ 约为 2),差别全在门槛:中性者 $c = 0$,$\beta = 1$;宽松者 $c = -0.70$,击中率高达 95%,代价是 40% 的虚报;严格者 $c = +1.00$,几乎从不虚报(2.3%),却漏掉一半的信号。如果只看击中率,宽松者「最厉害」;如果只看虚报率,严格者「最厉害」;其实三人看得一样清楚。第四位判断者击中率 90%,看起来不错,但虚报率高达 69%,$d'$ 只有 0.79:他主要是在什么都说「是」。 (https://5dlibrary.com/answers/zh/judgment-calibration-forecasting/c04/) - Ch.5 Reference-Class Forecasting and the Planning Fallacy — The planning fallacy is the tendency to take the inside view and underestimate time and cost systematically; students took longer to write their theses than even their "worst case" estimates. Almost nine in ten (86 %) of 258 transport projects overran their costs, by an average of 44.7 % for rail, 33.8 % for bridges and tunnels and 20.4 % for roads, owing both to optimism bias and to strategic misrepresentation. Reference-class forecasting first finds the distribution of outcomes for similar projects and then places the case within it; a synthetic model shows that the uplift depends on the risk one will accept, and should come from a real empirical distribution. Empirical-Bayes shrinkage combines a team's own record with the reference class according to its reliability, and clearly beats relying on one's own record when records are short and noisy. The next chapter turns from individuals to groups. (https://5dlibrary.com/answers/en/judgment-calibration-forecasting/c05/) 第5章 参考类别预测与计划谬误 — 计划谬误是人们采用内部视角、系统地低估时间与成本的倾向;学生写论文的实际用时比「最坏情况」的估计还长。258 个交通项目中近九成(86%)成本超支,平均幅度铁路 44.7%、桥梁与隧道 33.8%、道路 20.4%,超支的来源既有乐观偏差,也有策略性误报。参考类别预测先找同类项目的结果分布,再定位个案;合成模型显示,上调值取决于愿意承担的风险,并且应当来自真实的经验分布。经验贝叶斯收缩把团队自己的记录按可信程度与参考类别合并,在记录少、波动大时明显优于只看自己。下一章从个人转向群体。 (https://5dlibrary.com/answers/zh/judgment-calibration-forecasting/c05/) - Ch.6 The Statistics of Collective Intelligence — The crowd median Galton reported in 1907 was 1,207 lb (it should have been 1,208); the mean of 1,197 lb came later and exactly equals the corrected actual weight. The diversity prediction theorem is an identity: crowd error equals average individual error minus diversity; it guarantees that the crowd is no worse than the average individual, not that it beats the best one. With correlated errors, averaging removes only the uncorrelated part; once the correlation exceeds the expert's relative error, no number of people beats the expert. Social influence reduces diversity and raises confidence without improving accuracy. Moderately extremising the average of many probabilities can improve scores, because each person has only part of the information; how far to go depends on how much is shared. The next chapter reads this modern research beside the "temple calculation" of the Chinese military classics. (https://5dlibrary.com/answers/en/judgment-calibration-forecasting/c06/) 第6章 群体智慧的统计理论 — 高尔顿 1907 年报告的群体中位数为 1,207 磅(应为 1,208 磅),平均数 1,197 磅是随后才给出的,恰好等于更正后的实际重量。多样性预测定理是一个恒等式:群体误差等于个人平均误差减去多样性;它保证群体不比个人平均差,却不保证群体胜过最好的个人。误差相关时,平均只能消除不相关的部分;相关系数一旦超过专家的相对误差,再多的人也胜不过专家。社会影响会降低多样性、提高信心而不提高准确性。对许多人的概率取平均后适度推向极端可以改善分数,因为每个人只掌握部分信息;推多远取决于信息共享的程度。下一章把这些现代研究与中国古代兵家的「庙算」并读。 (https://5dlibrary.com/answers/zh/judgment-calibration-forecasting/c06/) - Ch.8 Practice and a Pre-registered Trial — The four-step calibration checklist turns the conclusions of earlier chapters into four recordable steps: replace vague words with numbers, find the base rate of an outside reference class first, write down update triggers in advance, and at the end of the quarter review with the Brier score against the base rates. The fictional example shows that intuitive numbers can be worse than stating the base rates alone, and that updating as agreed in advance helps, though 12 questions in one quarter are not enough to conclude anything. The trial protocol compares a 15-minute calibration micro-training with time-matched reading on logical thinking; its single primary outcome is the calibration error $|H - 0.80|$ of 80 % intervals on 20 items, with 133 analysable and 148 recruited participants per arm; the decision rule, including thresholds for a manipulation check and for exclusions, is written as code and published in advance; the results will be published whether positive or negative. (https://5dlibrary.com/answers/en/judgment-calibration-forecasting/c08/) 第8章 实践与预注册试验 — 四步校准核查表把前几章的结论落到四个可记录的环节:用数字替代模糊词,先找外部参考类别的基础率,事前写下更新触发条件,季度结束后用布莱尔分数复盘并与基础率比较;虚构示例显示,直觉数字可能比只报基础率还差,而按事先约定更新后有所改善,但一个季度的 12 个问题不足以下结论。试验方案比较 15 分钟校准微训练与时长匹配的逻辑思维阅读,唯一的主要结局是 20 道题上 80% 区间的校准误差 $|H - 0.80|$,每组 133 人可分析、招募 148 人,判定规则包括操作检查与排除比例的门槛,写成代码并事先公开;无论结果正负都发表。 (https://5dlibrary.com/answers/zh/judgment-calibration-forecasting/c08/) ## 5DL-B-027 · Mechanism Design, Incentive Compatibility and the Game Theory of Negotiation 中文书名:机制设计、激励相容与谈判博弈的数学模型 DOI: https://doi.org/10.5281/zenodo.23765012 Cite: Fan, Y. (2026). Mechanism Design, Incentive Compatibility and the Game Theory of Negotiation (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23765012 Book page: https://5dlibrary.com/books/mechanism-design-negotiation/ Scope: This book is not legal advice; bid rigging, collusive tendering and other unlawful practices are discussed only as weaknesses of mechanisms, and no method for them is given. Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. Abstract: A negotiation ends with both sides satisfied, yet a large piece of value is often left on the table; a reasonable-looking performance metric produces beautiful numbers and a worse business a year later. This book uses mechanism design, auction theory, bargaining theory and the economics of information to examine how rules themselves shape behaviour, and where even good rules meet unavoidable limits. It covers mechanism design as "reverse game theory" and the revelation principle; the envy-freeness of "I cut, you choose" (which guarantees each person only half, with the outcome depending on who knows the other's preferences); free-riding in voluntary public-good contributions and the Clarke pivot mechanism (truthful reporting is dominant, but taxes cannot be returned and the budget cannot balance); the weak dominance of truthful bidding in Vickrey's second-price auction (verified by program across 1,030,301 combinations), bid shading in first-price auctions as a derivable equilibrium strategy b = (n−1)v/n rather than a gamble, a program check of revenue equivalence, and the weaknesses of VCG, whose revenue can be zero, can fall as bidders are added and is fragile against coordination among losers; the four conditions of the Myerson–Satterthwaite theorem and its overlap assumption, with the best mechanism in the uniform case trading only when v ≥ c + 1/4 and realising 9/64 ÷ 1/6 = 84.375 % of the gains from trade, a bound attained by the linear Chatterjee–Samuelson equilibrium, which is not the only equilibrium; the Nash bargaining solution, Rubinstein's alternating offers x₁* = (1 − δ₂)/(1 − δ₁δ₂) and their convergence, with a checked correction of "your leverage depends on your BATNA": a BATNA sets the bottom line and acts continuously as a Nash threat point, but in alternating offers with an outside option it matters only when it exceeds one's share anyway (the outside option principle), while patience and payoffs during the talks move shares continuously; the price spiral of the market for lemons, certification, the separating conditions for Spence signals (single crossing, a cost range, verifiability) and the social waste of pure signals; the difference between Goodhart's 1975 wording and Strathern's 1997 popular version, Campbell's law, a selection model of manipulable metrics and a multitask model in which random audits permit steeper incentives. The eight chapters are read beside the Han Feizi on "holding names to account for reality", the "Two Handles" chapter's punishment of both "large words, small results" and "small words, large results" and its story of Marquis Zhao punishing both the keeper of robes and the keeper of caps (from "The Two Handles", not "Defining the Law"), the Guanzi's "not letting intentions roam outside the law, nor granting favours within the law", the Analects' "to govern is to correct" and "lead them with virtue", and Mencius's "when superiors and inferiors compete for profit, the state is in danger", stating clearly that ancient Legalism served the ruler's centralised power while modern mechanism design serves voluntary contracts between equals and public welfare. They end with a four-step mechanism and bargaining checklist (quantifying both sides' BATNAs, auditing the cost of waiting and cash-flow patience, replacing verbal assurances with verifiable signals, red-teaming KPIs for Goodhart leaks) and a pre-registered dyad-randomised controlled trial in which both members of each dyad receive the same condition, comparing a 10-minute structured preparation form with reading standard negotiation tips; the single primary outcome is joint points in a five-issue simulated negotiation (impasse scored as the sum of both BATNAs), with 173 analysable and 193 randomised dyads per arm, 772 people in all, and a decision rule written as code. Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. This book is not legal advice; bid rigging, collusive tendering and other unlawful practices are discussed only as weaknesses of mechanisms, and no method for them is given. 摘要:谈判结束,双方都觉得满意,桌上却常常留着一大块谁都没拿到的价值;一个看似合理的考核指标,一年后数字漂亮、生意变差。本书用机制设计、拍卖理论、议价博弈与信息经济学考察规则本身如何塑造行为,以及好的规则在哪里必然受限。依次讨论:机制设计作为「逆向博弈论」与显示原理,「我切你选」的无嫉妒性(只保证每人一半,结果取决于谁掌握对方偏好的信息),公共品自愿捐献中的搭便车与克拉克枢纽机制(如实报告为占优策略,但税款不能退还、预算无法平衡);维克里二价拍卖中报真值的弱占优性(程序在 1,030,301 种组合上验证),一价拍卖的压价是可推导的均衡策略 b = (n−1)v/n 而不是赌运气,收益等价定理的程序验证,以及 VCG 机制收入可能为零、随竞争者增加而下降、对落败者协调脆弱的弱点;迈尔森—萨特思韦特定理的四个条件与估值重叠前提,均匀分布下最好的机制只在 v ≥ c + 1/4 时成交、实现 9/64 ÷ 1/6 = 84.375% 的交易收益,查特吉—萨缪尔森线性均衡恰好达到这一上限但并非唯一均衡;纳什议价解与鲁宾斯坦轮流出价 x₁* = (1 − δ₂)/(1 − δ₁δ₂) 及两者的收敛,并核对后修正「筹码取决于 BATNA」:BATNA 决定底线,作为纳什威胁点时持续起作用,但在带外部选择的轮流出价中只有高于本来份额时才起作用(外部选择原则),持续改变份额的是耐心与谈判期间的内部收益;柠檬市场的价格螺旋、认证与斯宾塞信号的分离条件(单交叉、成本区间、可核查)及纯信号的社会浪费;古德哈特 1975 年原话与斯特拉森 1997 年通俗表述的区别、坎贝尔定律,以及可操纵指标的选择模型与随机审计允许更陡峭激励的多任务模型。 八章与《韩非子·定法》「循名而责实」、《二柄》同时惩罚「言大而功小」与「言小而功大」及韩昭侯「兼罪典衣与典冠」(出自《二柄》而非《定法》)、《管子·明法》「不淫意于法之外,不为惠于法之内」、《论语》「政者,正也」与「道之以德」、《孟子》「上下交征利而国危矣」并读,并明确古代法家服务于君主集权、现代机制设计服务于平等主体的自愿契约与公共福利;最后给出一份四步机制与议价核查表(双方 BATNA 量化、等待成本与现金流耐心审计、从口头保证切换为可核查信号、KPI 古德哈特漏洞红队测试),以及一项预注册双人随机对照试验方案:以两人组为随机化单位、两人接受同一条件,比较谈判前 10 分钟结构化准备表格与常规谈判技巧阅读,唯一主要结局为五议题模拟谈判的联合点数(未达成协议计为双方 BATNA 之和),每组 173 组可分析、随机化 193 组,共 772 人,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序,全部核对)、原典与作者的假说。本书不是法律意见;拍卖串通、围标等违法行为只作为机制弱点讨论,不提供操作方法。 Chapter answers: - Ch.1 Reverse Game Theory: Incentive Compatibility and the Revelation Principle — The result shows three properties of the mechanism. First, the cutter cuts truthfully according to her own preferences (at 0.362) and gets exactly half in her own eyes; the chooser gets 0.777 in his eyes, and neither envies the other. Second, what it guarantees is only one half; it does not guarantee a "fair share" of the surplus. If the cutter knows the chooser's preferences, she moves the knife to the point where he is just indifferent (0.638), keeping 0.777 for herself and leaving him only 0.501: still envy-free, but the gains all go to the cutter. This is exactly the information problem that mechanism design cares about: the outcome of a rule depends on who knows what. Third, it is not necessarily efficient: the sum of the two valuations could reach 1.300 (cut in the middle, each taking the preferred end), whereas cut and choose yields 1.277. (https://5dlibrary.com/answers/en/mechanism-design-negotiation/c01/) 第1章 逆向博弈论:激励相容与显示原理 — 结果显示了这一机制的三个性质。第一,切的人如实地按自己的偏好切(切在 0.362 处),得到自己眼中恰好一半;挑的人得到自己眼中的 0.777,两人都不嫉妒对方。第二,它保证的只是一半,并不保证「公平地分享剩余」:如果切的人知道挑的人的偏好,她会把刀移到挑的人恰好无所谓的位置(0.638),自己拿到 0.777,挑的人只剩 0.501——仍然无嫉妒,但好处全归了切的人。这正是机制设计关心的信息问题:规则的结果取决于谁知道什么。第三,它不一定有效:两人估值之和最高可以达到 1.300(从正中间切开、各拿自己喜欢的一端),而「我切你选」只得到 1.277。 (https://5dlibrary.com/answers/zh/mechanism-design-negotiation/c01/) - Ch.2 Auctions: Vickrey, Revenue Equivalence and the Loneliness of VCG — The second-price auction makes the winner pay the second-highest bid, so that one's own bid determines only whether one wins, not the price; truthful bidding is therefore weakly dominant, as the program verified across more than a million combinations. Laboratory subjects nevertheless often overbid. In the first-price auction, shading is an equilibrium strategy that can be derived: with $n$ bidders and independent uniform private values, $b(v) = (n-1)v/n$; laboratory bids are often above this risk-neutral benchmark. Revenue equivalence means that first-price, second-price and all-pay auctions yield the same expected revenue, though with different risk. In theory VCG achieves truthful reporting and efficient allocation; in practice it can yield very low revenue, revenue that falls as bidders are added, and is fragile against coordination among losers and false identities. The next chapter asks how hard things become when both buyer and seller hold private information. (https://5dlibrary.com/answers/en/mechanism-design-negotiation/c02/) 第2章 拍卖:维克里、收益等价与 VCG 的孤独 — 二价拍卖让赢家支付第二高出价,使自己的出价只决定是否获胜、不决定价格,因此报真值是弱占优策略;程序在一百多万种组合上验证了这一点。但实验中的被试仍常常高报。一价拍卖中,压价是可以推导的均衡策略:在 $n$ 人均匀独立私人价值的模型中,$b(v) = (n-1)v/n$;实验中的出价常高于这一风险中性基准。收益等价定理说明一价、二价与全付拍卖的期望收入相同,但收入的风险不同。VCG 机制在理论上实现如实报告与有效配置,在实践中却可能收入很低、随竞争者增加反而下降,并且对落败者的协调与虚假身份很脆弱。下一章讨论,当买卖双方都有私人信息时,事情会难到什么程度。 (https://5dlibrary.com/answers/zh/mechanism-design-negotiation/c02/) - Ch.3 The Myerson–Satterthwaite Impossibility Theorem — The Myerson–Satterthwaite theorem shows that when buyer and seller both hold private information and their value ranges overlap, no mechanism can be at once Bayesian incentive compatible, interim individually rational, ex post efficient and budget balanced. In the uniform example, the best mechanism trades only when $v \ge c + 1/4$, realising $9/64 \div 1/6 = 84.375\,\%$ of the first-best gains; the linear equilibrium of the Chatterjee–Samuelson double auction attains this bound exactly, though the auction also has many worse equilibria. The loss of efficiency comes from the information structure, not from a lack of skill; the ways to reduce it are to make information verifiable, to bring in competition, or to have an outsider bear the deficit. The next chapter leaves mechanisms behind and turns to how two people facing each other divide a surplus. (https://5dlibrary.com/answers/en/mechanism-design-negotiation/c03/) 第3章 迈尔森—萨特思韦特不可能性定理 — 迈尔森—萨特思韦特定理说明,当买卖双方都有私人信息、估值区间重叠时,没有任何机制能同时做到贝叶斯激励相容、中期个人理性、事后有效与预算平衡。在均匀分布的例子中,最好的机制只在 $v \ge c + 1/4$ 时成交,实现最优交易收益的 $9/64 \div 1/6 = 84.375\%$;查特吉—萨缪尔森双边拍卖的线性均衡恰好达到这个上限,但该拍卖还有许多更差的均衡。效率损失来自信息结构,而不是来自技巧不足;缩小损失的办法是让信息可核查、引入竞争或由外部承担赤字。下一章离开机制,讨论两个人面对面讨价还价时的分配。 (https://5dlibrary.com/answers/zh/mechanism-design-negotiation/c03/) - Ch.4 Bargaining: Nash, Rubinstein and What a BATNA Really Does — Nash's axiomatic bargaining solution maximises $(x_1 - d_1)^{\alpha}(x_2 - d_2){1-\alpha}$; Rubinstein's alternating-offers model gives $x_1^* = (1 - \delta_2)/(1 - \delta_1\delta_2)$, so patience is bargaining power. As the interval between offers goes to zero the two coincide, with the Nash weight set by relative patience. A BATNA sets your bottom line, but whether it raises your share depends on the structure of the negotiation: as a Nash threat point it matters continuously; as an outside option in alternating offers it matters only when it exceeds the share you would get anyway; payoffs earned during the negotiation are what move shares continuously, like a threat point. The next chapter turns to another kind of information problem: one side knows the quality of the goods and the other does not. (https://5dlibrary.com/answers/en/mechanism-design-negotiation/c04/) 第4章 议价:纳什、鲁宾斯坦与 BATNA 的真正作用 — 纳什的公理化议价解最大化 $(x_1 - d_1)^{\alpha}(x_2 - d_2){1-\alpha}$;鲁宾斯坦的轮流出价模型给出 $x_1^* = (1 - \delta_2)/(1 - \delta_1\delta_2)$,耐心就是议价力量。当出价间隔趋于零时,两者重合,纳什权重由相对耐心决定。BATNA 决定了你的底线,但它是否提高你的份额取决于谈判的结构:作为纳什威胁点,它持续起作用;作为轮流出价中的外部选择,它只有在高于你本来就能得到的份额时才起作用;谈判期间的内部收益才像威胁点一样持续改变份额。下一章讨论另一类信息问题:一方知道商品的质量,另一方不知道。 (https://5dlibrary.com/answers/zh/mechanism-design-negotiation/c04/) - Ch.5 Lemons and Costly Signals — With $k = 1.5$ the price is multiplied by 0.75 each round: from 1.00 to 0.75, 0.56, 0.42…; after 17 rounds fewer than 1 % of cars remain on the market. With $k = 1.9$ the shrinking is much slower, but it still takes about 90 rounds to fall below 1 %: as long as $k < 2$, the spiral ends at zero. With $k = 2.5$ the buyers' premium for quality is high enough: at a price of 1 the average quality on offer is 0.5, and 0.5 times 2.5 is 1.25, still above the top quality, so the market does not shrink at all. (https://5dlibrary.com/answers/en/mechanism-design-negotiation/c05/) 第5章 柠檬市场与昂贵的信号 — $k = 1.5$ 时价格每一轮乘以 0.75:从 1.00 降到 0.75、0.56、0.42……17 轮后不到 1% 的车还在市场上。$k = 1.9$ 时萎缩慢得多,但仍要 90 轮左右才降到 1% 以下:只要 $k < 2$,螺旋最终都会走到零。$k = 2.5$ 时买方对质量的溢价足够高:价格为 1 时市场上车辆的平均车况为 0.5,乘以 2.5 得 1.25,仍高于最高车况,市场完全不萎缩。 (https://5dlibrary.com/answers/zh/mechanism-design-negotiation/c05/) - Ch.6 Goodhart's Law: When a Measure Becomes a Target — Goodhart's 1975 wording was "any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes"; the popular "when a measure becomes a target" comes from Strathern in 1997; Campbell in 1979 stressed that indicators corrupt the process they measure. A selection model shows that as long as gaming is cheaper than real improvement and people differ in their talent for gaming, the higher the bonus, the worse the metric becomes at telling people apart. The main value of random audits is that they allow steeper incentives; audits that announce which route they check merely shift gaming, while unannounced random audits can suppress several routes at once, though they may add risk and crowd out unmeasured effort. The next chapter reads these modern results beside ancient Chinese debates on reward, punishment, names and reality. (https://5dlibrary.com/answers/en/mechanism-design-negotiation/c06/) 第6章 古德哈特定律:当指标变成目标 — 古德哈特 1975 年的原话是「任何被观察到的统计规律,一旦为了控制的目的而对它施加压力,就会趋于崩溃」;「当一个指标变成目标」的通俗表述来自斯特拉森 1997;坎贝尔 1979 强调指标会腐蚀被测的过程。选择模型显示,只要操纵比真实改进更便宜、且人们的操纵才能不同,奖金越高,指标区分人的能力越差。随机审计的主要价值在于允许组织使用更陡峭的激励;公布查哪条途径的审计只会转移操纵,不公布的随机审计能同时遏制多条途径,但可能增加风险并挤出不被衡量的努力。下一章把这些现代研究与中国古代关于赏罚与名实的讨论并读。 (https://5dlibrary.com/answers/zh/mechanism-design-negotiation/c06/) - Ch.8 Practice and a Pre-registered Trial — With a smallest effect of interest of 0.35 standard deviations and 90 % power, each arm needs 173 analysable dyads; allowing for about 10 % of dyads being unusable because one member fails to appear or the protocol is breached, each arm randomises 193 dyads, 386 dyads and 772 people in all. If the true effect were only 0.25 standard deviations, each arm would need 338 dyads; this trial does not aim to detect an effect that small. (https://5dlibrary.com/answers/en/mechanism-design-negotiation/c08/) 第8章 实践与预注册试验 — 以 0.35 个标准差为最小关注效应、90% 把握度计算,每组需要 173 个可分析的两人组;考虑约 10% 的两人组因一方缺席或违反程序而无法使用,每组随机化 193 组,共 386 组、772 人。如果真实效应只有 0.25 个标准差,每组需要 338 组;本试验不打算检测那么小的效应。 (https://5dlibrary.com/answers/zh/mechanism-design-negotiation/c08/) ## 5DL-B-028 · Phase Transitions, Critical Slowing Down and Resilience in Complex Dynamical Systems 中文书名:复杂系统相变、临界慢化与生态韧性的非线性动力学 DOI: https://doi.org/10.5281/zenodo.23770082 Cite: Fan, Y. (2026). Phase Transitions, Critical Slowing Down and Resilience in Complex Dynamical Systems (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23770082 Book page: https://5dlibrary.com/books/critical-transitions-resilience/ Scope: The book is not medical advice and makes no investment or financial forecasts; its business and physiological examples are analogies. Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. Abstract: "Cross a threshold, flip suddenly, and find it hard to go back" is said of lakes, coral reefs, the climate, power grids and even a company's cash flow. This book uses models from nonlinear dynamics, runnable programs and empirical research to examine such critical transitions: what they are mathematically, what traces a system leaves as it approaches a threshold, how reliable those traces are in finite data, and what kinds of structure make a system less prone to flip. It covers the two saddle-node bifurcations of a grazing (harvesting) model (c1 = 1.79 and c2 = 2.60 with illustrative parameters), the finding that once c2 is crossed the pressure must fall about 31 % below the collapse point for recovery, and reversible, hysteretic and irreversible lakes in a phosphorus model; the evidence for multiple stable states in lakes and the dispute over "alternative stable states" on coral reefs (about 4 % of 1,851 reefs dominated by macroalgae); the square-root law of critical slowing down and Wissel's extrapolation, the derivation and simulation check of AC(1) = exp(−λΔt) and Var = σ²/(2λ) in the Ornstein–Uhlenbeck process, and evidence from palaeoclimate, a whole-lake experiment and yeast; ROC evaluation of Kendall's τ indicators (AUC 0.98 for variance and 0.94 for autocorrelation with dense records, a τ cut-off of about 0.8 to hold false alarms to 5 %, close to a coin toss with only the last short stretch), noise-induced flips, flickering and the "prosecutor's fallacy", and a test on 126 long-term ecological data sets in which true positives were rare; how Pimm's engineering resilience and Holling's ecological resilience can rank systems in opposite orders, and how suppressing small fires or removing an ungrazeable reserve shrinks a basin of attraction in models (presented as conclusions under model conditions and generalisations from cases); Motter and Lai's capacity–load model (in a 300-node scale-free network scanning in steps of 0.1, about 140 % spare capacity is needed to withstand an attack on the highest-load node, a figure specific to that illustrative network) and the first-order collapse of interdependent networks (p_c = 2.4554/⟨k⟩, becoming continuous when dependence is weakened); and Perrow's normal accident theory, high-reliability theory and the price of modularity as a firebreak. Chapter 7 reads this research beside "Treading on frost, the hard ice is coming" and its commentaries in the Book of Changes, the sayings of Confucius quoted in the "Great Commentary" ("danger comes to one who rests secure in his position", "to know the incipient is divine"), chapter 64 of the Laozi and the dyke and the story of Bian Que in the "Illustrations of Laozi" of the Han Feizi (noting that "a dyke a thousand li long" is a later popular form and that the Bian Que story is best read as a parable), and states that the ancients expressed political and moral foresight and experiential analogy, not differential equations or statistical tests. Chapter 8 gives a four-step resilience and early-warning checklist (with a fictional water system as example) and a pre-registered trial randomised by individual: in a simulated resource-management task with a hidden fold, it compares an added dashboard of rolling recovery time and variance with current stock and profit only (single primary outcome: the share of irreversible collapses; 203 analysable and 226 recruited per arm, 452 in all; decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. The book is not medical advice and makes no investment or financial forecasts; its business and physiological examples are analogies. The author's interests are disclosed in the book. 摘要:「越过门槛、突然翻转、难以回头」常被用来描述湖泊、珊瑚礁、气候、电网乃至企业的资金链。本书用非线性动力学模型、可运行的程序与经验研究考察这类临界转变:它在数学上是什么,系统接近门槛时会留下什么痕迹,这些痕迹在有限的数据中有多可靠,以及怎样的结构使系统更不容易翻转。依次讨论:放牧(收获)模型的两个鞍结分岔(示意参数下 c1 = 1.79、c2 = 2.60),越过 c2 之后压力须比崩溃点低约 31% 才能恢复,以及湖泊磷模型中可逆、滞后与不可逆三类湖泊;湖泊多稳态的证据与珊瑚礁「可替代稳态」的争议(1,851 个珊瑚礁中约 4% 由大型藻类主导);临界慢化的平方根定律与维塞尔外推,奥恩斯坦—乌伦贝克过程中 AC(1) = exp(−λΔt) 与 Var = σ²/(2λ) 的推导与验证,以及古气候、全湖实验与酵母实验的证据;用 ROC 曲线评估肯德尔 τ 指标(记录密集时方差与自相关的 AUC 为 0.98 与 0.94,误报率控制在 5% 时 τ 门槛约为 0.8,只看最后一小段时接近抛硬币),噪声诱导的翻转、闪烁与「检察官谬误」,以及 126 组长期生态数据中真阳性很少的检验结果;皮姆的工程韧性与霍林的生态韧性可以给出相反的排序,压制小火与去掉不可采食的储备在模型中怎样缩小吸引盆(写为模型条件下的结论与案例归纳);莫特—赖容量—负荷模型(300 个节点的无标度网络中,按 0.1 步长扫描,挡住对负荷最高节点的攻击需要约 140% 的冗余,只适用于该示意网络)与相互依赖网络的一阶崩溃(p_c = 2.4554/⟨k⟩,依赖减弱后变为连续);佩罗的常态事故理论、高可靠性组织理论,以及模块化作为防火墙的代价。 第七章把这些研究与《周易·坤》初六「履霜,坚冰至」及其《文言》、《系辞下》所引孔子之言「危者,安其位者也」与「知几其神乎」、《老子》第六十四章「其安易持,其未兆易谋」以及《韩非子·喻老》「千丈之堤以蝼蚁之穴溃」与扁鹊见蔡桓公并读(并说明「千里之堤」是后来的流行说法,扁鹊故事宜作寓言读),明确古人讲的是政治与道德上的忧患意识和经验类比,不是微分方程或统计检验;第八章给出「四步韧性与早期预警核查表」(以虚构的供水系统为例),以及一项以个人为单位随机化的预注册试验方案:在含隐藏折叠点的模拟资源管理任务中,比较「另加滚动恢复时间与方差仪表盘」与「只看当前存量与当期收益」,唯一主要结局为资源不可逆崩溃的比例,每组 203 名可分析、招募 226 名,共 452 人,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部核对)、原典与作者的假说。本书不是医疗建议,也不作投资或金融预测;企业与生理的例子都是类比。作者的利益关系已在书中披露。 Chapter answers: - Ch.1 Folds, Hysteresis and Flips That Are Hard to Undo — Systems with positive feedback can have two stable states over a range of conditions. In the grazing model, once the pressure exceeds $c_2 \approx 2.60$ the system jumps to the low state, and recovery requires a fall below $c_1 \approx 1.79$, about 31 % below the collapse point. In the lake phosphorus model the same structure yields reversible, hysteretic and irreversible behaviour, depending on the loss rate. The evidence for multiple stable states and slow recovery is good for lakes; whether coral reefs have alternative stable states remains disputed. The next chapter asks whether the approach to a fold can be seen in a system's response to small disturbances. (https://5dlibrary.com/answers/en/critical-transitions-resilience/c01/) 第1章 折叠分岔、滞后与难以回头的翻转 — 具有正反馈的系统在一段条件范围内可以有两个稳定状态。放牧模型中,压力越过 $c_2 \approx 2.60$ 后系统跳到低稳态,要恢复必须退到 $c_1 \approx 1.79$ 以下,比崩溃点低约 31%。湖泊磷模型中,同一结构可以给出可逆、滞后与不可逆三种行为。湖泊的多稳态与缓慢恢复有较好的证据;珊瑚礁是否有可替代的稳态仍有争议。下一章讨论,系统接近折叠点时能否从它对小扰动的反应中看出来。 (https://5dlibrary.com/answers/zh/critical-transitions-resilience/c01/) - Ch.2 Critical Slowing Down: Why Recovery Gets Slower — Near a fold the recovery rate falls to zero as a square root; in the grazing model the recovery time measured far from the threshold extrapolates the threshold fairly accurately. Under random disturbance, slower recovery appears as rising lag-1 autocorrelation $e^{-\lambda\Delta t}$ and variance $\sigma2/(2\lambda)$; rising variance alone may only mean more noise. Similar signals have been seen in an experimental lake, in yeast and in palaeoclimate records. But the closer the threshold, the more dangerous the measurement and the more data it takes. (https://5dlibrary.com/answers/en/critical-transitions-resilience/c02/) 第2章 临界慢化:恢复为什么越来越慢 — 接近折叠点时,恢复速率按平方根趋于零;在放牧模型中,用远离门槛处的恢复时间可以较准确地外推门槛的位置。在随机扰动下,恢复变慢表现为一阶自相关 $e^{-\lambda\Delta t}$ 与方差 $\sigma2/(2\lambda)$ 的上升;只有方差上升可能只是噪声变大。实验湖泊、酵母与古气候记录中观察到了类似的信号。但离门槛越近,测量本身越危险,估计所需的数据也越多。 (https://5dlibrary.com/answers/zh/critical-transitions-resilience/c02/) - Ch.3 The Limits of Early Warning: Signal to Noise, Selection Bias and Flips Without Warning — Warning indicators must be judged by detection and false-alarm rates together. In the grazing model, with dense records spanning the whole change, Kendall's $\tau$ for variance and autocorrelation separates well (AUC 0.98 and 0.94), but holding false alarms to 5 % needs a cut-off of about 0.8; with sparser records autocorrelation fails first, and with only the last short stretch both are close to a coin toss. Noise-induced flips and flickering can happen far from a threshold; selecting only series that have collapsed makes us "see" warnings in a system whose parameters never changed. In 126 real data sets true positives were rare and false positives commoner than false negatives. The next chapter takes another angle: instead of forecasting when a threshold will arrive, it asks how far away the threshold is and how that distance can be increased. (https://5dlibrary.com/answers/en/critical-transitions-resilience/c03/) 第3章 早期预警的局限:信噪比、选择偏差与没有预兆的翻转 — 预警指标的价值要用检出率与误报率同时衡量。在放牧模型中,记录密集、跨越整个变化过程时,方差与自相关的肯德尔 $\tau$ 区分得很好(AUC 0.98 与 0.94),但要把误报率控制在 5%,$\tau$ 的门槛要设在约 0.8;记录变稀时自相关先失效,只看最后一小段时两者都接近抛硬币。噪声诱导的翻转与闪烁可以发生在离门槛很远处;只挑选已经崩溃的序列,会在参数从未改变的系统中「看到」预兆。在 126 组真实数据中,真阳性很少,假阳性比假阴性更常见。下一章换一个角度:不预报门槛何时到来,而问系统离门槛有多远,怎样让这段距离变大。 (https://5dlibrary.com/answers/zh/critical-transitions-resilience/c03/) - Ch.4 Two Kinds of Resilience: Recovering Quickly or Withstanding the Blow — The reserve model: with a reserve $R = 1.0$, two stable states coexist only over a narrow range of pressures (2.624–2.676), and at high pressure the vegetation settles at 2.33; with the reserve cut to 0.5 the range widens to 2.308–2.632 and the collapsed state is 1.34; with no reserve the range is 1.788–2.604 and the collapsed state only 0.40. At $c = 2.3$ the system with a reserve has only one stable state, with no "other valley" to fall into. An "efficiency gain" that makes the unreachable part grazeable widens the range of hysteresis and deepens the collapse; the price of the reserve is a slightly smaller harvest in ordinary years. (https://5dlibrary.com/answers/en/critical-transitions-resilience/c04/) 第4章 两种韧性:恢复得快,还是承受得住 — 林火模型:完全不扑火时,每年烧掉的燃料波动较大(标准差 1.45),燃料维持在低水平(前 50 年平均 3.9),300 年内树冠火的概率只有 3%。扑灭 90% 与 95% 的起火时,每年的波动降到 1.10 与 0.56,系统看起来平静得多;但燃料不断积累(前 50 年平均 17.6 与 20.8),300 年内几乎必然发生树冠火。$s = 0.95$ 时 100 年内的概率(72%)低于 $s = 0.9$(87%):灾难被推迟了,没有被避免。$s = 0.5$ 时波动反而增大(2.17):火更少,但每一场更大。机制是一个慢变量:小火在不断消耗燃料,使系统远离门槛;压制小火之后,快变量变平稳,慢变量却向门槛逼近。这一结论依赖一个模型条件:小扰动确实在消耗那个把系统推向门槛的慢变量。没有这样的慢变量,削平小波动不一定有害。 (https://5dlibrary.com/answers/zh/critical-transitions-resilience/c04/) - Ch.5 Cascading Failure: Spare Capacity and Interdependence — In the capacity–load model, the more uneven the load, the more easily the failure of one key node brings the network down: in a 300-node scale-free network, 10 % spare capacity is enough for random faults, but withstanding an attack on the highest-load node takes about 140 %. The most important redundancy looks most like waste in ordinary times. Two one-to-one interdependent random networks collapse completely and suddenly when about 61 % of nodes are kept, while a single network vanishes continuously only at 25 %; with weaker dependence the collapse becomes continuous again. All these conclusions rest on explicit model assumptions. (https://5dlibrary.com/answers/en/critical-transitions-resilience/c05/) 第5章 级联失效:冗余容量与相互依赖 — 在容量—负荷模型中,负荷越不均匀,越容易被一个关键节点的失效拖垮:300 个节点的无标度网络中,10% 的冗余足以应付随机故障,但挡住对负荷最高节点的攻击需要约 140%。最关键的冗余平时最像浪费。两个一对一相互依赖的随机网络在保留约 61% 的节点时突然全部瘫痪,单个网络要到 25% 才连续地消失;依赖减弱后崩溃又变为连续的。这些结论都依赖明确的模型假设。 (https://5dlibrary.com/answers/zh/critical-transitions-resilience/c05/) - Ch.6 Tight Coupling and Modularity: Normal Accidents and High Reliability — The firebreak is real, but it is pronounced only when cross-module links are very few. With $\mu = 0.01$ (about 12 cross-module links in the whole network), a mean of 35 nodes fail, nearly two thirds (65 %) of failures stay within the starting module, and the chance that more than 100 nodes fail is only 6 %. With 5 % cross-module links, the mean rises to 136 failed nodes and the chance of more than 100 to 54 %; at 10 % the network differs little from a fully random one ($\mu = 0.9$). A handful of cross-module links is enough to carry the failure out. (https://5dlibrary.com/answers/en/critical-transitions-resilience/c06/) 第6章 紧耦合与模块化:常态事故与高可靠性 — 防火墙的作用是真实的,但只在跨模块的连接很少时才明显。$\mu = 0.01$(全网只有约 12 条跨模块连接)时,平均失效 35 个节点,近三分之二(65%)的失效停在起始模块里,超过 100 个节点失效的概率只有 6%。跨模块连接增加到 5%,平均失效升到 136 个,超过 100 个的概率升到 54%;到 10%,与完全随机的网络($\mu = 0.9$)已相差不大。几条跨模块的连接就足以把失效带出去。 (https://5dlibrary.com/answers/zh/critical-transitions-resilience/c06/) - Ch.8 Practice and a Pre-registered Trial — Chapter 8 gives a four-step resilience and early-warning checklist (with a fictional water system as example) and a pre-registered trial randomised by individual: in a simulated resource-management task with a hidden fold, it compares an added dashboard of rolling recovery time and variance with current stock and profit only (single primary outcome: the share of irreversible collapses; 203 analysable and 226 recruited per arm, 452 in all; decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. The book is not medical advice and makes no investment or financial forecasts; its business and physiological examples are analogies. The author's interests are disclosed in the book. (https://5dlibrary.com/answers/en/critical-transitions-resilience/c08/) 第8章 实践与预注册试验 — 第八章给出「四步韧性与早期预警核查表」(以虚构的供水系统为例),以及一项以个人为单位随机化的预注册试验方案:在含隐藏折叠点的模拟资源管理任务中,比较「另加滚动恢复时间与方差仪表盘」与「只看当前存量与当期收益」,唯一主要结局为资源不可逆崩溃的比例,每组 203 名可分析、招募 226 名,共 452 人,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部核对)、原典与作者的假说。本书不是医疗建议,也不作投资或金融预测;企业与生理的例子都是类比。作者的利益关系已在书中披露。 (https://5dlibrary.com/answers/zh/critical-transitions-resilience/c08/) ## 5DL-B-029 · Vocal Biomechanics, Non-Linear Source-Filter Coupling and Auditory-Motor Feedback 中文书名:嗓音生物力学、非线性源-滤波耦合与听觉-运动反馈的数学模型 DOI: https://doi.org/10.5281/zenodo.23770084 Cite: Fan, Y. (2026). Vocal Biomechanics, Non-Linear Source-Filter Coupling and Auditory-Motor Feedback (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23770084 Book page: https://5dlibrary.com/books/vocal-biomechanics-auditory-loop/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. Abstract: Why can the vocal folds vibrate by themselves in an airflow? How does the air column of the vocal tract push and pull back on them? And how do singers hear themselves through a loop that is delayed and runs through air and bone at once? This book is a sequel to Volume 10 of the series, The Biomechanics of Phonation and Attentional Shift. It covers why, in Titze's 1988 small-amplitude theory, the phase difference between the lower and upper margins is what lets the folds gain energy (an unchanging or synchronously moving shape does zero net work), with a check of the threshold pressure against the general form 2k_tBcξ01²/[T(ξ01+ξ02)], which reduces to k_tBcξ0/T for a rectangular glottis, whereas the framework's 2k_tBcξ0/T is twice too high; the evidence that "hydration lowers viscosity", which comes from excised canine larynges and small human experiments of 6 and 12 people, appears mainly at high pitch and never measured viscosity in people; the premise of linear source–filter theory and Titze's two levels of interaction, with vocal-tract inertance lowering the threshold of an illustrative model by about a third and an equal compliant load raising it by about 73 %, and with bifurcations more frequent when F0 crosses F1 although "voice breaks" are mainly a laryngeal register mechanism; straws, which lower the first resonance to about 200 Hz (208 Hz in the program), with a compliant band and then a wide inertive band above it, and create back pressure, while evidence of practical effects remains limited and inconsistent; the singer's formant, produced by the area ratio between the epilarynx (larynx tube) exit and the pharynx (about 1:6 in 1974, about 4.5:1 in 1995 MRI) rather than the laryngeal ventricle, and, under stated assumptions, a gain in 2.5–3.2 kHz band signal-to-noise ratio roughly equal to the height of the formant cluster (about 8.8 dB for a 10 dB cluster), so that "a gain of 10–15 dB" is not an established measurement; an asymmetric two-mass model moving from 1:1 synchrony to 2:1 and 3:1 locking, a generic Duffing oscillator showing period doubling on the way to chaos, regular period doubling that pushes shimmer to 35 %, and irregular vibration in some pathological voices that cannot be used for diagnosis; the pitch-shift reflex with its latency of about 100–150 ms (up to about 200 ms in some studies), the DIVA model, and how an internal model with a Kalman estimate removes the oscillation of high-gain slow feedback but cannot break the error floor of about 5.4 cents set by the delay; and the Lombard and sidetone effects, air and bone conduction summing in the same cochlea, ear plugs that leave the own voice heard almost entirely by bone, and monitor mixes that restore the natural proportion. The eight chapters are read beside Shi Yi on song in the Record of Music ("speech drawn out", "like a string of pearls"), the piping of man, earth and heaven in the Zhuangzi, the Yuan-dynasty Treatise on Singing and the Qing Transmitting the Sound of Yuefu (two sentences quoted in the framework could not be verified word for word and are only paraphrased), and the Śūraṅgama Sūtra's "turn the hearing back to hear one's own nature" (as a metaphor only, echoing Volume 2), stating clearly the boundary between religious practice and acoustic training. They end with a four-step checklist of voice biomechanics and auditory feedback and a pre-registered randomised controlled trial comparing 10 minutes of straw phonation (including a tube in water) plus bone-conduction self-monitoring calibration with 10 minutes of a usual warm-up of reading aloud; the single primary outcome is the smoothed cepstral peak prominence (CPPS) of a read passage immediately afterwards, with Borg CR10 self-rated effort as a secondary outcome, a blinded acoustic analyst, 98 analysable and 109 recruited participants per arm, 218 in all, and a decision rule written as code. Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. The book does not diagnose or treat any voice disorder; hoarseness lasting more than about two weeks, pain on phonation or coughing up blood needs a doctor (an ENT specialist, and a speech-language therapist). 摘要:声带为什么能在气流中自己振动?声道里的空气柱怎样反过来推拉声带?歌者又是怎样通过一个有延迟、同时经由空气和骨头的回路听见自己的?本书是本系列第 10 本《发声的生物力学与注意力转移》的续篇,依次讨论:提策 1988 年小振幅理论中上下缘相位差为何是声带获得能量的关键(静止或同步运动的形状净功为零),并按原文一般式核对起振阈压 2k_tBcξ01²/[T(ξ01+ξ02)]——矩形声门时化为 k_tBcξ0/T,框架中的 2k_tBcξ0/T 会高估一倍;「水化降低黏度」的证据来自离体犬喉与 6 人、12 人的小样本人体实验,影响主要见于高音,黏度并未在人身上直接测得;线性源—滤波理论的前提与提策 2008 年的两级相互作用,声道惯性抗使示意模型的阈压降低约三分之一、等大的容性负载使之升高约 73%,f0 与 F1 交叉时分岔更常见但「破音」主要是喉部音区机制;吸管把第一共振峰降到约 200 Hz(程序 208 Hz),其上先有一段容性区再有一大段惯性区,并产生背压,实际效果的证据仍有限而不一致;歌手共振峰起作用的是喉前庭(喉管)出口与咽腔的面积比(1974 年约 1:6,1995 年磁共振约 4.5:1),不是喉室,明示假设下 2.5–3.2 kHz 频带信噪比的改善约等于共振峰簇的高度(10 dB 簇约 8.8 dB),「提升 10–15 dB」不是已证实的测量;不对称双质量模型从 1:1 同步走到 2:1、3:1 锁频,通用杜芬振子演示倍周期通向混沌,有规律的倍周期可使振幅微扰高达 35%,某些病理嗓音中的不规则振动不能用于诊断;变调反射约 100–150 毫秒(部分研究约 200 毫秒)的潜伏期,DIVA 模型,以及内部模型加卡尔曼估计如何消除高增益慢反馈的振荡、却无法突破延迟造成的约 5.4 音分误差下限;隆巴德效应与侧音效应,骨导与气导在同一耳蜗相加,耳塞使自身声音几乎全部来自骨导、返听混音可恢复自然比例。 八章与《乐记·师乙》「长言之」「累累乎端如贯珠」、《庄子·齐物论》人籁地籁天籁、元代《唱论》与清代《乐府传声》(作者所引两句未能逐字核对,只作转述)以及《楞严经》「反闻闻自性」(只作比喻,与第 2 本呼应)并读,并明确宗教修行与声学训练的界限;最后给出一份四步嗓音生物力学与听觉反馈核查表,以及一项比较 10 分钟「吸管发声(含水阻管)+ 骨导自我监听校准」与 10 分钟常规开嗓朗读的预注册随机对照试验方案:唯一主要结局为练习后立即朗读段落的平滑倒谱峰显著度(CPPS),Borg CR10 自评费力为次要结局,声学分析者设盲,每组 98 人可分析、招募 109 人,共 218 人,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序,全部核对)、原典与作者的假说。本书不作任何嗓音疾病的诊断或治疗;持续嘶哑超过约两周、发声疼痛或咳血请就医(耳鼻喉科、言语治疗师)。 Chapter answers: - Ch.1 Why the Vocal Folds Vibrate by Themselves: Phase Difference and Threshold Pressure — The energy of self-sustained vocal fold oscillation comes from the phase difference between the lower and upper margins: it makes the glottis more convergent while opening and more divergent while closing, so that the pressure "pushes at the right moment"; no unchanging or synchronously moving shape gains net energy. Checked against Titze's 1988 general form, the threshold pressure is $2k_tBc\xi_{01}^2/[T(\xi_{01}+\xi_{02})]$, reducing to $k_tBc\xi_0/T$ for a rectangular glottis as in Volume 10; the framework's $2k_tBc\xi_0/T$ is twice too high. The threshold rises with damping, mucosal-wave speed and convergence, and falls as the medial surface thickens. The evidence that hydration affects ease of onset comes from excised canine larynges and small human experiments, appears mainly at high pitch, and treats tissue viscosity as an explanation rather than a measurement. The next chapter turns to the last term of the theory: how the vocal tract changes the threshold in return. (https://5dlibrary.com/answers/en/vocal-biomechanics-auditory-loop/c01/) 第1章 声带为什么能自己振动:相位差与起振阈压 — 声带自激振荡的能量来自上下缘之间的相位差:它使声门在张开时偏收敛、闭合时偏发散,让压力「在恰当的时刻推」;任何静止或同步运动的形状都得不到净能量。按提策 1988 年的一般式核对,起振阈压为 $2k_tBc\xi_{01}^2/[T(\xi_{01}+\xi_{02})]$,矩形声门时化为 $k_tBc\xi_0/T$,与第 10 本一致,框架中的 $2k_tBc\xi_0/T$ 会高估一倍。阈压随阻尼、黏膜波速度、收敛程度增加而升高,随内侧面变厚而降低。水化影响起振难易的证据来自离体犬喉与小样本的人体实验,主要见于高音,组织黏度是解释而不是直接测量。下一章看理论中的最后一项:声道如何反过来改变阈压。 (https://5dlibrary.com/answers/zh/vocal-biomechanics-auditory-loop/c01/) - Ch.2 Nonlinear Source–Filter Coupling: Inertance, Compliance and the Semi-Occluded Tract — Linear source–filter theory assumes that the source is unaffected by the tract, which holds when glottal impedance far exceeds tract impedance; with a narrow epilarynx the two become comparable and nonlinear interaction appears. Titze (2008) divides it into a level that drives the airflow and a level that drives the folds. Below its first formant the tract is inertive, and inertance puts the pressure above the glottis in step with fold motion, lowering the threshold pressure (by about a third in the sketch), while a compliant load raises it. Bifurcations are more frequent when F0 crosses F1, but "voice breaks" are chiefly a laryngeal register mechanism, with acoustic loading as one factor. A straw lowers the first resonance to about 200 Hz, changes the reactance region in which each harmonic falls and creates back pressure; consistent evidence for its practical effects is still lacking. The next chapter returns to the singer's formant of Volume 10 and does the signal-to-noise accounting under stated assumptions. (https://5dlibrary.com/answers/en/vocal-biomechanics-auditory-loop/c02/) 第2章 非线性源—滤波耦合:惯性抗、容性抗与半阻塞声道 — 线性源—滤波理论假定声源不受声道影响,这在声门阻抗远大于声道阻抗时成立;喉前庭狭窄时二者相当,出现非线性相互作用。提策 2008 把它分为驱动气流与驱动声带两级。声道在第一共振峰以下呈惯性,惯性使声门上方的压力与声带运动同相,降低起振阈压(示意模型中约降低三分之一),容性负载则提高阈压。f0 与 F1 交叉时分岔更常见,但「破音」主要是喉部音区机制,声学负载只是因素之一。吸管把第一共振峰降到约 200 Hz、改变谐波所处的电抗区域并产生背压;它的实际效果仍缺乏一致的证据。下一章回到第 10 本讲过的歌手共振峰,在明示假设下算一笔信噪比的账。 (https://5dlibrary.com/answers/zh/vocal-biomechanics-auditory-loop/c02/) - Ch.3 The Singer's Formant and Orchestral Masking: An Account with Stated Assumptions — The singer's formant arises from a clustering of the third to fifth formants near 3 kHz; what produces it is the area mismatch between a narrow epilarynx and a wider pharynx (about 1:6 in the 1974 model, about 4.5:1 in the 1995 MRI measurements), not the laryngeal ventricle. Its frequency coincides with the resonance of the outer ear, while orchestral energy is concentrated in the few hundred hertz range. Under stated spectral assumptions, the gain in signal-to-noise ratio in the 2.5–3.2 kHz bands is about the number of decibels by which the formant cluster exceeds a speech-like spectrum (about 8.8 dB for a 10 dB cluster), and the absolute ratio depends on orchestration; "a gain of 10–15 dB" is not an established measurement. The next chapter turns to the nonlinearity of the source itself: registers, the two-mass model and bifurcations. (https://5dlibrary.com/answers/en/vocal-biomechanics-auditory-loop/c03/) 第3章 歌手共振峰与管弦乐掩蔽:一笔明示假设的账 — 歌手共振峰来自第三至五共振峰在约 3 kHz 附近的聚集,起作用的是狭窄的喉前庭与较宽咽腔之间的面积不匹配(1974 年模型约 1:6,1995 年磁共振约 4.5:1),不是喉室。它的频率与外耳共振重合,而管弦乐的能量集中在几百赫兹。在明示的频谱假设下,2.5–3.2 kHz 频带的信噪比改善约等于共振峰簇高出说话式频谱的分贝数(10 dB 簇约改善 8.8 dB),绝对信噪比则取决于配器;「提升 10–15 dB」不是已证实的测量结果。下一章转向声源本身的非线性:音区、双质量模型与分岔。 (https://5dlibrary.com/answers/zh/vocal-biomechanics-auditory-loop/c03/) - Ch.4 Registers, the Two-Mass Model and Nonlinear Bifurcations — Registers can be understood as switches between laryngeal vibratory mechanisms (M0–M3), and jumps and hysteresis are typical of nonlinear systems. The two-mass model produces the phase difference between the margins naturally from two coupled masses; with left–right asymmetry the sketch moves from 1:1 synchrony to 2:1 and 3:1 locking, producing subharmonics. A generic Duffing oscillator shows the classic period-doubling route to chaos, chaos being the precise property of sensitive dependence on initial conditions. Regular period doubling can push shimmer to 35 %, showing that such measures suit only nearly periodic voices. Irregular vibration is observed in some pathological voices but cannot be used for diagnosis, while kargyraa and rock distortion exploit nonlinearity deliberately. The next chapter turns from the source to the ear: how singers hear themselves through a delayed loop. (https://5dlibrary.com/answers/en/vocal-biomechanics-auditory-loop/c04/) 第4章 音区、双质量模型与非线性分岔 — 音区可以理解为喉部振动机制(M0–M3)的切换,跳变与迟滞是非线性系统的典型特征。双质量模型用两块耦合的质量自然地产生上下缘相位差;加入左右不对称后,示意模型从 1:1 同步走到 2:1、3:1 锁频,出现次谐波。用通用的杜芬振子可以看到经典的倍周期通向混沌,混沌是对初始条件敏感依赖的精确性质。有规律的倍周期会使振幅微扰高达 35%,说明这类指标只适合接近周期的嗓音。某些病理嗓音中观察到不规则振动,但它不能用于诊断;卡尔格拉与摇滚失真音则是对非线性的有意利用。下一章从声源转向耳朵:歌者怎样通过一个有延迟的回路听自己。 (https://5dlibrary.com/answers/zh/vocal-biomechanics-auditory-loop/c04/) - Ch.5 The Auditory-Motor Loop: The Pitch-Shift Reflex, DIVA and Delayed State Estimation — The pitch-shift reflex shows that people monitor themselves continuously while phonating: a shift of pitch in the headphones produces an opposing, partial compensation after about 100–150 ms, and trained singers can suppress it on instruction. The DIVA model organises speech and song as feedforward control plus auditory and somatosensory feedback, with feedback corrections gradually absorbed into feedforward commands; state feedback control theory emphasises internal models and prediction. The programs show that direct delayed feedback has only a narrow range of useful gain and oscillates when the gain is too high; with an internal model and a Kalman estimate a large gain stays stable, but drift during the delay sets a floor on the error that cannot be broken. "Attending to bone conduction and resonance stabilises pitch instantly" is rewritten as a hypothesis about somatosensory weighting under masking. The next chapter looks at how noise and ear plugs change this loop. (https://5dlibrary.com/answers/en/vocal-biomechanics-auditory-loop/c05/) 第5章 听觉—运动闭环:变调反射、DIVA 与带延迟的状态估计 — 变调反射显示,人在发声时持续监听自己:耳机中的音高偏移会在约 100–150 毫秒后引起反向的部分补偿,训练有素的歌者还能按指令抑制它。DIVA 模型把言语与歌唱组织为前馈控制加听觉与体感反馈,反馈纠正逐渐被吸收进前馈;状态反馈控制理论则强调内部模型与预测。程序显示,直接延迟反馈只有一个狭窄的好用增益范围,增益过高即振荡;加入内部模型与卡尔曼估计后,可以用大增益而保持稳定,但延迟期间的漂移形成一个无法突破的误差下限。「注意骨导与共鸣能瞬间稳住音准」改写为一个关于掩蔽条件下体感加权的假说。下一章看噪声与耳塞怎样改变这个回路。 (https://5dlibrary.com/answers/zh/vocal-biomechanics-auditory-loop/c05/) - Ch.6 The Lombard Effect and Mixed Air- and Bone-Conducted Monitoring — The Lombard effect is a robust, nearly automatic reflex: the louder the noise, the louder the voice, and the louder one hears oneself, the softer; instructions alone hardly suppress it, and feedback training is needed. A singer's own voice reaches the same cochlea by air and by bone (bone dominating at 1–2 kHz), and outside noise masks both. The illustrative calculation shows that with open ears in a loud band the own voice is 6–13 dB below the accompaniment, pitch estimation worsens and somatosensory weight rises; ear plugs restore the signal-to-noise ratio but leave the own voice heard almost entirely by bone; a monitor mix can restore the natural proportion of air and bone conduction. Isolation also makes singers sing more softly, lowering the voice-to-band ratio reaching the audience, a reliable direction whose size depends on assumptions. The next chapter leaves measurement behind to read some classical writing on singing and listening. (https://5dlibrary.com/answers/en/vocal-biomechanics-auditory-loop/c06/) 第6章 隆巴德效应与骨导—气导混合监听 — 隆巴德效应是稳健、近乎自动的反射:噪声越大说话越响,听到自己越响说话越轻;单靠指令难以抑制,需要反馈训练。歌者的自身声音经气导与骨导在同一个耳蜗相加(骨导在 1–2 kHz 占优),外界噪声同时掩蔽两者。示意计算显示:在响亮的乐队中张开耳朵,自身声音低于伴奏 6–13 dB,音高估计变差,体感权重上升;耳塞恢复了信噪比,却使听到的自身声音几乎全部来自骨导;返听混音可以恢复气导与骨导的自然比例。隔声还会使歌者唱得更轻,传到听众的歌声与乐队之比下降——方向可靠,大小取决于假设。下一章离开测量,读几段古人关于歌唱与聆听的文字。 (https://5dlibrary.com/answers/zh/vocal-biomechanics-auditory-loop/c06/) - Ch.8 Practice and a Pre-Registered Trial — Taking 0.4 dB as the smallest effect of interest (no minimal important difference in CPPS has been established for healthy speakers; this is an assumption for the author to confirm), a between-person standard deviation of 1.2 dB, a cautious baseline–outcome correlation of 0.7 and 90 % power, each arm needs 98 analysable participants, 196 in all; allowing about 10 % for unusable recordings or withdrawal, 109 are recruited per arm, 218 in all. If the correlation were only 0.6, 123 per arm would be needed. The standard deviation and correlation will be checked in a pilot used for this purpose only and not pooled into the main analysis. (https://5dlibrary.com/answers/en/vocal-biomechanics-auditory-loop/c08/) 第8章 实践与预注册试验 — 以 0.4 dB 为最小关注效应(对健康说话者,目前没有公认的 CPPS 最小重要差值,这是一个需要作者确认的假设)、个体间标准差 1.2 dB、基线与结局的相关取较保守的 0.7、90% 把握度,每组需要 98 名可分析的参与者,共 196 人;考虑约 10% 因录音不合格或退出而无法分析,每组招募 109 人,共 218 人。如果相关只有 0.6,每组需要 123 人。标准差与相关将在一个只用于此目的、不并入正式分析的预试验中核对。 (https://5dlibrary.com/answers/zh/vocal-biomechanics-auditory-loop/c08/) ## 5DL-B-030 · Ageing Kinetics, the mTOR-AMPK Autophagy Switch and Hormesis 中文书名:衰老动力学、mTOR-AMPK自噬开关与毒物兴奋效应的生物数学模型 DOI: https://doi.org/10.5281/zenodo.23770092 Cite: Fan, Y. (2026). Ageing Kinetics, the mTOR-AMPK Autophagy Switch and Hormesis (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23770092 Book page: https://5dlibrary.com/books/longevity-autophagy-hormesis/ Scope: It is not medical advice, makes no recommendation on any anti-ageing drug, supplement or extreme fasting, and gives no doses. Abstract: "Autophagy", "hormesis", "mitochondrial low-dose stress" and "Zone 2" are often combined into one story: eating too often locks cells into a building state, moderate hunger and exercise stress switch on repair, and so the mortality curve can be pushed back. This book takes the story apart with biomathematical models of ageing and the epidemiology of exercise: the Gompertz–Makeham law and a mortality doubling time of about 8 years (a typical value, not a constant), reliability theory as a qualitative account of the exponential rise and late-life deceleration, and a compression of morbidity that US data do not support; the two-input mTORC1–AMPK–ULK1 switch, and why "constant eating makes damaged protein pile up linearly" holds only with zero basal clearance; the definition of hormesis (which excludes "beneficial", with stimulation usually 30–60 %), the difficulty of detecting it, the worm evidence for mitohormesis, and the inconsistent replication of "antioxidants block the benefits of exercise"; a critical clearance rate and hysteresis in a model of proteostasis; the epidemiology of cardiorespiratory fitness (about 13 % lower mortality per MET; an adjusted 5.04 for low versus elite fitness among 122,007 Cleveland Clinic patients, where the comparison of like with smoking's 1.41 is below versus above average, 1.41), the assumptions behind converting hazard ratios into years on the Gompertz curve, reverse causation and genetic confounding, and the null Generation 100 trial; and a Banister model whose verdict on intensity distribution depends on chosen zone weights, with weak human evidence that "Zone 2 specifically raises mitochondria". The chapters are read beside the Lüshi chunqiu ("flowing water does not go foul, a door pivot is not eaten by worms"), Hua Tuo's "the body needs to move, but should not be pushed to the limit" and the Suwen, with a review of poisoning by ancient mineral elixirs, treating classical longevity practice as experience of the rhythms of life rather than mechanism or clinical evidence. Chapter 8 gives a four-step resilience and recovery checklist and a pre-registered randomised trial: sedentary adults receive 12 weeks of 45-minute incline walks three times a week at or below the first ventilatory threshold (grade individualised, 15 % only as an upper limit) or standard steps advice; the single primary outcome is one-minute heart-rate recovery, with 86 analysable and 102 randomised per arm, 204 in all. The book contains 16 runnable Python programs. It is not medical advice, makes no recommendation on any anti-ageing drug, supplement or extreme fasting, and gives no doses. 摘要:「自噬」「毒物兴奋」「线粒体低剂量应激」「Zone 2」常被组合成一个故事:吃得太勤把细胞锁在合成状态,适度的饥饿与运动应激开启修复,死亡率曲线因此可以推迟。本书用衰老的生物数学模型与运动流行病学把这个故事逐段拆开:冈珀茨—梅克姆定律与约 8 年的死亡率翻倍时间(典型值而非常数),可靠性理论对指数上升与高龄减速的定性解释,以及没有得到美国数据支持的「发病压缩」;mTORC1、AMPK 与 ULK1 的两输入开关,以及「持续进食使受损蛋白线性堆积」只在基础清除为零时成立;毒物兴奋效应的定义(不包含「有益」,刺激通常只有 30%–60%)、检测难题与线粒体低剂量应激的线虫证据,「运动期间补充抗氧化剂阻断益处」没有一致重复;蛋白稳态模型中的临界清除速率与滞后;心肺适能流行病学(每高 1 MET 全因死亡风险约低 13%;克利夫兰诊所 122,007 名患者中低对精英的调整风险比 5.04,与吸烟 1.41 可比的是「低于平均对高于平均」的 1.41),风险比换算为冈珀茨曲线上年数的假设,反向因果与遗传混杂,以及 Generation 100 随机试验的阴性结果;班尼斯特模型对强度分布的「判决」取决于人为的区权重,「Zone 2 特异地提高线粒体」在人体中证据不足。 各章与《吕氏春秋·尽数》「流水不腐,户枢不蝼」(通行本作「蝼」)、《三国志·华佗传》「人体欲得劳动,但不当使极尔」与《素问》并读,并检讨古代服食金石丹药的中毒,说明古典养生是生活节律的经验总结,不是分子机制或临床证据。第八章给出四步生理韧性与恢复核查表,以及一项预注册随机对照试验:久坐成年人随机接受 12 周、每周 3 次 45 分钟、强度不超过第一通气阈值的坡道快走(坡度个体化,15% 只作上限)或常规步数建议,唯一主要结局为 1 分钟心率恢复,每组 86 名可分析、随机化 102 名,共 204 人。书中有 16 段可运行的 Python 程序。本书不是医疗建议,不构成任何抗衰老药物、补充剂或极端禁食的建议,不给任何剂量。 Chapter answers: - Ch.1 The Gompertz–Makeham Law: the Shape of the Mortality Curve — The Gompertz–Makeham law $\mu(t) = A + R e^{\alpha t}$ is a robust empirical description of adult mortality. The human MRDT is about 8 years, but this is a typical value rather than a constant, and estimates depend on fitting choices. Reliability theory shows that non-ageing elements plus the exhaustion of redundancy can produce an exponential rise and late-life deceleration, but its simplest version is a qualitative explanation only; whether the late-life plateau is real or an artefact of data errors remains disputed. Fries's compression of morbidity is a testable hypothesis that US data have not supported; in the model, morbidity is compressed only if onset is postponed faster than death, and making deaths more concentrated requires changing the slope of the mortality curve. The next chapter moves from population curves to cells: how nutrient-sensing pathways switch between building and recycling. (https://5dlibrary.com/answers/en/longevity-autophagy-hormesis/c01/) 第1章 冈珀茨—梅克姆定律:死亡率曲线的形状 — 冈珀茨—梅克姆定律 $\mu(t) = A + R e^{\alpha t}$ 是对成年人死亡率的稳健经验描述,人类的 MRDT 约为 8 年,但它是典型值而不是常数,估计值还依赖拟合方法。可靠性理论说明,不老化的元件加上冗余的耗尽,就能产生指数型上升与高龄减速,但最简版本只是定性解释;高龄平台是真实现象还是数据误差仍有争论。弗里斯的发病压缩是一个可检验的假说,美国的数据没有支持它;在模型中,只有发病推迟得比死亡更快才会压缩,而要让死亡年龄更集中,必须改变死亡率曲线的斜率。下一章从群体曲线走进细胞:营养感应通路如何在合成与回收之间切换。 (https://5dlibrary.com/answers/zh/longevity-autophagy-hormesis/c01/) - Ch.2 mTORC1, AMPK and ULK1: the Nutrient-Sensing Switch — mTORC1 integrates nutrient and growth signals, promoting synthesis and inhibiting autophagy; rapamycin extends lifespan in mice, but this is neither human evidence nor a reason to take the drug. ULK1 is where the mTORC1 and AMPK signals converge: in the classical picture AMPK activates and mTORC1 inhibits ULK1, and mutual inhibition makes the switch almost all-or-none, while cell studies from 2023 suggest that AMPK may restrain ULK1 under energy stress, in which case the main signal opening autophagy is nutrient withdrawal. "Eating around the clock makes damaged protein pile up linearly" holds only when basal clearance is zero; in general damage tends to a plateau, and the fasting window changes the plateau's height by an amount that depends on parameters not yet measured in humans. The next chapter turns to another kind of "benefit of a low dose": hormesis. (https://5dlibrary.com/answers/en/longevity-autophagy-hormesis/c02/) 第2章 mTORC1、AMPK 与 ULK1:营养感应的开关 — mTORC1 整合营养与生长信号,促进合成、抑制自噬;雷帕霉素延长小鼠寿命,但这不是人体证据,也不是用药理由。ULK1 是 mTORC1 与 AMPK 两路信号的汇合点:在经典图景中 AMPK 激活、mTORC1 抑制 ULK1,相互抑制使开关接近全或无;而 2023 年的细胞研究提示 AMPK 在能量应激下可能约束 ULK1,此时开启自噬的主要信号是营养撤除。「全天候进食使受损蛋白线性堆积」只在基础清除为零时成立;一般情况下损伤趋向平台,空腹窗口改变的是平台高低,其大小取决于人体中尚未测量的参数。下一章讨论另一类「低剂量的好处」:毒物兴奋效应。 (https://5dlibrary.com/answers/zh/longevity-autophagy-hormesis/c02/) - Ch.3 Hormesis and Mitochondrial Low-Dose Stress — Hormesis is a biphasic dose–response shape whose definition leaves out "beneficial"; low-dose stimulation is usually only 30–60 %, easy both to miss and to invent. Mitohormesis is supported in worms, with no lifespan evidence in humans. "Antioxidants block the benefits of exercise" has not replicated consistently: some muscle markers were blunted, while gains in maximal oxygen uptake, performance and insulin sensitivity were not. Dose, frequency and recovery must be considered together. The next chapter turns to proteostasis and its critical point. (https://5dlibrary.com/answers/en/longevity-autophagy-hormesis/c03/) 第3章 毒物兴奋效应与线粒体低剂量应激 — 毒物兴奋效应是一种双相剂量—反应形状,定义不包含「有益」;低剂量刺激通常只有 30%–60%,容易被漏掉也容易被虚构。线粒体低剂量应激在线虫中有支持,人身上没有寿命证据。「抗氧化剂阻断运动益处」没有一致重复:某些肌肉标志物被削弱,最大摄氧量、成绩与胰岛素敏感性的改善不受影响。剂量、频率与恢复必须一起考虑。下一章讨论蛋白稳态及其临界点。 (https://5dlibrary.com/answers/zh/longevity-autophagy-hormesis/c03/) - Ch.4 Proteostasis, Chaperones and the Critical Clearance Rate — The hallmarks of ageing grew from nine in 2013 to twelve in 2023, adding disabled macroautophagy, chronic inflammation and dysbiosis; the criteria for a hallmark rest mostly on animal experiments. The proteostasis network consists of synthesis, chaperone-assisted folding and degradation, and the sequestration of chaperones by aggregates forms a positive feedback. A mass-balance model shows a critical clearance rate: crossing it makes aggregates jump, and recovery needs more clearance capacity than before the collapse (hysteresis); two declines that are each tolerable can together cross the critical point. These are structural conclusions of a model, and the corresponding rates have not been measured in humans. The next chapter leaves the cell and returns to populations: the epidemiology of cardiorespiratory fitness and mortality. (https://5dlibrary.com/answers/en/longevity-autophagy-hormesis/c04/) 第4章 蛋白稳态、伴侣蛋白与临界清除速率 — 衰老标志从 2013 年的九个扩展为 2023 年的十二个,新增巨自噬失能、慢性炎症与生态失调;「标志」的标准大多建立在动物实验上。蛋白稳态网络由合成、伴侣蛋白折叠与降解三部分组成,聚集体隔离伴侣蛋白构成正反馈。质量守恒模型显示存在临界清除速率:越过它时聚集体突然跳升,恢复则需要比崩溃前更高的清除能力(滞后);两种各自可以承受的衰退合在一起可以越过临界点。这些是模型的结构性结论,人体中的相应速率尚未测量。下一章离开细胞,回到人群:心肺适能与死亡率的流行病学。 (https://5dlibrary.com/answers/zh/longevity-autophagy-hormesis/c04/) - Ch.5 Cardiorespiratory Fitness and Mortality: the Epidemiological Evidence and Its Limits — Blair 1989, Kodama 2009 and Mandsager 2018 agree that cardiorespiratory fitness is strongly and inversely associated with all-cause mortality: about 13 % lower risk per MET, and adjusted hazard ratios of up to 5 between the lowest and highest fitness groups. As the paper puts it, the risk of reduced fitness is "comparable to or greater than" traditional risk factors, and the like-for-like comparison is below versus above average (1.41) against smoking (1.41). On the Gompertz curve, 0.87 is roughly a 1.6-year shift, but only if the hazard ratio is causal and proportional across ages; a synthetic cohort shows that reverse causation and genetic confounding can nearly double the observed log hazard ratio, and the only large randomised trial found no reduction in mortality from organised training over 5 years. Individual differences in the training response are real and about half heritable, but a single measurement cannot identify a "non-responder". The next chapter discusses how to train: intensity distribution, Zone 2 and impulse–response models. (https://5dlibrary.com/answers/en/longevity-autophagy-hormesis/c05/) 第5章 心肺适能与死亡率:流行病学证据及其边界 — 布莱尔 1989、小玉 2009 与曼萨格 2018 一致显示,心肺适能与全因死亡率强烈负相关:每高 1 MET 风险约低 13%,最低与最高适能组之间的调整风险比可达 5 倍。按原文,降低的适能的风险「与传统风险因素相当或更大」,可比的对比是「低于平均对高于平均」的 1.41 与吸烟的 1.41。在冈珀茨曲线上,0.87 约相当于 1.6 年的平移,但这要求风险比是因果的、各年龄成比例;合成队列显示反向因果与遗传混杂可以使观察到的效应放大约一倍,唯一的大型随机试验没有发现有组织的训练在 5 年内降低死亡率。训练反应的个体差异是真实的,并有约一半的遗传力,但单次测量不足以判定「无反应者」。下一章讨论怎样训练:强度分布、Zone 2 与脉冲—响应模型。 (https://5dlibrary.com/answers/zh/longevity-autophagy-hormesis/c05/) - Ch.6 Zone 2, Polarised Training and Banister's Impulse–Response Model — With the first and second lactate (or ventilatory) thresholds as boundaries, intensity falls into three zones; the popular "Zone 2" lies in the upper part of zone 1 of the three-zone scheme, and thresholds need to be measured individually. Banister's fitness–fatigue model describes the rebound in performance during a taper, but its "verdict" on intensity distribution depends entirely on the zone weights chosen, the linear model rewards unlimited volume, and in real data its key parameters have wide confidence intervals and are collinear. Polarised training describes how elite athletes train; randomised trials and a meta-analysis show only a small short-term advantage for maximal oxygen uptake and none for performance. "Zone 2 specifically raises mitochondria" is not supported in humans: volume is linked to mitochondrial content, and high intensity can bring similar fitness gains in less time. The next chapter reads these modern studies beside ancient writings on "movement" and "nourishing life". (https://5dlibrary.com/answers/en/longevity-autophagy-hormesis/c06/) 第6章 Zone 2、极化训练与班尼斯特脉冲—响应模型 — 以第一与第二乳酸(或通气)阈值为界,强度可以分为三个区;流行的「Zone 2」属于三区划分中区 1 的上部,阈值需要个人测量。班尼斯特的体能—疲劳模型能描述减量期的成绩回升,但它对强度分布的「判决」完全取决于人为设定的区权重,线性模型还奖励无限的训练量;用真实数据估计时,关键参数的置信区间很宽、彼此共线。极化训练是对优秀运动员训练方式的描述,随机试验与元分析只显示对最大摄氧量的小幅短期优势,对成绩没有差别。「Zone 2 特异地提高线粒体」在人体中没有得到支持:训练量与线粒体含量有关,高强度在更少的时间里也能带来相近的适能改善。下一章把这些现代研究与古代关于「动」与「养」的文字并读。 (https://5dlibrary.com/answers/zh/longevity-autophagy-hormesis/c06/) - Ch.8 Practice and a Pre-registered Trial — Chapter 8 gives a four-step resilience and recovery checklist and a pre-registered randomised trial: sedentary adults receive 12 weeks of 45-minute incline walks three times a week at or below the first ventilatory threshold (grade individualised, 15 % only as an upper limit) or standard steps advice; the single primary outcome is one-minute heart-rate recovery, with 86 analysable and 102 randomised per arm, 204 in all. The book contains 16 runnable Python programs. It is not medical advice, makes no recommendation on any anti-ageing drug, supplement or extreme fasting, and gives no doses. (https://5dlibrary.com/answers/en/longevity-autophagy-hormesis/c08/) 第8章 实践与预注册试验 — 第八章给出四步生理韧性与恢复核查表,以及一项预注册随机对照试验:久坐成年人随机接受 12 周、每周 3 次 45 分钟、强度不超过第一通气阈值的坡道快走(坡度个体化,15% 只作上限)或常规步数建议,唯一主要结局为 1 分钟心率恢复,每组 86 名可分析、随机化 102 名,共 204 人。书中有 16 段可运行的 Python 程序。本书不是医疗建议,不构成任何抗衰老药物、补充剂或极端禁食的建议,不给任何剂量。 (https://5dlibrary.com/answers/zh/longevity-autophagy-hormesis/c08/) ## 5DL-B-031 · Architectural Acoustics, Reverberation Time and the Environmental Psychology of Built Space 中文书名:建筑声学、混响时间与空间环境心理学的物理模型 DOI: https://doi.org/10.5281/zenodo.23782090 Cite: Fan, Y. (2026). Architectural Acoustics, Reverberation Time and the Environmental Psychology of Built Space (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23782090 Book page: https://5dlibrary.com/books/spatial-acoustics-architectural-biophilia/ Scope: This book is not medical advice; problems with hearing or sleep call for a doctor. Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. Abstract: The same people doing the same work feel very different in a meeting room of glass and concrete and in a study lined with books with trees outside the window; the difference is often put down to the "energy" of a place. This book takes it apart into measurable physical quantities and checks the research on each: the derivation of Sabine's formula RT60 = 0.161 V/A (the constant being 24 ln 10 / c) and Eyring's correction, with one illustrative meeting room falling from about 2.1 s to 0.42 s of reverberation and its speech transmission index (STI) rising from 0.43 to 0.75, while "about 0.6 s" for speech rooms is a standard's upper limit rather than an experimental optimum; normal modes of rectangular rooms, the Schroeder frequency and modal degeneracy in a cube (62 modes on only 15 frequencies), and why at a single seat listener position and low-frequency damping matter at least as much as proportions; how A-weighting hides low-frequency noise, and why the 2002 study by Persson Waye and colleagues is a 32-person experiment whose cortisol effect appeared only in highly noise-sensitive people; the irrelevant speech effect, which does not require understanding (foreign speech, nonsense words and reversed speech all disrupt; spectral change is the key), Hongisto's model, in which the performance decrement falls from about 7 % to about 1 % as STI drops from 0.7 to 0.3, and the failure of masking to protect the next desk; box-counting measurement of fractal dimension and its biases, the wide spread of natural dimensions with 1.3–1.5 only the peak of group preference, and Ulrich's 1984 study as a retrospective comparison of 23 pairs of patients; light measured as melanopic equivalent daylight illuminance at the eye rather than colour temperature (at least 250 lux by day, no more than 10 lux before bed), and a ceiling-height effect that appears only when the ceiling is salient and has not been directly replicated; and the mass balance of indoor CO₂ (six people in a 48.6 m³ room at one air change per hour pass 1,000 ppm in 18 minutes), the COGfx study's CO₂ conditions of about 945 and 1,400 ppm, the null effect of pure CO₂ up to 3,000 ppm when manipulated separately, and the near-impossibility of separating CO₂ from bioeffluents in field data. Chapter 7 reads orientation by shadows in "Gong Liu" and the Kaogongji, siting by the land in the Guanzi, the Book of Burial traditionally attributed to Guo Pu (whose authorship is doubtful), Ji Cheng's Yuanye ("skill lies in adapting and borrowing, refinement in fitness") and Li Yu's writing on dwellings, accepting the solar, wind and drainage reasons for "hills behind, water in front" in the northern-hemisphere monsoon climate while rejecting completely the superstition that burial orientation decides descendants' fortunes. Chapter 8 gives a four-step checklist of environmental physics and a pre-registered within-participant randomised crossover trial comparing intelligible conversation (STI about 0.65) with the same conversation masked to unintelligibility (STI below 0.25) at the same 55 dB(A); the single primary outcome is serial-recall error rate, with 64 analysable and 72 randomised participants and a decision rule written as code. Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. This book is not medical advice; problems with hearing or sleep call for a doctor. 摘要:同样的人做同样的工作,在玻璃与混凝土的会议室里和在满墙书架、窗外有树的书房里,感受很不一样;这种差别常被说成「气场」。本书把它拆成可以测量的物理量,并逐项核对相关研究:赛宾公式 RT60 = 0.161 V/A(常数即 24 ln10 / c)与艾林修正的推导,示意计算中同一会议室的混响时间从约 2.1 秒降到 0.42 秒、语言传输指数(STI)从 0.43 升到 0.75,而语言用房「0.6 秒左右」是标准规定的上限而非实验最适点;矩形房间的简正模式、施罗德频率与正方体的模式简并(62 个模式只落在 15 个频率上),以及在单个座位上听者位置与低频阻尼至少与房间比例同样重要;A 计权如何藏起低频噪声,佩尔松·韦等人 2002 年的研究是 32 人的实验、皮质醇效应只见于高噪声敏感者;无关言语效应不需要「听懂」(外语、伪词、倒放言语都有干扰,关键是频谱变化),洪伊斯托模型中 STI 从 0.7 降到 0.3 时工作表现下降从约 7% 降到约 1%,掩蔽声保护不了紧邻的工位;分形维数的盒计数测量及其偏差,自然形态的维数分布很宽、1.3–1.5 只是群体偏好的峰值,乌尔里希 1984 年是 23 对患者的回顾性比较;以眼睛位置的黑视素等效日光照度而不是色温衡量光照(白天至少 250 勒克斯,睡前不超过 10 勒克斯),层高效应只在天花板显眼时出现且尚无直接重复;室内 CO₂ 的质量守恒方程(6 人在 48.6 立方米、每小时换气 1 次的会议室里 18 分钟超过 1,000 ppm),哈佛 COGfx 研究的 CO₂ 条件约为 945 与 1,400 ppm,分开操纵后纯 CO₂ 到 3,000 ppm 未影响认知成绩,现场数据中 CO₂ 与人体散发物几乎无法分开。 第七章与《诗经·公刘》《考工记·匠人》的日影定向、《管子·乘马》的因地选址、旧题郭璞的《葬书》(作者归属为托名之疑)、计成《园冶》「巧于因借,精在体宜」与李渔《闲情偶寄·居室部》并读,承认「背山面水」在北半球季风气候下的日照、避风与排水理由,同时彻底摒弃葬地方位决定子孙吉凶的迷信。第八章给出四步环境物理核查表,以及一项预注册受试者内随机交叉试验:在相同的 55 dB(A) 下比较可理解对话(STI 约 0.65)与被掩蔽到不可理解(STI 低于 0.25)的条件,唯一主要结局为序列回忆错误率,64 名可分析、随机化 72 名,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部核对)、原典与作者的假说。本书不是医疗建议;听力、睡眠问题请就医。 Chapter answers: - Ch.1 Sabine and Eyring: Reverberation Time and Speech Clarity — Sabine's formula $RT_{60} = 0.161V/A$ follows from an energy balance in a diffuse field, the constant 0.161 being $24\ln 10/c$; Eyring's formula keeps $1-\bar\alpha$ of the energy at each reflection, and the more absorption, the more the two differ (about 22 % shorter for Eyring at a mean coefficient of 0.4). In the illustrative calculation, one meeting room falls from about 2.1 s in glass and concrete to 0.80 s with an absorbent ceiling and 0.42 s with books and fabrics, its STI rising from 0.43 to 0.75 and its reverberant level falling by about 7 dB. "About 0.6 s" for speech rooms is a standard's upper limit rather than an experimental optimum; about 2 s for concert halls is design convention; "1.5–2.2 s for meditation spaces" has no basis. The next chapter turns to where reverberation theory breaks down: low frequencies in small rooms. (https://5dlibrary.com/answers/en/spatial-acoustics-architectural-biophilia/c01/) 第1章 赛宾与艾林:混响时间与语言清晰度 — 赛宾公式 $RT_{60} = 0.161V/A$ 来自扩散声场中的能量平衡,常数 0.161 就是 $24\ln 10/c$;艾林公式考虑每次反射保留 $1-\bar\alpha$ 的能量,吸声越多两者差别越大(平均吸声系数 0.4 时艾林值短约 22%)。示意计算中,同一间会议室从玻璃与混凝土的约 2.1 秒,降到加吸声吊顶后的 0.80 秒、再到加书架与织物后的 0.42 秒,STI 从 0.43 升到 0.75,混响声级约低 7 dB。语言用房「0.6 秒左右」是标准规定的上限而非实验最适点;音乐厅约 2 秒是设计惯例;「冥想空间 1.5–2.2 秒」没有依据。下一章转向混响理论失效的地方:小房间里的低频。 (https://5dlibrary.com/answers/zh/spatial-acoustics-architectural-biophilia/c01/) - Ch.2 Standing Waves, Room Modes and Low-Frequency Noise — The normal frequencies of a rectangular room are given by $f = \frac{c}{2}\sqrt{(n_x/L)^2 + (n_y/W)2 + (n_z/H)^2}$, and the number of modes grows as $f3$. At equal volume, a cube's 62 modes below 200 Hz fall on only 15 frequencies, whereas a 1:1.4:1.9 room has almost no coincidences; the Schroeder frequency $2000\sqrt{T/V}$ marks the boundary between the modal and diffuse regions (about 185–262 Hz for the example room). At a single seat, though, listener position and low-frequency damping matter at least as much as proportions. A-weighting hides low-frequency noise: two ventilation noises at the same 40 dB(A) can differ by 13 dB in C − A. The 2002 experiment by Persson Waye and colleagues suggests that low-frequency noise may affect the cortisol rhythm of highly sensitive people, but the evidence is limited. The next chapter discusses a property of sound that matters more than decibels: the voices in the background. (https://5dlibrary.com/answers/en/spatial-acoustics-architectural-biophilia/c02/) 第2章 驻波、简正模式与低频噪声 — 矩形房间的简正频率由 $f = \frac{c}{2}\sqrt{(n_x/L)^2 + (n_y/W)2 + (n_z/H)^2}$ 给出,模式数随 $f3$ 增长。体积相同时,正方体在 200 赫兹以下的 62 个模式只落在 15 个频率上,1:1.4:1.9 则几乎没有重合;施罗德频率 $2000\sqrt{T/V}$ 标出模式区与扩散区的分界(示例房间约 185–262 赫兹)。但在单个座位上,听者位置与低频阻尼的作用至少与房间比例同样重要。A 计权会把低频噪声藏起来:同为 40 dB(A) 的两种通风噪声,C − A 可以相差 13 dB。佩尔松·韦等人 2002 年的实验提示低频噪声可能影响高敏感者的皮质醇节律,但证据有限。下一章讨论另一种比分贝更重要的声音性质:背景里的人声。 (https://5dlibrary.com/answers/zh/spatial-acoustics-architectural-biophilia/c02/) - Ch.3 Open-Plan Offices and the Irrelevant Speech Effect — The irrelevant speech effect is one of the most robust phenomena in cognitive psychology: background speech increases serial-recall errors by about half on average, almost regardless of level. But it does not require understanding: foreign speech, nonsense words and reversed speech all produce it, and the key is spectral change over time; for reading comprehension, meaning is what matters. At the scale of an office, Hongisto's model writes the performance decrement as an S-shaped function of STI (with a ceiling of about 7 %), and the distraction distance of ISO 3382-3 predicts how disturbed employees are. In the illustrative calculation, absorption, screens and masking shrink the distraction distance from about 25 m to 3–5 m, but do almost nothing for neighbours within 1–2 m. The next chapter turns from the ear to the eye: what is seen through the window. (https://5dlibrary.com/answers/en/spatial-acoustics-architectural-biophilia/c03/) 第3章 开放式办公室与无关言语效应 — 无关言语效应是认知心理学中最稳健的现象之一:背景言语使序列回忆错误平均增加约一半,而且几乎与声级无关。但它不需要「听懂」:外语、伪词、倒放言语都有效应,关键是声音的频谱随时间变化;对阅读理解,意义才重要。在办公室尺度上,洪伊斯托模型把工作表现下降写成 STI 的 S 形函数(上限约 7%),ISO 3382-3 的干扰距离能预测员工的受扰程度。示意计算中,吸声、隔断与掩蔽声把干扰距离从约 25 米缩到 3–5 米,但对 1–2 米内的邻座几乎无效。下一章从耳朵转向眼睛:窗外看到的东西。 (https://5dlibrary.com/answers/zh/spatial-acoustics-architectural-biophilia/c03/) - Ch.4 Trees outside the Window: Fractal Dimension and Biophilic Design — Ulrich's 1984 comparison of records (23 pairs; 7.96 versus 8.70 days in hospital) opened research on window views, but it was not a randomised trial and has not been directly replicated; a systematic review gives attention restoration theory only partial support. Fractal dimension can be measured by box counting, but at finite resolution rough curves are underestimated and the result depends on the range of scales. Natural forms have widely spread values of $D$; 1.3–1.5 is where group preference peaks, with clear individual differences; the skin-conductance and EEG evidence comes from preliminary studies of 24 and 32 people. In the illustrative calculation a box-like skyline has $D$ of about 1.0 and a rough tree line about 1.5. The next chapter discusses two other visible spatial variables: light and ceiling height. (https://5dlibrary.com/answers/en/spatial-acoustics-architectural-biophilia/c04/) 第4章 窗外的树:分形维数与亲生命设计 — 乌尔里希 1984 年的病历比较(23 对,住院 7.96 对 8.70 天)开启了对「窗外景观」的研究,但它不是随机试验,也没有直接重复;注意恢复理论的系统综述只给出部分支持。分形维数可以用盒计数法测量,但有限分辨率下会低估粗糙曲线,结果依赖尺度范围。自然形态的 $D$ 分布很宽,1.3–1.5 是群体偏好的峰值,个体差异明显;皮肤电与脑电的证据来自 24 人与 32 人的初步研究。示意计算中,方盒子楼房的窗景 $D$ 约 1.0,粗糙树冠线约 1.5。下一章讨论另外两个看得见的空间变量:光与层高。 (https://5dlibrary.com/answers/zh/spatial-acoustics-architectural-biophilia/c04/) - Ch.5 Illuminance, Spectrum and Ceiling Height — The effect of light on ipRGCs should be measured as melanopic EDI at the eye rather than by colour temperature; the 2022 consensus recommends at least 250 lux during the day, no more than 10 lux from 3 hours before bed and no more than 1 lux during sleep. The illustrative calculation shows that quantity matters as much as spectrum, and that seats far from a window may not reach the daytime recommendation. Evening light from screens delays sleep and circadian timing (a laboratory study of 12 people); bright daytime light has fairly consistent effects on subjective alertness and inconsistent ones on objective alertness. The ceiling-height effect of Meyers-Levy and Zhu (2007) appears only when the ceiling is salient, rests on small samples and has not been directly replicated; with 25 per cell the power to detect the interaction is only about 0.24. The next chapter discusses an invisible environmental variable: the air. (https://5dlibrary.com/answers/en/spatial-acoustics-architectural-biophilia/c05/) 第5章 照度、光谱与层高 — 光对 ipRGC 的作用应当用眼睛位置的黑视素 EDI 衡量,而不是色温;2022 年共识建议白天不少于 250 勒克斯、睡前 3 小时起不超过 10 勒克斯、睡眠时不超过 1 勒克斯。示意计算表明,光的数量与光谱同样重要,离窗较远的座位白天可能达不到建议值。晚间发光屏幕推迟入睡与节律(12 人的实验室研究),白天强光对主观警觉的作用较一致、对客观警觉不一致。迈耶斯-利维与朱 2007 年的层高效应只在天花板显眼时出现、样本小、尚无直接重复;每单元格 25 人的设计检验交互作用的把握度只有约 0.24。下一章讨论看不见的环境变量:空气。 (https://5dlibrary.com/answers/zh/spatial-acoustics-architectural-biophilia/c05/) - Ch.6 Carbon Dioxide, Ventilation and Decision-Making — Indoor CO~2~ is an indicator of ventilation, and 1,000 ppm is not a health limit. The well-mixed mass balance $V\,dC/dt = NG - Q(C - C_{out})$ shows that six people in a 48.6 m³ room at one air change per hour pass 1,000 ppm in 18 minutes, and that a steady state below 800 ppm needs about 13 L/s of outdoor air per person. Satish and colleagues (2012) and Allen and colleagues (2016) reported marked falls in decision-test scores after injecting pure CO~2~ (about 945–2,500 ppm), but with small samples and calls for replication from the authors themselves; when Zhang and colleagues (2017) manipulated them separately, pure CO~2~ up to 3,000 ppm had no effect while bioeffluents did. The illustrative calculation shows that in field data CO~2~ and bioeffluents can hardly be separated. The next chapter returns to the ancients: how they wrote about wind, water, shadows and houses. (https://5dlibrary.com/answers/en/spatial-acoustics-architectural-biophilia/c06/) 第6章 二氧化碳、通风与决策 — 室内 CO~2~ 是通风的指示物,1,000 ppm 不是健康限值。充分混合的质量守恒方程 $V\,dC/dt = NG - Q(C - C_{out})$ 给出:6 人在 48.6 立方米、每小时换气 1 次的会议室里,18 分钟就超过 1,000 ppm;稳态低于 800 ppm 需要每人约 13 升/秒新风。萨蒂什等人 2012 年与艾伦等人 2016 年报告了注入纯 CO~2~(约 945–2,500 ppm)后决策测验得分的明显下降,但样本小、作者呼吁重复;张等人 2017 年分开操纵后,纯 CO~2~ 到 3,000 ppm 没有影响,人体散发物有影响。示意计算显示,在现场数据中 CO~2~ 与人体散发物几乎无法分开。下一章回到古人:他们怎样谈论风、水、日影与房屋。 (https://5dlibrary.com/answers/zh/spatial-acoustics-architectural-biophilia/c06/) - Ch.8 Practice and a Pre-Registered Trial — Chapter 8 gives a four-step checklist of environmental physics and a pre-registered within-participant randomised crossover trial comparing intelligible conversation (STI about 0.65) with the same conversation masked to unintelligibility (STI below 0.25) at the same 55 dB(A); the single primary outcome is serial-recall error rate, with 64 analysable and 72 randomised participants and a decision rule written as code. Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. This book is not medical advice; problems with hearing or sleep call for a doctor. (https://5dlibrary.com/answers/en/spatial-acoustics-architectural-biophilia/c08/) 第8章 实践与预注册试验 — 第八章给出四步环境物理核查表,以及一项预注册受试者内随机交叉试验:在相同的 55 dB(A) 下比较可理解对话(STI 约 0.65)与被掩蔽到不可理解(STI 低于 0.25)的条件,唯一主要结局为序列回忆错误率,64 名可分析、随机化 72 名,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部核对)、原典与作者的假说。本书不是医疗建议;听力、睡眠问题请就医。 (https://5dlibrary.com/answers/zh/spatial-acoustics-architectural-biophilia/c08/) ## 5DL-B-032 · Algorithmic Game Theory, Braess's Paradox and the Price of Anarchy in Traffic and Attention Networks 中文书名:算法博弈论、布雷斯悖论与流量分配的无秩序代价 DOI: https://doi.org/10.5281/zenodo.23782094 Cite: Fan, Y. (2026). Algorithmic Game Theory, Braess's Paradox and the Price of Anarchy in Traffic and Attention Networks (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23782094 Book page: https://5dlibrary.com/books/algorithmic-game-theory-congestion/ Scope: The models of traffic pools and recommender algorithms are schematic and describe no real platform; the book is not investment, business or marketing advice. Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. Abstract: Why can building one more road make everyone slower? This book uses algorithmic game theory to examine the gap between selfish choices and the collective optimum: where it comes from, how large it can be, what prices or rules can close it, and in what form the same structure appears in attention markets, recommender systems and matching markets. It covers Wardrop's user equilibrium and system optimum; Braess's diamond network, in which a zero-delay shortcut raises travel time from 1.5 to 2.0 (the shortcut is weakly, not strictly, dominant, and the paradox appears only for demand between 2/3 and 2), and the same equations in networks of springs and strings and in direct-current circuits; the definition and origin of the price of anarchy, the 4/3 bound for nonatomic users with linear latencies and the bound for polynomial latencies of degree d (2.151 for d = 4, growing as Θ(d/ln d); the topology-independence result is Roughgarden's of 2003), numerical checks on random networks, and the worst case of 5/2 in games with finitely many players; marginal-cost tolls that make the selfish equilibrium optimal, although users' time plus toll need not fall unless the revenue is returned, and the potential-game structure of finite congestion games, in which improvement dynamics in a six-player Braess game stop at "everyone on the bridge"; Hotelling's linear city with the 1979 correction by d'Aspremont and colleagues, a congestion game of crowded "hot niches", and a "digital Braess paradox" that can arise only in networks that are not series-parallel; greedy, ε-greedy, UCB1 and Thompson-sampling allocation in the multi-armed bandit, the difference between rich-get-richer exposure and learning about quality, and the signal-to-noise ratio of a schematic three-tier traffic pool; and Gale and Shapley's deferred acceptance, proposer optimality and the dominance of truth-telling for proposers (but not for receivers). The eight chapters are read beside "he lived in humble rooms and spent his means on the ditches" from the Shiji's Annals of Xia (the received text reads gouxu, and the line is not continuous with "quick and diligent"), Prince Jin's contrast in the Guoyu between Gonggong's "damming the hundred rivers" and Yu's "dredging the rivers and leading off what was stuck", the Guanzi's "of the five harms, water is the greatest", and the historical records of Dujiangyan, stating that "dredge the shoals deep, build the weir low" is a later maxim for annual maintenance, that Dujiangyan's passive diversion depended on centuries of upkeep, and where classical thought on water control ends and modern network-flow theory begins. They end with a four-step checklist for congestion and diversion and a pre-registered group-randomised controlled experiment comparing a display of the delay each player adds to others, deducted as congestion points, with a display of route times only (six-person groups as the unit of randomisation, a single primary outcome of whether a group is trapped in the Braess equilibrium, 56 analysable and 63 recruited groups per arm, 756 participants, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. The models of traffic pools and recommender algorithms are schematic and describe no real platform; the book is not investment, business or marketing advice. 摘要:多修一条路,为什么每个人都更慢了?本书用算法博弈论考察自私的选择与集体最优之间的差距:它从哪里来、最坏有多大、什么价格或规则能消除它,以及同样的结构在注意力市场、推荐系统与撮合市场中以什么形式出现。依次讨论:瓦德罗普用户均衡与系统最优,布雷斯菱形网络中零延迟捷径使通行时间从 1.5 增至 2.0(捷径是弱占优而非严格占优,悖论只在需求介于 2/3 与 2 之间时出现),以及同一组方程在弹簧—绳子网络与直流电路中的同构;无秩序代价的定义与来历,非原子用户、线性延迟下的 4/3 上界与 d 次多项式延迟下的上界(d = 4 时 2.151,按 Θ(d/ln d) 增长,与拓扑无关的结论出自拉夫加登 2003),随机网络中的数值检验,以及有限人数博弈中最坏情形为 5/2;边际成本通行费使自私均衡与最优重合,但不退还收入时用户的「时间 + 费用」未必下降,有限人数的拥堵博弈是势博弈,6 人布雷斯博弈的改进过程停在「全走桥」;霍特林线性城市及达斯普勒蒙等人 1979 年的修正、热门赛道的拥堵博弈与只在非串并联结构中才可能出现的「数字布雷斯悖论」;多臂老虎机中贪婪、ε-贪婪、UCB1 与汤普森采样的比较,富者愈富的曝光与学习质量的区别,以及一个示意的三级流量池的筛选信噪比;盖尔—沙普利延迟接受算法、提议方最优与提议方说真话的占优性(接受方不成立)。 八章与《史记·夏本纪》「卑宫室,致费于沟淢」(通行本作「沟淢」,与「敏给克勤」两句不相连)、《国语·周语下》太子晋论共工「壅防百川,堕高堙庳」与禹「疏川导滞,钟水丰物」、《管子·度地》「五害之属,水最为大」及都江堰的历史记载并读,写明「深淘滩,低作堰」是后世总结的岁修口诀、都江堰的被动分流依赖历代维护,并明确古代治水观与现代网络流理论的界限;最后给出一份四步去拥堵与分流核查表,以及一项比较「显示自己给他人增加的延迟并扣减拥堵积分」与「只显示各路线通行时间」的预注册小组随机对照实验方案(以 6 人小组为随机化单位,唯一主要结局为小组是否陷入布雷斯均衡,每组 56 个小组可分析、招募 63 个,共 756 人,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序,全部核对)、原典与作者的假说。书中关于流量池与推荐算法的模型都是示意,不描述任何真实平台;本书不是投资、经营或营销建议。 Chapter answers: - Ch.1 Braess's Paradox: One More Road, Everyone Slower — The user equilibrium requires all used routes to be equally fast; the system optimum requires total cost to be as low as possible; the two differ by the marginal delay each person adds for others. In Braess's diamond network a zero-delay shortcut, weakly dominant for each person, raises everyone's travel time from 1.5 to 2.0, while the optimum ignores it; the paradox appears only in a range of demand. The same conditions describe springs and strings and direct-current circuits. The next chapter asks how much worse than the optimum selfish choices can be. (https://5dlibrary.com/answers/en/algorithmic-game-theory-congestion/c01/) 第1章 布雷斯悖论:多修一条路,所有人更慢 — 用户均衡要求被使用的路线一样快,系统最优要求总成本最低,两者相差每个人给别人增加的边际延迟。在布雷斯网络中,对每个人都弱占优的零延迟捷径使通行时间从 1.5 增至 2.0,而最优根本不用它;悖论只在一段需求范围内出现。同一组条件也描述弹簧—绳子网络与直流电路。下一章问:自私最多比最优差多少? (https://5dlibrary.com/answers/zh/algorithmic-game-theory-congestion/c01/) - Ch.2 The Price of Anarchy: How Much Can Selfishness Cost? — The price of anarchy is the ratio of the worst selfish equilibrium to the optimum. In networks with linear delays and nonatomic users it is at most 4/3, and the worst case is Pigou's two-road network whatever the topology; with polynomial delays of degree $d$ the bound is $[1 - d(d+1)^{-(d+1)/d}]{-1}$, which is 2.151 for $d = 4$ and grows as $\Theta(d/\ln d)$. In random networks the PoA is usually far below the bound and first rises, then falls, with demand; a BPR-type road approaches the worst case only under severe overload. The bound does not apply to games with finitely many players, whose worst case is 5/2. The next chapter asks whether prices can remove this loss. (https://5dlibrary.com/answers/en/algorithmic-game-theory-congestion/c02/) 第2章 无秩序代价:自私最多让我们损失多少 — 无秩序代价是最坏的自私均衡与最优之比。在线性延迟、非原子用户的网络中,它不超过 4/3,而且最坏情形就是庇古的两条路网络,与拓扑无关;$d$ 次多项式延迟下上界为 $[1 - d(d+1)^{-(d+1)/d}]{-1}$,$d = 4$ 时为 2.151,按 $\Theta(d/\ln d)$ 增长。随机网络中的 PoA 通常远低于上界,并随需求先升后降;BPR 型道路只有在严重超载时才接近最坏情形。上界不适用于有限人数的博弈,后者的最坏情形是 5/2。下一章问:能不能用价格消除这部分损失? (https://5dlibrary.com/answers/zh/algorithmic-game-theory-congestion/c02/) - Ch.3 Pigouvian Tolls and Potential Games: Pricing the Unseen Delay — Selfish users bear only the average delay, not the marginal delay they add for others. A toll of $x^c'(x)$ on each road at the optimum makes the selfish equilibrium coincide with the optimum; but unless the revenue is returned, users' own "time + toll" need not fall and can even rise. The Stockholm congestion charge shows that prices can change travel substantially. Congestion games with finitely many players are potential games: unilateral improvement must stop at an equilibrium, and in the six-player Braess network it stops at "everyone on the bridge, 120 each"; charging points equal to each player's externality turns the potential into the total cost, and everyone-on-the-bridge is no longer an equilibrium. The next chapter carries the same structure over to attention markets. (https://5dlibrary.com/answers/en/algorithmic-game-theory-congestion/c03/) 第3章 庇古税与势博弈:给看不见的延迟定价 — 在最优处按 $x^c'(x)$ 收费,自私均衡就与最优重合,但不退还收入时用户的「时间 + 费用」未必下降;斯德哥尔摩的拥堵费显示价格能显著改变出行。有限人数的拥堵博弈是势博弈,6 人布雷斯网络中的改进停在「全走桥、每人 120」;按外部性收取积分后,势函数变成总成本,全走桥不再是均衡。下一章把这一结构搬到注意力市场。 (https://5dlibrary.com/answers/zh/algorithmic-game-theory-congestion/c03/) - Ch.4 Congestion in Attention: Hot Niches, Hotelling and a Digital Braess Paradox — Hotelling's linear city predicts clustering when prices are fixed, but under price competition the sellers move apart, because moving towards a rival only invites a price war. In the niche congestion game the equilibrium crowds too many sellers into the hot niche, about 4 % below the optimal total; entering the hot niche imposes twice the loss on others as entering an edge niche, and charging that loss makes the equilibrium optimal. A "digital Braess paradox" requires two linked congestible stages and is an analogy awaiting test. The next chapter turns to how recommender systems allocate exposure. (https://5dlibrary.com/answers/en/algorithmic-game-theory-congestion/c04/) 第4章 注意力的拥堵:热门赛道、霍特林与数字布雷斯悖论 — 霍特林的线性城市在价格固定时预测卖家扎堆,但在价格竞争下这个结论不成立:线性成本下靠得太近时没有价格均衡,平方成本下卖家会尽量远离,因为靠近对手只会引来价格战。在热门赛道的拥堵博弈中,均衡把过多的卖家挤进热门赛道,总收益比最优低约 4%;进入热门赛道给他人造成的损失是进入边缘赛道的两倍,按这一损失收费可以使均衡变成最优。「数字布雷斯悖论」只在注意力的流动经过两个相连的拥堵环节时才可能出现,是一个有待检验的类比。下一章讨论推荐系统怎样分配有限的曝光。 (https://5dlibrary.com/answers/zh/algorithmic-game-theory-congestion/c04/) - Ch.5 Exploration and Exploitation: Bandits, the Matthew Effect and Traffic Pools in Recommendation — Exploiting without exploring locks exposure onto content that was lucky early: in 50 of 60 runs the greedy rule gave the best item almost no chance. Thompson sampling and $\varepsilon$-greedy lose only about a third as much as greedy, and the former is steadier; UCB1 over-explores when click rates are low. The "rich get richer" rule, allocating exposure by likes so far, makes outcomes depend partly on luck, while Thompson sampling makes the concentration of exposure agree with quality, without making exposure equal. In the schematic three-tier traffic pool, choosing the right first-tier metric matters more than adding tiers, and the second tier's sample size decides how accurately the final selection can be made. The empirical evidence that recommendation algorithms create filter bubbles is disputed. The next chapter turns to another allocation problem: two-sided matching. (https://5dlibrary.com/answers/en/algorithmic-game-theory-congestion/c05/) 第5章 探索与利用:推荐中的老虎机、马太效应与流量池 — 只利用不探索会把曝光锁定在早期运气好的内容上;汤普森采样与 $\varepsilon$-贪婪的平均损失只有贪婪的三分之一,前者更稳定,UCB1 在点击率很低时探索过度。「富者愈富」的曝光使结果部分取决于运气,汤普森采样使曝光的集中与质量一致,但不使曝光平均。在示意的三级流量池中,选对第一级指标比增加级数更重要。「信息茧房」的经验证据有争议。下一章讨论双边匹配。 (https://5dlibrary.com/answers/zh/algorithmic-game-theory-congestion/c05/) - Ch.6 Stable Matching: Deferred Acceptance, Early Contracting and Telling the Truth — The program checks these properties one by one on 500 random 6 × 6 markets: every result is stable; every proposer gets his best partner among all stable matchings (found by enumerating all 720 pairings); every receiver gets his worst; and shuffling the order of applications five times never changes the result. There is often more than one stable matching: only 183 of the 500 markets have a unique one, 204 have two, and one market has eight. (https://5dlibrary.com/answers/en/algorithmic-game-theory-congestion/c06/) 第6章 稳定匹配:延迟接受、提前签约与说真话 — 程序在 500 个随机的 6 × 6 市场上逐一检查了这些性质:结果全部稳定;每个提议者得到的都是他在全部稳定匹配(用穷举 720 种配对找出)中最好的伴侣;每个接受者得到的都是他最差的;随机打乱申请次序 5 次,结果都不变。稳定匹配常常不止一个:500 个市场中只有 183 个是唯一的,204 个有两个,最多的一个市场有 8 个。 (https://5dlibrary.com/answers/zh/algorithmic-game-theory-congestion/c06/) - Ch.8 Practice and a Pre-Registered Trial — The sample size is therefore based on much more conservative assumptions: 60 % of control groups trapped and 30 % of arm A groups, two-sided $\alpha = 0.05$, 90 % power. Each arm needs 56 analysable groups; allowing about 10 % of groups to be incomplete because someone drops out, each arm recruits 63 groups, that is, 378 people per arm and 756 in all; the simulated power is 0.89. If the true difference is only 25 percentage points (60 % against 35 %), each arm needs 82 groups. These parameters are assumptions and should be checked against pilot data before the trial is run. (https://5dlibrary.com/answers/en/algorithmic-game-theory-congestion/c08/) 第8章 实践与预注册试验 — 所以样本量按保守得多的假设计算:对照组陷入 60%,A 组 30%,双侧 $\alpha = 0.05$,把握度 90%:每组需要 56 个可分析的小组;考虑约 10% 的小组因有人中途退出而不完整,每组招募 63 个小组,即每组 378 人,共 756 人;模拟检验的把握度为 0.89。若真实差距只有 25 个百分点(60% 对 35%),每组需要 82 个小组。这些参数都是假设值,正式实施前应当用预试验的数据核对。 (https://5dlibrary.com/answers/zh/algorithmic-game-theory-congestion/c08/) ## 5DL-B-033 · Scientific Methodology, Falsifiability and the Mathematical Statistics of the Replication Crisis 中文书名:科学方法论、可证伪性与可重复性危机的数理统计 DOI: https://doi.org/10.5281/zenodo.23782104 Cite: Fan, Y. (2026). Scientific Methodology, Falsifiability and the Mathematical Statistics of the Replication Crisis (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23782104 Book page: https://5dlibrary.com/books/scientific-method-falsifiability/ Scope: This book is not medical, psychological or investment advice. Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses. Abstract: A supplement is "clinically proven", a case report says conversions tripled, a paper reports p < 0.05: why should we believe a claim? This book, the methodological charter of 5D Library, answers with the philosophy of science and mathematical statistics. It covers Popper's falsifiability and "empirical content" (the more possible worlds a statement forbids, the more it says, as in Bar-Hillel and Carnap's semantic information; a theory that forbids nothing in advance carries no information and can be supported by no observation), a Bayesian reading of the Duhem–Quine thesis and Lakatos's progressive and degenerating research programmes; a term-by-term check of Ioannidis's positive-predictive-value formula (the framework's form equals the original; with power 0.20, pre-study odds 0.1 and bias 0.20 the PPV is 13.0 %, not about 12.8 %; all nine rows of the paper's Table 4 are reproduced), with the criticisms of Goodman and Greenland and Jager and Leek's empirical estimate of about 14 %; a Monte Carlo re-run of the 2011 simulations of Simmons et al. (four researcher degrees of freedom together give 57.9 % false positives, or 61.2 % with gender coded 0/1, against 60.7 % in the paper), optional stopping, p-curves and the garden of forking paths; the file drawer, the fail-safe N and the limits of funnel-plot tests, and the winner's curse (with a true d = 0.20 and 20 % power, significant results average about 0.45, an exaggeration of about 2.2 times, not about 0.65 and 3 times); three large replication projects (psychology 35/97 = 36.1 %, economics 11/18, social-science experiments in Nature and Science 13/21), why a replication rate cannot be read directly as a false discovery rate, reasonable benchmarks for confidence-interval capture, and the criticism by Gilbert et al. with the reply; and pre-registration, Registered Reports (44 % positive first hypotheses against 96 % in the standard literature), equivalence tests, and the statistical properties of this series' decision rule (in fact five verdicts; "90 % power" means reaching some verdict of benefit, and the strongest verdict appears only 2.5 % of the time when the true effect equals the smallest effect of interest). The eight chapters are read beside "seek truth from facts" in the Book of Han's "Thirteen Sons of Emperor Jing" (with Yan Shigu's gloss), "no trust without evidence" in the Doctrine of the Mean, "the records are insufficient" in the Analects, Chen Di's criticism of harmonising pronunciations, Gu Yanwu's "copper mined from the mountain", Dai Zhen's "relying on a single piece of evidence to believe it holds generally", Wang Yinzhi's "hold when tested in other chapters" and Liang Qichao's summary "a single piece of evidence does not make a settled conclusion", stating clearly that textual scholarship dealt with historical texts, without experiment or probabilistic inference. They end with a four-step meta-science checklist for not being fooled, applied to the series itself as a live specimen, and a pre-registered randomised controlled trial comparing a 20-minute training in recognising the garden of forking paths and the small-sample winner's curse with reading conventional critical-thinking maxims (primary outcome the discrimination d′ in judging 24 news reports, 12 with a methodological flaw and 12 accurate; 176 analysable and 208 recruited per arm, 416 in all; decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses. This book is not medical, psychological or investment advice. 摘要:一种保健品「经临床研究证实」,一份案例报告说转化率提高了三倍,一篇论文写着 p < 0.05——一个说法,凭什么相信?本书是 5D Library 的方法论总纲,用科学哲学与数理统计回答这个问题。依次讨论:波普尔的可证伪性与「经验内容」(命题禁止的可能世界越多说得越多,巴-希勒尔与卡尔纳普的语义信息量;事先什么都不禁止的理论信息量为零,也不能从任何观察中得到支持),杜恒—蒯因论题的贝叶斯读法与拉卡托斯的进步与退化研究纲领;伊奥尼迪斯阳性预测值公式的逐项核对(框架写法与原文等价;功效 0.20、事前比值 0.1、偏倚 0.20 时为 13.0%,而非约 12.8%;原文表 4 九行全部重现)及古德曼与格林兰的批评、贾格尔与利克约 14% 的经验估计;对西蒙斯等人 2011 年模拟的蒙特卡洛重做(四种研究者自由度合计使假阳性率达 57.9%,性别改按 0/1 编码时为 61.2%,原文 60.7%)、可选停止与 p 曲线、分岔花园;抽屉问题、失效安全数与漏斗图检验的局限,以及赢家诅咒(真实 d = 0.20、功效 20% 时显著结果平均约 0.45、夸大约 2.2 倍,而非约 0.65、3 倍);三个大型重复项目(心理学 35/97 = 36.1%、经济学 11/18、《自然》《科学》社会科学实验 13/21)、重复率为何不能直接换算为假发现率与置信区间捕获率的合理基准,以及吉尔伯特等人的批评与回应;预注册、注册报告(第一个假说的阳性率 44% 对标准文献 96%)、等效性检验,以及本系列判定规则的统计性质(实为五种结论;「90% 把握度」指得到某种获益结论,最强结论在真实效应等于最小关注效应时只有 2.5% 的机会出现)。 八章与《汉书·景十三王传》「实事求是」(及颜师古注)、《中庸》「无征不信」、《论语·八佾》「文献不足」、陈第对叶音说的批评、顾炎武「采铜于山」、戴震「据于孤证以信其通」、王引之「验之他卷而通」及梁启超「孤证不为定说」的概括并读,并说明考据学处理的是历史文本,没有实验与概率推断;最后给出一份以本系列自身为活标本自检的四步防忽悠元科学核查表,以及一项比较 20 分钟「分岔花园与小样本赢家诅咒识别训练」与常规批判性思维格言阅读的预注册随机对照试验方案(主要结局为判断 24 篇新闻报道〔12 篇含方法学缺陷、12 篇如实〕的辨别力 d′,每组 176 名可分析、招募 208 名,共 416 人,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(14 段程序,全部核对)、原典与作者的假说。本书不是医疗、心理或投资建议。 Chapter answers: - Ch.1 Falsifiability and Research Programmes: What Counts as a Scientific Statement? — The first case resembles the historical example Dorling discussed: a well-tested theory (0.90 beforehand) with a less reliable auxiliary (0.60). After the failure the theory falls only to 0.783, while the auxiliary drops from 0.60 to 0.130: the blame falls mainly on the auxiliary, as it should. The second case is the reverse: a young theory (0.50) paired with a well-tested auxiliary (0.95) falls to 0.048. The third case shows that if "theory true but auxiliary wrong" would itself rarely produce such a failure, even a theory that was very credible (0.90) falls to 0.067. So the Duhem–Quine thesis describes a logical indeterminacy, but evidence can still discriminate between the parts quite sharply. (https://5dlibrary.com/answers/en/scientific-method-falsifiability/c01/) 第1章 可证伪性与研究纲领:什么才算科学命题 — 第一种情形与多林讨论的历史案例相似:一个久经考验的理论(事先 0.90)加上一个不太可靠的辅助假设(事先 0.60)。预测失败后,理论的可信度只降到 0.783,辅助假设却从 0.60 跌到 0.130:责任主要落在辅助假设上,这是合理的。第二种情形反过来:一个年轻的理论(0.50)配上一个久经检验的辅助假设(0.95),失败后理论跌到 0.048。第三种情形说明,如果「理论为真而辅助假设出错」本身就很难造成这种失败,即使理论事先很可信(0.90),它也会跌到 0.067。所以杜恒—蒯因论题说的是逻辑上的不确定,而证据可以在不同部分之间作出相当明确的区分。 (https://5dlibrary.com/answers/zh/scientific-method-falsifiability/c01/) - Ch.2 Why Most Published Findings May Be False: Ioannidis's Derivation — Ioannidis applied the Bayesian reasoning of diagnostic tests to scientific findings: the positive predictive value is fixed by the pre-study odds $R$, the power and the significance level, and bias $u$ lowers it further. The framework's way of writing the formula is equivalent to the original; with power 0.20, $R = 0.1$ and $u = 0.20$ the PPV is about 13.0 %. Goodman and Greenland pointed out that dichotomising $p$-values hides differences in the strength of evidence and that "most are false" depends on assumptions; Jager and Leek's empirical estimate (about 14 %) rests on contested assumptions of its own. The next chapter looks at where bias $u$ comes from: the researcher's freedom in analysing data. (https://5dlibrary.com/answers/en/scientific-method-falsifiability/c02/) 第2章 为什么大多数已发表的发现可能是错的:伊奥尼迪斯的推导 — 伊奥尼迪斯把诊断检验的贝叶斯推理用到科学发现上:阳性预测值由事前比值 $R$、功效与显著性水平决定,偏倚 $u$ 会使它进一步下降。框架中的公式写法与原文等价;功效 0.20、$R = 0.1$、$u = 0.20$ 时阳性预测值约为 13.0%。古德曼与格林兰指出,把 $p$ 值一刀切会掩盖证据强度的差异,且「大多数是错的」依赖于假设;贾格尔与利克的经验估计(约 14%)本身也依赖于有争议的假设。下一章看偏倚 $u$ 从哪里来:研究者在数据分析中的自由选择。 (https://5dlibrary.com/answers/zh/scientific-method-falsifiability/c02/) - Ch.3 p-Hacking, Researcher Degrees of Freedom and the Garden of Forking Paths — In data with no true effect, choosing among dependent variables, adding observations, adding covariates and dropping conditions each raise the false-positive rate from 5 % to roughly 8 % to 13 %, and together to about sixty per cent (60.7 % in the paper; in this book's re-run 57.9 % with gender effect-coded and 61.2 % with 0/1 coding). The more often one looks at the data, the more false positives. The shape of the $p$-curve can hint whether a literature has been manipulated. Gelman and Loken stressed that even without deliberate fishing, data-dependent choices of analysis rob $p$-values of their meaning. The next chapter looks at another layer of selection: which results get published. (https://5dlibrary.com/answers/en/scientific-method-falsifiability/c03/) 第3章 p 值操纵、研究者自由度与分岔花园 — 在真实效应为零的数据中,挑选因变量、追加样本、加入协变量与取舍条件四种做法,各自把假阳性率从 5% 抬高到约 8% 到 13%,合在一起达到约六成(原文 60.7%;本书重做,性别按效应编码为 57.9%,按 0/1 编码为 61.2%)。可选停止越频繁,假阳性越多。$p$ 曲线的形状可以提示一组文献是否被操纵。格尔曼与洛肯强调,即使没有故意钓鱼,依赖数据的分析选择也会让 $p$ 值失去意义。下一章讨论另一层筛选:什么样的结果会被发表。 (https://5dlibrary.com/answers/zh/scientific-method-falsifiability/c03/) - Ch.4 Publication Bias, the File Drawer and the Winner's Curse — Journals publish mainly significant results, and this filtering has barely changed since 1959. The fail-safe N answers only "does an effect exist?" and assumes the drawer averages zero effect; in this chapter's simulation it looked "extremely robust" while the published literature exaggerated the effect by 42 %. Egger's funnel-plot test has limited power, and its silence does not mean the absence of bias. In underpowered studies the effect sizes of significant results are systematically exaggerated: with a true $d = 0.20$ and 20 % power, significant results average about 0.45 (about 2.2 times too large), with a 0.5 % chance of the wrong sign. The next chapter looks at what large-scale replication actually found. (https://5dlibrary.com/answers/en/scientific-method-falsifiability/c04/) 第4章 发表偏倚、抽屉问题与赢家诅咒 — 期刊主要发表显著结果,这一筛选从 1959 年至今几乎没有变化。失效安全数只回答「效应是否存在」,而且假设抽屉里的研究平均效应为零;在本章的模拟中,它显示「极其稳健」,而发表文献把效应夸大了 42%。漏斗图的埃格检验把握度有限,没有报警不等于没有偏倚。在低功效研究中,显著结果的效应量被系统性夸大:真实 $d = 0.20$、功效 20% 时,显著结果平均约为 0.45(夸大约 2.2 倍),方向弄反的概率约 0.5%。下一章看大规模重复检验的实际结果。 (https://5dlibrary.com/answers/zh/scientific-method-falsifiability/c04/) - Ch.5 Large-Scale Replication: Results and Controversy — Three large replication projects found same-direction significance rates of 36.1 % (psychology, 35/97), 61 % (economics, 11/18) and 62 % (social-science experiments in Nature and Science, 13/21), with replication effect sizes about a half to two-thirds of the originals. A replication rate cannot be converted directly into a false discovery rate: a field with no false findings but small effects can also replicate poorly. Gilbert et al.'s criticisms about heterogeneity, single replications and protocol departures, and the Open Science Collaboration's reply, each have merit. When replication is judged by confidence intervals, the reasonable benchmark is far below 95 %. The next chapter discusses how to tackle these problems at their source. (https://5dlibrary.com/answers/en/scientific-method-falsifiability/c05/) 第5章 大规模重复检验:结果与争议 — 三个大型重复项目给出的同方向显著率分别为 36.1%(心理学,35/97)、61%(经济学,11/18)与 62%(《自然》《科学》社会科学实验,13/21),重复效应量约为原研究的一半到三分之二。重复率不能直接换算成假发现率:一个没有任何假发现、但研究的效应很小的领域,重复率也可能很低。吉尔伯特等人关于异质性、单次重复与方案偏离的批评,与开放科学协作组的回应,各有道理。用置信区间判断重复时,合理的基准远低于 95%。下一章讨论怎样从源头上解决这些问题。 (https://5dlibrary.com/answers/zh/scientific-method-falsifiability/c05/) - Ch.6 The Remedies of Meta-Science: Pre-Registration, Registered Reports and Equivalence Tests — Pre-registration writes down what a theory forbids before the data are seen, bringing the false-positive rate back to its nominal level; it does not forbid multiple outcomes or interim analyses, only requires that they be stated in advance and the criteria adjusted. Registered Reports make publication independent of the result; the positive rate for first hypotheses is about 44 % in Registered Reports and about 96 % in the standard literature. Equivalence tests let "no difference" be shown. This series' decision rule has five verdicts; it controls the Type I error, but "90 % power" refers to reaching some verdict of benefit, and the strongest verdict appears only 2.5 % of the time when the true effect exactly equals the SESOI. The next chapter returns to classical Chinese scholarship to see what "seek truth from facts" and "without evidence there is no trust" meant in their original texts. (https://5dlibrary.com/answers/en/scientific-method-falsifiability/c06/) 第6章 元科学的解药:预注册、注册报告与等效性检验 — 预注册在看到数据之前写下理论禁止什么,把假阳性率拉回名义水平;它不禁止多个结局或期中分析,只要求事先说清并调整标准。注册报告让发表与否不取决于结果;注册报告中第一个假说的阳性率约为 44%,标准文献约为 96%。等效性检验让「没有差别」也能被证明。本系列的判定规则有五种结论;它控制了第一类错误,但「90% 把握度」指的是得到某种获益结论,最强的结论在真实效应恰好等于最小关注效应时只有 2.5% 的机会出现。下一章回到中国古典学术,看「实事求是」「无征不信」在原文中说的是什么。 (https://5dlibrary.com/answers/zh/scientific-method-falsifiability/c06/) - Ch.8 Practice and a Pre-Registered Trial — The four-step checklist asks in turn whether a claim was falsifiable in advance, whether it went through multiple comparisons and optional stopping, whether the sample was so small that the effect must be exaggerated, and whether it has been independently replicated with interests disclosed. Applied to three fictional claims, it clearly separates the case report, the supplement advertisement and the pre-registered trial. Applied to the series itself, it shows that the series' protocols meet the requirements of pre-registration only on paper (they have not been registered), and that there are no data, the effect-size assumptions lack a prior basis, the text is AI-drafted, the author has a commercial interest, and there has been no independent replication. The pre-registered trial has $d'$ as its single primary outcome, 176 analysable and 208 recruited per arm, a decision rule written as code, and will be published whatever the result. (https://5dlibrary.com/answers/en/scientific-method-falsifiability/c08/) 第8章 实践与预注册试验 — 四步核查表依次检查一个说法是否事先可证伪、是否经过多重比较与可选停止、样本是否小到效应必然被夸大、是否经过独立重复并披露了利益关系。用于三个虚构的说法时,它能清楚地区分案例报告、保健品广告与预注册试验。用于本系列自身时,它显示本系列的方案只在写法上符合预注册的要求(尚未在登记平台登记),而且没有数据、效应量假设缺乏先验依据、由 AI 起草、作者有商业利益、没有独立重复。预注册试验以 $d'$ 为唯一主要结局,每组 176 名可分析、招募 208 名,判定规则写成代码,无论结果如何都发表。 (https://5dlibrary.com/answers/zh/scientific-method-falsifiability/c08/) ## 5DL-B-034 · Psychophysics, the Weber-Fechner Law and Optimal Bayesian Multisensory Integration 中文书名:心理物理学、韦伯-费希纳定律与多感官贝叶斯整合的数学模型 DOI: https://doi.org/10.5281/zenodo.23788548 Cite: Fan, Y. (2026). Psychophysics, the Weber-Fechner Law and Optimal Bayesian Multisensory Integration (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23788548 Book page: https://5dlibrary.com/books/psychophysics-multisensory-integration/ Scope: This book is not medical advice and cannot diagnose any sensory or neurological condition. Marked boxes separate published evidence, runnable Python calculations (17 programs, all checked), classical sources and the author's hypotheses. Abstract: In a quiet room you can hear a pin drop; in a noisy street you may miss a shout; and many people who see lips saying /ga/ while hearing /ba/ report /da/. How does physical intensity relate to sensation, and how does the brain combine several noisy senses into one judgement? This book answers with psychophysics and Bayesian models. It covers Weber's law and internal noise (a model with additive, Poisson-like and multiplicative noise reproduces Weber's law and its failure at low intensities; the framework's ΔI/(I + Ia) = k is an approximation, the corresponding form when noise variances add being ΔI = k√(I² + Ia²), at most about 40 % apart), the extra assumption that every just noticeable difference is subjectively equal needed to get from Weber to Fechner (Miller 1947: JNDs of about 0.41 dB do not correspond to equal loudness), and the separate histories of the decibel, the stellar magnitude and the Richter magnitude (no common psychophysical origin); Stevens's power law, the conditions behind its exponents (brightness 0.33 only for a 5° target in the dark, loudness 0.67 relative to sound pressure), the regression effect in magnitude estimation and Shepard's criticism, and cross-modality matching (slopes of nB/nA reproduced, but two logarithmic laws matched against each other also give a power function); maximum-likelihood integration (with the values of Ernst and Banks 2002, the weight on vision falls from 0.82 to 0.15 as noise rises), the ventriloquist effect (Alais and Burr 2004: an average gain of 1.425 against a predicted 1.414, but significant for only 2 of 6 observers) and the cases in children and some adult experiments where optimal integration fails; causal inference with the values fitted by Körding and colleagues in 2007 (the pull of a sound toward a flash falls from 20 % at 5° to 1 % at 40°) and individual and stimulus differences in the McGurk effect (17 %–58 % across stimuli); superadditivity in collicular neurons, the normalisation model and the cross-modal suppression it predicts, and "inverse effectiveness" manufactured by regression to the mean (with a true gain of 0.10 everywhere, post hoc grouping gives 0.149 against 0.052); and sensory substitution (the 1969 device stimulated the back, not the tongue; "acuity" with a device is largely bought with field of view and time) and the role of active movement in learning. The eight chapters are read beside Kangcangzi in the Liezi's "Zhongni" ("I can see and hear without using ears and eyes, but I cannot exchange their functions", a denial of swapping the senses), "the eye became like the ear, the ear like the nose, the nose like the mouth; there was no difference" in the Liezi's "Huangdi" (with a near-identical passage in "Zhongni"), the fasting of the mind in the Zhuangzi's "In the World of Men" (quoted from the received text, "hearing stops at the ear"), "thus the six faculties function interchangeably" in fascicle 4 of the Śūraṅgama Sūtra, and Qian Zhongshu's essay "Synaesthesia", stating that these texts describe experiences in cultivation, meditation and aesthetic appreciation, not psychophysical equations or experiments on cortical reorganisation. They end with a four-step multisensory precision checklist (the audiovisual synchrony window corrected from "about ±100 ms" to an asymmetric window of about 45 ms with sound early and 125 ms with sound late) and a pre-registered randomised controlled trial of vocal training that compares, under identical audiovisual conditions, "real-time pitch or formant error encoded as forearm vibration" with "the same device not vibrating" (single primary outcome the mean absolute pitch-imitation error in cents 24 hours later without feedback; an external pilot of 20 people, with the sample size re-estimated from the one-sided 80 % upper confidence limit of the standard deviation and the smallest effect of interest fixed at 20 cents; a plan of 191 analysable and 225 recruited per arm, 450 in all, bounded at 150 to 350 per arm; decision rule written as code; safety provisions for vocal load and vibration; "90 % power" means reaching some verdict of benefit, and the strongest verdict appears only about 2.5 % of the time when the true effect equals the smallest effect of interest). Marked boxes separate published evidence, runnable Python calculations (17 programs, all checked), classical sources and the author's hypotheses. This book is not medical advice and cannot diagnose any sensory or neurological condition. 摘要:安静的房间里针落地都听得见,喧闹的街头大喊也可能听不清;看嘴型 /ga/、听声音 /ba/,很多人会听成 /da/。物理强度与主观感觉之间是什么关系?大脑怎样把几种各有噪声的感觉合成一个判断?本书用心理物理学与贝叶斯模型回答这两个问题。依次讨论:韦伯定律与内部噪声(加性、泊松型与倍增性噪声的模型重现韦伯定律及其在低强度区的失效;框架中的 ΔI/(I + Ia) = k 是近似,噪声方差相加时对应的形式为 ΔI = k√(I² + Ia²),两者最多相差约 40%)、从韦伯到费希纳所需的「每个最小可觉差主观上等大」的额外假设(米勒 1947:约 0.41 分贝的最小可觉差不对应相等的响度),以及分贝、星等与里氏震级三种对数单位各自的来历(并无共同的心理物理学起源);史蒂文斯幂定律、各指数的测量条件(亮度 0.33 只适用于黑暗中 5° 的目标,响度 0.67 是相对声压)、量值估计的回归效应与谢泼德的批评,以及跨模态匹配(斜率 nB/nA 得到重现,但两条对数定律互相匹配同样给出幂函数);最大似然整合(按恩斯特与班克斯 2002 年的数值,视觉权重随噪声从 0.82 降到 0.15)、腹语术效应(阿莱斯与伯尔 2004:平均提高 1.425 倍对预测 1.414 倍,但 6 人中只有 2 人显著)与儿童及部分成人实验中最优整合不成立的情形;科丁等人 2007 年拟合值下的因果推断(声音被闪光拉过去的比例从 5° 时的 20% 降到 40° 时的 1%)与麦格克效应的个体与刺激差异(不同刺激 17%–58%);上丘神经元的超加性、归一化模型及其预测的交叉模态抑制,以及回归到均值造成的「逆向有效性」(真实增益处处为 0.10 时,事后分组得到 0.149 对 0.052);感官替代(1969 年的设备刺激背部而非舌面;设备上的「视力」主要用视野与时间换来)与主动运动在学习中的作用。 八章与《列子·仲尼》亢仓子(「我能视听不用耳目,不能易耳目之用」,否认互换耳目)、《列子·黄帝》「眼如耳,耳如鼻,鼻如口,无不同也」(《仲尼》篇另有几乎相同的一段)、《庄子·人间世》心斋(照通行本「听止于耳」)、《楞严经》卷四「由是六根互相为用」及钱锺书《通感》并读,并说明它们描述的是修道、禅定与审美中的体验,不是心理物理学方程或皮层重组实验;最后给出一份四步多感官精度核查表(视听同步窗口由「约 ±100 毫秒」修正为不对称的声音提前约 45 毫秒、延后约 125 毫秒),以及一项预注册随机对照试验方案:在相同的视听条件下,比较「把实时音高或共振峰误差编码为前臂振动」与「佩戴同款但不振动的设备」两种发音训练(唯一主要结局为 24 小时后无反馈时模仿音高的平均绝对误差,以音分计;先做 20 人的外部预试验,按标准差的单侧 80% 置信上限重估样本量,最小关注差异固定为 20 音分,计划每组 191 名可分析、招募 225 名,共 450 人,每组限定在 150 至 350 人之间;判定规则写成代码;含嗓音负荷与振动的安全条款;「90% 把握度」指得到某种获益结论,最强结论在真实效应等于最小关注效应时只有约 2.5% 的机会出现)。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序,全部核对)、原典与作者的假说。本书不是医疗建议,不能诊断任何感官或神经疾病。 Chapter answers: - Ch.1 A Logarithmic Bridge between the Physical and the Subjective: Weber's Law and Fechner's Law — Weber's law says that the JND is proportional to the intensity of the standard. A model of internal noise explains it, and also explains its failure at low intensities: floor noise stops the JND from falling near threshold, and Poisson-like noise creates a square-root region at intermediate intensities. The framework's $\Delta I/(I + I_a) = k$ is an approximation; when the noise variances add, the corresponding form is $\Delta I = k\sqrt{I^2 + I_a2}$, and the two differ by up to about 40 % near the knee. Fechner's law requires the extra assumption that every JND is subjectively the same size; Miller's observation and the loudness scale both show that this assumption fails. The logarithmic forms of stellar magnitude, the decibel and the Richter scale each have their own history; only the stellar magnitude has a forerunner in sensory grading, and it predates Fechner. The next chapter looks at how Stevens measured sensation another way and found power functions. (https://5dlibrary.com/answers/en/psychophysics-multisensory-integration/c01/) 第1章 物理世界与主观世界之间的对数桥梁:韦伯定律与费希纳定律 — 韦伯定律说最小可觉差与标准刺激的强度成正比;用内部噪声的模型可以解释它,也可以解释它在低强度区的失效:本底噪声使最小可觉差在阈限附近停止下降,泊松型噪声在中间强度产生平方根区。框架中的 $\Delta I/(I + I_a) = k$ 是近似,噪声方差相加时对应的形式是 $\Delta I = k\sqrt{I^2 + I_a2}$,两者在拐点附近最多相差约 40%。费希纳定律需要「每个最小可觉差在主观上一样大」的额外假设,米勒的观察与响度标度都表明它不成立。星等、分贝与里氏震级的对数形式各有来历,并无共同的心理物理学起源。 (https://5dlibrary.com/answers/zh/psychophysics-multisensory-integration/c01/) - Ch.2 Stevens's Power Law and Cross-Modality Matching — Stevens's magnitude estimation yields power functions $S = aI^{\,n}$ for many senses. The three exponents in the framework (brightness 0.33, length 1.0, shock 3.5) agree with the summary table, but each has specific conditions: the brightness exponent changes with target size and duration, and loudness 0.67 is relative to sound pressure. Magnitude estimation is affected by regression, range and context, and measures a blend of sensation and number use. The slope of a cross-modality match equals the ratio of two exponents, which the program reproduces exactly; but two logarithmic laws matched against each other also produce a power function, so matching cannot prove the power law superior to the logarithmic one. The next chapter turns from single senses to the combination of several. (https://5dlibrary.com/answers/en/psychophysics-multisensory-integration/c02/) 第2章 史蒂文斯幂定律与跨模态匹配 — 史蒂文斯的量值估计法在许多感觉上得到幂函数 $S = aI^{\,n}$。框架中的三个指数(亮度 0.33、长度 1.0、电击 3.5)都与汇总表一致,但各有特定的测量条件:亮度指数随目标大小与呈现时间改变,响度 0.67 是相对声压而言。量值估计受回归效应、范围与情境的影响,测到的是感觉与数字使用方式的混合。跨模态匹配的斜率等于两个指数之比,这一点程序完全重现;但两条对数定律互相匹配同样得到幂函数,所以它不能证明幂定律优于对数定律。下一章从单一感觉转向多种感觉的合成。 (https://5dlibrary.com/answers/zh/psychophysics-multisensory-integration/c02/) - Ch.3 Bayesian Optimal Multisensory Integration — Maximum-likelihood integration weights each sense by its precision; the combined estimate is more precise than any single sense, and the gain is largest, a standard deviation shrunk to $1/\sqrt 2$, when the senses are equally reliable. The visual–haptic experiment of Ernst and Banks and the audiovisual localisation experiment of Alais and Burr both came close to the model's predictions, and the latter showed that sound takes over the judgement of position when vision is blurred enough. But these experiments had very few observers (four and six), individual observers often fall short of the predicted gain, children do not integrate optimally before about 8 to 10 years, and adult experiments have also found over-reliance on vision and departures from maximum-likelihood weights. "Optimal" holds only within the model's assumptions, the most important of which is that the two signals share a source. (https://5dlibrary.com/answers/en/psychophysics-multisensory-integration/c03/) 第3章 多感官的贝叶斯最优整合 — 最大似然整合按精度分配权重,合成估计比任何单一感觉更精确,两种感觉同样可靠时收益最大,标准差缩小为 $1/\sqrt 2$。恩斯特与班克斯的视觉—触觉实验与阿莱斯与伯尔的视听定位实验都与这一模型的预测相近,后者还显示视觉足够模糊时声音会反过来主导位置判断。但这些实验的受试者很少(4 人与 6 人),单个观察者常常达不到预测的收益;儿童在 8 至 10 岁之前不按最优方式整合;成人实验中也有过度依赖视觉、不严格符合最大似然权重的结果。「最优」只在模型假设之内成立,其中最重要的一条是两个信号来自同一个来源。 (https://5dlibrary.com/answers/zh/psychophysics-multisensory-integration/c03/) - Ch.4 Causal Inference and the McGurk Effect — The causal-inference model makes "whether to combine" itself an object of inference: the farther apart two signals are, the lower the posterior probability of a common source and the weaker the integration. With the values fitted by Körding and colleagues, the pull of a sound toward a light falls gradually from 20 % at 5° to 1 % at 40°, which the forced-fusion model cannot reproduce. The McGurk effect can be understood in the same way, but fusion rates range from 17 % to 58 % across stimuli and from 0 to 100 % across people. Audiovisual speech is integrated within an asymmetric window, from about 30 ms with sound leading to about 170 ms with sound lagging. (https://5dlibrary.com/answers/en/psychophysics-multisensory-integration/c04/) 第4章 因果推断与麦格克效应 — 因果推断模型把「是否合在一起」也当作推断的对象:两个信号离得越远,同一来源的后验概率越低,整合也越弱。用科丁等人的拟合值,声音被光斑拉过去的比例从 5° 时的 20% 逐渐降到 40° 时的 1%,强制融合模型做不到这一点。麦格克效应可以用同样的思路理解,但它的强弱因人、因刺激差异极大:不同刺激的融合率从 17% 到 58%,个人从 0 到 100%,只用一段刺激的研究可能得出相差数倍的结论。视听言语在声音提前约 30 毫秒到延后约 170 毫秒的窗口内整合,窗口不对称。下一章从行为转向神经元。 (https://5dlibrary.com/answers/zh/psychophysics-multisensory-integration/c04/) - Ch.5 Multisensory Neurons in the Superior Colliculus: Superadditivity and Inverse Effectiveness — For a neuron that weights both inputs equally, the combined response at intensity 0.05 is 2.00 times the sum of the single responses (an enhancement of +299 %), 1.22 times at intensity 1, and only 0.60 times at intensity 20 (an enhancement of +21 %). The second part shows suppression: for a neuron driven mainly by input 1 ($d_2 = 0.25$), adding an equally strong input 2 when input 1 is strong (5) lowers the response to 0.50 of its value with input 1 alone, and a stronger input 2 (20) lowers it to 0.20, even though input 2 is excitatory for that neuron. When input 1 is weak (0.3), the same input 2 raises the response to 1.45 and 2.23 times. The same input enhances against a weak background and suppresses against a strong one. (https://5dlibrary.com/answers/en/psychophysics-multisensory-integration/c05/) 第5章 上丘多感官神经元:超加性与逆向有效性 — 对两种输入权重相等的神经元,强度为 0.05 时组合反应是两者之和的 2.00 倍(增强 +299%),强度为 1 时 1.22 倍,强度为 20 时只有 0.60 倍(增强 +21%)。第二部分显示了抑制:一个主要接受输入 1 的神经元($d_2 = 0.25$),在输入 1 较强(5)时加入同样强的输入 2,反应降到单独时的 0.50,输入 2 更强(20)时降到 0.20,尽管输入 2 对它本身是兴奋性的。而当输入 1 很弱(0.3)时,同样的输入 2 使反应增加到 1.45 倍与 2.23 倍。同一个输入,在弱背景下增强,在强背景下抑制。 (https://5dlibrary.com/answers/zh/psychophysics-multisensory-integration/c05/) - Ch.6 Sensory Substitution and Cross-Modal Plasticity — The tactile vision-substitution device of Bach-y-Rita in 1969 stimulated the skin of the back; electrotactile arrays on the tongue came later; the auditory device vOICe codes height as pitch and brightness as loudness. The visual cortex of blind people can be activated by Braille reading and contributes functionally to tactile tasks, and a shape-processing region of the lateral occipital cortex can be activated by sound-coded shapes. With a substitution device, congenitally blind people can score better than the blindness line on an acuity chart, but only by narrowing the field and lengthening the scan. Active control of the camera matters for learning the mapping between action and sensation. All these studies are small, and claims of "really seeing" must be qualified. The next chapter turns to ancient texts on the senses. (https://5dlibrary.com/answers/en/psychophysics-multisensory-integration/c06/) 第6章 感官替代与跨模态可塑性 — 巴赫-伊-丽塔 1969 年的触觉视觉替代设备刺激的是背部皮肤,舌面电刺激阵列是后来的发展;声音替代设备 vOICe 用音高表示高度、响度表示亮度。盲人的视觉皮层可以被盲文阅读激活,并对触觉任务有功能上的贡献;外侧枕叶的形状加工区可以被声音编码的形状激活。借助替代设备,先天全盲者的视力表成绩可以超过盲的界限,但这需要缩小视野、增加扫描时间。主动控制摄像机对建立动作与感觉的对应关系很重要。这些研究的样本都很小,「真正看见」的说法须加限定。下一章转向古代文献中关于感官的论述。 (https://5dlibrary.com/answers/zh/psychophysics-multisensory-integration/c06/) - Ch.8 Practice and a Pre-Registered Trial — The four-step checklist asks in turn whether a design depends on a single channel that may fail, whether weak signals have synchronous support from a second channel, whether the audiovisual timing difference lies within the asymmetric window (sound no more than about 45 ms early or 125 ms late), and how much of the time the learner explores actively. The pre-registered trial compares, under identical audiovisual conditions, "real-time pitch or formant error encoded as forearm vibration" with "the same device not vibrating", measured by pitch-imitation error without feedback 24 hours later. It has a single primary outcome in cents; an external pilot of 20 people followed by a sample-size rule written as code (standard deviation at its 80 % upper confidence limit, smallest effect of interest fixed at 20 cents, 150 to 350 per arm), with a plan of 191 analysable participants per arm (225 recruited); a decision rule written as code; safety provisions for vocal load and vibration; and it will be published whatever the result. (https://5dlibrary.com/answers/en/psychophysics-multisensory-integration/c08/) 第8章 实践与预注册试验 — 四步核查表依次检查:是否依赖一个可能失效的单一通道、微弱信号是否有第二通道的同步支持、视听时间差是否在不对称的窗口之内(声音提前不超过约 45 毫秒、延后不超过约 125 毫秒),以及学习者有多少时间在主动探索。预注册试验在相同的视听条件下,比较「把实时音高或共振峰误差编码为前臂振动」与「佩戴同款但不振动的设备」两种训练对 24 小时后无反馈音高模仿误差的影响:唯一的主要结局以音分计,先做 20 人的外部预试验,再按写成代码的规则(标准差取 80% 置信上限、最小关注差异固定为 20 音分、每组 150 至 350 人)确定样本量,计划值为每组 191 名可分析(招募 225 名),判定规则写成代码,并有嗓音负荷与振动的安全条款,无论结果如何都发表。 (https://5dlibrary.com/answers/zh/psychophysics-multisensory-integration/c08/) ## 5DL-B-035 · Evolutionary Game Theory, Direct and Indirect Reciprocity and the Emergence of Cooperation 中文书名:演化博弈、直接与间接互惠及合作涌现的数学模型 DOI: https://doi.org/10.5281/zenodo.23788551 Cite: Fan, Y. (2026). Evolutionary Game Theory, Direct and Indirect Reciprocity and the Emergence of Cooperation (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23788551 Book page: https://5dlibrary.com/books/evolutionary-cooperation-reciprocity/ Scope: The book discusses mathematical models of evolutionary games; it is not moral instruction and not business advice. Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. Abstract: Natural selection favours individuals that leave more offspring; how, then, can behaviour that costs the actor and benefits others persist? This book uses evolutionary game theory to examine when cooperation emerges, is maintained and breaks down. It covers the replicator equation dx_i/dt = x_i[f_i(x) − f̄(x)] and evolutionarily stable strategies, with cooperators inevitably disappearing in the one-shot anonymous prisoner's dilemma while the two strategies coexist in the hawk–dove game and the outcome of the stag hunt depends on the starting point, together with fixation probabilities in finite populations and the one-third law; Hamilton's rule rb > c as a consequence of the Price equation (a Simpson's paradox in which cooperation falls within every group yet rises overall), the separate model assumptions behind Nowak's five thresholds (1/r, 1/w, 1/q, k, 1 + n/m), and the dispute over inclusive fitness; Axelrod's two tournaments (14 and 62 programs), tit for tat falling into retaliation cycles under 1 % execution errors so that its long-run payoff drops to the average of the four outcomes, the upper limit min{1 − (T − R)/(R − S), (R − P)/(T − P)} on generous tit for tat's forgiveness (1 − c/b in the donation game), win-stay, lose-shift needing b/c > 2, and zero-determinant extortioners losing out in large populations; the flaw in image scoring, which penalises even refusals to help bad individuals, and Ohtsuki and Iwasa's leading eight norms (with the difference between simple standing and stern judging); about 31 % cooperators in Nowak and May's spatial game, the b/c > k rule under death–birth updating with the exact threshold (N − 2)/(N/k − 2) for finite networks, and two large human network experiments with different conclusions; and altruistic punishment, the antisocial punishment found in 16 cities in 15 countries, the efficiency cost of punishment and Ostrom's eight design principles. The eight chapters are read beside "love one's kin, and so be humane to the people; be humane to the people, and so cherish creatures" from the Mencius and Mencius's criticism of Mohist impartial care, "whoever loves others will surely be loved by others in return" from the Mozi, and "repay resentment with uprightness, repay kindness with kindness" from the Analects, stating the boundary between classical moral philosophy and the calculation of gene frequencies and never claiming that the ancients "had already discovered" evolutionary game theory. They end with a four-step checklist for cooperation and protection against extortion and a pre-registered cluster-randomised controlled experiment comparing the display of second-order reputation labels with the display of first-order contribution records only (labels under a single rule, simple standing with strict justification: withholding from someone in bad standing is good, from someone in good standing bad, and contributing to someone in bad standing costs nothing, so that in groups of three withholding is justified only when both partners are bad; online communities of 12 as the unit of randomisation, a single primary outcome of community cooperation in rounds 21–30, 48 analysable and 54 recruited communities per arm, 1,296 participants, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. The book discusses mathematical models of evolutionary games; it is not moral instruction and not business advice. 摘要:自然选择偏爱留下更多后代的个体,为他人付出代价的行为怎样才能存续?本书用演化博弈论考察合作在什么条件下出现、维持与瓦解。依次讨论:复制子方程 dx_i/dt = x_i[f_i(x) − f̄(x)] 与演化稳定策略,单次匿名的囚徒困境中合作者必然消失,而鹰鸽博弈中两种策略共存、猎鹿博弈中结局取决于起点,以及有限种群中的固定概率与三分之一法则;汉密尔顿法则 rb > c 作为普莱斯方程的推论(每个群体内部合作都在减少、整体却在增加的辛普森悖论),诺瓦克归纳的五个门槛(1/r、1/w、1/q、k、1 + n/m)各自的模型假设,以及包容适合度之争;阿克塞尔罗德的两次锦标赛(14 与 62 个程序),一报还一报在 1% 执行错误下陷入报复循环、长期收益降到四种结果的平均值,慷慨一报还一报的原谅概率上限 min{1 − (T − R)/(R − S), (R − P)/(T − P)}(捐献博弈中为 1 − c/b),赢定输移需要 b/c > 2,以及零行列式勒索策略在大种群中被淘汰;形象分「拒绝坏人也扣分」的漏洞与大槻—岩佐的前八规范(简单持守与严格评判的区别);诺瓦克—梅空间博弈中约 31% 的合作者,「死亡—出生」更新下的 b/c > k 规则与有限网络的精确门槛 (N − 2)/(N/k − 2),以及两项结论不同的大型人类网络实验;利他惩罚、在 15 国 16 个城市中发现的反社会惩罚、惩罚的效率代价与奥斯特罗姆的八项设计原则。 八章与《孟子·尽心上》「亲亲而仁民,仁民而爱物」及孟子对墨家兼爱的批评、《墨子·兼爱中》「夫爱人者,人必从而爱之」、《论语·宪问》「以直报怨,以德报德」并读,明确古典道德哲学与基因频率计算的界限,不说古人「早已发现」演化博弈;最后给出一份四步合作与防敲诈核查表,以及一项比较「公布二阶声誉标签」与「只公布一阶历史捐献率」的预注册整群随机对照实验方案(标签只用一个口径:简单持守规则、严格正当化,对坏声誉者不捐献为好、对好声誉者不捐献为坏、向坏人捐献不扣分,3 人小组中同组另外两人都为坏时不捐献才算正当;以 12 人线上社群为随机化单位,唯一主要结局为第 21—30 轮的社群合作维持率,每组 48 个社群可分析、招募 54 个,共 1,296 人,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部核对)、原典与作者的假说。本书讨论演化博弈的数学模型,不作道德说教,也不是经营建议。 Chapter answers: - Ch.1 Replicator Dynamics: Why Cooperation Loses in One-Shot Anonymous Games — The last part of the program checks this law. With a stag payoff of 4, $x^ = 0.75$ and $\rho N = 0.65$; with a stag payoff of 12, $x = 0.25$ and $\rho N = 1.15$. The dividing line lies close to $x^ = 1/3$: at $x = 0.333$, $\rho N = 0.96$, and at $x^* = 0.30$ it is 1.02. That $\rho N$ falls slightly below 1 exactly at 1/3 when $N = 100$ is a finite-population correction, which shrinks as the population grows. Chapter 2 shows that the one-third law is one of the three criteria Nowak used to compare the five mechanisms. (https://5dlibrary.com/answers/en/evolutionary-cooperation-reciprocity/c01/) 第1章 复制子动力学:单次匿名博弈中合作为什么输 — 程序用莫兰过程(每一步按适合度挑一个个体繁殖,后代随机替换一个个体)计算固定概率,并以 $\rho N$ 表示相对于中性的倍数。在单次匿名的捐献博弈中,合作者的 $\rho N$ 总是小于 1,背叛者的总是大于 1。种群越大、选择越强,差距越大:$N = 10$、选择强度 $s = 0.01$ 时,合作者的固定概率只比中性低 6%(0.94);$N = 200$、$s = 0.1$ 时,它实际上为零,而背叛者是中性的 17 倍。对 $N = 20$、$s = 0.1$ 的模拟(40,000 次)得到 0.257,与公式的 0.253 相符。小种群中合作者偶尔能靠运气固定下来,但选择的方向始终不利于它。 (https://5dlibrary.com/answers/zh/evolutionary-cooperation-reciprocity/c01/) - Ch.2 Hamilton's Rule, the Price Equation and the Five Mechanisms — The network threshold $k$ is the limit for large populations. For a finite network in which everyone has $k$ neighbours the exact threshold is $(N - 2)/(N/k - 2)$ (Taylor 2007; Chen 2013): with $k = 4$ it is 6.00 for $N = 20$, 4.26 for $N = 100$ and 4.02 for $N = 1000$. In small networks the threshold lies well above $k$. The group-selection threshold $1 + n/m$ approaches 1 when groups are many and small: 1.2 with 50 groups of 10, and 3 with 5 groups of 10. (https://5dlibrary.com/answers/en/evolutionary-cooperation-reciprocity/c02/) 第2章 汉密尔顿法则、普莱斯方程与五种机制 — 网络互惠的门槛 $k$ 是大种群中的极限。对每人都有 $k$ 个邻居的有限网络,精确门槛是 $(N - 2)/(N/k - 2)$ (Taylor 2007; Chen 2013):$k = 4$ 时,$N = 20$ 需要 6.00,$N = 100$ 需要 4.26,$N = 1000$ 需要 4.02。小网络中门槛远高于 $k$。群体选择的门槛 $1 + n/m$ 在群体多、规模小时接近 1:10 人的群体有 50 个时为 1.2,有 5 个时为 3。 (https://5dlibrary.com/answers/zh/evolutionary-cooperation-reciprocity/c02/) - Ch.3 Direct Reciprocity: Tournaments, Noise and Forgiveness — Tit for tat won Axelrod's two tournaments, but it never beats an opponent in a single match, and its ranking depends on the field. With execution errors two tit-for-tat players fall into a retaliation cycle and their long-run payoff drops to the average of the four outcomes. Generous tit for tat's forgiveness has an upper limit $\min\{1 - (T - R)/(R - S), (R - P)/(T - P)\}$, which is $1 - c/b$ in the donation game; win-stay, lose-shift is robust when $b/c > 2$ and prevails in evolution. Zero-determinant extortioners never lose a match but are eliminated in evolution in large populations. Direct reciprocity needs the same pair to meet again and again; the next chapter turns to the case of helping someone one will never meet again. (https://5dlibrary.com/answers/en/evolutionary-cooperation-reciprocity/c03/) 第3章 直接互惠:锦标赛、噪声与宽容 — 一报还一报在阿克塞尔罗德的两次锦标赛中胜出,但它从不在单局中击败对手,名次依赖参赛构成。有执行错误时,两个一报还一报陷入报复循环,长期收益降到四种结果的平均值。慷慨一报还一报的原谅概率有上限 $\min\{1 - (T - R)/(R - S), (R - P)/(T - P)\}$,捐献博弈中为 $1 - c/b$;赢定输移在 $b/c > 2$ 时稳健,在演化中占优。零行列式的勒索策略在每局中都不吃亏,但在大种群的演化中被淘汰。直接互惠需要同一对个体反复相遇;下一章讨论另一种情形:帮助一个以后不会再见的人。 (https://5dlibrary.com/answers/zh/evolutionary-cooperation-reciprocity/c03/) - Ch.4 Indirect Reciprocity: Image Scores and Second-Order Reputation Norms — The first set of results shows the flaw in image scoring: with 90 % discriminators and 10 % defectors, under image scoring only 8 % of discriminators are rated good, discriminators help each other only 8 % of the time, and their average payoff per round is 0.14, almost as low as the defectors'. Refusing defectors makes discriminators bad, refusing bad discriminators makes more discriminators bad, and cooperation among discriminators collapses. Shunning gives almost the same result. Under simple standing and stern judging 97 % of discriminators are good, they help each other 95 % of the time, and they earn 1.71 and 1.70, far above the defectors' 0.28 and 0.30. With 30 % defectors the difference is starker. (https://5dlibrary.com/answers/en/evolutionary-cooperation-reciprocity/c04/) 第4章 间接互惠:形象分与二阶声誉规范 — 第一组结果显示了形象分的漏洞:种群中有 90% 的鉴别者与 10% 的背叛者时,在形象分规则下鉴别者中只有 8% 被评为好,鉴别者之间的帮助率只有 8%,每轮平均收益 0.14,几乎与背叛者一样低。拒绝背叛者使鉴别者变坏,而拒绝变坏的鉴别者又使更多鉴别者变坏,合作在鉴别者之间瓦解了。排斥规则的结果几乎一样。在简单持守与严格评判下,鉴别者 97% 为好,彼此之间的帮助率为 95%,收益 1.71 与 1.70,远高于背叛者的 0.28 与 0.30。背叛者 30% 时,差别更明显。 (https://5dlibrary.com/answers/zh/evolutionary-cooperation-reciprocity/c04/) - Ch.5 Network Reciprocity: Spatial Structure and Human Network Experiments — When individuals interact only with neighbours, cooperators can survive by forming clusters: in the Nowak–May model cooperators settle at about 31 % for $1.8 < T < 2$, whereas they all die out under random mixing. The rule of Ohtsuki and colleagues, $b/c > k$, gives the condition under which a network favours cooperation with death–birth updating and weak selection; the exact threshold for finite networks, $(N - 2)/(N/k - 2)$, is higher, and under birth–death updating a network never favours cooperation. In human experiments a large static network did not raise cooperation, while cooperation remained stable under the condition $b/c > k$; whether networks help cooperation depends on payoff structure and on how people actually respond. The next chapter discusses a mechanism that changes payoffs directly: punishment. (https://5dlibrary.com/answers/en/evolutionary-cooperation-reciprocity/c05/) 第5章 网络互惠:空间结构与人类网络实验 — 当个体只与邻居互动时,合作者可以结成团块而存活:诺瓦克—梅模型中,合作者在 $1.8 < T < 2$ 时稳定在约 31%,而随机相遇时全部灭绝。大槻等人的规则 $b/c > k$ 给出了「死亡—出生」更新、弱选择下网络偏爱合作的条件;有限网络的精确门槛 $(N - 2)/(N/k - 2)$ 更高,「出生—死亡」更新下网络从不偏爱合作。人类实验中,大型静态网络没有提高合作,而在 $b/c > k$ 的条件下合作保持稳定;网络能否帮助合作,取决于收益结构与人们实际的反应方式。下一章讨论一种直接改变收益的机制:惩罚。 (https://5dlibrary.com/answers/zh/evolutionary-cooperation-reciprocity/c05/) - Ch.6 Punishment: Altruistic Punishment, Antisocial Punishment and Governing the Commons — The last part of the program lists each person's payoff in a group of 5 to show the efficiency cost of punishment. When all cooperate each gets 2.0, the same as when all are punishers, because nobody needs to be punished. With 4 punishers and 1 defector the group's total payoff is −8.0, lower not only than the 10.0 of full cooperation but also than the 0 of full defection: each punisher gets only 0.4 and the defector is driven down to −9.6. When 3 punishers and 2 antisocial punishers punish each other, the group total is −42.0. Punishment when used always destroys income; its value lies before use: if its threat is enough to make people cooperate, it rarely needs to be used. This is the logic of the 50-period experiment of Gächter and colleagues. (https://5dlibrary.com/answers/en/evolutionary-cooperation-reciprocity/c06/) 第6章 惩罚:利他惩罚、反社会惩罚与公共池塘的治理 — 程序的最后一部分列出 5 人组中每人的收益,说明惩罚的效率代价。全组合作时每人得 2.0,与全是惩罚者时相同,因为没有人需要被惩罚。4 个惩罚者与 1 个背叛者时,全组总收益是 −8.0,不仅低于全组合作的 10.0,也低于全组背叛的 0:惩罚者每人只得 0.4,背叛者被扣到 −9.6。3 个惩罚者与 2 个反社会惩罚者互相惩罚时,全组总收益为 −42.0。惩罚在被使用时总是毁掉收入;它的价值在于使用之前:如果它的威胁足以让人合作,它就很少需要被使用。格希特等人的 50 期实验正是这个道理。 (https://5dlibrary.com/answers/zh/evolutionary-cooperation-reciprocity/c06/) - Ch.8 Practice and a Pre-Registered Trial — With an SD of community means of 0.15 (slightly more conservative than the illustrative model above), a smallest difference of interest of 0.10 (control 0.40, arm A 0.50), two-sided $\alpha = 0.05$ and 90 % power, 48 analysable communities are needed per arm (normal approximation; a t-distribution correction adds about one); allowing for about 10 % of communities being incomplete because someone drops out, 54 communities are recruited per arm, that is 648 people per arm and 1,296 in all. With an SD of 0.18, 69 communities per arm would be needed; with a smallest difference of interest of 0.08, 74. All these parameters are assumptions and should be checked against pilot data before the trial begins. (https://5dlibrary.com/answers/en/evolutionary-cooperation-reciprocity/c08/) 第8章 实践与预注册试验 — 按社群平均值的标准差 0.15(比上面的示意模型略保守)、最小关注差异 0.10(对照组 0.40,A 组 0.50)、双侧 $\alpha = 0.05$、把握度 90% 计算,每组需要 48 个可分析的社群(正态近似;若按 t 分布校正,约多需 1 个);考虑约 10% 的社群因有人中途退出而不完整,每组招募 54 个社群,即每组 648 人,共 1,296 人。若标准差为 0.18,每组需要 69 个社群;若最小关注差异为 0.08,需要 74 个。这些参数都是假设值,正式实施前应当用预试验的数据核对。 (https://5dlibrary.com/answers/zh/evolutionary-cooperation-reciprocity/c08/) ## 5DL-B-036 · Scaling Laws, Allometric Growth and the Statistical Physics of Organisational Metabolism 中文书名:规模法则、异速生长与组织新陈代谢的统计物理 DOI: https://doi.org/10.5281/zenodo.23788553 Cite: Fan, Y. (2026). Scaling Laws, Allometric Growth and the Statistical Physics of Organisational Metabolism (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23788553 Book page: https://5dlibrary.com/books/scaling-laws-organizational-metabolism/ Scope: The book discusses mathematical models of biological allometry and of the size of cities and firms; it is not investment advice and promises no results for the management of any business. Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. Abstract: A mouse burns far more energy per gram than an elephant, yet large cities produce more per person than small ones; companies often vanish, cities seldom do. This book uses biological allometry, urban scaling, survival analysis and organisation theory to examine how size changes the metabolism of a system, and how far the conclusions carry over to firms and teams. It covers Kleiber's law and the dispute between 2/3 and 3/4 (in synthetic data, one slightly curved relation fitted over different ranges of mass gives 0.652, 0.707 or 0.770); the derivation of the fractal network of West, Brown and Enquist and its finite-size corrections (area-preserving networks approach 3/4 from below, and with Poiseuille flow in the smallest vessels the exponent over an illustrative range is about 0.81), and heartbeats per mammalian lifetime, which Levine's compilation puts at about 0.73 billion on average and about 1 billion pooled, so that "1 to 1.5 billion" is too high and hides a tenfold spread between species, with humans, at about 2.9 billion, an exception; the universal curve of the ontogenetic growth equation (an identity for that equation, which cannot distinguish 3/4 from 2/3 at 10 % error) and "maintenance swallowing supply", with 84 % of metabolism spent on maintenance at half the asymptotic size; the sublinear scaling of urban infrastructure and superlinear scaling of socioeconomic output (doubling the population multiplies totals by 1.80 and 2.22, about −10 % and +11 % per person rather than the often-quoted 15 %), the effect of city boundaries and noise models on the exponents, and the finite-time singularity of the superlinear growth equation (each doubling about 10 % faster than the last; the singularity is a consequence of the model, not an observation); the roughly constant exit hazard of listed companies and their half-life of about a decade (most exits are acquisitions rather than failures, and heterogeneity can create the illusion that older means safer), with a schematic model of administrative overhead and survival; Ringelmann's table of declining output per person (often described as rope-pulling data; a fixed loss per communication channel, with an optimum of about seven people), the distinction between motivation and coordination losses drawn by Ingham and colleagues, and Brooks's law as a conditional rule of thumb; Simon's model of executive pay, Williamson's control loss and optimal hierarchy (where the first-order condition matches the level-by-level comparison), and the extrapolation behind Dunbar's 147.8 and the dispute over the interval of 4 to 520 found by Lindenfors and colleagues. The eight chapters are read beside "governing the many is like governing the few: it is a matter of divisions and numbers; fighting with the many is like fighting with the few: it is a matter of forms and names" from the "Momentum" chapter of the Sunzi, "the lesser ruler exhausts his own abilities, the middling ruler the strength of others, the greatest ruler the wisdom of others" from the Han Feizi's "Eight Canons", and, from the "Examining Divisions" section of the Lüshi Chunqiu, "common work is slow, for there is somewhere to hide one's strength" in "Examining Divisions" and "when the ruler likes to act himself, those who hold office abandon their offices and cater to what the ruler does" in "Guarding the Ruler's Position" (which the framework had assigned to "Examining Divisions" with a misquoted ending, corrected here from the received text), stating where classical military and political experience ends and modern scaling laws begin. They end with a four-step checklist for scale and organisational metabolism and a pre-registered group-randomised controlled experiment comparing "three squads of four plus standardised interface documents" (21 communication channels) with "twelve people in a flat structure communicating freely" (66 channels) on a modular programming task that is complete when a hidden integration test suite written in advance passes, with both arms limited to logged, structured text chat; the 12-person group is the unit of randomisation, the single primary outcome is completion time, secondary outcomes include rework, numbers of messages and bytes, information entropy and response latency, the primary plan at 90 % power has 75 analysable and 84 recruited groups per arm (2,016 participants), a pre-registered fallback at 80 % power has 1,512, and the decision rule is written as code. Marked boxes separate published evidence, runnable Python calculations (15 programs, all checked), classical sources and the author's hypotheses. The book discusses mathematical models of biological allometry and of the size of cities and firms; it is not investment advice and promises no results for the management of any business. 摘要:小鼠每克体重燃烧的能量比大象多得多,大城市的人均产出却比小城市高;公司常常消失,城市很少消亡。本书用生物异速生长、城市标度律、生存分析与组织理论考察「规模」怎样改变一个系统的新陈代谢,以及这些结论在多大程度上可以用于企业与团队。依次讨论:克莱伯定律与 2/3、3/4 之争(合成数据中,同一条略微弯曲的曲线按体重范围拟合可得 0.652、0.707 或 0.770),韦斯特—布朗—恩奎斯特分形管网的推导及其有限大小修正(面积保持型管网从下方趋近 3/4,末端为泊肃叶流动时示意范围内约 0.81),以及一生心跳数:莱文汇总的哺乳动物平均约 7.3 亿次、综合估计约 10 亿次,「10—15 亿次」偏高且掩盖了物种间十倍的离散,人约 29 亿次是例外;个体发育生长方程的普适曲线(对该方程是数学恒等式,10% 误差下无法区分 3/4 与 2/3)与「维护吞噬供给」:上限一半处维护已占 84% 的代谢;城市基础设施的亚线性与社会经济产出的超线性(人口翻倍时总量为 1.80 倍与 2.22 倍,人均约 −10% 与 +11%,不是常说的 15%),城市边界与涨落模型对指数的影响,以及超线性增长方程的有限时间奇点(翻倍时间每次缩短约 10%,奇点是模型推论而非观测);上市公司近似恒定的退出风险与约十年的半衰期(退出大多是被收购而非倒闭,人群异质性会制造「越老越安全」的假象)及行政管理费与存活的示意模型;林格曼的人均出力递减表(常被说成拉绳数据;每条沟通通道固定损失,最优约 7 人)、英厄姆等人区分的动机损失与协调损失、作为有条件经验法则的布鲁克斯定律;西蒙的高管薪酬模型、威廉姆森的控制损失与最优层级(一阶条件与逐级比较一致),以及邓巴数 147.8 的外推及林登福斯等人 4 到 520 的区间之争。 八章与《孙子兵法·势篇》「凡治众如治寡,分数是也;斗众如斗寡,形名是也」、《韩非子·八经》「下君尽己之能,中君尽人之力,上君尽人之智」与《吕氏春秋·审分览》的《审分》「公作则迟,有所匿其力也」和《君守》「人主好以己为,则守职者舍职而阿主之为矣」(框架误作《审分》「之己」,已据通行本改正)并读,并明确古代军事与统治经验与现代标度律的界限;最后给出一份四步规模与新陈代谢核查表,以及一项比较「3 个 4 人小队 + 标准化接口文档」(21 条沟通通道)与「12 人扁平全员自由沟通」(66 条)的预注册小组随机对照实验方案:任务为按模块分工编程,以通过一套预先写好的隐藏集成测试为「完成」,两组都只用有记录的结构化文字聊天;以 12 人小组为随机化单位,唯一主要结局为完成时间,次要结局包括返工、消息条数、字节数、信息熵与响应延迟;主方案把握度 90%,每组 75 个小组可分析、招募 84 个,共 2,016 人,80% 把握度(共 1,512 人)为预先登记的备选,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部核对)、原典与作者的假说。本书讨论生物异速生长与城市、企业规模的数学模型,不是投资建议,也不构成任何企业管理成效的承诺。 Chapter answers: - Ch.1 Kleiber's Law and Fractal Networks: The Three-Quarter Power and Its Disputes — Kleiber's law is a line that holds approximately on logarithmic axes, but its slope depends on the species, the range of mass, the physiological state and the fitting method, and fitting a single straight line may hide curvature. The WBE fractal-network model shows that a space-filling branching network with invariant terminal units produces sublinear scaling; 3/4 holds exactly only for an infinitely large area-preserving network, and finite size and the Poiseuille regime move the exponent away from it. The "invariant" number of heartbeats per lifetime is only a rough statistical trend: Levine's compiled average is about 0.7 billion and the order of magnitude about 1 billion, with species differing by up to tenfold, not a fixed quota of 1 to 1.5 billion. The next chapter asks: if supply grows sublinearly and maintenance linearly, where does growth stop? (https://5dlibrary.com/answers/en/scaling-laws-organizational-metabolism/c01/) 第1章 克莱伯定律与分形管网:四分之三次幂及其争议 — 克莱伯定律是一条在对数坐标上近似成立的直线,但它的斜率取决于物种、体重范围、生理状态与拟合方法;把数据整体画成一条直线,可能掩盖曲率。WBE 分形管网模型说明了末端单元不变、充满空间的分支网络会产生亚线性标度,3/4 只在无限大的面积保持型管网中严格成立,有限大小与泊肃叶区段会使指数偏离。一生心跳总数的「不变量」只是一个粗略的统计趋势:莱文汇总的平均约 7 亿次,数量级约为 10 亿次,物种之间相差可达十倍,不是一个 10—15 亿次的固定配额。下一章问:如果供给按亚线性增长而维护按线性增长,生长会在哪里停下来? (https://5dlibrary.com/answers/zh/scaling-laws-organizational-metabolism/c01/) - Ch.2 Ontogenetic Growth: Why Growth Stops — The WBE model of ontogenetic growth writes growth as sublinear supply minus linear maintenance, with an asymptotic mass of $(a/b)^4$; the universal curve after non-dimensionalisation is an identity for this equation, and existing data can hardly distinguish 3/4 from 2/3. The share of supply taken by maintenance rises with mass and reaches 84 % at half the ceiling. Applied to organisations, the structure implies that the closer supply is to linear, the higher and more efficiency-sensitive the ceiling, but the more of the supply maintenance absorbs along the way. The next chapter turns to a system that does not stop growing: the city. (https://5dlibrary.com/answers/en/scaling-laws-organizational-metabolism/c02/) 第2章 个体发育生长:为什么生长会停下来 — WBE 个体发育模型把生长写成「亚线性供给减去线性维护」,渐近体重为 $(a/b)^4$;无量纲化后的普适曲线对这个方程是恒等式,现有数据难以区分 3/4 与 2/3。维护所占的比例随体重上升,到上限一半时已占 84%。用于组织时,这一结构说明:供给越接近线性,上限越高、对效率越敏感,但维护在中途就吸收了大部分供给。下一章转向一种不会停止生长的系统:城市。 (https://5dlibrary.com/answers/zh/scaling-laws-organizational-metabolism/c02/) - Ch.3 Urban Scaling: Superlinearity, Finite-Time Singularities and Accelerating Innovation — Urban infrastructure grows sublinearly with population and socioeconomic output superlinearly; doubling the population multiplies the totals by 1.80 and 2.22, or about −10 % and +11 % per person. The exponents are sensitive to the definition of city boundaries and to the fluctuation model, and the synthetic model shows that commuting can "turn" a linear relation into a superlinear one of 1.09. The superlinear growth equation diverges in finite time and each doubling takes about 10 % less time than the last, so that "resetting by innovation" sustains growth only if innovation keeps accelerating; the singularity is a mathematical consequence, not an observation. The next chapter turns from cities back to firms: how long do companies live? (https://5dlibrary.com/answers/en/scaling-laws-organizational-metabolism/c03/) 第3章 城市标度律:超线性、有限时间奇点与加速的创新 — 城市的基础设施随人口亚线性增长,社会经济产出超线性增长;人口翻倍时总量为 1.80 倍与 2.22 倍,人均约 −10% 与 +11%。指数的大小对城市边界的定义与涨落模型很敏感,合成模型显示通勤可以把线性关系「变成」1.09 的超线性。超线性增长方程在有限时间内发散,翻倍时间每次缩短约 10%,所以「靠创新重置」只有在创新不断加速时才能延续增长;奇点是数学推论,不是观测。下一章从城市回到企业:企业的寿命有多长? (https://5dlibrary.com/answers/zh/scaling-laws-organizational-metabolism/c03/) - Ch.4 The Mortality of Companies: Constant Hazard, Half-Lives and Administrative Overhead — With $\delta = 0$, administrative costs are proportional to headcount and size brings only benefits: a 10-person company survives 20 years with probability 82.3 %, companies of 1,000 or more almost surely, and small companies die mainly of volatility. With $\delta = 0.15$, administrative costs eat the margin of a 100,000-person company down to about 0, and survival falls from 98.9 % at 10,000 people to 83.9 %; with $\delta = 0.20$, a 10,000-person company is already down to 70.7 %, and a 100,000-person company has a margin of −0.12 and always exits. Survival thus first rises and then falls with size: small companies die of volatility, large ones of coordination costs. Where the turning point lies depends entirely on $\delta$, and this book has no reliable estimate of $\delta$ for real companies. (https://5dlibrary.com/answers/en/scaling-laws-organizational-metabolism/c04/) 第4章 企业的死亡率:恒定风险、半衰期与行政管理费 — $\delta = 0$ 时,行政成本与人数成正比,规模只带来好处:10 人公司 20 年存活 82.3%,1,000 人以上接近 100%,小公司主要死于波动。$\delta = 0.15$ 时,10 万人的公司利润率被行政成本吃到接近 0,存活率从 1 万人时的 98.9% 降到 83.9%;$\delta = 0.20$ 时,1 万人的公司已降到 70.7%,10 万人的公司利润率为 −0.12,全部退出。于是存活率随规模先升后降:小公司死于波动,大公司死于协调成本。转折点的位置完全取决于 $\delta$,而真实企业的 $\delta$ 是多少,本书没有可靠的估计。 (https://5dlibrary.com/answers/zh/scaling-laws-organizational-metabolism/c04/) - Ch.5 The Ringelmann Effect and Brooks's Law: More People, Less Each? — Ringelmann's table shows output per person falling linearly with group size, equivalent to a fixed loss per communication channel, with an optimal size of about 7; but Ingham and colleagues found that the loss levels off after three people and that much of it is motivational rather than a matter of coordination. Brooks's law is a conditional rule of thumb: near the end of a project, with slow ramp-up and high communication costs, adding people delays completion; adding them early usually helps. The next chapter widens the view from a single team to a whole hierarchical organisation. (https://5dlibrary.com/answers/en/scaling-laws-organizational-metabolism/c05/) 第5章 林格曼效应与布鲁克斯定律:人越多,每人越少? — 林格曼的表显示人均出力随人数线性下降,等价于每条沟通通道固定损失,最优人数约为 7;但英厄姆等人发现损失在三人后趋于稳定,而且相当一部分来自动机而非协调。布鲁克斯定律是有条件的经验法则:项目末期、上手慢、沟通成本高时,加人会推迟完工;早期加人通常有利。下一章把视野从一个团队扩大到整个层级组织。 (https://5dlibrary.com/answers/zh/scaling-laws-organizational-metabolism/c05/) - Ch.6 Span of Control, Managerial Hierarchy and Dunbar's Number — Span and hierarchy reduce communication channels from order $N^2$ to order $N$, at the price of distortion from level to level and managerial pay. Simon's model explains why executive pay rises as a power of size; Williamson's control loss gives a finite optimal number of levels, and the optimal size is extremely sensitive to the fidelity of information. Nearly decomposable modular structures are the classic way to reduce coordination costs. Dunbar's 147.8 is an extrapolation with a wide interval; the reanalysis by Lindenfors and colleagues gives an interval of 4 to 520, and Dunbar replies that they used the wrong regression method; the observation that layers differ by a factor of about 3 is a separate matter from the specific number 150. The next chapter reads this modern research beside classical Chinese texts on "governing the many". (https://5dlibrary.com/answers/en/scaling-laws-organizational-metabolism/c06/) 第6章 控制跨度、管理层级与邓巴数 — 跨度与层级把沟通通道从 $N^2$ 量级降到 $N$ 量级,代价是逐级失真与管理薪酬。西蒙的模型解释了高管薪酬随规模的幂律增长,威廉姆森的控制损失给出有限的最优层级,而最优规模对信息保真度极为敏感。近可分解的模块化结构是降低协调成本的经典办法。邓巴的 147.8 是一个区间很宽的外推,林登福斯等人的重新分析给出 4 到 520 的区间,邓巴则认为对方用错了回归方法;分层比例约为 3 的观察与 150 这个具体数字是两回事。下一章把这些现代研究与中国古代关于「治众」的文献并读。 (https://5dlibrary.com/answers/zh/scaling-laws-organizational-metabolism/c06/) - Ch.8 Practice and a Pre-Registered Trial — Step 1: the exponent of net output is 0.91 and that of coordination costs 1.27; coordination costs rise from 8.9 % of net output at 20 people to 13.3 % at 85, and net output per person falls from 6.71 to 6.44. Headcount has more than quadrupled while each person's net output has fallen, a signal of the "maintenance swallowing supply" of Chapter 2 and of the $\delta > 0$ case of Chapter 4. With only 12 points spanning less than an order of magnitude, the confidence intervals of the two exponents are wide; they are warnings, not diagnoses. (https://5dlibrary.com/answers/en/scaling-laws-organizational-metabolism/c08/) 第8章 实践与预注册试验 — 所以样本量按保守得多的假设计算:组间对数完成时间的标准差取 0.35,最小关注效应为几何均数之比 0.85(分队组快 15%),基线协变量与结局的相关取 0.5(方差乘以 0.75),双侧 $\alpha = 0.05$。主方案的把握度取 90%:每组需要 75 个可分析的小组;考虑约 10% 的小组因有人缺席或中途退出而作废,每组招募 84 个小组,即每组 1,008 人,共 2,016 人。若标准差是 0.25 或 0.45,所需小组数分别为 39 与 122;若不用协变量,标准差 0.35 时需要 99 个。预先登记的次要备选方案只把把握度改为 80%,其余假设不变:每组 56 个可分析小组,招募 63 个,每组 756 人,共 1,512 人(程序最后两行)。这些参数都是假设值。正式试验之前先做一个小规模预试验,只用来检查任务难度、隐藏测试与聊天系统是否正常运行(例如 90 分钟内的完成比例),不用来改变样本量。 (https://5dlibrary.com/answers/zh/scaling-laws-organizational-metabolism/c08/) ## 5DL-B-037 · Two-Scale Renormalization, Occupancy Entropy and Thrust-Wave Superposition in Discrete Cyclical Networks 中文书名:离散周期网络中的双尺度重整化、占位数分布熵与推力波叠加定理 DOI: https://doi.org/10.5281/zenodo.23796761 Cite: Fan, Y. (2026). Two-Scale Renormalization, Occupancy Entropy and Thrust-Wave Superposition in Discrete Cyclical Networks (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23796761 Book page: https://5dlibrary.com/books/discrete-cyclical-networks-occupancy-entropy/ Scope: The book presents mathematical models for teaching and research; it is not investment, trading, tax or legal advice. Marked boxes separate published evidence, runnable Python calculations (17 programs, all checked), classical sources and the author's hypotheses. Abstract: Put fourteen particles into twelve cells arranged in a ring: the more evenly they are spread, the higher the entropy; the more they are crowded, the higher the peak, and the empty cells can borrow half the excitation of the cell opposite. The author's model is drawn from the reckoning rules of the traditional Chinese stem-and-branch calendar: the twelve Earthly Branches, the six branch oppositions (the antipodes i+6 of the ring) and the "five tigers" rule that gives the stem of the first month from the stem of the year. The book studies only the mathematical structure of these rules and neither asserts nor tests any prediction in the sense of fate calculation or divination. It covers the sixty-term cycle as the subgroup of same-parity pairs in ℤ₁₀ × ℤ₁₂, isomorphic to ℤ₆₀, with inverse k ≡ 6s − 5b (mod 60); the five-tigers rule, first-month stem ≡ 2a + 2 (mod 10), a two-to-one affine map rather than a permutation whose image holds only the five yang stems (the "five tigers"), and the year–month homomorphism Y ↦ 12Y, under which block decimation changes the period from 60 to 5, so that "self-similar fractal" fails and "renormalisation" is only an analogy; the entropy–peak bound theorem for occupancy entropy (H ≤ log₂(K − V), a decreasing upper bound for a given peak, peak ≥ N·2^(−H), and an identity linking variance and the Rényi entropy of order 2), with "low entropy, high peak" a theorem only for patterns comparable by majorisation and, under random allocation, an entropy–peak correlation of −0.708, a statistical tendency rather than a conservation law; the author's four-state excitation map and antipodal half-wave coupling written as axiomatic definitions, under which the total flux decomposes into six antipodal pairs, lies between 14 and 21, equals 14 exactly with no vacancy, and makes no use of the ring's neighbour structure; the Weber–Fechner law, adaptation level and habituation as an analogy and hypothesis for the "felt resistance of the zero-thrust state", whose model results come mainly from the concavity of the logarithm and an arbitrary floor; sector rotation, organisational vacancies and periodic staking as untested structural-isomorphism hypotheses, and twelve-tone equal temperament, whose correspondence with ℤ₁₂ and the dihedral group is strict mathematics but not the product of a resonant cavity; and the Barnum effect, Carlson's 1985 double-blind test (astrologers about 34 % correct, no different from the chance one-third), multiple testing and data snooping. Claims corrected on checking include: the five-tigers rule is not a permutation; the two-scale relation is not fractal self-similarity; the "inverse theorem" becomes bounds and a statistical tendency; "coherent superposition, +2.0" corresponds physically to incoherent addition of intensities; equal temperament is not a resonance phenomenon; "days have six jia, branches have five zi" in the Book of Han treatise is a numerological gloss on "Heaven six, Earth five", not modular arithmetic; the earliest reliable source of the five-tigers rhyme could not be found, the Ming Sanming tonghui, quoting the Sanche, already uses its rule, and the book treats it as a later divinatory mnemonic; Shen Kuo's twelve-qi calendar is paraphrased and cited by juan and entry (Supplement, juan 2, "Images and numbers"). The eight chapters are read beside the Book of Han treatise, chapter 11 of the Laozi, the Appended Statements of the Book of Changes, Shen Kuo's twelve-qi calendar and the Sanming tonghui, drawing the line between calendrical arithmetic and divinatory inference. They end with a four-step checklist and a pre-registered blind permutation backtest (year and month boundaries at the exact moments of the start of spring and the twelve jie terms, with excess returns cumulated over solar-term months from daily data; single primary outcome the within-year contrast of log absolute returns in opposition months, 96 years and 1,152 solar-term months, a one-sided stratified permutation test with about 90 % power for a 1.5-fold effect, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (17 programs, all checked), classical sources and the author's hypotheses. The book presents mathematical models for teaching and research; it is not investment, trading, tax or legal advice. 摘要:把十四颗粒子放进排成一圈的十二个格子:放得越匀,熵越高;放得越挤,峰值越高,空出来的格子还可以从对面借来一半的激发。作者的这一模型取材于中国传统干支历法的推算规则——十二地支、地支六冲(环上的对径点 i+6)与由年干推正月月干的「五虎遁」。本书只研究这些规则的数学结构,不主张、也不检验任何命理或占卜意义上的预测。依次讨论:六十甲子是 ℤ₁₀ × ℤ₁₂ 中同奇偶组合构成的子群、与 ℤ₆₀ 同构,逆映射为 k ≡ 6s − 5b (mod 60);五虎遁的正月月干 ≡ 2a + 2 (mod 10) 是二对一的仿射映射而不是置换,像集只有五个阳干(即「五虎」),年—月关系是同态 Y ↦ 12Y,块抽取使周期从 60 变为 5,所以「自相似分形」不成立,「重整化」只作类比;占位熵的熵—峰值界定理(H ≤ log₂(K − V)、给定峰值的递减上界、峰值 ≥ N·2^(−H)、方差与雷尼 2 阶熵的恒等式),「熵低峰高」只在优超可比时是定理,随机分配下熵与峰值的相关系数为 −0.708,是统计倾向而非守恒律;作者的四态激发映射与对径半波耦合写成公理式定义后,总通量可分解为六个对径对之和,在 14 到 21 之间,零空位时恰为 14,并且完全不使用环的相邻结构;用韦伯—费希纳定律、适应水平与习惯化解释「零推力态的主观阻力」只是类比与假说,模型结果主要来自对数的凹性与人为的下限;板块轮动、组织空编与周期质押写成尚待检验的结构同构假说,十二平均律与 ℤ₁₂、二面体群的对应是严格的数学,但不是共振腔的产物;巴纳姆效应、卡尔森 1985 年的双盲检验(占星师命中约 34%,与随机的三分之一无异)、多重检验与数据窥探。 核对后修正的要点包括:五虎遁不是置换;双尺度关系不是分形自相似;「反比定理」改为界与统计倾向;「相干叠加 +2.0」在物理上对应非相干的强度相加;十二平均律不是共振现象;《汉书·律历志上》「日有六甲,辰有五子」是「天六地五」的数术说明而非模运算;五虎遁口诀的最早可靠出处未能查到,明代《三命通会》引《三车》已用其规则,本书写为后世术数口诀;沈括「十二气历」按《补笔谈》卷二「象数」门的卷次与条目转述。全书八章与《汉书·律历志》、《老子》第十一章、《周易·系辞上》、沈括十二气历与《三命通会》并读,并划清历法算术与术数推断的界限;最后给出四步核查表与一项预注册盲法置换回测方案(年界与月界取立春与十二个「节」的精确交节时刻,用日度数据按节气月累计超额收益;唯一主要结局为冲月对数绝对收益的年内对比,96 年、1,152 个节气月,单侧分层置换检验,对 1.5 倍效应约有 90% 把握度,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序,全部核对)、原典与作者的假说。本书是教学与研究用的数学模型,不是投资、交易、税务或法律建议。 Chapter answers: - Ch.1 The Mod-12 Ring Lattice and the Five-Tigers Map: Two-to-One, the Chinese Remainder Theorem and Block Maps — The sixty-term cycle is the subgroup of same-parity pairs in $\mathbb{Z}_{10} \times \mathbb{Z}_{12}$, isomorphic to $\mathbb{Z}_{60}$, with inverse $k \equiv 6s - 5b \pmod{60}$. The five-tigers rule gives the first-month stem $\equiv 2a + 2 \pmod{10}$, a two-to-one affine map rather than a permutation: its image contains only the five yang stems, matching the five yin months, and the pairs of year stems sharing a first month are exactly the cosets of the kernel of $x \mapsto 2x$. Years and months are related by the homomorphism $Y \mapsto 12Y$, so month labels repeat every 5 years; days and double-hours have exactly the same structure. Block decimation changes the period from 60 to 5 and then cycles among four step sizes, so "self-similar fractal" does not hold and "renormalisation" can only be an analogy. The next chapter treats the twelve cells as containers and asks which relations between entropy and peak can be proved once particles are put in. (https://5dlibrary.com/answers/en/discrete-cyclical-networks-occupancy-entropy/c01/) 第1章 模 12 环形晶格与五虎遁映射:二对一、中国剩余定理与块映射 — 六十甲子是 $\mathbb{Z}_{10} \times \mathbb{Z}_{12}$ 中同奇偶组合构成的子群,与 $\mathbb{Z}_{60}$ 同构,逆映射为 $k \equiv 6s - 5b \pmod{60}$。五虎遁的正月月干 $\equiv 2a + 2 \pmod{10}$,是二对一的仿射映射而不是置换:像集只有五个阳干,对应五个寅月,同一正月的年干对恰好是同态 $x \mapsto 2x$ 的核的陪集。年与月之间的关系是同态 $Y \mapsto 12Y$,月的干支每 5 年重复;日与时之间有完全相同的结构。块抽取把周期从 60 变为 5,并在四种步长之间循环,所以「自相似分形」不成立,「重整化」只能作类比。下一章把十二个格子当作容器,看粒子放进去以后,熵与峰值之间有哪些能证明的关系。 (https://5dlibrary.com/answers/zh/discrete-cyclical-networks-occupancy-entropy/c01/) - Ch.2 Shannon Entropy and Peak of an Occupancy Distribution: The Entropy–Peak Bound Theorem — What can be proved are bounds: $H \le \log_2(K - V)$; for a given peak $H \le h(m/N)$, with $h$ decreasing for $m \ge N/K$; $m \ge N \cdot 2^{-H}$; and an identity linking variance and collision entropy. With 14 particles in 12 cells, the integer pattern of largest entropy is unique, "ten 1s and two 2s" (3.5216 bits). "The lower the entropy, the higher the peak" is a theorem only between patterns comparable by majorisation; of 8,778 pairs, 1,405 are not comparable, and in 65.9 % of those the direction is reversed. Under random allocation the mean number of vacancies is 3.55, no vacancy occurs with probability only 0.0013, and the correlation between entropy and peak is −0.708: a statistical tendency, not a conservation law. The next chapter adds the author's excitation rules on top of occupation and asks what "an empty cell borrows half from the one opposite" brings. (https://5dlibrary.com/answers/en/discrete-cyclical-networks-occupancy-entropy/c02/) 第2章 占位分布的香农熵与峰值:熵—峰值界定理 — 能严格证明的是界:$H \le \log_2(K - V)$;给定峰值时 $H \le h(m/N)$,且 $h$ 在 $m \ge N/K$ 时递减;$m \ge N \cdot 2^{-H}$;方差与碰撞熵之间有恒等式。在 14 个粒子、12 个格子中,熵最大的整数形态唯一,是「十个 1、两个 2」(3.5216 比特)。「熵越低、峰值越高」只在优超可比的形态之间是定理;8,778 对形态中有 1,405 对不可比,其中 65.9% 方向相反。在随机分配下,平均空位 3.55 个,零空位的概率只有 0.0013,熵与峰值的相关系数为 −0.708,这是统计倾向,不是守恒律。下一章在占位之上加入作者规定的激发规则,看「空位从对面借一半」会带来什么。 (https://5dlibrary.com/answers/zh/discrete-cyclical-networks-occupancy-entropy/c02/) - Ch.3 The Four-State Excitation Map, Antipodal Half-Wave Coupling and the Decomposition of Thrust — Once the author's rules are written as definitions, the total flux decomposes strictly into the contributions of six antipodal pairs (Theorem 3.1). Under linear excitation $\Phi = N + c S$, and the total flux lies between 14 and 21: exactly 14 with no vacancy, and 21 only with at least 6 vacancies; its mean under random allocation is 16.26. The +2.0 the author calls "coherent superposition" corresponds physically to incoherent addition of intensities; with a saturating or squared extension, the size and direction of the correlation between entropy and flux change. The excitation map is equivariant under $D_{12}$, but the total flux is in fact invariant under a much larger group and makes no use of the ring's neighbour structure. The next chapter leaves combinatorics for a psychological analogy: if a cell in the "zero-thrust" state could feel, what would it feel? (https://5dlibrary.com/answers/en/discrete-cyclical-networks-occupancy-entropy/c03/) 第3章 四态激发映射、对径半波耦合与推力分解 — 把作者的规则写成定义以后,总通量可以严格分解为六个对径对的贡献之和(定理 3.1)。在线性激发下,$\Phi = N + c S$,总通量在 14 到 21 之间:零空位时恰为 14,至少 6 个空位才能达到 21;随机分配下的平均值为 16.26。作者称为「相干叠加」的 +2.0 在物理上对应非相干的强度相加;如果改用饱和或平方延拓,熵与通量相关的大小与方向都会改变。激发映射在 $D_{12}$ 下等变,但总通量实际上在一个大得多的群下不变,完全不使用环的相邻结构。下一章离开组合数学,问一个心理学的类比问题:一个处于「零推力」状态的格子,如果它能「感觉」,会感觉到什么? (https://5dlibrary.com/answers/zh/discrete-cyclical-networks-occupancy-entropy/c03/) - Ch.4 The Zero-Thrust State and Felt Resistance: The Weber–Fechner Law, Adaptation Level and Habituation (Analogy and Hypothesis) — The Weber–Fechner law by itself says only that sensation is related to the logarithm of the stimulus and gives no "negative sensation" below threshold, so it does not imply that the zero-thrust state brings felt resistance. Helson's adaptation level and the reference point of prospect theory supply the missing mechanism: feeling depends on the gap between a stimulus and the recent level. Research on habituation shows that responses to repeated stimuli weaken and recover when the stimulus stops. Written as an illustrative model, Wave-Rider gives higher mean excitation than Fixed-Mode (1.59 against 1.17) yet a lower mean feeling (−0.595 against −0.023), with about 44 % of steps at "strong resistance"; but the result comes mainly from the concavity of the logarithm and an arbitrary floor, and raising the floor from 0.1 to 1.0 moves the mean feeling from −0.937 to −0.215. The next chapter discusses the author's four cross-domain applications and asks what is missing before a formal resemblance becomes a testable hypothesis. (https://5dlibrary.com/answers/en/discrete-cyclical-networks-occupancy-entropy/c04/) 第4章 零推力态与主观阻力:韦伯—费希纳定律、适应水平与习惯化(类比与假说) — 韦伯—费希纳定律本身只说明感觉与刺激的对数有关,在阈限以下给不出「负的感觉」,所以推不出零推力态会带来阻力感;赫尔森的适应水平与前景理论的参照点提供了所需的机制:感受取决于刺激相对于近期水平的差距。习惯化研究说明,对重复刺激的反应会减弱,停止后会恢复。把这些写成一个示意模型以后,乘波模式的平均激发比固定模式高(1.59 对 1.17),平均感受却更低(−0.595 对 −0.023),约 44% 的时段为「强阻力」;但这一结果主要来自对数的凹性与人为的下限,下限从 0.1 提高到 1.0 时,平均感受从 −0.937 变为 −0.215。下一章讨论作者提出的四个跨领域应用,并问:要把形式上的相似变成可检验的假说,还缺什么? (https://5dlibrary.com/answers/zh/discrete-cyclical-networks-occupancy-entropy/c04/) - Ch.5 Cross-Domain "Structural Isomorphism" Hypotheses: Sector Rotation, Flexible Vacancies, Twelve-Tone Equal Temperament and Periodic Staking — The theorems of Chapters 2 and 3 hold for every allocation over 12 categories, so "finding" them in data is no evidence; a null model with no mechanism at all moves the correlation between entropy and next-period change from −0.025 to +0.332. Sector rotation and organisational vacancies can be tested only with categories, pairings, a definition of the conserved quantity, a single outcome and a null model all fixed in advance; the known findings on industry momentum and organisational slack are not this book's model. The correspondence of twelve-tone equal temperament with $\mathbb{Z}_{12}$ and $D_{12}$ is a strict mathematical fact: the fifth is 1.96 cents narrow, the major third 13.69 cents wide, and the tritone is the antipode, $\sqrt{2}$; but it is not the product of a resonant cavity. Periodic staking corresponds strictly only when the protocol itself is designed that way, and it carries the risk of losing all capital. The next chapter discusses why people so often see regularities where there are none. (https://5dlibrary.com/answers/en/discrete-cyclical-networks-occupancy-entropy/c05/) 第5章 跨领域「结构同构」假说:板块轮动、机动空编、十二平均律与周期质押 — 第二、三章的定理对任何 12 类分配都成立,所以在数据中「发现」它们不是证据;一个没有任何机制的零模型就能让熵与下一期变动的相关系数从 −0.025 变到 +0.332。金融板块轮动与组织空编两项,需要事先固定的类别、配对、守恒量定义、唯一结局与零模型,才谈得上检验;已知的行业动量与组织冗余研究与本书的模型并不相同。十二平均律与 $\mathbb{Z}_{12}$、$D_{12}$ 的对应是严格的数学事实:五度低 1.96 音分,大三度高 13.69 音分,三全音是对径点 $\sqrt{2}$;但它不是共振腔的产物。周期质押只有在协议本身这样设计时才严格对应,并伴随本金全部损失的风险。下一章讨论,为什么人们常常在没有规律的地方看到规律。 (https://5dlibrary.com/answers/zh/discrete-cyclical-networks-occupancy-entropy/c05/) - Ch.6 Cleaning Out Counterexamples and Spurious Correlations: The Barnum Effect, Multiple Testing and Data Snooping — The Barnum effect shows that "feeling accurate" is not evidence: Forer's students gave a description identical for all of them a mean of about 4.3 out of 5. In Carlson's 1985 double-blind test the astrologers' first choices were right about 34 % of the time, no different from the chance one-third; Ertel's reanalysis is disputed, and Dean and Kelly's review and the time-twins study are likewise negative; for stem-and-branch fate calculation the author found no comparable test. In pure noise without any effect, 176 tests yield on average 8.8 "significant" patterns and every history has a "discovery"; the most significant pattern of the first half has a mean $t$ of 0.00 in the second. Data snooping must be corrected against a null distribution over "all models considered", and the null distribution must also respect time structure, or else for contiguous labels the false-positive rate can double. The next chapter returns to the documentary sources of these rules. (https://5dlibrary.com/answers/en/discrete-cyclical-networks-occupancy-entropy/c06/) 第6章 反例与假相关的清洗:巴纳姆效应、多重检验与数据窥探 — 巴纳姆效应说明「觉得准」不是证据:福勒的学生给一份人人相同的笼统描述打出约 4.3 分(满分 5 分)。卡尔森 1985 年的双盲检验中,占星师的第一选择命中约 34%,与随机的三分之一无异;埃特尔的重新分析有争议,迪恩与凯利的综述与时间双胞胎研究同样是阴性的;对干支命理,作者没有找到同类检验。在不含任何效应的纯噪声中,176 次检验平均产生 8.8 个「显著」模式,每一段历史都有「发现」;前半段最显著的模式在后半段平均 $t = 0.00$。数据窥探要用「全部被考虑过的模型」作零分布来校正;零分布还必须尊重时间结构,否则对连续的标签,假阳性率可以翻倍。下一章回到这些规则的文献出处。 (https://5dlibrary.com/answers/zh/discrete-cyclical-networks-occupancy-entropy/c06/) - Ch.8 Practice and a Pre-Registered Test: A Four-Step Checklist and a Blind Permutation Backtest — The simulations (with the same daily model as the power program) show that if the true $R = 1$ (no effect), 92.0 % of 600 simulated backtests conclude "an effect of 1.5 times or more is excluded", 3.0 % wrongly reach a "larger in opposition months" conclusion, 3.3 % are "unclear" and 1.7 % conclude "smaller in opposition months". If the true $R = 1.5$, 85.8 % conclude "larger in opposition months; 1.5 times not shown" and 2.3 % reach the strongest conclusion, 88.1 % together, consistent with the design power of 90 % within the simulation error of about ±1.3 percentage points for 600 runs and the use of only 199 permutations; another 9.3 % are "unclear" and 2.5 % wrongly exclude an effect of 1.5 times. (https://5dlibrary.com/answers/en/discrete-cyclical-networks-occupancy-entropy/c08/) 第8章 实践与预注册检验:四步核查表与盲法置换回测 — 程序按登记的做法生成数据:先用带波动聚集的日度 GARCH(1,1) 模型生成日超额收益(正态冲击与厚尾的 $t(5)$ 冲击两种,参数为示意值,每个节气月 21 个交易日),再按节气月累计;效应设为把冲月里每一天的收益乘以 $R$,不改变波动过程本身。年内对比的标准差约为 1.20 与 1.21,于是 $D$ 的标准误约为 0.122 与 0.123。在单侧 $\alpha = 0.025$ 下,对 $R = 1.5$ 的把握度两种模型都是 91%;对 $R = 1.25$ 只有 45% 与 44%。要达到 90% 的把握度,真实效应至少要为 $R \approx 1.49$;若要以 90% 的把握度检测 $R = 1.25$,需要 303 到 308 年的数据,这在任何现存的股市中都不存在。用登记的分层置换检验直接模拟,无效应时拒绝零假设的比例为 2.3%(名义水平 2.5%),$R = 1.5$ 时为 89.8%;模拟中每次只做 199 次置换,略低于正式分析 9,999 次置换时的把握度,所以「约 90%」是偏保守的说法。 (https://5dlibrary.com/answers/zh/discrete-cyclical-networks-occupancy-entropy/c08/) ## 5DL-B-038 · Implied Volatility Surfaces, Skew Premia and Quantitative Risk Control of Vertical Spreads 中文书名:期权隐含波动率曲面、偏度溢价与垂直价差结构的定量风控 DOI: https://doi.org/10.5281/zenodo.23796765 Cite: Fan, Y. (2026). Implied Volatility Surfaces, Skew Premia and Quantitative Risk Control of Vertical Spreads (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23796765 Book page: https://5dlibrary.com/books/implied-volatility-skew-vertical-spreads/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; every price and trade in the book is fictional or illustrative. Abstract: A trader opening a bear put vertical spread knows at once the most he can lose and the most he can make; but can the rule of thumb "if it hasn't worked in seven to ten trading days, close it" be proved? This book presents mathematical models for teaching and research and is not investment advice. It covers Black–Scholes–Merton pricing and the numerical inversion of implied volatility (the Manaster–Koehler starting value makes Newton's method converge in all 226 test cases, against 33 failures from a fixed start; the Brenner–Subrahmanyam approximation holds only at the money forward, where it is the first Newton step from zero); the volatility surface, the 25-delta risk reversal (only −1.3 vol points under crash-type jumps, while strike 85 sits about 12 points above at-the-money), three fat-tailed models (Student's t cannot be used directly for pricing), and the evidence, counter-evidence and time variation of the variance and skew premia; the capped gain and loss of the spread, the cost of bid–ask spreads (a round trip of up to 26 % of the maximum gain) and early assignment of American options; and the non-linear behaviour of the Greeks as expiry approaches. The author casts the time stop as an optimal stopping problem (Snell envelope, backward induction on a binomial tree): counting trading days, with the base parameters the best fixed exit day is trading day 9 or 10, but trading days 7 to 14 are almost as good, a rule depending on price is clearly better (+0.240 against +0.136), and across twelve scenarios the best day ranges from trading day 1 to holding to expiry; even if the model is right, telling a trading-day-9 exit from holding to expiry would take about 190,000 independent trades. On the Kelly criterion and the author's "three-tier ratchet": with capped losses and fixed-fraction betting, capital never reaches zero geometrically; the ratchet is applied only to incremental profit above the high-water mark, so withdrawals come only from new highs and never eat into existing capital, though losses still cause drawdowns; as a warning, skimming every winning trade would make trading capital decay geometrically whenever 0.4pb ≤ q; clustered losses, gaps and liquidity, all outside the model, raise the chance of a drawdown of more than half from 36.6 % to 89.6 %. Chapter 7 reads these beside Ji Ran's "not above eighty, not below thirty" and Bai Gui's "what others discard I take, what others take I give" in the "Money-Makers" chapter of the Records of the Grand Historian, its "Treatise on Balanced Standard", balanced purchase and the ever-normal granary in the Book of Han's "Treatise on Food and Money", and the art of "light and heavy" in the Guanzi's "State Reserves", using a buffer-stock model to show that these were mechanisms of price stabilisation and reserve management that resemble options only in structure, not "prototype grain options". Chapter 8 offers a four-step checklist and a pre-registered out-of-sample test of "a time stop at trading day 9 versus holding to expiry" (single primary outcome the mean paired difference in return per unit of maximum loss, smallest effect of interest 0.15, 50 underlyings, 112 entry dates, about 8.6 years of data, run by a third party, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; every price and trade in the book is fictional or illustrative. 摘要:一个熊市看跌垂直价差最多亏多少、最多赚多少,交易者在开仓时就知道;可「持仓 7—10 个交易日没有起色就平仓」这条经验规则,能被证明吗?本书是教学与研究用的数学模型,不是投资建议。它依次讨论:布莱克—斯科尔斯—默顿定价与隐含波动率的数值反解(马纳斯特与克勒的初值使牛顿法在 226 个算例上全部收敛,固定初值有 33 个失败;布伦纳与苏布拉马尼亚姆的近似只适用于平值远期期权,它恰是从零出发的第一步牛顿迭代);波动率曲面、25-Delta 风险逆转(崩盘型跳跃下只有 −1.3 个波动率点,而行权价 85 比平值高约 12 点)、厚尾的三种模型(学生 t 不能直接用于定价)与方差风险溢价和偏度溢价的证据、反证及时变性;价差的封顶收益与亏损、买卖价差的成本(往返最多占最大收益的 26%)与美式期权的提前指派;希腊值随剩余期限的非线性变化。作者把时间止损写成最优停时问题(斯内尔包络、二叉树倒推):以交易日计时,在基准参数下最优的固定平仓日为第 9 到 10 个交易日,但第 7 到 14 个交易日几乎一样好,依价格状态而定的规则明显更好(+0.240 对 +0.136),在 12 种情形中最优日从第 1 个交易日到持有到期不等;即使模型正确,分辨第 9 个交易日平仓与持有到期也需要约 19 万笔独立交易。凯利准则与作者的「三阶棘轮」分仓:在亏损封顶、按固定比例下注时资本几何上不归零;棘轮只对高水位以上的增量利润执行,提取只来自新高部分、不侵蚀已有本金,但亏损照样造成回撤;作为警示,若按笔提取盈利,只要 0.4pb ≤ q 交易资本就几何衰减;亏损扎堆、跳空与流动性等模型外风险使回撤一半以上的概率从 36.6% 升到 89.6%。 第七章与《史记·货殖列传》计然「上不过八十,下不减三十」与白圭「人弃我取,人取我与」、《史记·平准书》、《汉书·食货志》的平籴与常平仓及《管子·国蓄》的轻重之术并读,并用缓冲储备模型说明:这些是价格稳定与储备调节机制,与期权只有结构上的相似,不是「谷物期权雏形」。第八章给出四步核查表与一项「第 9 个交易日时间止损对持有到期」的预注册样本外检验方案(唯一主要结局为以最大亏损为单位的配对平均收益差,最小关注效应 0.15,50 个标的、112 个入场日、约 8.6 年数据,第三方执行,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有价格与交易都是虚构或示意的。 Chapter answers: - Ch.1 Black–Scholes–Merton Pricing and the Inversion of Implied Volatility — The Black–Scholes–Merton formula prices European options under constant volatility, lognormal returns and continuous frictionless trading, and can be implemented in the standard library with math.erf; the binomial tree converges to it as the number of steps grows. Implied volatility is a unique inverse with no closed form: Newton's method from the Manaster–Koehler start converges in every test case, about 15 % of cases fail from a fixed start, and bisection provides a fallback. The Brenner–Subrahmanyam approximation holds only at the money forward, where it is the first Newton step from zero. The next chapter looks at how markets price different strikes with different implied volatilities. (https://5dlibrary.com/answers/en/implied-volatility-skew-vertical-spreads/c01/) 第1章 布莱克—斯科尔斯—默顿定价与隐含波动率的反解 — 布莱克—斯科尔斯—默顿公式在常数波动率、对数正态、连续无摩擦交易的假设下给出欧式期权价格,标准库用 math.erf 即可实现;二叉树随步数增加收敛到公式。隐含波动率是唯一存在但没有解析式的反函数:牛顿法从马纳斯特—克勒初值出发在全部算例上收敛,从固定初值出发约 15% 失败,二分法可以兜底。布伦纳—苏布拉马尼亚姆近似只对平值远期期权成立,它就是从零出发的第一步牛顿迭代。下一章看市场怎样用不同的隐含波动率给不同行权价定价。 (https://5dlibrary.com/answers/zh/implied-volatility-skew-vertical-spreads/c01/) - Ch.2 The Volatility Surface, Skew and the Variance Risk Premium — Implied volatility varies with strike and expiry, forming a surface that for equity indexes usually slopes down to the left. Crash-type jumps can produce such a skew, but the 25-delta risk reversal reflects only the slope near the money. Fat tails are pronounced at the daily horizon, while monthly tails come mainly from volatility clustering; Student's t cannot be used directly for pricing. There is substantial evidence for variance and skew premia in index options, weaker evidence for single stocks; the premia vary over time and are disputed, and the seller's average gain comes with extreme negative skewness. (https://5dlibrary.com/answers/en/implied-volatility-skew-vertical-spreads/c02/) 第2章 波动率曲面、偏度与方差风险溢价 — 隐含波动率随行权价与期限变化,形成曲面;股票指数的曲面通常向左倾斜。崩盘型跳跃可以产生这种偏度,但 25-Delta 风险逆转只反映平值附近的斜率。厚尾在日频上明显,月度尾部主要来自波动率聚集;学生 t 分布不能直接用来定价。指数期权的方差风险溢价与偏度溢价有大量证据,但个股较弱,溢价随时期而变,而且有争议;卖方的平均收益伴随着极端的负偏。 (https://5dlibrary.com/answers/zh/implied-volatility-skew-vertical-spreads/c02/) - Ch.3 The Bear Put Vertical Spread: Geometry, Costs and Early Exercise — With a 3 % interest rate and 30 days left, the early-exercise premium of the 100-strike American put is 0.246 when the underlying is at 80 (where the American price is just the intrinsic value of 20) and only 0.015 at the money; immediate exercise is optimal below 88.6, and with seven days left the critical price rises to 93.4. At 8 % interest both the premium and the critical price are higher (0.052 at the money; critical prices 91.1 and 94.6), because receiving the strike early earns more interest. (https://5dlibrary.com/answers/en/implied-volatility-skew-vertical-spreads/c03/) 第3章 熊市看跌垂直价差:几何、成本与提前行权 — 利率为 3%、剩余 30 天时,行权价 100 的美式看跌期权在标的为 80 时的提前行权溢价为 0.246(美式价格就是内在价值 20),在平值时只有 0.015;立即行权在标的低于 88.6 时最优,剩余 7 天时临界价格升到 93.4。利率为 8% 时,溢价与临界价格都更高(平值时 0.052,临界价格 91.1 与 94.6),因为提前拿到行权价的利息更值钱。 (https://5dlibrary.com/answers/zh/implied-volatility-skew-vertical-spreads/c03/) - Ch.4 The Non-Linear Behaviour of the Greeks as Expiry Approaches — The Greeks are tied together by the Black–Scholes equation, and in the model time decay is the price of gamma. At-the-money gamma and decay accelerate roughly as $1/\sqrt{T}$, while in- and out-of-the-money Greeks rise and then fall. The net Greeks of a bear put spread change sign with the position of the underlying: near the long strike net decay is negative and grows quickly, near the short strike it is positive. Whether a time stop is reasonable therefore depends on the state, not only on the count of days. (https://5dlibrary.com/answers/en/implied-volatility-skew-vertical-spreads/c04/) 第4章 希腊值随剩余期限的非线性变化 — 希腊值由布莱克—斯科尔斯方程联系在一起,在模型中时间损耗是伽马的价格。平值期权的伽马与时间损耗随到期临近按 $1/\sqrt{T}$ 加速,实值与虚值期权的则先升后降。熊市看跌价差的净希腊值随标的位置改变符号:标的在多头行权价附近时净损耗为负并迅速放大,在空头行权价附近时为正。时间止损的合理性因此取决于状态,而不只是天数。 (https://5dlibrary.com/answers/zh/implied-volatility-skew-vertical-spreads/c04/) - Ch.5 The Time Stop as an Optimal Stopping Problem — "Seven to ten days is optimal" cannot be proved in general. Following the author's confirmation, the book counts these days as trading days. Written as an optimal stopping problem, in the author's model the best fixed exit day depends on how fast the signal decays, the equity premium, the variance premium, the skew and costs, and ranges from trading day 1 to holding to expiry across twelve scenarios; in the base case it is trading day 9 or 10, but days 7 to 14 are almost equally good and holding to expiry is only 0.011 worse, while a rule that depends on the price is clearly better. Even if the model is right, single trades are so noisy that a few hundred of them cannot distinguish one exit rule from another. "Seven to ten days" should be treated as a rule of thumb to be tested. (https://5dlibrary.com/answers/en/implied-volatility-skew-vertical-spreads/c05/) 第5章 时间止损作为最优停时问题 — 「7—10 天是最优的」不能在一般意义上证明。本书按作者的确认,把「天」统一为交易日。把平仓写成最优停时问题后,在作者的模型中,最优的固定平仓日取决于信号衰减速度、股权风险溢价、方差风险溢价、偏度与成本,在 12 种情形中从第 1 个交易日到持有到期不等;基准情形的最优日为第 9 到 10 个交易日,但第 7 到 14 个交易日几乎一样好,持有到期也只差 0.011,而依价格状态而定的规则明显更好。即使模型正确,单笔交易的噪声也大到几百笔交易无法分辨不同的平仓规则。「7—10 天」应当被看作需要检验的经验规则。 (https://5dlibrary.com/answers/zh/implied-volatility-skew-vertical-spreads/c05/) - Ch.6 The Ratchet, the Kelly Criterion and Risks outside the Model — If each loss is capped and a fixed fraction is bet, trading capital never reaches zero in the geometric sense, but that does not make it safe. The Kelly fraction maximises long-run growth and brings deep drawdowns; beyond about twice the Kelly fraction growth turns negative. Following the author's confirmation, the book applies the three-tier ratchet only to incremental profit above the high-water mark: withdrawals come only from new highs, trading capital stays above its previous peak after every withdrawal, and skimming neither eats into existing capital nor changes the sign of the growth rate; but drawdowns of trading capital are the same as under full compounding, and losses can still take it below the starting stake. Skimming every winning trade is a trap: whenever $0.4pb \le q$, trading capital has a negative long-run growth rate. Clustered losses, gaps, liquidity and margin, all outside the model, raise drawdowns substantially and can make growth negative. (https://5dlibrary.com/answers/en/implied-volatility-skew-vertical-spreads/c06/) 第6章 分仓棘轮、凯利准则与模型外风险 — 在每笔亏损有上限、按固定比例下注的假设下,交易资本在几何意义上不会归零,但这不等于没有风险。凯利比例使长期增长率最大,同时伴随很深的回撤;超过约两倍凯利比例,增长率变为负。本书按作者的确认,只对高水位以上的增量利润执行三阶棘轮:提取只来自新高部分,每次提取后交易资本仍高于此前的最高水位,提取本身不会侵蚀本金,也不会改变增长率的符号;但交易资本的回撤与全额复利相同,亏损照样会使它跌破初始本金。按笔提取则是一个陷阱:只要 $0.4pb \le q$,交易资本的长期增长率就为负。亏损扎堆、跳空、流动性与保证金等模型外风险会显著增加回撤,甚至使增长率变为负。 (https://5dlibrary.com/answers/zh/implied-volatility-skew-vertical-spreads/c06/) - Ch.7 Reading the Classics: the Balanced Standard, the Ever-Normal Granary and "Light and Heavy" — The program below writes Ji Ran's price band as the simplest possible buffer-stock model: harvests are lognormal, with a 5 % chance each year of a disaster that cuts the harvest by 40 %; the market price is proportional to the quantity of grain on the market raised to the power −2.5 (inelastic demand), with a normal-year price of 50; the granary buys when the price would fall below 30 (or 45 in the variant), restoring the price to that level, and sells from stock when the price would exceed 80, bringing it back to 80; 5 % of the stock spoils each year. Each row simulates 20,000 years. All parameters are illustrative. (https://5dlibrary.com/answers/en/implied-volatility-skew-vertical-spreads/c07/) 第7章 原典并读:平准、常平与轻重 — 没有粮仓时,77.2% 的年份价格落在 30 到 80 之间,10.4% 的年份超过 90(「病末」),0.6% 的年份低于 20(「病农」)。只在价格跌破 30 时收购的粮仓,库存容量从平均收成的 25% 增加到 50%,效果几乎不变:仍有 11% 以上的年份价格超过 80 而粮仓已空。原因是收购的机会太少:价格跌破 30 需要大丰收,而价格超过 80 的歉收与大灾更常见,库存来不及积累。改为在价格低于 45 时就收购(类似「谷贱时增其贾而籴」),容量为平均收成的一半时,价格超过 90 的年份降到 1.7%,粮仓空仓的年份降到 2.3%;但粮仓的买卖净收入变为负(每年 −0.10,以平年一季收成价值 50 计),容量更大时亏得更多(−0.59),因为它更常在中等价格收购,又要承担损耗。 (https://5dlibrary.com/answers/zh/implied-volatility-skew-vertical-spreads/c07/) - Ch.8 Practice and a Pre-Registered Test — Chapter 8 offers a four-step checklist and a pre-registered out-of-sample test of "a time stop at trading day 9 versus holding to expiry" (single primary outcome the mean paired difference in return per unit of maximum loss, smallest effect of interest 0.15, 50 underlyings, 112 entry dates, about 8.6 years of data, run by a third party, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; every price and trade in the book is fictional or illustrative. (https://5dlibrary.com/answers/en/implied-volatility-skew-vertical-spreads/c08/) 第8章 实践与预注册检验 — 第八章给出四步核查表与一项「第 9 个交易日时间止损对持有到期」的预注册样本外检验方案(唯一主要结局为以最大亏损为单位的配对平均收益差,最小关注效应 0.15,50 个标的、112 个入场日、约 8.6 年数据,第三方执行,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有价格与交易都是虚构或示意的。 (https://5dlibrary.com/answers/zh/implied-volatility-skew-vertical-spreads/c08/) ## 5DL-B-039 · Liquidation Cascades, Order-Book Depth Depletion and Cointegrated Statistical Arbitrage 中文书名:高杠杆衍生品清算瀑布、订单簿深度衰竭与协整统计套利 DOI: https://doi.org/10.5281/zenodo.23796776 Cite: Fan, Y. (2026). Liquidation Cascades, Order-Book Depth Depletion and Cointegrated Statistical Arbitrage (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23796776 Book page: https://5dlibrary.com/books/liquidation-cascades-cointegration-arbitrage/ Scope: This book presents mathematical models for teaching and research and is not investment, trading, tax or legal advice; leverage and derivatives can lose all of the money put in and more, and backtests and simulations do not represent future returns. Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. Abstract: A forced-liquidation sell order punches through a thin bid side, and the fall triggers the next batch of liquidations; another kind of trader looks for two prices that should move together and buys one and sells the other when their spread strays. This book examines both with one body of mathematics. It covers the derivation and simulated check of the price-impact coefficient λ = √Σ₀/(2σᵤ) in Kyle's model, the evidence for the square-root impact law (exponents of about 0.4–0.7 for metaorders, an average of 0.489 in complete account-level data from the Tokyo Stock Exchange) and its scope, and the gaps in a book whose depth decays with distance once its visible depth is used up; the general mechanism of funding in perpetual contracts, the derivation of the isolated-margin liquidation price P₀(1−1/L)/(1−m) (20 times leverage tolerates a fall of only about 4.5 %), and a toy model of chained liquidations driven by "leverage distribution × order-book depth" (with the same leverage mix, cutting depth from 400 to 100 turns a 2 % shock from a 2.5 % fall into a 17.0 % fall); Hawkes self-exciting point processes, the branching ratio and the criticality criterion, Ogata thinning and maximum-likelihood estimation, the dispute over whether markets are near criticality (data with no self-excitation, only baseline switches, can be fitted with a branching ratio of 0.92), and the author's derivation of liquidation cascades as an "effective branching ratio n = ρ/d"; the exact discretisation of the Ornstein–Uhlenbeck process, the half-life ln2/θ and its small-sample bias (with a true half-life of 40 days, the median estimate from one year of data is only 22 days); why the Engle–Granger cointegration test needs MacKinnon's residual critical values (with the Dickey–Fuller −2.86, 14.8 % of unrelated random walks look cointegrated), the decay of pairs-trading returns and broken relationships; and a comparison of Kalman-filter dynamic hedge ratios with rolling least squares (on synthetic pairs, the setting that tracks the hedge ratio best earns less after costs). Claims corrected after checking include: the jiaozi's "first issue of 1,256,340 strings against a reserve of 360,000" are, in the History of Song, the cap per issue period and the reserve per period (about 28.65 %); the "arbitrage character" of price stabilisation and equitable transport is only an analogy; "markets are at criticality" is a disputed estimate; the square-root law applies to metaorders, not to a static order book. The eight chapters are read beside the debate on price stabilisation and equitable transport in the Discourses on Salt and Iron and the record in the History of Song of the jiaozi reserve, of qianyin "worth a dozen-odd coins a string" in the Daguan era and of salt vouchers depreciating through "too many empty vouchers", compared structurally with the Diamond–Dybvig model of runs. They end with a four-step checklist and a pre-registered, prospective out-of-sample test of Kalman-filter hedging against 60-day rolling least squares (single primary outcome the pair-window difference in annualised net return, smallest effect of interest 1 percentage point a year, 513 analysable and 620 planned pair-windows from 31 monthly formation dates, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. This book presents mathematical models for teaching and research and is not investment, trading, tax or legal advice; leverage and derivatives can lose all of the money put in and more, and backtests and simulations do not represent future returns. 摘要:一笔强制平仓卖单打穿薄薄的买盘,价格下跌又触发下一批强平;另一类交易者则寻找本该一起走的两种价格,在价差偏离时买一卖一。本书用同一套数学考察这两件事。依次讨论:凯尔模型中价格冲击系数 λ = √Σ₀/(2σᵤ) 的推导与模拟核对,平方根冲击律的实证(元订单的指数约 0.4—0.7,东京证券交易所全账户数据的平均值 0.489)及其范围,以及深度随距离衰减的订单簿在可见深度被吃光时的跳空;永续合约资金费的一般机制、逐仓强平价 P₀(1−1/L)/(1−m) 的推导(20 倍杠杆只容许约 4.5% 的下跌),以及一个「杠杆分布 × 订单簿深度」的连环清算玩具模型(同一杠杆组合,深度从 400 降到 100,2% 的冲击从 2.5% 的跌幅变成 17.0%);霍克斯自激点过程、分支比与临界判据、奥加塔稀疏化模拟与极大似然估计,关于市场是否接近临界的争论(没有自激、只有基线切换的数据也能拟合出 0.92 的分支比),以及作者把清算瀑布写成「有效分支比 n = ρ/d」的推导;奥恩斯坦—乌伦贝克过程的精确离散化、半衰期 ln2/θ 与短样本偏差(真实半衰期 40 天时,一年数据的中位估计只有 22 天);恩格尔—格兰杰协整检验必须使用麦金农残差临界值(用迪基—富勒的 −2.86 会使 14.8% 的无关随机游走被判为协整),配对交易收益的衰减与关系断裂;卡尔曼滤波动态对冲比率与滚动最小二乘的比较(在合成配对上,追踪对冲比率最准的设定交易净收益反而更低)。 核对后修正的要点包括:交子「首届发行额 1,256,340 缗、本钱 36 万缗」按《宋史·食货志》原文是每界的限额与每界所备本钱(约 28.65%);「平准均输的套利性质」只是类比;「市场处于临界」是有争议的估计;平方根律适用于元订单而非静态订单簿。八章与《盐铁论·本议》的平准均输之辩、《宋史·食货志》所记交子本钱、大观年间钱引「一缗当钱十数」与盐钞「出多虚钞」并读,并与戴蒙德—迪布维格挤兑模型作结构对照;最后给出一份四步核查表,以及一项「卡尔曼滤波对冲与 60 天滚动最小二乘对冲」的预注册前瞻性样本外检验方案(唯一主要结局为逐个配对—窗口的年化净收益差,最小关注效应每年 1 个百分点,需 513 个可分析的配对—窗口、31 个月度形成日共 620 个,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序,全部核对)、原典与作者的假说。本书是教学与研究用的数学模型,不是投资、交易、税务或法律建议;杠杆与衍生品可能造成全部本金甚至超出本金的损失,回测与模拟不代表未来收益。 Chapter answers: - Ch.1 Order-Book Depth and Price Impact — Kyle's model gives the price-impact coefficient $\lambda = \sqrt{\Sigma_0}/(2\sigma_u)$: more noise trading makes a deeper market. Empirically, the impact of small order flow is roughly linear and inversely proportional to depth; the impact of large metaorders roughly follows a square-root law, with exponents between about 0.4 and 0.7, and recent complete account-level data support 1/2. The square-root law describes latent liquidity, while visible depth is finite; when a large order or a stream of liquidation orders uses it up, the price gaps. The next chapter asks where such streams of selling come from: leverage and forced liquidation in perpetual contracts. (https://5dlibrary.com/answers/en/liquidation-cascades-cointegration-arbitrage/c01/) 第1章 订单簿深度与价格冲击 — 凯尔模型给出价格冲击系数 $\lambda = \sqrt{\Sigma_0}/(2\sigma_u)$:噪声交易越多,市场越深。实证上,小订单流的冲击近似线性且与深度成反比;大额元订单的冲击近似服从平方根律,指数约在 0.4 到 0.7 之间,近年的完整账户数据支持 1/2。平方根律描述的是潜在流动性,而可见订单簿的深度有限,当一笔大单或一连串强平单吃光可见深度时,价格会跳空。下一章讨论这种连串卖单从哪里来:永续合约的杠杆与强制平仓。 (https://5dlibrary.com/answers/zh/liquidation-cascades-cointegration-arbitrage/c01/) - Ch.2 Perpetual Contracts, Funding and Forced Liquidation — Perpetual contracts use periodic funding payments to tie their price to spot; funding is charged on notional value and so erodes the margin of leveraged positions many times over. The liquidation price of an isolated long is about $P_0(1-1/L)/(1-m)$; 20 times leverage tolerates a fall of only about 4.5 %, which in volatile markets has a sizeable probability of being touched within a day. The toy model shows that when liquidation prices are concentrated near the current price and the bid side is thin, a small shock can be amplified into chained liquidations. The next chapter describes this amplification with self-exciting point processes and the branching ratio. (https://5dlibrary.com/answers/en/liquidation-cascades-cointegration-arbitrage/c02/) 第2章 永续合约、资金费率与强制平仓 — 永续合约用周期性的资金费把价格拴在现货附近;资金费按名义价值收取,对杠杆仓位的保证金是成倍的侵蚀。逐仓多头的强平价约为 $P_0(1-1/L)/(1-m)$,20 倍杠杆只容许约 4.5% 的下跌,在高波动下一天之内就有可观的概率被触及。玩具模型显示,当强平价集中在现价附近、而买盘较薄时,一个小冲击可以被放大成连环清算。下一章用自激点过程与分支比来描述这种放大。 (https://5dlibrary.com/answers/zh/liquidation-cascades-cointegration-arbitrage/c02/) - Ch.3 Liquidation Cascades, Branching Processes and Hawkes Self-Exciting Point Processes — The Hawkes process describes how far events trigger events through the branching ratio $n$: when $n < 1$ the mean intensity is $\mu/(1-n)$, and as $n$ approaches 1 events cluster strongly with occasional huge chains. The branching ratio can be estimated by recursive likelihood but is extremely sensitive to model specification: data without self-excitation, only with baseline switches, can be fitted with a branching ratio of 0.92, which is the heart of the dispute over whether markets are at criticality. The author writes liquidation cascades as an effective branching ratio in price, $n_{\text{eff}} = \rho/d$, set jointly by the leverage distribution and depth, and local in character. The next chapter turns to the other side of trading: mean reversion. (https://5dlibrary.com/answers/en/liquidation-cascades-cointegration-arbitrage/c03/) 第3章 清算瀑布、分支过程与霍克斯自激点过程 — 霍克斯过程用分支比 $n$ 描述事件引发事件的程度:$n < 1$ 时平均强度为 $\mu/(1-n)$,接近 1 时事件高度成簇、偶有巨大连锁。分支比可以用递推似然估计,但对模型设定极其敏感:没有自激、只有基线切换的数据也能拟合出 0.92 的分支比,这正是关于「市场是否处于临界」的争论的核心。作者把清算瀑布写成价格上的有效分支比 $n_{\text{eff}} = \rho/d$,它由杠杆分布与深度共同决定,并且是局部的。下一章转向交易的另一面:均值回复。 (https://5dlibrary.com/answers/zh/liquidation-cascades-cointegration-arbitrage/c03/) - Ch.4 The Ornstein–Uhlenbeck Process: Mean Reversion and Its Estimation — The OU process $\mathrm{d}X = \theta(\mu - X)\mathrm{d}t + \sigma\mathrm{d}W$ discretises exactly into an AR(1); its half-life is $\ln 2/\theta$ and its maximum-likelihood estimate is the AR(1) regression. The Euler shortcut underestimates the speed of reversion when $\theta h$ is not small. In short samples the AR(1) coefficient is underestimated by about $(1+3b)/n$, and the half-life is systematically underestimated: with a true half-life of 40 days, the median estimate from one year of data is only 22 days, and a pure random walk shows a half-life of under 30 days nearly four times in ten. The next chapter asks how to test whether two prices have a mean-reverting combination: cointegration. (https://5dlibrary.com/answers/en/liquidation-cascades-cointegration-arbitrage/c04/) 第4章 奥恩斯坦—乌伦贝克过程:均值回复与它的估计 — OU 过程 $\mathrm{d}X = \theta(\mu - X)\mathrm{d}t + \sigma\mathrm{d}W$ 可以精确离散化为 AR(1),半衰期为 $\ln 2/\theta$,极大似然估计就是 AR(1) 回归。欧拉捷径在 $\theta h$ 不小时会低估回复速率。短样本中,AR(1) 系数被低估约 $(1+3b)/n$,半衰期被系统地低估;真实半衰期 40 天时,一年数据的中位估计只有 22 天,而纯随机游走也有近四成的机会显示出不到 30 天的半衰期。下一章讨论如何检验「两个价格之间存在一个均值回复的组合」:协整。 (https://5dlibrary.com/answers/zh/liquidation-cascades-cointegration-arbitrage/c04/) - Ch.5 Cointegration, Pairs Trading and Broken Relationships — Regressions between random walks produce spurious relationships; cointegration requires a stationary linear combination, and the Engle–Granger two-step test must use residual-based critical values (two variables with a constant, about −3.34 at 5 %, about −3.36 with 250 observations); using the ordinary Dickey–Fuller −2.86 pushes the error rate close to 15 %. Even with correct critical values, large-scale screening produces hundreds of false pairs, while one year of data has little power to detect genuine cointegration. Historical pairs-trading returns were real but have kept decaying and shrink sharply after costs. When a cointegrating relationship breaks, the rule's returns go to zero or below, and stops only trim the tail. The next chapter takes a different approach: instead of assuming a fixed hedge ratio, let it change over time. (https://5dlibrary.com/answers/en/liquidation-cascades-cointegration-arbitrage/c05/) 第5章 协整、配对交易与关系断裂 — 两条随机游走之间的回归会产生伪关系;协整要求存在平稳的线性组合,用恩格尔—格兰杰两步法检验时必须使用残差检验的临界值(两变量、带常数,5% 约 −3.34,250 个观测约 −3.36),用普通迪基—富勒的 −2.86 会使错判率接近 15%。即使临界值正确,大规模筛选也会制造数以百计的假配对,而一年数据检出真协整的把握度很低。配对交易的历史收益确实存在,但持续衰减,计入成本后大幅缩小。协整关系断裂时,规则的收益趋于零或为负,止损只能截尾。下一章讨论另一种思路:不假设对冲比率固定,而让它随时间变化。 (https://5dlibrary.com/answers/zh/liquidation-cascades-cointegration-arbitrage/c05/) - Ch.6 The Kalman Filter and Dynamic Hedge Ratios — The Kalman filter treats the hedge ratio as a hidden random-walk state and corrects it each day by the innovation times the gain; $Q/R$ sets how quickly it reacts. On synthetic data where its settings are exactly right and observation noise is white, it tracks the hedge ratio far better than rolling regression; but faced with sudden jumps no setting is best everywhere. More importantly, when the spread itself is mean-reverting, the setting that tracks best absorbs the trading signal, and on synthetic pairs its net return falls below that of 60-day rolling regression. Whether the Kalman filter does better on real data must be answered by a pre-registered, out-of-sample test with realistic costs, which is the protocol of Chapter 8. First, the next chapter returns to Sichuan about a thousand years ago, to see how a paper money backed by a "reserve" lost its credit through over-issue. (https://5dlibrary.com/answers/en/liquidation-cascades-cointegration-arbitrage/c06/) 第6章 卡尔曼滤波与动态对冲比率 — 卡尔曼滤波把对冲比率当作随机游走的隐状态,每天用新息按增益修正;$Q/R$ 决定它的反应速度。在设定恰好正确、观测噪声为白噪声的合成数据上,它追踪对冲比率比滚动回归准得多;但面对突然的跳变,没有一种设定处处最好。更重要的是,当价差本身是均值回复的,追踪得最准的设定反而会吸收掉交易信号,在合成配对上的净收益低于 60 天滚动回归。卡尔曼滤波是否在真实数据上更好,必须用预先登记、样本外、扣除现实成本的检验来回答——这是第八章的试验方案。下一章先回到约一千年前的四川,看一种由「本钱」支撑的纸币如何在滥发中失去信用。 (https://5dlibrary.com/answers/zh/liquidation-cascades-cointegration-arbitrage/c06/) - Ch.8 Practice and a Pre-Registered Test — The meaning of "90 % power" needs stating (consistent with Book 33): it is the probability, when the true difference exactly equals the smallest effect of interest, of reaching some verdict that "the Kalman filter is better"; the strongest verdict in the decision rule, "better by at least 1 percentage point", then has only about a 2.5 % chance. The program below computes all these probabilities. (https://5dlibrary.com/answers/en/liquidation-cascades-cointegration-arbitrage/c08/) 第8章 实践与预注册检验 — 四步中共有六处被标记。第一步:15 倍杠杆的强平距离为 6.20%,在示意的波动率下只有 1.7 个日标准差,一天内触及的概率约 8%,三天内约 31%。第二步:一笔占日成交量 2% 的订单,按平方根律估计冲击在 0.26% 到 0.78% 之间,但它是现价 1% 以内可见深度的 1.3 倍:第一章的衰减订单簿说明,这种情况下平方根律给出的数字可能严重低估真实冲击。第三步最有教益:这个虚构配对在前 500 天确实是协整的(程序就是这样生成的),但第一个窗口的检验统计量 −3.27 没有达到 −3.36,第二个窗口勉强通过;第三个窗口中价差已经变成随机游走,统计量只有 −1.42,对冲比率从约 1.2 漂到 1.50,「半衰期」变成 41 天。第四步:对冲比率自开仓以来变了 9.2%,持仓 35 天已超过 3 个半衰期的时间止损。这份核查表并不告诉你该做什么,它只是把几个容易被忽视的数字摆到桌面上。 (https://5dlibrary.com/answers/zh/liquidation-cascades-cointegration-arbitrage/c08/) ## 5DL-B-040 · Macro-Liquidity Regime Switching, Extreme Tail Dependence and Barbell Asset Allocation 中文书名:宏观流动性体制转换、极值尾部相依与杠铃资产配置 DOI: https://doi.org/10.5281/zenodo.23828049 Cite: Fan, Y. (2026). Macro-Liquidity Regime Switching, Extreme Tail Dependence and Barbell Asset Allocation (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23828049 Book page: https://5dlibrary.com/books/tail-risk-copulas-macro-liquidity-regimes/ Scope: When a crisis comes, do assets fall together? How large must reserves and safety cushions be? This book presents mathematical models for teaching and research and is not investment advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. Abstract: When a crisis comes, do assets fall together? How large must reserves and safety cushions be? This book presents mathematical models for teaching and research and is not investment advice. It covers the identification of volatility regimes with Hamilton's Markov-switching model and hidden Markov models (forward–backward, Baum–Welch and Viterbi algorithms, all in the standard library; the expected duration 1/(1−p) holds only under the memoryless assumption, and even with the true states known, ten years of daily data leave a 90 % interval of 23 to 48 days for the duration of the stress regime; real-time filtered probabilities lag the smoothed ones by two days to a week, and with fat-tailed returns an information criterion adds a regime that does not exist); extreme value theory (generalized extreme value and generalized Pareto distributions, the Pickands–Balkema–de Haan theorem, the Hill estimator and the mean-excess plot), a backtest of the conditional EVT method (in 8,000 forecast days normal models break the 99 % VaR 165 to 183 times, conditional EVT 87 times, against 80 expected) and the power of Kupiec's test, only about one in four in a 250-day sample; copulas and tail dependence (the Gaussian copula's tail-dependence coefficient is zero, which is not "zero correlation": the conditional probability of a joint crash at the 1 % quantile is still 0.125, it merely tends to zero), the evidence on both sides, Forbes and Rigobon's "no contagion, only interdependence" and Longin and Solnik's extreme correlation in bear markets, and vine copulas: with the same pairwise rank correlations as a Gaussian copula, the probability that four assets all fall below their 1 % quantiles is about ten times higher. The author asked for a proof that a barbell of "90 % short-term bills + 10 % options" stochastically dominates; the book proves that it cannot be dominated by a stock and, when its options expire worthless more often than the stock falls below the floor (as in the book's model), cannot first-order dominate the stock either, and shows in a jump-diffusion model under four option prices that it neither first- nor second-order dominates the stock or a 90/10 linear mix, and that when options are rich it is itself second-order dominated by the 90/10 mix; CPPI with multiplier 6 ends below its floor in 6.4 % of years. The chapter on sovereign reserves checks the Greenspan–Guidotti rule and the IMF's ARA metric (fixed- and floating-rate weights differ, giving 108 % and 163 % for the same fictional economy) and uses a two-regime liquidity model to show the diminishing marginal protection of a buffer. Chapter 7 reads these beside the "Royal Regulations" of the Book of Rites ("in three years of tilling there must be one year's food; in nine years, three years' food; reckoning over thirty years, even with disasters of drought and flood the people do not look starved"), the Book of Zhou as quoted in the Mozi's "Seven Calamities", and the ever-normal granary of the Book of Han, with a program showing that if grain loses more than 1/27 (about 3.7 %) a year, a "nine-year store" can never be reached. Chapter 8 gives a four-step checklist (regime, tail, dependence, buffer) and a pre-registered backtest of conditional EVT against conditional normal 99 % VaR (single primary outcome the difference in violation rates, smallest effect of interest 0.25 percentage points, 20 series and about 9.2 years of locked data, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. 摘要:危机来临时,资产会一起下跌吗?储备与安全垫要留多少才够?本书是教学与研究用的数学模型,不是投资建议。它依次讨论:用汉密尔顿的马尔可夫切换模型与隐马尔可夫模型识别波动率体制(前向—后向、鲍姆—韦尔奇与维特比算法均用标准库实现;期望持续期 1/(1−p) 只在无记忆的假设下成立,十年日数据即使知道真实状态,危机体制持续期的 90% 区间也宽达 23 到 48 天;实时滤波概率比事后平滑概率晚两天到一周,收益本身厚尾时信息准则会多选出一个并不存在的体制);极值理论(广义极值与广义帕累托分布、皮坎兹—巴尔克马—德哈恩定理、希尔估计量与平均超额图),条件极值方法的回测(8,000 个预测日中正态模型突破 99% VaR 165 到 183 次,条件 EVT 为 87 次,期望 80 次),以及库皮克检验在 250 天样本中把握度只有约四分之一;Copula 与尾部相依(高斯 Copula 的尾部相依系数为零而非「相关为零」,在 1% 分位条件共同暴跌的概率仍有 0.125,只是趋于零),福布斯—里戈本的「没有传染,只有相互依存」与朗金—索尔尼克的熊市极值相关两方证据,以及藤 Copula:与高斯 Copula 两两秩相关相同时,四项资产同时跌破 1% 分位的概率约高十倍。作者要求证明「90% 短债 + 10% 期权」的杠铃配置随机占优;书中证明它不可能被股票占优,且在期权作废的概率高于股票跌破保底的概率时(本书模型即如此)也不可能一阶占优于股票,并在跳跃扩散模型的四种期权定价下显示它对股票、对 90/10 线性组合都不构成一阶或二阶占优,期权偏贵时反而被 90/10 线性组合二阶占优;CPPI 在乘数 6 时有 6.4% 的年份跌破保底线。主权储备一章核对了格林斯潘—吉多蒂规则与 IMF 的 ARA 指标(固定汇率与浮动汇率权重不同,同一虚构经济体分别为 108% 与 163%),并用两体制流动性模型说明缓冲的边际保护递减。 第七章与《礼记·王制》「三年耕,必有一年之食;九年耕,必有三年之食。以三十年之通,虽有凶旱水溢,民无菜色」、《墨子·七患》所引《周书》与《汉书·食货志》的常平仓并读,用程序显示:若粮食每年损耗超过 1/27(约 3.7%),「九年之蓄」永远达不到。第八章给出体制、尾部、相依与缓冲四步核查表,以及一项「条件 EVT 与条件正态 99% VaR」的预注册回测方案(唯一主要结局为两模型突破率之差,最小关注效应 0.25 个百分点,20 个序列、约 9.2 年锁定数据,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有数据都是虚构或模拟的。 Chapter answers: - Ch.1 Hidden Markov Models and Volatility Regimes — A hidden Markov model describes the market as switching between a few unseen regimes. The forward–backward algorithm gives filtered and smoothed probabilities, Baum–Welch estimates the parameters and Viterbi gives the most likely path of states. Hamilton's original 1989 model switched the mean of GNP growth; volatility regimes are an extension. The expected duration $1/(1-p_{ii})$ holds under the memoryless assumption, but durations estimated from ten years of data are very imprecise. After the fact, smoothed probabilities divide history into regimes quite clearly; in real time, filtered probabilities lag and give false alarms, and when returns are fat-tailed an information criterion selects a regime that does not exist. Regime identification is therefore an assumption-dependent probabilistic division, not a direct observation of the state of the market. (https://5dlibrary.com/answers/en/tail-risk-copulas-macro-liquidity-regimes/c01/) 第1章 隐马尔可夫模型与波动率体制 — 隐马尔可夫模型把市场描述为在几个看不见的体制之间切换。前向—后向算法给出滤波与平滑概率,鲍姆—韦尔奇算法估计参数,维特比算法给出最可能的状态路径。汉密尔顿 1989 年的原始模型切换的是 GNP 增长的均值,波动率体制是其推广。期望持续期 $1/(1-p_{ii})$ 在无记忆假设下成立,但从十年数据估计的持续期误差很大。事后的平滑概率能把体制划分得相当清楚,实时的滤波概率则有滞后与误报;收益厚尾时,信息准则会多选出一个不存在的体制。体制识别因此是一种依赖假设的概率划分,而不是对市场状态的直接观测。 (https://5dlibrary.com/answers/zh/tail-risk-copulas-macro-liquidity-regimes/c01/) - Ch.2 Extreme Value Theory: From Tail Index to Expected Shortfall — Finally, the program computes the power of Kupiec's test. Over 250 trading days (about a year, the window commonly used in regulatory backtesting), a model whose true violation rate is 2 % (twice the nominal 1 %) has only a 24 % chance of being rejected; with a true rate of 1.5 %, only 11 %. Rejecting a 2 % model with about 80 % probability takes about 1,000 days. At 250 days the actual size of the test is 9.5 %, not 5 %, because of the discreteness of the binomial distribution. A year of backtesting that "does not reject" says almost nothing about whether a model is right. (https://5dlibrary.com/answers/en/tail-risk-copulas-macro-liquidity-regimes/c02/) 第2章 极值理论:从尾部指数到期望损失 — 对自由度为 3 的学生 t 分布(真实 $\alpha = 3$),希尔估计在 $k = 25$ 到 100 时为 3.40、3.22、2.84,随后一路下降到 $k = 1{,}600$ 时的 1.14:取的数据太多,就把分布的中心部分也当成了尾部。对正态数据,希尔估计从 10.06 降到 1.29,其中 $k = 400$ 时的 3.74 看上去就像一个尾部指数约为 3.7 的幂律厚尾,与 t 分布在小 $k$ 处的估计(3.22 到 3.40)及其真值 3 相差无几,而正态分布根本没有幂律尾部。所以单看某一个 $k$ 的希尔估计,既可能把薄尾误认为厚尾,也说不出真正的尾部指数;需要看整条「希尔图」是否有稳定的平台,并结合平均超额图。图 2 给出了这两种图:t 分布的 $e(u)$ 从 1.08 升到 1.95,近似线性(理论斜率 0.5);正态分布的 $e(u)$ 从 0.54 降到 0.27。用 50 个块最大值估计 GEV 的形状参数(概率加权矩法 (Hosking 1985)),t 分布的中位数为 0.290、90% 区间 0.066 到 0.572,正态分布为 −0.109、区间 −0.277 到 0.071:方向是对的,但 50 个数据点的不确定性很大。 (https://5dlibrary.com/answers/zh/tail-risk-copulas-macro-liquidity-regimes/c02/) - Ch.3 Copulas and Tail Dependence — Sklar's theorem splits a joint distribution into margins and a copula, so that the dependence structure can be modelled separately; rank correlations and tail-dependence coefficients are measures at the level of the copula. The Gaussian copula ($\rho < 1$) has zero tail dependence, but only in the limit: with correlation 0.5, the conditional probability of a joint crash at the 1 % quantile is still 0.125. The Clayton copula has only lower-tail dependence, $2^{-1/\theta}$, the Gumbel copula only upper-tail dependence, $2 - 2{1/\theta}$, and the t copula symmetric dependence in both tails. With the same Kendall's τ, the four copulas give probabilities of both assets falling into their worst 1 % that differ by a factor of more than eight. The accurate core of the 2008 criticism of the Gaussian copula is zero tail dependence and calibration on calm periods, not that "a formula killed Wall Street". (https://5dlibrary.com/answers/en/tail-risk-copulas-macro-liquidity-regimes/c03/) 第3章 Copula 与尾部相依 — 斯克拉定理把联合分布分解为边际分布与 Copula,使相依结构可以单独建模;秩相关与尾部相依系数是 Copula 层面的度量。高斯 Copula($\rho < 1$)的尾部相依系数为零,但这只是极限:相关系数 0.5 时,在 1% 分位上的条件共同暴跌概率仍为 0.125。克莱顿 Copula 只有下尾相依 $2^{-1/\theta}$,古贝尔 Copula 只有上尾相依 $2 - 2{1/\theta}$,t Copula 两尾对称。在相同的肯德尔 τ 下,四种 Copula 给出的两项资产同时落入最坏 1% 的概率相差八倍以上。对 2008 年高斯 Copula 的批评,准确的核心是尾部相依为零与平静期校准,而不是「一个公式毁掉了华尔街」。 (https://5dlibrary.com/answers/zh/tail-risk-copulas-macro-liquidity-regimes/c03/) - Ch.4 Correlation in Crises and Vine Copulas — "Correlations go to one in a crisis" is not a law. Correlation conditional on high volatility rises mechanically, and after their adjustment Forbes and Rigobon found almost no contagion in three crises; but the adjustment itself depends on the model and gives false alarms in a common-factor world. Longin and Solnik found that extreme negative returns are far more correlated than a normal model allows, and since the normal model predicts that more extreme means less correlated, this asymmetry is weighty evidence. Vine copulas build high-dimensional dependence out of bivariate copulas and allow different tail behaviour for different pairs; with exactly the same pairwise rank correlations, a D-vine with Clayton copulas in its first tree makes the probability of all four assets falling below their 1 % quantiles about ten times higher than a Gaussian copula does. (https://5dlibrary.com/answers/en/tail-risk-copulas-macro-liquidity-regimes/c04/) 第4章 危机中的相关性与藤 Copula — 「危机时相关性趋近 1」不是定律。以高波动为条件的相关系数会机械地升高,福布斯与里戈本调整后在三次危机中几乎没有发现传染;但调整本身依赖模型,在共同因子的世界里会误报。朗金与索尔尼克发现极端负收益的相关远高于正态模型的预期,而正态模型预测的恰恰是越极端越不相关,所以这种不对称是有分量的证据。藤 Copula 用一组二元 Copula 搭出高维相依,可以为不同资产对选择不同的尾部行为;在两两秩相关完全相同的条件下,一个以克莱顿 Copula 为第一棵树的 D 藤,使四项资产同时跌破 1% 分位的概率比高斯 Copula 高约十倍。 (https://5dlibrary.com/answers/zh/tail-risk-copulas-macro-liquidity-regimes/c04/) - Ch.5 Barbell Allocation and Stochastic Dominance — A barbell of "90 % bills + 10 % options" has a nominal floor and a convex upper tail, but "stochastic dominance" does not hold in general. The book proves that a barbell with a floor cannot be first- or second-order dominated by a stock, and that when its options expire worthless more often than the stock falls below the floor (56.4 % against 20.5 % in this chapter's model) it cannot first-order dominate the stock either; second-order dominance requires an expected return that is no lower, which depends on option prices. In the author's jump-diffusion model, under four option prices, the barbell shows neither first- nor second-order dominance over the stock or the 90/10 linear portfolio; when options are rich, it is itself second-order dominated by the 90/10 portfolio. Options that are dear over the long run (the variance risk premium) mean the barbell's option sleeve is paying an insurance premium in most years. CPPI is a dynamic way to protect a floor; the larger the multiplier, the greater the risk that a gap breaches the floor. (https://5dlibrary.com/answers/en/tail-risk-copulas-macro-liquidity-regimes/c05/) 第5章 杠铃配置与随机占优 — 「90% 短债 + 10% 期权」的杠铃组合有名义下限与凸的上尾,但「随机占优」在一般意义上不成立。本书证明:有保底的杠铃组合不可能被股票一阶或二阶占优;当期权作废的概率高于股票跌破保底的概率时(本章模型中 56.4% 对 20.5%),它也不可能一阶占优股票;二阶占优要求期望收益不低,而这取决于期权的价格。在作者的跳跃扩散模型中,杠铃在四种期权定价下对股票、对 90/10 线性组合都不构成一阶或二阶占优;期权偏贵时,它反而被 90/10 线性组合二阶占优。期权长期偏贵(方差风险溢价)意味着杠铃的期权仓位多数年份是在交保险费。CPPI 是动态的保底方法,乘数越大,跳空击穿保底的风险越大。 (https://5dlibrary.com/answers/zh/tail-risk-copulas-macro-liquidity-regimes/c05/) - Ch.6 Sovereign Buffers and Liquidity Regimes — Foreign-exchange reserves are self-insurance against sudden stops of capital. The Greenspan–Guidotti rule requires usable reserves to cover external debt falling due within a year; the IMF's ARA metric applies regime-dependent weights (short-term external debt 30 %, other portfolio liabilities 20 % or 15 %, broad money and exports 10 % or 5 % each) and suggests reserves of 100–150 % of the metric. The same fictional economy scores 108 % under a fixed rate and 163 % under a floating one, and different measures can give different verdicts. The author's two-regime buffer model shows that more reserves reduce the crisis probability ever more slowly while the holding cost rises linearly; the "optimal" level depends on the cost of a crisis and the nature of liquidity regimes, and rules of thumb are only references. (https://5dlibrary.com/answers/en/tail-risk-copulas-macro-liquidity-regimes/c06/) 第6章 主权缓冲与流动性体制 — 外汇储备是对资本突然停止的自我保险。格林斯潘—吉多蒂规则要求可用储备覆盖未来一年到期的外债;IMF 的 ARA 指标按汇率制度给出不同权重(短期外债 30%,其他证券投资负债 20% 或 15%,广义货币与出口各 10% 或 5%),建议储备为指标的 100%–150%。同一个虚构经济体在固定与浮动汇率下的充足率分别为 108% 与 163%,不同指标可以给出不同结论。作者的两体制缓冲模型显示,储备越多,危机概率下降得越慢,而持有成本线性增加;「最优」水平取决于危机的代价与流动性体制的性质,经验规则只是参照。 (https://5dlibrary.com/answers/zh/tail-risk-copulas-macro-liquidity-regimes/c06/) - Ch.7 Reading the Classics: the "Nine-Year Store" and the Ever-Normal Granary — Chapter 7 reads these beside the "Royal Regulations" of the Book of Rites ("in three years of tilling there must be one year's food; in nine years, three years' food; reckoning over thirty years, even with disasters of drought and flood the people do not look starved"), the Book of Zhou as quoted in the Mozi's "Seven Calamities", and the ever-normal granary of the Book of Han, with a program showing that if grain loses more than 1/27 (about 3.7 %) a year, a "nine-year store" can never be reached. (https://5dlibrary.com/answers/en/tail-risk-copulas-macro-liquidity-regimes/c07/) 第7章 原典并读:「九年之蓄」与常平 — 第七章与《礼记·王制》「三年耕,必有一年之食;九年耕,必有三年之食。以三十年之通,虽有凶旱水溢,民无菜色」、《墨子·七患》所引《周书》与《汉书·食货志》的常平仓并读,用程序显示:若粮食每年损耗超过 1/27(约 3.7%),「九年之蓄」永远达不到。 (https://5dlibrary.com/answers/zh/tail-risk-copulas-macro-liquidity-regimes/c07/) - Ch.8 Practice and a Pre-Registered Test — Chapter 8 gives a four-step checklist (regime, tail, dependence, buffer) and a pre-registered backtest of conditional EVT against conditional normal 99 % VaR (single primary outcome the difference in violation rates, smallest effect of interest 0.25 percentage points, 20 series and about 9.2 years of locked data, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. (https://5dlibrary.com/answers/en/tail-risk-copulas-macro-liquidity-regimes/c08/) 第8章 实践与预注册检验 — 第八章给出体制、尾部、相依与缓冲四步核查表,以及一项「条件 EVT 与条件正态 99% VaR」的预注册回测方案(唯一主要结局为两模型突破率之差,最小关注效应 0.25 个百分点,20 个序列、约 9.2 年锁定数据,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有数据都是虚构或模拟的。 (https://5dlibrary.com/answers/zh/tail-risk-copulas-macro-liquidity-regimes/c08/) ## 5DL-B-041 · Power-Law Weighting and Bottom-Up Controllability in Multi-Tier Conditional Excitation Systems 中文书名:多层条件激发系统中的幂律权重与逆向控制可及性 DOI: https://doi.org/10.5281/zenodo.23810093 Cite: Fan, Y. (2026). Power-Law Weighting and Bottom-Up Controllability in Multi-Tier Conditional Excitation Systems (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23810093 Book page: https://5dlibrary.com/books/mechanism-design-bidirectional-hierarchy/ Scope: The book presents mathematical models for teaching and research; it is not investment, trading, career, tax or legal advice and must not be used to rate or screen anyone. Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. Abstract: Many people carry a "pyramid of success" in their heads, with endowment and fortune at the top and the learning and effort they control at the bottom. This book turns the author's "bidirectional pyramid of the five success factors" (endogenous prior, timing and cycles, environmental topology, reputational games, knowledge investment) into explicit system and probability models, keeping the author's model, published evidence and the author's hypotheses apart. It covers the weakest-link property, log-additivity and log-normality of multiplicative and conditional-probability-chain models (Shockley's 1957 product of factors), and the fact that the chain holds in any order and implies no hierarchy by itself; the author's 50/25/12.5/7.5/5 weights (a strict binary sequence sums to 96.875 %, and halving is a geometric sequence rather than a power law), with Shapley values (LMG) and Sobol' indices showing that variance shares depend on correlation, order and scale (in one model the fifth tier's share ranges from 2.0 % to 20.6 %); the population-level explanatory power of polygenic scores for educational attainment (11–16 %), its shrinkage within families, genetic nurture and portability, and the long-run earnings losses of graduating in a recession and their small variance share; the Moving to Opportunity experiment, the Matthew effect, the artificial music market and the "talent versus luck" model; and, with a linear control system, Kalman's rank condition and the controllability Gramian, a proof that under the author's assumption of an unchangeable endowment the top-tier prior is an uncontrollable state, so that bottom-tier investment can raise its realization but not its potential, with realization capped by potential; controllability depends only on the coupling structure, a "plastic prior" needs thousands of times the energy of the bottom tier, and under crowding-out coupling more investment lowers the overall outcome. The eight chapters are read beside Wang Chong's Lunheng ("Ming lu": "if one's fate is to be poor and lowly, then even if made rich and noble one will still run into calamity"), the Chan master Yungu's "fate is made by oneself; fortune is sought by oneself" in Yuan Liaofan's "Liming zhi xue", and the Xunzi's "the gentleman is not different by birth; he is good at making use of things", stating clearly the boundary between classical ideas of fate and modern models. They end with a four-step checklist (find the weakest tier, estimate controllable inputs, assess the coupling, design feedback) and a pre-registered randomized controlled trial comparing structured support aimed at the weakest reachable tier with the same hours of general employability training (primary outcome administrative earnings in months 7–18; 786 analysable and 874 recruited per arm; decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses. The book presents mathematical models for teaching and research; it is not investment, trading, career, tax or legal advice and must not be used to rate or screen anyone. 摘要:许多人心里都有一座「成功金字塔」:顶层是先天与时运,底层是自己能掌握的学习与努力。本书把作者的「成功五要素双向金字塔」(先天先验、时机与周期、环境拓扑、声誉博弈、知识投入)写成严格的系统与概率模型,并把作者的模型、已发表的证据与作者的假说分开。依次讨论:乘法与条件概率链模型的短板效应、对数可加与对数正态分布(肖克利 1957 年的多因子相乘),以及链式写法按任何顺序都成立、本身不蕴含层级;作者的 50/25/12.5/7.5/5 权重(严格的二进序列合计 96.875%,且逐层减半是几何级数而非幂律),以及用沙普利值(LMG)与索博尔指数说明方差份额依赖相关、顺序与尺度(同一模型中第五层的份额可从 2.0% 变到 20.6%);教育成就多基因评分的群体解释力(11%—16%)及其在家族内的缩小、遗传养育与可移植性,衰退期毕业的长期收入损失及其很小的方差份额;「搬向机会」实验、马太效应、人造音乐市场与「才能与运气」模型;用线性控制系统、卡尔曼秩判据与可控性格拉姆矩阵证明:在作者「先天不可改变」的设定下,顶层先验是不可控的状态,底层投入只能提高其「兑现」而不能提高「潜能」,兑现以潜能为上限;可控性只取决于耦合结构,「先验可塑」时所需能量是推动底层的数千倍,「挤出」耦合下加大投入反而降低总结果。 八章与王充《论衡·命禄》「命当贫贱,虽富贵之,犹涉祸患矣」、《了凡四训·立命之学》中云谷禅师的「命由我作,福自己求」、《荀子·劝学》「君子生非异也,善假于物也」并读,并明确古代命定观与现代模型的界限;最后给出一份四步核查表(识别短板层、估计可控输入、评估耦合、设计反馈),以及一项比较「针对最弱可及层的结构化支持」与同等时长一般就业技能培训的预注册随机对照试验方案(主要结局为第 7—18 个月行政记录收入,每组 786 人可分析、招募 874 人,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序,全部核对)、原典与作者的假说。本书是教学与研究用的数学模型,不是投资、交易、职业、税务或法律建议,也不能用于评价或筛选任何人。 Chapter answers: - Ch.1 Products and Cascades: The Conditional Excitation Equation — Multiplying five conditions gives three exact properties: any tier close to zero drags down the whole outcome; the model is additive in logarithms, so variance can be decomposed on the log scale; and a product of independent factors has a right-skewed, log-normal distribution. Shockley explained the skewed distribution of research output in 1957 by multiplying several hurdles, but his factors were an explanatory assumption, and the shape of the distribution of individual performance is itself disputed. The conditional-probability chain is an identity that holds in any order and implies no hierarchy. The next chapter turns to the weights the author assigns to the five tiers, and to why "how much does each tier explain" has no single answer. (https://5dlibrary.com/answers/en/mechanism-design-bidirectional-hierarchy/c01/) 第1章 乘积与级联:条件激发方程 — 把五个条件相乘,会得到三个严格的性质:任何一层接近零都会压低总结果;对数可加,所以在对数尺度上可以讨论方差分解;多个独立因子相乘产生右偏的对数正态分布。肖克利 1957 年用多道关卡相乘解释了科研产出的右偏分布,但他的因子是解释性的假设;个人绩效分布的形状本身也有争议。条件概率链是一个按任何顺序都成立的恒等式,它不蕴含层级。下一章讨论作者为五层指定的权重,以及「每层解释多少」这个问题为什么没有唯一答案。 (https://5dlibrary.com/answers/zh/mechanism-design-bidirectional-hierarchy/c01/) - Ch.2 Variance Shares: The Author's Weights and Three Ways to Split — The author's 50/25/12.5/7.5/5 is not a strict binary sequence (which sums to 96.875 %), and $2^{-k}$ is a geometric sequence rather than a power law; the numbers are the author's assumption, not an empirical estimate. With independent tiers every splitting method gives the same shares; once the tiers correlate, the shares depend on the rule, a top-down order systematically favours the upper tiers, and the Shapley value (LMG) is the convention of averaging over all orders. On the raw scale the multiplicative model has interactions; the greater the spread, the more they matter and the more the shares depend on scale and convention. The next chapter asks what published research can say about endowment and timing. (https://5dlibrary.com/answers/en/mechanism-design-bidirectional-hierarchy/c02/) 第2章 方差贡献:作者的权重与三种拆分 — 作者的 50/25/12.5/7.5/5 不是严格的二进序列(严格序列合计 96.875%),$2^{-k}$ 也不是幂律而是几何级数;这组数字是作者的假设,不是实证估计。各层独立时,所有拆分方法给出相同的份额;层与层一旦相关,份额就取决于拆分规则,自上而下的顺序会系统地偏向上层,沙普利值(LMG)是对所有顺序取平均的约定。乘法模型在原始尺度上含有交互作用,离散越大,交互越重要,份额越依赖尺度与约定。下一章看已发表的研究对先天与时机能说什么。 (https://5dlibrary.com/answers/zh/mechanism-design-bidirectional-hierarchy/c02/) - Ch.3 Endowment and Timing: The Published Evidence — Polygenic scores for educational attainment explain 11–16 % of the variance in the population, but the figure includes genetic nurture and stratification; within families the direct effect is about half the magnitude of the association and about a quarter in terms of variance; the scores predict markedly less well in other ancestry groups and cannot be used to judge individuals. Graduating into a recession brings real and lasting earnings losses, yet entry timing may account for only a small share of the variance in earnings across everyone. The evidence on neither tier supports "endowment and timing together are 75 %", and neither supports "they do not matter". The next chapter turns to environment, reputation and luck. (https://5dlibrary.com/answers/en/mechanism-design-bidirectional-hierarchy/c03/) 第3章 先天与时机:已发表的证据 — 教育成就的多基因评分在群体中可解释 11%—16% 的变异,但其中混入了遗传养育与人口分层;家族内的直接效应只有关联强度的一半左右,按方差约为四分之一;评分在其他祖源人群中的预测力明显下降,不能用于评价个人。在衰退中毕业会带来真实而持久的收入损失,但在全体人的收入差异中,入职时机所占的方差份额可能很小。两层的已有证据都不支持「先天与时机合计占 75%」,也都不支持「它们无关紧要」。下一章讨论环境、声誉与运气。 (https://5dlibrary.com/answers/zh/mechanism-design-bidirectional-hierarchy/c03/) - Ch.4 Environment, Reputation and Luck — The Moving to Opportunity experiment shows that under random assignment the environment changed children's adult earnings, but the effect depends on age at the move: environment and timing interact. Studies of the Matthew effect, the artificial music market, randomly given early success and funding thresholds together show that early, partly random success is amplified, raising inequality and lowering predictability. "Talent versus luck" is a model whose conclusions follow from its assumed rules. Deliberate practice explains between about 26 % and under 1 % of the differences in performance in different domains, but these are variance shares across people, not the effect of practice on an individual. The next chapter turns to the second direction of the author's framework: how far one can push from the bottom. (https://5dlibrary.com/answers/en/mechanism-design-bidirectional-hierarchy/c04/) 第4章 环境、声誉与运气 — 「搬向机会」实验显示,环境在随机分配下改变了孩子的成年收入,但效应依赖搬迁时的年龄,环境与时机存在交互作用。马太效应、人造音乐市场、随机早期成功与资助门槛的研究共同说明:早期的、部分随机的成功会被放大,既提高不平等,也降低可预测性。「才能与运气」是一个模型,它的结论来自假设的规则。刻意练习在不同领域解释的表现差异从约 26% 到不足 1%,但这是人与人之间的方差份额,不是练习对个人的作用。下一章转向作者框架的第二个方向:从底层出发,能推动多少。 (https://5dlibrary.com/answers/zh/mechanism-design-bidirectional-hierarchy/c04/) - Ch.5 Bottom-Up Controllability — Written as a linear control system, the author's assumption that "endowment cannot change" implies directly that $C_1$ is an uncontrollable mode: the Gramian is zero in that direction and no input from the bottom can move it. Which tiers an input can reach depends only on the coupling structure: with top-down couplings alone, the input at the bottom goes nowhere; with bottom-up feedback, $C_2$–$C_5$ are controllable; only under the assumption of a plastic prior is $C_1$ controllable, and then at thousands of times the energy needed for $C_5$. "The higher, the harder" holds under attenuating couplings but changes with coupling strength and can even reverse under strong feedback. The next chapter separates "potential" from "realization" and asks what the author's "unlocking" is in the model. (https://5dlibrary.com/answers/en/mechanism-design-bidirectional-hierarchy/c05/) 第5章 自下而上的可控性 — 把五层写成线性控制系统后,作者设定的「先天不可改变」直接推出:$C_1$ 是不可控的模态,格拉姆矩阵在这一方向为零,任何来自底层的输入都推不动它。输入能到达哪些层只取决于耦合结构:只有自上而下的耦合时,底层的输入哪里也去不了;有了自下而上的反馈,$C_2$—$C_5$ 可控;只有在「先验可塑」的假设下 $C_1$ 才可控,但所需能量是推动 $C_5$ 的数千倍。「越上层越难推动」在逐层衰减的耦合下成立,但随耦合强度变化,在强反馈下甚至会颠倒。下一章区分「潜能」与「兑现」,看作者所说的「解锁」在模型中究竟是什么。 (https://5dlibrary.com/answers/zh/mechanism-design-bidirectional-hierarchy/c05/) - Ch.6 Potential and Realization — There are four results. First, the potential $P$ is the same in every row: raising the input from 0 to 16 changes $P$ in none of them. Second, the realized fraction $g$ rises from 0.12 to nearly 1: 0.96 at $u = 4$ and 0.9998 at $u = 8$. For the person with potential 0.5, realization rises from 0.060 to nearly 0.5; for the person with potential 0.9, from 0.107 to nearly 0.9. Realization can approach potential but not exceed it. The author's "$C_5 = 100\%$" corresponds in the model to a large input with $g$ close to 1: it pushes realization towards the ceiling, but the ceiling itself does not move. (https://5dlibrary.com/answers/en/mechanism-design-bidirectional-hierarchy/c06/) 第6章 潜能与兑现 — 结果有四点。第一,潜能 $P$ 在每一行都不变:输入从 0 增加到 16,没有任何一行改变了 $P$。第二,兑现比例 $g$ 从 0.12 上升到接近 1:$u = 4$ 时为 0.96,$u = 8$ 时为 0.9998。对潜能 0.5 的人,兑现从 0.060 上升到接近 0.5;对潜能 0.9 的人,从 0.107 上升到接近 0.9。兑现可以逼近潜能,但不能超过潜能。作者框架中的「$C_5 = 100\%$」,在模型中对应输入很大、$g$ 接近 1 的情形:它把兑现推向天花板,但天花板本身不动。 (https://5dlibrary.com/answers/zh/mechanism-design-bidirectional-hierarchy/c06/) - Ch.8 Practice and a Pre-Registered Trial — In this invented example L.'s weakest tier is environment (0.30), whose marginal value of 0.154 is far above the others'. The available inputs are an evening course (acting on $C_5$), a mentor network ($C_4$) and relocation support ($C_3$); nothing acts on $C_1$ or $C_2$. Step 3 shows that the evening course changes $\ln S$ by +2.60 under UP and by −0.85 under crowding-out, so its direction can flip; the mentor network and relocation support are positive under both. In step 4, "always take the course" raises the overall outcome only 1.13-fold in 12 months ($C_5$ from 0.80 to 0.90), while "invest in the weakest reachable tier each month" raises it 2.33-fold ($C_3$ from 0.30 to 0.60, $C_4$ from 0.50 to 0.58). (https://5dlibrary.com/answers/en/mechanism-design-bidirectional-hierarchy/c08/) 第8章 实践与预注册试验 — 在这个虚构的例子中,L. 最弱的是环境层(0.30),其边际价值 0.154 远高于其他各层。可用的输入是夜间课程(作用于 $C_5$)、导师网络($C_4$)与迁移支持($C_3$),没有任何输入作用于 $C_1$ 或 $C_2$。第三步显示,夜间课程在 UP 假设下对 $\ln S$ 的作用是 +2.60,在挤出假设下是 −0.85,方向会翻转;导师网络与迁移支持在两种假设下都为正。第四步中,「一直上课」12 个月后总结果只提高到 1.13 倍($C_5$ 从 0.80 升到 0.90),「每月投入最弱的可及层」提高到 2.33 倍($C_3$ 从 0.30 升到 0.60,$C_4$ 从 0.50 升到 0.58)。 (https://5dlibrary.com/answers/zh/mechanism-design-bidirectional-hierarchy/c08/) ## 5DL-B-042 · Bayesian Search in Unknown Landscapes, Optimal Stopping and Gittins Indices 中文书名:未知地形中的贝叶斯搜索、秘书问题与吉廷斯指数 DOI: https://doi.org/10.5281/zenodo.23810102 Cite: Fan, Y. (2026). Bayesian Search in Unknown Landscapes, Optimal Stopping and Gittins Indices (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23810102 Book page: https://5dlibrary.com/books/bayesian-search-optimal-stopping-exploration/ Scope: Marked boxes separate published evidence, runnable Python calculations (17 programs, all checked), classical sources and the author's hypotheses. Abstract: Look at about 37 % of the applicants without hiring, then hire the first one better than everyone before, and you choose the best with probability about 37 %: this "37 % rule" is widely repeated, but the conditions under which it holds are demanding. This book uses statistical decision theory to examine three related questions: when to stop observing and act (stopping times), where to look (Bayesian search), and how much immediate cost to accept for the sake of information (the exploration–exploitation trade-off). It covers the exact optimal threshold and success probability of the secretary problem for finite n (skip 3 of 10 with success probability 0.399; 0.371 for n = 100; limit 1/e; flat near the optimum, and for n = 40 the rule ends with nobody in 37.5 % of cases) and Ferguson's 1989 history (Gardner's original game is a two-person game, and Kepler's story violates "no recall"); variants with full information (about 0.58), minimal expected rank (limit about 3.8695, against 19.6 for the 37 % rule at n = 100), offers that may be refused, recall and an unknown number of applicants (best threshold about 13.5 %, success probability about 0.27), and how people stop too early in experiments yet quickly approach the optimum in repeated games whose distribution can be learned; Koopman's search theory, Bayesian updating and the greedy rule "posterior probability × detection probability", and the searches for the wreckage of Scorpion and AF 447 (where the key step was re-examining the assumption that the beacons worked, and a joint model is robust to both truths on an illustrative grid), together with Charnov's marginal value theorem from foraging ecology (the farther the travel, the longer the stay, and a patch is left before it is exhausted) and the dispute over the Lévy-flight foraging hypothesis (μ = 2 is near-optimal only for sparse, revisitable targets, the optimum in a one-dimensional model approaches 2 as targets become sparser, straight-line motion is better for one-off targets, and the albatross power law of 1996 vanished in a 2007 re-analysis); the Gittins index computed by calibration and by restart-in-state (0.7029 for Beta(1, 1) at discount 0.9) and the conditions for its optimality (in a two-armed problem with discount 0.9 a greedy policy loses only about 0.3 %), Thompson sampling, UCB1 and the asymptotic lower bound of Lai and Robbins (at finite T the regret of Thompson sampling can lie below c ln T); the evidence and disputes on sunk costs (Arkes and Blumer 1985; a pooled effect of about 0.5 in the meta-analysis of Roth and colleagues, mostly from hypothetical scenarios but larger in field studies of project decisions; small and erratic effects in laboratory experiments with real incentives), the difference between rational and fallacious persistence in a one-armed bandit (with illustrative parameters, myopic early quitting loses about 29 % and adding a sunk-cost tendency to the optimal rule about 6 %), and stop-loss boundaries set in advance by Wald's sequential probability ratio test; the two objectives of earning while learning and learning then earning in testing start-up lanes, and a best cutoff of about √n when selection is scored by actual quality. The eight chapters are read beside "in war victory is prized, not duration" from the "Waging War" chapter of the Sunzi, "speed is what the army prizes" from the biography of Guo Jia in the Records of the Three Kingdoms (not from the Sunzi), "there are roads one does not follow" from the "Nine Variations" (the framework's 涂 noted against the received 途), the "Manoeuvring" chapter, and "try them in office and assess their accomplishments" from the "Eminent Teachings" and "examining whether forms and names tally" from the "Two Handles" of the Han Feizi (the received reading 言异事也 noted), stating where classical military and political experience ends and the modern theory of stopping begins. They end with a four-step checklist for searching and stopping and a pre-registered online randomised controlled trial comparing "learning a stopping-threshold rule in advance" with control instructions (20 games of the classical secretary problem with 40 virtual applicants each; single primary outcome the share of games in which the overall best is chosen; smallest effect of interest 4 percentage points; 90 % power; 190 analysable and 224 recruited per arm; decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (17 programs, all checked), classical sources and the author's hypotheses. The book is a teaching and research text on mathematical models, not investment, start-up, career or human-resources advice. 摘要:先看过约 37% 的人而不录用、之后录用第一个比之前所有人都好的人,选中最佳者的概率约为 37%——这条「37% 法则」流传很广,但它成立的条件相当苛刻。本书用统计决策论考察三类相关的问题:什么时候停止观察并行动(停时),往哪里找(贝叶斯搜索),以及为了信息愿意付出多少眼前的代价(探索—利用权衡)。依次讨论:秘书问题在有限 n 时的精确最优阈值与成功概率(n = 10 时跳过 3 人、成功概率 0.399,n = 100 时 0.371,极限 1/e;最优点附近很平坦,n = 40 时有 37.5% 的情形空手而归),弗格森 1989 年的考证(加德纳的原始游戏是两人博弈,开普勒的故事违反「不能回头」);完全信息版本(约 0.58)、最小期望名次(极限约 3.8695,而 37% 规则在 n = 100 时为 19.6)、录用可能被拒绝、可以回头与人数未知(最佳阈值约 13.5%,成功概率约 0.27)等变体,以及实验中人们停得太早与在可学习的重复游戏中迅速接近最优;库普曼的搜索论、贝叶斯更新与「后验概率 × 发现概率」的贪婪规则,「天蝎号」与法航 447 的残骸搜寻(后者的关键在于重新检查信标是否工作的假设,联合模型在示意网格中对两种真相都稳健),以及觅食生态学中查诺夫的边际价值定理(路程越远停留越久,离开时斑块并未吃空)与列维飞行觅食假说之争(μ = 2 只在目标稀疏、可重复访问时近似最优,一维模型中目标越稀疏最优值越接近 2,一次性目标时近直线运动更好;1996 年信天翁的幂律在 2007 年的重新分析中消失);吉廷斯指数的校准法与重启法计算(Beta(1, 1)、折扣 0.9 时为 0.7029)及其成立条件(折扣 0.9 的两臂问题中贪婪策略只少约 0.3%),汤普森抽样、UCB1 与莱—罗宾斯渐近下界(有限 T 下汤普森抽样的后悔可低于 c ln T);沉没成本的证据与争议(阿克斯与布卢默 1985、罗特等人 2015 年元分析的合并效应约 0.5、多为假想情境,但项目推进类决策的现场研究中效应更大;有真实激励的实验室实验中效应小而不稳定),单臂老虎机中「理性的坚持」与「谬误的坚持」的区别(示意参数下短视的过早放弃损失约 29%,最优规则加上沉没成本倾向损失约 6%),以及沃尔德序贯概率比检验给出的预设止损边界;赛道测试中边学边赚与先学后赚两种目标,以及人才筛选中按实际水平计分时最佳阈值约为 √n。 八章与《孙子兵法·作战篇》「兵贵胜,不贵久」、《三国志·魏书·郭嘉传》「兵贵神速」(不出自《孙子兵法》)、《九变篇》「途有所不由」(框架作「涂」,已注明)、《军争篇》以及《韩非子·显学》「试之官职,课其功伐」与《二柄》「审合刑名」(今本作「言异事也」,已注明)并读,并明确古代军事与统治经验与现代停时理论的界限;最后给出一份四步搜索与停止核查表,以及一项比较「事先学习停止阈值规则」与对照说明的预注册在线随机对照试验方案(任务为 20 局各 40 名虚拟求职者的经典秘书问题,唯一主要结局为选中全体最佳者的局数比例,最小关注效应 4 个百分点,把握度 90%,每组 190 人可分析、招募 224 人,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序,全部核对)、原典与作者的假说。本书是教学与研究用的数学模型,不是投资、创业、职业或人力资源建议。 Chapter answers: - Ch.1 The Secretary Problem and 1/e: When to Stop If You See Only Ranks and Cannot Go Back — With a known number of applicants, random order, relative ranks only, no recall and only the best counting as success, the optimal rule is a cutoff rule: skip about $n/e$ applicants, then hire the first relative best; the success probability tends to $1/e \approx 0.368$. For finite $n$ the optimal share skipped and the success probability differ slightly and can be computed exactly; the success probability is flat near the optimum. About 63 % of the time the rule does not choose the best, and for $n = 40$ it ends with nobody in 37.5 % of cases. Ferguson's history shows that Gardner's original game was a two-person game and that the standard solutions answer a narrower question; Kepler's story violates the "no recall" assumption. The next chapter relaxes these conditions. (https://5dlibrary.com/answers/en/bayesian-search-optimal-stopping-exploration/c01/) 第1章 秘书问题与 1/e:只看名次、不能回头时何时停下 — 在人数已知、随机顺序、只看相对名次、不能回头、只有第一名算成功的条件下,最优规则是阈值规则:先跳过约 $n/e$ 人,之后录用第一个相对最佳者,成功概率趋于 $1/e \approx 0.368$;有限 $n$ 时最优跳过比例与成功概率略有不同,可以精确计算,最优点附近成功概率很平坦。约 63% 的时候规则没有选中第一名,$n = 40$ 时有 37.5% 的情形空手而归。弗格森的考证说明,加德纳最初的游戏是一个两人博弈,至今的标准解法回答的是更窄的问题;开普勒的故事违反了「不能回头」的假设。下一章放宽这些条件。 (https://5dlibrary.com/answers/zh/bayesian-search-optimal-stopping-exploration/c01/) - Ch.2 Variants of the Secretary Problem and How People Stop — When scores are visible and their distribution is known, the optimal threshold changes with the number of applicants remaining and the success probability is about 0.58. With expected rank as the objective, the optimal rule achieves an expected rank of only about 3.87, whereas under the 37 % rule the expected rank for $n = 100$ is about 19.6. When offers may be refused, one should act earlier; when recall is possible, later. When the number of applicants is unknown, the best single threshold skips only about 13.5 %, with success probability about 0.27. In experiments, people tend to stop too early in tasks that give only ranks, but quickly approach the optimum in repeated games in which the distribution can be learned. The next chapter turns from "when to stop" to "where to look". (https://5dlibrary.com/answers/en/bayesian-search-optimal-stopping-exploration/c02/) 第2章 秘书问题的变体与人的停止行为 — 看得到分数并知道分布时,最优阈值随剩余人数变化,成功概率约为 0.58;以期望名次为目标时,最优规则的期望名次只有约 3.87,而 37% 规则下 $n = 100$ 的期望名次约为 19.6。录用可能被拒绝时应当更早出手,可以回头时应当更晚出手;人数未知时,最佳单一阈值只跳过约 13.5%,成功概率约 0.27。实验中人们在只给名次的任务里倾向于停得太早,而在可以学习分布的重复游戏中很快接近最优。下一章从「何时停」转向「往哪里找」。 (https://5dlibrary.com/answers/zh/bayesian-search-optimal-stopping-exploration/c02/) - Ch.3 Bayesian Search: Where to Look, and What Not Finding Means — Search theory writes "where to look" in terms of three ingredients: the prior, the detection function and Bayesian updating; the greedy rule for the next search picks the place with the largest product of posterior probability and detection probability. Not finding is information, pushing probability from easy places towards hard ones. On the illustrative grid, searching by the product is faster than looking only at probability or sweeping outwards. The Scorpion and AF 447 cases show that a good prior and honest detection probabilities are crucial; AF 447 in particular shows that assumptions about the effectiveness of past searches must themselves be put into the model and updated. In foraging ecology the marginal value theorem writes "when to leave a patch" as stopping when the marginal rate of gain falls to the average rate; "Lévy flights are optimal" holds only under conditions such as sparse, revisitable targets, and whether animals really search this way is still disputed. The next chapter turns to another kind of search: allocating a limited number of tries among several options with uncertain payoffs. (https://5dlibrary.com/answers/en/bayesian-search-optimal-stopping-exploration/c03/) 第3章 贝叶斯搜索:往哪里找,没找到又说明了什么 — 搜索论把「往哪里找」写成先验、发现函数与贝叶斯更新三个要素;下一次搜索的贪婪规则是选后验概率与发现概率乘积最大的地方。没有发现也是信息,它把概率从容易搜的地方推向难搜的地方。在示意的网格中,按乘积搜索比只看概率或由近及远扫过更快。「天蝎号」与法航 447 的经验说明,好的先验与诚实的发现概率是关键;法航 447 一例尤其说明,对过去搜索效果的假设本身需要放进模型里更新。觅食生态学的边际价值定理把「何时离开一个斑块」写成边际收获率降到平均收获率时停止;「列维飞行最优」则只在目标稀疏、可重复访问等条件下成立,动物是否真的这样搜索仍有争议。下一章讨论另一种搜索:在几个收益不确定的选项之间分配有限的尝试。 (https://5dlibrary.com/answers/zh/bayesian-search-optimal-stopping-exploration/c03/) - Ch.4 Multi-Armed Bandits: Gittins Indices, Thompson Sampling and the Regret Lower Bound — The Gittins index breaks the multi-armed bandit into one-armed problems: each machine's index is the reward that leaves one indifferent between "continue" and "retire on a fixed reward"; the index exceeds the posterior mean, and the exploration bonus shrinks as knowledge grows and grows as the future carries more weight. The optimality of the index policy requires independent arms, frozen arms when not chosen, fixed discounting, an infinite horizon and no switching costs. In the example with discount factor 0.9, the greedy policy loses only about 0.3 % relative to the Gittins policy, while Thompson sampling and UCB1 explore too much. The Lai–Robbins bound is an asymptotic statement; at finite $T$ the regret of Thompson sampling can lie below $c\ln T$, while UCB1 lies far above it. The next chapter connects "whether to keep pulling a machine" with sunk costs. (https://5dlibrary.com/answers/en/bayesian-search-optimal-stopping-exploration/c04/) 第4章 多臂老虎机:吉廷斯指数、汤普森抽样与后悔下界 — 吉廷斯指数把多臂老虎机问题拆成单臂问题:每台机器的指数等于让人在「继续」与「拿固定收益退休」之间无差别的那个收益;指数高于后验均值,高出的探索奖励随了解增多而减小、随看重未来而增大。指数策略的最优性要求独立的臂、未被选的臂状态不变、固定折扣、无限期限、没有切换成本。在折扣因子 0.9 的示例中,贪婪策略只比吉廷斯策略少约 0.3%,而汤普森抽样与 UCB1 探索得太多。莱—罗宾斯下界是渐近结论;汤普森抽样在有限 $T$ 下的后悔可以低于 $c\ln T$,UCB1 远高于它。下一章把「是否继续拉一台机器」与沉没成本联系起来。 (https://5dlibrary.com/answers/zh/bayesian-search-optimal-stopping-exploration/c04/) - Ch.5 Sunk Costs: Rational Persistence, Fallacious Persistence and Stop-Loss Boundaries Set in Advance — The sunk-cost effect is robust in hypothetical scenarios, with a pooled meta-analytic effect size of about 0.5, but in laboratory experiments with real incentives it is small and erratic (surveys and field studies of project decisions show larger effects), and it is often entangled with closeness to completion and personal responsibility. In the normative model, money already spent does not enter the decision; the evidence accumulated and the remaining value of information do, so that continuing below break-even early on can be rational. With illustrative parameters, myopic early quitting loses about 29 %, while adding to the optimal rule the sunk-cost tendency "never quit after 10 trials" loses about 6 %. Wald's sequential probability ratio test gives stop-loss and go-ahead boundaries that can be written down in advance and needs on average only half to six-tenths of a fixed sample at the two hypotheses, but it can run long when the true value lies between them, so a cap must be set. The next chapter applies these tools to start-up experiments and to the selection of people. (https://5dlibrary.com/answers/en/bayesian-search-optimal-stopping-exploration/c05/) 第5章 沉没成本:理性的坚持、谬误的坚持与预设的止损边界 — 沉没成本效应在假想情境中稳健存在,元分析的合并效应量约为 0.5,但有真实激励的实验室实验中效应小而不稳定(项目推进类决策的调查与现场研究中效应反而较大),而且常与「接近完成」与个人责任混在一起。规范模型中,已经花掉的钱不进入决策,已经积累的证据与剩余的信息价值才进入决策;所以早期在盈亏平衡点以下继续可以是理性的。在示意参数下,短视的过早放弃损失约 29%,最优规则加上「做满 10 次不再放弃」的沉没成本倾向损失约 6%。沃尔德的序贯概率比检验给出可以事先写下的止损与推进边界,在两个假设处平均只需固定样本的五到六成,但真实值落在中间时可能很长,需要预设上限。下一章把这些工具用于创业试验与人才筛选。 (https://5dlibrary.com/answers/zh/bayesian-search-optimal-stopping-exploration/c05/) - Ch.6 Applications: Testing Lanes and Selecting People — In lane testing, "earning while learning" and "learning, then earning" are different objectives: Thompson sampling obtains the most conversions during the test, top-two Thompson sampling keeps more trials for challengers to raise the probability of choosing correctly, and successive halving is a form of early stopping that needs no probability calculation; with the illustrative parameters, differences in the probability of choosing correctly lie within simulation error, while differences in conversions are clearer. Real applications must attend to delayed outcomes, the choice of measure and the winner's curse. In selecting people, when the payoff is actual quality the best cutoff is about $\sqrt{n}$, so that with 40 applicants only 5 need be skipped; when quality is visible, its distribution known and interviews costly, the optimal rule often hires among the first few applicants. The next chapter reads these ideas beside several classical Chinese texts. (https://5dlibrary.com/answers/en/bayesian-search-optimal-stopping-exploration/c06/) 第6章 应用:赛道测试与人才筛选 — 赛道测试中,「边学边赚」与「先学后赚」是两个不同的目标:汤普森抽样在测试期间得到最多的转化,前二汤普森抽样把更多试验留给挑战者以提高选对的概率,逐轮淘汰是一种不需要概率计算的提前停止;在示意的参数下,选对概率的差别在模拟误差之内,转化数的差别更清楚。真实应用中需要注意结果延迟、指标选择与赢家的诅咒。人才筛选中,报酬是实际水平时最好的阈值约为 $\sqrt{n}$,40 人时只需跳过 5 人;看得到水平、知道分布并且面试有成本时,最优规则往往在最初几个人中就录用。下一章把这些想法与几段中国古代文献并读。 (https://5dlibrary.com/answers/zh/bayesian-search-optimal-stopping-exploration/c06/) - Ch.8 Practice and a Pre-Registered Trial — The sample size rests on the following assumptions: a smallest effect of interest of 4 percentage points (0.04); a standard deviation of each participant's share of 0.12 (the binomial sampling error over 20 games is about 0.107, plus differences between people; the model's standard deviations are 0.107 to 0.120); two-sided $\alpha = 0.05$; and power of 90 %. Each arm needs 190 analysable participants; allowing for about 15 % excluded for not finishing or failing the attention checks, each arm recruits 224, 448 in all. With a standard deviation of 0.10 or 0.14, each arm would need 132 or 258; with a smallest effect of interest of 0.03 and standard deviation 0.12, 337. All these parameters are assumptions. Before the main trial, a small pilot checks only that the task software, the length of the explanations and the attention checks work properly; it is not used to change the sample size. (https://5dlibrary.com/answers/en/bayesian-search-optimal-stopping-exploration/c08/) 第8章 实践与预注册试验 — 样本量按以下假设计算:最小关注效应为 4 个百分点(0.04);每名参与者比例的标准差取 0.12(20 局的二项抽样误差约为 0.107,加上个人之间的差异;模型中的标准差为 0.107 到 0.120),双侧 $\alpha = 0.05$,把握度 90%。每组需要 190 名可分析的参与者;考虑约 15% 的人因未完成或注意力检查失败被排除,每组招募 224 名,共 448 名。标准差为 0.10 或 0.14 时,每组分别需要 132 与 258 名;最小关注效应为 0.03 时,标准差 0.12 下需要 337 名。这些参数都是假设值。正式试验之前先做一个小规模预试验,只用来检查任务程序、讲解时长与注意力检查是否正常,不用来改变样本量。 (https://5dlibrary.com/answers/zh/bayesian-search-optimal-stopping-exploration/c08/) ## 5DL-B-043 · Embodied Cognition, Bernstein's Degrees-of-Freedom Problem and the Control Theory of Motor Synergies 中文书名:具身认知、伯恩斯坦自由度问题与运动协同的控制理论 DOI: https://doi.org/10.5281/zenodo.23828057 Cite: Fan, Y. (2026). Embodied Cognition, Bernstein's Degrees-of-Freedom Problem and the Control Theory of Motor Synergies (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23828057 Book page: https://5dlibrary.com/books/embodied-cognition-motor-synergies/ Scope: Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses; all data in the book are fictional or simulated. Abstract: The human arm has seven joint degrees of freedom; placing a fingertip at a point in space needs only three, and dozens of muscles drive those joints, so every action can be performed in endlessly many ways. How does the nervous system choose? This book examines this "Bernstein problem" with the mathematical models of motor control. It covers the degrees-of-freedom problem and "repetition without repetition" in Bernstein's 1967 English volume The Co-ordination and Regulation of Movements (secondary sources give p. 134; the page was not seen), the "principle of abundance" proposed by Gelfand and Latash in 1998 and reviewed by Latash in 2012 (a reformulation of the problem, not a measurable "degree of abundance"), and an illustrative planar three-link arm (one fingertip position corresponds to a whole curve of solutions with elbow angles from 18° to 95°; choosing a new solution every trial gives a fingertip SD of 6.5 mm, whereas joint variability of the same size, if independent, gives 121.4 mm); the uncontrolled-manifold (UCM) analysis of Scholz and Schöner 1999 (nine people standing up from a chair, variance compared per degree of freedom), the synergy index ΔV and its bounds, a log transform that is not zero at ΔV = 0, joint versus segment coordinates giving different answers (−0.94 instead of 0.00 for independent noise), and how with 20 trials 9.4 % of estimates from independent noise exceed +0.5; optimal feedback control and the "minimal intervention principle" of Todorov and Jordan 2002 (correct only deviations that interfere with the task): in a two-effector "sum" task with additive noise, at equal corrective effort, a task SD of 1.14 mm against 2.84 mm for servo control, ln(V_UCM/V_ORT) of +4.68, and a 10 mm push on one effector compensated by 4.87 mm in the other — a demonstration in one model, not a "proof"; Fitts's original log₂(2A/W) versus MacKenzie's Shannon form log₂(A/W + 1) (16 participants; about 10.3–11.5 bits/s in the best tapping conditions), movement time almost perfectly linear in √(A/W) in the optimised-submovement model of Meyer and colleagues 1988 (R² = 0.99996 against 0.908 for the log form), and minimum jerk and minimum variance as two different theories, the latter's velocity peak moving from 0.67T to 0.39T as limb stiffness rises and the book's open-loop model giving a speed–accuracy relation closer to a power law; the sensorimotor-integration internal model of Wolpert, Ghahramani and Jordan 1995 (not localised to the cerebellum; for cerebellar internal models see Miall and colleagues 1993 and Wolpert, Miall and Kawato 1998), feedback delays of the order of 100 ms that differ by modality (monosynaptic stretch reflex 10–40 ms; visual responses about 110–150 ms), a Kalman estimate with a 100 ms delay erring by 3.18 mm against 32.37 mm for the delayed reading alone, and an unknown external force causing a steady bias of about 5 mm; Blakemore and colleagues 1998 as an fMRI study of six people, the delay-and-rotation tickle psychophysics of 1999, and the 38 % mean escalation per turn of Shergill and colleagues 2003 corresponding to a self-attenuation factor of about 0.725 (3.04 N over eight turns against the reported 3.2 N); the non-negative matrix factorisation of Lee and Seung, the time-varying synergies of d'Avella and colleagues 2003 in kicking frogs, VAF thresholds that undercount synergies (a 90 % threshold picks 2 when there are 4), eight independently controlled muscles whose minimum-effort solutions need only 4 components for 95 %, 3 components reaching 95.9 % when only 5 muscles are recorded, and the biomechanical argument of Kutch and Valero-Cuevas 2012; "freezing and freeing" degrees of freedom (Vereijken and colleagues 1992) and its counter-examples (Newell and Vaillancourt 2001; Konczak and colleagues 2009), an illustrative model in which freezing is a response to internal-model error, Wulf's external-focus research, the meta-analysis of Chua and colleagues 2021 (performance g = 0.264, retention 0.583, transfer 0.584) and the bias-corrected 0.01–0.15 of McKay and colleagues (discussed in Book 10; only transfer and the design implications are added here), and a model result: an "internal" cue that constrains only the task-irrelevant direction can move ln(V_UCM/V_ORT) from +4.69 to −1.94 with the task SD unchanged (1.11 mm). The eight chapters are read beside Cook Ding carving the ox in the "Nourishing the Lord of Life" chapter of the Zhuangzi (the received text says "what I care about is the Way, which goes beyond skill"; "skill advancing into the Way" is a later summary), Wheelwright Bian in the "Way of Heaven" chapter ("I get it in my hand and respond with my mind; my mouth cannot put it into words; there is a knack in it") and Ji Chang learning archery in the "Tang Wen" chapter of the Liezi, stating that these concern nourishing life, words and meaning and the passing on of a craft, not motor-control theory. They end with a four-step embodied-synergy checklist (write down the task variable and redundancy; decompose variance per degree of freedom with intervals; check coordinates and a null model; return to the task outcome) and a pre-registered randomised controlled trial in a redundant tracking task where two joints drive a one-dimensional cursor, comparing "external task-outcome focus instructions + redundancy-manifold feedback" with "internal fixed-joint-angle instructions" (single primary outcome the RMS tracking error on a new trajectory without feedback 24 h later — transfer accuracy — by ANCOVA on pre-practice error; ln(V_UCM/V_ORT) secondary; smallest effect of interest 1.5 mm, SD 4.0 mm, correlation 0.5, 90 % power, 113 analysable and 133 recruited per arm, 266 in all; decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (14 programs, all checked), classical sources and the author's hypotheses; all data in the book are fictional or simulated. The book is not sports-medicine, physiotherapy or rehabilitation advice; anyone with an injury or a neurological condition should consult a qualified professional. 摘要:人的手臂有 7 个关节自由度,把指尖放到空间中一点只需要 3 个;几十块肌肉驱动这些关节,每次动作都可以用无数种方式完成。神经系统怎样在这么多可能性中做出选择?本书用运动控制的数学模型考察这个「伯恩斯坦问题」。依次讨论:伯恩斯坦 1967 年英文版《动作的协调与调节》中的自由度问题与「没有重复的重复」(二手资料引作第 134 页,本书未见原页),格尔方德与拉塔什 1998 年提出、拉塔什 2012 年综述的「丰裕原则」(是对问题的重新表述,不是一个可测的「丰裕度」),以及平面三连杆手臂的示意计算(同一个指尖位置对应肘角 18°—95° 的一整条解曲线;每次换一个解时指尖标准差 6.5 毫米,而同样大小的关节变异若彼此独立则为 121.4 毫米);肖尔茨与舍纳 1999 年的非受控流形(UCM)分析(原研究为 9 人的坐—站任务,方差按每个自由度比较),协同指数 ΔV 与其上下界、对数变换在 ΔV = 0 时不为零、关节坐标与节段坐标给出不同结论(独立噪声的对数比由 0.00 变为 −0.94),以及每次 20 个试次时独立噪声也有 9.4% 的估计超过 +0.5;托多罗夫与乔丹 2002 年的最优反馈控制与「最小干预原则」(只纠正妨碍任务的偏差),在加性噪声的双效应器「求和」任务中,同等纠正努力下任务标准差 1.14 毫米对伺服控制的 2.84 毫米,ln(V_UCM/V_ORT) 为 +4.68,推开一个效应器 10 毫米后另一个反向补偿 4.87 毫米——这是特定模型中的演示,不是「证明」;费茨 1954 年原式 log₂(2A/W) 与麦肯齐的香农式 log₂(A/W + 1) 的区别(16 人、敲击任务最好条件约 10.3—11.5 比特/秒),迈耶等人 1988 年的最优子运动模型中运动时间几乎完全与 √(A/W) 成线性关系(R² = 0.99996,对数式 0.908),以及最小急动度与最小方差是两种理论:后者的速度曲线随肢体刚度从峰值在 0.67T 变到 0.39T,本书开环模型的速度—精度关系更接近幂函数;沃尔珀特、加赫拉玛尼与乔丹 1995 年的感觉运动整合内部模型(没有定位到小脑;小脑内部模型见米亚尔等 1993、沃尔珀特等 1998),延迟约 100 毫秒量级且随通道而异(单突触牵张反射 10—40 毫秒,视觉约 110—150 毫秒),延迟 100 毫秒时卡尔曼估计误差 3.18 毫米对只看延迟读数的 32.37 毫米,未知外力造成约 5 毫米的稳定偏差;布莱克莫尔等 1998 年是 6 人的功能磁共振研究,延迟与旋转的挠痒心理物理见 1999 年,以及舍吉尔等 2003 年每回合平均升级 38% 对应约 0.725 的自感衰减系数(8 回合升高 3.04 牛,原文 3.2 牛);李与承的非负矩阵分解、达韦拉等 2003 年牛蛙踢腿的时变协同、VAF 阈值会低估协同数(真有 4 个时 90% 阈值只选 2 个)、8 块独立控制的肌肉在最小努力解下 4 个成分即可解释 95%、只记录 5 块肌肉时 3 个成分 95.9%,以及库奇与巴列罗—奎瓦斯 2012 年的生物力学约束论证;维赖肯等 1992 年的「冻结—释放」自由度及其反例(纽厄尔与瓦扬古 2001、孔恰克等 2009),一个把冻结看作对内部模型误差的回应的示意模型,沃尔夫的外部焦点研究、蔡等 2021 年元分析(表现 g = 0.264,保持 0.583,迁移 0.584)与麦凯等人校正报告偏倚后的 0.01—0.15(第 10 本已讨论,此处只补充迁移与试验设计的含义),以及一个模型结果:只约束任务无关方向的「内部」指令可以把 ln(V_UCM/V_ORT) 从 +4.69 降到 −1.94 而任务标准差不变(1.11 毫米)。 八章与《庄子·养生主》庖丁解牛(通行本作「臣之所好者道也,进乎技矣」,「技进乎道」是后人的概括)、《庄子·天道》轮扁斲轮「得之于手而应于心,口不能言,有数存焉于其间」以及《列子·汤问》纪昌学射并读,说明它们讲的是养生、言意与技艺传授,不是运动控制理论;最后给出一份四步具身协同核查表(写下任务变量与冗余度、按自由度分解变异并给出区间、检查坐标与零模型、回到任务结果),以及一项预注册随机对照试验方案:在两关节控制一维光标的冗余轨迹追踪任务中,比较「外部任务结果聚焦指令 + 冗余流形反馈」与「内部关节角度固定指令」(唯一主要结局为 24 小时后新轨迹上无反馈的追踪均方根误差,即迁移准确度,以协方差分析校正练习前误差;ln(V_UCM/V_ORT) 为次要结局;最小关注差异 1.5 毫米,标准差 4.0 毫米,相关 0.5,90% 把握度,每组 113 人可分析、招募 133 人,共 266 人,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(14 段程序,全部核对)、原典与作者的假说;书中所有数据都是虚构或模拟的。本书不是运动医学、物理治疗或康复建议;有伤病或神经系统疾病者应咨询专业人员。 Chapter answers: - Ch.1 Bernstein's Problem: Redundancy, Abundance and "Repetition without Repetition" — Bernstein defined the coordination of movement as the process of mastering redundant degrees of freedom. In a planar three-link arm, one spare degree of freedom is enough for a single fingertip position to correspond to a whole curve of solutions, with the elbow angle varying between 18° and 95°. The "redundancy" framing treats extra degrees of freedom as a nuisance to be eliminated, the "abundance" framing as a resource for stabilising the task; the latter is a reformulation of the problem, not a measurable quantity. "Repetition without repetition" is first a claim about practice; when it is written as a kinematic model, the key is the direction of variability: joint variability along the solution curve (SD up to 14.1°) spreads the fingertip by only 6.5 mm, whereas independent variability of the same size spreads it by 121.4 mm. The next chapter shows how this directionality can be measured in data, and what kind of controller produces it. (https://5dlibrary.com/answers/en/embodied-cognition-motor-synergies/c01/) 第1章 伯恩斯坦问题:冗余、丰裕与「没有重复的重复」 — 伯恩斯坦把动作的协调定义为掌握多余自由度的过程。在一个平面三连杆手臂中,一个多余的自由度就让同一个指尖位置对应着一整条解曲线,肘角可以在 18° 到 95° 之间变化。「冗余」的表述把多余自由度看作要消除的麻烦,「丰裕」的表述把它看作稳定任务的资源;后者是对问题的重新表述,不是一个可测量的量。「没有重复的重复」首先是关于练习的论断;把它写成运动学模型时,关键在于变异的方向:落在解曲线上的关节变异(标准差最大 14.1°)只让指尖散布 6.5 毫米,同样大小的独立变异则让指尖散布 121.4 毫米。下一章介绍怎样从数据里测出这种方向性,以及什么样的控制器会产生它。 (https://5dlibrary.com/answers/zh/embodied-cognition-motor-synergies/c01/) - Ch.2 The Uncontrolled Manifold and the Minimal Intervention Principle: Control Only What Matters — UCM analysis divides variability in joint space into a part that does not affect the task (UCM) and a part that does (ORT), compared per degree of freedom; the original study was a sit-to-stand task in nine people, in which the centre of mass was strongly stabilised. The synergy index $\Delta V$ is bounded, and a log transform of it need not be zero at $\Delta V = 0$; the conclusions depend on coordinates (the same independent noise gives −0.94 in segment coordinates), and estimates from 20 trials carry sizeable error. The optimal feedback control of Todorov and Jordan offers one explanation of this structure: the minimal intervention principle. In the two-effector sum task at equal corrective effort, the task SD is 1.14 mm under OFC and 2.84 mm under servo control, and when one effector is pushed the other compensates in the opposite direction. This is a demonstration in a particular model, not a proof. The next chapter turns to another classical problem: the trade-off between the speed and the accuracy of movement. (https://5dlibrary.com/answers/en/embodied-cognition-motor-synergies/c02/) 第2章 非受控流形与最小干预原则:只管该管的 — UCM 分析把关节空间的变异分为不影响任务(UCM)与影响任务(ORT)两部分,按每个自由度比较;原研究是 9 人的坐—站任务,质心被强烈稳定。协同指数 $\Delta V$ 有上下界,常用的对数变换在 $\Delta V = 0$ 时不一定为零;结论依赖坐标(同一组独立噪声在节段坐标下给出 −0.94),20 个试次的估计误差也不小。托多罗夫与乔丹的最优反馈控制给出了这种结构的一种解释:最小干预原则。在双效应器求和任务中,同等纠正努力下 OFC 的任务标准差为 1.14 毫米,伺服控制为 2.84 毫米;一个效应器被推开时,另一个反向补偿。这是特定模型中的演示,不是证明。下一章转向另一个经典问题:动作的速度与精度之间的权衡。 (https://5dlibrary.com/answers/zh/embodied-cognition-motor-synergies/c02/) - Ch.3 Fitts's Law and the Speed–Accuracy Trade-off: The Limits of an Information-Theoretic Analogy — Fitts's 1954 tapping experiment (16 participants, amplitudes 2–16 inches, widths 0.25–2 inches) gave $MT = a + b\log_2(2A/W)$, with an index of performance of about 10.3–11.5 bits/s in the best conditions. The Shannon form $\log_2(A/W + 1)$ is another index, promoted by MacKenzie, which differs greatly from the original at low difficulty. The logarithmic form is not the only possibility: in a Meyer-type optimised-submovement model, movement time is almost perfectly linear in $\sqrt{A/W}$ ($R^2 = 0.99996$), yet the logarithmic form still fits with $R2 = 0.908$. Minimum jerk and minimum variance are different theories: the former is a kinematic smoothness cost, the latter is derived from signal-dependent noise and its velocity profile changes with limb stiffness; in the book's simplified open-loop model, minimum variance gives a speed–accuracy relation closer to a power function. The next chapter discusses another limitation — the delay of sensory feedback — and how the nervous system gets around it with internal models. (https://5dlibrary.com/answers/en/embodied-cognition-motor-synergies/c03/) 第3章 费茨定律与速度—精度权衡:信息论类比的边界 — 费茨 1954 年的敲击实验(16 人,幅度 2—16 英寸,宽度 0.25—2 英寸)给出了 $MT = a + b\log_2(2A/W)$,最好条件下的表现指数约为 10.3—11.5 比特/秒。香农式 $\log_2(A/W + 1)$ 是麦肯齐推广的另一种指数,难度低时与原式差别很大。对数形式不是唯一的可能:在迈耶式最优子运动模型中,运动时间几乎完全与 $\sqrt{A/W}$ 成线性关系($R^2 = 0.99996$),对数式仍可拟合出 $R2 = 0.908$。最小急动度与最小方差是两种不同的理论:前者是运动学的平滑性代价,后者由信号依赖噪声推出,其速度曲线随肢体刚度改变;在本书的开环简化模型中,最小方差给出的速度—精度关系更接近幂函数。下一章讨论另一个限制:感觉反馈的延迟,以及神经系统怎样用内部模型绕过它。 (https://5dlibrary.com/answers/zh/embodied-cognition-motor-synergies/c03/) - Ch.4 Internal Forward Models and Efference Copy: Getting Around Sensory Delay — Delays in sensorimotor loops differ by modality: about 10–40 ms for the monosynaptic reflex and about 110–150 ms for visual responses, so "50–200 ms" is only a rough range. A forward model predicts the consequences of an action from the efference copy, and Kalman estimation merges the prediction with late feedback according to their uncertainties: in the illustrative model with a 100 ms delay the error falls from 32.37 mm (late reading alone) to 3.18 mm; an unknown external force leaves a steady bias of about 5 mm that only adaptation of the model itself can remove. The 1995 paper of Wolpert and colleagues concerns an internal model in sensorimotor integration; the cerebellar internal model goes back to Miall and colleagues 1993 and Wolpert and colleagues 1998. Predictive attenuation explains why you cannot tickle yourself (the fMRI of Blakemore and colleagues 1998 and the psychophysics of 1999) and the escalation of force (38 % per turn, corresponding to an attenuation factor of about 0.725). The next chapter turns from joints and senses to muscles: how muscle synergies were "discovered", and whether they really come from the nervous system. (https://5dlibrary.com/answers/en/embodied-cognition-motor-synergies/c04/) 第4章 内部正向模型与传出副本:绕过感觉延迟 — 感觉运动回路的延迟随通道而异:单突触反射约 10—40 毫秒,视觉反应约 110—150 毫秒,「50—200 毫秒」只是粗略范围。正向模型用传出副本预测动作的后果,卡尔曼估计把预测与迟到的反馈按不确定性加权合并:在示意模型中,延迟 100 毫秒时误差从 32.37 毫米(只看迟到读数)降到 3.18 毫米;未知外力造成约 5 毫米的稳定偏差,只有模型本身适应才能消除。沃尔珀特等 1995 年的论文讨论的是感觉运动整合中的内部模型,小脑内部模型的提法见米亚尔等 1993 年与沃尔珀特等 1998 年。预测性衰减解释了自己挠自己不痒(布莱克莫尔等 1998 年的功能磁共振与 1999 年的心理物理)与力的升级(每回合 38%,对应约 0.725 的衰减系数)。下一章从关节与感觉转向肌肉:肌肉协同是怎样被「发现」的,又是否真的来自神经系统。 (https://5dlibrary.com/answers/zh/embodied-cognition-motor-synergies/c04/) - Ch.5 Muscle Synergies: Non-Negative Matrix Factorisation and the Biomechanical-Constraint Dispute — The muscle-synergy hypothesis holds that movements are built by superimposing a few groups of muscles activated in fixed proportions, usually extracted from EMG by non-negative matrix factorisation; d'Avella and colleagues in 2003 analysed time-varying synergies in kicking frogs with a generalisation of NMF. The number of synergies given by a VAF threshold mainly reflects the threshold and the noise: with four true synergies, a 90 % threshold picks only two. Kutch and Valero-Cuevas showed that task and biomechanical constraints alone can produce low-dimensional EMG: eight independently controlled muscles with minimum-effort solutions need only four components for 95.0 %, and with only five muscles recorded three components reach 95.9 %. These results do not refute neural synergies, but they require constraint-and-cost explanations to be ruled out before concluding. The next chapter turns to learning: how practice changes the use of degrees of freedom, and what the direction of attention does. (https://5dlibrary.com/answers/en/embodied-cognition-motor-synergies/c05/) 第5章 肌肉协同:非负矩阵分解与生物力学约束之争 — 肌肉协同假说认为动作由少数几组按固定比例激活的肌肉叠加而成,常用非负矩阵分解从肌电中提取;达韦拉等人 2003 年分析的是蛙踢腿的时变协同,用的是 NMF 的推广算法。VAF 阈值给出的协同数主要反映阈值与噪声:真有 4 个协同时,90% 阈值只选出 2 个。库奇与巴列罗—奎瓦斯说明,任务与生物力学约束本身就能产生低维肌电:8 块独立控制的肌肉在最小努力解下,4 个成分解释 95.0%;只记录 5 块肌肉时 3 个成分解释 95.9%。这些结果不否定神经协同,但要求在下结论之前排除约束与代价的解释。下一章转向学习:练习怎样改变自由度的使用,注意的方向又有什么影响。 (https://5dlibrary.com/answers/zh/embodied-cognition-motor-synergies/c05/) - Ch.6 Skill Learning: Freezing and Freeing Degrees of Freedom, External Focus and "Instructions as Costs" — "Freezing and freeing" is the learning process described by Vereijken and colleagues on a ski simulator in 1992, but it is not a general law: dimension can rise or fall with practice, and violin learners with more practice show less shoulder motion, not more. In an illustrative model, freezing (less task correction, more posture stiffness) is a reasonable response to internal-model error, and as the model error shrinks $\ln(V_{\mathrm{UCM}}/V_{\mathrm{ORT}})$ rises from +0.78 to +1.47. The meta-analytic external-focus effect (0.584 for transfer) falls below about 0.1 after correction for reporting bias. Treating instructions as cost functions, an "internal" instruction that constrains only the task-irrelevant direction can move the log ratio from +4.69 to −1.94 with the task SD unchanged; only an imposed stiffness that also acts on the task direction makes the task worse. The next chapter reads these studies beside ancient texts on skill. (https://5dlibrary.com/answers/en/embodied-cognition-motor-synergies/c06/) 第6章 技能学习:冻结与释放自由度、外部焦点与「指令即代价」 — 「冻结—释放」是维赖肯等人 1992 年在滑雪模拟器上描述的学习过程,但不是普遍规律:维数可以随练习增加也可以减少,练习量较多的小提琴学习者肩部活动反而较小。在一个示意模型中,冻结(少用任务纠正、多用姿势刚度)是对内部模型误差的合理回应,模型误差减小时 $\ln(V_{\mathrm{UCM}}/V_{\mathrm{ORT}})$ 从 +0.78 升到 +1.47。外部焦点的元分析效应(迁移 0.584)在校正报告偏倚后降到约 0.1 以下。把指令看作代价函数时,只约束任务无关方向的「内部」指令可以让对数比从 +4.69 降到 −1.94 而任务标准差不变;强加同时影响任务方向的刚度才使任务变差。下一章把这些研究与古代关于技艺的文本并读。 (https://5dlibrary.com/answers/zh/embodied-cognition-motor-synergies/c06/) - Ch.8 Practice and a Pre-Registered Trial — The sample size rests on the following assumptions (all assumed values, not estimates from any study): an SD of transfer error of 4.0 mm; a correlation of 0.5 between pre-practice error and transfer error (ANCOVA multiplies the variance by $1 - 0.5^2$); a smallest effect of interest of 1.5 mm (about 0.375 SD), two-sided $\alpha = 0.05$ and 90 % power. Each arm needs 113 analysable participants; allowing for about 15 % missing data, 133 are recruited per arm, 266 in all. The 80 % power alternative is 84 analysable per arm. Sensitivity: with an SD of 3.5–4.5 mm and a correlation of 0.3–0.6, 74 to 173 per arm are needed; with a smallest effect of interest of 1.0 mm, 165 to 388. ANCOVA on 2,000 simulated trials gives a power of 0.896, in agreement with the formula; with no true difference, "intervention better" is wrongly concluded in 2.4 %. (https://5dlibrary.com/answers/en/embodied-cognition-motor-synergies/c08/) 第8章 实践与预注册试验 — 样本量按以下假设计算(都是假设值,不是任何研究的估计):迁移误差的标准差 4.0 毫米;练习前误差与迁移误差的相关 0.5(协方差分析把方差乘以 $1 - 0.5^2$);最小关注差异 1.5 毫米(约 0.375 个标准差),双侧 $\alpha = 0.05$,把握度 90%。每组需要 113 名可分析的参与者;考虑约 15% 的缺失,每组招募 133 名,共 266 名。80% 把握度的备选方案为每组 84 名可分析。敏感性:标准差 3.5—4.5 毫米、相关 0.3—0.6 时,每组需要 74 到 173 名;最小关注差异为 1.0 毫米时需要 165 到 388 名。2,000 次模拟的协方差分析给出把握度 0.896,与公式一致;真实差异为 0 时,错误地得出「干预更好」的比例为 2.4%。 (https://5dlibrary.com/answers/zh/embodied-cognition-motor-synergies/c08/) ## 5DL-B-044 · Collective Dynamics, Coupled-Oscillator Synchrony and Distributed Consensus 中文书名:集群动力学、耦合振子同步与分布式共识的数学模型 DOI: https://doi.org/10.5281/zenodo.23828063 Cite: Fan, Y. (2026). Collective Dynamics, Coupled-Oscillator Synchrony and Distributed Consensus (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23828063 Book page: https://5dlibrary.com/books/collective-dynamics-consensus-synchrony/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses; all data in the book are fictional or simulated. Abstract: Why does applause fall into step and then fall apart? Why can a footbridge be set swaying by the people walking across it? Why do many individuals who interact only with neighbours sometimes move as one and sometimes split into pieces? This book uses statistical physics, dynamical systems and network science to examine three families of models of collective dynamics — coupled-oscillator synchronisation, the collective motion of self-propelled particles, and distributed consensus — stating for each conclusion the assumptions under which it holds. It covers the complex order parameter and self-consistency equation of the Kuramoto model; the critical coupling K_c = 2/(π g(0)), valid only for global coupling, N → ∞ and a symmetric unimodal frequency distribution; the exact r = √(1 − K_c/K) for the Lorentzian, a continuous transition with a square-root law for the Gaussian (which already overestimates r by about 10 % at K/K_c = 1.10), and a first-order jump to π/4 for the uniform distribution; the smearing of the threshold at finite size (mean r of 0.552 for 50 oscillators at K = K_c); the applause studies of Néda and colleagues, in which doubling the period shrinks the frequency dispersion so that K_c falls and synchrony appears, while synchrony halves the average loudness (69.7 to 34.7 claps per second in the simulation); the synchronisation model of the London Millennium Bridge by Strogatz and colleagues, for which the book derives the critical crowd N_c = 2BΩ0/(πGCP(Ω0)) — "negative damping per walker times the number of walkers equals the bridge's damping" — 149 walkers with illustrative parameters against observed values of about 155 to 165 in different sources, with simulations showing amplitude and synchrony growing together, and the 2021 argument of Belykh and colleagues that synchrony is a result of the instability, not its cause; the master stability function and the eigenratio criterion λ_N/λ_2 (which follows directly from the master stability function of Pecora and Carroll 1998 and was used by Barahona and Pecora in 2002 as an index of network synchronisability, valid only for bounded stable regions), with an exact coupled-logistic-map example requiring a ratio below 3 in which a star synchronises under normalised coupling; the noise scan of the Vicsek model and the discontinuous transition of Grégoire and Chaté; the four collective states of Couzin's three-zone model (swarm, torus, dynamic parallel, highly parallel) and its hysteretic "collective memory", reproduced in a two-dimensional simplification; Couzin and colleagues' 2005 finding that the larger the group, the smaller the share of informed individuals needed (accuracy depends mainly on their number in a simplified model), and the 2011 fish experiments in which untrained individuals return control to the numerical majority (an effect that depends on the interaction rules: the uninformed have no effect in the book's copy-one-other model, and in the model of Hartnett and colleagues a highly nonlinear response reverses its direction); the influence vector, spectral gap and rounds to convergence of DeGroot consensus, the theorem of Golub and Jackson that a crowd is wise if and only if the largest influence tends to 0, a star whose error stays at 1/3 and transmits the centre's bias in proportion to influence, and the 19 % rise in error in the experiment of Becker and colleagues when the central person deviated from the truth; the consensus threshold of about 0.2 in the Hegselmann–Krause model, which holds only for large populations (all runs reach consensus at ε = 0.21 with 5,000 people, 3 % with 50), the roughly ⌊1/(2ε)⌋ major clusters of the Deffuant model with outliers left at the ends, and the gap between the "echo chamber" analogy and the evidence of field experiments. The eight chapters are read beside "the gentleman harmonises but does not conform; the petty man conforms but does not harmonise" from the Analects ("Zilu"), Shi Bo's "harmony truly generates things, sameness does not continue; to balance one thing with another is called harmony" from the Discourses of the States ("Discourses of Zheng"), Yanzi's "harmony is like a broth" from the Zuo Tradition (Duke Zhao, year 20) and "music makes for sameness, ritual makes for difference" from the "Record of Music" in the Book of Rites, stating the boundary between ancient ethical and political discussion and modern models (the "difference" of the "Record of Music" is rank) and never claiming that the ancients "already discovered" synchronisation or consensus theory. They end with a four-step checklist against echo chambers and for calibrating synchrony, and a pre-registered cluster-randomised controlled trial comparing a "centralised hub topology" with an "equal-degree decentralised regular ring" when a core seat (a programmed seat disclosed to participants, showing twice the true value) carries a bias in repeated numerical estimation; the single primary outcome is the absolute error of the group's final mean log estimate; a model shows that with a smaller bias the two arms hardly differ; smallest effect of interest 0.02, 90 % power, 49 analysable and 58 randomised groups per arm (2,204 participants), decision rule written as code. Marked boxes separate published evidence, runnable Python calculations (16 programs, all checked), classical sources and the author's hypotheses; all data in the book are fictional or simulated. The book is a teaching and research text on mathematical models, not management consulting, and makes no political claim of any kind. 摘要:掌声为什么时齐时散?一座人行桥为什么会被人群走得左右摇晃?许多只与邻居交互的个体,为什么有时整齐划一,有时分裂成几块?本书用统计物理、动力系统与网络科学考察三类集体动力学模型:耦合振子同步、自驱动粒子的集体运动与分布式共识,并逐条说明它们的结论在什么假设下成立。依次讨论:藏本(库拉莫托)模型的复序参量与自洽方程,临界耦合 K_c = 2/(π g(0)) 只在全局耦合、N → ∞、对称单峰频率分布时成立,洛伦兹分布下 r = √(1 − K_c/K) 精确成立,高斯分布为平方根律的连续相变(K/K_c = 1.10 时平方根律已高估约 10%),均匀分布则在 K_c 处跳到 π/4,是一级相变;有限人数时阈值被涨落抹平(K = K_c 时 50 人的平均 r 为 0.552);内达等人的掌声研究中,周期加倍使频率离散度变小、K_c 下降才出现同步,同步又使平均声响减半(模拟中每秒拍手 69.7 次降到 34.7 次);伦敦千禧桥的斯特罗加茨同步模型,本书推导出临界人数 N_c = 2BΩ0/(πGCP(Ω0)),即「人均负阻尼乘以人数等于桥的阻尼」,示意参数下为 149 人,观测值在不同文献中为约 155 至 165 人,模拟显示振幅与行人同步一起增长,别雷赫等人 2021 年认为同步是失稳的结果而非原因;主稳定函数与特征值比 λ_N/λ_2 判据(由佩科拉—卡罗尔 1998 年的主稳定函数直接得到,巴拉奥纳与佩科拉 2002 年用作网络可同步性指标,只在稳定区有界时适用),耦合逻辑斯谛映射的精确例子中比值须小于 3,星形网络在归一化耦合下可以同步;维切克模型的噪声扫描与格雷瓜尔—沙泰的不连续相变,库赞三区模型的四种集体状态(群集、环面、动态平行、高度平行)与「集体记忆」滞后(二维简化版复现);库赞等人 2005 年「群体越大所需知情者比例越小」(简化模型中准确度主要取决于知情者人数)与 2011 年鱼群实验中未训练个体把控制权交还数量上的多数(这一效应取决于交互规则:本书逐一模仿随机他人的模型中未知情者不起作用,哈特内特等人的模型中反应高度非线性时效应方向反而相反);德格鲁特共识的影响力向量、谱隙与收敛轮数,戈卢布与杰克逊「群体明智当且仅当最大影响力趋于 0」,星形网络的误差停在 1/3、中心人物的偏差按影响力传递,以及贝克尔等人实验中中心人物偏离真值时误差上升 19%;黑格泽尔曼—克劳泽模型的共识阈值约 0.2 只在人数很多时成立(5,000 人时 ε = 0.21 全部共识,50 人时只有 3%),德富昂模型的主要簇数约为 ⌊1/(2ε)⌋ 并常留下两端的离群者,以及「回音室」类比与现场实验证据的差距。 八章与《论语·子路》「君子和而不同,小人同而不和」、《国语·郑语》史伯「夫和实生物,同则不继。以他平他谓之和」、《左传·昭公二十年》晏子「和如羹焉」与《礼记·乐记》「乐者为同,礼者为异」并读,明确古代伦理与治理论述同现代模型的界限(《乐记》的「异」是贵贱等级),不说古人「早已发现」同步或共识理论;最后给出一份四步防回音室与同步校准表,以及一项预注册的整群随机对照试验方案:比较「中心化辐射拓扑」与「等度去中心化正则环网」在核心席位(事先告知参与者的程序席位,显示真值的 2 倍)带偏差时对多轮数值估计最终误差的影响,唯一主要结局为小组最终平均对数估计的绝对误差,模型显示偏差较小时两者几乎无差别,最小关注效应 0.02,把握度 90%,每组 49 个可分析小组、随机化 58 个(共 2,204 人),判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序,全部核对)、原典与作者的假说;书中所有数据都是虚构或模拟的。本书是教学与研究用的数学模型,不是管理咨询,也不构成任何政治主张。 Chapter answers: - Ch.1 The Kuramoto Model: the Synchronisation Transition of Coupled Oscillators, and Applause — The Kuramoto model uses the complex order parameter to turn an $N$-body problem into a mean-field problem and, under global coupling, $N \to \infty$ and a symmetric unimodal frequency distribution, gives the critical coupling $K_c = 2/(\pi g(0))$. For the Lorentzian, $r = \sqrt{1 - K_c/K}$ holds exactly; general smooth distributions give a continuous transition with a square-root law near threshold; the uniform distribution gives a first-order transition in which $r$ jumps to $\pi/4$. At finite size the threshold is smeared by fluctuations. The key to synchronised applause is that doubling the period shrinks the frequency dispersion, while synchrony lowers the average loudness, the audience speeds up, and synchrony collapses. The next chapter turns to a risk that synchrony can bring: the London Millennium Bridge. (https://5dlibrary.com/answers/en/collective-dynamics-consensus-synchrony/c01/) 第1章 藏本(库拉莫托)模型:耦合振子的同步相变与掌声 — 藏本(库拉莫托)模型用复序参量把 $N$ 体问题化为平均场问题,在全局耦合、$N \to \infty$、对称单峰频率分布的条件下给出临界耦合 $K_c = 2/(\pi g(0))$。洛伦兹分布时 $r = \sqrt{1 - K_c/K}$ 精确成立;一般光滑分布在阈值附近为平方根律的连续相变;均匀分布为一级相变,$r$ 跳到 $\pi/4$。有限人数时阈值被涨落抹平。掌声同步的关键是周期加倍使频率离散度变小,同步又使平均声响下降,观众加快拍手后同步瓦解。下一章讨论同步可能带来的风险:伦敦千禧桥。 (https://5dlibrary.com/answers/zh/collective-dynamics-consensus-synchrony/c01/) - Ch.2 The Wobbling Millennium Bridge and the Master Stability Function of Network Synchrony — The sway of the Millennium Bridge can be described by the synchronisation model of Strogatz and colleagues, whose critical crowd $N_c = 2B\Omega_0/(\pi GCP(\Omega_0))$ is equivalent to "negative damping per walker times the number of walkers equals the bridge's damping"; the observed critical number is reported as 155 to 165 in different sources. Both the model and later research show that synchrony and sway grow together, and that the onset of the sway does not need everyone to be in step; the 2021 study argues that synchrony is a result of the instability. The master stability function separates dynamics from network structure; the criterion $\lambda_N/\lambda_2$ follows directly from the method of Pecora and Carroll in 1998 and was used by Barahona and Pecora in 2002 as an index of network synchronisability; it applies only when the stable region is bounded, and its verdict depends on whether coupling is normalised. The next chapter turns from oscillators to individuals that move. (https://5dlibrary.com/answers/en/collective-dynamics-consensus-synchrony/c02/) 第2章 千禧桥的摇晃与网络同步的主稳定函数 — 千禧桥的晃动可以用斯特罗加茨等人的同步模型描述,其临界人数 $N_c = 2B\Omega_0/(\pi GCP(\Omega_0))$ 等价于「人均负阻尼乘以人数等于桥的阻尼」;不同文献转述的观测临界人数在 155 到 165 之间。模型与后来的研究都表明,同步与晃动一起增长,晃动之初并不需要全体同步;2021 年的研究认为同步是失稳的结果。网络同步的主稳定函数方法把动力学与网络结构分开;$\lambda_N/\lambda_2$ 判据由佩科拉与卡罗尔 1998 年的方法直接得到,巴拉奥纳与佩科拉 2002 年把它用作网络可同步性的指标,只在稳定区有界时适用,结果还取决于耦合是否归一化。下一章从振子转向会移动的个体。 (https://5dlibrary.com/answers/zh/collective-dynamics-consensus-synchrony/c02/) - Ch.3 Self-Propelled Particles and the Three-Zone Model: Noise, Alignment and Collective Memory — At low noise both sizes are almost perfectly ordered ($v_a \approx 0.99$ at $\eta = 0.5$); as the noise grows the order falls gradually, to 0.326 at $\eta = 4.0$ and 0.093 at $\eta = 6.0$ for $N = 100$. The curve of the small system is flatter: at $\eta = 5.0$, $v_a$ is 0.233 for $N = 25$ against 0.127 for $N = 100$, partly because of finite-size fluctuations ($N^{-1/2}$ is 0.20 and 0.10 respectively). Systems this small show a smooth crossover, which does not contradict the discontinuous transition Grégoire and Chaté found in large systems — the discontinuity appears only at large enough scales. (https://5dlibrary.com/answers/en/collective-dynamics-consensus-synchrony/c03/) 第3章 自驱动粒子与三区模型:噪声、对齐与集体记忆 — 噪声小时两种规模都几乎完全有序($\eta = 0.5$ 时 $v_a$ 约为 0.99);噪声增大,有序度逐渐下降,$N = 100$ 时 $\eta = 4.0$ 为 0.326,$\eta = 6.0$ 为 0.093。小系统的曲线更平缓:$\eta = 5.0$ 时 $N = 25$ 的 $v_a$ 为 0.233,$N = 100$ 为 0.127,这部分来自有限人数的涨落($N^{-1/2}$ 分别为 0.20 与 0.10)。在这样小的系统中看到的是平滑过渡,与格雷瓜尔与沙泰在大系统中发现的不连续转变并不矛盾——不连续性只有在足够大的尺度上才显现。 (https://5dlibrary.com/answers/zh/collective-dynamics-consensus-synchrony/c03/) - Ch.4 Leadership without Signals: Informed Minorities, Uninformed Individuals and Group Consensus — A few informed individuals can guide a group without signals; the conclusion of Couzin and colleagues in 2005 is that the larger the group, the smaller the proportion of informed individuals needed, but there is no universal proportion. The book's simplified model shows that accuracy depends mainly on the number of informed individuals, and also exposes its limitation: it can compromise but not choose. The fish experiments of Couzin and colleagues in 2011 show that untrained individuals can weaken the control of a strongly opinionated minority and return the choice to the numerical majority; in the book's model in which individuals copy one random other, the uninformed have no effect at all, and the effect appears only under a "majority of three" rule, while the model of Hartnett and colleagues in 2016 shows that with highly nonlinear responses the direction of the effect is reversed — whether the effect appears, and in which direction, depends on the specific form of the interaction rules. The next chapter turns to the averaging of opinions: the DeGroot model. (https://5dlibrary.com/answers/en/collective-dynamics-consensus-synchrony/c04/) 第4章 没有信号的领导:知情少数、未知情者与群体共识 — 少数知情者可以在没有信号的情况下引导群体,库赞等人 2005 年的结论是:群体越大,所需知情者的比例越小,但没有普适的比例;本书的简化模型显示准确度主要取决于知情者人数,同时也暴露出它只能折中、不能选择的局限。库赞等人 2011 年的鱼群实验表明,未训练个体可以削弱偏好强烈的少数的控制,把选择交还数量上的多数;在本书逐一模仿随机他人的模型中未知情者完全不起作用,改用「三人取多数」的规则后才出现这一效应,而哈特内特等人 2016 年的模型显示,反应高度非线性时效应的方向反而相反——效应的有无与方向都取决于交互规则的具体形式。下一章转向意见的平均:德格鲁特模型。 (https://5dlibrary.com/answers/zh/collective-dynamics-consensus-synchrony/c04/) - Ch.5 DeGroot Consensus: Influence, Spectral Gap and the Wisdom of Crowds — In the DeGroot model a strongly connected aperiodic network always reaches consensus, and the consensus is the influence-weighted average $\pi$ of initial opinions; disagreement decays like $|\lambda_2|^t$, and the spectral gap describes only the asymptotic rate. Golub and Jackson proved that a crowd is wise if and only if the largest influence tends to 0 as the population grows, and that speed and wisdom are unrelated. Stars and "elite" networks converge fast but are not wise, and the central person's bias is passed to the whole group in proportion to influence. Experiments show that most people update close to DeGroot averaging; whether social influence improves estimates in centralised networks depends on whether the central person deviates from the truth. The next chapter asks what happens when people listen only to opinions close to their own. (https://5dlibrary.com/answers/en/collective-dynamics-consensus-synchrony/c05/) 第5章 德格鲁特共识:影响力、谱隙与群体智慧 — 德格鲁特模型中,强连通且非周期的网络必然达成共识,共识是初始意见按影响力 $\pi$ 的加权平均;分歧按 $|\lambda_2|^t$ 衰减,谱隙只描述渐近速率。戈卢布与杰克逊证明,群体明智当且仅当最大影响力随人数趋于 0;收敛快慢与是否明智无关。星形与「精英」网络收敛快却不明智,中心人物的偏差按其影响力传递给整个群体。实验表明多数人的更新接近德格鲁特式平均;中心化网络中的社会影响能否改善估计,取决于中心人物是否偏离真值。下一章讨论人们只听取与自己相近的意见时会发生什么。 (https://5dlibrary.com/answers/zh/collective-dynamics-consensus-synchrony/c05/) - Ch.6 Bounded Confidence: from Consensus to Clusters, and the Evidence on "Echo Chambers" — In bounded-confidence models opinions are influenced only by nearby opinions. The Hegselmann–Krause model passes from clusters to consensus at about $\varepsilon \approx 0.2$ in large populations; in small populations the threshold is blurred, convergence near the threshold is slow, and the share reaching consensus need not be monotone. The Deffuant model gives about $\lfloor 1/(2\varepsilon)\rfloor$ major clusters, often leaves outliers at the ends, and always reaches consensus for $\varepsilon \ge 0.5$. "Polarisation" in the models is just two clusters, not necessarily at the extremes; "echo chamber" is an analogy, and field experiments do not support the simple inference that more exposure to opposing views reduces polarisation. The next chapter reads this research beside ancient discussions of "harmony" and "sameness". (https://5dlibrary.com/answers/en/collective-dynamics-consensus-synchrony/c06/) 第6章 有界信任:从共识到分簇,以及「回音室」的证据 — 有界信任模型中,意见只受相近意见的影响。黑格泽尔曼—克劳泽模型在人数很多时约在 $\varepsilon \approx 0.2$ 处由分簇转为共识,人数少时阈值模糊,阈值附近收敛很慢,共识比例也不一定单调;德富昂模型的主要簇数约为 $\lfloor 1/(2\varepsilon)\rfloor$,两端常留下离群者,$\varepsilon \ge 0.5$ 时必定共识。模型中的「两极化」只是两个簇,不一定在两端;「回音室」是类比,现场实验并不支持「多接触对立观点就能减少极化」的简单推论。下一章把这些研究与古代关于「和」与「同」的论述并读。 (https://5dlibrary.com/answers/zh/collective-dynamics-consensus-synchrony/c06/) - Ch.8 Practice and a Pre-Registered Trial — Step 1: in the network before, the convener's influence is 0.214 (the average is 0.083) and the effective number 9.8; after, influence is perfectly equal and the effective number 12.0. Step 2: if the convener has a bias of +15, the root-mean-square error of the consensus is 7.21 before and 5.91 after, the same as an equally weighted mean. The second half of step 3: both networks have 6 links between the camps, but before, all of them run through the convener (share 1.00), while after, at most a third run through any one person. Step 4: the network before converges faster — after 5 rounds the disagreement falls from 61 to 10.0, against 12.3 after; the consensus values are 100.6 and 102.0. The network before reached, faster, a conclusion that depends more on the convener — an example of Chapter 5's "speed and accuracy are unrelated". (https://5dlibrary.com/answers/en/collective-dynamics-consensus-synchrony/c08/) 第8章 实践与预注册试验 — 样本量按以下假设计算:最小关注效应为 0.02(模型中去中心化组的误差约为 0.12,0.02 相当于相对增加约 17%,与贝克尔等人报告的中心人物偏离真值时误差增加 19% 同一量级);小组结局的标准差取 0.030(模型中为 0.028 至 0.030);双侧 $\alpha = 0.05$,把握度 90%,按 $t$ 分布加 1。每组需要 49 个可分析的小组;考虑约 15% 的小组因参与者流失(第 3 轮完成者少于 15 人)而不可分析,每组随机化 58 个,共 116 组、2,204 名参与者。标准差为 0.025 或 0.040 时,每组分别需要 34 或 86 个小组;最小关注效应为 0.03 时,标准差 0.030 下需要 23 个。如需较小规模的备选方案,80% 把握度下每组需要 37 个可分析的小组,应在登记时一并写明。正式试验之前先做一次预试验,只用来检查程序、界面与题目,不用来改变样本量。 (https://5dlibrary.com/answers/zh/collective-dynamics-consensus-synchrony/c08/) ## 5DL-B-045 · High-Frequency Limit Order Book Microstructure, Hawkes Self-Exciting Point Processes and Adverse-Selection Flow Toxicity 中文书名:高频限价订单簿微观结构、霍克斯自激点过程与逆向选择毒流 DOI: https://doi.org/10.5281/zenodo.23828065 Cite: Fan, Y. (2026). High-Frequency Limit Order Book Microstructure, Hawkes Self-Exciting Point Processes and Adverse-Selection Flow Toxicity (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23828065 Book page: https://5dlibrary.com/books/order-book-microstructure-hawkes-toxicity/ Scope: Why does a market maker quote a spread? Why do orders arrive in clusters? Can "toxic" order flow be measured from trade data? This book presents mathematical models for teaching and research and is not investment or trading advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. Abstract: Why does a market maker quote a spread? Why do orders arrive in clusters? Can "toxic" order flow be measured from trade data? This book presents mathematical models for teaching and research and is not investment or trading advice. It covers the Glosten–Milgrom sequential-trade model (the "Grossman" of the author's framework is corrected; with a prior of one-half the spread equals the informed share times the value difference, and a dealer who underestimates the informed share loses systematically) and Kyle's sequential and continuous auctions (computed from the difference equations of his Theorem 2: the continuous-auction price impact is twice the one-period value, 0.9963 with 1,024 auctions, and information enters the price linearly); the Avellaneda–Stoikov reservation price s − qγσ²(T−t) and approximately optimal spread, compared with the exact ODE solution of Guéant, Lehalle and Fernandez-Tapia (with the paper's parameters and one unit of time remaining, the approximation puts the reservation-price shift at inventory 1 at −0.400 against −0.052 exactly; re-running the paper's simulation with γ = 1, quoting by the exact solution earns 63.9 on average against 31.2); the cluster representation of Hawkes processes, the stationarity condition α/β < 1, Fano factors and time-rescaling tests, the closed-form signature plot of mutually exciting processes and the Epps effect, and the Hayashi–Yoshida correction for asynchronous trading; the sampling error and floating-point overflow of PIN estimation, VPIN and the 2010 flash-crash dispute (bulk classification can read a volatility shock with no informed trading as toxic flow), and spread decomposition with a threshold at which the market shuts down; the difference between order flow imbalance (an average contemporaneous R² of 65 % in the CKS evidence) and queue imbalance (non-linearly related to the direction of the next move, depending on absolute queue sizes), and the adverse selection of queue position; and the Almgren–Chriss efficient frontier, with VWAP optimal only when temporary impact is inversely proportional to volume. Chapter 7 reads these beside the market director and price officers of the Rites of Zhou, the "conspicuous mound" of the Mencius (which reads 古之为市也) and the market-exchange law in the History of Song; the records of brokers or the market-exchange office "quoting and matching" in the Dream Pool Essays mentioned in the author's framework could not be found and are not cited. Chapter 8 gives a four-step checklist (adverse selection, order flow, inventory, execution) and a pre-registered simulated-matching experiment comparing the cross-session Sharpe ratios of inventory-skewed Avellaneda–Stoikov quotes and fixed symmetric quotes under toxic pulses on a simulator frozen by an independent third party (single primary outcome their difference, smallest effect of interest 0.31, 8,600 paired sessions, double-blind running, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. 摘要:做市商为什么要留出买卖价差?订单为什么成簇到来?能否从成交数据中度量「有毒」的订单流?本书是教学与研究用的数学模型,不是投资或交易建议。它依次讨论:格罗斯滕—米尔格罗姆序贯交易模型(作者框架中的「格罗斯曼」已改正;在先验 1/2 时价差等于知情比例乘以价值之差,做市商低估知情比例会系统性亏损)与凯尔的序贯及连续拍卖(用定理 2 的差分方程逐次计算:连续拍卖的价格冲击系数为单期的两倍,1,024 次拍卖时为 0.9963,信息线性进入价格);阿韦拉内达—斯托伊科夫的保留价格 s − qγσ²(T−t) 与近似最优价差,并与格扬、勒阿勒与费尔南德斯—塔皮亚的精确常微分方程解对照(在原文参数下,剩余时间为 1 时近似公式把库存为 1 的保留价格偏移估为 −0.400,精确解为 −0.052;重做原文模拟,γ = 1 时按精确解报价的平均利润为 63.9,近似公式为 31.2);霍克斯自激过程的分支表示、平稳性判据 α/β < 1、法诺因子与时间重标度检验,互激过程的签名图闭式解与埃普斯效应,以及林—吉田估计量对非同步交易偏差的修正;知情交易概率 PIN 的估计误差与浮点溢出问题、VPIN 与 2010 年闪崩之争(批量分类可能把不含知情交易的波动率冲击读成订单流毒性),以及价差分解与市场关闭的门槛;订单流不平衡(CKS 实证中同期平均 R² 为 65%)与挂单量不平衡(与下一次变动方向的非线性关系,依赖队列绝对大小)的区别与排队位置的逆向选择;阿尔姆格伦—克里斯有效前沿,以及 VWAP 只在临时冲击与成交量成反比时才最优。 第七章与《周礼·地官》的司市与贾师、《孟子·公孙丑下》的「龙断」(原文为「古之为市也」)以及《宋史·食货志下八》的市易法并读;作者框架所称《梦溪笔谈》中牙人或市易司「报价撮合」的记载未能找到,未予引用。第八章给出逆向选择、订单流、库存与执行四步核查表,以及一项预注册的模拟撮合实验方案:在独立第三方冻结的模拟器中,比较库存偏置的阿韦拉内达—斯托伊科夫报价与固定对称报价在毒性脉冲下的跨会话夏普比率(唯一主要结局为二者之差,最小关注效应 0.31,8,600 个配对会话,双盲运行,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有数据都是虚构或模拟的。 Chapter answers: - Ch.1 Information Asymmetry and the Bid–Ask Spread: Glosten–Milgrom and Kyle — Part of the spread is the cost of trading with the informed. In the Glosten–Milgrom model the spread comes entirely from adverse selection; with a prior of one-half it equals the informed share times the difference in values, and it narrows as trading reveals information. The dealer's gains from noise traders exactly offset her losses to the informed, and underestimating the informed share brings systematic losses. In Kyle's model the insider hides his orders in the noise; the price-impact coefficient of the continuous auction, $\sqrt{\Sigma_0}/\sigma_u$, is twice the one-period value, and information enters the price linearly. The name in the author's framework has been corrected to Glosten. (https://5dlibrary.com/answers/en/order-book-microstructure-hawkes-toxicity/c01/) 第1章 信息不对称与买卖价差:格罗斯滕—米尔格罗姆与凯尔 — 价差的一部分是与知情者交易的代价。格罗斯滕—米尔格罗姆模型中,价差完全来自逆向选择,在先验 1/2 时等于知情比例乘以价值的差,并随交易中信息的揭示而收窄;做市商对噪声交易者的盈利恰好抵消对知情者的亏损,低估知情比例则带来系统性亏损。凯尔模型中,知情交易者把订单藏在噪声里;连续拍卖的价格冲击系数 $\sqrt{\Sigma_0}/\sigma_u$ 是单期的两倍,信息线性地进入价格。作者框架中的人名已改正为格罗斯滕。 (https://5dlibrary.com/answers/zh/order-book-microstructure-hawkes-toxicity/c01/) - Ch.2 Inventory Risk and Optimal Market Making: the Avellaneda–Stoikov Equations — The Avellaneda–Stoikov reservation price $s - q\gamma\sigma^2(T-t)$ and total-spread formula check out, but the spread formula depends on exponential arrival intensities, an expansion in inventory and a first-order approximation of the order-arrival term. The exact solution of Guéant and co-authors shows that optimal quotes hardly depend on time away from the horizon; with the paper's parameters and one unit of time remaining, the approximation already overstates the reservation-price shift at inventory 1 about eightfold (−0.400 against −0.052). Re-running the paper's simulation gives results close to the paper's; quoting by the exact solution raises mean profit at $\gamma = 1$ from 31.2 to 63.9. There is no adverse selection in the model, which is the starting point of the experiment designed in Chapter 8. (https://5dlibrary.com/answers/en/order-book-microstructure-hawkes-toxicity/c02/) 第2章 库存风险与最优做市:阿韦拉内达—斯托伊科夫方程 — 阿韦拉内达—斯托伊科夫的保留价格 $s - q\gamma\sigma^2(T-t)$ 与总价差公式核对无误,但价差公式依赖指数到达强度、按库存的展开与对订单到达项的一阶近似。格扬等人的精确解显示,最优报价在远离终点时几乎与时间无关;在原文参数下,剩余时间为 1 时近似公式已把库存为 1 的保留价格偏移高估约八倍(−0.400 对 −0.052)。重做原文模拟得到与原文接近的结果;按精确解报价时,$\gamma = 1$ 的平均利润从 31.2 升到 63.9。模型中没有逆向选择,这是第八章实验设计的出发点。 (https://5dlibrary.com/answers/zh/order-book-microstructure-hawkes-toxicity/c02/) - Ch.3 Why Orders Arrive in Clusters: Hawkes Processes, the Signature Plot and the Epps Effect — Hawkes processes describe order clustering by letting past events raise the current intensity; the branching representation gives the mean intensity $\mu/(1-\eta)$, the cluster size $1/(1-\eta)$ and the long-window Fano factor $1/(1-\eta)^2$, and stationarity requires $\eta = \alpha/\beta < 1$, as the framework states. Time-rescaling tests distinguish Poisson and Hawkes models clearly. Whether real markets have a branching ratio near 1 remains disputed (recapped from Book 39). Mutually exciting processes reproduce both the signature plot and the Epps effect; but the Epps effect can also come entirely from asynchronous trading, in which case the Hayashi–Yoshida estimator removes it. "Hawkes processes as the explanation of the Epps effect" in the author's framework should be qualified to "Hawkes processes as one explanation". (https://5dlibrary.com/answers/en/order-book-microstructure-hawkes-toxicity/c03/) 第3章 订单为什么成簇到来:霍克斯过程、签名图与埃普斯效应 — 霍克斯过程用「过去的事件抬高当前强度」描述订单成簇;分支表示给出平均强度 $\mu/(1-\eta)$、簇大小 $1/(1-\eta)$ 与长窗口法诺因子 $1/(1-\eta)^2$,平稳性要求 $\eta = \alpha/\beta < 1$,核对无误。时间重标度检验能清楚地区分泊松与霍克斯模型。真实市场的分支比是否接近 1 仍有争议(只回顾第 39 本)。互激过程能同时再现签名图与埃普斯效应;但埃普斯效应也可以完全来自非同步交易,此时林—吉田估计量可以消除它。作者框架中「霍克斯过程即埃普斯效应的解释」应限定为「霍克斯过程是解释之一」。 (https://5dlibrary.com/answers/zh/order-book-microstructure-hawkes-toxicity/c03/) - Ch.4 Measuring Toxicity: the Probability of Informed Trading, VPIN and the Flash-Crash Dispute — PIN estimates the probability of informed trading with a mixture model of daily buy and sell counts; even when the model is right, estimates from 60 days carry a sampling error of about ±0.04, and the textbook form of the likelihood overflows for active stocks, which, Lin and Ke estimated, may have biased about 44 % of recent estimates downward. Whether PIN is priced as a measure of information asymmetry is contested. VPIN computes order imbalance in volume time and is simple to compute, but bulk classification can misread a rise in volatility as one-sided order flow, consistent with the critique of Andersen and Bondarenko. "Adverse-selection withdrawal" before the flash crash is a disputed mechanistic interpretation, not an established cause. Spread decomposition shows that the realized spread turns negative as toxicity rises and that, with price-elastic noise trading, the market shuts down beyond a threshold. (https://5dlibrary.com/answers/en/order-book-microstructure-hawkes-toxicity/c04/) 第4章 度量毒性:知情交易概率、VPIN 与闪崩之争 — PIN 用每日买卖笔数的混合模型估计知情交易的概率;即使模型正确,60 天数据的估计也有约 ±0.04 的抽样误差,而教科书形式的似然在活跃股票上会溢出,林与柯估计近期约 44% 的估计可能因此向下偏倚。PIN 是否以信息不对称的身份被定价,有相反的证据。VPIN 在成交量时间中计算订单不平衡,计算简单,但批量分类可能把波动率的上升误读为订单流的一边倒,这与安德森与邦达连科的批评一致。闪崩前「逆向选择撤单」是一种有争议的机制解释,不是已确立的因果。价差分解显示,毒性上升时已实现价差转负,在价格弹性的噪声交易下,超过一个门槛后市场会关闭。 (https://5dlibrary.com/answers/zh/order-book-microstructure-hawkes-toxicity/c04/) - Ch.5 Book Imbalance and Queues: Order Flow Imbalance, Queue Imbalance and Queue Position — Order flow imbalance (OFI) is the net effect of events at the best quotes over an interval; it is close to linear in contemporaneous mid-price changes, with an average $R^2$ of 65 % in CKS and a slope roughly inversely proportional to depth; it is not queue imbalance. Queue imbalance is related to the direction of the next price move, but non-linearly, and the relation depends on the absolute sizes of the two queues. Queue position determines how adversely selected a fill is: in the author's model a buy order at the back of the queue sees the price fall after its fill with probability 0.551, and earns about 20 % less on average than one at the front. (https://5dlibrary.com/answers/en/order-book-microstructure-hawkes-toxicity/c05/) 第5章 盘口不平衡与排队:订单流不平衡、挂单量不平衡与排队位置 — 订单流不平衡(OFI)是一段时间内最优价位各类事件的净效果,与同期中间价变化近似线性,CKS 报告的平均 $R^2$ 为 65%,斜率约与深度成反比;它不是挂单量不平衡。挂单量不平衡与下一次价格变动的方向有关,但关系是非线性的,并且依赖两边队列的绝对大小。排队位置决定了成交时的逆向选择程度:在作者的模型中,排在队尾的买单成交后价格下跌的概率为 0.551,平均收益比队首低约 20%。 (https://5dlibrary.com/answers/zh/order-book-microstructure-hawkes-toxicity/c05/) - Ch.6 How to Split a Large Order: Almgren–Chriss Optimal Execution — Under linear impact and a mean–variance criterion, the Almgren–Chriss model gives a hyperbolic-sine optimal trajectory; the risk aversion $\lambda$ slides between expected cost and variance, tracing the efficient frontier, with TWAP and aggressive execution at its two ends. In the program, Monte Carlo agrees with the formulas. VWAP is the risk-neutral optimum only when temporary impact is inversely proportional to volume; with constant impact it costs 5.3 % more than TWAP. Iceberg orders are a question of how orders are displayed, handled in a separate layer from the splitting schedule. Linear impact is a convenient approximation; the empirical impact of metaorders is close to a square-root law (Book 39). (https://5dlibrary.com/answers/en/order-book-microstructure-hawkes-toxicity/c06/) 第6章 大单怎样拆:阿尔姆格伦—克里斯最优执行 — 阿尔姆格伦—克里斯模型在线性冲击与均值—方差准则下给出双曲正弦形状的最优轨迹,风险厌恶系数 $\lambda$ 在期望成本与方差之间滑动,形成有效前沿;TWAP 与激进执行是前沿的两端。程序中的蒙特卡洛与公式一致。VWAP 只有在临时冲击与成交量成反比时才是风险中性下的最优策略,在常数冲击下它的成本比 TWAP 高 5.3%。冰山订单是挂单显示的问题,与拆单时间表分层处理。线性冲击是便于求解的近似,元订单的实证冲击近似为平方根律(第 39 本)。 (https://5dlibrary.com/answers/zh/order-book-microstructure-hawkes-toxicity/c06/) - Ch.7 Reading the Classics: the Market Director, the Conspicuous Mound and the Market Exchange — The call auction executes 96 units, with traders' surplus of 4,858. The monopolist middleman cuts volume to 52 units, bidding 22.9 and asking 71.8; he earns 2,543, traders keep only 1,335, and the total surplus lost is 980. After a tax of 10 on the middleman, volume falls further to 40 and the loss rises to 1,625, with only 400 in tax: in this model, taxing the monopolist does not correct the monopoly but makes trade even thinner. Extending the tax to all trades lowers volume from 96 to 83, raises 830 in tax and loses only 60. The middleman who sees every reservation price executes all 96 units with no loss of efficiency, but takes the whole surplus of 4,858. This last row comes closest to the theme of this book: an information advantage need not reduce trade, but it decides who gets the gains from it. (https://5dlibrary.com/answers/en/order-book-microstructure-hawkes-toxicity/c07/) 第7章 原典并读:司市、龙断与市易 — 集合竞价成交 96 个单位,交易者剩余 4,858。垄断中间商把成交量压到 52 个单位,买价 22.9、卖价 71.8,自己获利 2,543,交易者只剩 1,335,损失的总剩余为 980。对中间商征税 10 之后,成交量进一步降到 40,损失升到 1,625,税收只有 400:在这个模型中,对垄断者征税并不能纠正垄断,反而使交易更少。把税推广到所有交易时,成交量从 96 降到 83,税收 830,损失只有 60。能看见每个人保留价格的中间商则成交全部 96 个单位,没有效率损失,但全部 4,858 的剩余都归他所有。最后这一行最接近本书的主题:信息优势本身不一定减少交易,但它决定了交易的收益归谁。 (https://5dlibrary.com/answers/zh/order-book-microstructure-hawkes-toxicity/c07/) - Ch.8 Practice and a Pre-Registered Test — Chapter 8 gives a four-step checklist (adverse selection, order flow, inventory, execution) and a pre-registered simulated-matching experiment comparing the cross-session Sharpe ratios of inventory-skewed Avellaneda–Stoikov quotes and fixed symmetric quotes under toxic pulses on a simulator frozen by an independent third party (single primary outcome their difference, smallest effect of interest 0.31, 8,600 paired sessions, double-blind running, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. (https://5dlibrary.com/answers/en/order-book-microstructure-hawkes-toxicity/c08/) 第8章 实践与预注册检验 — 第八章给出逆向选择、订单流、库存与执行四步核查表,以及一项预注册的模拟撮合实验方案:在独立第三方冻结的模拟器中,比较库存偏置的阿韦拉内达—斯托伊科夫报价与固定对称报价在毒性脉冲下的跨会话夏普比率(唯一主要结局为二者之差,最小关注效应 0.31,8,600 个配对会话,双盲运行,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有数据都是虚构或模拟的。 (https://5dlibrary.com/answers/zh/order-book-microstructure-hawkes-toxicity/c08/) ## 5DL-B-046 · Computational Aesthetics, Visual Saliency Maps and Cognitive Ergonomics in User Interface Design 中文书名:计算美学、视觉显著性图谱与人机界面认知工程 DOI: https://doi.org/10.5281/zenodo.23836258 Cite: Fan, Y. (2026). Computational Aesthetics, Visual Saliency Maps and Cognitive Ergonomics in User Interface Design (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23836258 Book page: https://5dlibrary.com/books/computational-aesthetics-visual-saliency-ui/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all pages, brands and data in the book are fictional or simulated. Abstract: Where should a button go, how big should it be and what colour; how many options should an interface offer, in how many levels; and what makes a page "look good"? This book examines these questions with psychophysics, information theory and computational models of vision; it is neither a promise of conversion gains nor a design standard. It covers the evidence for two popular claims, that "the golden ratio 1.618 universally raises conversion" and that "everything must be reachable within three clicks" (rectangle-preference experiments since Fechner give weak, method-dependent results; a random layout of only 8 boxes contains a ratio within 1 % of 1.618 with probability 0.795; in Porter's 2003 industry study the number of clicks was unrelated to task completion, and the author's menu-tree model shows information scent and hierarchy shape mattering more than click counts); Hick's log₂(n+1) and Hyman's entropy H, and the non-standard Σp log₂(1/p+1) of the author's framework (equal to log₂(n+1) for equal probabilities, and neither law otherwise); Fitts's original and Shannon forms, and the optimal widths of a "search–decide–point" chain under stated assumptions (with the logarithmic form the optimal widths are proportional to click probabilities and the expected difficulty equals a constant plus Hyman's entropy); set-size effects in feature, serial and guided search; a standard-library implementation of the Itti–Koch–Niebur saliency model (nine-level pyramids, centre–surround differences, normalisation, winner-take-all and inhibition of return, 42 feature maps), its limits on web pages (centre bias, task and text; the F-pattern is a conditional description) and a fixation-timing model; Campbell and Robson's spatial-frequency channels and the Mannos–Sakrison contrast sensitivity function (peak near 7.9 cycles per degree), and WCAG 2.x relative luminance and contrast (the thresholds 0.03928 and 0.04045 give identical results for 8-bit colour; APCA was removed from the WCAG 3 drafts in 2023); modular type scales as conventions rather than a law derived from Weber's law; Birkhoff's aesthetic measure, which gives every rectangle of ordinary (not too elongated) proportions the same score of 1.25; Kolmogorov complexity, which is uncomputable and can only be bounded above by compressed length; and the evidence on visual complexity and first impressions (including the 50-millisecond study of Lindgaard and colleagues, 2006). Chapter 7 reads these beside the Kaogongji of the Rites of Zhou ("examining curves, aspects and dispositions to work the five materials"; "heaven has its seasons, earth its qi, materials their excellence and craftsmen their skill") and juan 4 of Li Jie's Yingzao fashi ("timber sections come in eight grades"; "the depth of each section is divided into fifteen fen, and ten fen are its breadth"): all eight grades have a 3:2 section, not the golden ratio; the arithmetic comparison of the human visual field with screen proportions is an analogy only. Chapter 8 gives a four-step checklist (salience, legibility, choice and pointing, evidence) and a pre-registered online randomised trial to be run by independent researchers (single primary outcome the log time to a correct click on the target, smallest effect of interest a 10 % shorter geometric-mean time, 158 analysable and 190 recruited per arm for 90 % power, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all pages, brands and data in the book are fictional or simulated. 摘要:一个按钮放在哪里、做多大、用什么颜色,界面上的选项该有几个、几层,页面怎样才「好看」?本书用心理物理学、信息论与计算视觉模型审视这些问题,不是转化率承诺,也不是设计规范。它依次讨论:两条流行说法——「黄金分割比 1.618 普遍提高转化」与「任何内容都必须在三次点击内到达」——的证据(费希纳以来的长方形偏好实验结论微弱且依赖方法;随机布局中只要有 8 个盒子,就有 0.795 的概率找到与 1.618 相差不到 1% 的比例;波特 2003 年的行业研究中点击次数与完成率无关,作者的菜单树模型显示信息气味与层级结构比点击数更重要);席克(log₂(n+1))与海曼(信息熵 H)两种写法,以及作者框架中 Σp log₂(1/p+1) 这一非标准式(等概时与 log₂(n+1) 相同,不等概时两者都不是),菲茨原式与香农式,以及「搜索—决策—指向」链在给定假设下的最优宽度(对数式下最优宽度与点击概率成正比,期望难度等于一个常数加海曼熵);特征搜索、序列搜索与引导搜索的集合大小效应;用纯标准库实现的伊蒂—科赫—尼布尔显著性模型(九级金字塔、中央—周边差分、归一化、胜者全得与返回抑制,42 张特征图)及其在网页上的局限(中央偏差、任务与文字的作用,F 形浏览只是条件性描述)与注视时序模型;坎贝尔—罗布森空间频率通道与曼诺斯—萨克里森对比敏感度函数(峰值约 7.9 周/度)、WCAG 2.x 相对亮度与对比度(0.03928 与 0.04045 两个阈值在 8 位色上结果完全相同,APCA 已于 2023 年移出 WCAG 3 草案);模块化字号比例不是由韦伯定律推出的定律,伯克霍夫审美度量对一般比例(不过分细长)的长方形给出同一分数 1.25,柯尔莫哥洛夫复杂度不可计算、只能用压缩长度作上界,以及视觉复杂度与第一印象(林德高等 2006 年的 50 毫秒研究)的实证。 第七章与《周礼·考工记》「审曲面埶,以饬五材」「天有时,地有气,材有美,工有巧」以及李诫《营造法式》卷四「材有八等」「各以其材之广,分为十五分,以十分为其厚」并读:八等材的广厚比都是 3:2,不是黄金比;人类视场与屏幕比例的算术对照只作类比。第八章给出显著性、可读性、选择与指向、证据四步核查表,以及一项由独立研究人员实施的预注册在线随机对照试验方案(唯一主要结局为找到并正确点击目标的对数完成时间,最小关注效应为几何平均时间缩短 10%,90% 把握度下每组 158 名可分析、招募 190 名,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有页面、品牌与数据都是虚构或模拟的。 Chapter answers: - Ch.1 Two Popular Rules: the Golden Ratio and Three Clicks — The golden ratio 1.618 is a weak, method-dependent peak in rectangle-preference experiments, and no peer-reviewed evidence shows it raising web conversion; in any moderately complex layout a ratio close to 1.618 is almost certain to be found by chance, and chance "significance" across many layout tests is common. The main evidence against the three-click rule comes from an industry study (Porter 2003); peer-reviewed menu studies show that depth and breadth each have costs and that information scent matters as much. The author's menu-tree model shows that the best number of levels depends on whether users are novices or experts, and that the correlation between clicks and failure comes from being lost, not from clicking. Both rules should be rewritten: proportion is a design choice, not a conversion lever; the number of clicks is a symptom to monitor, not a design constraint. (https://5dlibrary.com/answers/en/computational-aesthetics-visual-saliency-ui/c01/) 第1章 两条流行规则:黄金分割与三次点击 — 黄金分割比 1.618 在长方形偏好实验中是一个微弱、依赖方法的峰值,没有同行评审证据表明它提高网页转化;在稍复杂的布局中,碰巧找到一个接近 1.618 的比例几乎是必然的,多次版式测试中碰巧「显著」也很常见。三次点击法则的主要反证来自一项行业研究(波特 2003),同行评审的菜单研究表明深度与广度各有代价、信息气味同样重要。作者的菜单树模型显示,最优层数取决于用户是新手还是熟练者,而点击数与失败的相关来自迷路,不是点击本身。两条规则都应改写:比例是设计选择,不是转化杠杆;点击次数是需要监测的症状,不是设计约束。 (https://5dlibrary.com/answers/zh/computational-aesthetics-visual-saliency-ui/c01/) - Ch.2 Deciding and Pointing: the Hick–Hyman Law and Fitts's Law — The $+1$ in Hick's 1952 $\log(n+1)$ stands for the "no signal" alternative; Hyman's 1953 entropy $H = \sum p_i\log_2(1/p_i)$ has no $+1$. The framework's $\sum p_i\log_2(1/p_i + 1)$ equals $\log_2(n+1)$ with equal probabilities and is neither law otherwise; Hyman's entropy should be used instead. The Hick–Hyman law describes deciding when the target is known, not searching among unfamiliar labels, and its slope changes with practice and compatibility. Fitts's original form is negative for $D/W < 1/2$; interfaces need the Shannon form. "Minimising" the sum of search, decision and pointing makes sense only once the variables, the probability distribution and the type of user are stated: with the logarithmic form the optimal widths are proportional to click probabilities and the expected difficulty equals a constant plus Hyman's entropy; total time for novices is roughly linear and for experts roughly logarithmic, and the two call for different designs. (https://5dlibrary.com/answers/en/computational-aesthetics-visual-saliency-ui/c02/) 第2章 决定与指向:席克—海曼定律与菲茨定律 — 席克 1952 年的 $\log(n+1)$ 中的 $+1$ 对应「无信号」备择;海曼 1953 年的熵 $H = \sum p_i\log_2(1/p_i)$ 不加 1。作者框架中的 $\sum p_i\log_2(1/p_i + 1)$ 在等概时等于 $\log_2(n+1)$,不等概时两种定律都不是,应改用海曼熵。席克—海曼定律描述已知目标时的决策,不描述在陌生标签中的搜索,其斜率随练习与相容性而变。菲茨原式在 $D/W < 1/2$ 时为负,界面中应用香农式。把搜索、决策与指向相加的「最小化」只在写明变量、概率分布与用户类型之后才有意义:在对数式下,最优宽度与点击概率成正比,期望难度等于一个常数加海曼熵;新手的总时间近似线性、熟练用户近似对数,二者的最优设计不同。 (https://5dlibrary.com/answers/zh/computational-aesthetics-visual-saliency-ui/c02/) - Ch.3 What Pops Out First: Visual Search and Saliency Maps — The speed of visual search depends on the relative difference between target and distractors on features that can be computed in parallel: a large enough difference makes the target pop out, no difference reduces search to inspecting items one by one, and guided search lies in between; the simulated slopes agree with the prediction $50\,\Phi(-d'/\sqrt2)$. The structure of the Itti–Koch–Niebur model (nine-level pyramids, $c\in\{2,3,4\}$, $\delta\in\{3,4\}$, 42 feature maps, $\mathcal N(\cdot)$, winner-take-all with inhibition of return) checks out; a standard-library implementation shows that, in this model, a button's colour uniqueness can move it from the 18th shift of attention to the first. But the model describes only low-level saliency, and the paper itself says its trajectories are not necessarily similar to human fixations; on web pages, centre and top-left biases, the task, text and faces are often more important. (https://5dlibrary.com/answers/en/computational-aesthetics-visual-saliency-ui/c03/) 第3章 什么先跳出来:视觉搜索与显著性图谱 — 视觉搜索的速度取决于目标在可并行计算的特征上与干扰项的相对差异:差异足够大时目标「跳出」,没有差异时退化为逐项搜索,引导搜索介于两者之间,模拟中的斜率与 $50\,\Phi(-d'/\sqrt2)$ 的预测一致。伊蒂—科赫—尼布尔模型的结构(九级金字塔、$c\in\{2,3,4\}$、$\delta\in\{3,4\}$、42 张特征图、$\mathcal N(\cdot)$、胜者全得加返回抑制)核对无误;纯标准库的实现显示,在这个模型中,一个按钮的颜色唯一性可以把它从第 18 次注意转移提前到第 1 次。但模型只描述底层显著性,原文自己说明其轨迹不一定与人的注视相似;网页上的中央与左上偏差、任务、文字与人脸往往更重要。 (https://5dlibrary.com/answers/zh/computational-aesthetics-visual-saliency-ui/c03/) - Ch.4 How Gaze Moves: Centre Bias, Saccades and Inhibition of Return — Fixation durations depend on the task, commonly about 180–330 ms (Rayner 2009), and the framework's "200–300 ms" should be qualified accordingly; saccade amplitude, duration and velocity follow the main sequence. On web pages, centre bias, the top-left and reading order, the task and text all influence fixations; the F-pattern is a conditional description of scanning text-heavy pages. The author's mixture model shows that when fixations are driven by several sources together, predicting them from low-level saliency alone gives mediocre scores, and a content-free centre bias can do about as well. Inhibition of return moves attention away from locations already inspected; in the simulation it hardly affects a button that is the most salient element anyway, but for a button that is not salient, the shorter the inhibition, the harder it is to find. (https://5dlibrary.com/answers/en/computational-aesthetics-visual-saliency-ui/c04/) 第4章 注视怎样移动:中央偏差、扫视与返回抑制 — 注视时长依任务而异,常见范围约 180–330 毫秒(雷纳 2009),作者框架的「200–300 毫秒」应如此限定;扫视的幅度、时长与速度遵循主序列。在网页上,中央偏差、左上与阅读顺序、任务与文字都影响注视;F 形浏览是以文字为主、扫读时的条件性描述。作者的混合模型说明,当注视由几种来源共同决定时,只用底层显著性预测成绩平庸,与内容无关的中央偏差可以与之相当。返回抑制使注意离开已检查的位置;在模拟中,它对一个本来就最显著的按钮几乎没有影响,但对一个不显著的按钮,抑制越短,越难被找到。 (https://5dlibrary.com/answers/zh/computational-aesthetics-visual-saliency-ui/c04/) - Ch.5 Seeing Clearly: Spatial-Frequency Channels, Contrast Sensitivity and WCAG — Campbell and Robson's experiments support the existence of relatively independent spatial-frequency channels in vision: a square wave is visible when its fundamental reaches threshold and looks "square" only when its third harmonic does. The Mannos–Sakrison contrast sensitivity function peaks near 7.9 cycles per degree; the letter-identification band of common body-text sizes lies close to this region, while 12-pixel text at the desktop reference distance falls below the critical print size of about 0.2 degrees. The WCAG relative-luminance and contrast formulas are reproduced; the two thresholds 0.03928 and 0.04045 give identical results for 8-bit colour, as the program proves value by value. APCA has been removed from the WCAG 3 drafts, and WCAG 3's contrast algorithm has not been decided. (https://5dlibrary.com/answers/en/computational-aesthetics-visual-saliency-ui/c05/) 第5章 看得清:空间频率通道、对比敏感度与 WCAG — 坎贝尔与罗布森的实验支持视觉系统中存在相对独立的空间频率通道:方波在基频达到阈值时被看见,在三次谐波达到阈值时才显得「方」。曼诺斯—萨克里森的对比敏感度函数在约 7.9 周/度处最敏感;常见正文字号的字母识别频带落在这一区域附近,而 12 像素小字在桌面参考距离上低于约 0.2 度的临界字号。WCAG 的相对亮度与对比度公式照写;0.03928 与 0.04045 两个阈值在 8 位色上给出完全相同的结果,程序逐值证明。APCA 已移出 WCAG 3 草案,WCAG 3 的对比度算法尚未确定。 (https://5dlibrary.com/answers/zh/computational-aesthetics-visual-saliency-ui/c05/) - Ch.6 Can "Looking Good" Be Measured? Type Scales, Aesthetic Measure and Complexity — Modular type scales are design conventions: Weber's law requires only that the ratio exceed a threshold and cannot choose the best ratio; the choice is a trade-off between clarity of hierarchy and number of levels, and rounding to whole pixels makes small ratios uneven. Birkhoff's $M = O/C$ gives every rectangle of ordinary (not too elongated) proportions the same score of 1.25 and the square more (1.50), so no special status for the golden ratio follows. Kolmogorov complexity is uncomputable, and compressed length is only an upper-bound proxy. Visual complexity and prototypicality affect aesthetic ratings within tens of milliseconds (Lindgaard and colleagues 2006, Tuch and colleagues 2012, Reinecke and colleagues 2013), but these studies measure brief ratings, not behavioural outcomes. (https://5dlibrary.com/answers/en/computational-aesthetics-visual-saliency-ui/c06/) 第6章 能否度量「好看」:字号阶梯、审美度量与复杂度 — 模块化字号比例是设计惯例:韦伯定律只要求比例超过一个阈值,不能选出最优比例;比例的选择是层级清晰度与级数之间的取舍,整数取整还会让小比例的步长不均匀。伯克霍夫的 $M = O/C$ 对一般比例(不过分细长)的长方形给出同一分数 1.25,正方形更高(1.50),所以推不出黄金比的特殊地位。柯尔莫哥洛夫复杂度不可计算,压缩长度只是上界代理。视觉复杂度与原型性在几十毫秒内影响审美评分(林德高等 2006、图赫等 2012、赖内克等 2013),但这些研究测量的是短时评分,不是行为结局。 (https://5dlibrary.com/answers/zh/computational-aesthetics-visual-saliency-ui/c06/) - Ch.7 Reading the Classics: Examining Curves and Aspects, and Timber as the Root — Chapter 7 reads these beside the Kaogongji of the Rites of Zhou ("examining curves, aspects and dispositions to work the five materials"; "heaven has its seasons, earth its qi, materials their excellence and craftsmen their skill") and juan 4 of Li Jie's Yingzao fashi ("timber sections come in eight grades"; "the depth of each section is divided into fifteen fen, and ten fen are its breadth"): all eight grades have a 3:2 section, not the golden ratio; the arithmetic comparison of the human visual field with screen proportions is an analogy only. (https://5dlibrary.com/answers/en/computational-aesthetics-visual-saliency-ui/c07/) 第7章 原典并读:审曲面埶与以材为祖 — 第七章与《周礼·考工记》「审曲面埶,以饬五材」「天有时,地有气,材有美,工有巧」以及李诫《营造法式》卷四「材有八等」「各以其材之广,分为十五分,以十分为其厚」并读:八等材的广厚比都是 3:2,不是黄金比;人类视场与屏幕比例的算术对照只作类比。 (https://5dlibrary.com/answers/zh/computational-aesthetics-visual-saliency-ui/c07/) - Ch.8 Practice and a Pre-Registered Test — Chapter 8 gives a four-step checklist (salience, legibility, choice and pointing, evidence) and a pre-registered online randomised trial to be run by independent researchers (single primary outcome the log time to a correct click on the target, smallest effect of interest a 10 % shorter geometric-mean time, 158 analysable and 190 recruited per arm for 90 % power, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all pages, brands and data in the book are fictional or simulated. (https://5dlibrary.com/answers/en/computational-aesthetics-visual-saliency-ui/c08/) 第8章 实践与预注册检验 — 第八章给出显著性、可读性、选择与指向、证据四步核查表,以及一项由独立研究人员实施的预注册在线随机对照试验方案(唯一主要结局为找到并正确点击目标的对数完成时间,最小关注效应为几何平均时间缩短 10%,90% 把握度下每组 158 名可分析、招募 190 名,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有页面、品牌与数据都是虚构或模拟的。 (https://5dlibrary.com/answers/zh/computational-aesthetics-visual-saliency-ui/c08/) ## 5DL-B-047 · Choice Architecture, Cumulative Prospect Theory and Computational Behavioural Economics of Pricing Interfaces 中文书名:选择架构、累积前景理论与销售界面价格锚定的计算行为学 DOI: https://doi.org/10.5281/zenodo.23836260 Cite: Fan, Y. (2026). Choice Architecture, Cumulative Prospect Theory and Computational Behavioural Economics of Pricing Interfaces (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23836260 Book page: https://5dlibrary.com/books/choice-architecture-prospect-sales-interfaces/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. Abstract: Do more options sell more or less? How far can a struck-through price raise what a customer is willing to pay? Does an unattractive option placed next to the middle of three tiers make the middle tier sell better? This book presents mathematical models for teaching and research; it is not marketing, legal or compliance advice. It covers the evidence on choice overload and its limits (in the jam study 2.8 % against 29.8 % of those who stopped bought, 1.7 % against 11.9 % of those who passed; a meta-analysis found a mean effect near zero with large heterogeneity, and a later one identified four moderators), and how the publication filter inflates nudge effects (with a true effect of 1.4 percentage points and about 240 people per arm, significant results average about 8.7); the credibility ceiling of struck-through and reference prices (in the author's model the gain in willingness to pay is bounded and an exaggerated reference price can backfire) and the US and EU price-indication rules (the EU defines the "prior price" as the lowest price applied in at least the 30 days before a reduction); the value and probability-weighting functions of cumulative prospect theory (Tversky and Kahneman's 1992 weighting function and Prelec's 1998 function are different forms, and 0.65 is only an illustrative value of the latter; weighting each probability separately violates stochastic dominance, cumulative weights do not); how the loss-aversion coefficient depends on the model and the stakes (one fictitious decision-maker's simulated choices at small, medium and large stakes, fitted with the same linear-value model, give 1.43, 1.13 and 0.88 against a true 1.5), and why prospect theory does not by itself yield a checkout conversion rate: it needs a link function whose scale parameter sets the size of any "framing gap"; the IIA property of the multinomial logit and the regularity theorem (in no random utility model can adding an option raise the choice probability of an existing one, so the decoy effect violates every random utility model), how Tversky and Simonson's context-dependent model and multialternative decision field theory produce the attraction effect, and how that effect often vanishes with perceptual or experienced attributes; and quasi-hyperbolic discounting, naive and sophisticated procrastination, and the author's model of multi-step forms (with 3 to 14 steps the naive person waits until the last day and completes with probability 0.488). Chapter 7 reads these beside "goods that are plentiful are cheap, scarce dear; dispersed light, concentrated heavy" in the "Guo xu" chapter of the Guanzi (located, after checking, in "Guo xu" rather than "Qingzhong jia"), "the balance has no fixed number" in "Qingzhong yi", and Jiran's "no higher than eighty, no lower than thirty" in the "Biographies of the Money-Makers" of the Shiji, as statements about state price management and price bands rather than a modern theory of price anchoring. Chapter 8 gives a four-step checklist (options, reference prices, decoys, friction and defaults) and a pre-registered online randomised experiment: on a three-tier pricing page for a fictitious product, run by independent researchers, it compares the share choosing the target tier with and without an asymmetrically dominated decoy (single primary outcome, smallest effect of interest 5 percentage points, 90 % power with 1,842 analysable and 2,200 recruited per arm, decision rule written as code). The three-tier example prices of US$9.99, 49.90 and 149.00 match 5D Library's own prices; this is stated in the disclosure, and nothing in the book validates those prices. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. 摘要:选项多一些会卖得更好还是更差?划掉一个原价能把顾客愿意支付的价格抬多高?三档定价中间那一档旁边放一个「不划算」的选项,会不会让中间档卖得更多?本书是教学与研究用的数学模型,不是营销、法律或合规建议。它依次讨论:选择过载的证据与边界(果酱研究中购买的是停留者的 2.8% 对 29.8%,按路过者计算为 1.7% 对 11.9%;梅塔分析的平均效应接近零而异质性大,后续梅塔分析找到四个调节变量),以及助推效应在发表筛选下的放大(真实效应 1.4 个百分点、每组约 240 人时,显著结果的平均值约为 8.7);划线价与参照价格的可信度上限(作者的模型中支付意愿的增量有上限,夸大的参照价可能适得其反)与美国、欧盟的价格标示规则(欧盟以降价前不少于 30 天内的最低价为「此前价格」);累积前景理论的价值函数与概率权重(特沃斯基与卡尼曼 1992 年的权重函数与普雷莱克 1998 年的函数是两种不同的形式,0.65 只是后者的示意值;逐项加权会违反随机占优,累积权重不会);损失厌恶系数随模型设定与赌注大小而变(同一位虚构决策者在小、中、大三档赌注下的模拟选择,用同一个线性模型估计,分别得到 1.43、1.13 与 0.88,真值 1.5),以及前景理论不直接给出结账转化率:它需要一个连接函数,其尺度参数决定了「框架差」的大小;多项 Logit 的 IIA 与正则性定理(任何随机效用模型都不能让新增选项提高原有选项的被选概率,所以诱饵效应违反的是所有随机效用模型),特沃斯基—西蒙森的情境依赖模型与多属性决策场理论如何产生吸引效应,以及吸引效应在知觉型、体验型属性下常常消失;准双曲贴现、天真与老练的拖延,以及作者的分步表单模型(步数为 3 至 14 步时,天真者拖到最后一天、完成率 0.488)。 第七章与《管子·国蓄》「夫物多则贱,寡则贵,散则轻,聚则重」(核对后在《国蓄》而非《轻重甲》)、《轻重乙》「衡无数也」及《史记·货殖列传》计然「上不过八十,下不减三十」并读,说明它们是君主调控与价格区间的论述,不是现代的价格锚定理论。第八章给出四步核查表(选项、参照价、诱饵、摩擦与默认)与一项预注册的在线随机实验方案:在由独立研究人员实施、使用虚构产品的三档价格页面上,比较有无非对称占优诱饵时选择目标档的比例(唯一主要结局,最小关注效应 5 个百分点,90% 把握度每组可分析 1,842 人、招募 2,200 人,判定规则写成代码)。书中三档价格示例 9.99、49.90、149.00 美元与 5D Library 自己的定价相同,已在利益披露中说明,书中分析不构成对该定价的验证。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有数据都是虚构或模拟的。 Chapter answers: - Ch.1 How Many Options: Choice Overload and the Size of Nudge Effects — The "3 % against 30 %" of the jam study is the purchase share among those who stopped; per passer-by it is 1.7 % against 11.9 %, and the 24-jam display drew more people in. Meta-analysis shows a mean choice-overload effect near zero with large heterogeneity, and identifies four moderators: complexity, task difficulty, preference uncertainty and an effort-minimising goal. The author's model shows the direction of the option-count effect set by comparison cost and preference uncertainty, favouring more options in 8 of 12 settings and fewer in 4. Large nudge-unit trials average 1.4 percentage points against 8.7 in academic journals; with only significant results published, a true effect of 1.4 points appears on average as 8.7 in studies of about 240 per arm. So neither "more options are better" nor "choice overload is universal" holds, and any large effect reported from a single experiment calls first for the question of what filter it passed through. (https://5dlibrary.com/answers/en/choice-architecture-prospect-sales-interfaces/c01/) 第1章 选项的多少:选择过载与助推效应的规模 — 果酱研究中「3% 对 30%」是停留者中的购买比例,按路过者计算是 1.7% 对 11.9%;24 种一组吸引了更多人停留。梅塔分析显示选择过载的平均效应接近零、异质性很大,并找到复杂性、任务难度、偏好不确定性与省力目标四个调节变量。作者的模型显示选项数量的效应方向由比较成本与偏好不确定性决定,12 种设定中 8 种对多选项有利、4 种不利。助推机构的大规模试验平均效应为 1.4 个百分点,学术期刊为 8.7;只发表显著结果时,一个 1.4 个百分点的真实效应在每组约 240 人的研究中平均显示为 8.7。所以「选项越多越好」与「选择过载普遍存在」都不成立,而任何单次实验报告的大效应都需要先问它经过了怎样的筛选。 (https://5dlibrary.com/answers/zh/choice-architecture-prospect-sales-interfaces/c01/) - Ch.2 Struck-Through and Reference Prices: Transaction Utility, Credibility and Price-Indication Rules — Thaler's transaction utility explains a struck-through price as raising the reference price. The evidence checked shows that reference-price effects exist and that exaggerated reference prices lose credibility, without reversing in those studies; Krishna and co-authors' meta-analysis shows the regular price helping large plausible deals and implausible ones, but not small plausible ones. The author's model shows that as long as credibility saturates with exaggeration, the gain in WTP is bounded (3.00, or 7.5 %, in the example); with suspicion added, very high struck-through prices lower WTP, but that is a hypothesis. The US Guides require the former price to be bona fide and regularly offered; EU Article 6a requires the prior price to be the lowest price in at least the 30 days before a reduction, which blocks "raise briefly, then cut" but cannot by itself exclude a long-standing inflated list price. "Striking through a former price raises willingness to pay without limit" therefore fails in theory, in evidence and in law. (https://5dlibrary.com/answers/en/choice-architecture-prospect-sales-interfaces/c02/) 第2章 划线价与参照价格:交易效用、可信度与标示规则 — 塞勒的交易效用把划线价的作用解释为提高参照价格。已核对的证据显示参照价格效应存在,夸大的参照价格可信度下降,但并未在这些研究中反转;克里希纳等人的梅塔分析显示原价对大额可信折扣与不可信折扣有作用、对小额可信折扣没有作用。作者的模型说明,只要可信度随夸大而饱和,WTP 的增量就有上限(示例中为 3.00,即 7.5%);加入怀疑后,过高的划线价会降低 WTP,但这是假说。美国指南要求原价是真实、善意、经常报出的价格;欧盟第 6a 条要求此前价格是降价前不少于 30 天内的最低价,它禁止「短期抬价再降价」,但不能单独排除长期虚高的挂牌价。所以「划掉原价即可无限抬高支付意愿」在理论、证据与法律三个层面都不成立。 (https://5dlibrary.com/answers/zh/choice-architecture-prospect-sales-interfaces/c02/) - Ch.3 Cumulative Prospect Theory: Value Functions, Probability Weights and Rank Dependence — The value-function parameters of 1992 cumulative prospect theory, $\alpha = \beta = 0.88$ and $\lambda = 2.25$, match the author's framework, but they are medians from 25 graduate students not paid according to their choices. The 1992 weighting function is $p^\gamma/(p\gamma + (1-p)^\gamma){1/\gamma}$ ($\gamma = 0.61$, $\delta = 0.69$); $\exp(-(-\ln p)^\alpha)$ is Prelec's different 1998 function, 0.65 is an illustrative value, and the parameter closest to the 1992 gain curve is 0.501. Cumulative prospect theory weights cumulative probabilities and takes differences; separable weighting prefers a dominated lottery and lifts the certainty equivalent to 316.63 when one outcome is split 50 ways, cumulative weights do not. (https://5dlibrary.com/answers/en/choice-architecture-prospect-sales-interfaces/c03/) 第3章 累积前景理论:价值函数、概率权重与秩依赖 — 1992 年累积前景理论的价值函数参数 $\alpha = \beta = 0.88$、$\lambda = 2.25$ 与作者框架一致,但它们是 25 名研究生、不按选择付酬的中位数。1992 年的权重函数是 $p^\gamma/(p\gamma + (1-p)^\gamma){1/\gamma}$($\gamma = 0.61$,$\delta = 0.69$);$\exp(-(-\ln p)^\alpha)$ 是普雷莱克 1998 年的另一个函数,0.65 是示意值,与 1992 年收益域曲线最接近的参数是 0.501。累积前景理论对累积概率加权再取差分;逐项加权会偏好被占优的彩票,把一个结果拆成 50 份可把确定性等价抬到 316.63,累积权重不会。 (https://5dlibrary.com/answers/zh/choice-architecture-prospect-sales-interfaces/c03/) - Ch.4 Loss Aversion and Framing: From Individual Choice to a "Conversion Gap" — The meta-analytic mean of the loss-aversion coefficient is about 1.96, but estimates vary with the task, the stakes and the model specification, and some researchers dispute that losses generally outweigh gains. The program shows a decision-maker with a true $\lambda = 1.5$ estimated at 1.43, 1.13 and 0.88 by a linear model depending on the stakes. Framing effects are reliable but small to moderate overall. Cumulative prospect theory yields values, not shares; a "conversion gap" requires a link function and a scale parameter; the 1992 medians with one common scale cannot fit both shares in the Asian disease problem, and in the checkout example the framing gap ranges from 4.3 to 26.7 percentage points. The framework's "conversion gap" can only be illustrative. (https://5dlibrary.com/answers/en/choice-architecture-prospect-sales-interfaces/c04/) 第4章 损失厌恶与框架:从个体选择到「转化率差」 — 损失厌恶系数的梅塔分析均值约为 1.96,但估计值随任务、赌注与模型设定变化,也有研究者质疑损失普遍重于收益。程序显示,同一个真实 $\lambda = 1.5$ 的决策者,在线性模型下按赌注大小被估计为 1.43、1.13 与 0.88。框架效应可靠但总体为小到中等。累积前景理论给出价值而不是比例,要得到「转化率差」必须加上连接函数与尺度参数;1992 年中位数参数加上一个共同尺度不能同时拟合亚洲疾病问题的两个比例,结账页面例子中的框架差可以从 4.3 到 26.7 个百分点不等。作者框架中的「转化率差」只能作为示意。 (https://5dlibrary.com/answers/zh/choice-architecture-prospect-sales-interfaces/c04/) - Ch.5 Decoys and Compromises: Regularity in Discrete-Choice Models and Its Violation — The multinomial logit satisfies IIA, so adding a decoy lowers the shares of all existing options proportionally. More generally, every random utility model satisfies regularity, so the decoy effect violates all random utility models; correlated errors can at most stop the decoy from taking share, never make the target rise. In the Tversky–Simonson model the attraction effect appears once the weight on relative advantage passes a threshold (between 0.01 and 0.03 in the example), while the compromise effect also needs a convex disadvantage function; in MDFT both depend on lateral inhibition, the compromise effect needing broader and stronger inhibition. Empirically the attraction effect replicates in abstract choices with numerical attributes but often vanishes or even reverses with perceptual and experiential attributes, and only 11 % to 25 % notice dominance in the field. The example prices are the same as 5D Library's own, and nothing in this chapter validates them. (https://5dlibrary.com/answers/en/choice-architecture-prospect-sales-interfaces/c05/) 第5章 诱饵与折中:离散选择模型的正则性及其违反 — 多项 Logit 满足无关选项独立性,加入诱饵只会按比例降低所有原有选项的份额。更一般地,任何随机效用模型都满足正则性,所以诱饵效应违反的是所有随机效用模型;误差相关最多使诱饵不带走份额,不能使目标档上升。特沃斯基—西蒙森模型中吸引效应在相对优势权重超过一个阈值(示例中 0.01 至 0.03 之间)后出现,折中效应还需要凸的劣势函数;MDFT 中两者都依赖侧抑制,折中效应需要更宽更强的抑制。实证上,吸引效应在数字属性的抽象选择中可以重现,在知觉与体验型属性中常常消失甚至反转,现场察觉率只有 11% 至 25%。示例价格与 5D Library 自家定价相同,本章结果不构成对该定价的验证。 (https://5dlibrary.com/answers/zh/choice-architecture-prospect-sales-interfaces/c05/) - Ch.6 Time: Quasi-Hyperbolic Discounting, Procrastination and the Friction of Forms — The quasi-hyperbolic discount function originates in Phelps and Pollak's intergenerational model; Laibson applied it to one person's dates; it and the hyperbolic function both produce preference reversals, the exponential does not. Present bias is clearer in effort tasks than in choices over money. In the author's form model the threshold for today beating tomorrow is 2 steps; with 3 to 14 steps the naive person waits for the last day and completes with probability 0.488, while the sophisticated person completes on day 2 with probability 0.902 for 4 to 7 steps, her day of action becoming non-monotone beyond that; when cost exceeds $\beta V$ both give up. Defaults make no difference for exponential discounters, change the outcome for about half of naive people and for about a tenth of sophisticated ones. The framework's "procrastination–default threshold" is, in this model, two thresholds, and it is the author's model, not a published result. (https://5dlibrary.com/answers/en/choice-architecture-prospect-sales-interfaces/c06/) 第6章 时间:准双曲贴现、拖延与表单的摩擦 — 准双曲贴现函数源于费尔普斯与波拉克的代际模型,莱布森把它用于个人的不同时点;它与双曲函数都会产生偏好反转,指数函数不会。现时偏差在努力任务中比在金钱选择中明显。作者的分步表单模型中,今天做胜过明天做的门槛为 2 步;步数为 3 至 14 时,天真者拖到最后一天、完成率 0.488,老练者在 4 至 7 步时于第 2 天完成、完成率 0.902,步数再多时行动日不单调;成本超过 $\beta V$ 时两者都放弃。默认选项对指数贴现者无影响,对天真者改变约一半人的结局,对老练者约一成。框架中的「拖延—违约临界值」在这个模型中是两个门槛,而且是作者的模型,不是已发表的结果。 (https://5dlibrary.com/answers/zh/choice-architecture-prospect-sales-interfaces/c06/) - Ch.7 Reading the Classics: Weight and Lightness in the Guanzi and the Policies of Jiran — With no persistence in harvest shocks ($\rho = 0$), without a granary the mean price is 54.0 with standard deviation 22.8, and of 4,000 years 399 fall below 30 and 490 rise above 80; the granary brings the years above 80 down to 0 and those below 30 to 168 (its storage cap stops it absorbing all of a glut), lowers the price standard deviation to 16.9, and makes a cumulative profit of 1,744.2 — buying low and selling high is itself "the ruler gains the profit". With persistent shocks ($\rho = 0.5$), runs of bad harvests exhaust the stock, and in 38 years the price breaks through 80, exactly the 38 years in which the stock ran out: the credibility of a band policy rests on stocks, not just on ordinances. (https://5dlibrary.com/answers/en/choice-architecture-prospect-sales-interfaces/c07/) 第7章 原典并读:《管子》轻重、衡数与计然之策 — 收成冲击不持续($\rho = 0$)时,没有常平仓的价格均值为 54.0、标准差 22.8,4,000 年中有 399 年低于 30、490 年高于 80;常平仓把高于 80 的年份降到 0,低于 30 的年份降到 168(库存上限使它在丰年收不完),价格标准差降到 16.9,并且累计盈利 1,744.2——低买高卖本身就是「君得其利」。收成冲击持续($\rho = 0.5$)时,连续的歉收会耗尽库存,有 38 年价格突破 80,而这 38 年正是库存用完的年份:区间政策的可信度取决于库存,而不只是政令。 (https://5dlibrary.com/answers/zh/choice-architecture-prospect-sales-interfaces/c07/) - Ch.8 Practice and Pre-Registered Testing — Chapter 8 gives a four-step checklist (options, reference prices, decoys, friction and defaults) and a pre-registered online randomised experiment: on a three-tier pricing page for a fictitious product, run by independent researchers, it compares the share choosing the target tier with and without an asymmetrically dominated decoy (single primary outcome, smallest effect of interest 5 percentage points, 90 % power with 1,842 analysable and 2,200 recruited per arm, decision rule written as code). The three-tier example prices of US$9.99, 49.90 and 149.00 match 5D Library's own prices; this is stated in the disclosure, and nothing in the book validates those prices. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. (https://5dlibrary.com/answers/en/choice-architecture-prospect-sales-interfaces/c08/) 第8章 实践与预注册检验 — 第八章给出四步核查表(选项、参照价、诱饵、摩擦与默认)与一项预注册的在线随机实验方案:在由独立研究人员实施、使用虚构产品的三档价格页面上,比较有无非对称占优诱饵时选择目标档的比例(唯一主要结局,最小关注效应 5 个百分点,90% 把握度每组可分析 1,842 人、招募 2,200 人,判定规则写成代码)。书中三档价格示例 9.99、49.90、149.00 美元与 5D Library 自己的定价相同,已在利益披露中说明,书中分析不构成对该定价的验证。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有数据都是虚构或模拟的。 (https://5dlibrary.com/answers/zh/choice-architecture-prospect-sales-interfaces/c08/) ## 5DL-B-048 · Algorithmic Marketing, Multi-Touch Shapley Attribution and Stochastic Control of Acquisition Funnels 中文书名:算法营销、多触点沙普利归因与获客漏斗的随机控制 DOI: https://doi.org/10.5281/zenodo.23836264 Cite: Fan, Y. (2026). Algorithmic Marketing, Multi-Touch Shapley Attribution and Stochastic Control of Acquisition Funnels (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23836264 Book page: https://5dlibrary.com/books/algorithmic-marketing-shapley-attribution-funnels/ Scope: Which advertising channel should a sale be credited to? When may an A/B test stop? What is a customer still worth? This book presents mathematical models for teaching and research and is not marketing, investment or legal advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. Abstract: Which advertising channel should a sale be credited to? When may an A/B test stop? What is a customer still worth? This book presents mathematical models for teaching and research and is not marketing, investment or legal advice. It covers the errors of last-click, first-click and linear attribution in a fictional world whose true incrementality is known (last click credits 7,299 conversions to ads, which caused about 3,636); Shapley-value attribution (Shapley 1953; the versions of Shao and Li and of Dalessandro and co-authors) and the removal effect of an absorbing Markov chain (Anderl and co-authors 2016), their dependence on the characteristic function v(S) and their disagreement on synergistic channels — and shows that on observational data both allocate correlation, not causal increments: with targeting, observational Shapley shares are off by 13.1 percentage points, the same order as last click; only randomised holdouts and ghost-ad experiments measure increments (Lewis and Rao 2015, Gordon and co-authors 2019, Johnson and co-authors 2017, Blake and co-authors 2015); the exact type I error of optional stopping (at a nominal 5 %, 14.2 % with 5 looks, 24.8 % with 20 and 37.4 % with 100, so "25 %+" holds only from about 20 looks), Pocock and O'Brien–Fleming group-sequential boundaries and the mixture sequential probability ratio test (always-valid inference); the regret of Thompson sampling against an A/B/C test, the negative bias of sample means under adaptive allocation and its inverse-propensity correction; standard-library implementations of the BG/NBD and Gamma-Gamma models (₂F₁ by its series and by Euler's integral after Pfaff's transformation, each checking the other), the independence assumption of Gamma-Gamma, and "retention rising over time", which is a sorting effect of heterogeneity (the sBG model reproduces Fader and Hardie's 2007 estimates exactly), not Simpson's paradox, with a genuine Simpson's paradox built separately; the Bass diffusion model, the Dorfman–Steiner condition (advertising-to-sales ratio = advertising elasticity / price elasticity, a static rule) and the author's model of "whole-life optimal acquisition spend" (front-loaded under concave response; on one S-shaped response curve, alternate-month pulsing is better at lower monthly budgets, with a switch at about 30–32.5 thousand that does not coincide with the inflection point). Chapter 7 reads these beside the Sunzi's "Initial Estimates" ("shi is to shape one's plans according to advantage") and "Strategic Dispositions" ("the swiftness of rushing water, to the point of tossing stones about, is shi; the swiftness of a hawk, to the point of breaking its prey, is jie"; "therefore his shi is steep, his jie short"), following Yang Bing'an's collated Eleven Commentaries (the 是 missing from the author's framework is restored; according to Giles the Taiping yulan quotes 疾 as 击, while the main text follows the Eleven Commentaries in reading 疾, pending the printed collation notes), and uses the author's goodwill-stock model to show what "storing force" and "a short release" mean as an analogy, and where the analogy ends. Chapter 8 gives a four-step checklist (attribution, experiments, lifetime value, budget) and a pre-registered simulation benchmark: on a synthetic multichannel data generator with known true increments, frozen by an independent third party, observational Shapley attribution is compared with last click on their error against the true incremental shares (single primary outcome the difference in mean absolute errors, smallest effect of interest 2 percentage points, 79 synthetic worlds for 90 % power, 90 planned, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. 摘要:一笔成交该算在哪个广告渠道头上?一个 A/B 测试什么时候可以停?一位客户还值多少钱?本书是教学与研究用的数学模型,不是营销、投资或法律建议。它依次讨论:末次点击、首次点击与线性归因在一个已知真实增量的虚构世界中的误差(末次点击把 7,299 次成交记在广告名下,广告实际造成约 3,636 次),沙普利值归因(Shapley 1953;Shao 与 Li、Dalessandro 等人的版本)与吸收马尔可夫链移除效应(Anderl 等 2016)的定义、它们对特征函数 v(S) 的依赖与在协同渠道上的分歧——并说明二者在观测数据上分配的是相关性而不是因果增量:有定向投放时,观测沙普利值的份额误差为 13.1 个百分点,与末次点击同一量级;只有随机化的持出、幽灵广告实验能给出增量(Lewis 与 Rao 2015、Gordon 等 2019、Johnson 等 2017、Blake 等 2015);可选停止的精确第一类错误(名义 5% 时看 5 次为 14.2%、20 次为 24.8%、100 次为 37.4%,「25%+」只在约 20 次以上成立)、波科克与奥布赖恩—弗莱明群序贯边界及混合序贯概率比检验(始终有效推断);汤普森采样与 A/B/C 测试的后悔比较,以及自适应分配下样本均值的负偏差与逆倾向加权修正;BG/NBD 与 Gamma-Gamma 模型的纯标准库实现(₂F₁ 用级数与普法夫变换后的欧拉积分互相核对)、Gamma-Gamma 的独立性假设,以及「留存率随时间上升」——这是异质性造成的分选效应(sBG 模型精确重现 Fader 与 Hardie 2007 年的参数),不是辛普森悖论,书中另造一个真正的辛普森悖论例子;巴斯扩散模型、多夫曼—斯坦纳条件(广告/销售额 = 广告弹性/价格弹性,静态规则)以及作者的「全周期最优买量」模型(凹响应下前重后轻;在一条 S 形响应曲线上,月均预算较低时隔月脉冲更好,分界约在 30–32.5 千之间,并不等于拐点)。 第七章与《孙子兵法》的《计篇》「势者,因利而制权也」和《势篇》「激水之疾,至于漂石者,势也;鸷鸟之疾,至于毁折者,节也」「是故善战者,其势险,其节短」并读(据杨丙安《十一家注孙子校理》,作者框架脱「是」字已补;「疾」字据翟林奈注《太平御览》引作「击」,正文从十一家注本作「疾」,待纸本校记确认),用作者的商誉存量模型说明「蓄势」与「短节」作为类比的含义与限度。第八章给出归因、实验、生命周期价值与预算四步核查表,以及一项预注册的模拟基准研究:在独立第三方冻结、已知真实增量的合成多渠道数据生成器上,比较观测沙普利归因与末次点击对真实增量份额的误差(唯一主要结局为二者平均绝对误差之差,最小关注效应 2 个百分点,90% 把握度需要 79 个合成世界、计划 90 个,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有数据都是虚构或模拟的。 Chapter answers: - Ch.1 Attribution Is Not Incrementality: Last Click, Potential Outcomes and Randomised Holdouts — The world has 7,676 conversions, of which about 4,000 would have happened without any advertising (high-intent users buy anyway); the ads caused only about 3,636. Last click credits 7,299 conversions to the ads, twice the true incrementality. The shares matter more: social's true incremental share is 51.2 %, and last click gives it 21.0 %; search's true share is 11.0 %, and last click gives it 38.5 %. There are two reasons, independent of each other. First, search usually comes at the end of a path (people search once they have decided to buy), so the last-touch rule naturally favours it. Second, search and email mainly reach people who were going to buy anyway, so they "appear" on paths much more often than they "cause" conversions. First click and linear do little better, with mean absolute errors of 12.3 and 11.9 percentage points (last click 18.1). (https://5dlibrary.com/answers/en/algorithmic-marketing-shapley-attribution-funnels/c01/) 第1章 归因不是增量:末次点击、潜在结果与随机持出 — 这个世界里共有 7,676 次转化,其中约 4,000 次即使没有任何广告也会发生(高意向者本来就会买),广告真正造成的只有约 3,636 次。末次点击把 7,299 次转化记在了广告名下,是真实增量的两倍。更重要的是份额:社交的真实增量份额是 51.2%,末次点击只给它 21.0%;搜索的真实份额是 11.0%,末次点击给了 38.5%。原因有两个,而且彼此独立:第一,搜索通常出现在路径末端(用户决定买了才去搜),所以末次规则天然偏向它;第二,搜索与邮件主要触达本来就要买的人,它们在路径上「出现」的频率远高于它们「造成」转化的频率。首次点击与线性规则也好不到哪里去,平均绝对误差分别为 12.3 与 11.9 个百分点(末次点击 18.1)。 (https://5dlibrary.com/answers/zh/algorithmic-marketing-shapley-attribution-funnels/c01/) - Ch.2 Shapley Values and Markov Removal Effects: Exact Allocations of an Inexact Question — The Shapley value is the only allocation satisfying efficiency, symmetry, null player and additivity, and under synergy it splits the interaction effect equally among participants; its meaning depends entirely on the characteristic function $v(S)$. When $v(S)$ is estimated from observed paths, it inherits the selection bias of targeted delivery: in the fictional world of Chapter 1, its share error is 13.1 points with targeting and 2.0 points without. The Markov removal effect has a similar error in that world (13.7), but in a world in which only two channels work together it gives more than half the credit to a wide-reach channel that does nothing. "Exact multi-touch attribution" in the author's framework should be qualified as "exact axiomatic allocation"; the end point of attribution is the incrementality experiment, not a more refined allocation formula. (https://5dlibrary.com/answers/en/algorithmic-marketing-shapley-attribution-funnels/c02/) 第2章 沙普利值与马尔可夫移除效应:精确的分配,不精确的问题 — 沙普利值是唯一满足效率、对称、零贡献与可加四条公理的分配,在协同时把交互效应平分给参与者;它的含义完全取决于特征函数 $v(S)$。用观测路径估计 $v(S)$ 时,它继承了定向投放的选择偏差:在第一章的虚构世界中,有定向时份额误差 13.1 个百分点,无定向时 2.0 个百分点。马尔可夫移除效应在同一世界中误差相近(13.7),但在一个只有两个渠道协同起作用的世界里,它把一半以上的功劳给了一个毫无作用的高覆盖渠道。作者框架中「精确的多触点归因」应限定为「精确的公理化分配」;归因的终点是增量实验,不是更精细的分配公式。 (https://5dlibrary.com/answers/zh/algorithmic-marketing-shapley-attribution-funnels/c02/) - Ch.3 The Price of Peeking: Optional Stopping, Group-Sequential Designs and Always-Valid Inference — With equally spaced looks and the unadjusted 1.96 at each, a nominal 5 % test has an actual type I error of 14.2 % with 5 looks, 24.8 % with 20 and 37.4 % with 100; "soaring to 25 % or more" holds only from about 20 looks. The problem is not looking but failing to fix a stopping rule in advance. Group-sequential designs adjust the critical values for a fixed number of looks (with 5 looks, Pocock 2.413, O'Brien–Fleming from 4.562 down to 2.040) and save a quarter to a third of the sample on average; always-valid inference allows looking and stopping at any time, at the cost of somewhat larger samples for small effects. (https://5dlibrary.com/answers/en/algorithmic-marketing-shapley-attribution-funnels/c03/) 第3章 偷看的代价:可选停止、群序贯设计与始终有效推断 — 在每次都用 1.96 作判据的等间隔查看下,名义 5% 的检验在看 5 次时实际第一类错误率为 14.2%,20 次为 24.8%,100 次为 37.4%;「飙升至 25% 以上」只在约 20 次以上查看时成立。问题不在于查看,而在于没有事先规定停止规则。群序贯设计在事先固定的查看次数下调整临界值(5 次查看时波科克 2.413、奥布赖恩—弗莱明从 4.562 降到 2.040),平均可节省四分之一到三分之一的样本;始终有效推断允许随时查看与停止,以对小效应稍多的样本为代价。 (https://5dlibrary.com/answers/zh/algorithmic-marketing-shapley-attribution-funnels/c03/) - Ch.4 Learning While Earning: Thompson Sampling and the Bias of Adaptive Experiments — Thompson sampling allocates traffic by "the posterior probability of being the best variant"; in the fictional three-creative example it cuts regret from 33.9 conversions for the A/B/C test to 26.1, but lowers the share of runs that pick the best creative from 83 % to 79 %. When two variants are identical, Thompson allocation leaves the mean sample conversion rate 0.33 points too low and nominal intervals cover only 82.5 % of the time; inverse-propensity weighting with the known allocation probabilities removes the bias. The author's framework is right to present bandits as a tool for "optimising conversion", but they should not be used as a tool for measuring effects; the two goals differ, and which one is wanted should be stated when the experiment is designed. (https://5dlibrary.com/answers/en/algorithmic-marketing-shapley-attribution-funnels/c04/) 第4章 边做边学:汤普森采样与自适应实验的估计偏差 — 汤普森采样按「是最好版本的后验概率」分配流量,在虚构的三创意例子中把后悔从 A/B/C 测试的 33.9 次降到 26.1 次,但选对最好创意的比例从 83% 降到 79%。在两个版本完全相同时,汤普森分配使样本转化率平均偏低 0.33 个百分点,名义置信区间的覆盖率降到 82.5%;用已知分配概率做逆倾向加权可以消除偏差。作者框架中把赌博机写成「优化转化」的工具是对的,但不能把它当作测量效果的工具;两者的目标不同,应当在实验设计时就说清楚要哪一个。 (https://5dlibrary.com/answers/zh/algorithmic-marketing-shapley-attribution-funnels/c04/) - Ch.5 What Is a Customer Still Worth? BG/NBD, Gamma-Gamma and the Illusion of Rising Retention — BG/NBD describes the purchasing and dropout of non-contractual customers with four parameters and depends only on frequency and recency; the series for the $_2F_1$ in its expectation converges very slowly for long horizons, and Euler's integral after Pfaff's transformation and a change of variable computes it stably, the two algorithms checking each other. The Gamma-Gamma model gives a shrunken expected spend but assumes spend and frequency are independent; in a fictional example in which frequent buyers spend less, violating the assumption overstates revenue by 49 %. Period retention rising over time is usually a sorting effect of heterogeneity: the sBG reproduces Fader and Hardie's parameters exactly and shows that "multiplying retention rates" understates expected lifetime by about a third. Simpson's paradox is a different thing: every subgroup gets worse while the pooled figure improves. (https://5dlibrary.com/answers/en/algorithmic-marketing-shapley-attribution-funnels/c05/) 第5章 一位客户还值多少钱:BG/NBD、Gamma-Gamma 与留存率的假象 — BG/NBD 用四个参数描述非契约客户的购买与流失,只依赖频率与最近一次购买;它的期望公式中的 $_2F_1$ 在长期预测时级数收敛很慢,用普法夫变换加变量代换后的欧拉积分可以稳定计算,两种算法互相核对。Gamma-Gamma 模型给出收缩后的期望客单价,但假设金额与频率独立;在常客花得少的虚构例子中,违反这一假设使收入高估 49%。逐期留存率随时间上升,通常是异质性造成的分选效应:sBG 精确重现了费德与哈迪的参数,并显示「留存率连乘」把期望存续年数低估了约三分之一。辛普森悖论是另一件事:子群都变差而合并后变好。 (https://5dlibrary.com/answers/zh/algorithmic-marketing-shapley-attribution-funnels/c05/) - Ch.6 Diffusion and Spend: the Bass Model, the Dorfman–Steiner Condition and Whole-Life Budgets — The Bass equation $dF/dt = (p+qF)(1-F)$ checks out, with peak time $\ln(q/p)/(p+q)$; before the peak, market size and the imitation coefficient are hard to separate, and early forecasts can be off by a factor of two or more. The Dorfman–Steiner condition (advertising/sales = advertising elasticity/price elasticity) checks out, but it is a static condition and cannot by itself give a whole-life allocation. The author's whole-life model shows that with concave response and word-of-mouth diffusion the optimal spend is front-loaded and earns 8.3 % more than even spending of the same budget, while a fixed share of sales does worst of the three policies with that budget (though still better than no advertising); under the concave response pulsing is always worse than even spending, and under the example's S-shaped response pulsing helps only at relatively low monthly budgets (up to 30 thousand in this example), with a switch that does not coincide with the inflection point. (https://5dlibrary.com/answers/en/algorithmic-marketing-shapley-attribution-funnels/c06/) 第6章 扩散与买量:巴斯模型、多夫曼—斯坦纳条件与全周期支出 — 巴斯方程 $dF/dt = (p+qF)(1-F)$ 核对无误,峰值时刻 $\ln(q/p)/(p+q)$;在峰值之前,市场规模与模仿系数难以分开估计,早期预测可能差一倍以上。多夫曼—斯坦纳条件(广告/销售额 = 广告弹性/价格弹性)核对无误,但它是静态条件,不能直接给出全周期分配。作者的全周期模型显示,在凹响应与口碑扩散下,最优支出前重后轻,比同预算的平均投放多获利 8.3%,而按销售额比例投放在同预算的三种投放中最差(仍优于不投放);在凹响应下脉冲投放总不如平均投放,在示例的 S 形响应下,脉冲只在月均预算相对较低(本例不超过 30 千)时有利,分界不等于拐点。 (https://5dlibrary.com/answers/zh/algorithmic-marketing-shapley-attribution-funnels/c06/) - Ch.7 Reading the Classics: Shi and Jie in the Sunzi — With goodwill depreciating 1 % a day (half-life 69 days), the last 7 days take only 29.0 % of the budget, and the result hardly differs from even spending (goodwill on the event day is only 0.4 % higher); at 5 % a day (half-life 13.5 days), the last 7 days take 53.7 % and goodwill is 9.1 % higher than with even spending; at 15 %, the last 7 days take 89.7 % and goodwill is 57.2 % higher; at 40 % (half-life 1.4 days), almost the whole budget falls in the last week and goodwill is 2.739 times that of even spending. A numerical search agrees with the analytic solution at both depreciation rates checked (ratio 1.0000). (https://5dlibrary.com/answers/en/algorithmic-marketing-shapley-attribution-funnels/c07/) 第7章 原典并读:《孙子兵法》的「势」与「节」 — 商誉每天折旧 1%(半衰期 69 天)时,最后 7 天只占预算的 29.0%,与平均投放几乎没有差别(活动日的商誉只高 0.4%);每天折旧 5%(半衰期 13.5 天)时,最后 7 天占 53.7%,商誉比平均投放高 9.1%;折旧 15% 时,最后 7 天占 89.7%,商誉高 57.2%;折旧 40%(半衰期 1.4 天)时几乎全部预算都在最后一周,商誉是平均投放的 2.739 倍。数值搜索在两种折旧率下都与解析解一致(比值 1.0000)。 (https://5dlibrary.com/answers/zh/algorithmic-marketing-shapley-attribution-funnels/c07/) - Ch.8 Practice and a Pre-Registered Test — Chapter 8 gives a four-step checklist (attribution, experiments, lifetime value, budget) and a pre-registered simulation benchmark: on a synthetic multichannel data generator with known true increments, frozen by an independent third party, observational Shapley attribution is compared with last click on their error against the true incremental shares (single primary outcome the difference in mean absolute errors, smallest effect of interest 2 percentage points, 79 synthetic worlds for 90 % power, 90 planned, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. (https://5dlibrary.com/answers/en/algorithmic-marketing-shapley-attribution-funnels/c08/) 第8章 实践与预注册检验 — 第八章给出归因、实验、生命周期价值与预算四步核查表,以及一项预注册的模拟基准研究:在独立第三方冻结、已知真实增量的合成多渠道数据生成器上,比较观测沙普利归因与末次点击对真实增量份额的误差(唯一主要结局为二者平均绝对误差之差,最小关注效应 2 个百分点,90% 把握度需要 79 个合成世界、计划 90 个,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有数据都是虚构或模拟的。 (https://5dlibrary.com/answers/zh/algorithmic-marketing-shapley-attribution-funnels/c08/) ## 5DL-B-049 · Information Geometry, the Fisher Information Metric and Natural-Gradient Optimisation on Statistical Manifolds 中文书名:信息几何、费希尔信息度量与自然梯度优化的统计流形 DOI: https://doi.org/10.5281/zenodo.23836268 Cite: Fan, Y. (2026). Information Geometry, the Fisher Information Metric and Natural-Gradient Optimisation on Statistical Manifolds (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23836268 Book page: https://5dlibrary.com/books/information-geometry-fisher-natural-gradient/ Abstract: A family of probability distributions can be treated as a space. How "far apart" are two nearby distributions? Does that distance change if the same family is described with different parameters? In which direction should an optimiser descend so that the choice of coordinates does not matter? This book answers these questions with mathematical statistics, information theory and optimisation theory, and checks, qualifies or corrects the popular claims in the author's framework one by one. It covers the Fisher information as a Riemannian metric and as the second-order term of the KL divergence (verified numerically; KL is asymmetric and fails the triangle inequality, as does its square root), and the exact conditions of Čencov's uniqueness theorem (finite sample spaces, invariance under Markov embeddings, uniqueness up to a constant; extended to general sample spaces by Ay and co-authors in 2015), with statistics able only to lose Fisher information; the Cramér–Rao bound (Rao 1945, Cramér 1946, found independently by Fréchet and Darmois) and the mistaken idea that it is always attained: at finite n only the mean parameter of an exponential family is attained by an unbiased estimator, the biased SS/(n+1) and the James–Stein estimator have mean squared error below the bound, the error of the uniform-endpoint estimator falls like 1/n², and Hodges's superefficient estimator pays an unbounded local price; Jeffreys's 1946 prior, its invariance under reparametrisation, its improperness for a normal mean and a Poisson rate, its posterior concentrating at half the true variance in the Neyman–Scott problem, its different posteriors under binomial and negative-binomial sampling (a breach of the likelihood principle), and the good frequentist coverage of its intervals for a binomial proportion; Amari's 1998 natural gradient, whose invariance holds exactly only for the continuous-time flow, discrete steps in different coordinates differing by O(η) and greatly at step size 1; on ill-conditioned logistic regression the natural gradient (here Newton's method) takes the fewest steps but costs more than gradient descent when conditioning is good, costs more than Adam on a small neural network, the empirical Fisher mis-scales steps a hundredfold when the noise level is wrong, and the K-FAC approximation has relative error up to 0.67; the dually flat structure of exponential families, Legendre duality and Bregman divergences, the generalised Pythagorean theorem (whose direction cannot be reversed) and iterative proportional fitting as alternating projections; and Watanabe's singular learning theory: in a singular model such as y = abx the free-energy coefficient is the real log canonical threshold λ = 1/2, not BIC's d/2 = 1, and WBIC tracks the free energy without knowing λ, but singularity does not "explain" generalisation in deep learning. Chapter 7 reads these beside "changing and moving without rest, flowing through the six empty places" of the Great Commentary (Xici, part II) of the Book of Changes, Han Kangbo's gloss "the six empty places are the six positions" and Kong Yingda's subcommentary, and "the spirit has no fixed place and change has no fixed body" of part I, stating plainly that the parallel between positions and lines and between coordinates and objects is an analogy: the book does not claim that the Book of Changes contains differential geometry, nor treat hexagrams as statistical models. Chapter 8 gives a four-step checklist and a pre-registered computational benchmark: on a stratified suite of synthetic tasks frozen by an independent third party, natural-gradient optimisation (exact or K-FAC) is compared with Adam on wall-clock time to a target loss; the single primary outcome is the mean over tasks of the base-2 logarithm of the time ratio, the smallest effect of interest log₂1.25, 393 paired tasks for 90 % power and 470 planned after allowing for invalid tasks, with the decision rule written as code. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are synthetic or simulated and are no guarantee of the performance of any particular machine-learning system. 摘要:一族概率分布可以看作一个空间:两个相近的分布之间有多「远」?换一套参数来描述同一族分布,这个距离会不会变?优化算法沿着哪个方向下降才不依赖坐标的选择?本书用数理统计、信息论与优化理论回答这些问题,并把作者框架中的流行说法逐条核对、限定或改正。依次讨论:费希尔信息作为黎曼度量与 KL 散度的二阶项(程序逐级验证二阶展开;KL 不对称,也不满足三角不等式,连它的平方根也不满足),以及钦佐夫(Čencov)唯一性定理的准确条件(有限样本空间、在马尔可夫嵌入下不变、至多差一个常数;艾等人 2015 年推广到一般样本空间),统计量只能丢失而不能增加费希尔信息;克拉美—拉奥下界(拉奥 1945、克拉美 1946,弗雷歇与达尔穆瓦也独立得到)及「下界总能达到」之误:有限样本中只有指数族的均值参数可以被无偏估计恰好达到,有偏的 SS/(n+1) 与詹姆斯—斯坦估计量的均方误差低于下界,均匀分布端点的估计误差按 1/n² 下降,霍奇斯超有效估计量在局部付出无界的代价;杰弗里斯先验(1946)的重参数化不变性、它在正态均值与泊松率上不可积、在内曼—斯科特问题中后验集中在真值的一半、二项与负二项抽样下给出不同的后验(违反似然原理),以及它的区间在二项比例上有良好的频率覆盖;甘利俊一(Amari)1998 年的自然梯度:不变性只对连续时间的流严格成立,离散步长下不同坐标的结果相差 O(η),步长为 1 时差别很大;在条件数高的逻辑回归上自然梯度(此时即牛顿法)步数最少,但按计算量在条件良好时不如梯度下降,在一个小型神经网络上按计算量不如 Adam,经验费希尔在噪声估计不对时步长偏差百倍,K-FAC 近似的相对误差可达 0.67;指数族的对偶平坦结构、勒让德对偶与布雷格曼散度、广义勾股定理(方向不能颠倒)与迭代比例拟合的交替投影;渡边澄夫的奇异学习理论:在 y = abx 这样的奇异模型中自由能的系数是实对数典范阈 λ = 1/2 而不是 BIC 的 d/2 = 1,WBIC 无须知道 λ 就能跟上自由能,但奇异性并未「解释」深度学习的泛化。 第七章与《周易·系辞下》「变动不居,周流六虚」及韩康伯注「六虚,六位也」、孔颖达疏,以及《系辞上》「神无方而易无体」并读,明确这只是「位与爻」与「坐标与对象」之间的类比,不主张《周易》含有微分几何,也不把卦象当作统计模型。第八章给出四步核查表,以及一项预注册的计算基准试验:由独立第三方冻结一组分层的合成任务,比较自然梯度(精确或 K-FAC)与 Adam 达到目标损失的墙钟时间,唯一主要结局为二者时间比的以 2 为底的对数在任务上的均值,最小关注效应 log₂1.25,90% 把握度需 393 个配对任务、计入无效任务后计划 470 个,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中数据均为合成或模拟,不是对任何具体机器学习系统性能的保证。 Chapter answers: - Ch.1 Fisher Information: a Metric on Statistical Manifolds — The Fisher information $g_{ij} = E[\partial_i\ell\,\partial_j\ell]$ transforms as a tensor; it is a coordinate-free metric on parameter space and the second-order term of the KL divergence (a ratio of 0.9987 at $dp = 0.001$ in the program). The KL divergence is asymmetric and fails the triangle inequality, as does its square root; the Fisher–Rao geodesic distance is a distance. Čencov's theorem shows that on finite sample spaces the only metric invariant under Markov embeddings is the Fisher metric, up to a constant, and the result has been extended to general sample spaces; "unique" is relative to this invariance requirement. Statistics can only lose Fisher information; sufficient statistics lose none. (https://5dlibrary.com/answers/en/information-geometry-fisher-natural-gradient/c01/) 第1章 费希尔信息:统计流形上的度量 — 费希尔信息 $g_{ij} = E[\partial_i\ell\,\partial_j\ell]$ 按张量规律变换,是参数空间上与坐标无关的度量,也是 KL 散度的二阶项(程序中 $dp = 0.001$ 时比值为 0.9987)。KL 散度不对称、不满足三角不等式,平方根也不满足;费希尔—拉奥测地距离才是距离。钦佐夫定理说明在有限样本空间上,对马尔可夫嵌入不变的度量只有费希尔度量(至多差常数),这一结论已推广到一般样本空间;「唯一」是相对于这一不变性要求而言的。统计量只能丢失费希尔信息,充分统计量不丢失。 (https://5dlibrary.com/answers/zh/information-geometry-fisher-natural-gradient/c01/) - Ch.2 The Cramér–Rao Bound: Attained, Not Attained and "Beaten" — The Cramér–Rao bound (Rao 1945, Cramér 1946, found independently by Fréchet and Darmois) constrains the variance of unbiased estimators under regularity conditions. "Always attained" is false: exact attainment occurs only for the mean parameter of an exponential family (ratio 1.000 for the Bernoulli); the best unbiased estimator of an exponential rate is 1.250 times the bound at $n = 10$; the biased $SS/(n+1)$ is 0.909 times; James–Stein with $d = 10$ and $\theta = 0$ cuts total risk from about 10 to about 2; the "bound" for a uniform endpoint does not apply and the error falls like $1/n^2$; Hodges's estimator is superefficient at one point and pays an unbounded price nearby. (https://5dlibrary.com/answers/en/information-geometry-fisher-natural-gradient/c02/) 第2章 克拉美—拉奥下界:达到、达不到与「低于」 — 克拉美—拉奥下界(拉奥 1945、克拉美 1946,弗雷歇、达尔穆瓦独立得到)在正则条件下约束无偏估计的方差。「总能达到」是错的:恰好达到只发生在指数族的均值参数上(伯努利比值 1.000);指数率参数的最好无偏估计在 $n = 10$ 时是下界的 1.250 倍;有偏的 $SS/(n+1)$ 为 0.909 倍,詹姆斯—斯坦在 $d = 10$、$\theta = 0$ 时把总风险从约 10 降到约 2;均匀分布端点的「下界」不适用,误差按 $1/n^2$ 下降;霍奇斯估计在一点上超有效,却在邻近处付出无界的代价。 (https://5dlibrary.com/answers/zh/information-geometry-fisher-natural-gradient/c02/) - Ch.3 Jeffreys's Prior: Invariance and Its Price — Jeffreys's prior $\sqrt{\det G}$ is the Fisher–Rao volume element, so applying it in any coordinates gives the same distribution (an interval probability of 0.164177 in both charts in the program). But it is not "non-informative and always proper": the rule gives improper priors for a normal mean and a Poisson rate; the joint form for several parameters makes the posterior in the Neyman–Scott problem concentrate, with ever smaller standard deviation, at half the true value (0.4934), while the independent form Jeffreys recommended does not (0.9870); it depends on the sampling design, binomial and negative-binomial sampling giving different posteriors in breach of the likelihood principle. On the other hand, its intervals for a binomial proportion have frequentist coverage close to the nominal level. Reference priors are a more systematic generalisation. (https://5dlibrary.com/answers/en/information-geometry-fisher-natural-gradient/c03/) 第3章 杰弗里斯先验:不变性及其代价 — 杰弗里斯先验 $\sqrt{\det G}$ 是费希尔—拉奥体积元,因而在任何坐标下应用都得到同一个分布(程序中两种坐标的区间概率都是 0.164177)。但它不是「无信息、总是合适」的:正态均值与泊松率的规则先验不可积;多参数的联合形式在内曼—斯科特问题中使后验以越来越小的标准差集中在真值的一半(0.4934),杰弗里斯本人建议的独立形式没有这个问题(0.9870);它依赖抽样设计,二项与负二项给出不同的后验,违反似然原理。另一方面,它的区间在二项比例上有接近名义水平的频率覆盖。参考先验是更系统的推广。 (https://5dlibrary.com/answers/zh/information-geometry-fisher-natural-gradient/c03/) - Ch.4 The Natural Gradient: Invariance, Speed and Approximation — The natural gradient $G^{-1}\nabla L$ is the steepest-descent direction under the Fisher metric. Its invariance holds exactly for the continuous-time flow and only to first order for discrete steps: in the program the gap between charts is $2.5\times10{-7}$ at step size 0.004 and $2.1\times10^6$ at step size 1. The framework's "proof" should read: when the Fisher equals the Hessian, the natural gradient is Newton's method. On ill-conditioned logistic regression it takes the fewest steps but costs more than gradient descent when conditioning is good; on a small network the exact natural gradient takes the fewest steps at about six times Adam's cost, and K-FAC misses the target. The empirical Fisher is not the Fisher: its scale changes a hundredfold with the true noise, and it stalls far from the optimum. Published benchmarks show no optimiser dominating on every task. (https://5dlibrary.com/answers/en/information-geometry-fisher-natural-gradient/c04/) 第4章 自然梯度:不变性、速度与近似 — 自然梯度 $G^{-1}\nabla L$ 是费希尔度量下的最陡下降方向。它的不变性对连续时间的流严格成立,对离散步只在一阶成立:程序中步长 0.004 时坐标间差距为 $2.5\times10{-7}$,步长 1 时达到 $2.1\times10^6$。作者框架所说的「证明」应改为:费希尔等于黑塞矩阵时自然梯度就是牛顿法。在病态逻辑回归上它的步数最少,但按计算量在条件良好时不如梯度下降;在一个小网络上,精确自然梯度步数最少而计算量约为 Adam 的六倍,K-FAC 未达到目标。经验费希尔不是费希尔:它的尺度随真实噪声改变一百倍,远离最优点时停滞。已发表的基准显示没有一个优化器在所有任务上占优。 (https://5dlibrary.com/answers/zh/information-geometry-fisher-natural-gradient/c04/) - Ch.5 Dual Flatness: Exponential Families, Legendre Duality and the Generalised Pythagorean Theorem — An exponential family has two coordinate systems, the natural parameters $\theta$ and the expectation parameters $\eta$, linked by the convex function $\psi$ and its Legendre dual $\phi$ (the negative entropy); the Fisher metric is the Hessian of $\psi$, and the KL divergence is the Bregman divergence of $\psi$ (both 0.360388 in the program). e-geodesics are geometric mixtures and m-geodesics arithmetic mixtures; the α-connections join the two in a family. The generalised Pythagorean theorem holds for "m-projection onto an e-flat submanifold" (or its dual), and the direction of the divergence cannot be reversed: in the program the correct direction agrees digit for digit, while the reversed one gives 1.412998 against 1.208860. Iterative proportional fitting is alternating I-projection, each step satisfying the Pythagorean equation; its limit is the minimum-KL solution, not the least-squares one, which can contain negative cells. (https://5dlibrary.com/answers/en/information-geometry-fisher-natural-gradient/c05/) 第5章 对偶平坦:指数族、勒让德对偶与广义勾股定理 — 指数族有自然参数 $\theta$ 与期望参数 $\eta$ 两套坐标,由凸函数 $\psi$ 与其勒让德对偶 $\phi$(负熵)联系;费希尔度量是 $\psi$ 的黑塞矩阵,KL 散度是 $\psi$ 的布雷格曼散度(程序中都是 0.360388)。e-测地线是几何混合,m-测地线是算术混合;α-联络把两者连成一族。广义勾股定理在「m-投影到 e-平坦子流形」(或其对偶)时成立,散度方向不能颠倒:程序中正确方向逐位相等,颠倒后 1.412998 对 1.208860。迭代比例拟合是交替的 I-投影,每一步都满足勾股等式;它的极限是最小 KL 解而不是最小二乘解,后者可以出现负的格子。 (https://5dlibrary.com/answers/zh/information-geometry-fisher-natural-gradient/c05/) - Ch.6 Singular Learning Theory: When the Fisher Matrix Degenerates — When parameters are not identifiable and the Fisher matrix degenerates, BIC's $\tfrac d2\log n$ is wrong: the leading terms of the Bayes free energy are $\lambda\log n - (m-1)\log\log n$, with $\lambda \le d/2$ the real log canonical threshold. Direct integration in the program gives $\lambda \approx 1/2$ for $y = abx$ (BIC uses 1) and $\lambda = 1/4$ for $y = a^2x$; the $\log\log n$ term depresses the slope at finite $n$. WBIC does not need $\lambda$, and its error against the exact free energy (0.27–0.80) is far smaller than BIC's (3.5–6.4). The claim that singularity explains generalisation in deep learning is not established: the theory describes the asymptotics of the Bayesian posterior, and its connection to stochastic-gradient training is still being studied. (https://5dlibrary.com/answers/en/information-geometry-fisher-natural-gradient/c06/) 第6章 奇异学习理论:当费希尔矩阵退化 — 当参数不可辨识、费希尔矩阵退化时,BIC 的 $\tfrac d2\log n$ 不对:贝叶斯自由能的主项是 $\lambda\log n - (m-1)\log\log n$,$\lambda \le d/2$ 是实对数典范阈。程序直接积分得到 $y = abx$ 的 $\lambda \approx 1/2$(BIC 用 1),$y = a^2x$ 的 $\lambda = 1/4$;$\log\log n$ 项使有限 $n$ 下的斜率偏低。WBIC 无须知道 $\lambda$,它与精确自由能的误差(0.27–0.80)远小于 BIC(3.5–6.4)。奇异性解释深度学习泛化的说法尚未确立:理论描述的是贝叶斯后验的渐近行为,与随机梯度训练的联系仍在研究中。 (https://5dlibrary.com/answers/zh/information-geometry-fisher-natural-gradient/c06/) - Ch.8 Practice and a Pre-Registered Test — Chapter 8 gives a four-step checklist and a pre-registered computational benchmark: on a stratified suite of synthetic tasks frozen by an independent third party, natural-gradient optimisation (exact or K-FAC) is compared with Adam on wall-clock time to a target loss; the single primary outcome is the mean over tasks of the base-2 logarithm of the time ratio, the smallest effect of interest log₂1.25, 393 paired tasks for 90 % power and 470 planned after allowing for invalid tasks, with the decision rule written as code. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are synthetic or simulated and are no guarantee of the performance of any particular machine-learning system. (https://5dlibrary.com/answers/en/information-geometry-fisher-natural-gradient/c08/) 第8章 实践与预注册检验 — 第八章给出四步核查表,以及一项预注册的计算基准试验:由独立第三方冻结一组分层的合成任务,比较自然梯度(精确或 K-FAC)与 Adam 达到目标损失的墙钟时间,唯一主要结局为二者时间比的以 2 为底的对数在任务上的均值,最小关注效应 log₂1.25,90% 把握度需 393 个配对任务、计入无效任务后计划 470 个,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中数据均为合成或模拟,不是对任何具体机器学习系统性能的保证。 (https://5dlibrary.com/answers/zh/information-geometry-fisher-natural-gradient/c08/) ## 5DL-B-050 · Mechanism Design, Vickrey–Clarke–Groves Auctions and Incentive-Compatible Community Engineering 中文书名:机制设计、维克里—克拉克—格罗夫斯(VCG)拍卖与数字社群激励相容工程 DOI: https://doi.org/10.5281/zenodo.23836274 Cite: Fan, Y. (2026). Mechanism Design, Vickrey–Clarke–Groves Auctions and Incentive-Compatible Community Engineering (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23836274 Book page: https://5dlibrary.com/books/mechanism-design-vcg-auctions-incentive-systems/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. Abstract: How can a rule get every participant to tell the truth, to put in effort and not to cheat? This fiftieth book of 5D Library uses mechanism design to examine auctions, advertising auctions and the points, badges and reputation rules of online communities, and separates what can be proved, what needs qualifying and the author's own models. The Vickrey second-price auction, revenue equivalence, the Myerson–Satterthwaite theorem and the Clarke pivot mechanism, treated in Book 27, are only recapped. The book covers the dominant-strategy and Bayesian versions of the revelation principle with proofs, and the commitment they rely on (a seller who cannot promise not to cut her price earns, over 1,000 periods with discount factor 0.99, only 18.2 % of the commitment revenue); multi-item VCG as pricing of externalities (in a three-item package auction 20,000 misreports are all unprofitable, yet the revenue of 12 lies below the core bound of 13), and false-name bidding, the Sybil attack of auctions: a bidder splitting into two identities cuts his payment from 8 to 0, profitable splits exist in 206 of 600 random markets with complements and package bidders and in 0 of 600 when all bidders are substitutes; the generalized second-price (GSP) auction, where truth-telling is not an equilibrium and locally envy-free equilibria earn between the VCG revenue and an upper bound (920 to 1,000 in the example); Myerson's optimal auction, virtual values, the reserve price that does not depend on the number of bidders, the Bulow–Klemperer theorem and ironing for non-regular distributions (in the example the optimal auction earns 1.4–3.3 % more than a second-price auction with the best reserve); Hurwicz's 1972 impossibility in exchange economies, the Green–Laffont uniqueness of Groves mechanisms and their budget imbalance, and Roberts' affine-maximiser theorem (the max-min rule violates weak monotonicity in 16 of 3,000 type pairs, so no payments can implement it truthfully); points inflation (the author's model, with the quantity theory of money as an analogy), behavioural evidence on badges, and why symmetric reputation functions cannot resist Sybils (five fake identities lift an attacker's PageRank rank from 161 to 9 while his max-flow from trusted seeds is unchanged). Chapter 7 reads these beside "one reward, one punishment, one teaching" in the "Rewards and Punishments" chapter of the Book of Lord Shang, "one hare runs, a hundred chase it" in its "Fixing Shares" chapter and in the Lüshi chunqiu, and the phrase "rewards reliable, punishments certain" (xin shang bi fa) in the Han Feizi and the History of the Former Han; it finds that the four-character phrase does not occur in the Book of Lord Shang and explains why Legalist reward and punishment and incentive compatibility share a name but not a meaning. Chapter 8 gives a four-step checklist and a pre-registered incentivised laboratory experiment: participants randomised to GSP or VCG position auctions, a single primary outcome (the share of rounds with truthful bids), a smallest effect of interest of 10 percentage points, 150 analysable and 168 randomised participants per arm for 90 % power, run by independent researchers, with the decision rule written as code. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. 摘要:一条规则怎样才能让每个参与者说真话、出力而不作弊?本书是 5D Library 第 50 本,用机制设计考察拍卖、广告竞价与线上社群的积分、徽章和声誉规则,并把能证明的、需要限定的与作者的模型分开。第 27 本已讨论的维克里二价拍卖、收益等价、迈尔森—萨特思韦特定理与克拉克枢纽机制只作回顾。本书依次讨论:显示原理的占优策略与贝叶斯两种版本及其证明,以及它依赖的承诺能力(一位卖家不能承诺不降价时,1,000 期、贴现因子 0.99 的收入只有承诺时的 18.2%);多物品 VCG 作为外部性定价(三件物品的组合拍卖中 20,000 次误报都无利可图,但收入 12 低于核心下限 13),以及假名竞价——拍卖中的女巫攻击:一个出价者分身为两个身份可把支付从 8 降到 0,在含互补与组合出价的随机市场中 206/600 个存在有利的分身,所有出价者都是替代品时为 0/600;广义二价拍卖(GSP)中说真话不是均衡,局部无嫉妒均衡的收入介于 VCG 与上界之间(示例 920 至 1,000);迈尔森最优拍卖的虚拟估值、与人数无关的最优保留价、布洛—克伦佩雷尔定理,以及非正则分布的熨平(示例中最优拍卖比最佳保留价二价拍卖多收 1.4%—3.3%);赫维茨 1972 年的交换经济不可能性、格林—拉丰的格罗夫斯唯一性与预算不平衡、罗伯茨仿射最大化定理(最大最小规则在 3,000 对类型中出现 16 次弱单调性违反,因而无论怎样定价都不能如实实施);积分通胀(作者的模型,借货币数量论作类比)、徽章的行为证据,以及对称声誉函数不能防女巫(5 个假身份把攻击者的 PageRank 排名从 161 提到第 9,种子最大流不变)。 第七章与《商君书·赏刑》「壹赏、壹刑、壹教」、《定分》与《吕氏春秋·慎势》的「一兔走,百人逐之」,以及《韩非子·外储说右上》与《汉书·宣帝纪》赞中的「信赏必罚」并读,核实「信赏必罚」四字不见于《商君书》,并写明法家赏罚与机制设计的激励相容同名不同义。第八章给出四步核查表与一项预注册的激励化实验室实验:随机分配到 GSP 或 VCG 广告位拍卖,唯一主要结局为如实出价轮次的比例,最小关注效应 10 个百分点,90% 把握度每组可分析 150 人、随机化 168 人,由独立研究人员实施,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有数据都是虚构或模拟的。 Chapter answers: - Ch.1 The Revelation Principle, Incentive Compatibility and Commitment — The revelation principle says that the equilibrium outcome of any mechanism can be obtained from a direct mechanism in which people report truthfully; the proofs of the dominant-strategy and Bayesian versions both "load" the equilibrium strategies into the machine. It lets mechanism design study only incentive-compatible direct mechanisms, but it does not make the equilibrium unique, relies on the designer's ability to commit, asks for complete types, and does not ensure that participants can see the dominant strategy. The program shows that the direct mechanism built from a first-price auction is only Bayesian incentive compatible, and that a seller who cannot commit not to cut her price earns, with many periods and cheap waiting, only a small part of the commitment revenue (18.2 % with $d = 0.99$ and 1,000 periods). (https://5dlibrary.com/answers/en/mechanism-design-vcg-auctions-incentive-systems/c01/) 第1章 显示原理、激励相容与承诺 — 显示原理说:任何机制的均衡结果,都可以由一个让人如实报告的直接机制得到;占优策略版本与贝叶斯版本的证明都是把均衡策略「装进」机器。它让机制设计可以只研究激励相容的直接机制,但它不保证均衡唯一,依赖设计者的承诺能力,要求报告全部类型,也不保证参与者看得出占优策略。程序显示,由一价拍卖组合成的直接机制只是贝叶斯激励相容;不能承诺不降价的卖家,在多期、低等待成本时只能得到承诺收入的一小部分($d = 0.99$、1,000 期时为 18.2%)。 (https://5dlibrary.com/answers/zh/mechanism-design-vcg-auctions-incentive-systems/c01/) - Ch.2 Multi-Item VCG: Pricing Externalities, the Core and False-Name Bids — VCG makes each person pay the externality he imposes on the others and thereby makes truth-telling a dominant strategy; the package auction in the program confirms this. But its revenue can fall below the core, giving the seller and losers a reason to bypass the auction (12 against 13 in the example); core-selecting auctions repair this at the price of dominant-strategy incentive compatibility. More seriously, VCG is not false-name-proof: one person splitting into two identities can cut his payment from 8 to 0. The concavity condition of Yokoo and co-authors is sufficient for VCG to be false-name-proof, but it must hold over the whole domain of declarable types; checking the real bidders is not enough. (https://5dlibrary.com/answers/en/mechanism-design-vcg-auctions-incentive-systems/c02/) 第2章 多物品 VCG:外部性定价、核心与假名竞价 — VCG 让每个人支付自己给别人造成的外部性,由此使说真话成为占优策略;程序中的组合拍卖确认了这一点。但它的收入可能低于核心,使卖家与落败者有绕开拍卖的动机(示例中 12 对 13),选择核心的拍卖可以修补这一点,代价是放弃占优策略激励相容。更严重的是,VCG 不防假名竞价:一个人分身为两个身份可以把支付从 8 降到 0;横尾等人的凹性条件给出了 VCG 防假名的充分条件,但它必须对允许申报的整个类型域成立,只检查真实出价者是不够的。 (https://5dlibrary.com/answers/zh/mechanism-design-vcg-auctions-incentive-systems/c02/) - Ch.3 Position Auctions: Equilibria of the Generalized Second-Price Auction and VCG — In position auctions GSP charges the next bid and looks like a second-price auction, but truth-telling is generally not an equilibrium: in the example, the top bidder raises his payoff from 200 to 300 by dropping to slot 2, and in random instances truth-telling is an equilibrium in only 0.0–4.2 % of cases. Locally envy-free equilibria earn between the VCG revenue and an upper bound (920 to 1,000 in the example), the lowest reproducing the VCG outcome exactly; "GSP earns at least as much as VCG" is guaranteed only for this class of equilibria; other Nash equilibria can earn as little as 605, or more than VCG (up to 1,355). Laboratory evidence shows similar revenue and efficiency under the two rules. (https://5dlibrary.com/answers/en/mechanism-design-vcg-auctions-incentive-systems/c03/) 第3章 广告位拍卖:广义二价拍卖的均衡与 VCG — 在广告位拍卖中,GSP 按下一名的出价收费,看起来像二价拍卖,但说真话一般不是均衡:示例中第 1 名下移到第 2 名可从 200 增到 300,随机实例中说真话为均衡的比例只有 0.0%—4.2%。局部无嫉妒均衡的收入介于 VCG 与一个上界之间(示例中 920 到 1,000),最低者恰好等于 VCG 的结果;「GSP 收入不低于 VCG」只在这类均衡中有保证;其他纳什均衡的收入可以低到 605,也可以高于 VCG(最高 1,355)。实验室证据显示两种规则的收入与效率相近。 (https://5dlibrary.com/answers/zh/mechanism-design-vcg-auctions-incentive-systems/c03/) - Ch.4 Myerson's Optimal Auction: Virtual Values, Reserve Prices and Ironing — Myerson's lemma reduces incentive compatibility to "monotone allocation plus utility given by the envelope formula", which yields "expected revenue = expected virtual surplus". For regular distributions the optimal auction is a second-price auction with reserve $\psi^{-1}(0)$, independent of the number of bidders (0.5 for the uniform, 1 for the exponential); the gain from the reserve shrinks quickly with more bidders, and the Bulow–Klemperer theorem says one more bidder beats the optimal reserve. Non-regular distributions require ironing: in the program's example the optimal auction earns 1.4–3.3 % more than a second-price auction with the best reserve, and allocating by raw virtual values is not monotone and cannot be implemented. (https://5dlibrary.com/answers/en/mechanism-design-vcg-auctions-incentive-systems/c04/) 第4章 迈尔森最优拍卖:虚拟估值、保留价与熨平 — 迈尔森引理把激励相容化为「分配单调 + 效用由包络公式决定」,由此得到「期望收入 = 期望虚拟剩余」。分布正则时,最优拍卖是保留价为 $\psi^{-1}(0)$ 的二价拍卖,保留价与人数无关(均匀分布 0.5、指数分布 1);保留价的增益随人数迅速变小,而布洛—克伦佩雷尔定理说多一名出价者胜过最优保留价。分布非正则时要熨平:程序中的例子里,最优拍卖比最佳保留价的二价拍卖多收 1.4%—3.3%,而按原始虚拟估值分配的规则不单调、不可实施。 (https://5dlibrary.com/answers/zh/mechanism-design-vcg-auctions-incentive-systems/c04/) - Ch.5 Impossibility: Hurwicz, Green–Laffont and Roberts — Hurwicz's 1972 impossibility is a precise statement about two-person exchange economies: truth-telling, efficiency and individual rationality cannot all hold; in the program the Walrasian mechanism, no trade and dictatorship each lack one property. Green and Laffont proved that on an unrestricted domain only Groves mechanisms are efficient and dominant-strategy incentive compatible, and Groves mechanisms cannot balance the budget; Clarke taxes on a marginal project take about 42–50 % of total welfare, and returning them equally breaks incentives. Roberts proved that with at least three alternatives and unrestricted valuations, the only implementable rules are affine maximisers; the max-min rule violates weak monotonicity and no pricing can implement it truthfully. (https://5dlibrary.com/answers/en/mechanism-design-vcg-auctions-incentive-systems/c05/) 第5章 不可能性:赫维茨、格林—拉丰与罗伯茨 — 赫维茨 1972 年的不可能性是关于两人交换经济的精确结论:说真话、有效与个体理性不能兼得;程序中瓦尔拉斯机制、不交易与独裁各缺一项。格林—拉丰证明在不受限制的估值域上只有格罗夫斯机制有效且占优策略激励相容,而格罗夫斯机制不能预算平衡;克拉克税在边缘项目中约占总福利的 42%—50%,平均退还则破坏激励。罗伯茨证明在至少三个方案、估值不受限制时,可实施的规则只有仿射最大化;最大最小规则违反弱单调性,无论怎样定价都不能如实实施。 (https://5dlibrary.com/answers/zh/mechanism-design-vcg-auctions-incentive-systems/c05/) - Ch.6 Community Incentives: Points Inflation, Badges and Sybil Attacks — Points are a community currency: in the author's model, when issuance grows with membership and rewards do not, points lose value at the rate of membership growth (4.00 % a month in the example); expiry lowers the level but not the growth rate, a fixed total dilutes each member's share, and farming accelerates inflation. Badges change behaviour: observational research shows users accelerating as they near a threshold, and meta-analytic effects of gamification on learning are small (close to medium for cognitive outcomes), with only cognitive outcomes stable in rigorous studies; the simplified model shows that the best threshold depends on the badge's value and on retention. Symmetric reputation functions are not sybilproof: five fake identities lift an attacker from 161st to 9th under PageRank, while the maximum flow from trusted seeds is unaffected by fake identities. The "cost boundary" is the author's model: each fake identity must cost more than the average gain it brings for an attack to be deterred. (https://5dlibrary.com/answers/en/mechanism-design-vcg-auctions-incentive-systems/c06/) 第6章 社群激励:积分通胀、徽章与女巫攻击 — 积分是一种社群货币:按作者的模型,发放随人数增长而奖励不增长时,积分按人数增长的速度贬值(示例中月通胀 4.00%),过期只降低水平、不改变增长率,固定总量会稀释每人所得,刷分使通胀加速。徽章能改变行为:观测研究显示用户在接近阈值时加速,游戏化对学习的元分析效应为小(认知结局接近中等)、严格研究中只有认知结局稳定;简化模型显示最好的阈值取决于徽章价值与留存。对称的声誉函数不能防女巫:5 个假身份即可把攻击者的 PageRank 从第 161 位推到第 9 位,而从可信种子出发的最大流不受假身份影响。「成本边界」是作者的模型:每个假身份的成本要高于假身份带来的平均收益,才能阻止攻击。 (https://5dlibrary.com/answers/zh/mechanism-design-vcg-auctions-incentive-systems/c06/) - Ch.7 Reading the Classics: One Reward, Settled Shares and "Rewards Reliable, Punishments Certain" — What the "Rewards and Punishments" chapter of the Book of Lord Shang says is "one reward, one punishment, one teaching": rewards come only from military merit, punishment knows no ranks and merit does not offset faults, and "clear rewards need not be costly, clear punishments need not kill". "One hare runs, a hundred chase it" in "Fixing Shares" and in Shen Dao's words quoted in the Lüshi chunqiu differ in wording but agree in meaning: an unclaimed valuable thing provokes a contest, and the author's model (Tullock's contest) shows 99 % of its value dissipated when a hundred people compete, and nothing dissipated when ownership is settled. "Rewards reliable, punishments certain" does not occur in the Book of Lord Shang; its earliest source is Hu Yan's words in the Han Feizi, "Outer Collections of Sayings, Right, Upper", and it recurs in the eulogy of the "Annals of Emperor Xuan". Legalist reward and punishment serves the ruler, allows no exit and binds the subjects rather than the ruler; the incentive compatibility of mechanism design rests on the participant's own best choice and voluntary participation. The two share a name but not a meaning, and the similarity is only structural. (https://5dlibrary.com/answers/en/mechanism-design-vcg-auctions-incentive-systems/c07/) 第7章 原典并读:壹赏、定分与信赏必罚 — 《商君书·赏刑》的原话是「壹赏,壹刑,壹教」:赏只出于战功一途,刑不分等级、功过不相抵,「明赏不费,明刑不戮」。《定分》与《吕氏春秋·慎势》所引慎子的「一兔走,百人逐之」字句不同,意思相同:归属未定的有价物引起争夺,作者的模型(图洛克竞赛)显示 100 人争夺时 99% 的价值被耗散,界定归属则不耗散。「信赏必罚」不见于《商君书》,较早的出处是《韩非子·外储说右上》狐偃之语,又见《汉书·宣帝纪》赞。法家的赏罚服务于君主、参与不可退出、约束的是臣民而不是君主;机制设计的激励相容以参与者自身的最优选择与自愿参与为前提。二者同名不同义,相似只在结构层面。 (https://5dlibrary.com/answers/zh/mechanism-design-vcg-auctions-incentive-systems/c07/) - Ch.8 Practice and Pre-Registered Testing — Chapter 8 gives a four-step checklist and a pre-registered incentivised laboratory experiment: participants randomised to GSP or VCG position auctions, a single primary outcome (the share of rounds with truthful bids), a smallest effect of interest of 10 percentage points, 150 analysable and 168 randomised participants per arm for 90 % power, run by independent researchers, with the decision rule written as code. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; all data in the book are fictional or simulated. (https://5dlibrary.com/answers/en/mechanism-design-vcg-auctions-incentive-systems/c08/) 第8章 实践与预注册检验 — 第八章给出四步核查表与一项预注册的激励化实验室实验:随机分配到 GSP 或 VCG 广告位拍卖,唯一主要结局为如实出价轮次的比例,最小关注效应 10 个百分点,90% 把握度每组可分析 150 人、随机化 168 人,由独立研究人员实施,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中所有数据都是虚构或模拟的。 (https://5dlibrary.com/answers/zh/mechanism-design-vcg-auctions-incentive-systems/c08/) ## 5DL-B-051 · Neurobiology of Ritual, Spatial Boundary Transitions and Computational Biomedicine of Self-Transcendence 中文书名:仪式神经学、空间边界切换与自我转移的计算生物医学 DOI: https://doi.org/10.5281/zenodo.23839910 Cite: Fan, Y. (2026). Neurobiology of Ritual, Spatial Boundary Transitions and Computational Biomedicine of Self-Transcendence (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23839910 Book page: https://5dlibrary.com/books/ritual-neurorestoration-spatial-transition-embodied/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. Abstract: Can ritual reduce anxiety? Can walking through a door "clear" stress? Can slow breathing restore endocrine and immune function through the vagus nerve? This book presents mathematical models and a literature review for teaching and research; it is not medical advice, none of its practices is a treatment for any disease, and any symptom should first be assessed by a doctor. Keeping human evidence, animal and mechanistic evidence and the author's models apart, it covers: anxiety making spontaneous movement more repetitive and rigid (Lang and colleagues 2015: anxiety elicits ritual, which is not ritual reducing anxiety), small experiments on ritual and performance anxiety and the inflation of published effects (in simulation, a true effect of 0.2 with 20 per group is overstated 3.96-fold), and the author's precision-weighted cascade model (bias is damped when g(1−w) < 1) with its dark-room loophole; an illustrative HPA model solved by Runge–Kutta (in this model the diurnal cortisol slope flattens mainly with an evening load while feedback resistance steepens it; in human meta-analysis a flatter slope is associated with poorer health at r = 0.147), resonance breathing in the baroreflex loop (resonance moves between 4.92 and 8.12 breaths a minute with the delay; RMSSD is relatively insensitive to slow oscillations but still changes clearly with breathing during the recording), and evidence for the cholinergic anti-inflammatory pathway that comes mainly from animals and implanted stimulators; the doorway effect as reduced accessibility of object memory after an event boundary, not restored by returning and dependent on the number of new event models, and found in only one of four replication experiments (virtual reality with a working-memory load, as more false alarms), with grid cells, border cells and global remapping as mainly rodent findings and an attractor model showing abrupt switches of representation; a Wilson–Cowan-type triple-network model in which salient input must exceed a threshold to switch states and later task drive decides whether the switch lasts; open and hidden administration (Colloca, Lopiano, Lanotte and Benedetti 2004), a Bayesian placebo model, regression to the mean (a 1.42-point "improvement" without treatment), conditioned immunosuppression and Rescorla–Wagner learning; multisensory inverse effectiveness and its statistical artefacts, the non-thalamic main olfactory route and its thalamic side route, respiration-entrained limbic oscillations, and "significant" phase–amplitude coupling produced by a sharp waveform with no gamma oscillation (z = 103.4); and a screening model of costly signals (ritual cost helps only when free-riders are common enough, and the best cost is the smallest that screens). Chapter 7 reads these beside the Suwen chapter on shifting the essence and changing the qi (with its context, "in the present age … invocation can no longer cure"), the nine kinds of qi of the Suwen "On Pain", the Lingshu "Root of the Spirit", and the vigils of the Book of Rites ("complete vigil within, relaxed vigil without" in the Jiyi; "seven days … three days" in the Jitong), and gives the author's two-dimensional model of organic × functional load: symptoms alone cannot separate the two. Chapter 8 gives a four-step checklist and a pre-registered three-arm randomised controlled trial to be run by an independent institution (after the TSST, single primary comparison spatial transition + ritual + slow breathing versus slow breathing in the same room, single primary outcome ln RMSSD with spontaneous breathing over minutes 15–20, analysis of covariance, smallest effect of interest 0.15, 145 per arm for 90 % power and 495 recruited, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. 摘要:仪式能不能降低焦虑?走过一道门能不能「清空」压力?慢呼吸能不能经迷走神经修复内分泌与免疫?本书是教学与研究用的数学模型与文献评述,不是医疗建议,书中的练习都不是疾病的治疗方案,任何症状都应先由医生评估。它把人体证据、动物与机制证据和作者的模型分开,依次讨论:焦虑使自发动作更重复、更刻板(朗等人 2015,这是「焦虑引出仪式」而不是「仪式降低焦虑」),仪式降低表演焦虑的小样本实验与已发表效应的高估(模拟中真实效应 0.2、每组 20 人时被高估 3.96 倍),以及作者的精度加权级联模型(偏差被压住的条件 g(1−w) < 1)与它的暗室漏洞;一个用龙格—库塔法求解的示意 HPA 模型(在本模型中皮质醇日斜率变平主要来自晚间负荷,反馈抵抗反而使斜率变陡;人体元分析中斜率变平与较差健康的关联为 r = 0.147),压力反射环路的共振呼吸(共振频率随延迟在每分钟 4.92 至 8.12 次之间移动;RMSSD 对慢振荡相对不敏感,但仍会随记录时的呼吸方式明显改变),以及胆碱能抗炎通路主要来自动物与植入式刺激的证据;门口效应是事件边界后物体记忆可及性的下降,返回原房间不能恢复、成绩取决于新事件模型的数目,在四项重复实验中只有一项(虚拟现实加工作记忆负荷)出现效应且表现为误报增多,网格细胞、边界细胞与全局重映射主要是啮齿类动物的发现,作者的吸引子模型显示表征可以突然切换;威尔逊—考恩型三网络模型中突显输入超过门槛才能切换状态,切换能否保持取决于之后的任务驱动;公开—隐蔽给药(科洛卡、洛皮亚诺、拉诺特与贝内代蒂 2004)、贝叶斯安慰剂模型、回归均值(无治疗时「改善」1.42 分)、条件化免疫抑制与雷斯科拉—瓦格纳学习;多感官逆效应及其统计假象、嗅觉的非丘脑主通路与经丘脑旁路、呼吸夹带的边缘系统振荡,以及没有 γ 振荡的尖锐波形也能产生「显著」的相位—振幅耦合(z = 103.4);高代价信号的筛选模型(仪式代价只在搭便车者足够多时有用,最优代价是能筛选的最小代价)。 第七章与《素问·移精变气论》(连同「当今之世……祝由不能已也」的上下文)、《素问·举痛论》的九气、《灵枢·本神》以及《礼记·祭义》「致齐于内,散齐于外」与《祭统》「散齐七日……致齐三日」并读,并给出作者的器质负荷 × 功能负荷二维边界模型:仅凭症状无法分辨两种负荷。第八章给出四步核查表与一项由独立机构实施的预注册三臂随机对照试验方案(TSST 应激后,唯一主要比较为空间切换 + 仪式 + 慢呼吸对同室慢呼吸,唯一主要结局为第 15–20 分钟自然呼吸下的 ln RMSSD,协方差分析,最小关注效应 0.15,90% 把握度下每臂 145 人、共招募 495 人,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中数据除注明的文献数据外均为合成数据。 Chapter answers: - Ch.1 More Than Superstition, Less Than a Cure-All: Anxiety, Rigid Action and Precision Weighting — "Ritual is mere superstition" can be falsified: Lang and colleagues showed that anxiety makes movement more repetitive and rigid (partly replicated in the pre-registered study of Karl and Fischer), and Brooks and colleagues and Norton and Gino showed that ritual is sometimes followed by lower anxiety and better performance, though Karl and Fischer's pre-registered study found essentially null results in the "stress reduction" direction. But the first finding is "anxiety elicits ritual", and the second rests on small samples and depends on instructions, with no large pre-registered replication yet; simulation shows that in such conditions published effects may be several times too large. The author's precision-weighting model states the condition $g(1-w) < 1$ under which, in this model, a rigid sequence damps bias, and it also exposes the dark-room loophole: the same mechanism makes one underestimate the uncertainty of the world. Ritual is more than superstition, and less than a cure-all. (https://5dlibrary.com/answers/en/ritual-neurorestoration-spatial-transition-embodied/c01/) 第1章 仪式不只是迷信,也不是万能的:焦虑、刚性动作与精度加权 — 「仪式只是迷信」可以证伪:朗等人的实验显示焦虑使动作更重复、更刻板(卡尔与费希尔的预注册研究部分重复了这一方向),布鲁克斯等人与诺顿和吉诺的实验显示仪式有时伴随较低的焦虑与较好的表现,但卡尔与费希尔的预注册研究在「降低应激」方向上基本是零结果。但前者说的是「焦虑引出仪式」,后者的样本小、依赖说明语,尚缺大样本预注册重复;模拟显示,在这种条件下已发表的效应可能被高估数倍。作者的精度加权模型说明了在本模型中刚性序列压住偏差的条件 $g(1-w) < 1$,同时暴露了它的暗室漏洞:同样的机制也会让人低估外界的不确定性。仪式不只是迷信,也不是万能的。 (https://5dlibrary.com/answers/zh/ritual-neurorestoration-spatial-transition-embodied/c01/) - Ch.2 Stress Axis, Breathing and Inflammation: Which Parts Are Equations and Which Are Evidence — Allostatic load is a multi-marker index, not a variable in an equation; the book's HPA equations are an illustrative model, and the RK4 numerical error is negligible. In this model the diurnal cortisol slope flattens mainly with an evening load (from −0.137 to −0.030), while feedback resistance steepens it; in human meta-analysis a flatter slope is associated with poorer health with $r = 0.147$, an association. Resonance breathing comes from the delay of the baroreflex loop; it is close to, but not the same as, Mayer waves, and resonance frequencies differ between people; RMSSD is relatively insensitive to slow oscillations (it changes far less than SDNN), yet breathing during the recording still changes it clearly. Evidence for the cholinergic anti-inflammatory pathway comes mainly from animals and implanted stimulators; evidence that breathing practice lowers inflammatory cytokines in people is thin; the kinetics of endogenous peptides in this book are only illustrative. (https://5dlibrary.com/answers/en/ritual-neurorestoration-spatial-transition-embodied/c02/) 第2章 应激轴、呼吸与炎症:哪些是方程,哪些是证据 — 异态负荷是一个多指标指数,不是方程变量;本书的 HPA 方程是示意模型,RK4 数值误差可忽略。在本模型中,皮质醇日斜率变平主要来自晚间负荷(−0.137 变为 −0.030),而反馈抵抗使斜率变陡;人体元分析中斜率变平与较差健康的关联为 $r = 0.147$,是相关。共振呼吸来自压力反射环路的延迟,与迈耶波相近但不相同,共振频率因人而异;RMSSD 对慢振荡相对不敏感(变化远小于 SDNN),但记录时的呼吸方式仍能明显改变它。胆碱能抗炎通路的证据主要来自动物与植入式刺激,呼吸练习降低人体炎症因子的证据薄弱;内源性肽的动力学在本书只是示意。 (https://5dlibrary.com/answers/zh/ritual-neurorestoration-spatial-transition-embodied/c02/) - Ch.3 Walking Through a Door: the Doorway Effect, Event Boundaries and the Maps of the Hippocampus — Both models produce a doorway effect: accuracy falls from 0.943 to 0.838 in the reinstatement model and from 0.950 to 0.836 in the event model. They differ in the last two conditions: the reinstatement model predicts almost full recovery on returning (0.934), while the event model predicts that returning equals one shift (0.836); passing into two new rooms gives the lowest value in both models (0.697 and 0.796). The 2011 results of Radvansky and colleagues (no recovery on returning, performance close to one shift, the double shift worst) agree with the event model and not with the reinstatement model. The left panel of Figure 3 shows the contrast. (https://5dlibrary.com/answers/en/ritual-neurorestoration-spatial-transition-embodied/c03/) 第3章 走过一道门:门口效应、事件边界与海马的地图 — 两个模型都能产生门口效应:情境恢复模型中正确率从 0.943 降到 0.838,事件模型中从 0.950 降到 0.836。区别在后两个条件:情境恢复模型预测回到原房间几乎完全恢复(0.934),事件模型预测返回与换一次房间相同(0.836);连续进入两个新房间时,两个模型都给出最低值(0.697 与 0.796)。拉德万斯基等人 2011 年的结果——返回不恢复、与换一次房间相近、连续换房最差——与事件模型一致,与情境恢复模型不一致。图 3 左画出了这一对比。 (https://5dlibrary.com/answers/zh/ritual-neurorestoration-spatial-transition-embodied/c03/) - Ch.4 From Rumination to the Present: the Triple-Network Model and Salience-Network Switching — Menon's triple-network model is an integrative framework; Sridharan and colleagues inferred from effective-connectivity analysis that the right fronto-insular cortex plays a part in network switching, and Hamilton and colleagues reviewed correlational evidence linking rumination to increased DMN connectivity. The author's Wilson–Cowan-type model shows that, in this model, salient input must exceed a threshold (the $D$-dominated state disappears above about 0.312; the slow ramp scan gives 0.324) to switch states, that whether the switch lasts depends on later task drive, and that with noise the input acts by shortening dwell times. "Minimal embodied presence" is not in the model; it is the author's interpretation. The DMN and attention networks discussed in Books 8 and 24 are only recalled. (https://5dlibrary.com/answers/en/ritual-neurorestoration-spatial-transition-embodied/c04/) 第4章 从反刍到当下:三网络模型与突显网络的切换 — 梅农的三网络模型是一个整合性的框架;斯里达兰等人从有效连接分析推断右侧额—岛叶在网络切换中起作用,汉密尔顿等人综述了反刍与 DMN 连接增强的相关证据。作者的威尔逊—考恩型模型显示,在本模型中突显输入需要超过门槛($D$ 主导态在 $I_S$ 超过约 0.312 时消失;慢斜坡扫描测得 0.324)才能切换状态,切换能否保持取决于之后的任务驱动,噪声使输入的作用表现为缩短停留时间。「最小具身在场」不在模型之中,是作者的解释。第 8 本与第 24 本讨论过的 DMN 与注意网络只作回顾。 (https://5dlibrary.com/answers/zh/ritual-neurorestoration-spatial-transition-embodied/c04/) - Ch.5 Expectation and Learning: Open and Hidden Administration, Placebo and Conditioned Immunity — The classic open–hidden study is the 2004 joint work of Colloca, Lopiano, Lanotte and Benedetti; the author's Bayesian model shows that, in this model, the open-minus-hidden difference is $(1-w)$ times the change in expectation, and that an unchanged expectation also "discounts" the drug's effect. Effects of open-label placebos are seen mainly in self-reported outcomes, and the evidence is of limited quality. Regression to the mean can produce an "improvement" of 1.42 points with no treatment at all. Ader and Cohen discovered conditioned immunosuppression in rats, and Goebel and colleagues observed a similar phenomenon in a small sample of healthy people; the Rescorla–Wagner model illustrates the roles of cue novelty, extinction, partial reinforcement and blocking, but conditioning sustains only part of a drug's effect and is no substitute therapy. (https://5dlibrary.com/answers/en/ritual-neurorestoration-spatial-transition-embodied/c05/) 第5章 期望与学习:公开—隐蔽给药、安慰剂与条件化免疫 — 公开—隐蔽给药的经典研究是科洛卡、洛皮亚诺、拉诺特与贝内代蒂 2004 年的合作研究;作者的贝叶斯模型显示,在本模型中公开与隐蔽之差等于 $(1-w)$ 乘以期望的改变,并且不变的期望也会使药物的效应「打折扣」。开放标签安慰剂的效应主要见于自评结局,证据质量有限。回归均值可以在没有任何治疗时产生 1.42 分的「改善」。阿德与科恩在大鼠中发现条件化免疫抑制,戈贝尔等人在小样本健康人中观察到类似现象;雷斯科拉—瓦格纳模型说明了新奇线索、消退、部分强化与阻断的作用,但条件化只能维持药物效应的一部分,不是替代疗法。 (https://5dlibrary.com/answers/zh/ritual-neurorestoration-spatial-transition-embodied/c05/) - Ch.6 Incense, Sound and Cost: Multisensory Integration, Breathing Rhythms and Costly Signals — Inverse effectiveness has a real non-linear source in single neurons, but grouping by observed values can also manufacture it (in the book's simulation, true enhancement is 100 % throughout, yet the groups show 115.1 % and 57.9 %). The main olfactory pathway has no thalamic relay, but a side route through the mediodorsal thalamus exists; terpenes have no special pathway. The intracranial recordings of Zelano and colleagues show that nasal breathing entrains limbic oscillations; the modulation index of phase–amplitude coupling can be faked by non-sinusoidal waveforms ($z = 103.4$ with no gamma oscillation). Costly signals correlate with cooperation in historical and field studies; the author's screening model shows that ritual cost helps only when free-riders are common enough, and that the best cost is the smallest that screens. (https://5dlibrary.com/answers/en/ritual-neurorestoration-spatial-transition-embodied/c06/) 第6章 香、声与代价:多感官整合、呼吸节律与高代价信号 — 逆效应在单神经元中有真实的非线性来源,但按观察值分组也能制造出它(本书模拟中真实增强都是 100%,分组后为 115.1% 与 57.9%)。嗅觉的主要通路不经丘脑中继,但存在经丘脑背内侧核的旁路;萜烯没有特殊通路。泽拉诺等人的颅内记录显示经鼻呼吸夹带边缘系统振荡;相位—振幅耦合的调制指数可以被非正弦波形伪造(无 γ 振荡时 $z = 103.4$)。高代价信号在历史与田野研究中与合作相关;作者的筛选模型显示,仪式代价只在搭便车者足够多时有用,最优代价是能筛选的最小代价。 (https://5dlibrary.com/answers/zh/ritual-neurorestoration-spatial-transition-embodied/c06/) - Ch.8 Practice and a Pre-Registered Test — Chapter 8 gives a four-step checklist and a pre-registered three-arm randomised controlled trial to be run by an independent institution (after the TSST, single primary comparison spatial transition + ritual + slow breathing versus slow breathing in the same room, single primary outcome ln RMSSD with spontaneous breathing over minutes 15–20, analysis of covariance, smallest effect of interest 0.15, 145 per arm for 90 % power and 495 recruited, decision rule written as code). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. (https://5dlibrary.com/answers/en/ritual-neurorestoration-spatial-transition-embodied/c08/) 第8章 实践与预注册检验 — 第八章给出四步核查表与一项由独立机构实施的预注册三臂随机对照试验方案(TSST 应激后,唯一主要比较为空间切换 + 仪式 + 慢呼吸对同室慢呼吸,唯一主要结局为第 15–20 分钟自然呼吸下的 ln RMSSD,协方差分析,最小关注效应 0.15,90% 把握度下每臂 145 人、共招募 495 人,判定规则写成代码)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;书中数据除注明的文献数据外均为合成数据。 (https://5dlibrary.com/answers/zh/ritual-neurorestoration-spatial-transition-embodied/c08/) ## 5DL-B-052 · Finger–Brain Cortical Magnification, Distal Microneedle Afference and Computational Neurology of Post-Stroke Cognitive Network Reconstruction 中文书名:指—脑皮层放大、远端微针传入与卒中后认知网络重建的计算神经学 DOI: https://doi.org/10.5281/zenodo.23839915 Cite: Fan, Y. (2026). Finger–Brain Cortical Magnification, Distal Microneedle Afference and Computational Neurology of Post-Stroke Cognitive Network Reconstruction (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23839915 Book page: https://5dlibrary.com/books/jingwell-microneedle-cortical-alpha-stroke-cognition/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. Abstract: Can press needles 0.9 mm long, taped to the twelve well points of both hands and rubbed daily after a stroke, improve cognition through dense fingertip afference? This book presents mathematical models and methodological discussion for teaching and research and is not medical advice. Its main empirical source is Lu Shunkang's 2024 master's thesis at Tianjin University of Traditional Chinese Medicine, and it falsifies before it sets bounds: the thesis's acupoint network is a co-occurrence network of 172 prescriptions; about 50.2 of the 65 Baihui–Sishencong co-occurrences come from the two points' frequencies, and a simulated database with no pairing rule gives a clustering coefficient of about 0.72–0.75, a diameter of 3 and the same five "core points", so the network reflects prescription habits, not physiological coupling; in the author's schematic brain-region network the energy with a single driver is finite but huge and its controllability Gramian has a condition number of 10¹²–10¹⁴, too ill-conditioned for double precision, and the factor by which "basal-ganglia damage" changes the energy ranges from 0.0045 to 3,460 across perturbed networks, so "an order of magnitude less energy" is ungrounded; glabrous fingertip skin has about 241 receptor units per cm² (palm 58), "15–30-fold cortical magnification" has no traceable source, and vPAG P2RX7 is evidence from brain-injured rats; stochastic resonance helps only with a fixed threshold, and extra afference unrelated to activity does not strengthen connections under spike-timing-dependent plasticity; with 50 % overlap and a Hann window the Welch variance is 1.054 times 1/K, not 1/K; in the Jansen–Rit model more external input does not restore a collapsed alpha peak; and six uncontrolled EEG recordings cannot test "alpha restoration". The trial itself was an open-label add-on design, and the thesis gives two between-group results for the MoCA: response rates p = 0.287 ("comparable efficacy") and post-treatment scores P = 0.005, the latter unadjusted for baseline, with an inconsistent control cell and no sham, so still not enough for a conclusion. A table-by-table check found that 56 of 60 integer-scale cells cannot be medians of 34 scores, two cells exceed the item maximum, an MMSE row duplicates colour-word data, the response table's counts disagree while p = 0.287 matches 26/34 against 22/34 responders, the efficacy formula runs in the wrong direction, and the sample-size calculation does not match the analysis; the book invites the thesis author to release the raw data. The drift-diffusion model shows that faster timed tests cannot by themselves be read as better cognition; under the MoCA ceiling, analysis of covariance has clearly more power than a response-rate chi-square test. Chapter 7 reads these beside the Lingshu chapters "Nine Needles and Twelve Sources" (所注为腧), "Root Transport Points" (the heart issues at Zhongchong; Shaochong is absent) and "Channels"; later pricking of the twelve well points to bleed was emergency treatment of stroke collapse, the wording of the Qiankun Shengyi could not be checked in print, and the history of Sishencong is in doubt. Chapter 8 gives a four-step checklist and a pre-registered three-arm trial: single primary comparison genuine against sham press needles (both rubbed three times a day), single primary outcome the MoCA at week 4 by analysis of covariance, smallest effect of interest 1.5 points, 72 analysable and 85 randomised per group, 255 in all, frontal alpha only as a pre-specified mediation analysis, and a decision rule written as code. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. 摘要:卒中后在双手十二井穴贴上长 0.9 毫米的揿针并每天按揉,能否借指端密集的感觉传入改善认知?本书是教学与研究用的数学模型与方法学讨论,不是医疗建议。它以卢顺康 2024 年天津中医药大学硕士学位论文为主要经验来源,先证伪、再立边界:论文的腧穴网络是 172 条处方的共现网络,百会—四神聪的 65 次共现中约 50.2 次来自两穴的频率,没有任何配伍规则的模拟处方库同样给出约 0.72–0.75 的聚类系数、直径 3 与相同的前五位「核心穴」,所以它反映处方习惯而非生理耦合;在作者的示意性脑区网络中,单点驱动的能量有限但极大,可控性格拉姆矩阵条件数达 10¹²–10¹⁴、双精度下无法可靠求逆,「基底节受损」使能量变化的倍数在扰动网络中从 0.0045 到 3,460 不等,「降低一个数量级」没有根据;指尖无毛皮肤约 241 个感受单位/cm²(手掌 58),「皮层放大 15–30 倍」未找到出处,vPAG P2RX7 是大鼠脑外伤证据;随机共振只在阈值固定时有益,脉冲时序依赖可塑性下与活动无关的额外传入不加强连接;Welch 估计在 50% 重叠加 Hann 窗时方差为 1/K 的 1.054 倍而非 1/K;Jansen—Rit 模型中增加外部输入不能恢复塌陷的 α 峰;6 例无对照脑电不能检验「α 恢复」。试验本身是开放标签加载设计,论文对 MoCA 给出两种组间结果:有效率 p = 0.287(「疗效相当」)与治疗后评分 P = 0.005,后者未调整基线、对照组数值不自洽、又无假针对照,仍不足以下结论。逐表核查发现:60 个整数计分格中 56 个不可能是 34 例的中位数,两格超出分项满分,一行 MMSE 与色词数据重复,有效率表计数不一致而 p = 0.287 对应有效 26/34 对 22/34,疗效公式方向相反,样本量计算与分析不匹配;本书邀请论文作者公开原始数据。漂移扩散模型说明计时测验提速不能单独读作认知改善;MoCA 天花板下协方差分析的把握度明显高于有效率卡方检验。 第七章与《灵枢》的《九针十二原》(「所注为腧」)、《本输》(心出于中冲,无少冲)、《经脉》并读,后世刺十二井穴放血用于中风昏仆急救,《乾坤生意》原文未能核到纸本,四神聪出处的先后关系存疑。第八章给出四步核查表与一项预注册三臂试验:唯一主要比较为真揿针对假揿针(两组都每天按揉三次),唯一主要结局为第 4 周 MoCA 的协方差分析,最小关注效应 1.5 分,每组 72 例可分析、85 例随机,共 255 例,前额 α 只作预先指定的中介分析,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;除注明的文献数据外,书中数据都是合成的。 Chapter answers: - Ch.1 Acupoint Co-occurrence Networks and Pinning Control of Brain Networks — The acupoint network of the thesis is a prescription co-occurrence network. Its weights are determined mainly by the frequency of each point: independent choice alone produces about 50 of the 65 Baihui–Sishencong co-occurrences, and a simulated database with no pairing rule yields a clustering coefficient of about 0.72–0.75, a diameter of 3 and the same five "core points" as the thesis. Some structure remains after removing frequency, but its shape is that of prescription schools, not physiological coupling. Controllability and pinning control can only be discussed on a schematic brain-region network; in the author's model single-node energies are finite but huge and cannot be inverted reliably in double precision, an extra driver always lowers the energy by a factor set by the weights, and the factor by which "damage" changes the energy spans nearly six orders of magnitude across perturbed networks, so "an order of magnitude less energy" is not a grounded conclusion. 64.7 % is the share of infarcts involving the basal ganglia, not a measured share of impaired conduction. (https://5dlibrary.com/answers/en/jingwell-microneedle-cortical-alpha-stroke-cognition/c01/) 第1章 腧穴共现网络与脑网络的牵制控制 — 论文的腧穴网络是处方共现网络。它的权重主要由各穴的使用频率决定:独立选择就能产生百会—四神聪 65 次中的约 50 次;一个没有任何配伍规则的模拟处方库,同样产生约 0.72–0.75 的聚类系数、直径 3,以及与论文相同的前五位「核心穴」。扣除频率后仍有可辨认的结构,但其形态是处方流派,而不是生理耦合。可控性与牵制控制只能放在示意性的脑区网络中讨论;在作者的模型中,单点驱动的能量有限但极大、双精度下无法可靠求逆,加驱动点必然降低能量但倍数取决于权重,「受损后能量变化」的倍数在扰动网络中跨越近六个数量级,所以「降低一个数量级」不是一个有根据的结论。64.7% 是梗死部位涉及基底节的比例,不是测得的传导受损比例。 (https://5dlibrary.com/answers/zh/jingwell-microneedle-cortical-alpha-stroke-cognition/c01/) - Ch.2 Press Needles, Skin and Cortical Magnification of the Fingertip — The press-needle protocol of the thesis consists of a needle and rubbing three times a day; the control group had neither. In the author's viscoelastic-skin model, stress after taping relaxes to about 35 % of its peak within tens of seconds, fast-adapting afference appears only during rubbing, and the contribution of 24 hours of static wear depends on an unmeasurable threshold and can range from zero to a lot. Piezo2 is the main transduction channel for fine touch, with human genetic evidence; glabrous fingertip skin has about 241 units per cm², about 4.2 times the palm; cortical magnification exceeds what peripheral density explains, but "15–30 times" has no source, and the well points are not in the middle of the finger pad. The effect of adaptive gain is non-monotonic; vPAG P2RX7 is mechanistic evidence from rats with brain injury. The sham press needle passed a blinding test in healthy young men and must be validated again for stroke patients. (https://5dlibrary.com/answers/en/jingwell-microneedle-cortical-alpha-stroke-cognition/c02/) 第2章 揿针、皮肤与指尖的皮层放大 — 论文的揿针方案由针体与每天三次的按揉组成,对照组两者都没有。在作者的黏弹性皮肤模型中,贴敷后的应力在几十秒内松弛到峰值的约 35%,快适应传入只在按揉时出现,静置 24 小时的贡献取决于无法测得的阈值,可以从零到很多。Piezo2 是精细触觉的主要转导通道,有人体遗传学证据;指尖无毛皮肤的感受单位密度约 241 个/cm²,是手掌的约 4.2 倍;皮层放大超过外周密度所能解释的部分,但「15–30 倍」没有出处,井穴也不在指腹中央。适应性增益的效果是非单调的;vPAG P2RX7 是大鼠脑外伤的机制证据。假揿针在健康年轻人中通过了设盲检验,用于卒中患者时需要重新验证。 (https://5dlibrary.com/answers/zh/jingwell-microneedle-cortical-alpha-stroke-cognition/c02/) - Ch.3 Cortico-Striato-Thalamic Loops, Stochastic Resonance and Plasticity — In the thesis, 64.7 % (44/68) of patients had infarcts involving the basal ganglia, but without a group of cognitively normal stroke patients this cannot show that lesion location is "closely related" to cognitive impairment; "胼底体" is probably a slip for the corpus callosum. Parallel cortico-striato-thalamic loops give an anatomical framework for "basal-ganglia lesions affect executive function", while "afference bypassing the striatum" reaches only primary somatosensory cortex anatomically. In the author's models, stochastic resonance helps only when a threshold is fixed and too high, and the information near the optimal noise is modest; under spike-timing-dependent plasticity, adding input unrelated to activity slightly weakens connections, and only causally paired input strengthens them. Neither mechanism has been measured, and neither is evidence of press-needle efficacy. (https://5dlibrary.com/answers/en/jingwell-microneedle-cortical-alpha-stroke-cognition/c03/) 第3章 皮层—纹状体—丘脑环路、随机共振与可塑性 — 论文中 64.7%(44/68)的患者梗死部位涉及基底节,但没有卒中后认知正常的对照组,不能说明病灶部位与认知障碍「密切相关」;「胼底体」疑为「胼胝体」之误。皮层—纹状体—丘脑并行环路为「基底节病灶影响执行功能」提供了解剖框架,而「绕过纹状体的去传入」在解剖上只到达初级体感皮层。在作者的模型中,随机共振只在阈值固定而偏高时才有好处,最佳噪声附近的信息量也很有限;脉冲时序依赖可塑性下,增加与活动无关的传入反而使连接略微减弱,只有因果配对的输入才使连接增强。两个机制都没有被测量过,都不能作为揿针疗效的证据。 (https://5dlibrary.com/answers/zh/jingwell-microneedle-cortical-alpha-stroke-cognition/c03/) - Ch.4 Alpha Oscillations: the Welch Estimate, the Jansen–Rit Model and Six EEG Recordings — The EEG part of the thesis is a six-person before–after comparison without control that shows one participant's spectrum and a qualitative description; the text also contradicts itself on eyes-closed and eyes-open recording and leaves the epoching unclear. "The variance of the Welch estimate is 1/k of the periodogram's" holds only for non-overlapping segments; with 50 % overlap and a Hann window it is 1.054 times 1/K and with 75 % overlap 1.909 times; but the spectrum of a single 5-minute recording is already stable, and the problem is the sample and the design. The Jansen–Rit model can write the trajectory "connectivity falls, alpha collapses, connectivity returns, alpha returns", but more external input does not restore alpha, and the spectral shape does not determine the parameters uniquely. Without any treatment effect, the probability that at least five of six people show "more alpha" can reach about 40 % with a common time drift, and the Berger effect is about threefold under the simulation's settings. "Alpha restored after stimulation" was not tested in the thesis. (https://5dlibrary.com/answers/en/jingwell-microneedle-cortical-alpha-stroke-cognition/c04/) 第4章 α 振荡:Welch 谱估计、Jansen—Rit 模型与六例脑电 — 论文的脑电部分是 6 例、无对照的自身前后比较,只展示了一位受试者的频谱与定性描述,文中还有闭眼与睁眼记录的矛盾、分段方法不明等问题。「Welch 估计的方差是周期图的 1/k」只对不重叠的分段成立;50% 重叠加 Hann 窗时为 1/K 的 1.054 倍,75% 重叠时为 1.909 倍;但单个 5 分钟记录的谱估计已经足够稳定,问题在于样本与设计。Jansen—Rit 模型可以写出「连接强度下降、α 塌陷、连接恢复、α 回来」的轨迹,但增加外部输入并不能恢复 α,而且谱形状不能唯一确定参数。没有治疗效应时,6 人中至少 5 人「α 升高」的概率在有共同时间漂移时可达约 40%,在本模拟设定下伯杰效应约为 3 倍。「刺激后 α 恢复」在论文中没有得到检验。 (https://5dlibrary.com/answers/zh/jingwell-microneedle-cortical-alpha-stroke-cognition/c04/) - Ch.5 Proposed Analyses: Microstates, the Phase Lag Index and Transfer Entropy — The thesis did no connectivity analysis, so this chapter only proposes a plan. Synthetic data show that zero-lag mixing alone can push the coherence of two independent sources to 0.87 while PLI and wPLI are unaffected; a genuine 12-ms lagged drive raises wPLI to 0.95. Transfer entropy can be misled by unobserved common drivers, zero-lag mixing and too coarse an embedding: a hidden driver produces "information flow" almost as large as a real drive. Microstates use the spatial information of all electrodes, but the number of classes and the state (eyes open or closed, drowsiness) change conclusions. Any connectivity measure should be fixed in advance, computed in source space and tested against surrogates; none can by itself prove "cognitive network reconstruction". (https://5dlibrary.com/answers/en/jingwell-microneedle-cortical-alpha-stroke-cognition/c05/) 第5章 建议的分析:微状态、相位滞后指数与传递熵 — 论文没有做连接分析,本章因此只给出建议的方案。合成数据显示,零时滞混合本身就能把两个独立源的相干推到 0.87,而 PLI 与 wPLI 不受影响;真实的 12 毫秒时滞驱动则使 wPLI 达到 0.95。传递熵会被未观测的共同驱动、零时滞混合与过于粗糙的嵌入误导:一个隐藏的驱动源产生的「信息流」与真实驱动几乎一样大。微状态利用全部电极的空间信息,但类别数与状态(睁眼/闭眼、困倦)都会改变结论。任何连接指标都应事先写定,在源空间计算,并配合替代数据检验;它们不能单独证明「认知网络重建」。 (https://5dlibrary.com/answers/zh/jingwell-microneedle-cortical-alpha-stroke-cognition/c05/) - Ch.6 Drift Diffusion, Ceiling Effects and a Check of the Thesis Data — Study 2 was an open-label, add-on, single-centre trial with 34 per group. The thesis gives two between-group results for the MoCA: response rates p = 0.287 ("comparable efficacy") and post-treatment scores P = 0.005; the latter is unadjusted for baseline, its control cell is inconsistent and there was no sham, so it still does not support a conclusion; "better than the control group" in the abstract partly rests on within-group before–after tests, and comparisons with baseline within groups cannot replace a comparison between groups. Faster timed tests can come from drift rate, boundary or non-decision time. The MoCA has a ceiling of 30; a response-rate chi-square test wastes most of the information, analysis of covariance has clearly higher power, and a Tobit model can correct the underestimate caused by the ceiling. The table-by-table check found that 56 of 60 integer-scale cells cannot be medians; four cells are impossible or hard to accept even as means (two above the item maximum); an MMSE row duplicates the colour-word data; the response table's counts disagree, while p = 0.287 matches 26/34 against 22/34 responders; the efficacy formula runs in the wrong direction; the sample-size calculation does not match the analysis; and the p columns are unlabelled, with two conflicting baseline p values. These problems need checking by the thesis author against the raw data. (https://5dlibrary.com/answers/en/jingwell-microneedle-cortical-alpha-stroke-cognition/c06/) 第6章 漂移扩散、天花板效应与论文数据核查 — 研究内容二是开放标签、加载设计、单中心、每组 34 例的试验。论文对 MoCA 给出两种组间结果:有效率 p = 0.287(「疗效相当」),治疗后评分 P = 0.005;后者未调整基线,对照组数值不自洽,又无假针对照,仍不足以下结论;摘要「疗效优于对照组」部分依据组内前后检验,而组内与基线比较不能代替组间比较。计时测验的提速可以来自漂移率、边界或非决策时间中的任何一个。MoCA 有 30 分的上限,有效率卡方检验浪费了大部分信息,协方差分析的把握度明显更高,托宾模型可以校正上限造成的低估。逐表核查发现:60 个整数计分格中 56 个不可能是中位数;四格即使作为均数也不成立或很难成立(两格超出分项满分);一行 MMSE 与色词数据重复;有效率表计数不一致,而 p = 0.287 对应有效 26/34 对 22/34;疗效公式方向相反;样本量计算与分析方法不匹配;p 值列未注明含义,两处基线 p 矛盾。这些问题需要论文作者用原始数据核对。 (https://5dlibrary.com/answers/zh/jingwell-microneedle-cortical-alpha-stroke-cognition/c06/) - Ch.7 Reading the Classics: Well Points, the Five Transport Points and Pricking the Twelve Wells — The Lingshu chapter "Nine Needles and Twelve Sources" uses the image of flowing water; "where the qi issues is the well" places the well point in the flow of channel qi, and the received text reads 所注为腧. In "Root Transport Points" the well point of the heart channel is Zhongchong, now assigned to the pericardium channel, and Shaochong is absent; "the twelve well points of the hands" is a later systematisation; ancient locations say only "at the tip of a finger", whereas the modern standard puts all but Zhongchong 0.1 cun from the nail corner. Later records of pricking the twelve well points to bleed concern emergency treatment of stroke collapse; the original wording of the Qiankun Shengyi could not be checked in print in this round and is only paraphrased. The thesis's order for Sishencong — "first in the Yinhai Jingwei, later corrected in the Taiping Shenghui Fang" — runs against the usual dating of the two books and needs checking. These ancient statements share only structural resemblances with modern neuroscience. (https://5dlibrary.com/answers/en/jingwell-microneedle-cortical-alpha-stroke-cognition/c07/) 第7章 原典并读:井穴、五输与刺十二井 — 《灵枢·九针十二原》以水流为喻,「所出为井」说明井穴在经气流注中的位置,通行本作「所注为腧」。《灵枢·本输》中心经的井穴是今天归于心包经的中冲,没有少冲,「手十二井穴」是后世的归纳;古代定位只说「某指之端」,现代标准除中冲外定在甲角旁 0.1 寸。后世刺十二井穴放血的记载用于中风昏仆的急救,《乾坤生意》的原文本轮未能核到纸本,只作转述。论文关于四神聪「最早见于《银海精微》、后在《太平圣惠方》中修正」的先后关系,与两书通常认定的成书年代相反,需要核对。这些古代论述与现代神经科学之间只有结构上的相似。 (https://5dlibrary.com/answers/zh/jingwell-microneedle-cortical-alpha-stroke-cognition/c07/) - Ch.8 Practice and a Pre-registered Test — Chapter 8 gives a four-step checklist and a pre-registered three-arm trial: single primary comparison genuine against sham press needles (both rubbed three times a day), single primary outcome the MoCA at week 4 by analysis of covariance, smallest effect of interest 1.5 points, 72 analysable and 85 randomised per group, 255 in all, frontal alpha only as a pre-specified mediation analysis, and a decision rule written as code. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. (https://5dlibrary.com/answers/en/jingwell-microneedle-cortical-alpha-stroke-cognition/c08/) 第8章 实践与预注册检验 — 第八章给出四步核查表与一项预注册三臂试验:唯一主要比较为真揿针对假揿针(两组都每天按揉三次),唯一主要结局为第 4 周 MoCA 的协方差分析,最小关注效应 1.5 分,每组 72 例可分析、85 例随机,共 255 例,前额 α 只作预先指定的中介分析,判定规则写成代码。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;除注明的文献数据外,书中数据都是合成的。 (https://5dlibrary.com/answers/zh/jingwell-microneedle-cortical-alpha-stroke-cognition/c08/) ## 5DL-B-053 · Interoceptive Decoupling, Salience-Network Switching and Autonomic Visceral Restoration 中文书名:内感受解离、突显网络切换与内脏自主神经修复 DOI: https://doi.org/10.5281/zenodo.24020418 Cite: Fan, Y. (2026). Interoceptive Decoupling, Salience-Network Switching and Autonomic Visceral Restoration (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24020418 Book page: https://5dlibrary.com/books/interoceptive-decoupling-salience-switching/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. Abstract: Can attending to the inside of the body without interpreting it release sympathetic defence and let visceral and pelvic microcirculation restore itself? This book presents mathematical models and methodological discussion for teaching and research and is not medical advice. It breaks the claim into six testable links and falsifies before it sets bounds. The first result is negative: in a precision-weighted Bayesian framework, lowering the precision of the interoceptive channel lowers the posterior estimate only when the prior is itself benign; with an alarmed prior (set here at 3.0 units) the same move pushes the posterior up — lowering sensory precision from 2.0 to 0.25 raises it from 2.333 to 2.800, so the practice makes things worse the better it works. Second, written as a gain term in a signal-detection model, "not interpreting the inner picture" leaves d′ untouched: with the criterion re-tuned, the expected cost is identical at every gain (0.3857). Third, a triple-network Markov chain shows that lowering the probability that inner events recruit salience raises default-mode occupancy (0.362 to 0.596) and dwell time (2.06 to 5.54 s) while necessarily lowering the entropy rate and lengthening first passage to the executive state (4.92 to 6.66 s): a cost, not a side effect. Fourth, and most important: the vagus does not innervate the pelvic organs. The pelvic spinal outflow is sacral, and the "sacral parasympathetic" is itself contested. Heart-rate variability therefore indexes cardiac vagal control and cannot be read as a measure of pelvic blood flow. Four parameter sets reproduce the same cardiac trace while implying mean visceral conductances from 0.506 to 0.884, so "HRV rose, therefore visceral perfusion was restored" is not an inference but an assumption written as if it had been measured. Fifth, RMSSD and high-frequency power move with the breathing pattern: in synthetic records whose mean interbeat interval is held exactly constant, RMSSD ranges from 18.4 to 31.4 ms, and at six breaths a minute the high-frequency index collapses because the respiratory peak has left the band. Sixth, flow is constrained by the fourth-power law and by autoregulation: doubling flow needs a 18.9 % radius increase, flow-sensing autoregulation absorbs 34 % to 60 % of the steady-state response. Chapter 7 reads these beside the Suwen, the Lingshu and the Zhuangzi, set against the Stoic inner citadel, and finds that the compound neiguan cannot be traced to the Huangdi Neijing. Chapter 8 gives a four-step checklist and a pre-registered two-arm trial against a matched external-focus control (primary outcome the week-6 body-vigilance score; smallest effect of interest 0.35 SD; 195 randomised per arm, 390 in all), with a negative-control outcome for generic expectancy. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. 摘要:把注意力放到身体内部而不去解释它,能不能解除交感防御、让内脏与盆腔的微循环自行恢复?本书是教学与研究用的数学模型与方法学讨论,不是医疗建议。它把这一链条拆成六个可检验的环节,并先证伪、再立边界。第一个结果是否定的:在精度加权的贝叶斯框架里,「下调内感受通道的精度」只有在先验本身温和时才使后验估计下降;先验是警觉的(本书设为 3.0 个单位)时,同样的操作把后验推高——感觉精度从 2.0 降到 0.25 时后验从 2.333 升到 2.800,越练越糟。第二,把「不解释内视觉画面」写成信号检测模型中的增益项后,判别力 d′ 与增益无关:重新调定判准时各增益的期望代价完全相同(0.3857)。第三,三网络马尔可夫链表明,降低「内部事件招募突显网络」的概率在提高默认模式占位(0.362 → 0.596)与驻留时间(2.06 → 5.54 秒)的同时,必然降低熵率并延长到执行状态的首达时间(4.92 → 6.66 秒),这是代价而非副作用。 第四,也是最重要的一处修正:迷走神经不支配盆腔器官。盆腔的脊髓出路是骶部,而「骶部副交感」本身正受到质疑。因此心率变异度衡量的是心脏迷走控制,不能当作盆腔血流的指标。四组参数完全重现同一条心脏观测曲线,而它们蕴含的平均内脏传导度从 0.506 到 0.884,所以「心率变异度上升,故内脏灌注恢复」不是推论,而是把假设写成了测量。第五,RMSSD 与高频功率随呼吸方式大幅改变:在平均心搏间期完全不变的合成记录中,RMSSD 从 18.4 变到 31.4 毫秒,而 6 次/分时高频指标反而塌陷,因为呼吸峰已移出频带。第六,血流本身受 r⁴ 律与自动调节约束:流量翻倍需半径增加 18.9%,而流量敏感的自动调节吸收了 34% 到 60% 的稳态反应。 第七章与《素问》《灵枢》《庄子》并读,并与斯多亚学派的「内在堡垒」对勘,指出「内观」一词无法追溯到《黄帝内经》。第八章给出四步核查表与一项预注册两臂试验,对照为结构匹配的外部注意(主要结局为第 6 周身体警觉量表;最小关注效应 0.35 个标准差;每臂随机 195 例,共 390 例),并预先指定一个阴性对照结局,用以识别泛化的预期效应。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;除注明的文献数据外,书中数据都是合成的。 Chapter answers: - Ch.1 Interoception and Precision: The First Link Runs Backwards — Interoception has well-described anatomical pathways, and writing it as precision-weighted inference is a productive framework. But the framework's first expression refutes the first link of the original claim: lowering the precision of the interoceptive channel pushes the posterior toward the prior, so it lowers the estimate of bodily threat only when the prior is benign, and raises it when the prior is alarmed (program p01: from 2.333 to 2.800). Bringing the estimate down requires moving the prior, which is a different operation from declining to interpret, and about seventeen times more costly in distributional terms. The two-level model shows that what "decoupling" amounts to is the situation estimate ceasing to track the bodily estimate, not the bodily signal disappearing, and that it simultaneously makes the situation estimate slower (time constant from 0.688 s to 1.886 s). Finally, the measurement of interoception has unsolved problems of its own: heartbeat counting scores cannot be taken as interoceptive accuracy, and self-report scales measure a different construct from objective performance; Chapter 8 chooses its outcomes accordingly. (https://5dlibrary.com/answers/en/interoceptive-decoupling-salience-switching/c01/) 第1章 内感受与精度加权:第一个环节就反了 — 内感受有明确的解剖通路,把它写成精度加权的推断是一个有生产力的框架。但该框架的第一个式子就否证了原说法的第一个环节:下调内感受通道的精度把后验推向先验,所以只有先验温和时才使身体威胁的估计下降;先验警觉时它使估计上升(程序 p01:从 2.333 升到 2.800)。要让估计下降,必须移动先验,而这与「不作解释」不是同一个操作,其在分布上的代价也大约 17 倍。两层模型说明,「解离」的内容是处境估计不再跟踪身体估计,而不是身体信号消失,并且它同时使处境估计变得更迟钝(时间常数从 0.688 秒到 1.886 秒)。最后,内感受的测量本身存在未解决的问题:心跳计数任务的得分不能当作内感受准确度,自评量表与客观成绩是不同的构念,第八章据此选择结局。 (https://5dlibrary.com/answers/zh/interoceptive-decoupling-salience-switching/c01/) - Ch.2 Switching Between Networks: Dwell Time and Its Cost — Three networks can be separated in resting-state data and their temporal organisation can be characterised with state-sequence models; both are supported. Once "inner events no longer recruit the salience network" is written as a fall in a transition probability, three things happen together: default-mode occupancy rises from 0.285 to 0.596 and dwell time from 1.44 to 5.54 s, while the entropy rate falls from 1.8348 to 1.2472 bits per second and first passage to the executive state lengthens from 4.47 to 6.66 s. The first two are what the claim wants; the last two are its cost, and that cost is an arithmetic necessity rather than an empirical finding. The phrase "soothes the default mode network" needs correction in both direction and level: raising default-mode dwell time and lowering default-mode activity are different descriptions, and identifying "higher-order awareness" with the precision modulation of some network is not yet supported. Finally, entropy-rate estimates on short state sequences are systematically biased and are unsuitable as a primary outcome. (https://5dlibrary.com/answers/en/interoceptive-decoupling-salience-switching/c02/) 第2章 三网络的切换:驻留时间与它的代价 — 三个网络在静息态数据中可以分离,它们的时间组织可以用状态序列模型刻画,这些都有证据。把「内部事件不再招募突显网络」写成转移概率的下降之后,结果是三件事同时发生:默认模式占位从 0.285 升到 0.596、驻留时间从 1.44 升到 5.54 秒,而熵率从 1.8348 降到 1.2472 比特/秒、到执行状态的首达时间从 4.47 升到 6.66 秒。前两项是原说法想要的,后两项是它的代价,且这个代价是算术必然,不是经验发现。「安抚默认模式网络」这句话在方向与层次上都需要修正:提高默认模式驻留与降低默认模式活动是两个不同的描述,而把「高阶觉知」指认为某个网络的精度调节还缺证据。最后,状态序列上的熵率估计在短序列上有系统偏差,不宜作主要结局。 (https://5dlibrary.com/answers/zh/interoceptive-decoupling-salience-switching/c02/) - Ch.3 The Non-Judgmental Imagery Stream: Trading Errors, Not Gaining Information — Written as a gain term in a signal-detection model, "not interpreting the inner picture" leaves discriminability $d'$ untouched: with the criterion re-tuned, the expected cost is identical at every gain (0.3857); without re-tuning, a gain of 0.5 cuts the hit rate from 0.622 to 0.281 and raises the cost from 0.3857 to 0.4634. Non-judgmental attention trades one class of error for another rather than adding information; its benefit comes from the operating point having been wrong to begin with, which is exactly the situation described by the cognitive model of panic. The affect-labelling literature offers an equally supported practice running in the opposite direction. The dual-process model shows that spacing barely changes how much is learned but changes the fast process measured at the end of a session, so the primary measurement belongs on another day. Beta–Bernoulli updating supplies a number for "how long" (about 850 clean trials for a firmly held belief) and simultaneously fails to explain relapse, making it evidence against simple evidential updating. Finally, transition entropy on session-length sequences carries a bias two to five times the effect sought, and must be corrected, with a fixed sampling interval and per-person averaging. (https://5dlibrary.com/answers/en/interoceptive-decoupling-salience-switching/c03/) 第3章 无评判的表象流:交换错误,而非增加信息 — 把「不解释内在画面」写成信号检测模型中的增益项之后,判别力 $d'$ 与增益无关:判准重新调定时,各增益的期望代价完全相同(0.3857);判准不重调时,增益 0.5 把命中率从 0.622 压到 0.281、把代价从 0.3857 抬到 0.4634。所以「无评判」交换的是错误类型,不是信息量;它的好处来自原来的操作点偏了,这正是惊恐认知模型所描述的情形。情绪标注的文献给出方向相反而同样有支持的做法。双过程模型显示,间隔几乎不改变学到了多少,而改变课末测量到的快过程,所以主要测量应放在另一天。Beta—伯努利更新给出了「要多久」的数(牢固信念约 850 次干净试验),同时无法解释复发,因此是反对单纯证据性更新的证据。最后,表象流的转移熵在练习长度的序列上偏差约为效应的 2 到 5 倍,必须校正、固定取样间隔,并按个人均值分析。 (https://5dlibrary.com/answers/zh/interoceptive-decoupling-salience-switching/c03/) - Ch.4 Two Autonomic Outflows: The Vagus Does Not Reach the Pelvis — The fourth link of the original claim is wrong: the spinal outflow to the pelvic organs is sacral, the vagus does not innervate them, so "redistribution of vagal tone leading to the restoration of pelvic microcirculation" cannot hold anatomically, and heart-rate variability, which measures cardiac vagal control, cannot support a conclusion about the pelvis. Even the "sacral parasympathetic" is under dispute: Espinosa-Medina and colleagues and Sivori and colleagues argue from developmental lineage and single-cell transcriptomics that the sacral outflow is sympathetic, while Jänig and colleagues and Neuhuber and colleagues reply from transmitter and function, and the open peer review judged the relevant title to exceed the data. This book takes no side, because whatever it is called it is not the vagus. Sympathetic outflow is regionally specific and not a scalar. Finally, RMSSD and high-frequency power move a long way with breathing pattern: in synthetic records with the mean interbeat interval held exactly constant, RMSSD ranges from 18.4 to 31.4 ms, and at six breaths a minute the high-frequency index collapses because the respiratory peak has left the band. (https://5dlibrary.com/answers/en/interoceptive-decoupling-salience-switching/c04/) 第4章 自主神经的两条出路:迷走神经不到盆腔 — 原说法的第四个环节是错的:盆腔器官的脊髓出路是骶部的,迷走神经不支配盆腔器官,所以「迷走神经张力重分配 → 盆腔微循环修复」在解剖上无法成立,而心率变异度测的是心脏的迷走控制,不能用来支持关于盆腔的结论。连「骶部副交感」本身也在争论中:Espinosa-Medina 等人与 Sivori 等人依据发育谱系与单细胞转录组主张骶部出路是交感的,Jänig 等人与 Neuhuber 等人按递质与功能反驳,公开评审认为相关文章的标题超出了数据;本书不站边,因为无论它叫什么,它都不是迷走神经。交感输出有区域特异性,不是一个标量。最后,RMSSD 与高频功率随呼吸方式大幅改变:在平均心搏间期完全不变的合成记录中 RMSSD 从 18.4 变到 31.4 毫秒,而 6 次/分时高频指标塌陷,因为呼吸峰已离开频带。 (https://5dlibrary.com/answers/zh/interoceptive-decoupling-salience-switching/c04/) - Ch.5 How Much Can Blood Flow Change: The Fourth Power and Autoregulation — Even granting that vasoconstrictor drive really is withdrawn, the change in flow is doubly constrained. The fourth-power law: doubling flow needs a 18.9 % increase in radius, a 10 % rise needs 2.41 %, and paired measurement with a 12 % coefficient of variation over 20 pairs detects only a +11.2 % change in flow. Parallel beds: at fixed pressure, local dilation requires total output to rise; at fixed cardiac output, the same dilation raises that bed only to 1.840-fold and presses the others to 0.920-fold. Autoregulation: flow-sensing regulation absorbs 34 % to 60 % of the steady-state response in this model, the approach is monotone with no overshoot, and the time constant makes the same manoeuvre read as +9 % at 5 s and +37 % at 60 s, so the time of measurement must be specified in advance. Finally, "a change in flow" and "repair of tissue" are two different things, the second requiring a chain of premises this book cannot evidence; and a mechanical pathway through pelvic-floor and abdominal-wall tone is an alternative this book does not exclude, and which Chapter 8 must therefore match. (https://5dlibrary.com/answers/en/interoceptive-decoupling-salience-switching/c05/) 第5章 血流能变多少:四次方律与自动调节 — 即使假设血管收缩驱动真的撤除了,血流的变化也受两重约束。四次方律:流量翻倍需半径增加 18.9%,10% 的流量上升需 2.41%,而配对测量在 12% 变异系数、20 对下只能检出 +11.2% 的流量变化。并联床区:压力固定时局部扩张要求总输出上升,心输出量固定时同样的扩张只使该床区上升到 1.840 倍并把其他床区压到 0.920 倍。自动调节:流量敏感的调节在本书模型中吸收了稳态反应的 34% 到 60%,趋近是单调的、没有过冲,而时间常数使同一操作在 5 秒时读作 +9%、在 60 秒时读作 +37%,所以测量时点必须预先规定。最后,「血流变化」与「组织修复」是两件事,后者需要一整串本书无法引证的额外前提;而一条力学通路(盆底与腹壁张力)是本书未排除的替代解释,并因此成为第八章必须匹配的因素。 (https://5dlibrary.com/answers/zh/interoceptive-decoupling-salience-switching/c05/) - Ch.6 The Composite Model, Non-Identifiability and Case Evidence — Joined into a three-state model, "lowering higher-order precision lowers defensive drive and recovers conductance" holds inside the model — but that is arithmetic, since $\pi$ multiplies the only term that raises $S$, not a finding about bodies. The model has three known defects, the most important being that it cannot represent the situation of Chapter 1; the variant that admits this yields a testable boundary condition, namely that once a standing alarmed prior exceeds about 0.3, lowering $\pi$ goes from helping to harming. Structural identifiability gives the chapter's main result: four parameter sets reproduce the same cardiac trace to within floating-point error while implying mean visceral conductances from 0.506 to 0.884 and minima from 0.000 to 0.426, so "HRV rose, hence visceral perfusion was restored" writes an assumption as a measurement; what the cardiac channel identifies is only the timing of the response. Finally, under a generator with no effect at all, an AB case design yields an average "improvement" of +0.294 SD because of entry-time selection and autocorrelation, with a naive test declaring benefit in 28.5 % of cases against 61.6 % under a true 0.5 SD effect: a likelihood ratio of about 2.2. (https://5dlibrary.com/answers/en/interoceptive-decoupling-salience-switching/c06/) 第6章 复合模型、不可识别性与个案证据 — 把链条接成三状态模型后,模型内部「降低高阶精度降低防御驱动、收回传导度」成立,但这是算术——$\pi$ 乘在唯一抬高 $S$ 的项上——不是关于身体的发现。该模型有三处已知缺陷,其中最要紧的是它无法表示第一章的情形;承认这一点的变体给出一个可检验的边界条件:常驻警觉先验超过约 0.3 时,降低 $\pi$ 由有益转为有害。结构可识别性给出本章的主要结果:四组参数把同一条心脏观测曲线重现到浮点误差量级,而蕴含的平均内脏传导度从 0.506 到 0.884、最低值从 0.000 到 0.426,所以「心率变异度上升故内脏灌注恢复」是把假设写成了测量;心脏通道能识别的只是反应的时点。最后,在完全无效应的生成器下,AB 个案设计由于进入时点的选择与自相关,平均给出 +0.294 个标准差的「改善」,朴素检验在 28.5% 的个案中宣称获益,而真实 0.5 个标准差效应下为 61.6%,似然比约 2.2。 (https://5dlibrary.com/answers/zh/interoceptive-decoupling-salience-switching/c06/) - Ch.8 Practice and a Pre-Registered Test — Chapter 8 gives a four-step checklist and a pre-registered two-arm trial against a matched external-focus control (primary outcome the week-6 body-vigilance score; smallest effect of interest 0.35 SD; 195 randomised per arm, 390 in all), with a negative-control outcome for generic expectancy. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. (https://5dlibrary.com/answers/en/interoceptive-decoupling-salience-switching/c08/) 第8章 实践与预注册检验 — 第八章给出四步核查表与一项预注册两臂试验,对照为结构匹配的外部注意(主要结局为第 6 周身体警觉量表;最小关注效应 0.35 个标准差;每臂随机 195 例,共 390 例),并预先指定一个阴性对照结局,用以识别泛化的预期效应。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;除注明的文献数据外,书中数据都是合成的。 (https://5dlibrary.com/answers/zh/interoceptive-decoupling-salience-switching/c08/) ## 5DL-B-054 · Biomathematical Models of Sub-Perceptual Tryptamine Pharmacokinetics, 5-HT2A Receptor Desensitisation and Neuroplastic Window Dynamics 中文书名:亚知觉色胺药代动力学、5-HT2A 受体脱敏与神经可塑性窗口的生物数学模型 DOI: https://doi.org/10.5281/zenodo.24019347 Cite: Fan, Y. (2026). Biomathematical Models of Sub-Perceptual Tryptamine Pharmacokinetics, 5-HT2A Receptor Desensitisation and Neuroplastic Window Dynamics (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24019347 Book page: https://5dlibrary.com/books/tryptamine-pharmacokinetics-5ht2a-desensitisation/ Scope: It is not medical advice and is not a guide to using anything. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. Abstract: This book presents biomathematics and methodology for teaching and research. It is not medical advice and is not a guide to using anything. The substances it discusses are controlled in most jurisdictions, and the book gives no dose, mass, product, source, route, synthesis or course of use: every amount is a dimensionless ratio to an unspecified reference, the convention established in Volume 8. The book writes the pharmacological chain behind "sub-perceptual microdosing" as mathematics and tests it link by link. Receptor occupancy follows the Hill relation, whose log-log slope falls smoothly from 1 to 0, so no exposure is sub-perceptual by virtue of the receptor; a threshold can only be imposed by perception. Saturating a prodrug's conversion does not make total exposure jump: mass balance fixes the area under the curve and saturation changes only the shape — unsaturated, a tenfold amount raises the peak 7.84-fold; saturated, only 3.49-fold, while the time to peak stretches 6.40-fold. Fitting one exposure's curve pins the Michaelis constant only to within a factor of 13 yet predicts the peak at a fivefold exposure to within a factor of 1.31; the real obstacle is that no human conversion kinetics are published, which widens the peak uncertainty to a factor of 6.13. "Sub-perceptual" is a population distribution, not a dose. The receptor-downregulation "tolerance ratchet" depends on a recovery half-life never measured in humans, and a tighter schedule lowers the effect per dose while raising the cumulative effect. A rise in connectivity entropy is indistinguishable from added noise: 1 bit across 20 regions needs only about 4.6 % added noise variance. A plasticity "window" is a gate, not a drive. Chapter 7 treats the Shennong Bencao Jing's escalation rule as a dose-finding algorithm, compares it with a 3+3 design and a continual reassessment method, and finds that its final tenfold step carries most of the risk. Chapter 8 gives a four-step checklist and a pre-registered double-blind trial: smallest effect of interest 0.30 SD, 282 randomised per arm, 564 in all, with blind integrity written into the decision rule. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. 摘要:本书是教学与研究用的生物数学与方法学讨论,不是医疗建议,也不是任何物质的使用指南。本书讨论的物质在多数司法辖区受管制;书中不给出任何剂量、质量、产品、来源、途径、合成或疗程:全书的剂量一律以相对于一个未指定参考值的无量纲比值表示,这是第 8 本确立的写法。 本书把「亚知觉微剂量」所依赖的药理链条写成数学,并逐段检验。受体占据遵循希尔关系,其对数斜率从 1 平滑地降到 0,所以没有哪个暴露量「因为受体的缘故」是亚知觉的——阈值只能由知觉给出。前药的饱和转化并不使总暴露跳升:质量守恒把曲线下面积固定住,饱和改变的只是形状——未饱和时十倍剂量使峰值升高 7.84 倍;饱和时只升高 3.49 倍,而达峰时间拉长 6.40 倍。由单一暴露的曲线拟合,米氏常数只能定到 13 倍以内,五倍暴露下的峰值预测却只差 1.31 倍;真正的障碍是人体转化动力学根本没有发表,这使峰值的不确定度扩大到 6.13 倍。「亚知觉」是一个人群分布而不是一个剂量。受体下调的「耐受棘轮」取决于一个从未在人体测量过的恢复半衰期,而较密的日程使每次效应下降、累计效应上升。连接熵的上升与额外噪声不可区分:跨 20 个脑区使熵升高 1 比特只需约 4.6% 的额外噪声方差。可塑性「窗口」是闸门而非驱动。 第七章把《神农本草经》的递增规则当作一个剂量探索算法,与 3+3 设计及连续重估法比较,指出风险主要来自最后的「十之」一步。第八章给出四步核查表与一项预注册双盲试验:最小关注效应 0.30 个标准差,每臂随机 282 例,共 564 例,设盲完整性写入判定规则。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;除注明的文献数据外,书中数据都是合成的。 Chapter answers: - Ch.1 Receptor Occupancy: No Corner, Therefore No Threshold — 5-HT2A agonism is a necessary link in producing a marked subjective effect, with human evidence from pharmacological blockade. But receptor occupancy follows the Hill relation, whose log-log slope falls smoothly from 1 to 0 and never shows a corner or a threshold, so no exposure is sub-perceptual by virtue of the receptor; a threshold can only be imposed by perception. The relation between an exposure ratio and an occupancy ratio depends on position along the curve (a fifth of the exposure cuts occupancy by factors between 4.33 and 1.67), so "a tenth of the dose" has no definite meaning at receptor level. The three possible senses of "sub-perceptual" — a subjective threshold, a measurement detection limit, and an assumed effect threshold — are not equivalent, and conflating them produces empty arguments. Finally, if there is a route that bypasses 5-HT2A, as Hesselgrave and colleagues' and Moliner and colleagues' results suggest, then dose reasoning built on receptor occupancy has less support, not more. (https://5dlibrary.com/answers/en/tryptamine-pharmacokinetics-5ht2a-desensitisation/c01/) 第1章 受体占据:没有拐点,所以没有阈值 — 5-HT2A 的激动是这类化合物产生明显主观效应的必要环节,这有药理阻断的人体证据。但受体占据遵循希尔关系,其对数—对数斜率从 1 平滑地降到 0,从不出现拐点或阈值,所以没有哪个暴露量「因为受体的缘故」是亚知觉的;阈值只能由知觉给出。暴露比与占据比的关系取决于曲线上的位置(五分之一的暴露使占据降为 1/4.33 到 1/1.67 不等),所以「十分之一的剂量」在受体层面没有确定含义。「亚知觉」的三种可能意义——主观阈值、测量检出限、假定的效应阈值——互不等价,混用会产生空洞的论证。最后,若存在不经 5-HT2A 的通路(Hesselgrave 等人与 Moliner 等人的结果提示这一点),那么以受体占据为基础的剂量推理就更缺依据,而不是更有依据。 (https://5dlibrary.com/answers/zh/tryptamine-pharmacokinetics-5ht2a-desensitisation/c01/) - Ch.2 Prodrugs and Saturable Conversion: Saturation Blunts the Peak — The dephosphorylation of psilocybin has had tissue-homogenate evidence since 1961, its metabolism has been reviewed, and human pharmacokinetics has an escalating-dose study. No published human conversion kinetics for the acetylated prodrugs were found, so the Michaelis parameters here are assumed values. The model gives a result running opposite to the common account: saturating the prodrug step does not make total exposure jump — mass balance fixes the area under the curve — but blunts the peak and extends the time to peak (at a tenfold amount: unsaturated, peak ×7.84 and time to peak ×1.18; saturated, peak ×3.49 and time to peak ×6.40). Without knowing $K_m$, the peak at a fixed amount is uncertain by a factor of 13.68. But fitting from a single exposure's curve yields a positive result: although $K_m$ is pinned only to within a factor of 13, the predicted peak at fivefold exposure differs by only 1.31, because mass balance constrains total conversion. The obstacle is therefore not non-identifiability but the absence of a curve to fit, where the peak uncertainty is 6.13. (https://5dlibrary.com/answers/en/tryptamine-pharmacokinetics-5ht2a-desensitisation/c02/) 第2章 前药与饱和转化:饱和削平峰值 — 赛洛西宾的去磷酸化自 1961 年起就有组织匀浆证据,其代谢有综述,人体药代动力学有递增剂量研究。乙醯基前药的人体转化动力学,本书核查未找到已发表数据,所以本章的米氏参数是假设值。模型给出一个方向与常见说法相反的结果:饱和的前药步骤不使总暴露跳升——质量守恒把 AUC 固定住——而是削平峰值并拉长达峰时间(十倍剂量下,未饱和时峰值升 7.84 倍、达峰时间 1.18 倍;饱和时峰值升 3.49 倍、达峰时间 6.40 倍)。不知道 $K_m$ 时,同一剂量比例下峰值的不确定度达 13.68 倍。但从单一暴露的曲线拟合给出一个正面结果:$K_m$ 虽只定到 13 倍以内,五倍暴露下的峰值预测只差 1.31 倍,因为质量守恒约束了总转化。所以障碍不是不可识别性,而是没有数据可拟合——此时峰值的不确定度是 6.13 倍。 (https://5dlibrary.com/answers/zh/tryptamine-pharmacokinetics-5ht2a-desensitisation/c02/) - Ch.3 Coupling Kinetics to Effect: The Perceptual Threshold Is Not a Number — The effect compartment is a standard pharmacological tool, and it yields a conclusion central to this book: the time course of an effect is a property of $k_{e0}$ rather than of the pharmacokinetics. Under an identical plasma curve, lowering $k_{e0}$ from 0.3 to 0.1 changes "does it cross the perceptual threshold" from "nearly four half-lives above it" to "never", and $k_{e0}$ can only be fitted to effect data. The same model yields the hysteresis loop: at a given plasma concentration the falling limb shows far more effect than the rising limb, so a concentration-effect fit that does not model the lag gives a wrong $EC_{50}$. Finally, "sub-perceptual" is a population distribution rather than a dose: in a population with individual variation in both exposure and threshold, no dose fraction makes the proportion noticing zero, and between 2 % and 25 % noticing there is only a 2.6-fold margin. To explain an observed 0.20 SD difference by expectancy alone requires only a 0.50 SD expectancy effect at a 40 % correct-guess rate, so blind integrity must be measured. (https://5dlibrary.com/answers/en/tryptamine-pharmacokinetics-5ht2a-desensitisation/c03/) 第3章 药代—药效耦合:知觉阈不是一个数 — 效应室模型是药理学的标准工具,而它给出一个对本书主题关键的结论:效应的时程是 $k_{e0}$ 的性质而不是药代动力学的性质。在完全相同的血浆曲线下,$k_{e0}$ 从 0.3 降到 0.1 就使「是否越过知觉阈」从「将近四个半衰期在阈上」变为「从未越过」,而 $k_{e0}$ 只能由效应数据拟合。同一模型给出滞后回线:在同一血浆浓度下下降段的效应远大于上升段,所以不建模滞后的浓度—效应拟合会给出错误的 $EC_{50}$。最后,「亚知觉」是一个人群分布而不是一个剂量:在暴露与阈值各有个体差异的人群中,没有任何剂量比例使察觉比例为零,而从 2% 察觉到 25% 察觉之间只有 2.6 倍的余地。要用纯预期解释一个 0.20 个标准差的观测差别,在 40% 猜中率下只需 0.50 个标准差的预期效应,所以设盲完整性必须被测量。 (https://5dlibrary.com/answers/zh/tryptamine-pharmacokinetics-5ht2a-desensitisation/c03/) - Ch.4 Downregulation and the Tolerance Ratchet: Each Dose Less, the Total More — Repeated high exposure downregulates 5-HT2 receptors in rat brain, behavioural tolerance is associated with reduced cortical signalling, and human tolerance was observed in 1956. But those exposures are far above the range discussed here, and the human recovery half-life at sub-perceptual exposure has not been measured. The mathematics of the ratchet shows that what decides the outcome is the ratio of the recovery half-life to the dosing interval: the critical interval is 3.0, 8.0 and 18.2 days at $\tau_R$ of 1, 3 and 7 days. Since $\tau_R$ is unknown, the specific schedules in popular accounts cannot be derived from downregulation. The chapter's second result runs opposite to the usual inference: over a fixed calendar period a tighter schedule lowers the effect per dose while raising the cumulative effect, so downregulation alone cannot argue for spacing doses out; which quantity matters depends on the purpose, which pharmacology does not address. Finally, with the calendar period and dose count fixed, the ratio of last dose to first ranges with $\tau_R$ from near 1 to 0.389. (https://5dlibrary.com/answers/en/tryptamine-pharmacokinetics-5ht2a-desensitisation/c04/) 第4章 受体下调与耐受棘轮:每次下降,累计上升 — 反复高暴露使大鼠脑内 5-HT2 受体下调,行为耐受与皮层信号降低相关,人体耐受在 1956 年就有观察。但这些暴露远高于本书讨论的范围,而人体在亚知觉暴露下的恢复半衰期未被测量。棘轮的数学表明,决定结果的是恢复半衰期与给药间隔之比:临界间隔在 $\tau_R$ 为 1、3、7 天时分别是 3.0、8.0、18.2 天。由于 $\tau_R$ 未知,流行说法中的具体日程不能从受体下调推出。本章的第二个结果与通常的推论相反:在固定的日历期内,较密的日程使每次的效应下降而使累计效应上升,所以仅凭受体下调不能论证应当拉开间隔;哪一个量重要取决于目的,而药理学不回答这个问题。最后,在固定日历期与次数下,最后一次对第一次的效应比随 $\tau_R$ 从接近 1 变到 0.389。 (https://5dlibrary.com/answers/zh/tryptamine-pharmacokinetics-5ht2a-desensitisation/c04/) - Ch.5 Connectivity Entropy: Indistinguishable from Noise — The entropic-brain hypothesis and its associated imaging results are central to this field. This chapter handles only the measurement question and reaches two results. First, in a Gaussian graphical model an "increase in entropy" is indistinguishable from an increase in noise: true coupling 0.8 with noise and true coupling 0.6 without give the same observed correlations and the same entropy (differing by 0.009 bits), while raising entropy by 1 bit across 20 regions needs only about 4.6 % added noise variance; head motion is known to produce systematic connectivity artefacts, and no confound-regression strategy removes them completely. Second, short-window entropy estimates are systematically biased: with 12 regions and 30 time points the bias is $-1.978$ bits against a true difference of 2.676 bits between two quite different coupling strengths. The remedy is a design one: a difference between length-matched windows is close to the truth at every length. Chapter 8 therefore treats connectivity-entropy measures as exploratory outcomes with three stated requirements. (https://5dlibrary.com/answers/en/tryptamine-pharmacokinetics-5ht2a-desensitisation/c05/) 第5章 连接熵:与噪声不可区分 — 熵脑假说与相关影像学结果是这一领域的核心。本章只处理测量问题,得到两个结果。第一,在高斯图模型中,「熵升高」与「噪声增加」不可区分:真实耦合 0.8 加噪声与真实耦合 0.6 无噪声给出相同的观测相关与相同的熵(相差 0.009 比特),而跨 20 个脑区使熵升高 1 比特只需约 4.6% 的额外噪声方差;头动已知会造成系统性的连接伪迹,且没有混淆回归策略能完全消除它。第二,短窗口的熵估计有系统偏差:12 个脑区、30 个时间点时偏差 $-1.978$ 比特,而两种耦合强度之间的真实差别是 2.676 比特。解决办法是设计上的:匹配窗口长度的差值比较在每个长度下都接近真值。第八章因此把连接熵类指标列为探索性结局并附三项规定。 (https://5dlibrary.com/answers/zh/tryptamine-pharmacokinetics-5ht2a-desensitisation/c05/) - Ch.6 The Plasticity Window: A Gate, Not a Drive — The evidence for structural plasticity is real: increases in dendritic spines and dendritic complexity in cultured neurons and animals, and rapid, persistent spine growth on in vivo imaging. There are two mechanistic surprises: the synaptic and behavioural effects in mice are not blocked by a 5-HT2 antagonist, and these compounds bind TrkB directly — both pointing to a route that bypasses 5-HT2A and making dose reasoning built on receptor occupancy less applicable. The critical-period framing gives the chapter's main result: the window is a gate and not a drive — in the model, with no input the synaptic weight stays at 0 however wide the window, so a measurement of the window cannot predict whether anything was learned. The latency of the input matters far more than its duration: two days beginning at the exposure reach 78 % of the best case, delayed to days 4–6 only 6 %, while extending two days to seven adds about 28 %. The schedule that matters is therefore practice relative to exposure, not dosing. A model combining the window with the receptor ratchet would need the unmeasured recovery half-life, and the book does not write it. (https://5dlibrary.com/answers/en/tryptamine-pharmacokinetics-5ht2a-desensitisation/c06/) 第6章 可塑性窗口:闸门,不是驱动 — 结构可塑性的证据是实在的:培养神经元与动物中的树突棘与树突复杂度增加,在体成像显示快速且持久的棘生长。机制上有两个意外:小鼠的突触与行为效应不被 5-HT2 拮抗剂阻断,而这类化合物可直接结合 TrkB;两者都提示存在不经 5-HT2A 的通路,这使以受体占据为基础的剂量推理更不适用。关键期重开的框架给出本章的主结果:窗口是闸门而不是驱动——在模型中,没有输入时无论窗口多宽突触权重都停在 0,所以对窗口的测量不能预测是否学到了东西。输入的延迟远比其时长重要:紧接暴露的两天达到最佳的 78%,延到第 4–6 天只有 6%,而把两天延长到七天只增加约 28%。因此真正要紧的日程是练习相对于暴露的日程,不是给药日程。把窗口与受体棘轮合在一起的模型需要未被测量的恢复半衰期,本书不写。 (https://5dlibrary.com/answers/zh/tryptamine-pharmacokinetics-5ht2a-desensitisation/c06/) - Ch.7 Classical Sources Read Side by Side: "Begin Like a Millet Grain" and the Boundary of Dose — Chapter 7 treats the Shennong Bencao Jing's escalation rule as a dose-finding algorithm, compares it with a 3+3 design and a continual reassessment method, and finds that its final tenfold step carries most of the risk. (https://5dlibrary.com/answers/en/tryptamine-pharmacokinetics-5ht2a-desensitisation/c07/) 第7章 原典并读:「先起如黍粟」与剂量的边界 — 第七章把《神农本草经》的递增规则当作一个剂量探索算法,与 3+3 设计及连续重估法比较,指出风险主要来自最后的「十之」一步。 (https://5dlibrary.com/answers/zh/tryptamine-pharmacokinetics-5ht2a-desensitisation/c07/) - Ch.8 Practice and a Pre-Registered Test — Chapter 8 gives a four-step checklist and a pre-registered double-blind trial: smallest effect of interest 0.30 SD, 282 randomised per arm, 564 in all, with blind integrity written into the decision rule. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. (https://5dlibrary.com/answers/en/tryptamine-pharmacokinetics-5ht2a-desensitisation/c08/) 第8章 实践与预注册检验 — 第八章给出四步核查表与一项预注册双盲试验:最小关注效应 0.30 个标准差,每臂随机 282 例,共 564 例,设盲完整性写入判定规则。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;除注明的文献数据外,书中数据都是合成的。 (https://5dlibrary.com/answers/zh/tryptamine-pharmacokinetics-5ht2a-desensitisation/c08/) ## 5DL-B-055 · Imperfect-Information Stochastic Games, Bayesian Hand-Reading and Counterfactual Regret Minimisation in Four-Player Tile Dynamics 中文书名:四人不完全信息随机博弈、贝叶斯读牌与反事实遗憾最小化 DOI: https://doi.org/10.5281/zenodo.24020464 Cite: Fan, Y. (2026). Imperfect-Information Stochastic Games, Bayesian Hand-Reading and Counterfactual Regret Minimisation in Four-Player Tile Dynamics (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24020464 Book page: https://5dlibrary.com/books/imperfect-information-tile-games/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. Abstract: A four-person table is the least agreeable class of object in game theory: imperfect information, stochastic, four-sided and negative-sum in money. This book writes that class of game as mathematics and tests popular strategic reasoning segment by segment. It is not a gambling guide: Chapter 8 computes that a player with a plausible edge, playing 3,120 hands a year at a fee of 0.010, expects +8.6 buy-ins while 36 % of such players still finish down, and that with no edge any fee makes the expectation negative whatever the staking rule. Chapter 1 gives the scale: about 9.85 × 10¹⁰ distinct 13-tile hands and about 5.0 × 10⁴⁶ configurations of the other three hands, so enumerating information sets is a category error. Chapter 2 treats the tile pool as hypergeometric inference: a two-sided wait's advantage over a one-sided wait approaches 2 only late in the hand (1.94 with two draws left) and is 1.38 early. Chapter 3 derives suji as a Bayesian deduction that makes a tile 3.6 times safer, not safe. Chapter 4 computes shanten by dynamic programming and shows that a one-player race model overstates the win probability by factors of 1.36 to 1.79, the error growing with distance. Chapter 5 writes folding as optimal stopping: the opponent's self-draw term cancels, so the case for folding rests on the deal-in probability alone. Chapter 6 verifies counterfactual regret minimisation on a two-player zero-sum game (exploitability from 0.667 to 0.0013, value −0.055554 against −1/18) and honestly reports two negative results on a three-player game: exploitability still falls to 4 × 10⁻⁶, and four simple deviations failed to push the player below the equilibrium value. The Kelly growth curve is nearly symmetric (half-Kelly keeps 75.3 %, 1.5× Kelly 75.7 %), so the reason to stake less is estimation error, not the curve's shape. Chapter 8 gives a four-step checklist and a pre-registered duplicate-design policy comparison scored at the first differing decision: smallest effect of interest 0.010 buy-ins per scored decision, 12,000 scored decisions, the decision rule written as code; the conclusion holds only against the fixed opponent pool used. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. 摘要:一张四人牌桌是博弈论里最不讨人喜欢的那一类对象:不完全信息、随机、四方、而且在金钱上是负和。本书把这类博弈写成数学,并逐段检验流行的策略推理。它不是赌博指南:第八章算出,一个有合理优势的玩家一年打 3,120 手、抽水 0.010 时期望收益为 +8.6 个买入,却仍有 36% 的人年终亏损;而没有优势时任何抽水都使期望为负,无论用什么资金管理规则。 第一章给出规模:13 张手牌的不同多重集约有 9.85 × 10¹⁰ 个,另外三家的配置约 5.0 × 10⁴⁶ 种,所以枚举信息集是范畴错误。第二章把牌池写成超几何推断:两面听相对单骑的优势只在后期接近 2(剩 2 次摸牌时 1.94),前期只有 1.38。第三章把筋牌推导成贝叶斯演绎:它使某张牌安全 3.6 倍,而非完全安全。第四章用动态规划算向聴,并指出单人「赛跑」模型把胜率高估 1.36 到 1.79 倍,误差随距离增大。第五章把弃胡写成最优停时:对手自摸的项相消,所以弃牌的理由只落在放铳概率上。 第六章在两人零和博弈上验证反事实遗憾最小化(可利用度从 0.667 降到 0.0013,博弈值 −0.055554,理论值 −1/18),并在三人博弈上诚实地报告两个否定结果:可利用度照样降到 4 × 10⁻⁶,四种简单偏离也没能把玩家压到均衡值以下。凯利增长率曲线在最优点附近近乎对称(半凯利保 75.3%、1.5 倍凯利保 75.7%),所以降低注码的理由是估计误差而不是曲线形状。第八章给出四步核查表与一项预注册的重复对局策略比较,只在首次分歧的决策上计分:最小关注效应 0.010 个买入/计分决策,需 12,000 个计分决策,判定规则写成代码;结论只对所用的固定对手池成立。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;除注明的文献数据外,书中数据都是合成的。 Chapter answers: - Ch.1 Scale: Why the Information Sets Cannot Be Listed — Chapter 1 gives the scale: about 9.85 × 10¹⁰ distinct 13-tile hands and about 5.0 × 10⁴⁶ configurations of the other three hands, so enumerating information sets is a category error. (https://5dlibrary.com/answers/en/imperfect-information-tile-games/c01/) 第1章 规模:为什么不能把信息集列出来 — 第一章给出规模:13 张手牌的不同多重集约有 9.85 × 10¹⁰ 个,另外三家的配置约 5.0 × 10⁴⁶ 种,所以枚举信息集是范畴错误。 (https://5dlibrary.com/answers/zh/imperfect-information-tile-games/c01/) - Ch.2 The Tile Pool: Hypergeometric Inference and a Result That Runs Backwards — Chapter 2 treats the tile pool as hypergeometric inference: a two-sided wait's advantage over a one-sided wait approaches 2 only late in the hand (1.94 with two draws left) and is 1.38 early. (https://5dlibrary.com/answers/en/imperfect-information-tile-games/c02/) 第2章 牌池:超几何推断与一个方向相反的结果 — 第二章把牌池写成超几何推断:两面听相对单骑的优势只在后期接近 2(剩 2 次摸牌时 1.94),前期只有 1.38。 (https://5dlibrary.com/answers/zh/imperfect-information-tile-games/c02/) - Ch.3 Hand Reading: Suji Is a Deduction, Not a Regularity — Chapter 3 derives suji as a Bayesian deduction that makes a tile 3.6 times safer, not safe. (https://5dlibrary.com/answers/en/imperfect-information-tile-games/c03/) 第3章 读牌:筋是一条演绎,不是一条规律 — 第三章把筋牌推导成贝叶斯演绎:它使某张牌安全 3.6 倍,而非完全安全。 (https://5dlibrary.com/answers/zh/imperfect-information-tile-games/c03/) - Ch.4 Shanten, Acceptance and the Race: Ignoring Three Players Overstates the Win — Chapter 4 computes shanten by dynamic programming and shows that a one-player race model overstates the win probability by factors of 1.36 to 1.79, the error growing with distance. (https://5dlibrary.com/answers/en/imperfect-information-tile-games/c04/) 第4章 向聴、进张与赛跑:忽略三家会高估胜率 — 第四章用动态规划算向聴,并指出单人「赛跑」模型把胜率高估 1.36 到 1.79 倍,误差随距离增大。 (https://5dlibrary.com/answers/zh/imperfect-information-tile-games/c04/) - Ch.5 Folding: Optimal Stopping, and a Conclusion Stated Backwards — Writing folding as a finite-horizon Markov decision process and solving it by backward induction gives a threshold policy: the higher one's own win probability the longer one pushes, and the higher the deal-in probability the shorter. The main result is a structural conclusion that is often stated backwards: the opponent's self-draw term cancels between pushing and folding, because folding protects only against dealing in, so the indifference condition contains only the deal-in probability and one's own win probability (here $p_{\text{win}} \ge 1.333\,p_{\text{deal}}$). A nearly safe tile should therefore be pushed even with a weak hand, which is the opposite of "when in doubt, fold". The framing fails in three places: the deal-in probability depends on an opponent who adapts, so this is not a decision problem; there are three opponents with dependent hands, so "fold" is not a single action; and points are not utility, so in the last hand of a match the correct policy can be a negative-expectation push. (https://5dlibrary.com/answers/en/imperfect-information-tile-games/c05/) 第5章 弃胡:最优停时,与一个被说反的结论 — 把弃牌写成有限期马尔可夫决策过程并反向归纳求解,得到阈值策略:自己的和牌概率越高推得越久,放铳概率越高推得越短。主要结果是一个常被说反的结构性结论:对手自摸的项在推与弃的比较中相消,因为弃牌只防放铳不防自摸,所以无差异条件只含放铳概率与自己的和牌概率(这里是 $p_{\text{win}} \ge 1.333\,p_{\text{deal}}$)。于是一张接近安全的牌即使手牌很弱也该推,这与「拿不准就弃」相反。该框架有三处不成立:放铳概率依赖会适应的对手,所以这不是决策问题;有三家且手牌相依,所以「弃」不是单一行动;点数不是效用,最后一局追分时正确的策略可以是负期望的推。 (https://5dlibrary.com/answers/zh/imperfect-information-tile-games/c05/) - Ch.6 Solving the Game, and the Money — Chapter 6 verifies counterfactual regret minimisation on a two-player zero-sum game (exploitability from 0.667 to 0.0013, value −0.055554 against −1/18) and honestly reports two negative results on a three-player game: exploitability still falls to 4 × 10⁻⁶, and four simple deviations failed to push the player below the equilibrium value. The Kelly growth curve is nearly symmetric (half-Kelly keeps 75.3 %, 1.5× Kelly 75.7 %), so the reason to stake less is estimation error, not the curve's shape. (https://5dlibrary.com/answers/en/imperfect-information-tile-games/c06/) 第6章 博弈的解与钱 — 第六章在两人零和博弈上验证反事实遗憾最小化(可利用度从 0.667 降到 0.0013,博弈值 −0.055554,理论值 −1/18),并在三人博弈上诚实地报告两个否定结果:可利用度照样降到 4 × 10⁻⁶,四种简单偏离也没能把玩家压到均衡值以下。凯利增长率曲线在最优点附近近乎对称(半凯利保 75.3%、1.5 倍凯利保 75.7%),所以降低注码的理由是估计误差而不是曲线形状。 (https://5dlibrary.com/answers/zh/imperfect-information-tile-games/c06/) - Ch.7 Classics Read Side by Side: The Substantial and the Empty, the Orthodox and the Unorthodox, and the Fog of War — "Victory according to the enemy" and "no constant disposition" land on best response rather than equilibrium in this book's framework, which is the same side as Chapter 6's conclusion that multiplayer games have no value guarantee, so the useful question is a best response to the opponents present. "The direct joins battle, the indirect secures victory" corresponds to mixed strategies, and program p12 supplies the precise content: both deterministic policies are worse than the mix, the loss curve has a kink at the equilibrium, and the correct proportion (0.1667 here) is determined uniquely by the payoff matrix. But the value of mixing is conditional on the opponent's volume of observation, and reading a proportion takes thousands of observations, so when that volume is small the orthodox-to-unorthodox proportion hardly matters — a qualification the formalisation yields and reading the classics alone does not. Clausewitz's fog divides the labour differently from Sunzi's solid and void: it concerns the unreliability of one's own information, which in this book is "everything beyond one's own tiles is in the fog", and Chapter 4's calibration warning is its quantified version. The comparison extends only to structure, because the treatises have no fixed rules, determinate payoffs or repeated play, and every theorem here depends on all three. (https://5dlibrary.com/answers/en/imperfect-information-tile-games/c07/) 第7章 原典并读:虚实、奇正与战争的迷雾 — 《虚实篇》的「因敌制胜」「兵无常势」在本书框架里对应最优反应而非均衡,这与第六章的结论同侧——多人博弈没有价值保证,所以有用的问题是对在场对手的最优反应。《势篇》的「以正合,以奇胜」对应混合策略,而程序 p12 给出精确内容:两个确定性策略都比混合差,损失曲线在均衡处有折点,正确的比例(本例 0.1667)由支付矩阵唯一决定。但混合的价值以对手的观察量为条件,而读出一个比例需要数千次观察,所以在观察量小时奇正的比例几乎不重要——这是形式化能给出而单读古籍不能给出的限定。克劳塞维茨的迷雾与孙子的虚实分工不同:前者讲自己信息的不可靠,对应本书「手牌之外的一切都在迷雾里」,而第四章的校准警告是它的量化版本。对勘只到结构为止:兵书中没有固定规则、确定支付与重复对局,而本书的定理全依赖这三者。 (https://5dlibrary.com/answers/zh/imperfect-information-tile-games/c07/) - Ch.8 Practice and a Pre-Registered Test — Chapter 8 gives a four-step checklist and a pre-registered duplicate-design policy comparison scored at the first differing decision: smallest effect of interest 0.010 buy-ins per scored decision, 12,000 scored decisions, the decision rule written as code; the conclusion holds only against the fixed opponent pool used. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's models and hypotheses; apart from cited literature data, all data in the book are synthetic. (https://5dlibrary.com/answers/en/imperfect-information-tile-games/c08/) 第8章 实践与预注册检验 — 第八章给出四步核查表与一项预注册的重复对局策略比较,只在首次分歧的决策上计分:最小关注效应 0.010 个买入/计分决策,需 12,000 个计分决策,判定规则写成代码;结论只对所用的固定对手池成立。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的模型和假说;除注明的文献数据外,书中数据都是合成的。 (https://5dlibrary.com/answers/zh/imperfect-information-tile-games/c08/) ## 5DL-B-056 · Dealer Gamma Exposure, Zero-Day Options and Volatility Harvesting Dynamics 中文书名:做市商伽马敞口、零日到期期权与波动率收割动力学 DOI: https://doi.org/10.5281/zenodo.24090160 Cite: Fan, Y. (2026). Dealer Gamma Exposure, Zero-Day Options and Volatility Harvesting Dynamics (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090160 Book page: https://5dlibrary.com/books/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/ Scope: It is not investment advice. Boxes separate evidence, runnable Python (16 programs), sources and hypotheses; all data are synthetic unless cited. Abstract: Every option has two sides; if the other side is a dealer who does not want the direction, the dealer's hedge trades the underlying. This book writes that mechanism as mathematics, runs it, and uses it to test two popular claims: that a GEX chart predicts the day's tops and bottoms, and that selling a 0DTE iron condor every day, winning 85 % of the time, pays a steady income. It is not investment advice. Chapter 1 shows that gamma near expiry is a spike, not a wall: with thirty minutes left at-the-money gamma is 19.76 times its 30-day value, but the strike half-width over which it is large shrinks from +8.726 % to +0.413 %. Chapter 2 writes hedging feedback as one equation, dS = dS₀/(1 + κ): long gamma dampens, short gamma amplifies, and at κ = −1 the price must jump; with lagged hedging, long gamma produces mean reversion and short gamma momentum, while one-minute volatility rises in both cases (1.154 times) and the dampening appears only at the daily horizon (0.684 against 1.958). The same mechanism pins the expiry-day price to a strike dealers hold long (closes within 0.10 rise from 0.052 to 0.132) and repels it from one they hold short. Chapter 3 is the book's central negative result: the dealers' sign is not in the data a GEX chart uses. Identical open interest is consistent with dealer gamma of +18.9 or −94.5, and intraday 0DTE positions turn +18.9 into −260.7. Even when the hedging mechanism is true by construction, a correctly signed estimate correlates only −0.238 with the size of the next day's move, not at all with its direction, and the information fades to −0.010 as the sign convention fails. Chapter 4 shows under Merton jump diffusion that the same strangle wins 0.862 to 0.873 of days whether its expectation is negative, zero or positive — the win rate belongs to the strikes, the expectation to the price. "A few gaps eat all the premium" is also wrong: at the fair price jump days repay 0.232 of it. The real danger is size: the growth-optimal leverage of 12.40 sits next to the 13.6 that wipes the account out, and 13 of 25 ten-year records estimate an optimum beyond that cliff. Chapter 5 shows that overpriced tail protection raises optimal growth from +0.404 to +0.781 by raising what can be carried, while a cost of 0.005 per contract turns the best gross Sharpe ratio into the worst net one (−0.57); discrete hedging error matches √(π/4)·νσ/√N. Chapter 6 treats variance swaps: narrow strips understate variance, a down-jump pays the hedged strip only 0.9365 of the variance it adds, and there is a convexity gap of about 1.0 vol points between volatility and variance swaps. Chapter 7 reads the Guanzi's "gather when light, scatter when heavy" and the Shiji's "sell when high, buy when low" beside the long-gamma hedge, and shows with a toy model that the stabiliser earns under reverting shocks but accumulates a loss and stops damping under a lasting trend. Chapter 8 gives a four-step check and a pre-registered test (773 trading days), and states that testing "no directional effect" would need 17,316 days. Boxes separate evidence, runnable Python (16 programs), sources and hypotheses; all data are synthetic unless cited. 摘要:每张期权都有两边;如果对面是一个不想承担方向的做市商,他的对冲就会交易标的。本书把这个机制写成数学并实际运行,再用它检验两类流行说法:「看一张 GEX 图就能预测当天顶和底」,以及「每天卖 0DTE 铁鹰、85% 的时候赢、稳拿现金流」。本书不是投资建议。 第一章说明临近到期的伽马是尖峰而不是墙:剩三十分钟时平值伽马是 30 日的 19.76 倍,但伽马较大的行权价半宽从 +8.726 % 缩到 +0.413 %。第二章把对冲反馈写成一个方程 dS = dS₀/(1 + κ):多头伽马平抑、空头伽马放大、κ = −1 时价格必须跳空;滞后对冲下,多头伽马产生均值回复、空头伽马产生动量,而一分钟波动率在两种情况下都上升(1.154 倍),平抑只在日度时段出现(0.684 对 1.958)。同一机制在到期日把价格钉在做市商持有多头的行权价上(收盘落在 0.10 之内的比例从 0.052 升到 0.132),并把它从持有空头的行权价推开。 第三章是本书的核心否定结果:做市商的符号不在 GEX 图所用的数据里。相同的未平仓量与 +18.9 或 −94.5 的做市商伽马都相容,日内开平的 0DTE 头寸把 +18.9 变成 −260.7。即便对冲机制按构造成立,符号正确的估计与次日波动幅度的相关也只有 −0.238,与方向的相关与零无法区分,并随符号惯例失效而消退到 −0.010。 第四章在默顿跳跃扩散下证明:同一张宽跨式无论期望为负、为零还是为正,胜率都在 0.862 到 0.873 之间——胜率是行权价位置的属性,期望是价格的属性。「几次缺口吃光权利金」同样不对:公允价格下跳跃日只偿还 0.232 的权利金。真正的危险在规模:增长最优杠杆 12.40 紧挨着使账户清零的 13.6,25 份十年记录中有 13 份估出的最优点落在悬崖之外。第五章说明被高估的尾部保护通过提高可承载规模把最优增长率从 +0.404 提高到 +0.781,而每张合约 0.005 的成本又能把最好的毛夏普比率变成最差的净夏普比率(−0.57);离散对冲误差与 √(π/4)·νσ/√N 吻合。第六章讨论方差互换:窄条带低估方差,向下跳跃使已对冲的条带只得到所增方差的 0.9365,波动率互换与方差互换之间有约 1.0 个波动率点的凸性缺口。第七章把《管子·国蓄》「敛之以轻,散之以重」与《史记·平准书》「贵即卖之,贱则买之」与多头伽马并读,用玩具模型说明稳定者在回复冲击下获利、在持久趋势下累积亏损并停止平抑。第八章给出四步核查与一项预注册检验(773 个交易日),并如实说明「方向无效应」需要 17,316 天才能检验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说;除注明的文献外,书中数据都是合成的。 Chapter answers: - Ch.1 Gamma Near Expiry: a Spike, Not a Wall — Chapter 1 shows that gamma near expiry is a spike, not a wall: with thirty minutes left at-the-money gamma is 19.76 times its 30-day value, but the strike half-width over which it is large shrinks from +8.726 % to +0.413 %. (https://5dlibrary.com/answers/en/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c01/) 第1章 临近到期的伽马:是尖峰,不是墙 — 第一章说明临近到期的伽马是尖峰而不是墙:剩三十分钟时平值伽马是 30 日的 19.76 倍,但伽马较大的行权价半宽从 +8.726 % 缩到 +0.413 %。 (https://5dlibrary.com/answers/zh/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c01/) - Ch.2 The Hedging Feedback Loop and Strike Pinning — Chapter 2 writes hedging feedback as one equation, dS = dS₀/(1 + κ): long gamma dampens, short gamma amplifies, and at κ = −1 the price must jump; with lagged hedging, long gamma produces mean reversion and short gamma momentum, while one-minute volatility rises in both cases (1.154 times) and the dampening appears only at the daily horizon (0.684 against 1.958). The same mechanism pins the expiry-day price to a strike dealers hold long (closes within 0.10 rise from 0.052 to 0.132) and repels it from one they hold short. (https://5dlibrary.com/answers/en/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c02/) 第2章 对冲反馈环与行权价钉死 — 第二章把对冲反馈写成一个方程 dS = dS₀/(1 + κ):多头伽马平抑、空头伽马放大、κ = −1 时价格必须跳空;滞后对冲下,多头伽马产生均值回复、空头伽马产生动量,而一分钟波动率在两种情况下都上升(1.154 倍),平抑只在日度时段出现(0.684 对 1.958)。同一机制在到期日把价格钉在做市商持有多头的行权价上(收盘落在 0.10 之内的比例从 0.052 升到 0.132),并把它从持有空头的行权价推开。 (https://5dlibrary.com/answers/zh/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c02/) - Ch.3 What a GEX Chart Can Know: the Sign Is Not in the Data — Chapter 3 is the book's central negative result: the dealers' sign is not in the data a GEX chart uses. Identical open interest is consistent with dealer gamma of +18.9 or −94.5, and intraday 0DTE positions turn +18.9 into −260.7. Even when the hedging mechanism is true by construction, a correctly signed estimate correlates only −0.238 with the size of the next day's move, not at all with its direction, and the information fades to −0.010 as the sign convention fails. (https://5dlibrary.com/answers/en/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c03/) 第3章 一张 GEX 图能知道什么:符号不在数据里 — 第三章是本书的核心否定结果:做市商的符号不在 GEX 图所用的数据里。相同的未平仓量与 +18.9 或 −94.5 的做市商伽马都相容,日内开平的 0DTE 头寸把 +18.9 变成 −260.7。即便对冲机制按构造成立,符号正确的估计与次日波动幅度的相关也只有 −0.238,与方向的相关与零无法区分,并随符号惯例失效而消退到 −0.010。 (https://5dlibrary.com/answers/zh/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c03/) - Ch.4 Selling Volatility: the Win Rate Says Nothing, the Tail Says Everything — Chapter 4 shows under Merton jump diffusion that the same strangle wins 0.862 to 0.873 of days whether its expectation is negative, zero or positive — the win rate belongs to the strikes, the expectation to the price. "A few gaps eat all the premium" is also wrong: at the fair price jump days repay 0.232 of it. The real danger is size: the growth-optimal leverage of 12.40 sits next to the 13.6 that wipes the account out, and 13 of 25 ten-year records estimate an optimum beyond that cliff. (https://5dlibrary.com/answers/en/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c04/) 第4章 卖出波动率:胜率什么都没说,尾部说明一切 — 第四章在默顿跳跃扩散下证明:同一张宽跨式无论期望为负、为零还是为正,胜率都在 0.862 到 0.873 之间——胜率是行权价位置的属性,期望是价格的属性。「几次缺口吃光权利金」同样不对:公允价格下跳跃日只偿还 0.232 的权利金。真正的危险在规模:增长最优杠杆 12.40 紧挨着使账户清零的 13.6,25 份十年记录中有 13 份估出的最优点落在悬崖之外。 (https://5dlibrary.com/answers/zh/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c04/) - Ch.5 Buying the Tail Back, Hedging in Discrete Time, and the Net Sharpe Ratio — Chapter 5 shows that overpriced tail protection raises optimal growth from +0.404 to +0.781 by raising what can be carried, while a cost of 0.005 per contract turns the best gross Sharpe ratio into the worst net one (−0.57); discrete hedging error matches √(π/4)·νσ/√N. (https://5dlibrary.com/answers/en/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c05/) 第5章 买回尾部、离散时间对冲与净夏普比率 — 第五章说明被高估的尾部保护通过提高可承载规模把最优增长率从 +0.404 提高到 +0.781,而每张合约 0.005 的成本又能把最好的毛夏普比率变成最差的净夏普比率(−0.57);离散对冲误差与 √(π/4)·νσ/√N 吻合。 (https://5dlibrary.com/answers/zh/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c05/) - Ch.6 Variance Swaps: Replication, Skew, Jumps and Convexity — Chapter 6 treats variance swaps: narrow strips understate variance, a down-jump pays the hedged strip only 0.9365 of the variance it adds, and there is a convexity gap of about 1.0 vol points between volatility and variance swaps. (https://5dlibrary.com/answers/en/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c06/) 第6章 方差互换:复制、偏斜、跳跃与凸性 — 第六章讨论方差互换:窄条带低估方差,向下跳跃使已对冲的条带只得到所增方差的 0.9365,波动率互换与方差互换之间有约 1.0 个波动率点的凸性缺口。 (https://5dlibrary.com/answers/zh/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c06/) - Ch.7 Gathering and Scattering: the Guanzi, the Balanced Standard and the Limits of a Stabiliser — With reverting shocks the policy does what the Guanzi says. A store of 60 cuts the standard deviation of price changes from 0.0512 to 0.0103 and earns a trading profit of +8.30; even a store of 20, at its limit 0.274 of the time, halves the volatility and earns +7.65. "Gather when light, scatter when heavy" calms the price and pays. (https://5dlibrary.com/answers/en/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c07/) 第7章 敛散之权:《管子》、平准与稳定者的边界 — 冲击会回复时,这个政策做到了《管子》所说的。储备为 60 时,价格变动的标准差从 0.0512 降到 0.0103,交易利润为 +8.30;即使储备只有 20、有 0.274 的时间处于极限,也把波动率减半,并赚得 +7.65。「敛之以轻,散之以重」既平抑价格,又有利可图。 (https://5dlibrary.com/answers/zh/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c07/) - Ch.8 A Four-Step Check and a Pre-Registered Test — Chapter 8 gives a four-step check and a pre-registered test (773 trading days), and states that testing "no directional effect" would need 17,316 days. Boxes separate evidence, runnable Python (16 programs), sources and hypotheses; all data are synthetic unless cited. (https://5dlibrary.com/answers/en/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c08/) 第8章 四步核查与一项预注册检验 — 第八章给出四步核查与一项预注册检验(773 个交易日),并如实说明「方向无效应」需要 17,316 天才能检验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说;除注明的文献外,书中数据都是合成的。 (https://5dlibrary.com/answers/zh/dealer-gamma-exposure-pinned-strikes-volatility-harvesting/c08/) ## 5DL-B-057 · Two-Process Sleep Architecture, Slow-Wave Glymphatic Clearance and Thermoregulatory Phase Control 中文书名:双过程睡眠架构、慢波胶质淋巴清除与体温节律相位控制 DOI: https://doi.org/10.5281/zenodo.24090162 Cite: Fan, Y. (2026). Two-Process Sleep Architecture, Slow-Wave Glymphatic Clearance and Thermoregulatory Phase Control (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090162 Book page: https://5dlibrary.com/books/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/ Scope: It is not medical advice, gives no doses and recommends no product; for persistent insomnia the first-line treatment is cognitive behavioural therapy for insomnia. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses; apart from cited literature, all data in the book are synthetic. Abstract: Sleep attracts much popular advice and little arithmetic. This book writes the main models of sleep regulation as equations, runs them, and uses them to find where each popular sentence holds and where it breaks. It is not medical advice, gives no doses and recommends no product; for persistent insomnia the first-line treatment is cognitive behavioural therapy for insomnia. Chapter 1 writes Borbély's two-process model: lost sleep is repaid mainly in intensity — a full extra day awake lengthens recovery sleep only to 1.11 times usual while the slow-wave proxy reaches 1.47 times — and a morning recovery sleep is shorter than usual (0.66) because the clock ends it. Caffeine, a competitive antagonist, does not remove adenosine (bedtime pressure stays 0.626); it changes the share that is felt, and under a stated working hypothesis a slow metaboliser's morning cup cuts early slow-wave activity (−22.5 %) about as much as a fast metaboliser's cup at 17:00 (−20.3 %). Chapter 2 writes the mutual inhibition of the VLPO and the monoaminergic nuclei as a flip-flop: weakening it shrinks the bistable margin from 0.655 to nothing, and with noise time asleep falls only from 0.989 to 0.908 while awakenings per night rise from 0.12 to 35.40 — ageing fragments the night before it shortens it. Chapter 3's non-homogeneous Markov chain, with one decaying pressure and one ultradian timer, reproduces 0.69 of deep sleep in the first third of the night and a cycle of about 92 minutes; for a sleeper whose physiology never changes, deep sleep ranges from 24.0 to 131.0 minutes across the middle 90 % of nights, and detecting a real 10-minute change needs 165 nights per block. Chapter 4 tests "sleep washes the brain": a rise in extracellular volume fraction from 0.14 to 0.23 leaves effective diffusivity unchanged but raises permeability to flow by 5.53, so it matters only if flow dominates; depending on unmeasured flow speeds the predicted effect of sleep ranges from ×0.76 to ×2.16, and the 2024 finding of reduced clearance in sleep is reported. Chapter 5's three-store thermal model shows the evening fall in core temperature as a valve, the distal vessels, opening: opening it 90 minutes late delays the decline from 22:32 to 23:47, and a warm bath helps only if it ends well before lights-out. Chapter 6 shows that even a perfect device agrees with a human scorer only at kappa 0.815, that a device may find only 0.571 of deep-sleep epochs, and that the night-to-night change in a score (correlation 0.640) is noisier than the score (0.708), and writes "orthosomnia" as a feedback loop explicitly marked as a hypothesis. Chapter 7 reads the Lingshu's defensive qi beside the models: a circuit of 28.8 minutes has nothing to do with the 92-minute cycle, "by day not alert, by night not sleeping" precisely describes a weakened switch, and the longest single sleep the model allows is 11.2 hours. Chapter 8 gives a four-step check and a pre-registered crossover trial (score visible against hidden, 75 completers). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses; apart from cited literature, all data in the book are synthetic. 摘要:关于睡眠的流行建议很多,算术很少。本书把睡眠调节的主要模型写成方程并实际运行,用它们找出每一句流行说法在哪里成立、在哪里断裂。本书不是医疗建议,不给任何剂量,不推荐任何产品;持续失眠的一线治疗是失眠的认知行为疗法。 第一章写出 Borbély 双过程模型:缺失的睡眠主要以强度偿还——多醒整整一天,补觉只延长到 1.11 倍,慢波代理量却到 1.47 倍;早上补的觉反而更短(0.66),因为时钟把它结束了。咖啡因作为竞争性拮抗剂不清除腺苷(就寝压力恒为 0.626),只改变被感到的份额;在明确声明的工作假设下,慢代谢者早上一杯对早段慢波活动的削减(−22.5%)与快代谢者 17:00 的一杯(−20.3%)相当。第二章把 VLPO 与单胺能核团的相互抑制写成触发器:削弱抑制使双稳态余量从 0.655 缩到零;加入噪声后睡着的时间只从 0.989 降到 0.908,每夜觉醒却从 0.12 次增到 35.40 次——衰老先让夜晚破碎,再让它变短。第三章的非齐次马尔可夫链只用一个衰减压力与一个超日计时器,就重现了 0.69 的深睡眠落在前三分之一夜、约 92 分钟的周期;对一个生理从不改变的睡眠者,中间 90% 的夜晚深睡眠从 24.0 到 131.0 分钟不等,检测 10 分钟的真实变化每段需要 165 晚。 第四章检验「睡眠会清洗大脑」:细胞外体积分数从 0.14 升到 0.23 不改变有效扩散系数,却使对流动的渗透率提高 5.53 倍,所以只有在流动占主导时才要紧;按尚未测定的流速,睡眠的预测效果从 ×0.76 到 ×2.16,并如实报告 2024 年睡眠中清除降低的反例。第五章用三热库模型说明傍晚核心降温是远端血管这道阀门打开的结果:阀门晚开 90 分钟,核心下降从 22:32 推迟到 23:47;热水澡只有在熄灯前早早结束才有帮助。第六章说明完美设备与人工判读者的 kappa 也只有 0.815,设备只认出 0.571 的深睡眠时段,分数的逐晚变化(相关 0.640)比分数本身(0.708)更嘈杂,并把「正睡症」写成一个明确标为假说的反馈环。第七章与《灵枢》卫气之说并读:五十度一周为 28.8 分钟,与 92 分钟的周期无关;「昼不精,夜不瞑」精确描述了变弱的开关;双过程模型允许的最长单次睡眠为 11.2 小时。第八章给出四步核查与一项预注册交叉试验(分数可见与隐藏,需 75 名完成者)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说;除注明的文献外,书中数据都是合成的。 Chapter answers: - Ch.1 Sleep Pressure, the Clock, and What Caffeine Actually Does — Chapter 1 writes Borbély's two-process model: lost sleep is repaid mainly in intensity — a full extra day awake lengthens recovery sleep only to 1.11 times usual while the slow-wave proxy reaches 1.47 times — and a morning recovery sleep is shorter than usual (0.66) because the clock ends it. Caffeine, a competitive antagonist, does not remove adenosine (bedtime pressure stays 0.626); it changes the share that is felt, and under a stated working hypothesis a slow metaboliser's morning cup cuts early slow-wave activity (−22.5 %) about as much as a fast metaboliser's cup at 17:00 (−20.3 %). (https://5dlibrary.com/answers/en/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c01/) 第1章 睡眠压力、时钟,以及咖啡因到底做了什么 — 第一章写出 Borbély 双过程模型:缺失的睡眠主要以强度偿还——多醒整整一天,补觉只延长到 1.11 倍,慢波代理量却到 1.47 倍;早上补的觉反而更短(0.66),因为时钟把它结束了。咖啡因作为竞争性拮抗剂不清除腺苷(就寝压力恒为 0.626),只改变被感到的份额;在明确声明的工作假设下,慢代谢者早上一杯对早段慢波活动的削减(−22.5%)与快代谢者 17:00 的一杯(−20.3%)相当。 (https://5dlibrary.com/answers/zh/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c01/) - Ch.2 The Sleep Switch: Why Ageing Fragments the Night Before It Shortens It — Chapter 2 writes the mutual inhibition of the VLPO and the monoaminergic nuclei as a flip-flop: weakening it shrinks the bistable margin from 0.655 to nothing, and with noise time asleep falls only from 0.989 to 0.908 while awakenings per night rise from 0.12 to 35.40 — ageing fragments the night before it shortens it. (https://5dlibrary.com/answers/en/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c02/) 第2章 睡眠开关:为什么衰老先让夜晚破碎,再让它变短 — 第二章把 VLPO 与单胺能核团的相互抑制写成触发器:削弱抑制使双稳态余量从 0.655 缩到零;加入噪声后睡着的时间只从 0.989 降到 0.908,每夜觉醒却从 0.12 次增到 35.40 次——衰老先让夜晚破碎,再让它变短。 (https://5dlibrary.com/answers/zh/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c02/) - Ch.3 Sleep Architecture as a Markov Chain, and What One Night Can Tell You — Chapter 3's non-homogeneous Markov chain, with one decaying pressure and one ultradian timer, reproduces 0.69 of deep sleep in the first third of the night and a cycle of about 92 minutes; for a sleeper whose physiology never changes, deep sleep ranges from 24.0 to 131.0 minutes across the middle 90 % of nights, and detecting a real 10-minute change needs 165 nights per block. (https://5dlibrary.com/answers/en/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c03/) 第3章 作为马尔可夫链的睡眠架构,以及一晚能告诉你什么 — 第三章的非齐次马尔可夫链只用一个衰减压力与一个超日计时器,就重现了 0.69 的深睡眠落在前三分之一夜、约 92 分钟的周期;对一个生理从不改变的睡眠者,中间 90% 的夜晚深睡眠从 24.0 到 131.0 分钟不等,检测 10 分钟的真实变化每段需要 165 晚。 (https://5dlibrary.com/answers/zh/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c03/) - Ch.4 Does Sleep Wash the Brain? Diffusion, Flow and an Open Dispute — Chapter 4 tests "sleep washes the brain": a rise in extracellular volume fraction from 0.14 to 0.23 leaves effective diffusivity unchanged but raises permeability to flow by 5.53, so it matters only if flow dominates; depending on unmeasured flow speeds the predicted effect of sleep ranges from ×0.76 to ×2.16, and the 2024 finding of reduced clearance in sleep is reported. (https://5dlibrary.com/answers/en/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c04/) 第4章 睡眠会清洗大脑吗?扩散、流动与一场未决的争论 — 第四章检验「睡眠会清洗大脑」:细胞外体积分数从 0.14 升到 0.23 不改变有效扩散系数,却使对流动的渗透率提高 5.53 倍,所以只有在流动占主导时才要紧;按尚未测定的流速,睡眠的预测效果从 ×0.76 到 ×2.16,并如实报告 2024 年睡眠中清除降低的反例。 (https://5dlibrary.com/answers/zh/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c04/) - Ch.5 Warm Hands, Cool Core: the Thermal Gate of Sleep — Chapter 5's three-store thermal model shows the evening fall in core temperature as a valve, the distal vessels, opening: opening it 90 minutes late delays the decline from 22:32 to 23:47, and a warm bath helps only if it ends well before lights-out. (https://5dlibrary.com/answers/en/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c05/) 第5章 手暖,核心凉:睡眠的温度闸门 — 第五章用三热库模型说明傍晚核心降温是远端血管这道阀门打开的结果:阀门晚开 90 分钟,核心下降从 22:32 推迟到 23:47;热水澡只有在熄灯前早早结束才有帮助。 (https://5dlibrary.com/answers/zh/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c05/) - Ch.6 What a Sleep Tracker Can Know, and the Score That Keeps You Awake — Chapter 6 shows that even a perfect device agrees with a human scorer only at kappa 0.815, that a device may find only 0.571 of deep-sleep epochs, and that the night-to-night change in a score (correlation 0.640) is noisier than the score (0.708), and writes "orthosomnia" as a feedback loop explicitly marked as a hypothesis. (https://5dlibrary.com/answers/en/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c06/) 第6章 睡眠追踪器能知道什么,以及让你睡不着的那个分数 — 第六章说明完美设备与人工判读者的 kappa 也只有 0.815,设备只认出 0.571 的深睡眠时段,分数的逐晚变化(相关 0.640)比分数本身(0.708)更嘈杂,并把「正睡症」写成一个明确标为假说的反馈环。 (https://5dlibrary.com/answers/zh/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c06/) - Ch.7 The Defensive Qi by Day and by Night: the Lingshu and the Sleeping Immortal — Chapter 7 reads the Lingshu's defensive qi beside the models: a circuit of 28.8 minutes has nothing to do with the 92-minute cycle, "by day not alert, by night not sleeping" precisely describes a weakened switch, and the longest single sleep the model allows is 11.2 hours. (https://5dlibrary.com/answers/en/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c07/) 第7章 卫气昼夜:《灵枢》与睡仙 — 第七章与《灵枢》卫气之说并读:五十度一周为 28.8 分钟,与 92 分钟的周期无关;「昼不精,夜不瞑」精确描述了变弱的开关;双过程模型允许的最长单次睡眠为 11.2 小时。 (https://5dlibrary.com/answers/zh/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c07/) - Ch.8 A Four-Step Check and a Pre-Registered Trial — Chapter 8 gives a four-step check and a pre-registered crossover trial (score visible against hidden, 75 completers). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses; apart from cited literature, all data in the book are synthetic. (https://5dlibrary.com/answers/en/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c08/) 第8章 四步核查与一项预注册试验 — 第八章给出四步核查与一项预注册交叉试验(分数可见与隐藏,需 75 名完成者)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说;除注明的文献外,书中数据都是合成的。 (https://5dlibrary.com/answers/zh/two-process-sleep-architecture-glymphatic-clearance-thermoregulation/c08/) ## 5DL-B-058 · Subscription Unit Economics, Net Revenue Retention and AI Inference Scaling Laws 中文书名:订阅制单位经济学、净收入留存与人工智能推理标度律 DOI: https://doi.org/10.5281/zenodo.24090164 Cite: Fan, Y. (2026). Subscription Unit Economics, Net Revenue Retention and AI Inference Scaling Laws (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090164 Book page: https://5dlibrary.com/books/subscription-unit-economics-net-retention-ai-inference-scaling/ Scope: It is not investment or financial advice; all data are synthetic, generated by its programs. Chapter 7 reads the Shiji's "the greedy merchant makes three, the modest merchant five" and Jiran's "let no money lie idle", marking the margin-against-turnover reading as an interpretation. Abstract: Subscription businesses are run by ratios, and each ratio compresses a dynamic process into one number and hides an assumption. This book writes LTV, NRR, LTV/CAC and the margins of AI products back out as the processes they compress, runs them, and finds where each ratio holds and where it misleads. It is not investment or financial advice; all data are synthetic, generated by its programs. Chapter 1 replaces "LTV = margin / churn" with shifted-beta-geometric cohort survival: with heterogeneous customers, first-month churn understates discounted value by up to −59 %, while the undiscounted pitch-deck number using a mature base's churn overstates it by +16 % to +48 %; for Beta parameter a ≤ 1 the expected lifetime is infinite. From six months of data a single churn rate values a cohort at 1364 against a true 1966; a heterogeneity model is close when its structure holds and overstates by +33 % when hazards fall with tenure, so a range across models should be reported. Chapter 2 writes NRR as dR/dt = A + (e − c)R: with constant acquisition, revenue saturates below 100 % (a ceiling of 538 at 80 %), grows linearly at 100 % and compounds above it — a bifurcation at e = c — but expansion is bounded, so a cohort's NRR falls from 144.1 % in year one to 95.7 % in year five and a steady company's blended NRR falls from 134.6 % to 100.2 % as its base ages. Chapter 3 shows that LTV/CAC hides the cash: with identical unit economics the funding needed rises from 912,868 at no growth to 8,320,737 at 10 % monthly growth, and the business with the best slogan ratio (4.96) needs the most money and has the lowest discounted ratio (1.80). Chapter 4 computes token costs from first principles: at batch 1 the 8B model's output costs 3.012 dollars per million against 0.018 for input; batching cuts it to 0.072; long context removes the batch; agent loops re-send their history, so output is only 6.1 % of the cost of a hundred-turn conversation, and caching cuts 0.6137 to 0.0708. Chapter 5 shows that routing buys most of a large model's accuracy at a third to a half of its cost; that against a flat price the mean user sets the margin, which is −36.1 % at a log-sd of 2.5; and that faster churn of light users erodes margin from +81.3 % to +77.0 %. Chapter 6 compares Metcalfe's n² with Odlyzko and Tilly's n ln n (tenfold users giving 100.0-fold against 11.60-fold value), adds congestion so that value peaks and falls, and shows that with monetisation growing 50 % a year a revenue-on-users fit returns an exponent of 1.91 and "confirms" the wrong law. Chapter 7 reads the Shiji's "the greedy merchant makes three, the modest merchant five" and Jiran's "let no money lie idle", marking the margin-against-turnover reading as an interpretation. Chapter 8 gives a four-step check and a pre-registered price test (7,908 sign-ups per arm). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:订阅制企业靠比率经营,而每个比率都把一个动态过程压缩成一个数字、藏起一个假设。本书把 LTV、NRR、LTV/CAC 与 AI 产品的毛利还原为它们所压缩的过程并实际运行,找出比率在哪里成立、在哪里误导。本书不是投资或财务建议;全部数据为程序生成的合成数据。 第一章用移位贝塔-几何队列存活取代「LTV = 毛利 / 流失率」:客户异质时,首月流失率最多把折现价值低估 −59%,而不折现、用成熟基数流失率的路演数字高估 +16% 到 +48%;贝塔参数 a ≤ 1 时期望寿命为无穷大。用六个月数据,单一流失率给出的队列价值为 1364,真值为 1966;异质性模型在结构正确时接近、在风险率随使用时长下降时高估 +33%,所以应报告跨模型的区间。第二章把 NRR 写成 dR/dt = A + (e − c)R:获客恒定时,低于 100% 饱和(80% 的天花板为 538),等于 100% 线性,高于 100% 复利——e = c 处是分岔;但扩张有界,一个队列的 NRR 从第一年的 144.1% 降到第五年的 95.7%,一家稳定公司的混合 NRR 随基数老化从 134.6% 降到 100.2%。第三章说明 LTV/CAC 藏起了现金:单位经济学相同,所需资金从不增长时的 912,868 升到月增长 10% 时的 8,320,737;口号比率最好(4.96)的企业需要最多的钱,折现比率却最低(1.80)。 第四章从第一性原理计算 token 成本:批量为 1 时 8B 模型的输出为每百万 3.012 美元、输入 0.018;批处理降到 0.072;长上下文拿走批处理;智能体循环重发历史,一百轮对话中输出只占成本的 6.1%,缓存把 0.6137 降到 0.0708。第五章说明路由以三分之一到一半的成本买到大模型的大部分准确率;固定价格下决定毛利的是平均用户,对数标准差 2.5 时毛利率为 −36.1%,且轻度用户更快流失使毛利率从 +81.3% 侵蚀到 +77.0%。第六章比较梅特卡夫的 n² 与 Odlyzko-Tilly 的 n ln n(十倍用户分别带来 100.0 倍与 11.60 倍价值),加入拥堵后价值先升后降,并说明当变现能力每年增长 50% 时,收入对用户的拟合会给出 1.91 的指数而「证实」错误的定律。第七章与《史记·货殖列传》「贪贾三之,廉贾五之」与计然「无息币」并读,把利润率与周转率的读法明确标为诠释。第八章给出四步核查与一项预注册价格检验(每组 7,908 位注册用户)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Lifetime Value Is Not Margin Divided by Churn — Chapter 1 replaces "LTV = margin / churn" with shifted-beta-geometric cohort survival: with heterogeneous customers, first-month churn understates discounted value by up to −59 %, while the undiscounted pitch-deck number using a mature base's churn overstates it by +16 % to +48 %; for Beta parameter a ≤ 1 the expected lifetime is infinite. From six months of data a single churn rate values a cohort at 1364 against a true 1966; a heterogeneity model is close when its structure holds and overstates by +33 % when hazards fall with tenure, so a range across models should be reported. (https://5dlibrary.com/answers/en/subscription-unit-economics-net-retention-ai-inference-scaling/c01/) 第1章 客户终身价值不是毛利除以流失率 — 第一章用移位贝塔-几何队列存活取代「LTV = 毛利 / 流失率」:客户异质时,首月流失率最多把折现价值低估 −59%,而不折现、用成熟基数流失率的路演数字高估 +16% 到 +48%;贝塔参数 a ≤ 1 时期望寿命为无穷大。用六个月数据,单一流失率给出的队列价值为 1364,真值为 1966;异质性模型在结构正确时接近、在风险率随使用时长下降时高估 +33%,所以应报告跨模型的区间。 (https://5dlibrary.com/answers/zh/subscription-unit-economics-net-retention-ai-inference-scaling/c01/) - Ch.2 Net Revenue Retention: a Threshold at 100 %, and Why It Does Not Last — Chapter 2 writes NRR as dR/dt = A + (e − c)R: with constant acquisition, revenue saturates below 100 % (a ceiling of 538 at 80 %), grows linearly at 100 % and compounds above it — a bifurcation at e = c — but expansion is bounded, so a cohort's NRR falls from 144.1 % in year one to 95.7 % in year five and a steady company's blended NRR falls from 134.6 % to 100.2 % as its base ages. (https://5dlibrary.com/answers/en/subscription-unit-economics-net-retention-ai-inference-scaling/c02/) 第2章 净收入留存:100% 处的门槛,以及它为何不能持久 — 第二章把 NRR 写成 dR/dt = A + (e − c)R:获客恒定时,低于 100% 饱和(80% 的天花板为 538),等于 100% 线性,高于 100% 复利——e = c 处是分岔;但扩张有界,一个队列的 NRR 从第一年的 144.1% 降到第五年的 95.7%,一家稳定公司的混合 NRR 随基数老化从 134.6% 降到 100.2%。 (https://5dlibrary.com/answers/zh/subscription-unit-economics-net-retention-ai-inference-scaling/c02/) - Ch.3 The Cash Trough: Why Good Unit Economics Still Run Out of Money — Chapter 3 shows that LTV/CAC hides the cash: with identical unit economics the funding needed rises from 912,868 at no growth to 8,320,737 at 10 % monthly growth, and the business with the best slogan ratio (4.96) needs the most money and has the lowest discounted ratio (1.80). (https://5dlibrary.com/answers/en/subscription-unit-economics-net-retention-ai-inference-scaling/c03/) 第3章 现金低谷:为什么单位经济学很好也会把钱烧光 — 第三章说明 LTV/CAC 藏起了现金:单位经济学相同,所需资金从不增长时的 912,868 升到月增长 10% 时的 8,320,737;口号比率最好(4.96)的企业需要最多的钱,折现比率却最低(1.80)。 (https://5dlibrary.com/answers/zh/subscription-unit-economics-net-retention-ai-inference-scaling/c03/) - Ch.4 What a Token Costs: Prefill, Decode and the Memory Wall — Chapter 4 computes token costs from first principles: at batch 1 the 8B model's output costs 3.012 dollars per million against 0.018 for input; batching cuts it to 0.072; long context removes the batch; agent loops re-send their history, so output is only 6.1 % of the cost of a hundred-turn conversation, and caching cuts 0.6137 to 0.0708. (https://5dlibrary.com/answers/en/subscription-unit-economics-net-retention-ai-inference-scaling/c04/) 第4章 一个 Token 的成本:预填充、解码与内存墙 — 第四章从第一性原理计算 token 成本:批量为 1 时 8B 模型的输出为每百万 3.012 美元、输入 0.018;批处理降到 0.072;长上下文拿走批处理;智能体循环重发历史,一百轮对话中输出只占成本的 6.1%,缓存把 0.6137 降到 0.0708。 (https://5dlibrary.com/answers/zh/subscription-unit-economics-net-retention-ai-inference-scaling/c04/) - Ch.5 Heavy Users, Flat Prices and Routing — Chapter 5 shows that routing buys most of a large model's accuracy at a third to a half of its cost; that against a flat price the mean user sets the margin, which is −36.1 % at a log-sd of 2.5; and that faster churn of light users erodes margin from +81.3 % to +77.0 %. (https://5dlibrary.com/answers/en/subscription-unit-economics-net-retention-ai-inference-scaling/c05/) 第5章 重度用户、固定价格与模型路由 — 第五章说明路由以三分之一到一半的成本买到大模型的大部分准确率;固定价格下决定毛利的是平均用户,对数标准差 2.5 时毛利率为 −36.1%,且轻度用户更快流失使毛利率从 +81.3% 侵蚀到 +77.0%。 (https://5dlibrary.com/answers/zh/subscription-unit-economics-net-retention-ai-inference-scaling/c05/) - Ch.6 Network Effects: n², n log n, and the Point Where More Users Hurt — Chapter 6 compares Metcalfe's n² with Odlyzko and Tilly's n ln n (tenfold users giving 100.0-fold against 11.60-fold value), adds congestion so that value peaks and falls, and shows that with monetisation growing 50 % a year a revenue-on-users fit returns an exponent of 1.91 and "confirms" the wrong law. (https://5dlibrary.com/answers/en/subscription-unit-economics-net-retention-ai-inference-scaling/c06/) 第6章 网络效应:n²、n log n,以及用户越多越糟的那一点 — 第六章比较梅特卡夫的 n² 与 Odlyzko-Tilly 的 n ln n(十倍用户分别带来 100.0 倍与 11.60 倍价值),加入拥堵后价值先升后降,并说明当变现能力每年增长 50% 时,收入对用户的拟合会给出 1.91 的指数而「证实」错误的定律。 (https://5dlibrary.com/answers/zh/subscription-unit-economics-net-retention-ai-inference-scaling/c06/) - Ch.7 "The Greedy Merchant Makes Three, the Modest Merchant Five": Margin, Turnover and Money That Flows — The Shiji reports that the greedy merchant makes three and the modest five, and that money should flow like water and never lie idle. Read as margin against turnover — an interpretation, not the text's stated meaning — a simple model reproduces "three" and "five" with the greedy merchant charging more than twice the markup and turning his stock a quarter as often, and ten years of compounding widens the gap to 57.7-fold against 13.8-fold. The same arithmetic sets a subscription's best price where churn's price sensitivity balances the margin, below the price that maximises first-year revenue. (https://5dlibrary.com/answers/en/subscription-unit-economics-net-retention-ai-inference-scaling/c07/) 第7章 「贪贾三之,廉贾五之」:利润率、周转率与「财币欲其行如流水」 — 《史记》记载贪贾三之、廉贾五之,并说财币欲其行如流水、无息币。若读作利润率与周转率之间的权衡——这是诠释,不是文本明说的意思——一个简单模型就能重现「三」与「五」:贪贾的加价是廉贾的两倍多,存货周转却只有其四分之一;十年复利把差距拉大到 57.7 倍对 13.8 倍。同样的算术把订阅产品的最佳价格定在流失率的价格敏感度与利润率相平衡之处,低于首年收入最大的价格。 (https://5dlibrary.com/answers/zh/subscription-unit-economics-net-retention-ai-inference-scaling/c07/) - Ch.8 A Four-Step Check and a Pre-Registered Price Test — Chapter 8 gives a four-step check and a pre-registered price test (7,908 sign-ups per arm). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/subscription-unit-economics-net-retention-ai-inference-scaling/c08/) 第8章 四步核查与一项预注册的价格检验 — 第八章给出四步核查与一项预注册价格检验(每组 7,908 位注册用户)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/subscription-unit-economics-net-retention-ai-inference-scaling/c08/) ## 5DL-B-059 · Prediction Market Microstructure, Bregman Information Projection and Cross-Book Arbitrage Bounds 中文书名:预测市场微观结构、布雷格曼信息投影与跨盘口无套利边界 DOI: https://doi.org/10.5281/zenodo.24090166 Cite: Fan, Y. (2026). Prediction Market Microstructure, Bregman Information Projection and Cross-Book Arbitrage Bounds (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090166 Book page: https://5dlibrary.com/books/prediction-market-microstructure-bregman-projection-arbitrage/ Scope: It is not investment, trading or betting advice; all data are synthetic, generated by its programs. Chapter 7 reads the Guanzi's "the clever gain tenfold" and the Shiji's Bai Gui, "what others discard I take, what others take I let go", as information rent and the cost of carry, marked as interpretation. Abstract: Prediction-market prices are often read as probabilities: a price of 0.20 means a 20 % chance. This book splits that reading into three questions — how beliefs aggregate into a price, how the market mechanism pays for information, and how far a price can stray when capital is locked up — writes each as a runnable model, and computes how much of an apparent arbitrage can actually be executed. It is not investment, trading or betting advice; all data are synthetic, generated by its programs. Chapter 1 uses the models of Manski and of Wolfers and Zitzewitz to show that a price is a quantile or weighted mean of the belief distribution: a population whose mean belief is 0.100 trades at 0.167; under log utility the price is exactly the wealth-weighted mean, 0.0989, and when risk attitude goes with belief it can move to 0.1444; with dispersed beliefs and bounded prices alone, contracts priced from 0.05 to 0.10 average 0.084 against a true probability of only 0.024 — the favourite-longshot bias needs no one who loves risk. Chapter 2 checks the convex duality of Hanson's logarithmic market scoring rule: prices are a softmax, a trader's expected gain is exactly the liquidity parameter times a KL divergence (a Bregman projection), and the market maker's worst-case loss is exactly the parameter times the log of the number of outcomes, 69.31 with two; subsidy times price impact tends to a constant, 34.66. Chapter 3 builds the no-arbitrage band of capital lock-up: with a true probability of 0.99 and a 14.5 % cost of capital its lower edge is 0.9252 at six months and 0.8646 at one year, which explains why near-certain contracts sit at 0.92 to 0.96 and lets the favourite-longshot bias survive in long-dated contracts. Chapter 4 computes the executable arbitrage in a book of exclusive outcomes: for a five-outcome book whose asks sum to 0.96, the best case is +7.0 % a year, and with a fee of 0.07·P(1−P) every size loses; arbitrage created by stale quotes loses money when executed leg by leg. Chapter 5 puts logically related contracts into a convex polytope and uses a hand-written Frank-Wolfe projection to find a certificate that gains in every state; the plain variant is unconverged after 400 steps while the pairwise variant converges exactly; pairwise logical rules catch only 92.9 % of incoherent quotes, and the best executed return is only +4.6 % a year. Chapter 6 shows that the bias is hard to confirm: with 1,000 independent contracts a test has a power of only 0.575, with a within-event correlation of 0.3 1,000 contracts are worth 139, and a flexible correction makes forecasts worse. Chapter 7 reads the Guanzi's "the clever gain tenfold" and the Shiji's Bai Gui, "what others discard I take, what others take I let go", as information rent and the cost of carry, marked as interpretation. Chapter 8 gives a four-step check and a pre-registered test of the lock-up discount (600 events, power 0.950). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:预测市场的价格常被读作概率:价格 0.20,事件就有 20% 的可能。本书把这个读法拆成三件事——信念如何聚合成价格、做市机制如何为信息付钱、资本被锁住时价格还能偏离多远——逐一写成可运行的模型,并计算看上去的套利还剩多少能执行。本书不是投资、交易或博彩建议;全部数据为程序生成的合成数据。 第一章用 Manski 与 Wolfers-Zitzewitz 的模型说明价格是信念分布的分位数或加权均值:平均只信 0.100 的人群,价格是 0.167;对数效用下价格恰为财富加权均值 0.0989,风险态度与信念相关时可偏离到 0.1444;只要信念分散、价格有界,价格 0.05 到 0.10 的合约平均价 0.084,真实概率只有 0.024——冷热门偏差不需要任何人爱好风险。第二章核对 Hanson 对数市场评分规则的凸对偶:价格是 Softmax,交易者的期望收益恰为流动性参数乘以 KL 散度(布雷格曼投影),做市商最坏亏损恰为参数乘以结果数的对数,两个结果时 69.31;补贴与价格冲击之积趋于常数 34.66。第三章建立资金锁仓的无套利带:真实概率 0.99、资金成本 14.5% 时,半年期合约的下沿是 0.9252,一年期是 0.8646,这解释了几乎确定的合约为何停在 0.92 到 0.96,也让冷热门偏差在长期合约中存活。 第四章计算互斥结果盘口的可执行套利:卖价之和 0.96 的五结果盘口,最好的情形年化 +7.0%,按 0.07·P(1−P) 收费时任何规模都亏钱;陈旧报价制造的套利逐腿执行后平均亏损。第五章把逻辑相关的合约放进凸多面体,手写 Frank-Wolfe 投影求出在所有状态下都赚钱的套利证书;普通变体走 400 步仍未收敛,成对变体精确收敛;两两逻辑规则只抓住 92.9% 的不一致,执行后最好的年化回报只有 +4.6%。第六章说明偏差小到难以确认:1,000 份独立合约的检验把握度只有 0.575,事件内相关系数 0.3 时 1,000 份合约只相当于 139 份,灵活的修正使预测变差。第七章把《管子·国蓄》的「智者有什倍人之功」与《史记》白圭的「人弃我取,人取我与」读作信息租与持有成本,并明确标为诠释。第八章给出四步核查与一项预注册的锁仓检验(600 个事件,把握度 0.950)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A Price Is Not a Probability: Belief Dispersion, Risk Attitudes and the Favourite-Longshot Bias — Chapter 1 uses the models of Manski and of Wolfers and Zitzewitz to show that a price is a quantile or weighted mean of the belief distribution: a population whose mean belief is 0.100 trades at 0.167; under log utility the price is exactly the wealth-weighted mean, 0.0989, and when risk attitude goes with belief it can move to 0.1444; with dispersed beliefs and bounded prices alone, contracts priced from 0.05 to 0.10 average 0.084 against a true probability of only 0.024 — the favourite-longshot bias needs no one who loves risk. (https://5dlibrary.com/answers/en/prediction-market-microstructure-bregman-projection-arbitrage/c01/) 第1章 价格不是概率:信念分布、风险态度与冷热门偏差 — 第一章用 Manski 与 Wolfers-Zitzewitz 的模型说明价格是信念分布的分位数或加权均值:平均只信 0.100 的人群,价格是 0.167;对数效用下价格恰为财富加权均值 0.0989,风险态度与信念相关时可偏离到 0.1444;只要信念分散、价格有界,价格 0.05 到 0.10 的合约平均价 0.084,真实概率只有 0.024——冷热门偏差不需要任何人爱好风险。 (https://5dlibrary.com/answers/zh/prediction-market-microstructure-bregman-projection-arbitrage/c01/) - Ch.2 The Logarithmic Market Scoring Rule: Softmax, KL Divergence and the Subsidy Cap — Chapter 2 checks the convex duality of Hanson's logarithmic market scoring rule: prices are a softmax, a trader's expected gain is exactly the liquidity parameter times a KL divergence (a Bregman projection), and the market maker's worst-case loss is exactly the parameter times the log of the number of outcomes, 69.31 with two; subsidy times price impact tends to a constant, 34.66. (https://5dlibrary.com/answers/en/prediction-market-microstructure-bregman-projection-arbitrage/c02/) 第2章 对数市场评分规则:Softmax、KL 散度与补贴上限 — 第二章核对 Hanson 对数市场评分规则的凸对偶:价格是 Softmax,交易者的期望收益恰为流动性参数乘以 KL 散度(布雷格曼投影),做市商最坏亏损恰为参数乘以结果数的对数,两个结果时 69.31;补贴与价格冲击之积趋于常数 34.66。 (https://5dlibrary.com/answers/zh/prediction-market-microstructure-bregman-projection-arbitrage/c02/) - Ch.3 Capital Lock-Up: Why Near-Certain Contracts Sit at 0.92 to 0.96 — Chapter 3 builds the no-arbitrage band of capital lock-up: with a true probability of 0.99 and a 14.5 % cost of capital its lower edge is 0.9252 at six months and 0.8646 at one year, which explains why near-certain contracts sit at 0.92 to 0.96 and lets the favourite-longshot bias survive in long-dated contracts. (https://5dlibrary.com/answers/en/prediction-market-microstructure-bregman-projection-arbitrage/c03/) 第3章 资金锁仓:为什么几乎确定的合约停在 0.92 到 0.96 — 第三章建立资金锁仓的无套利带:真实概率 0.99、资金成本 14.5% 时,半年期合约的下沿是 0.9252,一年期是 0.8646,这解释了几乎确定的合约为何停在 0.92 到 0.96,也让冷热门偏差在长期合约中存活。 (https://5dlibrary.com/answers/zh/prediction-market-microstructure-bregman-projection-arbitrage/c03/) - Ch.4 Books of Exclusive Outcomes: Executable Arbitrage When Prices Do Not Sum to 1 — Chapter 4 computes the executable arbitrage in a book of exclusive outcomes: for a five-outcome book whose asks sum to 0.96, the best case is +7.0 % a year, and with a fee of 0.07·P(1−P) every size loses; arbitrage created by stale quotes loses money when executed leg by leg. (https://5dlibrary.com/answers/en/prediction-market-microstructure-bregman-projection-arbitrage/c04/) 第4章 互斥结果的盘口:价格之和不等于 1 时的可执行套利 — 第四章计算互斥结果盘口的可执行套利:卖价之和 0.96 的五结果盘口,最好的情形年化 +7.0%,按 0.07·P(1−P) 收费时任何规模都亏钱;陈旧报价制造的套利逐腿执行后平均亏损。 (https://5dlibrary.com/answers/zh/prediction-market-microstructure-bregman-projection-arbitrage/c04/) - Ch.5 Logically Related Events: Convex Polytopes, Frank-Wolfe Projection and Arbitrage Certificates — Chapter 5 puts logically related contracts into a convex polytope and uses a hand-written Frank-Wolfe projection to find a certificate that gains in every state; the plain variant is unconverged after 400 steps while the pairwise variant converges exactly; pairwise logical rules catch only 92.9 % of incoherent quotes, and the best executed return is only +4.6 % a year. (https://5dlibrary.com/answers/en/prediction-market-microstructure-bregman-projection-arbitrage/c05/) 第5章 逻辑相关的事件:凸多面体、Frank-Wolfe 投影与套利证书 — 第五章把逻辑相关的合约放进凸多面体,手写 Frank-Wolfe 投影求出在所有状态下都赚钱的套利证书;普通变体走 400 步仍未收敛,成对变体精确收敛;两两逻辑规则只抓住 92.9% 的不一致,执行后最好的年化回报只有 +4.6%。 (https://5dlibrary.com/answers/zh/prediction-market-microstructure-bregman-projection-arbitrage/c05/) - Ch.6 Reading Probabilities from Prices: Calibration Tests, Correlated Resolutions and the Cost of Correction — Chapter 6 shows that the bias is hard to confirm: with 1,000 independent contracts a test has a power of only 0.575, with a within-event correlation of 0.3 1,000 contracts are worth 139, and a flexible correction makes forecasts worse. (https://5dlibrary.com/answers/en/prediction-market-microstructure-bregman-projection-arbitrage/c06/) 第6章 从价格读概率:校准检验、相关结算与修正的代价 — 第六章说明偏差小到难以确认:1,000 份独立合约的检验把握度只有 0.575,事件内相关系数 0.3 时 1,000 份合约只相当于 139 份,灵活的修正使预测变差。 (https://5dlibrary.com/answers/zh/prediction-market-microstructure-bregman-projection-arbitrage/c06/) - Ch.7 "The Wise Gain Tenfold" and "Watching the Turn of the Times": Information Rent and the Cost of Carry — The Guanzi's "Guo xu" explains price swings by "decrees urgent or slack, goods heavy or light" and notes that those who know first "gain tenfold"; in a prediction market an insider who buys a sure contract at 0.0909 makes exactly ten times, and the market maker passes that information rent on to uninformed traders through the spread — 0.0739 when one buyer in five is informed. The Shiji records Bai Gui's "what others discard I take, what others take I let go"; read as liquidity provision and the cost of carry — an interpretation, not what the text says outright — the seasonal spread remains substantial after spoilage and the cost of capital, but the profit lies in a narrow window of time, for the same reason as Chapter 3's lock-up band. (https://5dlibrary.com/answers/en/prediction-market-microstructure-bregman-projection-arbitrage/c07/) 第7章 「智者有什倍人之功」与「乐观时变」:信息租与持有成本 — 《管子·国蓄》以「令有缓急,故物有轻重」解释物价的波动,并指出先知者「有什倍人之功」;在预测市场中,先知者以 0.0909 的价格买入一份必然兑现的合约,恰得十倍,而做市商以价差把这笔信息租转嫁给不知情的交易者——五个买家中有一个知情时,价差为 0.0739。《史记》记白圭「人弃我取,人取我与」;把它读作流动性提供与持有成本——这是诠释,不是文本明说的意思——季节价差在扣除损耗与资金成本后仍可观,但利润集中在一个狭窄的时间窗口内,这与第三章的锁仓带出于同一道理。 (https://5dlibrary.com/answers/zh/prediction-market-microstructure-bregman-projection-arbitrage/c07/) - Ch.8 A Four-Step Check and a Pre-Registered Test of the Lock-Up Discount — Chapter 8 gives a four-step check and a pre-registered test of the lock-up discount (600 events, power 0.950). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/prediction-market-microstructure-bregman-projection-arbitrage/c08/) 第8章 四步核查与一项预注册的锁仓检验 — 第八章给出四步核查与一项预注册的锁仓检验(600 个事件,把握度 0.950)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/prediction-market-microstructure-bregman-projection-arbitrage/c08/) ## 5DL-B-060 · Five-Node Cyclic Network Dynamics, Solar-Longitude Ephemeris and Multi-Scale Oscillator Perturbations 中文书名:五节点循环网络动力学、黄经历表与多尺度振子扰动 DOI: https://doi.org/10.5281/zenodo.24090168 Cite: Fan, Y. (2026). Five-Node Cyclic Network Dynamics, Solar-Longitude Ephemeris and Multi-Scale Oscillator Perturbations (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090168 Book page: https://5dlibrary.com/books/five-node-cyclic-networks-solar-ephemeris-perturbation/ Abstract: This is a book of applied mathematics that sets astronomical reckoning beside dynamical systems and asks one question of both: in a result, what comes from the structure, what from the parameters, and what from a convention? Its five-node ring borrows only the graph structure of the traditional "mutual generation" and "mutual restraint", and makes no claim about its meaning in medicine, divination or any real system; apart from the Sun's position and calendar constants, all data are synthetic, generated by the book's programs. Chapter 1 solves Kepler's equation by Newton's method (at most three steps to machine precision at Earth's eccentricity) and finds the equation of time running from −14.2 to +16.4 minutes, with an eccentricity part of amplitude 7.66 minutes and an obliquity part of 9.86; of the 2,400 solar terms of 2001-2100, the share dated differently by the UTC+8 clock and by local solar time runs from 0.54 % on the zone meridian to 9.17 % at Urumqi. Chapter 2 writes down the circulant spectrum of the five-node generation and skip-one-inhibition ring, shows that without self-limitation the uniform state is always unstable, finds the Hopf threshold c1·cos72° + c2·cos36°, and shows that no oscillation is born when the coupling ratio equals the golden ratio, 1.618; the bifurcation is supercritical; the log-sum is conserved only in the zero-sum ring (drift 4.71e-11 over 2,000 time units) and changes by 1.89 within a cycle of the generation-inhibition ring. Chapter 3 shows the limit cycle to be a wave travelling in the direction of generation, whose amplitude envelope does not depend on the start in a dissipative ring (0.3149) and is set by the initial weights only in the zero-sum ring. Chapter 4 forces the ring with a decadal square wave and a yearly sine: what drives it is the square wave's fifth harmonic (gain 1.8654), not the yearly sine (0.1648); the symmetric ring's Jacobian is normal and cannot amplify transiently, and even ±40 % weights amplify only 1.245 times. Chapter 5 shows that a one-node record cannot resolve the golden-ratio direction the symmetry predicts (cosine 0.999996, condition number 11361.2); in a forced, damped ring the dates of extremes are locked to the forcing, while a free oscillation's date 100 years ahead is uncertain by about 0.23 years. Chapter 6 finds mean and true solar terms up to 2.386 days apart, shows that equal- and unequal-interval quadratic interpolation both cut the even-motion error about sevenfold, and gives the drift per century of the Santong, Dayan and Shoushi constants. Chapter 7 sets the Huainanzi's "forty-six days from the winter solstice comes Spring Begins" against the true terms and shows the extra day serves to close the year's count. Chapter 8 gives four questions and two software-only pre-registered tests, and reports that one of the book's own rules failed (33/40, Wilson lower bound 0.680). 摘要:本书是一本应用数学书,把天文历算与动力系统放在一起,追问同一个问题:一个结果中,哪些来自结构,哪些来自参数,哪些来自约定。书中的五节点环只借用传统「相生」「相克」的图结构,不对它在医学、命理或任何真实系统中的含义作任何主张;除太阳位置与历法常数外,全部数据为程序生成的合成数据。 第一章用牛顿法解开普勒方程(地球偏心率下最多三步到机器精度),算出全年均时差在 −14.2 到 +16.4 分钟之间,其中偏心率分量幅度 7.66 分钟、黄赤交角分量 9.86 分钟;2001 至 2100 年的 2,400 个节气中,按东八区钟表与按当地太阳时记日不同的比例,从时区基准经线的 0.54% 到乌鲁木齐的 9.17%。第二章写出五节点生成-隔一抑制环的循环矩阵谱,证明没有自身饱和时均匀状态总不稳定,Hopf 阈值为 c1·cos72° + c2·cos36°,耦合比等于黄金比 1.618 时没有振荡诞生;分岔是超临界的;对数和只在零和环中守恒(2,000 个时间单位漂移 4.71e-11),在生成-抑制环中一个周期变化 1.89。第三章说明极限环是沿生成方向传播的行波,振幅包络在耗散环中与初值无关(0.3149),只在零和环中由初始权重决定。第四章给环加上十年方波与一年正弦两种强迫:推动环的是方波的第五次谐波(增益 1.8654),而不是一年正弦(0.1648);对称环的雅可比矩阵是正规的,没有瞬态放大,权重偏离 ±40% 时放大也只有 1.245 倍。 第五章说明单节点记录分辨不了由对称性预言的黄金比方向(夹角余弦 0.999996,条件数 11361.2);受强迫的阻尼环,其极值时刻被强迫锁定,而自由振荡 100 年后的时刻误差约 0.23 年。第六章算出平气与定气最多相差 2.386 天、等间距与不等间距二次插值都把匀速假定的误差缩小约七倍,并给出三统历、大衍历、授时历常数每世纪的漂移。第七章把《淮南子·天文训》「距日冬至四十六日而立春」与真实节气对照,说明多出的一日服务于全年计数的闭合。第八章给出四个问题与两项纯软件的预注册检验,并如实报告本书自己的一条规则没有通过(33/40,Wilson 下限 0.680)。 Chapter answers: - Ch.1 Kepler's Equation and the Equation of Time: Clock Time, Solar Time and the Dates of the Solar Terms — Chapter 1 solves Kepler's equation by Newton's method (at most three steps to machine precision at Earth's eccentricity) and finds the equation of time running from −14.2 to +16.4 minutes, with an eccentricity part of amplitude 7.66 minutes and an obliquity part of 9.86; of the 2,400 solar terms of 2001-2100, the share dated differently by the UTC+8 clock and by local solar time runs from 0.54 % on the zone meridian to 9.17 % at Urumqi. (https://5dlibrary.com/answers/en/five-node-cyclic-networks-solar-ephemeris-perturbation/c01/) 第1章 开普勒方程与均时差:钟表时间、太阳时间与节气的日期 — 第一章用牛顿法解开普勒方程(地球偏心率下最多三步到机器精度),算出全年均时差在 −14.2 到 +16.4 分钟之间,其中偏心率分量幅度 7.66 分钟、黄赤交角分量 9.86 分钟;2001 至 2100 年的 2,400 个节气中,按东八区钟表与按当地太阳时记日不同的比例,从时区基准经线的 0.54% 到乌鲁木齐的 9.17%。 (https://5dlibrary.com/answers/zh/five-node-cyclic-networks-solar-ephemeris-perturbation/c01/) - Ch.2 The Five-Node Cyclic Digraph: Circulant Spectrum, Hopf Bifurcation and the Log-Sum — Chapter 2 writes down the circulant spectrum of the five-node generation and skip-one-inhibition ring, shows that without self-limitation the uniform state is always unstable, finds the Hopf threshold c1·cos72° + c2·cos36°, and shows that no oscillation is born when the coupling ratio equals the golden ratio, 1.618; the bifurcation is supercritical; the log-sum is conserved only in the zero-sum ring (drift 4.71e-11 over 2,000 time units) and changes by 1.89 within a cycle of the generation-inhibition ring. (https://5dlibrary.com/answers/en/five-node-cyclic-networks-solar-ephemeris-perturbation/c02/) 第2章 五节点循环有向图:循环矩阵谱、Hopf 分岔与对数和 — 第二章写出五节点生成-隔一抑制环的循环矩阵谱,证明没有自身饱和时均匀状态总不稳定,Hopf 阈值为 c1·cos72° + c2·cos36°,耦合比等于黄金比 1.618 时没有振荡诞生;分岔是超临界的;对数和只在零和环中守恒(2,000 个时间单位漂移 4.71e-11),在生成-抑制环中一个周期变化 1.89。 (https://5dlibrary.com/answers/zh/five-node-cyclic-networks-solar-ephemeris-perturbation/c02/) - Ch.3 The Limit Cycle: Travelling Wave, Drifting Period and Amplitude Envelope — Chapter 3 shows the limit cycle to be a wave travelling in the direction of generation, whose amplitude envelope does not depend on the start in a dissipative ring (0.3149) and is set by the initial weights only in the zero-sum ring. (https://5dlibrary.com/answers/en/five-node-cyclic-networks-solar-ephemeris-perturbation/c03/) 第3章 极限环:行波、周期漂移与振幅包络 — 第三章说明极限环是沿生成方向传播的行波,振幅包络在耗散环中与初值无关(0.3149),只在零和环中由初始权重决定。 (https://5dlibrary.com/answers/zh/five-node-cyclic-networks-solar-ephemeris-perturbation/c03/) - Ch.4 Two-Scale Forcing: Resonant Gain, Peaks and Troughs, and the Jacobian's Condition Number — Chapter 4 forces the ring with a decadal square wave and a yearly sine: what drives it is the square wave's fifth harmonic (gain 1.8654), not the yearly sine (0.1648); the symmetric ring's Jacobian is normal and cannot amplify transiently, and even ±40 % weights amplify only 1.245 times. (https://5dlibrary.com/answers/en/five-node-cyclic-networks-solar-ephemeris-perturbation/c04/) 第4章 双尺度强迫:共振增益、波峰波谷与雅可比条件数 — 第四章给环加上十年方波与一年正弦两种强迫:推动环的是方波的第五次谐波(增益 1.8654),而不是一年正弦(0.1648);对称环的雅可比矩阵是正规的,没有瞬态放大,权重偏离 ±40% 时放大也只有 1.245 倍。 (https://5dlibrary.com/answers/zh/five-node-cyclic-networks-solar-ephemeris-perturbation/c04/) - Ch.5 A Finite Record: Identifiability, the Golden-Ratio Direction and Forecasting the Dates of Extremes — Chapter 5 shows that a one-node record cannot resolve the golden-ratio direction the symmetry predicts (cosine 0.999996, condition number 11361.2); in a forced, damped ring the dates of extremes are locked to the forcing, while a free oscillation's date 100 years ahead is uncertain by about 0.23 years. (https://5dlibrary.com/answers/en/five-node-cyclic-networks-solar-ephemeris-perturbation/c05/) 第5章 有限观测:可辨识性、黄金比方向与极值时刻的预报 — 第五章说明单节点记录分辨不了由对称性预言的黄金比方向(夹角余弦 0.999996,条件数 11361.2);受强迫的阻尼环,其极值时刻被强迫锁定,而自由振荡 100 年后的时刻误差约 0.23 年。 (https://5dlibrary.com/answers/zh/five-node-cyclic-networks-solar-ephemeris-perturbation/c05/) - Ch.6 Mean and True Solar Terms, Unequal-Interval Interpolation and the Drift of Calendar Constants — Chapter 6 finds mean and true solar terms up to 2.386 days apart, shows that equal- and unequal-interval quadratic interpolation both cut the even-motion error about sevenfold, and gives the drift per century of the Santong, Dayan and Shoushi constants. (https://5dlibrary.com/answers/en/five-node-cyclic-networks-solar-ephemeris-perturbation/c06/) 第6章 平气与定气、不等间距插值与历法常数的漂移 — 第六章算出平气与定气最多相差 2.386 天、等间距与不等间距二次插值都把匀速假定的误差缩小约七倍,并给出三统历、大衍历、授时历常数每世纪的漂移。 (https://5dlibrary.com/answers/zh/five-node-cyclic-networks-solar-ephemeris-perturbation/c06/) - Ch.7 "Fifteen Days Make One Node": the Huainanzi's Counting Rule and the Long Coexistence of Mean and True Terms — The Huainanzi's "Tianwen" counts fifteen days a node and puts Spring Begins forty-six days after the winter solstice. Against the true terms, plain fifteen-day nodes fall 5.24 days short by the year's end, and a day at each corner brings the largest error down to 2.78 days (3.59 in the Han sky); but in neither today's sky nor the Han one is day 46 closer to the true Spring Begins than day 45: the extra day serves to close the year's count, not to follow the Sun's uneven motion. Mean and true terms coexisted in the Chinese calendar for over a thousand years until the Shixian calendar made true terms official — the same structure as Chapter 1's split between clock time and solar time. (https://5dlibrary.com/answers/en/five-node-cyclic-networks-solar-ephemeris-perturbation/c07/) 第7章 「十五日为一节」:《淮南子·天文训》的节点规则与平气、定气之争 — 《淮南子·天文训》以十五日为一节、立春在冬至后四十六日。与真实节气相比,单纯十五日一节到年终少 5.24 日,四维各加一日后最大误差降到 2.78 日(汉代天象下 3.59 日),但第 46 日在今天与汉代的天空中都不比第 45 日更接近真实的立春:多出的一日服务于全年计数的闭合,而非太阳运动的不均匀。平气与定气在中国历法中并存千余年,直到时宪历才统一为定气;这与第一章钟表时间和太阳时间的分歧是同一种结构。 (https://5dlibrary.com/answers/zh/five-node-cyclic-networks-solar-ephemeris-perturbation/c07/) - Ch.8 Four Questions and Two Software-Only Pre-Registered Tests — Chapter 8 gives four questions and two software-only pre-registered tests, and reports that one of the book's own rules failed (33/40, Wilson lower bound 0.680). (https://5dlibrary.com/answers/en/five-node-cyclic-networks-solar-ephemeris-perturbation/c08/) 第8章 四个问题与两项纯软件的预注册检验 — 第八章给出四个问题与两项纯软件的预注册检验,并如实报告本书自己的一条规则没有通过(33/40,Wilson 下限 0.680)。 (https://5dlibrary.com/answers/zh/five-node-cyclic-networks-solar-ephemeris-perturbation/c08/) ## 5DL-B-061 · Retrieval-Augmented Generation Geometry, Reciprocal Rank Fusion and Multi-Agent Error Cascades 中文书名:检索增强生成几何、倒数秩融合与多智能体误差级联 DOI: https://doi.org/10.5281/zenodo.24090170 Cite: Fan, Y. (2026). Retrieval-Augmented Generation Geometry, Reciprocal Rank Fusion and Multi-Agent Error Cascades (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090170 Book page: https://5dlibrary.com/books/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/ Scope: All data are synthetic, generated by the book's programs; the numbers show mechanisms and orders of magnitude and are not measurements of any product. Abstract: Retrieval-augmented generation (RAG) and multi-agent pipelines can be split into three links: retrieval, assembly and the chain. Each has its popular intuition — higher dimension and more chunks are better; a longer context is better; more agents are better. This book writes each intuition as a runnable model and finds where it holds and where it fails. All data are synthetic, generated by the book's programs; the numbers show mechanisms and orders of magnitude and are not measurements of any product. Chapter 1 uses concentration of measure, effective dimension and hubness to show that the larger the corpus, the higher the top-1 error (from 0.0730 to 0.5030 at 256 dimensions); that with 768 nominal dimensions and a power-law spectrum only 31.7 effective dimensions remain, putting the wrong chunk first among a hundred thousand with probability 0.9875; and that in a hundred dimensions 1 % of hub chunks intrude on the top ten of 77.7 % of unrelated questions. Chapter 2 shows that dense retrieval cannot recognise identifiers and lexical retrieval cannot recognise paraphrases; fusing complete rankings leaves both query types at a recall@1 of 0.055, and fusing short lists leaves identifier queries at 0.475; reciprocal rank fusion's smoothing constant k = 60 is not the best for recall@1 in any of three scenarios, and with one strong and one weak retriever no k beats the strong one alone (0.604). Chapter 3 uses a stylised reader model to show "lost in the middle" and evidence dilution: a gold chunk in the middle is answered correctly only about a third as often as at either end (0.098 against 0.288), and the best number of chunks falls from 8 to 2 as distractors grow plausible. Chapter 4 confirms that 0.92 to the tenth power is 0.4344, shows that three-way voting helps only while errors are independent, and writes verifier gating as an absorbing Markov chain: a deterministic, independent check lifts a ten-step chain to 0.9949 for only 10.84 attempts. Chapter 5 shows that a verifier's value lies in independence: if hard steps stay hard even a deterministic check reaches only 0.886, and a judge sharing the generator's blind spot only 0.6950; reliability is paid for in tail latency. Chapter 6 shows that a 3-point improvement needs about three thousand queries to be detected reliably (power 0.989 at 3,000) and that picking the best of 256 equally good configurations inflates it by more than seven points. Chapter 7 reads the Qilue as routing by category (routing must be more than 0.656 accurate to beat searching everything) and the Han Feizi's canyan as checking against several sources (sources sharing an origin are worth one however many). Chapter 8 gives a five-step check and a pre-registered offline evaluation, and shows as it fell a wrong conclusion reached through bad luck, with its error rate (0.045). 摘要:检索增强生成(RAG)与多智能体流水线可以拆成三个环节:检索、拼接与链条。每个环节都有一条流行的直觉——维度越高、块越多越好;上下文越长越好;智能体越多越好。本书把每条直觉写成可运行的模型,找出它在哪里成立、在哪里失效。全部数据为程序生成的合成数据,数字说明机制与量级,不是对任何产品的测量。 第一章用测度集中、有效维度与枢纽现象说明:语料越大,第一名错误率越高(256 维时从 0.0730 升到 0.5030);名义 768 维、方差按幂律衰减时,有效维度只有 31.7,十万块中排错第一名的概率 0.9875;一百维时 1% 的枢纽块闯入 77.7% 的无关问题的前十名。第二章说明稠密检索认不出编号、词法检索认不出改写,融合完整排名时两类查询的 recall@1 都只剩 0.055,只融合短列表时编号查询停在 0.475;倒数秩融合的平滑常数 k = 60 在三种情形下都不是 recall@1 的最优值,一强一弱时任何 k 都不如强检索器单用(0.604)。第三章用程式化读者模型说明「中间遗失」与证据稀释:正确块在中间时回答正确的概率只有两端的三分之一左右(0.098 对 0.288),最佳块数随干扰强度从 8 降到 2。 第四章证实 0.92 的十次方等于 0.4344,三票表决只在错误独立时有效,并把验证器门控写成吸收马尔可夫链:确定、独立的检查把十步链提到 0.9949,只需 10.84 次尝试。第五章说明验证器的价值在于独立:困难步骤一直困难时,即使确定性检查也只到 0.886,与生成器共享盲区的评判只到 0.6950;可靠性的代价付在尾部延迟上。第六章说明三个百分点的改进需要约三千条查询才能可靠检出(3,000 条时把握度 0.989),从 256 个同样好的配置中挑最好的会虚高七个多百分点。第七章把《七略》读作分类路由(路由准确率须高于 0.656 才胜过全库检索),把《韩非子》的「参验」读作多源核验(来源同出一源时多少个都只相当于一个)。第八章给出五步核查与一项预注册的离线评测,并如实展示一次「坏运气」得出的错误结论及其错误率(0.045)。 Chapter answers: - Ch.1 High-Dimensional Geometry: Concentration of Measure, Effective Dimension and Hub Chunks — Chapter 1 uses concentration of measure, effective dimension and hubness to show that the larger the corpus, the higher the top-1 error (from 0.0730 to 0.5030 at 256 dimensions); that with 768 nominal dimensions and a power-law spectrum only 31.7 effective dimensions remain, putting the wrong chunk first among a hundred thousand with probability 0.9875; and that in a hundred dimensions 1 % of hub chunks intrude on the top ten of 77.7 % of unrelated questions. (https://5dlibrary.com/answers/en/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c01/) 第1章 高维几何:测度集中、有效维度与枢纽块 — 第一章用测度集中、有效维度与枢纽现象说明:语料越大,第一名错误率越高(256 维时从 0.0730 升到 0.5030);名义 768 维、方差按幂律衰减时,有效维度只有 31.7,十万块中排错第一名的概率 0.9875;一百维时 1% 的枢纽块闯入 77.7% 的无关问题的前十名。 (https://5dlibrary.com/answers/zh/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c01/) - Ch.2 Lexical and Dense Retrieval: Their Blind Spots and the Smoothing Constant of Reciprocal Rank Fusion — Chapter 2 shows that dense retrieval cannot recognise identifiers and lexical retrieval cannot recognise paraphrases; fusing complete rankings leaves both query types at a recall@1 of 0.055, and fusing short lists leaves identifier queries at 0.475; reciprocal rank fusion's smoothing constant k = 60 is not the best for recall@1 in any of three scenarios, and with one strong and one weak retriever no k beats the strong one alone (0.604). (https://5dlibrary.com/answers/en/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c02/) 第2章 词法与稠密检索:各自的盲区与倒数秩融合的平滑常数 — 第二章说明稠密检索认不出编号、词法检索认不出改写,融合完整排名时两类查询的 recall@1 都只剩 0.055,只融合短列表时编号查询停在 0.475;倒数秩融合的平滑常数 k = 60 在三种情形下都不是 recall@1 的最优值,一强一弱时任何 k 都不如强检索器单用(0.604)。 (https://5dlibrary.com/answers/zh/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c02/) - Ch.3 The Context Window: Position Decay, Evidence Dilution and the Best Number of Chunks — Chapter 3 uses a stylised reader model to show "lost in the middle" and evidence dilution: a gold chunk in the middle is answered correctly only about a third as often as at either end (0.098 against 0.288), and the best number of chunks falls from 8 to 2 as distractors grow plausible. (https://5dlibrary.com/answers/en/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c03/) 第3章 上下文窗口:位置衰减、证据稀释与最佳块数 — 第三章用程式化读者模型说明「中间遗失」与证据稀释:正确块在中间时回答正确的概率只有两端的三分之一左右(0.098 对 0.288),最佳块数随干扰强度从 8 降到 2。 (https://5dlibrary.com/answers/zh/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c03/) - Ch.4 Multi-Agent Pipelines: Error Cascades and the Verifier-Gated Absorbing Markov Chain — Chapter 4 confirms that 0.92 to the tenth power is 0.4344, shows that three-way voting helps only while errors are independent, and writes verifier gating as an absorbing Markov chain: a deterministic, independent check lifts a ten-step chain to 0.9949 for only 10.84 attempts. (https://5dlibrary.com/answers/en/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c04/) 第4章 多智能体流水线:误差级联与验证器门控的吸收马尔可夫链 — 第四章证实 0.92 的十次方等于 0.4344,三票表决只在错误独立时有效,并把验证器门控写成吸收马尔可夫链:确定、独立的检查把十步链提到 0.9949,只需 10.84 次尝试。 (https://5dlibrary.com/answers/zh/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c04/) - Ch.5 Verifier Independence and the Cost of Retries — Chapter 5 shows that a verifier's value lies in independence: if hard steps stay hard even a deterministic check reaches only 0.886, and a judge sharing the generator's blind spot only 0.6950; reliability is paid for in tail latency. (https://5dlibrary.com/answers/en/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c05/) 第5章 验证器的独立性与重试的代价 — 第五章说明验证器的价值在于独立:困难步骤一直困难时,即使确定性检查也只到 0.886,与生成器共享盲区的评判只到 0.6950;可靠性的代价付在尾部延迟上。 (https://5dlibrary.com/answers/zh/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c05/) - Ch.6 The Statistics of Evaluation Itself: Sample Size, Judge Bias and the Winner's Curse — Chapter 6 shows that a 3-point improvement needs about three thousand queries to be detected reliably (power 0.989 at 3,000) and that picking the best of 256 equally good configurations inflates it by more than seven points. (https://5dlibrary.com/answers/en/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c06/) 第6章 评测本身的统计:样本量、评判偏差与赢家诅咒 — 第六章说明三个百分点的改进需要约三千条查询才能可靠检出(3,000 条时把握度 0.989),从 256 个同样好的配置中挑最好的会虚高七个多百分点。 (https://5dlibrary.com/answers/zh/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c06/) - Ch.7 The Routing of the Qilue and Han Fei's Canyan: Two Ancient Methods of Retrieval and Verification — Chapter 7 reads the Qilue as routing by category (routing must be more than 0.656 accurate to beat searching everything) and the Han Feizi's canyan as checking against several sources (sources sharing an origin are worth one however many). (https://5dlibrary.com/answers/en/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c07/) 第7章 《七略》的分类路由与《韩非子》的参验:检索与核验的两种古法 — 第七章把《七略》读作分类路由(路由准确率须高于 0.656 才胜过全库检索),把《韩非子》的「参验」读作多源核验(来源同出一源时多少个都只相当于一个)。 (https://5dlibrary.com/answers/zh/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c07/) - Ch.8 A Five-Step Check and a Pre-Registered Offline Evaluation — Chapter 8 gives a five-step check and a pre-registered offline evaluation, and shows as it fell a wrong conclusion reached through bad luck, with its error rate (0.045). (https://5dlibrary.com/answers/en/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c08/) 第8章 五步核查与一项预注册的离线评测 — 第八章给出五步核查与一项预注册的离线评测,并如实展示一次「坏运气」得出的错误结论及其错误率(0.045)。 (https://5dlibrary.com/answers/zh/rag-vector-geometry-reciprocal-rank-fusion-agent-cascades/c08/) ## 5DL-B-062 · Dyadic Phase Coupling, Bilinear Compatibility Tensors and Spatiotemporal Grid Games 中文书名:双人相位耦合、双线性相容张量与时空网格博弈 DOI: https://doi.org/10.5281/zenodo.24090172 Cite: Fan, Y. (2026). Dyadic Phase Coupling, Bilinear Compatibility Tensors and Spatiotemporal Grid Games (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090172 Book page: https://5dlibrary.com/books/dyadic-phase-coupling-bilinear-tensors-grid-games/ Scope: All data are synthetic, generated by its programs, and wherever physics or game theory is applied to people it is marked as an analogy. Chapter 7 reads the Tuan commentary on Xian and Heng, the Rhapsody of the Old Fisherman and Yang Hui's magic squares; that Xian upside down is Heng is read, as marked interpretation, as a transposed matrix. Abstract: Stripped of their names, rules for matching charts, choosing dates and choosing directions on the nine palaces are bilinear forms, coupled oscillators, first-passage times, permutation groups and games. This book studies only these mathematical structures; it neither claims nor tests any divinatory prediction. All data are synthetic, generated by its programs, and wherever physics or game theory is applied to people it is marked as an analogy. Chapter 1 writes a match score as a bilinear form and splits its matrix uniquely into a symmetric and an antisymmetric part: the larger the antisymmetric part, the more one-way pairs of opposite sign, and at a ratio of 1.0 "the person I give most to" and "the person who gives most to me" differ in 0.985 of cases; from an outcome the two share, only 10 of 16 parameters are identified, so one-sidedness can be seen only by measuring each person separately. Chapter 2 uses the asymmetric Adler-Kuramoto equations to show that how easily a pair locks depends only on the sum of the two couplings, while the asymmetry decides who adapts (a common rhythm of 1.2700 or 1.0300). Chapter 3 adds amplitude: between equal oscillators coupling costs both sides, and only when natural amplitudes differ does the weaker side gain (+0.0517); pursuit against flight still locks, with a common rhythm of 1.7500 faster than either. Chapter 4 writes the breaking of a lock as a first-passage time: with noise alone the hazard is constant and there is no danger window; a window comes from an assumed yearly shock, the hazard ratio rising from 1.09 to 11.22; testing a window with hazard ratio 2 needs about 130 dated separations, and choosing the month afterwards manufactures a window in 0.432 of studies. Chapter 5 proves that the magic square of order three is unique up to symmetry, and that the flying and turning motions generate all 362,880 arrangements, of which only 8 stay magic. Chapter 6 solves the choice of time and direction on the turning nine palaces: with the direction free, timing adds no value (a mean of 0.900); with it fixed, waiting raises the mean from −0.1500 to +0.6625; against a rival the value of the game does not change with the hour, and in equilibrium neutral doors (0.200) are used more than good ones (0.133). Chapter 7 reads the Tuan commentary on Xian and Heng, the Rhapsody of the Old Fisherman and Yang Hui's magic squares; that Xian upside down is Heng is read, as marked interpretation, as a transposed matrix. Chapter 8 gives a four-step check and a pre-registered test: picking among twenty conventions afterwards manufactures a "significant" effect in 0.517 of studies with no effect; a hazard ratio of 1.5 needs 191 separations from 764 couples. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:合盘、择日与九宫方位的规则,去掉名称以后是双线性形式、耦合振子、首达时间、置换群与博弈。本书只研究这些数学结构本身,不主张、也不检验任何术数意义上的预测;全部数据为程序生成的合成数据,物理与博弈概念用于人际关系之处一律标为类比。 第一章把合盘评分写成双线性形式,并把矩阵唯一地分成对称部分与反对称部分:反对称部分越大,符号相反的「单向」配对越多,比例为 1.0 时一个人「最适合给予的对象」与「最能给予自己的对象」在 0.985 的情形下不是同一个人;只观察两人共有的结局时,16 个参数中只有 10 个可以识别,单向性必须对每个人分别测量才能看见。第二章用非对称的阿德勒—藏本方程说明,锁相的难易只取决于两个耦合之和,不对称只决定谁迁就谁(共同节律为 1.2700 或 1.0300)。第三章加入振幅:相同的振子之间耦合让双方都付出,只有自然振幅不同时较弱的一方才获益(+0.0517);一追一躲仍能锁相,共同节律 1.7500 快于双方。第四章把锁相的打断写成首达时间:只有噪声时风险率恒定,没有危险窗口;窗口来自假设的年度冲击,风险比从 1.09 升到 11.22;检验风险比为 2 的窗口需要约 130 次带日期的分离,事后挑月份会在 0.432 的研究中制造窗口。 第五章证明三阶幻方在对称意义下唯一,飞宫与转盘两种动作生成全部 362,880 种排列,其中只有 8 种保持幻方。第六章在转动的九宫上求解择时择向:方向自由时择时不增加价值(均值 0.900),方向固定时等待把均值从 −0.1500 提高到 +0.6625;有对手时博弈值不随时辰改变,均衡中中平之门(0.200)比吉门(0.133)用得更多。第七章与咸、恒两卦的《彖传》、《烟波钓叟赋》与杨辉纵横图并读;咸卦颠倒即为恒卦,被读作矩阵的转置,并标为诠释。第八章给出四步核查与一项预注册检验:事后在二十种口径中挑选,会在 0.517 的无效应研究中制造「显著」效应;风险比 1.5 时需要 191 次分离、764 对伴侣。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A Match Score Is a Bilinear Form — Chapter 1 writes a match score as a bilinear form and splits its matrix uniquely into a symmetric and an antisymmetric part: the larger the antisymmetric part, the more one-way pairs of opposite sign, and at a ratio of 1.0 "the person I give most to" and "the person who gives most to me" differ in 0.985 of cases; from an outcome the two share, only 10 of 16 parameters are identified, so one-sidedness can be seen only by measuring each person separately. (https://5dlibrary.com/answers/en/dyadic-phase-coupling-bilinear-tensors-grid-games/c01/) 第1章 合盘评分是一个双线性形式 — 第一章把合盘评分写成双线性形式,并把矩阵唯一地分成对称部分与反对称部分:反对称部分越大,符号相反的「单向」配对越多,比例为 1.0 时一个人「最适合给予的对象」与「最能给予自己的对象」在 0.985 的情形下不是同一个人;只观察两人共有的结局时,16 个参数中只有 10 个可以识别,单向性必须对每个人分别测量才能看见。 (https://5dlibrary.com/answers/zh/dyadic-phase-coupling-bilinear-tensors-grid-games/c01/) - Ch.2 Locking, Arnold Tongues and "Who Adapts to Whom" — Chapter 2 uses the asymmetric Adler-Kuramoto equations to show that how easily a pair locks depends only on the sum of the two couplings, while the asymmetry decides who adapts (a common rhythm of 1.2700 or 1.0300). (https://5dlibrary.com/answers/en/dyadic-phase-coupling-bilinear-tensors-grid-games/c02/) 第2章 锁相、阿诺德舌与「谁迁就谁」 — 第二章用非对称的阿德勒—藏本方程说明,锁相的难易只取决于两个耦合之和,不对称只决定谁迁就谁(共同节律为 1.2700 或 1.0300)。 (https://5dlibrary.com/answers/zh/dyadic-phase-coupling-bilinear-tensors-grid-games/c02/) - Ch.3 Amplitude, "Siphoning", and Pursuit Against Flight — Chapter 3 adds amplitude: between equal oscillators coupling costs both sides, and only when natural amplitudes differ does the weaker side gain (+0.0517); pursuit against flight still locks, with a common rhythm of 1.7500 faster than either. (https://5dlibrary.com/answers/en/dyadic-phase-coupling-bilinear-tensors-grid-games/c03/) 第3章 振幅、「虹吸」与一追一躲 — 第三章加入振幅:相同的振子之间耦合让双方都付出,只有自然振幅不同时较弱的一方才获益(+0.0517);一追一躲仍能锁相,共同节律 1.7500 快于双方。 (https://5dlibrary.com/answers/zh/dyadic-phase-coupling-bilinear-tensors-grid-games/c03/) - Ch.4 The Stopping Time of Separation and the "Danger Window" — Chapter 4 writes the breaking of a lock as a first-passage time: with noise alone the hazard is constant and there is no danger window; a window comes from an assumed yearly shock, the hazard ratio rising from 1.09 to 11.22; testing a window with hazard ratio 2 needs about 130 dated separations, and choosing the month afterwards manufactures a window in 0.432 of studies. (https://5dlibrary.com/answers/en/dyadic-phase-coupling-bilinear-tensors-grid-games/c04/) 第4章 分离的停时与「高危窗口」 — 第四章把锁相的打断写成首达时间:只有噪声时风险率恒定,没有危险窗口;窗口来自假设的年度冲击,风险比从 1.09 升到 11.22;检验风险比为 2 的窗口需要约 130 次带日期的分离,事后挑月份会在 0.432 的研究中制造窗口。 (https://5dlibrary.com/answers/zh/dyadic-phase-coupling-bilinear-tensors-grid-games/c04/) - Ch.5 The Luoshu, the Eight Doors and Permutation Groups — Chapter 5 proves that the magic square of order three is unique up to symmetry, and that the flying and turning motions generate all 362,880 arrangements, of which only 8 stay magic. (https://5dlibrary.com/answers/en/dyadic-phase-coupling-bilinear-tensors-grid-games/c05/) 第5章 洛书、八门与置换群 — 第五章证明三阶幻方在对称意义下唯一,飞宫与转盘两种动作生成全部 362,880 种排列,其中只有 8 种保持幻方。 (https://5dlibrary.com/answers/zh/dyadic-phase-coupling-bilinear-tensors-grid-games/c05/) - Ch.6 When and Where to Act on a Turning Grid — Chapter 6 solves the choice of time and direction on the turning nine palaces: with the direction free, timing adds no value (a mean of 0.900); with it fixed, waiting raises the mean from −0.1500 to +0.6625; against a rival the value of the game does not change with the hour, and in equilibrium neutral doors (0.200) are used more than good ones (0.133). (https://5dlibrary.com/answers/en/dyadic-phase-coupling-bilinear-tensors-grid-games/c06/) 第6章 转动的网格上何时何方行事 — 第六章在转动的九宫上求解择时择向:方向自由时择时不增加价值(均值 0.900),方向固定时等待把均值从 −0.1500 提高到 +0.6625;有对手时博弈值不随时辰改变,均衡中中平之门(0.200)比吉门(0.133)用得更多。 (https://5dlibrary.com/answers/zh/dyadic-phase-coupling-bilinear-tensors-grid-games/c06/) - Ch.8 A Four-Step Check and a Pre-Registered Test of Chart Matching — Chapter 8 gives a four-step check and a pre-registered test: picking among twenty conventions afterwards manufactures a "significant" effect in 0.517 of studies with no effect; a hazard ratio of 1.5 needs 191 separations from 764 couples. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/dyadic-phase-coupling-bilinear-tensors-grid-games/c08/) 第8章 四步核查与一项预注册的合盘检验 — 第八章给出四步核查与一项预注册检验:事后在二十种口径中挑选,会在 0.517 的无效应研究中制造「显著」效应;风险比 1.5 时需要 191 次分离、764 对伴侣。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/dyadic-phase-coupling-bilinear-tensors-grid-games/c08/) ## 5DL-B-063 · Supercritical Viral Cascades, Information-Gap Hazard Rates and Multi-Tier Paywall Optimization 中文书名:超临界病毒级联、信息缺口风险率与多层付费墙优化 DOI: https://doi.org/10.5281/zenodo.24090174 Cite: Fan, Y. (2026). Supercritical Viral Cascades, Information-Gap Hazard Rates and Multi-Tier Paywall Optimization (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090174 Book page: https://5dlibrary.com/books/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/ Scope: It is not business, legal or investment advice; all data are synthetic, generated by its programs. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: The growth engine of a paid reading website consists of a sharing mechanism, a free hook and a ladder of paid tiers, and behind each stands a slogan. This book writes the slogans back out as models of branching processes, hazard rates, signal detection and price discrimination, runs them, and sees where the slogans hold. It is not business, legal or investment advice; all data are synthetic, generated by its programs. Chapter 1 writes "invite friends to unlock free" and "pay and earn a commission" as a Galton-Watson branching process: a sharer stops at the threshold, so the share rate caps K, and at share rates of 30 % or 50 % no threshold takes K past 1; even at K = 1.15 a single seed dies out with probability 0.729, and the median cascade is 3 users. Chapter 2 shows that in a finite audience K = 1.15 reaches only 0.255, that in a clustered network the measured K is 0.934 against a tree estimate of 1.35, and that acquisition costs do not fall to zero: 3.22 per paying user, 0.325 of the price. Chapter 3 shows that commissions hardly change who pays but raise K and total revenue per seed; when fake accounts are cheap the platform measures K at 1.992 while the real K is 0.065, and a 50 % commission becomes a 50 % discount. Chapter 4 writes the information gap as a purchase hazard: revealing half gives the highest purchase probability, 0.451, falling to 0.078 after a day's delay; a generic reading is judged 0.838 accurate, close to an informed one, yet scores only the chance level of 0.333 in a swap test, and an identity label raises sharing from 0.583 to 0.745. Chapter 5 shows under Pareto willingness to pay that at a tail index of 1.5 the ceiling of perfect discrimination is 3.00 times one price, but that for one good no menu beats one price (2.961 against 3.000); a multi-tier paywall takes more because it sells several goods, raising the share of the ceiling from 0.150 to 0.339. Chapter 6 shows that bundling independent questions collects 0.678 of their value at 20 questions, that a decoy tier depends on a parameter that must be measured, and that offering per-question pricing alongside an annual pass is best. Chapter 7 reads the "Rubbing" and "Flying and Clamping" chapters of the Guiguzi and the "State Reserves" chapter of the Guanzi; homophily raises the growth factor from 0.757 to 1.297 but confines growth to one community. Chapter 8 gives a four-step check, a compliance checklist and a pre-registered paywall test (166,788 referral trees per arm). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:付费测算网站的增长引擎由分享机制、免费钩子与付费阶梯组成,每个部件背后都有一句口号。本书把口号还原为分支过程、风险率、信号检测与价格歧视的模型,运行模型,看口号在哪里成立。本书不是经营、法律或投资建议;全部数据为程序生成的合成数据。 第一章把「拉人免费解锁」与「付费加分佣」写成高尔顿—沃森分支过程:分享者凑够门槛就停,所以 K 被分享率封顶,分享率为 30% 或 50% 时任何门槛都不能让 K 越过 1;即使 K 为 1.15,单个种子仍有 0.729 的概率熄灭,级联规模的中位数只有 3 人。第二章说明,在有限受众中 K 为 1.15 只能触达 0.255,聚集网络中实测的 K 为 0.934 而树状估计为 1.35,获客成本没有归零:每个付费用户为 3.22,占价格的 0.325。第三章说明分佣几乎不改变谁付费,却提高 K 与每个种子的总收入;假账号便宜时平台测得的 K 为 1.992 而真实 K 为 0.065,50% 的分佣变成 50% 的折扣。第四章把信息缺口写成购买风险率:揭示一半时购买概率 0.451 最高,延迟一天降到 0.078;通用解读被认为的准确度 0.838 接近有信息的解读,却在调包检验中只有随机水平 0.333,身份标签使转发率从 0.583 升到 0.745。 第五章在帕累托支付意愿下说明,尾指数 1.5 时完全价格歧视的上界是单一价格的 3.00 倍,但对一件商品任何菜单都胜不过单一价格(2.961 对 3.000);多层付费墙多收钱是因为它卖多件商品,收入占上界的比例从 0.150 升到 0.339。第六章说明捆绑独立的问题在 20 题时收走 0.678 的价值,诱饵档的效果取决于一个必须测量的参数,按题付费与年卡并行最好。第七章与《鬼谷子》摩篇、飞箝篇及《管子·国蓄》并读,同质性把增长因子从 0.757 提高到 1.297 但把增长局限在一个社群。第八章给出四步核查、合规清单与一项预注册的付费墙检验(每组 166,788 棵推荐树)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Branching Process of "Invite to Unlock" — Chapter 1 writes "invite friends to unlock free" and "pay and earn a commission" as a Galton-Watson branching process: a sharer stops at the threshold, so the share rate caps K, and at share rates of 30 % or 50 % no threshold takes K past 1; even at K = 1.15 a single seed dies out with probability 0.729, and the median cascade is 3 users. (https://5dlibrary.com/answers/en/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c01/) 第1章 拉人解锁的分支过程 — 第一章把「拉人免费解锁」与「付费加分佣」写成高尔顿—沃森分支过程:分享者凑够门槛就停,所以 K 被分享率封顶,分享率为 30% 或 50% 时任何门槛都不能让 K 越过 1;即使 K 为 1.15,单个种子仍有 0.729 的概率熄灭,级联规模的中位数只有 3 人。 (https://5dlibrary.com/answers/zh/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c01/) - Ch.2 A Finite Audience, a Clustered Network and an Acquisition Cost That Is Not Zero — Chapter 2 shows that in a finite audience K = 1.15 reaches only 0.255, that in a clustered network the measured K is 0.934 against a tree estimate of 1.35, and that acquisition costs do not fall to zero: 3.22 per paying user, 0.325 of the price. (https://5dlibrary.com/answers/en/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c02/) 第2章 有限的受众、聚集的网络与没有归零的获客成本 — 第二章说明,在有限受众中 K 为 1.15 只能触达 0.255,聚集网络中实测的 K 为 0.934 而树状估计为 1.35,获客成本没有归零:每个付费用户为 3.22,占价格的 0.325。 (https://5dlibrary.com/answers/zh/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c02/) - Ch.3 Commissions, Share-to-Unlock and the Incentives of Fake Accounts — Chapter 3 shows that commissions hardly change who pays but raise K and total revenue per seed; when fake accounts are cheap the platform measures K at 1.992 while the real K is 0.065, and a 50 % commission becomes a 50 % discount. (https://5dlibrary.com/answers/en/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c03/) 第3章 分佣、分享解锁与假账号的激励相容 — 第三章说明分佣几乎不改变谁付费,却提高 K 与每个种子的总收入;假账号便宜时平台测得的 K 为 1.992 而真实 K 为 0.065,50% 的分佣变成 50% 的折扣。 (https://5dlibrary.com/answers/zh/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c03/) - Ch.4 Information Gaps, Purchase Hazards and the Feeling of "So Accurate" — Chapter 4 writes the information gap as a purchase hazard: revealing half gives the highest purchase probability, 0.451, falling to 0.078 after a day's delay; a generic reading is judged 0.838 accurate, close to an informed one, yet scores only the chance level of 0.333 in a swap test, and an identity label raises sharing from 0.583 to 0.745. (https://5dlibrary.com/answers/en/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c04/) 第4章 信息缺口、购买风险率与「说得准」的感觉 — 第四章把信息缺口写成购买风险率:揭示一半时购买概率 0.451 最高,延迟一天降到 0.078;通用解读被认为的准确度 0.838 接近有信息的解读,却在调包检验中只有随机水平 0.333,身份标签使转发率从 0.583 升到 0.745。 (https://5dlibrary.com/answers/zh/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c04/) - Ch.5 Pareto Willingness to Pay, One Price and the Revenue Ceiling — Chapter 5 shows under Pareto willingness to pay that at a tail index of 1.5 the ceiling of perfect discrimination is 3.00 times one price, but that for one good no menu beats one price (2.961 against 3.000); a multi-tier paywall takes more because it sells several goods, raising the share of the ceiling from 0.150 to 0.339. (https://5dlibrary.com/answers/en/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c05/) 第5章 帕累托支付意愿、单一价格与收益上界 — 第五章在帕累托支付意愿下说明,尾指数 1.5 时完全价格歧视的上界是单一价格的 3.00 倍,但对一件商品任何菜单都胜不过单一价格(2.961 对 3.000);多层付费墙多收钱是因为它卖多件商品,收入占上界的比例从 0.150 升到 0.339。 (https://5dlibrary.com/answers/zh/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c05/) - Ch.6 Bundles, Decoy Tiers and Paying per Question — Chapter 6 shows that bundling independent questions collects 0.678 of their value at 20 questions, that a decoy tier depends on a parameter that must be measured, and that offering per-question pricing alongside an annual pass is best. (https://5dlibrary.com/answers/en/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c06/) 第6章 捆绑、诱饵档与按题付费 — 第六章说明捆绑独立的问题在 20 题时收走 0.678 的价值,诱饵档的效果取决于一个必须测量的参数,按题付费与年卡并行最好。 (https://5dlibrary.com/answers/zh/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c06/) - Ch.7 "Rub It with Its Own Kind", "Lift Then Clamp" and "Gather When Light" — Chapter 7 reads the "Rubbing" and "Flying and Clamping" chapters of the Guiguzi and the "State Reserves" chapter of the Guanzi; homophily raises the growth factor from 0.757 to 1.297 but confines growth to one community. (https://5dlibrary.com/answers/en/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c07/) 第7章 「摩之以其类」「飞而箝之」与「敛之以轻」 — 第七章与《鬼谷子》摩篇、飞箝篇及《管子·国蓄》并读,同质性把增长因子从 0.757 提高到 1.297 但把增长局限在一个社群。 (https://5dlibrary.com/answers/zh/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c07/) - Ch.8 A Four-Step Check, a Compliance Checklist and a Pre-Registered Paywall Test — Chapter 8 gives a four-step check, a compliance checklist and a pre-registered paywall test (166,788 referral trees per arm). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c08/) 第8章 四步核查、合规清单与一项预注册的付费墙检验 — 第八章给出四步核查、合规清单与一项预注册的付费墙检验(每组 166,788 棵推荐树)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/supercritical-viral-cascades-curiosity-gaps-paywall-pricing/c08/) ## 5DL-B-064 · Six-Dimensional Hypercube Walks, Hexagram Mutation Markov Chains and Modular Star-Palace Lattices 中文书名:六维超立方体游走、卦变马尔可夫链与模运算星宫晶格 DOI: https://doi.org/10.5281/zenodo.24090176 Cite: Fan, Y. (2026). Six-Dimensional Hypercube Walks, Hexagram Mutation Markov Chains and Modular Star-Palace Lattices (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090176 Book page: https://5dlibrary.com/books/hypercube-hexagram-transitions-modular-astrological-lattices/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: The six-line hexagram method and Zi Wei Dou Shu are systems of rules that compute on finite sets. This book writes the sixty-four hexagrams as the six-dimensional Boolean cube, casting and changing as probability and Markov chains, and the twelve palaces as the integers modulo 12 with the dihedral group, and computes the probabilities, symmetries and losses of information of the rules themselves. It neither claims nor tests that any reading comes true; all data are generated by its programs. Chapter 1 shows that complement and reversal generate a group of four that divides the hexagrams into 20 orbits, and that the received order pairs them into 32 pairs by reversal and complement; the nuclear hexagram is a sliding window on the middle four lines that folds 64 hexagrams onto 16 and then onto 4, of which Qian and Kun are fixed points and Ji Ji and Wei Ji a 2-cycle, not four attractors. Chapter 2 computes the two casting methods exactly: the number of moving lines is binomial under both, with no moving line at 0.1780; with coins old yin and old yang are 1/8 each, with yarrow old yang at 3/16 is three times old yin at 1/16, so the changed hexagram leans yin; the factor of three holds to four decimals under a natural split of the stalks and becomes 4.082 under an evenly spread split. Chapter 3 writes the change from cast to changed hexagram as a product Markov chain on the cube: both methods have spectral gap 1/2 and forget their start within three steps, against 8 for a one-line walk; the yarrow chain is not symmetric, Qian changing into Kun 729 times as often as Kun into Qian. Chapter 4 computes the chance of reading each of 450 texts under Zhu Xi's rules: the two methods' main-text distributions differ by 0.178; a good-or-bad score built on line positions is a low-degree Boolean function, and the correlation between cast and changed scores is exactly 0.4167. Chapter 5 writes the Life palace, the element bureau and the Zi Wei placement verse as piecewise affine maps modulo 12: Tian Fu is the reflection of Zi Wei across the Yin-Shen axis, and the fourteen main stars have only 12 layouts. Chapter 6 writes "three sides and four squares" as a convolution on the ring of twelve palaces: equal trines lose 8 of 12 harmonics, and two defensible sets of star values rank 144 cases with a rank correlation of only 0.606. Chapter 7 reads the number of the great expansion, Zhu Xi's divination procedure, Shao Yong's Before-Heaven diagrams and the textual history of the Zi Wei Dou Shu; the Before-Heaven diagrams are exactly binary counting, and the King Wen order has a rank correlation of −0.004 with binary value. Chapter 8 gives a four-step check and a pre-registered blind matching test (349 participants). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:六爻起卦与紫微斗数排盘都是在有限集合上运算的规则系统。本书把六十四卦写成六维布尔立方体,把起卦与变卦写成概率与马尔可夫链,把十二宫写成模 12 的整数与二面体群,计算这些规则本身的概率、对称与信息损失。本书不主张、也不检验任何占断是否应验;全部数据为程序生成。 第一章说明错卦与综卦生成一个四元群,把六十四卦分成 20 个轨道,通行卦序按综卦与错卦配成 32 对;互卦是中间四爻上的滑动窗口,一次把 64 卦折叠为 16 卦,二次折叠为 4 卦,其中乾坤为不动点、既济未济为二周期,而不是四个吸引子。第二章精确计算两种起卦法:动爻的个数在两者下同为二项分布,无动爻的概率为 0.1780;铜钱法下老阴老阳各 1/8,蓍草法下老阳 3/16 是老阴 1/16 的三倍,所以之卦偏阴;这个三倍在自然分蓍下精确到四位小数,在均匀分蓍下变为 4.082。第三章把本卦到之卦写成立方体上的乘积马尔可夫链:两种方法的谱隙都是 1/2,三步之内遗忘出发点,一次只变一爻的游走则需要 8 步;蓍草法的链不对称,乾变坤是坤变乾的 729 倍。第四章计算朱熹变占规则下 450 段文字被读到的概率:两种方法的主读文字分布相差 0.178;以爻位为基础的吉凶分数是低阶布尔函数,本卦与之卦分数的相关系数恰为 0.4167。 第五章把命宫、五行局与安紫微星诀写成模 12 的分段仿射映射:天府是紫微关于寅申轴的反射,十四主星只有 12 种排布。第六章把三方四正写成十二宫环上的卷积:等权三合丢失 12 个谐波中的 8 个;两套都说得通的星曜赋值对 144 种情形排序的秩相关只有 0.606。第七章与《系辞》大衍之数、朱熹《筮仪》、邵雍先天图与《紫微斗数》的版本问题并读,先天图恰为二进制计数,文王卦序与二进制的秩相关为 −0.004。第八章给出四步核查与一项预注册的盲法配对检验(349 名参与者)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Sixty-Four Hexagrams Are a Six-Dimensional Cube — Chapter 1 shows that complement and reversal generate a group of four that divides the hexagrams into 20 orbits, and that the received order pairs them into 32 pairs by reversal and complement; the nuclear hexagram is a sliding window on the middle four lines that folds 64 hexagrams onto 16 and then onto 4, of which Qian and Kun are fixed points and Ji Ji and Wei Ji a 2-cycle, not four attractors. (https://5dlibrary.com/answers/en/hypercube-hexagram-transitions-modular-astrological-lattices/c01/) 第1章 六十四卦是六维立方体 — 第一章说明错卦与综卦生成一个四元群,把六十四卦分成 20 个轨道,通行卦序按综卦与错卦配成 32 对;互卦是中间四爻上的滑动窗口,一次把 64 卦折叠为 16 卦,二次折叠为 4 卦,其中乾坤为不动点、既济未济为二周期,而不是四个吸引子。 (https://5dlibrary.com/answers/zh/hypercube-hexagram-transitions-modular-astrological-lattices/c01/) - Ch.2 Yarrow Stalks and Coins: the Exact Odds of the Two Methods — Chapter 2 computes the two casting methods exactly: the number of moving lines is binomial under both, with no moving line at 0.1780; with coins old yin and old yang are 1/8 each, with yarrow old yang at 3/16 is three times old yin at 1/16, so the changed hexagram leans yin; the factor of three holds to four decimals under a natural split of the stalks and becomes 4.082 under an evenly spread split. (https://5dlibrary.com/answers/en/hypercube-hexagram-transitions-modular-astrological-lattices/c02/) 第2章 蓍草与铜钱:两种起卦法的精确概率 — 第二章精确计算两种起卦法:动爻的个数在两者下同为二项分布,无动爻的概率为 0.1780;铜钱法下老阴老阳各 1/8,蓍草法下老阳 3/16 是老阴 1/16 的三倍,所以之卦偏阴;这个三倍在自然分蓍下精确到四位小数,在均匀分蓍下变为 4.082。 (https://5dlibrary.com/answers/zh/hypercube-hexagram-transitions-modular-astrological-lattices/c02/) - Ch.3 The Markov Chain of Changing Hexagrams and Its Spectral Gap — Chapter 3 writes the change from cast to changed hexagram as a product Markov chain on the cube: both methods have spectral gap 1/2 and forget their start within three steps, against 8 for a one-line walk; the yarrow chain is not symmetric, Qian changing into Kun 729 times as often as Kun into Qian. (https://5dlibrary.com/answers/en/hypercube-hexagram-transitions-modular-astrological-lattices/c03/) 第3章 卦变的马尔可夫链与谱隙 — 第三章把本卦到之卦写成立方体上的乘积马尔可夫链:两种方法的谱隙都是 1/2,三步之内遗忘出发点,一次只变一爻的游走则需要 8 步;蓍草法的链不对称,乾变坤是坤变乾的 729 倍。 (https://5dlibrary.com/answers/zh/hypercube-hexagram-transitions-modular-astrological-lattices/c03/) - Ch.4 Which Text Is Read: the Rules for Moving Lines and a Good-or-Bad Function — Chapter 4 computes the chance of reading each of 450 texts under Zhu Xi's rules: the two methods' main-text distributions differ by 0.178; a good-or-bad score built on line positions is a low-degree Boolean function, and the correlation between cast and changed scores is exactly 0.4167. (https://5dlibrary.com/answers/en/hypercube-hexagram-transitions-modular-astrological-lattices/c04/) 第4章 读哪一段文字:变占规则与吉凶判别函数 — 第四章计算朱熹变占规则下 450 段文字被读到的概率:两种方法的主读文字分布相差 0.178;以爻位为基础的吉凶分数是低阶布尔函数,本卦与之卦分数的相关系数恰为 0.4167。 (https://5dlibrary.com/answers/zh/hypercube-hexagram-transitions-modular-astrological-lattices/c04/) - Ch.5 The Placement Rules of Zi Wei Dou Shu Are Maps Modulo 12 — Chapter 5 writes the Life palace, the element bureau and the Zi Wei placement verse as piecewise affine maps modulo 12: Tian Fu is the reflection of Zi Wei across the Yin-Shen axis, and the fourteen main stars have only 12 layouts. (https://5dlibrary.com/answers/en/hypercube-hexagram-transitions-modular-astrological-lattices/c05/) 第5章 紫微斗数的安星法则是模 12 的映射 — 第五章把命宫、五行局与安紫微星诀写成模 12 的分段仿射映射:天府是紫微关于寅申轴的反射,十四主星只有 12 种排布。 (https://5dlibrary.com/answers/zh/hypercube-hexagram-transitions-modular-astrological-lattices/c05/) - Ch.6 "Three Sides and Four Squares" Is a Convolution on the Ring of Twelve Palaces — Chapter 6 writes "three sides and four squares" as a convolution on the ring of twelve palaces: equal trines lose 8 of 12 harmonics, and two defensible sets of star values rank 144 cases with a rank correlation of only 0.606. (https://5dlibrary.com/answers/en/hypercube-hexagram-transitions-modular-astrological-lattices/c06/) 第6章 三方四正是十二宫环上的卷积 — 第六章把三方四正写成十二宫环上的卷积:等权三合丢失 12 个谐波中的 8 个;两套都说得通的星曜赋值对 144 种情形排序的秩相关只有 0.606。 (https://5dlibrary.com/answers/zh/hypercube-hexagram-transitions-modular-astrological-lattices/c06/) - Ch.7 The Number of the Great Expansion, the Before-Heaven Diagrams and the Origins of the Zi Wei Dou Shu — Chapter 7 reads the number of the great expansion, Zhu Xi's divination procedure, Shao Yong's Before-Heaven diagrams and the textual history of the Zi Wei Dou Shu; the Before-Heaven diagrams are exactly binary counting, and the King Wen order has a rank correlation of −0.004 with binary value. (https://5dlibrary.com/answers/en/hypercube-hexagram-transitions-modular-astrological-lattices/c07/) 第7章 大衍之数、先天图与《紫微斗数》的来历 — 第七章与《系辞》大衍之数、朱熹《筮仪》、邵雍先天图与《紫微斗数》的版本问题并读,先天图恰为二进制计数,文王卦序与二进制的秩相关为 −0.004。 (https://5dlibrary.com/answers/zh/hypercube-hexagram-transitions-modular-astrological-lattices/c07/) - Ch.8 A Four-Step Check and a Pre-Registered Matching Test — Chapter 8 gives a four-step check and a pre-registered blind matching test (349 participants). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/hypercube-hexagram-transitions-modular-astrological-lattices/c08/) 第8章 四步核查与一项预注册的配对检验 — 第八章给出四步核查与一项预注册的盲法配对检验(349 名参与者)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/hypercube-hexagram-transitions-modular-astrological-lattices/c08/) ## 5DL-B-065 · Modular Numerology, Character-Stroke Hashes and Residue-Class Distributions 中文书名:姓名笔画、生命灵数与手机号码的同余映射和分布论 DOI: https://doi.org/10.5281/zenodo.24090178 Cite: Fan, Y. (2026). Modular Numerology, Character-Stroke Hashes and Residue-Class Distributions (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090178 Book page: https://5dlibrary.com/books/modular-numerology-stroke-hashes-residue-distributions/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: Life path numbers, the five grids of Chinese name analysis and phone-number numerology all compress a very large set onto a few residues and put a label on each. This book writes the three sets of rules as congruences, maps, convolutions and combinatorial counts, and computes their base rates among people, their sensitivity to convention and their loss of information. It neither claims nor tests that any verdict comes true; all data are generated by its programs. Chapter 1 proves that the digital root is the residue modulo 9, a ring homomorphism that can be computed in any order; the stopping rule for the master numbers 11, 22 and 33 reads digits rather than residues, so four published methods give different numbers, pair by pair, on 0.0956 to 0.1608 of the dates from 1900 to 2099, yet never disagree about the residue. Chapter 2 computes base rates: with the same number of births on every day, the plain digital root is even to one decimal place across several years of births; the bias comes from master numbers, which make 2, 4 and 6 the rarest, and under the common method no one born in the 1990s has the life path number 2. Chapter 3 writes the Pythagorean letter table as the alphabet modulo 9: random letters are almost even after two, vowels alone far from even; two common letter tables agree no more often than independent numbers, and three spellings of one Chinese name give three different numbers. Chapter 4 writes Kangxi stroke counting and the restoration of radicals as a fixed map; in a working table of 168 common characters about half have two different counts, differing by −1 to +18. Chapter 5 writes the five grids as convolutions of stroke counts and the eighty-one numbers as a digital root modulo 80: with Kangxi counts the total grid takes 61 values, but 18 of them hold eight names in ten; the good numbers crowd the low range where stroke sums fall, so each grid is good more often than the 0.432 of 35/81; another stroke convention flips each grid's label for 0.244 to 0.473 of names. Chapter 6 shows that the modulo-80 rule for phone numbers reads only the last four digits, because 10,000 is a multiple of 80, with a base rate of 0.425 for a good number; the "eight stars" are a three-bit XOR, and good or bad is the single bit of whether two digits share the East or West group; eight-digit endings with no bad pair are only 0.0336 to 0.0894 of all. Chapter 7 reads the eighty-one parts of the Han calendar treatise, the five hundred and forty radicals of the Shuowen jiezi and the stroke conventions of the Kangxi Dictionary. Chapter 8 gives a four-step check and a pre-registered matching test (419 participants). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:生命灵数、姓名五格与手机号码数理,都把一个很大的集合压缩成少数几个余数,再给每个余数贴上标签。本书把这三套规则写成同余、映射、卷积与组合计数,计算它们在人群中的基率、对约定的敏感度与信息损失。本书不主张、也不检验任何断语是否应验;全部数据为程序生成。 第一章证明数字根就是模 9 的余数,是一个可以按任意顺序计算的环同态;大师数 11、22、33 的停止规则读的是数字而不是余数,所以四种通行算法在 1900 年到 2099 年的日期上两两有 0.0956 到 0.1608 给出不同的数,却从不在余数上矛盾。第二章计算基率:按每日出生人数相同的假设,纯数字根在多年的出生人群中均匀到小数点后一位;偏倚来自大师数,它让 2、4、6 成为最少见的数,1990 年代出生的人按通行算法没有一个人的生命灵数是 2。第三章把毕达哥拉斯字母表写成字母序号模 9:随机字母两个之后就几乎均匀,只取元音则远不均匀;两种通行字母表给出相同数的比例与互相独立无异,同一个中文名字的三种拼法给出三个不同的数。 第四章把康熙笔画与部首还原写成确定的映射,在 168 个常用字的工作表中约一半的字两种笔画不同,差值从 −1 到 +18。第五章把五格写成笔画的卷积、八十一数理写成模 80 的数字根:用康熙笔画,总格能取到 61 个数,但 18 个数容纳了八成名字;吉数集中在笔画之和常落的低段,所以每一格为吉的比例都高于 35/81 的 0.432;换一种笔画约定,各格标签对 0.244 到 0.473 的名字翻转。第六章说明手机号码模 80 只读后四位,因为 10,000 是 80 的倍数,吉号基率为 0.425;「八星」是三位二进制的异或,吉凶只是两个数字是否同属东四或西四的一位信息,没有凶对的八位尾号只占 0.0336 到 0.0894。第七章与《汉书·律历志》的八十一分、《说文解字》的五百四十部与《康熙字典》的笔画体例并读。第八章给出四步核查与一项预注册的配对检验(419 名参与者)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Digital Root Is the Residue Modulo 9 — Chapter 1 proves that the digital root is the residue modulo 9, a ring homomorphism that can be computed in any order; the stopping rule for the master numbers 11, 22 and 33 reads digits rather than residues, so four published methods give different numbers, pair by pair, on 0.0956 to 0.1608 of the dates from 1900 to 2099, yet never disagree about the residue. (https://5dlibrary.com/answers/en/modular-numerology-stroke-hashes-residue-distributions/c01/) 第1章 数字根就是模 9 的余数 — 第一章证明数字根就是模 9 的余数,是一个可以按任意顺序计算的环同态;大师数 11、22、33 的停止规则读的是数字而不是余数,所以四种通行算法在 1900 年到 2099 年的日期上两两有 0.0956 到 0.1608 给出不同的数,却从不在余数上矛盾。 (https://5dlibrary.com/answers/zh/modular-numerology-stroke-hashes-residue-distributions/c01/) - Ch.2 The Base Rates of Life Path Numbers — Chapter 2 computes base rates: with the same number of births on every day, the plain digital root is even to one decimal place across several years of births; the bias comes from master numbers, which make 2, 4 and 6 the rarest, and under the common method no one born in the 1990s has the life path number 2. (https://5dlibrary.com/answers/en/modular-numerology-stroke-hashes-residue-distributions/c02/) 第2章 生命灵数在人群中的基率 — 第二章计算基率:按每日出生人数相同的假设,纯数字根在多年的出生人群中均匀到小数点后一位;偏倚来自大师数,它让 2、4、6 成为最少见的数,1990 年代出生的人按通行算法没有一个人的生命灵数是 2。 (https://5dlibrary.com/answers/zh/modular-numerology-stroke-hashes-residue-distributions/c02/) - Ch.3 The Alphabet Modulo 9: Name Numbers and Two Letter Tables — Chapter 3 writes the Pythagorean letter table as the alphabet modulo 9: random letters are almost even after two, vowels alone far from even; two common letter tables agree no more often than independent numbers, and three spellings of one Chinese name give three different numbers. (https://5dlibrary.com/answers/en/modular-numerology-stroke-hashes-residue-distributions/c03/) 第3章 字母表模 9:名字数与两种字母表 — 第三章把毕达哥拉斯字母表写成字母序号模 9:随机字母两个之后就几乎均匀,只取元音则远不均匀;两种通行字母表给出相同数的比例与互相独立无异,同一个中文名字的三种拼法给出三个不同的数。 (https://5dlibrary.com/answers/zh/modular-numerology-stroke-hashes-residue-distributions/c03/) - Ch.4 Kangxi Strokes: Radical Restoration Is a Fixed Map — Chapter 4 writes Kangxi stroke counting and the restoration of radicals as a fixed map; in a working table of 168 common characters about half have two different counts, differing by −1 to +18. (https://5dlibrary.com/answers/en/modular-numerology-stroke-hashes-residue-distributions/c04/) 第4章 康熙笔画:部首还原是一个确定的映射 — 第四章把康熙笔画与部首还原写成确定的映射,在 168 个常用字的工作表中约一半的字两种笔画不同,差值从 −1 到 +18。 (https://5dlibrary.com/answers/zh/modular-numerology-stroke-hashes-residue-distributions/c04/) - Ch.5 The Five Grids and the Eighty-One Numbers: Convolution and Modulo 80 — Chapter 5 writes the five grids as convolutions of stroke counts and the eighty-one numbers as a digital root modulo 80: with Kangxi counts the total grid takes 61 values, but 18 of them hold eight names in ten; the good numbers crowd the low range where stroke sums fall, so each grid is good more often than the 0.432 of 35/81; another stroke convention flips each grid's label for 0.244 to 0.473 of names. (https://5dlibrary.com/answers/en/modular-numerology-stroke-hashes-residue-distributions/c05/) 第5章 五格与八十一数理:卷积与模 80 — 第五章把五格写成笔画的卷积、八十一数理写成模 80 的数字根:用康熙笔画,总格能取到 61 个数,但 18 个数容纳了八成名字;吉数集中在笔画之和常落的低段,所以每一格为吉的比例都高于 35/81 的 0.432;换一种笔画约定,各格标签对 0.244 到 0.473 的名字翻转。 (https://5dlibrary.com/answers/zh/modular-numerology-stroke-hashes-residue-distributions/c05/) - Ch.6 Phone Numbers: Modulo 80, XOR and Compressed Numbers — Chapter 6 shows that the modulo-80 rule for phone numbers reads only the last four digits, because 10,000 is a multiple of 80, with a base rate of 0.425 for a good number; the "eight stars" are a three-bit XOR, and good or bad is the single bit of whether two digits share the East or West group; eight-digit endings with no bad pair are only 0.0336 to 0.0894 of all. (https://5dlibrary.com/answers/en/modular-numerology-stroke-hashes-residue-distributions/c06/) 第6章 手机号码:模 80、异或与被压缩的号码 — 第六章说明手机号码模 80 只读后四位,因为 10,000 是 80 的倍数,吉号基率为 0.425;「八星」是三位二进制的异或,吉凶只是两个数字是否同属东四或西四的一位信息,没有凶对的八位尾号只占 0.0336 到 0.0894。 (https://5dlibrary.com/answers/zh/modular-numerology-stroke-hashes-residue-distributions/c06/) - Ch.7 Eighty-One Parts, Five Hundred and Forty Radicals and the Kangxi Dictionary — The Han calendar treatise takes 81 parts to the day from the 9 times 9 of the yellow-bell pipe; the month and year it gives are about 24 seconds and 11.5 minutes too long. The 540 sections of the Shuowen jiezi are a way of grouping by form, from 一 to 亥, with no numerical meaning. The radical-plus-remainder of the Kangxi Dictionary is a convention for looking characters up; name analysis uses it as a rule of counting, which is a later use. The eighty-one numbers share 81 with the Han day method, but this book has found no evidence of a line of descent between them. (https://5dlibrary.com/answers/en/modular-numerology-stroke-hashes-residue-distributions/c07/) 第7章 八十一分、五百四十部与康熙字典 — 《汉书·律历志》以 81 为日法,取自黄钟律管的 9 乘 9;它给出的月长与年长分别多出约 24 秒与 11.5 分钟。《说文解字》的 540 部是据形归类的方法,始一终亥,不带数的意义。《康熙字典》的部首加余画是检索约定;姓名学把它用作计数规则,这是后来的用法。八十一数理与汉代日法共用 81,但本书没有找到两者之间有传承的证据。 (https://5dlibrary.com/answers/zh/modular-numerology-stroke-hashes-residue-distributions/c07/) - Ch.8 A Four-Step Check and a Pre-Registered Matching Test — Chapter 8 gives a four-step check and a pre-registered matching test (419 participants). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/modular-numerology-stroke-hashes-residue-distributions/c08/) 第8章 四步核查与一项预注册的配对检验 — 第八章给出四步核查与一项预注册的配对检验(419 名参与者)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/modular-numerology-stroke-hashes-residue-distributions/c08/) ## 5DL-B-066 · Four-Pillar Calendrical Computation and Formal Rule Engines 中文书名:四柱排盘的历法计算与规则引擎 DOI: https://doi.org/10.5281/zenodo.24090180 Cite: Fan, Y. (2026). Four-Pillar Calendrical Computation and Formal Rule Engines (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090180 Book page: https://5dlibrary.com/books/four-pillar-calendrical-computation-rule-engine/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: A four-pillar chart first does calendar work and then judges. The first half is astronomy and calendar calculation, with a single answer once the conventions are stated; the second half is a rule engine whose results depend on parameters. This book writes both halves as programs and asks three questions. Do different programs give the same birth different charts? Does a birth time a few minutes off change the chart? Do different parameters give the same chart opposite judgements? It neither claims nor tests that any reading comes true; all data are generated by its programs. Chapter 1 writes the day pillar as the residue of the Julian day plus 49 modulo 60, continuous across the reform of 1582, and the five-tigers and five-rats rules as affine maps modulo 10 that agree with the verses line by line; only four of the eight characters are free, giving 518,400 possible charts. Chapter 2 compares three conventions for the zi hour, which differ only between 23:00 and 24:00 and affect 0.0417 of births; the gap between clock time and true solar time changes the hour branch for 0.069 of births in Shanghai and 0.610 in Kuala Lumpur. Chapter 3 follows errors through: a quarter of an hour of error makes the hour pillar wrong with probability 0.06250 to 0.09974, the day, month and year pillars one to three orders of magnitude less often; it gives the conditional probabilities a "boundary check" needs; and the starting age of the luck cycles is a straight line of one year per three days between solar terms that jumps by as much as 10.48 years at a term. Chapter 4 proves that the ten gods are the ten elements of Z5 × Z2, their table a Latin square and not a function of distance along the order of the stems; three common weightings of the hidden stems agree on the main element, and 0.6, 0.3 and 0.1 share their origin with the 18, 9 and 3 days in command. Chapter 5 writes the judgement of strength as scoring on a weighted graph: over all charts, three defensible parameter sets reach a pairwise Cohen's kappa of only 0.553 to 0.622, all three agree on 0.684 of charts, and the share of "follow" charts lies between 0.022 and 0.121; one double-hour of difference changes the verdict for 0.127 of charts. Chapter 6 compares two chart engines, clock time with the day changing at 23:00 against true solar time with the day changing at midnight: at least one pillar differs for 0.1043 of births in Shanghai and 0.6463 in Kuala Lumpur; it also gives test vectors and a fingerprint. Chapter 7 reads the Yuanhai ziping, the Sanming tonghui and the Ditian sui. Chapter 8 gives a four-step check, the sample size of an agreement study (255 charts fix kappa to within 0.100) and a pre-registered matching test (359 analysed participants). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:四柱排盘先做历法,再做判断。前一半是天文与历法的计算,约定说清楚后有唯一的答案;后一半是一套规则引擎,结果取决于参数。本书把两半都写成程序,问三个问题:同一个出生,不同的软件会不会排出不同的盘?出生时间差几分钟,盘会不会变?同一张盘,不同的参数会不会给出相反的判断?本书不主张、也不检验任何断语是否应验;全部数据为程序生成。 第一章用儒略日把日柱写成(儒略日 + 49)模 60 的余数,跨越 1582 年改历仍然连续,并把五虎遁、五鼠遁写成模 10 的仿射映射,与歌诀逐条相符;八个字中只有四个自由,可能的四柱共 518,400 种。第二章比较子时的三种约定,它们只在 23 点到 24 点之间不同,影响 0.0417 的出生;钟表时间与真太阳时之差使时支在上海对 0.069 的出生不同,在吉隆坡对 0.610 的出生不同。第三章计算误差的传播:一刻钟的误差让时柱出错的概率为 0.06250 到 0.09974,日柱、月柱、年柱小一到三个数量级;给出「时辰边界校准」的条件概率表;起运岁数在节气之间是三天一岁的直线,在节气处跳变可达 10.48 年。 第四章证明十神是 Z5 × Z2 的十个元素,十神表是拉丁方,而不是天干次序距离的函数;藏干的三种通行权重在主五行上一致,0.6、0.3、0.1 与人元司令的 18、9、3 天同源。第五章把旺衰判断写成加权图上的打分,三套说得通的参数对全部四柱两两的 Cohen's kappa 只有 0.553 到 0.622,三者一致的盘占 0.684,从格的比例在 0.022 与 0.121 之间;时支差一格,判定对 0.127 的盘改变。第六章比较两个排盘引擎:钟表时间与 23 点换日对真太阳时与零点换日,在上海对 0.1043 的出生、在吉隆坡对 0.6463 的出生至少一柱不同,并给出一套测试向量与指纹。第七章与《渊海子平》《三命通会》《滴天髓》并读。第八章给出四步核查、一致性研究的样本量(255 张盘把 kappa 定在 0.100 之内)与一项预注册的配对检验(359 名纳入分析的参与者)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Julian Day and the Congruences of the Four Pillars — Chapter 1 writes the day pillar as the residue of the Julian day plus 49 modulo 60, continuous across the reform of 1582, and the five-tigers and five-rats rules as affine maps modulo 10 that agree with the verses line by line; only four of the eight characters are free, giving 518,400 possible charts. (https://5dlibrary.com/answers/en/four-pillar-calendrical-computation-rule-engine/c01/) 第1章 儒略日与四柱的同余公式 — 第一章用儒略日把日柱写成(儒略日 + 49)模 60 的余数,跨越 1582 年改历仍然连续,并把五虎遁、五鼠遁写成模 10 的仿射映射,与歌诀逐条相符;八个字中只有四个自由,可能的四柱共 518,400 种。 (https://5dlibrary.com/answers/zh/four-pillar-calendrical-computation-rule-engine/c01/) - Ch.2 Three Conventions for the Zi Hour, and True Solar Time — Chapter 2 compares three conventions for the zi hour, which differ only between 23:00 and 24:00 and affect 0.0417 of births; the gap between clock time and true solar time changes the hour branch for 0.069 of births in Shanghai and 0.610 in Kuala Lumpur. (https://5dlibrary.com/answers/en/four-pillar-calendrical-computation-rule-engine/c02/) 第2章 子时的三种约定与真太阳时 — 第二章比较子时的三种约定,它们只在 23 点到 24 点之间不同,影响 0.0417 的出生;钟表时间与真太阳时之差使时支在上海对 0.069 的出生不同,在吉隆坡对 0.610 的出生不同。 (https://5dlibrary.com/answers/zh/four-pillar-calendrical-computation-rule-engine/c02/) - Ch.3 Timing Errors and the Edges of the Double-Hours — Chapter 3 follows errors through: a quarter of an hour of error makes the hour pillar wrong with probability 0.06250 to 0.09974, the day, month and year pillars one to three orders of magnitude less often; it gives the conditional probabilities a "boundary check" needs; and the starting age of the luck cycles is a straight line of one year per three days between solar terms that jumps by as much as 10.48 years at a term. (https://5dlibrary.com/answers/en/four-pillar-calendrical-computation-rule-engine/c03/) 第3章 出生时间误差与时辰边界 — 第三章计算误差的传播:一刻钟的误差让时柱出错的概率为 0.06250 到 0.09974,日柱、月柱、年柱小一到三个数量级;给出「时辰边界校准」的条件概率表;起运岁数在节气之间是三天一岁的直线,在节气处跳变可达 10.48 年。 (https://5dlibrary.com/answers/zh/four-pillar-calendrical-computation-rule-engine/c03/) - Ch.4 The Ten Gods Are Z5 × Z2, the Hidden Stems Weight Vectors — Chapter 4 proves that the ten gods are the ten elements of Z5 × Z2, their table a Latin square and not a function of distance along the order of the stems; three common weightings of the hidden stems agree on the main element, and 0.6, 0.3 and 0.1 share their origin with the 18, 9 and 3 days in command. (https://5dlibrary.com/answers/en/four-pillar-calendrical-computation-rule-engine/c04/) 第4章 十神是 Z5 × Z2,藏干是权重向量 — 第四章证明十神是 Z5 × Z2 的十个元素,十神表是拉丁方,而不是天干次序距离的函数;藏干的三种通行权重在主五行上一致,0.6、0.3、0.1 与人元司令的 18、9、3 天同源。 (https://5dlibrary.com/answers/zh/four-pillar-calendrical-computation-rule-engine/c04/) - Ch.5 Strength Is Scoring on a Weighted Graph — Chapter 5 writes the judgement of strength as scoring on a weighted graph: over all charts, three defensible parameter sets reach a pairwise Cohen's kappa of only 0.553 to 0.622, all three agree on 0.684 of charts, and the share of "follow" charts lies between 0.022 and 0.121; one double-hour of difference changes the verdict for 0.127 of charts. (https://5dlibrary.com/answers/en/four-pillar-calendrical-computation-rule-engine/c05/) 第5章 旺衰是加权图上的打分 — 第五章把旺衰判断写成加权图上的打分,三套说得通的参数对全部四柱两两的 Cohen's kappa 只有 0.553 到 0.622,三者一致的盘占 0.684,从格的比例在 0.022 与 0.121 之间;时支差一格,判定对 0.127 的盘改变。 (https://5dlibrary.com/answers/zh/four-pillar-calendrical-computation-rule-engine/c05/) - Ch.6 Two Chart Engines and a Set of Test Vectors — Chapter 6 compares two chart engines, clock time with the day changing at 23:00 against true solar time with the day changing at midnight: at least one pillar differs for 0.1043 of births in Shanghai and 0.6463 in Kuala Lumpur; it also gives test vectors and a fingerprint. (https://5dlibrary.com/answers/en/four-pillar-calendrical-computation-rule-engine/c06/) 第6章 两个排盘引擎与一套测试向量 — 第六章比较两个排盘引擎:钟表时间与 23 点换日对真太阳时与零点换日,在上海对 0.1043 的出生、在吉隆坡对 0.6463 的出生至少一柱不同,并给出一套测试向量与指纹。 (https://5dlibrary.com/answers/zh/four-pillar-calendrical-computation-rule-engine/c06/) - Ch.7 The Yuanhai ziping, the Sanming tonghui and the Ditian sui — The Yuanhai ziping takes the day as master, and the ten gods are coordinates on $\mathbb{Z}_5 \times \mathbb{Z}_2$ with the day stem at the origin. The days in command of the Sanming tonghui and other books fill thirty days each month but disagree with the hidden-stem list in five branches; the 18, 9 and 3 days of the storage branches are the 0.6, 0.3 and 0.1 common in software. The working rules for the start of luck assume a 360-day year. The Ditian sui treats strength as the key of the art, and Chapter 5 shows that it is also the step most dependent on convention. (https://5dlibrary.com/answers/en/four-pillar-calendrical-computation-rule-engine/c07/) 第7章 《渊海子平》《三命通会》与《滴天髓》 — 《渊海子平》以日为主,十神是以日干为原点的 $\mathbb{Z}_5 \times \mathbb{Z}_2$ 坐标。《三命通会》等书的人元司令天数每月分足三十天,但与藏干表在五个地支上不一致;墓库之支的 18、9、3 天就是软件常用的 0.6、0.3、0.1。起运口诀以 360 日为一年。《滴天髓》把衰旺视为命理的关键,而第五章显示它也正是最依赖约定的一步。 (https://5dlibrary.com/answers/zh/four-pillar-calendrical-computation-rule-engine/c07/) - Ch.8 A Four-Step Check, an Agreement Study and a Pre-Registered Matching Test — Chapter 8 gives a four-step check, the sample size of an agreement study (255 charts fix kappa to within 0.100) and a pre-registered matching test (359 analysed participants). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/four-pillar-calendrical-computation-rule-engine/c08/) 第8章 四步核查、一致性研究与一项预注册的配对检验 — 第八章给出四步核查、一致性研究的样本量(255 张盘把 kappa 定在 0.100 之内)与一项预注册的配对检验(359 名纳入分析的参与者)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/four-pillar-calendrical-computation-rule-engine/c08/) ## 5DL-B-067 · Najia Hexagram Encoding and Qimen Dunjia Combinatorial Engines 中文书名:纳甲装卦与奇门遁甲排盘的组合计算 DOI: https://doi.org/10.5281/zenodo.24090182 Cite: Fan, Y. (2026). Najia Hexagram Encoding and Qimen Dunjia Combinatorial Engines (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090182 Book page: https://5dlibrary.com/books/najia-hexagram-encoding-qimen-combinatorial-engines/ Scope: Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. Abstract: Najia six-line divination and Qimen Dunjia are deterministic algorithms on finite sets: the first dresses the sixty-four hexagrams with stems, branches, six relatives, six spirits and void branches; the second turns a double-hour into an arrangement on three turning plates. This book writes both in full as programs, studies the groups, cosets and permutations inside their rules, and computes how often different calendar and school conventions give the same moment different charts. It neither claims nor tests that any reading comes true; all data are generated by its programs. Chapter 1 writes Jing Fang's eight palaces as pure hexagrams XOR eight fixed masks: upper XOR lower tells a hexagram's generation and world line, and the first six generations form a Gray code; the najia table is generated by one starting branch, one direction and one stem per trigram. Chapter 2 computes how the six relatives fall on the 384 lines: 44 of the 64 hexagrams lack at least one relative, so hidden spirits are routine; the six spirits start by the day stem's element, with earth split between wu and ji, not by the stem modulo 6; the void branches are the two residues outside a decade's ten consecutive branches, and the six decades' void pairs share out all twelve branches. Chapter 3 shows that the 72 bureau numbers of the solar terms come from eight starting palaces and two step rules, and that 1,080 is 18 bureaus times 60 double-hours; compared double-hour by double-hour from 2000 to 2029, the split-supplement and intercalation calendars give different bureaus for 0.3067 of double-hours and different duns for 0.0219. Chapter 4 shows that the earth plate is a dihedral group and the heaven plate a rotation of the eight outer palaces; fu yin takes 0.2222 of all bureau-hours, more than twice the 0.1000 of jia hours, and fan yin 0.1111. Chapter 5 adds the human and spirit plates: the 1,080 bureau-hours give only 727 distinct combinations of the three plates; composed freely, the plate moves generate the whole symmetric group S9 of 362,880 permutations, yet charts use only 72 of them, because every yin earth plate is a yang plate turned half way round. Chapter 6 computes the patterns: a punishment is on the heaven plate in 0.5296 of bureau-hours, and three common tomb rules disagree about the same chart, pair by pair, in 0.1944 to 0.4046 of cases; the verse of the five non-meeting hours omits the second solution on ji and geng days; between neighbouring double-hours the fu yin, punishment and tomb flags change in two fifths to more than half of all pairs. Chapter 8 gives a budget of conventions (two engines with different choices disagree on 0.4800 of double-hours in Shanghai) and a pre-registered matching test (447 participants). Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. 摘要:纳甲六爻与奇门遁甲是两套在有限集合上运算的确定性算法:前者给六十四卦装上天干、地支、六亲、六神与旬空,后者把一个时辰变成三层转盘上的一个排列。本书把两者完整地写成程序,研究规则内部的群、陪集与置换结构,并计算不同历法与流派约定让同一时刻得到不同排盘的比例。本书不主张、也不检验任何占断是否应验;全部数据为程序生成。 第一章把京房八宫写成纯卦与八个固定掩码的异或:上卦异或下卦决定一卦的世代与世爻,前六代构成格雷码;纳甲表由每卦一个起支、一个方向与一个天干生成。第二章计算六亲在 384 爻中的分布,64 卦中有 44 卦缺少至少一种六亲,伏神是常态;六神按日干五行起、土分戊己,不是日干模 6;旬空是每旬十个连续地支之外的两个余数,六旬的空亡恰好分完十二地支。第三章证明奇门 72 个节气局数由八个起宫与两条步长规则生成,1,080 是 18 局乘 60 时辰;逐时比较 2000 年到 2029 年的拆补法与置闰法,两者在 0.3067 的时辰给出不同的局,0.0219 连阴阳遁都不同。 第四章证明地盘是二面体群,天盘是外圈八宫的旋转,伏吟占全部局时的 0.2222,是甲时所占 0.1000 的两倍多,反吟占 0.1111。第五章加上人盘与神盘:1,080 个局时只给出 727 种不同的三盘组合;转盘运算复合起来生成整个对称群 S9,共 362,880 个排列,但排盘只用到其中 72 种,因为每个阴遁地盘都是某个阳遁地盘转半圈。第六章计算格局:天盘击刑出现在 0.5296 的局时中,三种通行的入墓规则对同一张盘两两意见不一的比例在 0.1944 与 0.4046 之间;五不遇时的歌诀漏列了己日与庚日的第二个解;相邻时辰之间,伏吟、击刑、入墓的标记有四成到五成半会改变。第八章给出一份约定预算(两个取法不同的引擎在上海对 0.4800 的时辰意见不一)与一项预注册的配对检验(447 名参与者)。书中以标注框区分证据、16 段可运行的 Python 程序、原典与假说。 Chapter answers: - Ch.1 Jing Fang's Eight Palaces and the Najia: XOR Masks and a Closed Form — The program checks that each of the 64 hexagrams belongs to exactly one palace; in King Wen numbers the Qian palace is 1, 44, 33, 12, 20, 23, 35, 14 (Qian, Gou, Dun, Pi, Guan, Bo, Jin, Da You) and the Kun palace 2, 24, 19, 11, 34, 43, 5, 8, as in the usual table of the eight palaces. (https://5dlibrary.com/answers/en/najia-hexagram-encoding-qimen-combinatorial-engines/c01/) 第1章 京房八宫与纳甲:异或掩码与闭式 — 程序验证了 64 卦恰好每卦属于一宫,乾宫依文王卦序为 1、44、33、12、20、23、35、14(乾、姤、遁、否、观、剥、晋、大有),坤宫为 2、24、19、11、34、43、5、8,与通行的八宫表一致。 (https://5dlibrary.com/answers/zh/najia-hexagram-encoding-qimen-combinatorial-engines/c01/) - Ch.2 Six Relatives, Six Spirits and the Void Branches — Chapter 2 computes how the six relatives fall on the 384 lines: 44 of the 64 hexagrams lack at least one relative, so hidden spirits are routine; the six spirits start by the day stem's element, with earth split between wu and ji, not by the stem modulo 6; the void branches are the two residues outside a decade's ten consecutive branches, and the six decades' void pairs share out all twelve branches. (https://5dlibrary.com/answers/en/najia-hexagram-encoding-qimen-combinatorial-engines/c02/) 第2章 六亲、六神与旬空 — 第二章计算六亲在 384 爻中的分布,64 卦中有 44 卦缺少至少一种六亲,伏神是常态;六神按日干五行起、土分戊己,不是日干模 6;旬空是每旬十个连续地支之外的两个余数,六旬的空亡恰好分完十二地支。 (https://5dlibrary.com/answers/zh/najia-hexagram-encoding-qimen-combinatorial-engines/c02/) - Ch.3 The Seventy-Two Bureau Numbers and Two Calendars — Chapter 3 shows that the 72 bureau numbers of the solar terms come from eight starting palaces and two step rules, and that 1,080 is 18 bureaus times 60 double-hours; compared double-hour by double-hour from 2000 to 2029, the split-supplement and intercalation calendars give different bureaus for 0.3067 of double-hours and different duns for 0.0219. (https://5dlibrary.com/answers/en/najia-hexagram-encoding-qimen-combinatorial-engines/c03/) 第3章 七十二个局数与两种定局历法 — 第三章证明奇门 72 个节气局数由八个起宫与两条步长规则生成,1,080 是 18 局乘 60 时辰;逐时比较 2000 年到 2029 年的拆补法与置闰法,两者在 0.3067 的时辰给出不同的局,0.0219 连阴阳遁都不同。 (https://5dlibrary.com/answers/zh/najia-hexagram-encoding-qimen-combinatorial-engines/c03/) - Ch.4 Earth Plate and Heaven Plate: a Dihedral Group and a Ring Rotation — Chapter 4 shows that the earth plate is a dihedral group and the heaven plate a rotation of the eight outer palaces; fu yin takes 0.2222 of all bureau-hours, more than twice the 0.1000 of jia hours, and fan yin 0.1111. (https://5dlibrary.com/answers/en/najia-hexagram-encoding-qimen-combinatorial-engines/c04/) 第4章 地盘与天盘:二面体群与外圈旋转 — 第四章证明地盘是二面体群,天盘是外圈八宫的旋转,伏吟占全部局时的 0.2222,是甲时所占 0.1000 的两倍多,反吟占 0.1111。 (https://5dlibrary.com/answers/zh/najia-hexagram-encoding-qimen-combinatorial-engines/c04/) - Ch.5 The Human and Spirit Plates, and the Group the Plates Generate — Chapter 5 adds the human and spirit plates: the 1,080 bureau-hours give only 727 distinct combinations of the three plates; composed freely, the plate moves generate the whole symmetric group S9 of 362,880 permutations, yet charts use only 72 of them, because every yin earth plate is a yang plate turned half way round. (https://5dlibrary.com/answers/en/najia-hexagram-encoding-qimen-combinatorial-engines/c05/) 第5章 人盘、神盘与转盘生成的置换群 — 第五章加上人盘与神盘:1,080 个局时只给出 727 种不同的三盘组合;转盘运算复合起来生成整个对称群 S9,共 362,880 个排列,但排盘只用到其中 72 种,因为每个阴遁地盘都是某个阳遁地盘转半圈。 (https://5dlibrary.com/answers/zh/najia-hexagram-encoding-qimen-combinatorial-engines/c05/) - Ch.6 How Often the Patterns Occur, How Conventions Disagree, and Neighbouring Hours — Chapter 6 computes the patterns: a punishment is on the heaven plate in 0.5296 of bureau-hours, and three common tomb rules disagree about the same chart, pair by pair, in 0.1944 to 0.4046 of cases; the verse of the five non-meeting hours omits the second solution on ji and geng days; between neighbouring double-hours the fu yin, punishment and tomb flags change in two fifths to more than half of all pairs. (https://5dlibrary.com/answers/en/najia-hexagram-encoding-qimen-combinatorial-engines/c06/) 第6章 格局的频率、约定的分歧与相邻时辰 — 第六章计算格局:天盘击刑出现在 0.5296 的局时中,三种通行的入墓规则对同一张盘两两意见不一的比例在 0.1944 与 0.4046 之间;五不遇时的歌诀漏列了己日与庚日的第二个解;相邻时辰之间,伏吟、击刑、入墓的标记有四成到五成半会改变。 (https://5dlibrary.com/answers/zh/najia-hexagram-encoding-qimen-combinatorial-engines/c06/) - Ch.7 The Jing shi yizhuan and the Rhapsody of the Old Fisherman — The palace order of the Jing shi yizhuan puts the world line on the two middle lines for half of all hexagrams and lets the generation be read directly from upper XOR lower. The 1,080, 72 and 18 of the Rhapsody are three units of counting in one system — bureau-hours, term-yuan and bureaus — not one table cut down twice. (https://5dlibrary.com/answers/en/najia-hexagram-encoding-qimen-combinatorial-engines/c07/) 第7章 《京氏易传》与《烟波钓叟赋》 — 《京氏易传》的八宫序使一半的卦世爻落在中间两爻,并使世代可由上下卦之异或直接读出。《烟波钓叟赋》的 1,080、72、18 是同一系统的三种计数单位:局时、节气元与局,不是同一张表的两次删减。 (https://5dlibrary.com/answers/zh/najia-hexagram-encoding-qimen-combinatorial-engines/c07/) - Ch.8 A Budget of Conventions and a Pre-Registered Matching Test — Chapter 8 gives a budget of conventions (two engines with different choices disagree on 0.4800 of double-hours in Shanghai) and a pre-registered matching test (447 participants). Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. (https://5dlibrary.com/answers/en/najia-hexagram-encoding-qimen-combinatorial-engines/c08/) 第8章 约定预算与一项预注册的配对检验 — 第八章给出一份约定预算(两个取法不同的引擎在上海对 0.4800 的时辰意见不一)与一项预注册的配对检验(447 名参与者)。书中以标注框区分证据、16 段可运行的 Python 程序、原典与假说。 (https://5dlibrary.com/answers/zh/najia-hexagram-encoding-qimen-combinatorial-engines/c08/) ## 5DL-B-068 · Unbundling Economics, Anti-Arbitrage Identity Locking and Multi-Tier Licensing Ladders 中文书名:分拆经济学、防套利身份锁定与多阶授权阶梯 DOI: https://doi.org/10.5281/zenodo.24090184 Cite: Fan, Y. (2026). Unbundling Economics, Anti-Arbitrage Identity Locking and Multi-Tier Licensing Ladders (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090184 Book page: https://5dlibrary.com/books/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/ Scope: Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. Abstract: A digital service that sells four modules - cycle readings, acoustic interventions, metabolic meal plans and business scripts - has to answer three pricing questions: should it offer a cheap all-access pass, may a personal annual card carry transferable extra readings, and how should the price steps run from a lead-in reading to a five-figure business licence? This book writes the three as runnable programs: multiproduct bundling, resale arbitrage and nonlinear pricing (screening). Its demand curves and values are model parameters, not data from any business; it neither claims nor tests that any reading comes true. Chapters 1 and 2 show that separate selling does not generally beat the optimal bundle: under independent, single-scale orthogonal and heavy-tailed demand with tail index 1.5, separate selling earns only 0.825, 0.873 and 0.730 of optimal bundle revenue; only when demand is orthogonal and studios' values are three orders of magnitude above individual users' does separate selling earn 1.479 times the optimal bundle. The loss from a cheap all-access pass comes from high-value buyers coming back down: at a tenth of the sum of single prices, revenue falls to 0.439 of separate selling. Chapters 3 and 4 model resale arbitrage: with a 50% affiliate commission, resale beats referral when the extra slots satisfy k + 1 > 4.020; with self-referral, arbitrage costs half as much and six extra slots make the firm lose money on every reading it serves; imitation among channel partners spreads resale to 0.877 of them within two years. Under an identity lock, an extra-chart price of at least half the single price keeps resale from beating referral, and full price also protects direct sales; birth data are not an identity, as among 100,000 card holders 14,259.5 pairs of strangers share a date-and-double-hour key. Chapter 5 solves the six-tier menu exactly as a screening problem: the optimal prices are 12, 117, 287, 567, 12537 and 34537, and the business licence and agency seat bring about four fifths of profit; with the licence held at 3,000, profit falls to 0.434. Chapter 6 shows the menu's fragility when demand is misspecified (the worst 5% of worlds earn 0.315 of their optimum), that add-on discounts are right only for negatively correlated values, and a 0.112 chance that a quarter's profit falls below 80% of plan. Chapter 7 reads the Han tax on hoarded cash and its informer rule in the Shiji, the Guanzi and the Song salt vouchers. Chapter 8 gives a four-flag menu audit and a pre-registered experiment (188 channel partners per arm). Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. 摘要:一家同时出售周期测算、声学干预、代谢配餐与商业脚本四个板块的数字服务,要回答三个定价问题:要不要出廉价全通卡,个人年卡能不能附赠转让名额,从引流测算到五位数商业牌照的价格台阶怎样排。本书把三者写成可运行的程序:多产品捆绑、转售套利与非线性定价(甄别)。书中的需求与估值都是模型参数,不是任何企业的数据;本书不主张、也不检验任何占断是否应验。 第一、二章证明,分拆并不普遍胜过最优捆绑:在独立、单一尺度正交与尾指数 1.5 的重尾需求下,分拆只得到最优捆绑收入的 0.825、0.873 与 0.730;只有当需求正交、且工作室与个人用户的估值相差三个数量级时,分拆才是最优捆绑的 1.479 倍。廉价全通卡的损失来自高估值客户的回流:卡价为单品价总和的十分之一时,收入只剩分拆的 0.439。第三、四章建立转售套利模型:在 50% 的分销佣金下,附赠名额满足 k + 1 > 4.020 时转售胜过推荐;允许自购返佣时,套利成本减半,附赠六个名额就使公司每服务一次都亏钱;渠道伙伴的模仿使转售在两年内扩散到 0.877 的伙伴。身份锁之下,额外命盘不低于单价的一半即可防止转售胜过推荐,全价则同时保护直销;出生资料本身不是身份,100,000 名持卡人中会有 14,259.5 对陌生人共享「日期加时辰」的键。 第五章把六级菜单写成精确求解的甄别问题,最优价格为 12、117、287、567、12537 与 34537,商业牌照与总代席位贡献约八成利润;牌照被压到 3,000 时利润只剩 0.434。第六章显示菜单在需求设错时的脆弱(最差 5% 的世界只拿到最优的 0.315)、加购折扣只在估值负相关时成立,以及季度利润低于计划八成的概率 0.112。第七章与《史记·平准书》的算缗告缗、《管子·乘马》与宋代盐钞并读。第八章给出四项菜单审计与一项预注册实验(每组 188 位渠道伙伴)。书中以标注框区分证据、16 段可运行的 Python 程序、原典与假说。 Chapter answers: - Ch.1 When Bundling Pays: from Two Modules to Four — Chapters 1 and 2 show that separate selling does not generally beat the optimal bundle: under independent, single-scale orthogonal and heavy-tailed demand with tail index 1.5, separate selling earns only 0.825, 0.873 and 0.730 of optimal bundle revenue; only when demand is orthogonal and studios' values are three orders of magnitude above individual users' does separate selling earn 1.479 times the optimal bundle. The loss from a cheap all-access pass comes from high-value buyers coming back down: at a tenth of the sum of single prices, revenue falls to 0.439 of separate selling. (https://5dlibrary.com/answers/en/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c01/) 第1章 捆绑何时有利:从两个板块到四个板块 — 第一、二章证明,分拆并不普遍胜过最优捆绑:在独立、单一尺度正交与尾指数 1.5 的重尾需求下,分拆只得到最优捆绑收入的 0.825、0.873 与 0.730;只有当需求正交、且工作室与个人用户的估值相差三个数量级时,分拆才是最优捆绑的 1.479 倍。廉价全通卡的损失来自高估值客户的回流:卡价为单品价总和的十分之一时,收入只剩分拆的 0.439。 (https://5dlibrary.com/answers/zh/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c01/) - Ch.2 How a Cheap All-Access Pass Cannibalizes, and Heavy Tails — The loss from a cheap all-access pass comes from high-value buyers coming back down: once the pass is cheaper than what studios pay for the business module, they all switch; at a tenth of the sum of single prices, revenue falls to 0.439. Heavy tails shrink the bundle's edge from about a third to about 5%, but with this book's parameters do not let separate selling overtake it. (https://5dlibrary.com/answers/en/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c02/) 第2章 廉价全通卡的蚕食与重尾估值 — 廉价全通卡的损失来自高估值客户的回流:卡价一旦低于工作室为商业板块所付的价格,工作室全部转卡;卡价为单品总和的十分之一时,收入只剩 0.439。重尾估值使捆绑的优势从约三分之一降到约 5%,但在本书参数下没有使分拆反超。 (https://5dlibrary.com/answers/zh/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c02/) - Ch.3 Resale Arbitrage and an Identity Hash Lock — Chapters 3 and 4 model resale arbitrage: with a 50% affiliate commission, resale beats referral when the extra slots satisfy k + 1 > 4.020; with self-referral, arbitrage costs half as much and six extra slots make the firm lose money on every reading it serves; imitation among channel partners spreads resale to 0.877 of them within two years. Under an identity lock, an extra-chart price of at least half the single price keeps resale from beating referral, and full price also protects direct sales; birth data are not an identity, as among 100,000 card holders 14,259.5 pairs of strangers share a date-and-double-hour key. (https://5dlibrary.com/answers/en/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c03/) 第3章 转售套利与身份哈希锁 — 第三、四章建立转售套利模型:在 50% 的分销佣金下,附赠名额满足 k + 1 > 4.020 时转售胜过推荐;允许自购返佣时,套利成本减半,附赠六个名额就使公司每服务一次都亏钱;渠道伙伴的模仿使转售在两年内扩散到 0.877 的伙伴。身份锁之下,额外命盘不低于单价的一半即可防止转售胜过推荐,全价则同时保护直销;出生资料本身不是身份,100,000 名持卡人中会有 14,259.5 对陌生人共享「日期加时辰」的键。 (https://5dlibrary.com/answers/zh/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c03/) - Ch.4 Self-Referral, Channel Blow-Through and Imitating Partners — Self-referral halves the cost of arbitrage: two extra slots make resale beat referral, and at six the firm pays for every resold slot it serves. Imitation among partners spreads profitable resale: with 3 extra slots and self-referral, 0.877 of partners resell after two years. The identity lock and "no commission on a partner's own purchase" together make resale unprofitable from the start. (https://5dlibrary.com/answers/en/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c04/) 第4章 自购返佣、渠道穿仓与伙伴的模仿 — 允许自购返佣会让套利成本减半:附赠 2 个名额,转售就胜过推荐;附赠 6 个名额,公司开始为每个转售名额倒贴。渠道伙伴之间的模仿会把有利可图的转售扩散开来:附赠 3 个名额并允许自购返佣时,两年后 0.877 的伙伴在转售。身份锁与「不对本人购买付佣金」两条规则合起来,使转售在一开始就无利可图。 (https://5dlibrary.com/answers/zh/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c04/) - Ch.5 The Six-Tier Menu: an Exactly Solved Screening Problem — Chapter 5 solves the six-tier menu exactly as a screening problem: the optimal prices are 12, 117, 287, 567, 12537 and 34537, and the business licence and agency seat bring about four fifths of profit; with the licence held at 3,000, profit falls to 0.434. (https://5dlibrary.com/answers/en/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c05/) 第5章 六级菜单:一个精确求解的甄别问题 — 第五章把六级菜单写成精确求解的甄别问题,最优价格为 12、117、287、567、12537 与 34537,商业牌照与总代席位贡献约八成利润;牌照被压到 3,000 时利润只剩 0.434。 (https://5dlibrary.com/answers/zh/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c05/) - Ch.6 How Robust Is the Menu: Misspecified Demand, Add-on Prices and Whale Risk — Chapter 6 shows the menu's fragility when demand is misspecified (the worst 5% of worlds earn 0.315 of their optimum), that add-on discounts are right only for negatively correlated values, and a 0.112 chance that a quarter's profit falls below 80% of plan. (https://5dlibrary.com/answers/en/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c06/) 第6章 菜单的稳健性:需求设错、加购定价与大客户风险 — 第六章显示菜单在需求设错时的脆弱(最差 5% 的世界只拿到最优的 0.315)、加购折扣只在估值负相关时成立,以及季度利润低于计划八成的概率 0.112。 (https://5dlibrary.com/answers/zh/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c06/) - Ch.7 The Han Tax on Hoarded Cash, the Guanzi and Song Salt Vouchers — The Pingzhun shu's tax on hoarded cash graded rates by occupation and gave half of the confiscation to informers, a structure that mirrors this book's segment pricing and 50% commission; it also shows how a large share can change the behaviour of a whole society. The Guanzi's "the market is the standard of goods" is a reminder that prices must rest on market values. The over-issue and depreciation of Song salt vouchers are the historical version of resold extra slots. (https://5dlibrary.com/answers/en/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c07/) 第7章 算缗告缗、《管子》与宋代盐钞 — 《平准书》的算缗令按经营性质分级征收,并把没收财产的一半分给告发者,与本书的分级定价和 50% 佣金同构;它也显示了高额分成如何改变整个社会的行为。《管子》「市者,货之准也」提醒定价必须以市场估值为准。宋代盐钞的超发与贬值,是附赠名额被转售的历史版本。 (https://5dlibrary.com/answers/zh/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c07/) - Ch.8 A Menu Audit and a Pre-Registered Experiment — Chapter 8 gives a four-flag menu audit and a pre-registered experiment (188 channel partners per arm). Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. (https://5dlibrary.com/answers/en/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c08/) 第8章 菜单审计与一项预注册的实验 — 第八章给出四项菜单审计与一项预注册实验(每组 188 位渠道伙伴)。书中以标注框区分证据、16 段可运行的 Python 程序、原典与假说。 (https://5dlibrary.com/answers/zh/unbundling-economics-anti-arbitrage-licensing-upsell-ladders/c08/) ## 5DL-B-069 · Retrodictive Belief Locking, Sequential Pain-Point Agitation and Markovian Upsell Automata 中文书名:回溯信念锁定、序贯痛点唤醒与马尔可夫升单自动机 DOI: https://doi.org/10.5281/zenodo.24090186 Cite: Fan, Y. (2026). Retrodictive Belief Locking, Sequential Pain-Point Agitation and Markovian Upsell Automata (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090186 Book page: https://5dlibrary.com/books/retrodictive-belief-locking-sequential-upsell-conversion-automata/ Scope: Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. Abstract: This book studies three mechanisms of reading services that are often treated as "conversion techniques": winning trust with retrodictive statements about the client's past, pushing upsells with price steps and sunk cost, and the conversion chain from a single reading to an academy course to a business licence. All parameters are model parameters, not data from any business; the book neither claims nor tests that any reading comes true, and it computes the costs of each mechanism together with its gains. Chapter 1 shows that the "trust avalanche" is a transition in feeling, not in evidence: a generic retrodictive statement has a true likelihood ratio of only 1.19, and a sceptic needs 27 hits to be half convinced, but if its base rate of 0.80 is taken for 0.20, 3 suffice; among 20,000 simulated users a reader with no insight at all convinces 0.5263 within 6 generic statements, while users who update with the true ratios are almost never convinced. A "trough year" that is a shared low for a quarter of the population has a likelihood ratio of 3.60, not 13.50. Chapter 2 gives a three-way sequential test and shows that selective recall turns statements right half the time into a remembered accuracy of 0.7493, against 0.4995 when recorded beforehand. Chapter 3 finds the best upsell price ratio under a resistance threshold and sunk cost (1.99 at a first purchase, 1.86 with refunds) and the best timing (an offer on day 7 earns several times one made the next day). Chapter 4 computes pain-point agitation: with refunds and churn counted the best intensity is still 0.63, and the share of buyers in acute distress rises with intensity from 0.100 to 0.206; a cooling-off period and a distress screen together bring it down to 0.056. Chapter 5 writes "reading, academy, licence" as an absorbing Markov chain: a reader becomes a licensee with probability 0.0267, and her LTV is 4.35 times her first purchase, about three quarters of it from the licence path. Chapter 6 shows that a rare cross-module transition estimated from 1,000 users still has an interval 0.74 times as wide as the true value, that the reading income cannibalized by "learning to read it yourself" is smaller than the course fee, and that pooling two kinds of user into one chain badly underestimates long-tail retention. Chapter 7 reads the "Mo" and "Fei qian" chapters of the Guiguzi and "Shuo nan" of the Han Feizi. Chapter 8 gives a calculator for users and a pre-registered blinded test (152 participants). Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. 摘要:本书研究测算服务中三个常被当作「转化技巧」的机制:以说中过去的回溯断言赢得信任,以价格台阶与沉没成本推动升单,以及从单次测算到命理学堂课程、再到商业牌照的转化链。全部参数都是模型参数,不是任何企业的数据;本书不主张、也不检验任何占断是否应验,并把每一种机制的代价与收益一起计算。 第一章证明「信任雪崩」是感受上的相变而不是证据上的:泛泛的回溯断言真实似然比只有 1.19,怀疑者需要 27 次命中才半信,但若把基率 0.80 误判为 0.20,3 次就够;在 20,000 位模拟用户中,一位没有任何洞察力的测算师在 6 条泛泛断言之内使 0.5263 的用户确信,而按真实似然比更新的用户几乎无人确信。「波谷年份」若是四分之一人群的共同低谷,似然比只有 3.60 而非 13.50。第二章给出三选一的序贯检验,并显示选择性回忆把对错各半的断言记成 0.7493 的准确率,事先记录则为 0.4995。第三章求含抗拒阈值与沉没成本的最优升单跳升比(首单 1.99,计入退款 1.86)与时机(第 7 天的报价收入是次日的数倍)。第四章计算痛点唤醒:计入退款与流失后最优强度仍为 0.63,买家中急性困境者的比例随强度从 0.100 升到 0.206;冷静期与困境筛查两道护栏使这一比例降到 0.056。 第五章把「测算、学堂、牌照」写成吸收马尔可夫链:测算用户成为牌照持有者的概率为 0.0267,LTV 为首单的 4.35 倍,约四分之三来自牌照之路。第六章显示稀有的跨板块转移在 1,000 位用户时区间宽度仍为真值的 0.74 倍;「自己学会算」蚕食的测算收入小于学费;两类用户合成一条链会严重低估长尾留存。第七章与《鬼谷子》的摩篇、飞箝及《韩非子·说难》并读。第八章给出一个用户核算器与一项预注册的盲法检验(152 位参与者)。书中以标注框区分证据、16 段可运行的 Python 程序、原典与假说。 Chapter answers: - Ch.1 Retrodictive Statements and the "Trust Avalanche" — Chapter 1 shows that the "trust avalanche" is a transition in feeling, not in evidence: a generic retrodictive statement has a true likelihood ratio of only 1.19, and a sceptic needs 27 hits to be half convinced, but if its base rate of 0.80 is taken for 0.20, 3 suffice; among 20,000 simulated users a reader with no insight at all convinces 0.5263 within 6 generic statements, while users who update with the true ratios are almost never convinced. A "trough year" that is a shared low for a quarter of the population has a likelihood ratio of 3.60, not 13.50. (https://5dlibrary.com/answers/en/retrodictive-belief-locking-sequential-upsell-conversion-automata/c01/) 第1章 回溯断言与「信任雪崩」 — 第一章证明「信任雪崩」是感受上的相变而不是证据上的:泛泛的回溯断言真实似然比只有 1.19,怀疑者需要 27 次命中才半信,但若把基率 0.80 误判为 0.20,3 次就够;在 20,000 位模拟用户中,一位没有任何洞察力的测算师在 6 条泛泛断言之内使 0.5263 的用户确信,而按真实似然比更新的用户几乎无人确信。「波谷年份」若是四分之一人群的共同低谷,似然比只有 3.60 而非 13.50。 (https://5dlibrary.com/answers/zh/retrodictive-belief-locking-sequential-upsell-conversion-automata/c01/) - Ch.2 Testing Retrodiction Fairly: Sequential Tests and Selective Recall — Chapter 2 gives a three-way sequential test and shows that selective recall turns statements right half the time into a remembered accuracy of 0.7493, against 0.4995 when recorded beforehand. (https://5dlibrary.com/answers/en/retrodictive-belief-locking-sequential-upsell-conversion-automata/c02/) 第2章 公平地检验回溯:序贯检验与选择性回忆 — 第二章给出三选一的序贯检验,并显示选择性回忆把对错各半的断言记成 0.7493 的准确率,事先记录则为 0.4995。 (https://5dlibrary.com/answers/zh/retrodictive-belief-locking-sequential-upsell-conversion-automata/c02/) - Ch.3 The Commitment Ratchet: Price Ratios and Upsell Timing — Chapter 3 finds the best upsell price ratio under a resistance threshold and sunk cost (1.99 at a first purchase, 1.86 with refunds) and the best timing (an offer on day 7 earns several times one made the next day). (https://5dlibrary.com/answers/en/retrodictive-belief-locking-sequential-upsell-conversion-automata/c03/) 第3章 承诺棘轮:价格跳升比与升单时机 — 第三章求含抗拒阈值与沉没成本的最优升单跳升比(首单 1.99,计入退款 1.86)与时机(第 7 天的报价收入是次日的数倍)。 (https://5dlibrary.com/answers/zh/retrodictive-belief-locking-sequential-upsell-conversion-automata/c03/) - Ch.4 What Pain-Point Agitation Buys and Costs, and Two Guardrails — Chapter 4 computes pain-point agitation: with refunds and churn counted the best intensity is still 0.63, and the share of buyers in acute distress rises with intensity from 0.100 to 0.206; a cooling-off period and a distress screen together bring it down to 0.056. (https://5dlibrary.com/answers/en/retrodictive-belief-locking-sequential-upsell-conversion-automata/c04/) 第4章 痛点唤醒的收益与代价,以及两道护栏 — 第四章计算痛点唤醒:计入退款与流失后最优强度仍为 0.63,买家中急性困境者的比例随强度从 0.100 升到 0.206;冷静期与困境筛查两道护栏使这一比例降到 0.056。 (https://5dlibrary.com/answers/zh/retrodictive-belief-locking-sequential-upsell-conversion-automata/c04/) - Ch.5 The Three-Stage Rocket: an Absorbing Markov Chain of Reading, Academy and Licence — Chapter 5 writes "reading, academy, licence" as an absorbing Markov chain: a reader becomes a licensee with probability 0.0267, and her LTV is 4.35 times her first purchase, about three quarters of it from the licence path. (https://5dlibrary.com/answers/en/retrodictive-belief-locking-sequential-upsell-conversion-automata/c05/) 第5章 三级火箭:测算、学堂与牌照的吸收马尔可夫链 — 第五章把「测算、学堂、牌照」写成吸收马尔可夫链:测算用户成为牌照持有者的概率为 0.0267,LTV 为首单的 4.35 倍,约四分之三来自牌照之路。 (https://5dlibrary.com/answers/zh/retrodictive-belief-locking-sequential-upsell-conversion-automata/c05/) - Ch.6 Cross-Module Transition Matrices: Estimation, Cannibalization and Heterogeneity — Chapter 6 shows that a rare cross-module transition estimated from 1,000 users still has an interval 0.74 times as wide as the true value, that the reading income cannibalized by "learning to read it yourself" is smaller than the course fee, and that pooling two kinds of user into one chain badly underestimates long-tail retention. (https://5dlibrary.com/answers/en/retrodictive-belief-locking-sequential-upsell-conversion-automata/c06/) 第6章 跨板块转移矩阵:估计、蚕食与异质性 — 第六章显示稀有的跨板块转移在 1,000 位用户时区间宽度仍为真值的 0.74 倍;「自己学会算」蚕食的测算收入小于学费;两类用户合成一条链会严重低估长尾留存。 (https://5dlibrary.com/answers/zh/retrodictive-belief-locking-sequential-upsell-conversion-automata/c06/) - Ch.7 "Probing" and "Flying and Clamping" in the Guiguzi, "The Difficulties of Persuasion" in the Han Feizi — Han Fei says that the difficulty of persuasion lies not in one's own knowledge and eloquence but in knowing the other's mind so that one's words match it; and that the dragon has a reverse scale under its throat, death to whoever touches it, and so has the ruler. Item (2) of the program writes the "reverse scale" as each person's tolerance (spread uniformly between 0.3 and 1.0 on the agitation scale of Chapter 4): agitation above it turns the client against the seller. At the best intensity of Chapter 4 without guardrails, 0.63, the reverse scale of 0.471 of clients is touched; at the best intensity with both guardrails, 0.43, the share is 0.186. (https://5dlibrary.com/answers/en/retrodictive-belief-locking-sequential-upsell-conversion-automata/c07/) 第7章 《鬼谷子》的摩与飞箝,《韩非子》的说难 — 韩非说,游说之难,不在自己的知识与口才,而在了解对方的心,使自己的说辞与之相当;又说龙喉下有逆鳞,触之者必死,人主也有逆鳞。程序的第二项把「逆鳞」写成每个人的一个承受阈值(在第四章的唤醒强度上均匀分布于 0.3 到 1.0 之间):唤醒强度超过阈值,客户就会转而反对卖家。在第四章不设护栏的最优强度 0.63 下,0.471 的客户被触及逆鳞;在两道护栏下的最优强度 0.43 下,比例为 0.186。 (https://5dlibrary.com/answers/zh/retrodictive-belief-locking-sequential-upsell-conversion-automata/c07/) - Ch.8 A Calculator for the User and a Pre-Registered Blinded Test — Chapter 8 gives a calculator for users and a pre-registered blinded test (152 participants). Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. (https://5dlibrary.com/answers/en/retrodictive-belief-locking-sequential-upsell-conversion-automata/c08/) 第8章 一个用户自己的核算器与一项预注册的盲法检验 — 第八章给出一个用户核算器与一项预注册的盲法检验(152 位参与者)。书中以标注框区分证据、16 段可运行的 Python 程序、原典与假说。 (https://5dlibrary.com/answers/zh/retrodictive-belief-locking-sequential-upsell-conversion-automata/c08/) ## 5DL-B-070 · Multimodal Calendrical Tensor Fusion, Modulo-90 Harmonic Dials and Cross-System Resonance Diagnostics 中文书名:多模态历法张量融合、模九十谐波拨盘与跨系统共振诊断 DOI: https://doi.org/10.5281/zenodo.24090189 Cite: Fan, Y. (2026). Multimodal Calendrical Tensor Fusion, Modulo-90 Harmonic Dials and Cross-System Resonance Diagnostics (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090189 Book page: https://5dlibrary.com/books/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/ Scope: Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. Abstract: Multi-system reading software often puts bone weight, Ziping, Zi Wei Dou Shu, Qimen Dunjia and Western astrology into one report and announces resonance when "several systems point at the same year". This book studies only the mathematical structure of these five calendrical symbol systems and the arithmetic of combining them, and checks two implicit premises: that each system carries independent information, and that agreement among systems deserves more trust than one system. It neither claims nor tests that any reading comes true; all data are generated by its programs. Chapter 1 writes bone weight as a sum of four lookup tables: 259,200 births fall on only 51 totals, from 2.1 to 7.1 liang, and the verse "seven liang two qian" cannot be reached; a birth hour off by one double-hour changes the total for 0.8333 of births. Chapter 2 writes Ziping climate balance as two-dimensional vectors of warmth and moisture: "fire in winter, water in summer" follows directly from geometry, and changing the 70/30 weighting to 0.6 changes the verdict in 0.0500 of cells. Chapter 3 builds all 43,200 Zi Wei charts: single edges of flying ji are close to random (1.0078 self-ji palaces per chart), but the twelve ji edges reach only 7.458 palaces on average; double stacked ji appears in 0.4137 of charts, and a lunar day off by one changes the stacked palaces in 0.4735. Chapter 4 shows that the midpoint equation on the 90-degree dial is unambiguous, but 18 points with a 1.5-degree orb give 81.60 planetary pictures per chart by chance, and solar arcs bring about 135.4 contacts a year. Chapter 5 puts the five systems on one set of births: Qimen bureaus and the Sun's dial sector share 0.585 bits, while Zi Wei is almost independent; for the binary verdicts defined in the book the effective number of independent votes is 5.17, so independence depends on the question. Chapter 6 uses Condorcet's theorem and a Gaussian copula to show that with correlation 0.6 the majority accuracy of 9 systems rises only from 0.6014 to 0.6231, and that two unconnected systems "resonate" by chance in 10.00 years of 40 on average. Chapter 7 reads the Qiongtong baojian, the common table of four transformations and Witte's Rules for Planetary Pictures. Chapter 8 gives a convention budget (two school conventions make 0.1387 of combined reports differ) and a pre-registered test of resonance years (127 participants). Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. 摘要:多系统测算软件常把称骨、子平、紫微斗数、奇门遁甲与西方占星放进同一份报告,并在「几套系统同时指向同一年」时宣布共振。本书只研究这五套历法符号系统的数学结构与多模型集成的算术,检查两个隐含的前提:每套系统各有独立的信息,以及多系统的一致比单系统更可信。本书不主张、也不检验任何占断是否应验;全部数据为程序生成。 第一章把称骨写成四张查表之和:259,200 种出生只落在 51 种总重上,从 2.1 到 7.1 两,通行歌诀中的「七两二钱」无法达到;出生时辰差一格,0.8333 的总重改变。第二章把子平调候写成寒暖燥湿的二维向量,「冬用火、夏用水」由几何直接推出,70/30 加权改为 0.6 时 0.0500 的格子改变结论。第三章构造全部 43,200 种紫微命盘:飞宫化忌的单条边接近随机(每盘 1.0078 个自化忌宫),但十二条忌边平均只到达 7.458 个宫;双重叠忌出现在 0.4137 的命盘中,农历日差一天使 0.4735 的命盘改变叠忌宫位。第四章证明 90 度拨盘上的中点方程没有歧义,但 18 个点在 1.5 度容许度下每盘偶然出现 81.60 个行星图形,太阳弧每年带来约 135.4 次接触。 第五章把五套系统放在同一批出生上:奇门局数与太阳的拨盘扇区共享 0.585 比特,紫微几乎独立;按本书定义的二元判断,有效独立票数为 5.17,独立与否取决于问题。第六章用孔多塞定理与高斯 Copula 显示,相关系数为 0.6 时,9 套系统的多数票正确率只从 0.6014 升到 0.6231;两套毫无关联的系统在 40 年中平均有 10.00 年偶然「共振」。第七章与《穷通宝鉴》、四化通行表及 Witte 的《行星图形规则集》并读。第八章给出约定预算(两项流派约定使 0.1387 的合盘报告不同)与一项预注册的共振年份检验(127 位参与者)。书中以标注框区分证据、16 段可运行的 Python 程序、原典与假说。 Chapter answers: - Ch.1 Bone Weight: a Sum of Four Lookup Tables — Chapter 1 writes bone weight as a sum of four lookup tables: 259,200 births fall on only 51 totals, from 2.1 to 7.1 liang, and the verse "seven liang two qian" cannot be reached; a birth hour off by one double-hour changes the total for 0.8333 of births. (https://5dlibrary.com/answers/en/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c01/) 第1章 称骨:四张查表之和 — 第一章把称骨写成四张查表之和:259,200 种出生只落在 51 种总重上,从 2.1 到 7.1 两,通行歌诀中的「七两二钱」无法达到;出生时辰差一格,0.8333 的总重改变。 (https://5dlibrary.com/answers/zh/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c01/) - Ch.2 Ziping Climate Balance: Warmth and Moisture as Vectors — Chapter 2 writes Ziping climate balance as two-dimensional vectors of warmth and moisture: "fire in winter, water in summer" follows directly from geometry, and changing the 70/30 weighting to 0.6 changes the verdict in 0.0500 of cells. (https://5dlibrary.com/answers/en/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c02/) 第2章 子平调候:寒暖燥湿的二维向量 — 第二章把子平调候写成寒暖燥湿的二维向量,「冬用火、夏用水」由几何直接推出,70/30 加权改为 0.6 时 0.0500 的格子改变结论。 (https://5dlibrary.com/answers/zh/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c02/) - Ch.3 Flying Four Transformations: a Directed Graph and Three-Layer Stacked Ji — Chapter 3 builds all 43,200 Zi Wei charts: single edges of flying ji are close to random (1.0078 self-ji palaces per chart), but the twelve ji edges reach only 7.458 palaces on average; double stacked ji appears in 0.4137 of charts, and a lunar day off by one changes the stacked palaces in 0.4735. (https://5dlibrary.com/answers/en/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c03/) 第3章 飞宫四化:有向图与三层叠忌 — 第三章构造全部 43,200 种紫微命盘:飞宫化忌的单条边接近随机(每盘 1.0078 个自化忌宫),但十二条忌边平均只到达 7.458 个宫;双重叠忌出现在 0.4137 的命盘中,农历日差一天使 0.4735 的命盘改变叠忌宫位。 (https://5dlibrary.com/answers/zh/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c03/) - Ch.4 The Hamburg School's 90-Degree Dial: Midpoints, Pictures and Solar Arcs — Chapter 4 shows that the midpoint equation on the 90-degree dial is unambiguous, but 18 points with a 1.5-degree orb give 81.60 planetary pictures per chart by chance, and solar arcs bring about 135.4 contacts a year. (https://5dlibrary.com/answers/en/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c04/) 第4章 汉堡学派的 90 度拨盘:中点、图形与太阳弧 — 第四章证明 90 度拨盘上的中点方程没有歧义,但 18 个点在 1.5 度容许度下每盘偶然出现 81.60 个行星图形,太阳弧每年带来约 135.4 次接触。 (https://5dlibrary.com/answers/zh/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c04/) - Ch.5 A Multimodal Tensor: How Much Do the Five Systems Share? — Chapter 5 puts the five systems on one set of births: Qimen bureaus and the Sun's dial sector share 0.585 bits, while Zi Wei is almost independent; for the binary verdicts defined in the book the effective number of independent votes is 5.17, so independence depends on the question. (https://5dlibrary.com/answers/en/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c05/) 第5章 多模态张量:五套系统共享多少信息 — 第五章把五套系统放在同一批出生上:奇门局数与太阳的拨盘扇区共享 0.585 比特,紫微几乎独立;按本书定义的二元判断,有效独立票数为 5.17,独立与否取决于问题。 (https://5dlibrary.com/answers/zh/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c05/) - Ch.6 The Limits of Combining: Condorcet, a Gaussian Copula and the Base Rate of "Resonance" — Chapter 6 uses Condorcet's theorem and a Gaussian copula to show that with correlation 0.6 the majority accuracy of 9 systems rises only from 0.6014 to 0.6231, and that two unconnected systems "resonate" by chance in 10.00 years of 40 on average. (https://5dlibrary.com/answers/en/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c06/) 第6章 集成的上限:孔多塞、高斯 Copula 与「共振」的基率 — 第六章用孔多塞定理与高斯 Copula 显示,相关系数为 0.6 时,9 套系统的多数票正确率只从 0.6014 升到 0.6231;两套毫无关联的系统在 40 年中平均有 10.00 年偶然「共振」。 (https://5dlibrary.com/answers/zh/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c06/) - Ch.7 The Qiongtong baojian, the Common Table of Four Transformations, and Witte's Rules for Planetary Pictures — The Qiongtong baojian is arranged by ten stems and twelve months, 120 sections, exactly the structure of Chapter 2's climate table. The common transformation table has 40 assignments over 15 stars, with a different ji star for each stem; repeated palace stems mean that twelve palaces use only 10 stems. Witte's rulebook is characterised by midpoints and 8 hypothetical points, its core rules published in 1928; its geometry on the dial is unambiguous, but each chart has 2,448 possible pictures to test. (https://5dlibrary.com/answers/en/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c07/) 第7章 《穷通宝鉴》、四化通行表与 Witte 的《行星图形规则集》 — 《穷通宝鉴》按十干乘十二月分论,共 120 节,正是第二章调候表的结构。四化通行表有 40 个映射、15 颗星,化忌之星十干各异;宫干的重复使十二宫只用到 10 个天干。Witte 的规则集以中点与 8 个假设点为特征,其核心规则 1928 年出版;它在拨盘上的几何没有歧义,但每张星盘要检验 2,448 个可能的图形。 (https://5dlibrary.com/answers/zh/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c07/) - Ch.8 A Convention Budget and a Pre-Registered Test of "Resonance Years" — Chapter 8 gives a convention budget (two school conventions make 0.1387 of combined reports differ) and a pre-registered test of resonance years (127 participants). Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. (https://5dlibrary.com/answers/en/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c08/) 第8章 约定预算与一项预注册的「共振年份」检验 — 第八章给出约定预算(两项流派约定使 0.1387 的合盘报告不同)与一项预注册的共振年份检验(127 位参与者)。书中以标注框区分证据、16 段可运行的 Python 程序、原典与假说。 (https://5dlibrary.com/answers/zh/multimodal-calendrical-tensor-fusion-harmonic-dials-resonance/c08/) ## 5DL-B-071 · Tarot Hypergeometric Sampling, Lenormand Grid Topology and Kabbalistic Tree Graphs 中文书名:塔罗超几何抽样、雷诺曼网格拓扑与卡巴拉生命树图论 DOI: https://doi.org/10.5281/zenodo.24090191 Cite: Fan, Y. (2026). Tarot Hypergeometric Sampling, Lenormand Grid Topology and Kabbalistic Tree Graphs (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090191 Book page: https://5dlibrary.com/books/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: A tarot spread is sampling without replacement, a Lenormand Grand Tableau is a random permutation on a grid, and the Kabbalistic Tree of Life is a graph. This book writes these three historical symbol systems as mathematical objects and studies only their combinatorial, probabilistic and graph-theoretic structure, with an analysis of base rates and the Barnum effect. It neither claims nor tests that any card meaning or reading comes true; all data are generated by its programs. Chapter 1 computes spreads exactly with hypergeometric and multivariate hypergeometric laws: in the ten cards of a Celtic Cross the expected number of majors is 2.8205 and the variance 1.7883, below the binomial 2.0250; three or more majors occur with probability 0.57847, four or more of one suit with 0.28678, and two minors of the same rank with 0.75189. Chapter 2 computes mixing times: 78 cards need eight GSR riffles before the total variation falls below 0.5 (0.3065); riffles produce no reversals, and the correlation of reversals depends on the turning action: splitting, turning half and riffling finely gives a lag-1 correlation of −0.013, no different from independent turning, while turning packets raises it to 0.774. Chapter 3 gives exact base rates for the Lenormand Grand Tableau: two named cards touch on the 9 × 4 grid with probability 0.1698, their mean Manhattan distance is 13/3, and at least one card lands in its own house with probability 0.63212. Chapter 4 writes the Tree as a graph of 10 vertices and 22 edges: algebraic connectivity 1.2805, 54349 spanning trees and an automorphism group containing only the left-right reflection; Tiphereth leads in degree, eigenvector, betweenness and closeness centrality and holds 0.5321 of all betweenness; 533 of the 573 simple paths from Malkuth to Kether pass through it, and a random walk needs 21.0358 steps on average to first reach Kether. Chapter 5 computes information and multiple comparisons: two Celtic Crosses share at least one card with probability 0.7689, and a random spread shows at least one of eight common features with probability 0.9099. Chapter 6 analyses the Barnum effect with a coverage model and a model of selective recall. Chapter 7 reads the Sefer Yetzirah, Kircher's tree and Waite's numbering of 1910: Kircher lettered the paths in alphabetical order, so only 11 letters fall on paths matching their classes of three, seven and twelve. Chapter 8 gives a four-step check and two pre-registered studies: a blind matching test needing 359 participants and a shuffle audit needing 263 spreads. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:塔罗牌阵是不放回的抽样,雷诺曼大桌阵是网格上的随机排列,卡巴拉生命树是一张图。本书把这三套历史符号系统写成数学对象,只研究它们的组合、概率与图论结构,并附基率与巴纳姆效应分析。本书不主张、也不检验任何牌义或解读是否应验;全部数据由书中程序生成。 第一章用超几何与多元超几何分布精确计算牌阵:十张凯尔特十字中大阿卡纳的期望为 2.8205,方差 1.7883,小于二项近似的 2.0250;三张以上大阿卡纳的概率为 0.57847,某个花色四张以上为 0.28678,两张小阿卡纳同点为 0.75189。第二章计算交叉洗牌的混合时间:78 张牌需要八次 GSR 交叉洗牌才使全变差降到 0.5 以下(0.3065);交叉洗牌不产生逆位,逆位的相关结构取决于翻牌动作:对切转半后细致交叉洗牌的相邻相关系数为 −0.013,与独立翻牌无异,成叠转牌则升到 0.774。第三章给出雷诺曼大桌阵上的精确基率:两张指定的牌在 9 × 4 网格上相邻的概率为 0.1698,平均曼哈顿距离为 13/3,至少一张牌落在本宫的概率为 0.63212。 第四章把生命树写成 10 个顶点、22 条边的图:代数连通度 1.2805,生成树 54349 棵,自同构群只有左右反射;Tiphereth 在度、特征向量、介数与紧密度四种中心性上都居首位,独占介数的 0.5321;从 Malkuth 到 Kether 的 573 条简单路径中有 533 条经过它,随机游走的首达时间为 21.0358 步。第五章计算信息量与多重比较:两次凯尔特十字至少共享一张牌的概率为 0.7689,随机牌阵具有八种常见特征中至少一种的概率为 0.9099。第六章用覆盖模型与选择性记忆模型分析巴纳姆效应。第七章与《形成之书》、基歇尔的生命树与韦特 1910 年的牌序并读:基歇尔按字母表顺序分配字母,只有 11 个落在与其三、七、十二分类相符的路径上。第八章给出四步核查与两项预注册研究:盲法配对检验需 359 名参与者,洗牌审计需 263 个牌阵。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Hypergeometric Sampling of Seventy-Eight Cards — Chapter 1 computes spreads exactly with hypergeometric and multivariate hypergeometric laws: in the ten cards of a Celtic Cross the expected number of majors is 2.8205 and the variance 1.7883, below the binomial 2.0250; three or more majors occur with probability 0.57847, four or more of one suit with 0.28678, and two minors of the same rank with 0.75189. (https://5dlibrary.com/answers/en/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c01/) 第1章 七十八张牌的超几何抽样 — 第一章用超几何与多元超几何分布精确计算牌阵:十张凯尔特十字中大阿卡纳的期望为 2.8205,方差 1.7883,小于二项近似的 2.0250;三张以上大阿卡纳的概率为 0.57847,某个花色四张以上为 0.28678,两张小阿卡纳同点为 0.75189。 (https://5dlibrary.com/answers/zh/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c01/) - Ch.2 Shuffles, Mixing Times and the Correlation of Reversals — Chapter 2 computes mixing times: 78 cards need eight GSR riffles before the total variation falls below 0.5 (0.3065); riffles produce no reversals, and the correlation of reversals depends on the turning action: splitting, turning half and riffling finely gives a lag-1 correlation of −0.013, no different from independent turning, while turning packets raises it to 0.774. (https://5dlibrary.com/answers/en/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c02/) 第2章 洗牌、混合时间与正逆位的相关结构 — 第二章计算交叉洗牌的混合时间:78 张牌需要八次 GSR 交叉洗牌才使全变差降到 0.5 以下(0.3065);交叉洗牌不产生逆位,逆位的相关结构取决于翻牌动作:对切转半后细致交叉洗牌的相邻相关系数为 −0.013,与独立翻牌无异,成叠转牌则升到 0.774。 (https://5dlibrary.com/answers/zh/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c02/) - Ch.3 The Grid Combinatorics of the Lenormand Grand Tableau — Chapter 3 gives exact base rates for the Lenormand Grand Tableau: two named cards touch on the 9 × 4 grid with probability 0.1698, their mean Manhattan distance is 13/3, and at least one card lands in its own house with probability 0.63212. (https://5dlibrary.com/answers/en/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c03/) 第3章 雷诺曼大桌阵的网格组合学 — 第三章给出雷诺曼大桌阵上的精确基率:两张指定的牌在 9 × 4 网格上相邻的概率为 0.1698,平均曼哈顿距离为 13/3,至少一张牌落在本宫的概率为 0.63212。 (https://5dlibrary.com/answers/zh/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c03/) - Ch.4 The Tree of Life as a Graph: Spectrum, Centrality and Paths — Chapter 4 writes the Tree as a graph of 10 vertices and 22 edges: algebraic connectivity 1.2805, 54349 spanning trees and an automorphism group containing only the left-right reflection; Tiphereth leads in degree, eigenvector, betweenness and closeness centrality and holds 0.5321 of all betweenness; 533 of the 573 simple paths from Malkuth to Kether pass through it, and a random walk needs 21.0358 steps on average to first reach Kether. (https://5dlibrary.com/answers/en/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c04/) 第4章 生命之树的图论:谱、中心性与路径 — 第四章把生命树写成 10 个顶点、22 条边的图:代数连通度 1.2805,生成树 54349 棵,自同构群只有左右反射;Tiphereth 在度、特征向量、介数与紧密度四种中心性上都居首位,独占介数的 0.5321;从 Malkuth 到 Kether 的 573 条简单路径中有 533 条经过它,随机游走的首达时间为 21.0358 步。 (https://5dlibrary.com/answers/zh/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c04/) - Ch.5 Information, Repeats and the Base Rate of "Something Special" — Chapter 5 computes information and multiple comparisons: two Celtic Crosses share at least one card with probability 0.7689, and a random spread shows at least one of eight common features with probability 0.9099. (https://5dlibrary.com/answers/en/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c05/) 第5章 信息量、重复与「总有特别之处」的基率 — 第五章计算信息量与多重比较:两次凯尔特十字至少共享一张牌的概率为 0.7689,随机牌阵具有八种常见特征中至少一种的概率为 0.9099。 (https://5dlibrary.com/answers/zh/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c05/) - Ch.6 The Barnum Effect: a Coverage Model and Selective Recall — A statement with $f = 0.10$ is accepted with probability 0.100 when checked against one episode, 0.651 against 10 and 0.958 against 30; with $f = 0.20$, ten episodes already give 0.893. A person's memory holds far more than 30 episodes to check against. A reading of 10 statements with $f = 0.10$ has 6.51 statements accepted on average at $M = 10$, and exactly the same number when the reading is given to anyone else. So if people are asked to pick out their own reading from several, vague statements can only produce chance: 0.5000 out of two, 0.3333 out of three, 0.2500 out of four. (https://5dlibrary.com/answers/en/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c06/) 第6章 巴纳姆效应:覆盖模型与选择性记忆 — $f = 0.10$ 的陈述,对照 1 个经历时只有 0.100 的机会被接受,对照 10 个经历时为 0.651,对照 30 个时为 0.958;$f = 0.20$ 时对照 10 个经历已达 0.893。人的记忆里可供对照的经历远不止 30 个。一份由 10 句 $f = 0.10$ 的陈述组成的解读,在 $M = 10$ 时平均被接受 6.51 句;而把这份解读交给任何一个别人,被接受的句数完全相同。因此,若请当事人从若干份解读中认出自己的那一份,笼统陈述只能给出机会水平:两份中挑一为 0.5000,三份中挑一为 0.3333,四份中挑一为 0.2500。 (https://5dlibrary.com/answers/zh/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c06/) - Ch.7 The Sefer Yetzirah, Kircher's Tree and Waite's Numbering — Chapter 7 reads the Sefer Yetzirah, Kircher's tree and Waite's numbering of 1910: Kircher lettered the paths in alphabetical order, so only 11 letters fall on paths matching their classes of three, seven and twelve. (https://5dlibrary.com/answers/en/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c07/) 第7章 《形成之书》、基歇尔之树与韦特牌序 — 第七章与《形成之书》、基歇尔的生命树与韦特 1910 年的牌序并读:基歇尔按字母表顺序分配字母,只有 11 个落在与其三、七、十二分类相符的路径上。 (https://5dlibrary.com/answers/zh/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c07/) - Ch.8 A Four-Step Check and Two Pre-Registered Studies — Chapter 8 gives a four-step check and two pre-registered studies: a blind matching test needing 359 participants and a shuffle audit needing 263 spreads. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c08/) 第8章 四步核查与两项预注册研究 — 第八章给出四步核查与两项预注册研究:盲法配对检验需 359 名参与者,洗牌审计需 263 个牌阵。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/tarot-hypergeometric-sampling-lenormand-grids-kabbalah-graphs/c08/) ## 5DL-B-072 · Mayan Tzolkin Modular Gears, Oracle-Cross Involutions and Sixty-Four-Gate Network Percolation 中文书名:玛雅卓尔金双模齿轮、神谕十字对合变换与六十四闸门网络渗流 DOI: https://doi.org/10.5281/zenodo.24090193 Cite: Fan, Y. (2026). Mayan Tzolkin Modular Gears, Oracle-Cross Involutions and Sixty-Four-Gate Network Percolation (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090193 Book page: https://5dlibrary.com/books/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This book writes three calendar and symbol systems as mathematical objects: the congruence gears of the Maya tzolkin and haab, the "fifth-force oracle" of the Dreamspell galactic signature, and the network of 64 gates, 36 channels and 9 centres of Human Design. It studies only their algebraic, combinatorial and graph-theoretic structure, with an analysis of base rates and the Barnum effect; it neither claims nor tests that any signature, type or reading comes true; all data are generated by its programs. Chapter 1 writes the tzolkin with the Chinese remainder theorem: 13 and 20 are coprime, so all 260 combinations occur, whereas the sexagenary cycle, with greatest common divisor 2 of 10 and 12, uses only 60 of 120; the Calendar Round is 18980 days, 52 haab years; 365 leaves 5 modulo 20, so the year bearers can only be one residue class modulo 5, Ik, Manik, Eb and Kaban when 0 Pop is new year. Chapter 2 compares the GMT correlation constants 584283 and 584285 and shows that Dreamspell does not advance on 29 February, so between 1950 and 2029 it never gives the same kin as the traditional count. Chapter 3 proves that analog, antipode and occult are involutions without fixed points, generating a regular action of order 20 on the day signs and the Klein four-group on the colours; the guide table equals the affine map s + 12(t − 1); two random people are linked by the oracle with probability 0.01455. Chapter 4 proves that the Human Design gate wheel is Shao Yong's Xiantian circle read from hexagram 41, and that at the level of gates the channel graph is one complete graph on four vertices plus 30 isolated edges. Chapter 5 computes exactly in a model of independent uniform activations: a single chart activates 21.5033 gates and conducts 3.9603 channels on average, with probability 0.00462 that none conducts; a two-person composite conducts 11.1618. Chapter 6 recomputes on real sky positions with a low-precision ephemeris checked against observed events: 19.759 active gates on average and a share of Reflectors of about 0.0084; 88 degrees is 93.8667 lines, so there are only 12 profiles, six with probability 0.1444 each and six with 0.0222; an hour's delay in birth time changes the Moon's line with probability 0.581. Chapter 7 reads the Venus table of the Dresden Codex (the periods 584, 365 and 260 closing together in 37960 days) and Leibniz's binary numbers. Chapter 8 gives the base rates of coincidence, the entropy of the labels and two pre-registered studies: blind recognition needing 359 participants and an oracle-tie study needing about 8561 couples. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本书把三套历法与符号系统写成数学对象:玛雅卓尔金历与哈布历的同余齿轮、Dreamspell 星系印记的「五大神谕」、人类图(Human Design)的 64 闸门、36 通道、9 中心网络。本书只研究它们的代数、组合与图论结构,并附基率与巴纳姆效应分析;不主张、也不检验任何印记、类型或解读是否应验;全部数据由书中程序生成。 第一章用中国剩余定理写出卓尔金历:13 与 20 互素,260 种组合全部出现,而干支因 10 与 12 的最大公约数为 2 只出现 120 种中的 60 种;历法轮为 18980 日,等于 52 个哈布年;365 模 20 余 5,所以年承载者只能是模 5 的一个剩余类,以 0 Pop 为新年时为 Ik、Manik、Eb、Kaban。第二章比较 GMT 关联常数 584283 与 584285,并说明 Dreamspell 在 2 月 29 日不前进,1950 年到 2029 年间与传统计数从未给出相同的 kin。第三章证明支持、挑战、隐藏推动是没有不动点的对合,在日符上生成 20 阶的正则作用,在颜色上构成克莱因四元群;指引表等价于仿射映射 s + 12(t − 1);两个随机的人之间存在神谕关系的概率为 0.01455。 第四章证明人类图的闸门轮就是邵雍先天圆图从第 41 卦起的读法,通道图在闸门层面是一个四点完全图加 30 条孤立边。第五章在独立均匀模型下精确计算:单人盘平均激活 21.5033 个闸门、导通 3.9603 条通道,零导通的概率为 0.00462;两人合盘平均导通 11.1618 条。第六章用经过天象检验的低精度星历在真实天空位置上重算:平均激活 19.759 个闸门,反映者的比例约为 0.0084;88 度等于 93.8667 条爻线,因此只有 12 种人生角色,其中 6 种各占 0.1444、另 6 种各占 0.0222;出生时间晚一小时,月亮的爻线有 0.581 的机会改变。第七章与《德累斯顿抄本》金星表(584、365、260 三个周期在 37960 日上同时闭合)和莱布尼茨的二进制并读。第八章给出巧合的基率、标签的熵与两项预注册研究:盲法辨认需 359 名参与者,神谕关系研究需招募约 8561 对伴侣。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Tzolkin as Two Coprime Gears — Chapter 1 writes the tzolkin with the Chinese remainder theorem: 13 and 20 are coprime, so all 260 combinations occur, whereas the sexagenary cycle, with greatest common divisor 2 of 10 and 12, uses only 60 of 120; the Calendar Round is 18980 days, 52 haab years; 365 leaves 5 modulo 20, so the year bearers can only be one residue class modulo 5, Ik, Manik, Eb and Kaban when 0 Pop is new year. (https://5dlibrary.com/answers/en/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c01/) 第1章 卓尔金历的双模同余齿轮 — 第一章用中国剩余定理写出卓尔金历:13 与 20 互素,260 种组合全部出现,而干支因 10 与 12 的最大公约数为 2 只出现 120 种中的 60 种;历法轮为 18980 日,等于 52 个哈布年;365 模 20 余 5,所以年承载者只能是模 5 的一个剩余类,以 0 Pop 为新年时为 Ik、Manik、Eb、Kaban。 (https://5dlibrary.com/answers/zh/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c01/) - Ch.2 The Long Count, Two Correlation Constants and the Dreamspell Count — Chapter 2 compares the GMT correlation constants 584283 and 584285 and shows that Dreamspell does not advance on 29 February, so between 1950 and 2029 it never gives the same kin as the traditional count. (https://5dlibrary.com/answers/en/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c02/) 第2章 长纪年、两个关联常数与 Dreamspell 计数 — 第二章比较 GMT 关联常数 584283 与 584285,并说明 Dreamspell 在 2 月 29 日不前进,1950 年到 2029 年间与传统计数从未给出相同的 kin。 (https://5dlibrary.com/answers/zh/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c02/) - Ch.3 The Five-Part Oracle: Involutions and a Group Action on Signs and Tones — Chapter 3 proves that analog, antipode and occult are involutions without fixed points, generating a regular action of order 20 on the day signs and the Klein four-group on the colours; the guide table equals the affine map s + 12(t − 1); two random people are linked by the oracle with probability 0.01455. (https://5dlibrary.com/answers/en/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c03/) 第3章 五大神谕:印记与音阶上的对合与群作用 — 第三章证明支持、挑战、隐藏推动是没有不动点的对合,在日符上生成 20 阶的正则作用,在颜色上构成克莱因四元群;指引表等价于仿射映射 s + 12(t − 1);两个随机的人之间存在神谕关系的概率为 0.01455。 (https://5dlibrary.com/answers/zh/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c03/) - Ch.4 The Sixty-Four-Gate Wheel and the Channel-Centre Graph — Chapter 4 proves that the Human Design gate wheel is Shao Yong's Xiantian circle read from hexagram 41, and that at the level of gates the channel graph is one complete graph on four vertices plus 30 isolated edges. (https://5dlibrary.com/answers/en/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c04/) 第4章 六十四闸门轮、通道与中心 — 第四章证明人类图的闸门轮就是邵雍先天圆图从第 41 卦起的读法,通道图在闸门层面是一个四点完全图加 30 条孤立边。 (https://5dlibrary.com/answers/zh/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c04/) - Ch.5 Gate Activation and Channel Percolation in the Independence Model — Chapter 5 computes exactly in a model of independent uniform activations: a single chart activates 21.5033 gates and conducts 3.9603 channels on average, with probability 0.00462 that none conducts; a two-person composite conducts 11.1618. (https://5dlibrary.com/answers/en/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c05/) 第5章 独立模型下的闸门激活与通道渗流 — 第五章在独立均匀模型下精确计算:单人盘平均激活 21.5033 个闸门、导通 3.9603 条通道,零导通的概率为 0.00462;两人合盘平均导通 11.1618 条。 (https://5dlibrary.com/answers/zh/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c05/) - Ch.6 Base Rates in the Real Sky: an Ephemeris, Correlated Activations and Twelve Profiles — Chapter 6 recomputes on real sky positions with a low-precision ephemeris checked against observed events: 19.759 active gates on average and a share of Reflectors of about 0.0084; 88 degrees is 93.8667 lines, so there are only 12 profiles, six with probability 0.1444 each and six with 0.0222; an hour's delay in birth time changes the Moon's line with probability 0.581. (https://5dlibrary.com/answers/en/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c06/) 第6章 真实天空中的基率:星历、相关激活与十二种人生角色 — 第六章用经过天象检验的低精度星历在真实天空位置上重算:平均激活 19.759 个闸门,反映者的比例约为 0.0084;88 度等于 93.8667 条爻线,因此只有 12 种人生角色,其中 6 种各占 0.1444、另 6 种各占 0.0222;出生时间晚一小时,月亮的爻线有 0.581 的机会改变。 (https://5dlibrary.com/answers/zh/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c06/) - Ch.7 The Venus Table of the Dresden Codex and the Xiantian Circle — Chapter 7 reads the Venus table of the Dresden Codex (the periods 584, 365 and 260 closing together in 37960 days) and Leibniz's binary numbers. (https://5dlibrary.com/answers/en/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c07/) 第7章 《德累斯顿抄本》金星表与先天圆图 — 第七章与《德累斯顿抄本》金星表(584、365、260 三个周期在 37960 日上同时闭合)和莱布尼茨的二进制并读。 (https://5dlibrary.com/answers/zh/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c07/) - Ch.8 Base Rates, the Information in Labels and Two Pre-Registered Studies — Chapter 8 gives the base rates of coincidence, the entropy of the labels and two pre-registered studies: blind recognition needing 359 participants and an oracle-tie study needing about 8561 couples. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c08/) 第8章 基率、标签的信息量与两项预注册研究 — 第八章给出巧合的基率、标签的熵与两项预注册研究:盲法辨认需 359 名参与者,神谕关系研究需招募约 8561 对伴侣。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/mayan-tzolkin-modular-gears-oracle-involutions-gate-percolation/c08/) ## 5DL-B-073 · Horary Dignity Scoring, Sidereal Nakshatra Precession and Vimshottari Dasha Chains 中文书名:古典卜卦尊贵度积分、恒星二十七宿岁差与维姆肖塔里运限链 DOI: https://doi.org/10.5281/zenodo.24090195 Cite: Fan, Y. (2026). Horary Dignity Scoring, Sidereal Nakshatra Precession and Vimshottari Dasha Chains (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090195 Book page: https://5dlibrary.com/books/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: Classical horary astrology scores the planets, and Indian astrology turns the Moon's nakshatra into a chain of periods. This book writes these rules as programs and studies only their algebraic, combinatorial and kinematic structure, with an analysis of base rates and the Barnum effect; it neither claims nor tests that any judgement comes true; all data are generated by its programs, with planetary positions from a low-precision ephemeris checked against observed events. Chapter 1 writes Lilly's essential dignities of 1647 (domicile 5, exaltation 4, triplicity 3, term 2, face 1, detriment −5, fall −4, peregrine −5) as a step function of longitude: the Egyptian terms are 2 to 12 degrees wide and total 57, 79, 66, 82 and 76 degrees by planet; for a uniform longitude the planets' expected scores are unequal, by day from −0.586 for Venus to −3.611 for the Moon. Chapter 2 writes mutual reception as a 2-cycle of the dispositor map, with 61/144 receptions expected in the uniform model and at least one in 0.3596 of real charts, and prohibition as a first-passage race, a third planet getting in first while the Moon applies in 0.1283 to 0.1886 of cases. Chapter 3 computes the frequencies of accidental dignities (Saturn retrograde 0.367 of the time, Mercury combust 0.2396) and shows that three scoring conventions agree on "the strongest planet" in only 0.4223 to 0.9688 of charts. Chapter 4 computes precession: 50.29 arcseconds a year, one degree in 71.6 years; in 2026 the Lahiri ayanamsha is about 24.220 degrees, putting 0.8073 of longitudes in different signs in the two zodiacs, and differences of 1.349 to 2.330 degrees between ayanamsha systems change 0.405 to 0.699 of padas. Chapter 5 proves that the navamsha is the pada index modulo 12, the classical rule of starting points for the four elements following from 9 ≡ −3 (mod 12); each minute of birth time moves the dasha balance in a Venus nakshatra by about 5.01 days. Chapter 6 writes out the self-similar nested partition of the 120-year Vimshottari dasha and proves exactly that in a population of fixed age the share in each major period is 0.1111 at birth and then varies with a period of 120 years, differing from the period share T/120 by up to 0.1190: at age 30 Rahu's period holds 0.2056, not 0.1500, and only the average over age equals T/120. Chapter 7 reads Ptolemy's Tetrabiblos, Lilly's Christian Astrology and the Brihat Parashara Hora Shastra, and shows that taking the dasha year as 360 or 365.25 days moves late dasha dates by 486.5 days. Chapter 8 gives the base rate of "something notable" (at least one such mark at a random moment with probability 0.9908) and two pre-registered studies. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:古典卜卦占星给行星打分,印度占星把月亮的宿位换算成一串运限。本书把这些规则写成程序,只研究它们的代数、组合与运动学结构,并附基率与巴纳姆效应分析;不主张、也不检验任何判断是否应验;全部数据由书中程序生成,行星位置取自一个经过天象检验的低精度星历。 第一章把莉莉 1647 年的先天尊贵度(庙 5、旺 4、三分 3、界 2、面 1、陷 −5、落 −4、游走 −5)写成黄经上的阶梯函数:埃及界宽 2 到 12 度,各行星的界合计 57、79、66、82、76 度;在均匀黄经下各行星的期望得分并不相等,白天从金星的 −0.586 到月亮的 −3.611。第二章把互溶写成定位星映射的 2-循环,均匀模型下期望 61/144 对,真实天空中至少一对的概率为 0.3596;把截夺写成首达竞争,月亮入相位期间被第三颗行星抢先的比例为 0.1283 到 0.1886。第三章计算后天尊贵度的频率(土星 0.367 的时间逆行,水星 0.2396 的时间焦伤),并显示三种计分约定下「最强行星」的一致率只有 0.4223 到 0.9688。 第四章计算岁差:每年 50.29 角秒,71.6 年一度;2026 年拉希里岁差约 24.220 度,使 0.8073 的黄经在两种黄道中属于不同星座;岁差体系之间 1.349 到 2.330 度的差别使 0.405 到 0.699 的宿足改变。第五章证明九分盘就是宿足序号模 12,古典的四元素起点规则来自 9 ≡ −3 (mod 12);出生时间每差一分钟,金星宿中的运限余额移动约 5.01 天。第六章写出 120 年维姆肖塔里运限的自相似嵌套切分,并精确证明:固定年龄人群中处于各大运的比例在出生时都是 0.1111,此后以 120 年为周期变化,与周期比例 T/120 的差距可达 0.1190,30 岁时罗睺大运占 0.2056 而非 0.1500,只有对年龄取平均才等于 T/120。第七章与托勒密《四书》、莉莉《基督教占星学》和《帕拉夏拉大典》并读,并说明运限年取 360 天还是 365.25 天会使晚年运限相差 486.5 天。第八章给出「值得注意之处」的基率(随机时刻至少出现一项的概率为 0.9908)与两项预注册研究。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Essential Dignity: a Step Function of Longitude — Chapter 1 writes Lilly's essential dignities of 1647 (domicile 5, exaltation 4, triplicity 3, term 2, face 1, detriment −5, fall −4, peregrine −5) as a step function of longitude: the Egyptian terms are 2 to 12 degrees wide and total 57, 79, 66, 82 and 76 degrees by planet; for a uniform longitude the planets' expected scores are unequal, by day from −0.586 for Venus to −3.611 for the Moon. (https://5dlibrary.com/answers/en/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c01/) 第1章 先天尊贵度:黄经上的阶梯函数 — 第一章把莉莉 1647 年的先天尊贵度(庙 5、旺 4、三分 3、界 2、面 1、陷 −5、落 −4、游走 −5)写成黄经上的阶梯函数:埃及界宽 2 到 12 度,各行星的界合计 57、79、66、82、76 度;在均匀黄经下各行星的期望得分并不相等,白天从金星的 −0.586 到月亮的 −3.611。 (https://5dlibrary.com/answers/zh/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c01/) - Ch.2 Reception, the Dispositor Map and the First-Passage Race of Prohibition — Chapter 2 writes mutual reception as a 2-cycle of the dispositor map, with 61/144 receptions expected in the uniform model and at least one in 0.3596 of real charts, and prohibition as a first-passage race, a third planet getting in first while the Moon applies in 0.1283 to 0.1886 of cases. (https://5dlibrary.com/answers/en/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c02/) 第2章 互溶、定位星映射与截夺的首达竞争 — 第二章把互溶写成定位星映射的 2-循环,均匀模型下期望 61/144 对,真实天空中至少一对的概率为 0.3596;把截夺写成首达竞争,月亮入相位期间被第三颗行星抢先的比例为 0.1283 到 0.1886。 (https://5dlibrary.com/answers/zh/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c02/) - Ch.3 Accidental Dignity and the Convention Budget of "the Strongest Planet" — Chapter 3 computes the frequencies of accidental dignities (Saturn retrograde 0.367 of the time, Mercury combust 0.2396) and shows that three scoring conventions agree on "the strongest planet" in only 0.4223 to 0.9688 of charts. (https://5dlibrary.com/answers/en/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c03/) 第3章 后天尊贵度与「最强行星」的约定预算 — 第三章计算后天尊贵度的频率(土星 0.367 的时间逆行,水星 0.2396 的时间焦伤),并显示三种计分约定下「最强行星」的一致率只有 0.4223 到 0.9688。 (https://5dlibrary.com/answers/zh/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c03/) - Ch.4 Precession Drift and Sign Changes between Tropical and Sidereal Zodiacs — Chapter 4 computes precession: 50.29 arcseconds a year, one degree in 71.6 years; in 2026 the Lahiri ayanamsha is about 24.220 degrees, putting 0.8073 of longitudes in different signs in the two zodiacs, and differences of 1.349 to 2.330 degrees between ayanamsha systems change 0.405 to 0.699 of padas. (https://5dlibrary.com/answers/en/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c04/) 第4章 岁差漂移与回归、恒星黄道的星座翻转 — 第四章计算岁差:每年 50.29 角秒,71.6 年一度;2026 年拉希里岁差约 24.220 度,使 0.8073 的黄经在两种黄道中属于不同星座;岁差体系之间 1.349 到 2.330 度的差别使 0.405 到 0.699 的宿足改变。 (https://5dlibrary.com/answers/zh/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c04/) - Ch.5 Twenty-Seven Nakshatras, 108 Padas and the Congruence Structure of the Navamsha — Chapter 5 proves that the navamsha is the pada index modulo 12, the classical rule of starting points for the four elements following from 9 ≡ −3 (mod 12); each minute of birth time moves the dasha balance in a Venus nakshatra by about 5.01 days. (https://5dlibrary.com/answers/en/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c05/) 第5章 二十七宿、一百零八宿足与九分盘的同余结构 — 第五章证明九分盘就是宿足序号模 12,古典的四元素起点规则来自 9 ≡ −3 (mod 12);出生时间每差一分钟,金星宿中的运限余额移动约 5.01 天。 (https://5dlibrary.com/answers/zh/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c05/) - Ch.6 The Vimshottari Dasha: Nested Partitions and a Cross-Sectional Bias — Chapter 6 writes out the self-similar nested partition of the 120-year Vimshottari dasha and proves exactly that in a population of fixed age the share in each major period is 0.1111 at birth and then varies with a period of 120 years, differing from the period share T/120 by up to 0.1190: at age 30 Rahu's period holds 0.2056, not 0.1500, and only the average over age equals T/120. (https://5dlibrary.com/answers/en/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c06/) 第6章 维姆肖塔里运限:嵌套切分与横截面的先验偏差 — 第六章写出 120 年维姆肖塔里运限的自相似嵌套切分,并精确证明:固定年龄人群中处于各大运的比例在出生时都是 0.1111,此后以 120 年为周期变化,与周期比例 T/120 的差距可达 0.1190,30 岁时罗睺大运占 0.2056 而非 0.1500,只有对年龄取平均才等于 T/120。 (https://5dlibrary.com/answers/zh/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c06/) - Ch.7 The Tetrabiblos, Christian Astrology and the Brihat Parashara Hora Shastra — Chapter 7 reads Ptolemy's Tetrabiblos, Lilly's Christian Astrology and the Brihat Parashara Hora Shastra, and shows that taking the dasha year as 360 or 365.25 days moves late dasha dates by 486.5 days. (https://5dlibrary.com/answers/en/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c07/) 第7章 《四书》、《基督教占星学》与《帕拉夏拉大典》 — 第七章与托勒密《四书》、莉莉《基督教占星学》和《帕拉夏拉大典》并读,并说明运限年取 360 天还是 365.25 天会使晚年运限相差 486.5 天。 (https://5dlibrary.com/answers/zh/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c07/) - Ch.8 The Base Rate of "Something Notable" and Two Pre-Registered Studies — Chapter 8 gives the base rate of "something notable" (at least one such mark at a random moment with probability 0.9908) and two pre-registered studies. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c08/) 第8章 「值得注意之处」的基率与两项预注册研究 — 第八章给出「值得注意之处」的基率(随机时刻至少出现一项的概率为 0.9908)与两项预注册研究。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/horary-dignity-scoring-sidereal-nakshatras-vimshottari-chains/c08/) ## 5DL-B-074 · Diagnostic-to-Pedagogy Conversion Funnels, Mastery-Gated Curriculum Architecture and Certification Network Economics 中文书名:诊学一体转化漏斗、掌握度门控课程与持牌人网络经济学 DOI: https://doi.org/10.5281/zenodo.24090197 Cite: Fan, Y. (2026). Diagnostic-to-Pedagogy Conversion Funnels, Mastery-Gated Curriculum Architecture and Certification Network Economics (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090197 Book page: https://5dlibrary.com/books/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/ Scope: Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. Abstract: A digital reading service that turns paying readers into students, and students into licensed practitioners, has to answer three questions: does the course eat the revenue of retail readings, how should the course be cut along prerequisites, forgetting and dropout, and can an AI examiner with five questions and a pass mark of 16 hold the quality of a practitioner network? This book writes the three as runnable programs. Its prices, valuations and mastery distributions are model parameters, not data from any business; it neither claims nor tests that any reading comes true. Chapter 1 shows that a course of 480 is by itself worth only 1.60 times the retail revenue it ends; whether lifetime value rises several times depends on the licence funnel (4.87 times at exam take-up 0.6 and pass rate 0.50), and the course is bought by heavy readers, who give up a retail stream worth 499.25 against 203.38 for the average reader. Chapter 2 writes 12 subjects as a prerequisite graph with a critical path of 98 hours and 543,510 admissible study orders; exponential and power-law forgetting give almost the same spaced-review schedule. Chapter 3 shows that the five-question examiner tells learners apart poorly near mastery 0.80: 0.4930 of passers in the population are below 0.80 and 0.1953 of the competent fail on the day; with a bank of 40 questions, a weak candidate who learns nothing passes within five attempts with chance 0.627. Chapter 4 finds a flat optimum of about 79 chapters of 2.58 hours, and twelve subject gates cut completion from 0.5307 to 0.1656. Chapters 5 and 6 treat the pass mark as a quality filter: one static year's profit is highest at a mark of 10; with reputation feedback the five-year optimum moves to 12-14; in a two-sided market it is 18, and a mark of 16 earns 0.9918 of the optimum. AI grading noise mainly fails strong candidates and borderline review repairs it; teaching to the test lowers passers' mean mastery from 0.8177 to 0.7176, and three transfer questions bring it back to 0.8150. Chapter 7 reads the Analects on opening the way only for the eager and on one corner and three, and the Hanshu on examinations encouraged by official salaries. Chapter 8 gives a five-flag audit and a pre-registered experiment (234 per arm for the primary outcome). Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. 摘要:一家数字测算服务把付费读者转化为学员、再把学员转化为持牌执业者,要回答三个问题:课程会不会吃掉零售测算的收入,课程该如何按先修关系、遗忘规律与流失风险切分,五道题、16 分及格的 AI 考官又能否为执业者网络把住质量。本书把三者写成可运行的程序。书中的价格、估值与掌握度分布都是模型参数,不是任何企业的数据;本书不主张、也不检验任何占断是否应验。 第一章证明,480 的课程本身只抵得上它所终止的零售收入的 1.60 倍;终身价值能否提高数倍取决于牌照漏斗(参考率 0.6、通过率 0.50 时为 4.87 倍),而买课程的恰是重度读者,他们放弃的零售现值是 499.25,平均读者只有 203.38。第二章把 12 个科目写成先修图,关键路径 98 学时,合法学习顺序 543,510 种;指数与幂律两种遗忘律给出几乎相同的间隔复习计划。第三章显示五题考官在掌握度 0.80 附近分辨力很弱:人群中 0.4930 的及格者低于 0.80,0.1953 的胜任者当天不及格;题库只有 40 题时,不学习的弱考生补考五次有 0.627 的机会通过。第四章求出约 79 章、每章 2.58 小时的平坦最优粒度,12 道科目门槛把完课率从 0.5307 降到 0.1656。 第五、六章把及格线当作质量过滤器:单年静态利润在及格线 10 时最高;加入声誉反馈后五年最优移到 12 至 14;加入双边市场后最优为 18,而 16 分的利润是最优的 0.9918。AI 阅卷噪声主要拒掉强考生,边界复核可以修补;应试教学使及格者的平均掌握度从 0.8177 降到 0.7176,三道迁移题使之回到 0.8150。第七章与《论语·述而》「不愤不启,不悱不发,举一隅不以三隅反」及《汉书·儒林传》「设科射策,劝以官禄」并读。第八章给出五项自动审计与一项预注册实验(主要结局每组 234 人)。书中以标注框区分证据、16 段可运行的 Python 程序、原典与假说。 Chapter answers: - Ch.1 The Self-Cannibalization Bound: When Does a Course Pay? — Chapter 1 shows that a course of 480 is by itself worth only 1.60 times the retail revenue it ends; whether lifetime value rises several times depends on the licence funnel (4.87 times at exam take-up 0.6 and pass rate 0.50), and the course is bought by heavy readers, who give up a retail stream worth 499.25 against 203.38 for the average reader. (https://5dlibrary.com/answers/en/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c01/) 第1章 自蚕食边界:课程何时多赚 — 第一章证明,480 的课程本身只抵得上它所终止的零售收入的 1.60 倍;终身价值能否提高数倍取决于牌照漏斗(参考率 0.6、通过率 0.50 时为 4.87 倍),而买课程的恰是重度读者,他们放弃的零售现值是 499.25,平均读者只有 203.38。 (https://5dlibrary.com/answers/zh/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c01/) - Ch.2 Prerequisite Graphs and Forgetting Curves: The Skeleton of a Course — Chapter 2 writes 12 subjects as a prerequisite graph with a critical path of 98 hours and 543,510 admissible study orders; exponential and power-law forgetting give almost the same spaced-review schedule. (https://5dlibrary.com/answers/en/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c02/) 第2章 先修图与遗忘曲线:课程的骨架 — 第二章把 12 个科目写成先修图,关键路径 98 学时,合法学习顺序 543,510 种;指数与幂律两种遗忘律给出几乎相同的间隔复习计划。 (https://5dlibrary.com/answers/zh/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c02/) - Ch.3 A Five-Question AI Examiner: What Can It Tell Apart? — Chapter 3 shows that the five-question examiner tells learners apart poorly near mastery 0.80: 0.4930 of passers in the population are below 0.80 and 0.1953 of the competent fail on the day; with a bank of 40 questions, a weak candidate who learns nothing passes within five attempts with chance 0.627. (https://5dlibrary.com/answers/en/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c03/) 第3章 五道题的 AI 考官:它能分辨什么 — 第三章显示五题考官在掌握度 0.80 附近分辨力很弱:人群中 0.4930 的及格者低于 0.80,0.1953 的胜任者当天不及格;题库只有 40 题时,不学习的弱考生补考五次有 0.627 的机会通过。 (https://5dlibrary.com/answers/zh/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c03/) - Ch.4 Chapter Granularity, Dropout and Twelve Gates — Chapter 4 finds a flat optimum of about 79 chapters of 2.58 hours, and twelve subject gates cut completion from 0.5307 to 0.1656. (https://5dlibrary.com/answers/en/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c04/) 第4章 章节粒度、流失风险与十二道门槛 — 第四章求出约 79 章、每章 2.58 小时的平坦最优粒度,12 道科目门槛把完课率从 0.5307 降到 0.1656。 (https://5dlibrary.com/answers/zh/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c04/) - Ch.5 The Pass Mark as a Quality Filter: Static Optimum and Reputation — Chapters 5 and 6 treat the pass mark as a quality filter: one static year's profit is highest at a mark of 10; with reputation feedback the five-year optimum moves to 12-14; in a two-sided market it is 18, and a mark of 16 earns 0.9918 of the optimum. AI grading noise mainly fails strong candidates and borderline review repairs it; teaching to the test lowers passers' mean mastery from 0.8177 to 0.7176, and three transfer questions bring it back to 0.8150. (https://5dlibrary.com/answers/en/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c05/) 第5章 及格线是质量过滤器:静态最优与声誉动态 — 第五、六章把及格线当作质量过滤器:单年静态利润在及格线 10 时最高;加入声誉反馈后五年最优移到 12 至 14;加入双边市场后最优为 18,而 16 分的利润是最优的 0.9918。AI 阅卷噪声主要拒掉强考生,边界复核可以修补;应试教学使及格者的平均掌握度从 0.8177 降到 0.7176,三道迁移题使之回到 0.8150。 (https://5dlibrary.com/answers/zh/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c05/) - Ch.6 Two-Sided Networks, Grading Noise and Teaching to the Test — Once the client side is treated as a finite pool, the best mark in the two-sided equilibrium is 18, and 16 is less than one per cent short. AI grading noise mainly hurts strong candidates, and borderline review repairs it. Teaching to the test empties the pass mark of its filtering power; unseen transfer questions bring it back. (https://5dlibrary.com/answers/en/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c06/) 第6章 双边网络、阅卷噪声与应试教学 — 把客户一侧当作有限的池子之后,双边均衡的最优及格线是 18,16 分只少不到百分之一。AI 阅卷噪声主要伤害强考生,边界复核可以修补。应试教学会让及格线失去过滤作用,题库外的迁移题可以把它找回来。 (https://5dlibrary.com/answers/zh/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c06/) - Ch.7 "No Opening for One Not Eager", Graded Examinations and Guild Heads — "No opening for one not eager, no prompting for one not struggling" is a readiness gate; "one corner and three" is a transfer gate: three unseen corners cut a guesser's pass chance from 0.25 to 0.0156, while one who has mastered the principle still passes 0.729 of the time. The Hanshu records how an examination system encouraged by official salaries, in little more than a century, raised the great masters to more than a thousand and the exposition of one classic to more than a million words. The guild heads serve only as a structural analogy, without quotation. (https://5dlibrary.com/answers/en/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c07/) 第7章 「不愤不启」、设科射策与行老 — 「不愤不启,不悱不发」是准备度门槛,「举一隅不以三隅反」是迁移门槛:三个没见过的角把猜测者的通过率从 0.25 降到 0.0156,而真正掌握者仍有 0.729。《汉书·儒林传》记录了以官禄劝勉的考核制度如何在百余年间使大师增至千余人、使一经的解说增至百余万言。行老只作为结构类比,不引原文。 (https://5dlibrary.com/answers/zh/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c07/) - Ch.8 An Audit of Course and Licence, and a Pre-Registered Experiment — Chapter 8 gives a five-flag audit and a pre-registered experiment (234 per arm for the primary outcome). Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. (https://5dlibrary.com/answers/en/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c08/) 第8章 课程与牌照的审计,以及一项预注册实验 — 第八章给出五项自动审计与一项预注册实验(主要结局每组 234 人)。书中以标注框区分证据、16 段可运行的 Python 程序、原典与假说。 (https://5dlibrary.com/answers/zh/diagnostic-to-pedagogy-funnels-mastery-gating-certification-economics/c08/) ## 5DL-B-075 · Zero-Touch Multi-Site Fault Tolerance, Payment Anti-Fraud State Machines and Passive Cashflow Queuing 中文书名:零人工多站点容错架构、支付防欺诈状态机与被动现金流排队论 DOI: https://doi.org/10.5281/zenodo.24090201 Cite: Fan, Y. (2026). Zero-Touch Multi-Site Fault Tolerance, Payment Anti-Fraud State Machines and Passive Cashflow Queuing (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090201 Book page: https://5dlibrary.com/books/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/ Scope: Marked boxes separate evidence, 16 runnable Python programs, historical sources and hypotheses. Abstract: A digital service network of dozens of sites that share one program kernel and try to run without human work has to answer four questions: how to stop manual transfer receipts from being reused or forged, how self-referral commissions and "invite friends to unlock" offers are arbitraged, how the order state machine and message retries stay correct, and how much headroom the shared kernel and the cash flow need under bursts and shocks. This book writes these questions as runnable programs. Its traffic, price, fraud and failure parameters are model parameters, not data from any business; it gives no method of attack, and it neither claims nor tests that any reading comes true. Chapter 1 shows that checking only 4 trailing digits gives an even chance of a collision by 119 orders and, after 5,000 orders, wrongly refuses 0.3935 of honest receipts; a perceptual hash recognises re-saved copies but matches honest receipts from the same app; release-then-audit deters only when the person caught loses a stake, and with a stake of 200 auditing 0.485 of orders suffices. Chapter 2's association graph shows that devices and payment accounts identify most cheating rings, while IP subnets join 4692 accounts into one component and wrongly block 0.817 of referrals between strangers; "invite 3 to unlock" lets a self-referrer get a level listed at 299 for 67.50. Chapter 3's exhaustive model check shows that handlers blind to state double-release with two events, and that idempotency keys must outlive the 511-minute retry schedule and be written atomically. Chapter 4 shows by queueing theory and simulation that bursty arrivals push a single worker's T99 to tens of seconds; at a mean load of 0.80, bursts need 4 workers, or 3 with load shedding (refusing 0.0005 of requests). Chapter 5 shows that a standby node and canary releases cut the time all 30 sites are down together to 6.53 minutes a year, and that the drain time of a blue-green switch is set by the longest streamed requests. Chapter 6 allocates traffic by Thompson sampling (470.0 conversions short of the ideal), gives a posterior stopping rule for promoting a site to its own domain, and shows that a three-month halving of orders needs a cash reserve of 90,000. Chapter 7 reads the matched tallies of Tang flying cash, the multiple seals and issue cap of Song jiaozi, and Liu Yan's clearing of the Bian River. Chapter 8 gives a nine-flag audit and a pre-registered experiment. Marked boxes separate evidence, 16 runnable Python programs, historical sources and hypotheses. 摘要:一个由几十个站点组成、共用一套程序内核、尽量不需要人工操作的数字服务网络,要回答四个问题:人工转账回单如何防止重复与伪造,自购返佣与「邀请好友解锁」如何被套利,订单状态机与消息重试如何保持正确,共享内核与现金流在突发冲击下需要多少余量。本书把这些问题写成可运行的程序。书中的流量、价格、欺诈率与故障参数都是模型参数,不是任何企业的数据;本书不提供任何攻击方法,也不主张、不检验任何占断是否应验。 第一章证明只核对 4 位尾号在 119 单时就有一半的碰撞机会,累计 5,000 单后会误拒 0.3935 的诚实回单;感知哈希能认出重新保存的副本,却会把同一 App 的诚实回单误配;「先放行、后抽查」只有在被抓者失去抵押时才有威慑力,抵押为 200 时抽查 0.485 的订单即可。第二章的关联图显示,设备与支付账户能识别多数作弊团伙,加入 IP 子网则把 4692 个账户连成一片、误拦 0.817 的陌生人推荐;「邀请 3 人解锁」让自购者以 67.50 拿到标价 299 的二级。第三章的穷举模型检验显示,不看状态的处理函数两个事件就能造成重复放行,幂等键必须比 511 分钟的重试计划活得更久且原子写入。 第四章用排队论与模拟显示,突发到达把单个工作进程的 T99 推到数十秒;平均负载 0.80 时,突发需要 4 个工作进程,或 3 个加限流(拒绝 0.0005 的请求)。第五章显示主备节点与金丝雀发布把 30 个站点一起停机的时间降到每年 6.53 分钟,蓝绿切换的排空时间由最长的流式请求决定。第六章用汤普森抽样分配流量(与理想只差 470.0 次转化),给出升级独立域名的后验停止规则,并显示订单减半三个月的冲击需要 90,000 的现金储备。第七章与唐代飞钱合券、宋代交子多重印押与发行上限、刘晏疏浚汴河并读。第八章给出九项自动审计与一项预注册实验。书中以标注框区分证据、16 段可运行的 Python 程序、史料与假说。 Chapter answers: - Ch.1 Transfer Receipts: Deduplication, Image Hashes and Release-Then-Audit — Chapter 1 shows that checking only 4 trailing digits gives an even chance of a collision by 119 orders and, after 5,000 orders, wrongly refuses 0.3935 of honest receipts; a perceptual hash recognises re-saved copies but matches honest receipts from the same app; release-then-audit deters only when the person caught loses a stake, and with a stake of 200 auditing 0.485 of orders suffices. (https://5dlibrary.com/answers/en/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c01/) 第1章 转账回单:去重、图像哈希与先放行后抽查 — 第一章证明只核对 4 位尾号在 119 单时就有一半的碰撞机会,累计 5,000 单后会误拒 0.3935 的诚实回单;感知哈希能认出重新保存的副本,却会把同一 App 的诚实回单误配;「先放行、后抽查」只有在被抓者失去抵押时才有威慑力,抵押为 200 时抽查 0.485 的订单即可。 (https://5dlibrary.com/answers/zh/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c01/) - Ch.2 Self-Referral and "Invite Friends to Unlock" — Chapter 2's association graph shows that devices and payment accounts identify most cheating rings, while IP subnets join 4692 accounts into one component and wrongly block 0.817 of referrals between strangers; "invite 3 to unlock" lets a self-referrer get a level listed at 299 for 67.50. (https://5dlibrary.com/answers/en/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c02/) 第2章 自购返佣与「邀请好友解锁」 — 第二章的关联图显示,设备与支付账户能识别多数作弊团伙,加入 IP 子网则把 4692 个账户连成一片、误拦 0.817 的陌生人推荐;「邀请 3 人解锁」让自购者以 67.50 拿到标价 299 的二级。 (https://5dlibrary.com/answers/zh/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c02/) - Ch.3 The Order State Machine: Exhaustive Checking and the Idempotency Window — Chapter 3's exhaustive model check shows that handlers blind to state double-release with two events, and that idempotency keys must outlive the 511-minute retry schedule and be written atomically. (https://5dlibrary.com/answers/en/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c03/) 第3章 订单状态机:穷举检验与幂等窗口 — 第三章的穷举模型检验显示,不看状态的处理函数两个事件就能造成重复放行,幂等键必须比 511 分钟的重试计划活得更久且原子写入。 (https://5dlibrary.com/answers/zh/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c03/) - Ch.4 Thirty Sites on One Kernel: Queues and Tail Latency — Chapter 4 shows by queueing theory and simulation that bursty arrivals push a single worker's T99 to tens of seconds; at a mean load of 0.80, bursts need 4 workers, or 3 with load shedding (refusing 0.0005 of requests). (https://5dlibrary.com/answers/en/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c04/) 第4章 三十个站点共用一个内核:排队与尾延迟 — 第四章用排队论与模拟显示,突发到达把单个工作进程的 T99 推到数十秒;平均负载 0.80 时,突发需要 4 个工作进程,或 3 个加限流(拒绝 0.0005 的请求)。 (https://5dlibrary.com/answers/zh/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c04/) - Ch.5 Availability: Shared Kernels, Standby Nodes and Blue-Green Draining — Chapter 5 shows that a standby node and canary releases cut the time all 30 sites are down together to 6.53 minutes a year, and that the drain time of a blue-green switch is set by the longest streamed requests. (https://5dlibrary.com/answers/en/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c05/) 第5章 可用性:共享内核、主备切换与蓝绿排空 — 第五章显示主备节点与金丝雀发布把 30 个站点一起停机的时间降到每年 6.53 分钟,蓝绿切换的排空时间由最长的流式请求决定。 (https://5dlibrary.com/answers/zh/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c05/) - Ch.6 Allocating Traffic, Promoting Domains and Holding Cash — Chapter 6 allocates traffic by Thompson sampling (470.0 conversions short of the ideal), gives a posterior stopping rule for promoting a site to its own domain, and shows that a three-month halving of orders needs a cash reserve of 90,000. (https://5dlibrary.com/answers/en/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c06/) 第6章 流量分配、升级独立域名与现金储备 — 第六章用汤普森抽样分配流量(与理想只差 470.0 次转化),给出升级独立域名的后验停止规则,并显示订单减半三个月的冲击需要 90,000 的现金储备。 (https://5dlibrary.com/answers/zh/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c06/) - Ch.7 Matched Tallies, Banknote Seals and Liu Yan's Canal — Matching the tallies of flying cash is a check of the whole voucher: 4 marks of 10 forms leave a blind forger one chance in ten thousand. The multiple seals of jiaozi multiply independent checks, and the issue cap and two-year term bound the vouchers outstanding. After Liu Yan cleared the Bian River the food supply of Chang'an became plentiful; Little's law shows that shortening time in transit reduces the goods on the road. Everything quoted in this chapter is the wording of English Wikipedia articles; no classical text is quoted. (https://5dlibrary.com/answers/en/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c07/) 第7章 飞钱合券、交子印押与刘晏漕运 — 飞钱的合券是一种完整凭证的核对:4 处、每处 10 种记号就让盲目伪造只剩万分之一的机会。交子的多重印押让独立的检查相乘,发行上限与两年兑换期限制了在途的凭证总量。刘晏疏浚汴河之后长安粮食充足;利特尔定律说明,缩短在途时间就减少了压在路上的物资。本章所引均为英文维基百科条目的文字,未引古籍原文。 (https://5dlibrary.com/answers/zh/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c07/) - Ch.8 A Nine-Flag Audit and a Pre-Registered Experiment — Chapter 8 gives a nine-flag audit and a pre-registered experiment. Marked boxes separate evidence, 16 runnable Python programs, historical sources and hypotheses. (https://5dlibrary.com/answers/en/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c08/) 第8章 九项自动审计与一项预注册实验 — 第八章给出九项自动审计与一项预注册实验。书中以标注框区分证据、16 段可运行的 Python 程序、史料与假说。 (https://5dlibrary.com/answers/zh/zero-touch-multi-site-fault-tolerance-payment-antifraud-queuing/c08/) ## 5DL-B-076 · Spiritual Quotient Quantification, Prior-Conditioned Generation and Retrieval-Augmented Cognitive Agents 中文书名:五维灵商量化架构、先验条件化生成与检索增强认知体 DOI: https://doi.org/10.5281/zenodo.24090203 Cite: Fan, Y. (2026). Spiritual Quotient Quantification, Prior-Conditioned Generation and Retrieval-Augmented Cognitive Agents (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090203 Book page: https://5dlibrary.com/books/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/ Scope: Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. Abstract: This book analyses all 82 volumes of 5D Library. A system that hands a language model both a personal profile and a whole library has to answer three questions: does a five-dimensional spiritual-quotient scale give more information than one total; how much uncertainty can a profile of about 8 KB remove from answers, and how should the context window be divided; and are the volumes of the library joined into one, and how is synthesis across them possible? This book writes the three as runnable programs. Its loadings, correlations, reliabilities and retrieval parameters are model parameters, not data from any assessment or user; the library graph comes from the volumes' public English keywords. It neither claims nor tests that any reading comes true. Chapters 1 and 2 show that whether five dimensions are "orthogonal" depends on their correlation: at 0.9, only 1 eigenvalue of 20 items exceeds 1. Five scores are never worse than a total: equal to it when the outcome depends on all five dimensions (0.3968), while a total explains only 0.0794 when the outcome depends on one; every gain is under half a bit. Four items per dimension give an alpha of 0.7935, and the reliability of a profile's shape falls to 0.2776 at a dimension correlation of 0.9; a blind-pick test needs at most a few hundred users to tell reports from generic text. Chapters 3 and 4 compute exactly how much conditional entropy a profile removes: a profile of 243 states carries 7.9248 bits but removes only 0.8519 bits of uncertainty from answers, and 0 when answers ignore it; with a quarter of the profile wrong only 0.5350 of the gain is left. The Lagrangian water-filling of the context shows the profile at its 8 KB cap from 32k on, with the rest going to retrieval and history; a good retriever needs only 7 passages. Chapter 5 shows that synthesis questions spanning three volumes succeed only 0.060 of the time with direct retrieval in the top 40, while splitting the question first makes success depend on the accuracy of the split. Chapter 6 builds a graph from the keywords of all 82 volumes: once volumes 3, 9, 24 and 31, the islands of the 73-volume edition, carry keywords their own text supports and share topics with volumes 080 to 082, the whole library is connected with no islands and a Fiedler value of 0.5475, a small world (clustering 0.591 against 0.208 at random); without "pre-registration" volumes 3 and 9 hang on 3 edges each. Chapter 7 reads the Zhouyi Xici on looking up at the patterns of heaven and down at the order of earth, and Bol's study of Shao Yong's method of observing things. Chapter 8 gives a seven-flag audit and a pre-registered blind comparison with a stranger's profile as control (1,047 users to detect a preference of 0.55). Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. 摘要:本书的书库分析覆盖 5D Library 全库 82 卷(第 001 卷至第 082 卷)。一个把个人档案与整座书库同时交给语言模型的系统,要回答三个问题:五个维度的灵商量表是否比一个总分提供更多信息,一份约 8 KB 的个人档案最多能让回答少掉多少不确定性、上下文窗口又该如何分配,以及书库各卷是否连成一体、跨书综合如何可能。本书把三者写成可运行的程序。书中的载荷、相关、信度与检索参数都是模型参数,不是任何测评或用户的数据;书库关联图来自各卷公开的英文关键词。本书不主张、也不检验任何占断是否应验。 第一、二章显示,五维是否「正交」取决于维度相关:相关为 0.9 时 20 道题只剩 1 个大于 1 的特征值。五个分数相对于总分「从不更差」:结果依赖全部五维时与总分相同(0.3968),依赖单一维度时总分只解释 0.0794;全部增益都不到半个比特。每维 4 题的 α 为 0.7935,档案「形状」的信度在维度相关为 0.9 时只有 0.2776;盲选测验最多只需几百人即可区分报告与泛泛之词。第三、四章精确计算档案带来的条件熵降低:243 种状态的档案有 7.9248 比特,却只能让回答少 0.8519 比特的不确定性,回答不依赖档案时为 0;档案四分之一出错时收益只剩 0.5350。上下文的拉格朗日注水解显示,32k 起档案停在 8 KB 上限,其余归检索与历史;好的检索器只需 7 段。 第五章显示跨三卷的综合问题直接检索在前 40 段中只有 0.060 的成功率,先拆分问题后成功率由拆分准确率决定。第六章用全库 82 卷的关键词建图:73 卷版本中的孤岛第 3、9、24、31 卷补上正文支持的关键词,并与第 080 卷至第 082 卷共享主题后,全书库连成一体、没有孤岛,菲德勒值 0.5475,是聚类系数远高于随机图(0.591 对 0.208)的小世界;去掉「预注册」后第 3、9 卷各只靠 3 条边相连。第七章与《周易·系辞上》「仰以观于天文,俯以察于地理」及 Bol 对邵雍观物之法的研究并读。第八章给出七项自动审计与一项以「换入陌生人档案」为对照的预注册盲比实验(检出 0.55 的偏好需 1,047 人)。书中以标注框区分证据、16 段可运行的 Python 程序、原典与假说。 Chapter answers: - Ch.1 Five Dimensions or One: The Spectrum and Information of an SQ Scale — Chapters 1 and 2 show that whether five dimensions are "orthogonal" depends on their correlation: at 0.9, only 1 eigenvalue of 20 items exceeds 1. Five scores are never worse than a total: equal to it when the outcome depends on all five dimensions (0.3968), while a total explains only 0.0794 when the outcome depends on one; every gain is under half a bit. Four items per dimension give an alpha of 0.7935, and the reliability of a profile's shape falls to 0.2776 at a dimension correlation of 0.9; a blind-pick test needs at most a few hundred users to tell reports from generic text. (https://5dlibrary.com/answers/en/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c01/) 第1章 五维还是一维:灵商量表的特征值谱与信息增益 — 第一、二章显示,五维是否「正交」取决于维度相关:相关为 0.9 时 20 道题只剩 1 个大于 1 的特征值。五个分数相对于总分「从不更差」:结果依赖全部五维时与总分相同(0.3968),依赖单一维度时总分只解释 0.0794;全部增益都不到半个比特。每维 4 题的 α 为 0.7935,档案「形状」的信度在维度相关为 0.9 时只有 0.2776;盲选测验最多只需几百人即可区分报告与泛泛之词。 (https://5dlibrary.com/answers/zh/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c01/) - Ch.2 Reliability, the "Shape" of a Profile and a Blind-Pick Test — Four items per dimension give an $\alpha$ near 0.79; 0.90 needs 10. A profile's shape is less reliable than its scores: at a dimension correlation of 0.9 the difference of two dimensions has reliability 0.2776. Users' ratings of accuracy cannot show validity; a blind-pick test can, with a few hundred users at most. (https://5dlibrary.com/answers/en/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c02/) 第2章 信度、档案的「形状」与盲选测验 — 每维 4 题的 $\alpha$ 约为 0.79,达到 0.90 需要 10 题。档案的形状比单个分数更不可靠,维度相关为 0.9 时两维之差的信度只有 0.2776。用户评价「准不准」不能证明效度;盲选测验可以,最多只需几百人。 (https://5dlibrary.com/answers/zh/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c02/) - Ch.3 Prior Conditioning: How Much Uncertainty Can a Profile Remove? — Chapters 3 and 4 compute exactly how much conditional entropy a profile removes: a profile of 243 states carries 7.9248 bits but removes only 0.8519 bits of uncertainty from answers, and 0 when answers ignore it; with a quarter of the profile wrong only 0.5350 of the gain is left. The Lagrangian water-filling of the context shows the profile at its 8 KB cap from 32k on, with the rest going to retrieval and history; a good retriever needs only 7 passages. (https://5dlibrary.com/answers/en/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c03/) 第3章 先验条件化:一份档案能让回答少多少不确定 — 第三、四章精确计算档案带来的条件熵降低:243 种状态的档案有 7.9248 比特,却只能让回答少 0.8519 比特的不确定性,回答不依赖档案时为 0;档案四分之一出错时收益只剩 0.5350。上下文的拉格朗日注水解显示,32k 起档案停在 8 KB 上限,其余归检索与历史;好的检索器只需 7 段。 (https://5dlibrary.com/answers/zh/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c03/) - Ch.4 The Context Budget: Lagrangian Water-Filling and the Best Number of Passages — Under diminishing returns, the optimal division of the context is a water-filling solution: from 32k the profile stays at its 8 KB cap, and the rest goes to retrieval and history by weight. The best number of passages is set jointly by retrieval quality and distraction: a good retriever needs only 7, and a poor one cannot catch up however many it takes. (https://5dlibrary.com/answers/en/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c04/) 第4章 上下文预算:拉格朗日注水与最优检索段数 — 在边际递减的收益下,上下文的最优分配是一个注水解:档案在 32k 起就停在 8 KB 的上限,其余按权重分给检索与历史。最优检索段数由检索质量与注意力分散共同决定:好的检索器取 7 段即可,差的检索器取再多也追不上。 (https://5dlibrary.com/answers/zh/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c04/) - Ch.5 Retrieval Augmentation: One-Book Recall and Cross-Book Synthesis — Chapter 5 shows that synthesis questions spanning three volumes succeed only 0.060 of the time with direct retrieval in the top 40, while splitting the question first makes success depend on the accuracy of the split. (https://5dlibrary.com/answers/en/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c05/) 第5章 检索增强:单书召回与跨书综合 — 第五章显示跨三卷的综合问题直接检索在前 40 段中只有 0.060 的成功率,先拆分问题后成功率由拆分准确率决定。 (https://5dlibrary.com/answers/zh/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c05/) - Ch.6 The Library Graph: Islands, the Fiedler Value and a Small World — Chapter 6 builds a graph from the keywords of all 82 volumes: once volumes 3, 9, 24 and 31, the islands of the 73-volume edition, carry keywords their own text supports and share topics with volumes 080 to 082, the whole library is connected with no islands and a Fiedler value of 0.5475, a small world (clustering 0.591 against 0.208 at random); without "pre-registration" volumes 3 and 9 hang on 3 edges each. (https://5dlibrary.com/answers/en/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c06/) 第6章 书库关联图:孤岛、菲德勒值与小世界 — 第六章用全库 82 卷的关键词建图:73 卷版本中的孤岛第 3、9、24、31 卷补上正文支持的关键词,并与第 080 卷至第 082 卷共享主题后,全书库连成一体、没有孤岛,菲德勒值 0.5475,是聚类系数远高于随机图(0.591 对 0.208)的小世界;去掉「预注册」后第 3、9 卷各只靠 3 条边相连。 (https://5dlibrary.com/answers/zh/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c06/) - Ch.7 Looking Up and Looking Down, and Shao Yong's Observing of Things — The Xici's "looking up and looking down" sets out two independent directions of observation; Shao Yong divides any topic into four aspects and counts the generations of a cycle as 1x12x30x12 = 4320. Four fourfold divisions give 256 classes, 8 bits, almost the 7.9248 bits of a five-by-three profile. Classifying is easy; showing that a classification predicts anything is the hard part. (https://5dlibrary.com/answers/en/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c07/) 第7章 仰观俯察与邵雍观物 — 《系辞》的「仰观俯察」提出了两个独立的观察方向;邵雍把任何题目分为四个方面,并以 1×12×30×12 = 4320 推算一元之世。四次四分得 256 类、8 比特,与五维三档档案的 7.9248 比特几乎相同。分类容易,证明分类能预测什么才是难处。 (https://5dlibrary.com/answers/zh/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c07/) - Ch.8 A Seven-Flag Audit and a Blind Comparison — Chapter 8 gives a seven-flag audit and a pre-registered blind comparison with a stranger's profile as control (1,047 users to detect a preference of 0.55). Marked boxes separate evidence, 16 runnable Python programs, classical sources and hypotheses. (https://5dlibrary.com/answers/en/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c08/) 第8章 七项审计与一项盲比实验 — 第八章给出七项自动审计与一项以「换入陌生人档案」为对照的预注册盲比实验(检出 0.55 的偏好需 1,047 人)。书中以标注框区分证据、16 段可运行的 Python 程序、原典与假说。 (https://5dlibrary.com/answers/zh/spiritual-quotient-quantification-prior-conditioned-rag-cognitive-agents/c08/) ## 5DL-B-077 · Persuasive Copywriting Algebra, Loss-Framing Asymmetry and Checkout Micro-Architecture 中文书名:说服文案代数、损失框架不对称性与结账微界面工程 DOI: https://doi.org/10.5281/zenodo.24090205 Cite: Fan, Y. (2026). Persuasive Copywriting Algebra, Loss-Framing Asymmetry and Checkout Micro-Architecture (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090205 Book page: https://5dlibrary.com/books/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/ Scope: Conversion rates, refund rates and costs not attributed to a source are illustrative; all data are generated by its programs; the book is not legal advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: A sales page for a paid report breaks into questions that can be computed: how differently is the same thing valued when described as a gain or as a loss? Should the price read 49 or 50? Why are precise numbers more credible? How many more buyers stay with one checkout field fewer? This book writes these rules of thumb as runnable arithmetic and then asks when they hold and when they fail. Conversion rates, refund rates and costs not attributed to a source are illustrative; all data are generated by its programs; the book is not legal advice. Chapter 1 computes with cumulative prospect theory: the subjective value of a sure loss to a sure gain of the same size is 2.25; "avoid a one-in-ten chance of losing RM 1,000" feels like a sure gain of RM 335.85. Turned into conversion, even inflating all valuations 2.25-fold raises conversion at RM 49 by only 100.6%; meta-analyses put framing effects near zero, so plans should use multipliers of 1.05 to 1.10, worth +5.6% to +11.1%. Chapter 2 computes the left-digit effect with an inattention parameter of 0.3: RM 49 is perceived as 46.3, a perceived gap of 7.4% from RM 50; a Bayesian model shows that the credibility of precise numbers comes entirely from checkability (posterior 0.500 without checking, 0.667 with a checking probability of 0.50). Chapter 3 shows that the Hick-Hyman law is logarithmic and that checkout loss is a product of survival rates: each field removed multiplies completion by 1.0204, and going from three pages with 12 fields to one modal with 3 fields raises completion by 56.8%. Chapter 4 measures the fluency of copy and computes A/B sample sizes (13,914 visitors per arm for 3.0% against 3.6%) and the cost of peeking (false-positive rate 0.256 with 20 looks). Chapter 5 puts the fear frame on a full ledger: copy implying "loss of luck" nets 16.8% less after refunds, chargebacks and complaints, and a year of repeat purchases only brings it close (RM 6,364 against RM 6,629). Chapter 6 decomposes the funnel, with the largest room in "checkout opened" (a ceiling of 6.67 times), and the left-digit effect moves the best price from RM 41 to RM 39. Chapter 7 reads Han Feizi's Shuonan, the Wenxin diaolong and the Strategists, writing "knowing the listener's heart" as a value of information (+34.5%). Chapter 8 gives the winner's curse (only 0.624 of significant winners are real), the Barnum effect on free report pages and a pre-registered experiment with a refund guardrail (about 30,000 visitors per arm). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:一张付费报告的销售页可以拆成几个可以计算的问题:同一件事说成「得到」还是「失去」,估值差多少?价格写 49 还是 50?精确数字为什么更可信?结账少一个输入框能多留住多少人?本书把这些经验法则写成可以运行的算术,再看它们在什么条件下成立、在什么条件下失效。书中的转化率、退款率与成本参数凡未注明出处的均为示意;全部数据由书中程序生成;本书不是法律意见。 第一章用累积前景理论计算:同额确定损失与确定收益的主观值之比为 2.25;「避免一成机会损失 RM 1,000」在主观上相当于确定得到 RM 335.85。换算成转化率,即使估值整体放大 2.25 倍,RM 49 的转化率也只提高 100.6%;框架效应的元分析接近于零,规划时应取 1.05 到 1.10 倍,对应 +5.6% 到 +11.1%。第二章以注意力不足参数 0.3 计算左位数字效应:RM 49 被感知为 46.3,与 RM 50 的感知差距为 7.4%;并用贝叶斯模型证明精确数字的可信度完全来自可核查性(没有核查时后验为 0.500,核查概率 0.50 时为 0.667)。第三章说明希克-海曼定律是对数关系,结账损失是各环节存活率的乘积:每去掉一个字段完成率乘以 1.0204,从三页 12 个字段压到一个弹窗 3 个字段,完成率提高 56.8%。 第四章度量文案流畅度,并计算 A/B 测试的样本量(3.0% 对 3.6% 每组 13,914 名访客)与中途偷看的代价(看 20 次时假阳性率 0.256)。第五章把恐惧框架放进总账:暗示「失去好运」的写法计入退款、拒付与投诉后净收入低 16.8%,一年复购也只是相近(RM 6,364 对 RM 6,629)。第六章分解漏斗,最大的空间在「打开结账」(上限 6.67 倍),最优价格在左位数字效应下由 RM 41 移到 RM 39。第七章与《韩非子·说难》《文心雕龙》和纵横家并读,把「知所说之心」写成信息价值(+34.5%)。第八章给出赢家诅咒(显著胜出中只有 0.624 为真)、免费报告页的巴纳姆效应与一项以退款率为护栏的预注册实验(每组约 30,000 名访客)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Prospect-Theory Arithmetic of Loss Frames — Chapter 1 computes with cumulative prospect theory: the subjective value of a sure loss to a sure gain of the same size is 2.25; "avoid a one-in-ten chance of losing RM 1,000" feels like a sure gain of RM 335.85. Turned into conversion, even inflating all valuations 2.25-fold raises conversion at RM 49 by only 100.6%; meta-analyses put framing effects near zero, so plans should use multipliers of 1.05 to 1.10, worth +5.6% to +11.1%. (https://5dlibrary.com/answers/en/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c01/) 第1章 损失框架的前景理论算术 — 第一章用累积前景理论计算:同额确定损失与确定收益的主观值之比为 2.25;「避免一成机会损失 RM 1,000」在主观上相当于确定得到 RM 335.85。换算成转化率,即使估值整体放大 2.25 倍,RM 49 的转化率也只提高 100.6%;框架效应的元分析接近于零,规划时应取 1.05 到 1.10 倍,对应 +5.6% 到 +11.1%。 (https://5dlibrary.com/answers/zh/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c01/) - Ch.2 Signalling Models of Price Endings and Numerical Precision — Chapter 2 computes the left-digit effect with an inattention parameter of 0.3: RM 49 is perceived as 46.3, a perceived gap of 7.4% from RM 50; a Bayesian model shows that the credibility of precise numbers comes entirely from checkability (posterior 0.500 without checking, 0.667 with a checking probability of 0.50). (https://5dlibrary.com/answers/en/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c02/) 第2章 价格尾数与数字精确性的信号模型 — 第二章以注意力不足参数 0.3 计算左位数字效应:RM 49 被感知为 46.3,与 RM 50 的感知差距为 7.4%;并用贝叶斯模型证明精确数字的可信度完全来自可核查性(没有核查时后验为 0.500,核查概率 0.50 时为 0.667)。 (https://5dlibrary.com/answers/zh/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c02/) - Ch.3 The Friction Equations of Checkout Micro-Architecture — Chapter 3 shows that the Hick-Hyman law is logarithmic and that checkout loss is a product of survival rates: each field removed multiplies completion by 1.0204, and going from three pages with 12 fields to one modal with 3 fields raises completion by 56.8%. (https://5dlibrary.com/answers/en/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c03/) 第3章 结账微界面的摩擦方程 — 第三章说明希克-海曼定律是对数关系,结账损失是各环节存活率的乘积:每去掉一个字段完成率乘以 1.0204,从三页 12 个字段压到一个弹窗 3 个字段,完成率提高 56.8%。 (https://5dlibrary.com/answers/zh/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c03/) - Ch.4 Cognitive Fluency and the Statistical Limits of Testing — Chapter 4 measures the fluency of copy and computes A/B sample sizes (13,914 visitors per arm for 3.0% against 3.6%) and the cost of peeking (false-positive rate 0.256 with 20 looks). (https://5dlibrary.com/answers/en/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c04/) 第4章 认知流畅度与测试的统计边界 — 第四章度量文案流畅度,并计算 A/B 测试的样本量(3.0% 对 3.6% 每组 13,914 名访客)与中途偷看的代价(看 20 次时假阳性率 0.256)。 (https://5dlibrary.com/answers/zh/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c04/) - Ch.5 The Full Ledger of a Fear Frame: Refunds, Chargebacks and Repeat Purchases — Chapter 5 puts the fear frame on a full ledger: copy implying "loss of luck" nets 16.8% less after refunds, chargebacks and complaints, and a year of repeat purchases only brings it close (RM 6,364 against RM 6,629). (https://5dlibrary.com/answers/en/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c05/) 第5章 恐惧框架的总账:退款、拒付与复购 — 第五章把恐惧框架放进总账:暗示「失去好运」的写法计入退款、拒付与投诉后净收入低 16.8%,一年复购也只是相近(RM 6,364 对 RM 6,629)。 (https://5dlibrary.com/answers/zh/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c05/) - Ch.6 The Ceiling of the Funnel and the Best Price — Chapter 6 decomposes the funnel, with the largest room in "checkout opened" (a ceiling of 6.67 times), and the left-digit effect moves the best price from RM 41 to RM 39. (https://5dlibrary.com/answers/en/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c06/) 第6章 转化漏斗的上限与最优价格 — 第六章分解漏斗,最大的空间在「打开结账」(上限 6.67 倍),最优价格在左位数字效应下由 RM 41 移到 RM 39。 (https://5dlibrary.com/answers/zh/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c06/) - Ch.7 Han Feizi, the Wenxin diaolong and the Rhetoric of Gain and Harm — Chapter 7 reads Han Feizi's Shuonan, the Wenxin diaolong and the Strategists, writing "knowing the listener's heart" as a value of information (+34.5%). (https://5dlibrary.com/answers/en/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c07/) 第7章 《韩非子·说难》《文心雕龙》与纵横家的利害之辞 — 第七章与《韩非子·说难》《文心雕龙》和纵横家并读,把「知所说之心」写成信息价值(+34.5%)。 (https://5dlibrary.com/answers/zh/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c07/) - Ch.8 The Winner's Curse, the Barnum Effect on Free Report Pages and a Pre-Registered Experiment — Chapter 8 gives the winner's curse (only 0.624 of significant winners are real), the Barnum effect on free report pages and a pre-registered experiment with a refund guardrail (about 30,000 visitors per arm). Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c08/) 第8章 赢家诅咒、免费报告页的巴纳姆效应与预注册实验 — 第八章给出赢家诅咒(显著胜出中只有 0.624 为真)、免费报告页的巴纳姆效应与一项以退款率为护栏的预注册实验(每组约 30,000 名访客)。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/persuasive-copywriting-algebra-loss-framing-checkout-microarchitecture/c08/) ## 5DL-B-078 · Order-Bump Friction Elasticity, One-Click Upsell Kinetics and Cross-Domain Recommendation Matrices 中文书名:结账顺手加购弹性、一键升单动力学与跨领域推荐矩阵 DOI: https://doi.org/10.5281/zenodo.24090207 Cite: Fan, Y. (2026). Order-Bump Friction Elasticity, One-Click Upsell Kinetics and Cross-Domain Recommendation Matrices (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090207 Book page: https://5dlibrary.com/books/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/ Scope: Valuation distributions, acceptance and refund rates and the needs structure of 30 sites are illustrative or synthetic; all data are generated by its programs; the book is not legal advice. Chapter 3 routes each of 30 synthetic sites to its top two by conditional mutual information and finds that even with 60,000 users only 14 to 17 sites have stable top twos. Abstract: When a buyer opens checkout, adding one more item takes a single tick; in the first second after payment, upgrading by paying the difference needs no re-entry of payment details; and the needs a buyer reveals on one site are often related to another. This book writes order bumps, one-click upsells and cross-site recommendation as runnable arithmetic and then asks when they hold. Valuation distributions, acceptance and refund rates and the needs structure of 30 sites are illustrative or synthetic; all data are generated by its programs; the book is not legal advice. Chapter 1 shows that the optimal bump price ratio is not a fixed band: in the illustrative model with zero friction it is 0.451, adding +22.5% to average order value, and across valuation distributions it ranges from 0.246 to 0.812, adding +15.0% to +31.0%; two bumps add +34.7%. Chapter 2 compares upsells in the first second after payment and on day 7 (acceptance 0.1300 against 0.0264, net RM 33.56 against RM 7.46 per buyer); counting refunds, the first second still wins. Chapter 3 routes each of 30 synthetic sites to its top two by conditional mutual information and finds that even with 60,000 users only 14 to 17 sites have stable top twos. Chapter 4 decomposes the order-value multiplier (22% needs a take rate of 0.489) and finds that a product-page bundle (RM 88.85) can beat a checkout bump (RM 72.48). Chapter 5 computes the cost of consent for using personal data across sites and the best number of recommendations at the foot of a report (3). Chapter 6 uses Thompson sampling to cut the loss of a price test from RM 48,005 to RM 1,831 and shows Simpson's paradox in comparing recommendations. Chapter 7 reads Bai Gui and Jiran in the Shiji and Shan Mugong's "mother and child weighing each other" in the Guoyu, writing a price ladder as a chain of ratios (geometric mean 5.10, largest gap 24.62). Chapter 8 separates attributed from incremental revenue (the increment is only 0.111 to 0.833 of it), uses refunds and chargebacks (11.1%) as a guardrail and gives a pre-registered holdout experiment. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:买家打开收银台的那一刻,再加一件东西只需要勾一下;付款成功后的第一秒,补差价升级不需要重新填写支付信息;买家在一个站点暴露的需要,常常与另一个站点相关。本书把结账加购、一键升单与跨站点交叉推荐写成可以运行的算术,再看它们在什么条件下成立。书中的估值分布、接受率、退款率与 30 个站点的需求结构均为示意或合成模型;全部数据由书中程序生成;本书不是法律意见。 第一章证明最优加购价格比不是固定区间:示意模型中摩擦为零时最优比为 0.451、对平均客单价贡献 +22.5%,随估值分布变化可从 0.246 到 0.812,贡献 +15.0% 到 +31.0%;两个加购合计 +34.7%。第二章比较付款后第一秒与第七天的升单(接受率 0.1300 对 0.0264,每位买家净收入 RM 33.56 对 RM 7.46),计入退款后第一秒仍然更优。第三章在合成的 30 站点模型上用条件互信息求前两条推荐路由,并发现即使有 60,000 名用户也只有 14 到 17 个站点的前两条稳定。第四章分解客单价乘数(22% 需要 0.489 的接受率),并发现商品页组合价(RM 88.85)可能优于结账加购(RM 72.48)。 第五章计算跨站点使用个人数据所需同意的代价与报告底部推荐的最优条数(3 条)。第六章以汤普森抽样把定价测试的损失从 RM 48,005 降到 RM 1,831,并展示推荐比较中的辛普森悖论。第七章与《史记·货殖列传》的白圭、计然及《国语·周语下》单穆公的「子母相权」并读,把价格阶梯写成比例链(几何平均 5.10,最大缺口 24.62 倍)。第八章区分归因收入与增量收入(增量只占 0.111 到 0.833),以退款与拒付(11.1%)为护栏,并给出一项预注册的留出实验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Zero-Friction Arbitrage of the Order Bump — Chapter 1 shows that the optimal bump price ratio is not a fixed band: in the illustrative model with zero friction it is 0.451, adding +22.5% to average order value, and across valuation distributions it ranges from 0.246 to 0.812, adding +15.0% to +31.0%; two bumps add +34.7%. (https://5dlibrary.com/answers/en/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c01/) 第1章 结账加购的零摩擦套利 — 第一章证明最优加购价格比不是固定区间:示意模型中摩擦为零时最优比为 0.451、对平均客单价贡献 +22.5%,随估值分布变化可从 0.246 到 0.812,贡献 +15.0% 到 +31.0%;两个加购合计 +34.7%。 (https://5dlibrary.com/answers/zh/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c01/) - Ch.2 Second 1 and Day 7: the Two Peaks of the Upsell — Chapter 2 compares upsells in the first second after payment and on day 7 (acceptance 0.1300 against 0.0264, net RM 33.56 against RM 7.46 per buyer); counting refunds, the first second still wins. (https://5dlibrary.com/answers/en/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c02/) 第2章 付款后第一秒与第七天:升单的双峰排程 — 第二章比较付款后第一秒与第七天的升单(接受率 0.1300 对 0.0264,每位买家净收入 RM 33.56 对 RM 7.46),计入退款后第一秒仍然更优。 (https://5dlibrary.com/answers/zh/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c02/) - Ch.3 Cross-Fleet Routing by Conditional Mutual Information — Chapter 3 routes each of 30 synthetic sites to its top two by conditional mutual information and finds that even with 60,000 users only 14 to 17 sites have stable top twos. (https://5dlibrary.com/answers/en/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c03/) 第3章 跨舰队推荐的条件互信息路由 — 第三章在合成的 30 站点模型上用条件互信息求前两条推荐路由,并发现即使有 60,000 名用户也只有 14 到 17 个站点的前两条稳定。 (https://5dlibrary.com/answers/zh/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c03/) - Ch.4 Order-Value Multipliers and Bundle Pricing — Chapter 4 decomposes the order-value multiplier (22% needs a take rate of 0.489) and finds that a product-page bundle (RM 88.85) can beat a checkout bump (RM 72.48). (https://5dlibrary.com/answers/en/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c04/) 第4章 客单价乘数与组合定价 — 第四章分解客单价乘数(22% 需要 0.489 的接受率),并发现商品页组合价(RM 88.85)可能优于结账加购(RM 72.48)。 (https://5dlibrary.com/answers/zh/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c04/) - Ch.5 The Cost of Consent and the Number of Recommendations — Chapter 5 computes the cost of consent for using personal data across sites and the best number of recommendations at the foot of a report (3). (https://5dlibrary.com/answers/en/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c05/) 第5章 同意的代价与推荐的数量 — 第五章计算跨站点使用个人数据所需同意的代价与报告底部推荐的最优条数(3 条)。 (https://5dlibrary.com/answers/zh/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c05/) - Ch.6 Learning the Bump Price Online, and Simpson's Paradox — Chapter 6 uses Thompson sampling to cut the loss of a price test from RM 48,005 to RM 1,831 and shows Simpson's paradox in comparing recommendations. (https://5dlibrary.com/answers/en/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c06/) 第6章 加购定价的在线学习与辛普森悖论 — 第六章以汤普森抽样把定价测试的损失从 RM 48,005 降到 RM 1,831,并展示推荐比较中的辛普森悖论。 (https://5dlibrary.com/answers/zh/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c06/) - Ch.7 The Shiji on Money-Makers and the Guoyu on Mother and Child Coins — Chapter 7 reads Bai Gui and Jiran in the Shiji and Shan Mugong's "mother and child weighing each other" in the Guoyu, writing a price ladder as a chain of ratios (geometric mean 5.10, largest gap 24.62). (https://5dlibrary.com/answers/en/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c07/) 第7章 《史记·货殖列传》与《国语》的子母相权 — 第七章与《史记·货殖列传》的白圭、计然及《国语·周语下》单穆公的「子母相权」并读,把价格阶梯写成比例链(几何平均 5.10,最大缺口 24.62 倍)。 (https://5dlibrary.com/answers/zh/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c07/) - Ch.8 Incremental Revenue, Guardrails and a Pre-Registered Holdout Experiment — Chapter 8 separates attributed from incremental revenue (the increment is only 0.111 to 0.833 of it), uses refunds and chargebacks (11.1%) as a guardrail and gives a pre-registered holdout experiment. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c08/) 第8章 增量收入、护栏指标与预注册的留出实验 — 第八章区分归因收入与增量收入(增量只占 0.111 到 0.833),以退款与拒付(11.1%)为护栏,并给出一项预注册的留出实验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/order-bump-elasticity-one-click-upsells-cross-domain-recommendation/c08/) ## 5DL-B-079 · Cart-Abandonment Recovery Kinetics, Sunk-Data Retention Ratchets and High-Ticket B2B Qualification Mechanisms 中文书名:弃单挽回动力学、沉没数据续费棘轮与高客单顾问筛选机制 DOI: https://doi.org/10.5281/zenodo.24090209 Cite: Fan, Y. (2026). Cart-Abandonment Recovery Kinetics, Sunk-Data Retention Ratchets and High-Ticket B2B Qualification Mechanisms (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090209 Book page: https://5dlibrary.com/books/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/ Scope: Return rates, unsubscribe rates, renewal parameters, client valuations and close rates are illustrative or synthetic; all data are generated by its programs; the book is not legal advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: When, and how many times, to follow up after a buyer abandons a cart? Does the data subscribers accumulate make renewal rise month by month? How can a business licence at RM 9,800 or RM 29,800 screen clients with menus, application forms and deposits? This book writes these rules of thumb as runnable arithmetic and then asks when they hold. Return rates, unsubscribe rates, renewal parameters, client valuations and close rates are illustrative or synthetic; all data are generated by its programs; the book is not legal advice. Chapter 1 compares follow-up schedules with an externally excited return-intensity model (not strictly a Hawkes process): "15 minutes, 24 hours, 72 hours" reaches 91% of the best gain on the grid, 33 of 220 schedules reach 90%, and the optimum is flat; net of unsubscribes, messages after the fourth add almost nothing. Chapter 2 finds that discount hooks teach repeat buyers to abandon on purpose (at a learning probability of 0.6 a year's revenue is RM 880,971, below the RM 882,196 of plain reminders), and that only 0.43 of attributed recovery is incremental. Chapter 3 writes down a sunk-data ratchet and shows that a selection model in which everyone's renewal rate is constant (sBG) fits the same cohort curve to within a largest gap of 0.011. Chapter 4 computes the cost of data export (present only in the ratchet model) and the self-selection of annual plans (-4.0% when buyers know their churn). Chapter 5 corrects a source: the two-licence problem is the monopoly menu of Mussa and Rosen, not the insurance market of Rothschild and Stiglitz; the menu separates the two kinds of client, while the friction of an application form screens only under the single-crossing condition. Chapter 6 computes the allocation of advisory calls by score (+48% to +100%) and when a refundable deposit is a signal. Chapter 7 reads the Sunzi's "Nine Grounds", the Guiguzi's "Chuai" and "Mo" chapters and the Mozi's "Exalting the Worthy I", writing "gauging the dispositions" as the information in a question ("budget approved", 0.149 bits). Chapter 8 gives the full recovery ledger (net only 36% of the attributed figure), a randomised experiment that tells a ratchet from selection, and a pre-registered holdout experiment. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:买家放弃购物车之后何时跟进、发几次?订阅者积累的数据会不会让续费逐月上升?一笔 RM 9,800 或 RM 29,800 的企业牌照,怎样用菜单、申请表与订金筛选客户?本书把这些经验法则写成可以运行的算术,再看它们在什么条件下成立。书中的回访率、退订率、续费参数、客户估值与成交率均为示意或合成模型;全部数据由书中程序生成;本书不是法律意见。 第一章用受外部激发的回访强度模型(不是严格的霍克斯过程)比较跟进时序:「15 分钟、24 小时、72 小时」达到网格最优增益的 91%,220 个时序中有 33 个达到 90%,最优面很平;扣除退订后,第四封以后几乎没有净收益。第二章发现折扣钩子会教会回头客故意弃单(学习概率 0.6 时一年收入 RM 880,971,低于普通提醒的 RM 882,196),而归因挽回中只有 0.43 是增量。第三章写出沉没数据棘轮,并证明一个人人续费率不变的筛选模型(sBG)能把同一条群组曲线拟合到最大差距 0.011。第四章计算数据导出的代价(只在棘轮模型中存在)与年付方案的自我选择(买家清楚自己的流失时收入 -4.0%)。 第五章改正出处:两档牌照的问题是穆萨-罗森的垄断菜单,不是罗斯柴尔德-斯蒂格利茨的保险市场;菜单把两类客户分开,而申请表的摩擦只有满足单交叉条件时才有甄别作用。第六章计算按评分分配顾问通话(+48% 到 +100%)与可退订金的信号条件。第七章与《孙子·九地》、《鬼谷子》揣篇与摩篇、《墨子·尚贤上》并读,把「揣情」写成问题的信息量(「预算已批准」0.149 比特)。第八章给出完整挽回账本(净收入只有归因的 36%)、区分棘轮与筛选的随机实验与一项预注册的留出实验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Timing and Number of Follow-Ups after Abandonment — Chapter 1 compares follow-up schedules with an externally excited return-intensity model (not strictly a Hawkes process): "15 minutes, 24 hours, 72 hours" reaches 91% of the best gain on the grid, 33 of 220 schedules reach 90%, and the optimum is flat; net of unsubscribes, messages after the fourth add almost nothing. (https://5dlibrary.com/answers/en/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c01/) 第1章 弃单跟进的时序与次数 — 第一章用受外部激发的回访强度模型(不是严格的霍克斯过程)比较跟进时序:「15 分钟、24 小时、72 小时」达到网格最优增益的 91%,220 个时序中有 33 个达到 90%,最优面很平;扣除退订后,第四封以后几乎没有净收益。 (https://5dlibrary.com/answers/zh/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c01/) - Ch.2 Discount Hooks and Overstated Recovery — Chapter 2 finds that discount hooks teach repeat buyers to abandon on purpose (at a learning probability of 0.6 a year's revenue is RM 880,971, below the RM 882,196 of plain reminders), and that only 0.43 of attributed recovery is incremental. (https://5dlibrary.com/answers/en/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c02/) 第2章 折扣钩子与被高估的挽回 — 第二章发现折扣钩子会教会回头客故意弃单(学习概率 0.6 时一年收入 RM 880,971,低于普通提醒的 RM 882,196),而归因挽回中只有 0.43 是增量。 (https://5dlibrary.com/answers/zh/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c02/) - Ch.3 A Sunk-Data Renewal Ratchet, and a Selection Curve that Rises the Same Way — Chapter 3 writes down a sunk-data ratchet and shows that a selection model in which everyone's renewal rate is constant (sBG) fits the same cohort curve to within a largest gap of 0.011. (https://5dlibrary.com/answers/en/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c03/) 第3章 沉没数据的续费棘轮,与同样上升的筛选曲线 — 第三章写出沉没数据棘轮,并证明一个人人续费率不变的筛选模型(sBG)能把同一条群组曲线拟合到最大差距 0.011。 (https://5dlibrary.com/answers/zh/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c03/) - Ch.4 The Cost of Data Export and the Self-Selection of Annual Plans — Chapter 4 computes the cost of data export (present only in the ratchet model) and the self-selection of annual plans (-4.0% when buyers know their churn). (https://5dlibrary.com/answers/en/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c04/) 第4章 数据导出的代价与年付方案的自我选择 — 第四章计算数据导出的代价(只在棘轮模型中存在)与年付方案的自我选择(买家清楚自己的流失时收入 -4.0%)。 (https://5dlibrary.com/answers/zh/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c04/) - Ch.5 A Two-Licence Menu and Screening by Application Form — "Keep unsuitable clients out with a long application form" works under one condition only: each extra question makes unsuitable clients more likely to give up. This is the single-crossing condition of screening theory. If both kinds give up at 0.04 per question, after 10 questions high-fit clients fall from 200 to 133.0 and low-fit from 800 to 531.9, and the share stays at 0.200: the friction only drives clients away. If high-fit clients give up at 0.03 per question and low-fit at 0.08, the share rises to 0.298 after 10 questions; if the former is 0.06, it rises only to 0.237, while nearly half of the high-fit clients are lost (107.7 remain). (https://5dlibrary.com/answers/en/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c05/) 第5章 两档牌照的菜单与申请表的甄别 — 「用一份长申请表把不合适的客户挡在门外」,只有在一个条件下成立:不合适的客户每多一道问题就更容易放弃。这就是甄别理论中的单交叉条件。若两类客户每道问题都有 0.04 的人放弃,10 道问题之后高合适度客户从 200 人降到 133.0 人,低合适度从 800 人降到 531.9 人,比例仍然是 0.200:摩擦只是赶走了客户。若高合适度客户每道题放弃 0.03、低合适度 0.08,10 道题后比例升到 0.298;若前者是 0.06,只升到 0.237,而高合适度客户损失近半(剩 107.7 人)。 (https://5dlibrary.com/answers/zh/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c05/) - Ch.6 Allocating Advisor Time, and Deposits as Signals — Chapter 6 computes the allocation of advisory calls by score (+48% to +100%) and when a refundable deposit is a signal. (https://5dlibrary.com/answers/en/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c06/) 第6章 顾问时间的分配与订金的信号 — 第六章计算按评分分配顾问通话(+48% 到 +100%)与可退订金的信号条件。 (https://5dlibrary.com/answers/zh/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c06/) - Ch.7 The Sunzi on the Nine Grounds, the Guiguzi on Probing, and the Mozi on Paying by Merit — Chapter 7 reads the Sunzi's "Nine Grounds", the Guiguzi's "Chuai" and "Mo" chapters and the Mozi's "Exalting the Worthy I", writing "gauging the dispositions" as the information in a question ("budget approved", 0.149 bits). (https://5dlibrary.com/answers/en/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c07/) 第7章 《孙子·九地》《鬼谷子》揣摩与《墨子》量功分禄 — 第七章与《孙子·九地》、《鬼谷子》揣篇与摩篇、《墨子·尚贤上》并读,把「揣情」写成问题的信息量(「预算已批准」0.149 比特)。 (https://5dlibrary.com/answers/zh/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c07/) - Ch.8 The Full Ledger, an Experiment to Tell a Ratchet from Selection, and a Pre-Registered Holdout — Chapter 8 gives the full recovery ledger (net only 36% of the attributed figure), a randomised experiment that tells a ratchet from selection, and a pre-registered holdout experiment. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c08/) 第8章 完整账本、区分棘轮与筛选的实验,以及预注册的留出实验 — 第八章给出完整挽回账本(净收入只有归因的 36%)、区分棘轮与筛选的随机实验与一项预注册的留出实验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/abandonment-recovery-kinetics-retention-ratchets-b2b-qualification/c08/) ## 5DL-B-080 · Rhythmic Acoustic Stimulation, Auditory Steady-State Responses and Modulation Signal Engineering 中文书名:节律声刺激、听觉稳态反应与双耳等时调制工程 DOI: https://doi.org/10.5281/zenodo.24090211 Cite: Fan, Y. (2026). Rhythmic Acoustic Stimulation, Auditory Steady-State Responses and Modulation Signal Engineering (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090211 Book page: https://5dlibrary.com/books/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/ Scope: Its phase-locking curves, response amplitudes, EEG noise and arousal probabilities are illustrative model parameters, not data from any listener or product; it gives no medical, hearing-care or sleep-treatment advice. Under an illustrative phase-locking curve, an isochronic tone drives 17.6 dB more power than a binaural beat on a 400 Hz carrier and 84.4 dB more at 2,000 Hz. Abstract: Most rhythmic audio rests on three claims: that binaural beats make the EEG follow the beat rate, that static pure tones such as 432 Hz or 528 Hz resonate with brain waves or even repair DNA, and that pink noise and 7.83 Hz "Schumann resonance" tracks deepen sleep. This book turns them into 16 runnable programs and tests each. Its phase-locking curves, response amplitudes, EEG noise and arousal probabilities are illustrative model parameters, not data from any listener or product; it gives no medical, hearing-care or sleep-treatment advice. Chapter 1 shows with Hilbert envelopes that each ear of a binaural pair receives a steady tone of modulation depth 0.0000, the beat existing only in the brainstem, while the same two tones mixed in the air make a monaural beat of depth 1.0000. Under an illustrative phase-locking curve, an isochronic tone drives 17.6 dB more power than a binaural beat on a 400 Hz carrier and 84.4 dB more at 2,000 Hz. Chapter 2 draws the Arnold tongue of Adler's equation and computes how the signal-to-noise ratio of the auditory steady-state response grows with recording time: a 40 Hz isochronic tone becomes visible in 8.0 s, a binaural beat at the same rate needs 7.8 min, 58.1 times longer, and a static tone has drive 0. Chapter 3 synthesises pink noise by the Voss-McCartney method and a filter bank, with slopes of 3.08 dB and 2.99 dB per octave (target 3.01 dB); brown noise at the same unweighted level reads 12.8 dB lower on an A-weighted meter; in an illustrative model a masking background lowers the expected micro-arousals in a night from 5.68 to 2.01. Chapter 4 shows that loudspeakers and bone conduction turn a binaural beat into a monaural one (leaked depth 0.7079 through loudspeakers) and compares gating shapes for isochronic tones. Chapter 5 computes hearing dose and shows that the Schumann resonance, ideally 10.59 Hz and observed near 7.83 Hz, is an electromagnetic mode wrapping once around the planet with a magnetic field 50,000,000 times weaker than the static geomagnetic field, sharing no mechanism with audio at the same frequency. Chapter 6 shows that study design can decide the result: in 60 s of recording a 40 Hz isochronic tone is visible in 0.999 of listeners and a binaural beat in 0.298; an uncontrolled before-and-after study of 60 people finds a "significant" result with chance 0.873; 285 tests produce 14.27 false positive cells on average. Chapter 7 reads the Mengxi Bitan on the sixteen pitches, the comma of 23.46 cents in sanfen sunyi, Zhu Zaiyu's equal temperament and Helmholtz on combination tones, and summarises the pitch-pipe legend of the Lüshi Chunqiu. Chapter 8 gives a ten-flag audit and a pre-registered crossover trial (262 sleepers to complete, 328 to enrol). Marked boxes separate evidence, runnable Python programs, classical sources and hypotheses. 摘要:节律音频产品大多建立在三个主张上:双耳拍能让脑电跟随拍频,432 Hz、528 Hz 一类的静态纯音能共振脑波甚至修复 DNA,粉红噪声与 7.83 Hz 的「舒曼共振」音轨能加深睡眠。本书把这些主张写成 16 段可运行的程序逐条检验。书中的锁相曲线、反应幅度、脑电噪声与觉醒概率都是示意模型参数,不是任何受试者或产品的数据;本书不提供医疗、听力保健或睡眠治疗建议。 第一章用希尔伯特包络证明,双耳拍的每只耳朵收到的是调制深度为 0.0000 的稳态音,拍只存在于脑干;同样两个音在空气中混合,就成了深度 1.0000 的单耳拍。在示意锁相曲线下,400 Hz 载波时等时音的驱动功率比双耳拍高 17.6 dB,2,000 Hz 时高 84.4 dB。第二章画出 Adler 方程的阿诺德舌,并计算听觉稳态反应的信噪比随记录时长增长:40 Hz 等时音只需 8.0 s 即可看见,同频率的双耳拍需要 7.8 min,长 58.1 倍;静态纯音的驱动为 0。第三章用 Voss-McCartney 法与滤波器组合成粉红噪声,斜率为每倍频程 3.08 dB 与 2.99 dB(目标 3.01 dB);同样未加权声级的布朗噪声,A 计权读数低 12.8 dB;在示意模型中,掩蔽背景把一夜的预期微觉醒从 5.68 次降到 2.01 次。 第四章说明扬声器与骨传导把双耳拍变成单耳拍(扬声器的串音深度 0.7079),并比较等时音的门控波形。第五章计算听力剂量,并说明舒曼共振的理想值为 10.59 Hz、观测值约 7.83 Hz,是一个绕地球一圈的电磁模态,磁场比地磁静场弱 50,000,000 倍,与同频率的音频没有共同机制。第六章显示研究设计足以决定结论:40 Hz 等时音在 60 s 记录中对 0.999 的听者可见,双耳拍只有 0.298;没有对照的前后对比在 60 人时有 0.873 的机会得到「显著」结果;285 次检验平均产生 14.27 个假阳性格子。第七章读《梦溪笔谈》论十六律、三分损益的 23.46 音分音差、朱载堉的十二平均律与亥姆霍兹的组合音,并概述《吕氏春秋·古乐》的律管传说。第八章给出十项自动审计与一项预注册交叉试验(完成 262 人,招募 328 人)。书中以标注框区分证据、可运行的 Python 程序、原典与假说。 Chapter answers: - Ch.1 Where Is the Envelope? Binaural Beats and Isochronic Tones — Chapter 1 shows with Hilbert envelopes that each ear of a binaural pair receives a steady tone of modulation depth 0.0000, the beat existing only in the brainstem, while the same two tones mixed in the air make a monaural beat of depth 1.0000. Under an illustrative phase-locking curve, an isochronic tone drives 17.6 dB more power than a binaural beat on a 400 Hz carrier and 84.4 dB more at 2,000 Hz. (https://5dlibrary.com/answers/en/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c01/) 第1章 包络在哪里:双耳拍与等时音 — 第一章用希尔伯特包络证明,双耳拍的每只耳朵收到的是调制深度为 0.0000 的稳态音,拍只存在于脑干;同样两个音在空气中混合,就成了深度 1.0000 的单耳拍。在示意锁相曲线下,400 Hz 载波时等时音的驱动功率比双耳拍高 17.6 dB,2,000 Hz 时高 84.4 dB。 (https://5dlibrary.com/answers/zh/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c01/) - Ch.2 Auditory Steady-State Responses and the Arnold Tongue — Chapter 2 draws the Arnold tongue of Adler's equation and computes how the signal-to-noise ratio of the auditory steady-state response grows with recording time: a 40 Hz isochronic tone becomes visible in 8.0 s, a binaural beat at the same rate needs 7.8 min, 58.1 times longer, and a static tone has drive 0. (https://5dlibrary.com/answers/en/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c02/) 第2章 听觉稳态反应与阿诺德舌 — 第二章画出 Adler 方程的阿诺德舌,并计算听觉稳态反应的信噪比随记录时长增长:40 Hz 等时音只需 8.0 s 即可看见,同频率的双耳拍需要 7.8 min,长 58.1 倍;静态纯音的驱动为 0。 (https://5dlibrary.com/answers/zh/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c02/) - Ch.3 Pink Noise, A-Weighting and Sleep Masking — Chapter 3 synthesises pink noise by the Voss-McCartney method and a filter bank, with slopes of 3.08 dB and 2.99 dB per octave (target 3.01 dB); brown noise at the same unweighted level reads 12.8 dB lower on an A-weighted meter; in an illustrative model a masking background lowers the expected micro-arousals in a night from 5.68 to 2.01. (https://5dlibrary.com/answers/en/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c03/) 第3章 粉红噪声、A 计权与睡眠掩蔽 — 第三章用 Voss-McCartney 法与滤波器组合成粉红噪声,斜率为每倍频程 3.08 dB 与 2.99 dB(目标 3.01 dB);同样未加权声级的布朗噪声,A 计权读数低 12.8 dB;在示意模型中,掩蔽背景把一夜的预期微觉醒从 5.68 次降到 2.01 次。 (https://5dlibrary.com/answers/zh/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c03/) - Ch.4 Delivery Routes and Gating Shapes — Chapter 4 shows that loudspeakers and bone conduction turn a binaural beat into a monaural one (leaked depth 0.7079 through loudspeakers) and compares gating shapes for isochronic tones. (https://5dlibrary.com/answers/en/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c04/) 第4章 播放途径与门控波形 — 第四章说明扬声器与骨传导把双耳拍变成单耳拍(扬声器的串音深度 0.7079),并比较等时音的门控波形。 (https://5dlibrary.com/answers/zh/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c04/) - Ch.5 Hearing Dose and the Schumann Resonance — Chapter 5 computes hearing dose and shows that the Schumann resonance, ideally 10.59 Hz and observed near 7.83 Hz, is an electromagnetic mode wrapping once around the planet with a magnetic field 50,000,000 times weaker than the static geomagnetic field, sharing no mechanism with audio at the same frequency. (https://5dlibrary.com/answers/en/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c05/) 第5章 听力剂量与舒曼共振 — 第五章计算听力剂量,并说明舒曼共振的理想值为 10.59 Hz、观测值约 7.83 Hz,是一个绕地球一圈的电磁模态,磁场比地磁静场弱 50,000,000 倍,与同频率的音频没有共同机制。 (https://5dlibrary.com/answers/zh/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c05/) - Ch.6 Evidence: Detection Power, Placebo and Forking Paths — Chapter 6 shows that study design can decide the result: in 60 s of recording a 40 Hz isochronic tone is visible in 0.999 of listeners and a binaural beat in 0.298; an uncontrolled before-and-after study of 60 people finds a "significant" result with chance 0.873; 285 tests produce 14.27 false positive cells on average. (https://5dlibrary.com/answers/en/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c06/) 第6章 证据:检出力、安慰剂与分岔路径 — 第六章显示研究设计足以决定结论:40 Hz 等时音在 60 s 记录中对 0.999 的听者可见,双耳拍只有 0.298;没有对照的前后对比在 60 人时有 0.873 的机会得到「显著」结果;285 次检验平均产生 14.27 个假阳性格子。 (https://5dlibrary.com/answers/zh/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c06/) - Ch.7 Pitch Pipes, Sixteen Pitches, the Exact Root and Combination Tones — Chapter 7 reads the Mengxi Bitan on the sixteen pitches, the comma of 23.46 cents in sanfen sunyi, Zhu Zaiyu's equal temperament and Helmholtz on combination tones, and summarises the pitch-pipe legend of the Lüshi Chunqiu. (https://5dlibrary.com/answers/en/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c07/) 第7章 律管、十六律、密率与组合音 — 第七章读《梦溪笔谈》论十六律、三分损益的 23.46 音分音差、朱载堉的十二平均律与亥姆霍兹的组合音,并概述《吕氏春秋·古乐》的律管传说。 (https://5dlibrary.com/answers/zh/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c07/) - Ch.8 A Ten-Flag Audit and a Pre-Registered Crossover Trial — Chapter 8 gives a ten-flag audit and a pre-registered crossover trial (262 sleepers to complete, 328 to enrol). Marked boxes separate evidence, runnable Python programs, classical sources and hypotheses. (https://5dlibrary.com/answers/en/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c08/) 第8章 十项审计与一项预注册交叉试验 — 第八章给出十项自动审计与一项预注册交叉试验(完成 262 人,招募 328 人)。书中以标注框区分证据、可运行的 Python 程序、原典与假说。 (https://5dlibrary.com/answers/zh/rhythmic-acoustic-stimulation-assr-binaural-isochronic-modulation/c08/) ## 5DL-B-081 · Short-Video Completion Hazards, Recommender Exploration Loops and Automated Acquisition Funnels 中文书名:短视频完播率风险函数、推荐反馈环与自动化获客漏斗 DOI: https://doi.org/10.5281/zenodo.24090213 Cite: Fan, Y. (2026). Short-Video Completion Hazards, Recommender Exploration Loops and Automated Acquisition Funnels (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090213 Book page: https://5dlibrary.com/books/short-video-completion-hazards-recommender-feedback-acquisition-funnels/ Scope: Marked boxes separate evidence, runnable Python programs, classical sources and hypotheses. Abstract: People who acquire customers through short video widely believe three rules of thumb: the first three seconds decide everything and completion rate is the only metric that matters; a platform first shows a new video to about 300 people, and a video that fails that gate is bad; more views are better, and broader hooks are better. This book turns the three rules into 16 runnable programs and tests them one by one. Its Weibull parameters, pool sizes, conversion rates, prices (Malaysian ringgit, RM) and ban probabilities are illustrative model parameters, not data from any platform, account or business; it describes no platform's real algorithm and is not marketing or business advice. Chapter 1 describes swiping with a Weibull hazard: under a falling hazard 0.320 of all swipes on a 60-second video happen in the first 3 seconds, while completion falls with length far more slowly than an exponential model says (0.145 at 60 seconds). Chapter 2 simulates tiered traffic pools: with a sample of 300, even when audiences are alike, 0.169 of good videos are stopped at the first gate, and 0.308 to 0.478 with audience differences; the right hedge is a re-edit, not a repost. Chapter 3 shows recommender feedback magnifying luck into gaps (the Gini coefficient of likes rises from 0.214 to 0.964) and proves that both completion and watch time can be gamed by editing length: the 7-second video with the weakest holding power has the highest completion (0.544), and only normalising by the same-length baseline restores the true order. Chapter 4 finds a cash-optimal hook breadth of 0.314 (RM 2,534 a day), while the broadest hook brings 10 times the views of the narrowest and leaves RM 545 a day; an attention-residue model gives a best offer delay of 7.59 seconds, and letting the video finish its story first lifts conversion 1.246 times. Chapter 5 simulates the cash flow from views to settled payments (median break-even on day 35) and shows that an account matrix sharing devices raises the chance of losing eight or more accounts in a year from 0.0019 to 0.2353. Chapter 6 shows that an A/B test of paid conversion needs 42,691 viewers per arm, that peeking declares 0.239 of null tests significant, that studying only top accounts recommends a shock hook with lower mean revenue (RM 1,250 against RM 1,647), and that forecasting a year from the first three days falls short by 0.602. Chapter 7 reads the Guiguzi chapter "Chuai", the Wenxin Diaolong chapter "Zhangju" and the Hanfeizi story of the casket without the pearl. Chapter 8 gives a ten-flag audit and a pre-registered experiment randomised by video (88 videos per arm). Marked boxes separate evidence, runnable Python programs, classical sources and hypotheses. 摘要:以短视频获客的人普遍相信三条经验法则:前三秒决定生死,完播率是唯一重要的指标;平台先把新视频推给约 300 人,没过第一关的就是坏视频;播放越多越好,钩子越宽越好。本书把三条法则写成 16 段可运行的程序逐条检验。书中的威布尔参数、流量池规模、转化率、价格(令吉 RM)与封号概率都是示意模型参数,不是任何平台、账号或企业的数据;本书不描述任何平台的真实算法,也不构成营销或经营建议。 第一章用威布尔风险函数描述划走:在递减风险下,60 秒视频的划走有 0.320 发生在前 3 秒,而完播率随时长下降得比指数模型慢得多(60 秒为 0.145)。第二章模拟分级流量池:300 人样本下,即使各池观众相同,好视频也有 0.169 在第一关被拦下,观众差异下为 0.308 至 0.478;对冲噪声应重剪而不是重发。第三章说明推荐反馈把运气放大为差距(点赞基尼系数从 0.214 升到 0.964),并证明完播率与观看时长都能被剪辑长度刷高:留存力最弱的 7 秒视频完播率最高(0.544),按同时长基线归一才能恢复真实次序。第四章求出钩子宽度的现金最优 0.314(每天 RM 2,534),最宽的钩子带来 10 倍于最窄钩子的播放,却只剩每天 RM 545;注意力残留模型给出 7.59 秒的最佳出价延迟,而先讲完故事可使转化提高到 1.246 倍。 第五章模拟从播放到到账的现金流(回本日中位数为第 35 天),并说明共享设备的账号矩阵一年失去八个以上账号的概率从 0.0019 升到 0.2353。第六章显示付费转化的 A/B 测试每组需要 42,691 名观众,偷看式停止使 0.239 的无效测试被宣告显著;只学头部账号会推荐平均收入更低的惊悚钩子(RM 1,250 对 RM 1,647);以前三天预测全年会低估 0.602。第七章读《鬼谷子·揣篇》、《文心雕龙·章句》与《韩非子》「买椟还珠」。第八章给出十项自动审计与一项以视频为单位随机化的预注册实验(每组 88 条视频)。书中以标注框区分证据、可运行的 Python 程序、原典与假说。 Chapter answers: - Ch.1 The Kinetics of Swiping: a Weibull Hazard — Chapter 1 describes swiping with a Weibull hazard: under a falling hazard 0.320 of all swipes on a 60-second video happen in the first 3 seconds, while completion falls with length far more slowly than an exponential model says (0.145 at 60 seconds). (https://5dlibrary.com/answers/en/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c01/) 第1章 划走的动力学:威布尔风险函数 — 第一章用威布尔风险函数描述划走:在递减风险下,60 秒视频的划走有 0.320 发生在前 3 秒,而完播率随时长下降得比指数模型慢得多(60 秒为 0.145)。 (https://5dlibrary.com/answers/zh/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c01/) - Ch.2 Tiered Traffic Pools and the First Gate — Chapter 2 simulates tiered traffic pools: with a sample of 300, even when audiences are alike, 0.169 of good videos are stopped at the first gate, and 0.308 to 0.478 with audience differences; the right hedge is a re-edit, not a repost. (https://5dlibrary.com/answers/en/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c02/) 第2章 分级流量池与第一关的误杀 — 第二章模拟分级流量池:300 人样本下,即使各池观众相同,好视频也有 0.169 在第一关被拦下,观众差异下为 0.308 至 0.478;对冲噪声应重剪而不是重发。 (https://5dlibrary.com/answers/zh/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c02/) - Ch.3 Feedback Loops, the Matthew Effect and Gameable Metrics — Chapter 3 shows recommender feedback magnifying luck into gaps (the Gini coefficient of likes rises from 0.214 to 0.964) and proves that both completion and watch time can be gamed by editing length: the 7-second video with the weakest holding power has the highest completion (0.544), and only normalising by the same-length baseline restores the true order. (https://5dlibrary.com/answers/en/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c03/) 第3章 反馈环、马太效应与可被「刷」的指标 — 第三章说明推荐反馈把运气放大为差距(点赞基尼系数从 0.214 升到 0.964),并证明完播率与观看时长都能被剪辑长度刷高:留存力最弱的 7 秒视频完播率最高(0.544),按同时长基线归一才能恢复真实次序。 (https://5dlibrary.com/answers/zh/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c03/) - Ch.4 Reverse Selection by Hook Breadth and the Timing of the Offer — Chapter 4 finds a cash-optimal hook breadth of 0.314 (RM 2,534 a day), while the broadest hook brings 10 times the views of the narrowest and leaves RM 545 a day; an attention-residue model gives a best offer delay of 7.59 seconds, and letting the video finish its story first lifts conversion 1.246 times. (https://5dlibrary.com/answers/en/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c04/) 第4章 钩子宽度的逆向选择与落地时机 — 第四章求出钩子宽度的现金最优 0.314(每天 RM 2,534),最宽的钩子带来 10 倍于最窄钩子的播放,却只剩每天 RM 545;注意力残留模型给出 7.59 秒的最佳出价延迟,而先讲完故事可使转化提高到 1.246 倍。 (https://5dlibrary.com/answers/zh/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c04/) - Ch.5 From Views to Cash: Lags, Heavy Tails and Linked Bans — Chapter 5 simulates the cash flow from views to settled payments (median break-even on day 35) and shows that an account matrix sharing devices raises the chance of losing eight or more accounts in a year from 0.0019 to 0.2353. (https://5dlibrary.com/answers/en/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c05/) 第5章 从播放到现金:滞后、厚尾与关联封号 — 第五章模拟从播放到到账的现金流(回本日中位数为第 35 天),并说明共享设备的账号矩阵一年失去八个以上账号的概率从 0.0019 升到 0.2353。 (https://5dlibrary.com/answers/zh/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c05/) - Ch.6 Evidence: A/B Tests, Survivorship and Early Forecasts — Chapter 6 shows that an A/B test of paid conversion needs 42,691 viewers per arm, that peeking declares 0.239 of null tests significant, that studying only top accounts recommends a shock hook with lower mean revenue (RM 1,250 against RM 1,647), and that forecasting a year from the first three days falls short by 0.602. (https://5dlibrary.com/answers/en/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c06/) 第6章 证据:A/B 测试、幸存者偏差与早期预测 — 第六章显示付费转化的 A/B 测试每组需要 42,691 名观众,偷看式停止使 0.239 的无效测试被宣告显著;只学头部账号会推荐平均收入更低的惊悚钩子(RM 1,250 对 RM 1,647);以前三天预测全年会低估 0.602。 (https://5dlibrary.com/answers/zh/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c06/) - Ch.7 Gauging Desire, Opening Lines and the Casket Without the Pearl — The man of Chu made the casket so beautiful that the buyer kept the casket and gave back the pearl. A broad hook is just such a casket. The third part of the program uses the model of Chapter 4 to compute the share of "casket buyers", viewers the hook brings in who, judged by the conversion of the narrowest hook, will not buy: 0.256 at the cash-optimal breadth 0.314, 0.660 at breadth 0.6, and 0.950 at the broadest. Of the viewers the broadest hook brings, nineteen in twenty came for the casket. (https://5dlibrary.com/answers/en/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c07/) 第7章 揣情、章句与买椟还珠 — 楚人把匣子装饰得过于华美,郑人买下匣子,退还了珍珠:这可以说善于卖匣子,却不能说善于卖珍珠。宽钩子就是一只华美的匣子。程序的第三部分以第四章的模型计算「买椟者」的比例,即被钩子吸引来、但按最窄钩子的转化率衡量不会购买的观众:现金最优宽度 0.314 时为 0.256,宽度 0.6 时为 0.660,最宽时为 0.950。最宽的钩子带来的观众,二十个里有十九个是来买匣子的。 (https://5dlibrary.com/answers/zh/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c07/) - Ch.8 A Ten-Flag Audit and a Pre-Registered Experiment Randomised by Video — Chapter 8 gives a ten-flag audit and a pre-registered experiment randomised by video (88 videos per arm). Marked boxes separate evidence, runnable Python programs, classical sources and hypotheses. (https://5dlibrary.com/answers/en/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c08/) 第8章 十项审计与一项以视频为单位的预注册实验 — 第八章给出十项自动审计与一项以视频为单位随机化的预注册实验(每组 88 条视频)。书中以标注框区分证据、可运行的 Python 程序、原典与假说。 (https://5dlibrary.com/answers/zh/short-video-completion-hazards-recommender-feedback-acquisition-funnels/c08/) ## 5DL-B-082 · Xuankong Flying-Star Permutations, Twenty-Four-Mountain Compass Discretization and Craniofacial Ratio Metrology 中文书名:玄空飞星置换代数、二十四山罗盘离散化与骨相比例计量 DOI: https://doi.org/10.5281/zenodo.24090215 Cite: Fan, Y. (2026). Xuankong Flying-Star Permutations, Twenty-Four-Mountain Compass Discretization and Craniofacial Ratio Metrology (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090215 Book page: https://5dlibrary.com/books/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/ Scope: Compass errors, declinations, facial dimensions and landmark errors are illustrative parameters; chart rules follow conventions stated in the book. Marked boxes separate evidence, runnable Python programs, classical sources and hypotheses. Abstract: This book does not discuss whether any feng shui configuration or face reading comes true. It studies only the structures of Xuankong flying stars and craniofacial proportions that arithmetic can test, written as 16 runnable programs. Compass errors, declinations, facial dimensions and landmark errors are illustrative parameters; chart rules follow conventions stated in the book. Chapter 1 proves that flying-star charts are affine permutations of the cyclic group of order nine: forward and backward flights make 18 charts, exactly the dihedral group of order 18, of which only the two with 5 at the centre are magic squares. The 9 periods × 24 mountains = 216 ordinary charts fall into four classes by where the period star lands: 48 each for prosperous mountain, prosperous facing and for mountain up, water down, and 60 each for double stars at the facing and at the sitting; the class depends only on the flight directions of the mountain and facing stars, and period 9 has no prosperous ordinary chart. Chapter 2 counts sum-to-ten charts (12 on each side) and three-combination charts (16, all of the mountain up, water down class), and shows that two replacement conventions each change 182 charts and agree on only half. Chapter 3 computes the effects of compass error: at a standard deviation of 3 degrees, 0.1594 of readings cross into another mountain, the period-9 chart changes for 0.1063, and the near-the-edge call flips for 0.3118; with no stated north, a declination of 5 degrees moves 0.3333 of houses to another mountain. Chapter 4 computes base rates for equal thirds (0.0988 or less), five eyes and golden faces (about 0.0014), and shows that a selfie at 30 cm inflates nose width against cheekbone width by 1.273. Chapter 5 shows that 2 pixels of landmark error make two annotators disagree in 0.3513 of cases even for a face at the centre of an ideal band, and that with 12 measured ratios 0.7176 of people have at least one rare ratio. Chapter 6 discusses the royal cubit and canons of proportion, the Bayes factor of testimonials (only 1.0667 per vague hit), and a false configuration effect manufactured by orientation in a grid city (z = 7.02, gone after stratification). Chapter 7 reads Shen Kuo's Mengxi Bitan on the needle that inclines slightly east and four ways to mount it, and explains the place of the Dili Bianzheng, the Shenshi Xuankong Xue and the Mayi Shenxiang. Chapter 8 gives a ten-flag audit and a pre-registered blind paired test (204 pairs at a true hit rate of 0.60). Marked boxes separate evidence, runnable Python programs, classical sources and hypotheses. 摘要:本书不讨论任何风水格局或面相是否应验,只研究玄空飞星与骨相比例中可以用算术检验的结构,写成 16 段可运行的程序。罗盘误差、磁偏角、面部尺寸与标点误差都是示意参数;排盘规则按书中写明的约定执行。 第一章证明飞星排盘是九元循环群上的仿射置换:顺飞与逆飞共 18 张盘,恰为 18 阶二面体群,其中只有中宫为 5 的两张是幻方。9 个元运 × 24 山 = 216 个下卦盘按当运之星的落点分为四类,旺山旺向与上山下水各 48 个,双星会向与双星会坐各 60 个;格局只由山星与向星的飞行方向决定,九运没有旺山旺向的下卦盘。第二章计算合十(向星、山星各 12 个)与三般卦(16 个,全在上山下水之中),并显示两种替卦约定各改变 182 个盘,彼此只在一半的盘上一致。第三章计算罗盘误差的后果:误差标准差 3 度时 0.1594 的读数跨山,九运盘面改变 0.1063,兼向的判断翻转 0.3118;不说明以哪一个北为准时,磁偏角 5 度使 0.3333 的房子换山。 第四章计算三停均等(0.0988 以下)、五眼与黄金脸(约 0.0014)的基础比率,以及 30 厘米自拍使鼻宽对颧宽放大 1.273 倍。第五章显示 2 像素的标点误差使两位标注者对位于理想区间中点的脸也有 0.3513 的分歧,测量 12 个比例时 0.7176 的人至少有一个罕见比例。第六章讨论王室腕尺与比例法则、见证的贝叶斯因子(笼统应验每次只有 1.0667),以及棋盘格城市中由朝向制造的虚假格局效应(z = 7.02,分层后消失)。第七章读沈括《梦溪笔谈》论磁针「常微偏东」与四种安针法,并说明《地理辨正》《沈氏玄空学》《麻衣神相》在本书中的位置。第八章给出十项自动审计与一项预注册的盲配对检验(真实命中率 0.60 时需 204 对)。书中以标注框区分证据、可运行的 Python 程序、原典与假说。 Chapter answers: - Ch.1 The Lo Shu, the Flight Path and the Dihedral Group — Chapter 1 proves that flying-star charts are affine permutations of the cyclic group of order nine: forward and backward flights make 18 charts, exactly the dihedral group of order 18, of which only the two with 5 at the centre are magic squares. The 9 periods × 24 mountains = 216 ordinary charts fall into four classes by where the period star lands: 48 each for prosperous mountain, prosperous facing and for mountain up, water down, and 60 each for double stars at the facing and at the sitting; the class depends only on the flight directions of the mountain and facing stars, and period 9 has no prosperous ordinary chart. (https://5dlibrary.com/answers/en/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c01/) 第1章 洛书、飞星路径与二面体群 — 第一章证明飞星排盘是九元循环群上的仿射置换:顺飞与逆飞共 18 张盘,恰为 18 阶二面体群,其中只有中宫为 5 的两张是幻方。9 个元运 × 24 山 = 216 个下卦盘按当运之星的落点分为四类,旺山旺向与上山下水各 48 个,双星会向与双星会坐各 60 个;格局只由山星与向星的飞行方向决定,九运没有旺山旺向的下卦盘。 (https://5dlibrary.com/answers/zh/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c01/) - Ch.2 Sum to Ten, Three Combinations and the Conventions of Replacement — Chapter 2 counts sum-to-ten charts (12 on each side) and three-combination charts (16, all of the mountain up, water down class), and shows that two replacement conventions each change 182 charts and agree on only half. (https://5dlibrary.com/answers/en/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c02/) 第2章 合十、三般卦与替卦的约定 — 第二章计算合十(向星、山星各 12 个)与三般卦(16 个,全在上山下水之中),并显示两种替卦约定各改变 182 个盘,彼此只在一半的盘上一致。 (https://5dlibrary.com/answers/zh/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c02/) - Ch.3 Discretising Twenty-Four Mountains and Compass Error — Chapter 3 computes the effects of compass error: at a standard deviation of 3 degrees, 0.1594 of readings cross into another mountain, the period-9 chart changes for 0.1063, and the near-the-edge call flips for 0.3118; with no stated north, a declination of 5 degrees moves 0.3333 of houses to another mountain. (https://5dlibrary.com/answers/en/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c03/) 第3章 二十四山的离散化与罗盘误差 — 第三章计算罗盘误差的后果:误差标准差 3 度时 0.1594 的读数跨山,九运盘面改变 0.1063,兼向的判断翻转 0.3118;不说明以哪一个北为准时,磁偏角 5 度使 0.3333 的房子换山。 (https://5dlibrary.com/answers/zh/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c03/) - Ch.4 Thirds, Five Eyes, the Golden Ratio and Selfie Perspective — Chapter 4 computes base rates for equal thirds (0.0988 or less), five eyes and golden faces (about 0.0014), and shows that a selfie at 30 cm inflates nose width against cheekbone width by 1.273. (https://5dlibrary.com/answers/en/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c04/) 第4章 三停五眼、黄金比例与自拍的透视 — 第四章计算三停均等(0.0988 以下)、五眼与黄金脸(约 0.0014)的基础比率,以及 30 厘米自拍使鼻宽对颧宽放大 1.273 倍。 (https://5dlibrary.com/answers/zh/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c04/) - Ch.5 Landmark Error and the Barnum Effect — Chapter 5 shows that 2 pixels of landmark error make two annotators disagree in 0.3513 of cases even for a face at the centre of an ideal band, and that with 12 measured ratios 0.7176 of people have at least one rare ratio. (https://5dlibrary.com/answers/en/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c05/) 第5章 特征点误差与巴纳姆效应 — 第五章显示 2 像素的标点误差使两位标注者对位于理想区间中点的脸也有 0.3513 的分歧,测量 12 个比例时 0.7176 的人至少有一个罕见比例。 (https://5dlibrary.com/answers/zh/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c05/) - Ch.6 The Royal Cubit, Canons of Proportion and Evidence — Chapter 6 discusses the royal cubit and canons of proportion, the Bayes factor of testimonials (only 1.0667 per vague hit), and a false configuration effect manufactured by orientation in a grid city (z = 7.02, gone after stratification). (https://5dlibrary.com/answers/en/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c06/) 第6章 王室腕尺、比例法则与证据 — 第六章讨论王室腕尺与比例法则、见证的贝叶斯因子(笼统应验每次只有 1.0667),以及棋盘格城市中由朝向制造的虚假格局效应(z = 7.02,分层后消失)。 (https://5dlibrary.com/answers/zh/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c06/) - Ch.7 Shen Kuo on the Compass, and the Texts of Xuankong and Physiognomy — The program gives each method an illustrative reading error: 6 degrees floated on water, 4 on a fingernail or a bowl rim, 1.5 hung from silk. For a uniform facing, the floating needle names the wrong mountain with chance 0.3192 and the wrong palace with 0.1064; the hanging needle only 0.0798 and 0.0266. If "slightly east" is 3 degrees, 0.2000 of facings change mountain when it is not corrected. In choosing the silk thread Shen Kuo chose the smallest measurement error; his note on the eastward lean reminds every compass user to state which north is meant. (https://5dlibrary.com/answers/en/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c07/) 第7章 沈括论指南针与玄空、相法的文本 — 程序给四种方法各设一个示意的读数误差:水浮 6 度,指甲与碗沿各 4 度,丝悬 1.5 度。对均匀分布的朝向,水浮法读错山的概率为 0.3192,读错宫的概率为 0.1064;丝悬法分别只有 0.0798 与 0.0266。「微偏东」若为 3 度,未经改正就有 0.2000 的朝向换山。沈括选择丝悬,正是选择了最小的测量误差;他对偏东的记录,提醒每一位用罗盘的人说明以哪一个北为准。 (https://5dlibrary.com/answers/zh/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c07/) - Ch.8 A Ten-Flag Audit and a Pre-Registered Blind Paired Test — Chapter 8 gives a ten-flag audit and a pre-registered blind paired test (204 pairs at a true hit rate of 0.60). Marked boxes separate evidence, runnable Python programs, classical sources and hypotheses. (https://5dlibrary.com/answers/en/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c08/) 第8章 十项审计与一项预注册的盲配对检验 — 第八章给出十项自动审计与一项预注册的盲配对检验(真实命中率 0.60 时需 204 对)。书中以标注框区分证据、可运行的 Python 程序、原典与假说。 (https://5dlibrary.com/answers/zh/xuankong-flying-star-permutations-compass-discretization-craniofacial-ratios/c08/) ## 5DL-B-083 · Enterprise Bottleneck Diagnostics, SKU Cannibalization Audits and Automation Return Bounds 中文书名:企业瓶颈诊断、产品线自蚕食审计与自动化改造回报边界 DOI: https://doi.org/10.5281/zenodo.24090217 Cite: Fan, Y. (2026). Enterprise Bottleneck Diagnostics, SKU Cannibalization Audits and Automation Return Bounds (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090217 Book page: https://5dlibrary.com/books/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/ Scope: Team capacities, demand parameters, margins, noise and contract parameters are illustrative or synthetic; all data are generated by its programs; the book is not investment, legal or tax advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: Where is a company's bottleneck? Why does automation or staff added at a non-bottleneck not bring in more business? How do sub-line and budget products cannibalize the flagship's margin, and when does delisting SKUs raise profit? How should an outside consultant's "fixed base fee plus a share of the incremental profit" be designed? This book writes these rules of thumb as runnable arithmetic and then asks when they hold. Team capacities, demand parameters, margins, noise and contract parameters are illustrative or synthetic; all data are generated by its programs; the book is not investment, legal or tax advice. Chapter 1 writes content, acquisition, conversion and fulfilment as a Jackson queueing network: doubling the content team's capacity leaves the maximum throughput at 24.0 jobs a day, and only adding capacity at the conversion bottleneck raises it to 28.0. Chapter 2 corrects the claim of a "quadratic blow-up": work in progress pushed into the bottleneck grows along a hyperbola (49.00 jobs at a load of 0.98), and what grows quadratically is the number of communication channels; pushing in too much lets leads go cold, and sales fall from 11.06 to 5.22 a day; reducing a non-bottleneck's variability helps the bottleneck more than speeding it up (-22% against -5%). Chapter 3 computes diversion ratios with a nested logit model: delisting the lowest-contributing 30% of SKUs raises profit by 5.3% in the baseline model, mainly through demand diverted to higher-margin products. Chapter 4 writes down the margin dilution matrix and finds that for each unit the sub-line A3 no longer sells, RM 51.4 is recaptured, more than its own margin; over 400 random worlds, a 30% cut by naive contribution loses money in 218, so "cut 30%" is not a law. Chapter 5 solves for the share under double moral hazard (about 0.214) and shows that only annual reconciliation deters hiding the gain (audit probability 0.893). Chapter 6 compares linear and stepped shares with flat fees and computes the months needed to measure the gain. Chapter 7 reads the Guanzi's "Cheng Ma" and "Qing Zhong II", the market exchange in the History of Song's treatise on food and money, and the Han Feizi's "Ding Fa", correcting two attributions. Chapter 8 finds the bottleneck from event logs (named correctly with probability 0.976 after 30 days), and gives a contract ledger and a pre-registered switchback experiment. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:一家公司的瓶颈在哪里?在非瓶颈上投入自动化或人手为什么不会多做生意?产品线中的副线与低价品如何蚕食旗舰品的毛利,删减 SKU 什么时候能提高利润?外部顾问的「固定底座费 + 增量利润分成」应当怎样设计?本书把这些经验法则写成可以运行的算术,再看它们在什么条件下成立。书中的团队产能、需求参数、毛利、噪声与合约参数均为示意或合成模型;全部数据由书中程序生成;本书不是投资、法律或税务意见。 第一章把内容、投流、转化与履约写成杰克逊排队网络:把内容团队的产能翻倍,最大吞吐量仍是每天 24.0 件,只有转化瓶颈扩容才把它提高到 28.0 件。第二章改正「二次方膨胀」的说法:被推入瓶颈的在制品按双曲线增长(负荷 0.98 时 49.00 件),按二次方增长的是沟通通道;推入过多会让线索变冷,成交从 11.06 单降到 5.22 单;非瓶颈降低波动比提速更能帮助瓶颈(-22% 对 -5%)。 第三章用嵌套 Logit 模型计算转移分流率:砍掉贡献最低的三成 SKU 在基准模型中使利润上升 5.3%,主要来自转向高毛利商品的分流。第四章写出毛利稀释矩阵,发现副线 A3 每少卖一件被接回 RM 51.4,高于它自己的毛利;在 400 个随机世界中,按表面贡献砍三成有 218 个亏损,「砍三成」不是定律。第五章求解双边道德风险下的分成比例(约 0.214),并说明按年度合计核对才能威慑隐瞒增量(被审计概率 0.893)。第六章比较线性、阶梯分成与固定费,并计算测量增量所需的月数。第七章与《管子》乘马篇与轻重乙篇、《宋史·食货志》的市易、《韩非子·定法》并读,改正两处出处。第八章从事件日志找瓶颈(30 天认对的概率 0.976),给出合约账本与一项预注册的切换实验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A Queueing Network of Four Teams and Its Bottleneck — Chapter 1 writes content, acquisition, conversion and fulfilment as a Jackson queueing network: doubling the content team's capacity leaves the maximum throughput at 24.0 jobs a day, and only adding capacity at the conversion bottleneck raises it to 28.0. (https://5dlibrary.com/answers/en/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c01/) 第1章 四个团队的排队网络与瓶颈 — 第一章把内容、投流、转化与履约写成杰克逊排队网络:把内容团队的产能翻倍,最大吞吐量仍是每天 24.0 件,只有转化瓶颈扩容才把它提高到 28.0 件。 (https://5dlibrary.com/answers/zh/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c01/) - Ch.2 Work Pushed into the Bottleneck, Communication Channels and Variability — Chapter 2 corrects the claim of a "quadratic blow-up": work in progress pushed into the bottleneck grows along a hyperbola (49.00 jobs at a load of 0.98), and what grows quadratically is the number of communication channels; pushing in too much lets leads go cold, and sales fall from 11.06 to 5.22 a day; reducing a non-bottleneck's variability helps the bottleneck more than speeding it up (-22% against -5%). (https://5dlibrary.com/answers/en/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c02/) 第2章 推入瓶颈的工作、沟通通道与波动 — 第二章改正「二次方膨胀」的说法:被推入瓶颈的在制品按双曲线增长(负荷 0.98 时 49.00 件),按二次方增长的是沟通通道;推入过多会让线索变冷,成交从 11.06 单降到 5.22 单;非瓶颈降低波动比提速更能帮助瓶颈(-22% 对 -5%)。 (https://5dlibrary.com/answers/zh/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c02/) - Ch.3 Self-Cannibalization in a Product Line: Nested Logit and Diversion Ratios — Chapter 3 computes diversion ratios with a nested logit model: delisting the lowest-contributing 30% of SKUs raises profit by 5.3% in the baseline model, mainly through demand diverted to higher-margin products. (https://5dlibrary.com/answers/en/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c03/) 第3章 产品线自蚕食:嵌套 Logit 与转移分流率 — 第三章用嵌套 Logit 模型计算转移分流率:砍掉贡献最低的三成 SKU 在基准模型中使利润上升 5.3%,主要来自转向高毛利商品的分流。 (https://5dlibrary.com/answers/zh/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c03/) - Ch.4 The Margin Dilution Matrix and the Limits of "Cut 30%" — Chapter 4 writes down the margin dilution matrix and finds that for each unit the sub-line A3 no longer sells, RM 51.4 is recaptured, more than its own margin; over 400 random worlds, a 30% cut by naive contribution loses money in 218, so "cut 30%" is not a law. (https://5dlibrary.com/answers/en/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c04/) 第4章 毛利稀释矩阵与「砍三成」的边界 — 第四章写出毛利稀释矩阵,发现副线 A3 每少卖一件被接回 RM 51.4,高于它自己的毛利;在 400 个随机世界中,按表面贡献砍三成有 218 个亏损,「砍三成」不是定律。 (https://5dlibrary.com/answers/zh/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c04/) - Ch.5 A Fixed Base Fee plus a Share of the Gain: Contracts under Double Moral Hazard — Chapter 5 solves for the share under double moral hazard (about 0.214) and shows that only annual reconciliation deters hiding the gain (audit probability 0.893). (https://5dlibrary.com/answers/en/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c05/) 第5章 固定底座费加增量利润分成:双边道德风险下的契约 — 第五章求解双边道德风险下的分成比例(约 0.214),并说明按年度合计核对才能威慑隐瞒增量(被审计概率 0.893)。 (https://5dlibrary.com/answers/zh/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c05/) - Ch.6 Paying for Results: Stepped Shares, Attitudes to Risk and Measuring the Baseline — A linear share gives a steady incentive; a stepped share gives its strongest incentive near the threshold, almost none below it, and rewards risk-taking. Before sharing, the gain must be measured well: a before-and-after comparison often needs more than two years, and only a control with a correlation above 0.5 really shortens the measurement period. (https://5dlibrary.com/answers/en/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c06/) 第6章 按效计费:阶梯分成、风险偏好与基线测量 — 线性分成的激励稳定,阶梯分成在门槛附近激励最强、在门槛以下几乎为零,并且奖励冒险。分成之前要先把增量量准:只做前后比较往往需要两年以上,一个相关系数高于 0.5 的对照组才能真正缩短测量期。 (https://5dlibrary.com/answers/zh/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c06/) - Ch.7 The Guanzi's "Cheng Ma" and "Qing Zhong II", Song Market Exchange and the Monopoly Goods Bureau, and the Han Feizi's "Holding the Name to the Reality" — "Technique means granting office according to ability, following the name and demanding the reality, holding the power of life and death, and examining the abilities of the ministers; it is what the ruler holds." In business diagnosis the "name" is how a team presents itself at review, most often the hours each person logs per week; the "reality" is its true load in the flow of work. In the illustrative data the acquisition team logs the most hours (57 a week), while the real bottleneck is the conversion team, which logs the fewest (47 hours, load 0.833). Giving resources to the team that calls loudest adds no capacity at all (0.00). Meetings, rework and hand-offs fill hours too, so hours are the name and load is the reality. (https://5dlibrary.com/answers/en/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c07/) 第7章 《管子》乘马与轻重乙、宋代市易与榷货务、《韩非子》循名而责实 — 「循名而责实」是按职位的名分去考核实际的作为。用在企业诊断上,「名」是一个团队在例会上呈现的样子,最常见的就是每人每周登记的工时;「实」是它在工作流中的真实负荷。在示意数据中,登记工时最多的是投流团队(每周 57 小时),真正的瓶颈却是登记工时最少的转化团队(47 小时,负荷 0.833)。把资源给喊得最响的团队,产能一件也不会增加(0.00)。会议、返工与交接同样会填满工时,所以工时只是「名」,负荷才是「实」。 (https://5dlibrary.com/answers/zh/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c07/) - Ch.8 Finding the Bottleneck in the Logs, a Contract Ledger and a Pre-Registered Switchback Experiment — Chapter 8 finds the bottleneck from event logs (named correctly with probability 0.976 after 30 days), and gives a contract ledger and a pre-registered switchback experiment. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c08/) 第8章 从日志找瓶颈、合约账本与预注册的切换实验 — 第八章从事件日志找瓶颈(30 天认对的概率 0.976),给出合约账本与一项预注册的切换实验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/enterprise-bottleneck-diagnostics-sku-cannibalization-automation-roi/c08/) ## 5DL-B-084 · Conversational Closing Automata, Objection-Handling Information Gain and High-Ticket Sales Kinetics 中文书名:对话式成交自动机、异议消解信息增益与高客单销售动力学 DOI: https://doi.org/10.5281/zenodo.24090221 Cite: Fan, Y. (2026). Conversational Closing Automata, Objection-Handling Information Gain and High-Ticket Sales Kinetics (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090221 Book page: https://5dlibrary.com/books/conversational-closing-automata-objection-information-gain-high-ticket-sales/ Scope: Priors, answer probabilities, close rates, refund rates and costs are illustrative or synthetic; all data are generated by its programs; the book is not legal advice and offers no scripts for manipulating buyers. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: When a prospect asks in a private channel about a high-ticket offer at RM 9,800 or RM 29,800, what should the seller (or the AI speaking for the seller) ask first? How should "too expensive" be answered? When should payment be requested, and when should a human take over? This book writes these rules of thumb as runnable arithmetic and then asks when they hold. Priors, answer probabilities, close rates, refund rates and costs are illustrative or synthetic; all data are generated by its programs; the book is not legal advice and offers no scripts for manipulating buyers. Chapter 1 writes the buyer's budget and urgency as hidden states of a partially observable decision process and corrects an intuition: an early pitch is not empty of information (0.229 bits); its cost is the buyers lost after a refusal. Chapter 2 solves for the optimal question tree by dynamic programming: ask the situation question first, only 1.73 questions on average, worth 9.0% more than pitching at once; the full-information bound is RM 3,466. Chapter 3 decodes the four classic objections with Bayes' rule: answering by the posterior recovers 0.219 of the price, one script only 0.163; asking a clarifying question after "too expensive" raises the recovery by 9.9%. Chapter 4 finds that high pressure closes more (27.0 sales against 21.9) but, after refunds, chargebacks and bad reviews, nets RM 113,605 less per 100 objectors. Chapter 5 allocates human time among sources of leads with Gittins indices (an exploration bonus of +0.102 for a new source) and finds that handing over the "hottest" conversations uses only 55% of the humans' potential. Chapter 6 solves for a trust-threshold stopping rule: keep talking below a belief of 0.138, hand over between 0.138 and 0.500, let the AI ask for payment above 0.500; and staffs the human desk with the Erlang C formula. Chapter 7 reads the Guiguzi's "Fan Ying" and "Nei Jian" and the Zhanguo Ce's "Qin I", correcting the quotation to "follow his words, listen to his phrasing" (an open question, 0.509 bits). Chapter 8 gives a ledger of ten thousand conversations, a calibration check and a pre-registered experiment. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:一位潜在客户在私域询问 RM 9,800 或 RM 29,800 的高客单方案时,卖方(或替卖方说话的 AI)该先问什么?怎样回应「太贵了」?何时请求付款,何时交给人工?本书把这些经验法则写成可以运行的算术,再看它们在什么条件下成立。书中的先验、回答概率、成交率、退款率与成本均为示意或合成模型;全部数据由书中程序生成;本书不是法律意见,也不提供任何操纵买家的话术。 第一章把买家的预算与紧迫度写成部分可观测决策过程的隐藏状态,改正一个直觉:过早报价并非没有信息(0.229 比特),它的代价是被拒绝之后的流失。第二章用动态规划求解最优提问树:先问情境问题,平均只问 1.73 题,价值比直接报价高 9.0%;完全信息的上限是 RM 3,466。第三章用贝叶斯公式解码四大异议,按后验回应收回原价的 0.219,一套话术只有 0.163;「太贵了」先问澄清问题再回应,收回提高 9.9%。第四章发现高压话术多成交(27.0 单对 21.9 单),计入退款、拒付与差评之后每 100 位提出异议的买家净少 RM 113,605。 第五章用吉廷斯指数在广告来源之间分配人工时间(新来源的探索奖励 +0.102),并发现按「最热」交接只用上人工潜力的 55%。第六章求解信任阈值停时规则:信念低于 0.138 继续聊,0.138 至 0.500 交给人,0.500 以上由 AI 请求付款;并以爱尔朗 C 公式配置人工坐席。第七章与《鬼谷子》反应篇、内揵篇与《战国策·秦策一》并读,改正「因其言,听其辞」的引文(开放问题 0.509 比特)。第八章给出一万段对话的账本、概率校准的检查与一项预注册实验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 What the Buyer Hides, and How Much Each Question Tells — Chapter 1 writes the buyer's budget and urgency as hidden states of a partially observable decision process and corrects an intuition: an early pitch is not empty of information (0.229 bits); its cost is the buyers lost after a refusal. (https://5dlibrary.com/answers/en/conversational-closing-automata-objection-information-gain-high-ticket-sales/c01/) 第1章 买家隐藏的状态与每个问题的信息量 — 第一章把买家的预算与紧迫度写成部分可观测决策过程的隐藏状态,改正一个直觉:过早报价并非没有信息(0.229 比特),它的代价是被拒绝之后的流失。 (https://5dlibrary.com/answers/zh/conversational-closing-automata-objection-information-gain-high-ticket-sales/c01/) - Ch.2 Diagnostic Question Trees: Which Question, and How Many — Chapter 2 solves for the optimal question tree by dynamic programming: ask the situation question first, only 1.73 questions on average, worth 9.0% more than pitching at once; the full-information bound is RM 3,466. (https://5dlibrary.com/answers/en/conversational-closing-automata-objection-information-gain-high-ticket-sales/c02/) 第2章 诊断式提问树:问哪一题、问几题 — 第二章用动态规划求解最优提问树:先问情境问题,平均只问 1.73 题,价值比直接报价高 9.0%;完全信息的上限是 RM 3,466。 (https://5dlibrary.com/answers/zh/conversational-closing-automata-objection-information-gain-high-ticket-sales/c02/) - Ch.3 Decoding the Four Classic Objections with Bayes — Chapter 3 decodes the four classic objections with Bayes' rule: answering by the posterior recovers 0.219 of the price, one script only 0.163; asking a clarifying question after "too expensive" raises the recovery by 9.9%. (https://5dlibrary.com/answers/en/conversational-closing-automata-objection-information-gain-high-ticket-sales/c03/) 第3章 四大异议的贝叶斯解码 — 第三章用贝叶斯公式解码四大异议,按后验回应收回原价的 0.219,一套话术只有 0.163;「太贵了」先问澄清问题再回应,收回提高 9.9%。 (https://5dlibrary.com/answers/zh/conversational-closing-automata-objection-information-gain-high-ticket-sales/c03/) - Ch.4 The Books of High-Pressure Closing: Refunds, Chargebacks and Reputation — Chapter 4 finds that high pressure closes more (27.0 sales against 21.9) but, after refunds, chargebacks and bad reviews, nets RM 113,605 less per 100 objectors. (https://5dlibrary.com/answers/en/conversational-closing-automata-objection-information-gain-high-ticket-sales/c04/) 第4章 高压逼单的账:退款、拒付与声誉 — 第四章发现高压话术多成交(27.0 单对 21.9 单),计入退款、拒付与差评之后每 100 位提出异议的买家净少 RM 113,605。 (https://5dlibrary.com/answers/zh/conversational-closing-automata-objection-information-gain-high-ticket-sales/c04/) - Ch.5 When an AI Sales Agent Hands Over to a Human: Gittins Indices and Ranking by Uplift — Chapter 5 allocates human time among sources of leads with Gittins indices (an exploration bonus of +0.102 for a new source) and finds that handing over the "hottest" conversations uses only 55% of the humans' potential. (https://5dlibrary.com/answers/en/conversational-closing-automata-objection-information-gain-high-ticket-sales/c05/) 第5章 AI 销售智能体何时交给人:吉廷斯指数与增量排序 — 第五章用吉廷斯指数在广告来源之间分配人工时间(新来源的探索奖励 +0.102),并发现按「最热」交接只用上人工潜力的 55%。 (https://5dlibrary.com/answers/zh/conversational-closing-automata-objection-information-gain-high-ticket-sales/c05/) - Ch.6 Trust-Threshold Stopping Rules and Planning the Human Desk — Chapter 6 solves for a trust-threshold stopping rule: keep talking below a belief of 0.138, hand over between 0.138 and 0.500, let the AI ask for payment above 0.500; and staffs the human desk with the Erlang C formula. (https://5dlibrary.com/answers/en/conversational-closing-automata-objection-information-gain-high-ticket-sales/c06/) 第6章 信任阈值停时规则与人工坐席规划 — 第六章求解信任阈值停时规则:信念低于 0.138 继续聊,0.138 至 0.500 交给人,0.500 以上由 AI 请求付款;并以爱尔朗 C 公式配置人工坐席。 (https://5dlibrary.com/answers/zh/conversational-closing-automata-objection-information-gain-high-ticket-sales/c06/) - Ch.7 The Guiguzi's "Fan Ying" and "Nei Jian", and Su Qin's Persuasion of Qin in the Zhanguo Ce — Chapter 7 reads the Guiguzi's "Fan Ying" and "Nei Jian" and the Zhanguo Ce's "Qin I", correcting the quotation to "follow his words, listen to his phrasing" (an open question, 0.509 bits). (https://5dlibrary.com/answers/en/conversational-closing-automata-objection-information-gain-high-ticket-sales/c07/) 第7章 《鬼谷子》反应与内揵、《战国策》苏秦说秦 — 第七章与《鬼谷子》反应篇、内揵篇与《战国策·秦策一》并读,改正「因其言,听其辞」的引文(开放问题 0.509 比特)。 (https://5dlibrary.com/answers/zh/conversational-closing-automata-objection-information-gain-high-ticket-sales/c07/) - Ch.8 A Ledger of Ten Thousand Conversations, Probability Calibration and a Pre-Registered Experiment — Chapter 8 gives a ledger of ten thousand conversations, a calibration check and a pre-registered experiment. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/conversational-closing-automata-objection-information-gain-high-ticket-sales/c08/) 第8章 一万段对话的账本、概率校准与预注册实验 — 第八章给出一万段对话的账本、概率校准的检查与一项预注册实验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/conversational-closing-automata-objection-information-gain-high-ticket-sales/c08/) ## 5DL-B-085 · Cross-Border Payment Routing, Chargeback Defense Games and Treasury Ruin Bounds 中文书名:跨境支付智能路由、拒付抗辩博弈与资金池破产边界 DOI: https://doi.org/10.5281/zenodo.24090225 Cite: Fan, Y. (2026). Cross-Border Payment Routing, Chargeback Defense Games and Treasury Ruin Bounds (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090225 Book page: https://5dlibrary.com/books/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/ Scope: Chapter 3 writes friendly fraud as a game of deterrence: once the win probability reaches 0.75, false disputes vanish in the illustrative model; in an evidence package the "considered purchase" tick counts least. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: Should a cross-border order go by card or by local instant transfer? Below which line must the chargeback ratio be kept? How can "friendly fraud" be met with evidence? When a payment gateway withholds a reserve or freezes settlements, how much cash must a firm hold not to run dry? This book writes these rules of thumb as runnable arithmetic and then asks when they hold. Order volumes, event rates, margins, refund rates, freeze frequencies and win probabilities are illustrative or synthetic; all data are generated by its programs; card network rules are stated as found in public sources at the time of writing; the book is not legal, financial or compliance advice. Chapter 1 corrects where the red line comes from: 0.9% was the standard threshold of Visa's retired VDMP, while the current VAMP counts fraud reports and disputes together, with an Asia Pacific threshold of 1.5%; steering the riskiest orders to QR lowers the card ratio from 0.94% to 0.59%; chargebacks arrive late, so reacting to last month's ratio still leaves 2.74 months a year above the line, against 1.33 when predicting from scores at checkout. Chapter 2 shows that a ratio is a noisy count, and finds that refunding within 1 day brings 227 fewer disputes than within 14 days. Chapter 3 writes friendly fraud as a game of deterrence: once the win probability reaches 0.75, false disputes vanish in the illustrative model; in an evidence package the "considered purchase" tick counts least. Chapter 4 builds a hash-chained evidence log and points out that disputes that are won still count in the ratio. Chapter 5 uses the Cramér-Lundberg model to show that "zero ruin" cannot be bought: a 1% ruin probability needs RM 175,281, and a reserve rate above the net margin rate drains cash continuously. Chapter 6 gives an approximate formula for the minimum safe reserve and finds that the greatest risk is a gateway freeze: one gateway needs RM 3,114,683, four only RM 1,258,468. Chapter 7 reads the maritime levies in the History of Song's treatise on food and money and the secret codes and transfers of the Shanxi remittance banks, noting that the levy was a tax while a reserve is only a delay, and correcting the source of "zhuo ying ji xu". Chapter 8 gives a year's ledger, stress tests and a pre-registered experiment. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:跨境订单该走信用卡还是本地即时转账?拒付比率要压在哪条线下?怎样用证据对付「友好欺诈」?支付网关扣留保证金或冻结结算时,企业要留多少现金才不会断流?本书把这些经验法则写成可以运行的算术,再看它们在什么条件下成立。书中的订单量、事件率、毛利、退款率、冻结频率与胜诉概率均为示意或合成模型;全部数据由书中程序生成;卡组织规则以写作时的公开资料为准;本书不是法律、财务或合规意见。 第一章改正红线的来源:0.9% 是 Visa 已退役的 VDMP 标准门槛,现行的 VAMP 把欺诈报告与争议一起计算,亚太门槛是 1.5%;把风险最高的订单改走扫码,可以把卡通道比率从 0.94% 压到 0.59%;拒付会迟到,看上月比率再反应一年仍有 2.74 个月超线,在结账时按评分预测只有 1.33 个月。第二章说明比率是带噪声的计数,并发现 1 天内退款比 14 天少 227 笔争议。第三章把友好欺诈写成威慑博弈:胜诉概率达到 0.75 时虚假争议在示意模型中完全消失;证据包中「理性消费勾选」的分量最小。第四章构造哈希链证据日志,并指出赢下的争议仍计入比率。 第五章用克拉美-伦德伯格模型说明「零破产」买不到:1% 的破产概率需要 RM 175,281;保证金比例高于净毛利率时会持续消耗现金。第六章给出最小安全储备的近似公式,并发现最大的风险是网关冻结:一个网关需要 RM 3,114,683,分给四个网关只需 RM 1,258,468。第七章与《宋史·食货志》的市舶抽解、明清山西票号的密押与调拨并读,指出抽解是税而保证金只是延后,并改正「酌盈剂虚」的出处。第八章给出全年账本、压力测试与一项预注册实验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Multi-Acquirer Routing and the Chargeback Red Line — Chapter 1 corrects where the red line comes from: 0.9% was the standard threshold of Visa's retired VDMP, while the current VAMP counts fraud reports and disputes together, with an Asia Pacific threshold of 1.5%; steering the riskiest orders to QR lowers the card ratio from 0.94% to 0.59%; chargebacks arrive late, so reacting to last month's ratio still leaves 2.74 months a year above the line, against 1.33 when predicting from scores at checkout. (https://5dlibrary.com/answers/en/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c01/) 第1章 多通道路由与拒付红线 — 第一章改正红线的来源:0.9% 是 Visa 已退役的 VDMP 标准门槛,现行的 VAMP 把欺诈报告与争议一起计算,亚太门槛是 1.5%;把风险最高的订单改走扫码,可以把卡通道比率从 0.94% 压到 0.59%;拒付会迟到,看上月比率再反应一年仍有 2.74 个月超线,在结账时按评分预测只有 1.33 个月。 (https://5dlibrary.com/answers/zh/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c01/) - Ch.2 A Ratio Is a Noisy Count, and Refunds Must Beat Disputes — Chapter 2 shows that a ratio is a noisy count, and finds that refunding within 1 day brings 227 fewer disputes than within 14 days. (https://5dlibrary.com/answers/en/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c02/) 第2章 比率是带噪声的计数,退款要快于拒付 — 第二章说明比率是带噪声的计数,并发现 1 天内退款比 14 天少 227 笔争议。 (https://5dlibrary.com/answers/zh/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c02/) - Ch.3 The Defence Game against Friendly Fraud, and Scoring the Evidence — Chapter 3 writes friendly fraud as a game of deterrence: once the win probability reaches 0.75, false disputes vanish in the illustrative model; in an evidence package the "considered purchase" tick counts least. (https://5dlibrary.com/answers/en/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c03/) 第3章 友好欺诈的抗辩博弈与证据评分 — 第三章把友好欺诈写成威慑博弈:胜诉概率达到 0.75 时虚假争议在示意模型中完全消失;证据包中「理性消费勾选」的分量最小。 (https://5dlibrary.com/answers/zh/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c03/) - Ch.4 A Cryptographic Evidence Chain, and "Fight or Pre-Empt" — With a preparation cost of RM 15 and a win probability of 0.60, disputes above RM 25 are worth fighting; with RM 60 of manual preparation and a win probability of 0.35, the threshold rises to RM 171. For an illustrative merchant with 10,800 card transactions a month and a dispute rate of 0.60%, there are 64.8 disputes a month; if alerts stop 30% of them, those counted fall to 45.4 and the ratio to 0.42%, at RM 3,402 a month in alert refunds and fees; stopping 60% brings the ratio to 0.24% at RM 6,804 a month. (https://5dlibrary.com/answers/en/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c04/) 第4章 加密证据链与「抗辩还是预警」 — 准备成本 RM 15、胜诉概率 0.60 时,金额高于 RM 25 的争议值得抗辩;人工准备成本 RM 60、胜诉概率 0.35 时,门槛升到 RM 171。在每月 10,800 笔卡交易、争议率 0.60% 的示意商户中,每月有 64.8 笔争议;预警服务拦下 30% 时,计入比率的降到 45.4 笔,比率 0.42%,预警退款与服务费每月 RM 3,402;拦下 60% 时比率降到 0.24%,每月 RM 6,804。 (https://5dlibrary.com/answers/zh/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c04/) - Ch.5 The Cramér-Lundberg Model and the Rolling Reserve — Chapter 5 uses the Cramér-Lundberg model to show that "zero ruin" cannot be bought: a 1% ruin probability needs RM 175,281, and a reserve rate above the net margin rate drains cash continuously. (https://5dlibrary.com/answers/en/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c05/) 第5章 克拉美-伦德伯格模型与滚动保证金 — 第五章用克拉美-伦德伯格模型说明「零破产」买不到:1% 的破产概率需要 RM 175,281;保证金比例高于净毛利率时会持续消耗现金。 (https://5dlibrary.com/answers/zh/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c05/) - Ch.6 A Formula for the Minimum Safe Reserve, and Spreading over Gateways — Chapter 6 gives an approximate formula for the minimum safe reserve and finds that the greatest risk is a gateway freeze: one gateway needs RM 3,114,683, four only RM 1,258,468. (https://5dlibrary.com/answers/en/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c06/) 第6章 最小安全储备公式与多网关分散 — 第六章给出最小安全储备的近似公式,并发现最大的风险是网关冻结:一个网关需要 RM 3,114,683,分给四个网关只需 RM 1,258,468。 (https://5dlibrary.com/answers/zh/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c06/) - Ch.7 Song Maritime Customs Levies, and the Secret Codes and Transfers of the Shanxi Remittance Banks — A forger who has never seen a genuine draft guesses all six characters with a chance of 1 in 3,600,000. But the table is fixed: after seeing 10 genuine drafts, the probability that he already knows every character needed to forge a random draft is 0.155; after 50, 0.806; after 200, 0.998. The security of the code lies not in the table but in changing it regularly and keeping each draft in few hands. A modern keyed hash (truncated to 32 bits) is different: however many genuine documents the attacker has seen, his chance of guessing remains 1 in 4,294,967,296, provided the key itself does not leak. The evidence log of Chapter 4 works this way. (https://5dlibrary.com/answers/en/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c07/) 第7章 宋代市舶抽解与明清山西票号的密押、调拨 — 一个从未见过真汇票的伪造者,盲猜全部六个字的成功机会是 3,600,000 分之一。但字表是固定的:见过 10 张真汇票之后,他已经认得伪造一张随机汇票所需全部字的概率是 0.155;见过 50 张是 0.806;见过 200 张是 0.998。密押的安全不在字表本身,而在于定期更换字表、并让每张汇票只在少数人手中经过。现代的带密钥哈希(截短为 32 位)则不同:无论攻击者见过多少张真单据,猜中的机会都是 4,294,967,296 分之一,前提是密钥本身不外泄。第四章的证据日志用的就是这种做法。 (https://5dlibrary.com/answers/zh/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c07/) - Ch.8 A Year's Ledger, Stress Tests and a Pre-Registered Experiment — Chapter 8 gives a year's ledger, stress tests and a pre-registered experiment. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c08/) 第8章 全年账本、压力测试与预注册实验 — 第八章给出全年账本、压力测试与一项预注册实验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cross-border-payment-routing-chargeback-defense-treasury-ruin-bounds/c08/) ## 5DL-B-086 · Autonomic Sympathovagal Balance, Six-Meridian State-Space Models and Multimodal Biomarker Inversion 中文书名:自主神经交感迷走平衡、六经辨证状态空间与多模态标志物反演 DOI: https://doi.org/10.5281/zenodo.24090227 Cite: Fan, Y. (2026). Autonomic Sympathovagal Balance, Six-Meridian State-Space Models and Multimodal Biomarker Inversion (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090227 Book page: https://5dlibrary.com/books/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: Is LF/HF the balance between sympathetic and vagal activity? Can yang hyperactivity, yin deficiency, qi sinking and yang collapse be matched to autonomic states? Can the six-stage progression of the Shanghan Lun be written as a state space that wearables could track? Can a three-minute phone screen of questionnaire, camera pulse and voice tell mind-body states apart? This book writes these claims as runnable arithmetic and then asks when they hold. It is not medical advice and not a diagnostic tool; its physiological parameters, pattern mappings and population settings are illustrative or synthetic, all data are generated by its programs, and every correspondence between Chinese-medicine patterns and autonomic states is treated as a hypothesis. Chapter 1 shows that, without changing any autonomic setting, slowing breathing alone raises LF/HF from 0.33 to 85.14. Chapter 2 builds a two-compartment model of sympathetic and vagal activity: under sympathetic overload RMSSD can even rise, and five regimes are classified correctly 0.783 of the time from heart rate and RMSSD, reaching 0.876 only with recovery after a stressor. Chapter 3 computes the day-to-day wandering of the temperature-rhythm peak: one day of data has only a 0.652 chance of revealing a one-hour drift, seven days 0.836; if room temperature is not measured with the core-to-fingertip gradient, the error in working back almost triples. Chapter 4 writes the six stages as a hidden Markov model: with the model exactly right, days are decoded correctly 0.830 of the time, taiyang and shaoyang and shaoyin and jueyin are hardest to separate, and modelling "combined disease" on its own needs 21 states. Chapter 5 writes a three-modality screen as a Bayesian inverse problem: the error falls from 0.532 with the questionnaire alone to 0.298, but with correlated modalities naive Bayes claims 0.707 certainty while being right only 0.655 of the time. Chapter 6 shows that camera timing error inflates low RMSSD by 86%, and that at a prevalence of 2% the positive predictive value is only 0.098. Chapter 7 reads the six-stage outlines of the Shanghan Lun, the twenty-four pulses of Wang Shuhe's Maijing and the fifty circuits of the Lingshu, noting that the Maijing counts rapid and slow against the examiner's own breath. Chapter 8 gives the sample size of an accuracy study, test-retest reliability and a pre-registered paired comparison. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:LF/HF 是不是交感与迷走的平衡?阳亢、阴虚、气陷、亡阳能否对应几种自主神经状态?《伤寒论》的六经传变能否写成可以用可穿戴设备追踪的状态空间?手机三分钟的问卷、摄像头脉搏与嗓音,能否分出不同的身心状态?本书把这些说法写成可以运行的算术,再看它们在什么条件下成立。本书不是医疗建议,也不是诊断工具;书中的生理参数、证候映射与人群设定均为示意或合成模型,全部数据由书中程序生成;中医证候与自主神经状态的对应一律作为假说处理。 第一章说明不改变任何自主神经设定、只把呼吸放慢,LF/HF 就从 0.33 升到 85.14。第二章建立交感与迷走的双室模型:交感过载时 RMSSD 反而可以升高,五种状态只用心率与 RMSSD 的正确分类率是 0.783,加上应激恢复时间才到 0.876。第三章计算体温节律峰值的日间游走:一天的数据只有 0.652 的机会看出一小时的漂移,七天升到 0.836;核心至指尖温差不连同室温一起量,反推误差几乎增加三倍。第四章把六经写成隐马尔可夫模型:模型完全正确时按天解码的正确率是 0.830,太阳与少阳、少阴与厥阴最难分辨,「合病」单独建模需要 21 个状态。 第五章把三模态筛查写成贝叶斯反问题:误判率从单用问卷的 0.532 降到 0.298,但模态相关时朴素贝叶斯报出 0.707 的把握,实际只对 0.655。第六章说明摄像头计时误差会把低 RMSSD 放大 86%,而在患病率 2% 的人群中阳性预测值只有 0.098。第七章与《伤寒论》六经提纲、王叔和《脉经》的二十四脉与《灵枢·营卫生会》的五十周并读,指出《脉经》的数与迟以医者的呼吸为单位。第八章给出准确性研究的样本量、重测信度与一项预先登记的配对比较。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Heart-Rate Variability Spectrum and the Limits of "Sympathovagal Balance" — Chapter 1 shows that, without changing any autonomic setting, slowing breathing alone raises LF/HF from 0.33 to 85.14. (https://5dlibrary.com/answers/en/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c01/) 第1章 心率变异谱与「交感迷走平衡」的局限 — 第一章说明不改变任何自主神经设定、只把呼吸放慢,LF/HF 就从 0.33 升到 85.14。 (https://5dlibrary.com/answers/zh/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c01/) - Ch.2 A Two-Compartment Model of Sympathetic and Vagal Tone, and Five Regimes — Chapter 2 builds a two-compartment model of sympathetic and vagal activity: under sympathetic overload RMSSD can even rise, and five regimes are classified correctly 0.783 of the time from heart rate and RMSSD, reaching 0.876 only with recovery after a stressor. (https://5dlibrary.com/answers/en/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c02/) 第2章 交感与迷走的双室模型与五种状态 — 第二章建立交感与迷走的双室模型:交感过载时 RMSSD 反而可以升高,五种状态只用心率与 RMSSD 的正确分类率是 0.783,加上应激恢复时间才到 0.876。 (https://5dlibrary.com/answers/zh/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c02/) - Ch.3 Drift in the Temperature Rhythm and the Core-to-Fingertip Gradient — Chapter 3 computes the day-to-day wandering of the temperature-rhythm peak: one day of data has only a 0.652 chance of revealing a one-hour drift, seven days 0.836; if room temperature is not measured with the core-to-fingertip gradient, the error in working back almost triples. (https://5dlibrary.com/answers/en/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c03/) 第3章 体温节律漂移与核心至末梢温差 — 第三章计算体温节律峰值的日间游走:一天的数据只有 0.652 的机会看出一小时的漂移,七天升到 0.836;核心至指尖温差不连同室温一起量,反推误差几乎增加三倍。 (https://5dlibrary.com/answers/zh/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c03/) - Ch.4 The Six Stages as a Hidden Markov State Space — Chapter 4 writes the six stages as a hidden Markov model: with the model exactly right, days are decoded correctly 0.830 of the time, taiyang and shaoyang and shaoyin and jueyin are hardest to separate, and modelling "combined disease" on its own needs 21 states. (https://5dlibrary.com/answers/en/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c04/) 第4章 六经作为隐马尔可夫状态空间 — 第四章把六经写成隐马尔可夫模型:模型完全正确时按天解码的正确率是 0.830,太阳与少阳、少阴与厥阴最难分辨,「合病」单独建模需要 21 个状态。 (https://5dlibrary.com/answers/zh/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c04/) - Ch.5 A Three-Modality Screen as a Bayesian Inverse Problem — Chapter 5 writes a three-modality screen as a Bayesian inverse problem: the error falls from 0.532 with the questionnaire alone to 0.298, but with correlated modalities naive Bayes claims 0.707 certainty while being right only 0.655 of the time. (https://5dlibrary.com/answers/en/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c05/) 第5章 三模态快速筛查作为贝叶斯反问题 — 第五章把三模态筛查写成贝叶斯反问题:误判率从单用问卷的 0.532 降到 0.298,但模态相关时朴素贝叶斯报出 0.707 的把握,实际只对 0.655。 (https://5dlibrary.com/answers/zh/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c05/) - Ch.6 Limits of Measurement: Camera Jitter and the Predictive Value of a Screen — Chapter 6 shows that camera timing error inflates low RMSSD by 86%, and that at a prevalence of 2% the positive predictive value is only 0.098. (https://5dlibrary.com/answers/en/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c06/) 第6章 测量的极限:摄像头抖动与筛查的阳性预测值 — 第六章说明摄像头计时误差会把低 RMSSD 放大 86%,而在患病率 2% 的人群中阳性预测值只有 0.098。 (https://5dlibrary.com/answers/zh/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c06/) - Ch.7 The Six-Stage Outlines, the Pulse Forms of the Maijing and the Fifty Circuits of the Lingshu — The six-stage outlines summarise symptoms and pulses, and Chapter 4's measurement mapping is the author's translation, not the text; the Maijing's rapid and slow are counts relative to the examiner's breath, and turning them into beats per minute requires stating that breath; the Lingshu's fifty circuits correspond to a cycle of about 28.8 minutes within the day, a different time scale from the twenty-four-hour temperature rhythm. (https://5dlibrary.com/answers/en/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c07/) 第7章 六经提纲、《脉经》的脉形与《灵枢》的五十周 — 六经提纲是症状与脉象的归纳,第四章的测量映射是作者的翻译而非原文;《脉经》的数与迟是相对于医者呼吸的计数,换成每分钟心跳数时要说明医者的呼吸;《灵枢》的五十周对应约 28.8 分钟一周的日内循环,与二十四小时的体温节律属于不同时间尺度。 (https://5dlibrary.com/answers/zh/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c07/) - Ch.8 Practice: Accuracy Studies, Test-Retest Reliability and Pre-Registration — Chapter 8 gives the sample size of an accuracy study, test-retest reliability and a pre-registered paired comparison. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c08/) 第8章 实践:准确性研究、重测信度与预先登记 — 第八章给出准确性研究的样本量、重测信度与一项预先登记的配对比较。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/autonomic-sympathovagal-balance-six-meridian-state-space-biomarker-inversion/c08/) ## 5DL-B-087 · Affiliate Network Power Laws, Influencer Portfolio Optimization and Incentive Stability Bounds 中文书名:分销网络幂律激活、网红矩阵组合优化与激励稳定边界 DOI: https://doi.org/10.5281/zenodo.24090229 Cite: Fan, Y. (2026). Affiliate Network Power Laws, Influencer Portfolio Optimization and Incentive Stability Bounds (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090229 Book page: https://5dlibrary.com/books/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/ Scope: Its numbers of affiliates, sales distributions, conversion rates, customer counts, commission rates and recruitment speeds are illustrative or synthetic; all data are generated by its programs; the book is not legal, financial or investment advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: Do the best hundred of ten thousand affiliates really sell half? Whom does a first-sale activation bonus buy? How much does putting 20% of the budget on head creators and 80% on micro creators cut the swings in acquisition cost? Where is the line between single-tier and multi-level commission? This book writes these rules of thumb as runnable arithmetic and then asks when they hold. Its numbers of affiliates, sales distributions, conversion rates, customer counts, commission rates and recruitment speeds are illustrative or synthetic; all data are generated by its programs; the book is not legal, financial or investment advice. Chapter 1 shows that Price's law is not a constant: at a tail index of 1.5 the share of the top square root of N falls from 0.584 for a hundred people to 0.299 for ten thousand, and counting dormant registrants makes concentration look higher. Chapter 2 finds that 0.867 of a 20-ringgit first-sale bonus goes to people who would have sold anyway and that large bonuses clearly lose money; with the same budget, a sales contest yields 2,570 and pay by volume 3,023. Chapter 3 tests "20% head, 80% micro": with uncorrelated creators the variance falls by 0.833, at a correlation of 0.3 by only 0.333; the minimum-variance share is about 20%, but a year of data can hardly pin it down. Chapter 4 compares rules for learning while spending and shows that of three ways of computing CAC (723, 500 and 304 ringgit) only pooled CAC matches profit and loss. Chapter 5 computes the geometry of multi-level commission: a full twelve-level tree with five recruits each needs 305,175,781 people, and once a finite market saturates 0.724 of members have a net loss; it also sets out the line drawn by Malaysia's Act 500 and the US Federal Trade Commission's guidance. Chapter 6 follows the decay of an affiliate cohort, compares how three channels divide a sale, and shows how the share of income from recruiting can rise across the line as a network grows. Chapter 7 reads Tao Zhu Gong in the Shiji, the labour shares of the Shanxi merchants and the Wei Liaozi's "rewards are prized when they reach the small". Chapter 8 gives a year's ledger, concentration monitoring and a pre-registered activation experiment. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:一万名分销员里,最好的一百人真的卖出一半吗?首单激活奖励买到的是谁?二成预算给头部达人、八成给中小达人,获客成本的波动能降多少?单层分佣与多层分佣的界线在哪里?本书把这些经验法则写成可以运行的算术,再看它们在什么条件下成立。书中的分销员人数、销售分布、转化率、获客量、佣金比例与招募速度均为示意或合成模型;全部数据由书中程序生成;本书不是法律、财务或投资意见。 第一章说明普赖斯定律不是常数:尾指数 1.5 时,前根号 N 人的份额从一百人的 0.584 降到一万人的 0.299;把沉睡者算进分母会让集中度看起来更高。第二章发现 20 令吉的首单奖励中 0.867 发给了本来就会成交的人,大额奖励明显亏损;同样预算下,销售竞赛的产出是 2,570,按量计酬是 3,023。第三章检验「二成头部、八成中小」:达人之间不相关时方差降低 0.833,相关系数 0.3 时只降 0.333;最小方差份额约为二成,但一年的数据很难把它估准。第四章比较在投放中学习的规则,并指出 CAC 的三种算法(723、500、304 令吉)中只有合并 CAC 对应盈亏。 第五章计算多层分佣的几何:每人招五人的十二层满树需要 305,175,781 人,有限市场饱和后 0.724 的成员净亏损,并说明马来西亚第 500 号法令与美国联邦贸易委员会指引的界线。第六章追踪分销队列的衰减,比较三种渠道的分配,并说明来自招募的收入比例会随网络增长越过界线。第七章与《史记·货殖列传》的陶朱公、晋商身股与《尉缭子》「赏之贵小」并读。第八章给出全年账本、集中度监测与一项预先登记的激活实验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Heavy-Tailed Output in Affiliate Networks and Price's Law — Chapter 1 shows that Price's law is not a constant: at a tail index of 1.5 the share of the top square root of N falls from 0.584 for a hundred people to 0.299 for ten thousand, and counting dormant registrants makes concentration look higher. (https://5dlibrary.com/answers/en/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c01/) 第1章 分销网络的重尾产出与普赖斯定律 — 第一章说明普赖斯定律不是常数:尾指数 1.5 时,前根号 N 人的份额从一百人的 0.584 降到一万人的 0.299;把沉睡者算进分母会让集中度看起来更高。 (https://5dlibrary.com/answers/zh/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c01/) - Ch.2 First-Sale Activation Bonuses and Sales Contests — Chapter 2 finds that 0.867 of a 20-ringgit first-sale bonus goes to people who would have sold anyway and that large bonuses clearly lose money; with the same budget, a sales contest yields 2,570 and pay by volume 3,023. (https://5dlibrary.com/answers/en/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c02/) 第2章 首单激活奖励与销售竞赛 — 第二章发现 20 令吉的首单奖励中 0.867 发给了本来就会成交的人,大额奖励明显亏损;同样预算下,销售竞赛的产出是 2,570,按量计酬是 3,023。 (https://5dlibrary.com/answers/zh/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c02/) - Ch.3 Mean-Variance Allocation across a Creator Portfolio — Chapter 3 tests "20% head, 80% micro": with uncorrelated creators the variance falls by 0.833, at a correlation of 0.3 by only 0.333; the minimum-variance share is about 20%, but a year of data can hardly pin it down. (https://5dlibrary.com/answers/en/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c03/) 第3章 网红矩阵的均值方差配置 — 第三章检验「二成头部、八成中小」:达人之间不相关时方差降低 0.833,相关系数 0.3 时只降 0.333;最小方差份额约为二成,但一年的数据很难把它估准。 (https://5dlibrary.com/answers/zh/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c03/) - Ch.4 Learning While Spending: Thompson Sampling and the Tail of CAC — Chapter 4 compares rules for learning while spending and shows that of three ways of computing CAC (723, 500 and 304 ringgit) only pooled CAC matches profit and loss. (https://5dlibrary.com/answers/en/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c04/) 第4章 在投放中学习:汤普森抽样与 CAC 的尾部 — 第四章比较在投放中学习的规则,并指出 CAC 的三种算法(723、500、304 令吉)中只有合并 CAC 对应盈亏。 (https://5dlibrary.com/answers/zh/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c04/) - Ch.5 The Geometry of Multi-Level Commissions and the Saturation of Pyramids — Chapter 5 computes the geometry of multi-level commission: a full twelve-level tree with five recruits each needs 305,175,781 people, and once a finite market saturates 0.724 of members have a net loss; it also sets out the line drawn by Malaysia's Act 500 and the US Federal Trade Commission's guidance. (https://5dlibrary.com/answers/en/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c05/) 第5章 多层分佣的几何与金字塔的饱和 — 第五章计算多层分佣的几何:每人招五人的十二层满树需要 305,175,781 人,有限市场饱和后 0.724 的成员净亏损,并说明马来西亚第 500 号法令与美国联邦贸易委员会指引的界线。 (https://5dlibrary.com/answers/zh/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c05/) - Ch.6 Cohort Decay and the Wholesale Margin of Licensed Principals — After six months only 0.350 are still selling, yet revenue is still 0.600 of month one; after twelve months 0.122 of the people remain with 0.354 of the revenue; after twenty-four months only 0.025 remain, and revenue is still 0.140. The reason is that those who stay are mostly the strong: the initial top 1% of sellers bring 0.274 of revenue in month one and 0.457 by month twenty-four. So a programme's long-run revenue depends mainly on keeping the top few, not on the number activated. (https://5dlibrary.com/answers/en/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c06/) 第6章 分销队列的衰减与持牌主理人的批发差价 — 六个月后只剩 0.350 的人还在卖,但收入还有第一个月的 0.600;十二个月后人数剩 0.122,收入剩 0.354;二十四个月后只剩 0.025 的人,收入仍有 0.140。原因是留下来的多半是强者:最初前 1% 的销售者,在第一个月贡献 0.274 的收入,到第二十四个月贡献 0.457。这意味着一个分销计划的长期收入主要取决于能否留住顶尖的少数人,而不是激活人数。 (https://5dlibrary.com/answers/zh/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c06/) - Ch.7 Tao Zhu Gong's Giving Away, the Labour Shares of the Shanxi Merchants and the Rewards of the Wei Liaozi — With a profit noise standard deviation of 100, the best labour share is about 0.833; at 400 it is only 0.238, and the manager's effort falls with it from 8.33 to 2.38. The harder it is to attribute results to one person's effort, the less suitable it is to put all pay into a share. For an affiliate, commission on a single sale has little noise (a sale is a sale), so 50% can be given; for a licensed principal's annual profit share the noise is large, so the share should be lower and paired with a fixed wholesale margin. (https://5dlibrary.com/answers/en/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c07/) 第7章 陶朱公的散财、晋商的身股与《尉缭子》的赏罚 — 利润噪声的标准差为 100 时,最优身股比例约 0.833;标准差 400 时只有 0.238,经营者的努力也随之从 8.33 降到 2.38。业绩越难归因于个人努力,就越不适合把报酬全部押在分成上。对分销员而言,单件佣金的噪声小(卖出就算),所以可以给到 50%;对持牌主理人的年度分红,噪声大,分成比例应当低一些,并配合固定的批发差价。 (https://5dlibrary.com/answers/zh/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c07/) - Ch.8 Practice: A Year's Ledger, Monitoring Concentration and a Pre-Registered Activation Experiment — Chapter 8 gives a year's ledger, concentration monitoring and a pre-registered activation experiment. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c08/) 第8章 实践:全年账本、集中度监测与预先登记的激活实验 — 第八章给出全年账本、集中度监测与一项预先登记的激活实验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/affiliate-network-power-laws-influencer-portfolio-optimization-incentive-bounds/c08/) ## 5DL-B-088 · Slow-Crash Liquidity Reserves, Real-Asset Option Valuation and Intergenerational Wealth Preservation 中文书名:慢崩盘流动性储备、实物资产期权估值与跨代财富保全动力学 DOI: https://doi.org/10.5281/zenodo.24090231 Cite: Fan, Y. (2026). Slow-Crash Liquidity Reserves, Real-Asset Option Valuation and Intergenerational Wealth Preservation (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090231 Book page: https://5dlibrary.com/books/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/ Scope: Its model parameters are illustrative or synthetic; historical index data are used only to show the shape of declines; all calculations are generated by its programs; the book is not investment, legal or tax advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: Does a great crash really always take thirty months? Why does a bear who is right on direction still fail to profit? When, and at what discount, can a holder of cash sensibly buy distressed assets? How much can a family fund meant to last for generations take each year? This book writes these claims as runnable arithmetic and then asks when they hold. Its model parameters are illustrative or synthetic; historical index data are used only to show the shape of declines; all calculations are generated by its programs; the book is not investment, legal or tax advice. Chapter 1 checks several famous declines: the Dow fell 89.2% from 1929 to 1932 over 34.1 months and the Nasdaq took 31.0 months from 2000 to 2002, but the S&P 500 took only 17.0 months from 2007 to 2009; a model with damping, a debt maturity ladder and injections shows the mechanism of a slow crash. Chapter 2 finds that rolling puts all the way down leaves a median cumulative result of only 0.002 after the index has halved, and that forced sales cluster in the second half of a decline. Chapter 3 uses real options to show that the discount needed depends on competition: 0.606 if an offer never goes away, only 0.163 if it is lost to someone else once a year; buying at a further 30% off in month 27 gives a median value multiple of 1.85, above the 1.58 from buying at the true low. Chapter 4 compares four rules for spending a cash reserve and shows that when industry buyers are short of cash in a crisis the discount can reach 0.443. Chapter 5 uses Peters' coin to show an ensemble average of +0.050 a round against time-average growth of -0.0513, and shows how crash years make the Kelly leverage computed under a normal law dangerous. Chapter 6 uses Merton's rule and a hundred-year simulation to show that a fixed withdrawal of 4 runs the fund out with chance 0.926, and that only 1.5% of the three-year average brings this down to 0.024. Chapter 7 reads Bai Gui and Ji Ran in the Shiji, the Guanzi's "four autumns" and the rules of the Fan clan estate, and corrects the source of "when others discard, I take". Chapter 8 gives a cash runway and historical stress tests, and shows that four declines can only place the typical duration between 18.7 and 39.6 months. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:大崩盘真的总要走三十个月吗?看对方向的空头为什么仍然赚不到钱?持有现金的一方何时、以多大的折价收购困境资产才合理?一笔要延续几代人的家族基金每年能提取多少?本书把这些说法写成可以运行的算术,再看它们在什么条件下成立。书中的模型参数均为示意或合成模型;历史指数数据只用于说明下跌的形态;全部计算由书中程序生成;本书不是投资、法律或税务建议。 第一章核对几次著名下跌:道琼斯 1929 至 1932 年跌 89.2%,用了 34.1 个月,纳斯达克 2000 至 2002 年 31.0 个月,但标普 500 在 2007 至 2009 年只用了 17.0 个月;并用一个带阻尼、债务到期阶梯与注资的模型说明慢崩盘的机制。第二章发现一路滚动买入看跌期权,在指数跌掉一半之后累计盈亏的中位数只有 0.002,而被迫卖出集中在下跌的后半段。第三章用实物期权说明所需折价取决于竞争:报价永不消失时是 0.606,每年被别人拿走一次时只需 0.163;在第 27 个月额外折价三成买入的价值倍数中位数是 1.85,高于在真正低点买入的 1.58。第四章比较现金储备的四种投入规则,并说明危机中同行缺现金时折价可达 0.443。 第五章用 Peters 的硬币说明群体平均每轮 +0.050 而时间平均是 -0.0513,并说明崩盘年会让正态分布下的凯利杠杆变得危险。第六章用默顿规则与一百年的模拟说明固定提取 4 的耗尽机会是 0.926,按三年平均值的 1.5% 提取才降到 0.024。第七章与《史记·货殖列传》的白圭、计然,《管子·轻重乙》的「岁有四秋」与范氏义庄的规矩并读,并改正「人弃我取」的出处。第八章给出现金跑道、历史压力测试,并说明四次下跌只能把典型持续时间定在 18.7 至 39.6 个月之间。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Kinetics of a Slow Crash: Where Thirty Months Come From — Chapter 1 checks several famous declines: the Dow fell 89.2% from 1929 to 1932 over 34.1 months and the Nasdaq took 31.0 months from 2000 to 2002, but the S&P 500 took only 17.0 months from 2007 to 2009; a model with damping, a debt maturity ladder and injections shows the mechanism of a slow crash. (https://5dlibrary.com/answers/en/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c01/) 第1章 慢崩盘的动力学:三十个月从哪里来 — 第一章核对几次著名下跌:道琼斯 1929 至 1932 年跌 89.2%,用了 34.1 个月,纳斯达克 2000 至 2002 年 31.0 个月,但标普 500 在 2007 至 2009 年只用了 17.0 个月;并用一个带阻尼、债务到期阶梯与注资的模型说明慢崩盘的机制。 (https://5dlibrary.com/answers/zh/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c01/) - Ch.2 Bear-Market Rallies, the Time Decay of Shorts, and the Timetable of Forced Sellers — Chapter 2 finds that rolling puts all the way down leaves a median cumulative result of only 0.002 after the index has halved, and that forced sales cluster in the second half of a decline. (https://5dlibrary.com/answers/en/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c02/) 第2章 熊市反弹、空头的时间损耗与被迫卖出的时间表 — 第二章发现一路滚动买入看跌期权,在指数跌掉一半之后累计盈亏的中位数只有 0.002,而被迫卖出集中在下跌的后半段。 (https://5dlibrary.com/answers/zh/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c02/) - Ch.3 Real Options on Distressed Assets: How Deep a Discount Before Buying — Chapter 3 uses real options to show that the discount needed depends on competition: 0.606 if an offer never goes away, only 0.163 if it is lost to someone else once a year; buying at a further 30% off in month 27 gives a median value multiple of 1.85, above the 1.58 from buying at the true low. (https://5dlibrary.com/answers/en/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c03/) 第3章 困境资产的实物期权:要多大的折价才值得出手 — 第三章用实物期权说明所需折价取决于竞争:报价永不消失时是 0.606,每年被别人拿走一次时只需 0.163;在第 27 个月额外折价三成买入的价值倍数中位数是 1.85,高于在真正低点买入的 1.58。 (https://5dlibrary.com/answers/zh/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c03/) - Ch.4 Putting a Cash Reserve to Work in Stages, and Where Distress Discounts Come From — Chapter 4 compares four rules for spending a cash reserve and shows that when industry buyers are short of cash in a crisis the discount can reach 0.443. (https://5dlibrary.com/answers/en/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c04/) 第4章 现金储备怎样分批投入,与困境折价从哪里来 — 第四章比较现金储备的四种投入规则,并说明危机中同行缺现金时折价可达 0.443。 (https://5dlibrary.com/answers/zh/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c04/) - Ch.5 Ergodicity: Averages That Mislead — Chapter 5 uses Peters' coin to show an ensemble average of +0.050 a round against time-average growth of -0.0513, and shows how crash years make the Kelly leverage computed under a normal law dangerous. (https://5dlibrary.com/answers/en/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c05/) 第5章 遍历性:平均数会骗人 — 第五章用 Peters 的硬币说明群体平均每轮 +0.050 而时间平均是 -0.0513,并说明崩盘年会让正态分布下的凯利杠杆变得危险。 (https://5dlibrary.com/answers/zh/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c05/) - Ch.6 The Family Fund: Merton's Rule and a Dynamic Cap on Withdrawals — Chapter 6 uses Merton's rule and a hundred-year simulation to show that a fixed withdrawal of 4 runs the fund out with chance 0.926, and that only 1.5% of the three-year average brings this down to 0.024. (https://5dlibrary.com/answers/en/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c06/) 第6章 家族基金:默顿规则与动态提取上限 — 第六章用默顿规则与一百年的模拟说明固定提取 4 的耗尽机会是 0.926,按三年平均值的 1.5% 提取才降到 0.024。 (https://5dlibrary.com/answers/zh/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c06/) - Ch.7 Bai Gui's "Take What Others Discard", Ji Ran's "Stock Boats in Drought", the Guanzi's "Four Autumns" and the Fan Clan Estate — If demand turns after one year, the yearly return is 0.905; after two years 0.348, after three 0.203, and after six only 0.074. Profit against the cycle depends mainly on how long one waits, which ties directly to the length of the slow crash in Chapter 1: carrying costs eat most of the return on a long wait. (https://5dlibrary.com/answers/en/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c07/) 第7章 白圭的「人弃我取」、计然的「旱则资舟」、《管子》的「岁有四秋」与范氏义庄 — 等一年就反转,年化回报是 0.905;等两年是 0.348,三年 0.203,六年只有 0.074。逆周期的利润主要取决于等待多久,这与第一章慢崩盘的长度直接相关:持有成本会把漫长等待的回报吃掉大半。 (https://5dlibrary.com/answers/zh/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c07/) - Ch.8 Practice: The Cash Runway, Historical Stress Tests and the Problem of Too Few Cases — Chapter 8 gives a cash runway and historical stress tests, and shows that four declines can only place the typical duration between 18.7 and 39.6 months. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c08/) 第8章 实践:现金跑道、历史压力测试与样本太少的问题 — 第八章给出现金跑道、历史压力测试,并说明四次下跌只能把典型持续时间定在 18.7 至 39.6 个月之间。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/slow-crash-liquidity-reserves-real-asset-options-intergenerational-wealth/c08/) ## 5DL-B-089 · Dream Motif Bipartite Graphs, Deterministic Oracle Decks and Subconscious Prior Activation 中文书名:梦境意象二分图、确定性哈希牌卡与潜意识先验激活 DOI: https://doi.org/10.5281/zenodo.24090233 Cite: Fan, Y. (2026). Dream Motif Bipartite Graphs, Deterministic Oracle Decks and Subconscious Prior Activation (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090233 Book page: https://5dlibrary.com/books/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/ Scope: It does not claim that dreams or cards foretell anything; the motif graph, event rates, memory rates and behavioural parameters are illustrative or synthetic; all data are generated by its programs; the book is not medical or psychological advice. In Chapter 6 an illustrative model shows that a card helps with action paralysis (the chance of starting work rises from 0.584 to 0.740) exactly as much as a coin, and less than an if-then plan. Abstract: When a dream "comes true", is the subconscious sending a message? Is a card reading seeded from the user's question made for that person? Do the coincidences that follow a daily card show that it works? This book treats dream dictionaries, oracle decks and coincidences as objects of graph theory, cryptographic shuffling and cognitive psychology, writes these claims as runnable arithmetic, and then asks when they hold. It does not claim that dreams or cards foretell anything; the motif graph, event rates, memory rates and behavioural parameters are illustrative or synthetic; all data are generated by its programs; the book is not medical or psychological advice. Chapter 1 writes dream dictionaries and Hall-Van de Castle dream coding as a weighted bipartite graph of motifs and life domains: with loose readings allowed, 0.942 of motif pairs are neighbours, and a spreading-activation toy in the spirit of the REM findings ends far from its start in 0.570 of chains. Chapter 2 computes the base rates of direct and reverse dreams: accepting both, a dream comes true by chance 0.918 of the time within a week, and a month-long verdict carries 0.000007 bits; day residue meeting a weekly routine makes residues from exactly seven days earlier come true the next day 0.735 of the time. Chapter 3 builds a 54-card deck from SHA-256 and the Fisher-Yates shuffle; its 148,824 three-card spreads pass a chi-square test over 54,000 users, and three common slips are shown. Chapter 4 shows that one changed character gives a new reading, and that 455 users have an even chance of sharing a three-card spread. Chapter 5 counts by Littlewood's law: a daily card fits at least one event on 0.395 of days by chance, and selective memory turns this into 0.746. In Chapter 6 an illustrative model shows that a card helps with action paralysis (the chance of starting work rises from 0.584 to 0.740) exactly as much as a coin, and less than an if-then plan. Chapter 7 reads the six dreams of the Rites of Zhou, Wang Fu's ten kinds of dream, the Dunhuang dream book and Jung's synchronicity, noting that the six dreams describe the dreamer's state. Chapter 8 gives an audit method and a pre-registered trial. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:梦「应验」了,是潜意识在传话吗?用用户的问题作种子洗出的牌,是为这个人量身定做的吗?每日一张牌之后接连出现的巧合,说明它灵验吗?本书把解梦词典、神谕牌卡与巧合当作图论、密码学洗牌与认知心理学的研究对象,把这些说法写成可以运行的算术,再看它们在什么条件下成立。本书不主张梦或牌能预示任何事;意象图、事件率、记忆率与行为参数均为示意或合成模型;全部数据由书中程序生成;本书不是医疗或心理建议。 第一章把解梦词典与霍尔-范德卡斯尔梦境编码写成意象与生活领域的加权二分图:允许宽松解读后,0.942 的意象对互为邻居;仿照快速眼动期研究的语义扩散玩具模型有 0.570 的联想链终止在远处。第二章计算直梦与反梦的基率:同时接受直接与反向应验,一周之内凭偶然就有 0.918 的应验率,一个月内的判定只携带 0.000007 比特信息;日间残留遇上每周作息,恰好七天前的残留第二天应验 0.735。第三章用 SHA-256 与费希尔-耶茨洗牌构造五十四张牌,148,824 种三张牌排列在 54,000 个用户上通过卡方检验,并列出三种常见错法。第四章说明改一个字符就换一副牌,455 名用户就有一半机会撞上相同的三张牌。 第五章按利特尔伍德定律计算:每日一张牌凭偶然在 0.395 的日子里吻合至少一件事,选择性记忆把它变成 0.746。第六章的示意模型显示,牌对「行动瘫痪」的帮助(开始工作的概率从 0.584 到 0.740)与硬币相同,且不如「如果-就」计划。第七章读《周礼》六梦、王符《潜夫论·梦列》、敦煌《新集周公解梦书》与荣格的共时性,指出六梦描述的是做梦者的状态。第八章给出审计方法与一项预注册试验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Dream Motifs as a Bipartite Graph, and Spreading Activation in REM Sleep — Chapter 1 writes dream dictionaries and Hall-Van de Castle dream coding as a weighted bipartite graph of motifs and life domains: with loose readings allowed, 0.942 of motif pairs are neighbours, and a spreading-activation toy in the spirit of the REM findings ends far from its start in 0.570 of chains. (https://5dlibrary.com/answers/en/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c01/) 第1章 梦境意象二分图与快速眼动期的语义扩散 — 第一章把解梦词典与霍尔-范德卡斯尔梦境编码写成意象与生活领域的加权二分图:允许宽松解读后,0.942 的意象对互为邻居;仿照快速眼动期研究的语义扩散玩具模型有 0.570 的联想链终止在远处。 (https://5dlibrary.com/answers/zh/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c01/) - Ch.2 Reverse Dreams, Direct Dreams and Day Residue — Chapter 2 computes the base rates of direct and reverse dreams: accepting both, a dream comes true by chance 0.918 of the time within a week, and a month-long verdict carries 0.000007 bits; day residue meeting a weekly routine makes residues from exactly seven days earlier come true the next day 0.735 of the time. (https://5dlibrary.com/answers/en/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c02/) 第2章 反梦、直梦与日间残留 — 第二章计算直梦与反梦的基率:同时接受直接与反向应验,一周之内凭偶然就有 0.918 的应验率,一个月内的判定只携带 0.000007 比特信息;日间残留遇上每周作息,恰好七天前的残留第二天应验 0.735。 (https://5dlibrary.com/answers/zh/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c02/) - Ch.3 A Deterministic 54-Card Deck: SHA-256 and the Fisher-Yates Shuffle — Chapter 3 builds a 54-card deck from SHA-256 and the Fisher-Yates shuffle; its 148,824 three-card spreads pass a chi-square test over 54,000 users, and three common slips are shown. (https://5dlibrary.com/answers/en/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c03/) 第3章 确定性的五十四张牌:SHA-256 与费希尔-耶茨洗牌 — 第三章用 SHA-256 与费希尔-耶茨洗牌构造五十四张牌,148,824 种三张牌排列在 54,000 个用户上通过卡方检验,并列出三种常见错法。 (https://5dlibrary.com/answers/zh/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c03/) - Ch.4 Avalanche, Collisions and How Unique a Reading Is — Chapter 4 shows that one changed character gives a new reading, and that 455 users have an even chance of sharing a three-card spread. (https://5dlibrary.com/answers/en/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c04/) 第4章 雪崩效应、碰撞与一次解读有多独特 — 第四章说明改一个字符就换一副牌,455 名用户就有一半机会撞上相同的三张牌。 (https://5dlibrary.com/answers/zh/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c04/) - Ch.5 Littlewood's Law and Selective Memory — Chapter 5 counts by Littlewood's law: a daily card fits at least one event on 0.395 of days by chance, and selective memory turns this into 0.746. In Chapter 6 an illustrative model shows that a card helps with action paralysis (the chance of starting work rises from 0.584 to 0.740) exactly as much as a coin, and less than an if-then plan. (https://5dlibrary.com/answers/en/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c05/) 第5章 利特尔伍德定律与选择性记忆 — 第五章按利特尔伍德定律计算:每日一张牌凭偶然在 0.395 的日子里吻合至少一件事,选择性记忆把它变成 0.746。 (https://5dlibrary.com/answers/zh/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c05/) - Ch.6 The Card as an Action Anchor — In an illustrative model, a daily card followed 80% of the time raises the chance of starting a work block from 0.584 to 0.740, the same as a coin, and less than an if-then plan (0.851). A random pointer is a good rule when the options are nearly equal and a bad one when they differ. (https://5dlibrary.com/answers/en/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c06/) 第6章 作为行动锚点的牌 — 在示意模型中,每日一张牌、80% 的时候被遵从,把一个工作时段开始的概率从 0.584 提高到 0.740,与硬币相同,且低于「如果-就」计划(0.851)。随机指针在选项几乎相等时是好规则,在选项有差别时是坏规则。 (https://5dlibrary.com/answers/zh/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c06/) - Ch.7 The Six Dreams of the Rites of Zhou, Wang Fu's Ten Kinds of Dream, the Dunhuang Dream Book and Jung's Synchronicity — An analyst with 1,000 sessions a year for 25 years hears about 100,000 motif-hours of dreams. If the chance that a given motif turns up in the room or at the window during the hour is 1 in 20,000, about 5.00 such events are expected in one career and the probability of at least one is 0.993; at 1 in 1,000,000 the probability is still 0.095, and among 1,000 analysts about 100 such stories arise. A single remarkable case does not distinguish an acausal principle from a large number of opportunities. (https://5dlibrary.com/answers/en/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c07/) 第7章 《周礼》六梦、王符「十梦」、敦煌梦书与荣格的共时性 — 一位分析师每年 1,000 次会谈,持续 25 年,大约听到 100,000 个「意象-小时」。如果某个意象在那一小时里出现在室内或窗外的概率是 20,000 分之一,一个职业生涯里预计有 5.00 次这样的事件,至少发生一次的概率是 0.993;即使是 1,000,000 分之一,概率仍有 0.095,而在 1,000 位分析师中会出现约 100 个这样的故事。单独一个惊人案例,无法把非因果原则与大量的机会区分开来。 (https://5dlibrary.com/answers/zh/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c07/) - Ch.8 Auditing a Card App and a Pre-Registered Trial — Chapter 8 gives an audit method and a pre-registered trial. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c08/) 第8章 牌卡应用的审计与预注册试验 — 第八章给出审计方法与一项预注册试验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/dream-motif-bipartite-graphs-deterministic-oracle-decks-subconscious-priors/c08/) ## 5DL-B-090 · One-Person Enterprise Automation, Avatar Livestream Funnels and Task-Exposure Arbitrage 中文书名:一人公司自动化架构、数字人直播漏斗与任务级人工智能套利 DOI: https://doi.org/10.5281/zenodo.24090235 Cite: Fan, Y. (2026). One-Person Enterprise Automation, Avatar Livestream Funnels and Task-Exposure Arbitrage (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090235 Book page: https://5dlibrary.com/books/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/ Scope: Revenues, hours, costs, conversion and exit rates, refund rates and job-loss hazards are illustrative or synthetic; all data are generated by its programs; the book is not financial, legal, tax or career advice, and promises no income. Chapter 3 writes a looping run-of-show as a Markov chain: with illustrative exit rates the best cycle is about 4.0 minutes, with 32.0 orders per 1,000 viewers. Abstract: Can AI agents shrink a firm to one person? Where is the boundary of "zero employees, a million in revenue"? Does a 24-hour digital avatar outsell a human host? How should break-even ROAS for paid traffic be computed? When should someone whose job is exposed to AI move income into a business of their own? This book writes these questions as runnable arithmetic and then asks when the popular answers hold. Revenues, hours, costs, conversion and exit rates, refund rates and job-loss hazards are illustrative or synthetic; all data are generated by its programs; the book is not financial, legal, tax or career advice, and promises no income. Chapter 1 builds a production model whose coordination costs grow as n(n-1)/2: at an AI share of 0.80 the team shrinks from 21 people to 4, but with 0.05 review hours per AI hour a one-person firm with a million in revenue needs an AI share of at least 0.977, and at 0.10 it is out of reach. Chapter 2 shows that errors compound along agent chains (10 steps right 0.95 of the time deliver 0.599), and that the founder answering escalations is a queue that saturates. Chapter 3 writes a looping run-of-show as a Markov chain: with illustrative exit rates the best cycle is about 4.0 minutes, with 32.0 orders per 1,000 viewers. Chapter 4 derives the exact break-even ROAS, which is 3.31 where the short rule gives 4.55; an avatar converting 0.6 as well as a human beats the human even at the peak, but not below 0.551. Chapter 5 writes an occupation as a graph of tasks, with an exposure score of 0.46 for an illustrative marketing manager, and runs the Acemoglu-Autor task model. Chapter 6 compares transition paths: quitting now runs out of cash with probability 0.399, while quitting after the side business has covered living costs for 3 months gives the best mean. Chapter 7 reads Bai Gui in the Records of the Grand Historian, the hundred crafts of the Kaogongji and Coase's Nature of the Firm, noting that "lean troops and simple administration" is not in the Records and that "take what others discard" works only where prices swing back. Chapter 8 gives a year's ledger and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:AI 智能体能把企业缩小到一个人吗?「零员工、百万营收」的边界在哪里?24 小时数字人直播比真人卖得多吗?投流的盈亏平衡 ROAS 怎样算?工作暴露于 AI 的人,什么时候该把收入转到自己的生意上?本书把这些问题写成可以运行的算术,再看流行的答案在什么条件下成立。书中的营收、工时、成本、转化率、离开率、退款率与失业风险均为示意或合成模型;全部数据由书中程序生成;本书不是财务、法律、税务或职业建议,也不承诺任何收入。 第一章建立协调成本按 n(n-1)/2 增长的生产模型:AI 比例 0.80 时团队从 21 人缩到 4 人,但每个 AI 小时审核 0.05 小时时,百万营收的一人公司要求 AI 比例至少 0.977,审核 0.10 小时则无法实现。第二章说明误差沿智能体链条累积(10 步、每步 0.95 只交付 0.599),处理上交的创始人是一个会饱和的排队系统。第三章把循环话术写成马尔可夫链:在示意离开率下最佳周期约 4.0 分钟,每 1,000 名观众 32.0 单。第四章推导精确的盈亏平衡 ROAS,简式在精确值 3.31 时给出 4.55;转化率为真人 0.6 的数字人即使在高峰也胜过真人,低于 0.551 则不然。 第五章把职业写成任务图:示意的营销经理暴露度 0.46,并运行阿西莫格鲁-奥托任务模型。第六章比较过渡路径:现在辞职有 0.399 的概率现金耗尽,等副业连续 3 个月覆盖生活费再辞职平均结果最好。第七章读《史记·货殖列传》的白圭、《考工记》的百工与科斯《企业的性质》,指出「精兵简政」不在《史记》里,「人弃我取」只在价格会回归的市场有效。第八章给出全年账本与一项预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Coase's Boundary Reversed: Team Size, Coordination and the One-Person Ceiling — Chapter 1 builds a production model whose coordination costs grow as n(n-1)/2: at an AI share of 0.80 the team shrinks from 21 people to 4, but with 0.05 review hours per AI hour a one-person firm with a million in revenue needs an AI share of at least 0.977, and at 0.10 it is out of reach. (https://5dlibrary.com/answers/en/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c01/) 第1章 科斯边界的逆转:团队规模、协调成本与一人公司的上限 — 第一章建立协调成本按 n(n-1)/2 增长的生产模型:AI 比例 0.80 时团队从 21 人缩到 4 人,但每个 AI 小时审核 0.05 小时时,百万营收的一人公司要求 AI 比例至少 0.977,审核 0.10 小时则无法实现。 (https://5dlibrary.com/answers/zh/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c01/) - Ch.2 Agent Chains and the Founder's Queue — Chapter 2 shows that errors compound along agent chains (10 steps right 0.95 of the time deliver 0.599), and that the founder answering escalations is a queue that saturates. (https://5dlibrary.com/answers/en/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c02/) 第2章 智能体链条与创始人的排队 — 第二章说明误差沿智能体链条累积(10 步、每步 0.95 只交付 0.599),处理上交的创始人是一个会饱和的排队系统。 (https://5dlibrary.com/answers/zh/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c02/) - Ch.3 The Cyclic Run-of-Show as a Markov Chain — Chapter 3 writes a looping run-of-show as a Markov chain: with illustrative exit rates the best cycle is about 4.0 minutes, with 32.0 orders per 1,000 viewers. (https://5dlibrary.com/answers/en/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c03/) 第3章 循环话术的马尔可夫模型 — 第三章把循环话术写成马尔可夫链:在示意离开率下最佳周期约 4.0 分钟,每 1,000 名观众 32.0 单。 (https://5dlibrary.com/answers/zh/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c03/) - Ch.4 Break-Even ROAS and the 24-Hour Avatar — Chapter 4 derives the exact break-even ROAS, which is 3.31 where the short rule gives 4.55; an avatar converting 0.6 as well as a human beats the human even at the peak, but not below 0.551. (https://5dlibrary.com/answers/en/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c04/) 第4章 投流盈亏平衡 ROAS 与 24 小时数字人 — 第四章推导精确的盈亏平衡 ROAS,简式在精确值 3.31 时给出 4.55;转化率为真人 0.6 的数字人即使在高峰也胜过真人,低于 0.551 则不然。 (https://5dlibrary.com/answers/zh/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c04/) - Ch.5 Task Exposure and the Acemoglu-Autor Task Model — Chapter 5 writes an occupation as a graph of tasks, with an exposure score of 0.46 for an illustrative marketing manager, and runs the Acemoglu-Autor task model. (https://5dlibrary.com/answers/en/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c05/) 第5章 任务暴露度与阿西莫格鲁-奥托任务模型 — 第五章把职业写成任务图:示意的营销经理暴露度 0.46,并运行阿西莫格鲁-奥托任务模型。 (https://5dlibrary.com/answers/zh/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c05/) - Ch.6 Moving Income Across: Transition Paths and a Solo Portfolio — Chapter 6 compares transition paths: quitting now runs out of cash with probability 0.399, while quitting after the side business has covered living costs for 3 months gives the best mean. (https://5dlibrary.com/answers/en/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c06/) 第6章 现金流的转移:过渡路径与个体组合 — 第六章比较过渡路径:现在辞职有 0.399 的概率现金耗尽,等副业连续 3 个月覆盖生活费再辞职平均结果最好。 (https://5dlibrary.com/answers/zh/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c06/) - Ch.7 Bai Gui in the Records of the Grand Historian, the Division of Crafts in the Kaogongji, and Coase's Nature of the Firm — Bai Gui's "take what others discard" works where prices swing back and fails on a random walk; the frugality the Records praise is sharing hardship with workers, not having none. The Kaogongji's 30 crafts would need 435 channels without standards, and 30 with them. Coase's boundary moves toward one person when software becomes the cheapest way to carry out a transaction, and stops where the founder's review time binds. (https://5dlibrary.com/answers/en/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c07/) 第7章 《史记·货殖列传》的白圭、《考工记》的百工分职与科斯《企业的性质》 — 白圭的「人弃我取」在价格会回归的市场有效,在随机游走的市场失效;《史记》称道的节俭,是与做事的僮仆同甘共苦,而不是不用人。《考工记》的 30 门工匠,没有标准时需要 435 条渠道,有标准时只要 30 条。当软件成为完成交易最便宜的方式,科斯的边界就向一个人收缩,并在创始人的审核时间成为约束时止步。 (https://5dlibrary.com/answers/zh/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c07/) - Ch.8 A Year's Ledger and a Pre-Registered Test — Chapter 8 gives a year's ledger and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c08/) 第8章 全年账本与一项预注册检验 — 第八章给出全年账本与一项预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/one-person-enterprise-automation-avatar-livestream-funnels-task-exposure-arbitrage/c08/) ## 5DL-B-091 · Almanac Auspicious Boolean Lattices, Jianchu Twelve-Officer Congruences and Runic Combinatorics 中文书名:黄历择吉布尔格、建除十二直同余与卢恩符文组合学 DOI: https://doi.org/10.5281/zenodo.24090239 Cite: Fan, Y. (2026). Almanac Auspicious Boolean Lattices, Jianchu Twelve-Officer Congruences and Runic Combinatorics (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090239 Book page: https://5dlibrary.com/books/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/ Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: Can an expert almanac always find a perfectly clean day? Do the twelve officers simply cycle? Is a lucky date lucky again next year? Is a drawn rune reversed half the time? This book treats Chinese almanac day selection and modern rune casting as calendar algebra and combinatorial probability, writes these claims as runnable arithmetic, and then asks when they hold. It does not claim that any day is lucky or unlucky, or that runes foretell anything; stem-branch days, solar terms and lodges are computed from published astronomical data, while the choice of rules and the personal parameters are illustrative; all data are generated by its programs. Chapter 1 writes almanac rules as Boolean predicates on the days of 2027: stacking seven rule sets (officers, yellow and black roads, year breaker, three sha, unlucky lodges and two personal clashes) leaves 33 pure days (0.090), the share falling roughly as the product of the single-rule shares, and two schools built from the same components agree at a kappa of only 0.050. Chapter 2 writes the officers as a piecewise congruence that repeats at the twelve jie: a plain cycle is wrong on 336 days, and since Lidong 2027 falls at 23:39 in Beijing but after midnight in Tokyo, almanacs for Tokyo and Los Angeles differ from Beijing's on 1 and 8 days. Chapter 3 shows that stem-branch days and the 28 lodges have only 420 combinations, repeating every 420 days, and that a date receives the same verdict in two years only 0.685 of the time (0.619 by chance). Chapter 4 treats day selection as satisfiability: 0.496 of weeks have no acceptable day, and 0.690 of days are both good and bad. Chapter 5 derives the exact law of rune reversals: only 15 of the 24 Elder Futhark runes can be reversed, so a drawn rune is reversed with probability 0.3125, and a Norns reading shows no reversal with probability 0.320 rather than 0.125; Chapter 6 follows the consequences and the effect of the rune set. Chapter 7 reads the Shuihudi daybooks, the Qing imperial compilation on day selection and the Havamal, noting that the daybook already holds two officer systems and that any ranking of marks fixes the verdict on only 113 days. Chapter 8 gives an audit of almanac apps and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:高明的通书总能找到完全无凶的日子吗?建除十二神只是循环吗?今年的吉日明年还吉吗?抽出的卢恩符文一半时候是逆位吗?本书把中国通书的择日与现代卢恩抽签当作历法代数与组合概率来研究,把这些说法写成可以运行的算术,再看它们在什么条件下成立。本书不主张任何一天吉或凶,也不主张卢恩能预示任何事;干支、节气与星宿由公布的天文数据算出,规则的选取与人物参数为示意模型;全部数据由书中程序生成。 第一章把通书规则写成 2027 年各天上的布尔谓词:建除、黄黑道、岁破、三煞、凶宿与两人冲日七组规则叠加后只剩 33 个纯吉日(0.090),比例大致按单条规则的乘积下降;由相同成分构成的两个流派一致程度只有 kappa 0.050。第二章把建除写成在十二个节重复的分段同余式:不做重复的简单循环有 336 天出错,2027 年立冬在北京 23:39、在东京已是次日,所以东京与洛杉矶的通书与北京版分别有 1 天与 8 天不同。第三章说明干支日与二十八宿只有 420 种组合、每 420 天重复,同一日期两年判定相同的比例只有 0.685(偶然 0.619)。第四章把择日写成可满足性问题:0.496 的星期没有可用日,0.690 的日子既吉又凶。 第五章推导卢恩逆位的精确分布:古弗萨克 24 个符文中只有 15 个可逆,抽出的符文逆位概率是 0.3125,命运三女神阵没有逆位的概率是 0.320 而不是 0.125;第六章讨论推论与套组差异。第七章读睡虎地秦简《日书》、清代《协纪辨方书》与《高人的箴言》,指出《日书》已并列两套建除系统,任何标记排序只能确定 113 天的判定。第八章给出黄历应用的审计方法与一项预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Almanac Rules as Boolean Predicates, and the Vanishing Pure Day — Chapter 1 writes almanac rules as Boolean predicates on the days of 2027: stacking seven rule sets (officers, yellow and black roads, year breaker, three sha, unlucky lodges and two personal clashes) leaves 33 pure days (0.090), the share falling roughly as the product of the single-rule shares, and two schools built from the same components agree at a kappa of only 0.050. (https://5dlibrary.com/answers/en/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c01/) 第1章 黄历规则的布尔谓词与消失的纯吉日 — 第一章把通书规则写成 2027 年各天上的布尔谓词:建除、黄黑道、岁破、三煞、凶宿与两人冲日七组规则叠加后只剩 33 个纯吉日(0.090),比例大致按单条规则的乘积下降;由相同成分构成的两个流派一致程度只有 kappa 0.050。 (https://5dlibrary.com/answers/zh/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c01/) - Ch.2 The Twelve Officers as a Piecewise Congruence, and Which Day Is the Jie Day — Chapter 2 writes the officers as a piecewise congruence that repeats at the twelve jie: a plain cycle is wrong on 336 days, and since Lidong 2027 falls at 23:39 in Beijing but after midnight in Tokyo, almanacs for Tokyo and Los Angeles differ from Beijing's on 1 and 8 days. (https://5dlibrary.com/answers/en/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c02/) 第2章 建除十二直的分段同余与交节日的归属 — 第二章把建除写成在十二个节重复的分段同余式:不做重复的简单循环有 336 天出错,2027 年立冬在北京 23:39、在东京已是次日,所以东京与洛杉矶的通书与北京版分别有 1 天与 8 天不同。 (https://5dlibrary.com/answers/zh/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c02/) - Ch.3 Joint Periods: Stem-Branch Days, the Twenty-Eight Lodges and the Week — Chapter 3 shows that stem-branch days and the 28 lodges have only 420 combinations, repeating every 420 days, and that a date receives the same verdict in two years only 0.685 of the time (0.619 by chance). (https://5dlibrary.com/answers/en/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c03/) 第3章 联合周期:干支日、二十八宿与七曜 — 第三章说明干支日与二十八宿只有 420 种组合、每 420 天重复,同一日期两年判定相同的比例只有 0.685(偶然 0.619)。 (https://5dlibrary.com/answers/zh/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c03/) - Ch.4 Satisfiability: When No Day Is Acceptable, and When Every Day Is Both — Chapter 4 treats day selection as satisfiability: 0.496 of weeks have no acceptable day, and 0.690 of days are both good and bad. (https://5dlibrary.com/answers/en/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c04/) 第4章 可满足性:没有一天可用,与每一天都既吉又凶 — 第四章把择日写成可满足性问题:0.496 的星期没有可用日,0.690 的日子既吉又凶。 (https://5dlibrary.com/answers/zh/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c04/) - Ch.5 Reversed Runes: A Hypergeometric-Binomial Law — Chapter 5 derives the exact law of rune reversals: only 15 of the 24 Elder Futhark runes can be reversed, so a drawn rune is reversed with probability 0.3125, and a Norns reading shows no reversal with probability 0.320 rather than 0.125; (https://5dlibrary.com/answers/en/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c05/) 第5章 卢恩符文的逆位:超几何-二项复合分布 — 第五章推导卢恩逆位的精确分布:古弗萨克 24 个符文中只有 15 个可逆,抽出的符文逆位概率是 0.3125,命运三女神阵没有逆位的概率是 0.320 而不是 0.125; (https://5dlibrary.com/answers/zh/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c05/) - Ch.6 What a Reversal Can Say, and How Much the Rune Set Matters — A Norns reading shows no reversal with probability 0.3197, 2.56 times the shortcut. Reversals come mostly from the first and third aettir, because most second-aett runes are symmetric. Variants of the rune set move the numbers a little (mean 0.8750 to 1.0000), and none of them comes close to the shortcut. (https://5dlibrary.com/answers/en/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c06/) 第6章 逆位能说明什么,以及符文套组差异的影响 — 命运三女神阵没有逆位的概率是 0.3197,是简化算法的 2.56 倍。逆位大多来自第一族与第三族,因为第二族的大多数符文是对称的。套组的变体会使数字略有变化(均值 0.8750 到 1.0000),但没有一种接近简化算法。 (https://5dlibrary.com/answers/zh/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c06/) - Ch.7 The Shuihudi Daybooks, the Qing Imperial Compilation on Day Selection, and the Runes of Havamal — Chapter 7 reads the Shuihudi daybooks, the Qing imperial compilation on day selection and the Havamal, noting that the daybook already holds two officer systems and that any ranking of marks fixes the verdict on only 113 days. (https://5dlibrary.com/answers/en/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c07/) 第7章 睡虎地秦简《日书》、清代《协纪辨方书》与《埃达·高人的箴言》中的卢恩 — 第七章读睡虎地秦简《日书》、清代《协纪辨方书》与《高人的箴言》,指出《日书》已并列两套建除系统,任何标记排序只能确定 113 天的判定。 (https://5dlibrary.com/answers/zh/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c07/) - Ch.8 Auditing an Almanac App and a Pre-Registered Test — Chapter 8 gives an audit of almanac apps and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c08/) 第8章 黄历应用的审计与一项预注册检验 — 第八章给出黄历应用的审计方法与一项预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/almanac-auspicious-boolean-lattices-jianchu-congruences-runic-combinatorics/c08/) ## 5DL-B-092 · Relationship Rupture Survival Analysis, No-Contact Memory Decay and Reconciliation Game Windows 中文书名:关系断裂生存分析、断联期记忆衰减与复联博弈窗口 DOI: https://doi.org/10.5281/zenodo.24090241 Cite: Fan, Y. (2026). Relationship Rupture Survival Analysis, No-Contact Memory Decay and Reconciliation Game Windows (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090241 Book page: https://5dlibrary.com/books/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/ Scope: Its parameters are illustrative or synthetic; all calculations are generated by its programs; the book is not psychological counselling or therapy and not a manual for winning anyone back, and the other person's decision is theirs. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: How many former partners get back together after a rupture, and how soon? Why does frequent early contact so often backfire? Is there a "golden window" for reaching out? What signal of change can be believed? How long do reunions last? This book writes these claims as runnable arithmetic and then asks when they hold. Its parameters are illustrative or synthetic; all calculations are generated by its programs; the book is not psychological counselling or therapy and not a manual for winning anyone back, and the other person's decision is theirs. Chapter 1 treats reconciliation as survival analysis with a competing event: in a synthetic cohort, Kaplan-Meier puts the year-end reunion rate at 0.456 while the share that really reunited is only 0.311, and even if no couple's own chance changes, selection lowers the group hazard to 0.355 of its first value. Chapter 2 uses a two-exponential model (negative arousal fading over 14 to 21 days, positive memory over 60 to 90) to find the peak of net pull: between days 34.0 and 51.0 for typical values, but spread over days 19.5 to 67.0 for eight people in ten once individual differences are allowed. Chapter 3 adds a reactance ratchet: three contacts a week in the first month cut the chance that a calm message on day 60 is welcome from 0.803 to 0.017. Chapter 4 uses a signalling game and Bayes' rule to show that a happy post moves the belief that someone has changed only from 0.300 to 0.325, while four months of steady conduct move it to 0.696. Chapter 5 uses a mixture model to show that reunions are not doomed, but that each quick repeat of the cycle is strong evidence that the cause was not addressed (likelihood ratio 5.46). Chapter 6 uses the Gottman and Murray difference equations to show that the same couple can rest in a warm or a hostile state, and that the ratio of positive to negative interactions is not a threshold. Chapter 7 reads the Book of Songs poems "Mang" and "Gu feng" with "Bai tou yin", and notes that the attribution of "Bai tou yin" to Zhuo Wenjun comes from the Xijing zaji. Chapter 8 compares three ways of spending the year after a rupture and shows that neither a few weeks of self-ratings nor a few relationships suffice to estimate a person's own numbers. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:分手之后有多少前任复合,又有多快?为什么早期频繁联系常常适得其反?是否存在一个「黄金复联窗口」?什么样的改变信号才可信?复合之后能维持多久?本书把这些说法写成可以运行的算术,再看它们在什么条件下成立。书中的参数均为示意或合成;全部计算由书中程序生成;本书不是心理咨询或治疗,也不是挽回任何人的操作手册,对方的决定属于对方。 第一章把复合当作带竞争事件的生存分析:在合成队列中,Kaplan-Meier 估计年底复合率为 0.456,真正复合的比例只有 0.311;并说明即使每对伴侣自身的机会不变,群体风险率也会因筛选降到最初的 0.355。第二章用双指数模型(负面情绪 14 至 21 天消退,积极记忆 60 至 90 天消退)求出净牵引力的峰值:典型参数下在第 34.0 至 51.0 天,但考虑个体差异后十分之八散布在第 19.5 至 67.0 天。第三章加入抗拒棘轮:第一个月每周联系三次,会让第 60 天一条平静消息被接受的机会从 0.803 降到 0.017。第四章用信号博弈与贝叶斯法则说明,一条快乐的动态只把「已经改变」的信念从 0.300 提高到 0.325,四个月稳定的表现提高到 0.696。 第五章用混合模型说明复合并非注定失败,但每一次很快重演的分合都是原因未被处理的有力证据(似然比 5.46)。第六章用戈特曼与默里的差分方程说明同一对伴侣可以停在温暖或敌对的状态,而正负互动之比不是门槛。第七章与《诗经》的《氓》《谷风》及《白头吟》并读,并指出《白头吟》归于卓文君的说法来自《西京杂记》。第八章比较分手后一年的三种过法,并说明几周的自评与几段关系都不足以估计一个人自己的数字。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Reconciliation as a Survival Problem — Chapter 1 treats reconciliation as survival analysis with a competing event: in a synthetic cohort, Kaplan-Meier puts the year-end reunion rate at 0.456 while the share that really reunited is only 0.311, and even if no couple's own chance changes, selection lowers the group hazard to 0.355 of its first value. (https://5dlibrary.com/answers/en/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c01/) 第1章 把复合当作生存分析问题 — 第一章把复合当作带竞争事件的生存分析:在合成队列中,Kaplan-Meier 估计年底复合率为 0.456,真正复合的比例只有 0.311;并说明即使每对伴侣自身的机会不变,群体风险率也会因筛选降到最初的 0.355。 (https://5dlibrary.com/answers/zh/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c01/) - Ch.2 Two Memories Fading at Two Speeds — Chapter 2 uses a two-exponential model (negative arousal fading over 14 to 21 days, positive memory over 60 to 90) to find the peak of net pull: between days 34.0 and 51.0 for typical values, but spread over days 19.5 to 67.0 for eight people in ten once individual differences are allowed. (https://5dlibrary.com/answers/en/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c02/) 第2章 以两种速度消退的两种记忆 — 第二章用双指数模型(负面情绪 14 至 21 天消退,积极记忆 60 至 90 天消退)求出净牵引力的峰值:典型参数下在第 34.0 至 51.0 天,但考虑个体差异后十分之八散布在第 19.5 至 67.0 天。 (https://5dlibrary.com/answers/zh/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c02/) - Ch.3 The Reactance Ratchet: Why Early Pressure Backfires — Chapter 3 adds a reactance ratchet: three contacts a week in the first month cut the chance that a calm message on day 60 is welcome from 0.803 to 0.017. (https://5dlibrary.com/answers/en/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c03/) 第3章 抗拒的棘轮:为什么早期施压适得其反 — 第三章加入抗拒棘轮:第一个月每周联系三次,会让第 60 天一条平静消息被接受的机会从 0.803 降到 0.017。 (https://5dlibrary.com/answers/zh/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c03/) - Ch.4 Signals That Can Be Believed — Chapter 4 uses a signalling game and Bayes' rule to show that a happy post moves the belief that someone has changed only from 0.300 to 0.325, while four months of steady conduct move it to 0.696. (https://5dlibrary.com/answers/en/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c04/) 第4章 可信的信号 — 第四章用信号博弈与贝叶斯法则说明,一条快乐的动态只把「已经改变」的信念从 0.300 提高到 0.325,四个月稳定的表现提高到 0.696。 (https://5dlibrary.com/answers/zh/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c04/) - Ch.5 How Long Reunions Last — Chapter 5 uses a mixture model to show that reunions are not doomed, but that each quick repeat of the cycle is strong evidence that the cause was not addressed (likelihood ratio 5.46). (https://5dlibrary.com/answers/en/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c05/) 第5章 复合能维持多久 — 第五章用混合模型说明复合并非注定失败,但每一次很快重演的分合都是原因未被处理的有力证据(似然比 5.46)。 (https://5dlibrary.com/answers/zh/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c05/) - Ch.6 Two Resting States: Gottman's Difference Equations — Without repair, 0.184 of talks end in the hostile state; with repair succeeding 30% of the time, 0.128; at 60%, 0.100; at 90%, 0.096. The ratios behave less simply. Talks that end well have a ratio of 4.02 without repair and 6.82 with frequent repair; talks that end badly have 2.75 and 5.10. A bad talk with frequent repair can thus show a higher ratio than a good talk without it. The ratio is an outcome of a couple's dynamics and depends on how the counting is done; it is not a threshold that by itself decides a couple's fate. (https://5dlibrary.com/answers/en/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c06/) 第6章 两个静止状态:戈特曼的差分方程 — 没有修复时,0.184 的对话以敌对状态结束;修复三成成功时 0.128;六成时 0.100;九成时 0.096。比例的表现没有那么简单。结局良好的对话,没有修复时比例是 4.02,修复频繁时 6.82;结局不好的对话分别是 2.75 与 5.10。于是,修复频繁的坏对话可以显示出比没有修复的好对话更高的比例。这个比例是一对伴侣动态的结果,并且取决于怎样计数;它不是一个本身就决定伴侣命运的门槛。 (https://5dlibrary.com/answers/zh/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c06/) - Ch.7 "Mang", "Gu feng" and "Bai tou yin" — "I meant to grow old with you; growing old with you has made me bitter. The Qi has its banks, the marsh has its shore. In the feasts of our youth we talked and laughed so gaily; you swore your oaths so earnestly, and I never thought you would turn. That you would turn I never thought: then let it end!" The river has banks: there is a limit. Part (c) of p13 writes the decision to try again as an expected value. If a reunion is worth 10 when the partner has changed and -6 when he has not, trying again is worth it only when the belief that he has changed exceeds 0.375. At a belief of 0.200 trying again is worth -2.80; at 0.600 it is worth +3.60. The speaker of "Mang" has seen enough to put her belief low, and her "let it end" is the decision rule at work. (https://5dlibrary.com/answers/en/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c07/) 第7章 《氓》《谷风》与《白头吟》 — 河有岸,湿地有边:凡事有界限。p13 的 (c) 部分把再试一次的决定写成期望值。如果对方改变了,复合值 10;没有改变,值 -6;那么只有当「他已经改变」的信念超过 0.375 时,再试一次才值得。信念为 0.200 时,再试一次的价值是 -2.80;为 0.600 时是 +3.60。《氓》的叙述者已经看够了,把信念放得很低,她的「亦已焉哉」正是这条决策规则在起作用。 (https://5dlibrary.com/answers/zh/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c07/) - Ch.8 Practice: Three Ways to Spend the Year, Measuring Oneself, and Too Few Cases — Chapter 8 compares three ways of spending the year after a rupture and shows that neither a few weeks of self-ratings nor a few relationships suffice to estimate a person's own numbers. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c08/) 第8章 实践:一年的三种过法、测量自己,以及样本太少 — 第八章比较分手后一年的三种过法,并说明几周的自评与几段关系都不足以估计一个人自己的数字。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/relationship-rupture-survival-no-contact-memory-decay-reconciliation-games/c08/) ## 5DL-B-093 · Shareholder Free-Riding Games, Dynamic Vesting Mechanisms and Defection Early-Warning Models 中文书名:股东搭便车博弈、动态股权成熟与背刺早期预警模型 DOI: https://doi.org/10.5281/zenodo.24090243 Cite: Fan, Y. (2026). Shareholder Free-Riding Games, Dynamic Vesting Mechanisms and Defection Early-Warning Models (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090243 Book page: https://5dlibrary.com/books/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/ Scope: Its parameters are illustrative or synthetic; all calculations are generated by its programs; the book is not legal, tax or investment advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: Is a 50/50 split really the fair default? What does four-year vesting with a one-year cliff solve? Why is a shotgun clause not neutral? Can equity follow contribution? Can a partner's defection be seen early? This book writes these claims as runnable arithmetic and then asks when they hold. Its parameters are illustrative or synthetic; all calculations are generated by its programs; the book is not legal, tax or investment advice. Chapter 1 computes the free-riding loss when effort cannot be verified: two equal founders reach only 0.750 of the first best and three 0.556; Holmström's group target restores 0.990 under small noise but falls to 0.010 under large noise. Chapter 2 shows that vesting cuts the mean equity held by a departed founder from 0.500 to 0.094 but repairs retention and not effort: the value created rises only from 0.016 to 0.312. Chapter 3 solves the shotgun clause: with private values the offerer shades toward the middle and the chooser gains; a partner who can raise only 0.2 can be bought out at 0.21 instead of 0.5. Chapter 4 computes Shapley values for three founders and shows that an alarm based on drift in contributions also sounds in 0.823 of cases in which nothing changed. Chapter 5 uses Wald's sequential test to watch for the quiet transfer of key resources, deciding after about 57.8 weeks on average where a fixed test needs 116; with a base rate of 2%, only 0.279 of alarms from a test with a 5% false-alarm rate are true. Chapter 6 shows that a bad-leaver clause keeps deterring after vesting has run, while a clause that is not enforced deters no one. Chapter 7 reads the Han Feizi's "Bei nei" and eight villainies with pooled capital and the Shanxi merchants' body shares, and notes that the "Song dynasty pooled-capital contract" could not be traced to a Song text. Chapter 8 gives a cap table, dual coverage of key clients (cutting the expected loss from 0.0594 to 0.0152) and the rule that an alarm is not proof. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:五五分真的是公平的默认吗?四年成熟加一年悬崖期解决了什么?互买条款为什么并不中性?股权能否跟着贡献走?合伙人的背叛能否早早看出?本书把这些说法写成可以运行的算术,再看它们在什么条件下成立。书中的参数均为示意或合成;全部计算由书中程序生成;本书不是法律、税务或投资建议。 第一章计算努力无法核实时的搭便车损失:两位均分创始人只达到最优的 0.750,三位 0.556;霍姆斯特罗姆的团队目标在小噪声下恢复到 0.990,但在大噪声下跌到 0.010。第二章说明成熟机制把离开创始人平均持有的股份从 0.500 降到 0.094,却只修补留任、不修补努力:共同创造的价值从 0.016 提高到 0.312。第三章求解互买条款:在私人估值下报价方向中间压价,做选择的一方获益;只能筹到 0.2 的合伙人可能以 0.21 而不是 0.5 被买断。第四章计算三位创始人的沙普利值,并说明以贡献漂移作警报时,在什么都没变的情形中也有 0.823 会响。 第五章用沃尔德序贯检验监测关键资源的悄悄转移,平均约 57.8 周作出决定,而固定检验需要 116 周;基础比率 2% 时,误报率 5% 的检验警报中只有 0.279 是真的。第六章说明坏离职条款在成熟期之后仍有威慑力,而不被执行的条款毫无威慑。第七章与《韩非子·备内》、八奸、合本集资与晋商身股并读,并指出「宋代合本连财契约」未能追溯到宋代原文。第八章给出股权表、关键客户的双人覆盖(预期损失从 0.0594 降到 0.0152),以及警报不是证据的规则。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Free Riding When Effort Cannot Be Verified — Chapter 1 computes the free-riding loss when effort cannot be verified: two equal founders reach only 0.750 of the first best and three 0.556; Holmström's group target restores 0.990 under small noise but falls to 0.010 under large noise. (https://5dlibrary.com/answers/en/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c01/) 第1章 努力无法核实时的搭便车 — 第一章计算努力无法核实时的搭便车损失:两位均分创始人只达到最优的 0.750,三位 0.556;霍姆斯特罗姆的团队目标在小噪声下恢复到 0.990,但在大噪声下跌到 0.010。 (https://5dlibrary.com/answers/zh/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c01/) - Ch.2 Vesting: What It Repairs and What It Does Not — Chapter 2 shows that vesting cuts the mean equity held by a departed founder from 0.500 to 0.094 but repairs retention and not effort: the value created rises only from 0.016 to 0.312. (https://5dlibrary.com/answers/en/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c02/) 第2章 股权成熟:它修补什么、不修补什么 — 第二章说明成熟机制把离开创始人平均持有的股份从 0.500 降到 0.094,却只修补留任、不修补努力:共同创造的价值从 0.016 提高到 0.312。 (https://5dlibrary.com/answers/zh/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c02/) - Ch.3 The Shotgun Clause: Who Gains When Information and Cash Are Unequal — Chapter 3 solves the shotgun clause: with private values the offerer shades toward the middle and the chooser gains; a partner who can raise only 0.2 can be bought out at 0.21 instead of 0.5. (https://5dlibrary.com/answers/en/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c03/) 第3章 互买条款:信息与现金不对等时谁获益 — 第三章求解互买条款:在私人估值下报价方向中间压价,做选择的一方获益;只能筹到 0.2 的合伙人可能以 0.21 而不是 0.5 被买断。 (https://5dlibrary.com/answers/zh/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c03/) - Ch.4 Shapley Values and the Drift of Contributions — Chapter 4 computes Shapley values for three founders and shows that an alarm based on drift in contributions also sounds in 0.823 of cases in which nothing changed. (https://5dlibrary.com/answers/en/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c04/) 第4章 沙普利值与贡献的漂移 — 第四章计算三位创始人的沙普利值,并说明以贡献漂移作警报时,在什么都没变的情形中也有 0.823 会响。 (https://5dlibrary.com/answers/zh/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c04/) - Ch.5 Early Warning with Wald's Sequential Test — Chapter 5 uses Wald's sequential test to watch for the quiet transfer of key resources, deciding after about 57.8 weeks on average where a fixed test needs 116; with a base rate of 2%, only 0.279 of alarms from a test with a 5% false-alarm rate are true. (https://5dlibrary.com/answers/en/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c05/) 第5章 用沃尔德序贯检验做早期预警 — 第五章用沃尔德序贯检验监测关键资源的悄悄转移,平均约 57.8 周作出决定,而固定检验需要 116 周;基础比率 2% 时,误报率 5% 的检验警报中只有 0.279 是真的。 (https://5dlibrary.com/answers/zh/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c05/) - Ch.6 When Clauses Deter — Without the clause, the loss from leaving to compete falls from 12.00 in month 6 to 9.00 in month 12, 6.00 in month 24, 3.00 in month 36 and 0.00 in month 48; against a gain of 6.0 it deters only in the first two years. With the clause, even if it is enforced only seven times in ten, the loss is 10.68 in month 12, 9.36 in month 24, 8.04 in month 36 and 6.72 in month 48, enough to deter throughout. The clause matters most precisely when vesting has done its work. (https://5dlibrary.com/answers/en/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c06/) 第6章 条款何时有威慑力 — 没有这一条款时,离开去竞争的损失从第 6 个月的 12.00 降到第 12 个月的 9.00、第 24 个月的 6.00、第 36 个月的 3.00、第 48 个月的 0.00;对于 6.0 的收益,只在头两年有威慑力。有这一条款时,即使十次只有七次得到执行,损失在第 12 个月是 10.68,第 24 个月 9.36,第 36 个月 8.04,第 48 个月 6.72,自始至终都足以威慑。这一条款恰恰在成熟机制完成任务之后最重要。 (https://5dlibrary.com/answers/zh/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c06/) - Ch.7 "Bei nei", the Eight Villainies, Pooled Capital and Body Shares — The Shanxi merchant houses are known for giving managers "body shares" (shen gu), a share in profits without any capital, alongside the owners' "silver shares" (yin gu) (weijuxian). Part (b) of p13 asks what share maximises the owner's take when the manager's effort follows the share. The owner's take, $(1-s)s$, is largest at $s = 0.5$, where it is 0.250 and total surplus is 0.375; a share of 0.7 raises total surplus to 0.455 but lowers the owner's take to 0.210. An owner who sets the share for his own take leaves value unrealised, the same free-riding loss as in Chapter 1. (https://5dlibrary.com/answers/en/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c07/) 第7章 《备内》、八奸、合本集资与身股 — 山西商号以给掌柜「身股」著称,即不出资而分享利润,与东家的「银股」并列 (weijuxian)。p13 的 (b) 部分问:当掌柜的努力随份额而定时,什么份额使东家所得最大。东家所得 $(1-s)s$ 在 $s = 0.5$ 时最大,为 0.250,总剩余为 0.375;份额 0.7 把总剩余提高到 0.455,却把东家所得降到 0.210。按自身所得来定份额的东家,会让一部分价值无法实现,这与第一章的搭便车损失相同。 (https://5dlibrary.com/answers/zh/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c07/) - Ch.8 Practice: A Cap Table, a Key-Person Test and the Rule That an Alarm Is Not Proof — Chapter 8 gives a cap table, dual coverage of key clients (cutting the expected loss from 0.0594 to 0.0152) and the rule that an alarm is not proof. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c08/) 第8章 实践:股权表、关键人物测试,以及警报不是证据 — 第八章给出股权表、关键客户的双人覆盖(预期损失从 0.0594 降到 0.0152),以及警报不是证据的规则。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/shareholder-free-riding-dynamic-vesting-mechanisms-defection-early-warning/c08/) ## 5DL-B-094 · Multistage Ultimatum Bargaining, Accounts-Receivable Decay Kinetics and Optimal Escalation Sequences 中文书名:多阶段最后通牒议价、应收账款衰减动力学与最优施压序列 DOI: https://doi.org/10.5281/zenodo.24090245 Cite: Fan, Y. (2026). Multistage Ultimatum Bargaining, Accounts-Receivable Decay Kinetics and Optimal Escalation Sequences (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090245 Book page: https://5dlibrary.com/books/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/ Scope: Its parameters are illustrative or synthetic; all calculations are generated by its programs; the book is not legal, financial or employment advice, and empty or unlawful threats are not tactics it endorses. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: Does the Ackerman sequence of shrinking offers really beat equal concessions? Are invoices 90 days overdue really less than 40% collectible? On which day should collection move from reminders to stopping service and a lawyer's letter? When is a threat to sue credible? What wins a raise? This book writes these claims as runnable arithmetic and then asks when they hold. Its parameters are illustrative or synthetic; all calculations are generated by its programs; the book is not legal, financial or employment advice, and empty or unlawful threats are not tactics it endorses. Chapter 1 reads a sequence of concessions as evidence about the buyer's limit: after 65, 85 and 95 the seller's estimate is 110.9, against 115.8 after equal steps, and a precise final number raises the belief that an offer is final from 0.333 to 0.750. Chapter 2 solves Rubinstein's alternating offers and an ultimatum game with fairness-minded responders, and shows how much depends on who makes the last offer before a deadline. Chapter 3 shows that the collectible share of overdue invoices falls month by month even when no debtor changes: of invoices still open on day 90, 0.697 are eventually paid in one book and 0.478 in a riskier one. Chapter 4 finds the best days to escalate and shows that escalating fully on day 1 cuts net recovery from 0.9450 to 0.5936. Chapter 5 shows that in ordinary litigation a threat to sue is credible only for invoices above 4,762, that a small-claims track lowers the threshold to 476, and computes the value of a reputation for following through. Chapter 6 treats a raise as a Nash bargain and shows that bluffing about an outside offer is worth -4,000 over a year, while a real search with an even chance of success is worth +1,500. Chapter 7 reads Zhu Zhiwu's speech to the Duke of Qin in the Zuo zhuan, the Sunzi's "Mou gong" and the Tang code's ladder for unpaid debt, and notes that the Zhu Zhiwu episode is not from the Zhanguo ce. Chapter 8 gives a collection playbook, a way to choose an opening offer and the reason a client cannot be known from a few invoices. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:阿克曼递减出价真的胜过等幅让步吗?逾期 90 天的账款真的只剩不到四成能收回吗?催收应当在哪一天从提醒升级到停止服务与律师函?起诉的威胁何时可信?加薪靠的是什么?本书把这些说法写成可以运行的算术,再看它们在什么条件下成立。书中的参数均为示意或合成;全部计算由书中程序生成;本书不是法律、财务或劳动关系建议,空洞或违法的威胁不是本书认可的手段。 第一章把让步序列当作关于买方底线的证据:出过 65、85、95 之后,卖方对底线的估计是 110.9,等幅出价之后是 115.8;精确的最后数字把「这是最后报价」的信念从 0.333 提高到 0.750。第二章求解鲁宾斯坦轮流出价与在乎公平的最后通牒博弈,并说明截止期限前由谁最后出价的重要性。第三章说明即使没有任何债务人改变,逾期账款的可收回比例也逐月下降:第 90 天仍未付的发票,在一种组合中最终收回 0.697,在风险更高的组合中 0.478。第四章求出升级的最佳日期,并说明第 1 天就全面升级会把净回收从 0.9450 降到 0.5936。 第五章说明普通诉讼下起诉的威胁只对 4,762 以上的发票可信,小额诉讼把门槛降到 476,并计算贯彻到底的信誉价值。第六章把加薪看作纳什议价,并说明谎称有录用一年下来值 -4,000,而有一半机会成功的真实求职值 +1,500。第七章与《左传》烛之武说秦伯、《孙子·谋攻》及《唐律疏议·杂律》的负债阶梯并读,并指出烛之武一事不出自《战国策》。第八章给出催收手册、开价方法,以及凭几张发票无法认识一个客户的道理。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 What a Pattern of Concessions Tells the Other Side — Chapter 1 reads a sequence of concessions as evidence about the buyer's limit: after 65, 85 and 95 the seller's estimate is 110.9, against 115.8 after equal steps, and a precise final number raises the belief that an offer is final from 0.333 to 0.750. (https://5dlibrary.com/answers/en/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c01/) 第1章 一串让步对对方透露了什么 — 第一章把让步序列当作关于买方底线的证据:出过 65、85、95 之后,卖方对底线的估计是 110.9,等幅出价之后是 115.8;精确的最后数字把「这是最后报价」的信念从 0.333 提高到 0.750。 (https://5dlibrary.com/answers/zh/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c01/) - Ch.2 Patience, Deadlines and Fairness — With equal patience of 0.9 the proposer gets 0.526; with 0.99, 0.503, almost an even split. Patience is power: a proposer with 0.95 facing an impatient side with 0.8 gets 0.833, and the reverse gives only 0.208. A risk that talks break down after a rejection acts like impatience on both sides and favours whoever moves first: with a breakdown risk of 0.05 the proposer's share rises to 0.515, with 0.2 to 0.558. In a collection or a price negotiation, the side that can afford to wait, a supplier with other customers or a debtor with no urgent need for the supplier, holds the stronger hand. (https://5dlibrary.com/answers/en/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c02/) 第2章 耐心、截止期限与公平 — 双方耐心都是 0.9 时,先出价方得到 0.526;都是 0.99 时得到 0.503,几乎平分。耐心就是力量:耐心 0.95 的先出价方面对耐心只有 0.8 的一方,得到 0.833,反过来只得到 0.208。拒绝之后谈判破裂的风险,对双方都像缺乏耐心,并有利于先出手的一方:破裂风险 0.05 时先出价方的份额升到 0.515,0.2 时升到 0.558。在催收或价格谈判中,等得起的一方,例如有别的客户的供应商,或者不急需这家供应商的债务人,握着较强的牌。 (https://5dlibrary.com/answers/zh/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c02/) - Ch.3 Why Overdue Invoices Look Less Collectible Every Month — Chapter 3 shows that the collectible share of overdue invoices falls month by month even when no debtor changes: of invoices still open on day 90, 0.697 are eventually paid in one book and 0.478 in a riskier one. (https://5dlibrary.com/answers/en/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c03/) 第3章 为什么逾期发票看起来每个月都更难收回 — 第三章说明即使没有任何债务人改变,逾期账款的可收回比例也逐月下降:第 90 天仍未付的发票,在一种组合中最终收回 0.697,在风险更高的组合中 0.478。 (https://5dlibrary.com/answers/zh/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c03/) - Ch.4 When to Escalate — Chapter 4 finds the best days to escalate and shows that escalating fully on day 1 cuts net recovery from 0.9450 to 0.5936. (https://5dlibrary.com/answers/en/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c04/) 第4章 何时升级 — 第四章求出升级的最佳日期,并说明第 1 天就全面升级会把净回收从 0.9450 降到 0.5936。 (https://5dlibrary.com/answers/zh/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c04/) - Ch.5 Credible Threats and the Value of Following Through — Chapter 5 shows that in ordinary litigation a threat to sue is credible only for invoices above 4,762, that a small-claims track lowers the threshold to 476, and computes the value of a reputation for following through. (https://5dlibrary.com/answers/en/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c05/) 第5章 可信的威胁与贯彻到底的价值 — 第五章说明普通诉讼下起诉的威胁只对 4,762 以上的发票可信,小额诉讼把门槛降到 476,并计算贯彻到底的信誉价值。 (https://5dlibrary.com/answers/zh/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c05/) - Ch.6 Raises and Price Increases: Building the Alternative — Chapter 6 treats a raise as a Nash bargain and shows that bluffing about an outside offer is worth -4,000 over a year, while a real search with an even chance of success is worth +1,500. (https://5dlibrary.com/answers/en/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c06/) 第6章 加薪与提价:建立替代选择 — 第六章把加薪看作纳什议价,并说明谎称有录用一年下来值 -4,000,而有一半机会成功的真实求职值 +1,500。 (https://5dlibrary.com/answers/zh/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c06/) - Ch.7 Zhu Zhiwu, "Mou gong" and the Tang Code's Ladder for Unpaid Debt — "To win a hundred battles out of a hundred is not the highest skill; to subdue the enemy without fighting is the highest skill. So the best is to attack the enemy's plans, the next to attack his alliances, the next to attack his army, and the worst to besiege walled cities." Part (c) of p13 gives the four levels illustrative costs of 1, 3, 10 and 30 and rising chances of success. Trying them in order and stopping at the first success costs 12.34 on average and fails 0.034 of the time; going straight to the walls costs 30 and fails 0.200 of the time. The ladder of Chapter 4 has the same shape: begin with what is cheap and preserves the relationship, and climb only as far as needed. (https://5dlibrary.com/answers/en/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c07/) 第7章 烛之武、《谋攻》与《唐律》的负债阶梯 — p13 的 (c) 部分给四个层次示意的成本 1、3、10、30,以及逐级上升的成功机会。依次尝试、第一次成功就停止,平均成本 12.34,失败的机会 0.034;直接攻城,成本 30,失败的机会 0.200。第四章的阶梯有同样的形状:从便宜而且保全关系的手段开始,只爬到必要的高度。 (https://5dlibrary.com/answers/zh/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c07/) - Ch.8 Practice: A Collection Playbook, an Opening Offer and Knowing a Client — Chapter 8 gives a collection playbook, a way to choose an opening offer and the reason a client cannot be known from a few invoices. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c08/) 第8章 实践:催收手册、开价与认识一个客户 — 第八章给出催收手册、开价方法,以及凭几张发票无法认识一个客户的道理。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/multistage-ultimatum-bargaining-receivables-decay-escalation-sequences/c08/) ## 5DL-B-095 · Insulin-Glucagon Bistability, Visceral Lipolysis Kinetics and Linear-Programming Diet Optimization 中文书名:胰岛素胰高血糖素双稳态、内脏脂解动力学与线性规划配餐 DOI: https://doi.org/10.5281/zenodo.24090248 Cite: Fan, Y. (2026). Insulin-Glucagon Bistability, Visceral Lipolysis Kinetics and Linear-Programming Diet Optimization (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090248 Book page: https://5dlibrary.com/books/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/ Scope: Its parameters are illustrative or synthetic; the samples of Chapter 4 are generated; all figures are produced by the programs printed in the book; the book is not medical advice and does not diagnose. Chapter 4 reaches an area of 0.9645 for the index against 0.9244 for the waist ratio on the same synthetic sample, so the blood test adds 0.0400, with a best index cutoff of 8.705. Abstract: Does liver glycogen really run out at a definite hour? Is the change between insulin and the fuel signals a threshold or a switch with hysteresis? Eating the same energy in two meals rather than six, how differently does visceral fat behave? Can twenty-eight days of distinct low-carbohydrate menus be solved for under electrolyte floors and a cost constraint? This book writes these claims as runnable arithmetic and then asks when they hold. Its parameters are illustrative or synthetic; the samples of Chapter 4 are generated; all figures are produced by the programs printed in the book; the book is not medical advice and does not diagnose. Chapter 1 computes glycogen depletion over a grid of plausible stores and drain rates: only 0.640 of the cells fall inside the fourteen to twenty hour window, the median is 18.2 hours and the spread runs from 12.3 to 26.7 hours. It then writes insulin and the fuel signals as a mutual-inhibition switch in which the fed state appears at a drive of 0.3554 while the fasted state survives to 1.1142, thresholds that differ by 3.14. Chapter 2 shows that hormone-sensitive lipase needs its activating sites phosphorylated and its restraining site dephosphorylated, and that the different timescales make activity overshoot to 1.874 times its steady fasted value; the visceral depot gives up 2.26 times its share of mass when fasted and 3.47 times after a meal. Chapter 3 compares two patterns carrying the same energy and carbohydrate whose mean insulin differs by 0.8 pmol/L, and finds the eight-hour window spends 7.68 more hours below the threshold and releases 0.122 more. Chapter 4 reaches an area of 0.9645 for the index against 0.9244 for the waist ratio on the same synthetic sample, so the blood test adds 0.0400, with a best index cutoff of 8.705. Chapter 5 finds the cheapest day meeting all three electrolyte floors costs 1.44, or 9.95 from whole food alone, a ratio of 6.91. Chapter 6 solves the cheapest compliant day at 8.2888 by two-phase simplex, with sodium the only tight constraint and no solution at all below an energy target of 1655.0 kcal, then builds twenty-eight distinct menus from the feasible subset of 210 candidates. Chapter 7 reads the textual crux in the Suwen's warning about rich food alongside Sun Simiao's chapter on food therapy: the plain household day carries 318.8 g of carbohydrate, more than the rich one, and the extremes of energy density differ by 3.62 times; the thermodynamic accounting shows 0.709 of burnt fat leaves through the lungs. Chapter 8 shows that only 0.202 of the day-seven fall on the scale is fat, and that the safety gate stops or refers 0.5818 of adults. Marked boxes separate published evidence, runnable Python calculations (sixteen programs), classical sources and the author's hypotheses. 摘要:肝糖原真的会在某个确定的小时数耗尽吗?胰岛素与燃料信号之间的切换是一个阈值,还是一个带回差的开关?同样的热量分两餐吃与分六餐吃,内脏脂肪的去向差多少?一份二十八天不重样的低碳水食谱,在电解质下界与成本约束下解得出来吗?本书把这些说法写成可以运行的算术,再看它们在什么条件下成立。书中参数均为示意或合成;第四章的样本是生成的;全部数字由书中程序产生;本书不是医疗建议,不作诊断。 第一章在合理的储量与消耗速率格子上算出肝糖原耗竭时程:落在十四到二十小时窗口内的格子只占 0.640,中位数 18.2 小时,跨度从 12.3 到 26.7 小时;并把胰岛素与燃料信号写成互抑开关,进食态在驱动 0.3554 处出现而空腹态直到 1.1142 才消失,两门槛相差 3.14 倍。第二章说明激素敏感性脂肪酶需要激活位点被磷酸化而抑制位点被去磷酸化,两者时间尺度不同导致活性过冲到稳定空腹值的 1.874 倍;内脏脂库在空腹时交出其质量份额的 2.26 倍,餐后则是 3.47 倍。第三章在热量与碳水相同、平均胰岛素只差 0.8 pmol/L 的两种模式间,量出八小时窗口多 7.68 小时处于阈值之下、内脏释放多 0.122。 第四章在同一份合成样本上得到指数曲线下面积 0.9645 与腰高比 0.9244,血检只加 0.0400,指数最佳界值 8.705。第五章算出满足三条电解质下界的最便宜一天为 1.44,只用全食物则为 9.95,比值 6.91。第六章用两阶段单纯形解出最便宜的合规一天 8.2888,只有钠是紧的,而能量目标低于 1655.0 kcal 时无解;再从 210 个候选中的可行菜单排出二十八天不重样。第七章与《素问·生气通天论》「高梁之变,足生大丁」的字句之争及孙思邈《千金要方·食治》并读:寻常人家那一天带 318.8 g 碳水化合物,比厚味那一天更多,两极的能量密度相差 3.62 倍;并给出热力学账目,烧掉的脂肪有 0.709 从肺里呼出。第八章显示第七天秤上的下降只有 0.202 是脂肪,而安全闸拦下或转诊 0.5818 的成年人。书中以标注框区分已发表的证据、可运行的 Python 计算(十六段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Switch: Where the Fourteen to Twenty Hours Come From — Chapter 1 computes glycogen depletion over a grid of plausible stores and drain rates: only 0.640 of the cells fall inside the fourteen to twenty hour window, the median is 18.2 hours and the spread runs from 12.3 to 26.7 hours. It then writes insulin and the fuel signals as a mutual-inhibition switch in which the fed state appears at a drive of 0.3554 while the fasted state survives to 1.1142, thresholds that differ by 3.14. (https://5dlibrary.com/answers/en/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c01/) 第1章 开关:十四到二十小时从哪里来 — 第一章在合理的储量与消耗速率格子上算出肝糖原耗竭时程:落在十四到二十小时窗口内的格子只占 0.640,中位数 18.2 小时,跨度从 12.3 到 26.7 小时;并把胰岛素与燃料信号写成互抑开关,进食态在驱动 0.3554 处出现而空腹态直到 1.1142 才消失,两门槛相差 3.14 倍。 (https://5dlibrary.com/answers/zh/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c01/) - Ch.2 The Lipase: Two Conditions, and an Overshoot — Chapter 2 shows that hormone-sensitive lipase needs its activating sites phosphorylated and its restraining site dephosphorylated, and that the different timescales make activity overshoot to 1.874 times its steady fasted value; the visceral depot gives up 2.26 times its share of mass when fasted and 3.47 times after a meal. (https://5dlibrary.com/answers/en/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c02/) 第2章 脂肪酶:两个条件,以及一次过冲 — 第二章说明激素敏感性脂肪酶需要激活位点被磷酸化而抑制位点被去磷酸化,两者时间尺度不同导致活性过冲到稳定空腹值的 1.874 倍;内脏脂库在空腹时交出其质量份额的 2.26 倍,餐后则是 3.47 倍。 (https://5dlibrary.com/answers/zh/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c02/) - Ch.3 Same Food, Two Patterns — Chapter 3 compares two patterns carrying the same energy and carbohydrate whose mean insulin differs by 0.8 pmol/L, and finds the eight-hour window spends 7.68 more hours below the threshold and releases 0.122 more. (https://5dlibrary.com/answers/en/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c03/) 第3章 同样的食物,两种模式 — 第三章在热量与碳水相同、平均胰岛素只差 0.8 pmol/L 的两种模式间,量出八小时窗口多 7.68 小时处于阈值之下、内脏释放多 0.122。 (https://5dlibrary.com/answers/zh/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c03/) - Ch.4 Two Cheap Measurements — On one synthetic sample the waist ratio reaches an area of 0.9244 and the index 0.9645, so the blood test adds 0.0400. A two-step rule scores 0.7858 against 0.7970 for the index alone while ordering blood on a minority. The best index cutoff, 8.705, sits where the validation literature puts it. Chapter 5 turns from measuring people to feeding them, starting with the mineral the low-carbohydrate week takes away. (https://5dlibrary.com/answers/en/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c04/) 第4章 两项便宜的测量 — 在一份合成样本上,腰高比的曲线下面积达到 0.9244,指数达到 0.9645,所以血检添加了 0.0400。一条两步规则得分 0.7858,单用指数得分 0.7970,而前者只对少数人开血检。指数的最佳界值 8.705 正落在验证文献给出的位置。第五章从量人转向喂人,先从低碳水的一周所拿走的那种矿物质开始。 (https://5dlibrary.com/answers/zh/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c04/) - Ch.5 The Mineral the Diet Takes Away — Chapter 5 finds the cheapest day meeting all three electrolyte floors costs 1.44, or 9.95 from whole food alone, a ratio of 6.91. (https://5dlibrary.com/answers/en/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c05/) 第5章 这套饮食拿走的那种矿物质 — 第五章算出满足三条电解质下界的最便宜一天为 1.44,只用全食物则为 9.95,比值 6.91。 (https://5dlibrary.com/answers/zh/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c05/) - Ch.6 The Diet Problem, Solved Twice — Chapter 6 solves the cheapest compliant day at 8.2888 by two-phase simplex, with sodium the only tight constraint and no solution at all below an energy target of 1655.0 kcal, then builds twenty-eight distinct menus from the feasible subset of 210 candidates. (https://5dlibrary.com/answers/en/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c06/) 第6章 配餐问题,解两遍 — 第六章用两阶段单纯形解出最便宜的合规一天 8.2888,只有钠是紧的,而能量目标低于 1655.0 kcal 时无解;再从 210 个候选中的可行菜单排出二十八天不重样。 (https://5dlibrary.com/answers/zh/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c06/) - Ch.7 What the Classics Were Warning About — Chapter 7 reads the textual crux in the Suwen's warning about rich food alongside Sun Simiao's chapter on food therapy: the plain household day carries 318.8 g of carbohydrate, more than the rich one, and the extremes of energy density differ by 3.62 times; the thermodynamic accounting shows 0.709 of burnt fat leaves through the lungs. (https://5dlibrary.com/answers/en/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c07/) 第7章 古典到底在警告什么 — 第七章与《素问·生气通天论》「高梁之变,足生大丁」的字句之争及孙思邈《千金要方·食治》并读:寻常人家那一天带 318.8 g 碳水化合物,比厚味那一天更多,两极的能量密度相差 3.62 倍;并给出热力学账目,烧掉的脂肪有 0.709 从肺里呼出。 (https://5dlibrary.com/answers/zh/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c07/) - Ch.8 The Month, and the Gate — Chapter 8 shows that only 0.202 of the day-seven fall on the scale is fat, and that the safety gate stops or refers 0.5818 of adults. Marked boxes separate published evidence, runnable Python calculations (sixteen programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c08/) 第8章 这个月,以及那道闸 — 第八章显示第七天秤上的下降只有 0.202 是脂肪,而安全闸拦下或转诊 0.5818 的成年人。书中以标注框区分已发表的证据、可运行的 Python 计算(十六段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/insulin-glucagon-bistability-visceral-lipolysis-linear-programming-diets/c08/) ## 5DL-B-096 · Polypharmacy Supplement Interactions, Chronotherapy Dosing Windows and Cytochrome Enzyme Competition 中文书名:复合补剂相互作用、时间疗法给药窗口与代谢酶竞争 DOI: https://doi.org/10.5281/zenodo.24090250 Cite: Fan, Y. (2026). Polypharmacy Supplement Interactions, Chronotherapy Dosing Windows and Cytochrome Enzyme Competition (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090250 Book page: https://5dlibrary.com/books/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/ Scope: Its parameters are illustrative or synthetic; all figures are produced by the programs printed in the book; the book is not medical advice, does not diagnose, and must not be used to start, stop or change any supplement or medicine. Marked boxes separate published evidence, runnable Python calculations (sixteen programs), classical sources and the author's hypotheses. Abstract: How much does a handful of capsules swallowed at once actually lose? And how much can be recovered by splitting them into groups taken at different hours? This book writes these questions as runnable arithmetic and then asks when the answers hold. Its parameters are illustrative or synthetic; all figures are produced by the programs printed in the book; the book is not medical advice, does not diagnose, and must not be used to start, stop or change any supplement or medicine. Chapter 1 starts from the Michaelis equation: a competitor present at its own inhibition constant takes 0.4762 of the rate when the substrate is a tenth of the affinity constant but only 0.1429 when it is five times that constant, a ratio of 3.33, so a saturable site is weakest exactly where the dose is small. A matrix of eleven common items shows seven losing more than one part in a hundred when taken together, with calcium worst hit at 0.7537 and zinc at 0.7498, while curcumin is the one gainer at 0.5560. Chapter 2 proves an identity about a shared carrier: a new competitor multiplies every existing occupant's absorption by exactly the same factor, so the calcium dose that halves iron absorption and the one that halves zinc absorption are both 573.3 mg; at calcium 500 and magnesium 400 mg, iron keeps only 0.5341. Losses inside the gut lumen are separate: a proton pump inhibitor costs iron 0.6697 through the pH term alone, and iron solubility halves at a stomach pH of 3.826. Chapter 3 multiplies site competition by gut condition: the handful loses 0.5096, eight of eleven items lose at least 0.35, and iron keeps only 0.0856. Chapter 4 locates the best moment on a one-minute grid: 04:32 for a morning cofactor and 23:42 for an evening calmer, a phase difference of 19.17 hours or 287.5 degrees, with a six-hour error costing 0.6156 and 0.7652. Chapter 5 shows the caffeine half-life differing by 7.50 times between the fastest and slowest states, so that one cup at 14:00 raises sleep onset from 18.4 to 40.5 minutes in the slowest, and shows the exposure multiplier of metabolic inhibition bounded by one over one minus the cleared fraction. Chapter 6 enumerates all 177147 three-slot schedules: the optimum scores 0.8605 against 0.5394 for everything at breakfast, and, more to the point, 0.8355 of arbitrary schedules are worse than everything at breakfast. Chapter 7 writes the seven relations of the Shennong pharmacopoeia as a matrix, finding 47 of the 55 pairs acting alone, 7 in opposition and 1 assisting, and compares the four dosing-hour rules quoted in the preface to the Bencao gangmu with the arithmetic, which agrees on all four while the book argues that the agreement comes from the author's own mapping. Chapter 8 shows adherence pulling the three-slot schedule from 0.8605 down to 0.7362, and the safety gate stopping or referring 0.5363 of adults. Marked boxes separate published evidence, runnable Python calculations (sixteen programs), classical sources and the author's hypotheses. 摘要:一把胶囊同时吞下,到底损失了多少?把它们分成几堆、分别在什么时辰吃,又能收回多少?本书把这些问题写成可以运行的算术,再看答案在什么条件下成立。书中参数均为示意或合成;全部数字由书中程序产生;本书不是医疗建议,不作诊断,也不可用于开始、停止或改动任何补剂或药物。 第一章从米氏方程出发:一个与底物等量的竞争者,在底物只有亲和常数十分之一时夺去 0.4762 的速率,而在底物为其五倍时只夺去 0.1429,相差 3.33 倍,所以可饱和位点恰在剂量小处最脆弱。十一种常见补剂的矩阵显示,同服时七项的损失超过百分之一,被打击最重的钙损失 0.7537,锌损失 0.7498,唯一的获益者姜黄素增加 0.5560。第二章给出共享载体的一条恒等式:新的竞争者把同一位点上每个既有占位者的吸收乘以完全相同的因子,所以使铁吸收减半与使锌吸收减半的钙剂量同为 573.3 毫克;而在钙 500 与镁 400 毫克之下,铁只保住 0.5341。腔内损耗另算:质子泵抑制剂使铁的 pH 项损失 0.6697,铁溶解度减半的胃内 pH 为 3.826。 第三章把位点竞争与肠道条件相乘:一把同吞的损失为 0.5096,其中八项损失至少 0.35,最重的铁只留下 0.0856。第四章在一分钟的网格上定出最佳时刻:晨间辅因子在 04:32,夜间镇静剂在 23:42,相位差 19.17 小时即 287.5 度;六小时的偏差分别损失 0.6156 与 0.7652。第五章显示咖啡因半衰期在最快与最慢两种状态间差 7.50 倍,14:00 一杯在最慢状态下使入睡时间从 18.4 分钟升到 40.5 分钟;代谢抑制的暴露倍数被一除以一减清除分数所封顶。 第六章穷举 177147 种三槽排程,最优为 0.8605,而全部放进早餐为 0.5394;更要紧的是 0.8355 的随意排程比全部放进早餐更差。第七章把《神农本草经》七情和合写成矩阵:55 对中 47 对各行其是,7 对相反,1 对相使;并把《本草纲目》序例所引服药时辰的四条规则与算术对照,四条全合,而本书指出这个一致性来自作者自己的对应方式。第八章显示依从性把三槽排程从 0.8605 拉到 0.7362,而安全闸拦下或转诊 0.5363 的成年人。书中以标注框区分已发表的证据、可运行的 Python 计算(十六段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Competition at a Saturable Site — Chapter 1 starts from the Michaelis equation: a competitor present at its own inhibition constant takes 0.4762 of the rate when the substrate is a tenth of the affinity constant but only 0.1429 when it is five times that constant, a ratio of 3.33, so a saturable site is weakest exactly where the dose is small. A matrix of eleven common items shows seven losing more than one part in a hundred when taken together, with calcium worst hit at 0.7537 and zinc at 0.7498, while curcumin is the one gainer at 0.5560. (https://5dlibrary.com/answers/en/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c01/) 第1章 可饱和位点上的竞争 — 第一章从米氏方程出发:一个与底物等量的竞争者,在底物只有亲和常数十分之一时夺去 0.4762 的速率,而在底物为其五倍时只夺去 0.1429,相差 3.33 倍,所以可饱和位点恰在剂量小处最脆弱。十一种常见补剂的矩阵显示,同服时七项的损失超过百分之一,被打击最重的钙损失 0.7537,锌损失 0.7498,唯一的获益者姜黄素增加 0.5560。 (https://5dlibrary.com/answers/zh/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c01/) - Ch.2 One Carrier, and the Losses Before It — Chapter 2 proves an identity about a shared carrier: a new competitor multiplies every existing occupant's absorption by exactly the same factor, so the calcium dose that halves iron absorption and the one that halves zinc absorption are both 573.3 mg; at calcium 500 and magnesium 400 mg, iron keeps only 0.5341. Losses inside the gut lumen are separate: a proton pump inhibitor costs iron 0.6697 through the pH term alone, and iron solubility halves at a stomach pH of 3.826. (https://5dlibrary.com/answers/en/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c02/) 第2章 一个载体,以及它之前的损耗 — 第二章给出共享载体的一条恒等式:新的竞争者把同一位点上每个既有占位者的吸收乘以完全相同的因子,所以使铁吸收减半与使锌吸收减半的钙剂量同为 573.3 毫克;而在钙 500 与镁 400 毫克之下,铁只保住 0.5341。腔内损耗另算:质子泵抑制剂使铁的 pH 项损失 0.6697,铁溶解度减半的胃内 pH 为 3.826。 (https://5dlibrary.com/answers/zh/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c02/) - Ch.3 The Handful — Chapter 3 multiplies site competition by gut condition: the handful loses 0.5096, eight of eleven items lose at least 0.35, and iron keeps only 0.0856. (https://5dlibrary.com/answers/en/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c03/) 第3章 那一把 — 第三章把位点竞争与肠道条件相乘:一把同吞的损失为 0.5096,其中八项损失至少 0.35,最重的铁只留下 0.0856。 (https://5dlibrary.com/answers/zh/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c03/) - Ch.4 When the Receptor Is Listening — Chapter 4 locates the best moment on a one-minute grid: 04:32 for a morning cofactor and 23:42 for an evening calmer, a phase difference of 19.17 hours or 287.5 degrees, with a six-hour error costing 0.6156 and 0.7652. (https://5dlibrary.com/answers/en/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c04/) 第4章 受体在听的时候 — 第四章在一分钟的网格上定出最佳时刻:晨间辅因子在 04:32,夜间镇静剂在 23:42,相位差 19.17 小时即 287.5 度;六小时的偏差分别损失 0.6156 与 0.7652。 (https://5dlibrary.com/answers/zh/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c04/) - Ch.5 Two Enzymes in the Liver — Chapter 5 shows the caffeine half-life differing by 7.50 times between the fastest and slowest states, so that one cup at 14:00 raises sleep onset from 18.4 to 40.5 minutes in the slowest, and shows the exposure multiplier of metabolic inhibition bounded by one over one minus the cleared fraction. (https://5dlibrary.com/answers/en/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c05/) 第5章 肝脏里的两个酶 — 第五章显示咖啡因半衰期在最快与最慢两种状态间差 7.50 倍,14:00 一杯在最慢状态下使入睡时间从 18.4 分钟升到 40.5 分钟;代谢抑制的暴露倍数被一除以一减清除分数所封顶。 (https://5dlibrary.com/answers/zh/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c05/) - Ch.6 The Schedule, Solved Exactly — Chapter 6 enumerates all 177147 three-slot schedules: the optimum scores 0.8605 against 0.5394 for everything at breakfast, and, more to the point, 0.8355 of arbitrary schedules are worse than everything at breakfast. (https://5dlibrary.com/answers/en/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c06/) 第6章 精确解出的排程 — 第六章穷举 177147 种三槽排程,最优为 0.8605,而全部放进早餐为 0.5394;更要紧的是 0.8355 的随意排程比全部放进早餐更差。 (https://5dlibrary.com/answers/zh/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c06/) - Ch.7 The Seven Relations — Chapter 7 writes the seven relations of the Shennong pharmacopoeia as a matrix, finding 47 of the 55 pairs acting alone, 7 in opposition and 1 assisting, and compares the four dosing-hour rules quoted in the preface to the Bencao gangmu with the arithmetic, which agrees on all four while the book argues that the agreement comes from the author's own mapping. (https://5dlibrary.com/answers/en/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c07/) 第7章 七情和合 — 第七章把《神农本草经》七情和合写成矩阵:55 对中 47 对各行其是,7 对相反,1 对相使;并把《本草纲目》序例所引服药时辰的四条规则与算术对照,四条全合,而本书指出这个一致性来自作者自己的对应方式。 (https://5dlibrary.com/answers/zh/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c07/) - Ch.8 Adherence, and the Gate — Chapter 8 shows adherence pulling the three-slot schedule from 0.8605 down to 0.7362, and the safety gate stopping or referring 0.5363 of adults. Marked boxes separate published evidence, runnable Python calculations (sixteen programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c08/) 第8章 依从性,以及那道闸 — 第八章显示依从性把三槽排程从 0.8605 拉到 0.7362,而安全闸拦下或转诊 0.5363 的成年人。书中以标注框区分已发表的证据、可运行的 Python 计算(十六段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/polypharmacy-supplement-interactions-chronotherapy-windows-enzyme-competition/c08/) ## 5DL-B-097 · HPA-Axis Coupled Oscillations, Cortisol Awakening Response Flattening and Burnout Recovery Kinetics 中文书名:下丘脑垂体肾上腺轴耦合振荡、皮质醇觉醒反应扁平化与倦怠恢复动力学 DOI: https://doi.org/10.5281/zenodo.24090252 Cite: Fan, Y. (2026). HPA-Axis Coupled Oscillations, Cortisol Awakening Response Flattening and Burnout Recovery Kinetics (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090252 Book page: https://5dlibrary.com/books/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/ Scope: Its parameters are illustrative or synthetic; all figures are produced by the programs printed in the book; the book is not medical advice and does not diagnose. Marked boxes separate published evidence, runnable Python calculations (sixteen programs), classical sources and the author's hypotheses. Abstract: Why does chronic pressure first raise cortisol and later let it collapse? Why does catching up on sleep for two days fail to restore a receptor that has been downregulated? How bright does morning light have to be, and at what hour, to actually pull the clock back? This book writes these questions as runnable arithmetic and then asks when the answers hold. Its parameters are illustrative or synthetic; all figures are produced by the programs printed in the book; the book is not medical advice and does not diagnose. A low cortisol with a high corticotropin is both the end state of this book's model and what real adrenal failure looks like, so a flat profile is a reason to be tested and not a conclusion to be reached alone. Chapter 1 builds a delayed negative feedback loop out of the published half-lives of three hormones: above a delay of 9.8 minutes it oscillates, and at a fifteen-minute delay its period of 171.5 minutes is 2.29 times the observed sixty to ninety. Chapter 2 shows why the pulses exist: the same daily total in the sharpest pulses holds mean occupancy to 0.1132 of a flat infusion, while peak over mean rises from 1.000 to 16.667. Chapter 3 writes the receptor as a variable that counts itself down: at 2.5 times the normal mean, 0.7955 of the receptors are left, recovery has a time constant of 11.20 days, closing 0.90 of the gap takes 25.8 days and 0.95 takes 33.6, while two days of rest close only 0.1635. Chapter 4 couples receptor downregulation to adrenal wear and gets three stages: cortisol above 1.30 on days 0 to 62, between them on days 63 to 178, and collapse below 0.80 from day 179, ending at 0.5680 with corticotropin raised 3.19 times. Chapter 5 shows the five-on-two-off cycle settling at 0.8598 of rested and recovering only 0.3144 of the gap, while cutting the load by a third every day reaches only 0.8357. Chapter 6 shows light fixed at seven in the morning making a three-hour delay worse, while light placed against the person's own temperature minimum closes it in eight to fourteen days; ten thousand lux for thirty minutes is worth 1.4713 times three hundred lux for thirty minutes but only 1.0299 times that three hundred lux left on for two hours. Chapter 7 reads the Suwen's complaint about taking the reckless as the rule alongside Zhang Jingyue on the water and fire of the gate of life: a bedtime spread must stay within 21 minutes for 0.95 of days to fall inside an hour-wide window, and the steady capacity is one minus 3.6000 times the demanded output, so above an output of 0.2778 no steady state exists at all. Chapter 8 shows realistic adherence giving up 0.5014 of the gap and the safety gate stopping or referring 0.6315 of adults. Marked boxes separate published evidence, runnable Python calculations (sixteen programs), classical sources and the author's hypotheses. 摘要:为什么慢性高压先让皮质醇升高,后来又让它坍塌?为什么周末补睡两天恢复不了已经下调的受体?晨间强光要多亮、要在哪个时辰才真正把生物钟拉回来?本书把这些问题写成可以运行的算术,再看答案在什么条件下成立。书中参数均为示意或合成;全部数字由书中程序产生;本书不是医疗建议,不作诊断。皮质醇低而促肾上腺皮质激素高,既是本书模型的终局,也是真实肾上腺衰竭的样子,所以扁平的曲线是去检查的理由,不是自我诊断的结论。 第一章用三种激素的公开半衰期搭出带延迟的负反馈环:延迟超过 9.8 分钟它开始振荡,延迟为十五分钟时周期 171.5 分钟,比观测到的六十至九十分钟慢 2.29 倍。第二章说明脉冲为什么存在:同样的日总量以最陡的脉冲送达,可以把平均受体占据压到平铺输注的 0.1132,而峰值与均值之比从 1.000 升到 16.667。第三章把受体写成会自己计数下去的变量:皮质醇为常态 2.5 倍时留下 0.7955 的受体,恢复的时间常数是 11.20 天,闭合缺口的 0.90 需要 25.8 天,0.95 需要 33.6 天,而两天休息只闭合 0.1635。 第四章把受体下调与肾上腺损耗耦合起来,得到三个阶段:皮质醇高于 1.30 的第 0 至 62 天、介于两者之间的第 63 至 178 天,以及自第 179 天起低于 0.80 的坍塌;负荷结束时皮质醇为 0.5680 而促肾上腺皮质激素升到 3.19 倍。第五章显示五天上班两天休息的循环稳定在静息值的 0.8598,只收回缺口的 0.3144,而把负荷减三分之一反而只到 0.8357。第六章指出固定在早上七点的强光会让已经延迟三小时的人更延迟,而按本人体温低点定时则八到十四天可以归位;一万勒克斯三十分钟只值三百勒克斯三十分钟的 1.4713 倍,却只值同样三百勒克斯两小时的 1.0299 倍。第七章与《素问·上古天真论》「以妄为常」「不知持满」及张介宾论命门水火并读:就寝时刻的标准差必须在 21 分钟以内,才有九成五的日子落在一小时的窗口内;而稳定能力等于一减去 3.6000 乘以所求产出,产出超过 0.2778 时没有任何稳态。第八章显示现实依从性放弃缺口的 0.5014,而安全闸拦下或转诊 0.6315 的成年人。书中以标注框区分已发表的证据、可运行的 Python 计算(十六段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A Loop with a Delay — Chapter 1 builds a delayed negative feedback loop out of the published half-lives of three hormones: above a delay of 9.8 minutes it oscillates, and at a fifteen-minute delay its period of 171.5 minutes is 2.29 times the observed sixty to ninety. (https://5dlibrary.com/answers/en/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c01/) 第1章 一个带延迟的环 — 第一章用三种激素的公开半衰期搭出带延迟的负反馈环:延迟超过 9.8 分钟它开始振荡,延迟为十五分钟时周期 171.5 分钟,比观测到的六十至九十分钟慢 2.29 倍。 (https://5dlibrary.com/answers/zh/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c01/) - Ch.2 What the Pulses Are For — Chapter 2 shows why the pulses exist: the same daily total in the sharpest pulses holds mean occupancy to 0.1132 of a flat infusion, while peak over mean rises from 1.000 to 16.667. (https://5dlibrary.com/answers/en/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c02/) 第2章 这些脉冲是为了什么 — 第二章说明脉冲为什么存在:同样的日总量以最陡的脉冲送达,可以把平均受体占据压到平铺输注的 0.1132,而峰值与均值之比从 1.000 升到 16.667。 (https://5dlibrary.com/answers/zh/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c02/) - Ch.3 The Receptor Counts Itself Down — Chapter 3 writes the receptor as a variable that counts itself down: at 2.5 times the normal mean, 0.7955 of the receptors are left, recovery has a time constant of 11.20 days, closing 0.90 of the gap takes 25.8 days and 0.95 takes 33.6, while two days of rest close only 0.1635. (https://5dlibrary.com/answers/en/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c03/) 第3章 受体把自己数下去 — 第三章把受体写成会自己计数下去的变量:皮质醇为常态 2.5 倍时留下 0.7955 的受体,恢复的时间常数是 11.20 天,闭合缺口的 0.90 需要 25.8 天,0.95 需要 33.6 天,而两天休息只闭合 0.1635。 (https://5dlibrary.com/answers/zh/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c03/) - Ch.4 Three Stages, and a Measure That Disagrees with Itself — Chapter 4 couples receptor downregulation to adrenal wear and gets three stages: cortisol above 1.30 on days 0 to 62, between them on days 63 to 178, and collapse below 0.80 from day 179, ending at 0.5680 with corticotropin raised 3.19 times. (https://5dlibrary.com/answers/en/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c04/) 第4章 三个阶段,以及一个自相矛盾的测量 — 第四章把受体下调与肾上腺损耗耦合起来,得到三个阶段:皮质醇高于 1.30 的第 0 至 62 天、介于两者之间的第 63 至 178 天,以及自第 179 天起低于 0.80 的坍塌;负荷结束时皮质醇为 0.5680 而促肾上腺皮质激素升到 3.19 倍。 (https://5dlibrary.com/answers/zh/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c04/) - Ch.5 Five On, Two Off — Chapter 5 shows the five-on-two-off cycle settling at 0.8598 of rested and recovering only 0.3144 of the gap, while cutting the load by a third every day reaches only 0.8357. (https://5dlibrary.com/answers/en/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c05/) 第5章 五天上,两天下 — 第五章显示五天上班两天休息的循环稳定在静息值的 0.8598,只收回缺口的 0.3144,而把负荷减三分之一反而只到 0.8357。 (https://5dlibrary.com/answers/zh/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c05/) - Ch.6 The Variable That Moves in Days — Chapter 6 shows light fixed at seven in the morning making a three-hour delay worse, while light placed against the person's own temperature minimum closes it in eight to fourteen days; ten thousand lux for thirty minutes is worth 1.4713 times three hundred lux for thirty minutes but only 1.0299 times that three hundred lux left on for two hours. (https://5dlibrary.com/answers/en/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c06/) 第6章 那个以天计移动的变量 — 第六章指出固定在早上七点的强光会让已经延迟三小时的人更延迟,而按本人体温低点定时则八到十四天可以归位;一万勒克斯三十分钟只值三百勒克斯三十分钟的 1.4713 倍,却只值同样三百勒克斯两小时的 1.0299 倍。 (https://5dlibrary.com/answers/zh/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c06/) - Ch.7 Taking the Reckless as the Rule — Chapter 7 reads the Suwen's complaint about taking the reckless as the rule alongside Zhang Jingyue on the water and fire of the gate of life: a bedtime spread must stay within 21 minutes for 0.95 of days to fall inside an hour-wide window, and the steady capacity is one minus 3.6000 times the demanded output, so above an output of 0.2778 no steady state exists at all. (https://5dlibrary.com/answers/en/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c07/) 第7章 以妄为常 — 第七章与《素问·上古天真论》「以妄为常」「不知持满」及张介宾论命门水火并读:就寝时刻的标准差必须在 21 分钟以内,才有九成五的日子落在一小时的窗口内;而稳定能力等于一减去 3.6000 乘以所求产出,产出超过 0.2778 时没有任何稳态。 (https://5dlibrary.com/answers/zh/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c07/) - Ch.8 The Month, and the Gate — Chapter 8 shows realistic adherence giving up 0.5014 of the gap and the safety gate stopping or referring 0.6315 of adults. Marked boxes separate published evidence, runnable Python calculations (sixteen programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c08/) 第8章 这个月,以及那道闸 — 第八章显示现实依从性放弃缺口的 0.5014,而安全闸拦下或转诊 0.6315 的成年人。书中以标注框区分已发表的证据、可运行的 Python 计算(十六段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/hpa-axis-oscillations-cortisol-awakening-flattening-burnout-recovery-kinetics/c08/) ## 5DL-B-098 · Regime-Switching Fractional Kelly, Maximum-Drawdown Constraints and Lifecycle Leverage Allocation 中文书名:体制转换分数凯利、最大回撤约束与生命周期杠杆配置 DOI: https://doi.org/10.5281/zenodo.24090256 Cite: Fan, Y. (2026). Regime-Switching Fractional Kelly, Maximum-Drawdown Constraints and Lifecycle Leverage Allocation (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090256 Book page: https://5dlibrary.com/books/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/ Scope: Its model parameters are illustrative or synthetic; the only historical figure is a one-day fall in 1987; all calculations are generated by its programs; the book is not investment, legal or tax advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: Why does a full Kelly bet computed in good times so often end in a deep drawdown when the market turns? What fraction of Kelly keeps a ten-year maximum drawdown within a stated limit? How should leverage change as human capital is turned into financial wealth? This book writes these claims as runnable arithmetic and then asks when they hold. Its model parameters are illustrative or synthetic; the only historical figure is a one-day fall in 1987; all calculations are generated by its programs; the book is not investment, legal or tax advice. Chapter 1 shows that the standard error of a drift estimated from ten years of data is 0.0506, as large as the drift itself, and that a Kelly estimated after a three-year bull run has a median of 5.69 against a whole-history 1.46. Chapter 2 shows that at the true Kelly wealth ever halves with chance 0.500 and only from twice Kelly is every low certain; but a full Kelly computed from bull months alone is 7.10 times wealth, and in a two-regime market it halved on every path within ten years and was wiped out on 0.640 of them. Chapter 3 finds that even with the true parameters a bear regime takes about two months to recognize and 0.193 of bear spells are missed. Chapter 4 turns "a ten-year drawdown above 20% with chance at most 0.05" into a fraction of Kelly: 0.20 in an i.i.d. market but only 0.10 in the regime market; a floor rule that bets the cushion is more reliable and grows faster. Chapter 5 checks Ralph Vince's optimal f: growth turns negative above f = 0.594, twice the optimum of 0.300, and when it is estimated from 30 trades half of the estimate does better. Chapter 6 builds a balance sheet with human capital: when layoffs come with bear markets, early leverage of two to one raises median wealth at 65 to 1,725 but gives 0.783 of lives a drawdown of half before 40. Chapter 7 reads the Shiji's "when things flourish they decline" and Feng Xuan's three burrows: three refuges cut risk a hundredfold only if they fail independently. Chapter 8 gives a ten-year leverage table, gap risk of the kind seen in 1987 and a drawdown warning rule. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:在顺风时算出的全凯利仓位,为什么常在市场转向时以深度回撤收场?多大比例的凯利能把十年最大回撤控制在既定界限之内?随着人力资本逐渐转成金融财富,杠杆应当怎样变化?本书把这些说法写成可以运行的算术,再看它们在什么条件下成立。书中的模型参数均为示意或合成模型;唯一的历史数字是 1987 年的单日跌幅;全部计算由书中程序生成;本书不是投资、法律或税务建议。 第一章说明用十年数据估计漂移的标准误是 0.0506,与漂移本身一样大,而在三年牛市之后估计,凯利的中位数是 5.69,全历史只有 1.46。第二章证明在真实凯利下财富曾经腰斩的机会是 0.500,只有从两倍凯利起任何低点才必然到达;但只用牛市月份算出的全凯利是财富的 7.10 倍,在两体制市场中十年内每条路径都腰斩,0.640 的路径归零。第三章发现即使知道真实参数,熊市体制也要约两个月才被识别,0.193 的熊市段落被漏掉。第四章把「十年回撤超过 20% 的机会不超过 0.05」换算成凯利比例:独立同分布市场中是 0.20,体制市场中只有 0.10;对缓冲下注的地板规则更可靠也增长更快。 第五章核对拉尔夫·文斯的最优 f:f 超过 0.594 后增长转负,恰为最优值 0.300 的两倍;用 30 笔交易估计时,用估计值的一半反而更好。第六章建立含人力资本的资产负债表:裁员与熊市同时到来时,二比一的早期杠杆使 65 岁财富中位数升到 1,725,却让 0.783 的人生在 40 岁前经历腰斩。第七章与《史记·平准书》的「物盛则衰」和冯谖的「狡兔三窟」并读:三窟只有在独立失效时才把风险降到百分之一。第八章给出十年杠杆表、1987 年那样的跳空风险与一条回撤预警规则。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Full Kelly and the Optimistic Drift: Why a Good Year Breeds Overbetting — Chapter 1 shows that the standard error of a drift estimated from ten years of data is 0.0506, as large as the drift itself, and that a Kelly estimated after a three-year bull run has a median of 5.69 against a whole-history 1.46. (https://5dlibrary.com/answers/en/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c01/) 第1章 全凯利与乐观的漂移:好年景为什么催生过度下注 — 第一章说明用十年数据估计漂移的标准误是 0.0506,与漂移本身一样大,而在三年牛市之后估计,凯利的中位数是 5.69,全历史只有 1.46。 (https://5dlibrary.com/answers/zh/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c01/) - Ch.2 Halving and Worse: First Passage of a Kelly Bettor — Chapter 2 shows that at the true Kelly wealth ever halves with chance 0.500 and only from twice Kelly is every low certain; but a full Kelly computed from bull months alone is 7.10 times wealth, and in a two-regime market it halved on every path within ten years and was wiped out on 0.640 of them. (https://5dlibrary.com/answers/en/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c02/) 第2章 腰斩及更糟:凯利下注者的首达概率 — 第二章证明在真实凯利下财富曾经腰斩的机会是 0.500,只有从两倍凯利起任何低点才必然到达;但只用牛市月份算出的全凯利是财富的 7.10 倍,在两体制市场中十年内每条路径都腰斩,0.640 的路径归零。 (https://5dlibrary.com/answers/zh/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c02/) - Ch.3 Hidden Regimes: How Fast a Bear Can Be Recognized and What a Regime-Aware Kelly Buys — Chapter 3 finds that even with the true parameters a bear regime takes about two months to recognize and 0.193 of bear spells are missed. (https://5dlibrary.com/answers/en/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c03/) 第3章 隐藏的体制:熊市能多快被识别,体制感知的凯利能带来什么 — 第三章发现即使知道真实参数,熊市体制也要约两个月才被识别,0.193 的熊市段落被漏掉。 (https://5dlibrary.com/answers/zh/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c03/) - Ch.4 Betting Under a Drawdown Limit: The Fraction of Kelly That Keeps a Promise — Chapter 4 turns "a ten-year drawdown above 20% with chance at most 0.05" into a fraction of Kelly: 0.20 in an i.i.d. market but only 0.10 in the regime market; a floor rule that bets the cushion is more reliable and grows faster. (https://5dlibrary.com/answers/en/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c04/) 第4章 在回撤界限下下注:能守住承诺的凯利比例 — 第四章把「十年回撤超过 20% 的机会不超过 0.05」换算成凯利比例:独立同分布市场中是 0.20,体制市场中只有 0.10;对缓冲下注的地板规则更可靠也增长更快。 (https://5dlibrary.com/answers/zh/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c04/) - Ch.5 Ralph Vince's Optimal f: Kelly in Trader's Units, and Its Drawdowns — Chapter 5 checks Ralph Vince's optimal f: growth turns negative above f = 0.594, twice the optimum of 0.300, and when it is estimated from 30 trades half of the estimate does better. (https://5dlibrary.com/answers/en/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c05/) 第5章 拉尔夫·文斯的最优 f:交易者单位下的凯利及其回撤 — 第五章核对拉尔夫·文斯的最优 f:f 超过 0.594 后增长转负,恰为最优值 0.300 的两倍;用 30 笔交易估计时,用估计值的一半反而更好。 (https://5dlibrary.com/answers/zh/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c05/) - Ch.6 The Lifecycle Balance Sheet: Human Capital, Early Leverage and the Glide Path — Chapter 6 builds a balance sheet with human capital: when layoffs come with bear markets, early leverage of two to one raises median wealth at 65 to 1,725 but gives 0.783 of lives a drawdown of half before 40. (https://5dlibrary.com/answers/en/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c06/) 第6章 生命周期资产负债表:人力资本、早期杠杆与下滑路径 — 第六章建立含人力资本的资产负债表:裁员与熊市同时到来时,二比一的早期杠杆使 65 岁财富中位数升到 1,725,却让 0.783 的人生在 40 岁前经历腰斩。 (https://5dlibrary.com/answers/zh/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c06/) - Ch.7 "When Things Flourish They Decline", Feng Xuan's Three Burrows, and Vince Reconsidered — In the Markov model a bull run ends within the next twelve months with chance 0.215 whether it is six months old or ten years old. With a Weibull run length of the same mean and shape 2, the chance rises from 0.087 at six months to 0.344 at 50 months and 0.613 at 120 months: "when a time reaches its extreme it turns". Whether market regimes have this property is an empirical question; for business-cycle expansions in the United States the evidence of duration dependence has been mixed (Diebold 1990). A bettor who believes Sima Qian should cut leverage as a bull run ages; a bettor who believes the Markov model should not. Part (b) shows how much the two beliefs differ: by a factor of almost three for a run of 120 months. (https://5dlibrary.com/answers/en/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c07/) 第7章 「物盛则衰」、冯谖的三窟与文斯再考 — 在马尔可夫模型中,无论牛市已持续六个月还是十年,它在接下来十二个月内结束的机会都是 0.215。若持续期服从均值相同、形状参数为 2 的韦布尔分布,这个机会从六个月时的 0.087 升到 50 个月时的 0.344 和 120 个月时的 0.613:「时极而转」。市场体制是否具有这一性质是一个经验问题;就美国经济周期的扩张期而言,持续期依赖的证据有正有反 (Diebold 1990)。相信司马迁的下注者应随牛市变老而降低杠杆;相信马尔可夫模型的下注者则不必。(b) 部分显示两种信念的差别有多大:对于持续 120 个月的牛市,相差将近三倍。 (https://5dlibrary.com/answers/zh/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c07/) - Ch.8 Practice: A Ten-Year Leverage Table, Gap Risk and a Drawdown Warning Rule — Chapter 8 gives a ten-year leverage table, gap risk of the kind seen in 1987 and a drawdown warning rule. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c08/) 第8章 实践:十年杠杆表、跳空风险与回撤预警规则 — 第八章给出十年杠杆表、1987 年那样的跳空风险与一条回撤预警规则。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/regime-switching-fractional-kelly-drawdown-constraints-lifecycle-leverage/c08/) ## 5DL-B-099 · Singer Formant Clustering, Glottal Closure Quotients and Acoustic Authority Psychophysics 中文书名:歌手共振峰簇、声门闭合商与声学权威度心理物理学 DOI: https://doi.org/10.5281/zenodo.24090258 Cite: Fan, Y. (2026). Singer Formant Clustering, Glottal Closure Quotients and Acoustic Authority Psychophysics (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090258 Book page: https://5dlibrary.com/books/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/ Scope: Its acoustic and rating parameters are illustrative or synthetic; all calculations are generated by its programs; the book is not medical advice, and its measures are not a clinical or diagnostic tool. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: How much intelligibility in noise does a voice gain when its third to fifth formants cluster? How reliably do jitter, shimmer, final falls and pause ratios predict listeners' trust? What can fourteen days of straw phonation change, and what evidence would show it? This book writes these claims as runnable arithmetic and then asks when they hold. Its acoustic and rating parameters are illustrative or synthetic; all calculations are generated by its programs; the book is not medical advice, and its measures are not a clinical or diagnostic tool. Chapter 1 uses a source-filter model to show that a formant cluster adds 14.2 dB in the 2 to 4 kHz band while leaving the overall level almost unchanged. Chapter 2 scores it in 75 dB of speech-shaped noise: it is worth 6.7 dB of broadband level, not the popular 12 dB; its real benefit is that the subglottal pressure needed falls from 1.77 to 0.99 kPa. Chapter 3 shows that the often-quoted 1.04% and 3.81% are clinical thresholds for disordered voices, and that a voice with a true jitter of 0.34% measures 1.00% when sampled at 8 kHz without interpolation. Chapter 4 shows that the same recordings have a pause ratio of 0.241 or 0.127 depending on the definition of the shortest pause. Chapter 5 finds that in a synthetic study of 120 recordings the weight of jitter is estimated at -0.123 with standard error 0.318 against a true -0.30, and that with a flatter optimum 60 recordings place the best pause ratio inside 0.18 to 0.25 only 0.402 of the time. Chapter 6 computes straw back pressure, 474 Pa at 3 mm and 30 Pa at 6 mm, and shows that a mean closed-quotient gain of 0.015 needs 69 persons to test. Chapter 7 reads the Record of Music, the five tones of the Guanzi and the "Quan" chapter of the Guiguzi, whose received text contains neither "slow and fast" nor "light and heavy". Chapter 8 gives a measurement protocol for reliability, microphone distance and baseline days. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:第三至第五共振峰聚集时,声音在噪声中能多出多少可懂度?基频微扰、振幅微扰、句末降调与停顿占比,能以多可靠的权重预测听众的信任?十四天的吸管发声能改变什么、需要什么样的证据?本书把这些说法写成可以运行的算术,再看它们在什么条件下成立。书中的声学与评分参数均为示意或合成模型;全部计算由书中程序生成;本书不是医疗建议,书中的指标也不是临床或诊断工具。 第一章用源滤波模型说明,共振峰簇在 2 至 4 kHz 频带增加 14.2 dB,而总声级几乎不变。第二章在 75 dB 的言语谱噪声中为它打分:它只相当于 6.7 dB 的宽带声级,而不是流行说法中的 12 dB;真正的好处是所需声门下压从 1.77 降到 0.99 kPa。第三章说明常被引用的 1.04% 与 3.81% 是区分病理嗓音的临床阈值,而以 8 kHz 采样、不插值时,一个真实基频微扰为 0.34% 的嗓音会测得 1.00%。第四章说明同一批录音的停顿占比可以是 0.241 也可以是 0.127,取决于最短停顿的定义。 第五章发现,在 120 段录音的合成研究中,基频微扰的权重估计为 -0.123、标准误 0.318,真实值却是 -0.30;最优点较平时,60 段录音只有 0.402 的机会把最佳停顿占比定位在 0.18 至 0.25 之内。第六章计算吸管背压:3 毫米时 474 Pa,6 毫米时 30 Pa;闭合商平均增益为 0.015 时需要 69 人才能检验。第七章读《礼记·乐记》、《管子·地员》的三分损益与《鬼谷子·权篇》,并指出通行本《权篇》中并无「缓急」「轻重」二词。第八章给出信度、麦克风距离与基线天数的测量规程。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Speaker's Formant Cluster: What Moving F3, F4 and F5 Together Does to a Spectrum — Chapter 1 uses a source-filter model to show that a formant cluster adds 14.2 dB in the 2 to 4 kHz band while leaving the overall level almost unchanged. (https://5dlibrary.com/answers/en/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c01/) 第1章 演说家共振峰簇:F3、F4、F5 聚到一起时频谱怎样变化 — 第一章用源滤波模型说明,共振峰簇在 2 至 4 kHz 频带增加 14.2 dB,而总声级几乎不变。 (https://5dlibrary.com/answers/zh/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c01/) - Ch.2 Heard Without Shouting? The Cluster in 75 dB of Noise and the Price in Subglottal Pressure — Chapter 2 scores it in 75 dB of speech-shaped noise: it is worth 6.7 dB of broadband level, not the popular 12 dB; its real benefit is that the subglottal pressure needed falls from 1.77 to 0.99 kPa. (https://5dlibrary.com/answers/en/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c02/) 第2章 不必喊叫就能被听见?75 dB 噪声中的共振峰簇与声门下压的代价 — 第二章在 75 dB 的言语谱噪声中为它打分:它只相当于 6.7 dB 的宽带声级,而不是流行说法中的 12 dB;真正的好处是所需声门下压从 1.77 降到 0.99 kPa。 (https://5dlibrary.com/answers/zh/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c02/) - Ch.3 Jitter and Shimmer: What the Numbers Measure and What the Recorder Adds — Chapter 3 shows that the often-quoted 1.04% and 3.81% are clinical thresholds for disordered voices, and that a voice with a true jitter of 0.34% measures 1.00% when sampled at 8 kHz without interpolation. (https://5dlibrary.com/answers/en/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c03/) 第3章 基频微扰与振幅微扰:这些数字测的是什么,录音设备又加进了什么 — 第三章说明常被引用的 1.04% 与 3.81% 是区分病理嗓音的临床阈值,而以 8 kHz 采样、不插值时,一个真实基频微扰为 0.34% 的嗓音会测得 1.00%。 (https://5dlibrary.com/answers/zh/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c03/) - Ch.4 Final Falls and Strategic Pauses: Measuring Two Prosodic Signals — Chapter 4 shows that the same recordings have a pause ratio of 0.241 or 0.127 depending on the definition of the shortest pause. (https://5dlibrary.com/answers/en/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c04/) 第4章 句末降调与战略停顿:两个韵律信号的测量 — 第四章说明同一批录音的停顿占比可以是 0.241 也可以是 0.127,取决于最短停顿的定义。 (https://5dlibrary.com/answers/zh/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c04/) - Ch.5 From Measures to Ratings: What a Regression of Trust on Voice Can Recover — Chapter 5 finds that in a synthetic study of 120 recordings the weight of jitter is estimated at -0.123 with standard error 0.318 against a true -0.30, and that with a flatter optimum 60 recordings place the best pause ratio inside 0.18 to 0.25 only 0.402 of the time. (https://5dlibrary.com/answers/en/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c05/) 第5章 从指标到评分:以信任评分对嗓音做回归能还原什么 — 第五章发现,在 120 段录音的合成研究中,基频微扰的权重估计为 -0.123、标准误 0.318,真实值却是 -0.30;最优点较平时,60 段录音只有 0.402 的机会把最佳停顿占比定位在 0.18 至 0.25 之内。 (https://5dlibrary.com/answers/zh/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c05/) - Ch.6 Semi-Occluded Vocal Tract Exercises: Straw Physics and a Fourteen-Day Closed-Quotient Curve — Chapter 6 computes straw back pressure, 474 Pa at 3 mm and 30 Pa at 6 mm, and shows that a mean closed-quotient gain of 0.015 needs 69 persons to test. (https://5dlibrary.com/answers/en/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c06/) 第6章 半阻塞声道练习:吸管物理与一条十四天的闭合商曲线 — 第六章计算吸管背压:3 毫米时 474 Pa,6 毫米时 30 Pa;闭合商平均增益为 0.015 时需要 69 人才能检验。 (https://5dlibrary.com/answers/zh/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c06/) - Ch.7 "Moved by Things, It Takes Form in Sound": The Record of Music, the Guanzi's Five Tones and the Guiguzi's "Quan" — The Record of Music describes how feeling shapes the voice in listeners' categories; the Guanzi's five tones give intervals within 8 cents of equal temperament; the Guiguzi's "Quan" asks the speaker to adapt to the listener, not to speak slowly or heavily. Computed in the terms of this book, adapting the pause ratio to the audience is worth little compared with pitch. (https://5dlibrary.com/answers/en/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c07/) 第7章 「感于物而动,故形于声」:《乐记》、《管子》的五音与《鬼谷子·权篇》 — 《乐记》以听者的类别描述情感如何塑造声音;《管子》的五音给出与平均律相差不到 8 音分的音程;《鬼谷子·权篇》要求说话者因听者而调整,而不是讲得慢或讲得重。用本书的语言计算,按听众调整停顿占比的价值远小于音高。 (https://5dlibrary.com/answers/zh/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c07/) - Ch.8 Practice: A Measurement Protocol for an Authority Voice Diagnostic — Chapter 8 gives a measurement protocol for reliability, microphone distance and baseline days. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c08/) 第8章 实践:权威嗓音诊断的测量规程 — 第八章给出信度、麦克风距离与基线天数的测量规程。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/singer-formant-clustering-glottal-closure-quotients-acoustic-authority-metrics/c08/) ## 5DL-B-100 · Hundred-Volume Knowledge Hypergraphs, Multimodal Personal Prior Compilation and Autonomous Empire Scaling 中文书名:百卷知识超图拓扑、多模态个人先验编译与自主知识帝国标度律 DOI: https://doi.org/10.5281/zenodo.24090260 Cite: Fan, Y. (2026). Hundred-Volume Knowledge Hypergraphs, Multimodal Personal Prior Compilation and Autonomous Empire Scaling (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090260 Book page: https://5dlibrary.com/books/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/ Scope: Its synthetic library, rates and hazards are illustrative models; all calculations are generated by its programs; the book is not investment, legal, medical or technical advice. Chapter 3 finds that a shared library of primitives shortens a naive description 6.10 times but only 1.04 times against a good general compressor, and that fifteen core primitives cover 82 of the 100 volumes of a synthetic library, full coverage needing 28. Abstract: The hundredth volume of 5D Library takes the library itself as its subject. In what sense does the same mathematics recur across very different domains? How fast can a loop of diagnostic stations, aggregated data and revised books learn? What would it take for the library to stay readable and checkable for a hundred years? This book writes these claims as runnable arithmetic and then asks when they hold. Its synthetic library, rates and hazards are illustrative models; all calculations are generated by its programs; the book is not investment, legal, medical or technical advice. Chapters 1 and 2 show that a cross-disciplinary isomorphism shares a form of calculation, not a cause: the same phase-locking program describes a neural rhythm and a circadian clock whose drives differ by a factor of 3,456,000, and Markov chains, Weibull hazards, the log-sum-exp duality and Bayesian updating all run unchanged across domains while their parameters never transfer. Chapter 3 finds that a shared library of primitives shortens a naive description 6.10 times but only 1.04 times against a good general compressor, and that fifteen core primitives cover 82 of the 100 volumes of a synthetic library, full coverage needing 28. Chapter 4 describes the library as a hypergraph. Chapter 5 shows that a feedback flywheel learns linearly: with a constant 1,000 users a month the standard error after a year is 0.00418; privacy noise with buckets of 30 users keeps only 0.441 of the information; greedy routing ends favouring the worse volume in 0.385 of runs. Chapter 6 computes a century of preservation: digital copies survive a common shock with chance 0.5421, and with printed copies in libraries 0.9980; programs that use several random algorithms reproduce byte for byte after a hundred years with chance only 0.017, so the printed output is what keeps them checkable. Chapter 7 checks that "store it in a famous mountain, with a duplicate in the capital" is from the Shiji's postface and "pass it on to the right people" from the letter to Ren An, that two copies which are recopied survive 2,000 years with chance 0.951229, and that the Taiping yulan has a thousand juan, not a hundred. Chapter 8 shows that zero marginal cost applies only to copies, as consistency checks grow from 30 pairs for 10 volumes to 3,620 for 100, and gives a preservation checklist with yearly checks. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:作为 5D Library 的第一百卷,本书把整个书库当作研究对象:同样的数学在什么意义上反复出现于相去甚远的领域?由诊断站、聚合数据与修订的专著构成的反馈回路能学得多快?书库要在一百年后仍可阅读、可核验需要什么?本书把这些说法写成可以运行的算术,再看它们在什么条件下成立。书中的合成书库、比率与风险率均为示意模型;全部计算由书中程序生成;本书不是投资、法律、医疗或技术建议。 第一、二章说明跨学科同构共享的是计算形式而不是原因:同一段锁相程序描述驱动频率相差 3,456,000 倍的神经节律与昼夜节律;马尔可夫链、威布尔风险率、对数求和指数的对偶与贝叶斯更新都能原封不动地跨领域运行,参数却从不迁移。第三章发现共享原语库使朴素描述缩短 6.10 倍,但与好的通用压缩器相比只有 1.04 倍;十五个核心原语覆盖合成书库 100 卷中的 82 卷,完整覆盖需要 28 个。第四章把书库描述为超图。第五章说明反馈飞轮线性地学习:每月固定 1,000 名用户,一年后标准误是 0.00418;隐私噪声在每桶 30 人时只留下 0.441 的信息;贪婪路由在 0.385 的运行中偏向较差的一卷。 第六章计算一个世纪的保存:数字副本遇共同冲击时存续机会是 0.5421,加上存于图书馆的纸本是 0.9980;使用多种随机算法的程序一百年后逐字节复现的机会只有 0.017,所以印出的输出才是可核验性的保障。第七章核对「藏之名山,副在京师」出自《太史公自序》、「传之其人」出自《报任少卿书》,两份会被重抄的副本存续 2,000 年的机会是 0.951229,并指出《太平御览》是一千卷而非一百卷。第八章说明零边际成本只适用于副本,一致性核对从 10 卷的 30 对增长到 100 卷的 3,620 对,并给出每年校验的保存清单。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 One Equation, Many Worlds: What a Cross-Disciplinary Isomorphism Is and Is Not — Chapters 1 and 2 show that a cross-disciplinary isomorphism shares a form of calculation, not a cause: the same phase-locking program describes a neural rhythm and a circadian clock whose drives differ by a factor of 3,456,000, and Markov chains, Weibull hazards, the log-sum-exp duality and Bayesian updating all run unchanged across domains while their parameters never transfer. (https://5dlibrary.com/answers/en/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c01/) 第1章 一个方程,许多世界:跨学科同构是什么、不是什么 — 第一、二章说明跨学科同构共享的是计算形式而不是原因:同一段锁相程序描述驱动频率相差 3,456,000 倍的神经节律与昼夜节律;马尔可夫链、威布尔风险率、对数求和指数的对偶与贝叶斯更新都能原封不动地跨领域运行,参数却从不迁移。 (https://5dlibrary.com/answers/zh/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c01/) - Ch.2 Three More Primitives: Weibull Hazards, Legendre-Bregman Duality and Bayesian Updating — The fits recover the shapes (0.60, 0.91 and 1.81 against true 0.60, 0.90 and 1.80) from 2,000 cases each. The hazard of the video viewer falls from 0.096 at one second to 0.038 at ten; that of the subscriber falls a little; that of the storage medium rises from 0.020 to 0.131 a year. "Retention" and "bit rot" are the same calculation with opposite shapes. (https://5dlibrary.com/answers/en/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c02/) 第2章 另外三个原语:威布尔风险率、勒让德-布雷格曼对偶与贝叶斯更新 — 拟合各从 2,000 个案例中还原了形状参数(0.60、0.91 与 1.81,真实值为 0.60、0.90 与 1.80)。视频观众的风险率从第一秒的 0.096 降到第十秒的 0.038;订阅者略有下降;存储介质则从每年 0.020 升到 0.131。「留存」与「比特腐败」是形状相反的同一个计算。 (https://5dlibrary.com/answers/zh/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c02/) - Ch.3 Compression and Coverage: How Much Do Fifteen Primitives Really Explain? — Chapter 3 finds that a shared library of primitives shortens a naive description 6.10 times but only 1.04 times against a good general compressor, and that fifteen core primitives cover 82 of the 100 volumes of a synthetic library, full coverage needing 28. (https://5dlibrary.com/answers/en/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c03/) 第3章 压缩与覆盖:十五个原语究竟解释了多少? — 第三章发现共享原语库使朴素描述缩短 6.10 倍,但与好的通用压缩器相比只有 1.04 倍;十五个核心原语覆盖合成书库 100 卷中的 82 卷,完整覆盖需要 28 个。 (https://5dlibrary.com/answers/zh/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c03/) - Ch.4 The Library as a Hypergraph: Degrees, Co-Use and Connected Volumes — The mean hyperedge joins 3.44 primitives. Degrees are very unequal: the top three primitives appear in 65, 41 and 34 volumes, the median one in 6, and eight are used once. The co-use graph, which joins two primitives if some volume uses both, has a density of 0.376. The s-line graph shows how volumes connect through shared primitives: sharing one primitive links 3,620 pairs of volumes into a single component; sharing two still leaves 99 volumes connected through 1,269 pairs; sharing three breaks the library into 41 components, the largest with 59 volumes. (https://5dlibrary.com/answers/en/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c04/) 第4章 作为超图的书库:度数、共用与相连的卷 — 平均每条超边连接 3.44 个原语。度数极不均衡:最常用的三个原语分别出现在 65、41 与 34 卷中,中位数原语只出现在 6 卷中,另有八个只用过一次。共用图把被同一卷同时用到的两个原语连起来,其密度是 0.376。s-线图显示各卷如何通过共用原语相连:共用一个原语时,3,620 对卷连成一个整体;共用两个时,仍有 99 卷经由 1,269 对相连;共用三个时,书库分裂成 41 个连通分量,最大的有 59 卷。 (https://5dlibrary.com/answers/zh/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c04/) - Ch.5 The Feedback Flywheel: Fisher Information, Privacy Noise and Self-Confirming Routing — Chapter 5 shows that a feedback flywheel learns linearly: with a constant 1,000 users a month the standard error after a year is 0.00418; privacy noise with buckets of 30 users keeps only 0.441 of the information; greedy routing ends favouring the worse volume in 0.385 of runs. (https://5dlibrary.com/answers/en/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c05/) 第5章 反馈飞轮:费希尔信息、隐私噪声与自我确认的路由 — 第五章说明反馈飞轮线性地学习:每月固定 1,000 名用户,一年后标准误是 0.00418;隐私噪声在每桶 30 人时只留下 0.441 的信息;贪婪路由在 0.385 的运行中偏向较差的一卷。 (https://5dlibrary.com/answers/zh/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c05/) - Ch.6 A Hundred Years: Copies, Common Shocks and What Reproducibility Can Promise — Chapter 6 computes a century of preservation: digital copies survive a common shock with chance 0.5421, and with printed copies in libraries 0.9980; programs that use several random algorithms reproduce byte for byte after a hundred years with chance only 0.017, so the printed output is what keeps them checkable. (https://5dlibrary.com/answers/en/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c06/) 第6章 一百年:副本、共同冲击与可复现性能承诺什么 — 第六章计算一个世纪的保存:数字副本遇共同冲击时存续机会是 0.5421,加上存于图书馆的纸本是 0.9980;使用多种随机算法的程序一百年后逐字节复现的机会只有 0.017,所以印出的输出才是可核验性的保障。 (https://5dlibrary.com/answers/zh/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c06/) - Ch.7 "Store It in a Famous Mountain": Sima Qian, the Library of Alexandria and the Taiping yulan — Chapter 7 checks that "store it in a famous mountain, with a duplicate in the capital" is from the Shiji's postface and "pass it on to the right people" from the letter to Ren An, that two copies which are recopied survive 2,000 years with chance 0.951229, and that the Taiping yulan has a thousand juan, not a hundred. (https://5dlibrary.com/answers/en/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c07/) 第7章 「藏之名山」:司马迁、亚历山大图书馆与《太平御览》 — 第七章核对「藏之名山,副在京师」出自《太史公自序》、「传之其人」出自《报任少卿书》,两份会被重抄的副本存续 2,000 年的机会是 0.951229,并指出《太平御览》是一千卷而非一百卷。 (https://5dlibrary.com/answers/zh/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c07/) - Ch.8 Practice: Zero Marginal Cost, Rising Consistency Cost and a Fixity Schedule — Chapter 8 shows that zero marginal cost applies only to copies, as consistency checks grow from 30 pairs for 10 volumes to 3,620 for 100, and gives a preservation checklist with yearly checks. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c08/) 第8章 实践:零边际成本、上升的一致性成本与校验时间表 — 第八章说明零边际成本只适用于副本,一致性核对从 10 卷的 30 对增长到 100 卷的 3,620 对,并给出每年校验的保存清单。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/hundred-volume-knowledge-hypergraphs-personal-prior-compilation-empire-scaling/c08/) ## 5DL-B-101 · Computational Bibliometrics, Falsification-First Rhetoric and Self-Verifying Monograph Compilers 中文书名:计算著作学、先证伪后立界修辞与自校验专著编译器 DOI: pending Cite: Fan, Y. (2026). Computational Bibliometrics, Falsification-First Rhetoric and Self-Verifying Monograph Compilers (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its readers, books, error rates and costs are illustrative models; all calculations are generated by its programs; the book is not legal, financial or publishing advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This book turns the methods of 5D Library on the library itself. Why do most books stop being read within a few years while a few keep being read and cited? Does opening with a refuted overclaim earn more trust than self-praise? Can a bilingual manuscript be compiled so that every printed number is checked against the program that produced it? The book writes these claims as runnable arithmetic and then asks when they hold. Its readers, books, error rates and costs are illustrative models; all calculations are generated by its programs; the book is not legal, financial or publishing advice. Chapter 1 shows that "three years to waste paper" is a parameter, not a law: a trend book with a rising hazard keeps 0.0055 of its first-year attention at year 3 and a reference book with a falling hazard 0.3330, yet a trend book that starts more than 11.08 times larger is never overtaken; and citation totals are finite. Chapter 2 measures surprise in bits: a correction from 10 times to 1.35 carries 12.65 bits; the best steady dose in a toy reader is 1.2 bits a section; and sales data overstate the value of anchors 3.39 times. Chapter 3 gives the Bayesian case for "falsify, then bound": with a likelihood ratio of 12, tripling trust needs a prior at or below 0.273; when only significant small studies are published, a true ratio of 1.35 has a median published value of 11.89. Chapter 4 shows that one found error cancels 17.8 good checks. Chapter 5 shows that when errors differ in visibility three human passes still leave 0.14950 of them, and that a whitelist cannot catch swaps between true numbers. Chapter 6 surveys the code points that print like ideographs (all 214 Kangxi radicals) and computes that the whole pipeline takes a draft with 81 wrong numbers to an expected 1.322, with a chance of no error of 0.2665, not zero. Chapter 7 checks Liu Xie's six virtues and his chapter on cutting, "without evidence, no belief" in the Doctrine of the Mean, Dai Zhen's letter on learning and Knuth's literate programming. Chapter 8 prices the gates, shows that ten errors leave a reference book 0.608 of its long-run attention, and gives a checklist. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本书把 5D Library 的方法用在书库自身:为什么大多数书几年后就无人再读,而少数书一直被阅读和引用?以证伪一个流行的夸大说法开篇,是否比自我夸耀赢得更多信任?一部双语书稿能否编译成每个印出的数字都对照产生它的程序核对过?本书把这些说法写成可以运行的算术,再看它们在什么条件下成立。书中的读者、书、错误率与成本均为示意模型;全部计算由书中程序生成;本书不是法律、财务或出版建议。 第一章说明「三年沦为废纸」是参数而不是定律:风险率上升的潮流书第 3 年只剩第一年注意力的 0.0055,风险率下降的参考书剩 0.3330,但起点大于 11.08 倍的潮流书永远不会被追上;引用的总量是有限的。第二章以比特度量惊奇:一次把 10 倍改为 1.35 倍的修正带来 12.65 比特;玩具读者模型中最佳的稳定剂量是每节 1.2 比特;销售数据把锚点的价值高估 3.39 倍。第三章给出「先证伪,后立界」的贝叶斯理由:似然比为 12 时,信任变成三倍需要先验不高于 0.273;只发表显著小型研究时,真实比值 1.35 的发表中位数是 11.89 倍。第四章说明发现一处错误抵消 17.8 次核对无误。 第五章说明可见度各不相同时,三遍人工校对后仍剩 0.14950 的错误,而白名单抓不到真实数字之间的调换。第六章清点印出来像汉字的码位(全部 214 个康熙部首),并计算整条流水线把含 81 个错误数字的草稿降到预期 1.322 个,全无错误的机会是 0.2665,而不是零。第七章核对《文心雕龙》的宗经六义与熔裁、《中庸》「无征不信」、戴震的论学书与克努特的文学编程。第八章为门禁算账,说明十处错误使参考书只剩长期注意力的 0.608,并给出一份清单。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Two Survival Laws: Why Most Books Fade and a Few Keep Being Read — Chapter 1 shows that "three years to waste paper" is a parameter, not a law: a trend book with a rising hazard keeps 0.0055 of its first-year attention at year 3 and a reference book with a falling hazard 0.3330, yet a trend book that starts more than 11.08 times larger is never overtaken; and citation totals are finite. (https://5dlibrary.com/answers/en/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c01/) 第1章 两种存续规律:为什么多数书褪色,少数书一直被读 — 第一章说明「三年沦为废纸」是参数而不是定律:风险率上升的潮流书第 3 年只剩第一年注意力的 0.0055,风险率下降的参考书剩 0.3330,但起点大于 11.08 倍的潮流书永远不会被追上;引用的总量是有限的。 (https://5dlibrary.com/answers/zh/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c01/) - Ch.2 Surprise, Anchors and Completion — Chapter 2 measures surprise in bits: a correction from 10 times to 1.35 carries 12.65 bits; the best steady dose in a toy reader is 1.2 bits a section; and sales data overstate the value of anchors 3.39 times. (https://5dlibrary.com/answers/en/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c02/) 第2章 惊奇、锚点与完读 — 第二章以比特度量惊奇:一次把 10 倍改为 1.35 倍的修正带来 12.65 比特;玩具读者模型中最佳的稳定剂量是每节 1.2 比特;销售数据把锚点的价值高估 3.39 倍。 (https://5dlibrary.com/answers/zh/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c02/) - Ch.3 Falsify, Then Bound: The Bayesian Case and Its Limits — Chapter 3 gives the Bayesian case for "falsify, then bound": with a likelihood ratio of 12, tripling trust needs a prior at or below 0.273; when only significant small studies are published, a true ratio of 1.35 has a median published value of 11.89. (https://5dlibrary.com/answers/en/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c03/) 第3章 先证伪,后立界:贝叶斯理由及其限度 — 第三章给出「先证伪,后立界」的贝叶斯理由:似然比为 12 时,信任变成三倍需要先验不高于 0.273;只发表显著小型研究时,真实比值 1.35 的发表中位数是 11.89 倍。 (https://5dlibrary.com/answers/zh/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c03/) - Ch.4 The Trust Ratchet Turns Both Ways — Chapter 4 shows that one found error cancels 17.8 good checks. (https://5dlibrary.com/answers/en/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c04/) 第4章 信任棘轮两个方向都会转 — 第四章说明发现一处错误抵消 17.8 次核对无误。 (https://5dlibrary.com/answers/zh/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c04/) - Ch.5 Proofreading Floors and Whitelist Gates — Chapter 5 shows that when errors differ in visibility three human passes still leave 0.14950 of them, and that a whitelist cannot catch swaps between true numbers. (https://5dlibrary.com/answers/en/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c05/) 第5章 校对的下限与白名单门禁 — 第五章说明可见度各不相同时,三遍人工校对后仍剩 0.14950 的错误,而白名单抓不到真实数字之间的调换。 (https://5dlibrary.com/answers/zh/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c05/) - Ch.6 The Self-Verifying Compiler: Code Points and the Whole Pipeline — Chapter 6 surveys the code points that print like ideographs (all 214 Kangxi radicals) and computes that the whole pipeline takes a draft with 81 wrong numbers to an expected 1.322, with a chance of no error of 0.2665, not zero. (https://5dlibrary.com/answers/en/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c06/) 第6章 自校验编译器:码位与整条流水线 — 第六章清点印出来像汉字的码位(全部 214 个康熙部首),并计算整条流水线把含 81 个错误数字的草稿降到预期 1.322 个,全无错误的机会是 0.2665,而不是零。 (https://5dlibrary.com/answers/zh/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c06/) - Ch.7 Liu Xie, Dai Zhen and Knuth: Old Rules for Checkable Writing — A true gloss fits each passage with chance 0.95, a plausible false one with chance 0.50. Starting from a prior of 0.30, one fitting passage moves belief only to 0.4488; four passages to 0.8481; eight to 0.9864; and 9 are needed for 0.99. A false gloss survives eight passages with chance 0.00391. The bilingual gate of Chapter 6 is the same test with two "passages" for every number: the Chinese and English text must print the same set. (https://5dlibrary.com/answers/en/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c07/) 第7章 刘勰、戴震与克努特:可核验写作的古老规矩 — 真确的释义在每一处吻合的机会是 0.95,貌似有理的错误释义是 0.50。从 0.30 的先验出发,一处吻合只把信念提到 0.4488;四处到 0.8481;八处到 0.9864;要到 0.99 需要 9 处。错误的释义通过八处检验的机会是 0.00391。第六章的双语门禁是同样的检验,每个数字都有两「处」:中文与英文正文必须印出同一个数字集合。 (https://5dlibrary.com/answers/zh/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c07/) - Ch.8 Practice: Costs, the Long Tail and a Checklist — Chapter 8 prices the gates, shows that ten errors leave a reference book 0.608 of its long-run attention, and gives a checklist. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c08/) 第8章 实践:成本、长尾与清单 — 第八章为门禁算账,说明十处错误使参考书只剩长期注意力的 0.608,并给出一份清单。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/computational-bibliometrics-falsification-first-rhetoric-self-verifying-monographs/c08/) ## 5DL-B-102 · Cross-Disciplinary Structural Holes, Functorial Knowledge Transfer and Inspiration Stochastic Resonance 中文书名:跨学科结构洞桥接、函子知识迁移与灵感随机共振 DOI: pending Cite: Fan, Y. (2026). Cross-Disciplinary Structural Holes, Functorial Knowledge Transfer and Inspiration Stochastic Resonance (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its networks, fields, scores and potentials are illustrative models; all calculations are generated by its programs; the book is not investment, medical or psychological advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This volume asks whether the leaps that join distant fields, the moments Koestler called bisociation, can be found on purpose and at scale. It studies the network positions from which such leaps come, the scores a machine might use to propose them, and the role of rest and noise in reaching them, writing popular claims as runnable arithmetic and asking when they hold. Its networks, fields, scores and potentials are illustrative models; all calculations are generated by its programs; the book is not investment, medical or psychological advice. Chapter 1 computes Burt's constraint: spanning a structural hole lowers it from 0.629 to 0.360, and in a clustered network of researchers rewiring 0.01 of ties cuts the mean distance from 17.09 steps to 9.98. Chapter 2 shows that distance buys variance, not a better average: in the model the breakthrough chance peaks at distance 5 (0.02275), with 23.46 failures per breakthrough at distance 4, and the best project puts only 0.3 of its parts at a distance. Chapter 3 finds that percolation on the 6-cube formed by the 64 hexagrams behaves almost like a random graph with the same edges, so an exact shared operator can predict nothing new, and that equal eigenvalues do not prove isomorphism. Chapter 4 shows that Jensen-Shannon span is bounded at 1 bit and saturates, and that the winner among 200 analogies chosen by in-sample fit has an R-squared of 0.283 that shrinks to 0.046 on fresh data. Chapter 5 shows that once the score becomes a target the true value of the top ten falls from 1.88 to 0.48 standard deviations, and that with 1 real pair in 500 one test has a precision of only 0.031. Chapter 6 checks Kramers' formula by simulation (within a factor of 1.32) and finds that too little noise is the main risk: with low noise only 0.017 of runs reach the better state, while the incubate-then-focus rhythm (0.730) does no better than steady moderate noise (0.750). Chapter 7 checks Su Shi's "observe broadly, take sparingly; accumulate richly, release thinly" and the Appended Phrases on extending by analogy, and shows that the four stages of creativity were named by Wallas, not Poincaré. Chapter 8 prices a monthly pipeline at 2.11 review hours per real analogy kept. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷问:把相距遥远的领域连接起来的跳跃,库斯勒称之为「双联」,能否有意地、成批地找到?本书研究这类跳跃出自网络中的什么位置、机器可以用什么评分来提出它们,以及休息与噪声在其中的作用,把流行的说法写成可以运行的算术,再看它们在什么条件下成立。书中的网络、领域、评分与势阱均为示意模型;全部计算由书中程序生成;本书不是投资、医疗或心理建议。 第一章计算伯特的约束度:跨越结构洞使约束度从 0.629 降到 0.360,而在聚类的研究者网络中,重连 0.01 的关系就把平均距离从 17.09 步缩短到 9.98。第二章说明距离提高的是方差而不是平均:模型中突破的机会在距离 5 处最高(0.02275),距离 4 处每次突破伴随 23.46 次失败;最好的项目只把 0.3 的部件放在远处。第三章发现六十四卦构成的六维立方体上的渗流,几乎与边数相同的随机图一样,精确的共同算子可能什么新东西也预测不了;特征值相同也不证明同构。第四章说明詹森-香农跨度以 1 比特为界、会饱和,从 200 个类比中按样本内拟合挑出的胜者,决定系数从 0.283 缩到新数据上的 0.046。 第五章说明评分成为目标之后,前十名的真实价值从 1.88 个标准差降到 0.48;真实对应每 500 对只有 1 对时,一道测试的精确率只有 0.031。第六章用模拟检验克莱默斯公式(相差不超过 1.32 倍),并发现噪声太少是主要风险:低噪声下只有 0.017 的运行找到更好的状态,而先孵化后专注的节奏(0.730)并不比稳定的中等噪声(0.750)更好。第七章核对苏轼「博观而约取,厚积而薄发」、《系辞》「触类而长之」,并说明创造四阶段是沃拉斯而不是庞加莱命名的。第八章为一条按月运行的流水线算账,每留下一个真类比需要 2.11 个评审小时。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Structural Holes: Where Bridging Ideas Come From — Chapter 1 computes Burt's constraint: spanning a structural hole lowers it from 0.629 to 0.360, and in a clustered network of researchers rewiring 0.01 of ties cuts the mean distance from 17.09 steps to 9.98. (https://5dlibrary.com/answers/en/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c01/) 第1章 结构洞:桥接的想法从哪里来 — 第一章计算伯特的约束度:跨越结构洞使约束度从 0.629 降到 0.360,而在聚类的研究者网络中,重连 0.01 的关系就把平均距离从 17.09 步缩短到 9.98。 (https://5dlibrary.com/answers/zh/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c01/) - Ch.2 Distance Buys Variance, Not a Better Average — Chapter 2 shows that distance buys variance, not a better average: in the model the breakthrough chance peaks at distance 5 (0.02275), with 23.46 failures per breakthrough at distance 4, and the best project puts only 0.3 of its parts at a distance. (https://5dlibrary.com/answers/en/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c02/) 第2章 距离买到的是方差,不是更好的平均 — 第二章说明距离提高的是方差而不是平均:模型中突破的机会在距离 5 处最高(0.02275),距离 4 处每次突破伴随 23.46 次失败;最好的项目只把 0.3 的部件放在远处。 (https://5dlibrary.com/answers/zh/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c02/) - Ch.3 Functors and Their Limits: What a Shared Operator Transfers — The cube has 192 edges and every node has degree 6. Keeping each edge with chance $p$, the largest cluster grows from 0.053 of the nodes at $p$ = 0.05 to 0.488 at 0.25 and 0.980 at 0.50. A random graph with the same edges gives 0.056, 0.558 and 0.946. The two curves are close, and both turn upward near 1/6 = 0.167, the threshold that mean degree alone predicts. The shared operator is real, but almost everything it predicts follows from a number, the degree, that any graph with six links per node has. It says nothing about the meaning of the hexagrams, and the hexagrams add nothing to percolation theory. That is the usual fate of an exact but shallow correspondence: it transfers a calculation, not an insight. (https://5dlibrary.com/answers/en/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c03/) 第3章 函子及其限度:共同算子迁移了什么 — 立方体有 192 条边,每个节点的度都是 6。每条边以机会 $p$ 保留时,最大团簇从 $p$ = 0.05 时占节点的 0.053,增长到 0.25 时的 0.488 与 0.50 时的 0.980。边数相同的随机图给出 0.056、0.558 与 0.946。两条曲线很接近,都在 1/6 = 0.167 附近转而上升,这正是只靠平均度就能预测的阈值。共同算子是真实的,但它所预测的几乎一切,都来自一个数字,即度,而任何每个节点有六条边的图都有这个数字。它对卦的意义一无所说,卦对渗流理论也毫无增益。这是精确而肤浅的对应常有的命运:它迁移的是一种计算,而不是一种洞见。 (https://5dlibrary.com/answers/zh/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c03/) - Ch.4 Two Parts of the Score: Semantic Span and Predictive Gain — Chapter 4 shows that Jensen-Shannon span is bounded at 1 bit and saturates, and that the winner among 200 analogies chosen by in-sample fit has an R-squared of 0.283 that shrinks to 0.046 on fresh data. (https://5dlibrary.com/answers/en/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c04/) 第4章 评分的两部分:语义跨度与预测力增益 — 第四章说明詹森-香农跨度以 1 比特为界、会饱和,从 200 个类比中按样本内拟合挑出的胜者,决定系数从 0.283 缩到新数据上的 0.046。 (https://5dlibrary.com/answers/zh/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c04/) - Ch.5 When the Score Becomes the Target — Chapter 5 shows that once the score becomes a target the true value of the top ten falls from 1.88 to 0.48 standard deviations, and that with 1 real pair in 500 one test has a precision of only 0.031. (https://5dlibrary.com/answers/en/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c05/) 第5章 当评分变成目标 — 第五章说明评分成为目标之后,前十名的真实价值从 1.88 个标准差降到 0.48;真实对应每 500 对只有 1 对时,一道测试的精确率只有 0.031。 (https://5dlibrary.com/answers/zh/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c05/) - Ch.6 Noise, Escape and Keeping: Kramers and the Incubation Claim — Chapter 6 checks Kramers' formula by simulation (within a factor of 1.32) and finds that too little noise is the main risk: with low noise only 0.017 of runs reach the better state, while the incubate-then-focus rhythm (0.730) does no better than steady moderate noise (0.750). (https://5dlibrary.com/answers/en/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c06/) 第6章 噪声、逃逸与留住:克莱默斯与孵化之说 — 第六章用模拟检验克莱默斯公式(相差不超过 1.32 倍),并发现噪声太少是主要风险:低噪声下只有 0.017 的运行找到更好的状态,而先孵化后专注的节奏(0.730)并不比稳定的中等噪声(0.750)更好。 (https://5dlibrary.com/answers/zh/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c06/) - Ch.7 Su Shi, the Appended Phrases, Poincaré and Wallas — The four named stages of creativity are Wallas's, not Poincaré's; Poincaré's own point, that almost all combinations are sterile, is confirmed by counting: a stock of 1000 ideas has 166,167,000 triples. Su Shi's "accumulate richly, release thinly" is a selection ratio: releasing the best tenth of a notebook gives about 1.75 standard deviations whatever its size. (https://5dlibrary.com/answers/en/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c07/) 第7章 苏轼、《系辞》、庞加莱与沃拉斯 — 创造的四个有名称的阶段属于沃拉斯,而不属于庞加莱;庞加莱自己的论点,即几乎所有组合都是不育的,可以用计数证实:1000 个想法有 166,167,000 个三元组。苏轼的「厚积而薄发」是一个筛选比例:从笔记中发表最好的十分之一,无论笔记多大,都得到约 1.75 个标准差。 (https://5dlibrary.com/answers/zh/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c07/) - Ch.8 Practice: A Monthly Pipeline and Its Calibration — Chapter 8 prices a monthly pipeline at 2.11 review hours per real analogy kept. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c08/) 第8章 实践:按月运行的流水线及其校准 — 第八章为一条按月运行的流水线算账,每留下一个真类比需要 2.11 个评审小时。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cross-disciplinary-structural-holes-functorial-knowledge-transfer-inspiration-resonance/c08/) ## 5DL-B-103 · Human-AI Symbiotic Cognition, Parallel Agent Pipelines and Prompt Information Bounds 中文书名:人机共生认知架构、多智能体并行流水线与指令信息论下界 DOI: pending Cite: Fan, Y. (2026). Human-AI Symbiotic Cognition, Parallel Agent Pipelines and Prompt Information Bounds (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its agents, rates, hours and values are illustrative models; all calculations are generated by its programs; the book is not legal, financial or employment advice and makes no claim about any particular AI product. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This volume studies how one person can direct many AI agents well: why default use of a language model tends to produce average work, how much a person's instructions, taste and checks must add, how far parallel agents can speed up a project, and which part of the work cannot be handed over. It writes popular claims as runnable arithmetic and asks when they hold. Its agents, rates, hours and values are illustrative models; all calculations are generated by its programs; the book is not legal, financial or employment advice and makes no claim about any particular AI product. Chapter 1 shows that drift to the average comes from feeding on one's own output: refitted to its own samples for 100 generations, a model keeps a median spread of 0.340; and that taste is worth more than volume, a judge correlating 0.9 with quality picking from 4 drafts getting 0.926, more than a judge at 0.3 picking from 64 (0.703). Chapter 2 counts the bits in an instruction: with 20 attributes that must all be right, 1.0 bit each gives acceptance 0.000484 and 3.0 bits 0.998734; a strong test cuts wrong deliveries from 0.4049 to 0.0134. Chapter 3 applies the Universal Scalability Law to parallel windows: ten isolated windows with a serial share of 0.02 reach 8.47, and ten agents writing to three shared files produce 2.95 useful results per round. Chapter 4 finds the commander to be the serial part: at 1.5 hours per volume, windows beyond 2.0 add nothing. Chapter 5 separates share of time from share of value and shows that an honest sign-off of an error rate below 0.010 takes 299 checked items. Chapter 6 treats an instruction as a noisy channel: at a misreading rate of 0.15, checks covering 0.9 of decisions give 0.6355 at 2.06 uses per decision, better than restating five times (0.4452). Chapter 7 reads the Huainanzi on the ruler who knows by relying on others, the Shiji on Han Xin's "the more the better" and the emperor who "leads leaders", and Shannon: with 100 units a tree of command reaches a speedup of 84.1, a flat one 16.5. Chapter 8 plans a commander's week and a new operator's training, and gives a checklist. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷研究一个人怎样把许多 AI 智能体指挥好:为什么默认地使用语言模型往往得到平庸的作品;一个人的指令、品味与检查必须加上多少;并行智能体能把项目加快到什么程度;哪一部分工作不能交出去。本书把流行的说法写成可以运行的算术,再看它们在什么条件下成立。书中的智能体、比率、工时与价值均为示意模型;全部计算由书中程序生成;本书不是法律、财务或雇佣建议,也不对任何具体的 AI 产品作出断言。 第一章说明向均值漂移来自以自身输出为食:反复对自己的样本重新拟合 100 代后,离散的中位数只剩 0.340;品味比数量更值钱,相关为 0.9 的人从 4 份草稿中挑选得到 0.926,高于相关为 0.3 的人从 64 份中挑选的 0.703。第二章计算指令的比特:20 个属性必须全对时,每个 1.0 比特的接受率是 0.000484,3.0 比特是 0.998734;一道强测试把错误交付从 0.4049 降到 0.0134。第三章把通用可扩展定律用于并行窗口:串行份额为 0.02 时十个隔离窗口达到 8.47,十个智能体写三个共享文件每轮只有 2.95 个有效结果。第四章发现指挥官就是串行部分:每卷需要 1.5 小时时,超过 2.0 个窗口就不再增加产出。 第五章区分时间份额与价值份额,并说明诚实地签发错误率低于 0.010 需要核对 299 项。第六章把指令看作有噪声的信道:误读率为 0.15 时,覆盖 0.9 的决定的检查以每个决定 2.06 次的用量得到 0.6355,胜过重复陈述五次的 0.4452。第七章读《淮南子·主术训》「因物以识物,因人以知人」、《史记》韩信「多多益善」与「善将将」,以及香农:100 个单位时树状指挥的加速比是 84.1,扁平指挥只有 16.5。第八章为指挥官的一周与新操作者的训练做计划,并给出一份清单。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Drift to the Middle, and What Taste Adds — Chapter 1 shows that drift to the average comes from feeding on one's own output: refitted to its own samples for 100 generations, a model keeps a median spread of 0.340; and that taste is worth more than volume, a judge correlating 0.9 with quality picking from 4 drafts getting 0.926, more than a judge at 0.3 picking from 64 (0.703). (https://5dlibrary.com/answers/en/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c01/) 第1章 向中间漂移,以及品味加上了什么 — 第一章说明向均值漂移来自以自身输出为食:反复对自己的样本重新拟合 100 代后,离散的中位数只剩 0.340;品味比数量更值钱,相关为 0.9 的人从 4 份草稿中挑选得到 0.926,高于相关为 0.3 的人从 64 份中挑选的 0.703。 (https://5dlibrary.com/answers/zh/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c01/) - Ch.2 How Many Bits an Instruction Must Carry — Chapter 2 counts the bits in an instruction: with 20 attributes that must all be right, 1.0 bit each gives acceptance 0.000484 and 3.0 bits 0.998734; a strong test cuts wrong deliveries from 0.4049 to 0.0134. (https://5dlibrary.com/answers/en/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c02/) 第2章 一条指令必须携带多少比特 — 第二章计算指令的比特:20 个属性必须全对时,每个 1.0 比特的接受率是 0.000484,3.0 比特是 0.998734;一道强测试把错误交付从 0.4049 降到 0.0134。 (https://5dlibrary.com/answers/zh/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c02/) - Ch.3 Amdahl, Gunther and Parallel Agents — Chapter 3 applies the Universal Scalability Law to parallel windows: ten isolated windows with a serial share of 0.02 reach 8.47, and ten agents writing to three shared files produce 2.95 useful results per round. (https://5dlibrary.com/answers/en/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c03/) 第3章 阿姆达尔、冈瑟与并行智能体 — 第三章把通用可扩展定律用于并行窗口:串行份额为 0.02 时十个隔离窗口达到 8.47,十个智能体写三个共享文件每轮只有 2.95 个有效结果。 (https://5dlibrary.com/answers/zh/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c03/) - Ch.4 The Serial Part Is the Commander — Chapter 4 finds the commander to be the serial part: at 1.5 hours per volume, windows beyond 2.0 add nothing. (https://5dlibrary.com/answers/en/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c04/) 第4章 串行的部分是指挥官 — 第四章发现指挥官就是串行部分:每卷需要 1.5 小时时,超过 2.0 个窗口就不再增加产出。 (https://5dlibrary.com/answers/zh/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c04/) - Ch.5 Share of Time, Share of Value, and Signing Off — Chapter 5 separates share of time from share of value and shows that an honest sign-off of an error rate below 0.010 takes 299 checked items. (https://5dlibrary.com/answers/en/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c05/) 第5章 时间份额、价值份额与签发 — 第五章区分时间份额与价值份额,并说明诚实地签发错误率低于 0.010 需要核对 299 项。 (https://5dlibrary.com/answers/zh/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c05/) - Ch.6 An Instruction as a Noisy Channel — Chapter 6 treats an instruction as a noisy channel: at a misreading rate of 0.15, checks covering 0.9 of decisions give 0.6355 at 2.06 uses per decision, better than restating five times (0.4452). (https://5dlibrary.com/answers/en/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c06/) 第6章 作为有噪声信道的指令 — 第六章把指令看作有噪声的信道:误读率为 0.15 时,覆盖 0.9 的决定的检查以每个决定 2.06 次的用量得到 0.6355,胜过重复陈述五次的 0.4452。 (https://5dlibrary.com/answers/zh/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c06/) - Ch.7 The Huainanzi, Han Xin and Shannon — Chapter 7 reads the Huainanzi on the ruler who knows by relying on others, the Shiji on Han Xin's "the more the better" and the emperor who "leads leaders", and Shannon: with 100 units a tree of command reaches a speedup of 84.1, a flat one 16.5. (https://5dlibrary.com/answers/en/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c07/) 第7章 《淮南子》、韩信与香农 — 第七章读《淮南子·主术训》「因物以识物,因人以知人」、《史记》韩信「多多益善」与「善将将」,以及香农:100 个单位时树状指挥的加速比是 84.1,扁平指挥只有 16.5。 (https://5dlibrary.com/answers/zh/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c07/) - Ch.8 Practice: A Commander's Week and a New Operator's Training — Chapter 8 plans a commander's week and a new operator's training, and gives a checklist. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c08/) 第8章 实践:指挥官的一周与新操作者的训练 — 第八章为指挥官的一周与新操作者的训练做计划,并给出一份清单。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/human-ai-symbiotic-cognition-parallel-agent-pipelines-prompt-information-bounds/c08/) ## 5DL-B-104 · Melanopic Spectral Integration, Blue-Light Melatonin Suppression and Lens Transmittance Optimization 中文书名:黑视蛋白光谱积分、蓝光褪黑素抑制与镜片透射率优化 DOI: pending Cite: Fan, Y. (2026). Melanopic Spectral Integration, Blue-Light Melatonin Suppression and Lens Transmittance Optimization (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its spectra, lenses, people and days are illustrative models; all calculations are generated by its programs; the book is not medical advice, diagnoses or treats no condition, and tinted lenses must never be worn to drive at night. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This volume studies how light at night reaches the body clock through the eye, and what clear or tinted "blue-light" lenses can change about it. It writes popular claims as runnable arithmetic and asks when they hold. Its spectra, lenses, people and days are illustrative models; all calculations are generated by its programs; the book is not medical advice, diagnoses or treats no condition, and tinted lenses must never be worn to drive at night. Chapter 1 builds the photopic curve and the CIE S 026 melanopic curve: the opsin peaks at 480 nm but the action spectrum referred to the cornea peaks at 489 nm, and near 440 nm the melanopic weight is 16.32 times the photopic. Chapter 2 models phones and lamps: a day-mode phone at 40 lux gives 37.3 lux melanopic EDI, night mode 21.5, and violet below 420 nm carries 0.000 to 0.002 of the signal. Chapter 3 turns dose into melatonin suppression: anchored on a 90-minute half-maximal dose of 208 lux, two hours of a phone suppresses a typical person by 0.088 and one five times more sensitive by 0.708, about 1.0 or 8.0 extra minutes of sleep latency when anchored on a controlled trial. Chapter 4 tests lenses: a clear violet-cut lens removes 0.0000 to 0.0003 more than plain glass, an amber lens protects 0.587 and a red one 0.954. Chapter 5 traces the front between protection and colour: the best long-pass lens with a fidelity index of 70 protects 0.364, while a notch lens blocking 440 to 514 nm reaches 0.633 at 70.3. Chapter 6 lays out a day: in 150 lux of warm light an amber lens still passes 29.6, and only a red lens or dimmer light meets the evening ceiling; wearing a red lens all day cuts the toy model's morning advance from 17.9 to 5.5 minutes. Chapter 7 reads the Suwen and the Lingshu: at an equinox noon the eye receives about 45725 lux, 758 times an evening room. Chapter 8 gives a light-meter routine and a self-test plan: seeing a 10-minute effect needs about 36 nights per arm. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷研究夜间的光怎样经由眼睛到达生物钟,以及透明或有色的「防蓝光」镜片对此能改变什么。本书把流行的说法写成可以运行的算术,再看它们在什么条件下成立。书中的光谱、镜片、人与日程均为示意模型;全部计算由书中程序生成;本书不是医疗建议,不诊断也不治疗任何疾病,夜间驾驶时绝不可佩戴有色镜片。 第一章建立明视觉曲线与 CIE S 026 黑视素曲线:视蛋白峰值在 480 nm,但以角膜为基准的作用光谱峰值在 489 nm;在 440 nm 附近,黑视素权重是明视觉权重的 16.32 倍。第二章为手机与灯具建模:日间模式手机在眼前 40 lux 时给出 37.3 lux 的黑视素等效日光照度,夜间模式 21.5;420 nm 以下的紫光只占信号的 0.000 至 0.002。第三章把剂量换算成褪黑素抑制:以 90 分钟半数剂量 208 lux 为锚,两小时手机使典型的人被抑制 0.088,敏感五倍的人 0.708,以对照试验为锚约合入睡多花 1.0 或 8.0 分钟。第四章检验镜片:透明防紫光镜片比普通玻璃多挡的只有 0.0000 至 0.0003,琥珀色镜片防护 0.587,红色 0.954。 第五章描出防护与显色的前沿:满足显色保真指数 70 的最佳长通镜片只防护 0.364,阻挡 440 至 514 nm 的陷波镜片以 70.3 达到 0.633。第六章排出一天:在 150 lux 暖光下琥珀色镜片仍放过 29.6,只有红色镜片或调暗灯光能达到傍晚上限;整天戴红色镜片使玩具模型的早晨前移从 17.9 分钟降到 5.5 分钟。第七章读《素问·生气通天论》与《灵枢·大惑论》:春秋分正午眼前约 45725 lux,是傍晚房间的 758 倍。第八章给出测光步骤与自我检验方案:要看到 10 分钟的效果,每组约需 36 晚。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Two Ways of Seeing Light — Chapter 1 builds the photopic curve and the CIE S 026 melanopic curve: the opsin peaks at 480 nm but the action spectrum referred to the cornea peaks at 489 nm, and near 440 nm the melanopic weight is 16.32 times the photopic. (https://5dlibrary.com/answers/en/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c01/) 第1章 看光的两种方式 — 第一章建立明视觉曲线与 CIE S 026 黑视素曲线:视蛋白峰值在 480 nm,但以角膜为基准的作用光谱峰值在 489 nm;在 440 nm 附近,黑视素权重是明视觉权重的 16.32 倍。 (https://5dlibrary.com/answers/zh/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c01/) - Ch.2 Screens, Lamps and the 440 to 490 nm Band — Chapter 2 models phones and lamps: a day-mode phone at 40 lux gives 37.3 lux melanopic EDI, night mode 21.5, and violet below 420 nm carries 0.000 to 0.002 of the signal. (https://5dlibrary.com/answers/en/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c02/) 第2章 屏幕、灯具与 440 至 490 nm 波段 — 第二章为手机与灯具建模:日间模式手机在眼前 40 lux 时给出 37.3 lux 的黑视素等效日光照度,夜间模式 21.5;420 nm 以下的紫光只占信号的 0.000 至 0.002。 (https://5dlibrary.com/answers/zh/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c02/) - Ch.3 From Melanopic Lux to Melatonin and Minutes — Chapter 3 turns dose into melatonin suppression: anchored on a 90-minute half-maximal dose of 208 lux, two hours of a phone suppresses a typical person by 0.088 and one five times more sensitive by 0.708, about 1.0 or 8.0 extra minutes of sleep latency when anchored on a controlled trial. (https://5dlibrary.com/answers/en/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c03/) 第3章 从黑视素照度到褪黑素与分钟数 — 第三章把剂量换算成褪黑素抑制:以 90 分钟半数剂量 208 lux 为锚,两小时手机使典型的人被抑制 0.088,敏感五倍的人 0.708,以对照试验为锚约合入睡多花 1.0 或 8.0 分钟。 (https://5dlibrary.com/answers/zh/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c03/) - Ch.4 Testing the Clear "Blue-Blocker" — Chapter 4 tests lenses: a clear violet-cut lens removes 0.0000 to 0.0003 more than plain glass, an amber lens protects 0.587 and a red one 0.954. (https://5dlibrary.com/answers/en/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c04/) 第4章 检验透明的「防蓝光」镜片 — 第四章检验镜片:透明防紫光镜片比普通玻璃多挡的只有 0.0000 至 0.0003,琥珀色镜片防护 0.587,红色 0.954。 (https://5dlibrary.com/answers/zh/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c04/) - Ch.5 Protection Against Colour — Chapter 5 traces the front between protection and colour: the best long-pass lens with a fidelity index of 70 protects 0.364, while a notch lens blocking 440 to 514 nm reaches 0.633 at 70.3. (https://5dlibrary.com/answers/en/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c05/) 第5章 防护与颜色 — 第五章描出防护与显色的前沿:满足显色保真指数 70 的最佳长通镜片只防护 0.364,阻挡 440 至 514 nm 的陷波镜片以 70.3 达到 0.633。 (https://5dlibrary.com/answers/zh/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c05/) - Ch.6 Lenses On, Lenses Off — Chapter 6 lays out a day: in 150 lux of warm light an amber lens still passes 29.6, and only a red lens or dimmer light meets the evening ceiling; wearing a red lens all day cuts the toy model's morning advance from 17.9 to 5.5 minutes. (https://5dlibrary.com/answers/en/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c06/) 第6章 戴上镜片,摘下镜片 — 第六章排出一天:在 150 lux 暖光下琥珀色镜片仍放过 29.6,只有红色镜片或调暗灯光能达到傍晚上限;整天戴红色镜片使玩具模型的早晨前移从 17.9 分钟降到 5.5 分钟。 (https://5dlibrary.com/answers/zh/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c06/) - Ch.7 The Huangdi Neijing and the Melanopic Standard — Chapter 7 reads the Suwen and the Lingshu: at an equinox noon the eye receives about 45725 lux, 758 times an evening room. (https://5dlibrary.com/answers/en/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c07/) 第7章 《黄帝内经》与黑视素标准 — 第七章读《素问·生气通天论》与《灵枢·大惑论》:春秋分正午眼前约 45725 lux,是傍晚房间的 758 倍。 (https://5dlibrary.com/answers/zh/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c07/) - Ch.8 A Light Meter, a Lens and a Notebook — Chapter 8 gives a light-meter routine and a self-test plan: seeing a 10-minute effect needs about 36 nights per arm. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c08/) 第8章 一台照度计、一片镜片与一本笔记 — 第八章给出测光步骤与自我检验方案:要看到 10 分钟的效果,每组约需 36 晚。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/melanopic-spectral-integration-blue-light-melatonin-suppression-lens-transmittance/c08/) ## 5DL-B-105 · Stratum Corneum Fickian Diffusion, Follicular Ion Shunts and Transdermal Patch Kinetics 中文书名:角质层菲克扩散、毛囊离子旁路与透皮贴片动力学 DOI: pending Cite: Fan, Y. (2026). Stratum Corneum Fickian Diffusion, Follicular Ion Shunts and Transdermal Patch Kinetics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its molecules, doses, rooms, patches and people are illustrative models; all calculations are generated by its programs; the book is not medical advice and is no guide to dosing any medicine or supplement. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This volume studies what can cross human skin, how fast and by which route: magnesium brines rubbed onto the skin, fragrance sprays breathed in at night, and patches that release a substance slowly. It writes popular claims as runnable arithmetic and asks when they hold. Its molecules, doses, rooms, patches and people are illustrative models; all calculations are generated by its programs; the book is not medical advice and is no guide to dosing any medicine or supplement. Chapter 1 quantifies the stratum corneum with the Potts-Guy equation and Fick's law: linalool permeates at 29340.668 millionths of a centimetre per hour, a magnesium ion, extrapolated on assumed values, at only 0.449 to 11.813; the lag is 625.0 minutes through the lipids and 3.9 down an aqueous follicle. Chapter 2 bounds magnesium through the follicles: 500 cm2 of magnesium oil for 20 minutes passes at most 3.950 mg, 0.0108 of a day's need; a whole-body Epsom bath at most 3.206 mg, a glass of mineral water 12.5. Chapter 3 shows that warmth multiplies the lipid route by at most 11.2, and that a trial without a plain-oil control credits the rub's 0.40 standard deviations to magnesium; seeing a magnesium effect of 0.1 needs 1570 people per arm. Chapter 4 follows a pillow spray: the breathing zone reaches 2178.3 micrograms per cubic metre, but plasma at most 0.0287 micromolar, 348 times below an optimistic GABA-A threshold; the route is smell. Chapter 5 compares patches: a Higuchi matrix releases 5.10 times as much in the first hour as in the seventh, and a membrane four times as resistant as skin cuts the between-person coefficient of variation from 0.534 to 0.104. Chapter 6 shows that a flat night needs a short half-life: at 0.5 hours 0.836 of the window is near plateau, at 4.0 hours none; after removal at 06:00, 0.533 can remain at 08:00. Chapter 7 reads the Suwen's "rub it, bathe it" and Wu Shiji's Liyue pianwen: tripling blood flow multiplies magnesium transport by 1.001; a nicotine patch needs 6 cm2, a day of magnesium through the lipids 326545 cm2. Chapter 8 teaches label reading and test design: ten pumps apply 145.2 mg, of which at most 0.0272 crosses. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷研究什么能穿过人的皮肤、多快、经由哪条路:涂在皮肤上的镁卤水、夜里吸入的香氛喷雾,以及缓慢释放物质的贴片。本书把流行的说法写成可以运行的算术,再看它们在什么条件下成立。书中的分子、剂量、房间、贴片与人均为示意模型;全部计算由书中程序生成;本书不是医疗建议,不是任何药物或补充剂的用量指南。 第一章用波茨-盖伊方程与菲克定律量化角质层:芳樟醇的渗透系数为每小时 29340.668 个百万分之一厘米,镁离子按假定值外推只有 0.449 至 11.813;经脂质的滞后为 625.0 分钟,经充水毛囊为 3.9 分钟。第二章为经毛囊的镁设定上界:500 cm2 镁油涂 20 分钟至多通过 3.950 mg,是每日需要量的 0.0108;全身爱普生盐浴至多 3.206 mg,一杯矿泉水含 12.5。第三章说明温热至多使脂质通路乘以 11.2,而没有普通油对照的试验会把揉搓带来的 0.40 个标准差记在镁的名下,看到 0.1 的镁效应每组需要 1570 人。第四章跟随枕头喷雾:呼吸区达到每立方米 2178.3 微克,血浆上界只有 0.0287 微摩尔,比乐观的 GABA-A 阈值低 348 倍,作用经由嗅觉。 第五章比较贴片:Higuchi 基质第一小时的释放是第七小时的 5.10 倍,阻力为皮肤四倍的控释膜把人与人之间的变异系数从 0.534 降到 0.104。第六章说明平稳的一夜需要短半衰期:半衰期 0.5 小时时窗口中 0.836 的时间接近平台,4.0 小时时一刻也没有;06:00 揭下后 08:00 仍可留有 0.533。第七章读《素问·至真要大论》「摩之浴之」与吴师机《理瀹骈文》:血流增至三倍只使镁的转运乘以 1.001;尼古丁贴片需 6 cm2,经脂质供给一天的镁需 326545 cm2。第八章教人读标签与设计检验:十泵施加 145.2 mg,至多 0.0272 穿过。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Brick Wall in Numbers — Chapter 1 quantifies the stratum corneum with the Potts-Guy equation and Fick's law: linalool permeates at 29340.668 millionths of a centimetre per hour, a magnesium ion, extrapolated on assumed values, at only 0.449 to 11.813; the lag is 625.0 minutes through the lipids and 3.9 down an aqueous follicle. (https://5dlibrary.com/answers/en/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c01/) 第1章 砖墙的数字 — 第一章用波茨-盖伊方程与菲克定律量化角质层:芳樟醇的渗透系数为每小时 29340.668 个百万分之一厘米,镁离子按假定值外推只有 0.449 至 11.813;经脂质的滞后为 625.0 分钟,经充水毛囊为 3.9 分钟。 (https://5dlibrary.com/answers/zh/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c01/) - Ch.2 Magnesium Through the Follicles — Chapter 2 bounds magnesium through the follicles: 500 cm2 of magnesium oil for 20 minutes passes at most 3.950 mg, 0.0108 of a day's need; a whole-body Epsom bath at most 3.206 mg, a glass of mineral water 12.5. (https://5dlibrary.com/answers/en/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c02/) 第2章 经由毛囊的镁 — 第二章为经毛囊的镁设定上界:500 cm2 镁油涂 20 分钟至多通过 3.950 mg,是每日需要量的 0.0108;全身爱普生盐浴至多 3.206 mg,一杯矿泉水含 12.5。 (https://5dlibrary.com/answers/zh/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c02/) - Ch.3 Warmth, Rubbing and the C-Tactile Confound — Chapter 3 shows that warmth multiplies the lipid route by at most 11.2, and that a trial without a plain-oil control credits the rub's 0.40 standard deviations to magnesium; seeing a magnesium effect of 0.1 needs 1570 people per arm. (https://5dlibrary.com/answers/en/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c03/) 第3章 温热、揉搓与 C 触觉纤维的混淆 — 第三章说明温热至多使脂质通路乘以 11.2,而没有普通油对照的试验会把揉搓带来的 0.40 个标准差记在镁的名下,看到 0.1 的镁效应每组需要 1570 人。 (https://5dlibrary.com/answers/zh/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c03/) - Ch.4 A Pillow Spray from Pillow to Nose — Chapter 4 follows a pillow spray: the breathing zone reaches 2178.3 micrograms per cubic metre, but plasma at most 0.0287 micromolar, 348 times below an optimistic GABA-A threshold; the route is smell. (https://5dlibrary.com/answers/en/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c04/) 第4章 枕头喷雾:从枕头到鼻子 — 第四章跟随枕头喷雾:呼吸区达到每立方米 2178.3 微克,血浆上界只有 0.0287 微摩尔,比乐观的 GABA-A 阈值低 348 倍,作用经由嗅觉。 (https://5dlibrary.com/answers/zh/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c04/) - Ch.5 Matrix Against Membrane — Chapter 5 compares patches: a Higuchi matrix releases 5.10 times as much in the first hour as in the seventh, and a membrane four times as resistant as skin cuts the between-person coefficient of variation from 0.534 to 0.104. (https://5dlibrary.com/answers/en/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c05/) 第5章 基质与控释膜 — 第五章比较贴片:Higuchi 基质第一小时的释放是第七小时的 5.10 倍,阻力为皮肤四倍的控释膜把人与人之间的变异系数从 0.534 降到 0.104。 (https://5dlibrary.com/answers/zh/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c05/) - Ch.6 One Night in the Blood — Chapter 6 shows that a flat night needs a short half-life: at 0.5 hours 0.836 of the window is near plateau, at 4.0 hours none; after removal at 06:00, 0.533 can remain at 08:00. (https://5dlibrary.com/answers/en/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c06/) 第6章 血液中的一夜 — 第六章说明平稳的一夜需要短半衰期:半衰期 0.5 小时时窗口中 0.836 的时间接近平台,4.0 小时时一刻也没有;06:00 揭下后 08:00 仍可留有 0.533。 (https://5dlibrary.com/answers/zh/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c06/) - Ch.7 "Rub It, Bathe It", Wu Shiji and Fick — Chapter 7 reads the Suwen's "rub it, bathe it" and Wu Shiji's Liyue pianwen: tripling blood flow multiplies magnesium transport by 1.001; a nicotine patch needs 6 cm2, a day of magnesium through the lipids 326545 cm2. (https://5dlibrary.com/answers/en/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c07/) 第7章 「摩之浴之」、吴师机与菲克 — 第七章读《素问·至真要大论》「摩之浴之」与吴师机《理瀹骈文》:血流增至三倍只使镁的转运乘以 1.001;尼古丁贴片需 6 cm2,经脂质供给一天的镁需 326545 cm2。 (https://5dlibrary.com/answers/zh/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c07/) - Ch.8 Reading a Label, Testing a Claim — Chapter 8 teaches label reading and test design: ten pumps apply 145.2 mg, of which at most 0.0272 crosses. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c08/) 第8章 读标签,验说法 — 第八章教人读标签与设计检验:十泵施加 145.2 mg,至多 0.0272 穿过。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/stratum-corneum-fickian-diffusion-follicular-ion-shunts-transdermal-patch-kinetics/c08/) ## 5DL-B-106 · Logarithmic-Spiral Planar Inductors, Golden-Angle Phyllotaxis Packing and Human RC Grounding Circuits 中文书名:对数螺旋平面电感、黄金角叶序堆积与人体接地电路 DOI: pending Cite: Fan, Y. (2026). Logarithmic-Spiral Planar Inductors, Golden-Angle Phyllotaxis Packing and Human RC Grounding Circuits (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its coils, rooms, bodies and faults are illustrative models; all calculations are generated by its programs; the book is not medical or electrical-safety advice and not an installation guide. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This volume applies classical electromagnetism and elementary number theory to three objects sold for wellbeing: flat spiral copper coils, golden-angle patterns and grounding sheets. It writes popular claims as runnable arithmetic and asks when they hold. Its coils, rooms, bodies and faults are illustrative models; all calculations are generated by its programs; the book is not medical or electrical-safety advice and not an installation guide. Chapter 1 computes the inductance of flat coils with two published formulas and checks them against an exact sum of Maxwell's ring inductances: they agree within a factor of 1.010 to 1.026, and 24 turns give 21.44 microhenries. Chapter 2 compares spirals: between radii of 5 and 50 mm a golden spiral makes only 1.20 turns and 0.11 microhenries, an Archimedean spiral 22.50 turns and 21.08; as an antenna at 1 MHz the golden spiral radiates 0.0012 parts per million of its input. Chapter 3 shows the inverse-cube fall of the field, from 715.0 microtesla at the centre to 0.0123 at one metre for 1 A, and that a passive coil with open ends carries no current and makes no field. Chapter 4 builds Vogel's sunflower: the golden angle gives a nearest-neighbour coefficient of variation of only 0.015, 137.6 degrees 0.130, and the golden angle ranks 1 of 302, for the reason given by Hurwitz's theorem. Chapter 5 treats the body as a capacitive divider: on 230 V mains it carries 1.52 to 3.77 V; grounding through 100 kilohms with moist skin removes 0.993, through 1 megohm with dry skin only 0.873. Chapter 6 shows the resistor is a safety device: in a wiring fault a wet sleeper with no resistor can carry 230.000 mA, 100 kilohms limit it to 2.072 mA and 1 megohm to 0.227 mA; in normal use only 0.217 microamperes flow. Chapter 7 reads Wang Chong's Lunheng on amber and Shen Kuo's Mengxi bitan on the needle that points slightly east: amber lifts damp chaff across 4.10 mm, and with copper's skin depth of 9.23 mm at 50 Hz a thin sheet passes 0.995 of the magnetic field. Chapter 8 shows that a multimeter reading of 1.570 V is only part of the true 3.40 V, and that a fair sleep trial needs 175 people per arm. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷把经典电磁学与初等数论用于三种以养生为名出售的东西:平面螺旋铜线圈、黄金角图案与接地床单。本书把流行的说法写成可以运行的算术,再看它们在什么条件下成立。书中的线圈、房间、身体与故障均为示意模型;全部计算由书中程序生成;本书不是医疗或电气安全建议,也不是安装指南。 第一章用两个已发表的公式计算平面线圈的电感,并与麦克斯韦圆环电感的精确求和比较:两者相差在 1.010 至 1.026 倍之内,24 匝为 21.44 微亨。第二章比较螺旋:在半径 5 至 50 mm 之间,黄金螺旋只绕 1.20 圈、电感 0.11 微亨,阿基米德螺旋 22.50 圈、21.08 微亨;作为天线,1 MHz 时黄金螺旋只辐射输入功率的百万分之 0.0012。第三章说明场按立方反比下降:通 1 A 时中心 715.0 微特斯拉,一米处 0.0123;两端开路的无源线圈不带电流、不产生场。第四章建立沃格尔的向日葵:黄金角使最近邻距离的变异系数只有 0.015,137.6 度为 0.130;在 302 个角度中黄金角排第 1,原因是胡尔维茨定理。 第五章把人体看作电容分压器:230 V 市电下人体带 1.52 至 3.77 V;100 千欧接地加湿润皮肤去掉 0.993,1 兆欧加干燥皮肤只去掉 0.873。第六章说明电阻是安全装置:接线故障时无电阻、皮肤潮湿可流过 230.000 mA,100 千欧限制在 2.072 mA,1 兆欧限制在 0.227 mA;正常使用时只有 0.217 微安。第七章读王充《论衡》「顿牟掇芥」与沈括《梦溪笔谈》「常微偏东」:琥珀能隔 4.10 mm 吸起潮湿的草屑;铜在 50 Hz 的趋肤深度为 9.23 mm,薄床单让磁场的 0.995 通过。第八章说明万用表读数 1.570 V 只是真实 3.40 V 的一部分,公平的睡眠试验每组需要 175 人。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Inductance of a Flat Coil — Chapter 1 computes the inductance of flat coils with two published formulas and checks them against an exact sum of Maxwell's ring inductances: they agree within a factor of 1.010 to 1.026, and 24 turns give 21.44 microhenries. (https://5dlibrary.com/answers/en/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c01/) 第1章 平面线圈的电感 — 第一章用两个已发表的公式计算平面线圈的电感,并与麦克斯韦圆环电感的精确求和比较:两者相差在 1.010 至 1.026 倍之内,24 匝为 21.44 微亨。 (https://5dlibrary.com/answers/zh/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c01/) - Ch.2 Archimedean and Golden Spirals — Chapter 2 compares spirals: between radii of 5 and 50 mm a golden spiral makes only 1.20 turns and 0.11 microhenries, an Archimedean spiral 22.50 turns and 21.08; as an antenna at 1 MHz the golden spiral radiates 0.0012 parts per million of its input. (https://5dlibrary.com/answers/en/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c02/) 第2章 阿基米德螺旋与黄金螺旋 — 第二章比较螺旋:在半径 5 至 50 mm 之间,黄金螺旋只绕 1.20 圈、电感 0.11 微亨,阿基米德螺旋 22.50 圈、21.08 微亨;作为天线,1 MHz 时黄金螺旋只辐射输入功率的百万分之 0.0012。 (https://5dlibrary.com/answers/zh/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c02/) - Ch.3 The Field of a Coil, and of No Current — Chapter 3 shows the inverse-cube fall of the field, from 715.0 microtesla at the centre to 0.0123 at one metre for 1 A, and that a passive coil with open ends carries no current and makes no field. (https://5dlibrary.com/answers/en/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c03/) 第3章 线圈的场,以及没有电流时 — 第三章说明场按立方反比下降:通 1 A 时中心 715.0 微特斯拉,一米处 0.0123;两端开路的无源线圈不带电流、不产生场。 (https://5dlibrary.com/answers/zh/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c03/) - Ch.4 Vogel's Sunflower and the Golden Angle — Chapter 4 builds Vogel's sunflower: the golden angle gives a nearest-neighbour coefficient of variation of only 0.015, 137.6 degrees 0.130, and the golden angle ranks 1 of 302, for the reason given by Hurwitz's theorem. (https://5dlibrary.com/answers/en/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c04/) 第4章 沃格尔的向日葵与黄金角 — 第四章建立沃格尔的向日葵:黄金角使最近邻距离的变异系数只有 0.015,137.6 度为 0.130;在 302 个角度中黄金角排第 1,原因是胡尔维茨定理。 (https://5dlibrary.com/answers/zh/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c04/) - Ch.5 The Body as a Node in a Divider — Chapter 5 treats the body as a capacitive divider: on 230 V mains it carries 1.52 to 3.77 V; grounding through 100 kilohms with moist skin removes 0.993, through 1 megohm with dry skin only 0.873. (https://5dlibrary.com/answers/en/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c05/) 第5章 作为分压器节点的人体 — 第五章把人体看作电容分压器:230 V 市电下人体带 1.52 至 3.77 V;100 千欧接地加湿润皮肤去掉 0.993,1 兆欧加干燥皮肤只去掉 0.873。 (https://5dlibrary.com/answers/zh/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c05/) - Ch.6 Why the Resistor Is There — Chapter 6 shows the resistor is a safety device: in a wiring fault a wet sleeper with no resistor can carry 230.000 mA, 100 kilohms limit it to 2.072 mA and 1 megohm to 0.227 mA; in normal use only 0.217 microamperes flow. (https://5dlibrary.com/answers/en/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c06/) 第6章 电阻为什么在那里 — 第六章说明电阻是安全装置:接线故障时无电阻、皮肤潮湿可流过 230.000 mA,100 千欧限制在 2.072 mA,1 兆欧限制在 0.227 mA;正常使用时只有 0.217 微安。 (https://5dlibrary.com/answers/zh/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c06/) - Ch.7 Wang Chong's Amber, Shen Kuo's Needle and Faraday's Cage — Chapter 7 reads Wang Chong's Lunheng on amber and Shen Kuo's Mengxi bitan on the needle that points slightly east: amber lifts damp chaff across 4.10 mm, and with copper's skin depth of 9.23 mm at 50 Hz a thin sheet passes 0.995 of the magnetic field. (https://5dlibrary.com/answers/en/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c07/) 第7章 王充的琥珀、沈括的磁针与法拉第的笼子 — 第七章读王充《论衡》「顿牟掇芥」与沈括《梦溪笔谈》「常微偏东」:琥珀能隔 4.10 mm 吸起潮湿的草屑;铜在 50 Hz 的趋肤深度为 9.23 mm,薄床单让磁场的 0.995 通过。 (https://5dlibrary.com/answers/zh/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c07/) - Ch.8 Measuring and Testing — Chapter 8 shows that a multimeter reading of 1.570 V is only part of the true 3.40 V, and that a fair sleep trial needs 175 people per arm. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c08/) 第8章 测量与检验 — 第八章说明万用表读数 1.570 V 只是真实 3.40 V 的一部分,公平的睡眠试验每组需要 175 人。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/logarithmic-spiral-planar-inductors-golden-angle-phyllotaxis-rc-grounding-circuits/c08/) ## 5DL-B-107 · Appointment Studio Overbooking Queues, Prepaid Float Dynamics and High-Ticket Ascension Ladders 中文书名:预约制工作室超售排队、储值沉淀资金池与高客单升单阶梯 DOI: pending Cite: Fan, Y. (2026). Appointment Studio Overbooking Queues, Prepaid Float Dynamics and High-Ticket Ascension Ladders (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its clients, prices, rates and costs are illustrative models; all calculations are generated by its programs; the book is not financial, accounting, legal or medical advice and makes no claim about the effect of any treatment or service. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This volume, written for owners of physical studios in aesthetics and wellness in Taiwan and Southeast Asia, studies three things: no-shows and overbooking, the float and breakage of prepaid cards, and the ladder from an introductory session to an annual high-ticket plan. It writes popular claims as runnable arithmetic and asks when they hold. Its clients, prices, rates and costs are illustrative models; all calculations are generated by its programs; the book is not financial, accounting, legal or medical advice and makes no claim about the effect of any treatment or service. Chapter 1 computes the elasticity of no-shows with respect to a deposit: a deposit of 20 on a slot priced at 100 cuts the no-show rate from 0.20 to 0.062, but if a deposit buys away the duty to come, a deposit of 5 raises it to 0.259. Chapter 2 writes overbooking as a newsvendor problem: with a bump cost of 3 prices the best is 1 extra booking, adding 0.033 of a full day, and overbooking only on days with two or more risky clients lifts daily net takings from 6.717 to 6.801. Chapter 3 computes the float with Little's law and Buffett's definition: it is only 2.53 to 5.42 months of card sales, and when the chair would be full a card at 20% off is a loan at 118.5% a year; following IFRS 15, breakage may be recognised only in the pattern in which clients use their cards. Chapter 4 shows the "fourth-redemption cliff" to be a parameter: a first gap of 30 days rather than 25 moves it from session 5 to session 4, and a studio with only 30 cards has a 0.639 chance of seeing a cliff that is not there. Chapter 5 finds where net present values reverse: a card without reminders is worth 663.5, an activated card without plan sales 468.6, and the order reverses once plan uptake reaches 0.0632; with reminders the best moment for the offer is session 8, without them session 3. Chapter 6 shows that when price screens clients the best intro price is 12.5, and a free intro is worth only 72.4. Chapter 7 reads the Guanzi on labour that tires and desires without end, the Song jiaozi's reserve ratio of 0.2865 and its chengti, and Littlewood's seat-protection rule. Chapter 8 combines these in a week of trading and one test protocol, with a checklist. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷写给台湾与东南亚的线下工作室、医美与身心灵主理人,研究三件事:爽约与超售,储值卡的浮存金与沉睡残值,以及从体验课到年度高客单方案的阶梯。本书把流行的说法写成可以运行的算术,再看它们在什么条件下成立。书中的客户、价格、比率与成本均为示意模型;全部计算由书中程序生成;本书不是财务、会计、法律或医疗建议,不对任何疗程或服务的效果作出主张。 第一章计算定金对爽约率的弹性:价格 100 的档期收定金 20,爽约率从 0.20 降到 0.062;但定金若买走了赴约的责任感,定金 5 反而使爽约率升到 0.259。第二章把超售写成报童问题:让到场客户扑空的代价为价格 3 倍时最优只多排 1 位,每天多 0.033 个满日收入;只在两位以上高风险客户的日子超售,净收入从 6.717 升到 6.801。第三章用利特尔定律与巴菲特的定义计算浮存金:它只有 2.53 至 5.42 个月的售卡额,椅子坐满时八折卡是年利率 118.5% 的借款;并按 IFRS 15 说明沉睡残值只能按客户用卡的模式确认。第四章说明「第四次核销断崖」是参数:第一次间隔从 25 天变为 30 天,断崖就从第 5 次移到第 4 次;只有 30 张卡的工作室有 0.639 的机会看见不存在的断崖。 第五章找出净现值的反转点:不提醒的卡值 663.5,催用而没有升单只值 468.6,年度方案成交率达到 0.0632 就反转;有提醒时最好在第 8 次提出方案,没有提醒时在第 3 次。第六章说明体验课能筛选客户时最好的价格是 12.5,免费只值 72.4。第七章读《管子·权修》「民之用力有倦,而人君之欲无穷」、宋代交子 0.2865 的准备金比率与称提,以及利特尔伍德的舱位规则。第八章把这些合成一周的经营与一份检验方案,并给出一份清单。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 No-Shows, Deposits and the Price of an Empty Chair — Chapter 1 computes the elasticity of no-shows with respect to a deposit: a deposit of 20 on a slot priced at 100 cuts the no-show rate from 0.20 to 0.062, but if a deposit buys away the duty to come, a deposit of 5 raises it to 0.259. (https://5dlibrary.com/answers/en/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c01/) 第1章 爽约、定金与空椅子的价格 — 第一章计算定金对爽约率的弹性:价格 100 的档期收定金 20,爽约率从 0.20 降到 0.062;但定金若买走了赴约的责任感,定金 5 反而使爽约率升到 0.259。 (https://5dlibrary.com/answers/zh/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c01/) - Ch.2 Overbooking: Airline Arithmetic in the Treatment Room — Chapter 2 writes overbooking as a newsvendor problem: with a bump cost of 3 prices the best is 1 extra booking, adding 0.033 of a full day, and overbooking only on days with two or more risky clients lifts daily net takings from 6.717 to 6.801. (https://5dlibrary.com/answers/en/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c02/) 第2章 超售:航空公司的算术搬进美容室 — 第二章把超售写成报童问题:让到场客户扑空的代价为价格 3 倍时最优只多排 1 位,每天多 0.033 个满日收入;只在两位以上高风险客户的日子超售,净收入从 6.717 升到 6.801。 (https://5dlibrary.com/answers/zh/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c02/) - Ch.3 The Float and the Breakage of Prepaid Cards — Chapter 3 computes the float with Little's law and Buffett's definition: it is only 2.53 to 5.42 months of card sales, and when the chair would be full a card at 20% off is a loan at 118.5% a year; following IFRS 15, breakage may be recognised only in the pattern in which clients use their cards. (https://5dlibrary.com/answers/en/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c03/) 第3章 储值卡的浮存金与沉睡残值 — 第三章用利特尔定律与巴菲特的定义计算浮存金:它只有 2.53 至 5.42 个月的售卡额,椅子坐满时八折卡是年利率 118.5% 的借款;并按 IFRS 15 说明沉睡残值只能按客户用卡的模式确认。 (https://5dlibrary.com/answers/zh/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c03/) - Ch.4 Is the "Fourth-Session Cliff" a Law or a Parameter? — Chapter 4 shows the "fourth-redemption cliff" to be a parameter: a first gap of 30 days rather than 25 moves it from session 5 to session 4, and a studio with only 30 cards has a 0.639 chance of seeing a cliff that is not there. (https://5dlibrary.com/answers/en/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c04/) 第4章 「第四次核销断崖」是定律还是参数 — 第四章说明「第四次核销断崖」是参数:第一次间隔从 25 天变为 30 天,断崖就从第 5 次移到第 4 次;只有 30 张卡的工作室有 0.639 的机会看见不存在的断崖。 (https://5dlibrary.com/answers/zh/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c04/) - Ch.5 Harvest the Breakage or Get the Sessions Used: Where NPV Reverses — Chapter 5 finds where net present values reverse: a card without reminders is worth 663.5, an activated card without plan sales 468.6, and the order reverses once plan uptake reaches 0.0632; with reminders the best moment for the offer is session 8, without them session 3. (https://5dlibrary.com/answers/en/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c05/) 第5章 竭泽而渔还是催人用完:净现值的反转点 — 第五章找出净现值的反转点:不提醒的卡值 663.5,催用而没有升单只值 468.6,年度方案成交率达到 0.0632 就反转;有提醒时最好在第 8 次提出方案,没有提醒时在第 3 次。 (https://5dlibrary.com/answers/zh/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c05/) - Ch.6 Intro Sessions and Annual Plans: The Arithmetic of a Two-Stage Funnel — Chapter 6 shows that when price screens clients the best intro price is 12.5, and a free intro is worth only 72.4. (https://5dlibrary.com/answers/en/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c06/) 第6章 体验课与年度方案:两段漏斗的算术 — 第六章说明体验课能筛选客户时最好的价格是 12.5,免费只值 72.4。 (https://5dlibrary.com/answers/zh/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c06/) - Ch.7 The Guanzi's Tiring Labour, the Reserve of the Jiaozi and Airline Revenue Management — Chapter 7 reads the Guanzi on labour that tires and desires without end, the Song jiaozi's reserve ratio of 0.2865 and its chengti, and Littlewood's seat-protection rule. (https://5dlibrary.com/answers/en/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c07/) 第7章 《管子》的有倦之力、交子的称提与航空收益管理 — 第七章读《管子·权修》「民之用力有倦,而人君之欲无穷」、宋代交子 0.2865 的准备金比率与称提,以及利特尔伍德的舱位规则。 (https://5dlibrary.com/answers/zh/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c07/) - Ch.8 Practice: One Studio's Week and One Test Protocol — Chapter 8 combines these in a week of trading and one test protocol, with a checklist. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c08/) 第8章 实践:一家工作室的一周与一份检验方案 — 第八章把这些合成一周的经营与一份检验方案,并给出一份清单。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/appointment-studio-overbooking-queues-prepaid-float-dynamics-high-ticket-ascension/c08/) ## 5DL-B-108 · Multicurrency FX Spread Microstructure, Dynamic Currency Conversion Traps and Tax-Residency Day-Counting Thresholds 中文书名:多币种换汇点差微观结构、动态货币转换陷阱与税务居留天数规划 DOI: pending Cite: Fan, Y. (2026). Multicurrency FX Spread Microstructure, Dynamic Currency Conversion Traps and Tax-Residency Day-Counting Thresholds (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its exchange rates, fees and itineraries are illustrative models; all calculations are generated by its programs; the book is not tax, immigration, legal or investment advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This volume, written for cross-border founders who shuttle between Malaysia, Taiwan, Vietnam, mainland China and Japan, studies the spreads on foreign banknotes and on the routes for changing money, dynamic currency conversion (DCC) at the till, and counting days across several jurisdictions. It writes popular claims as runnable arithmetic and asks when they hold. Its exchange rates, fees and itineraries are illustrative models; all calculations are generated by its programs; the book is not tax, immigration, legal or investment advice. Chapter 1 treats banknotes as inventory: for a thin currency with a yearly volume of 500,000 the break-even half-spread is 0.1438, at 50,000,000 only 0.0033, and volatility barely matters; taking 30,000 at once from a destination ATM with a multicurrency card costs 0.0063, below the home airport booth's 0.0450. Chapter 2 shows that the weekend markup prices Monday's gap: with a gap of 0.004 the break-even markup is 0.0016. Chapter 3 uses the economic order quantity: with a fee of 25 the best withdrawal is 10954, and when leftovers are changed back at 0.06 the best is 12000, not the maximum. Chapter 4 computes DCC: paying in the local currency costs 0.0020 to 0.0321 extra, DCC 0.0350 to 0.0866, and the DCC markup is at least 648 times the value of the exchange-rate risk it removes. Chapter 5 compares counting rules: under the substantial presence test, someone who visits the US every year has a steady limit of 121 days; the same itinerary counts 120 days by calendar year and 240 in some twelve months. Chapter 6 writes a year as a knapsack: the day limits cost 16.5, a day more in Taipei has a shadow price of 0.508, and four entries a year call for a 5-day buffer. Chapter 7 reads the Rites of Zhou on contracts and the Money Office and the New Book of Tang on flying cash: monthly netting saves 0.455 of spreads, and at an interest gap of 0.015 twelve months' forward discount is 0.01456. Chapter 8 gives a year's friction budget (10239 for the casual habit, 3390 for the fee-free one), a day ledger and a checklist. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷写给穿梭于马来西亚、台湾、越南、中国大陆与日本之间的跨境创业者,研究外币现钞与各种换汇路线的点差、收银台前的动态货币转换(DCC),以及在多个司法管辖区之间数天数。本书把流行的说法写成可以运行的算术,再看它们在什么条件下成立。书中的汇率、费率与行程均为示意模型;全部计算由书中程序生成;本书不是税务、移民、法律或投资建议。 第一章把外币现钞当作库存:年交易量 500,000 的冷门现钞保本半点差是 0.1438,50,000,000 时只有 0.0033,波动率几乎不起作用;在目的地 ATM 用多币种卡一次取 30,000,摩擦率 0.0063,低于本国机场柜台的 0.0450。第二章说明周末加价在为周一的跳空定价:跳空标准差 0.004 时保本加价 0.0016。第三章用经济批量公式回答每次取多少:固定费用 25 时最优每次 10954;剩余现金要以 0.06 换回时,最优是 12000 而不是取满。第四章计算 DCC:用当地货币结算多付 0.0020 至 0.0321,DCC 多付 0.0350 至 0.0866,而 DCC 加价至少是它所消除的汇率风险价值的 648 倍。 第五章比较计数规则:每年都去美国的人,实质居住测试下的稳定上限是 121 天;同一份行程,日历年计 120 天,任何十二个月计 240 天。第六章把一年写成背包问题:天数上限的成本是 16.5,台北多一天的影子价格 0.508,一年进出 4 次需要预留 5 天。第七章读《周礼》的质剂与泉府、《新唐书》的飞钱:按月轧差省下 0.455 的点差;利差 0.015 时十二个月的远期折扣是 0.01456。第八章给出一年的摩擦预算(随手的习惯 10239,免手续费的习惯 3390)、一本天数账与一份清单。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Why the Airport Booth's Spread Is So Wide — Chapter 1 treats banknotes as inventory: for a thin currency with a yearly volume of 500,000 the break-even half-spread is 0.1438, at 50,000,000 only 0.0033, and volatility barely matters; taking 30,000 at once from a destination ATM with a multicurrency card costs 0.0063, below the home airport booth's 0.0450. (https://5dlibrary.com/answers/en/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c01/) 第1章 机场柜台的点差为什么这么宽 — 第一章把外币现钞当作库存:年交易量 500,000 的冷门现钞保本半点差是 0.1438,50,000,000 时只有 0.0033,波动率几乎不起作用;在目的地 ATM 用多币种卡一次取 30,000,摩擦率 0.0063,低于本国机场柜台的 0.0450。 (https://5dlibrary.com/answers/zh/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c01/) - Ch.2 The Weekend Markup on Multicurrency Cards — Chapter 2 shows that the weekend markup prices Monday's gap: with a gap of 0.004 the break-even markup is 0.0016. (https://5dlibrary.com/answers/en/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c02/) 第2章 多币种卡的周末加价 — 第二章说明周末加价在为周一的跳空定价:跳空标准差 0.004 时保本加价 0.0016。 (https://5dlibrary.com/answers/zh/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c02/) - Ch.3 How Much to Take Out of the ATM: The Economic Order Quantity — Chapter 3 uses the economic order quantity: with a fee of 25 the best withdrawal is 10954, and when leftovers are changed back at 0.06 the best is 12000, not the maximum. (https://5dlibrary.com/answers/en/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c03/) 第3章 每次从 ATM 取多少:经济批量公式 — 第三章用经济批量公式回答每次取多少:固定费用 25 时最优每次 10954;剩余现金要以 0.06 换回时,最优是 12000 而不是取满。 (https://5dlibrary.com/answers/zh/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c03/) - Ch.4 Dynamic Currency Conversion at the Till — Chapter 4 computes DCC: paying in the local currency costs 0.0020 to 0.0321 extra, DCC 0.0350 to 0.0866, and the DCC markup is at least 648 times the value of the exchange-rate risk it removes. (https://5dlibrary.com/answers/en/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c04/) 第4章 收银台前的动态货币转换 — 第四章计算 DCC:用当地货币结算多付 0.0020 至 0.0321,DCC 多付 0.0350 至 0.0866,而 DCC 加价至少是它所消除的汇率风险价值的 648 倍。 (https://5dlibrary.com/answers/zh/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c04/) - Ch.5 Counting Days: 183, Rolling Windows and a Weighted Formula — Chapter 5 compares counting rules: under the substantial presence test, someone who visits the US every year has a steady limit of 121 days; the same itinerary counts 120 days by calendar year and 240 in some twelve months. (https://5dlibrary.com/answers/en/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c05/) 第5章 数天数:183 天、滚动窗口与加权公式 — 第五章比较计数规则:每年都去美国的人,实质居住测试下的稳定上限是 121 天;同一份行程,日历年计 120 天,任何十二个月计 240 天。 (https://5dlibrary.com/answers/zh/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c05/) - Ch.6 Planning a Year as a Knapsack — Chapter 6 writes a year as a knapsack: the day limits cost 16.5, a day more in Taipei has a shadow price of 0.508, and four entries a year call for a 5-day buffer. (https://5dlibrary.com/answers/en/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c06/) 第6章 把一年排成背包问题 — 第六章把一年写成背包问题:天数上限的成本是 16.5,台北多一天的影子价格 0.508,一年进出 4 次需要预留 5 天。 (https://5dlibrary.com/answers/zh/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c06/) - Ch.7 The Rites of Zhou on Contracts and the Money Office, Tang Flying Cash and the Forward Discount — Chapter 7 reads the Rites of Zhou on contracts and the Money Office and the New Book of Tang on flying cash: monthly netting saves 0.455 of spreads, and at an interest gap of 0.015 twelve months' forward discount is 0.01456. (https://5dlibrary.com/answers/en/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c07/) 第7章 《周礼》的质剂与泉府、唐代飞钱与远期贴水 — 第七章读《周礼》的质剂与泉府、《新唐书》的飞钱:按月轧差省下 0.455 的点差;利差 0.015 时十二个月的远期折扣是 0.01456。 (https://5dlibrary.com/answers/zh/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c07/) - Ch.8 Practice: A Year's Friction Budget and a Day Ledger — Chapter 8 gives a year's friction budget (10239 for the casual habit, 3390 for the fee-free one), a day ledger and a checklist. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c08/) 第8章 实践:一年的摩擦预算与一本天数账 — 第八章给出一年的摩擦预算(随手的习惯 10239,免手续费的习惯 3390)、一本天数账与一份清单。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/multicurrency-fx-spread-microstructure-dcc-traps-tax-residency-thresholds/c08/) ## 5DL-B-109 · Wuge Name Numerology Congruences, Kangxi Radical Stroke Traps and Zodiac Demographic Shocks 中文书名:五格剖象数理同余、康熙部首笔画陷阱与生肖人口队列冲击 DOI: pending Cite: Fan, Y. (2026). Wuge Name Numerology Congruences, Kangxi Radical Stroke Traps and Zodiac Demographic Shocks (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its good-and-bad table is one widely circulated version, population figures are sourced, and other parameters are illustrative; all calculations are generated by its programs; the book is not naming, marriage, family-planning or investment advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This volume treats the five-grid name numerology most popular in Hong Kong, Taiwan and Southeast Asia as a set of integer rules and computes it exactly with combinatorics; treats the three ways of counting strokes as an encoding problem; and treats zodiac birth booms as a question of population economics. It makes no divinatory claim: it computes the rules themselves and shows how their effects could be tested. Its good-and-bad table is one widely circulated version, population figures are sourced, and other parameters are illustrative; all calculations are generated by its programs; the book is not naming, marriage, family-planning or investment advice. Chapter 1 writes the five grids as five integer maps: every surname of 2 to 25 strokes can be given a two-character name with four good grids, while a one-character name's outer grid is 2 by convention and never good. Chapter 2 shows that the 81-number table is mostly last digits: a two-part last-digit rule agrees with it on a share of 0.778, and none of two thousand random tables shows such regularity. Chapter 3 writes the three talents as arithmetic modulo 5; only 486 of 13824 stroke combinations are perfect, and the most "universal" combination suits only 8 of 24 surnames; the true "never" is the heaven grid, never good for 9 surname stroke counts. Chapter 4 compares simplified, Taiwan standard and restored Kangxi counts: of 26 common surnames, 12 differ by 1 to 5 strokes between the last two, and one stroke more on the surname leaves only 42 of 486 perfect names perfect. Chapter 5 computes zodiac birth swings: Taiwan's Dragon year 2012 had 1.1598 times the mean of its neighbours' births and Japan's fire-horse 1966 only 0.7240; a simple crowding model cuts the Dragon cohort's chance of a place from 0.40 to 0.345, but the parents' belief also brings more investment. Chapter 6 shows that a zodiac year is not a school year: the Dragon excess is split into Taiwanese school cohorts of 1.0974 and 1.0705, and for one family it explains 0.000208 of the variance of wages. Chapter 7 reads Wang Chong's Lunheng, "water overcomes fire: why does the rat not chase the horse?", and his rebuttal of surname tones: a zodiac clash is a 6-year age gap. Chapter 8 gives three audits of one name and a test protocol: detecting an income difference of 0.02 for perfect names takes 208,253 people. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷把港台与东南亚最流行的五格姓名学当作一套整数规则,用组合数学精确计算;把笔画的三种数法当作编码问题;把生肖出生潮当作人口经济学问题。本书不主张任何占验,只计算规则本身,并说明怎样检验其效果。吉凶表采用一种广泛流传的版本,人口数据注明来源,其余参数均为示意;全部计算由书中程序生成;本书不是命名、婚姻、生育或投资建议。 第一章把五格写成五个整数映射:每个 2 至 25 画的姓都配得出四格全吉的双字名,而单名的外格按约定恒为 2,永远不全吉。第二章说明 81 数理表大半是尾数:一条两段的尾数规则与表的一致比例是 0.778,两千张随机表中没有一张有这样的规律。第三章把三才写成模五的算术,13824 种笔画组合中只有 486 种完美,最「万能」的组合也只适合 24 个姓中的 8 个;真正「永远」的是天格,9 种笔画的姓天格永远不吉。第四章比较简体、台湾正体与康熙还原三种数法:26 个常见姓中 12 个在后两者之间相差 1 至 5 画,姓多数 1 画时 486 个完美名字中只有 42 个仍然完美。 第五章计算生肖出生潮:台湾 2012 年龙年出生数是邻年平均的 1.1598 倍,日本 1966 年丙午年只有 0.7240 倍;简单的拥挤模型使龙年同龄人的录取机会从 0.40 降到 0.345,但父母的信念也带来更多投资。第六章说明生肖年不是学年:龙年的超额在台湾的学校里分成 1.0974 与 1.0705 两届,对一个家庭只解释工资差异的 0.000208。第七章读王充《论衡》「水胜火,鼠何不逐马」与姓之五音的驳论:生肖相冲就是相差 6 岁。第八章给出一个名字的三种审计与一份检验方案:看出完美名字带来 0.02 的收入差距需要 208,253 人。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Five Grids Are Five Integer Maps — Chapter 1 writes the five grids as five integer maps: every surname of 2 to 25 strokes can be given a two-character name with four good grids, while a one-character name's outer grid is 2 by convention and never good. (https://5dlibrary.com/answers/en/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c01/) 第1章 五格是五个整数映射 — 第一章把五格写成五个整数映射:每个 2 至 25 画的姓都配得出四格全吉的双字名,而单名的外格按约定恒为 2,永远不全吉。 (https://5dlibrary.com/answers/zh/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c01/) - Ch.2 The 81-Number Table Is Mostly Last Digits — Chapter 2 shows that the 81-number table is mostly last digits: a two-part last-digit rule agrees with it on a share of 0.778, and none of two thousand random tables shows such regularity. (https://5dlibrary.com/answers/en/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c02/) 第2章 八十一数理表大半是尾数 — 第二章说明 81 数理表大半是尾数:一条两段的尾数规则与表的一致比例是 0.778,两千张随机表中没有一张有这样的规律。 (https://5dlibrary.com/answers/zh/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c02/) - Ch.3 The Three Talents and a Sparse Look-Up Table — Chapter 3 writes the three talents as arithmetic modulo 5; only 486 of 13824 stroke combinations are perfect, and the most "universal" combination suits only 8 of 24 surnames; the true "never" is the heaven grid, never good for 9 surname stroke counts. (https://5dlibrary.com/answers/en/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c03/) 第3章 三才配置与一张稀疏的查表 — 第三章把三才写成模五的算术,13824 种笔画组合中只有 486 种完美,最「万能」的组合也只适合 24 个姓中的 8 个;真正「永远」的是天格,9 种笔画的姓天格永远不吉。 (https://5dlibrary.com/answers/zh/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c03/) - Ch.4 Three Ways of Counting Strokes and the Kangxi Radical Conversion — Chapter 4 compares simplified, Taiwan standard and restored Kangxi counts: of 26 common surnames, 12 differ by 1 to 5 strokes between the last two, and one stroke more on the surname leaves only 42 of 486 perfect names perfect. (https://5dlibrary.com/answers/en/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c04/) 第4章 笔画的三种数法与康熙部首的折算 — 第四章比较简体、台湾正体与康熙还原三种数法:26 个常见姓中 12 个在后两者之间相差 1 至 5 画,姓多数 1 画时 486 个完美名字中只有 42 个仍然完美。 (https://5dlibrary.com/answers/zh/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c04/) - Ch.5 Dragon-Year Birth Booms and the Fire-Horse Hole — Chapter 5 computes zodiac birth swings: Taiwan's Dragon year 2012 had 1.1598 times the mean of its neighbours' births and Japan's fire-horse 1966 only 0.7240; a simple crowding model cuts the Dragon cohort's chance of a place from 0.40 to 0.345, but the parents' belief also brings more investment. (https://5dlibrary.com/answers/en/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c05/) 第5章 龙年出生潮与丙午年的空洞 — 第五章计算生肖出生潮:台湾 2012 年龙年出生数是邻年平均的 1.1598 倍,日本 1966 年丙午年只有 0.7240 倍;简单的拥挤模型使龙年同龄人的录取机会从 0.40 降到 0.345,但父母的信念也带来更多投资。 (https://5dlibrary.com/answers/zh/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c05/) - Ch.6 Zodiac Years and School Years Do Not Line Up — Chapter 6 shows that a zodiac year is not a school year: the Dragon excess is split into Taiwanese school cohorts of 1.0974 and 1.0705, and for one family it explains 0.000208 of the variance of wages. (https://5dlibrary.com/answers/en/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c06/) 第6章 生肖年与学年对不齐 — 第六章说明生肖年不是学年:龙年的超额在台湾的学校里分成 1.0974 与 1.0705 两届,对一个家庭只解释工资差异的 0.000208。 (https://5dlibrary.com/answers/zh/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c06/) - Ch.7 Wang Chong against Zodiac Conflicts and the Five Tones of Surnames — Chapter 7 reads Wang Chong's Lunheng, "water overcomes fire: why does the rat not chase the horse?", and his rebuttal of surname tones: a zodiac clash is a 6-year age gap. (https://5dlibrary.com/answers/en/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c07/) 第7章 王充驳生肖相克与姓之五音 — 第七章读王充《论衡》「水胜火,鼠何不逐马」与姓之五音的驳论:生肖相冲就是相差 6 岁。 (https://5dlibrary.com/answers/zh/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c07/) - Ch.8 Practice: Three Audits of One Name, and a Test Protocol — Chapter 8 gives three audits of one name and a test protocol: detecting an income difference of 0.02 for perfect names takes 208,253 people. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c08/) 第8章 实践:给一个名字做三种审计,以及一份检验方案 — 第八章给出一个名字的三种审计与一份检验方案:看出完美名字带来 0.02 的收入差距需要 208,253 人。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/wuge-name-numerology-congruences-kangxi-stroke-traps-zodiac-demographic-shocks/c08/) ## 5DL-B-110 · Cross-Border Buyer Dynamic Pricing, Fee-Threshold Payment Routing and Accessory Attach Economics 中文书名:跨境买手动态定价、阶跃费率支付路由与配件连带加购经济学 DOI: pending Cite: Fan, Y. (2026). Cross-Border Buyer Dynamic Pricing, Fee-Threshold Payment Routing and Accessory Attach Economics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This volume, written for cross-border personal buyers and boutique shops that buy luxury goods abroad for customers in Taiwan and Southeast Asia, studies three things: how to quote when the payment fee changes in steps, how much an accessory in the same box really earns, and how large a buffer the currency and the tax refund call for. It writes popular claims as runnable arithmetic and asks when they hold. Its customers, prices, fees, freight rates, refund rates and volatilities are illustrative models; all calculations are generated by its programs; the book is not financial, tax, customs, legal or investment advice and describes no particular payment service's terms. Chapter 1 proves that when card payments below a threshold pay a fee and payments at the threshold go by bank transfer, pricing back from an equal net margin must invert across a band of costs. At a threshold of 13,000 and a fee of 0.028 a bag costing 9,686 is quoted 12,808, below a bag costing 9,650 (12,953), and earns only 0.1884; the least price that keeps the margin is a monotone envelope that quotes exactly 13,000 across the deadzone, where 0.0279 of bag-days fall. Chapter 2 shows that a transfer discount is worth only 7.4 per order at the threshold and that plain rounding across a step cuts the margin to 0.1742. Chapter 3 computes accessories that ride in the bag's box at zero marginal freight: at an attach rate of 0.45 with two items they are 0.0593 of revenue and 0.2533 of net profit, and mixed bundling adds only 1.028 beyond the attach. Chapter 4 shows the checkout drawer earning 156.2 times the product page at full price, and discounts only lowering the net. Chapter 5 computes the currency risk of a five-day buying window: under fat tails the 99% loss is 0.0253 to 0.0395 of the cost, and at a normal margin a buffer costs more than it saves. Chapter 6 shows a promised refund paying 0.9550 of its face by card and 0.8775 in cash, and a joint buffer of 0.0276 pushing a bag into the deadzone. Chapter 7 reads "mother and child weigh each other" in the Discourses of the States (correcting its attribution to the Guanzi), the Han treatise on the equable transport and levelling office, and the Chicago school on tying: a strap sent alone pays 0.1276 of its value in freight, inside a bag box nothing. Chapter 8 combines these in a year of trading and one test protocol. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷写给在海外采购精品、服务台湾与东南亚客户的跨境代购与买手店,研究三件事:金流抽成按阶梯变化时怎样报价,同箱出货的配件究竟赚多少,汇率与退税需要多大的缓冲。本书把流行的说法写成可以运行的算术,再看它们在什么条件下成立。书中的客户、价格、抽成、运费、退税率与汇率波动率均为示意模型;全部计算由书中程序生成;本书不是财务、税务、海关、法律或投资建议,不描述任何特定金流服务的条款。 第一章证明:低于阈值刷卡抽成、达到阈值改收汇款时,等净利率倒推必然在一段成本区间内倒挂。阈值 13,000、抽成 0.028 时,成本 9,686 的包报 12,808,低于成本 9,650 的包(12,953),利润率只有 0.1884;保住利润率的价格下限是在死区内一律报 13,000 的单调包络线,有 0.0279 的「包·日」落在那里。第二章说明汇款折扣在阈值处每笔只值 7.4,简单进位越过阶梯会把利润率压到 0.1742。第三章计算同箱零边际运费的配件:附着率 0.45、每次两件时,配件占营收 0.0593、占净利 0.2533;混合捆绑在加购之外只补充 1.028 倍。第四章说明结账抽屉原价时是产品页的 156.2 倍,折扣只会降低净利。 第五章计算五天连线窗口的汇率风险:厚尾下 99% 的损失是成本的 0.0253 至 0.0395,正常利润率下缓冲得不偿失。第六章说明承诺的退税刷卡实得 0.9550、现金实得 0.8775,联合缓冲 0.0276 可能把一个包推进死区。第七章读《国语·周语下》「子母相权」(并更正它出自《国语》而非《管子·轻重甲》)、《汉书·食货志》的均输平准与芝加哥学派的搭售理论:背带单独寄送运费是货值的 0.1276,放进包的箱子里是 0。第八章把这些合成一年的经营与一份检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Deadzone: When a Dearer Bag Gets a Cheaper Quote — Chapter 1 proves that when card payments below a threshold pay a fee and payments at the threshold go by bank transfer, pricing back from an equal net margin must invert across a band of costs. At a threshold of 13,000 and a fee of 0.028 a bag costing 9,686 is quoted 12,808, below a bag costing 9,650 (12,953), and earns only 0.1884; the least price that keeps the margin is a monotone envelope that quotes exactly 13,000 across the deadzone, where 0.0279 of bag-days fall. (https://5dlibrary.com/answers/en/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c01/) 第1章 报价死区:进价越高,报价反而越低 — 第一章证明:低于阈值刷卡抽成、达到阈值改收汇款时,等净利率倒推必然在一段成本区间内倒挂。阈值 13,000、抽成 0.028 时,成本 9,686 的包报 12,808,低于成本 9,650 的包(12,953),利润率只有 0.1884;保住利润率的价格下限是在死区内一律报 13,000 的单调包络线,有 0.0279 的「包·日」落在那里。 (https://5dlibrary.com/answers/zh/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c01/) - Ch.2 The Envelope: Menus, Many Steps and Safe Rounding — Chapter 2 shows that a transfer discount is worth only 7.4 per order at the threshold and that plain rounding across a step cuts the margin to 0.1742. (https://5dlibrary.com/answers/en/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c02/) 第2章 包络线:让客户选择、多段阶梯与安全进位 — 第二章说明汇款折扣在阈值处每笔只值 7.4,简单进位越过阶梯会把利润率压到 0.1742。 (https://5dlibrary.com/answers/zh/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c02/) - Ch.3 The Accessory in the Box: Zero Marginal Freight and Mixed Bundling — Chapter 3 computes accessories that ride in the bag's box at zero marginal freight: at an attach rate of 0.45 with two items they are 0.0593 of revenue and 0.2533 of net profit, and mixed bundling adds only 1.028 beyond the attach. (https://5dlibrary.com/answers/en/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c03/) 第3章 同箱里的配件:零边际运费与混合捆绑 — 第三章计算同箱零边际运费的配件:附着率 0.45、每次两件时,配件占营收 0.0593、占净利 0.2533;混合捆绑在加购之外只补充 1.028 倍。 (https://5dlibrary.com/answers/zh/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c03/) - Ch.4 The Checkout Drawer: Exposure, Discounts and Small Tests — Chapter 4 shows the checkout drawer earning 156.2 times the product page at full price, and discounts only lowering the net. (https://5dlibrary.com/answers/en/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c04/) 第4章 结账抽屉:曝光、折扣与小样本测试 — 第四章说明结账抽屉原价时是产品页的 156.2 倍,折扣只会降低净利。 (https://5dlibrary.com/answers/zh/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c04/) - Ch.5 Five Days Open: Currency Risk in a Live Buying Window — Chapter 5 computes the currency risk of a five-day buying window: under fat tails the 99% loss is 0.0253 to 0.0395 of the cost, and at a normal margin a buffer costs more than it saves. (https://5dlibrary.com/answers/en/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c05/) 第5章 五天的敞口:海外连线期间的汇率风险 — 第五章计算五天连线窗口的汇率风险:厚尾下 99% 的损失是成本的 0.0253 至 0.0395,正常利润率下缓冲得不偿失。 (https://5dlibrary.com/answers/zh/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c05/) - Ch.6 The Refund That Never Quite Arrives — Chapter 6 shows a promised refund paying 0.9550 of its face by card and 0.8775 in cash, and a joint buffer of 0.0276 pushing a bag into the deadzone. (https://5dlibrary.com/answers/en/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c06/) 第6章 永远差一点才到的退税 — 第六章说明承诺的退税刷卡实得 0.9550、现金实得 0.8775,联合缓冲 0.0276 可能把一个包推进死区。 (https://5dlibrary.com/answers/zh/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c06/) - Ch.7 Mother and Child Coins, the Equable Transport and Block Booking — Chapter 7 reads "mother and child weigh each other" in the Discourses of the States (correcting its attribution to the Guanzi), the Han treatise on the equable transport and levelling office, and the Chicago school on tying: a strap sent alone pays 0.1276 of its value in freight, inside a bag box nothing. (https://5dlibrary.com/answers/en/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c07/) 第7章 子母相权、均输平准与整批订片 — 第七章读《国语·周语下》「子母相权」(并更正它出自《国语》而非《管子·轻重甲》)、《汉书·食货志》的均输平准与芝加哥学派的搭售理论:背带单独寄送运费是货值的 0.1276,放进包的箱子里是 0。 (https://5dlibrary.com/answers/zh/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c07/) - Ch.8 Practice: One Year of a Buyer Shop and One Test Protocol — Chapter 8 combines these in a year of trading and one test protocol. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c08/) 第8章 实践:一家买手店的一年与一份检验方案 — 第八章把这些合成一年的经营与一份检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cross-border-buyer-pricing-fee-threshold-routing-accessory-attach-economics/c08/) ## 5DL-B-111 · Computational Aesthetics of Boutique Luxury Web Design, Haptic Mental Imagery and Female Conversion Psychology 中文书名:女性精品买手店计算美学、触觉心理模拟与稀缺转化心理学 DOI: pending Cite: Fan, Y. (2026). Computational Aesthetics of Boutique Luxury Web Design, Haptic Mental Imagery and Female Conversion Psychology (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its shoppers, prices, rates and parameters are illustrative models; all calculations are generated by its programs; the book is not legal, financial or psychological advice. Chapter 3 treats photographs as readings of size and feel: under illustrative assumptions a hand-held photograph and a capacity video cut the doubt about size from 4.0 cm to 0.80 cm and abandonment by 0.4710. Abstract: This volume, written for boutiques and personal buyers who sell luxury bags online to customers in Taiwan and Southeast Asia, studies three things: how a page's whitespace, type and colour signal a boutique rather than a reseller; how photographs let a shopper simulate touching what she cannot touch; and whether deliberate friction and a memorable unboxing lower regret and bring customers back. None of its models depends on the customer's sex. Its shoppers, prices, rates and parameters are illustrative models; all calculations are generated by its programs; the book is not legal, financial or psychological advice. Chapter 1 writes whitespace as a costly signal that resellers cannot afford: the value index is best at a whitespace ratio of 0.550 and within 0.95 of that from 0.512 to 0.597, so the often quoted 0.45 to 0.68 is too wide; whitespace alone does not separate a collage from a grid, edge density does. Chapter 2 shows that a type ratio of 1.414 breaks a 12-character Chinese title into 3 lines on a phone and that gold on warm cream has a contrast of only 2.75. Chapter 3 treats photographs as readings of size and feel: under illustrative assumptions a hand-held photograph and a capacity video cut the doubt about size from 4.0 cm to 0.80 cm and abandonment by 0.4710. Chapter 4 shows that order beats number: the same pictures convert 0.5768 with the hand first and 0.1935 with white shots first. Chapter 5 shows benign friction moving regret before the money is spent: costly cancellations per sale fall from 0.1336 to 0.0437, but friction does not pay by itself. Chapter 6 applies the peak-end rule: a doubling of repurchase needs a slope of 0.1918 per point of remembered value, and small cohorts show false doublings 0.3347 of the time. Chapter 7 reads the Record of Trades on combining four factors, with O-ring arithmetic, the detailed likenesses of the Shuo Ren poem in the Book of Songs, and Veblen and Leibenstein: a Veblen effect moves the profit-maximising price from 15000 to 21000, but demand never rises with price. Chapter 8 builds one month layer by layer and one test protocol. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷写给在线上把精品包袋卖给台湾与东南亚客户的精品店与代购,研究三件事:页面的留白、字体与颜色怎样表明这是精品店而不是转售商;照片怎样让客户模拟触摸她摸不到的东西;刻意的摩擦与令人难忘的开箱,是否降低后悔并让客户回来。书中没有任何模型取决于客户的性别。书中的客户、价格、比率与参数均为示意模型;全部计算由书中程序生成;本书不是法律、财务或心理建议。 第一章把留白写成转售商负担不起的昂贵信号:价值指数在留白比例 0.550 处最高,从 0.512 到 0.597 都在最高值的 0.95 以内,而常被引用的 0.45 至 0.68 太宽;单靠留白分不开拼贴与格子,边缘密度分得开。第二章说明字号比例 1.414 会把 12 字的中文标题在手机上拆成 3 行,金字在暖奶油色上的对比度只有 2.75。第三章把照片当作尺寸与手感的读数:在示意假设下,手拿照与容量短片把尺寸疑虑从 4.0 公分降到 0.80 公分,弃单率降低 0.4710。第四章说明次序胜过数量:同样的照片,手拿照在前成交 0.5768,白底图在前只有 0.1935。 第五章说明良性摩擦让后悔发生在花钱之前:每笔销售的赔钱取消从 0.1336 降到 0.0437,但单靠它并不划算。第六章用峰终定律说明,复购翻倍需要每点记忆值 0.1918 的斜率,而小样本有 0.3347 的机会出现假的翻倍。第七章读《考工记》「合此四者,然后可以为良」与 O 形环算术、《诗经·卫风·硕人》的赋体细节,以及凡勃伦与莱本斯坦:凡勃伦效应把最佳价格从 15000 推到 21000,但需求从不随价格上升。第八章一层一层建出一个月,并给出一份检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Whitespace as a Signal: Clutter, Trust and the First Three Seconds — Chapter 1 writes whitespace as a costly signal that resellers cannot afford: the value index is best at a whitespace ratio of 0.550 and within 0.95 of that from 0.512 to 0.597, so the often quoted 0.45 to 0.68 is too wide; whitespace alone does not separate a collage from a grid, edge density does. (https://5dlibrary.com/answers/en/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c01/) 第1章 留白是信号:拥挤度、信任与最初三秒 — 第一章把留白写成转售商负担不起的昂贵信号:价值指数在留白比例 0.550 处最高,从 0.512 到 0.597 都在最高值的 0.95 以内,而常被引用的 0.45 至 0.68 太宽;单靠留白分不开拼贴与格子,边缘密度分得开。 (https://5dlibrary.com/answers/zh/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c01/) - Ch.2 Type Scales and Warm Colour: Ratios, Lines and Contrast — Chapter 2 shows that a type ratio of 1.414 breaks a 12-character Chinese title into 3 lines on a phone and that gold on warm cream has a contrast of only 2.75. (https://5dlibrary.com/answers/en/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c02/) 第2章 字号阶梯与暖色:比例、行数与对比度 — 第二章说明字号比例 1.414 会把 12 字的中文标题在手机上拆成 3 行,金字在暖奶油色上的对比度只有 2.75。 (https://5dlibrary.com/answers/zh/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c02/) - Ch.3 Touch at a Distance: Haptic Imagery and the Doubt About Size — Chapter 3 treats photographs as readings of size and feel: under illustrative assumptions a hand-held photograph and a capacity video cut the doubt about size from 4.0 cm to 0.80 cm and abandonment by 0.4710. (https://5dlibrary.com/answers/en/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c03/) 第3章 隔空的触觉:触觉想象与对尺寸的疑虑 — 第三章把照片当作尺寸与手感的读数:在示意假设下,手拿照与容量短片把尺寸疑虑从 4.0 公分降到 0.80 公分,弃单率降低 0.4710。 (https://5dlibrary.com/answers/zh/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c03/) - Ch.4 The Order of the Gallery and the Limits of a Test — Chapter 4 shows that order beats number: the same pictures convert 0.5768 with the hand first and 0.1935 with white shots first. (https://5dlibrary.com/answers/en/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c04/) 第4章 图组的次序与测试的极限 — 第四章说明次序胜过数量:同样的照片,手拿照在前成交 0.5768,白底图在前只有 0.1935。 (https://5dlibrary.com/answers/zh/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c04/) - Ch.5 Benign Friction: Moving Regret Before the Money Is Spent — Chapter 5 shows benign friction moving regret before the money is spent: costly cancellations per sale fall from 0.1336 to 0.0437, but friction does not pay by itself. (https://5dlibrary.com/answers/en/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c05/) 第5章 良性摩擦:让后悔发生在花钱之前 — 第五章说明良性摩擦让后悔发生在花钱之前:每笔销售的赔钱取消从 0.1336 降到 0.0437,但单靠它并不划算。 (https://5dlibrary.com/answers/zh/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c05/) - Ch.6 Peak and End: What the Second Purchase Remembers — Chapter 6 applies the peak-end rule: a doubling of repurchase needs a slope of 0.1918 per point of remembered value, and small cohorts show false doublings 0.3347 of the time. (https://5dlibrary.com/answers/en/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c06/) 第6章 峰值与终点:二次购买记住了什么 — 第六章用峰终定律说明,复购翻倍需要每点记忆值 0.1918 的斜率,而小样本有 0.3347 的机会出现假的翻倍。 (https://5dlibrary.com/answers/zh/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c06/) - Ch.7 Season, Place, Material, Skill; the Poem of Detail; and Veblen — Chapter 7 reads the Record of Trades on combining four factors, with O-ring arithmetic, the detailed likenesses of the Shuo Ren poem in the Book of Songs, and Veblen and Leibenstein: a Veblen effect moves the profit-maximising price from 15000 to 21000, but demand never rises with price. (https://5dlibrary.com/answers/en/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c07/) 第7章 天时地气材美工巧、赋体的细节与凡勃伦 — 第七章读《考工记》「合此四者,然后可以为良」与 O 形环算术、《诗经·卫风·硕人》的赋体细节,以及凡勃伦与莱本斯坦:凡勃伦效应把最佳价格从 15000 推到 21000,但需求从不随价格上升。 (https://5dlibrary.com/answers/zh/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c07/) - Ch.8 Practice: One Month, Layer by Layer, and One Test Protocol — Chapter 8 builds one month layer by layer and one test protocol. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c08/) 第8章 实践:一个月,一层一层,以及一份检验方案 — 第八章一层一层建出一个月,并给出一份检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/luxury-boutique-computational-aesthetics-haptic-imagery-female-conversion-psychology/c08/) ## 5DL-B-112 · Live-Shopping Flash Inventory Locks, Bayesian Bank-Feed Reconciliation and Personalized Styling CRM 中文书名:海外连线脉冲库存锁、银行流水贝叶斯核销与个性化买手推荐系统 DOI: pending Cite: Fan, Y. (2026). Live-Shopping Flash Inventory Locks, Bayesian Bank-Feed Reconciliation and Personalized Styling CRM (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its customers, delays, rates and parameters are illustrative models; all calculations are generated by its programs; the book is not legal, financial, banking or data-protection advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This volume, written for personal buyers who sell by live stream from boutiques abroad to customers in Taiwan and Southeast Asia, take deposits by bank transfer and keep a private pool of past customers, studies three things: how long to lock a scarce piece and when to buy it; how to match bank rows to orders when account digits collide or are missing; and how to recommend new arrivals and time messages without driving customers away. Five-element colour names are treated only as labels customers choose for themselves, with no claim about luck. Its customers, delays, rates and parameters are illustrative models; all calculations are generated by its programs; the book is not legal, financial, banking or data-protection advice. Chapter 1 shows that with two-hour locks only 0.7465 of pieces are paid for during the session, against 0.9899 with 10-minute locks, and that with a long queue buying on the lock leaves 0.0010 unpaid while waiting 15 minutes loses 0.1730 to walk-ins. Chapter 2 solves the cascade in closed form: the best lock is 12 minutes (0.9909), and ordering by payment record shortens the wait but shuts out new customers. Chapter 3 computes that 200 transfers of the same deposit share their last five digits somewhere with chance 0.1806, and that masked rows leave 50.3 rows ambiguous. Chapter 4 matches rows to orders with likelihood scores and the Hungarian method: with a unique offset per order only 0.0258 of rows need a person, no automatic match among 5090 is wrong, and checking falls from 348.3 to 9.0 minutes a session. Chapter 5's hybrid recommender reaches 0.1897 against 0.1620 for collaborative filtering alone; a colour label helps only when it predicts choices and costs 0.0260 when it does not. Chapter 6 shows a blast lifting the returning share from 0.2100 to 0.3848 while 2308 customers block the shop, and targeting twice a month reaching 0.4240 with 849 blocked; the best day to write is about day 103 after a purchase. Chapter 7 reads the permits of the Song maritime trade offices, the secret codes of the Rishengchang draft bank and Fellegi and Sunter's theory of record linkage: a row agreeing only in amount and time reaches a posterior of only 0.71663. Chapter 8 puts these into a year's ledger and one test protocol. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷写给在海外专柜以直播向台湾与东南亚客户销售、以银行转账收定金、并经营过去客户私域池的代购,研究三件事:一件稀缺的货该锁多久、何时买下;账号后五码碰撞或缺失时,怎样把银行流水对上订单;怎样推荐新到货、安排讯息时机,而不把客户赶走。书中的五行色系名称只当作客户自选的标签,不对运势作任何主张。书中的客户、延迟、比率与参数均为示意模型;全部计算由书中程序生成;本书不是法律、财务、银行或个人资料保护方面的建议。 第一章说明两小时的锁只有 0.7465 的货在直播期间收到款,10 分钟的锁是 0.9899;队列长时锁了就买只有 0.0010 的货没人付款,等 15 分钟则有 0.1730 被现场顾客买走。第二章求出递补的闭式解:最佳的锁是 12 分钟(0.9909);按付款记录排队缩短等待,却把新客户关在门外。第三章计算 200 笔同额定金后五码碰撞的机会是 0.1806,被遮蔽的行使 50.3 行模糊。第四章用似然分数与匈牙利算法对账:每张订单一个唯一尾数时,只剩 0.0258 的行需要人工检查,5090 个自动匹配没有错误,每场工作从 348.3 分钟降到 9.0 分钟。 第五章的混合推荐达到 0.1897,单用协同过滤是 0.1620;颜色标签只有能预测选择时才有帮助,没有信息时损失 0.0260。第六章说明群发把回头客比例从 0.2100 提高到 0.3848,却使 2308 位客户封锁;每月两次精准推送达到 0.4240,封锁 849 位;写信的最好时机约在购买后第 103 天。第七章读宋代市舶司的公凭、日昇昌的密押与费勒吉和森特的记录连接理论:只有金额与时间相符的一行,后验只有 0.71663。第八章把这些放进一年的账与一份检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Locks on a Live Shelf: How Long to Hold a Piece — Chapter 1 shows that with two-hour locks only 0.7465 of pieces are paid for during the session, against 0.9899 with 10-minute locks, and that with a long queue buying on the lock leaves 0.0010 unpaid while waiting 15 minutes loses 0.1730 to walk-ins. (https://5dlibrary.com/answers/en/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c01/) 第1章 直播货架上的锁:一件货该锁多久 — 第一章说明两小时的锁只有 0.7465 的货在直播期间收到款,10 分钟的锁是 0.9899;队列长时锁了就买只有 0.0010 的货没人付款,等 15 分钟则有 0.1730 被现场顾客买走。 (https://5dlibrary.com/answers/zh/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c01/) - Ch.2 The Cascade in Closed Form, and Who Goes First — Chapter 2 solves the cascade in closed form: the best lock is 12 minutes (0.9909), and ordering by payment record shortens the wait but shuts out new customers. (https://5dlibrary.com/answers/en/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c02/) 第2章 递补的闭式解,以及谁排第一 — 第二章求出递补的闭式解:最佳的锁是 12 分钟(0.9909);按付款记录排队缩短等待,却把新客户关在门外。 (https://5dlibrary.com/answers/zh/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c02/) - Ch.3 The Same Five Digits Twice: Collisions, Masks and Offsets — Chapter 3 computes that 200 transfers of the same deposit share their last five digits somewhere with chance 0.1806, and that masked rows leave 50.3 rows ambiguous. (https://5dlibrary.com/answers/en/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c03/) 第3章 同样的五位数出现两次:碰撞、遮蔽与尾数微调 — 第三章计算 200 笔同额定金后五码碰撞的机会是 0.1806,被遮蔽的行使 50.3 行模糊。 (https://5dlibrary.com/answers/zh/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c03/) - Ch.4 Matching Rows to Orders: Likelihoods and the Hungarian Method — Chapter 4 matches rows to orders with likelihood scores and the Hungarian method: with a unique offset per order only 0.0258 of rows need a person, no automatic match among 5090 is wrong, and checking falls from 348.3 to 9.0 minutes a session. (https://5dlibrary.com/answers/en/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c04/) 第4章 把流水对上订单:似然比与匈牙利算法 — 第四章用似然分数与匈牙利算法对账:每张订单一个唯一尾数时,只剩 0.0258 的行需要人工检查,5090 个自动匹配没有错误,每场工作从 348.3 分钟降到 9.0 分钟。 (https://5dlibrary.com/answers/zh/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c04/) - Ch.5 A Hybrid Recommender: Taste in Shapes and Colours People Name — Chapter 5's hybrid recommender reaches 0.1897 against 0.1620 for collaborative filtering alone; a colour label helps only when it predicts choices and costs 0.0260 when it does not. (https://5dlibrary.com/answers/en/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c05/) 第5章 混合推荐:轮廓品味与客户自己说的颜色 — 第五章的混合推荐达到 0.1897,单用协同过滤是 0.1620;颜色标签只有能预测选择时才有帮助,没有信息时损失 0.0260。 (https://5dlibrary.com/answers/zh/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c05/) - Ch.6 Waking the Pool: Blasts, Targets and the Right Day — Chapter 6 shows a blast lifting the returning share from 0.2100 to 0.3848 while 2308 customers block the shop, and targeting twice a month reaching 0.4240 with 849 blocked; the best day to write is about day 103 after a purchase. (https://5dlibrary.com/answers/en/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c06/) 第6章 唤醒私域池:群发、精准与对的那一天 — 第六章说明群发把回头客比例从 0.2100 提高到 0.3848,却使 2308 位客户封锁;每月两次精准推送达到 0.4240,封锁 849 位;写信的最好时机约在购买后第 103 天。 (https://5dlibrary.com/answers/zh/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c06/) - Ch.7 Permits, Secret Codes and Record Linkage — Chapter 7 reads the permits of the Song maritime trade offices, the secret codes of the Rishengchang draft bank and Fellegi and Sunter's theory of record linkage: a row agreeing only in amount and time reaches a posterior of only 0.71663. (https://5dlibrary.com/answers/en/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c07/) 第7章 公凭、密押与记录连接 — 第七章读宋代市舶司的公凭、日昇昌的密押与费勒吉和森特的记录连接理论:只有金额与时间相符的一行,后验只有 0.71663。 (https://5dlibrary.com/answers/zh/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c07/) - Ch.8 Practice: One Year of Live Sessions and One Test Protocol — Chapter 8 puts these into a year's ledger and one test protocol. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c08/) 第8章 实践:一年的直播与一份检验方案 — 第八章把这些放进一年的账与一份检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/live-shopping-inventory-locks-bayesian-bank-reconciliation-personalized-styling-crm/c08/) ## 5DL-B-113 · Unmanned Shared-Space Revenue Management, Cross-Venue Prepaid Float and Zero-Take-Rate Trojan Onboarding 中文书名:无人共享空间收益管理、跨店储值浮存金与木马收编博弈 DOI: pending Cite: Fan, Y. (2026). Unmanned Shared-Space Revenue Management, Cross-Venue Prepaid Float and Zero-Take-Rate Trojan Onboarding (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its clients, demands, rates and parameters are illustrative models; all calculations are generated by its programs; the book is not legal, financial, accounting or tax advice. Chapter 7 reads the New Book of Tang on flying cash, the Song office that let official buildings and the theory of two-sided markets: Little's law gives 17500 against a simulated 17554. Abstract: This volume, written for operators of unmanned photo and creative studios rented by the hour and for platforms that gather independent venues into one network with a shared stored-value pass, studies three things: how to price twenty open hours; what a cross-venue pass does to sales, to the money held and to the platform's margin; and when a zero-commission offer makes joining the network the best choice for an independent owner. Its clients, demands, rates and parameters are illustrative models; all calculations are generated by its programs; the book is not legal, financial, accounting or tax advice. Chapter 1 shows that bands alone reach 1.0576 in an hour-by-hour model, and that with bookings, segments, a commercial rate and a consecutive-hours ladder the gain is 1.1721 against a tuned single rate and 1.3165 against a habit price, so the claim of more than 40% does not hold. Chapter 2 shows that a 15-minute jump after a five-minute grace earns 143.9 a booking, while a steep headcount ladder loses money when groups are price-sensitive. Chapter 3 follows one pass to the end: breakage is 0.1808, and the platform keeps 2610.7 of TWD 10000 but only 980.0 if all credit is used; a ten-venue pass sells 9.3775 times as well as a one-venue pass, yet its buyers use 0.9962 of the credit. Chapter 4 shows that breakage estimated after twelve months is 0.2358 against a true 0.3409, and that keeping 0.7 of the liability in cash fails on 0.2225 of paths. Chapter 5 shows that joining beats staying alone unless more than 0.8105 of platform bookings are cannibalised, but beats a flat-commission rival only while platform bookings stay below 1.2500 times an owner's own; a lifetime commission invites leaking unless detection exceeds 0.1907. Chapter 6 shows monthly settlement costing venues 31806 a year and flat credit turning a margin of 104.9 an hour into a loss of 204.5. Chapter 7 reads the New Book of Tang on flying cash, the Song office that let official buildings and the theory of two-sided markets: Little's law gives 17500 against a simulated 17554. Chapter 8 puts these into a year's ledger: the designed network earns 1.3111 times the plain one. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷写给按小时出租的无人值守摄影棚与创作空间的经营者,以及把独立场馆收编进同一网络、发行跨店通用储值卡的平台,研究三件事:二十个营业小时怎样定价;跨店通用卡对销售、所持资金与平台毛利有什么影响;以及零抽成方案何时会让加入网络成为独立主理人的最佳选择。书中的客人、需求、费率与参数均为示意模型;全部计算由书中程序生成;本书不是法律、财务、会计或税务方面的建议。 第一章说明只分时段在逐时模型中增加到 1.0576 倍;加入预约、客群、商用费率与连租阶梯后,对调好的单一时薪为 1.1721 倍,对凭习惯定的低价为 1.3165 倍,所以「40% 以上」的说法不成立。第二章说明宽限五分钟的 15 分钟跳表每笔预约收入 143.9,而陡峭的人头阶梯在团体对价格敏感时亏钱。第三章跟着一张卡走到底:残值是 0.1808,平台在 10000 元中留下 2610.7,但全部额度都被使用时只有 980.0;十间店的卡卖出单店卡的 9.3775 倍,买家却用掉 0.9962 的额度。第四章说明十二个月后估出的残值是 0.2358,真值是 0.3409;以现金保留负债的 0.7 时有 0.2225 的路径付不出款。 第五章说明只要平台派单被蚕食的比例不超过 0.8105,加入就胜过自己经营,但只有平台派单低于自带流量的 1.2500 倍时才胜过统一抽成的竞争者;永久抽成会招来带客,除非发现机会超过 0.1907。第六章说明每月结算让场馆每年付出 31806,统一额度把每小时 104.9 的毛利变成 204.5 的亏损。第七章读《新唐书》的飞钱、宋代出租官房的店宅务与双边市场理论:利特尔定律给出 17500,模拟为 17554。第八章把这些放进一年的账:设计过的网络赚到朴素网络的 1.3111 倍。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Twenty Hours, Three Bands and a Ladder — Chapter 1 shows that bands alone reach 1.0576 in an hour-by-hour model, and that with bookings, segments, a commercial rate and a consecutive-hours ladder the gain is 1.1721 against a tuned single rate and 1.3165 against a habit price, so the claim of more than 40% does not hold. (https://5dlibrary.com/answers/en/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c01/) 第1章 二十小时、三个时段与一道阶梯 — 第一章说明只分时段在逐时模型中增加到 1.0576 倍;加入预约、客群、商用费率与连租阶梯后,对调好的单一时薪为 1.1721 倍,对凭习惯定的低价为 1.3165 倍,所以「40% 以上」的说法不成立。 (https://5dlibrary.com/answers/zh/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c01/) - Ch.2 The Fifteen-Minute Jump and the Headcount Ladder — Chapter 2 shows that a 15-minute jump after a five-minute grace earns 143.9 a booking, while a steep headcount ladder loses money when groups are price-sensitive. (https://5dlibrary.com/answers/en/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c02/) 第2章 十五分钟跳表与超额人头阶梯 — 第二章说明宽限五分钟的 15 分钟跳表每笔预约收入 143.9,而陡峭的人头阶梯在团体对价格敏感时亏钱。 (https://5dlibrary.com/answers/zh/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c02/) - Ch.3 The Universal Pass: Float, Breakage and Fatigue — Chapter 3 follows one pass to the end: breakage is 0.1808, and the platform keeps 2610.7 of TWD 10000 but only 980.0 if all credit is used; a ten-venue pass sells 9.3775 times as well as a one-venue pass, yet its buyers use 0.9962 of the credit. (https://5dlibrary.com/answers/en/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c03/) 第3章 跨店通用卡:浮存金、残值与审美疲劳 — 第三章跟着一张卡走到底:残值是 0.1808,平台在 10000 元中留下 2610.7,但全部额度都被使用时只有 980.0;十间店的卡卖出单店卡的 9.3775 倍,买家却用掉 0.9962 的额度。 (https://5dlibrary.com/answers/zh/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c03/) - Ch.4 Breakage Bounds and the Cost of Spending the Float — Chapter 4 shows that breakage estimated after twelve months is 0.2358 against a true 0.3409, and that keeping 0.7 of the liability in cash fails on 0.2225 of paths. (https://5dlibrary.com/answers/en/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c04/) 第4章 残值的精算边界与花掉浮存金的代价 — 第四章说明十二个月后估出的残值是 0.2358,真值是 0.3409;以现金保留负债的 0.7 时有 0.2225 的路径付不出款。 (https://5dlibrary.com/answers/zh/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c04/) - Ch.5 The Zero-Take Trojan: When Joining Dominates — Chapter 5 shows that joining beats staying alone unless more than 0.8105 of platform bookings are cannibalised, but beats a flat-commission rival only while platform bookings stay below 1.2500 times an owner's own; a lifetime commission invites leaking unless detection exceeds 0.1907. (https://5dlibrary.com/answers/en/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c05/) 第5章 零抽成木马:何时加入是占优策略 — 第五章说明只要平台派单被蚕食的比例不超过 0.8105,加入就胜过自己经营,但只有平台派单低于自带流量的 1.2500 倍时才胜过统一抽成的竞争者;永久抽成会招来带客,除非发现机会超过 0.1907。 (https://5dlibrary.com/answers/zh/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c05/) - Ch.6 Multi-Tenant Clearing and Graded Credit — Chapter 6 shows monthly settlement costing venues 31806 a year and flat credit turning a margin of 104.9 an hour into a loss of 204.5. (https://5dlibrary.com/answers/en/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c06/) 第6章 多租户清算与分等额度 — 第六章说明每月结算让场馆每年付出 31806,统一额度把每小时 104.9 的毛利变成 204.5 的亏损。 (https://5dlibrary.com/answers/zh/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c06/) - Ch.7 Flying Cash, Graded Rents and Two-Sided Markets — Chapter 7 reads the New Book of Tang on flying cash, the Song office that let official buildings and the theory of two-sided markets: Little's law gives 17500 against a simulated 17554. (https://5dlibrary.com/answers/en/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c07/) 第7章 飞钱、分等赁值与双边市场 — 第七章读《新唐书》的飞钱、宋代出租官房的店宅务与双边市场理论:利特尔定律给出 17500,模拟为 17554。 (https://5dlibrary.com/answers/zh/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c07/) - Ch.8 Practice: One Year of Ten Studios and One Test Protocol — Chapter 8 puts these into a year's ledger: the designed network earns 1.3111 times the plain one. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c08/) 第8章 实务:十间棚的一年与一份检验方案 — 第八章把这些放进一年的账:设计过的网络赚到朴素网络的 1.3111 倍。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/unmanned-space-revenue-management-prepaid-float-platform-onboarding-games/c08/) ## 5DL-B-114 · Algorithmic Diagnostic Paywalls, Curiosity-Gap Calibration and Tiered Report Unlock Economics 中文书名:算法诊断收费站、悬念缺口校准与多阶解锁经济学 DOI: pending Cite: Fan, Y. (2026). Algorithmic Diagnostic Paywalls, Curiosity-Gap Calibration and Tiered Report Unlock Economics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its users, rates and parameters are illustrative models; all calculations are generated by its programs; the book is not legal, financial or consumer-protection advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This volume, written for sellers of self-serve diagnostic tools (chart readings, quantitative scores, personality and signal reports), studies three things: how much of a report to show free; what asking for input before the paywall really does; and how a price menu, a subscription and routing among thirty sites add up. The book makes no claim that any chart, sign, reading or signal predicts anything; its models apply to tools whose statements a user can check. Its users, rates and parameters are illustrative models; all calculations are generated by its programs; the book is not legal, financial or consumer-protection advice. Chapter 1 shows that trust times the remaining gap has an interior optimum, 0.25 in the base case, but that across 60 cases it runs from 0.000 to 0.425, with only 15 inside the claimed 0.28 to 0.42; that interval appears reliably when free statements are Barnum statements and users do not notice, and then 0.5364 of buyers buy on a mistake. Chapter 2 shows that input before the paywall helps only if each step raises willingness to pay by more than 1.0753, and that selection alone raises paywall conversion by 1.4080 while purchases per visitor fall to 0.6439. Chapter 3 shows three prices earning 4.9029 times one price, the middle tier adding only 1.0065 and a subscription a further 1.3575. Chapter 4 solves thirty sites as one Markov chain: routing within themes raises purchases by 1.4087, and four links a page is best. Chapter 5 shows that a free quota is best at one reading a month, and that a user who re-asks a sensitive tool six times ends on the answer she wanted with chance 0.8840. Chapter 6 shows a refund promise paying while disappointment stays below 0.1904, and dark patterns ending the year at 0.6856 of the honest page's revenue. Chapter 7 reads the hexagram Meng on the second asking and the Guiguzi on giving first: re-asking drives the likelihood ratio from 2.3333 to 1.1325, and giving first is a signal only an accurate tool can afford. Chapter 8 puts these into a year's ledger: the honest matrix earns 1.6700 times the dark one. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷写给自助式诊断工具(排盘、量化评分、性格与信号报告)的卖家,研究三件事:报告要免费展示多少;收银台前要求输入究竟起什么作用;以及价格菜单、订阅与三十个站点之间的导流怎样加总。本书不主张任何命盘、星座、解读或信号能预测任何事情;书中的模型适用于用户能够核对其陈述的工具。书中的用户、比率与参数均为示意模型;全部计算由书中程序生成;本书不是法律、财务或消费者保护方面的建议。 第一章说明「信任乘以剩余缺口」有内部最优点,基准情形为 0.25,但在 60 种情形中从 0.000 到 0.425 不等,只有 15 种落在说法中的 0.28 至 0.42;这个区间在免费陈述是巴纳姆式、用户又没有察觉时才稳定出现,那时有 0.5364 的买家基于误判购买。第二章说明收银台前的输入,只有在每一步让支付意愿提高超过 1.0753 时才有帮助,而单靠筛选就能让收银台转化提高到 1.4080 倍,同时每位访客的购买降到 0.6439。第三章说明三个价格赚到单一价格的 4.9029 倍,中档只增加到 1.0065 倍,订阅再增加到 1.3575 倍。第四章把三十个站点解成一条马尔可夫链:同主题导流使购买提高到 1.4087 倍,每页四条链接最好。 第五章说明免费额度在每月一次时最好,而对敏感的工具追问六次的用户,以 0.8840 的机会停在想要的答案上。第六章说明失望率低于 0.1904 时退款承诺有利,暗黑模式一年下来只有诚实页面收入的 0.6856。第七章读蒙卦「初筮告,再三渎,渎则不告」与《鬼谷子》「欲取,反与」:反复问策把似然比从 2.3333 推到 1.1325,而先给是只有准确工具才负担得起的信号。第八章把这些放进一年的账:诚实的矩阵赚到暗黑矩阵的 1.6700 倍。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Curiosity Gap and the Free/Paid Cut — Chapter 1 shows that trust times the remaining gap has an interior optimum, 0.25 in the base case, but that across 60 cases it runs from 0.000 to 0.425, with only 15 inside the claimed 0.28 to 0.42; that interval appears reliably when free statements are Barnum statements and users do not notice, and then 0.5364 of buyers buy on a mistake. (https://5dlibrary.com/answers/en/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c01/) 第1章 悬念缺口与付费墙的切割点 — 第一章说明「信任乘以剩余缺口」有内部最优点,基准情形为 0.25,但在 60 种情形中从 0.000 到 0.425 不等,只有 15 种落在说法中的 0.28 至 0.42;这个区间在免费陈述是巴纳姆式、用户又没有察觉时才稳定出现,那时有 0.5364 的买家基于误判购买。 (https://5dlibrary.com/answers/zh/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c01/) - Ch.2 Sunk Input: Commitment or Selection? — Chapter 2 shows that input before the paywall helps only if each step raises willingness to pay by more than 1.0753, and that selection alone raises paywall conversion by 1.4080 while purchases per visitor fall to 0.6439. (https://5dlibrary.com/answers/en/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c02/) 第2章 沉没输入:承诺,还是筛选? — 第二章说明收银台前的输入,只有在每一步让支付意愿提高超过 1.0753 时才有帮助,而单靠筛选就能让收银台转化提高到 1.4080 倍,同时每位访客的购买降到 0.6439。 (https://5dlibrary.com/answers/zh/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c02/) - Ch.3 Three Prices and a Subscription — Chapter 3 shows three prices earning 4.9029 times one price, the middle tier adding only 1.0065 and a subscription a further 1.3575. (https://5dlibrary.com/answers/en/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c03/) 第3章 三个价格与一个订阅 — 第三章说明三个价格赚到单一价格的 4.9029 倍,中档只增加到 1.0065 倍,订阅再增加到 1.3575 倍。 (https://5dlibrary.com/answers/zh/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c03/) - Ch.4 Thirty Sites as One Markov Chain — Chapter 4 solves thirty sites as one Markov chain: routing within themes raises purchases by 1.4087, and four links a page is best. (https://5dlibrary.com/answers/en/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c04/) 第4章 三十个站点构成一条马尔可夫链 — 第四章把三十个站点解成一条马尔可夫链:同主题导流使购买提高到 1.4087 倍,每页四条链接最好。 (https://5dlibrary.com/answers/zh/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c04/) - Ch.5 Free Quotas and the Second Asking — Chapter 5 shows that a free quota is best at one reading a month, and that a user who re-asks a sensitive tool six times ends on the answer she wanted with chance 0.8840. (https://5dlibrary.com/answers/en/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c05/) 第5章 免费额度与再三问策 — 第五章说明免费额度在每月一次时最好,而对敏感的工具追问六次的用户,以 0.8840 的机会停在想要的答案上。 (https://5dlibrary.com/answers/zh/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c05/) - Ch.6 Refund Promises and Dark Patterns — Chapter 6 shows a refund promise paying while disappointment stays below 0.1904, and dark patterns ending the year at 0.6856 of the honest page's revenue. (https://5dlibrary.com/answers/en/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c06/) 第6章 退款承诺与暗黑模式 — 第六章说明失望率低于 0.1904 时退款承诺有利,暗黑模式一年下来只有诚实页面收入的 0.6856。 (https://5dlibrary.com/answers/zh/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c06/) - Ch.7 The Second Asking, Giving First and Costly Signals — Chapter 7 reads the hexagram Meng on the second asking and the Guiguzi on giving first: re-asking drives the likelihood ratio from 2.3333 to 1.1325, and giving first is a signal only an accurate tool can afford. (https://5dlibrary.com/answers/en/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c07/) 第7章 再三渎、欲取反与及昂贵信号 — 第七章读蒙卦「初筮告,再三渎,渎则不告」与《鬼谷子》「欲取,反与」:反复问策把似然比从 2.3333 推到 1.1325,而先给是只有准确工具才负担得起的信号。 (https://5dlibrary.com/answers/zh/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c07/) - Ch.8 Practice: One Year of the Matrix and One Test Protocol — Chapter 8 puts these into a year's ledger: the honest matrix earns 1.6700 times the dark one. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c08/) 第8章 实务:矩阵的一年与一份检验方案 — 第八章把这些放进一年的账:诚实的矩阵赚到暗黑矩阵的 1.6700 倍。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/algorithmic-diagnostic-paywalls-curiosity-gap-calibration-tiered-unlock-economics/c08/) ## 5DL-B-115 · Large-Scale Monograph Catalog Bundling, Platform Upload Throttling and Sovereign Direct-to-Reader Arbitrage 中文书名:百卷数字专著库混合捆绑、平台限流装箱与主权直售套利数学 DOI: pending Cite: Fan, Y. (2026). Large-Scale Monograph Catalog Bundling, Platform Upload Throttling and Sovereign Direct-to-Reader Arbitrage (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its readers, demands, rates and parameters are illustrative models; all calculations are generated by its programs; platform rules quoted were checked on 11 November 2026; the book is not legal, financial or tax advice. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. Abstract: This volume, written for authors and small publishers who hold a large catalogue of digital monographs (here a hundred volumes with runnable code), studies three things: how much value a complete-set bundle can collect; how to schedule releases under a weekly upload limit and royalty bands; and how direct sales, a platform and open access complement each other. Its readers, demands, rates and parameters are illustrative models; all calculations are generated by its programs; platform rules quoted were checked on 11 November 2026; the book is not legal, financial or tax advice. Chapter 1 shows a bundle of 100 independently valued volumes collecting 0.6839 of value against 0.3356 for singles, with the coefficient of variation falling to 0.2003, but only 0.4595 with a moderate common taste, so "nearly all consumer surplus" is too strong. Chapter 2 shows that within one shop no mixed menu beats the full catalogue, while singles on the platform raise total revenue by 1.4026 with no overlap. Chapter 3 shows that under a limit of two new titles per format a week, releasing by demand earns 1.7491 times releasing by volume number and the limit costs 0.2069 of the first year, and that the cliff at USD 12.99 cuts a sale's royalty from 8.8830 to 4.5500. Chapter 4 shows omnibus editions changing first-year royalties by a few per cent, while three format queues in parallel add 10835. Chapter 5 shows a catalogue sold at USD 299 netting 283.55, which is 41.80 ebooks at 9.99 but only 31.92 at the new ceiling of 12.99, and an own checkout beating a merchant of record only above 283.9 sales a year. Chapter 6 shows that open access, if a fifth of buyers switch to free copies, pays only if each download leads to a paid sale with chance above 0.0060, and that citation sales grow from 1701 in year one to 16729 in year five. Chapter 7 reads the History of Song on the Directorate of Education's printing blocks and the collection Baichuan xuehai: printing beats hand copying above 20.83 copies. Chapter 8 puts these into a year's ledger: the designed plan earns 3.8953 times the plain one. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. 摘要:本卷写给拥有大型数字专著库(这里是一百卷、附可运行代码)的作者与小出版者,研究三件事:全集捆绑能拿到多少价值;在每周上架限制与版税档位下怎样排程;以及直售、平台与开放获取怎样相互补充。书中的读者、需求、比率与参数均为示意模型;全部计算由书中程序生成;书中引用的平台规则于 2026 年 11 月 11 日核对;本书不是法律、财务或税务方面的建议。 第一章说明一百卷独立估值的书组成的捆绑拿到 0.6839 的价值,单卖是 0.3356,变异系数降到 0.2003;品味有中等共同因子时只有 0.4595,所以「几乎全部消费者剩余」说得太重。第二章说明在同一家店里没有混合菜单胜过全集,而平台上的单行本在没有重叠时让总收入增加到 1.4026 倍。第三章说明在每种格式每周两本的限制下,按需求上架赚到按卷号上架的 1.7491 倍,这条限制让第一年少了 0.2069;12.99 美元处的断崖把每笔版税从 8.8830 降到 4.5500。第四章说明合订本只让第一年版税改变几个百分点,而三条格式队列并行增加 10835。 第五章说明一套 299 美元的全集净赚 283.55,在 9.99 美元时等于 41.80 本电子书,在新上限 12.99 美元时只有 31.92 本;自建收银只有在每年超过 283.9 笔时才比记录商户划算。第六章说明开放获取在五分之一买家转读免费时,需要每次下载带来付费销售的机会超过 0.0060 才有利,引用带来的销售从第一年的 1701 增加到第五年的 16729。第七章读《宋史·邢昺传》记国子监书版「国初不及四千,今十余万」与丛书《百川学海》:超过 20.83 本时印刷胜过手抄。第八章把这些放进一年的账:设计过的方案赚到朴素方案的 3.8953 倍。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Law of Large Bundles — Chapter 1 shows a bundle of 100 independently valued volumes collecting 0.6839 of value against 0.3356 for singles, with the coefficient of variation falling to 0.2003, but only 0.4595 with a moderate common taste, so "nearly all consumer surplus" is too strong. (https://5dlibrary.com/answers/en/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c01/) 第1章 大数捆绑定律 — 第一章说明一百卷独立估值的书组成的捆绑拿到 0.6839 的价值,单卖是 0.3356,变异系数降到 0.2003;品味有中等共同因子时只有 0.4595,所以「几乎全部消费者剩余」说得太重。 (https://5dlibrary.com/answers/zh/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c01/) - Ch.2 Mixed Bundling and Where Singles Belong — Chapter 2 shows that within one shop no mixed menu beats the full catalogue, while singles on the platform raise total revenue by 1.4026 with no overlap. (https://5dlibrary.com/answers/en/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c02/) 第2章 混合捆绑与单行本该去哪里 — 第二章说明在同一家店里没有混合菜单胜过全集,而平台上的单行本在没有重叠时让总收入增加到 1.4026 倍。 (https://5dlibrary.com/answers/zh/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c02/) - Ch.3 Two Titles a Week and the Royalty Cliff — Chapter 3 shows that under a limit of two new titles per format a week, releasing by demand earns 1.7491 times releasing by volume number and the limit costs 0.2069 of the first year, and that the cliff at USD 12.99 cuts a sale's royalty from 8.8830 to 4.5500. (https://5dlibrary.com/answers/en/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c03/) 第3章 每周两本与版税断崖 — 第三章说明在每种格式每周两本的限制下,按需求上架赚到按卷号上架的 1.7491 倍,这条限制让第一年少了 0.2069;12.99 美元处的断崖把每笔版税从 8.8830 降到 4.5500。 (https://5dlibrary.com/answers/zh/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c03/) - Ch.4 Omnibus Packing and Three Queues — Chapter 4 shows omnibus editions changing first-year royalties by a few per cent, while three format queues in parallel add 10835. (https://5dlibrary.com/answers/en/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c04/) 第4章 合订本装箱与三条队列 — 第四章说明合订本只让第一年版税改变几个百分点,而三条格式队列并行增加 10835。 (https://5dlibrary.com/answers/zh/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c04/) - Ch.5 Direct Sales: Forty-One Ebooks and the Tax Bill — Chapter 5 shows a catalogue sold at USD 299 netting 283.55, which is 41.80 ebooks at 9.99 but only 31.92 at the new ceiling of 12.99, and an own checkout beating a merchant of record only above 283.9 sales a year. (https://5dlibrary.com/answers/en/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c05/) 第5章 直售:四十一本电子书与税单 — 第五章说明一套 299 美元的全集净赚 283.55,在 9.99 美元时等于 41.80 本电子书,在新上限 12.99 美元时只有 31.92 本;自建收银只有在每年超过 283.9 笔时才比记录商户划算。 (https://5dlibrary.com/answers/zh/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c05/) - Ch.6 Open Access: Substitute or Advertisement? — Chapter 6 shows that open access, if a fifth of buyers switch to free copies, pays only if each download leads to a paid sale with chance above 0.0060, and that citation sales grow from 1701 in year one to 16729 in year five. (https://5dlibrary.com/answers/en/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c06/) 第6章 开放获取:替代品还是广告? — 第六章说明开放获取在五分之一买家转读免费时,需要每次下载带来付费销售的机会超过 0.0060 才有利,引用带来的销售从第一年的 1701 增加到第五年的 16729。 (https://5dlibrary.com/answers/zh/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c06/) - Ch.7 State Blocks, a Hundred Works and the Economics of Printing — Chapter 7 reads the History of Song on the Directorate of Education's printing blocks and the collection Baichuan xuehai: printing beats hand copying above 20.83 copies. (https://5dlibrary.com/answers/en/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c07/) 第7章 监本、百川学海与印刷的经济学 — 第七章读《宋史·邢昺传》记国子监书版「国初不及四千,今十余万」与丛书《百川学海》:超过 20.83 本时印刷胜过手抄。 (https://5dlibrary.com/answers/zh/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c07/) - Ch.8 Practice: One Year of the Catalogue and One Test Protocol — Chapter 8 puts these into a year's ledger: the designed plan earns 3.8953 times the plain one. Marked boxes separate published evidence, runnable Python calculations (16 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c08/) 第8章 实务:书库的一年与一份检验方案 — 第八章把这些放进一年的账:设计过的方案赚到朴素方案的 3.8953 倍。书中以标注框区分已发表的证据、可运行的 Python 计算(16 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/monograph-catalog-mixed-bundling-platform-throttling-direct-sovereign-arbitrage/c08/) ## 5DL-B-116 · The Zi Wei Holographic Personality and Monetization Atlas 中文书名:紫微星宫全息人格与商业变现图谱 DOI: pending Cite: Fan, Y. (2026). The Zi Wei Holographic Personality and Monetization Atlas (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All calculations are generated by the programs printed in the book; it is not investment, career, psychological or marital advice. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. Abstract: This volume treats the placement rules of the fourteen main stars of Zi Wei Dou Shu as a fixed arrangement, computes the structure of the self-palace archetypes exactly by enumeration, and writes the traditional reading of every archetype as a sharp profile of personality and business: its drive, its built-in way of winning, its light face, its shadow face, how its money comes and where its money leaks. It makes no divinatory claim: it states the rules precisely, computes the structure exactly and shows how the readings themselves could be tested; the readings are specific so that they can be refuted. All calculations are generated by the programs printed in the book; it is not investment, career, psychological or marital advice. Chapter 1 proves that the twelve Zi Wei layouts produce only 38 self-palace archetypes (14 single stars and 24 pairs), and every pair is read as one archetype; the twelve layouts contain 24 empty palaces, all of which face pairs, only 12 pairs are ever borrowed, and the other 26 archetypes are borrowed by no chart at all. Over 43200 equally weighted birth combinations the self palace is empty in a share of 0.1656, and in 0.4550 of charts at least one of the self, wealth and career palaces is empty. The book distinguishes three kinds of drive: solid occupation is constant internal output, an empty palace borrowing across is resonance amplified by the outside wind, and projection from across is an outside mirror; every archetype is written in seven parts, and its leaks are written only as three kinds of checkable structure: contract wounds, paying for face and cash-flow holes. Chapter 2 covers the Zi Wei family: Zi Wei occupies the self palace in exactly one chart in twelve, a share of 0.0833, and its wealth palace is always led by Wu Qu and its career palace by Lian Zhen; Zi Wei + Tan Lang is the only one of the six that can be borrowed. Chapters 3 to 7 cover Tian Ji, Tai Yin and Tian Tong; Tai Yang, Ju Men and Tian Liang; the Wu Qu and Lian Zhen families; and the five stars that sit alone, Tian Fu, Tian Xiang, Qi Sha, Po Jun and Tan Lang. Chapter 8 demonstrates the look-up on a sample chart that contains no birth data and gives a pre-registered blind-matching protocol: with three choices and a true hit rate of 0.40, it needs 325 people. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. 摘要:本卷把紫微斗数十四主星的安星规则当作一个确定的排列,用穷举精确计算命宫原型的结构,再把每一种原型的传统读法写成一份锋利的人格与商业读本:驱动力、出厂赢法、阳之面、阴之面、搞钱赢法与漏财死穴。本书不主张任何占验:它精确写出规则、精确计算结构,并说明怎样检验读法本身;读法写得具体,是为了让它可以被证伪。全部计算由书中程序生成;本书不是投资、职业、心理或婚姻建议。 第一章证明十二种紫微排法只产生 38 种命宫原型(14 种独坐、24 种双星),双星一律作为一个原型来读;十二种排法共有 24 个空宫,全部正对双星,被借到的只有 12 组,其余 26 种原型没有任何一张盘会借到。在 43200 种等权的出生组合中,命宫为空的比例是 0.1656,命、财、官至少一宫为空的盘占 0.4550。全书区分三种驱动力:实星坐守是内生恒定输出,空宫借对宫是外部风口共振放大,对宫照入是外部之镜;每一种原型都写成七段,漏财只写合约暗伤、为面子付账与现金流窟窿三类可以核对的结构。第二章写紫微一族:紫微坐命恰好占十二分之一,比例 0.0833,其财帛宫必是武曲领头、官禄宫必是廉贞领头;紫微贪狼是六种之中唯一可借的一种。第三至第七章写天机、太阴、天同一组,太阳、巨门、天梁一组,武曲与廉贞两族,以及天府、天相、七杀、破军、贪狼五颗独坐的星。第八章用一张不含任何出生资料的样盘示范查法,并给出一份预先登记的盲配检验方案:三选一、真实命中率 0.40 时需要 325 人。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 Chapter answers: - Ch.1 Method: Thirty-Eight Archetypes, Solid Stars, Borrowed Stars and the Opposite Palace — Chapter 1 proves that the twelve Zi Wei layouts produce only 38 self-palace archetypes (14 single stars and 24 pairs), and every pair is read as one archetype; the twelve layouts contain 24 empty palaces, all of which face pairs, only 12 pairs are ever borrowed, and the other 26 archetypes are borrowed by no chart at all. Over 43200 equally weighted birth combinations the self palace is empty in a share of 0.1656, and in 0.4550 of charts at least one of the self, wealth and career palaces is empty. The book distinguishes three kinds of drive: solid occupation is constant internal output, an empty palace borrowing across is resonance amplified by the outside wind, and projection from across is an outside mirror; every archetype is written in seven parts, and its leaks are written only as three kinds of checkable structure: contract wounds, paying for face and cash-flow holes. (https://5dlibrary.com/answers/en/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c01/) 第1章 方法:三十八种原型、实星、借星与对宫 — 第一章证明十二种紫微排法只产生 38 种命宫原型(14 种独坐、24 种双星),双星一律作为一个原型来读;十二种排法共有 24 个空宫,全部正对双星,被借到的只有 12 组,其余 26 种原型没有任何一张盘会借到。在 43200 种等权的出生组合中,命宫为空的比例是 0.1656,命、财、官至少一宫为空的盘占 0.4550。全书区分三种驱动力:实星坐守是内生恒定输出,空宫借对宫是外部风口共振放大,对宫照入是外部之镜;每一种原型都写成七段,漏财只写合约暗伤、为面子付账与现金流窟窿三类可以核对的结构。 (https://5dlibrary.com/answers/zh/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c01/) - Ch.2 The Zi Wei Family: Six Kinds of Sovereign — Chapter 2 covers the Zi Wei family: Zi Wei occupies the self palace in exactly one chart in twelve, a share of 0.0833, and its wealth palace is always led by Wu Qu and its career palace by Lian Zhen; Zi Wei + Tan Lang is the only one of the six that can be borrowed. (https://5dlibrary.com/answers/en/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c02/) 第2章 紫微一族:六种坐镇者 — 第二章写紫微一族:紫微坐命恰好占十二分之一,比例 0.0833,其财帛宫必是武曲领头、官禄宫必是廉贞领头;紫微贪狼是六种之中唯一可借的一种。第三至第七章写天机、太阴、天同一组,太阳、巨门、天梁一组,武曲与廉贞两族,以及天府、天相、七杀、破军、贪狼五颗独坐的星。 (https://5dlibrary.com/answers/zh/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c02/) - Ch.3 Ji Yue Tong Liang: Nine Personalities of Planning, Care and Harmony — Ji Yue Tong Liang is a family that lives by mind, patience and goodwill: Tian Ji plans, Tai Yin and Tian Liang take care, Tian Tong harmonises, and Ju Men adds doubt and eloquence in two of the pairs. The family holds 30 of the 120 palaces with main stars, yet accounts for 12 of the 24 empty palaces; exactly half of empty self palaces (0.5000) borrow one of its pairs. The three single stars are never borrowed and hold self-palace shares of 0.0411 to 0.0428; each of the six pairs is half solid and half borrowed, with shares between 0.0128 and 0.0144. Which pair the wealth palace borrows fixes the self palace uniquely: whoever borrows Tian Tong with Tian Liang has Ju Men in the self palace, which is sample chart A. The solid reading is an internal constant output; the borrowed reading is ignition by the outside wind, explosion inside the window and leakage in the flat periods, and the latter is the falsifiable prediction this chapter stakes. (https://5dlibrary.com/answers/en/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c03/) 第3章 机月同梁:谋划、照料与调和的九种人格 — 机月同梁是一族靠脑子、耐心和人缘吃饭的人:天机谋划,太阴与天梁照料,天同调和,巨门在两对里添上怀疑与口才。这一族在 120 个有主星的宫里占 30 个,却承包了 24 个空宫中的 12 个;空命宫里恰好一半(0.5000)借的是本族对星。三颗单星从不被借,实坐命宫的份额是 0.0411 至 0.0428;六种对星实与借各占一半,份额在 0.0128 到 0.0144 之间。财帛宫借哪一对,命宫就被唯一确定:借天同天梁的人,命宫必坐巨门,正是样盘 A。实坐者的读法是内生的恒定输出,借入者的读法是被外部风口点燃、在窗口里爆发、在平淡期漏财;后者是本章押上的可证伪预测。 (https://5dlibrary.com/answers/zh/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c03/) - Ch.4 Sun, Gate and Beam: Six Archetypes of Shining, Doubting and Covering — The Sun, the Great Gate and the Roof Beam lock into one another across the twelve layouts: the three singles appear 6 times each and always face another single star, never borrowed; the three Sun pairs appear 2 times each and always face an empty palace. Across 43200 equally weighted cases the self palace is empty in 0.1656, and the family fills 0.2094 of self palaces; Tian Liang alone is the commonest (0.0428), Tai Yang with Tian Liang the rarest (0.0128). As readings: the Sun's money comes from being seen and leaks through face; the Sun with the Moon earns on two tracks and leaks when the two ledgers patch each other; the Sun with the Great Gate sells the right to explain and leaks through unlimited promises made in public; the Sun with the Roof Beam collects rent on credibility and leaks through personal guarantees; the Great Gate earns by explaining what others cannot and leaks through one public sentence and the clause it never argued; the Roof Beam is paid for being there when it happens and leaks through rescue loans without a note. In sample chart A the Great Gate in Wu forms a closed loop of transformations with the career palace. These are interpretations of rules; the tests are in the hypothesis box above and in Chapter 8. (https://5dlibrary.com/answers/en/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c04/) 第4章 日、巨、梁:发光、质疑与兜底的六种原型 — 太阳、巨门、天梁三颗星在十二种布局中互相咬合:三种单星各出现 6 次,对面永远是另一颗单星,从不被借;三个太阳双星组合各出现 2 次,对面永远是空宫。在 43200 种等权情形中,命宫空的比例是 0.1656,本族合计占命宫的 0.2094;天梁单星最常见(0.0428),太阳天梁最少见(0.0128)。读法上,太阳的钱来自被看见,漏在面子;太阳太阴靠两班倒的双轨,漏在两本账互补;太阳巨门卖解释权,漏在当众说出的无限承诺;太阳天梁收信誉的租金,漏在个人担保;巨门靠解释别人解释不了的东西赚钱,漏在一句当众的话和那条没争的条款;天梁收「出事时你在」的钱,漏在不立字据的救急。样盘 A 的巨门坐午,与官禄之间形成一个四化闭环。这些是规则的诠释,检验方法见上面的假说框与第八章。 (https://5dlibrary.com/answers/zh/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c04/) - Ch.5 Wu Qu: One Money Star and Its Six Partners — Wu Qu stands four branches behind Zi Wei; in only 2 of the 12 layouts does it sit alone, and in the other 10 it has a partner. Its five pairs face one another two by two, and only Wu Qu with Tan Lang faces an empty palace, the one archetype of the family that gets borrowed. Across 43200 equally weighted cases the six Wu Qu archetypes together fill 0.0972 of self palaces and 0.0972 of wealth palaces; Wu Qu with Tian Xiang is the rarest (0.0128). As readings: Wu Qu alone earns through the compounding of discipline and leaks through having no reserve and skipping professional fees; with Tian Fu it manages a pool of money and leaks by locking money away for safety; with Tan Lang it hauls in the net when the wind blows and leaks by treating the window as normal; with Tian Xiang it manages other people's money and leaks through vouching and nominee holding; with Qi Sha it collects an assault premium and leaks through unlimited personal guarantees; with Po Jun it earns the restructuring spread and leaks by clearing out the last round every time. p27's illustrative model shows that lumpy income spent at the peak rate, plus payment terms that someone else holds, can turn the same income into a cash gap three times as deep. These are interpretations of rules; the tests are in the hypothesis box above and in Chapter 8. (https://5dlibrary.com/answers/en/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c05/) 第5章 武曲:一颗财星和它的六种搭档 — 武曲站在紫微后面第四个地支上,在 12 种布局里只有 2 种是单独坐守,其余 10 种都有搭档;五个双星组合两两对望,只有武曲贪狼对着空宫,是这一族唯一会被借走的原型。在 43200 种等权情形中,六种武曲原型合计占命宫的 0.0972,财帛宫也是 0.0972;武曲天相最少见(0.0128)。读法上,武曲单星靠纪律的复利赚钱,漏在没有储备和省掉专业费用;武曲天府管资金池,漏在为了安全把钱锁死;武曲贪狼在风口里收网,漏在把窗口期当常态;武曲天相替人管钱,漏在替人背书与代持;武曲七杀收攻坚溢价,漏在个人无限担保;武曲破军赚重组的差价,漏在每次重来都清空上一轮。p27 的示意模型说明,块状收入按峰值花、再加上被人拿走的账期,同样的收入可以变成三倍深的现金缺口。这些是规则的诠释,检验方法见上面的假说框与第八章。 (https://5dlibrary.com/answers/zh/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c05/) - Ch.6 The Six Lian Zhen Types: Rules Locked Around a Vault, a Desire, a Seal, a Blade and a Hammer — Lian Zhen is a "prisoner" star: it does not decide what is locked up; the star in the same palace does. In all twelve layouts Zi Wei, Lian Zhen and Wu Qu stand on one triangle (12 of 12), so whenever Lian Zhen sits in the self palace as a real star, the wealth palace is Zi Wei and the career palace is Wu Qu (3600 cases). The self-palace shares of the six types are: alone 0.0144, with Tian Xiang 0.0144, with Tan Lang 0.0139, with Qi Sha 0.0139, with Po Jun 0.0139, with Tian Fu 0.0128, together 0.0833; only Lian Zhen with Tan Lang can be borrowed by an empty palace (0.0139), and then Zi Wei lands in the fortune palace and Wu Qu in the spouse palace. The switch for Lian Zhen's obstacle is the stem Bing: among people with Lian Zhen in the self palace, the share with a self-transforming ji is 0.1067. In reading, the six types share one fatal leak, and it is written in the contract: verbal understandings, personal guarantees, valuation bets, signatures for others, exits without settlement. The master of rules most often loses on the one rule they never wrote down. (https://5dlibrary.com/answers/en/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c06/) 第6章 廉贞六型:把规矩装进库、欲望、印、刀与锤 — 廉贞是一颗「囚」星:它不决定关住什么,同宫的星决定。十二种布局里,紫微、廉贞、武曲永远站在同一个三角上(12 of 12),所以廉贞实坐命宫时财帛必是紫微、官禄必是武曲(3600 个案例)。六个原型的命宫份额是:独坐 0.0144、天相 0.0144、贪狼 0.0139、七杀 0.0139、破军 0.0139、天府 0.0128,合计 0.0833;只有廉贞贪狼会被空宫借走(0.0139),借入时紫微落福德、武曲落夫妻。廉贞化忌的开关在丙干:廉贞坐命者中,命宫自化忌的比例是 0.1067。读法上,六型的共同死穴都写在合同里:口头默契、个人担保、对赌、代签、无清算的退出——规则高手最常输在自己没写下来的那一条规则上。 (https://5dlibrary.com/answers/zh/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c06/) - Ch.7 The Vault, the Seal and the Three Movers: Five Lone Stars and Their Always-Solid Opposites — Tian Fu, Tian Xiang, Qi Sha, Po Jun and Tan Lang each stand alone in six layouts: Fu and Xiang only on yin branches, Sha, Po and Lang only on yang branches. Qi Sha always faces Tian Fu, Po Jun always faces Tian Xiang, and a lone Tan Lang always faces a lone Zi Wei, Wu Qu or Lian Zhen, so the palace opposite these five lone stars is never empty and they are never borrowed (borrowed share 0.0000). Each stands alone in the self palace in a share of 0.0417, together 0.2083. The three movers always stand on one triangle (12 of 12); with one of them in the self palace, the wealth and career palaces hold the other two (3600 cases each), together a share of 0.2500 of all cases; with Tian Fu in the self palace the career palace is Tian Xiang, and with Tian Xiang in the self palace the wealth palace is Tian Fu. In reading, the fatal leaks of Fu-Xiang lie in "covering for others" and "a split never written down"; those of Sha-Po-Lang lie in "the next round", "revenue before cash" and "paying for face to stay at the centre". Those who hold leak through relationships; those who change leak through rhythm. (https://5dlibrary.com/answers/en/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c07/) 第7章 府相与杀破狼:五颗独坐的星和它们永远实心的对面 — 天府、天相、七杀、破军、贪狼各在六种布局里独坐:府相只落阴支,杀破狼只落阳支。七杀永远对着天府,破军永远对着天相,独坐贪狼永远对着独坐的紫微、武曲或廉贞,所以这五颗独坐星的对面从不是空宫,也从不被借(被借份额 0.0000)。每一颗独坐命宫的份额都是 0.0417,合计 0.2083。杀破狼三颗星永远在同一个三角上(12 of 12),命宫有其中一颗,财帛与官禄就是另外两颗(各 3600 个案例),合计占全部案例的 0.2500;命宫有天府,官禄是天相,命宫有天相,财帛是天府。读法上,府相的死穴在「替别人托底」和「没写下来的分配权」,杀破狼的死穴在「追加投入」「收入先于现金」和「为留在中心而付的面子钱」——守的人漏在关系里,变的人漏在节奏里。 (https://5dlibrary.com/answers/zh/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c07/) - Ch.8 Practice: Reading a Sample Chart, and a Blind-Matching Protocol — Chapter 8 demonstrates the look-up on a sample chart that contains no birth data and gives a pre-registered blind-matching protocol: with three choices and a true hit rate of 0.40, it needs 325 people. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c08/) 第8章 实践:读一张样盘,以及一份盲配检验方案 — 第八章用一张不含任何出生资料的样盘示范查法,并给出一份预先登记的盲配检验方案:三选一、真实命中率 0.40 时需要 325 人。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 (https://5dlibrary.com/answers/zh/ziwei-38-archetypes-solid-borrowed-drive-yang-yin-faces-monetization-leaks/c08/) ## 5DL-B-117 · The Qintian Three Knives and the Flying-Ji Codex of Psychological Turning Points 中文书名:钦天三刀与十二宫飞忌心理转折法典 DOI: pending Cite: Fan, Y. (2026). The Qintian Three Knives and the Flying-Ji Codex of Psychological Turning Points (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. Abstract: This volume takes the obstacle-transformation (ji) rules of the Taiwanese Qintian line of Zi Wei Dou Shu as a definite set of rules, states them completely and sharply, and enumerates their structure to the end. The Qintian reading fits in one line: ji is clinging, owing, retracting and hiding; the palace a ji enters shows the cause, the palace it strikes takes the damage, and a self-transforming ji comes fast and leaves fast. On this basis the author's reading engine takes "three knives" from every chart: the birth-year ji as the source of a lifelong fixation, the ji flown by the Self palace's stem as the knot that bites back the harder you push, and the self-transforming ji as the switch by which a person goes cold and pulls away at the breaking point of a setback. The book makes no divinatory claim: it presents the rules precisely, computes their structure and shows how their readings could be tested; all calculations are generated by its programs; it is not psychotherapeutic, marital, investment or medical advice. Chapter 1 enumerates 43200 equally weighted cases: the birth-year ji falls in the twelve palaces with shares from 0.0797 to 0.0867, almost uniform; 0.2777 of charts have no self-transforming ji, so the third knife falls back to the Yang-Tuo pincer; in 0.1567 the first two knives stack in one palace, 0.1000 of it only because the Self palace's stem equals the year stem; all 144 flying-ji trajectories are reachable, and the diagonal is the self-transforming ji. Chapter 2 reads the birth-year ji palace by palace: where the fixation began, how it shows at work, in money and in intimacy, the bill overpaid and the exit, and shows that for a Xin-year birth the place of the fixation is decided by the birth month alone. Chapter 3 reads the Self palace's flying ji into the twelve palaces, Chapter 4 the self-transforming ji in each palace, and Chapters 5 to 7 the 144 flying-ji trajectories cell by cell, each with structure, psychological subtext, modern scene, and backfire and exit. Chapter 8 reads the three knives of a sample chart and the cold-war structure of living under one roof, and gives a pre-registered test: a blind three-way choice needs 318 people to detect a hit rate of 0.40, and testing the cold-war prediction by behaviour coding needs 1152 recruited couples. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. 摘要:本卷把台湾钦天一系的紫微斗数化忌规则当作一套确定的规则,完整、锋利地写出来,并用穷举把它的结构算到底。钦天的读法压成一句话:忌是执、欠、收、藏,忌所入之宫说明原因,忌所冲之宫承受伤害,自化忌来得快、去得快。作者的测算引擎据此从每张盘取出「三把刀」:生年忌是一生执念的源头,命宫宫干飞忌是越用力越遭反噬的死结,自化忌是受挫临界点瞬间心冷抽离的开关。本书不主张任何占验,只精确呈现规则、计算其结构,并说明怎样检验其解读;全部计算由书中程序生成;本书不是心理治疗、婚姻、投资或医疗建议。 第一章在 43200 种等权情形上穷举:生年忌落在十二宫的比例从 0.0797 到 0.0867,几乎均匀;0.2777 的盘没有任何自化忌,第三刀回退到羊陀夹;0.1567 的盘前两把刀叠在同一宫,其中 0.1000 只因命宫宫干等于年干;一百四十四条飞忌轨迹全部可达,对角线就是自化忌。第二章逐宫解读生年忌落十二宫:执念源头,它在工作、钱与亲密关系里的样子,一直多付的账与出口,并说明辛年生人的执念落处只由出生月决定。第三章解读命宫飞忌入十二宫,第四章解读十二宫自化忌,第五至第七章逐格解读一百四十四条飞忌轨迹,每格写结构、心理潜台词、现代场景、反噬与出口。第八章读一张样盘的三把刀与同一屋檐下的冷战结构,并给出预先登记的检验:三选一盲选要看出 0.40 的命中率需要 318 人,以行为编码检验冷战预测需要招募 1152 对伴侣。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 Chapter answers: - Ch.1 Method: Qintian Transformations, Three Knives and 144 Cells — Chapter 1 enumerates 43200 equally weighted cases: the birth-year ji falls in the twelve palaces with shares from 0.0797 to 0.0867, almost uniform; 0.2777 of charts have no self-transforming ji, so the third knife falls back to the Yang-Tuo pincer; in 0.1567 the first two knives stack in one palace, 0.1000 of it only because the Self palace's stem equals the year stem; all 144 flying-ji trajectories are reachable, and the diagonal is the self-transforming ji. (https://5dlibrary.com/answers/en/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c01/) 第1章 方法:钦天四化、三把刀与一百四十四格 — 第一章在 43200 种等权情形上穷举:生年忌落在十二宫的比例从 0.0797 到 0.0867,几乎均匀;0.2777 的盘没有任何自化忌,第三刀回退到羊陀夹;0.1567 的盘前两把刀叠在同一宫,其中 0.1000 只因命宫宫干等于年干;一百四十四条飞忌轨迹全部可达,对角线就是自化忌。 (https://5dlibrary.com/answers/zh/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c01/) - Ch.2 The Birth-Year Ji in the Twelve Palaces: Where a Lifelong Fixation Begins — The birth-year ji falls almost evenly over the twelve palaces, from 0.0797 in Health to 0.0867 in Self and Siblings; every palace can be reached by all ten ji stars, so the information is in the palace, not the star. In 0.1567 of charts the Self palace's ji strikes into the same palace, and in 0.1638 the place of fixation carries its own self-transforming switch. For a Xin-year birth, the location of the fixation is decided by the birth month alone. The twelve sections share one structure: the palace the ji enters is the account you spend a lifetime collecting or repaying, the palace it strikes is the domain that takes the blow for it, and the exit is always an act of letting go that can be written down and checked. (https://5dlibrary.com/answers/en/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c02/) 第2章 生年忌落十二宫:一生执念的源头 — 生年忌几乎均匀地落在十二宫,从疾厄的 0.0797 到命宫与兄弟的 0.0867;每一宫都能被十颗忌星到达,信息在宫不在星。0.1567 的盘命宫飞忌也打进同一宫,0.1638 的盘执念之处自带自化开关。辛年生人的执念落处只由出生月决定。十二节的共同结构是:忌入之宫是你一生在讨或在还的那笔账,被冲之宫是替这笔账挨打的领域,出口永远是一个可以写下来、可以检查的松手动作。 (https://5dlibrary.com/answers/zh/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c02/) - Ch.3 The Self Palace's Flying Ji: Twelve Knots Where Effort Bites Back — The Self palace's flying Ji is your own hand: the palace it enters is where your force goes, the palace it strikes is where the bill is sent. Across 43200 cases the twelve landing places share almost equally (8.22% to 8.50%), so no knot is one that everyone has. The ten Ji stars carry a tenth each; Wen Qu and Wen Chang, placed by the hour, land in an empty palace about a sixth of the time. The boomerang is uneven: force thrown into Wealth or Career most often comes back at you (35.91% and 32.35%), while force thrown into Spouse or Mind comes back least but most easily piles up where it fell as a debt no one says aloud. The twelve exits share one skeleton: write the standard before you start, hand the relationship to terms, and replace rumination with time-limited action. (https://5dlibrary.com/answers/en/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c03/) 第3章 命宫飞忌:越用力越遭反噬的十二个死结 — 命宫飞忌是你自己的手:入哪一宫,力气就往哪里扔;冲哪一宫,账单就寄往哪里。在 43200 种情形中,十二个落点几乎均分(8.22% 到 8.50%),没有一种死结是人人都有的。十颗忌星各占一成;文曲、文昌按时辰排布,约六分之一落进空宫。回旋镖并不均匀:扔进财帛与官禄的力最容易弹回自己(35.91% 与 32.35%),扔进夫妻与福德的力最少回弹,却最容易在原地积成说不出口的账。十二个出口有同一副骨架:把标准写在事前,把关系交给条款,把反刍换成限时的行动。 (https://5dlibrary.com/answers/zh/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c03/) - Ch.4 Self-Transforming Ji: Breaking Points and Twelve Cold Switches — A self-transforming Ji is a switch, not a debt: investment comes fast, withdrawal comes fast, and between them lies an invisible breaking point. Across 43200 cases, 72.23% of charts carry at least one switch and 49.28% exactly one; the probability that any given palace carries one lies between 8.28% and 8.61%. Palaces on yang branches can be switched only by Tai Yang, Lian Zhen, Tian Ji, Tian Tong or Wu Qu, palaces on yin branches only by Tai Yin, Ju Men, Wen Qu, Wen Chang or Tan Lang. The twelve switches trip in twelve different ways, yet the rule for keeping them safe is one: write the exit conditions while calm, give separation a procedure, and say it aloud before the power is cut. (https://5dlibrary.com/answers/en/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c04/) 第4章 自化忌:受挫临界点与十二个心冷开关 — 自化忌是一个开关,而不是一笔债:投入得快,撤出得也快,中间隔着一个看不见的临界点。在 43200 种情形中,72.23% 的盘至少装着一个开关,49.28% 恰好一个;每一宫装开关的概率都在 8.28% 到 8.61% 之间。阳支上的宫只能由太阳、廉贞、天机、天同、武曲扳动,阴支上的宫只能由太阴、巨门、文曲、文昌、贪狼扳动。十二个开关跳闸的方式各不相同,防误触的原则却一致:在平静时写下退出条件,让分开有程序可走,在断电之前先说出口。 (https://5dlibrary.com/answers/zh/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c04/) - Ch.5 Flight Trajectories (I): The Rows of Self, Siblings, Spouse and Children — This chapter has written out the first four rows of the matrix, forty-eight cells. Every cell is read in three steps: the motive of the source palace (who is pushing), the debt in the target palace (where the force went), the damage on the opposite palace (who pays the bill). Exact enumeration of 43200 cases shows that every cell of the four rows has a frequency between 0.0794 and 0.0861, close to the uniform 0.0833: the Self row most often enters Siblings and Friends, the Spouse row most often enters Self, the Children row most often enters Mind; the least frequent are Children into Wealth and Children into Health. On one chart, at least two of the four arrows sharing a palace is the norm (0.5363), while mutual ji (0.0161) and four arrows in one (0.0096) are rare. These readings are interpretations; the chapter has given two tests that can be pre-registered, with the sample sizes they need. (https://5dlibrary.com/answers/en/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c05/) 第5章 飞忌轨迹(上):命宫、兄弟、夫妻、子女四行 — 本章写完了矩阵的前四行四十八格。读每一格分三步:来源宫的动机(谁在用力),目标宫的债(力用在哪里),对宫的伤(谁在付账)。精确枚举 43200 个情形表明,四行中每一格的频率都在 0.0794 到 0.0861 之间,接近均匀的 0.0833:命宫行最多是入兄弟与入交友,夫妻行最多是入命宫,子女行最多是入福德;最少的是子女入财帛与子女入疾厄。一张盘上四支箭至少两支同宫是常态(0.5363),互忌(0.0161)与四箭归一(0.0096)才少见。这些读法是诠释;本章给出了两项可以预先登记的检验与所需样本量。 (https://5dlibrary.com/answers/zh/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c05/) - Ch.6 Flights II: The Forty-Eight Trajectories of Wealth, Health, Travel and Friends — Wealth, Health, Travel and Friends: four rows, forty-eight trajectories. Exact enumeration of 43200 charts shows that the share of every trajectory lies between 0.0800 and 0.0872; there are no rare configurations, and the weight of each reading lies entirely in its structure. Where money's ji lands decides who pays for your pricing, payment terms and guarantees; where the body's ji lands decides where burnout is settled; where the outside's ji lands decides the distance between a fine figure outside and a collapse at home; where the many's ji lands decides what clients, colleagues and followers take from you. Read together, more than half of all charts (0.5390) have at least two rows pledging their debts to the same place, and in a quarter of all charts (0.2500) this is forced by a repeated stem. (https://5dlibrary.com/answers/en/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c06/) 第6章 飞忌之二:财帛、疾厄、迁移、交友的四十八条轨迹 — 财帛、疾厄、迁移、交友四行,共四十八条轨迹。对 43200 张盘的精确枚举显示,任何一条轨迹的比例都在 0.0800 至 0.0872 之间,没有稀有的格局,读法的分量全在结构。钱之忌的落点决定谁替你的定价、账期与担保付账;身体之忌决定倦怠在哪里结算;外面之忌决定在外风光与回家崩塌之间的距离;众人之忌决定客户、同事与追随者从你这里拿走什么。四行合看,超过一半的盘(0.5390)至少有两行把债押在同一处,其中四分之一的盘(0.2500)是宫干重复所强制。 (https://5dlibrary.com/answers/zh/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c06/) - Ch.7 Flight Trajectories III: Forty-Eight Paths from Career, Property, Mind and Parents — The forty-eight trajectories of the Career, Property, Mind and Parents rows are spread almost evenly over 43200 equally weighted cases: the highest, Parents into Self, takes 0.0861, the lowest, Parents into Children, 0.0800, and every cell can be reached by any of the ten ji stars. The self-transforming cells of the four rows take 0.0839, 0.0828, 0.0850 and 0.0839. The weight of reading therefore lies not in rarity but in structure: the palace entered shows where the fixation strikes from and where it is owed, the palace struck shows the wound; a self-transforming cell is a switch that snaps at one touch, and a boomerang cell is a force that loops round and wounds its own source. The Career row writes how work runs up debts, the Property row writes territory, the closed door and the silent ledger, the Mind row writes how attention sticks and ruminates, and the Parents row writes how rules, documents and face leak into life. The reading of Property on the Spouse-Career axis has been written as a pre-registered test that can fail. (https://5dlibrary.com/answers/en/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c07/) 第7章 飞忌轨迹(三):官禄、田宅、福德、父母出发的四十八条 — 官禄、田宅、福德、父母四行共四十八条轨迹,在 43200 个等权个案中几乎均匀分布:最高的父母入命宫占 0.0861,最低的父母入子女占 0.0800,每一格都能由十颗忌星中的任何一颗走到。四行的自化格分别占 0.0839、0.0828、0.0850、0.0839。读法的重心因此不在稀有度,而在结构:入的宫看执念从哪里出手、欠在哪里,冲的宫看伤口;自化格是一按就断的开关,回旋格是绕一圈伤回自己的力量。官禄一行写工作如何欠账,田宅一行写领地、关上的门与无声的账,福德一行写注意力如何黏住与反刍,父母一行写规则、文书与面子如何漏进生活。田宅落在夫妻—官禄轴上的读法,已经写成一项可以失败的预先登记检验。 (https://5dlibrary.com/answers/zh/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c07/) - Ch.8 Practice: Reading the Three Knives of Sample Chart A, and a Pre-Registered Test — Chapter 8 reads the three knives of a sample chart and the cold-war structure of living under one roof, and gives a pre-registered test: a blind three-way choice needs 318 people to detect a hit rate of 0.40, and testing the cold-war prediction by behaviour coding needs 1152 recruited couples. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c08/) 第8章 实践:读样盘 A 的三把刀,以及一份预先登记的检验 — 第八章读一张样盘的三把刀与同一屋檐下的冷战结构,并给出预先登记的检验:三选一盲选要看出 0.40 的命中率需要 318 人,以行为编码检验冷战预测需要招募 1152 对伴侣。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 (https://5dlibrary.com/answers/zh/qintian-three-knives-flying-ji-144-trajectories-psychological-turning-points/c08/) ## 5DL-B-118 · Vedic Nakshatra Shadows and East-West Somatic Mark Decoding 中文书名:吠陀星宿潜意识与东西合参体相印记解码 DOI: pending Cite: Fan, Y. (2026). Vedic Nakshatra Shadows and East-West Somatic Mark Decoding (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. Abstract: Starting from two traditions, the twenty-seven nakshatras and drekkana body zones of Vedic astrology and the twelve branch body regions of Zi Wei Dou Shu with its malefic and romance stars, this volume builds a verification table of birthmarks, moles and scars precise to zone, side, form and age of onset, and shows how to test it. It does not claim that the table's accuracy is established: it writes each tradition's rules precisely and unambiguously, computes the chance hit rate of every claim, and gives pre-registered test protocols; dermatological base rates are sourced, other parameters are illustrative, all calculations are generated by its programs, and the book is not medical advice. Chapter 1 computes base rates: an illustrative adult carries 24.66 countable marks on average, "you have a mark on your arm" hits by chance with probability 0.991, and only narrowing to "a scar on the outer right upper arm from before age ten" brings this down to 0.030, a hit worth 5.07 bits; branch regions differ widely, Wu at 0.252 and Xu at 0.973; loosening after the fact turns a 0.047 claim into a 0.585 one. The chapter also gives the likelihood-ratio scoring rule used throughout. Chapter 2 writes out each nakshatra's deity, lord, symbol, subconscious shadow, trigger and mark hint: the Nakshatra-purusha lays the twenty-seven on 24 body parts, with base rates of 0.994 for the thighs and 0.024 for the nails, and a birth time off by an hour gives a 0.0824 chance of the wrong nakshatra. Chapter 3 writes the thirty-six drekkana body zones; Chapter 4 refines the twelve branch regions for the self, health and parents palaces; Chapters 5 and 6 write the marks of seven malefic and five romance stars in the twelve branches, with the obstacle-transformation overlay; Chapter 7 joins the two traditions into one table in which zones confirm or veto each other. Chapter 8 works through sample chart A: the chance that at least one of three strict claims hits is 0.333, and the card's wording widens 0.141 to 0.266; a group comparison needs 188 people in all if the risk doubles and 672 if it rises by half, and a one-in-three design needs 133 people at a pick rate of 0.45. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. 摘要:本卷从两套传统出发——吠陀占星的二十七宿与三分盘(drekkana)体表分区,紫微斗数的十二地支部位与刑煞、桃花星曜——编出一张精确到部位、左右、形态与出现年龄的胎记、痣、疤核验表,并说明怎样检验它。本书不主张这张表的准确性已经成立:它把各家规则写得精确而不含糊,计算每条判断碰巧命中的概率,并给出预先登记的检验方案;皮肤学基准率注明来源,其余参数为示意,全部计算由书中程序生成;本书不是医疗建议。 第一章计算基准率:示意的成年人平均有 24.66 个可数的印记,「你手臂上有个印记」碰巧命中 0.991,收窄到「右上臂外侧、十岁前的疤」才降到 0.030,命中值 5.07 比特;地支部位宽窄悬殊,午位 0.252,戌位 0.973;允许事后放宽,会把 0.047 的判断变成 0.585。本章并给出全书共用的似然比计分规则。第二章逐一写出二十七宿的神祇、宿主、象征、潜意识阴影、触发与印记线索:星宿人身图把二十七宿放在 24 个部位上,大腿的基准率 0.994,指甲 0.024;出生时间差一小时,就有 0.0824 的概率把宿算错。第三章写三分盘的三十六个体表分区;第四章细分十二地支部位在命宫、疾厄宫、父母宫的读法;第五章写七颗煞星、第六章写五颗桃花星在十二地支的印记与化忌叠加;第七章把东西两套体系合成一张互相印证、互相否决的核验表。第八章用样盘 A 演示:三条严格判断至少中一条的碰巧概率是 0.333,卡片的措辞把 0.141 放宽到 0.266;分组检验在风险比加倍时共需 188 人、增加一半时 672 人,三选一设计在选中率 0.45 时需 133 人。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 Chapter answers: - Ch.1 Base Rates: Why "You Have a Mark on Your Arm" Is True of Almost Everyone — Chapter 1 computes base rates: an illustrative adult carries 24.66 countable marks on average, "you have a mark on your arm" hits by chance with probability 0.991, and only narrowing to "a scar on the outer right upper arm from before age ten" brings this down to 0.030, a hit worth 5.07 bits; branch regions differ widely, Wu at 0.252 and Xu at 0.973; loosening after the fact turns a 0.047 claim into a 0.585 one. The chapter also gives the likelihood-ratio scoring rule used throughout. (https://5dlibrary.com/answers/en/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c01/) 第1章 基准率:为什么「你手臂上有个印记」几乎对每个人都成立 — 第一章计算基准率:示意的成年人平均有 24.66 个可数的印记,「你手臂上有个印记」碰巧命中 0.991,收窄到「右上臂外侧、十岁前的疤」才降到 0.030,命中值 5.07 比特;地支部位宽窄悬殊,午位 0.252,戌位 0.973;允许事后放宽,会把 0.047 的判断变成 0.585。本章并给出全书共用的似然比计分规则。 (https://5dlibrary.com/answers/zh/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c01/) - Ch.2 The Twenty-Seven Nakshatras: Subconscious Shadows and the Nakshatra-Purusha — Chapter 2 writes out each nakshatra's deity, lord, symbol, subconscious shadow, trigger and mark hint: the Nakshatra-purusha lays the twenty-seven on 24 body parts, with base rates of 0.994 for the thighs and 0.024 for the nails, and a birth time off by an hour gives a 0.0824 chance of the wrong nakshatra. (https://5dlibrary.com/answers/en/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c02/) 第2章 二十七宿:潜意识阴影与星宿人身图 — 第二章逐一写出二十七宿的神祇、宿主、象征、潜意识阴影、触发与印记线索:星宿人身图把二十七宿放在 24 个部位上,大腿的基准率 0.994,指甲 0.024;出生时间差一小时,就有 0.0824 的概率把宿算错。 (https://5dlibrary.com/answers/zh/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c02/) - Ch.3 Thirty-Six Body Zones: The Drekkana Map of the Body — The drekkana cuts each sign into three parts of ten degrees, and the rising part lays the twelve houses on the head and face, on neck to navel, or on below the navel to the feet; malefics read as scars and benefics as moles (brihatjataka), and since translations disagree about left and right, side must be a parameter of any test. p10 shows that 0.200 of ascendants lie within one degree of a boundary; p11 shows that equal and whole-sign houses can move 4 of 7 planets of one chart; p12 shows that the prior for each third of the body is 0.33333 at any latitude. The thirty-six zones differ enormously in size, the largest 56.7 times the smallest; the scar zones a chart names cover 0.0172 of the skin with the first drekkana rising and 0.0907 with the third. Hits in the face zones are evidence with weight; hits in the large zones of the legs say almost nothing. (https://5dlibrary.com/answers/en/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c03/) 第3章 三十六个体表分区:德雷卡纳的身体地图 — 德雷卡纳把每个星座切成三段十度,上升段决定十二宫映射到头面、颈到脐、或脐下到足;凶星读作疤,吉星读作痣 (brihatjataka),左右读法有译本分歧,必须在检验中当作参数。p10 显示 0.200 的上升点离段界不到一度;p11 显示等宫与整宫在一张盘上可以让 4 of 7 颗行星换宫;p12 显示三截身体的先验概率在任何纬度都是 0.33333。三十六区面积悬殊,最大是最小的 56.7 倍;一张盘点名的疤区,第一段只占皮肤的 0.0172,第三段占 0.0907。所以头面区的命中才是有分量的证据,下肢大区的命中几乎不说明什么。 (https://5dlibrary.com/answers/zh/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c03/) - Ch.4 Twelve Branch Regions of the Body: Reading the Self, Health and Parents Palaces — Where the three palaces fall depends only on month and hour: each branch hosts the Self, Health and Parents palaces with share 0.0833, Self and Parents are always adjacent, and Health sits five branches behind Self. With the current eight mark stars the card says "high" for 0.9148 of charts, and 0.9749 with twelve, so "high" carries almost no information; all of it lies in the region named. Qing Yang never stands on Yin, Si, Shen or Hai and Tuo Luo never on Zi, Mao, Wu or You, which makes Chen (0.1488) and Wei (0.1457) the most named regions. The regions differ greatly in size, Xu about 14.0% and Wu about 1.2%; when the card says "high", five random marks confirm it by luck with chance 0.3211. The region-by-region subdivisions and the three context readings written out in this chapter prepare the east-west table of Chapter 7 and the pre-registered test of Chapter 8. (https://5dlibrary.com/answers/en/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c04/) 第4章 十二地支的体表区域:命宫、疾厄、父母三宫的读法 — 三宫的位置只由月份和时辰决定,每一支承载命宫、疾厄、父母的份额都是 0.0833,命宫与父母宫永远相邻,疾厄在命宫后第五支。按现行的八颗印记星,卡片在 0.9148 的星盘上说「高」,十二颗星时是 0.9749,所以「高」几乎不携带信息,信息全在被点名的区域。擎羊永不落寅、巳、申、亥,陀罗永不落子、卯、午、酉,使辰(0.1488)与未(0.1457)成为被点名最多的区域。十二区面积悬殊,戌约 14.0%,午约 1.2%;卡片说「高」时,五个随机印记偶然证实它的机会是 0.3211。本章逐区写出的细分与三种语境读法,是给第七章的东西合参表与第八章的预登记检验准备的。 (https://5dlibrary.com/answers/zh/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c04/) - Ch.5 The Seven Harsh Stars: Eighty-Four Marks on the Twelve Branches — The birth-year Ji falls in the Self palace with share 0.0867, in Health 0.0797 and in Parents 0.0856. The three are not one twelfth because the Ji star is usually a main star, and the main stars are not evenly spread relative to the Self palace. Given that a harsh star is already in one of the three palaces, the conditional share of the Ji in the same palace is: Qing Yang 0.0944 (the highest, because the year stem decides both where Qing Yang stands and which star carries the Ji), Tuo Luo 0.0833, Ling Xing and Tian Xing 0.0840, Huo Xing, Di Kong and Di Jie 0.0673. Over all charts, 0.1103 have at least one harsh star sitting with the birth-year Ji in one of the three palaces. In other words, the "with Ji" part of the cells lights up in about one chart in nine. (https://5dlibrary.com/answers/en/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c05/) 第5章 七颗刑煞星:十二地支上的八十四种印记 — 生年化忌落在命宫的比例是 0.0867,疾厄宫 0.0797,父母宫 0.0856。三个数字不等于十二分之一,因为化忌星多是主星,而主星相对于命宫的分布并不均匀。给定一颗刑煞星已经在三宫之一,化忌也在同一宫的条件比例是:擎羊 0.0944(最高,因为年干同时决定擎羊的位置与化忌星是哪一颗),陀罗 0.0833,铃星与天刑 0.0840,火星、地空、地劫 0.0673。所有命盘中,有 0.1103 至少有一颗刑煞星与生年化忌同坐三宫之一。也就是说,每一格的「化忌叠加」大约在九张命盘里亮一张。 (https://5dlibrary.com/answers/zh/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c05/) - Ch.6 The Five Peach Stars: Marks That Bleed, Cluster, Show, Hide and Pair — Placing the five peach stars on the skin is this book's reading of the Zi Wei tradition: Hong Luan shows, Tian Xi comes paired and pale, Tian Yao hides in a charming spot, Tan Lang clusters and favours the mouth and hands, Lian Zhen carries blood. The minor stars are placed simply: Tian Yao's palace depends on the hour alone, and the Chou, Yin and Shen hours send it into a body palace; Hong Luan and Tian Xi always face each other and light the Health-Parents axis together. Lian Zhen and Tan Lang are even by palace (0.0833 each) but uneven by branch. Of the sixty lit cells, at least one is lit in 0.7147 of charts, so each cell's reading must name a precise place, a precise form and a precise age. The ji overlay is read as raised, darkened, on a scar, hidden, and is most common in the cells of Lian Zhen and Tan Lang. Feeding this table to the body-mark card lowers the charts left at "medium" from 0.0852 to 0.0251, and makes the card all the more in need of a table that can veto it. (https://5dlibrary.com/answers/en/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c06/) 第6章 桃花五星:带血色、成簇、显眼、藏起与成对的记号 — 桃花五星落到皮肤上,是本书对斗数传统的读法:红鸾显眼,天喜成对而浅,天姚藏在有风情的位置,贪狼成簇并偏爱口与手,廉贞带血色。三颗小星的落位很简单:天姚所在宫只由时辰决定,丑、寅、申三个时辰把它送进体表宫;红鸾与天喜永远相对,在疾厄—父母一轴上同时照亮。廉贞、贪狼按宫位均匀(各 0.0833),按地支却不均匀。六十个亮格中,至少亮一格的命盘占 0.7147,所以每一格的读法必须写到具体位置、具体形态、具体年龄。化忌叠加读作凸起、变暗、压疤、藏起,在廉贞、贪狼的格子上最常见。把这张表喂给体表卡,会让退回「中等」的命盘从 0.0852 降到 0.0251,也让这张卡更需要一张能否决它的表。 (https://5dlibrary.com/answers/zh/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c06/) - Ch.7 The East-West Verification Table: How Two Body Maps Confirm and Veto Each Other — The western map has sides but lights only a third of the body per chart; the eastern map has no sides and lights up for almost everyone. Between the twelve branch regions and the thirty-six drekkana zones there are only 43 overlaps; 6 zones are out of the east's reach, the back and the back of the head are blind to the west, and only 0.485 of the skin is seen by both. The scoring rule is Z x S x F x A: confirmed by both 2 points, the west alone 1, the east alone 0.5, a western contradiction of side vetoes, a wrong form scores nothing, a wrong age halves. Under the base rates of Chapter 1 and the assumptions stated here, an unrelated chart scores on average only 0.240 out of 6, and the chance rate of a confirmation by both maps is only 0.017. The pre-registered test uses 150 people with 9 decoy charts each, and the mean percentile of own charts must exceed 0.5496; if it does not, this combined table is only a very finely written table. (https://5dlibrary.com/answers/en/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c07/) 第7章 东西合参核验表:两张身体地图如何互相确认、互相否决 — 西方地图分左右,但每张盘只照亮身体的三分之一;东方地图不分左右,而且几乎对谁都亮。十二个地支区域与三十六个 drekkana 区之间只有 43 个重叠,6 个区东方碰不到,背部与后脑是西方的盲区,皮肤只有 0.485 是两方都看得到的。评分规则是 Z × S × F × A:两方确认 2 分,西方单独 1 分,东方单独 0.5 分,西方左右相反即否决,形态不对不计分,年龄不对减半。在第一章的基准率与本章写明的假设下,一张无关的盘平均只能拿 0.240 分(满分 6),两方同时确认的偶然率只有 0.017。预先登记的检验用 150 人、每人 9 张诱饵盘,自己的盘的平均百分位必须超过 0.5496;做不到,这张合参表就只是一张写得很细的表。 (https://5dlibrary.com/answers/zh/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c07/) - Ch.8 Practice: The Mark Claims of Sample Chart A and a Pre-Registered Test — Chapter 8 works through sample chart A: the chance that at least one of three strict claims hits is 0.333, and the card's wording widens 0.141 to 0.266; a group comparison needs 188 people in all if the risk doubles and 672 if it rises by half, and a one-in-three design needs 133 people at a pick rate of 0.45. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c08/) 第8章 实践:样盘 A 的印记判断与一份预先登记的检验 — 第八章用样盘 A 演示:三条严格判断至少中一条的碰巧概率是 0.333,卡片的措辞把 0.141 放宽到 0.266;分组检验在风险比加倍时共需 188 人、增加一半时 672 人,三选一设计在选中率 0.45 时需 133 人。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 (https://5dlibrary.com/answers/zh/vedic-nakshatra-shadows-drekkana-body-zones-east-west-birthmark-verification/c08/) ## 5DL-B-119 · The Ziping Climate-Tension and Ten-Gods Codex of Human Games 中文书名:子平调候张力与十神人性博弈母典 DOI: pending Cite: Fan, Y. (2026). The Ziping Climate-Tension and Ten-Gods Codex of Human Games (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. Abstract: This volume treats two sets of Ziping rules as fixed structures: the one hundred and twenty climate-adjusting cells of the Qiongtong baojian, ten day stems by twelve month branches, and the ten gods computed from the producing and controlling cycle of the five phases and from equal or opposite polarity. It computes the structure of both rule sets exactly by enumerating the full calendar, and writes the traditional reading of every cell, every ten god and every classic combination as a sharp profile of personality and business: climate tension, light-side aura, shadow-side soft spot, how the money comes, contract blind spot, subtext of relationships and climate lever. It makes no divinatory claim: it states the rules precisely, computes the structure exactly and shows how the readings themselves could be tested; the readings are specific so that they can be refuted. All calculations are generated by the programs printed in the book; it is not investment, career, psychological, legal or marital advice. Chapter 1 enumerates the 43830 days from 1924 to 2043: each of the one hundred and twenty cells takes 353 to 378 days, a ratio of 1.0708 between the fullest and the emptiest, so no cell is rare. There are only 9 different first climate keys, Bing, Ren and Jia take most of the cells, and Bing comes first in a share of 0.2250; read as ten gods, the keys are mostly forces that discipline or hold up the day stem. Over 525960 charts under the book's counting rule the dominant ten god is almost evenly distributed, but in 0.7062 of charts the top two are less than one visible stem apart, and the first key is shown on a stem in only 0.2732. Chapters 2 to 6 write out the one hundred and twenty combinations of day stem and month cell by cell, and Chapter 7 the ten gods and twenty-four classic combinations with their detection rules. Chapter 8 demonstrates the look-up on a sample chart that contains no birth data: the climate cell does not depend on the hour, the dominant god is the same in all twelve hours, and the combination layer is the most fragile; it gives a pre-registered blind-matching protocol run in parallel with a Barnum control arm: with three choices and a true hit rate of 0.40 it needs 325 people, and 415 with two primary endpoints. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. 摘要:本卷把子平法的两套规则当作确定的结构来处理:《穷通宝鉴》按十个日主乘十二个月令排出的一百二十格调候用神,以及由五行生克与阴阳异同算出来的十神。它用全历法穷举精确计算这两套规则的结构,再把每一格、每一个十神、每一种经典组合的传统读法写成一份锋利的人格与商业读本:气候张力、阳面气场、阴面软肋、钱怎么来、契约盲区、关系潜台词与调候杠杆。本书不主张任何占验:它精确写出规则、精确计算结构,并说明怎样检验读法本身;读法写得具体,是为了让它可以被证伪。全部计算由书中程序生成;本书不是投资、职业、心理、法律或婚姻建议。 第一章穷举 1924 年至 2043 年共 43830 天:一百二十格每格 353 至 378 天,最多与最少之比 1.0708,没有哪一格稀少。首选调候用神只有 9 种,丙、壬、甲合占大半,丙为首选的比例是 0.2250;读成十神,多是管束或托住日主的力量。按全书的计数规则穷举 525960 张盘,主导十神接近均匀,但前两名差距不到一个透干的盘占 0.7062;首选用神透干的盘只占 0.2732。第二至第六章逐格写一百二十种日主与月令,第七章写十个十神与二十四种经典组合及其检测规则。第八章用一张不含任何出生资料的样盘示范查法:调候格不受时辰影响,主导十神在十二个时辰里都不变,组合一层最脆弱;并给出一份与巴纳姆对照臂并行、预先登记的盲配检验方案:三选一、真实命中率 0.40 时需要 325 人,两个主要终点时需要 415 人。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 Chapter answers: - Ch.1 Method: Climate Tension, Ten-God Arithmetic and Exact Distributions — Chapter 1 enumerates the 43830 days from 1924 to 2043: each of the one hundred and twenty cells takes 353 to 378 days, a ratio of 1.0708 between the fullest and the emptiest, so no cell is rare. There are only 9 different first climate keys, Bing, Ren and Jia take most of the cells, and Bing comes first in a share of 0.2250; read as ten gods, the keys are mostly forces that discipline or hold up the day stem. Over 525960 charts under the book's counting rule the dominant ten god is almost evenly distributed, but in 0.7062 of charts the top two are less than one visible stem apart, and the first key is shown on a stem in only 0.2732. (https://5dlibrary.com/answers/en/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c01/) 第1章 方法:调候张力、十神算术与精确分布 — 第一章穷举 1924 年至 2043 年共 43830 天:一百二十格每格 353 至 378 天,最多与最少之比 1.0708,没有哪一格稀少。首选调候用神只有 9 种,丙、壬、甲合占大半,丙为首选的比例是 0.2250;读成十神,多是管束或托住日主的力量。按全书的计数规则穷举 525960 张盘,主导十神接近均匀,但前两名差距不到一个透干的盘占 0.7062;首选用神透干的盘只占 0.2732。第二至第六章逐格写一百二十种日主与月令,第七章写十个十神与二十四种经典组合及其检测规则。 (https://5dlibrary.com/answers/zh/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c01/) - Ch.2 Jia and Yi Wood: Climate Tension in Twenty-Four Month Commands — Jia and Yi are both wood, but they want different weather: across Jia's twelve cells Geng and Ding each appear 11 times, because the great tree must be trimmed and set alight; across Yi's twelve cells Bing appears 10 times and Gui 9, because flowers and grasses need sunlight and rain. The first climate key already hides in the month branch in only 2 months for each stem; for the other twenty cells, what is most lacking has to be found outside the month branch. In the enumeration of the sixty-year cycle, winter wood with no fire in the year and month pillars and summer wood with no water in them both come to a share of 0.3500. Each of the twenty-four cells has been written out with its yang face, yin soft spot, money path, contract blind spot, relationship subtext and climate lever; these are interpretations, and the chapter gives two pre-registrable tests with the sample sizes they need. (https://5dlibrary.com/answers/en/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c02/) 第2章 甲木与乙木:二十四个月令里的气候张力 — 甲与乙同是木,要的天气却不同:甲的十二格里,庚与丁各出现 11 次,大树要被修整、被点燃;乙的十二格里,丙出现 10 次、癸出现 9 次,花草要阳光与雨露。第一调候字本来就藏在月支里的,甲乙各只有 2 个月,其余二十格的「最缺」都要到月支以外去找。在六十甲子的枚举里,冬季木年月两柱无火、夏季木年月两柱无水的比例都是 0.3500。二十四格各自写出了阳面、阴面、钱路、契约盲区、关系潜台词和调候杠杆;这些是诠释,本章给出了两项可以预先登记的检验和所需的样本量。 (https://5dlibrary.com/answers/zh/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c02/) - Ch.3 Bing and Ding: Twenty-Four Climate Tensions of the Sun and the Lamp — Bing is the sun and wants a lake; Ding is the lamp and wants both wood and an axe. p15 shows that among Bing's twelve cells Seven Killings (Ren) serves as a climate key 12 times, and among Ding's twelve cells Direct Wealth (Geng) serves 12 times and Direct Resource (Jia) 11 times; for both day stems the first key already hides in the month branch in only 2 months. p16, over 4320 equally weighted charts per cell, finds the mean share holding the first key to be 0.6398 (Bing) and 0.6528 (Ding); in Bing-Wei 0.2836 of charts hold not a single key, while eight other cells have a no-key share of 0.0000. The yang face, yin side, money, contracts, relationships and levers of all twenty-four cells are interpretations derived from two questions, "who will be the lake" and "who supplies and splits the wood", and p17 gives the means to test them: 300 people, 88 hits. (https://5dlibrary.com/answers/en/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c03/) 第3章 丙丁:太阳与灯烛的二十四种气候张力 — 丙是太阳,要一片湖;丁是灯烛,要柴也要斧。p15 显示,丙火十二格的调候用神里七杀(壬)出现 12 次,丁火十二格里正财(庚)出现 12 次、正印(甲)11 次,两个日干都只有 2 个月份的第一钥匙本来就藏在月令里。p16 在每格 4320 个等权的盘中数出:第一钥匙出现率平均为 0.6398(丙)与 0.6528(丁),丙未有 0.2836 的盘一把钥匙也没有,另有八格全缺的比例是 0.0000。二十四格的阳面、阴面、钱、契约、关系与杠杆,都是从「谁来做那片湖」「谁来供柴劈柴」这两个问题推出来的诠释,p17 给出了检验它们的办法:300 人,88 次命中。 (https://5dlibrary.com/answers/zh/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c03/) - Ch.4 Wu and Ji: Earth in the Twelve Month Commands — Earth's twelve months do not gain and lose by whole seasons; each season has one month in which earth runs the house. The year-round ten-god mix of Wu and Ji is symmetric: peer qi 0.2500, resource stars 0.1667, officer stars 0.1917. Bing fire appears in all twelve climate lists of both stems, so an earth day master spends a lifetime looking for the one who gives warmth. In the real calendar the first climate key is shown on a stem in fewer than four charts in ten for most cells, lowest for Ji born in You at 0.1734; Wu born in Chou has neither Bing nor Jia anywhere in 0.3229 of charts. The twenty-four contract blind spots are all different: performance clauses, change orders, deadlock, intellectual property, joint guarantees, vesting, payment terms, termination rights, prepayment and inventory, valuation adjustment, closed periods, joint accounts, non-competes, liability caps, drag-along and dilution, audit rights, exclusivity, limited liability, verbal side promises, consignment, profit-share inspection, full-discretion mandates, peak-season spending, and pooled-fund shares. They have only one thing in common: earth holds on to what it thinks it must keep, and signs the real risk over to the other side. (https://5dlibrary.com/answers/en/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c04/) 第4章 戊己:土在十二个月令里 — 土的十二个月不分季节整齐得失,而是每季一个当家月。戊与己全年的十神配比结构对称:同辈之气 0.2500,印 0.1667,官杀 0.1917。丙火出现在两干全部十二张调候清单里,所以土日干的人一生都在找那个给温度的人。真实日历里,首位调候之神透干的比例多在四成以下,最低是己土生酉月的 0.1734;戊土生丑月整盘无丙无甲的比例达 0.3229。二十四格的契约盲区各不相同:考核条款、变更单、僵局、知识产权、连带担保、兑现期、账期、解约权、预付与库存、对赌、封闭期、共有账户、竞业、责任上限、领售与稀释、审计权、独家条款、有限责任、口头附加承诺、寄售、分成查账、全权委托、旺季开支、资金池份额。共同点只有一个:土守住了它以为该守的,却把真正的风险签给了对方。 (https://5dlibrary.com/answers/zh/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c04/) - Ch.5 Geng and Xin: Twenty-Four Climates of the Metal Day Masters — All twenty-four metal cells come down to three questions: is there enough forge fire, is there enough water, is the earth too thick? Geng is raw iron never forged, its climate keys mostly Ding and Jia, read as rules and a target; Xin is a finished gem, its climate keys mostly Ren and Jia, read as a voice and clients. p25 shows that the month ten gods of Geng and Xin are perfectly symmetrical, with mean month support 0.4602 for both, strongest in You (1.0000) and weakest in Mao, Hai and Zi (0.0000); the four earth months give high support but carry the risk of buried metal. p26, over 525960 real calendar charts, shows that among Geng days the share with neither key is highest in Xu (0.1704) and among Xin days in Si (0.3042); Xin's keys are harder to assemble because Ren hides only in Shen and Hai; summer metal days with no water make up 0.0049 of all charts, winter metal days with no fire 0.0038. These readings are interpretations; the chapter gives two pre-registrable tests and the sample sizes they need. (https://5dlibrary.com/answers/en/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c05/) 第5章 庚辛:金日主的二十四种气候 — 金日主的二十四格都可以归结为三个问题:炉火够不够,水够不够,土是不是太厚。庚是未经锻打的生铁,调候钥匙多为丁、甲,读出来是规则与目标;辛是已成器的珠玉,调候钥匙多为壬、甲,读出来是发声与客户。p25 表明庚、辛的月令十神分布完全对称,平均月令支撑都是 0.4602,最强在酉(1.0000),最弱在卯、亥、子(0.0000);四个土月支撑虽高,却是埋金之忧。p26 在 525960 张真实日历盘上表明,庚日两钥皆无最高在戌月(0.1704),辛日最高在巳月(0.3042),辛的钥匙更难凑齐,因为壬只藏在申、亥两支;夏季无水的金日主占全部盘的 0.0049,冬季无火的占 0.0038。这些读法是诠释;本章给出了两项可以预先登记的检验与所需样本量。 (https://5dlibrary.com/answers/zh/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c05/) - Ch.6 Ren and Gui: The Twenty-Four Month Commands of the Water Day Masters — This chapter has written out the twenty-four cells of the water day masters. Ren is the river, and what it mostly lacks is banks and an outlet: among its twelve first keys Wu earth takes four and Jia wood three. Gui is rain, and what it mostly lacks is a source: among its twelve first keys metal takes eight. Across the twenty-four cells there are 62 climate keys, and a resource appears in 18 cells, the most common need of water. The calendar enumeration shows that the chance a key is actually in the chart varies widely: for Ren the first key is shown or rooted in between 0.5312 and 1.0000 of charts, for Gui between 0.4792 and 1.0000; the share with no key anywhere is 0.1782 for Ren in Xu and as high as 0.3704 for Gui in Shen. Sample chart A falls within the 0.1315 of Ren-in-Xu charts where Jia is only hidden and Bing is shown. These readings are interpretations; the chapter has given two tests that can be pre-registered, with the sample sizes they need. (https://5dlibrary.com/answers/en/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c06/) 第6章 壬癸:水日主的二十四个月令 — 本章写完水日主二十四格。壬是江河,缺的多是堤岸与出口:十二个第一用神里戊土占四、甲木占三。癸是雨露,缺的多是来源:十二个第一用神里金占八。二十四格合计 62 个调候用神,印在 18 格出现,是水最普遍的需要。日历枚举表明,钥匙在盘里的概率差别很大:壬水第一用神「透或藏」在 0.5312 到 1.0000 之间,癸水在 0.4792 到 1.0000 之间;所有钥匙都不在的比例,壬生戌月为 0.1782,癸生申月高达 0.3704。样盘 A 则落在壬生戌月里「甲只藏、丙透出」的那 0.1315 之中。这些读法是诠释,本章给出了可以预先登记的两项检验及其样本量。 (https://5dlibrary.com/answers/zh/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c06/) - Ch.7 The Ten Gods and the Twenty-Four Combinations: Who Speaks in a Chart, and Who Travels Together — Under the 80-point counting rule the ten gods dominate with nearly equal chances, the Eating God most at 0.1078, the Direct Resource least at 0.0942, and the dominant god equals the month branch's main-qi god in only 0.5278 of charts. Written as arithmetic, the twenty-four combinations hold 1.2074 times per chart on average, 0.7880 of charts have at least one, the most common is the Officer and seal generating each other, and the rarest, sole dominance, has only 0.0012. Of 276 pairs, 138 never hold together; the most frequent companions are Officer and Killings mixed with Officer and seal generating each other. Sample chart A is dominated by the Seven Killings and passes self and Killings evenly matched and the blade riding the Killings. Every dominance and every combination is written in six parts: structure, bright face, soft spot, how the money comes, contract blind spot, relationship subtext; they are readings of rules, not validated predictions, and the way to test them is set out in the hypothesis box above. (https://5dlibrary.com/answers/en/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c07/) 第7章 十神与二十四组合:谁在盘里说话,谁和谁结伴 — 按 80 点的计数规则,十个十神主导的机会几乎均等,最多的食神 0.1078,最少的正印 0.0942,主导十神等于月支本气的只有 0.5278。二十四种组合写成算术之后,平均每盘成立 1.2074 种,0.7880 的盘至少成立一种,最常见的是官印相生,最稀有的专旺格只有 0.0012。276 对组合里 138 对从不同时成立,最常结伴的是官杀混杂与官印相生。样盘 A 由七杀主导,成立身杀两停与羊刃驾杀。每一种主导与组合都写成了六段:结构、阳面气场、阴面软肋、钱怎么来、契约盲区、关系潜台词;它们是规则的读法,不是已经验证的预测,检验的方法写在上面的假说框里。 (https://5dlibrary.com/answers/zh/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c07/) - Ch.8 Practice: The Reading Card of Sample Chart A, and a Blind-Matching Test against Barnum Controls — Chapter 8 demonstrates the look-up on a sample chart that contains no birth data: the climate cell does not depend on the hour, the dominant god is the same in all twelve hours, and the combination layer is the most fragile; it gives a pre-registered blind-matching protocol run in parallel with a Barnum control arm: with three choices and a true hit rate of 0.40 it needs 325 people, and 415 with two primary endpoints. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c08/) 第8章 实践:样盘 A 的解读卡,以及一份对照巴纳姆的盲配检验 — 第八章用一张不含任何出生资料的样盘示范查法:调候格不受时辰影响,主导十神在十二个时辰里都不变,组合一层最脆弱;并给出一份与巴纳姆对照臂并行、预先登记的盲配检验方案:三选一、真实命中率 0.40 时需要 325 人,两个主要终点时需要 415 人。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 (https://5dlibrary.com/answers/zh/ziping-tiaohou-tension-120-month-commands-ten-gods-human-games/c08/) ## 5DL-B-120 · The Qintian Synastry Codex of Cross-Flying Transformations 中文书名:钦天双人合盘与宿命情缘飞化母典 DOI: pending Cite: Fan, Y. (2026). The Qintian Synastry Codex of Cross-Flying Transformations (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All calculations are generated by the programs printed in the book; it is not marital, psychological, legal or investment advice. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. Abstract: This volume writes the flights of the Qintian four transformations between two Zi Wei Dou Shu charts as one fixed rule: take the stem of a palace in one chart, look up the table for the four stars it turns into Lu, Quan, Ke and Ji, and the palace in which each star sits in the other chart is the part of the partner's life that this transformation enters. The book uses only the self, spouse, children and property palaces, computes the structure of the rule exactly by enumeration, and writes every cross cell, every mutual pattern and every month window as a sharp reading of a relationship: who holds the leash, the asymmetry of wanting, the trigger of a cold war under one roof, the exit, and month instructions for breaking contact and getting back together. It makes no divinatory claim: it states the rules precisely, computes the structure exactly and shows how the readings themselves could be tested; the readings are specific so that they can be refuted. All calculations are generated by the programs printed in the book; it is not marital, psychological, legal or investment advice. Chapter 1 gives the rules and shares used throughout: each of the sixty-four cross cells occurs in exactly one twelfth of all pairs of charts, 0.0833, so no cell is a rare bond, and one direction lights 5.3333 cells on average. The ten mutual patterns look only at what passes between the self and spouse palaces in each direction, and each has an executable detection rule: one-way Lu occurs in 0.4198 of pairs and one-way Ji in 0.4142, the commonest mutual pattern, Lu-Lu, in only 0.0865, and none of the ten holds in 0.2305. Month windows are fixed by the five-tiger month stems and repeat every five years; a couple has 7.6667 windows a year on average, and years with at least one window on both sides are only 0.0988. Chapters 2 to 5 write the four rows of the cross table cell by cell, Chapter 6 the ten mutual patterns and Chapter 7 the twelve month windows. Chapter 8 demonstrates the full look-up on two sample charts that contain no birth data and gives pre-registered protocols for blind choice, behavioural coding and dated events: a three-way blind choice with a true hit rate of 0.40 needs 318 couples. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. 摘要:本卷把钦天四化在两张紫微斗数命盘之间的飞化写成一条确定的规则:取一方某宫的宫干,查四化表得到它化出的禄、权、科、忌四颗星,这些星坐在另一方盘里的哪一宫,这一化就进了对方生活的哪一块。本书只取命宫、夫妻、子女、田宅四宫,用穷举精确计算这条规则的结构,再把每一个交叉格、每一种互飞格局、每一个流月窗口写成一份锋利的关系读本:权力拿捏、激情宿命差、同屋冷战触发点、出口,以及断联与复合的月份指令。本书不主张任何占验:它精确写出规则、精确计算结构,并说明怎样检验读法本身;读法写得具体,是为了让它可以被证伪。全部计算由书中程序生成;本书不是婚姻、心理、法律或投资建议。 第一章给出全书的规则与比例:六十四个交叉格每一格都恰好出现在十二分之一的盘对里,即 0.0833,没有一格是稀有的缘分;一个方向平均点亮 5.3333 格。十种互飞格局只看命宫与夫妻宫之间两个方向递了什么,判定规则逐条可执行:单向禄无回占 0.4198,单向忌无回占 0.4142,最多的互入格局禄禄互入也只有 0.0865,十格全不成立的配对占 0.2305。流月窗口按五虎遁的月干判定,每五年重复一次,一对人平均每年有 7.6667 个窗口,至少有一个双方窗口的年份只占 0.0988。第二至第五章逐格写出交叉表的四行,第六章写十种互飞格局,第七章写十二种流月窗口。第八章用两张不含任何出生资料的样盘示范完整的查法,并给出预先登记的盲选、行为编码与定月检验方案:三选一的盲选、真实命中率 0.40 时需要 318 对伴侣。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 Chapter answers: - Ch.1 Method: Cross Flights, Ten Mutual Patterns and Month Windows — Chapter 1 gives the rules and shares used throughout: each of the sixty-four cross cells occurs in exactly one twelfth of all pairs of charts, 0.0833, so no cell is a rare bond, and one direction lights 5.3333 cells on average. The ten mutual patterns look only at what passes between the self and spouse palaces in each direction, and each has an executable detection rule: one-way Lu occurs in 0.4198 of pairs and one-way Ji in 0.4142, the commonest mutual pattern, Lu-Lu, in only 0.0865, and none of the ten holds in 0.2305. Month windows are fixed by the five-tiger month stems and repeat every five years; a couple has 7.6667 windows a year on average, and years with at least one window on both sides are only 0.0988. (https://5dlibrary.com/answers/en/qintian-synastry-cross-flying-transformations-fated-romance-windows/c01/) 第1章 方法:交叉飞化、十种互飞格局与流月窗口 — 第一章给出全书的规则与比例:六十四个交叉格每一格都恰好出现在十二分之一的盘对里,即 0.0833,没有一格是稀有的缘分;一个方向平均点亮 5.3333 格。十种互飞格局只看命宫与夫妻宫之间两个方向递了什么,判定规则逐条可执行:单向禄无回占 0.4198,单向忌无回占 0.4142,最多的互入格局禄禄互入也只有 0.0865,十格全不成立的配对占 0.2305。流月窗口按五虎遁的月干判定,每五年重复一次,一对人平均每年有 7.6667 个窗口,至少有一个双方窗口的年份只占 0.0988。第二至第五章逐格写出交叉表的四行,第六章写十种互飞格局,第七章写十二种流月窗口。 (https://5dlibrary.com/answers/zh/qintian-synastry-cross-flying-transformations-fated-romance-windows/c01/) - Ch.2 The Self Palace Throws: Sixteen Ways Your Very Person Lands in Your Partner's Chart — This chapter has written out the sixteen cells of A's Self row: the four lu cells read where the pull your person gives away for free lands in your partner; the four quan cells read where you take the wheel; the four ke cells read the decency you lend your partner; the four ji cells read the debt you run up with your partner and collect. Exact computation shows that every one of the sixteen cells is exactly one twelfth (0.0833); no cell is a rare bond. What is rare is stacking: all four transformations landing in your partner's four palaces occurs in only 0.0039, lu and ji in the same palace of your partner in only 0.0111. Pairs that light at least one cell make up 0.8361, so "having a cross-link" says nothing by itself. These readings are interpretations; the chapter has given two tests that can be pre-registered, with the sample sizes they need. (https://5dlibrary.com/answers/en/qintian-synastry-cross-flying-transformations-fated-romance-windows/c02/) 第2章 命宫出手:你这个人落进对方盘里的十六种方式 — 本章写完了 A 命宫一行的十六格:禄四格读你这个人白给的引力落在对方哪里,权四格读你在对方哪里掌舵,科四格读你给对方的体面,忌四格读你在对方身上欠下、也讨着的债。精确计算表明,十六格每一格都恰好是十二分之一(0.0833),没有哪一格是稀有的缘分;稀有的是叠加:四化全落进对方四宫只有 0.0039,禄忌落进对方同一宫只有 0.0111。至少亮一格的盘对占 0.8361,所以「有交叉」本身不说明任何事。这些读法都是诠释;本章给出了两项可以预先登记的检验与所需样本量。 (https://5dlibrary.com/answers/zh/qintian-synastry-cross-flying-transformations-fated-romance-windows/c02/) - Ch.3 The Spouse Palace Throws: Sixteen Ways Your Way of Loving Lands in Your Partner's Chart — This chapter has written out the sixteen cells of A's Spouse row: the four lu cells read where what you give away for free in love lands in your partner; the four quan cells read the wheel you reach for when you love; the four ke cells read the decency you give in love; the four ji cells read the debt you run up in love and collect. As in the Self row, every one of the sixteen cells is exactly one twelfth (0.0833). The difference lies between the rows: in 0.1667 of charts the Spouse palace shares the Self palace's stem and the two rows coincide completely, so such a person's way of loving is simply who they are; the two rows together light 2.6667 cells on average, and the two ji hooks catch in the same palace of the partner in 0.0648. These readings are interpretations; the chapter has given two tests that can be pre-registered, with the sample sizes they need. (https://5dlibrary.com/answers/en/qintian-synastry-cross-flying-transformations-fated-romance-windows/c03/) 第3章 夫妻出手:你爱人的方式落进对方盘里的十六种方式 — 本章写完了 A 夫妻一行的十六格:禄四格读你爱人时白给的一份落在对方哪里,权四格读你爱人时伸手去掌的舵,科四格读你在爱里给出的体面,忌四格读你在爱里欠下、也讨着的债。与命宫一行一样,十六格每一格都恰好是十二分之一(0.0833)。不同在两行之间:0.1667 的盘里夫妻宫与命宫同干,两行完全重合,这样的人爱人的方式就是他这个人本身;两行合计平均点亮 2.6667 格,两支忌钩进对方同一宫的占 0.0648。这些读法都是诠释;本章给出了两项可以预先登记的检验与所需样本量。 (https://5dlibrary.com/answers/zh/qintian-synastry-cross-flying-transformations-fated-romance-windows/c03/) - Ch.4 The Children Row: Sixteen Cells of Desire, Children and Shared Making — This chapter has written out the sixteen cells of the Children row. Between two people the Children palace is desire, children and shared making at once; lu is the pull A gives freely, quan is A's steering, ke is the good face A lends, ji is the hook A leaves. Exact enumeration shows that each of the sixteen cells is exactly one twelfth (0.0833) and none is rare; but the four arrows of one row fall in clusters: in only 0.1639 of pairs does none of them land in these four palaces of B, the lone hook takes 0.1118, and lu and ji land in these four palaces together in 0.0972. The key to reading a cell is not rarity but whether its content really happens between the two of you; the chapter gives two tests that can be pre-registered, with the sample sizes they need. (https://5dlibrary.com/answers/en/qintian-synastry-cross-flying-transformations-fated-romance-windows/c04/) 第4章 子女一行:情欲、孩子与共同造物的十六格 — 本章写完了子女一行的十六格。子女宫在两人之间同时是情欲、孩子与共同造物;禄是 A 白给的拉力,权是 A 的掌舵,科是 A 给的体面,忌是 A 留下的钩子。精确枚举表明,十六格每一格都恰好是十二分之一(0.0833),没有哪一格罕见;但同一行的四支箭会成团落下:一支都不落进 B 这四个宫的只有 0.1639,孤钩占 0.1118,禄忌同落进这四个宫的占 0.0972。读每一格的关键,不在稀有,而在内容是否真的发生在你们之间;本章给出了两项可以预先登记的检验与所需样本量。 (https://5dlibrary.com/answers/zh/qintian-synastry-cross-flying-transformations-fated-romance-windows/c04/) - Ch.5 The Property Row: The Shared Home, Territory and the Closed Door — This chapter has written out the sixteen cells of the Property row. Between two people the Property palace is the shared home: territory, routine, rules, and the door shut after a quarrel. Each of the sixteen cells is exactly one twelfth (0.0833). The stems of the Property and Children rows are always different, four places apart in half of all charts and six in the other half, so A's bed and A's home speak two languages; at least one of A's eight arrows reaches B's Property palace in 0.4965 of pairs, both of A's ji land in B's Property palace in only 0.0028, and the two-sided closed door takes 0.0070. The key to reading this row is to read "home" as territory and order, and the cold war as a particular door and a particular ledger; the chapter gives one test that can be pre-registered, and shows the scale needed to test the most dramatic cell. (https://5dlibrary.com/answers/en/qintian-synastry-cross-flying-transformations-fated-romance-windows/c05/) 第5章 田宅一行:共同的家、领地与关上的门 — 本章写完了田宅一行的十六格。田宅宫在两人之间是共同的家:领地、作息、规矩,以及吵架后关上的那扇门。十六格每一格都恰好是十二分之一(0.0833)。田宅与子女两行的宫干永远不同,相差四位与六位各占一半,所以 A 的床与家用两种语言说话;A 的八支箭至少有一支进到 B 田宅宫的占 0.4965,两只忌同落 B 的田宅只有 0.0028,双向关门占 0.0070。读这一行的关键,是把「家」读成领地与秩序,把冷战读成具体的门、具体的账;本章给出了一项可以预先登记的检验,以及检验最戏剧化的一格需要多大的规模。 (https://5dlibrary.com/answers/zh/qintian-synastry-cross-flying-transformations-fated-romance-windows/c05/) - Ch.6 Ten Mutual Patterns: Reading Both Directions Together — This chapter has read the two directions together and obtained ten mutual patterns: only the lu, quan, ke and ji sent from the stems of both partners' Self and Spouse palaces are taken, and only whether they land in the other's Self and Spouse palaces is asked. Exact enumeration of 43200 × 43200 pairs shows: the four mutual patterns are rarest (quan-quan 0.0779 to lu-lu 0.0865), the four crosses nearly twice as common (ke-ji 0.1499 to lu-quan 0.1591), the two one-way patterns most common (one-way lu 0.4198, one-way ji 0.4142); pairs in which none of the ten holds make up 0.2305, and silent pairs that hand nothing in either direction 0.0822. Each pattern is written out in five parts: structure, power and leverage, the wanting gap, cold-war triggers under one roof, and the exit. These readings are interpretations; this chapter has given three tests that can be pre-registered and the sample sizes they need. (https://5dlibrary.com/answers/en/qintian-synastry-cross-flying-transformations-fated-romance-windows/c06/) 第6章 十种互飞格局:两个方向合起来读 — 本章把两个方向合起来,读出十种互飞格局:只取双方命宫与夫妻宫干发出的禄权科忌,只看它们是否落进对方的命宫与夫妻宫。精确枚举 43200 × 43200 个配对表明:四种互入最少(权权 0.0779 到禄禄 0.0865),四种交叉多出将近一倍(科忌 0.1499 到禄权 0.1591),两种单向最多(单向禄 0.4198,单向忌 0.4142),十格全不成立的占 0.2305,两个方向什么都不递的静默配对占 0.0822。每一格按结构、权力拿捏、激情宿命差、同屋冷战触发点与出口五段写完。这些读法是诠释;本章给出了三项可以预先登记的检验与所需的样本量。 (https://5dlibrary.com/answers/zh/qintian-synastry-cross-flying-transformations-fated-romance-windows/c06/) - Ch.7 Twelve Month Windows: When to Go Silent and When to Return — This chapter has brought time into the two-person reading: of the twelve months of a year, with month stems laid out by the five-tiger rule, a month opens one of twelve windows when the lu, quan, ke or ji it produces lands in your Spouse palace, his, or both. Exact enumeration shows that for every kind, on average exactly one month a year lands in your Spouse palace, and that whether a couple has a given window depends on the couple, not on the year. "Both" windows are rare: "lu, both" belongs to only 0.0670 of couples, and those who have it wait 5.4342 months on average; a lu window on either side belongs to 0.8802 of couples, with an average wait of 4.2257 months; from a "ji, both" month to the next lu month takes 4.2286 months on average. For every window the text states what the month does to contact, distance, cutting contact and getting back together, the action to take, and the thing never to do. These are interpretations; this chapter has given one pre-registrable test and its sample size. (https://5dlibrary.com/answers/en/qintian-synastry-cross-flying-transformations-fated-romance-windows/c07/) 第7章 十二种流月窗口:断联与复合的时机 — 本章把时间放进合盘:一年十二个月,月干按五虎遁排出,化出的禄、权、科、忌若落在你的夫妻宫、他的夫妻宫或两边同时,就打开十二种窗口中的一种。精确枚举表明,每一种化平均每年恰好有一个月落进你的夫妻宫;一对人有没有某种窗口取决于这对人自己,不取决于年份。双方窗口很少:禄-双方只属于 0.0670 的配对,拥有者平均等 5.4342 个月;任何一方的禄窗口属于 0.8802 的配对,平均等 4.2257 个月;从忌-双方的月份到下一个禄月平均 4.2286 个月。每一种窗口都写明了这个月对联系、距离、断联与复合做了什么,该做的动作,以及绝不要做的事。这些是诠释;本章给出了一项可以预先登记的检验及其样本量。 (https://5dlibrary.com/answers/zh/qintian-synastry-cross-flying-transformations-fated-romance-windows/c07/) - Ch.8 Practice: Reading Sample Charts A and B Together, and a Pre-Registered Test — Chapter 8 demonstrates the full look-up on two sample charts that contain no birth data and gives pre-registered protocols for blind choice, behavioural coding and dated events: a three-way blind choice with a true hit rate of 0.40 needs 318 couples. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/qintian-synastry-cross-flying-transformations-fated-romance-windows/c08/) 第8章 实践:样盘 A 与样盘 B 的合盘查法,以及一份预先登记的检验 — 第八章用两张不含任何出生资料的样盘示范完整的查法,并给出预先登记的盲选、行为编码与定月检验方案:三选一的盲选、真实命中率 0.40 时需要 318 对伴侣。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 (https://5dlibrary.com/answers/zh/qintian-synastry-cross-flying-transformations-fated-romance-windows/c08/) ## 5DL-B-121 · The Rotating Qimen Chaibu and Liuyao Najia Codex of Business Verdicts 中文书名:转盘奇门拆补法与六爻纳甲商业断事母典 DOI: pending Cite: Fan, Y. (2026). The Rotating Qimen Chaibu and Liuyao Najia Codex of Business Verdicts (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. Abstract: This volume treats the casting rules of rotating-plate Qimen with Chaibu bureaus and the najia rules of liuyao as fixed algorithms, computes their structure exactly by enumeration, and writes the traditional reading of every symbol, stem overlay, named pattern and line move as one plain business verdict and one plain verdict on love. It makes no divinatory claim: it states the rules precisely, computes the structure exactly and shows how the readings themselves could be tested; the verdicts are specific so that they can be refuted. All calculations are generated by the programs printed in the book; it is not investment, legal, career, psychological or marital advice. Chapter 1 sets out Chaibu bureaus after Zhang Zhichun's rotating-plate practice: the bureau changes at the exact moment of each solar term, and the yuan is set by the fu tou. Over the 87,660 double-hours from 2024 to 2043 each of the eighteen bureaus takes between 0.0552 and 0.0559, only about a third of the terms begin with an upper yuan, and 0.0055 of double-hours are cut into two charts by a term moment. Jia hides under the six yi, so only nine stems are visible and there are exactly 81 overlays; all 81 occur among 9,720 overlays, and a stem over itself takes 0.1975. The rule for each of the 19 named patterns is written out: Heaven Escape appears in only 0.0083 of bureau-hours, Six Yi Punishment in 0.5296 and Door Pressed in 0.6667, and a chart carries 3.3556 named patterns on average. On the liuyao side, a single moving line never clashes. Chapter 2 covers the nine stars, eight doors and eight spirits, Chapters 3 to 5 the 81 overlays, Chapter 6 the 19 named patterns, and Chapter 7 the 48 cells of six relations times eight line moves. Chapter 8 demonstrates the look-up on a sample hour chart that contains no birth data and on one business cast, and gives a pre-registered business forecast test: forecasts are scored by the Brier score against the base rate, and raising the direction hit rate from 0.50 to 0.60 needs 158 forecasts. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. 摘要:本卷把转盘奇门拆补法的排盘规则与六爻纳甲的装卦规则当作确定的算法,用穷举精确计算它们的结构,再把每一个符号、每一种干加、每一个名格、每一种动变的传统读法写成一条商业白话直断与一条情场白话直断。本书不主张任何占验:它精确写出规则、精确计算结构,并说明怎样检验读法本身;断语写得具体,是为了让它可以被证伪。全部计算由书中程序生成;本书不是投资、法律、职业、心理或婚姻建议。 第一章按张志春一脉的转盘实务写拆补定局:节气以交节的确切时刻换局,上中下元以符头定。在 2024 年到 2043 年的 87,660 个时辰中,十八局的份额都在 0.0552 与 0.0559 之间,只有约三分之一的节气从上元开始,0.0055 的时辰被交节切成两张盘。甲藏于六仪,所以可见的干只有九个,干加恰好八十一种;在 9,720 个干加中八十一种全部出现,同干相加占 0.1975。十九个名格逐条写出判定规则:天遁只出现在 0.0083 的局时里,六仪击刑在 0.5296 里,门迫在 0.6667 里,一张盘平均带 3.3556 个名格。六爻一侧,单爻发动永远不会化冲。第二章写九星、八门、八神,第三至第五章写八十一种干加,第六章写十九个名格,第七章写六种关系乘八种动变的四十八格。第八章用一张不含出生资料的样盘时盘与一次商业起卦示范查法,并给出一份预先登记的商业预测检验:以 Brier 分数与基础率比较,方向命中率从 0.50 提高到 0.60 需要 158 次预测。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 Chapter answers: - Ch.1 Method: Chaibu Bureaus, Four Rotating Plates and the Hundred Patterns Defined — Chapter 1 sets out Chaibu bureaus after Zhang Zhichun's rotating-plate practice: the bureau changes at the exact moment of each solar term, and the yuan is set by the fu tou. Over the 87,660 double-hours from 2024 to 2043 each of the eighteen bureaus takes between 0.0552 and 0.0559, only about a third of the terms begin with an upper yuan, and 0.0055 of double-hours are cut into two charts by a term moment. Jia hides under the six yi, so only nine stems are visible and there are exactly 81 overlays; all 81 occur among 9,720 overlays, and a stem over itself takes 0.1975. The rule for each of the 19 named patterns is written out: Heaven Escape appears in only 0.0083 of bureau-hours, Six Yi Punishment in 0.5296 and Door Pressed in 0.6667, and a chart carries 3.3556 named patterns on average. On the liuyao side, a single moving line never clashes. (https://5dlibrary.com/answers/en/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c01/) 第1章 方法:拆补定局、四层转盘与百格的定义 — 第一章按张志春一脉的转盘实务写拆补定局:节气以交节的确切时刻换局,上中下元以符头定。在 2024 年到 2043 年的 87,660 个时辰中,十八局的份额都在 0.0552 与 0.0559 之间,只有约三分之一的节气从上元开始,0.0055 的时辰被交节切成两张盘。甲藏于六仪,所以可见的干只有九个,干加恰好八十一种;在 9,720 个干加中八十一种全部出现,同干相加占 0.1975。十九个名格逐条写出判定规则:天遁只出现在 0.0083 的局时里,六仪击刑在 0.5296 里,门迫在 0.6667 里,一张盘平均带 3.3556 个名格。六爻一侧,单爻发动永远不会化冲。 (https://5dlibrary.com/answers/zh/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c01/) - Ch.2 Nine Stars, Eight Doors, Eight Spirits: Twenty-Five Symbols Read for Business and Love — This chapter has written out the three layers of symbols in one palace of a rotating chart: the nine stars give the temperament of the people and the setting, the eight doors the nature of the matter, the eight spirits the method working behind the scenes. Exact counts over all 1,080 bureau-hours: every door is at home in 0.2222 of them; on average 1.4444 doors per chart stand in door pressure, and the Rest, Life and Death doors are pressed only half as often as the other five. Tian Peng, Tian Ren and Tian Xin meet the good doors most often (0.4944, 0.4787, 0.4787), Tian Ying least often (0.2944). The Zhi Fu spirit always sits on the duty star, and its palace holds a qi in only 0.1111 of bureau-hours: authority and money do not share a palace. These verdicts are interpretations; the chapter gives two tests that can be pre-registered. (https://5dlibrary.com/answers/en/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c02/) 第2章 九星、八门、八神:二十五个符号的商业与情场直断 — 本章写完了转盘一宫之内的三层符号:九星定人与环境的气质,八门定事的性质,八神定背后起作用的手法。精确统计全部 1,080 个局时:每扇门归位的比例都是 0.2222;门迫平均每盘 1.4444 扇,休门、生门、死门受迫的机会只有其他五门的一半。天蓬、天任、天心与吉门同宫最多(0.4944、0.4787、0.4787),天英最少(0.2944)。值符神永远坐在值符星上,所在宫带三奇只占 0.1111,权威与钱并不同宫。这些断语是诠释;本章给出了两项可以预先登记的检验。 (https://5dlibrary.com/answers/zh/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c02/) - Ch.3 The Ten-Stem Responses, Part One: Wu, Ji and Geng over the Nine Stems — This chapter has written the first 27 ten-stem responses: heaven-plate Wu, Ji and Geng over each of the nine stems. Wu is capital, Ji is private intent, Geng is resistance; each overlay is read through its classical name and image, the production, control or harmony of its two stems, a business verdict and a love verdict. A check of all 131,496 double-hour charts of 2000-2029 shows that the three same-stem overlays each occur in 0.2222 of charts, and only in double-hours when the heaven plate does not turn; the 24 cross overlays all lie between 0.1074 and 0.1250, with a mean of 0.1148, so none is rare. By palace, Kun 2, because the centre lodges there, takes at least 0.225 of every overlay, and the six overlays whose stems are three steps apart in the earth-plate order go there more often (mean 0.308); the other seven palaces only move between 0.089 and 0.121, so no overlay has a fixed home. The verdicts are interpretations; the chapter gives one test that can be registered in advance and the number of contracts it needs. (https://5dlibrary.com/answers/en/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c03/) 第3章 十干克应(上):戊、己、庚临九干 — 本章写完了十干克应的前二十七格:天盘戊、己、庚各临九干。戊是本钱,己是私心,庚是阻力;每一格读古名与意象、两干的生克比和、商业直断与情场直断。对 2000—2029 年 131,496 张时辰盘的逐一排查表明:同干相叠的三格各占 0.2222,且只出现在天盘不转的时辰;二十四个异干格都在 0.1074 到 0.1250 之间,平均 0.1148,没有一格稀有。宫位上,坤二因寄宫而在每一格里都占到 0.225 以上,地盘次序上相隔三步的六格更常去坤二(平均 0.308);其余七宫只在 0.089 到 0.121 之间摆动,没有固定的老家。直断是诠释;本章给出了一项可以预先登记的检验与所需的合同数量。 (https://5dlibrary.com/answers/zh/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c03/) - Ch.4 Stem Overlays (II): Xin, Ren and Gui over the Nine Stems — This chapter has written out the twenty-seven cells in which heaven-plate xin, ren or gui sits over the nine earth-plate stems. Xin is the Heavenly Court and reads where a fault lands; ren is the Heavenly Jail and reads what catches the flow or snares it; gui is the Heavenly Net and reads where concealment falls. Enumerating the 131,496 double-hour charts from 2000 to 2029 shows that any cross cell occurs about once every eight or nine double-hours (0.1074 to 0.1249), that same-stem cells occur only in double-hours when the plate does not turn (0.2222), and that the six three-step cells are slightly more frequent because of the lodging rule. By palace, as charts are usually drawn all twenty-seven cells cluster in Kun 2 (0.224 to 0.316), but counting main stems only Kun 2 falls to 0.074 to 0.155 and the eight palaces are almost even: "always lands in a certain palace" is mainly an artefact of the lodging convention. These verdicts are interpretations; the chapter has given a pre-registrable test with contract disputes as its outcome, and the sample size it needs. (https://5dlibrary.com/answers/en/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c04/) 第4章 十干克应(中):辛、壬、癸加临九干 — 本章写完了天盘辛、壬、癸加临地盘九干的二十七格。辛为天庭,读错处落在哪里;壬为天牢,读流动被什么接住或缠住;癸为天网,读遮蔽落在哪里。逐时辰枚举 2000 年至 2029 年的 131,496 个时辰盘表明,任何一个异干格都大约每八到九个时辰出现一次(0.1074 到 0.1249),同干格只出现在不转盘的时辰(0.2222);相隔三步的六格因寄宫而略多。落宫上,照通常的画法二十七格都聚在坤二(0.224 到 0.316),但只算主干时坤二降到 0.074 到 0.155,八宫几乎均匀:所谓「专落某宫」,主要是寄宫约定造出来的。这些断语是诠释;本章给出了一项以合同涉诉为结果、可以预先登记的检验,以及它需要的样本量。 (https://5dlibrary.com/answers/zh/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c04/) - Ch.5 The Three Wonders Arrive: The 27 Overlays of Ding, Bing and Yi on the Heaven Plate — This chapter has written out the twenty-seven overlays of heaven-plate ding, bing and yi on the nine earth-plate stems. Reading takes three steps: what light the wonder on the heaven plate is (ding is documents and messages, bing is authority and momentum, yi is goodwill and the ability to bend), what ground the earth-plate stem is (principal, the private door, the checkpoint, the mistake, the net, the cover, or another beam of light), and whether the two generate, control or combine. Over the 131,496 double-hour charts of 2000 to 2029, p25 shows that the three same-stem fu yin overlays occur in 0.2222 of charts each; each of the other twenty-four occurs in 0.1055 to 0.1251 of charts, 0.1130 on average; no overlay is rare, and the whole excess comes from lodging the centre in Kun. p26 shows that all twenty-seven overlays land most often in Kun 2 (0.196 to 0.288), and that the other seven palaces are nearly flat. These verdicts are interpretations; the chapter gives two tests that can be pre-registered and the sample sizes they need. (https://5dlibrary.com/answers/en/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c05/) 第5章 三奇加临:丁、丙、乙在天盘的二十七种克应 — 本章写完了天盘丁、丙、乙落在九个地盘干上的二十七种克应。读法分三步:天盘的奇是什么光(丁是文书与消息,丙是权威与声势,乙是人情与身段),地盘的干是什么场(本钱、私门、关卡、错误、罗网、遮蔽,或另一束光),两者之间是生、是克、还是合。p25 在 2000 至 2029 年的 131,496 张时辰盘上表明:三种同干伏吟各占 0.2222;其余二十四种每一种出现在 0.1055 到 0.1251 的时辰盘里,平均 0.1130,没有罕见的叠法,多出来的部分全部来自中五寄坤。p26 表明:二十七种叠法都最常落在坤二宫(0.196 到 0.288),其余七宫几乎均匀。这些直断是诠释;本章给出了两项可以预先登记的检验与所需的样本量。 (https://5dlibrary.com/answers/zh/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c05/) - Ch.6 The Nineteen Named Patterns: Detection Rules, Real Frequencies and Verdicts — This chapter has written out, pattern by pattern, the detection rules, classical images and verdicts of the nineteen named patterns. With all 1,080 bureau-hours weighted equally: Heaven Escape is the rarest (0.0083), Door Pressed the most common (0.6667), Six Yi Punishment occurs in more than half (0.5296), Qi Meets the Envoy in more than four in ten (0.4315), Fu Yin in 0.3204, star Fu Yin and door Fu Yin each in 0.2222. Over the 87,660 hour charts of 2024 to 2043 in Beijing time, the shares differ from the equal-weight ones only in the fourth decimal. Nearly every chart mixes favourable and adverse patterns (0.7056), and a chart free of adverse patterns occurs in only 0.0259; so a named pattern is read by the palace it falls in, not by whether it appears. These verdicts are interpretations; the chapter gives one test that can be pre-registered. (https://5dlibrary.com/answers/en/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c06/) 第6章 十九个名格:判定规则、真实频率与直断 — 本章逐条写出十九个名格的判定规则、古象与直断。全部 1,080 个局时等权统计:天遁最少(0.0083),门迫最多(0.6667),六仪击刑过半(0.5296),三奇得使四成多(0.4315),伏吟 0.3204,其中星伏吟与门伏吟各 0.2222。按北京时间起 2024 至 2043 年的 87,660 张时盘,比例与等权统计只差小数第四位。几乎每张盘都吉凶并存(0.7056),没有凶格的盘只占 0.0259;所以名格要按它落在哪一宫来读,而不是按它有没有出现来读。这些断语是诠释;本章给出了一项可以预先登记的检验。 (https://5dlibrary.com/answers/zh/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c06/) - Ch.7 Six-Line Najia: Forty-Eight Cells of Relation and Move — Casting is a fixed piece of arithmetic: najia assigns branches by two-step strides, forward for yang trigrams and backward for yin ones; the eight palaces divide the hexagrams by eight masks; the six relations are the difference between line element and palace element modulo 5. Each of the eight moves has an exact rule, and one moving line may satisfy several. Because yang trigrams carry only yang branches and yin trigrams only yin branches, a single moving line produces just 48 kinds of branch change, only four of the eight advancing pairs are reachable, and a clash needs two or more moving lines. Over the 23040 single-moving-line cases, turning back to control (0.2292) is the most common and the tomb (0.0625) the rarest; among the forty-eight cells the Officer turning back to control (0.0625) is the largest. A time cast reaches only 192 of the 384 combinations of hexagram and moving line, and Siblings turning into cutting-off and every clash cell can never be produced by it. The verdicts of the forty-eight cells are interpretations, and the chapter gives one pre-registrable test. (https://5dlibrary.com/answers/en/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c07/) 第7章 六爻纳甲:六亲八动四十八格 — 装卦是一套确定的算术:纳甲按阳卦顺、阴卦逆隔位配支,八宫按八种掩码分卦,六亲是爻支五行与宫五行在模 5 下的差。八种变化各有一条精确的判定规则,一条动爻可以同时满足几种。由于阳卦只配阳支、阴卦只配阴支,单一动爻只产生 48 种支的变化,八对进神只有四对可得,化冲必须两爻以上同动才会出现。在 23040 个单动爻情形中,回头克(0.2292)最常见,化墓(0.0625)最少;四十八格中官鬼回头克(0.0625)最多。时间起卦只能起出 384 种「卦乘动爻」中的 192 种,兄弟化绝与全部化冲格永远起不出来。四十八格的直断是诠释,本章给出了一项可以预先登记的检验。 (https://5dlibrary.com/answers/zh/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c07/) - Ch.8 Practice: Reading an Hour Chart, a Liuyao Cast, and a Pre-Registered Business Forecast Test — Chapter 8 demonstrates the look-up on a sample hour chart that contains no birth data and on one business cast, and gives a pre-registered business forecast test: forecasts are scored by the Brier score against the base rate, and raising the direction hit rate from 0.50 to 0.60 needs 158 forecasts. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c08/) 第8章 实践:读一张时盘、一次起卦,与一份预先登记的商业预测检验 — 第八章用一张不含出生资料的样盘时盘与一次商业起卦示范查法,并给出一份预先登记的商业预测检验:以 Brier 分数与基础率比较,方向命中率从 0.50 提高到 0.60 需要 158 次预测。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 (https://5dlibrary.com/answers/zh/rotating-qimen-chaibu-hundred-patterns-liuyao-najia-business-verdicts/c08/) ## 5DL-B-122 · The Xuankong Flying-Star Sound-and-Light Field and Hamburg Midpoint Timing Codex 中文书名:玄空飞星空间声光场与汉堡中点择时母典 DOI: pending Cite: Fan, Y. (2026). The Xuankong Flying-Star Sound-and-Light Field and Hamburg Midpoint Timing Codex (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Both traditions are stated exactly and computed to the end by their own arithmetic; the combination of the three is the book's prescription and a testable hypothesis, not a demonstrated effect. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. Abstract: This volume puts two traditional rule systems and two pieces of physiological guidance into one "room by hour" prescription table: xuankong flying stars decide what each palace of a home is used for, the Hamburg School's 90-degree dial and solar arcs decide in which weeks of a year the schedule leans towards which palace, and how bright and how loud each palace should be at each hour comes from the consensus recommendations on melanopic equivalent daylight illuminance and from the WHO environmental noise guidelines. Both traditions are stated exactly and computed to the end by their own arithmetic; the combination of the three is the book's prescription and a testable hypothesis, not a demonstrated effect. All calculations are generated by the programs printed in the book; it is not medical, building-safety, investment or psychotherapeutic advice. Chapter 1 sets out the method: the twenty-four mountains are 15 degrees each, and a compass 1 degree off puts 0.0667 of bearings in the neighbouring mountain; periods 9 and 8 give 48 charts, period 9 only the two double-star types with 12 mountains each, period 8 the four types with 6 mountains each; the light targets of at least 250, at most 10 and at most 1 melanopic EDI for day, evening and sleep mean, with a 2700 K warm light, 556.8 lux at the eye by day and no more than 22.3 in the evening; thirteen factors form 78 axes and 156 solar-arc contacts, on average 17.08 per decade. Chapters 2 and 3 go mountain by mountain through periods 9 and 8, Chapter 4 writes the nine star numbers as nine sound-and-light fields, Chapter 5 covers the eighty-one mountain-facing combinations, Chapter 6 the timing of the 78 axes, and Chapter 7 merges a palace's star pair and a time window into a schedule. Chapter 8 takes sample house A and sample chart A through the whole procedure and gives a pre-registered test: two primary sleep outcomes and an effect size of 0.3 need 109 households; with only three declared axes, 0.25-degree core windows cover 0.0328 of weeks, while counting all 156 contacts they cover 0.5796. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. 摘要:本卷把两套传统规则与两份生理学指引放进同一张「房间乘时段」的处方表:玄空飞星决定房子的哪一宫做什么,汉堡学派的九十度盘与日弧决定一年中哪几周把日程往哪一宫倾斜,而每一宫在每个时段有多亮、多响,数字来自黑视素等效照度的共识建议与世界卫生组织的环境噪声指引。两套传统按原有算法写准、算尽;三者的组合是本书的处方,也是一个可检验的假说,不是已被证实的效果。全部计算由书中程序生成;本书不构成医疗、建筑安全、投资或心理方面的专业建议。 第一章写方法:二十四山每山 15 度,罗盘偏 1 度就有 0.0667 的方位判进隔壁的山;九运与八运共 48 盘,九运只有双星会向与双星会坐各 12 山,八运四类各 6 山;白天至少 250、傍晚至多 10、睡眠至多 1 的黑视素等效照度,换成 2700 K 的暖光,白天眼睛处要 556.8 勒克斯,傍晚不能超过 22.3;十三个因子组成 78 条轴线、156 种日弧接触,十年平均 17.08 次。第二、三章逐山写九运与八运的二十四山,第四章把九个星数写成九种声光场,第五章写八十一种山向组合,第六章写 78 条轴线的择时,第七章把一宫的星组合与一段时间窗口合成为日程。第八章用样宅 A 与样盘 A 走完全程,并给出预先登记的检验:两个主要睡眠结局、效应量 0.3 时需要 109 户;只申报三条轴线时,0.25 度的核心窗口只覆盖 0.0328 的周,而把 156 种接触全部算上则覆盖 0.5796。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 Chapter answers: - Ch.1 Method: Twenty-Four Mountains, Three Layers of Stars, Light and Sound Targets, and the 90-Degree Dial — Chapter 1 sets out the method: the twenty-four mountains are 15 degrees each, and a compass 1 degree off puts 0.0667 of bearings in the neighbouring mountain; periods 9 and 8 give 48 charts, period 9 only the two double-star types with 12 mountains each, period 8 the four types with 6 mountains each; the light targets of at least 250, at most 10 and at most 1 melanopic EDI for day, evening and sleep mean, with a 2700 K warm light, 556.8 lux at the eye by day and no more than 22.3 in the evening; thirteen factors form 78 axes and 156 solar-arc contacts, on average 17.08 per decade. (https://5dlibrary.com/answers/en/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c01/) 第1章 方法:二十四山、三层星盘、声光目标与九十度盘 — 第一章写方法:二十四山每山 15 度,罗盘偏 1 度就有 0.0667 的方位判进隔壁的山;九运与八运共 48 盘,九运只有双星会向与双星会坐各 12 山,八运四类各 6 山;白天至少 250、傍晚至多 10、睡眠至多 1 的黑视素等效照度,换成 2700 K 的暖光,白天眼睛处要 556.8 勒克斯,傍晚不能超过 22.3;十三个因子组成 78 条轴线、156 种日弧接触,十年平均 17.08 次。 (https://5dlibrary.com/answers/zh/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c01/) - Ch.2 The Twenty-Four Houses of Period 9: A Sound and Light Prescription for Each — Chapters 2 and 3 go mountain by mountain through periods 9 and 8, Chapter 4 writes the nine star numbers as nine sound-and-light fields, Chapter 5 covers the eighty-one mountain-facing combinations, Chapter 6 the timing of the 78 axes, and Chapter 7 merges a palace's star pair and a time window into a schedule. (https://5dlibrary.com/answers/en/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c02/) 第2章 九运二十四山:逐宅的声光场处方 — 第二、三章逐山写九运与八运的二十四山,第四章把九个星数写成九种声光场,第五章写八十一种山向组合,第六章写 78 条轴线的择时,第七章把一宫的星组合与一段时间窗口合成为日程。 (https://5dlibrary.com/answers/zh/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c02/) - Ch.3 The Twenty-Four Houses of Period 8: Sound and Light Fields of the Older Stock, and the Re-layout for Period 9 — The twenty-four charts of period 8 fall six into each of the four classes, and no sitting mountain shares its class with period 9. With the same rule as Chapter 2 (p16) this chapter gave every house its four zones: in the six prosperous houses the traffic is at the front and the bed at the back, in the six reversed houses the other way round. p17 gives the most practical verdict: in no house can the bed stay put on entering period 9, in sixteen houses all four zones must change place, and the traffic route holds only in the six houses sitting chou, gen, yin, chen, xun and si. (https://5dlibrary.com/answers/en/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c03/) 第3章 八运二十四山:旧宅的声光场与入九运的重排 — 八运的二十四盘四类各占六盘,与九运没有一个坐山同类。本章用与第二章相同的规则(p16)给每一宅分了四区:旺山旺向的六宅动线在前、床在后,上山下水的六宅反过来。p17 给出了最实际的结论:没有一宅的床可以在入九运后留在原地,十六宅的四个区全部要换位置,动线留得住的只有坐丑、坐艮、坐寅、坐辰、坐巽、坐巳六宅。 (https://5dlibrary.com/answers/zh/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c03/) - Ch.4 The Nine Numbers as Sound and Light Fields: Illuminance, Colour Temperature and Sound Level — This chapter reads the nine numbers as nine sound and light fields. The ruler for light is melanopic EDI at the eye: at least 250 in the daytime, at most 10 in the evening, at most 1 during sleep (Brown 2022); converted to lux, warm light needs more lux to reach the daytime dose (about 557 lux at 2700 K) but is allowed more in the evening (22.3 lux). The ruler for sound is the A-weighted equivalent level: outdoors, the WHO 2018 guidelines are the reference (who2018noise); indoors, levels add as energies and the loudest source sets the room. By the timeliness arithmetic, the ruling and rising numbers are 9, 1 and 2 in period 9, and were 8, 9 and 1 in period 8 (shenshi). Each of the nine numbers receives a prescription: one white cool and good for thinking, two black ventilated and dry, three jade with its edge dulled, four green with soft light for reading, five yellow left undisturbed, six white bright and decisive, seven red keeping the tongue in check, eight white steady and holding, nine purple the brightest and hottest yet dark at night too. The concrete values in the prescriptions are this book's translation, and the chapter gives a design that tests them separately from the numbers themselves. (https://5dlibrary.com/answers/en/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c04/) 第4章 九数声光场:每个数字的照度、色温与声级 — 本章把九个数字读成九种声光场。光的尺是眼睛处的 melanopic EDI:白天至少 250,傍晚不超过 10,睡眠不超过 1 (Brown 2022);换算成勒克斯时,暖光需要更多的勒克斯才能达到白天剂量(2700 K 约 557 勒克斯),却在傍晚允许更多(22.3 勒克斯)。声的尺是 A 计权等效声级:室外参照世卫 2018 指南 (who2018noise),室内按能量相加,最响的声源决定房间。按当运算术,九运里旺生的是九、一、二,八运里是八、九、一 (shenshi)。九个数字各得一份处方:一白清冷利思考,二黑要通风干爽,三碧压锋芒,四绿柔光读书,五黄禁动,六白明亮决断,七赤收口舌,八白稳守,九紫最亮最热但夜里也要暗下来。处方里的具体数值是本书的翻译,本章给出了一项能把它与数字本身分开检验的设计。 (https://5dlibrary.com/answers/zh/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c04/) - Ch.5 The Eighty-One Pairs: Stillness and Movement in Every Palace — This chapter has written out all eighty-one mountain-facing pairs. Each pair is read on three levels: the five-phase relation (quiet feeds active, active feeds quiet, quiet checks active, active checks quiet, same phase), the classical name (one and four in one palace, two and five combined, the fighting-bull hazard, crossed swords, nine and seven in one rut, fire burning the gate of heaven, purple and yellow poison, and the four River Chart couples), and timeliness. p25 shows that the standard charts of period 9 contain only nine pairs, each used once per chart, every one summing to a multiple of 9; period 8 has 41. p26 shows that two thirds of the slots of period 9 are checking pairs, and that across all nine periods only 5-5 never occurs. p27 lists the source periods of every pair. The principle behind the verdicts fits in one sentence: in the quiet part, ask whether the mountain star lets people sleep and sit in peace; in the active part, ask whether the facing star is worth opening the door and walking for; put the numbers in season where the room is used, and the numbers out of season and the five yellow where it is used least. (https://5dlibrary.com/answers/en/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c05/) 第5章 八十一组数对:每一宫的静与动 — 本章写完了八十一组山向数对。读每一组看三样:五行生克(静生动、动生静、静克动、动克静、比和),古诀的名目(一四同宫、二五交加、斗牛煞、交剑煞、九七合辙、火烧天门、紫黄毒药,以及河图四对),以及当运时令。p25 表明,九运的标准盘只有九组数对,每张盘各用一次,两数之和都是 9 的倍数;八运有 41 组。p26 表明九运盘三分之二的宫位是相克的组合,而且全部九个运里只有 5-5 永不出现。p27 列出了每一组的来源运。直断的原则可以压成一句:静处看山星能不能让人安睡久坐,动处看向星值不值得开门走动;当令的数放在要用的地方,失令的数和五黄放在少用的地方。 (https://5dlibrary.com/answers/zh/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c05/) - Ch.6 The Seventy-Eight Axes: What to Schedule, What to Avoid, How Long the Window Lasts — The 78 axes fall into three classes: 10 among personal points, 40 between personal points and planets, 28 between planets; each is touched at least once before the age of 45.66, and the median gap between its two touches is 45.64 years. The solar arc gains one arc-minute every 6.18 days; the core window of 0.25 deg runs about 93 days either side, preparation at 1.00 deg opens about 371 days ahead, settling at 0.50 deg takes about 185 days, and the peak is about 37 days either side. Four minutes of error in the birth time moves the dates of an MC axis by about a year; an hour moves the dates of a Moon axis by about 203 days. Declaring three axes, core windows cover only 0.0328 of weeks; counting all 78 gives 0.5796, an "it came true" almost everywhere, which is why axes must be declared in advance and at most three at a time. The Do and Avoid of each axis are this book's prescription: slow axes are prepared early and released late, sudden axes are trusted only at the peak and given a buffer, hazy axes are verified before anything is signed, and the three generational axes serve only as reminders. (https://5dlibrary.com/answers/en/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c06/) 第6章 七十八条轴线:被碰到时宜做什么、忌做什么、窗口多长 — 78 条轴线分三类:个人点之间 10 条,个人点与行星之间 40 条,行星之间 28 条;每一条在 45.66 岁之前至少被碰一次,两次接触的间隔中位数是 45.64 年。日弧每 6.18 天走 1 角分,0.25 度的核心窗口前后各约 93 天,1.00 度的准备期提前约 371 天,0.50 度的收尾约 185 天,峰值是前后各约 37 天。出生时间差 4 分钟,涉及天顶的日期就挪约一年;差一小时,涉及月亮的日期挪约 203 天。申报三条轴线,核心窗口只占 0.0328 的周;78 条全算是 0.5796,几乎处处「应验」,所以必须预先申报、最多三条。每一条的宜与忌是本书的处方:慢的轴线提前准备、延后恢复,突变的轴线只信峰值并留缓冲,模糊的轴线先核实再签字,代际的三条轴线只作提醒。 (https://5dlibrary.com/answers/zh/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c06/) - Ch.7 Synthesis: A Zone's Star Pair, a Timing Window and One Week's Prescription — The synthesis is ranked by authority: physiology first, space second, field third, time fourth; a lower layer cannot overturn a higher one. Seven of the nine zones of sample house A are open, with 357 to 477 lux at the eye in the daytime, at most 19.5 to 29.9 lux in the evening depending on each zone's colour temperature, and at most 4.2 lux from the night lamp; the sound level is the stricter of field and use. A window tilts only time: Do becomes five classes of activity placed by class, Avoid becomes prohibitions; daytime activities end by 20:00, palaces holding a 5 take no activity, talks on hard axes end by 15:00, and a core window moves at most 10 hours a week. The five plans used 10, 6, 8, 4 and 10 hours, and in each the blocks that broke a rule were refused or moved. When two axes overlap, core ranks above settling and settling above preparation; in sample chart A's 122 weeks, 42 have both axes open, with a mean tilt of 5.6 hours a week. All of this is this book's prescription; whether the time layer contributes anything is for the crossover test above to answer. (https://5dlibrary.com/answers/en/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c07/) 第7章 合成:一宫的星组合、一段时间窗口与一周的处方 — 合成按权限排序:生理第一,空间第二,场第三,时间第四;下层不能推翻上层。样宅 A 的九宫里七宫开放,白天眼睛处 357 到 477 勒克斯,傍晚按各宫色温不超过 19.5 到 29.9 勒克斯,起夜的小灯不超过 4.2 勒克斯;声级取场与用途中更严的一个。窗口只倾斜时间:宜翻成五类活动、按类找宫,忌写成禁令;白天的事 20:00 前结束,含 5 的宫不进活动,硬轴线的谈话 15:00 前结束,核心窗口每周最多 10 小时。五份周表分别用了 10、6、8、4、10 小时,每一份里违规的块都被拒绝或挪走。两条轴线重叠时,核心高于收尾、收尾高于准备;样盘 A 的 122 周里有 42 周两线同开,平均每周倾斜 5.6 小时。这些都是本书的处方,时间层是否有贡献,要靠上面的交叉检验回答。 (https://5dlibrary.com/answers/zh/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c07/) - Ch.8 Practice: Sample House A, Sample Chart A and a Pre-Registered Test — Chapter 8 takes sample house A and sample chart A through the whole procedure and gives a pre-registered test: two primary sleep outcomes and an effect size of 0.3 need 109 households; with only three declared axes, 0.25-degree core windows cover 0.0328 of weeks, while counting all 156 contacts they cover 0.5796. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c08/) 第8章 实践:样宅 A、样盘 A 与一份预先登记的检验 — 第八章用样宅 A 与样盘 A 走完全程,并给出预先登记的检验:两个主要睡眠结局、效应量 0.3 时需要 109 户;只申报三条轴线时,0.25 度的核心窗口只覆盖 0.0328 的周,而把 156 种接触全部算上则覆盖 0.5796。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 (https://5dlibrary.com/answers/zh/xuankong-flying-stars-indoor-sound-light-fields-hamburg-midpoint-timing/c08/) ## 5DL-B-123 · The Qintian Codex of Decade and Year Overlays and Thrust-Wave Events 中文书名:钦天大限流年叠宫与推力波事件显化母典 DOI: pending Cite: Fan, Y. (2026). The Qintian Codex of Decade and Year Overlays and Thrust-Wave Events (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. Abstract: This volume treats the rules by which the author's reading engine scores, year by year, the thrust behind money (M), the thrust behind work (W) and the relationship signals as a fixed structure, computes its distribution over all birth combinations exactly by enumeration, and writes every "decade theme x year kind", "decade theme x relationship signal" and "month kind x season" as a sharp handbook of advance and retreat: structure, real-world scene and play. It makes no divinatory claim: it states the rules precisely, computes the structure exactly and shows how the readings themselves could be tested; the readings are specific so that they can be refuted. All calculations are generated by the programs printed in the book; it is not investment, business, legal, psychological or marital advice. Chapter 1 sets out how the decade stem re-stems the twelve palaces by the five-tiger rule, how the new stems of the wealth and career palaces give the thrust sources, how a year is scored by its branch with half steps borrowed into empty palaces, and the order in which six year kinds (hard, roller, mirror, attack, steady, hold) and four relationship signals (true love, spark, hidden thunder, quiet) are assigned. Over 518400 equally weighted birth combinations each scanned for sixty years, 31104000 year cells, the most common year is the zero-thrust hold year at 0.3749; hard years take 0.1639 and attack years only 0.0252. Hard years arrive by quota: a decade has at most two, decades with exactly two make up 0.6630 against 0.2887 for independent years; the decade theme hardly changes the frequency of year kinds, only their reading; the travel decade lies outside the sixty-year scan. True love and spark each take 0.3218 and quiet years 0.4195; a sixty-year cycle holds only 20 month patterns. Chapters 2 to 5 write the six year kinds and four signals for each of the twelve decade themes, Chapter 6 the twenty-four month keys, Chapter 7 the stacking of the three layers and the assembly of a reading card, and Chapter 8 demonstrates the look-up on a sample chart that contains no birth data and gives a pre-registered testing protocol: with three choices and a true hit rate of 0.40, it needs 325 people. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. 摘要:本卷把作者测算引擎中逐年计算搞钱推力 M、事业推力 W 与关系信号的规则当作一个确定的结构,用穷举精确计算它在全部出生组合上的分布,再把每一种「大限主题 × 年型」「大限主题 × 关系信号」与「月型 × 季节」写成一份锋利的进退手册:结构、现实场景与进退打法。本书不主张任何占验:它精确写出规则、精确计算结构,并说明怎样检验读法本身;读法写得具体,是为了让它可以被证伪。全部计算由书中程序生成;本书不是投资、经营、法律、心理或婚姻建议。 第一章写大限干按五虎遁重排十二宫、财帛与官禄新干化出推力源、流年按地支计分与借进空宫的半格,以及六种年型(硬仗、过山车、镜像、进攻、稳进、蓄力)与四种关系信号(正缘、桃花、外缘暗雷、静默)的判定顺序。在 518400 种等权出生组合各扫六十年、共 31104000 个年格中,最常见的是零推力的蓄力年,占 0.3749;硬仗年占 0.1639,进攻年只有 0.0252。硬仗年按配额出现:一个大限最多两个,恰好两个的大限占 0.6630,十年若彼此独立只会是 0.2887;大限主题几乎不改变年型频率,只改变读法;迁移大限不在六十年扫描之内。正缘与桃花各占 0.3218,静默年占 0.4195;一个甲子里流月只有 20 种排列。第二至第五章逐一写十二种大限主题的六种年型与四种信号,第六章写二十四种月键,第七章写三层叠加与解读卡的拼法,第八章用一张不含出生资料的样盘示范查法,并给出预先登记的检验方案:三选一、真实命中率 0.40 时需要 325 人。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 Chapter answers: - Ch.1 Method: How Thrust Is Computed, Six Year Kinds and Four Relationship Signals — Chapter 1 sets out how the decade stem re-stems the twelve palaces by the five-tiger rule, how the new stems of the wealth and career palaces give the thrust sources, how a year is scored by its branch with half steps borrowed into empty palaces, and the order in which six year kinds (hard, roller, mirror, attack, steady, hold) and four relationship signals (true love, spark, hidden thunder, quiet) are assigned. Over 518400 equally weighted birth combinations each scanned for sixty years, 31104000 year cells, the most common year is the zero-thrust hold year at 0.3749; hard years take 0.1639 and attack years only 0.0252. Hard years arrive by quota: a decade has at most two, decades with exactly two make up 0.6630 against 0.2887 for independent years; the decade theme hardly changes the frequency of year kinds, only their reading; the travel decade lies outside the sixty-year scan. True love and spark each take 0.3218 and quiet years 0.4195; a sixty-year cycle holds only 20 month patterns. (https://5dlibrary.com/answers/en/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c01/) 第1章 方法:推力怎样算出来,六种年型与四种关系信号 — 第一章写大限干按五虎遁重排十二宫、财帛与官禄新干化出推力源、流年按地支计分与借进空宫的半格,以及六种年型(硬仗、过山车、镜像、进攻、稳进、蓄力)与四种关系信号(正缘、桃花、外缘暗雷、静默)的判定顺序。在 518400 种等权出生组合各扫六十年、共 31104000 个年格中,最常见的是零推力的蓄力年,占 0.3749;硬仗年占 0.1639,进攻年只有 0.0252。硬仗年按配额出现:一个大限最多两个,恰好两个的大限占 0.6630,十年若彼此独立只会是 0.2887;大限主题几乎不改变年型频率,只改变读法;迁移大限不在六十年扫描之内。正缘与桃花各占 0.3218,静默年占 0.4195;一个甲子里流月只有 20 种排列。第二至第五章逐一写十二种大限主题的六种年型与四种信号,第六章写二十四种月键,第七章写三层叠加与解读卡的拼法,第八章用一张不含出生资料的样盘示范查法,并给出预先登记的检验方案:三选一、真实命中率 0.40 时需要 325 人。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 (https://5dlibrary.com/answers/zh/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c01/) - Ch.2 The Self, Siblings and Spouse Decades — This chapter placed the six thrust kinds and the four relationship signals into the Self, Siblings and Spouse decades and wrote out all thirty cells, each in three parts: structure, real-world scene and play. p10 gives the overlay: the Self decade is the first decade, and its decade palaces coincide with the natal ones; the Siblings and Spouse decades appear only on backward-running charts, 259200 of all 518400 charts for each, with their thrust supplied instead by natal Health and Property, and by natal Travel and Fortune. p11 shows that the decade theme does not change the frequencies of the year kinds; the largest gap between the three themes is only 0.0005. p12 shows that a decade holds on average 1.64 hard-battle years, that 0.9908 of decades have at least one hard-battle or roller-coaster year, and that attack years average only 0.25. p13 shows that every decade has at least one true-love year, 3.22 on average. The readings are interpretations; the design and sample sizes of the tests are set out in p14. (https://5dlibrary.com/answers/en/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c02/) 第2章 命宫、兄弟、夫妻三种大限 — 本章把六种推力年份与四种关系信号放进命宫、兄弟、夫妻三种大限,写完三十格,每格分结构、现实场景、进退打法三段。p10 给出叠宫关系:命宫大限是第一个大限,大限各宫与本命重合;兄弟与夫妻大限只在逆行的盘上出现,在全部 518400 张盘里各占 259200 张,推力分别改由本命疾厄、田宅与本命迁移、福德供给。p11 表明大限主题不改变年份种类的频率,三种主题之间最大的差距只有 0.0005。p12 表明一个十年平均有 1.64 个硬仗年,0.9908 的十年至少有一个硬仗年或过山车年,全力进攻年平均只有 0.25 个。p13 表明正缘进场年每个十年至少一个,平均 3.22 个。读法是诠释,检验的设计与样本量已在 p14 列出。 (https://5dlibrary.com/answers/zh/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c02/) - Ch.3 The Children, Wealth and Health Decades: Ten Years of Handing Over, Ten Years of Money, Ten Years of the Body — This chapter read the fourth to sixth decades of backward-running charts: Children, Wealth and Health. The exact enumeration of 43200 charts shows that the proportions of year kinds within the three decades are almost identical: hard battles about 0.1639, roller coasters about 0.0583, full attacks about 0.0252, steady advances about 0.3013, half-step mirrors about 0.0763 and defend-and-store years about 0.3749, no different from all twelve decade themes taken together. A decade has at most two hard-battle years, and exactly two in 0.6630 of decades; full-attack years average only around 0.253, and more than three decades in four have none. The relationship signals do not vary with the theme either, but they stack on the year kinds in a clear way: true-love entries are commonest in hard-battle years, attack years most often carry a spark, and hidden thunder prefers borrowed years. The difference between the three decades lies in the overlays: in a Children decade money comes from Friends and work from Parents; in a Wealth decade money comes from Career and work from the Self; in a Health decade money comes from Property and work from Siblings. The thirty readings are written on that basis, each giving structure, the real-world scene and advance and retreat. All of these are interpretations; the chapter gives three tests that can be pre-registered. (https://5dlibrary.com/answers/en/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c03/) 第3章 子女、财帛、疾厄三限:交出去的十年、钱的十年、身体的十年 — 本章读了逆行盘第四到第六个大限:子女、财帛、疾厄。精确枚举 43200 张盘表明,三种大限内的年份比例几乎相同:硬仗约 0.1639,过山车约 0.0583,全力进攻约 0.0252,稳步推进约 0.3013,镜像半格约 0.0763,防守蓄力约 0.3749,与十二种大限的合计没有差别。一个十年最多两个硬仗年,恰有两个的占 0.6630;全力进攻年平均只有 0.253 个左右,四分之三以上的十年一个都没有。关系信号同样不随主题变化,但与年份种类有明显的叠法:硬仗年里正缘进场的比例最高,全力进攻年最常带火花,外缘暗雷偏爱借来的年份。三种大限的差别在叠宫:子女限的钱从交友来、事从父母来;财帛限的钱从官禄来、事从命宫来;疾厄限的钱从田宅来、事从兄弟来。三十节读法据此写成,每节给出结构、现实场景与进退打法。这些都是诠释,本章给出了三项可以预先登记的检验。 (https://5dlibrary.com/answers/zh/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c03/) - Ch.4 Travel, Friends, Career: Going Out, Going Among People, Running Your Own Engine — The three decades have almost the same year frequencies: hard 0.1639, hold around three in eight, true love about a third, thunder about one in nine. What differs is the wiring. The Travel decade wires money to your state of mind and work to your partner; it is always the seventh decade, outside the engine's sixty years. The Friends decade wires money to paperwork and credit and work to the people you trained, and buries its thunder around your marriage. The Career decade wires money to you personally and work to your cash, and its two hard years often come close together. (https://5dlibrary.com/answers/en/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c04/) 第4章 迁移、交友、官禄:向外走、向人群、向自己的引擎 — 三段大限的年份频率几乎一样:硬仗 0.1639,守藏三成七上下,正缘三成二,暗雷一成一。不同的是接线。迁移十年把钱接到你的心境上、把事业接到伴侣身上,它永远是第七个大限,在引擎的六十年之外。交友十年把钱接到文书与授信上、把事业接到你带出来的人身上,暗雷埋在婚姻周围。官禄十年把钱接到你本人、把事业接到现金上,两场硬仗常常挨得很近。 (https://5dlibrary.com/answers/zh/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c04/) - Ch.5 Property, Fortune, Parents: Three Decades of Territory, Inner Reserve and Authority — This chapter has written the six thrust kinds and four relationship signals of the Property, Fortune and Parents decades, thirty sections in all. Only forward charts reach these three decades within sixty years, at twelve to twenty-five, twenty-two to thirty-five and thirty-two to forty-five. p25 and p26 show that the shares of each kind of year are almost the same in all three themes, hard years 0.1639 and true-love years 0.3218 in each; a decade holds at most two hard years, and no decade lacks a true-love year. p28 gives the order: a hard year is most often followed by a hold year, an attack year is often followed straight away by a hard year, and more than half of mirror years are followed by a hold year. Sample A's Property decade takes two blows in a row at thirty-nine and forty, one on the money of the intimate relationship and one on the work of partners, preceded by three quiet years, with thirty-eight the only hold year in which to prepare. The chapter has set out a pre-registrable test and shown how large a study must be to test its central claim, that the theme changes the play. (https://5dlibrary.com/answers/en/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c05/) 第5章 田宅、福德、父母:地盘、心力与授权的三个十年 — 本章写完了田宅、福德、父母三个大限的六种推力年与四种情缘信号,共三十节。这三个十年只有顺行盘在六十年里走得到,依次落在十二到二十五岁、二十二到三十五岁、三十二到四十五岁。p25 与 p26 证明,三个主题里各种年份的比例几乎相同,硬仗年都是 0.1639,正缘进场年都是 0.3218;一个大限最多两个硬仗年,没有一个大限缺少正缘进场年。p28 给出先后:硬仗之后最常是防守,进攻之后常紧跟硬仗,镜像之后一半以上是防守。样盘 A 的田宅十年在三十九、四十岁连挨两拳,一拳在亲密关系的钱上,一拳在合伙的事上,前面是三个情缘静默年,三十八岁是唯一可以提前备战的防守年。本章给出了一项可以预先登记的检验,并说明了检验「主题改变剧本」这一核心说法需要多大的规模。 (https://5dlibrary.com/answers/zh/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c05/) - Ch.6 Twenty-Four Month Cells: Six Kinds of Thrust in Four Seasons — This chapter cut one layer below the year cell: restem the twelve palaces with the year's own stem, take the four transformations thrown by that year's Wealth and Career stems, place them on the twelve months with month 1 on Yin, classify six month kinds and cross them with four seasons to get twenty-four cells. Exact enumeration shows that the hold month is the most common (0.3749) and the attack month the rarest (0.0253); every year has at least one hard month (1.0000), 1.97 on average; the mirror month peaks in autumn (0.1174) and is rarest in winter (0.0433). Month kind and year kind are almost independent: they coincide with probability 0.2791 against 0.2681 under independence, so a year must be read on both layers. Each of the twenty-four cells was written out with its structure, real-world scene and advance-and-retreat play; the chapter gives a test that can be pre-registered and the sample it needs. (https://5dlibrary.com/answers/en/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c06/) 第6章 二十四种月格:六种推力落进四个季节 — 本章往年格下面再切一层:用流年自己的干换出十二宫干,取这一年的财帛与官禄宫干化出的四化,按正月建寅落到十二个月上,判出六种月格,再乘以四个季节,得到二十四格。精确枚举表明:防守蓄力月最多(0.3749),全力进攻月最少(0.0253);每一年都至少有一个硬仗月(1.0000),平均 1.97 个;镜像半格月在秋季最多(0.1174)、冬季最少(0.0433)。月格与年格几乎彼此独立,同格比例 0.2791,独立时应为 0.2681,所以读一年必须两层都看。二十四格逐格写了结构、现实场景与进退打法;本章给出了一项可以预先登记的检验与所需的样本量。 (https://5dlibrary.com/answers/zh/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c06/) - Ch.7 Three Layers Stacked: Decade Theme, Year Cell and Month Cell as One Instruction — A year is read on three layers: the natal palace gives the decade's battlefield, the decade stem gives the year's posture and relationship signals, and the year stem gives the beat of the twelve months. The three layers are three independent restemmings, and stacking means three forces meeting in the same month. The half-step mirror occurs only in empty palaces (year cells land on one 0.1656 of the time) and never in a palace with a main star; its mean thrust of 0.733 is below the steady 1.144, and the borrowed plot often carries someone else's ji. Nine merging rules arrange the layers into one instruction: the theme sets the battlefield, the year sets the cap, the month sets the timing, hard months are kept for collection and defence, an attack year looks first for its attack month (0.4545 of attack years have one), chances in a hold year are probes only, borrowed years run half exposure, and when true love and thunder share a year both are taken seriously (they coincide 0.0367 of the time and are independent). The card assembles the overall instruction, the relationship line, the month strip and the merged hint in that order. The four stacks of sample chart A show the full reading; the chapter gives two tests that can be pre-registered and the samples they need. (https://5dlibrary.com/answers/en/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c07/) 第7章 三层叠加:大限主题、年格与月格合成一条指令 — 一年的读数有三层:本命宫位给出十年的战场,大限的干给出这一年的姿态和关系信号,流年的干给出十二个月的节拍。三层是三次独立的换干,叠加的意思是三股力在同一个月相遇。镜像半格只在空宫出现(年格落空宫的比例 0.1656),有主星的宫永远不会出现镜像;它的平均推力 0.733,低于稳进的 1.144,而且借来的那块地上常常连着别人的忌。九条合成规则把三层排成一条指令:主题定战场,年格定上限,月格定时点,硬仗月留给收款和防守,全力进攻年先找全力进攻月(0.4545 的全力进攻年有全力进攻月),防守蓄力年的机会只做试探,借力年敞口减半,正缘与暗雷同年两个都当真(同年比例 0.0367,二者彼此独立)。卡片按总指令、关系、月格条、合成提示的顺序装配。样盘 A 的四个叠加给出了完整的读法;本章给出了两项可以预先登记的检验和所需的样本量。 (https://5dlibrary.com/answers/zh/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c07/) - Ch.8 Practice: Ten Years of Sample Chart A and a Pre-Registered Test — Sample chart A is fixed completely by its structure. Over sixty years it has 23 steady, 17 hold, 10 hard, 5 mirror, 4 roller and 1 attack year; 2026 is "田宅-steady" plus a "田宅-true_love" carried by a Ji, 2029 opens the career decade, and 2030 and 2032 are the two hard years the decade's quota hands out. Whether a year reading is accurate has to be answered by two kinds of test, blind matching and event clustering: 325 people for three choices at a hit rate of 0.40, and 243 events for a hard-year effect of one and a half times. Until such tests are run, every year key in this book is only a clearly written rule that can be overturned. (https://5dlibrary.com/answers/en/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c08/) 第8章 实践:样盘 A 的十年与一份预先登记的检验 — 样盘 A 只用结构就能完全确定。它六十年里稳进 23、蓄力 17、硬仗 10、镜像 5、过山车 4、进攻 1;2026 年是「田宅-steady」加一个由忌撑起的「田宅-true_love」,2029 年进入官禄大限,2030 与 2032 年是这个大限按配额发下的两个硬仗年。一份年运读法准不准,要用盲配与事件集中两类检验回答:三选一、命中率 0.40 时 325 人,硬仗年效应一倍半时 243 件事件。在这些检验做出来之前,本书的每一个年键都只是一条写得清楚、可以被推翻的规则。 (https://5dlibrary.com/answers/zh/qintian-decadal-annual-overlays-thrust-wave-event-manifestation/c08/) ## 5DL-B-124 · The Boundary-Hour Rectification, Twin-Chart and Bone-Weight Baseload Codex 中文书名:交界时辰反推校时、双胞胎分盘与六十纳音称骨定盘母典 DOI: pending Cite: Fan, Y. (2026). The Boundary-Hour Rectification, Twin-Chart and Bone-Weight Baseload Codex (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. Abstract: Which chart should a person use whose birth time lies close to the boundary between two double-hours, or who cannot say exactly when they were born? This volume writes the question as a procedure that can be computed: the clock reading is converted to true solar time through the longitude correction and the equation of time, its distance in minutes from the nearest boundary is set against the recording error, and three checkable facts from the person's past are combined into one likelihood ratio; the chart is fixed only at a posterior of 0.90, and otherwise reported as "undetermined". The same approach is applied to splitting the charts of twins and to two old methods that look only at the year, month, day and hour of birth: the fifty-two bone weights attributed to Yuan Tiangang and the sixty-year nayin cycle, read as a pressure baseload and a wealth-carrying capacity. The book makes no divinatory claim: it states the rules precisely, computes the structure exactly and shows how the readings themselves could be tested; all calculations are generated by the programs printed in the book; it is not medical, psychological, legal or investment advice. Chapter 1 gives the method and exact shares: births less than 30 minutes from some boundary are 0.5000 of the total; taking the double-hour straight from a UTC+8 clock puts 0.5300 of clock times in Chengdu into the wrong double-hour over the year; across one boundary the self palace always moves back one place and its main stars change in 0.9428 of cases; the two day-change conventions of the Zi hour decide 0.0417 of day pillars; and the 259200 combinations of the four bone-weight tables average 42.3833 qian and never reach the seven liang two qian of the verses. Chapters 2 and 3 hold thirty-nine questions about the past for the thirteen boundaries, Chapter 4 the eight twin-splitting rules, Chapters 5 and 6 read the fifty-two bone weights, Chapter 7 the sixty nayin years, and Chapter 8 demonstrates the procedure on a sample chart that contains no birth data and gives a pre-registered blind-judgement test: with a true accuracy of 0.65 it needs only 67 people. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. 摘要:出生时刻离时辰交界很近、或者根本说不准的人,该用哪一张盘?本卷把这个问题写成一套可以计算的程序:先把钟面时间经过经度修正与时差换成真太阳时,量出它离最近交界的分钟数与记录误差,再用三件可以查证的往事合成一个似然比,后验到 0.90 才定盘,到不了就写「未定」。同一套写法用于双胞胎分盘,以及两种只看出生年月日时的古法:袁天罡称骨的五十二级重量与六十甲子纳音,读成抗压底盘与财富承载力。本书不主张任何占验:它精确写出规则、精确计算结构,并说明怎样检验读法本身;全部计算由书中程序生成;本书不是医疗、心理、法律或投资建议。 第一章给出方法与精确比例:离某条交界不到 30 分钟的出生占 0.5000;按东八区钟面直接取时辰,成都全年有 0.5300 的钟面时刻落进错的时辰;跨过一条交界,命宫必退一格,命宫主星在 0.9428 的情形下改变;早晚子时的两种换日法牵动 0.0417 的日柱;称骨四张表的 259200 种组合平均 42.3833 钱,称不出歌诀里的七两二钱。第二、三章是十三条交界的三十九道往事题,第四章是双胞胎的八条分盘规则,第五、六章读五十二级称骨,第七章读六十甲子纳音,第八章用一张不含出生资料的样盘演示查法,并给出一份预先登记的盲判定检验:真实准确率为 0.65 时只需 67 人。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 Chapter answers: - Ch.1 Method: True Solar Time, Distance to a Boundary, Likelihood Ratios, Bone Weight and Nayin — Chapter 1 gives the method and exact shares: births less than 30 minutes from some boundary are 0.5000 of the total; taking the double-hour straight from a UTC+8 clock puts 0.5300 of clock times in Chengdu into the wrong double-hour over the year; across one boundary the self palace always moves back one place and its main stars change in 0.9428 of cases; the two day-change conventions of the Zi hour decide 0.0417 of day pillars; and the 259200 combinations of the four bone-weight tables average 42.3833 qian and never reach the seven liang two qian of the verses. (https://5dlibrary.com/answers/en/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c01/) 第1章 方法:真太阳时、交界距离、似然比、称骨与纳音 — 第一章给出方法与精确比例:离某条交界不到 30 分钟的出生占 0.5000;按东八区钟面直接取时辰,成都全年有 0.5300 的钟面时刻落进错的时辰;跨过一条交界,命宫必退一格,命宫主星在 0.9428 的情形下改变;早晚子时的两种换日法牵动 0.0417 的日柱;称骨四张表的 259200 种组合平均 42.3833 钱,称不出歌诀里的七两二钱。 (https://5dlibrary.com/answers/zh/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c01/) - Ch.2 Five Night Boundaries: 11 p.m. to 5 a.m. — Chapters 2 and 3 hold thirty-nine questions about the past for the thirteen boundaries, Chapter 4 the eight twin-splitting rules, Chapters 5 and 6 read the fifty-two bone weights, Chapter 7 the sixty nayin years, and Chapter 8 demonstrates the procedure on a sample chart that contains no birth data and gives a pre-registered blind-judgement test: with a true accuracy of 0.65 it needs only 67 people. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c02/) 第2章 夜半五道交界:晚上十一点到清晨五点 — 第二、三章是十三条交界的三十九道往事题,第四章是双胞胎的八条分盘规则,第五、六章读五十二级称骨,第七章读六十甲子纳音,第八章用一张不含出生资料的样盘演示查法,并给出一份预先登记的盲判定检验:真实准确率为 0.65 时只需 67 人。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 (https://5dlibrary.com/answers/zh/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c02/) - Ch.3 Eight Daytime Boundaries: 7 a.m. to 9 p.m. — On the eight daytime boundaries the day pillar never changes and only the hour does, so the changes on each line can be written as a fixed card. The bureau still changes in 0.5000 of cases and Zi Wei moves in 0.4672; the main-star combination of the Self palace changes in 0.9471 to 0.9603 of cases, and on average 11.84 to 11.93 of the twelve palaces differ. Tian Xing enters Self in the Si hour, falls back into Parents in the Wu hour and into Fortune in the Wei hour, so on the 9 a.m., 11 a.m. and 1 p.m. lines the mark-star set of Self or Parents always changes. On the 3 p.m. and 9 p.m. lines bone weight does not move, and rectification has to rely on the shifts of the Body palace, Tian Xing, Tian Yao and Zuo Fu. If sample chart A had been born thirty-eight minutes or more earlier, its Self palace would lose Qing Yang, Huo Xing and Tian Xing. All twenty-four questions ask about past events that can be checked against records; this chapter gives three tests that can be pre-registered and the number of people each group needs. (https://5dlibrary.com/answers/en/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c03/) 第3章 白昼八道交界:早上七点到晚上九点 — 白天八道交界的日柱永远不变,变的只有时辰,因此每条线上的变化可以写成固定的卡片。五行局改变率仍是 0.5000,紫微换位率仍是 0.4672,命宫主星组合改变的比例在 0.9471 到 0.9603 之间,十二宫平均有 11.84 到 11.93 个宫内容不同。天刑在巳时进入命宫、午时退入父母、未时退入福德,使早上九点、上午十一点、下午一点三条线上命宫或父母宫的煞星组合必然改变。下午三点与晚上九点两条线称骨不动,只能靠身宫、天刑、天姚与左辅的位移校时。样盘 A 若早生三十八分钟以上,命宫就失去擎羊、火星与天刑。二十四道题全部问有记录可查的往事;本章给出了三项可以预先登记的检验与每组所需人数。 (https://5dlibrary.com/answers/zh/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c03/) - Ch.4 Splitting Twin Charts: Eight Rules and What Twin Research Shows — This chapter arranges the splitting of twin charts as eight rules. Inside one hour slot the first-born takes the original chart, and there are three mutually contradictory ways of casting the second-born's chart: Self palace from the Travel palace, Self palace from the Siblings palace, or the hour pillar moved one hour later; p22 shows that the Siblings-palace method agrees with a birth one hour later on the main stars of the Self palace in 0.5483 of cases. Twins cut by a boundary have different charts by nature: by p21 the bureau changes in half of the pairs, the main stars of the Self palace differ in 0.9428 of the pairs, Wen Chang and Di Kong always stay under the same palace name, and the other four hour-driven stars shift by two palaces. Same-sex twins run their decades in the same direction, opposite-sex twins one forward and one backward; caesarean twins are almost always in the same hour; the early/late Zi school and the day-changes-at-Zi school give opposite judgements on twins born around midnight. Twin research shows that similarity follows genes, not the moment of birth (Bouchard 1990) (Dean 2003). The chapter gives two pre-registrable tests and uses p23 to compute the sample sizes they need. (https://5dlibrary.com/answers/en/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c04/) 第4章 双胞胎分盘:八条规则与双生子研究 — 本章把双胞胎分盘整理成八条规则。同一时辰内,先出生者用原盘,后出生者的另起一盘有三种互相矛盾的办法:迁移宫起命、兄弟宫起命、时柱推后一时;p22 表明,兄弟宫起命在命宫主星上与晚一个时辰出生一致的比例是 0.5483。被分界切开的双胞胎,盘天然不同:按 p21,五行局在一半的配对里改变,命宫主星在 0.9428 的配对里不同,文昌与地空始终留在同一宫名之下,其余四颗随时辰走的星挪动两宫。同性双胞胎大限同向,异性双胞胎一顺一逆;剖腹产双胞胎几乎总在同一时辰;早晚子分派与子初换日派对午夜前后出生的双胞胎给出相反的判断。双生子研究显示相似程度随基因而不随出生时刻 (Bouchard 1990) (Dean 2003)。本章给出两项可以预先登记的检验,并用 p23 算出了它们所需的样本量。 (https://5dlibrary.com/answers/zh/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c04/) - Ch.5 Bone Weight, the Lighter Half: 2 Liang 1 Qian to 4 Liang 6 Qian — This chapter read the twenty-six lighter weights of the bone-weight table, from 2 liang 1 qian to 4 liang 6 qian. Exact enumeration of 259200 combinations shows: the mean is 42.38 qian, the median 4 liang 2 qian, and the most common weight 3 liang 9 qian (5.37%); the twenty-six cells of this chapter hold 69.47% together, the weights below 3 liang only 4.30%, and 2 liang 1 qian occurs once in 7200 combinations. The light end is pressed down mainly by the month, the twelfth month being the only five-qian month in the table; from 4 liang 2 qian on, all four parts return to near their means. Bone weight is very sensitive to hour boundaries: the Zi-Chou and Hai-Zi boundaries jump ten qian and the Chen-Si boundary seven; sample chart A weighs 3 liang 8 qian in the Si hour and only 3 liang 1 qian if moved back to Chen. Every cell is written out in four parts, gist of the verse, structure, pressure baseline and wealth capacity; these readings are interpretations, and the chapter gives two pre-registrable tests with the sample sizes they need. (https://5dlibrary.com/answers/en/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c05/) 第5章 称骨上半程:二两一钱到四两六钱 — 本章读了称骨表较轻的二十六个重量,从二两一钱到四两六钱。精确枚举 259200 种组合表明:平均 42.38 钱,中位数四两二钱,最常见的是三两九钱(5.37%);本章二十六格合占 69.47%,三两以下只占 4.30%,二两一钱每 7200 种组合才有一种。轻的那一端主要由月份压出来,腊月是全表唯一五钱的月份;四两二钱以后四部分都回到均值附近。称骨对时辰交界很敏感:子丑与亥子之交跳十钱,辰巳之交跳七钱,样盘 A 在巳时称三两八钱,退到辰时就只剩三两一钱。每一格按古歌大意、结构、抗压底盘与财富承载力四段写完;这些读法是诠释,本章给出了两项可以预先登记的检验与所需的样本量。 (https://5dlibrary.com/answers/zh/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c05/) - Ch.6 Bone Weight, the Heavier Half: 4 Liang 7 Qian to 7 Liang 2 Qian — This chapter read the twenty-six heavier weights of the bone-weight table, from 4 liang 7 qian to 7 liang 2 qian. Exact enumeration shows: 5 liang and above is 18.06%, 6 liang and above only 1.64%, 4244 combinations; the heaviest weight that can be produced is 7 liang 1 qian, with 16 combinations, and 7 liang 2 qian does not exist in this set of tables, so the limits of the verses and of the tables do not line up. The heavy end is pushed by the four parts in turn; from 6 liang upwards it depends on a heavy year (ten years such as Ji Mao and Wu Wu) plus one of the two sixteen-qian hours, Zi and Si, and every combination from 6 liang 6 qian upwards was born in the Zi or the Si hour. Heavy weights are therefore the most sensitive to rectification: 0.837 of the combinations of 6 liang and above drop out of the system band with a single hour's error. Every cell is written out in four parts, gist of the verse, structure, pressure baseline and wealth capacity, with the heavy-cell readings focused on how such people overstep their limits in good times; the chapter gives one pre-registrable test with the sample sizes it needs. (https://5dlibrary.com/answers/en/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c06/) 第6章 称骨下半程:四两七钱到七两二钱 — 本章读了称骨表较重的二十六个重量,从四两七钱到七两二钱。精确枚举表明:五两及以上占 18.06%,六两及以上只占 1.64%,共 4244 种组合;最重只能称到七两一钱,16 种组合,七两二钱在这套表里不存在,歌诀与数表的边界并未对齐。重的一端由四个部分轮流推动,到六两以上,靠的是一个重年(己卯、戊午等十个年份)加上子、巳两个十六钱的重时辰;六两六钱以上的组合全部出生在子时或巳时。因此重格对校时最敏感:0.837 的六两以上组合只差一个时辰就会跌出系统档。每一格按古歌大意、结构、抗压底盘与财富承载力四段写完,重格的读法着重于它在顺境里如何越过边界;本章给出了一项可以预先登记的检验与所需的样本量。 (https://5dlibrary.com/answers/zh/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c06/) - Ch.7 The Sixty Sound-Elements: Pressure Base and Wealth-Carrying Capacity of the Year Cell — This chapter translates the sound-elements of the sixty year cells into two modern indices: the pressure index P, the sum of the life-stage tier of the sound-element's element on the year branch and the branch's relation to it; and the carrying type W, which combines strong or weak with whether money is in sight. By p36 the mean P is 2.20, and only 6 cells are S+, able to carry large sums; only 10 of the pairs sharing a sound-element have the same P, and in 8 one cell is strong and the other weak, so every cell has been written separately with its structure, pressure base and wealth-carrying capacity. The year cell coincides completely with the birth cohort; by p37 the effective sample size of any test is the number of birth years, and a forty-year window is far from enough. The chapter gives three tests that can be pre-registered, of which the Start of Spring comparison strips out the cohort effect and is the one most worth doing first. (https://5dlibrary.com/answers/en/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c07/) 第7章 六十纳音:年份格的抗压底盘与财富承载力 — 本章把六十年份格的纳音翻译成两个现代指数:抗压指数 P,由纳音五行在年支上的十二长生档位与年支的生克相加而得;承载类型 W,由身旺身弱与财是否在眼前组合而得。按 p36,P 的平均是 2.20,真正能扛大钱的 S+ 只有 6 格;同一纳音的两格只有 10 对 P 相同,有 8 对一格身旺、一格身弱,所以每一格都单独写了结构、抗压底盘与财富承载力。年份格与出生队列完全重合,按 p37,检验的有效样本量是出生年份的个数,四十年的窗口远远不够。本章给出三项可以预先登记的检验,其中立春断点的比较能把队列效应剥掉,是最值得先做的一项。 (https://5dlibrary.com/answers/zh/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c07/) - Ch.8 Practice: Rectifying, Weighing and Reading Sample Chart A, and a Pre-Registered Test — Sample chart A's true solar time is 09:38, 38 minutes after the Chen | Si boundary. On the two sides the main stars stay put; what moves is the palace labels and four hour stars, Huo Xing, Ling Xing, Wen Qu and Di Jie; the self palace changes from Ju Men to Tian Xiang, the bone weight drops from three liang eight qian to three liang one qian without changing band, and the nayin Fire in the Furnace does not move. The card first prints the three general questions of Chapter 3; this chapter refines them into three follow-ups: where the scar is, whether a parent was away for a long time, and whether there was a single large loss. With a recording error of plus or minus 60 minutes, even three answers all as expected will usually only allow "undetermined". Whether these readings hold must be decided by blind-judgement tests: at least 153 people against chance, and at least 305 per group against the sham group. (https://5dlibrary.com/answers/en/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c08/) 第8章 实践:样盘 A 的校时、称骨与纳音,以及一份预先登记的检验 — 样盘 A 的真太阳时 09:38,离辰巳之交 38 分钟。两侧主星不动,动的是宫名与四颗时辰星:火星、铃星、文曲、地劫;命宫从巨门换成天相,称骨从三两八钱降到三两一钱,档位不变,纳音炉中火不动。卡片先印第三章的三道通用题,本章再把它们细化成三道追问:疤的位置、父母是否长期离家、有没有一次性的大损失。按前后 60 分钟的记录误差,即使三题全中,也多半只能写「未定」。这些读法是否属实,要靠盲判定检验:对机会水平至少 153 人,对假题组每组至少 305 人。 (https://5dlibrary.com/answers/zh/boundary-hour-rectification-twin-charts-nayin-bone-weight-baseload/c08/) ## 5DL-B-125 · The Liuyao Codex of Multiple Moving Lines, Hidden Motion, Hidden Spirits and Timing 中文书名:六爻多爻同动暗动伏神与精准应期母典 DOI: pending Cite: Fan, Y. (2026). The Liuyao Codex of Multiple Moving Lines, Hidden Motion, Hidden Spirits and Timing (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All calculations are generated by the programs printed in the book; it is not investment, legal, psychological or marital advice. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. Abstract: This volume treats the casting rules of Jing Fang's najia, the eight palaces with their world and response lines and the six relations, together with the definitions the whole book shares (into clash, into combination, lone moving, lone still, hidden motion, day break, hidden and flying spirits, void, out of void) and twelve timing rules, as fixed algorithms; it computes their structure in three-coin casts exactly by enumeration, and writes the traditional reading of each situation as one plain business verdict and one plain verdict on love. It makes no divinatory claim: it states the rules precisely, computes the structure exactly and shows how the readings themselves could be tested; the verdicts are specific so that they can be refuted. All calculations are generated by the programs printed in the book; it is not investment, legal, psychological or marital advice. Chapter 1 finds that casts with two or more moving lines take 0.4661 and a lone still line only 0.0044; a clash has a single route, Kun and Xun changing into each other with the middle and top lines moving, so one moving line never clashes, only 0.0312 of all moving lines turn into a clash, and casts with a clash are 0.0233; of still lines struck by the day, hidden motion takes 0.0347 and day break 0.0486, and the two never share a cast; only 0.3125 of the 64 hexagrams carry all five relations, and two of the twenty hidden-flying pairs never occur; a cast switches on 2.9642 entries on average, and only 0.0090 switch on none. Chapter 2 covers the 24 cells of multiple motion, Chapter 3 hidden motion and day break, Chapter 4 the 20 hidden-flying cells, Chapter 5 void and out of void, Chapter 6 the twelve timing rules, and Chapter 7 reads several multi-line casts from start to finish. Chapter 8 demonstrates the look-up and the timing on a business cast with three moving lines that switches on ten entries, and gives a pre-registered forecasting test: with one day of tolerance each, the timing rules that apply to that cast cover 0.9286 of the following two weeks; comparing timing distributions by the logarithmic score needs 39 casts if the reading is right every time, and 203387 if only half the casts carry real information. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. 摘要:本卷把京房纳甲、八宫世应与六亲的装卦规则,以及全书统一采用的化冲、化合、独发、独静、暗动、日破、伏神飞神、旬空、出空与十二条应期法则,当作确定的算法,用穷举精确计算它们在三枚铜钱起卦中的结构,再把每一种情形的传统读法写成一条商业白话直断与一条情场白话直断。本书不主张任何占验:它精确写出规则、精确计算结构,并说明怎样检验读法本身;断语写得具体,是为了让它可以被证伪。全部计算由书中程序生成;本书不是投资、法律、心理或婚姻建议。 第一章算出:两爻以上齐动的起卦占 0.4661,独静只占 0.0044;化冲只有坤巽互变、中上两爻齐动一条来路,所以单爻永远不化冲,全部动爻中化冲只占 0.0312,带化冲的卦占 0.0233;静爻被日辰冲到时,暗动占 0.0347,日破占 0.0486,二者从不在一卦并见;六十四卦中六亲俱全的只占 0.3125,二十种伏神飞神组合中有两种从不出现;一卦平均点亮 2.9642 个条目,一个都不点亮的只占 0.0090。第二章写多爻同动的二十四格,第三章写暗动与日破,第四章写伏神飞神二十格,第五章写旬空与出空,第六章写十二条应期法则,第七章把几次多爻齐动的起卦从头读到尾。第八章用一次三爻齐动、点亮十个条目的商业起卦示范查法与定期,并给出一份预先登记的预测检验:那一卦适用的应期规则若各允许差一天,便覆盖两周之内 0.9286 的日子;以对数分数比较应期分布,若读法每次都对需要 39 次起卦,若只有一半起卦带着真信息则需要 203387 次。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 Chapter answers: - Ch.1 Method: Three Coins, the Month and the Day, and the Definitions the Whole Book Shares — Chapter 1 finds that casts with two or more moving lines take 0.4661 and a lone still line only 0.0044; a clash has a single route, Kun and Xun changing into each other with the middle and top lines moving, so one moving line never clashes, only 0.0312 of all moving lines turn into a clash, and casts with a clash are 0.0233; of still lines struck by the day, hidden motion takes 0.0347 and day break 0.0486, and the two never share a cast; only 0.3125 of the 64 hexagrams carry all five relations, and two of the twenty hidden-flying pairs never occur; a cast switches on 2.9642 entries on average, and only 0.0090 switch on none. (https://5dlibrary.com/answers/en/liuyao-multi-line-hidden-motion-fushen-void-timing/c01/) 第1章 方法:三枚铜钱、月建日辰与全书共用的定义 — 第一章算出:两爻以上齐动的起卦占 0.4661,独静只占 0.0044;化冲只有坤巽互变、中上两爻齐动一条来路,所以单爻永远不化冲,全部动爻中化冲只占 0.0312,带化冲的卦占 0.0233;静爻被日辰冲到时,暗动占 0.0347,日破占 0.0486,二者从不在一卦并见;六十四卦中六亲俱全的只占 0.3125,二十种伏神飞神组合中有两种从不出现;一卦平均点亮 2.9642 个条目,一个都不点亮的只占 0.0090。 (https://5dlibrary.com/answers/zh/liuyao-multi-line-hidden-motion-fushen-void-timing/c01/) - Ch.2 Several Lines Moving: Clash, Combine, Lone Mover and Lone Still Line — A clash requires exactly two lines of the same trigram to move together, and is impossible with one moving line or with all six; a combine can arise from a single line. Of the 4096 combinations of hexagram and moving lines, 252 contain a clash (0.0615) and 496 a combine (0.1211); weighted by coins, a clash appears in only 0.0233 of casts. Wealth turning into a combine is the most common cell (0.0740), Siblings and the World line turning into a clash the rarest (0.0195). The lone mover and the lone still line are symmetric in combinations but far apart in real casting: 0.3560 against 0.0044. In every row the key line is prosperous in the month about four times in ten. The verdicts and timing hints of the twenty-four cells are interpretations, and this chapter has set out a test that can be pre-registered. (https://5dlibrary.com/answers/en/liuyao-multi-line-hidden-motion-fushen-void-timing/c02/) 第2章 多爻同动:化冲、化合、独发与独静 — 化冲要求同一经卦恰好两爻同动,单爻与六爻全动都不可能;化合一爻即可成立。在 4096 种卦与动爻组合中,化冲 252 种(0.0615),化合 496 种(0.1211);按摇钱权重,化冲只有 0.0233。妻财化合最多(0.0740),兄弟化冲与世爻化冲最少(0.0195)。独发与独静在组合数上对称,在真实摇卦里却悬殊:独发 0.3560,独静 0.0044。键爻在月建中旺相的比例六行都约四成。二十四格的直断与应期都是诠释,本章给出了一项可预先登记的检验。 (https://5dlibrary.com/answers/zh/liuyao-multi-line-hidden-motion-fushen-void-timing/c02/) - Ch.3 Hidden Motion and Day Break: Still Lines Stirred by the Day — Hidden motion and day break both happen on still lines: when the day clashes a line, prosperous means hidden motion and not prosperous means day break. Over 2949120 equally weighted cases, the prosperous share among still lines clashed by the day is 0.4167, so day break is the more common of the two. Cases with at least one line in hidden motion make up 0.0998 and cases with at least one line in day break 0.1391; in hexagrams with no moving line they rise to 0.1910 and 0.2648. Among the twelve cells, Siblings in day break is the most common (0.0323) and Offspring in hidden motion the rarest (0.0155). Hidden motion is an unseen push that answers on a combining day or the line's own day; day break is a wound of the day, filled on the line's own day or restored by a combine. The verdicts of the twelve cells are interpretations, and this chapter has set out a test that can be pre-registered. (https://5dlibrary.com/answers/en/liuyao-multi-line-hidden-motion-fushen-void-timing/c03/) 第3章 暗动与日破:被日辰冲起的静爻 — 暗动与日破都发生在静爻上:日辰冲之,旺相为暗动,不旺相为日破。在 2949120 个等权情形中,被日冲的静爻里旺相的比例是 0.4167,所以日破比暗动常见。至少一爻暗动的情形占 0.0998,至少一爻日破的占 0.1391;六爻皆静的卦更高,分别为 0.1910 与 0.2648。十二格中兄弟日破最多(0.0323),子孙暗动最少(0.0155)。暗动是看不见的推动,应在逢合或值日;日破是当日之伤,待值日填实或逢合补起。十二格的直断是诠释,本章给出了一项可预先登记的检验。 (https://5dlibrary.com/answers/zh/liuyao-multi-line-hidden-motion-fushen-void-timing/c03/) - Ch.4 Flying and Hidden Spirits: Twenty Cells of What Is Missing and What Sits on It — The rule for hidden spirits is a single one: take the missing relation from the same line of the palace's head hexagram, and the cast hexagram's line on top of it is the flying spirit. The 64 hexagrams hold 56 hidden spirits in 44 hexagrams; the missing relation always appears exactly once in the head hexagram, so each hidden spirit has a unique seat, and the top line never hides one. Eighteen of the twenty cells actually occur; Offspring hidden under Parents is the most frequent (8 times), and Siblings hidden under Wealth and Officer hidden under Offspring are zero under this book's rule. The bond between flying and hidden is fixed by the two relations, not by the hexagram. At the day level every hidden spirit is opened and injured on equally many days out of sixty, and true out days take only 0.0899; Offspring hidden under Siblings comes out most easily (0.1333), Parents hidden under Officer least easily (0.0333). A good structure and a way out are two different things; that is the sentence to take from this chapter. Its timing readings are interpretation, and a way to test them has been given above. (https://5dlibrary.com/answers/en/liuyao-multi-line-hidden-motion-fushen-void-timing/c04/) 第4章 飞神伏神二十格:缺席的那一亲藏在谁的身下 — 伏神的规则只有一条:缺席的六亲到本宫首卦同位去取,压在它上面的本卦之爻就是飞神。64 卦里有 56 个伏神,分布在 44 卦;缺席的亲在首卦中永远只出现一次,所以每个伏神的位置唯一,而上爻从不藏伏神。二十格中十八格实际出现,子孙伏父母下最多(8 次),兄弟伏妻财下与官鬼伏子孙下在本书取法下为 0。飞伏之间的生克由两亲决定,与卦无关。日辰层面,所有伏神每六十日被打开与被伤的日子一样多,真正的出日只占 0.0899,其中子孙伏兄弟下最易出(0.1333),父母伏官鬼下最难(0.0333)。结构好与出得来是两回事,这是本章最该带走的一句话;本章的应期读法是诠释,上面给出了检验它的办法。 (https://5dlibrary.com/answers/zh/liuyao-multi-line-hidden-motion-fushen-void-timing/c04/) - Ch.5 Void at Rest and Void in Motion: Twelve Cells of What Is Absent and What Is on Its Way — Void at rest is a still line whose branch falls in the day's void; void in motion is a moving line whose branch does. A hexagram has at most two void lines, cases with at least one take 0.7969, and the world line is void exactly one time in six. Among the relations, Wealth is most often touched by the void yet least often void as a whole; Officer is the opposite. The clashing days are always the wu and ji days of the current xun and the filling days always the jia and yi days of the next; if the cast falls on the ji day or later, the next clash comes later than the fill. Reading the twelve cells comes down to three questions: is the void line the relation sought or the one that harms it, is it still or moving, and on which day is it clashed up or filled. These timings are readings, not confirmed predictions, and the test that would check them has been registered above. (https://5dlibrary.com/answers/en/liuyao-multi-line-hidden-motion-fushen-void-timing/c05/) 第5章 旬空与出空十二格:不在场的事与正在路上的事 — 旬空是静爻的地支落在当日之旬的空亡里,出空是动爻的地支落在空亡里。一卦最多两条空爻,至少一条空爻的情形占 0.7969,世爻落空恰为六分之一。各亲之中妻财最常被碰到空、却最少整亲全空,官鬼正好相反。冲空之日永远是本旬的戊日与己日,填实之日永远是下一旬的甲日与乙日;起卦在己日或更晚,下一次冲空反而比填实晚。十二格的读法归结为三问:空的是用神还是忌神,静而空还是动而空,冲起与填实在哪一天。这些应期是解读而非已证实的预测,检验方案已在上文登记。 (https://5dlibrary.com/answers/zh/liuyao-multi-line-hidden-motion-fushen-void-timing/c05/) - Ch.6 Twelve Timing Rules: the State of the Key Line and the Day It Lands — Before timing, settle four things: which line is the key line, strong or weak, void or not, moving or still. Each of the twelve rules is triggered by one state of the key line and fulfilled by one branch: a prosperous line meets its own branch, a bound line meets its clash, a scattered line meets its combination, a strong void waits to be clashed, a weak void waits to be filled, a moving void waits to leave the xun, a line in its tomb waits for the tomb to be clashed, an excessive line waits to enter its tomb, a weak line waits for birth or peak, a cut-off line waits for its birth branch, a still strong line waits to be clashed, a moving line waits to be combined. p30 shows that the twelve rules cover every case, with 1.8307 rules applying to a key line on average, birth and peak (0.3713) most often. p31 shows that the call chosen by "remove the obstacle first, then take the nearest day" comes 4.35 days after the cast on average, and that a day within 12 days is always available, a property of the branch cycle and no evidence of a hit; what deserves testing is whether, in the 0.3690 of casts where candidate days conflict, the choice rule beats simply taking the earliest day. p32 converts stem-branch days into calendar dates, and Chapter 7 uses it to date four actual casts. (https://5dlibrary.com/answers/en/liuyao-multi-line-hidden-motion-fushen-void-timing/c06/) 第6章 应期十二法:用爻的状态与兑现之日 — 应期之前先定四件事:用爻是哪一爻,旺还是衰,空不空,动还是静。十二法各由一个用爻状态触发、由一个地支兑现:旺相逢值,合住逢冲,冲散逢合,旺空待冲,衰空等填实,动空待出旬,入墓待冲墓,太过待入墓,休囚待生旺,临绝待长生,静而旺者待冲,动者待合。p30 显示十二法覆盖了所有情形,平均每个用爻同时有 1.8307 条法则成立,最常见的是生旺(0.3713)。p31 显示按先去病、后取近定出的应期平均在 4.35 天之后,并且 12 天之内必然有日子可报,这是地支轮回的性质,不是命中的证据;真正值得检验的,是候选日子互相冲突的那 0.3690 的卦里,取舍规则是否比随手取最早一天更准。p32 把干支日换成公历日,第七章用它给四个实际起卦写出日期。 (https://5dlibrary.com/answers/zh/liuyao-multi-line-hidden-motion-fushen-void-timing/c06/) - Ch.7 Four Casts Read Through: Several Lines Moving, from Layout to Landing Day — Reading a cast takes seven steps: write the question, fix the key line, lay out the cast, list every key, read them one by one, settle the four facts of the key line and run the twelve rules, and fix the day by removing the obstacle first and then taking the nearest day. In the four demonstration casts, the money of cast one is scattered by a clash and waits for the combination day, 20 November 2026; the key line of cast two is hidden but without obstacle, so the earliest birth-and-peak day is taken, 25 December; the lease of cast three, the sole still line among five moving ones, is void and waits for the filling day, 26 March 2027; the key line of cast four has entered its tomb with the day and waits for the day that clashes the tomb, 25 May. In every cast the cross-check lines give an earlier day that tells you whom to deal with before the call. These dates are readings derived from rules, not prophecies; p39 shows that only when at least 16 exact-day hits appear among 120 or more pre-registered casts does the timing beat the calendar itself. (https://5dlibrary.com/answers/en/liuyao-multi-line-hidden-motion-fushen-void-timing/c07/) 第7章 四卦实读:多爻齐动,从排卦到落日 — 一卦读完要经过七步:写问题、定用爻、装卦、列出全部条目、逐条读、判定用爻四件事并套十二法、按先去病后取近定日。四个示范起卦里,卦一的钱化冲散局,等逢合之日 2026 年 11 月 20 日;卦二的用爻伏而无病,取最早的生旺之日 12 月 25 日;卦三五爻乱动、独静的租约落空,等填实之日 2027 年 3 月 26 日;卦四的用爻随日入墓,等冲墓之日 5 月 25 日。每一卦的旁证爻都给出一个更早的日子,告诉你在应期之前该先处理谁。这些日期是规则推出的读法,不是预言;p39 说明,只有在 120 个以上预先登记的卦里恰好命中当日的次数达到 16 次,应期才算胜过了日历本身。 (https://5dlibrary.com/answers/zh/liuyao-multi-line-hidden-motion-fushen-void-timing/c07/) - Ch.8 Practice: One Cast with Three Moving Lines, the Ten Entries It Switches On, Its Timing, and a Pre-registered Forecasting Test — Chapter 8 demonstrates the look-up and the timing on a business cast with three moving lines that switches on ten entries, and gives a pre-registered forecasting test: with one day of tolerance each, the timing rules that apply to that cast cover 0.9286 of the following two weeks; comparing timing distributions by the logarithmic score needs 39 casts if the reading is right every time, and 203387 if only half the casts carry real information. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/liuyao-multi-line-hidden-motion-fushen-void-timing/c08/) 第8章 实践:一次三爻齐动的起卦、它点亮的十个条目、应期,与一份预先登记的预测检验 — 第八章用一次三爻齐动、点亮十个条目的商业起卦示范查法与定期,并给出一份预先登记的预测检验:那一卦适用的应期规则若各允许差一天,便覆盖两周之内 0.9286 的日子;以对数分数比较应期分布,若读法每次都对需要 39 次起卦,若只有一半起卦带着真信息则需要 203387 次。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 (https://5dlibrary.com/answers/zh/liuyao-multi-line-hidden-motion-fushen-void-timing/c08/) ## 5DL-B-126 · The Human Design, Tzolkin and Seventy-Eight Tarot Projection Codex 中文书名:人类图通道、玛雅卓尔金历与七十八塔罗投射母典 DOI: pending Cite: Fan, Y. (2026). The Human Design, Tzolkin and Seventy-Eight Tarot Projection Codex (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All calculations are generated by the book's programs; it is not investment, legal, medical or psychological advice. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. Abstract: This volume reads Human Design, the Maya Tzolkin and the seventy-eight tarot cards as three projective decision instruments: a type, a seal or a card is a screen onto which a person projects their situation and with whose help they put a choice into words. The book states the rules of the three systems precisely, computes their structure exactly with the programs printed in it, and writes the thirty-six channels, the nine centres defined and open, the five types, the twelve profiles, the twenty seals, the thirteen tones and the seventy-eight cards one by one as sharp readings: gift and shadow, or upright, reversed and projection, each ending with a decision rule that can be checked afterwards. The accuracy of these readings has not been established, and the book does not claim it; it shows how to test them. All calculations are generated by the book's programs; it is not investment, legal, medical or psychological advice. Chapter 1 is the method: each gate of the gate wheel spans 5.6250 degrees and each line 0.9375 degrees, and 88 degrees equal 93.8667 lines, so there are only 12 profiles, the six common ones at 0.1444 each and the six rare ones at 0.0222 each; the type shares the author's reading station measured on 11305 birth moments are Generator 0.3711, Manifesting Generator 0.3236, Projector 0.2153, Manifestor 0.0798 and Reflector 0.0103. The Tzolkin is two coprime gears of thirteen and twenty, and the traditional count (GMT 584283) and the Dreamspell count agree on not one of the 36525 days from 1950 to 2049, so every person has two kins. At least one major appears in a three-card draw with probability 0.6356, and a 78-card deck needs 10 riffle shuffles before its total-variation distance from random falls to 0.0780. Chapter 2 covers the thirty-six channels, Chapter 3 the eighteen cases of the nine centres, the five types and the twelve profiles, Chapter 4 the twenty seals and thirteen tones, and Chapters 5 to 7 the seventy-eight cards. Chapter 8 runs a sample chart that contains no birth data through the station, finds a Generator with single definition, profile 5/1, five conducting channels and Dreamspell kin 30, draws the Lovers reversed, the Hanged Man reversed and the King of Pentacles from a stated seed, turns them into a contract rule (write the exit clause first, negotiate the term afterwards), and gives a pre-registered blind-matching test with Barnum controls: with three choices and a true hit rate of 0.40, it needs 447 people. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. 摘要:本卷把人类图、玛雅卓尔金历与七十八张塔罗当作三件投射式的决策工具来读:一个类型、一个印记、一张牌,是人把自己的处境投上去、再借它把选择说清楚的屏幕。本书精确写出三套系统的规则,用书中印出的程序精确计算它们的结构,再把三十六条通道、九个中心的定义与开放、五个类型、十二种人生角色、二十个印记、十三个调性与七十八张牌逐一写成锋利的读本:天赋与阴影,或正位、逆位与投射,每一段都以一条事后可以核对的决策规则收尾。这些读法的准确性没有得到证实,本书也不主张它;本书说明怎样检验。全部计算由书中程序生成;本书不是投资、法律、医疗或心理建议。 第一章是方法:闸门轮每个闸门 5.6250 度、每条爻线 0.9375 度,88 度等于 93.8667 条爻线,所以人生角色只有 12 种,常见的六种各占 0.1444,少见的六种各占 0.0222;作者测算站在 11305 个出生时刻上实测的类型比例是生产者 0.3711、显示生产者 0.3236、投射者 0.2153、显示者 0.0798、反映者 0.0103。卓尔金历是十三与二十两个互素的齿轮,传统计数(GMT 584283)与 Dreamspell 计数在 1950 到 2049 年的 36525 天里没有一天相同,所以每个人都有两个 kin。三张牌里至少一张大牌的概率是 0.6356,七十八张牌要鸽尾洗牌 10 次,与完全随机的总变差距离才降到 0.0780。第二章写三十六条通道,第三章写九个中心的十八种情形、五个类型与十二种人生角色,第四章写二十个印记与十三个调性,第五至第七章写七十八张牌。第八章把一张不含任何出生资料的样盘送进测算站,得到单一定义的生产者、人生角色 5/1、五条导通通道与 Dreamspell kin 30,按写明的种子抽出恋人逆位、倒吊人逆位与星币国王,把它们落成一条先写退出条款、再谈租期的合约规则,并给出一份带巴纳姆对照的预先登记盲选检验:三选一、真实命中率 0.40 时需要 447 人。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 Chapter answers: - Ch.1 Method: The Gate Wheel, the 260 Days and the 78 Cards — Chapter 1 is the method: each gate of the gate wheel spans 5.6250 degrees and each line 0.9375 degrees, and 88 degrees equal 93.8667 lines, so there are only 12 profiles, the six common ones at 0.1444 each and the six rare ones at 0.0222 each; the type shares the author's reading station measured on 11305 birth moments are Generator 0.3711, Manifesting Generator 0.3236, Projector 0.2153, Manifestor 0.0798 and Reflector 0.0103. The Tzolkin is two coprime gears of thirteen and twenty, and the traditional count (GMT 584283) and the Dreamspell count agree on not one of the 36525 days from 1950 to 2049, so every person has two kins. At least one major appears in a three-card draw with probability 0.6356, and a 78-card deck needs 10 riffle shuffles before its total-variation distance from random falls to 0.0780. (https://5dlibrary.com/answers/en/human-design-channels-tzolkin-tarot-78-projection/c01/) 第1章 方法:闸门轮、二百六十日与七十八张牌 — 第一章是方法:闸门轮每个闸门 5.6250 度、每条爻线 0.9375 度,88 度等于 93.8667 条爻线,所以人生角色只有 12 种,常见的六种各占 0.1444,少见的六种各占 0.0222;作者测算站在 11305 个出生时刻上实测的类型比例是生产者 0.3711、显示生产者 0.3236、投射者 0.2153、显示者 0.0798、反映者 0.0103。卓尔金历是十三与二十两个互素的齿轮,传统计数(GMT 584283)与 Dreamspell 计数在 1950 到 2049 年的 36525 天里没有一天相同,所以每个人都有两个 kin。三张牌里至少一张大牌的概率是 0.6356,七十八张牌要鸽尾洗牌 10 次,与完全随机的总变差距离才降到 0.0780。 (https://5dlibrary.com/answers/zh/human-design-channels-tzolkin-tarot-78-projection/c01/) - Ch.2 The Thirty-Six Channels: Fixed Lines of Energy Once Two Gates Connect — This chapter has gone through the thirty-six channels one by one. Each channel is connected by two gates, joins two centres and belongs to one group: Individual, Collective, Tribal or Integration (Ra 2011). Sampling 11305 birth moments from 1950 to 2010 with a self-contained ecliptic model, the definition rates run from 5.97% for 35-36 to 15.79% for 37-40, and a chart connects 3.57 channels on average. In graph terms the 36 channels are a complete graph on gates 10, 20, 34 and 57 plus 30 one-to-one pairings; only four channels, 12-22, 20-34, 21-45 and 35-36, wire a motor directly to the throat. For every channel the structure, the gift, the shadow and the decision rules for money, partners and commitments have been written out. These rules are interpretations within the system; the chapter gives a pre-registered test with a Barnum control, and the sample sizes it needs. (https://5dlibrary.com/answers/en/human-design-channels-tzolkin-tarot-78-projection/c02/) 第2章 三十六条通道:两个闸门接通之后的固定能量线 — 本章逐条写了三十六条通道。每条通道由两个闸门接通,连接两个中心,属于个体人、集体人、部落或整合中的一组 (Ra 2011)。用自给自足的黄道模型对 1950 到 2010 年的 11305 个出生时刻抽样,定义率从 35-36 的 5.97% 到 37-40 的 15.79% 不等,平均每张盘接通 3.57 条。图论上,36 条通道是闸门 10、20、34、57 上的一个完全图加 30 条一对一的配对;只有 12-22、20-34、21-45、35-36 四条把动力直接接到喉咙。每条通道都写出了结构、天赋、阴影,以及钱、合伙人与承诺上的决策规则。这些规则是体系内的诠释;本章给出了一项带巴纳姆对照的预先登记检验,以及所需的样本量。 (https://5dlibrary.com/answers/zh/human-design-channels-tzolkin-tarot-78-projection/c02/) - Ch.3 Nine Centres, Five Types and Twelve Profiles — The definition of the nine centres is decided by the channels; Sacral and Throat are defined most often, Head least. The five types are fixed entirely by how the centres connect, and the station's base rates can be reproduced almost exactly on a light sky. The twelve profiles are locked by the 88-degree design offset, half common and half rare. The thirty-five readings give rules and ways of deciding, for readers to compare against their own real decisions, not a label to accept. (https://5dlibrary.com/answers/en/human-design-channels-tzolkin-tarot-78-projection/c03/) 第3章 九中心、五类型与十二人生角色 — 九个中心的定义由通道决定,荐骨与喉咙被定义得最多,头脑最少;五种类型完全由中心的连通方式判定,本站基率在轻量天空上可以几乎原样重现;十二种人生角色由 88 度的设计偏移锁定,一半常见、一半稀有。三十五节读法给出的是规则和决策方式,读者用来对照自己的真实决策,而不是用来接受一个标签。 (https://5dlibrary.com/answers/zh/human-design-channels-tzolkin-tarot-78-projection/c03/) - Ch.4 Twenty Seals and Thirteen Tones: One Table, Two Ways of Counting — Because they are coprime, twenty day names and thirteen numbers combine into 260 kins; each seal opens one thirteen-day wave and gains 7 in its number at every return. The Dreamspell stops on 29 February and so drifts ever further from the traditional highland count: between 1900 and 2099 the two give the same seal on only 2922 days, all within two four-year windows in the early 1960s and the early 2040s, and the same kin on not a single day. The four oracle positions are additions and subtractions mod 20; they come from the Dreamspell and do not belong to the traditional count. Each of the twenty seals and thirteen numbers is read with a gift, a shadow and one decision rule that can be carried out; whether they hold for a given person is for the three-way blind test at the end to answer. (https://5dlibrary.com/answers/en/human-design-channels-tzolkin-tarot-78-projection/c04/) 第4章 二十印记与十三调性:一张表,两种数法 — 二十个日名与十三个数字因为互素而组合成 260 个 kin,每个印记开启一个十三日波,每回来一次数字加 7。Dreamspell 在 2 月 29 日停跳,因而与高地的传统计数渐行渐远:1900 年到 2099 年间,两者给出同一个印记的只有 2922 天,全在 1960 年代初与 2040 年代初的两段四年窗口里,同一个 kin 一天也没有。神谕的四个位置是模 20 的加减,出自 Dreamspell,不属于传统计数。二十个印记与十三个数字的读法各有天赋、阴影与一条可执行的决策规则;它们是否对本人成立,要靠末尾那个三选一的盲测来回答。 (https://5dlibrary.com/answers/zh/human-design-channels-tzolkin-tarot-78-projection/c04/) - Ch.5 The Twenty-Two Majors: Upright, Reversed and Projected — This chapter wrote out the 22 majors one by one in the engine's numbering, each in five parts: structure, upright, reversed, projection and decision rule. p25 gives the base rates: one named major appears in a ten-card spread with probability 0.1282, but a ten-card spread holds at least one major with probability 0.9717, and 2.8205 on average. p26 simulates orientation under the engine's riffle shuffle: the reversed share is 0.4987, the variance of the number reversed among ten is 2.5663, almost the same as the independent model, and reversals do not come in runs. p27 shows why the feeling "there it is again" is so common: over five readings, the chance that some major appears three or more times is 0.3220, while for one major named in advance it is only 0.01723. The projection parts ask you, before pinning a card on someone else, to check whether it is you; the decision rules ask that every reading end in one thing you can write down clearly and check later. (https://5dlibrary.com/answers/en/human-design-channels-tzolkin-tarot-78-projection/c05/) 第5章 二十二张大牌:正向、逆向与投射 — 本章按引擎的编号把 22 张大牌逐张写完,每张五段:结构、正向、逆向、投射、决策规则。p25 给出基率:一张指定的大牌出现在十张牌阵里的概率是 0.1282,但十张牌阵至少有一张大牌的概率是 0.9717,平均 2.8205 张。p26 按引擎的鸽尾洗牌模拟了正逆:倒着的比例 0.4987,十张里倒着张数的方差 2.5663,与独立模型几乎相同,逆向不成串。p27 说明「它又来了」的感觉为什么常见:连读五次,总有某张大牌出现三次以上的概率是 0.3220,而事先指定的那一张只有 0.01723。投射段要求你在把一张牌安到别人身上之前,先检查一遍那是不是你自己;决策规则要求每一次读牌都落到一件写得清、查得到的事上。 (https://5dlibrary.com/answers/zh/human-design-channels-tzolkin-tarot-78-projection/c05/) - Ch.6 Wands and Cups: Twenty-Eight Cards of Fire for Action and Water for Ties — This chapter has read the twenty-eight wands and cups. Each minor card is built from two parts: the suit gives the domain (wands action and enterprise, cups feeling and relationship), the rank gives how far the force has travelled, and Waite's picture then separates the same rank in the two suits. Exact enumeration shows that the expected number of wands in a ten-card spread is 1.7949 and three or more occur 0.2531 of the time, so "three wands" is ordinary; the engine favours the earlier element when counts tie, reporting fire as dominant 0.3549 of the time, 2.0679 times earth's 0.1716. Court cards almost always appear: the chance of at least one in ten cards is 1 minus 0.0854. Every card has its structure, upright, reversed, projection and decision rule; the projection paragraph pulls "who is this card about" back from the other person to yourself. The chapter registers two matching tests and gives the sample sizes they need. (https://5dlibrary.com/answers/en/human-design-channels-tzolkin-tarot-78-projection/c06/) 第6章 权杖与圣杯:行动之火与关系之水的二十八张 — 本章读完了权杖与圣杯二十八张。每张小牌由两部分拼成:花色给出领域(权杖是行动与事业,圣杯是情感与关系),位阶给出这股力走到哪一步,韦特的画面再把同一个位阶在两种花色里分开。精确枚举表明,十张牌阵里权杖张数的期望是 1.7949,三张或以上占 0.2531,所以「三张权杖」只是常态;引擎在并列时偏向排在前面的元素,火被报告为主导元素的比例 0.3549,是土 0.1716 的 2.0679 倍。宫廷牌几乎总会出现:十张里至少一张的机会是 1 减去 0.0854。每一张牌都写了结构、正向、逆向、投射与决策规则;投射一段把「这张牌说的是谁」从别人身上拉回自己。本章给出两项可以预先登记的匹配检验与所需的样本量。 (https://5dlibrary.com/answers/zh/human-design-channels-tzolkin-tarot-78-projection/c06/) - Ch.7 Swords and Pentacles: Twenty-Eight Cards of Air for Judgement and Earth for Money — This chapter has read the twenty-eight swords and pentacles. The swords bring their force down on judgement and, the further along, more easily become thoughts that trap the thinker; the pentacles bring theirs down on money, goods, craft and the body, slow and honest. Exact calculation shows that a ten-card spread repeats at least one rank 0.7519 of the time and that a sword and pentacle of the same rank meet 0.1967 of the time, so "the same number" is no omen in itself; the number of reversals depends on shuffling habit, and with the same net-polarity verdict "push ahead" is 0.2137 under the one-half model and 0.4381 under the one-third model. Every card has its structure, upright, reversed, projection and decision rule. The chapter registers a blind judging test of the upright and reversed paragraphs, with a control arm, and states how many draws it needs. (https://5dlibrary.com/answers/en/human-design-channels-tzolkin-tarot-78-projection/c07/) 第7章 宝剑与星币:判断之风与金钱之土的二十八张 — 本章读完了宝剑与星币二十八张。宝剑把力落在判断上,越往后越容易变成困住自己的念头;星币把力落在钱、货、手艺与身体上,慢而诚实。精确计算表明,十张牌阵里至少有一个位阶重复出现的情形占 0.7519,宝剑与星币同位阶相遇占 0.1967,所以「同数字」本身不是征兆;逆向的多少取决于洗牌习惯,同样用净极性速断,「可以推」在二分之一模型下是 0.2137,在三分之一模型下是 0.4381。每一张牌都写了结构、正向、逆向、投射与决策规则。本章登记了一项正逆两段文字的盲评检验,带一组对照,并给出了所需的抽牌次数。 (https://5dlibrary.com/answers/zh/human-design-channels-tzolkin-tarot-78-projection/c07/) - Ch.8 Practice: Sample Chart A Through the Station, Three Cards, and a Blind-Matching Test with Barnum Controls — Chapter 8 runs a sample chart that contains no birth data through the station, finds a Generator with single definition, profile 5/1, five conducting channels and Dreamspell kin 30, draws the Lovers reversed, the Hanged Man reversed and the King of Pentacles from a stated seed, turns them into a contract rule (write the exit clause first, negotiate the term afterwards), and gives a pre-registered blind-matching test with Barnum controls: with three choices and a true hit rate of 0.40, it needs 447 people. Marked boxes separate published evidence, runnable Python calculations, classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/human-design-channels-tzolkin-tarot-78-projection/c08/) 第8章 实践:样盘 A 过一遍工作站,三张牌,以及带巴纳姆对照的盲选检验 — 第八章把一张不含任何出生资料的样盘送进测算站,得到单一定义的生产者、人生角色 5/1、五条导通通道与 Dreamspell kin 30,按写明的种子抽出恋人逆位、倒吊人逆位与星币国王,把它们落成一条先写退出条款、再谈租期的合约规则,并给出一份带巴纳姆对照的预先登记盲选检验:三选一、真实命中率 0.40 时需要 447 人。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。 (https://5dlibrary.com/answers/zh/human-design-channels-tzolkin-tarot-78-projection/c08/) ## 5DL-B-127 · The Name-Number, Digit-Field and Brand-Naming Codex 中文书名:姓名数理、数字磁场与品牌改运命名母典 DOI: pending Cite: Fan, Y. (2026). The Name-Number, Digit-Field and Brand-Naming Codex (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: It is not legal, investment or other professional advice. Abstract: This volume reads Chinese name numerology (five grids, eighty-one numbers, three-talent triads, sound and glyph) and the eight digit-pair stars of phone and door numbers as a checklist for naming. It states the rules of the fleet-one name and number stations exactly, computes with printed Python programs how they fall across random names and numbers, and writes every number, triad, grid, tone pattern, glyph structure and star as a reading for business owners: business effect, hidden friction, naming rule and a decision rule that can be checked later. Whether these readings describe a person or foretell a business has not been established, and the book does not claim it; where names act through fluency, dialect sound-alikes and shared cultural preferences, published evidence exists and the book treats the mechanism as real. It is not legal, investment or other professional advice. Chapter 1 is the method: the character fan has stroke counts of 11, 12 and 15, and 0.7375 of neighbouring numbers differ in grade; among 2,536,800 equally weighted names, 0.8165 have at least one bad key grid and only 0.1063 have all four good, with a mean engine score of 59.90; two digit-to-phase mappings disagree on 0.456 of triads; and detecting a survival gap of 0.50 against 0.55 would take 3,130 shops. Chapters 2 and 3 read the eighty-one numbers: between 2 and 39 a one-stroke error flips the grade with mean chance 0.684, and three-character totals never pass 62, so 41 to 81 is brand territory, holding 71.64% of four-character strings. Chapter 4 covers the five grids and 125 triads; the engine's full filter for suggested names passes a median 0.074 of given-name pairs. Chapter 5 covers sixteen tone patterns (only fifteen audible melodies), eight glyph structures and three sound-alike gates; a heard two-character name has 2.57 written forms on average. Chapter 6 reads the plain, hidden and shown forms of the eight stars: random numbers average 48.13 points, and swapping the last four digits moves one number by 71.0 points from best to worst. Chapter 7 gives eleven brand-naming rules: the numerology filter passes about half (0.5436 with the trade character fang); the real clash risk is sound, at least 0.1175 with 20,000 marks in a class and 0.8647 when names crowd onto 100 favoured syllables; lowest-cost rejections first cut hours per candidate from 1.019 to 0.812. Chapter 8 runs a fictional founder through everything: the rename list passes 482 of 7,140 combinations with only 10 different first characters in its top twenty; every number in her block carries a malign pair; and four-choice blind matching at a true hit rate of 0.32 needs 359 people. 摘要:本书把姓名五格与八十一数、三才、音律字形,以及手机号与门牌上的八星数字,当作一张命名与改号的检查表来读。它精确写出第一舰队姓名站与数字站引擎实际使用的规则,用书中印出的 Python 程序计算这些规则在随机名字与随机号码上的分布,再把八十一个数、一百二十五种三才配置、五格分工、十六种声调组合、八种字形结构与八星的本位、隐、显逐条写成老板视角的读本:生意上的表现、人际暗伤、命名法则,以及一条事后可以核对的决策规则。这些读法能否描述一个人、能否预示一门生意的成败,没有得到证实,本书也不主张;名字与数字通过念起来是否流畅、方言谐音与共同的文化偏好影响生意,这一部分有已发表的证据,本书当作真实机制来写,其余一律放进假说框并写明怎样证伪。全部计算由书中程序生成;本书不是法律、投资或任何专业意见。 第一章是方法:同一个「范」字有 11、12、15 画三种数法,八十一数相邻两数等级不同的占 0.7375;在 2,536,800 个等权的三字名里,人格、地格、总格、外格至少一格凶的占 0.8165,四格全吉的只占 0.1063,引擎综合分平均 59.90;三才的两种尾数配法在 0.456 的组合上判词相反;若改名的作用只是三年存活率 0.50 对 0.55,检验就需要 3,130 家门店。第二、三章写八十一数:2 到 39 之间,多算或少算一笔使等级翻转的平均机会是 0.684;三字名的总格最大只到 62,41 到 81 主要属于品牌全称,四字串里有 71.64% 落在这一段。第四章写五格分工与三才一百二十五配置,引擎推荐名字时的完整筛选,通过率的中位数只有 0.074。第五章写十六种声调组合(听起来只有十五种旋律)、八种字形结构与粤语、闽南语、英文三道谐音关,一个听到的两字名平均有 2.57 种同音写法。第六章写八星数字的本位、隐与显:随机号码平均 48.13 分,同一个号码头换最后四位,最好与最差相差 71.0 分。第七章写十一条品牌命名法则:品牌模式的数理过滤只淘汰一半上下的候选,配「坊」字时通过 0.5436;真正的撞名风险在读音,同类别已有 20,000 个商标时,按不计声调的读音估算撞名概率至少 0.1175,名字挤在 100 个热门音节上时升到 0.8647;把淘汰成本低的关放在前面,每个候选的平均工时从 1.019 小时降到 0.812 小时。第八章让一位虚构的创业者走完全程:改名清单在 7,140 种组合里只通过 482 种,前二十名只用了 10 个不同的首字;她的虚构号段里每一个号码都含一对凶星;四选一盲选在真实命中率 0.32 时需要 359 人。 Chapter answers: - Ch.1 Method: Kangxi Strokes, the Five-Grid Formulas, the Cycle Above 81 and Base Rates — Chapter 1 is the method: the character fan has stroke counts of 11, 12 and 15, and 0.7375 of neighbouring numbers differ in grade; among 2,536,800 equally weighted names, 0.8165 have at least one bad key grid and only 0.1063 have all four good, with a mean engine score of 59.90; two digit-to-phase mappings disagree on 0.456 of triads; and detecting a survival gap of 0.50 against 0.55 would take 3,130 shops. (https://5dlibrary.com/answers/en/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c01/) 第1章 方法:康熙笔画、五格公式、八十一循环与基础率 — 第一章是方法:同一个「范」字有 11、12、15 画三种数法,八十一数相邻两数等级不同的占 0.7375;在 2,536,800 个等权的三字名里,人格、地格、总格、外格至少一格凶的占 0.8165,四格全吉的只占 0.1063,引擎综合分平均 59.90;三才的两种尾数配法在 0.456 的组合上判词相反;若改名的作用只是三年存活率 0.50 对 0.55,检验就需要 3,130 家门店。 (https://5dlibrary.com/answers/zh/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c01/) - Ch.2 The Eighty-One Numbers, Part One: 1 to 40 — Chapters 2 and 3 read the eighty-one numbers: between 2 and 39 a one-stroke error flips the grade with mean chance 0.684, and three-character totals never pass 62, so 41 to 81 is brand territory, holding 71.64% of four-character strings. (https://5dlibrary.com/answers/en/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c02/) 第2章 八十一数之一:从 1 到 40 — 第二、三章写八十一数:2 到 39 之间,多算或少算一笔使等级翻转的平均机会是 0.684;三字名的总格最大只到 62,41 到 81 主要属于品牌全称,四字串里有 71.64% 落在这一段。 (https://5dlibrary.com/answers/zh/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c02/) - Ch.3 The Eighty-One Numbers, Part Two: 41 to 81 — The previous chapter covered 1 to 40; this one covers 41 to 81. The rules are still the tollbooth engine's: whenever any of the five grids comes out above 81, subtract 80 repeatedly until it is no larger than 81, so a raw total of 121 reads as 41 and 161 reads as 81. The three classes, good, half-good and bad, are copied from the engine's two sets, and every number in neither set counts as bad. How radicals are restored, how the five grids are summed, where this reading comes from and where its credibility ends were all set out in Chapter 1, so one reminder is enough here: the Structure paragraph of each section states facts you can check item by item, and the remaining paragraphs are this book's reading for people who run a business. None of it has been validated as a prediction. (https://5dlibrary.com/answers/en/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c03/) 第3章 八十一数之二:从四十一到八十一 — 上一章写完了 1 到 40,本章接着写 41 到 81。规则仍按收费站引擎:五格里任何一格算出来超过 81,就连续减 80,直到不大于 81 为止,所以原始总数 121 读作 41,161 读作 81;吉、半吉、凶三类也照抄引擎里的两张集合,不在集合里的一律算凶。部首怎样还原、五格怎样相加、这套读法从哪里来、可信度的边界在哪里,第一章已经交代清楚,这里只提醒一句:每一节的「结构」段是可以逐项核对的事实,其余各段是本书写给经营者的读法,没有被验证为预测。 (https://5dlibrary.com/answers/zh/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c03/) - Ch.4 The 125 Three-Talent Configurations and the Division of Labour among the Five Grids — Chapter 4 covers the five grids and 125 triads; the engine's full filter for suggested names passes a median 0.074 of given-name pairs. (https://5dlibrary.com/answers/en/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c04/) 第4章 三才一百二十五配置与五格分工 — 第四章写五格分工与三才一百二十五配置,引擎推荐名字时的完整筛选,通过率的中位数只有 0.074。 (https://5dlibrary.com/answers/zh/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c04/) - Ch.5 Sound and Shape: Sixteen Tone Patterns, Eight Glyph Structures and Three Homophone Checks — Chapter 5 covers sixteen tone patterns (only fifteen audible melodies), eight glyph structures and three sound-alike gates; a heard two-character name has 2.57 written forms on average. (https://5dlibrary.com/answers/en/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c05/) 第5章 音律与字形:十六种声调组合、八种字形结构与三道谐音关 — 第五章写十六种声调组合(听起来只有十五种旋律)、八种字形结构与粤语、闽南语、英文三道谐音关,一个听到的两字名平均有 2.57 种同音写法。 (https://5dlibrary.com/answers/zh/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c05/) - Ch.6 The Eight Stars of Digit Pairs: Sixty-Four Pairs, Plain, Hidden and Shown, and the Weight of the Tail — Chapter 6 reads the plain, hidden and shown forms of the eight stars: random numbers average 48.13 points, and swapping the last four digits moves one number by 71.0 points from best to worst. (https://5dlibrary.com/answers/en/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c06/) 第6章 八星数字磁场:六十四个数字对、本位隐显与尾数加权 — 第六章写八星数字的本位、隐与显:随机号码平均 48.13 分,同一个号码头换最后四位,最好与最差相差 71.0 分。 (https://5dlibrary.com/answers/zh/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c06/) - Ch.7 Brand-Naming Rules: Trade Word, Trademark, Sound-Alikes, Domains, Dialects, Recall and Length — Chapter 7 gives eleven brand-naming rules: the numerology filter passes about half (0.5436 with the trade character fang); the real clash risk is sound, at least 0.1175 with 20,000 marks in a class and 0.8647 when names crowd onto 100 favoured syllables; lowest-cost rejections first cut hours per candidate from 1.019 to 0.812. (https://5dlibrary.com/answers/en/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c07/) 第7章 品牌命名法则:行业字、商标、同音、域名、谐音、记忆与长度 — 第七章写十一条品牌命名法则:品牌模式的数理过滤只淘汰一半上下的候选,配「坊」字时通过 0.5436;真正的撞名风险在读音,同类别已有 20,000 个商标时,按不计声调的读音估算撞名概率至少 0.1175,名字挤在 100 个热门音节上时升到 0.8647;把淘汰成本低的关放在前面,每个候选的平均工时从 1.019 小时降到 0.812 小时。 (https://5dlibrary.com/answers/zh/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c07/) - Ch.8 Practice: A Fictional Founder's Name, Studio Brand and Phone Number, and Three Pre-Registered Tests — Chapter 8 runs a fictional founder through everything: the rename list passes 482 of 7,140 combinations with only 10 different first characters in its top twenty; every number in her block carries a malign pair; and four-choice blind matching at a true hit rate of 0.32 needs 359 people. (https://5dlibrary.com/answers/en/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c08/) 第8章 实战:一位虚构创业者的名字、工作室品牌与手机号,以及三项预先登记的检验 — 第八章让一位虚构的创业者走完全程:改名清单在 7,140 种组合里只通过 482 种,前二十名只用了 10 个不同的首字;她的虚构号段里每一个号码都含一对凶星;四选一盲选在真实命中率 0.32 时需要 359 人。 (https://5dlibrary.com/answers/zh/name-numerology-81-sancai-phonetic-glyph-eight-star-digit-fields-brand-naming/c08/) ## 5DL-B-128 · The Life-Path, Lenormand-Pair and Horary-Timing Codex 中文书名:卡巴拉生命灵数、雷诺曼三十六牌组合与古典卜卦占星母典 DOI: pending Cite: Fan, Y. (2026). The Life-Path, Lenormand-Pair and Horary-Timing Codex (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Abstract: This volume treats three Western divination methods as tools that make a business owner state a question and turn a decision into contract clauses: the life path and personal year added up from a birth date, pairs from the thirty-six Lenormand cards, and classical horary, cast for the moment a question is asked. It states the rules exactly, computes their structure and base rates with printed programs, and writes the twelve life paths, nine missing numbers, nine personal years, 210 anchor-plus-modifier pairs, twelve houses, six considerations before judgement, seven modes of perfection and nine timing units as readings for owners, each ending in a checkable decision rule. No link to real outcomes has been established or is claimed here; the book shows how to test for one. It is not investment, legal, tax, medical or psychological advice. Chapter 1: a life path keeps a birth date modulo 9, cutting 14.42 bits to 3.42, and master numbers depend on the order of addition; the 21,915 days of 1950 to 2009 make only 220 birth grids; two cards give 630 symmetric or 1260 directional readings; in a stated sky model the station judges only 0.1356 of contract charts "leans yes"; a remark true of eight people in ten has a likelihood ratio of at most 1.25. Chapter 2: under the station's rule, life path 2 takes only 0.0149 of births and 33 takes 0.0594 (0.0073 when parts are reduced first); people miss 4.489 digits on average, and 0.7029 miss the 3. Chapter 3: calendar and birthday conventions disagree on 0.4986 of days, a three-year contract crosses step 9 with probability 0.4441, and on random revenue a "harvest year" comes true 0.7031 of the time loosely and 0.1103 strictly. Chapters 4 and 5: a named modifier touches a named anchor in a Grand Tableau with probability 0.1698, some risk card sits beside some anchor in 0.9985 of layouts, and the station's "close" verdict fires in 0.8084. Chapter 6: 0.5355 of question moments trip at least one consideration. Chapter 7: direct perfection occurs in 0.2848 of contract questions, uninterrupted perfection in 0.0910 or 0.1922 depending on whether the Moon may prohibit, and picking the timing unit after the fact lifts chance hits from 0.0730 to 0.4790. Chapter 8 walks a fictional studio owner and hotel contract through the three lines: life path 5, personal-year step 8, the pairs 25+32 and 34+22, a station score of 8 ("leans yes"), and a cash table turning them into payment milestones, a checkpoint and an intermediary clause worth RM 243,000 at the checkpoint. All three lines say yes by chance 0.0499 of the time, so agreement is not evidence; a real test needs a pre-registered blind judging of at least 213 contract pairs. 摘要:本卷把由出生日期加出来的生命灵数与个人年、三十六张雷诺曼牌两两相连的双牌,以及以问题被提出那一刻起盘的古典卜卦,当作帮助经营者把问题写清、把决定落成合约条款的三件工具。本书精确写出三套规则,用书中印出的程序计算它们的结构与基率,再把十二种生命灵数、九种缺数、九个个人年、两百一十组「锚牌+修饰牌」双牌、卜卦十二宫、六条判前慎审、七种成局方式与九种应期逐条写成老板视角的读本,每一条都以一条事后可以核对的决策规则收尾。这些读法与现实结果之间的对应没有得到证实,本书也不主张它,只说明怎样检验。本书不是投资、法律、税务、医疗或心理建议。 第一章是方法:生命灵数只保留出生日期除以 9 的余数,信息量从 14.42 比特降到 3.42 比特,主数取决于加法顺序;1950 到 2009 年的 21,915 个出生日只有 220 种九宫格;两张雷诺曼牌在对称约定下有 630 种读法、方向约定下有 1260 种;在写明假设的天空模型里,卜卦站只把 0.1356 的合约盘判为「倾向成」;一句对八成人都成立的话说中你,似然比最多 1.25。第二章:按工作站的规则,生命灵数 2 只占 0.0149,33 占 0.0594(分段化约下只有 0.0073);每人平均缺 4.489 个数字,缺 3 的人占 0.7029。第三章:历年约定与生日约定在 0.4986 的日子里给出不同格数,三年期合约跨过第 9 格的概率是 0.4441,在与个人年无关的随机营收里「收成年」按宽松标准应验 0.7031、按严格标准只有 0.1103。第四、五章:大桌阵里指定的修饰牌贴着指定锚牌的概率是 0.1698,0.9985 的摆法里总有风险牌贴着某张锚牌,工作站「贴身」的判词出现在 0.8084 的摆法里。第六章:0.5355 的提问时刻至少触发一条慎审。第七章:合约问题直接成局占 0.2848,未被打断的成局按月亮算不算禁止者在 0.0910 与 0.1922 之间摆动,应期单位一旦可以事后挑选,碰巧命中的机会从 0.0730 升到 0.4790。第八章以一位虚构的工作室老板与一份虚构的酒店合约走完三条线:生命灵数 5、个人年第 8 格,抽出 25+32 与 34+22,卜卦站总分 8 判「倾向成」,再用一张现金表把三条线压成付款节点、检查点与中间人条款,反要约在检查点的价值为 RM 243,000;三条线同时说「成」的机会本来就有 0.0499,所以不算证据,真正的检验需要至少 213 对合约的预先登记盲判。 Chapter answers: - Ch.1 Method: Reduction Arithmetic, Grid Base Rates, the Pair Space and the Sources of Horary Rules — Chapter 1: a life path keeps a birth date modulo 9, cutting 14.42 bits to 3.42, and master numbers depend on the order of addition; the 21,915 days of 1950 to 2009 make only 220 birth grids; two cards give 630 symmetric or 1260 directional readings; in a stated sky model the station judges only 0.1356 of contract charts "leans yes"; a remark true of eight people in ten has a likelihood ratio of at most 1.25. (https://5dlibrary.com/answers/en/life-path-missing-numbers-lenormand-pairs-horary-timing/c01/) 第1章 方法:化约算术、九宫基率、双牌空间与卜卦规则的来历 — 第一章是方法:生命灵数只保留出生日期除以 9 的余数,信息量从 14.42 比特降到 3.42 比特,主数取决于加法顺序;1950 到 2009 年的 21,915 个出生日只有 220 种九宫格;两张雷诺曼牌在对称约定下有 630 种读法、方向约定下有 1260 种;在写明假设的天空模型里,卜卦站只把 0.1356 的合约盘判为「倾向成」;一句对八成人都成立的话说中你,似然比最多 1.25。 (https://5dlibrary.com/answers/zh/life-path-missing-numbers-lenormand-pairs-horary-timing/c01/) - Ch.2 The Twelve Life Paths and the Nine Missing Digits — Chapter 2: under the station's rule, life path 2 takes only 0.0149 of births and 33 takes 0.0594 (0.0073 when parts are reduced first); people miss 4.489 digits on average, and 0.7029 miss the 3. (https://5dlibrary.com/answers/en/life-path-missing-numbers-lenormand-pairs-horary-timing/c02/) 第2章 十二种生命灵数与九种缺数 — 第二章:按工作站的规则,生命灵数 2 只占 0.0149,33 占 0.0594(分段化约下只有 0.0073);每人平均缺 4.489 个数字,缺 3 的人占 0.7029。 (https://5dlibrary.com/answers/zh/life-path-missing-numbers-lenormand-pairs-horary-timing/c02/) - Ch.3 The Nine-Year Personal-Year Cycle: How a Nine-Step Clock Sets the Annual Rhythm of a Business — Chapter 3: calendar and birthday conventions disagree on 0.4986 of days, a three-year contract crosses step 9 with probability 0.4441, and on random revenue a "harvest year" comes true 0.7031 of the time loosely and 0.1103 strictly. (https://5dlibrary.com/answers/en/life-path-missing-numbers-lenormand-pairs-horary-timing/c03/) 第3章 个人年九年周期:一座九格的时钟怎样排一门生意的年度节奏 — 第三章:历年约定与生日约定在 0.4986 的日子里给出不同格数,三年期合约跨过第 9 格的概率是 0.4441,在与个人年无关的随机营收里「收成年」按宽松标准应验 0.7031、按严格标准只有 0.1103。 (https://5dlibrary.com/answers/zh/life-path-missing-numbers-lenormand-pairs-horary-timing/c03/) - Ch.4 Lenormand Pairs I: Rider to Dog, Eighteen Modifiers Beside Six Anchors — Chapters 4 and 5: a named modifier touches a named anchor in a Grand Tableau with probability 0.1698, some risk card sits beside some anchor in 0.9985 of layouts, and the station's "close" verdict fires in 0.8084. (https://5dlibrary.com/answers/en/life-path-missing-numbers-lenormand-pairs-horary-timing/c04/) 第4章 雷诺曼双牌之一:骑士到狗,十八张修饰牌贴着六张锚牌 — 第四、五章:大桌阵里指定的修饰牌贴着指定锚牌的概率是 0.1698,0.9985 的摆法里总有风险牌贴着某张锚牌,工作站「贴身」的判词出现在 0.8084 的摆法里。 (https://5dlibrary.com/answers/zh/life-path-missing-numbers-lenormand-pairs-horary-timing/c04/) - Ch.5 Lenormand Pairs II: Cards 19 to 36 as Modifiers — This chapter's 104 keys are 0.4952 of the dictionary. Some pair from this chapter appears in about one three-card line in four, while a pair named in advance appears in about 1 in 315. In the grand tableau every anchor has about three of this chapter's cards around it; counting the engine's six risk cards, the chance of a risk card touching an anchor is 0.6841, and the engine's "close" verdict appears in 0.8084 of layouts. These base rates set the discipline: write down beforehand which pair you are looking for, and afterwards check only the event you wrote. Each of the eighteen modifiers has its own push: 19 Tower hands the matter to a system, 20 Garden puts it on public view, 21 Mountain stops it, 22 Crossroads forces a choice, 23 Mice nibble it away, 24 Heart adds warmth, 25 Ring binds it, 26 Book hides something unknown, 27 Letter turns it into paper, 28 Man and 29 Woman pin it on one person, 30 Lily slows and steadies it, 31 Sun adds energy, 32 Moon adds reputation and emotion, 33 Key gives the answer, 34 Fish brings money in, 35 Anchor makes it last, and 36 Cross adds weight. (https://5dlibrary.com/answers/en/life-path-missing-numbers-lenormand-pairs-horary-timing/c05/) 第5章 雷诺曼双牌之二:第十九到三十六张作修饰 — 本章的 104 个键占字典的 0.4952。三张一排里出现本章某一对的机会约四分之一,事先指定的某一对只有约三百一十五分之一;大桌阵里每张锚牌身边平均有三张左右本章的牌,按引擎六张风险牌的口径,风险牌贴身的机会是 0.6841,引擎「贴身」的判词出现在 0.8084 的摆法里。这些基率决定了读法的纪律:事前写下要找的那一对,事后只核对写下的那件事。十八张修饰牌各有各的推力:19 塔把事交给制度,20 花园把事摊到台面,21 山让事停住,22 十字路口逼你选,23 老鼠一点点啃,24 心添温度,25 戒指绑定,26 书藏着未知,27 信落成文件,28 男人与 29 女人把事钉在一个人身上,30 百合让事变慢变稳,31 太阳加能量,32 月亮加名声与情绪,33 钥匙给答案,34 鱼带进钱,35 锚让事长久,36 十字架加重量。 (https://5dlibrary.com/answers/zh/life-path-missing-numbers-lenormand-pairs-horary-timing/c05/) - Ch.6 The Twelve Houses of a Horary Question and Six Considerations Before Judgement — Chapter 6: 0.5355 of question moments trip at least one consideration. (https://5dlibrary.com/answers/en/life-path-missing-numbers-lenormand-pairs-horary-timing/c06/) 第6章 卜卦十二宫问事与六条判前慎审 — 第六章:0.5355 的提问时刻至少触发一条慎审。 (https://5dlibrary.com/answers/zh/life-path-missing-numbers-lenormand-pairs-horary-timing/c06/) - Ch.7 Seven Ways a Matter Perfects and Nine Units of Timing — Chapter 7: direct perfection occurs in 0.2848 of contract questions, uninterrupted perfection in 0.0910 or 0.1922 depending on whether the Moon may prohibit, and picking the timing unit after the fact lifts chance hits from 0.0730 to 0.4790. (https://5dlibrary.com/answers/en/life-path-missing-numbers-lenormand-pairs-horary-timing/c07/) 第7章 成局七法与应期九种 — 第七章:合约问题直接成局占 0.2848,未被打断的成局按月亮算不算禁止者在 0.0910 与 0.1922 之间摆动,应期单位一旦可以事后挑选,碰巧命中的机会从 0.0730 升到 0.4790。 (https://5dlibrary.com/answers/zh/life-path-missing-numbers-lenormand-pairs-horary-timing/c07/) - Ch.8 Practice: Will a Hotel Contract Be Signed by Month End? Three Lines, Walked Through — Chapter 8 walks a fictional studio owner and hotel contract through the three lines: life path 5, personal-year step 8, the pairs 25+32 and 34+22, a station score of 8 ("leans yes"), and a cash table turning them into payment milestones, a checkpoint and an intermediary clause worth RM 243,000 at the checkpoint. All three lines say yes by chance 0.0499 of the time, so agreement is not evidence; a real test needs a pre-registered blind judging of at least 213 contract pairs. (https://5dlibrary.com/answers/en/life-path-missing-numbers-lenormand-pairs-horary-timing/c08/) 第8章 实战:一份酒店合约能否在月底签成,三条线走一遍 — 第八章以一位虚构的工作室老板与一份虚构的酒店合约走完三条线:生命灵数 5、个人年第 8 格,抽出 25+32 与 34+22,卜卦站总分 8 判「倾向成」,再用一张现金表把三条线压成付款节点、检查点与中间人条款,反要约在检查点的价值为 RM 243,000;三条线同时说「成」的机会本来就有 0.0499,所以不算证据,真正的检验需要至少 213 对合约的预先登记盲判。 (https://5dlibrary.com/answers/zh/life-path-missing-numbers-lenormand-pairs-horary-timing/c08/) ## 5DL-B-129 · The Auditory Entrainment, Five-Tone Resonance and Sleep-Space Codex 中文书名:声频神经夹带、五音脏腑共振与睡眠空间处方母典 DOI: pending Cite: Fan, Y. (2026). The Auditory Entrainment, Five-Tone Resonance and Sleep-Space Codex (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: It is not medical or investment advice. Abstract: Written for owners who wake in the small hours drafting quotations, this volume asks what sound belongs in a bedroom, how loud, for how long, and how to test on yourself whether it works. Sleep-sound products, the author's own insomnia station included, bundle a measurable signal, a promise that brain waves follow it, and a traditional table pairing five tones with organs and nine constitutions. The book computes the signal, grades the promise on five levels of evidence (CBT-I has the guidelines' strong recommendation; binaural-beat effects are small and unstable) and uses the tradition only to rank equally safe options. Every plan is a self-testable routine and room plan, with added sound from 35 dB(A) to at most 40 dB(A) at the pillow and four signals that call for a professional. It is not medical or investment advice. The preface audits the author's own typing: on 19200 fictional answer sets, only 0.7083 of those ticking early waking at 3 to 5 a.m. are named "early waking". Chapter 1: a 6 Hz tone is inaudible (A-weighting -87.0 dB); two tones in one ear swing with depth 0.934, split between the ears each ear gets 0.000, so headphones are required; started in one's worst week, a useless track still "improves" by 13.1 minutes; four randomised weeks per condition give a smallest one-sided p of 0.0143. Chapter 2: a 40 Hz binaural beat splits the ears by 289.2 cents and is no longer heard as a beat; a one-month self-test detects a meta-analysis-sized effect with power 0.221. Chapter 3: a 35 dB(A) masker cuts a 40 dB(A) intrusion's emergence from 15 to 4.59 dB; a 1 cm gap drops a facade from 34.2 to 28.3 dB. Chapter 4: the ancient string rule sits up to 7.82 cents above equal temperament; A=432 is 31.77 cents below A=440. Chapter 5: random answers flag 6.16 biased constitutions on average. Chapter 6: going to bed 90 minutes early brings waking forward 57 minutes; a diary gives sleep efficiency 0.583 where the station says 0.865. Chapter 7: sealing a door gap buys 8.0 dB, a new door only 2.2. Every entry in Chapters 2 to 7 ends in a checkable decision rule. Chapter 8 follows a fictional design-firm owner: the station estimates 361 minutes of sleep, her diary shows 287.5, early waking costs 135.0 minutes rather than 55, the window is 330 minutes, a 54 dB(A) refuse truck still stands 18.03 dB above a masked floor, and an eighteen-week randomised crossover has power 0.801 to detect a 15-minute difference. 摘要:本卷写给自己扛现金流、凌晨醒来就开始排报价的老板,回答卧室里该放什么声音、放多大、放多久,以及怎样在自己身上查明它有没有用。市面上的助眠声音,包括作者自己的失眠测算站,都把三样东西打包出售:一段可测的信号,一句「脑波会被带着走」的承诺,一张五音配五脏、九种体质的传统对应表。本书把信号按物理算清,把承诺按五级证据标明强弱(失眠认知行为疗法有指南的强推荐,双耳拍频的效应小而不稳定),传统对应只用来给同样安全的候选排先后。所有方案都写成可以自测的作息与环境方案,枕边外加声音从 35 dB(A) 起步、不超过 40 dB(A),并写明四种必须先就医的信号。本书不是医疗或投资建议。 序先核对作者自己的分型:在 19200 份虚构答卷里,勾了三到五点早醒的只有 0.7083 被命名为「早醒」。第一章:6 赫兹听不见(A 计权 -87.0 分贝);两音同进一耳时包络起伏深度 0.934,分进两耳时每只耳朵是 0.000,所以必须戴耳机;在最糟的一周开始,一段无用的音轨也能「改善」13.1 分钟;每个条件四周、完全随机,最小单侧 p 值是 0.0143。第二章:40 赫兹做成双耳节拍,两耳相差 289.2 音分,已经听不成拍;一个月的自测看出元分析那样大小的效应,把握只有 0.221。第三章:35 dB(A) 的掩蔽声把 40 dB(A) 入侵声的突出量从 15 压到 4.59 分贝;一条 1 厘米的缝把外墙隔声从 34.2 拉到 28.3 分贝。第四章:三分损益的五音比十二平均律最多高 7.82 音分,A=432 比 A=440 低 31.77 音分。第五章:完全随机作答,平均会被判出 6.16 种偏颇体质。第六章:提前 90 分钟上床,醒来提前 57 分钟;日记算出的睡眠效率是 0.583,工作站却给出 0.865。第七章:封住门缝买回 8.0 分贝,换一扇门只买回 2.2 分贝。第二至第七章的每个条目都以一条可核对的决策规则收尾。第八章跟随一位虚构的设计公司老板:工作站估算她睡了 361 分钟,日记只有 287.5 分钟,早醒代价是 135.0 分钟而非 55 分钟,睡眠窗定为 330 分钟;54 dB(A) 的垃圾车在掩蔽下仍高出底噪 18.03 分贝;十八周随机交叉自测看出 15 分钟差别的把握是 0.801。 Chapter answers: - Ch.1 Method: The Physics of Sound, Where Beats Come From, Grading the Evidence, and a Test You Can Run Yourself — Chapter 1: a 6 Hz tone is inaudible (A-weighting -87.0 dB); two tones in one ear swing with depth 0.934, split between the ears each ear gets 0.000, so headphones are required; started in one's worst week, a useless track still "improves" by 13.1 minutes; four randomised weeks per condition give a smallest one-sided p of 0.0143. (https://5dlibrary.com/answers/en/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c01/) 第1章 方法:声音的物理、拍频的来历、证据的分级与一份能自己做的检验 — 第一章:6 赫兹听不见(A 计权 -87.0 分贝);两音同进一耳时包络起伏深度 0.934,分进两耳时每只耳朵是 0.000,所以必须戴耳机;在最糟的一周开始,一段无用的音轨也能「改善」13.1 分钟;每个条件四周、完全随机,最小单侧 p 值是 0.0143。 (https://5dlibrary.com/answers/zh/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c01/) - Ch.2 The Five Brainwave Bands: The Arithmetic of Three Kinds of Beat, Choosing a Carrier, and the Weight of the Evidence — Chapter 2: a 40 Hz binaural beat splits the ears by 289.2 cents and is no longer heard as a beat; a one-month self-test detects a meta-analysis-sized effect with power 0.221. (https://5dlibrary.com/answers/en/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c02/) 第2章 五种脑波频段:三种节拍的算术、载波的选择与证据的分量 — 第二章:40 赫兹做成双耳节拍,两耳相差 289.2 音分,已经听不成拍;一个月的自测看出元分析那样大小的效应,把握只有 0.221。 (https://5dlibrary.com/answers/zh/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c02/) - Ch.3 Five Noises and Room Acoustics: The Arithmetic of Adding Decibels, Masking and Reverberation — Chapter 3: a 35 dB(A) masker cuts a 40 dB(A) intrusion's emergence from 15 to 4.59 dB; a 1 cm gap drops a facade from 34.2 to 28.3 dB. (https://5dlibrary.com/answers/en/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c03/) 第3章 五种噪声与房间声学:分贝相加、掩蔽与混响的算术 — 第三章:35 dB(A) 的掩蔽声把 40 dB(A) 入侵声的突出量从 15 压到 4.59 分贝;一条 1 厘米的缝把外墙隔声从 34.2 拉到 28.3 分贝。 (https://5dlibrary.com/answers/zh/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c03/) - Ch.4 The Five Tones: The Two-Thirds Rule, Equal Temperament and Which Mode Belongs Before Sleep — Chapter 4: the ancient string rule sits up to 7.82 cents above equal temperament; A=432 is 31.77 cents below A=440. (https://5dlibrary.com/answers/en/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c04/) 第4章 五音调式:三分损益、十二平均律与睡前该放哪一种调子 — 第四章:三分损益的五音比十二平均律最多高 7.82 音分,A=432 比 A=440 低 31.77 音分。 (https://5dlibrary.com/answers/zh/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c04/) - Ch.5 The Nine Constitutions: Reading Each Imbalance as a Sound Field, a Beat and a Bedroom — Chapter 5: random answers flag 6.16 biased constitutions on average. (https://5dlibrary.com/answers/en/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c05/) 第5章 九种体质:把偏颇读成声场、拍频与一间卧室 — 第五章:完全随机作答,平均会被判出 6.16 种偏颇体质。 (https://5dlibrary.com/answers/zh/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c05/) - Ch.6 Three Early-Waking Bands by Nine Constitutions: Reading the Night in Thirty-Six Cells — Chapter 6: going to bed 90 minutes early brings waking forward 57 minutes; a diary gives sleep efficiency 0.583 where the station says 0.865. (https://5dlibrary.com/answers/en/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c06/) 第6章 早醒三时段乘九种体质:三十六格的夜间作息读法 — 第六章:提前 90 分钟上床,醒来提前 57 分钟;日记算出的睡眠效率是 0.583,工作站却给出 0.865。 (https://5dlibrary.com/answers/zh/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c06/) - Ch.7 The Acoustic Sleep Pod: Insulation, Absorption, Light, Temperature and the Six Stages of a Night — Chapter 7: sealing a door gap buys 8.0 dB, a new door only 2.2. Every entry in Chapters 2 to 7 ends in a checkable decision rule. (https://5dlibrary.com/answers/en/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c07/) 第7章 声学睡眠舱:隔声、吸声、光、温度与一夜的六个阶段 — 第七章:封住门缝买回 8.0 分贝,换一扇门只买回 2.2 分贝。第二至第七章的每个条目都以一条可核对的决策规则收尾。 (https://5dlibrary.com/answers/zh/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c07/) - Ch.8 Practice: An Early-Waking Design Firm Owner, from Questionnaire to an Eighteen-Week Crossover — Chapter 8 follows a fictional design-firm owner: the station estimates 361 minutes of sleep, her diary shows 287.5, early waking costs 135.0 minutes rather than 55, the window is 330 minutes, a 54 dB(A) refuse truck still stands 18.03 dB above a masked floor, and an eighteen-week randomised crossover has power 0.801 to detect a 15-minute difference. (https://5dlibrary.com/answers/en/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c08/) 第8章 实战:一位早醒的设计公司老板,从问卷到十八周交叉自测 — 第八章跟随一位虚构的设计公司老板:工作站估算她睡了 361 分钟,日记只有 287.5 分钟,早醒代价是 135.0 分钟而非 55 分钟,睡眠窗定为 330 分钟;54 dB(A) 的垃圾车在掩蔽下仍高出底噪 18.03 分贝;十八周随机交叉自测看出 15 分钟差别的把握是 0.801。 (https://5dlibrary.com/answers/zh/binaural-entrainment-five-tone-resonance-early-waking-sleep-pod/c08/) ## 5DL-B-130 · The Fasting-Rhythm, Targeted-Mineral and High-Ticket Business Override Codex 中文书名:断食代谢节律、靶向矿物质与高客单商业破局母典 DOI: pending Cite: Fan, Y. (2026). The Fasting-Rhythm, Targeted-Mineral and High-Ticket Business Override Codex (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: A "protocol library" is a set of checklists to follow and check item by item afterwards, not medical or legal advice. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it. Abstract: This volume writes two protocol libraries for people who run a body and a company at once: routine checklists for rolling fasting windows, fats, minerals and light, and clause checklists against partner betrayal and exit accidents. A "protocol library" is a set of checklists to follow and check item by item afterwards, not medical or legal advice. Both share one discipline: grade how far a claim deserves trust, do the arithmetic with the book's programs, run a controlled trial on yourself or your company, and write a stop-loss line before acting. Every entry ends with a decision rule that can be checked afterwards. Contraindications, official mineral ceilings and the signals that call for a doctor are gathered in Chapter 1; the equity clauses follow Malaysian and Taiwanese company law in general terms and need a lawyer's review before signing. Chapter 1 is the method. In a representative twelve-month randomised trial, time-restricted eating and calorie restriction at equal energy differed in weight by 1.80 kg (p = 0.109), while the smallest gap that trial could reliably detect was 3.14 kg, so "no difference shown" is not "no difference proven"; the static rule of 7700 kcal per kilogram promises 22.5 kg in a year where the dynamic model gives about 9.9 kg; 2300 mg of sodium is 5.85 g of salt; a one-person randomised trial needs at least six pairs of periods before its smallest p value falls to 0.03125; and with only a per-payment dual-signature line one person can pay out RM 149,970 in a month. Chapter 2 sets eight fasting windows side by side; five have under twelve weeks of randomised evidence. Chapter 3 finds that a popular "healthy high-fat day" puts 12.7% of the reference man's energy into saturated fat; Chapter 4 that a label reading "zinc 50 mg (as gluconate)" means 348 mg of compound; Chapter 5 that the sun in Kuala Lumpur rises after 07:00 on 316 days a year; Chapter 6 that a shotgun offer without instalments can cost the partner short of cash RM 1,395,000; Chapter 7 that if half the clients of a partner holding 60.5% of revenue follow him out, the expected loss is RM 552,000. Chapter 8 takes an invented founder through the whole method: 400 fasting hours over four weeks, and an eye limit of only 235 seconds for a near-infrared panel measured at 30 mW/cm2; in his three-person company the shareholder who invested RM 200,000 has a voting power of 0.000 on special resolutions, and a founder leaving in month 20 receives RM 167,500 as a good leaver and RM 30,000 as a bad leaver. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it. 摘要:本卷为同时经营身体与公司的人写两套处方库:一套是滚动断食窗口、油脂、矿物质与光照的作息清单,一套是合伙股权防背刺与退出事故的条款清单。「处方库」指可以照做、事后能逐项核对的清单,不是医疗或法律意见。两套清单共用同一种纪律:先按证据分级判断一句话该信几分,再用书中印出的程序把账算清,然后在自己身上或自己的公司里做一次有对照的试验,并在动手之前写下不许放宽的止损线。每个条目都以一条事后可以核对的决策规则收尾。禁忌人群、矿物质的官方上限与就医信号集中写在第一章;股权条款按马来西亚与台湾公司法的一般框架写,签约前须由律师审定。全部计算由书中程序生成。 第一章是方法:在一项十二个月的代表性随机试验里,限时进食与同等热量节食的体重差是 1.80 公斤(p 值 0.109),而该试验能可靠看出的最小差距是 3.14 公斤,所以「没有显示差别」不等于「证明没有差别」;「每公斤 7700 千卡」的静态算法说一年减 22.5 公斤,动态模型说一年约 9.9 公斤;2300 毫克钠等于 5.85 克食盐;单人随机试验至少要六对时段,最小 p 值才降到 0.03125;只设单笔双签线时,一个人一个月可付出 RM 149,970。第二章把八种断食窗口并排比较,其中五种的随机试验证据不到十二周。第三章算出,流行的「健康高脂日」让参照男性的饱和脂肪占到热量的 12.7%;第四章指出,标签上的「锌 50 毫克(以葡萄糖酸锌形式)」意味着要吞下 348 毫克化合物;第五章算出吉隆坡一年有 316 天日出晚于 07:00;第六章算出没有分期付款保护时,一次霰弹枪报价可以让现金少的一方损失 RM 1,395,000;第七章算出握有 60.5% 收入的合伙人离开、半数客户跟他走时,期望损失 RM 552,000。第八章让一位虚构创始人走完全流程:四周空腹合计 400 小时,近红外面板在每平方厘米 30 毫瓦时的眼睛上限只有 235 秒;他那家三人公司里出资 RM 200,000 的股东在特别决议上的表决力是 0.000,第 20 个月离开的创始人按好离开者与坏离开者分别拿到 RM 167,500 与 RM 30,000。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它。 Chapter answers: - Ch.1 Method: Grading Evidence, the Energy and Salt Ledgers, One-Person Trials and a Shared Stop-Loss — Chapter 1 is the method. In a representative twelve-month randomised trial, time-restricted eating and calorie restriction at equal energy differed in weight by 1.80 kg (p = 0.109), while the smallest gap that trial could reliably detect was 3.14 kg, so "no difference shown" is not "no difference proven"; the static rule of 7700 kcal per kilogram promises 22.5 kg in a year where the dynamic model gives about 9.9 kg; 2300 mg of sodium is 5.85 g of salt; a one-person randomised trial needs at least six pairs of periods before its smallest p value falls to 0.03125; and with only a per-payment dual-signature line one person can pay out RM 149,970 in a month. (https://5dlibrary.com/answers/en/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c01/) 第1章 方法:证据分级、能量与水盐的账、一个人的试验与共用的止损线 — 第一章是方法:在一项十二个月的代表性随机试验里,限时进食与同等热量节食的体重差是 1.80 公斤(p 值 0.109),而该试验能可靠看出的最小差距是 3.14 公斤,所以「没有显示差别」不等于「证明没有差别」;「每公斤 7700 千卡」的静态算法说一年减 22.5 公斤,动态模型说一年约 9.9 公斤;2300 毫克钠等于 5.85 克食盐;单人随机试验至少要六对时段,最小 p 值才降到 0.03125;只设单笔双签线时,一个人一个月可付出 RM 149,970。 (https://5dlibrary.com/answers/zh/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c01/) - Ch.2 Eight Fasting Windows and the Four-Week Roll: Putting Fasting Hours into the Calendar — On one clock, the gaps between the eight windows are much larger than their names suggest. From 12:12 to 16:8, awake fasting doubles from 4 hours to 8; the late placement pushes hunger into morning negotiations and the early placement pushes it into evening dinners. The shorter the window, the heavier each meal: 16:8 on two meals puts each meal at 1.4 times the top of the per-meal protein band, one meal a day at 2.8 times, and a 5:2 low-energy day leaves almost no room once protein is eaten. On evidence, five of the eight windows have randomised evidence shorter than twelve weeks; only eight-hour windows, alternate-day fasting and 5:2 run past six months, and when people eat the same amount the window itself has an effect too small to see. So the book arranges only four mild combinations into a four-week ladder: 12:12, 14:10, 16:8 on weekdays, then two days of 18:6. You climb one step a week, pass a gate only by holding six days of seven, and step back after two failed weeks in a row. Four weeks is less than half the median time a new habit takes to form, so at the end you choose to hold, step back or roll again, rather than keep adding. (https://5dlibrary.com/answers/en/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c02/) 第2章 八种断食窗口与四周滚动:把空腹时数排进日程 — 八种窗口放进同一张钟表之后,差距比名字显示的大得多:从 12:12 到 16:8,清醒空腹从 4 小时翻成 8 小时;晚摆法把饥饿推进上午的谈判,早摆法把饥饿推进傍晚的饭局。窗口越短,每一餐越重,16:8 只吃两餐,每餐蛋白质就是参照带上沿的 1.4 倍,一日一餐是上沿的 2.8 倍;5:2 的低能量日吃够蛋白质就几乎没有余地。证据上,八种窗口里五种的随机证据不到十二周,跨过半年的只有八小时窗口、隔日断食与 5:2,且在吃得一样多的前提下,窗口本身的作用小到难以看出。所以本书只把四档温和的组合排成四周阶梯:12:12、14:10、工作日 16:8、两天 18:6,每周只升一档,七天守六天才过闸门,连续两周不过就退档;四周只是新习惯形成中位时间的不到一半,结束时选维持、退档或再滚一轮,而不是一路加码。 (https://5dlibrary.com/answers/zh/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c02/) - Ch.3 Eight Quality Fats: The Fatty-Acid Ledger, the Saturated-Fat Budget and Discipline at the Oil Bottle — Chapter 3 finds that a popular "healthy high-fat day" puts 12.7% of the reference man's energy into saturated fat; (https://5dlibrary.com/answers/en/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c03/) 第3章 八种优质油脂:脂肪酸的账、饱和脂肪的预算与油瓶上的纪律 — 第三章算出,流行的「健康高脂日」让参照男性的饱和脂肪占到热量的 12.7%; (https://5dlibrary.com/answers/zh/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c03/) - Ch.4 Ten Targeted Minerals: Reference Intakes, Ceilings, the Milligrams on the Label and an Audit of the Supplement Shelf — Chapter 4 that a label reading "zinc 50 mg (as gluconate)" means 348 mg of compound; (https://5dlibrary.com/answers/en/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c04/) 第4章 十种靶向矿物质:参考量、上限、标签上的毫克与补剂架的审计 — 第四章指出,标签上的「锌 50 毫克(以葡萄糖酸锌形式)」意味着要吞下 348 毫克化合物; (https://5dlibrary.com/answers/zh/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c04/) - Ch.5 Six Segments of Light: Morning Light, Red, Near-Infrared, Midday, Evening Dimming and Night Darkness — Chapter 5 that the sun in Kuala Lumpur rises after 07:00 on 316 days a year; (https://5dlibrary.com/answers/en/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c05/) 第5章 光节律六段:晨光、红光、近红外、正午、傍晚调暗与夜间全暗 — 第五章算出吉隆坡一年有 316 天日出晚于 07:00; (https://5dlibrary.com/answers/zh/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c05/) - Ch.6 Sixteen Equity Clauses Against a Partner's Knife: Write Everyone's Way Out Before the Money Comes In — Chapter 6 that a shotgun offer without instalments can cost the partner short of cash RM 1,395,000; (https://5dlibrary.com/answers/en/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c06/) 第6章 十六条股权防背刺条款:在钱进门之前写好每个人的退路 — 第六章算出没有分期付款保护时,一次霰弹枪报价可以让现金少的一方损失 RM 1,395,000; (https://5dlibrary.com/answers/zh/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c06/) - Ch.7 The High-Ticket Breakthrough and Eight Exit Incidents — Chapter 7 that if half the clients of a partner holding 60.5% of revenue follow him out, the expected loss is RM 552,000. (https://5dlibrary.com/answers/en/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c07/) 第7章 高客单破局与八种退出事故 — 第七章算出握有 60.5% 收入的合伙人离开、半数客户跟他走时,期望损失 RM 552,000。 (https://5dlibrary.com/answers/zh/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c07/) - Ch.8 Practice: One Founder's Four Weeks and a Shareholders' Agreement for a Three-Person Company — Chapter 8 takes an invented founder through the whole method: 400 fasting hours over four weeks, and an eye limit of only 235 seconds for a near-infrared panel measured at 30 mW/cm2; in his three-person company the shareholder who invested RM 200,000 has a voting power of 0.000 on special resolutions, and a founder leaving in month 20 receives RM 167,500 as a good leaver and RM 30,000 as a bad leaver. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it. (https://5dlibrary.com/answers/en/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c08/) 第8章 实战:一位创始人的四周作息与一家三人公司的股东协议 — 第八章让一位虚构创始人走完全流程:四周空腹合计 400 小时,近红外面板在每平方厘米 30 毫瓦时的眼睛上限只有 235 秒;他那家三人公司里出资 RM 200,000 的股东在特别决议上的表决力是 0.000,第 20 个月离开的创始人按好离开者与坏离开者分别拿到 RM 167,500 与 RM 30,000。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它。 (https://5dlibrary.com/answers/zh/rolling-fasting-windows-electrolytes-near-infrared-equity-stop-loss/c08/) ## 5DL-B-131 · The Same Chart, Different Lives Codex 中文书名:同盘异象:事业角色、婚育阶段与行业赛道断语调制母典 DOI: pending Cite: Fan, Y. (2026). The Same Chart, Different Lives Codex (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: This is not investment, legal, tax or medical advice; all calculations come from the book's programs. Marked boxes separate evidence, calculations, sources and hypotheses, each with what would refute it. Abstract: This hub codex asks why the same chart, the same decade palace and the same thrust wave grow into different realities in different lives, and how a reading should be rewritten for the person who hears it. Before any chart is read, the book writes down a prior of three lines: role (founder, hired executive, independent creator), family stage (six, from single to divorced) and industry track (eight, from retail to property and finance). Writing the identity first separates base rates from the chart and blocks the Barnum effect. Family stage is pure bookkeeping (who runs the money and the school run, how family assets stay apart from company shares); no form of family is judged. The chart readings are traditions of unproven validity, and base rates marked as demos are set by the author. This is not investment, legal, tax or medical advice; all calculations come from the book's programs. Chapter 1 is the method: 144 identities times 72 reading units make 10368 cells, of which the dictionary keeps 227 entries; on demo base rates "your partnership changes this year" hits founders 2.00 times as often as managers with no reading, a manager's reading needs a likelihood ratio of 2.43 to catch up, ten Barnum sentences win eight nods with probability 0.9298, and a reading that hurts within every role pools to a risk ratio of 1.351. Chapter 2 profiles roles, stages and tracks: ignoring identity costs managers 10.3% in Brier score, and education runs on -166.0 net cash days against 185.2 in property and finance. Chapters 3 and 4 read the twelve decade palaces: a ten-year window turns a 0.10 annual event into a chance hit of 0.6513, a creator needs 12.12 months of reserve where a founder needs 3.73, and the Parents decade never meets adult children (0.000). Chapters 5 and 6 read the six thrust waves: the hard-year verdict hits founders at 0.839 and managers at 0.167, a skewed crowd fakes a lift of 1.32, and one attack year adds 8.81 months of fixed cost in education but costs 10.10 in property. Chapter 7 sets the rules: five rungs, three caps (household floor, signing authority, leveraged track), the strictest wins, rungs only fall; over 864 combinations 0.7454 keep their raw rung, only 32 of 144 attack cells stay all-in, 900000 yuan of savings leaves a pool of 816000 yuan when single but 396000 yuan with young children, and evidence dates run from month 2.0 (retail) to month 15.0 (property). Chapter 8 runs three invented people born in the same hour through one chart, removing 4, 1 and 2 rungs over a decade, and designs a four-way blind test with a Barnum check that needs 129 people at a true pick rate of 0.35. Marked boxes separate evidence, calculations, sources and hypotheses, each with what would refute it. 摘要:本卷是一部中枢母典,回答一个前面各卷没有回答的问题:同一张盘、同一个十年大限坐的宫位、同一种推力波,落在不同的人身上,现实里为什么长成不同的样子,断语又该怎样随人改写。本书把「角色 × 婚育阶段 × 行业赛道」当作读盘之前必须写下的先验:三种角色(创业老板、受聘高管即经理人、独立创作者),六种阶段(单身、有伴未婚、已婚无子女、子女未成年、子女已成年、离异),八条赛道(零售电商、专业服务、科技软件、制造、内容媒体、身心健康、教育培训、地产金融)。先写身份、再读盘,是为了把基础率与盘面分开,堵住巴纳姆效应。婚育阶段只用来记账:谁管钱、谁接送孩子、家族资产与公司股权怎样隔离,不评判任何一种家庭形态。各种盘面读法是传统,预测效度没有被证实,本书也不主张它;书中标明「演示」的基础率是作者设定的数。本书不是投资、法律、税务、医疗或心理意见,全部计算由书中程序生成。 第一章是方法:144 种身份乘 72 个读盘单元共 10368 格,字典只写主效应与两两组合 227 条;按演示基础率,「今年合伙关系有变」不读盘说给老板的命中率是说给经理人的 2.00 倍,经理人的读盘要有 2.43 的似然比才追上老板的基础率;十句巴纳姆句里至少认同八句的概率是 0.9298;一份在每个角色里都帮倒忙的读盘,合并后的风险比却是 1.351,所以命中率必须先按角色、再按阶段分层统计。第二章写角色、阶段与赛道的画像:不写身份时,经理人的 Brier 误差多出 10.3%;教育培训的净现金天数是 -166.0 天,地产金融是 185.2 天。第三、第四章读十二宫的十年大限:年发生率 0.10 的事放进十年窗口,碰巧应验的机会就有 0.6513;同一句「钱会一波一波地来」,创作者要 12.12 个月的储备,老板只要 3.73 个月;父母十年与子女已成年同时出现的比例是 0.000,报告里出现这一格就先核对资料。第五、第六章读六种推力波:硬仗年那句断语对老板的基础应验率是 0.839,对经理人只有 0.167;来咨询的人群稍有偏斜,就能造出 1.32 倍的假抬升;同一个全力进攻年,让教育培训多出 8.81 个月固定成本的现金,却让地产金融倒贴 10.10 个月。第七章是调制规则:五级行动阶梯,家庭底线、签字权与杠杆赛道三道上限取最严,只降不升,并列时按家庭、角色、赛道记账;跑遍 864 种组合,保持原档的占 0.7454,全力进攻年的 144 格默认只有 32 格保持全力进攻;同一笔 900000 元的家庭储备,单身时可动用的池子 816000 元,子女未成年时只有 396000 元;证据日期从零售电商的第 2.0 个月到地产金融的第 15.0 个月。第八章让三位同一个时辰出生的虚构人物共用一张财帛大限样盘,逐年调制,十年里规则分别压掉 4、1 与 2 档,并给出带巴纳姆检查的四选一盲测:真实挑中率 0.35 时需要 129 人。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它。 Chapter answers: - Ch.1 Method: Write Down Who It Is Before Reading the Chart — Chapter 1 is the method: 144 identities times 72 reading units make 10368 cells, of which the dictionary keeps 227 entries; on demo base rates "your partnership changes this year" hits founders 2.00 times as often as managers with no reading, a manager's reading needs a likelihood ratio of 2.43 to catch up, ten Barnum sentences win eight nods with probability 0.9298, and a reading that hurts within every role pools to a risk ratio of 1.351. (https://5dlibrary.com/answers/en/same-chart-different-lives-role-family-stage-industry-track-modulation/c01/) 第1章 方法:先写下身份,再去读盘 — 第一章是方法:144 种身份乘 72 个读盘单元共 10368 格,字典只写主效应与两两组合 227 条;按演示基础率,「今年合伙关系有变」不读盘说给老板的命中率是说给经理人的 2.00 倍,经理人的读盘要有 2.43 的似然比才追上老板的基础率;十句巴纳姆句里至少认同八句的概率是 0.9298;一份在每个角色里都帮倒忙的读盘,合并后的风险比却是 1.351,所以命中率必须先按角色、再按阶段分层统计。 (https://5dlibrary.com/answers/zh/same-chart-different-lives-role-family-stage-industry-track-modulation/c01/) - Ch.2 Three Roles, Six Family Stages and Eight Tracks: The Seventeen Lines of Identity You Write Before Reading a Chart — Chapter 2 profiles roles, stages and tracks: ignoring identity costs managers 10.3% in Brier score, and education runs on -166.0 net cash days against 185.2 in property and finance. (https://5dlibrary.com/answers/en/same-chart-different-lives-role-family-stage-industry-track-modulation/c02/) 第2章 三种角色、六种婚育阶段与八条赛道:读盘之前先写下的十七行身份 — 第二章写角色、阶段与赛道的画像:不写身份时,经理人的 Brier 误差多出 10.3%;教育培训的净现金天数是 -166.0 天,地产金融是 185.2 天。第三、第四章读十二宫的十年大限:年发生率 0.10 的事放进十年窗口,碰巧应验的机会就有 0.6513;同一句「钱会一波一波地来」,创作者要 12.12 个月的储备,老板只要 3.73 个月;父母十年与子女已成年同时出现的比例是 0.000,报告里出现这一格就先核对资料。第五、第六章读六种推力波:硬仗年那句断语对老板的基础应验率是 0.839,对经理人只有 0.167;来咨询的人群稍有偏斜,就能造出 1.32 倍的假抬升;同一个全力进攻年,让教育培训多出 8.81 个月固定成本的现金,却让地产金融倒贴 10.10 个月。 (https://5dlibrary.com/answers/zh/same-chart-different-lives-role-family-stage-industry-track-modulation/c02/) - Ch.3 One Decade, Nine Lives: The Self-to-Health Decades Under Role and Family Stage — This chapter read the six decades from Self to Health cell by cell across nine identities. The first step was age: the bureau is uniform over the 1440 combinations, each at 0.2000, so every decade has a fixed age window. Self belongs to childhood, Siblings runs from the early teens to twenty-five, and Spouse, Children, Wealth and Health each move ten years later, all except Self on backward charts only. When an identity and a window cannot coexist, the statement passes to the household member inside the window, or becomes a retrospective audit. The second step was the base rate: a ten-year window lifts an event with an annual hazard of 0.10 to a chance hit rate of 0.6513, and once the base rate is that high no chart, however good, can prove itself. Each of the six sections then carried an account: pairwise relationships and buy-out costs in the Siblings decade, post-separation shareholdings in the Spouse decade, the week's slack in the Children decade, reserves for three cash shapes in the Wealth decade, and revenue lost to a key person's absence in the Health decade. Each section ends with a decision rule that can be checked within a year, and the chapter's overall claim goes to a pre-registered matching test. (https://5dlibrary.com/answers/en/same-chart-different-lives-role-family-stage-industry-track-modulation/c03/) 第3章 同一个十年,九种活法:命宫到疾厄六限的身份调制 — 本章把命宫到疾厄六种十年放在九种身份上逐格读过。第一步是年龄:局数在 1440 种组合里均匀分布,各占 0.2000,所以每种十年的年龄窗口是确定的,命宫在童年,兄弟在少年到二十五岁,夫妻、子女、财帛、疾厄依次后移十年,而且除命宫外都只在逆行盘上出现;身份与窗口不相容时,断语转给家里正处在窗口里的人,或者变成一次回顾审计。第二步是底数:十年窗口会把年发生率 0.10 的事情抬到 0.6513 的碰巧命中,底数一高,再好的盘也证明不了自己。之后的六节各配一笔账:兄弟大限算合伙关系的对数与回购金额,夫妻大限算分开时的持股比例,子女大限算一周的余量,财帛大限算三种现金形状的储备,疾厄大限算关键人缺席的营收损失。每一节都以一条一年内可以核对的决策规则收尾,本章的整体说法则交给一项预先登记的匹配检验。 (https://5dlibrary.com/answers/zh/same-chart-different-lives-role-family-stage-industry-track-modulation/c03/) - Ch.4 Travel, Friends, Career, Property, Fortune, Parents: Six Outward Decades That Change Face with Identity — The six outward palaces govern the layer between you and the world: ground, circle, engine, household base, way out and mandate. p40 shows that their decades sit at fixed places in a life. Parents, Fortune, Property, Career and Friends are reached in the prime business years only by forward charts, while Travel is a stretch everyone shares in their sixties. Age decides first which family stages can occur; identity then decides what the palace becomes in you. (https://5dlibrary.com/answers/en/same-chart-different-lives-role-family-stage-industry-track-modulation/c04/) 第4章 迁移、交友、官禄、田宅、福德、父母:六个朝外的十年,换一种身份就换一张脸 — 六个朝外的宫位管的是你与世界之间的那一层:地盘、圈子、引擎、家底、退路与授权。p40 显示它们的十年在人生里的位置是固定的,父母、福德、田宅、官禄、交友只有顺行盘在经营主力年龄走得到,迁移则是所有人在六十多岁共同的一段。年龄先决定了哪些婚育阶段可能出现,身份再决定这一宫在你身上长成什么样子。 (https://5dlibrary.com/answers/zh/same-chart-different-lives-role-family-stage-industry-track-modulation/c04/) - Ch.5 One Kind of Year, Many Lives: Six Thrust Waves by Role and Family Stage — p54 measures "borrowed momentum" as income concentration. The Herfindahl index is the sum of the squared shares of the income sources, and its inverse is the equivalent number of sources. The founder's sample has a channel partner as its largest source, at 0.45, and 3.51 equivalent sources; the executive's largest support is the mandate from the chair, at 0.55; the creator's largest is one platform, at 0.70, with only 1.90 equivalent sources. When the largest source leaves, even if half of it is won back within six months, the creator still loses 0.423 of a year's income. To bring the platform's share under 0.30, all other income must grow to 5.44 times its present size: lowering dependence is a project of several years, not of one. (https://5dlibrary.com/answers/en/same-chart-different-lives-role-family-stage-industry-track-modulation/c05/) 第5章 同一种年份落在不同的人身上:六种推力波 × 角色与婚育阶段 — p54 用收入集中度来量化「借来的势头」。赫芬达尔指数是各收入来源份额的平方和,倒数是等效的来源个数。创业老板的样本最大来源是渠道合作方,占 0.45,等效来源 3.51 个;受聘高管最大的依靠是董事长给的授权,占 0.55;独立创作者最大的依靠是一个平台,占 0.70,等效来源只有 1.90 个。最大来源退出之后,即使半年内追回一半,创作者一年仍会损失 0.423 的收入。要把平台占比压到 0.30 以下,其余收入必须增长到原来的 5.44 倍:这说明降低依赖是多年的工程,不是一年能做完的事。 (https://5dlibrary.com/answers/zh/same-chart-different-lives-role-family-stage-industry-track-modulation/c05/) - Ch.6 Six Thrust Waves on Eight Tracks: The Cash Shape of the Same Year in Different Businesses — The frequency of the six kinds is set by B-123: the hard year comes at least once in nearly every decade and twice in two decades out of three, the attack year has an average wait of 39.7 years. The track's cash clock decides how much of the force reaches the books in the same year: 0.833 for retail, 0.159 for property and finance. p62 shows that the same kind of year can move cash in opposite directions on different tracks: an attack year gives education 8.81 months of fixed cost in cash while costing property and finance 10.10; a hold year burns 3.18 months for education while giving property and finance 2.49. Reading a thrust wave without writing down the track is like not knowing which line of the books the force will land on. (https://5dlibrary.com/answers/en/same-chart-different-lives-role-family-stage-industry-track-modulation/c06/) 第6章 六种推力波落在八条赛道:同一种年份在不同生意里的现金形状 — 六种年份的频率由 B-123 决定,硬仗年几乎每个十年都至少来一次、三分之二的十年来两次,进攻年平均要等 39.7 年。赛道的现金时钟决定这股力当年见账多少:零售电商 0.833,地产金融 0.159。p62 说明同一种年份在不同赛道上的现金方向可以相反:进攻年让教育培训多出 8.81 个月固定成本的现金,却让地产金融倒贴 10.10 个月;蓄力年让教育培训烧掉 3.18 个月,却让地产金融多出 2.49 个月。读推力波之前不写赛道,就等于不知道这股力会落在账本的哪一行。 (https://5dlibrary.com/answers/zh/same-chart-different-lives-role-family-stage-industry-track-modulation/c06/) - Ch.7 Modulation Rules: Who Decides When Role, Stage and Track Disagree — Chapter 7 sets the rules: five rungs, three caps (household floor, signing authority, leveraged track), the strictest wins, rungs only fall; over 864 combinations 0.7454 keep their raw rung, only 32 of 144 attack cells stay all-in, 900000 yuan of savings leaves a pool of 816000 yuan when single but 396000 yuan with young children, and evidence dates run from month 2.0 (retail) to month 15.0 (property). (https://5dlibrary.com/answers/en/same-chart-different-lives-role-family-stage-industry-track-modulation/c07/) 第7章 调制规则:角色、阶段与赛道冲突时谁说了算 — 第七章是调制规则:五级行动阶梯,家庭底线、签字权与杠杆赛道三道上限取最严,只降不升,并列时按家庭、角色、赛道记账;跑遍 864 种组合,保持原档的占 0.7454,全力进攻年的 144 格默认只有 32 格保持全力进攻;同一笔 900000 元的家庭储备,单身时可动用的池子 816000 元,子女未成年时只有 396000 元;证据日期从零售电商的第 2.0 个月到地产金融的第 15.0 个月。 (https://5dlibrary.com/answers/zh/same-chart-different-lives-role-family-stage-industry-track-modulation/c07/) - Ch.8 Practice: One Sample Chart, Three Lives — Chapter 8 runs three invented people born in the same hour through one chart, removing 4, 1 and 2 rungs over a decade, and designs a four-way blind test with a Barnum check that needs 129 people at a true pick rate of 0.35. Marked boxes separate evidence, calculations, sources and hypotheses, each with what would refute it. (https://5dlibrary.com/answers/en/same-chart-different-lives-role-family-stage-industry-track-modulation/c08/) 第8章 实战:一张样盘,三种人生 — 第八章让三位同一个时辰出生的虚构人物共用一张财帛大限样盘,逐年调制,十年里规则分别压掉 4、1 与 2 档,并给出带巴纳姆检查的四选一盲测:真实挑中率 0.35 时需要 129 人。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它。 (https://5dlibrary.com/answers/zh/same-chart-different-lives-role-family-stage-industry-track-modulation/c08/) ## 5DL-B-132 · The Multi-Oracle Conflict-Resolution and Chief-of-Staff Action Codex 中文书名:多维神谕冲突消解、权重裁决与总参谋行动处方母典 DOI: pending Cite: Fan, Y. (2026). The Multi-Oracle Conflict-Resolution and Chief-of-Staff Action Codex (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All figures come from the book's programs; it is not investment, legal, medical or psychological advice. Abstract: The hub volume of the destiny series asks how to decide when Ziping, Zi Wei and Qin Tian, Qimen, Liuyao and the western bundle (tarot, Human Design, the Tzolkin, astrology) give opposite signals on one matter: attack, defend or hold. Casting no charts, it adds the vote each earlier volume hands over in log-odds weighted by question type and time scale, discounts the three birth-data and two moment systems by the design effect 1 + (n - 1)ρ, and maps the result onto a nine-step action ladder with a client script. No system's validity is established; the weights are conventions to be tested with the Brier score, calibration and a capped update rule. All figures come from the book's programs; it is not investment, legal, medical or psychological advice. Chapter 1: a weight of 1.0 claims a hit rate of 0.7311 and needs 27 judged calls to confirm, the worst-fitted cell 6872. At correlation 0.5 three witnesses sharing a source are worth 1.5, and an undiscounted synthesis is right only 0.8534 of the time when it claims over nine in ten. Across 8,748 verdicts five votes top out at 1.9415, so all-in and retreat open only through an evidence gate: the log-odds of a hard fact (a signed order, cash received, a judgment or regulatory decision in force, a default that has happened), registered in advance, zero by default. A useless synthesis scores 0.2782 on Brier, worse than fifty-fifty, yet over 20 calls looks useful with probability 0.2352. Chapter 2: equal weights would change 1334 of the 8,748 verdicts. Chapter 3: the month is the handover point, where the birth group's share falls to 53.7%; the twenty-one tallies. Chapters 4 and 5: alone, a disagreeing pair never moves L beyond 1.00, and its direction survives the other votes in 0.537 of cases. Chapter 6: hedge is the widest rung at 0.4024; even at a probability of 0.701 all-in gains only 2.59 over staging but loses 48.30 more in a bad world. Chapter 7: the client script, with natural frequencies, no threats, no promises and a review date; 62 of 100 single-vote changes would move the verdict. Chapter 8: an invented manufacturer weighs a three-year exclusive supply deal; three votes attack, two defend, the month-scale value is exactly -0.5000, trim; a pre-registered buyer default, if confirmed, adds -1.0 and makes it fortify. With the weights in doubt, hedge and trim split almost evenly, so the deal becomes a one-year non-exclusive contract with a bank guarantee: RM 41,200 less if all goes well, RM 406,200 less lost if orders stop. A blind ledger needs 240 judged calls for 0.8795 power against a coin. 摘要:本卷是命理系列的中枢卷,回答一个同时看几种盘的老板迟早会遇到的问题:当子平、紫微与钦天、奇门、六爻与西洋一组(塔罗、人类图、卓尔金、占星)对同一件事给出相反的信号,宜攻、宜守或观望,应当怎样裁决。本书不重新起盘,只接收前面各卷交来的一张票,用对数几率把五张票按「问事类型 × 时间尺度」加权合成;读同一份出生资料的三个系统与读发问时刻的两个系统,各自按设计效应 1 + (n - 1)ρ 打折;合成值落到九级行动阶梯上,再配一段对客户说的合成话术。各系统的预测效度没有被证实:权重是可以检验、也可以被推翻的决策约定,书中用 Brier 分数、校准表与一条带上限的更新规则说明怎样事后检验与修订。全部数字由书中程序生成;本书不是投资、法律、医疗或心理意见。 第一章立方法:权重 1.0 等于宣称 0.7311 的命中率,要 27 次有结果的判定才能证实,适配最差的一格要 6872 次;相关系数 0.5 时,三个同源证人只值 1.5 个,不打折的合成在自称九成以上把握的判断里实际只对了 0.8534。穷举 8,748 种裁决,五票合成最高只到 1.9415,所以全力进攻与撤退只能经过「证据闸门」打开:盘外硬证据(已经签订的订单、已经到账的款项、已经生效的判决或监管决定、已经发生的违约)的对数几率,事先登记,默认为零。一套毫无信息的合成,期望 Brier 分数 0.2782,比一直说五五开更差,只记 20 次时却仍有 0.2352 的机会显得有用。第二章定权重,改用等权会让 8,748 种裁决里的 1334 种换档。第三章说明月是长线与短线的换手点,出生组在这一层的音量份额降到 53.7%,并逐一写出二十一种共识计数。第四、五章拆解六十种两两冲突:一对分歧单独最多把 L 推到 1.00,放回五家合议之后,它的方向只在 0.537 的情形里保留。第六章给每一级配上预算与复查日:对冲是最宽的一级,占 0.4024;即使概率到 0.701,全力进攻比分批进攻只多赚 2.59,坏世界里却多亏 48.30。第七章写合成话术:三段式、自然频数、点名分歧、不恐吓、不许诺、写清复盘日期;十个虚构问事里,一百种单票改动有 62 种会让裁决换档。第八章让一位虚构的代工厂老板面对一份三年独家供货合约:三票攻、两票守,月尺度上的合成恰好是 -0.5000,裁决为收缩;预先登记的那一项买方违约若被查实,记 -1.0,裁决变为加固防线。把权重放进合理的怀疑范围,对冲与收缩几乎各占一半,于是落成一份一年期、非独家、有银行保函的合约,顺利时少赚 RM 41,200,对方停单时少亏 RM 406,200;盲测账本要 240 次有结果的判定,才有 0.8795 的把握把合成与抛硬币分开。 Chapter answers: - Ch.1 Method: Log-Odds, the Correlation Discount, the Nine-Step Ladder and Scoring Afterwards — Chapter 1: a weight of 1.0 claims a hit rate of 0.7311 and needs 27 judged calls to confirm, the worst-fitted cell 6872. At correlation 0.5 three witnesses sharing a source are worth 1.5, and an undiscounted synthesis is right only 0.8534 of the time when it claims over nine in ten. Across 8,748 verdicts five votes top out at 1.9415, so all-in and retreat open only through an evidence gate: the log-odds of a hard fact (a signed order, cash received, a judgment or regulatory decision in force, a default that has happened), registered in advance, zero by default. A useless synthesis scores 0.2782 on Brier, worse than fifty-fifty, yet over 20 calls looks useful with probability 0.2352. (https://5dlibrary.com/answers/en/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c01/) 第1章 方法:对数几率、相关折扣、九级阶梯与事后计分 — 第一章立方法:权重 1.0 等于宣称 0.7311 的命中率,要 27 次有结果的判定才能证实,适配最差的一格要 6872 次;相关系数 0.5 时,三个同源证人只值 1.5 个,不打折的合成在自称九成以上把握的判断里实际只对了 0.8534。穷举 8,748 种裁决,五票合成最高只到 1.9415,所以全力进攻与撤退只能经过「证据闸门」打开:盘外硬证据(已经签订的订单、已经到账的款项、已经生效的判决或监管决定、已经发生的违约)的对数几率,事先登记,默认为零。一套毫无信息的合成,期望 Brier 分数 0.2782,比一直说五五开更差,只记 20 次时却仍有 0.2352 的机会显得有用。 (https://5dlibrary.com/answers/zh/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c01/) - Ch.2 Six Question Types and Their Weights: Which Voice Speaks for What — Chapter 2: equal weights would change 1334 of the 8,748 verdicts. (https://5dlibrary.com/answers/en/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c02/) 第2章 六类问事的权重表:五个声音各管哪一段 — 第二章定权重,改用等权会让 8,748 种裁决里的 1334 种换档。 (https://5dlibrary.com/answers/zh/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c02/) - Ch.3 Six Time Scales and Twenty-One Consensus Tallies: Who Overrides Whom, and When a Vote Counts — Chapter 3: the month is the handover point, where the birth group's share falls to 53.7%; the twenty-one tallies. (https://5dlibrary.com/answers/en/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c03/) 第3章 时间尺度六层与共识计数二十一种:谁覆盖谁,几票才算数 — 第三章说明月是长线与短线的换手点,出生组在这一层的音量份额降到 53.7%,并逐一写出二十一种共识计数。 (https://5dlibrary.com/answers/zh/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c03/) - Ch.4 Pairwise Conflicts I: Bazi Against Four Systems, and Ziwei Against Qimen — Chapters 4 and 5: alone, a disagreeing pair never moves L beyond 1.00, and its direction survives the other votes in 0.537 of cases. (https://5dlibrary.com/answers/en/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c04/) 第4章 两两冲突(上):子平对四家、紫微对奇门的三十种分歧 — 第四、五章拆解六十种两两冲突:一对分歧单独最多把 L 推到 1.00,放回五家合议之后,它的方向只在 0.537 的情形里保留。 (https://5dlibrary.com/answers/zh/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c04/) - Ch.5 Pairwise Conflicts, Part Two: the Decade, the Single Question, the Board of the Moment and the Outer Projection — This chapter covered five pairs, Zi Wei against Liu Yao, Zi Wei against the western bundle, Qimen against Liu Yao, Qimen against the western bundle and Liu Yao against the western bundle, each in six directions, thirty keys in all. p50 gives each pair's handover point and the four tied cells. p51 shows that no pair on its own goes beyond probe or trim, the largest |L| being 1.00, and that the twenty keys with one side holding are only four solo voices. p52 shows that the in-group discount ignores whether voices agree, and that the birth group is capped at 2.000 witnesses and the moment group at 3.333. p53 puts the thirty keys into 29160 full verdicts, where the pair's own direction survives in only 0.537 of them and neither end of the ladder ever appears. p54 walks five fictional owners through the whole procedure. p55 gives the way to test the ledger, which needs well over two hundred resolved records before it can say anything. The five decision rules come down to three sentences: write the time scale before choosing a side; when one source disagrees with itself, check the birth time first; when two pictures of one moment disagree, split the action into a reversible half and an irreversible half. (https://5dlibrary.com/answers/en/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c05/) 第5章 两两冲突之二:十年大势、单事一问、眼前之局与外部投射 — 本章写了紫微对六爻、紫微对西洋、奇门对六爻、奇门对西洋、六爻对西洋五对,每对六个方向,共三十个键。p50 给出每一对的交棒点与四个平手格;p51 显示,任何一对单独最多走到小额试探或收缩,|L| 的最大值是 1.00,而且二十个有一方观望的键只是四种单声部;p52 说明组内折扣不分同向反向,出生组的证人上限是 2.000 个,当下组是 3.333 个;p53 把三十个键放进 29160 个全局裁决,单对的方向只保留 0.537,而且两端的两级一次都没出现;p54 用五位虚构老板走完全流程;p55 给出台账的检验方法,需要两百多条已见分晓的记录才能开始说话。五条决策规则归结为三句:先写时间尺度再选边;同源相反先核对出生时刻;同一刻两张图打架,把行动拆成可撤回与不可撤回两半。 (https://5dlibrary.com/answers/zh/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c05/) - Ch.6 The Nine-Rung Action Ladder: Turning One Number into One Move — Chapter 6: hedge is the widest rung at 0.4024; even at a probability of 0.701 all-in gains only 2.59 over staging but loses 48.30 more in a bad world. (https://5dlibrary.com/answers/en/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c06/) 第6章 九级行动阶梯:把一个数变成一个动作 — 第六章给每一级配上预算与复查日:对冲是最宽的一级,占 0.4024;即使概率到 0.701,全力进攻比分批进攻只多赚 2.59,坏世界里却多亏 48.30。 (https://5dlibrary.com/answers/zh/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c06/) - Ch.7 Synthesis Scripts: Saying the Verdict to the Client — Chapter 7: the client script, with natural frequencies, no threats, no promises and a review date; 62 of 100 single-vote changes would move the verdict. (https://5dlibrary.com/answers/en/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c07/) 第7章 合成话术:把裁决说给客户听 — 第七章写合成话术:三段式、自然频数、点名分歧、不恐吓、不许诺、写清复盘日期;十个虚构问事里,一百种单票改动有 62 种会让裁决换档。 (https://5dlibrary.com/answers/zh/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c07/) - Ch.8 Practice: A Three-Year Exclusive Supply Contract and Five Voices at Odds — Chapter 8: an invented manufacturer weighs a three-year exclusive supply deal; three votes attack, two defend, the month-scale value is exactly -0.5000, trim; a pre-registered buyer default, if confirmed, adds -1.0 and makes it fortify. With the weights in doubt, hedge and trim split almost evenly, so the deal becomes a one-year non-exclusive contract with a bank guarantee: RM 41,200 less if all goes well, RM 406,200 less lost if orders stop. A blind ledger needs 240 judged calls for 0.8795 power against a coin. (https://5dlibrary.com/answers/en/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c08/) 第8章 实战:一份三年独家供货合约,五个声音两两打架 — 第八章让一位虚构的代工厂老板面对一份三年独家供货合约:三票攻、两票守,月尺度上的合成恰好是 -0.5000,裁决为收缩;预先登记的那一项买方违约若被查实,记 -1.0,裁决变为加固防线。把权重放进合理的怀疑范围,对冲与收缩几乎各占一半,于是落成一份一年期、非独家、有银行保函的合约,顺利时少赚 RM 41,200,对方停单时少亏 RM 406,200;盲测账本要 240 次有结果的判定,才有 0.8795 的把握把合成与抛硬币分开。 (https://5dlibrary.com/answers/zh/multi-oracle-conflict-resolution-bayesian-weighting-chief-of-staff-action/c08/) ## 5DL-B-133 · Zero-History Cold-Start Price Testing, Van Westendorp Sensitivity Bounds and Bayesian Adaptive Repricing 中文书名:零历史数据冷启动测价、范氏价格敏感度边界与贝叶斯自适应调价母典 DOI: pending Cite: Fan, Y. (2026). Zero-History Cold-Start Price Testing, Van Westendorp Sensitivity Bounds and Bayesian Adaptive Repricing (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Laws on price display and segment pricing are described only in general terms and are not legal or tax advice. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it; the source studied is the "Guoxu" chapter of the Guanzi on gathering and releasing goods and the "level" price. Abstract: This volume answers a question every launch faces: how to price a brand-new product that has no sales history at all. Its method runs in four stages. First, write a deliberately light prior from cost, competitors and a survey. Second, draw the test range and lay out three to five price arms with the Van Westendorp price-sensitivity questions and the cost floor. Third, learn while selling on real traffic with a multi-armed bandit that chooses arms by revenue or gross profit per visitor, not by conversion. Fourth, decide when to converge and when to stop with an expected-loss rule and a regret cap. Every price test registers ten items before the first visitor arrives. The ethical floor is that the same customer sees and pays the same price at the same moment, that claims about deadlines and stock are true, and that customers who paid more during a test can claim the difference. Laws on price display and segment pricing are described only in general terms and are not legal or tax advice. All brands, surveys and sales are invented samples, and every calculation is produced by the book's programs. Chapter 1 is the method. On an invented demand curve, the prices with the highest conversion, revenue and gross profit are three different prices, and choosing by conversion gives up 0.427 of the best gross profit; raising the effective sample size of a survey prior from 10 to 200 raises the bias left after the same 300 visitors from 0.0005 to 0.0060; the regret of "test evenly, then commit" is U-shaped in the test length; peeking at results daily pushes the false positives of an A/A test from 0.052 to 0.281, and telling apart by significance two prices that differ by RM 0.220 per visitor needs 117,531 visitors per arm. Chapters 2 to 7 cover the four Van Westendorp intersections, zero-history priors, price bandits, contextual bandits and compliance boundaries, Bayesian adaptive repricing, and launch playbooks for four kinds of product with a tool specification for a cold-start price tester. Chapter 8 takes an invented card deck through the whole method: the survey gives an acceptable range of RM 65.75 to RM 154.96, and the cost floor of RM 80.95 overrides the survey's lower edge; six weeks and 6,600 visitors bring revenue of RM 29,941 with an expected regret of RM 1936; across three invented worlds the worst-case regret of learning on revenue is RM 3491, below the RM 6732 of a price fixed from the survey; a confirmatory significance test would take 270.1 weeks at this traffic, so convergence relies on registered decision rules. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it; the source studied is the "Guoxu" chapter of the Guanzi on gathering and releasing goods and the "level" price. 摘要:本卷回答一个每位首发者都会遇到的问题:一件全新的商品没有任何历史销量,价格怎样定。本书的做法分四个阶段:先用成本、竞品与问卷写出一个分量很轻的先验;再用范氏价格敏感度四问与成本底线划出测价区间,排出三到五个价格臂;然后在真实流量上用多臂赌博机边卖边学,按每位访客的收入或毛利而不是转化率选臂;最后用期望损失与后悔值上限决定何时收敛、何时止损。每一次测价都在第一位访客进来之前登记十项内容。书中对测价伦理的底线是:同一位客户在同一时刻看到并支付同一个价格,限时与库存的说法必须真实,测价期间买贵了的客户可退还差价。价格展示与分群定价的法规只写一般框架,不是法律或税务意见。全部品牌、问卷与销量都是虚构样本,全部计算由书中程序生成。 第一章是方法:在一条虚构的需求曲线上,转化率最高、收入最高与毛利最高的价格是三个不同的价格,按转化率选价要少赚最佳毛利的 0.427;问卷先验的有效样本量从 10 加到 200,同样 300 位访客之后残留的偏差从 0.0005 升到 0.0060;「先平均测、再全押」的后悔值随测试长度呈 U 形;每天偷看一次结果,会把 A/A 测试的误报从 0.052 推到 0.281,而要用显著性分辨每位访客相差 RM 0.220 的两个价格,每臂需要 117,531 位访客。第二至第七章依次写范氏四交点、零历史先验、价格赌博机、上下文赌博机与合规边界、贝叶斯自适应调价,以及四类产品的首发手册与冷启动测价器的工具规格。第八章让一副虚构牌卡走完全程:问卷给出可接受区间 RM 65.75 到 RM 154.96,成本底线 RM 80.95 压过问卷下沿;六周 6,600 位访客带来 RM 29,941 的收入,期望后悔值 RM 1936;在三个虚构世界里,按收入学习的最坏后悔值 RM 3491,低于凭问卷固定一个价格的 RM 6732;确认性的显著性检验在这一流量上要 270.1 周,所以收敛靠登记好的决策规则。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它;原典考辨取《管子·国蓄》的轻重敛散与「准平」。 Chapter answers: - Ch.1 Method: The Structure of Cold-Start Pricing, Evidence Grades, Four Stages and Regret — Chapter 1 is the method. On an invented demand curve, the prices with the highest conversion, revenue and gross profit are three different prices, and choosing by conversion gives up 0.427 of the best gross profit; raising the effective sample size of a survey prior from 10 to 200 raises the bias left after the same 300 visitors from 0.0005 to 0.0060; the regret of "test evenly, then commit" is U-shaped in the test length; peeking at results daily pushes the false positives of an A/A test from 0.052 to 0.281, and telling apart by significance two prices that differ by RM 0.220 per visitor needs 117,531 visitors per arm. (https://5dlibrary.com/answers/en/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c01/) 第1章 方法:冷启动定价的结构、证据分级、四个阶段与后悔值 — 第一章是方法:在一条虚构的需求曲线上,转化率最高、收入最高与毛利最高的价格是三个不同的价格,按转化率选价要少赚最佳毛利的 0.427;问卷先验的有效样本量从 10 加到 200,同样 300 位访客之后残留的偏差从 0.0005 升到 0.0060;「先平均测、再全押」的后悔值随测试长度呈 U 形;每天偷看一次结果,会把 A/A 测试的误报从 0.052 推到 0.281,而要用显著性分辨每位访客相差 RM 0.220 的两个价格,每臂需要 117,531 位访客。第二至第七章依次写范氏四交点、零历史先验、价格赌博机、上下文赌博机与合规边界、贝叶斯自适应调价,以及四类产品的首发手册与冷启动测价器的工具规格。 (https://5dlibrary.com/answers/zh/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c01/) - Ch.2 Van Westendorp Price Sensitivity: Four Questions, Four Crossings and What They Cannot Tell You — The Van Westendorp trades four open questions for four cumulative curves. In Qitong Studio's fictional sample the four crossings under convention A are a PMC of 95.8 yuan, an OPP of 145.0 yuan, an IPP of 147.6 yuan and a PME of 212.1 yuan. Convention B squeezes the range from 116.3 yuan wide to 53.7 yuan wide, so ask about the convention before reading any report. The bootstrap shows the OPP to be the least stable, 17.9 yuan wide; quadrupling the sample roughly halves the widths, and at least 200 valid forms are needed. The NMS calibration links the bounds to revenue: the trial-maximising 105 yuan is not the revenue-maximising 140 yuan, and the calibration weights change the level of volume, not the ranking of prices. Anchoring, mixed populations and the absence of purchase intention are three known gaps. (https://5dlibrary.com/answers/en/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c02/) 第2章 范氏价格敏感度:四个问题、四个交点与它们管不到的地方 — 范氏量表用四个开放题换来四条累积曲线,在栖桐工作室的虚构样本里,约定 A 下的四个交点是边际低价点 95.8 元、最优价格点 145.0 元、无差异价格点 147.6 元、边际高价点 212.1 元。约定 B 会把区间从 116.3 元宽压到 53.7 元宽,拿到报告先问约定。自举区间显示最优价格点最不稳,宽 17.9 元;样本量翻两番,区间宽度大约减半,有效问卷不少于 200 份。NMS 校准把边界接到收入上:试购率最高的 105 元不是收入最高的 140 元,校准权重改变的是销量的高低,不是价格的排名。锚定、混合人群与零购买意愿是三处已知的缺口。 (https://5dlibrary.com/answers/zh/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c02/) - Ch.3 Priors Without History: Six Sources and One Synthesis — At zero history an owner holds six kinds of information. Cost gives the hard boundary: Qitong Studio's fictional unit cost of 32.50 yuan and fees of one tenth give a zero-contribution floor of 36.11 yuan and a 40%-margin floor of 65.00 yuan. The competitor band's geometric mean of 117.5 yuan, the adjacent-category anchor of 126.7 yuan, the Gabor–Granger ladder's best rung of 119 yuan, the experts' calibrated merge of 142.0 yuan and the previous chapter's NMS figure of 140.0 yuan combine into a log-normal prior centred on 128.6 yuan, with a conservative 80% interval of 100.6 to 164.3 yuan. The simplified conjoint gives no price level, only feature premiums: about 21.8 yuan for the gift box and about 26.7 yuan for the video guide. Each of the four price arms has a Beta prior with n0 = 60; ranked by revenue, 129 yuan leads, ranked by conversion, 99 yuan does. The fictional first week shows that a weak prior can be swayed by luck, so the posterior is first used to steer traffic, not to close arms. The next chapter hands these priors to the bandit to learn while selling on real traffic. (https://5dlibrary.com/answers/en/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c03/) 第3章 没有历史的先验:六种来源与一次合成 — 零历史时,老板手里有六种信息。成本给出硬边界:栖桐工作室虚构的单位成本 32.50 元、一成费用,零贡献下限 36.11 元,四成贡献率下限 65.00 元。竞品带的几何平均 117.5 元,相邻品类锚 126.7 元,Gabor–Granger 梯子的收入最高级 119 元,专家校准后的合成 142.0 元,加上上一章 NMS 的 140.0 元,合成为中心 128.6 元的对数正态先验,保守八成区间 100.6 到 164.3 元。简化联合分析不给价格水平,只给功能的价差:礼盒约 21.8 元,视频导读约 26.7 元。四个价格臂各自有一个 n0 = 60 的 Beta 先验,按收入排序时 129 元领先,按转化率排序时 99 元领先。第一周的虚构数据显示,弱先验会被运气左右,所以后验先用来调流量,不用来关臂。下一章把这些先验交给赌博机,在真实流量上边卖边学。 (https://5dlibrary.com/answers/zh/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c03/) - Ch.4 Prices as Arms: Discrete Price Ladders, Epsilon-Greedy, UCB1 and Revenue Sampling — Cutting Chapter 2's band into three to five geometrically spaced arms and defining the reward as revenue per visitor are the premises of everything in this chapter; in the invented 99, 129, 159 and 209 sample, 99 yuan converts best and 159 yuan earns most. Six thousand launch visitors are far too few for a fixed test to separate neighbouring arms, and a bandit's value is to make learning cheap: revenue Thompson sampling averages 5,949 yuan of regret, an even split 9,074 and conversion Thompson 12,739. UCB1's half-width is uselessly wide when purchases are rare, and epsilon-greedy carries a lock-in gamble. A prior built Chapter 3's way saves over a thousand yuan more, provided it is not overconfident, and daily batched updates cost almost nothing. The next chapter adds market, channel and new versus returning customers as context, and draws the lines that segment pricing must not cross. (https://5dlibrary.com/answers/en/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c04/) 第4章 价格就是臂:离散价格梯、ε-贪心、UCB1 与按收入抽样 — 把第二章的带子切成三到五条等比间距的臂,奖励写成每访客收入,是本章一切计算的前提;在 99、129、159、209 四档的虚构样本里,转化最高的是 99 元,收入最高的是 159 元。首发六千名访客远不够做一次分辨相邻两档的固定检验,赌博机的价值在于让学习少花钱:按收入的 Thompson 抽样平均后悔值 5,949 元,等分流量 9,074 元,按转化率的 Thompson 12,739 元。UCB1 的半宽在稀有成交场景里宽得没有意义,ε-贪心有锁死的赌性。照第三章的办法造出的先验能再省一千多元,前提是它不自信;按天批量更新几乎不花钱。下一章把市场、渠道与新老客这些上下文加进来,并划出分群定价不能越过的线。 (https://5dlibrary.com/answers/zh/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c04/) - Ch.5 Contextual Bandits: Market, Channel, New versus Returning, and the Limits of Segment Pricing — Writing market, channel and new versus returning as context brings prices closer to each group's price sensitivity, but at launch scale the gain is limited: in Beian Studio's fictional sample, the segmentation ceiling is 1.095 times the single price, 5,928 yuan more over six thousand visitors, and the best contextual rule earns only 879 yuan more than knowing the best single price from day one. LinUCB's theoretical α is absurdly wide and must be scaled to reward noise; logistic Thompson sampling fits a shop's batch reality better. The most important finding is that the algorithms teach themselves to charge returning customers more, and that in this sample doing so earns almost nothing: charging new and returning customers alike loses only 0.005 of revenue. So the limits sit outside the algorithm: same customer, same time, same price, guaranteed by hash assignment; consistent display prices, with channel prices carrying public conditions and a capped spread; only public discounts between new and returning customers, never a loyalty penalty; and a former price in a time-limited offer must be a price that was really applied. The next chapter turns to repricing over time. (https://5dlibrary.com/answers/en/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c05/) 第5章 上下文赌博机:市场、渠道、新老客与分群定价的边界 — 把市场、渠道与新老客写成上下文,能让定价更贴近不同人群的价格敏感度,但首发规模下收益有限:在北岸工作室的虚构样本里,分群上限是单一价格的 1.095 倍、六千名访客多 5,928 元;最好的上下文规则只比从第一天就知道单一最优价多赚 879 元。LinUCB 的理论 α 宽得离谱,要按奖励噪声缩放;逻辑回归 Thompson 更贴近店铺的批量现实。最要紧的发现是,算法会自己学会向回头客收更高的价,而这样做在本例里几乎不赚钱:新老客同价只损失 0.005 的收入。所以边界写在算法之外:同客同时同价,由散列分配保证;展示价一致,渠道价附带公开条件,价差有上限;新老客只许公开优惠,不许杀熟;限时活动的原价必须是真正用过的价格。下一章转向随时间的调价。 (https://5dlibrary.com/answers/zh/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c05/) - Ch.6 Bayesian Adaptive Repricing: Elasticity, Stock, Cooldowns, Reference Prices and Stop-Loss Lines — The bottom half is the report card of 4000 simulated launches checked once a week. With the threshold at 0.90 and at least 600 visitors before any verdict, an arm that actually makes money (true conversion 0.032) is wrongly killed in 0.013 of launches; an arm that actually loses money (true conversion 0.018) survives all six weeks in 0.155 of launches, and when it is stopped, the stop comes in week 3.10 on average. Raising the threshold to 0.95 cuts wrongful kills to 0.004 but lets the losing arm survive in 0.305 of launches. Left open, the losing arm burns RM 857 more in advertising over six weeks than it brings in margin. Where to set the threshold is not a statistical question but a business one: a wrongful kill costs the good arm's future margin, a missed kill costs these 857. (https://5dlibrary.com/answers/en/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c06/) 第6章 贝叶斯自适应调价:弹性、库存、冷却期、参考价格与止损线 — 下半部分是 4000 次模拟里每周查一次的成绩单。门槛定在 0.90、要求至少 600 位访客才开始判:一个其实赚钱的臂(真实转化率 0.032)被误杀的比例是 0.013;一个其实亏钱的臂(真实转化率 0.018)六周都没被止住的比例是 0.155,被止住时平均在第 3.10 周。门槛提到 0.95,误杀降到 0.004,漏杀却升到 0.305。这个亏钱的臂若一直开着,六周烧掉的广告费比它带来的毛利多 RM 857。门槛怎样定,不是统计问题,而是生意问题:误杀一个好臂损失的是它未来的毛利,漏杀一个坏臂损失的是这 857。 (https://5dlibrary.com/answers/zh/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c06/) - Ch.7 The Launch Playbook: Four Kinds of Product, the Ethics of Price Testing and the Distillation Spec — The four kinds of product take four different routes to a price. Diagnostic reports can be tested side by side online; decks move the test into the pre-order window; courses test by cohort, one price per cohort; boutique pieces skip traffic tests and rely on surveys, negotiation logs and a markdown ladder archived in advance. The three optima come apart: the price with the highest conversion is often not the one with the highest revenue, and that is often not the one with the highest margin, especially for high-cost goods. The ethics of price testing can be written as program checks: the split ratio, the 30-day rule for struck-through prices, price protection announced in advance, and a data field behind every scarcity claim. The last section hands all of this, together with the calculations of the previous six chapters, to the pricing calculator. (https://5dlibrary.com/answers/en/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c07/) 第7章 首发测价落地:四类商品的手册、测价伦理与四件套蒸馏规格 — 四类商品的测价走四条不同的路:诊断报告能在线上并排测;牌卡把测价搬到预售期;课程按期测、每期一个价;精品不测流量,靠问卷、谈判记录与事先存档的降价阶梯。三种「最优」会分离,转化率最高的价格常常不是收入最高的,收入最高的又常常不是毛利最高的,高成本商品尤其如此。测价伦理可以写成程序检查:分流比例、划线价的 30 天规则、事先公布的价格保护,以及每一句稀缺话术都能追溯到一个数据字段。最后一节把这些连同前六章的计算一起,交给定价试算器。 (https://5dlibrary.com/answers/zh/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c07/) - Ch.8 Practice: The Mistharbour Deck from Zero to Week Six — Chapter 8 takes an invented card deck through the whole method: the survey gives an acceptable range of RM 65.75 to RM 154.96, and the cost floor of RM 80.95 overrides the survey's lower edge; six weeks and 6,600 visitors bring revenue of RM 29,941 with an expected regret of RM 1936; across three invented worlds the worst-case regret of learning on revenue is RM 3491, below the RM 6732 of a price fixed from the survey; a confirmatory significance test would take 270.1 weeks at this traffic, so convergence relies on registered decision rules. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it; the source studied is the "Guoxu" chapter of the Guanzi on gathering and releasing goods and the "level" price. (https://5dlibrary.com/answers/en/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c08/) 第8章 实战:雾港牌卡从零到第六周的完整测价 — 第八章让一副虚构牌卡走完全程:问卷给出可接受区间 RM 65.75 到 RM 154.96,成本底线 RM 80.95 压过问卷下沿;六周 6,600 位访客带来 RM 29,941 的收入,期望后悔值 RM 1936;在三个虚构世界里,按收入学习的最坏后悔值 RM 3491,低于凭问卷固定一个价格的 RM 6732;确认性的显著性检验在这一流量上要 270.1 周,所以收敛靠登记好的决策规则。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它;原典考辨取《管子·国蓄》的轻重敛散与「准平」。 (https://5dlibrary.com/answers/zh/zero-history-cold-start-price-testing-van-westendorp-sensitivity/c08/) ## 5DL-B-134 · Purchasing-Power-Parity Distortions, Multi-Currency Psychological Price Endings and the Southeast Asia–Taiwan–Japan Pricing Matrix 中文书名:跨国购买力平价畸变、多币种心理尾数映射与东南亚台日定价矩阵母典 DOI: pending Cite: Fan, Y. (2026). Purchasing-Power-Parity Distortions, Multi-Currency Psychological Price Endings and the Southeast Asia–Taiwan–Japan Pricing Matrix (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Exchange rates and price levels in the book are stated sample assumptions, not live data; taxes, display rules and the law on price discrimination are described only in general terms, not as legal or tax advice. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it. Abstract: This volume answers one question for owners who sell across borders: when the same product goes on sale in Malaysia, Singapore, Taiwan, Japan and the United States at once, what price should each market display? The method is a four-step matrix: set a price tier from purchasing-power parity and the price level index, move the tier price onto the local psychological ending lattice, check it against a gross-margin floor after tax, fees and cost, and check the price gaps against a cross-market arbitrage band made of freight, currency spread and time cost. Every cell holds a price and a reason, and every price is written as a pre-registered test judged on gross profit per visitor. Exchange rates and price levels in the book are stated sample assumptions, not live data; taxes, display rules and the law on price discrimination are described only in general terms, not as legal or tax advice. All calculations come from the book's programs. Chapter 1 is the method. Of 80 prices that follow American ending habits, hard conversion puts none on the book's ringgit lattice; at the sample price levels a 39-dollar product weighs 2.326 times as heavily in a Malaysian wallet as in an American one, so hard conversion is in fact a price rise for low-price-level markets; at the consensus half-life of the PPP literature only 0.1591 of a deviation closes in a year, and 90% closes only after 13.29 years; at the share of neglect the book assumes, a cut of 1 inside one leading digit is felt as only 0.75, while a cut across the leading digit is magnified 25.75 times; for a deck shipped from Kuala Lumpur the American margin floor is 43.18 dollars, above the 39-dollar base; a reseller's arbitrage band breaches 8 of ten market pairs; and a lower price must lift conversion to 1.9218 times its former level just to tie on gross profit per visitor. Chapter 1 also grades evidence from A to D and reads equitable transport and the balanced standard in the Treatise on the Balanced Standard of the Records of the Grand Historian, the book's canonical source, for what the ancients got right and wrong about price gaps, transport and stock. Chapters 2 to 6 cover the pricing customs of five currencies and the Hong Kong dollar, PPP distortions, ending lattices, arbitrage bands, and tax display and exchange-rate drift; Chapter 7 applies the matrix to physical card decks, digital reports, courses and boutique goods and writes the distillation spec of a multi-currency pricing calculator. Chapter 8 takes an invented card brand from 49 dollars through the four steps: the final prices are 49.00 dollars in the United States, RM 149.00 in Malaysia (lifted by the band from RM 129.90), S$ 59.90 in Singapore, NT$ 1,080 in Taiwan and ¥5,580 in Japan; the collector's box has a ratio of only 1.1344 between its highest and lowest dollar-equivalent prices, against 1.4536 for the everyday deck; simulated exchange-rate paths push the Japanese margin down to 0.5433 and trigger the first review; and at invented traffic a single market needs 97 to 257 weeks to detect a ten per cent gap in gross profit, so the tests address only large decisions and pool the five markets. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it. 摘要:本卷为跨国卖货的老板回答一个问题:同一件商品在马来西亚、新加坡、台湾、日本与美国同时开卖,每个市场该标什么价。书中的方法是一个四步矩阵:先按购买力平价与价格水平指数定价格档位,再把档位价落到当地的心理尾数格点上,然后用扣除税费与成本之后的毛利底线检查,最后用运费、汇差与时间成本构成的跨区套利带检查价差。每一格写一个价格与一个理由,每个价格都写成预先登记、按每位访客毛利来判的检验。书中的汇率与价格水平都是写明的样本假设,不是实时数据;税费、标价与价格歧视的法规只写一般框架,不是法律或税务意见。全部计算由书中程序生成。 第一章是方法:80 个符合美国尾数习惯的价格硬换算成令吉,没有一个落在本书的令吉格点上;按样本价格水平,一件 39 美元的商品在马来西亚客人钱包里的分量是美国的 2.326 倍,所以硬换算对低价格水平市场其实是涨价;按购买力平价文献的共识半衰期,偏离一年只回归 0.1591,回归九成要 13.29 年;按书中假设的忽略比例,在同一个左位数里降价 1 元只被感受成 0.75 元,跨过最左位的一次降价则被放大到 25.75 倍;从吉隆坡寄出的一副牌卡,美国的毛利底线是 43.18 美元,高于 39 美元的基准价;转卖者的套利带让十对市场里的 8 对被击穿;低价若要在每位访客毛利上打平,转化率必须升到原来的 1.9218 倍。第一章还按 A 至 D 四级给证据分级,并以《史记·平准书》的均输与平准为原典,考辨古人对价差、运输与库存看对与看错之处。第二至第六章分别写五种货币与港币的标价习惯、购买力平价畸变、尾数格点、跨区套利带,以及税费展示与汇率漂移;第七章把矩阵落到实体牌卡、数字报告、课程与精品四类商品,并写出多币种定价试算器的蒸馏规格。第八章让一个虚构牌卡品牌从 49 美元出发走完四步,最终价格是美国 49.00 美元、马来西亚 RM 149.00(被套利带从 RM 129.90 提上来)、新加坡 S$ 59.90、台湾 NT$ 1,080、日本 ¥5,580;典藏礼盒五地美元等值价的最大最小之比只有 1.1344,日常牌卡是 1.4536;模拟的汇率路径让日本的毛利率跌到 0.5433,最先触发复核;在虚构流量下,单个市场要分辨一成的毛利差距需要 97 到 257 周,所以测价只测大的决定,五地合并分析。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它。 Chapter answers: - Ch.1 Method: The Hard-Conversion Trap, Purchasing-Power Parity and Psychological Endings — Chapter 1 is the method. Of 80 prices that follow American ending habits, hard conversion puts none on the book's ringgit lattice; at the sample price levels a 39-dollar product weighs 2.326 times as heavily in a Malaysian wallet as in an American one, so hard conversion is in fact a price rise for low-price-level markets; at the consensus half-life of the PPP literature only 0.1591 of a deviation closes in a year, and 90% closes only after 13.29 years; at the share of neglect the book assumes, a cut of 1 inside one leading digit is felt as only 0.75, while a cut across the leading digit is magnified 25.75 times; for a deck shipped from Kuala Lumpur the American margin floor is 43.18 dollars, above the 39-dollar base; a reseller's arbitrage band breaches 8 of ten market pairs; and a lower price must lift conversion to 1.9218 times its former level just to tie on gross profit per visitor. (https://5dlibrary.com/answers/en/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c01/) 第1章 方法:硬换算的陷阱、购买力平价与心理尾数 — 第一章是方法:80 个符合美国尾数习惯的价格硬换算成令吉,没有一个落在本书的令吉格点上;按样本价格水平,一件 39 美元的商品在马来西亚客人钱包里的分量是美国的 2.326 倍,所以硬换算对低价格水平市场其实是涨价;按购买力平价文献的共识半衰期,偏离一年只回归 0.1591,回归九成要 13.29 年;按书中假设的忽略比例,在同一个左位数里降价 1 元只被感受成 0.75 元,跨过最左位的一次降价则被放大到 25.75 倍;从吉隆坡寄出的一副牌卡,美国的毛利底线是 43.18 美元,高于 39 美元的基准价;转卖者的套利带让十对市场里的 8 对被击穿;低价若要在每位访客毛利上打平,转化率必须升到原来的 1.9218 倍。 (https://5dlibrary.com/answers/zh/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c01/) - Ch.2 Anchors of Six Currencies: Denominations, Small Change, Endings and Tax-Inclusive Labels — A currency's anchor is made of four things: the smallest circulating unit and the denominations, cash rounding, ending habits and tax-inclusive labelling. The ringgit has statutory five-sen rounding, so only 20 of the hundred sen endings survive a cash till unchanged; x.99 becomes a whole number in cash, multi-item totals round up and down in turn, and the average adjustment is zero. The Singapore dollar has no one-cent coin, so 9.99 cannot be settled at a cash till. Paying 9.99 dollars exactly takes 14 pieces; the 1,078 yen produced by adding tax to a pre-tax 980 returns 9 pieces of change, while an inclusive 1,100 returns only 5. Japan, Singapore and Taiwan require tax-inclusive labels, so the master data should hold the inclusive price: adding tax to whole-yen pre-tax prices leaves 364 inclusive totals between 1,000 and 5,000 yen unreachable, 2,980 among them. US sales tax is added at the register, and a shelf price of 9.99 becomes 10.79 at a sample rate of 0.08. Hong Kong dollar conversion varies by only 1.28% across the linked-rate band. Hard-convert 39 dollars into five currencies and not one result lands on a local anchor, with anchor spacing ranging from 0.017 for the yen to 0.175 for the Singapore dollar. The next chapter shows that a bigger distortion than the ending comes from the exchange rate itself: hard conversion ignores local price levels. (https://5dlibrary.com/answers/en/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c02/) 第2章 六种货币的锚:面额、零钱、尾数与含税标价 — 一种货币的锚由四件事组成:最小流通单位与面额、现金舍入、尾数习惯、含税标价。马币有法定的五仙取整,一百个仙位尾数只有 20 个在现金柜台上不被改写,x.99 付现变成整数,多件合计时上下抵消,平均调整为零;新元没有一分硬币,9.99 在现金柜台上付不清。美元的 9.99 恰好付清要 14 件,日元由税前 980 加税得到的 1,078 要找零 9 件,而含税定在 1,100 只要 5 件。日本、新加坡、台湾要求含税标价,主数据应当存含税价:以整数日元的税前价加税,1,000 到 5,000 日元之间有 364 个含税总额凑不出来,2,980 就是其中之一;美国的销售税在收银时另加,货架上的 9.99 在 0.08 的样本税率下变成 10.79。港币在联系汇率带内的换算只差 1.28%。把 39 美元硬换成五种货币,没有一个结果落在本地锚上,锚距从日元的 0.017 到新元的 0.175 不等。下一章说明,比尾数更大的偏差来自汇率本身:硬换算忽略了各地的价格水平。 (https://5dlibrary.com/answers/zh/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c02/) - Ch.3 Purchasing-Power-Parity Distortions: Why Hard Conversion Is a Price Rise in Low-Price-Level Markets — The distortion of purchasing power parity comes from an identity: the PLI equals the PPP factor divided by the market rate. With the sample values of p30, hard conversion makes Malaysian shoppers feel they pay 132.6% more, Taiwanese 92.3%, Japanese 56.2%, Hong Kong shoppers 28.2% and Singaporeans 16.3%; full parity, by contrast, leaves Malaysia contributing only 16.77 dollars. Choosing β between the two raises the high-to-low ratio of dollar-equivalent prices from 1.000 all the way to 2.326, and to 1.659 at β = 0.6. The law of one price fails widely even for finely classified tradables; absolute parity provides a ruler across markets; relative parity decides how often the price list needs review, and on the sample paths Taiwan and Japan both need re-tiering within five years; deviations have a half-life of three to five years, an annual persistence of 0.7937 to 0.8706, so prices should not chase the exchange rate month by month; the Balassa–Samuelson effect explains why imported decks and local courses need different conversion coefficients, with product-level PLIs of 0.914 and 0.601; and the raw and income-adjusted readings of a Big Mac-style index can reach opposite conclusions. Finally, p34 cuts the PLI into four tiers at β = 0.6 and backs them with a floor of 2.426 times landed cost: the floor binds in the high-freight United States, Taiwan and Japan, not in Malaysia with the lowest general price level, and the high-to-low ratio closes at 1.788. The next chapter moves these not-yet-landed prices onto each currency's ending grid. (https://5dlibrary.com/answers/en/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c03/) 第3章 购买力平价的畸变:汇率硬算为什么等于给低价格水平市场涨价 — 购买力平价的畸变来自一个恒等式:PLI 等于平价因子除以市场汇率。按 p30 的样本值,汇率硬算让马来西亚顾客体感多付 132.6%,台湾 92.3%,日本 56.2%,香港 28.2%,新加坡 16.3%;全额平价则让马来西亚的美元收入只剩 16.77。在两头之间取 β,市场之间以美元计的价格高低之比从 1.000 一路升到 2.326,β 为 0.6 时是 1.659。一价定律对细分的可贸易品也大面积失灵;绝对平价给出一把跨市场的尺子;相对平价决定价格表多久该复核,样本路径上台湾与日本五年内都要重新定档;偏离的半衰期是三到五年,对应每年 0.7937 到 0.8706 的保留率,所以价格不该每月追着汇率跑;巴拉萨–萨缪尔森效应解释了为什么进口牌卡与本地课程该用不同的折算系数,产品层 PLI 分别是 0.914 与 0.601;巨无霸式指数的原始读法与收入调整读法可以给出相反的结论。最后,p34 按 β 为 0.6 把 PLI 分成四档,并用到岸成本 2.426 倍的底价兜底:触底的是运费高的美国、台湾与日本,而不是价格水平最低的马来西亚,价格高低之比收在 1.788。下一章把这些尚未落地的价格挪到各币种的尾数格点上。 (https://5dlibrary.com/answers/zh/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c03/) - Ch.4 Ending Lattices: Landing a Tier Price on a Number Local Buyers Recognise — A lattice is a period plus a set of endings, and how far it can sit from a tier price is set roughly by the period divided by the price: the same x9 has a nearest-snap error of about 6.0% at ringgit prices above 100 and 14.7% at Singapore-dollar prices in the forties. The eight candidate classes each have a use: decimal endings barely move the price, single and double 9s signal a good deal, round prices suit gifts and collections, just-below points are the only class that changes the leading digit, the lucky 8 and the avoided 4 are constraints rather than evidence, and the yen's 80 and 98 barely move a price on a period of 100. Rounding checks the floor first and the break-even volume second: one step down needs 16.0% more units in ringgit and 40.7% more in Singapore dollars. The left-digit effect shrinks with the leading digit and is worth most at thresholds like 20, 30 and 50. The cost of avoiding 4 is concentrated in the yen's four-thousand stretch, where the move reaches 10.5%. The pipeline strings these rules together and gives prices with reasons; converted to dollars, the deck runs from 22.60 to 43.99 across the five markets, a gap left for Chapter 5 to check. (https://5dlibrary.com/answers/en/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c04/) 第4章 尾数格点:把档位价落到当地客人认得的数字上 — 格点是一个周期加一组尾数,它离档位价最远有多远,大致由周期除以价格决定:同样的 x9,在令吉一百多的价位上就近误差约 6.0%,在新元四十多的价位上是 14.7%。八类候选格点各有用处:小数尾几乎不动价格,个位 9 与双 9 给出划算信号,整数价适合礼品与收藏,门槛下方格点是唯一改变首位读数的一类,吉利 8 与回避 4 是约束而不是证据,日元的 80 与 98 在百元周期上几乎不偏离。取整先看底线,再看打平所需的销量:令吉往下取一格要多卖 16.0%,新元要多卖 40.7%。左位数效应随首位数字递减,在 20、30、50 这样的门槛上最值钱。回避 4 的代价集中在日元四千段,偏离会到 10.5%。流水线把这些规则串起来,给出带理由的价格,牌卡在五个市场折成美元从 22.60 到 43.99,这个价差留给第五章去检查。 (https://5dlibrary.com/answers/zh/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c04/) - Ch.5 Arbitrage Bands: How Wide a Price Gap Can Hold, and What Holds It — The arbitrage band is the full cost of moving a good from the cheap market to the dear one: freight, the currency spread, the reseller's cut, time and risk, and tax above the de minimis line. There are two leak routes: buyers purchasing for themselves are limited by single-item freight and are most dangerous between neighbouring markets; resellers are limited by their own cut, and their band hardly falls below about forty per cent. Under this book's sample parameters the deck's mean band is 37.4% and the box's only 32.0%: the higher the unit price, the less price-level gap the band can hold. All the leaks come from the ringgit market: a price rise alone would take the deck to 139.90, while order caps, local-only shipping and a price rise together take it only to 109.90. Digital goods have no freight, and their tiers stand or fall with a regional check applied equally to every customer. In a general framework, pricing by market is third-degree price discrimination, and what keeps it honest is transparency: a named currency, a total with no hidden fees, one price per market at one time, genuine "was" prices and published order caps. The leak share is watched against a weekly alarm line, and the effect of the levers is tested with a pre-registered comparison. (https://5dlibrary.com/answers/en/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c05/) 第5章 跨区套利带:价差能撑多宽,用什么守住 — 套利带是把一件商品从便宜的市场搬到贵的市场的全部成本:运费、汇差、转卖者的抽成、时间与风险、越过免税线的税。它有两条泄漏路线,客人自购受单件运费约束,在近邻市场最危险;代购受自身抽成约束,带宽很难低于四成左右。在本书的样本参数下,牌卡的平均带宽是 37.4%,礼盒只有 32.0%,单价越高,能容纳的价格水平落差越小。泄漏全部来自令吉市场:单靠提价,牌卡要涨到 139.90;限购、只寄本地与提价合用,只要涨到 109.90。数字商品没有运费,分档能否成立取决于对所有客人一致执行的地区检查。分市场定价在一般框架下属于三级价格歧视,守住它的是透明:币种写明,总价不藏费用,同市场同时同价,划线价真实,限购公开。泄漏比例用每周的警戒线监测,堵漏的效果用预先登记的比较检验。 (https://5dlibrary.com/answers/zh/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c05/) - Ch.6 Tax-Inclusive Display, Fees and Exchange-Rate Drift: How Endings Wear Down Before and After Checkout — Tax, fees and the exchange rate are the three gates a lattice price passes before it reaches the books. Markets with tax-inclusive display should keep the tax-inclusive lattice price as master data: books-first does collect one more slice of tax, but it turns RM99.00 into RM106.92 and ¥5,580 into ¥6,138, and a system whose master data is the pre-tax price leaves some lattice points unreachable under every rounding mode, about one in eleven of the yen points. A fixed fee weighs heavily on cheap price points; the fee burden on a RM9.90 digital report is more than four times that on RM199.00, and a platform commission pushes the lowest sellable lattice point up, in Japan beyond the current price. Against exchange-rate drift, the book's dense lattice leaves every price only a narrow dead zone, and a weekly-recomputed table changes prices more than eighteen times a year. This book's default is two hard triggers, the floor and the band, a drift trigger of 0.03 with a two-week waiting period, and a stress buffer of 0.03. Chapter 7 puts these rules into the matrix engine and writes them out as the specification of the calculator. (https://5dlibrary.com/answers/en/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c06/) 第6章 含税标价、手续费与汇率漂移:尾数在结账前后的损耗 — 税、手续费与汇率是格点价格到账之前的三道关口。含税标价的市场应以含税格点价为主数据,因为账本优先虽然多收一份税额,却让 RM99.00 变成 RM106.92、¥5,580 变成 ¥6,138,而以税前价为主数据的系统,在每种舍入方式下都有一批格点凑不出来,日元格点约十一分之一。手续费里的固定费对低价档格外重,RM9.90 的数字报告手续费负担是 RM199.00 的四倍多;平台抽成会把最低可卖的格点往上推,在日本会推过现行价格。汇率漂移面前,全书的密格点让每个价格只有很窄的死区,每周重算的价格表一年改价十八次以上;本书的默认做法是底线与套利带两条硬触发、0.03 阈值加两周持续期的漂移触发,再配上 0.03 的压力缓冲。第七章把这些规则装进矩阵引擎,并写成试算器的规格。 (https://5dlibrary.com/answers/zh/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c06/) - Ch.7 Rolling Out the Southeast Asia–Taiwan–Japan Pricing Matrix: Four Product Classes, Update Rhythm and the Calculator Specification — Avoid rule: if a market sets avoid, drop the lattice points whose whole part contains any of those digits. Golden slices (to be compared item by item with the slices p70 prints): USD between 40 and 70 is 49.00, 49.90, 49.99, 59.00, 59.90, 59.99, 69.00, 69.90, 69.99; MYR between 120 and 160 is 129.00, 129.90, 139.00, 139.90, 149.00, 149.90, 159.00, 159.90; TWD from 980 to 1,299 is 980, 988, 990, 999, 1,080, 1,088, 1,090, 1,099, 1,180, 1,188, 1,190, 1,199, 1,280, 1,288, 1,290, 1,299; JPY between 4,980 and 6,000 has one 80-ending point per hundred, from 4,980 to 5,980; the round MYR points run from 900.00 to 1,200.00 in steps of 50; the round JPY points run from 30,000 to 40,000 in steps of 1,000. (https://5dlibrary.com/answers/en/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c07/) 第7章 东南亚台日定价矩阵落地:四类产品、更新节奏与试算器规格 — 回避规则:若市场设了 avoid,删掉整数部分含有这些数字的格点。黄金切片(实现后必须逐项比对 p70 打印的切片):USD 在 40 到 70 之间是 49.00、49.90、49.99、59.00、59.90、59.99、69.00、69.90、69.99;MYR 在 120 到 160 之间是 129.00、129.90、139.00、139.90、149.00、149.90、159.00、159.90;TWD 从 980 到 1,299 是 980、988、990、999、1,080、1,088、1,090、1,099、1,180、1,188、1,190、1,199、1,280、1,288、1,290、1,299;JPY 在 4,980 到 6,000 之间每一百一个 80 尾,从 4,980 到 5,980;MYR 的整数格点从 900.00 到 1,200.00 每 50 一格;JPY 的整数格点从 30,000 到 40,000 每 1,000 一格。 (https://5dlibrary.com/answers/zh/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c07/) - Ch.8 Practice: Mistharbor Cards Launches in Five Markets, from One Dollar Price through the Four Steps — Chapter 8 takes an invented card brand from 49 dollars through the four steps: the final prices are 49.00 dollars in the United States, RM 149.00 in Malaysia (lifted by the band from RM 129.90), S$ 59.90 in Singapore, NT$ 1,080 in Taiwan and ¥5,580 in Japan; the collector's box has a ratio of only 1.1344 between its highest and lowest dollar-equivalent prices, against 1.4536 for the everyday deck; simulated exchange-rate paths push the Japanese margin down to 0.5433 and trigger the first review; and at invented traffic a single market needs 97 to 257 weeks to detect a ten per cent gap in gross profit, so the tests address only large decisions and pool the five markets. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it. (https://5dlibrary.com/answers/en/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c08/) 第8章 实战:雾港牌卡五地上市,从一个美元价走完矩阵四步 — 第八章让一个虚构牌卡品牌从 49 美元出发走完四步,最终价格是美国 49.00 美元、马来西亚 RM 149.00(被套利带从 RM 129.90 提上来)、新加坡 S$ 59.90、台湾 NT$ 1,080、日本 ¥5,580;典藏礼盒五地美元等值价的最大最小之比只有 1.1344,日常牌卡是 1.4536;模拟的汇率路径让日本的毛利率跌到 0.5433,最先触发复核;在虚构流量下,单个市场要分辨一成的毛利差距需要 97 到 257 周,所以测价只测大的决定,五地合并分析。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它。 (https://5dlibrary.com/answers/zh/purchasing-power-parity-distortions-multi-currency-psychological-price-endings/c08/) ## 5DL-B-135 · Hosted-Platform Retainers, Value-Anchored Revenue-Share Hybrids and Customer Lifetime Gross-Margin Lock-In 中文书名:技术底座托管月费、价值锚定抽成混合契约与客户终身毛利锁定母典 DOI: pending Cite: Fan, Y. (2026). Hosted-Platform Retainers, Value-Anchored Revenue-Share Hybrids and Customer Lifetime Gross-Margin Lock-In (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it. Abstract: This volume answers one question for owners who host technology for clients: when you run someone's order, reconciliation, booking or API system, how much should you charge each month, how should the bill move as the client's business rises and falls, and how do you make the client want to stay? The method has three steps. Write a value card from the client's own records, converting team time saved and errors avoided into money; split the money with a two-part contract of a retainer plus a tiered share above an included volume, with a cap on the bill, so that the supplier takes at most 0.45 of measured value; and compute lifetime gross margin with a shifted-beta-geometric retention model. Clients are held by data portability and service levels, not by being unable to leave. Every firm and figure is invented; contracts and law are covered only within the general frameworks of Malaysia and Taiwan, and nothing is legal, tax or investment advice. All calculations come from the book's programs. Chapter 1 sets the book's sample, an invented middle-office supplier, Bridgecalc, and a purchasing-agent house, Lanyu, with about fifteen hundred orders a month. The system gives Lanyu RM 5,568.75 of measurable value a month, of which the list prices of three feature modules cover only 0.116; a pure retainer stops covering the cost to serve above 5,681.8 orders, an 8% pure share passes the client's cost of building its own system at 4,062.5 orders, while a hybrid capped at RM 4,800 still leaves the client RM 17,475 a month at 6,000 orders; client lifetime is only 15.00 months under constant churn against 32.08 months under sBG; and a grown pure-share client need only churn 0.0338 more a month for the pure share to lose to the hybrid. Chapter 1 also grades evidence from A to D and reads the Shanxi merchants' labour-share system as the canonical source, using a moral-hazard model to show why a share takes only a fraction, why longer settlement allows a fuller share (0.806 under monthly settlement, 0.980 under yearly), and why renewal rates cannot be tested by a small supplier. Chapters 2 to 7 cover measuring value, two-part pricing, revenue-sharing contracts, retention and lifetime gross margin, contract risk, and four kinds of hosting business (purchasing-agent middle offices, enterprise systems, studio bookings and API gateways) with the distillation spec for a pricing calculator. Chapter 8 has Bridgecalc design a three-tier contract menu for Lanyu: four rules stop two tiers at the draft stage; three growth paths show that the tier a client picks states its own forecast, so a client that foresees its prospects costs Bridgecalc RM 22,236 and RM 28,421 of menu cost on the steady and surge paths; and the evaluation plan is registered before go-live, with a direct re-measurement on day 90 and an asymmetric clause that re-anchors the retainer. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it. 摘要:本卷为替客户托管技术底座的老板回答一个问题:订单、对账、预约或接口系统由你运行,每个月该向客户收多少、账单怎样随客户的生意涨落、怎样让客户愿意一直留下。书中的方法分三步:先用客户自己的记录写一张价值卡,把省下的人手时间与少犯的错折算成钱;再用「保底月费加超额阶梯抽成、账单封顶」的两部制合同分账,供应商最多分走量得到的价值的 0.45;最后用移位贝塔几何留存模型算终身毛利。锁定靠数据可携与服务水平,而不是让客户走不掉。书中的公司与数字全部虚构,合同与法规只写马来西亚与台湾的一般框架,不是法律、税务或投资意见。全部计算由书中程序生成。 第一章以虚构的代购中台供应商桥算与每月约一千五百单的代购商行澜屿为全书样本:系统每月替澜屿量得到的价值是 RM 5,568.75,三个功能模块的标价只占其中 0.116;纯月费在 5,681.8 单以上盖不住服务成本,8% 的纯抽成在 4,062.5 单越过客户自建系统的成本,而封顶 RM 4,800 的混合合同在 6,000 单时仍让客户每月留下 RM 17,475;按固定流失率算客户寿命只有 15.00 个月,按 sBG 是 32.08 个月;做大的纯抽成客户只要每月多流失 0.0338,就输给混合合同。第一章还按 A 至 D 四级给证据分级,并以晋商身股制为原典,用一个道德风险模型说明为什么分成只取一部分、为什么结算期越长分成可以给得越足(按月结算时最优分成 0.806,按年 0.980),以及为什么续约率对小供应商检验不起。第二至第七章分别写价值测量、两部制定价、收益共享合同、留存与终身毛利、合同风险,以及代购中台、企业系统、工作室预约与收费站接口四类托管的落地与定价试算器的蒸馏规格。第八章让桥算为澜屿设计三档合同菜单:四条规则在初稿拦下两档,三条增长路径显示客户选哪一档就是在说出自己的预期,看得准前景的客户让桥算在平稳与高速路径上各付出 RM 22,236 与 RM 28,421 的菜单代价;评估方案在上线前登记,第九十天直接复测并按不对称条款重新锚定保底月费。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它。 Chapter answers: - Ch.1 Method: Three Ways to Misprice Hosting, the Value Anchor and the Two-Part Contract — Chapter 1 sets the book's sample, an invented middle-office supplier, Bridgecalc, and a purchasing-agent house, Lanyu, with about fifteen hundred orders a month. The system gives Lanyu RM 5,568.75 of measurable value a month, of which the list prices of three feature modules cover only 0.116; a pure retainer stops covering the cost to serve above 5,681.8 orders, an 8% pure share passes the client's cost of building its own system at 4,062.5 orders, while a hybrid capped at RM 4,800 still leaves the client RM 17,475 a month at 6,000 orders; client lifetime is only 15.00 months under constant churn against 32.08 months under sBG; and a grown pure-share client need only churn 0.0338 more a month for the pure share to lose to the hybrid. (https://5dlibrary.com/answers/en/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c01/) 第1章 方法:托管定价的三种错、价值锚与两部制合同 — 第一章以虚构的代购中台供应商桥算与每月约一千五百单的代购商行澜屿为全书样本:系统每月替澜屿量得到的价值是 RM 5,568.75,三个功能模块的标价只占其中 0.116;纯月费在 5,681.8 单以上盖不住服务成本,8% 的纯抽成在 4,062.5 单越过客户自建系统的成本,而封顶 RM 4,800 的混合合同在 6,000 单时仍让客户每月留下 RM 17,475;按固定流失率算客户寿命只有 15.00 个月,按 sBG 是 32.08 个月;做大的纯抽成客户只要每月多流失 0.0338,就输给混合合同。 (https://5dlibrary.com/answers/zh/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c01/) - Ch.2 Value Anchoring: Team Hours Saved, Errors Avoided and Growth Absorbed — Team cost the client no longer spends, errors it no longer makes and growth it can absorb are the three sources of a hosted system's value, and each of them gets overstated. Time value must be realised in whole people: Lanyu's notional figure is RM 4,218.75, the realisable figure RM 3,600, and which one is larger changes with volume. Error value goes in at its conservative level: a one-month pilot gives a point estimate of RM 1,405.00 and a conservative value of RM 544.95, rising to RM 1,020.84 only after 3 months. Expansion value averages RM 2,437.50 a month, appears only when growth actually comes, and must not be counted again on top of time value. The baseline takes at least 246 timed orders and a six-month average, never the month that triggered the purchase. Most important is what to compare against: with average-month staffing and the rival software on the table, Lanyu's economic value is only RM 1,557.00 and the negotiating zone is empty; with busy-month staffing the zone runs from RM 1,180.00 to RM 2,505.94, the even-power midpoint is RM 1,842.97, and Chapter 1's RM 1,900 falls inside it. The next chapter splits the fee into a base and a usage-based part, so that the value that grows with volume also has somewhere to go. (https://5dlibrary.com/answers/en/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c02/) 第2章 价值锚定:省下的人手、少犯的错与接得住的增长 — 替客户省下的人手成本、少犯的错与接得住的增长,是托管系统价值的三个来源,每一个都会被高估。时间价值要按整人兑现:澜屿的名义值 RM 4,218.75,可兑现 RM 3,600,而且两者的大小关系随订单量忽高忽低。错误价值要按保守值入账:一个月试跑的点估计 RM 1,405.00,保守值 RM 544.95,试跑满 3 个月才升到 RM 1,020.84。扩张价值平均每月 RM 2,437.50,只在增长真来时出现,不能与时间价值重复计算。基线要量够 246 单,取六个月平均,不取触发采购的那个月。最关键的是拿什么作比较:按平均月份排人、对手软件摆在桌上时,澜屿的经济价值只有 RM 1,557.00,谈判区间是空的;按旺季排人时,区间是 RM 1,180.00 到 RM 2,505.94,对等议价的中点 RM 1,842.97,第一章的 RM 1,900 落在其中。下一章把月费拆成保底与按量两部分,让随增长放大的那部分价值也有地方安放。 (https://5dlibrary.com/answers/zh/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c02/) - Ch.3 Two-Part Tariffs: Base Fees, Usage Pricing, Tiers and Caps — A two-part tariff splits the monthly fee into two numbers, each with its own job: the entry fee covers fixed cost and divides the value both sides are sure of; the unit price divides growth, and the risk of shrinking too. Five contracts that all charge RM 1,900 at 1,500 orders spread at 4,500 orders from 0.114 to 0.341 of value. Oi's Disneyland dilemma shows that with clients of different sizes the unit price belongs above marginal cost, and that a single two-part tariff tempts the supplier to drop small clients; the answer is a menu. A base fee comes in three shapes, which differ in the slack season and in the marginal price. The meter must rise and fall with value, and seats do the opposite. The "Balanced" tier's declining schedule is the lower envelope of two two-part plans plus a base and a cap. The RM 4,800 cap costs Bridgecalc only 0.033 of revenue on fast-growth paths while cutting the largest one-month jump from RM 3,487.30 to RM 2,127.96. Clients choosing for themselves lean towards the flat rate, and a yearly best-bill true-up is the lowest-cost correction. The next chapter turns to hybrids that take a share of the service fee, and to how incentives line up, and fall out of line, inside them. (https://5dlibrary.com/answers/en/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c03/) 第3章 两部制:保底月费、按量计费、阶梯与封顶 — 两部制把月费拆成两个数,各管一件事:入门费覆盖固定成本、分配双方都确定的那部分价值;单价分配增长,也分配萎缩的风险。五份在 1,500 单时都收 RM 1,900 的合同,到了 4,500 单,价值占比从 0.114 散到 0.341。Oi 的迪士尼困境说明,客户规模不同时单价要高于边际成本,而统一的两部制会诱使供应商放弃小客户;解法是菜单。保底有三种形状,差别在淡季与边际价格。计量单位要与价值同涨同落,座位数恰恰相反。「衡」档的递减阶梯就是两份两部制的下包络,加一个保底与一个封顶。RM 4,800 的封顶只让桥算在快速增长的路径上放弃 0.033 的收入,却把最大的单月跳升从 RM 3,487.30 压到 RM 2,127.96。客户自选方案时会偏向包月,一年一次的最优账单追溯是成本最低的修正。下一章接着讨论按服务费抽成的混合合同,以及激励在其中怎样对齐、又怎样错位。 (https://5dlibrary.com/answers/zh/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c03/) - Ch.4 Revenue Sharing and Tiered Excess Shares: Floors, Caps, and the Arithmetic of Aligned and Misaligned Incentives — A share must first say what it is a share of. A per-order base is easiest to reconcile, a measured-value base best rewards the supplier's real improvements, and a base of the client's revenue or profit turns the client's own pricing and cost changes into disputes. When both sides must put in effort, the best share depends on whose effort matters more; in p42's sample the client's effort matters more and the share is below thirty percent, while the split of the pie is left to the fixed fee rather than patched with the share. A pure share runs into the client's alternative cost once the client grows; a pure retainer makes the supplier lose money once the client grows. A hybrid's cap belongs below the client's alternative cost and far from the supplier's loss point, its tiers belong in incremental form, and its floor belongs at the level of the client's slowest month. The next chapter connects these monthly bills to the client's whole life, and shows how retention, discounting and acquisition cost decide what a contract is finally worth. (https://5dlibrary.com/answers/en/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c04/) 第4章 收益共享与超额阶梯抽成:上下限、激励对齐与错位的算术 — 抽成先要问「抽的是什么」:按单计提最容易核对,按实测价值计提最能奖励供应商的真实改进,按营收或毛利计提会把客户自己的定价与成本变化变成双方的争执。两边都要出力时,最优抽成比例由谁的投入更关键决定;p42 的样本里客户的投入更关键,比例不到三成,分配的问题交给固定费去解决,不要拿比例去补。纯抽成在客户做大后撞上替代方案的成本线,纯月费在客户做大后让供应商亏钱;混合合同的封顶要低于客户的替代成本、远离供应商的亏损点,阶梯要逐段累进,保底要按客户最淡的月份定。下一章把这些月度账单接到客户的整个生命周期上,看留存、折现与获客成本怎样决定一份合同最终值多少钱。 (https://5dlibrary.com/answers/zh/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c04/) - Ch.5 Customer Lifetime Gross Margin: sBG Retention, Cohort Accounting, Discounting and Acquisition Payback — Lifetime gross margin is decided by three numbers unrelated to any monthly bill: how long the client stays, the time value of money, and what signing cost. Retention must be estimated with an sBG model that admits clients differ; constant churn is not just rougher but points the wrong way, and across p50's two hundred cohorts it systematically understates long-run retention. Seven months of data already give an unbiased estimate, though a single cohort can still be off by a few months. Retention must be read by signing cohort; book-wide churn mostly reflects the sales rhythm and overstates a new client's value over two years while understating it over ten. The discount rate shifts the answer by a third or so, the horizon by more, and acquisition and pricing decisions use a sixty-month horizon. Payback is counted per client signed and discounted, with the share that never pays back reported alongside. Among the four levers, on a common definition, price and churn are the strongest, and service spending pays for itself with a churn cut of less than a tenth. Lock-in comes from service and portability, not from locked doors. The next chapter writes the promises behind these numbers into contract clauses. (https://5dlibrary.com/answers/en/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c05/) 第5章 客户终身毛利:sBG 留存、按批次核算、折现与获客回本期 — 终身毛利由三个与单月账单无关的数决定:留多久、钱的时间价值、签下来花了多少。留存要用承认客户之间差异的 sBG 模型去估,常数流失率不是粗一点而是方向错,p50 的两百个批次里它系统性地低估长期留存;七个月的数据已经能给出无偏的估计,但单个批次的误差仍有几个月。留存要按签约批次看,全体流失率主要反映销售节奏,在头两年高估、在十年期限上低估新客户的价值。折现率改变结论三四成,核算期限的影响更大,获客与定价决策用六十个月期限。回本期要按每签一家、折现后计,同时报告收不回来的比例。四个杠杆里,按同样口径比较,提价与降流失最有力,而服务投入只需降低不到一成的流失就能回本;锁定靠的是服务与可携,而不是锁死退出。下一章把这些数字背后的承诺写成合同条款。 (https://5dlibrary.com/answers/zh/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c05/) - Ch.6 Contract Risk: Service Levels, Minimum Terms, Exit, Price Adjustment and Disputes — The first five chapters worked out the price. Lanyu can measure RM 5,568.75 of value a month; the "Balanced" contract charges a retainer of RM 1,900 covering 1,500 orders, a tiered share above that and a cap of RM 4,800; and we know its lifetime gross margin under sBG retention and how many months it takes to pay back. But the price lives on the quote, and the risk lives in the contract. The same RM 1,900 monthly bill, paired with different service-level, termination, price-adjustment and dispute clauses, can mean twice as much or half as much to the two firms. This chapter takes apart the seven places where hosting contracts most often go wrong. For each one it does the arithmetic first, then sketches how the clause is usually written, and ends with a decision rule that can be checked afterwards. (https://5dlibrary.com/answers/en/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c06/) 第6章 合同风险:服务水平、最短期限、退出、调价与争议 — 前五章把价格算出来了:澜屿每月量得到的价值 RM 5,568.75,「衡」档合同每月保底 RM 1,900、含 1,500 单,超出部分按阶梯抽成,封顶 RM 4,800,按 sBG 留存折现后的终身毛利是多少、几个月回本,都有了数字。可是价格写在报价单上,风险写在合同里。同一张 RM 1,900 的月账单,配上不同的服务水平条款、解约条款、调价条款与争议条款,对两家公司的真实含义可以差出一倍。本章把托管合同里最容易出事的七处逐一拆开,每一处都先算账,再写条款的一般写法,最后给一条事后能核对的决策规则。 (https://5dlibrary.com/answers/zh/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c06/) - Ch.7 The Hosting Playbook: Four Kinds of Client and the Hosting-Contract Calculator — For Lanyu the engine reproduces Chapter 1: value per order 3.7125; a first-month bill of RM 1,900.00 against value of RM 5,568.75, a share of 0.341, and a Bridgecalc gross margin of 0.484. Over ten years the client is expected to pay for 32.08 months, and after RM 4,800 of onboarding the lifetime gross margin is RM 32,562.83, which rounds to exactly the RM 32,563 of p13 in Chapter 1; with a constant churn rate of 0.0667 instead, the engine likewise returns Chapter 1's RM 11,958. Discounted payback comes in month 6. The monthly bills of the four contracts, joined line by line as "client|month|bill", hashed with SHA-256 and cut to the first 16 hex digits, give the fingerprint f3e804c0b4d48f8d; any parameter or line of code that changes any month's bill changes the fingerprint. (https://5dlibrary.com/answers/en/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c07/) 第7章 托管落地手册:四类客户与托管合同试算器 — 对澜屿,引擎复现了第一章的数字:每单价值 3.7125,第一个月账单 RM 1,900.00、价值 RM 5,568.75,分成比例 0.341,桥算毛利率 0.484;十年内期望付费 32.08 个月,扣掉 RM 4,800 的导入成本后,终身毛利 RM 32,562.83,四舍五入正是第一章 p13 的 RM 32,563;若改用固定流失率 0.0667,引擎同样得到第一章的 RM 11,958,第 6 个月折现回本。四份合同的逐月账单按「客户名|月份|账单」逐行拼接后做 SHA-256,取前 16 位,指纹是 f3e804c0b4d48f8d;任何一个参数、任何一行代码改动了任何一个月的账单,指纹就会变。 (https://5dlibrary.com/answers/zh/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c07/) - Ch.8 Practice: Bridgecalc Designs a Three-Tier Contract for Lanyu — Chapter 8 has Bridgecalc design a three-tier contract menu for Lanyu: four rules stop two tiers at the draft stage; three growth paths show that the tier a client picks states its own forecast, so a client that foresees its prospects costs Bridgecalc RM 22,236 and RM 28,421 of menu cost on the steady and surge paths; and the evaluation plan is registered before go-live, with a direct re-measurement on day 90 and an asymmetric clause that re-anchors the retainer. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it. (https://5dlibrary.com/answers/en/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c08/) 第8章 实战:桥算为澜屿设计三档合同 — 第八章让桥算为澜屿设计三档合同菜单:四条规则在初稿拦下两档,三条增长路径显示客户选哪一档就是在说出自己的预期,看得准前景的客户让桥算在平稳与高速路径上各付出 RM 22,236 与 RM 28,421 的菜单代价;评估方案在上线前登记,第九十天直接复测并按不对称条款重新锚定保底月费。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它。 (https://5dlibrary.com/answers/zh/hosted-platform-retainers-value-anchored-revenue-share-hybrids-customer/c08/) ## 5DL-B-136 · Luxury Scarcity Allocation Auctions, Early-Bird Price-Ladder Dynamics and Limited-Edition Physical Premiums 中文书名:高奢稀缺配额拍卖、早鸟阶梯涨价动力学与限量实体品溢价母典 DOI: pending Cite: Fan, Y. (2026). Luxury Scarcity Allocation Auctions, Early-Bird Price-Ladder Dynamics and Limited-Edition Physical Premiums (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Legal rules on such claims are described only in general terms, not as legal advice. Marked boxes separate evidence, calculations, sources and hypotheses, each hypothesis stating what would refute it; the classical source is the Han Feizi parable of buying the casket and returning the pearl, read as a lesson on packaging premiums. Abstract: This volume answers the questions every seller of limited goods faces: how to price a numbered card deck, a small workshop or a run of numbered prints; whether to offer an early-bird ladder; who gets the goods when there are not enough to go round; and what the word "limited" is actually worth. Its claim is that "limited" is worth something only when the seller can be held to it. The Coase conjecture says that a seller who can always make more of the same thing, and cannot commit not to, will find buyers waiting, and the scarcity premium disappears; serial numbers, a public counter, destroying the printing plates in public and a scoped no-reprint statement are the commitment devices that hold the seller to its word. Every early-bird step must be real, announced in advance and closed when sold out, and the allocation rule decides both revenue and how the community feels. Legal rules on such claims are described only in general terms, not as legal advice. All brands, buyers, bids and resales are invented samples, and every calculation is produced by the book's programs. Chapter 1 is the method. In the sample of 500 numbered card sets from an invented brand, Moontide, gross margin is RM 120,874 with no cap on quantity and falls to RM 103,556 when the quantity is fixed at 500; for the limit to pay, a believed limit must raise collectors' willingness to pay by more than about eleven percent (break-even 0.1174). When buyers put the chance of a reprint at one half, the clearing price drops to RM 282.76, below the base price; and showing by a test that one allocation rule halves the complaint rate needs 1272 orders per rule, about 2.54 sales. Chapters 2 to 7 cover commitment devices, ladders, six allocation mechanisms, scarcity psychology, resale and legal limits, and playbooks with a pricing-calculator specification. Chapter 8 takes Moontide through the whole method: on the same buyers, a uniform-price auction collects RM 16,297 more than a lottery at list price; under the registered prior, the three-step ladder is expected to earn RM 4,416 less and pays only if soft demand has a probability above 0.715; ordering by speed sends RM 4,795 of the early-bird discount to resale accounts, and identity checks with a random order cut that to RM 903; the day-ninety median resale premium is 0.184, and weekly readings flicker, so actions wait for the registered decision point; over the long run, the auction stops paying once it lowers repurchase by 0.0414, a gap that takes several sales to measure. Marked boxes separate evidence, calculations, sources and hypotheses, each hypothesis stating what would refute it; the classical source is the Han Feizi parable of buying the casket and returning the pearl, read as a lesson on packaging premiums. 摘要:本卷回答发售限量品的人都会遇到的问题:一件编号牌卡、一场只收几十人的工作坊、一批编号版画,怎样定价、要不要设早鸟阶梯、货不够分时由谁买到,以及「限量」两个字到底值多少钱。本书的主张是:「限量」只有在卖家能被约束住的时候才值钱。科斯猜想说,能够随时再造同样东西的卖家若不能承诺不再加印,买家就会等待,稀缺溢价随之消失;编号与证书、公开计数、当众销毁印版、写清范围的不再版声明,都是让卖家被约束住的承诺装置。早鸟阶梯的每一级必须真实、事先公布、卖完即关;配额怎样分,决定收入,也决定社群感受。「限量」「早鸟」宣称的消费者保护规范只写一般框架,不是法律意见。全部品牌、买家、出价与转售都是虚构样本,全部计算由书中程序生成。 第一章是方法:在虚构品牌「月汐」500 套编号牌卡的样本里,不设数量上限时毛利是 RM 120,874,把数量钉在 500 套反而降到 RM 103,556,限量要赚钱,被相信的限量必须让收藏者多愿付一成一以上(盈亏平衡点 0.1174);买家心中的再版概率升到一半,出清价就跌到 RM 282.76,低于基础价;要用检验分辨两种分配规则的投诉率减半,每种规则需要 1272 笔订单,约 2.54 次发售。第二至第七章依次写科斯猜想与承诺装置、早鸟阶梯、六种配额机制、稀缺心理与包装溢价、二级市场与合规边界,以及四类限量品的手册与定价试算器的新规格。第八章让月汐完整走一遍:同样的买家,统一价拍卖比按标价抽签多收 RM 16,297;在登记的先验下,三级阶梯期望少赚 RM 4,416,只有需求疲软的概率高于 0.715 才划算;按速度排队时 RM 4,795 的早鸟折扣落到转卖账户,实名核验加随机排序把它降到 RM 903;第九十天的挂牌溢价中位数 0.184,周度读数会闪烁,动作只在登记好的判定点上做;从长期看,拍卖只要让复购率下降 0.0414 就已不划算,这个差别要多次发售才量得出来。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它;原典考辨取《韩非子·外储说左上》的「买椟还珠」,讨论包装溢价与内容价值的分离。 Chapter answers: - Ch.1 Method: Three Sources of a Scarcity Premium, Four Indicators, Evidence Grades and "Buying the Casket, Returning the Pearl" — Chapter 1 is the method. In the sample of 500 numbered card sets from an invented brand, Moontide, gross margin is RM 120,874 with no cap on quantity and falls to RM 103,556 when the quantity is fixed at 500; for the limit to pay, a believed limit must raise collectors' willingness to pay by more than about eleven percent (break-even 0.1174). When buyers put the chance of a reprint at one half, the clearing price drops to RM 282.76, below the base price; and showing by a test that one allocation rule halves the complaint rate needs 1272 orders per rule, about 2.54 sales. (https://5dlibrary.com/answers/en/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c01/) 第1章 方法:稀缺溢价的三个来源、四个指标、证据分级与「买椟还珠」考辨 — 第一章是方法:在虚构品牌「月汐」500 套编号牌卡的样本里,不设数量上限时毛利是 RM 120,874,把数量钉在 500 套反而降到 RM 103,556,限量要赚钱,被相信的限量必须让收藏者多愿付一成一以上(盈亏平衡点 0.1174);买家心中的再版概率升到一半,出清价就跌到 RM 282.76,低于基础价;要用检验分辨两种分配规则的投诉率减半,每种规则需要 1272 笔订单,约 2.54 次发售。第二至第七章依次写科斯猜想与承诺装置、早鸟阶梯、六种配额机制、稀缺心理与包装溢价、二级市场与合规边界,以及四类限量品的手册与定价试算器的新规格。 (https://5dlibrary.com/answers/zh/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c01/) - Ch.2 The Coase Conjecture and Commitment Devices: A Limit Must First Bind the Seller — On Moontide's demand, the Coase conjecture is a concrete bill. p20 shows that for a seller able to reprint at any time, the more often it reprints the lower its 24-month present-value margin: from RM 121,226 for a single sale down to RM 96,921 with monthly reprints, RM 24,341 less than a believed limit. Bulow reminds us to separate consumable from persistent scarcity; workshop seats are bound by the calendar, numbered sets need devices. Stokey's result, checked by p24, says an announced schedule of cuts cannot beat a single price, and when everyone is present at once neither can a rising schedule, so if an early-bird ladder is useful the reason lies elsewhere. p21 stacks the four devices plus an optional pledge: numbering restores RM 23,518 of margin at RM 289, the later devices leave sales unchanged but push the clearing price up to RM 335.42. p23 shows that numbering deters through the registry; at a registration share of 0.4, twenty extra sets are almost certain to come out. p22 lays out the temptation of a breach: after plate destruction and the pledge RM 3,607 remains, and what really makes a limit credible is that the brand has a future. The next chapter turns to early-bird ladders and asks when they actually pay. (https://5dlibrary.com/answers/en/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c02/) 第2章 科斯猜想与承诺装置:限量要先把卖家自己绑住 — 科斯猜想在月汐的需求上是一张具体的账单。p20 显示,可以随时再印的卖家,再印越频繁,24 个月的现值毛利越低:从一次性的 RM 121,226 降到每月再印的 RM 96,921,比被相信的限量少 RM 24,341。Bulow 提醒我们区分消耗型与存留型的稀缺,工作坊席位天然被日历约束,编号牌卡则必须另设装置。Stokey 的结论经 p24 验证:事先公布的降价表赚不过一口价,人人同时到场时上升的价格表也赚不过,早鸟阶梯若有用,理由必然在别处。p21 把四件装置加一件可选的补差承诺排成一叠:编号把 RM 289 上的毛利补回 RM 23,518,后面的装置不再改变销量,却把出清价推高到 RM 335.42。p23 说明编号的威慑来自登记簿,登记率 0.4 时多印 20 套几乎一定败露。p22 则把违约的诱惑摊开:销毁印版与补差承诺之后仍剩 RM 3,607,真正让限量可信的,是品牌还有未来。下一章转到早鸟阶梯,看它在什么条件下真的赚钱。 (https://5dlibrary.com/answers/zh/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c02/) - Ch.3 Early-Bird Ladder Dynamics: Tiers, Steps, Slopes and Honest Announcement — An early-bird ladder does not make money out of thin air. p30 places Moontide's three-tier ladder in its geometry: an average price of RM 275.00, RM 7,000 below a single price, a largest relative step of 0.0803 and a slope of RM 16.00 per hundred serials. p31 shows that with myopic buyers and excess demand the ladder only moves stock from late high-value collectors to early low-value ones, earning RM 6,790 less and raising the number left out from 70.8 to 120.4. p32 shows the ladder's real use is to deter waiting: the same ladder lifts the tipping belief from 0.225 to 0.387. p33 shows a presale discount earns a little more when buyers are uncertain, only RM 1,037 in Moontide's sample, and about two in ten presale buyers regret it. p34 turns price protection into a sum that can be provided for in advance, and p35 turns "the ladder is real" into six checks outsiders can repeat. No more than three tiers, a last tier equal to the base price, only rises, closed when sold out, refunds whenever the price is cut: these five rules are what this chapter hands to Chapter 7's pricing calculator. The next chapter asks who should get the goods when there are not enough to go round. (https://5dlibrary.com/answers/en/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c03/) 第3章 早鸟阶梯动力学:级数、步长、斜率与真实公布 — 早鸟阶梯本身不会凭空赚钱。p30 把月汐的三级阶梯放回几何里:平均价 RM 275.00,比一口价少 RM 7,000,最大相对步长 0.0803,斜率每百号 RM 16.00。p31 显示,面对短视的买家与超额的需求,阶梯只是把货从晚到的高估值收藏者挪给早到的低估值收藏者,少赚 RM 6,790,还让落空的人从 70.8 位增加到 120.4 位。p32 显示阶梯真正的用处在于防等待:同一条阶梯把崩盘的临界预期从 0.225 抬到 0.387。p33 显示预售折扣在买家不确定时能多赚一点,月汐的样本里只有 RM 1,037,而且约两成预售买家会后悔。p34 把降价补差算成一笔事先可以准备的钱,p35 把「阶梯真实」变成六项可以由外人复核的检查。级数不超过三、末级等于基础价、只升不降、卖完即关、降价必补差,这五条是本章交给第七章定价试算器的规则。下一章讨论货不够分的时候,该由谁买到。 (https://5dlibrary.com/answers/zh/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c03/) - Ch.4 Who Gets the Allocation: First Come, Queues, Lotteries and Three Auctions — The design does not change how many sets there are; it changes three things: how much the brand collects, whose hands the sets reach, and what the community remembers. In this chapter's pool, the uniform-price auction earns RM 15,161 more than the fixed RM 289; the pay-as-bid auction earns almost the same in equilibrium but is harder to explain; the Vickrey auction differs from the uniform price only when one person takes several sets. First come hands the allocation to scripts, a queue burns the gap as time, and a lottery leaves the gap to the winners and the resale market. ID checks cut lottery resale from 108.1 sets to 32.4, and two tickets for members lift the core collectors' chance to 0.865. For a low-price ID lottery to beat the auction, each marginal collector it serves must bring back RM 129.57 of margin in the future, about two thirds of a set's margin; that is a bet on repeat buying, written here as a hypothesis and handed to the Chapter 8 launch to test. (https://5dlibrary.com/answers/en/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c04/) 第4章 配额怎么分:先到先得、排队、抽签与三种拍卖 — 分法不改变货的数量,改变的是三件事:品牌收多少、货落到谁手里、社群记住什么。在本章的需求池里,统一价拍卖比定价 RM 289 多收 RM 15,161;按出价付费拍卖在均衡下收得几乎一样多,却更难解释;维克里拍卖只在一人多套时才与统一价不同。先到先得把配额交给脚本,排队把差额烧成时间,抽签把差额留给中签者与转售市场。实名把抽签里的转卖量从 108.1 压到 32.4,会员两张签把核心收藏者的中签率推到 0.865。低价实名抽签要胜过拍卖,每一位被它服务的边际收藏者需要在将来带回 RM 129.57 的毛利,约合三分之二套牌卡的毛利;这是一个关于复购的赌注,本章把它写成假说,交给第八章的发售去检验。 (https://5dlibrary.com/answers/zh/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c04/) - Ch.5 Scarcity Psychology: Evidence, Replication Status, Bandwagons and Snobs, and the Modern Casket — Research on scarcity psychology gives directions but not sizes, and most of it has not been replicated at scale: the cookie jar, limited-quantity messages and reactance are all grade C evidence. p50 and p51 show that experiments with twenty or thirty people per group often cannot detect even a medium effect, and that "publish only if significant" inflates effects by more than three times. On the social side, the bandwagon makes demand steeper and, over a band of prices, produces two equilibria; snobbery makes demand sluggish; a Veblen effect should not be assumed for a RM 289 deck; and public counters across several items amplify luck into rankings. Packaging can hold up a price: with a median box value of 30, the wooden box lifts margin to RM 113,501, but it also turns 188 winners into casket buyers. False scarcity is a bet placed again and again that loses more the longer it runs. What Moontide can do is put both the counter and the box into designs that can be tested, full edition against no-box edition and counter against no counter, and get grade A answers from its own buyers. (https://5dlibrary.com/answers/en/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c05/) 第5章 稀缺心理:证据、复制状态、从众与势利,以及买椟还珠的现代版 — 稀缺心理的研究给了方向,没给大小,而且大多没有经过大样本复制:饼干罐、限量讯息、心理抗拒,都是丙级证据。p50 与 p51 说明,每组二三十人的实验连中等效应都常常测不出,「显著才发表」又会把效应吹大三倍以上。社会影响一侧,从众让需求变陡、在某段价位上出现两个均衡,势利让需求变钝,凡勃伦效应在一副 RM 289 的牌卡上不该被假设;公开计数器在多款商品上会把运气放大成排名。包装能撑起价格,木盒在盒子价值中位数为 30 时让毛利升到 RM 113,501,但也让 188 位赢家成了椟客。假稀缺是一笔会重复下注、越下越输的赌。月汐能做的,是把计数器与盒子都放进可以检验的设计里:完整版对无盒版、计数器对无计数器,用自己的买家给出甲级答案。 (https://5dlibrary.com/answers/zh/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c05/) - Ch.6 Secondary-Market Premiums, Name Binding, Transfer Fees, Anti-Bot Defences and the Compliance Edge of "Limited" Claims — The resale premium is the gap between face value and what latecomers will pay, and the early-bird tier is where the gap is widest, so identity controls belong first on the early-bird tier. Resale has a legitimate side, moving goods to people who want them more, but that move should cost close to face value. Name binding is valuable because it means "transfer only by the rules": official face-value transfer plus a waitlist cuts the workshop's empty seats from 4.789 on average to 0.596. A resale share on prints follows a Laffer curve, and the take-maximising rate squeezes out almost half of the trading; this book recommends 0.05 to 0.10. Purchase limits must fall on verified identities, and once scripts are suppressed, the filter's false blocks become the main problem. A "limited" claim has to survive five reconciliations, and its failures fall into two classes: the quantity itself is false, or the devices that make it believable are empty. Chapter 7 writes both classes, together with the ladder and allocation checks, into a new op for the pricing calculator. (https://5dlibrary.com/answers/en/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c06/) 第6章 二级市场溢价、实名绑定、转让费、反机器人与「限量」宣称的合规边界 — 转售溢价是面值与后来者愿付价格之间的缝,早鸟档是这条缝最宽的地方,所以身份约束首先要落在早鸟档。转售有正当的一面:它把货挪给更想要的人,但这次挪动应当以接近面值的成本完成。实名绑定的价值在于「只许按规则转」,官方按面值转让配合候补,能把工作坊的空座从平均 4.789 个压到 0.596 个。版画的转售分成存在一条拉弗曲线,分成收入最大的比例会挤掉近一半的流转,本书建议的区间是 0.05 到 0.10。限购必须落在核验过的身份上;脚本被压下去以后,过滤器的误伤会成为主要问题。「限量」的宣称要经得起五项对账,失败分成两类:数量本身不真实,以及让人相信它的装置是空的。第七章把这两类失败连同阶梯与配额的检查一起,写进定价试算器的新 op。 (https://5dlibrary.com/answers/zh/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c06/) - Ch.7 The Limited-Edition Playbook: Numbered Decks, Workshop Seats, Numbered Prints, Boutique Quotas and the Distillation Spec for the Four Kits — One model applied to four kinds of goods yields four different defaults. Decks keep their pre-announced ladder, but the early tiers are allocated by identity-verified registration and lottery; workshops have no early-bird tier and use a lottery at the regular price, named entry and face-value transfer; prints are issued by lottery and carry a resale share of 0.05 to 0.10 afterwards; collector editions are sized to demand first and then allocated by a published hybrid quota. In this model the auction leads almost everywhere and trails only on collector surplus; its real cost lies outside the model and has to be measured in complaints and repeat purchases. The price split of the wooden-box set reminds the seller that buyers may pay only for the casket, but the seller must say plainly what the casket holds. The last section writes all of this into a new specification for the pricing calculator. (https://5dlibrary.com/answers/en/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c07/) 第7章 限量品落地手册:编号牌卡、工作坊席位、编号版画、精品配额与四件套蒸馏规格 — 同一个模型放在四类东西上,得出四种不同的默认做法。牌卡保留事先公布的阶梯,但早鸟档改为核验身份后的登记抽签;工作坊不设早鸟,正价抽签、实名、按面值转让;版画发行时抽签,发行后收 0.05 到 0.10 的转售分成;典藏版先按需求定版数,再用公开的混合规则配额。拍卖在这套模型里几乎处处占优,只在收藏者剩余上略输,它真正的代价在模型之外,要用投诉与复购去量。木盒版的价格拆解提醒卖方:买家可以只为椟付钱,但卖方必须说清楚椟里装的是什么。最后一节把这一切写成定价试算器的新规格。 (https://5dlibrary.com/answers/zh/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c07/) - Ch.8 Practice: Moontide's 500 Numbered Card Sets, from Pricing, Ladder and Allocation to Resale Monitoring — Chapter 8 takes Moontide through the whole method: on the same buyers, a uniform-price auction collects RM 16,297 more than a lottery at list price; under the registered prior, the three-step ladder is expected to earn RM 4,416 less and pays only if soft demand has a probability above 0.715; ordering by speed sends RM 4,795 of the early-bird discount to resale accounts, and identity checks with a random order cut that to RM 903; the day-ninety median resale premium is 0.184, and weekly readings flicker, so actions wait for the registered decision point; over the long run, the auction stops paying once it lowers repurchase by 0.0414, a gap that takes several sales to measure. Marked boxes separate evidence, calculations, sources and hypotheses, each hypothesis stating what would refute it; the classical source is the Han Feizi parable of buying the casket and returning the pearl, read as a lesson on packaging premiums. (https://5dlibrary.com/answers/en/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c08/) 第8章 实战:月汐五百套编号牌卡,从定价、阶梯、配额到转售监控 — 第八章让月汐完整走一遍:同样的买家,统一价拍卖比按标价抽签多收 RM 16,297;在登记的先验下,三级阶梯期望少赚 RM 4,416,只有需求疲软的概率高于 0.715 才划算;按速度排队时 RM 4,795 的早鸟折扣落到转卖账户,实名核验加随机排序把它降到 RM 903;第九十天的挂牌溢价中位数 0.184,周度读数会闪烁,动作只在登记好的判定点上做;从长期看,拍卖只要让复购率下降 0.0414 就已不划算,这个差别要多次发售才量得出来。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它;原典考辨取《韩非子·外储说左上》的「买椟还珠」,讨论包装溢价与内容价值的分离。 (https://5dlibrary.com/answers/zh/luxury-scarcity-allocation-auctions-early-bird-price-ladder-dynamics/c08/) ## 5DL-B-137 · Mobile Thumb-Zone Kinetics, Dark-Luxury Micro-Interactions and Above-the-Fold Visual Retention Engineering 中文书名:移动端拇指热区动力学、暗黑奢品微交互与首屏视觉留存工程 DOI: pending Cite: Fan, Y. (2026). Mobile Thumb-Zone Kinetics, Dark-Luxury Micro-Interactions and Above-the-Fold Visual Retention Engineering (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All brands and figures are invented samples computed by the book's programs; laws on privacy consent and accessibility are described only in general terms and are not legal advice. Abstract: This volume answers a question that owners of mobile web shops meet every day: why does a customer open a product page, glance at it and leave? Most mobile conversion loss lies not in looks but in four places measurable with a ruler and a stopwatch: the viewport tax of fixed strips eating the first screen, key buttons placed beyond the reach of one thumb, a first screen that appears and settles too slowly, and text whose contrast is too low or too high. Each loss has a page definition, a live-traffic indicator and a checkable decision rule; interface evidence comes in five grades, and only randomised controlled experiments on your own traffic answer "did the change sell more?". Rejecting cookies must be as easy as accepting them, and no stock, viewer counts or countdowns are invented. All brands and figures are invented samples computed by the book's programs; laws on privacy consent and accessibility are described only in general terms and are not legal advice. Chapter 1 is the method and fixes the sample shared by the whole book: the product page of a RM 2,890 handbag at Nightamber, an invented curated boutique, on two viewports of 390×844 and 360×780. On a first visit the current layout pays a viewport tax of 0.289 and 0.313, with neither price nor button on the first screen; 0.714 of seven key actions lie outside the thumb's comfortable zone, and the chat bubble covers 0.116 of the button; when the true 75th percentile of LCP is 2.44 seconds, just inside the threshold, 50 sessions judge it good in only 0.624 of estimates and 3200 sessions are needed to reach 0.988; interface experiments belong on the template shared by forty handbags, where an add-to-bag lift of 0.20 is measurable in 2.3 weeks, whereas a single page with purchase as the metric would need 6.6 years. Chapters 2 to 7 cover thumb zones, the viewport tax of sticky headers, dark-luxury typography, micro-interactions, first-screen arithmetic and playbooks for four kinds of luxury page, with new tokens and checks for a front-end style kit. Chapter 8 takes Nightamber through the whole method: after the redesign the viewport tax falls to 0.258 and 0.267, product facts on the first screen grow to 164 pixels, the price and the button are always visible, and the actions of the purchase path fall in the comfortable zone for both the right-hand and the left-hand model; the pre-registered four-week test, with 19,200 sessions per arm, can detect relative lifts of 0.149 and above, while a lift of 0.10 has only a 0.477 chance of being seen. Marked boxes separate evidence, calculations, sources and hypotheses, each hypothesis stating what would refute it; the source studied is the Kaogongji and its four conditions of good work: "heaven has its seasons, earth has its qi, materials have their beauty, craftsmen have their skill". 摘要:本卷回答一个手机网店老板天天遇到的问题:客人点进单品页,为什么看了一眼就走。手机端的转化损失大多不在「好不好看」,而在四种可以用尺子和秒表量出来的地方:固定横条吃掉第一屏的视口税、关键按钮落在单手拇指够不到之处、首屏出现与稳定得太慢、文字对比不够或者过头。每一种损失都有页面上的定义、流量上的指标与可核对的决策规则;界面证据分五级,只有自己流量上的随机对照实验能回答「改了卖得更好没有」。拒绝 Cookie 必须与接受一样容易,不虚构库存、人数与倒计时。全部品牌与数字都是虚构样本,由书中程序算出;涉及隐私同意与无障碍的法规只写一般框架,不是法律意见。 第一章是方法,并定下全书共用的样本:虚构买手精品店「夜珀」售价 RM 2,890 的手袋单品页,390×844 与 360×780 两种视口。现行版面第一次来访的视口税是 0.289 与 0.313,价格与按钮都不在第一屏;七个关键操作里 0.714 不在拇指舒适区,聊天浮窗压住按钮的 0.116;LCP 第 75 百分位真值 2.44 秒、离阈值只差一点时,50 次会话只有 0.624 的估计判为良好,3200 次才到 0.988;界面实验要在四十只手袋共用的模板上做,加入购物袋提升 0.20 时模板流量 2.3 周可测,而单一页面拿购买当指标要 6.6 年。第二至第七章依次写拇指热区、置顶栏视口税、暗黑奢品排版、微交互、首屏算术与四类奢品页面手册,并给前端样式库写出新增令牌与检查规则。第八章让夜珀走完全程:重排后视口税降到 0.258 与 0.267,第一屏的商品信息增加到 164 像素,价格与按钮始终可见,购买路径的操作在左右手模型下都落在舒适区;预先登记的四周测试每组 19,200 次会话,能检测 0.149 以上的相对提升,0.10 的提升只有 0.477 的把握看出来。书中以标注框区分证据、计算、原典与假说,每条假说都写明什么结果会推翻它;原典考辨取《考工记》「天有时,地有气,材有美,工有巧」的良工四条件。 Chapter answers: - Ch.1 Method: Four Measurable Losses, the Book's Sample, Evidence Grades and A/B Design — Chapter 1 is the method and fixes the sample shared by the whole book: the product page of a RM 2,890 handbag at Nightamber, an invented curated boutique, on two viewports of 390×844 and 360×780. On a first visit the current layout pays a viewport tax of 0.289 and 0.313, with neither price nor button on the first screen; 0.714 of seven key actions lie outside the thumb's comfortable zone, and the chat bubble covers 0.116 of the button; when the true 75th percentile of LCP is 2.44 seconds, just inside the threshold, 50 sessions judge it good in only 0.624 of estimates and 3200 sessions are needed to reach 0.988; interface experiments belong on the template shared by forty handbags, where an add-to-bag lift of 0.20 is measurable in 2.3 weeks, whereas a single page with purchase as the metric would need 6.6 years. (https://5dlibrary.com/answers/en/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c01/) 第1章 方法:四种可测损失、全书样本、证据分级与 A/B 设计 — 第一章是方法,并定下全书共用的样本:虚构买手精品店「夜珀」售价 RM 2,890 的手袋单品页,390×844 与 360×780 两种视口。现行版面第一次来访的视口税是 0.289 与 0.313,价格与按钮都不在第一屏;七个关键操作里 0.714 不在拇指舒适区,聊天浮窗压住按钮的 0.116;LCP 第 75 百分位真值 2.44 秒、离阈值只差一点时,50 次会话只有 0.624 的估计判为良好,3200 次才到 0.988;界面实验要在四十只手袋共用的模板上做,加入购物袋提升 0.20 时模板流量 2.3 周可测,而单一页面拿购买当指标要 6.6 年。第二至第七章依次写拇指热区、置顶栏视口税、暗黑奢品排版、微交互、首屏算术与四类奢品页面手册,并给前端样式库写出新增令牌与检查规则。 (https://5dlibrary.com/answers/zh/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c01/) - Ch.2 The Thumb Zone: the Geometry of What a Finger Can Reach — Fitts' law puts "far" and "small" into one unit: layout A's 32-pixel header icons cost 4.17 bits or more and layout B's bar button 2.31 bits; enlarging the icon to 44 pixels saves less than half a bit, while moving it saves nearly two. Hoober's observations give the shares of the three grips: one hand 49%, cradled 36%, two hands 15%. This chapter's millimetre model shows that for a medium hand on the 390×844 phone the comfortable zone covers only 26.6% of the screen, shaped as a diagonal band; the top quarter of the screen is entirely out of reach and the area near the base of the thumb is awkward too; the larger the screen, the smaller the comfortable share. (https://5dlibrary.com/answers/en/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c02/) 第2章 拇指热区:手指够得着的几何 — Fitts 定律把「远」和「小」换算成同一个单位:A 版页头的 32 像素图标难度在 4.17 比特以上,B 版操作栏的按钮是 2.31 比特,把图标放大到 44 像素只省下不到半个比特,挪位置才省下将近两个比特。Hoober 的观察给出三种握法的比例,单手 49%、托握 36%、双手 15%。本章的毫米模型显示,中等的手握 390×844 时,舒适区只占屏幕的 26.6%,形状是一条斜向的弧带,屏幕上部四分之一完全够不着,拇指根部附近也不好按;屏幕越大,舒适区比例越低。 (https://5dlibrary.com/answers/zh/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c02/) - Ch.3 The Sticky-Header Tax: Browser Bars, Safe Areas, Stacked Banners and Floating Widgets — The viewport tax must be computed against the visible height. With this chapter's assumed browser-bar heights, a first visit on 390×844 has only 664 pixels to work with and on 360×780 only 628. Layout A's fixed layers on a first visit take 0.319 and 0.338 of the visible height, and while reading still 0.139 and 0.146; at 2.0 times text, the two pinned bars eat 0.277. 100vh hides the button at the foot of the hero behind the browser bars, svh does not; dvh makes the content jump by 86 pixels when the bars collapse. A hide-on-scroll header cuts the space taken while reading to less than half, provided the switching threshold is 24 pixels or more, otherwise the header flickers with every twitch of the finger. A late banner pushed into the flow causes layout shift: a 120-pixel cookie bar inserted at the top scores 0.181 on its own. A chat bubble left in its default position covers 0.197 of the action bar's button. Fixed layers also hide focused controls, which scroll-padding reservations fix. (https://5dlibrary.com/answers/en/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c03/) 第3章 置顶栏的视口税:浏览器、安全区、横幅与浮窗的叠加 — 视口税要按可视高度算。按本章假设的浏览器栏高度,390×844 第一次来访时只有 664 像素可用,360×780 只有 628 像素。A 版第一次来访的固定层占可视高度的 0.319 与 0.338,阅读时仍有 0.139 与 0.146;文字放大到 2.0 倍时,钉住的两条横条吃掉 0.277。100vh 会把首图底部的按钮藏到浏览器栏后面,svh 不会;dvh 会让内容在栏收起时跳动 86 像素。滚动隐藏的页头把阅读时的占用降到一半以下,切换阈值要在 24 像素以上,否则手指一抖页头就闪。晚到的横幅插进文档流会产生布局偏移,一条 120 像素的 Cookie 条从顶部插入就是 0.181。聊天浮窗若保留默认位置,会盖住操作栏按钮的 0.197。固定层还会遮住获得焦点的控件,用 scroll-padding 预留即可解决。 (https://5dlibrary.com/answers/zh/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c03/) - Ch.4 The Arithmetic of Gold on Black: Display Polarity, Halation, Contrast Floors and Ceilings, and Price Numerals — This chapter breaks "dark luxury" into a few accounts that can be computed. The first is display polarity: dark text on a light ground versus light text on a dark ground, which one is read faster and more accurately, and what the evidence says. The second is contrast: WCAG 2.1 sets floors and no ceiling, yet in dark themes excessive contrast is a common fault. The third is gold: a single gold hue has only one usable band of lightness on a dark ground and another on a light ground. The fourth is type: how many characters each step of the type scale fits on the 390 and 360 pixel viewports, and how many device pixels a hairline keeps on a 3x screen. The last is the price itself: how to write it, which numeral style to use, and how instalments add back to the shelf price. (https://5dlibrary.com/answers/en/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c04/) 第4章 黑底金字的算术:显示极性、光晕、对比度上下限与价格数字 — 本章把「暗黑奢品」拆成几笔可以算的账。第一笔是显示极性:深底浅字与浅底深字,哪一种读得更快、更准,证据怎么说。第二笔是对比度:WCAG 2.1 给了下限,没有给上限,而在暗色主题里,过高的对比度恰恰是常见的毛病。第三笔是金色:同一个色相的金,在深底和浅底上各自只有一段明度能用。第四笔是字:字阶在 390 与 360 两种视口宽度里能放下几个字,细笔画在三倍像素屏上还剩几个像素。最后一笔是价格数字本身:怎么写、用什么数字字形、分期金额怎样加回原价。 (https://5dlibrary.com/answers/zh/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c04/) - Ch.6 Above the Fold: Pixel Allocation, the Arithmetic of Three Web Metrics, the Evidence on Skeleton Screens and a Hero Byte Budget — The first screen is fixed land. On the visible height with browser bars expanded, Nightamber's layout B cannot fit a square hero and the in-flow price together; the price stays on the first screen only through the persistent action bar. Of LCP's four segments, A's most expensive is the time before the hero is discovered: compressing the image alone brings the 75th percentile from 4.72 s to 3.65 s, and only the new discovery path brings it to 1.71 s. The hero budget from p61's simulation is 185 KB with headroom; a twice-density AVIF needs only 96.0 KB, a three-times one 216.0 KB. Most of the CLS comes from one unsized hero, 0.5971 in a single shift; most of the INP from third-party long tasks during the first screen. Whether skeleton screens shorten perceived waiting cannot be settled on present evidence, and Nightamber treats it as a hypothesis to test rather than a default. (https://5dlibrary.com/answers/en/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c06/) 第6章 首屏:像素分配、三项网页指标的算术、骨架屏的证据与首图字节预算 — 首屏是一块定量的地。按工具栏展开的可见高度算,夜珀 B 版连 1:1 的首图加流内价格都放不下,价格是靠常驻操作栏才留在第一屏里的。LCP 的四段里,A 版最贵的是「发现首图之前」那一段:只压缩图片,第 75 百分位从 4.72 秒降到 3.65 秒,改了发现路径才到 1.71 秒。首图的预算按 p61 的模拟是 185 KB 留余量,两倍密度的 AVIF 只要 96.0 KB,三倍密度要 216.0 KB。CLS 的大头是一张没有尺寸的首图,一次就是 0.5971 分;INP 的大头是首屏期间的第三方长任务。骨架屏能否缩短主观等待,现有证据不足以下结论,夜珀把它当作一个待测的假说,而不是默认做法。 (https://5dlibrary.com/answers/zh/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c06/) - Ch.7 The Boutique Playbook: Product Page, Checkout, Private-Message Entry, Membership Page and the Distillation Spec — The sample is still Chapter 1's fictional Nightamber, the handbag at RM 2,890, on two viewports, 390×844 and 360×780. The thumb model is Chapter 2's teaching model of p21 and p22, medium hand, taking the best zone found inside an element inset by 8 px; browser-bar assumptions follow Chapter 3's p30. Method, evidence grades and disclaimers are as set out in Chapter 1. (https://5dlibrary.com/answers/en/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c07/) 第7章 精品店手册:单品页、结账页、私讯入口、会员页与四件套蒸馏规格 — 1. reach:拇指模型的参数,出自第二章 p21。每 CSS 像素对应的毫米数,390x844 为 0.16565、360x780 为 0.15875;拇指半径 S、M、L 为 64、72、80 毫米;转轴在侧边外 6 毫米、底边上 28 毫米;自然区为 0.45 到 0.80 倍半径,勉强区为 0.25 到 0.45 及 0.80 到 1.00 倍;元素内缩 8 像素、按 4 像素网格取最好的区。 2. viewport:reading_max 0.15(阅读时被钉住的全宽横条占可见高度的上限),first_visit_max 0.35(第一次来访、Cookie 层未处理时的上限)。 3. contrast.dark_text_max:18.0,出自第四章的工作上限,只用于暗色主题的正文与小字;长文本目标区间 13.0 到 17.0 作为说明写进令牌注释,不作拒绝条件。 4. fold:hero_kb_max 185(出自第六章 p61,留 0.5 秒余量的首图预算),cls_budget 0.1,audit_viewport 取较窄的 360x780。 5. bar:底部操作栏高 64 像素,底部内边距取 env(safe-area-inset-bottom)。 (https://5dlibrary.com/answers/zh/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c07/) - Ch.8 Practice: The Nightamber Product Page from Diagnosis and Redesign to a Pre-Registered A/B Test — Chapter 8 takes Nightamber through the whole method: after the redesign the viewport tax falls to 0.258 and 0.267, product facts on the first screen grow to 164 pixels, the price and the button are always visible, and the actions of the purchase path fall in the comfortable zone for both the right-hand and the left-hand model; the pre-registered four-week test, with 19,200 sessions per arm, can detect relative lifts of 0.149 and above, while a lift of 0.10 has only a 0.477 chance of being seen. Marked boxes separate evidence, calculations, sources and hypotheses, each hypothesis stating what would refute it; the source studied is the Kaogongji and its four conditions of good work: "heaven has its seasons, earth has its qi, materials have their beauty, craftsmen have their skill". (https://5dlibrary.com/answers/en/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c08/) 第8章 实战:夜珀单品页从诊断、重排到预先登记的 A/B 测试 — 第八章让夜珀走完全程:重排后视口税降到 0.258 与 0.267,第一屏的商品信息增加到 164 像素,价格与按钮始终可见,购买路径的操作在左右手模型下都落在舒适区;预先登记的四周测试每组 19,200 次会话,能检测 0.149 以上的相对提升,0.10 的提升只有 0.477 的把握看出来。书中以标注框区分证据、计算、原典与假说,每条假说都写明什么结果会推翻它;原典考辨取《考工记》「天有时,地有气,材有美,工有巧」的良工四条件。 (https://5dlibrary.com/answers/zh/mobile-thumb-zone-kinetics-dark-luxury-micro-interactions-above/c08/) ## 5DL-B-138 · Cognitive Unloading of Dense Data Workbenches, Conditional-Format Salience and Zero-Training Error-Proof Form Engineering 中文书名:高密度数据工作台视觉减负、条件格式显著性与零培训防呆表单工程 DOI: pending Cite: Fan, Y. (2026). Cognitive Unloading of Dense Data Workbenches, Conditional-Format Salience and Zero-Training Error-Proof Form Engineering (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: The book is not accounting, audit or legal advice. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it; the source studied is the reporting system of the Tang finance minister Liu Yan, given in paraphrase. Abstract: This volume answers a question every business with a back office meets: with more than a thousand ledger rows a day and several people watching, why are anomalies still missed, figures still misread and statuses still changed wrongly? The book's view is that back-office errors mostly happen not because people are careless but because the table makes anomalies invisible and makes the right action awkward. It sorts back-office losses into unseen, misread and mis-done, and pins them to four metrics: the miss rate, the false-alarm rate with the positive predictive value, seconds per row, and the rework rate. Misses are measured with planted rows, redesigns are tested by randomly alternating days within each week, and ten items are registered before any operator opens the new interface. Timing and miss data on operators are used only to compare interfaces, never to rank people. The book is not accounting, audit or legal advice. All firms, people, ledger rows and anomalies are invented samples, and every calculation is produced by the book's programs. Chapter 1 is the method. The invented buying agent Harbourbridge has 1,200 ledger rows and 18 true anomalies a day; six whole-row red rules paint 435 rows red with a positive predictive value of only 0.037; with a budget of 60 proper checks a day, loosening the threshold can push the miss rate from 0.500 to 0.853; and a before-and-after comparison run during a rising peak season recovers on average less than half of the true effect, while randomly alternating days within weeks brings the estimate back to −0.1040. Chapters 2 to 7 cover scan paths and table layout, preattentive features and the hierarchy of conditional formats, signal detection and alarm thresholds, progressive disclosure and default views, error-proof forms, and playbooks for reconciliation, order, stock and customer-record back offices with a tool specification. Chapter 8 takes Harbourbridge's reconciliation back office through the whole method: a two-week planted-row audit gives a baseline miss rate of 0.683; keeping red for four reconciliation rules cuts red rows from 435 to 40 with a positive predictive value of 0.425, and the modelled miss rate falls from 0.664 to 0.089, or to 0.339 if the new rules reach only six tenths of their set sensitivity; a state machine, an amount mask, a currency default and counted, undoable batch actions cut the rework rate from 0.0064 to 0.0013; seconds per row fall by only 0.49 and cannot be pinned down in eight weeks, so they serve only as a secondary metric. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it; the source studied is the reporting system of the Tang finance minister Liu Yan, given in paraphrase. 摘要:本卷回答一个每家有后台的生意都会遇到的问题:每天上千行流水、几个人盯着,为什么异常还是被漏掉、数字还是被看错、状态还是被改错。本书的看法是,后台出错多半不是人不细心,而是表格让异常看不见、让正确的操作做不顺。全书把后台的损失分成看不见、看错、做错三类,落到漏检率、误报率与阳性预测值、每行处理秒数、返工率四个指标上;漏检率用埋雷行来量,改版用每周之内随机交替的对照来检验,并在第一位操作人员打开新界面之前登记十项内容。操作人员的计时与漏检数据只用来比较界面,不用来给人排名。本书不是会计、审计或法律意见。全部商行、人物、流水与异常都是虚构样本,全部计算由书中程序生成。 第一章是方法:虚构的代购商行海桥每天 1,200 行流水、18 行真异常,六条整行标红规则让 435 行变红,阳性预测值只有 0.037;在每天只能认真查 60 行的预算下,阈值越放越松,漏检率可以从 0.500 升到 0.853;把前后对照放在业务量上升的旺季里,估计出的效果平均只有真实效果的一半不到,改成每周之内随机交替才回到 −0.1040。第二至第七章依次写扫视路径与表格排版、前注意特征与条件格式的层级、信号检测与警报阈值、渐进披露与默认视图、防错表单,以及对账、订单、库存、客户资料四种后台的手册与工具规格。第八章让海桥的对账后台走完全程:两周埋雷审计的基线漏检率 0.683;红色只留给四条对账规则后,红行从 435 降到 40,阳性预测值 0.425,模型预期的漏检率从 0.664 降到 0.089,即使新规则只有设定灵敏度的六成也是 0.339;状态机、金额掩码、币种默认值与批量操作的报数撤销,把返工率从 0.0064 降到 0.0013;每行处理秒数只少 0.49 秒,在八周内测不准,所以只作次要指标。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它;原典考辨取唐代刘晏的报数制度,原文只转述大意。 Chapter answers: - Ch.1 Method: Three Losses, Four Metrics, Evidence Grades and the Liu Yan Reporting System — Chapter 1 is the method. The invented buying agent Harbourbridge has 1,200 ledger rows and 18 true anomalies a day; six whole-row red rules paint 435 rows red with a positive predictive value of only 0.037; with a budget of 60 proper checks a day, loosening the threshold can push the miss rate from 0.500 to 0.853; and a before-and-after comparison run during a rising peak season recovers on average less than half of the true effect, while randomly alternating days within weeks brings the estimate back to −0.1040. (https://5dlibrary.com/answers/en/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c01/) 第1章 方法:后台的三类损失、四个指标、证据分级与刘晏考辨 — 第一章是方法:虚构的代购商行海桥每天 1,200 行流水、18 行真异常,六条整行标红规则让 435 行变红,阳性预测值只有 0.037;在每天只能认真查 60 行的预算下,阈值越放越松,漏检率可以从 0.500 升到 0.853;把前后对照放在业务量上升的旺季里,估计出的效果平均只有真实效果的一半不到,改成每周之内随机交替才回到 −0.1040。第二至第七章依次写扫视路径与表格排版、前注意特征与条件格式的层级、信号检测与警报阈值、渐进披露与默认视图、防错表单,以及对账、订单、库存、客户资料四种后台的手册与工具规格。 (https://5dlibrary.com/answers/zh/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c01/) - Ch.2 Scan Paths and the Skeleton of a Table: Column Order, Alignment, Frozen Panes and Density — The skeleton is a cost paid on every row. Ordering columns by task shortens the eyes' travel per row by 27.7% under this chapter's task weights, at the price of slowing minority tasks, which need views of their own. Right-aligned numbers, fixed decimals, comma separators and tabular figures let 43.3% of neighbouring rows be compared without reading a digit. Freezing the header costs 4.8% of the viewport height and puts all 1,200 rows on screen with it; freezing the first column costs 6.7% of the width and keeps every row's identity in view. Zebra striping most likely does little, so little that a team of four could hardly detect it within a week, so it is off by default and the testing budget goes to column order and row height. Row height trades screens against wrong clicks, and tables with in-row actions keep rows at 32 pixels or more. Colour comes only once the skeleton stands, and that is the next chapter. (https://5dlibrary.com/answers/en/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c02/) 第2章 扫视路径与表格骨架:列序、对齐、冻结与密度 — 骨架是每一行都要付的成本。按任务排列序,在本章设定的任务比重下,每行眼睛的路程短了 27.7%,代价是少数任务变慢,需要给它们单独的视图。数字右对齐、固定小数位、逗号千分位、等宽数字,让 43.3% 的相邻行不用读数字就能比较大小。冻结表头的代价是视口高度的 4.8%,换来全部 1,200 行与表头同屏;冻结首列的代价是 6.7% 的宽度,换来每一行永远带着身份。斑马纹的效果多半很小,小到四个人的团队一周之内几乎不可能测出来,所以默认关闭,测试预算留给列序与行高。行高是翻屏与点错之间的交换,有行内操作的表不低于 32 像素。颜色要等骨架立好之后再上,那是下一章的事。 (https://5dlibrary.com/answers/zh/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c02/) - Ch.3 Preattentive Salience: Let the Anomalies Jump Out, and Only the Anomalies — Preattentive features are the least costly force in a back-office table: one red cell on a screen is seen without searching. But the force can be spent. Colour is the strongest and must carry a back-up channel, and red and green must not be paired; length suits size, not anomalies; orientation and closure are weaker and suit the second tier. p33 shows that pale fills sit only between 1.10 and 1.24 against white paper and are visible through hue alone, so the top tier needs a dark fill with white text. Conditional formatting has three tiers: red only for rules with a positive predictive value of at least 0.5, amber for 0.1 to 0.5, neutral marks for everything else. At Harbourbridge's base rate of 0.0150, a red rule needs a positive likelihood ratio of 65.7; the rules in Chapter 1's p10 reach only single digits. Colouring by band takes red from 435 rows a day to fewer than 3, at the price that half of the anomalies must be caught by means other than colour, which is the account Chapter 4 settles. (https://5dlibrary.com/answers/en/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c03/) 第3章 前注意显著性:让异常自己跳出来,也只让异常跳出来 — 前注意特征是后台表格里成本最低的力量:一屏里只有一个红格,它不用寻找就会被看见。但这股力量会被花光。颜色最强,必须带备份通道,红绿不得成对使用;长度适合大小,不适合异常;方向与闭合较弱,适合第二层。p33 显示,浅底色与白纸的对比度只在 1.10 到 1.24 之间,能被看见全靠色相,所以最高一层要用深底白字。条件格式分三层:红色只给阳性预测值不低于 0.5 的规则,琥珀给 0.1 到 0.5,其余都用中性标记。在海桥 0.0150 的基础率下,红色规则的阳性似然比要达到 65.7;第一章 p10 现状里的规则都只有个位数。按区间定颜色之后,红行从每天 435 行降到不到 3 行,代价是一半异常要靠颜色以外的手段去抓,这正是第四章要算的账。 (https://5dlibrary.com/answers/zh/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c03/) - Ch.4 Alarm Thresholds: Signal Detection, ROC, the Cost Matrix and a Per-Screen Alarm Cap — Harbourbridge's reconciliation back office takes in about 1,200 ledger rows a day, and under the book-wide sample fixed in Chapter 1, 18 of them are real anomalies: amount mismatches, duplicate postings, wrong exchange rates, payments without orders, and orders marked paid with no bank credit. The system gives every row an "anomaly score", and when the score crosses a line, the row turns red. Chapter 3 dealt with how red gets seen before anything else and how many tiers of colour there should be; this chapter deals with where that line should be drawn. (https://5dlibrary.com/answers/en/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c04/) 第4章 警报阈值:信号检测、ROC、代价矩阵与每屏警报上限 — 海桥 Harbourbridge 的对账后台每天进来约 1,200 行流水,按第一章定下的全书样本,其中 18 行是真异常:金额不符、重复入账、汇率错用、有款无单、已付未到账。系统给每一行算一个「异常分」,分数越过某条线,这一行就被标红。第三章讲了红色怎样被眼睛先看见、红色该分几层;本章讲那条线该画在哪里。 (https://5dlibrary.com/answers/zh/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c04/) - Ch.6 Zero-Training Error-Proof Forms: Constraints, Defaults, Inline Validation, Confirm-or-Undo and Batch Guards — The order of an error-proof form is fixed: what can be derived is not typed; what can be chosen is not keyed; only what can be learned is pre-filled; format belongs to the mask; numbers carry their own check digits; amounts are checked for magnitude; feedback comes when the field is left; undo comes before confirmation; and a batch shows its blast radius first. p61 shows that a default saves hands without necessarily saving errors; p62 that checking the full format on every keystroke manufactures 46,800 false errors a day; p60 that one extra Damm digit catches every adjacent transposition; p63 that a tiered policy cuts dialogs from 4,411 a day to 29; and p64 that a typed-count threshold of 50 rows puts friction on 0.1037 of batches while covering 0.5866 of rows at risk. Chapter 7 builds these practices into four back offices, reconciliation, orders, stock and customer records, and in its last section writes them up as checks the style kit can run. (https://5dlibrary.com/answers/en/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c06/) 第6章 零培训防呆表单:约束、默认值、即时校验、确认撤销与批量保护 — 防错表单的顺序是固定的:能推导的不让填,能选择的不让敲,能学来的才预填,格式交给掩码,号码自带校验,金额看数量级,提示在离开字段时给,撤回比确认优先,批量先看半径。p61 说明默认值省手不一定省错;p62 说明每敲一键都查完整格式,一天会制造 46,800 次假错;p60 说明多一位 Damm 校验码就能查出全部相邻对调;p63 说明分层之后对话框从一天 4,411 个降到 29 个;p64 说明 50 行的键入门槛用 0.1037 的批次摩擦覆盖 0.5866 的风险行数。第七章把这些做法装进对账、订单、库存与客户资料四种后台,并在末节把它们写成样式库可以执行的检查。 (https://5dlibrary.com/answers/zh/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c06/) - Ch.7 The Workbench Playbook: Reconciliation, Orders, Stock and Customer Records, and the Distillation Spec — Each of the four workbenches has one pivot that connects seeing with doing. Reconciliation uses two matching passes to turn 1,089 red rows into an 18-row queue. Orders use a server-side state machine plus buttons so that 6.04 illegal options no longer appear on screen. Stock compares the inventory position, so red rows fall from 35.60 a day to 4.24 with stock-outs just as low. Customer records normalise on entry, so de-duplication goes from finding 66 duplicates to finding all 300. The next section writes these practices, together with the calculations of the earlier chapters, into new specifications for the front-end style kit. (https://5dlibrary.com/answers/en/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c07/) 第7章 工作台手册:对账、订单、库存、客户资料四种后台与四件套蒸馏规格 — 四种后台各有一个把「看」与「做」接起来的支点。对账靠两遍匹配把 1,089 行红色收成 18 行队列;订单靠服务器端状态机加按钮,让 6.04 个非法选项不再出现在屏幕上;库存靠比较库存位置,让红行从每天 35.60 行降到 4.24 行,断货却一样少;客户资料靠录入时规范化,让查重从 66 个找到全部 300 个。下一节把这些做法连同前六章的计算写成前端样式库的新增规格。 (https://5dlibrary.com/answers/zh/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c07/) - Ch.8 Practice: Rebuilding the Harbourbridge Reconciliation Back Office from Diagnosis to Redesign — Chapter 8 takes Harbourbridge's reconciliation back office through the whole method: a two-week planted-row audit gives a baseline miss rate of 0.683; keeping red for four reconciliation rules cuts red rows from 435 to 40 with a positive predictive value of 0.425, and the modelled miss rate falls from 0.664 to 0.089, or to 0.339 if the new rules reach only six tenths of their set sensitivity; a state machine, an amount mask, a currency default and counted, undoable batch actions cut the rework rate from 0.0064 to 0.0013; seconds per row fall by only 0.49 and cannot be pinned down in eight weeks, so they serve only as a secondary metric. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it; the source studied is the reporting system of the Tang finance minister Liu Yan, given in paraphrase. (https://5dlibrary.com/answers/en/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c08/) 第8章 实战:海桥对账后台从诊断到重排的完整案例 — 第八章让海桥的对账后台走完全程:两周埋雷审计的基线漏检率 0.683;红色只留给四条对账规则后,红行从 435 降到 40,阳性预测值 0.425,模型预期的漏检率从 0.664 降到 0.089,即使新规则只有设定灵敏度的六成也是 0.339;状态机、金额掩码、币种默认值与批量操作的报数撤销,把返工率从 0.0064 降到 0.0013;每行处理秒数只少 0.49 秒,在八周内测不准,所以只作次要指标。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它;原典考辨取唐代刘晏的报数制度,原文只转述大意。 (https://5dlibrary.com/answers/zh/cognitive-unloading-dense-data-workbenches-conditional-format-salience-zero/c08/) ## 5DL-B-139 · Suspense Breakpoints in Multi-Tier Diagnostic Paywalls, Sunk-Progress Ratchets and Frictionless Unlock Interface Design 中文书名:多阶诊断付费墙悬念断点、进度沉没棘轮与微交互丝滑解锁界面学 DOI: pending Cite: Fan, Y. (2026). Suspense Breakpoints in Multi-Tier Diagnostic Paywalls, Sunk-Progress Ratchets and Frictionless Unlock Interface Design (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: It is not legal advice; pricing economics is left to earlier volumes. Marked boxes separate published evidence, runnable calculations, classical sources and the author's hypotheses, each stating what would refute it; the "hook and clamp" of the Guiguzi chapter Feiqian is examined in Chapter 1. Abstract: This volume is for sellers of diagnostic reports and similar products that give part away free and charge for the rest. It studies only the page in front of the report: on which screen and at which pixel the free content stops, how the progress a reader has already made is shown, how locked content is previewed honestly, and how payment fits on one screen. The method splits the page into four measurable variables and four indicators (reach, unlock, refund and complaint rates), takes net revenue per result viewer as the primary metric, grades claims on five levels of evidence, computes with the book's programs, and tests under pre-registration. One invented sample runs through the book: the free summary and three paid steps at RM 29, RM 99 and RM 299 of the diagnostic brand Starbalance. It is not legal advice; pricing economics is left to earlier volumes. Chapter 1 sets Starbalance's baseline: 20,000 questionnaire starts a month, 12,400 result viewers, 8,804 arrivals at the paywall card, an expected 748.34 orders at RM 62.00, and RM 43,613.26 of net revenue. The unlock rate alone picks the wrong page: adding fake progress, a countdown that restarts and previews of uncomputed sections gives 1.224 times the honest unlock rate but only 1.063 times its first-month net, while dissatisfied orders rise from 5.86 to 20.28 per thousand arrivals; over twelve months the break-even exposure risk is 0.0353 a month with later revenue ignored, and with later revenue counted the deceptive page earns RM 32,417 less even if never exposed. On Starbalance's traffic a one-tenth rise in unlocks needs 17,657 people per arm and 17.4 weeks. Chapters 2 to 7 cover a hazard model of the breakpoint, endowed progress and goal gradients, honest and dishonest uses of sunk input, locked previews and the truth of "computed, waiting to unlock" labels, one-screen checkout and the limits of dark patterns, and a playbook for the four report steps with new components for a front-end style kit. Chapter 8 audits Starbalance's page: a step-choice screen that hid prices loses 257 people who had already tapped, and eight seconds of fake progress cost customers 25.5 hours a month; honesty problems are fixed at once and not tested, moving the break from after section 2 to after section 3 (y = 2,300 px) beats the current cut in eight of nine cases, the central case lifts net revenue per viewer by 0.285, the worst case costs RM 1,869 of testing, and a ten-week registered test can detect a rise of 0.202. Marked boxes separate published evidence, runnable calculations, classical sources and the author's hypotheses, each stating what would refute it; the "hook and clamp" of the Guiguzi chapter Feiqian is examined in Chapter 1. 摘要:本卷写给卖诊断报告、测算结果与评估清单这类「先免费给一部分、再收费解锁其余」的产品的人,只研究报告前面那一页界面:免费内容在第几屏、第几个像素停下,读者已经付出的进度怎样被看见,锁住的内容怎样诚实地预览,付款怎样在一屏内完成。方法是把这一页拆成四个可量的界面变量与四个可记账的指标(到达断点率、解锁率、退款率、投诉率),以「每位结果页访客的净收入」为主指标,按五级证据判断每条经验该信几分,用书中印出的程序把账算清,再在自己的页面上做预先登记的随机测试。全书只用一个虚构样本:诊断报告品牌星衡的免费摘要与 RM 29、RM 99、RM 299 三个付费阶。本书不是法律或消费者保护意见;价格与菜单的经济学见本馆旧书,本书只写界面。 第一章定下星衡的基线:每月 20,000 人开始答题,12,400 人看到结果页,8,804 人到达付费卡片,期望 748.34 单,每单 RM 62.00,当月净收入 RM 43,613.26。只看解锁率会选错:加入假进度、会重来的倒计时与未算先锁预览的页面,解锁率是诚实页面的 1.224 倍,当月净收入只有 1.063 倍,不满意的订单却从每千位到达者 5.86 单升到 20.28 单;把账拉长到十二个月,忽略后续收入时盈亏平衡的曝光概率是每月 0.0353,算上后续收入时,欺骗页面即使永不被曝光也少赚 RM 32,417。在星衡的流量上,要看出解锁率一成的提升需要每组 17,657 人、17.4 周。第二至第七章依次写断点的风险率模型、被赋予的进度与目标梯度、沉没投入的正当用法与越界、锁定预览与「已算好待解锁」的真实性规则、单屏付款与暗黑模式的边界,以及四阶报告的落地手册与前端样式库的新组件规格。第八章审计星衡的现行页面:价格藏在按钮之后的选阶一屏丢掉 257 位已经点击的人,八秒的假进度一个月花掉客户 25.5 小时;诚实问题当天改掉、不做测试,断点从第 2 节之后移到第 3 节之后(y = 2,300 px)在九种情形中有八种胜过现行位置,中间情形下每位访客净收入提升 0.285,最坏情形下测试代价 RM 1,869;登记十周测试,能看出 0.202 的提升。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它;《鬼谷子·飞箝》的「钩箝」在第一章考辨。 Chapter answers: - Ch.1 Method: Four Interface Variables, Four Indicators, Evidence Grades and Honest Previews — Chapter 1 sets Starbalance's baseline: 20,000 questionnaire starts a month, 12,400 result viewers, 8,804 arrivals at the paywall card, an expected 748.34 orders at RM 62.00, and RM 43,613.26 of net revenue. The unlock rate alone picks the wrong page: adding fake progress, a countdown that restarts and previews of uncomputed sections gives 1.224 times the honest unlock rate but only 1.063 times its first-month net, while dissatisfied orders rise from 5.86 to 20.28 per thousand arrivals; over twelve months the break-even exposure risk is 0.0353 a month with later revenue ignored, and with later revenue counted the deceptive page earns RM 32,417 less even if never exposed. On Starbalance's traffic a one-tenth rise in unlocks needs 17,657 people per arm and 17.4 weeks. (https://5dlibrary.com/answers/en/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c01/) 第1章 方法:四个界面变量、四个指标、证据分级与诚实预览 — 第一章定下星衡的基线:每月 20,000 人开始答题,12,400 人看到结果页,8,804 人到达付费卡片,期望 748.34 单,每单 RM 62.00,当月净收入 RM 43,613.26。只看解锁率会选错:加入假进度、会重来的倒计时与未算先锁预览的页面,解锁率是诚实页面的 1.224 倍,当月净收入只有 1.063 倍,不满意的订单却从每千位到达者 5.86 单升到 20.28 单;把账拉长到十二个月,忽略后续收入时盈亏平衡的曝光概率是每月 0.0353,算上后续收入时,欺骗页面即使永不被曝光也少赚 RM 32,417。在星衡的流量上,要看出解锁率一成的提升需要每组 17,657 人、17.4 周。第二至第七章依次写断点的风险率模型、被赋予的进度与目标梯度、沉没投入的正当用法与越界、锁定预览与「已算好待解锁」的真实性规则、单屏付款与暗黑模式的边界,以及四阶报告的落地手册与前端样式库的新组件规格。 (https://5dlibrary.com/answers/zh/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c01/) - Ch.2 The Truncation Breakpoint: Which Screen, Which Pixel — A breakpoint is a hazard problem. p20's open cohort measures with the product-limit estimator how fast readers leave screen by screen; the naive count treats lost beacons as exits and understates reach at the end of screen 6 by 0.103. p21 multiplies reach, curiosity and locked value and, under Starbalance's settings, finds an optimal breakpoint at 2.75 screens with a flat top from 2.50 to 3.50 screens; locking on screen 1 earns only 0.696 of the optimum, so hiding more does not always sell more. p22 turns screens into pixels: the same content boundary sits deeper on 360×780 than on 390×844, and only the boundary after section 3 falls inside the flat top on both phones, so the breakpoint is defined by content and accepted on the smallest viewport. p23 shows that which sections go before the breakpoint matters more than its position: the best choice is 1.147 times report order, and the two constraints "include a strength, do not end on the weakest" cost only 0.0126. p25 shows that cutting again inside paid content may earn more on paper but buys a multiple of dissatisfied buyers, and should be ruled out as a hard constraint. p24 insists that the denominator be the result viewer: at Starbalance's traffic, breakpoint differences inside the flat top cannot be measured within a year, and test traffic should be kept for changes with large differences, such as content choice. The next chapter turns to an earlier stretch of the road: the progress readers see while answering the questionnaire. (https://5dlibrary.com/answers/en/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c02/) 第2章 截断断点:免费内容停在第几屏、第几个像素 — 断点是一个风险率问题。p20 的开放观测组用乘积限估计量出读者逐屏离开的速度,朴素算法会把丢失的信标当成离开,在第 6 屏末端把到达率低估 0.103。p21 把到达、好奇与锁住的价值相乘,在星衡的设定下求出最优断点 2.75 屏,平顶区 2.50 到 3.50 屏;锁在第 1 屏只得到最优的 0.696,遮得越多不一定越好卖。p22 把屏数换成像素:同一个内容边界在 360×780 上比在 390×844 上更深,只有第 3 节之后的位置在两种手机上都落进平顶区,所以断点按内容定义、按最小视口验收。p23 显示断点之前放哪几节比断点位置更要紧,最好的选择是报告顺序的 1.147 倍,加上「包含长处、不以最弱项收尾」两条约束只损失 0.0126。p25 说明在付费内容内部再切一刀,账面上可能多赚,却换来成倍的不满意买家,应当作为硬约束排除。p24 提醒分母必须是结果页访客:平顶区内的断点差别以星衡的流量一年也测不出来,测试流量应当留给内容选择这类差别大的改动。下一章转向断点之前的另一段路:读者答问卷时看到的进度。 (https://5dlibrary.com/answers/zh/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c02/) - Ch.3 Endowed Progress: Where the Bar Starts, What It Counts, How Fast It Moves and Whether It Tells the Truth — The progress bar decides how many people reach the paywall. Three classic pieces of evidence point the way: endowed progress raises completion, people move faster near the finish, and the middle is where effort sags. In Starbalance's simulated log, p31 reads both a goal gradient of 0.956 per question and a mid-course bump around question 13, which gives the display rule "report answered in the first half, report remaining in the second". The unit of the bar should be time: counting questions while putting the open questions last makes the bar overstate by 0.202. The starting point may count only steps really completed: the p30 model gives the same 0.6470 for real and padded endowment, and only the log separates them. The pace curve must stay inside the 0.05 honest band: the fast-to-slow bar does most to cut breakoff and overstates by 0.250 at the midpoint, and this book does not use it. A fake "analysing" animation shows every reader the same flaw and, under p34's settings, is all but certain to be exposed within a month; p35's four audits make every tick of progress match the log. The progress bar is the easiest interface variable to test, but tests run only between versions that pass the audit. The next chapter takes up how the time and answers readers invest in the questionnaire should be treated at the paywall. (https://5dlibrary.com/answers/en/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c03/) 第3章 被赋予的进度:进度条的起点、单位、速度与真假 — 进度条决定了有多少人能走到付费墙。三条经典证据给出方向:被赋予的进度提高完成率,越接近终点走得越快,中段最容易松劲;p31 在星衡的模拟日志里同时读出了每题 0.956 倍的目标梯度与第 13 题附近的中段隆起,显示规则因此是「前半程报已答、后半程报剩余」。进度条的单位应当是时间:按题计数、又把开放题放在最后,会让进度条高估 0.202。起点只能计入真实完成的步骤:p30 的模型对真实赋予与填充赋予给出同样的 0.6470,区分它们的只有日志。速度曲线必须留在 0.05 的诚实带里:先快后慢的进度条最能降低放弃率,也在中点高估 0.250,本书不用它。一条假的「分析中」动画对每位读者都露出同样的破绽,按 p34 的设定一个月内几乎必然被曝光;p35 的四条审计让每一格进度都对得上日志。进度条是最容易测试的界面变量,但只在通过审计的版本之间测试。下一章接着讨论读者在问卷里投入的时间与答案,在付费墙前应当被怎样对待。 (https://5dlibrary.com/answers/zh/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c03/) - Ch.4 Sunk-Progress Ratchets: Saved Progress, Personalised Results and the Ethical Edge of the Ratchet — At the paywall the reader holds time, answers and a first result. The ledger in p40 shows that this investment is six minutes, worth about a fifth of the standard price, far too little to support copy that says "pay or it was all for nothing"; the free summary explains only 6 of the 24 answers, and that is the honest reason to upgrade. A legitimate ratchet turns towards the reader: answers are saved for at least 30 days by default and can be exported and deleted; "personalised" must survive the swap test, and below a mean difference of 0.5 the word "personal" is not used; upgrades carry full credit, so climbing from standard all the way to advisor costs RM 299 in total, with break-even multiples between 1.1074 and 1.4950. A reversed ratchet includes deletion threats, false "90% done" claims, undisclosed full-price upgrades and guilt-trip declines. The account in p41 shows that an empty deletion threat starts losing once more than 0.460 of buyers find it out, and that a real deletion destroys resumption, the most dependable road back. The next chapter turns to how locked content should be previewed, and to the conditions under which "already computed, waiting to unlock" may be said at all. (https://5dlibrary.com/answers/en/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c04/) 第4章 进度沉没棘轮:保存进度、个人化结果与棘轮的伦理边界 — 读者走到付费墙前,手里有时间、答案和一份初步结果。p40 的账本显示,这份投入是六分钟、约为标准版价格两成的时间价值,撑不起「不付就白费」的话术;免费摘要只解读了 24 题中的 6 题,这才是一条诚实的升级理由。正当的棘轮朝读者转:答案默认保存不少于 30 天、可导出可删除;「个人化」要经得起换答案测试,平均差异不到 0.5 就不写「专属」;升级全额抵扣,从标准版一路升到顾问版总共只付 RM 299,抵扣的盈亏倍数在 1.1074 到 1.4950 之间。掉头的棘轮包括删除威胁、虚假的「已完成九成」、不告知的全价升级与内疚式拒绝;p41 的账显示,空的删除威胁在发现比例超过 0.460 时就开始亏,而真的删除会毁掉续答这条最可靠的回头路。下一章讨论锁住的内容该怎样预览,以及「已算好、待解锁」这句话在什么条件下才说得出口。 (https://5dlibrary.com/answers/zh/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c04/) - Ch.5 Previewing Locked Content: Blur, Truncation, Card Titles, Sample Sections and the Truth of "Already Computed" — Locked content has four common preview styles. A blur over real text leaks all 6,500 characters and all 39.95 bits of score information in sections 3 to 8; a blur over fake text breaks "the preview is a sample"; and Starbalance's current page also blurs four sections that do not yet exist, leaving preview honesty at only 0.600. Truncation shows a quarter of the original text without the result: honest, but silent about the reader herself. Card titles are a table of contents whose duty is a true status on every section. Only band hints carry true information about her, and 20-point bands reveal 0.388 of the score information. A free sample given in report order is safer than the best sections: once refunds respond to the gap with a sensitivity above 0.0646, picking the best loses. "Computed" needs a receipt, must be finished before the card appears, and must be delivered as receipted; sections not yet computed say honestly "written after payment, about 15 seconds", with the promise set at the 95th percentile. The next chapter moves to the payment screen itself: fields, wallets, trust marks and error messages. (https://5dlibrary.com/answers/en/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c05/) 第5章 锁住的内容怎样预览:模糊、截断、卡片标题、样本段与「已算好」的真实性 — 锁住的内容有四种常见的预览方式。模糊若底下是真文字,第 3 至 8 节的 6,500 个字符与 39.95 比特的分数信息全漏;若底下是假文字,就违反「预览即样本」;星衡现行页面还把尚不存在的四节也画成模糊段落,预览诚实度只有 0.600。截断露出原文的四分之一而不露结果,诚实但不触及读者本人;卡片标题是一份目录,它的责任是给每一节标上真实状态;只有分档提示带着她本人的真实信息,20 分一档露出 0.388 的分数信息。免费样本按顺序给出比挑最好的更稳:一旦退款对落差的敏感度超过 0.0646,挑最好的就亏。「已算好」要有回执、要先于卡片完成、交付要与回执一致;还没算好的节如实写「付款后生成,约需 15 秒」,承诺时间取第 95 百分位。下一章进入付款这一屏本身:字段、钱包、信任标记与错误提示。 (https://5dlibrary.com/answers/zh/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c05/) - Ch.6 One-Screen Checkout: Fields, Wallets, Trust Marks, Error Messages, Latency and the Edge of Dark Patterns — The completion rate of the payment flow is the second half of the unlock rate, and its losses split into screen changes, fields, waiting and errors, each measurable from the log. p60 gives the threshold for deciding whether to merge screens or cut fields first, T = 0.0686 with F = 0.03, while Starbalance's largest single loss is a tier screen with no fields at all. p61 uses the keystroke-level model to show that moving from four screens and eleven fields to one screen and four fields takes the table-value time from 76.45 seconds to 29.91 seconds, and that the wallet path needs only 4.95 seconds, a gain that must be multiplied by the share of readers who can use a wallet. p62 shows that a browser-side Luhn check catches every single-digit error and 44/45 of adjacent swaps, but no jump swaps. p63 shows that computing before locking keeps the median from payment to report at 4.71 seconds, while generating after payment pushes 0.685 of payments past 10 seconds. p64 turns seven checkout transgressions into seven reconciliation checks. Chapter 7 carries these rules onto the pages of the four report tiers and hands them to the style kit. (https://5dlibrary.com/answers/en/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c06/) 第6章 单屏付款:字段、钱包、信任标记、错误提示、延迟与暗黑模式的边界 — 付款流程的完成率是解锁率的后一半,它的损耗可以拆成换屏、字段、等待与错误四种,每一种都能从日志量出来。p60 给出判断先并屏还是先砍栏的门槛,在 F = 0.03 时是 T = 0.0686,而星衡最大的一处流失是没有任何字段的选阶页;p61 用击键层级模型显示,把四屏十一栏改成单屏四栏,按表值从 76.45 秒降到 29.91 秒,钱包路径只要 4.95 秒,但收益要乘上能用钱包的读者比例;p62 显示浏览器端的 Luhn 校验挡得住全部单字错、44/45 的相邻对调,挡不住隔位对调;p63 显示先算后锁让付款到看到报告的中位数停在 4.71 秒,付款后才生成则有 0.685 的付款超过 10 秒;p64 把七种付款页上的越界写成七条对账检查。第七章把这些规矩落到四阶报告各自的页面上,并交给样式库。 (https://5dlibrary.com/answers/zh/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c06/) - Ch.7 The Paywall Playbook: Pages for Four Report Tiers, Honest Status Labels and the Distillation Spec for the Four Kits — The four report tiers sell three different things: results already computed, results a machine will produce later, and a person's time. p70 lays the tiers out as pages and shows the break sitting on a block boundary at 1688 pixels, while 4 of the 10 sections blurred on today's page do not yet exist when it is shown. The four page playbooks settle what each tier may say: the free summary uses a list of titles instead of half a sentence; the standard tier may say "ready"; the full tier has to state truthfully, side by side, the batch time, the measured generation time and the dependency; the advisor tier gives only the first open slot on the calendar. p71 shows that even at a utilisation of only 0.863, an advisor tier without a sales limit lets 0.060 of sessions run later than promised, and that governing sales by the calendar brings this share to zero. The next section hands these rules to the front-end style kit. (https://5dlibrary.com/answers/en/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c07/) 第7章 付费墙手册:四阶报告的页面、诚实标注与四件套蒸馏规格 — 四阶报告卖的是三种不同的东西:已经算好的结果、稍后由机器产出的结果、一个人的时间。p70 把四阶摆成页面,显示断点落在 1688 像素的块交界上,而现行页面模糊显示的 10 节里有 4 节在显示时还不存在。四份页面手册给每一阶定下各自能说的话:免费摘要用标题一览代替半句话;标准版可以说「已算好」;完整版要同时如实写出批处理时刻、实测的生成时间与依赖关系;顾问版只写日历上的第一个空档。p71 显示,即使利用率只有 0.863,不设上限的顾问版也会让 0.060 的解读迟于承诺,按日历控制销量则能把这个比例降到零。下一节把这些规则交给前端样式库。 (https://5dlibrary.com/answers/zh/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c07/) - Ch.8 Practice: Starbalance's Unlock Page from Audit to Rebuild and a Pre-Registered Test — Chapter 8 audits Starbalance's page: a step-choice screen that hid prices loses 257 people who had already tapped, and eight seconds of fake progress cost customers 25.5 hours a month; honesty problems are fixed at once and not tested, moving the break from after section 2 to after section 3 (y = 2,300 px) beats the current cut in eight of nine cases, the central case lifts net revenue per viewer by 0.285, the worst case costs RM 1,869 of testing, and a ten-week registered test can detect a rise of 0.202. Marked boxes separate published evidence, runnable calculations, classical sources and the author's hypotheses, each stating what would refute it; the "hook and clamp" of the Guiguzi chapter Feiqian is examined in Chapter 1. (https://5dlibrary.com/answers/en/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c08/) 第8章 实战:星衡解锁页从审计到重排与预先登记的测试 — 第八章审计星衡的现行页面:价格藏在按钮之后的选阶一屏丢掉 257 位已经点击的人,八秒的假进度一个月花掉客户 25.5 小时;诚实问题当天改掉、不做测试,断点从第 2 节之后移到第 3 节之后(y = 2,300 px)在九种情形中有八种胜过现行位置,中间情形下每位访客净收入提升 0.285,最坏情形下测试代价 RM 1,869;登记十周测试,能看出 0.202 的提升。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它;《鬼谷子·飞箝》的「钩箝」在第一章考辨。 (https://5dlibrary.com/answers/zh/suspense-breakpoints-multi-tier-diagnostic-paywalls-sunk-progress-ratchets/c08/) ## 5DL-B-140 · Design-Token Algebra across Thirty Sites, Perceptually Uniform Colour Spaces and the Millisecond Science of Brand-Authority First Impressions 中文书名:跨三十站设计令牌代数、色彩感知均匀空间与品牌权威第一印象毫秒学 DOI: pending Cite: Fan, Y. (2026). Design-Token Algebra across Thirty Sites, Perceptually Uniform Colour Spaces and the Millisecond Science of Brand-Authority First Impressions (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All brands, sites, colours and ratings are invented, every calculation comes from the book's programs, and legal matters are described only in general terms, not as legal advice. Each hypothesis states what would refute it. Abstract: This volume answers a question every brand running many websites faces: how can thirty sites look like one brand. Its answer is three things that can be computed. First, write colour, type and spacing as design tokens in three layers, primitive, semantic and component, treat a theme as a function from primitive to semantic tokens, and let sites take values only from the tokens. Second, manage colour in perceptually uniform spaces: measure differences with CIELAB and CIEDE2000, build ramps in OKLab, and register every tolerance with its formula. Third, measure first impressions honestly: a 50-millisecond rating is stable, but it says whether a page looks good, not whether it is trusted or bought. Drift is described by three measures, token deviations, colour difference and first-impression score, never merged into one score; evidence is sorted into five grades, and every unification project pre-registers ten items. The classical source studied is the cai-fen module of Li Jie's Yingzao Fashi, an ancient design-token system. All brands, sites, colours and ratings are invented, every calculation comes from the book's programs, and legal matters are described only in general terms, not as legal advice. Chapter 1 fixes the book-wide sample: across the 30 sites of the invented brand group Hengyu, four family primaries have become 16 live colours, 24 sites have a role beyond a ΔE00 tolerance of 2.0, and 2,210 of 9,400 colour declarations are hard-coded; deviation rate and colour difference have a rank correlation of only 0.608; and the best-looking design is also the most trusted with a probability of only 0.202 when the two are unrelated. Chapters 2 to 7 cover token graphs, colour arithmetic, perceptual ramps, first-impression research, multi-site governance and playbooks with a spec for the style kit. Chapter 8 takes five sites through the whole process: 17 roles out of tolerance and 746 of 1,655 declarations hard-coded; a theme function that fills four families with 15 colour values and derives a new token because gold reaches a contrast of only 2.69; a migration in which only 0.060 of hard-coded values can be replaced automatically and the group's colours shrink from 103 to 15 tokens plus 37 exceptions; and a pre-registered test of 90 raters, while detecting a 10% lift in sales would take 26.9 weeks. Each hypothesis states what would refute it. 摘要:本卷回答一个同时经营许多网站的品牌都会遇到的问题:三十个站点怎样才能看起来像同一个品牌。本书的回答是三件可以计算的事:第一,把颜色、字号、间距写成原始、语义、组件三层的设计令牌,主题是从原始令牌到语义令牌的函数,站点只从令牌取值;第二,在感知均匀的颜色空间里管颜色,用 CIELAB 与 CIEDE2000 量差别,用 OKLab 取色阶,容差连同公式一起登记;第三,诚实地测量第一印象,五十毫秒的评分稳定,但它说的是好不好看,不等于信不信任、买不买。多站视觉漂移用令牌偏离数、色差与第一印象评分三个量描述,不合成总分;证据分为五级,每一次统一工程预先登记十项内容。原典考辨取李诫《营造法式》的材分制,它是古代的设计令牌。全部品牌、站点、颜色与评分都是虚构样本,全部计算由书中程序生成;涉及法规之处只写一般框架,不是法律意见。 第一章定下全书样本:虚构品牌集团衡宇的 30 个站点里,四个家族的主色在线上变成了 16 种,24 个站点有角色超出 ΔE00 不超过 2.0 的容差,9,400 条颜色声明里写死的有 2,210 条;偏离率与色差的等级相关只有 0.608;在好看与可信无关时,最好看的方案同时最可信的概率只有 0.202。第二至第七章依次写令牌图、颜色算术、感知色阶、第一印象研究、多站治理,以及四类站点的手册与样式库规格。第八章让五个站点走完全程:17 个角色越界,1,655 条声明里 746 条写死;主题函数用 15 个颜色值填满四个家族,并因金色的对比度只有 2.69 而派生出新令牌;迁移时写死的值只有 0.060 能自动替换,集团的颜色从 103 个收拢到 15 个令牌加 37 个例外;预先登记的测试用 90 位评分人,而在成交上检出 10% 的提升要 26.9 周。每条假说都写明什么结果会推翻它。 Chapter answers: - Ch.1 Method: Three Measures of Multi-Site Visual Drift, Grades of Evidence and the Cai-Fen Module of the Yingzao Fashi — Chapter 1 fixes the book-wide sample: across the 30 sites of the invented brand group Hengyu, four family primaries have become 16 live colours, 24 sites have a role beyond a ΔE00 tolerance of 2.0, and 2,210 of 9,400 colour declarations are hard-coded; deviation rate and colour difference have a rank correlation of only 0.608; and the best-looking design is also the most trusted with a probability of only 0.202 when the two are unrelated. (https://5dlibrary.com/answers/en/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c01/) 第1章 方法:多站视觉漂移的三个量、证据分级与《营造法式》材分制 — 第一章定下全书样本:虚构品牌集团衡宇的 30 个站点里,四个家族的主色在线上变成了 16 种,24 个站点有角色超出 ΔE00 不超过 2.0 的容差,9,400 条颜色声明里写死的有 2,210 条;偏离率与色差的等级相关只有 0.608;在好看与可信无关时,最好看的方案同时最可信的概率只有 0.202。第二至第七章依次写令牌图、颜色算术、感知色阶、第一印象研究、多站治理,以及四类站点的手册与样式库规格。 (https://5dlibrary.com/answers/zh/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c01/) - Ch.2 Token Algebra: Three Layers, Themes as Functions and the Dependency Graph of Thirty Sites — p20 starts by counting the set as a graph. Of 86 tokens, 38 are primitives, and the semantic and component layers have 24 each. There are 48 alias edges: 20 from semantic to primitive, 23 from component to semantic, 4 from semantic to semantic and 1 from component to component. There are 38 literal values, all of them in the primitive layer. That is the single most important invariant of the whole structure, and the audits later in the chapter check it again and again. (https://5dlibrary.com/answers/en/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c02/) 第2章 令牌代数:三层结构、主题函数与三十站的依赖图 — p20 先把这套令牌当成一张图来数。86 个令牌里,原始层 38 个,语义层与组件层各 24 个;48 条别名边,其中语义指向原始 20 条、组件指向语义 23 条、语义指向语义 4 条、组件指向组件 1 条。字面值一共 38 个,全部落在原始层:这是整套结构最重要的一个不变量,后面的审计会反复检查它。 (https://5dlibrary.com/answers/zh/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c02/) - Ch.3 Perceptually Uniform Colour Spaces: From sRGB to CIELAB, CIEDE2000 and OKLab — The first intuition is "half the code is mid-grey". The grey at code 128, #808080, has a CIELAB lightness of 53.59, more than three units brighter than the midpoint of 50. The grey closest to lightness 50 is #777777, code 119, at 50.03. The reason is that sRGB codes are not linear light; they are encoded through a power-shaped curve, computed step by step in the next section. (https://5dlibrary.com/answers/en/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c03/) 第3章 感知均匀的颜色空间:从 sRGB 到 CIELAB、CIEDE2000 与 OKLab — 第一条直觉是「码值的一半就是中灰」。码值 128 的灰,也就是 #808080,CIELAB 明度是 53.59,比明度的中点 50 亮了三个多单位;真正最接近明度 50 的灰是 #777777,码值 119,明度 50.03。原因是 sRGB 的码值不是线性光强,它经过了一条幂函数形状的曲线编码,下一节会逐步算出来。 (https://5dlibrary.com/answers/zh/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c03/) - Ch.4 Perceptually Uniform Ramps: Lightness Steps, Contrast Constraints, Gamut Clipping and Dark-Theme Inversion — Ramps interpolated on sRGB codes are uneven, ramps interpolated in linear light are more uneven, and ramps with even OKLab lightness are the most even but squeeze the brand primary off the ramp. So this chapter anchors the primary as a stop, uses relative chroma, and adds a cap for hues with high cusps such as green. On contrast: every colour reaches at least 4.583 with black or white text; at equal lightness, different hues give different contrast, so each family must be computed; distance rules such as "six stops apart guarantees 4.5" can be written as token algebra without giving up perceptual evenness. Out-of-gamut colours keep lightness and hue and lose chroma; they are never clamped channel by channel. Dark themes are solved role by role, never by swapping stop numbers. Finally, thirty sites should reference stop names on one shared ramp: R07's primary is only 2.13 off, but the ramp derived from it is 6.76 off. The next chapter leaves colour itself and asks what impression these colours leave within 50 milliseconds, and how far that impression is from a sale. (https://5dlibrary.com/answers/en/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c04/) 第4章 感知均匀的色阶:明度阶、对比度约束、色域裁剪与暗色反转 — 在 sRGB 码值上插值的阶不均匀,在线性光上插值的阶更不均匀,OKLab 明度等距的阶最均匀,但它会把品牌主色挤出阶外,所以本章把主色锚定成阶上的一档,彩度取相对彩度,绿色这类尖点高的色相再加上限。对比度方面,任何颜色配黑字或白字至少一边有 4.583;同明度不同色相的对比度并不相同,必须逐族核算;「隔 6 档保证 4.5」这类档距规则可以写成令牌代数,而且不必为此牺牲感知均匀。越界颜色保持明度与色相降彩度,不做逐通道截断。暗色主题按角色重新求解,不做阶号对调。最后,三十个站点只该引用同一条阶上的档名:R07 的主色只差 2.13,派生出的阶却差到 6.76。下一章离开颜色本身,去看这些颜色在 50 毫秒之内给人留下什么印象,以及这个印象离成交还有多远。 (https://5dlibrary.com/answers/zh/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c04/) - Ch.5 The 50-Millisecond First Impression: Exposure, Complexity, Prototypicality and the Distance to a Sale — The 50-millisecond research is real and its findings are sturdy: judgements of a page's visual appeal form at very short exposures and agree closely across raters and across exposures. But 50 milliseconds on a screen has to be registered in frames, and a 75 Hz screen cannot show it at all. Consistency comes mostly from panel size: 40 raters push the correlation between exposures to 0.927 but cannot push past the validity ceiling set by a shared halo. Complexity and prototypicality are the two parts of a first impression the group can pull directly, the first through token caps and the second through family layout skeletons. From attractive to sold runs a chain of three discounted inferences; even when each link is strong, only a correlation of 0.180 survives at the sales end, and testing it takes thousands to tens of thousands of visitors, whereas a first-impression test needs a few dozen to a hundred raters. So first-impression tests guard the first screen against slipping back, conversion tests answer the business question, and the two are registered and reported separately. Chapter 6 turns to how tokens for thirty sites are released and how drift is detected, where first-impression ratings return as a weekly monitoring measure. (https://5dlibrary.com/answers/en/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c05/) 第5章 五十毫秒的第一印象:曝光、复杂度、原型性与离成交的距离 — 50 毫秒研究是真实的,结论也稳:页面视觉吸引力的判断在极短曝光下就已形成,并且在评审之间、在不同曝光之间高度一致。但屏幕上的 50 毫秒要按帧数登记,75 Hz 的屏幕根本显示不出来;一致性主要来自评审人数,40 名评审能把两种曝光的均分相关推到 0.927,却推不过光环造成的有效性上限。复杂度与原型性是第一印象里集团能直接拉动的两个变量,前者由令牌上限控制,后者由同族版式骨架控制。从好看到成交要经过三段打折的推断链,即便每段都很强,成交端也只剩 0.180 的相关,验证它需要数千到数万名访客,而第一印象测试只需要几十到一百名评审。所以,第一印象测试用来守住首屏不退步,转化测试用来回答生意问题,两者分开登记、分开报告。第六章接着讨论三十个站点的令牌怎样发布、怎样检测漂移,第一印象评分会在那里作为一项周度监测指标再次出现。 (https://5dlibrary.com/answers/zh/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c05/) - Ch.6 Governing Thirty Sites: Release, Drift Detection, Per-Brand Re-Theming, Audit and Rollback — Governing thirty sites rests on four things that can be computed. p60's site table shows which site is on which version, how far it deviates, and which sites sit on a withdrawn release; p61 shows that staged rollout can cut the exposure to a group-wide defect to under half, and that the first wave must cover every family to catch defects confined to one. p62 sorts out-of-tolerance live colours into separate queues and finds colours that have drifted into another role's place; p63 shows that first-impression monitoring needs a cumulative sum, or thirty sites will raise more than a dozen false alarms in half a year. p64's four guardrails stop a colour change from clashing with another family or a status colour; p65's two gates keep a major release from letting new broken links through without letting old problems block it; p66 shows that rollback is not always safe; p67 makes every release traceable. Chapter 7 turns these checks into playbooks for the four kinds of site and writes broken links, cycles and out-of-band colour differences into the style kit's refusal codes. (https://5dlibrary.com/answers/en/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c06/) 第6章 三十站的治理:发布、漂移检测、按品牌换主题、审计与回滚 — 三十个站点的治理,靠的是四件可以计算的事。p60 的站点表告诉你谁落在哪个版本、偏离了多少,并把撤回版本上的站点挂出来;p61 说明分批上线能把一次全面缺陷的暴露压到不到一半,而第一批要覆盖每个家族才能拦住只在某个家族出现的缺陷。p62 把线上的越界颜色分进不同的队列,并找出漂到别的角色上去的错位色;p63 说明第一印象监测要用累积和,否则三十个站点半年会凭空报出十几次警。p64 的四道护栏防止换色时与别的家族、与状态色撞色;p65 的两道闸让大版本既不放过新断链也不被旧问题卡住;p66 说明回滚并不总是安全的;p67 让每一版都能被追查。第七章把这些检查按四类站点落成手册,并把断链、循环与色差越界编进样式库的拒绝码。 (https://5dlibrary.com/answers/zh/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c06/) - Ch.7 Playbooks for the Four Kinds of Site, and the Distillation Spec for the Four Kits — The three states read differently too. On the sRGB mix, "in progress" reaches 5.00 on white, "done" 13.93 and "not started" only 1.75; on the chord ramp "in progress" reaches 3.43, still above the 3.0 floor for components, and "done" 7.97. On both ramps the three states are more than 20 apart, easy to tell apart. The equal-lightness method has 3 stops whose chroma is cut by the sRGB gamut, so the ends of the ramp are greyer than the anchor, the gamut clipping of Chapter 4. (https://5dlibrary.com/answers/en/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c07/) 第7章 四类站点的落地手册与四件套蒸馏规格 — GK01 是一套干净的小型令牌,五类检查都通过。GK02 至 GK06 各复现第二章 p22 的一类错误,GK07 复现一条四跳的别名链,只出警告。GK08 与 GK09 是本章 p73 的主题环与缺失的父主题。GK10 是 Sharma 等人的第一组测试数据,2.0425,用来证明实现本身没有写错。GK11 与 GK12 是第一章样本里 B07 与 B08 的主色:1.4971 在容差之内,12.272 越界。GK13 是第四章的次要文字修正,0.7686,在修订版的 1.0 之内。GK14 与 GK15 是第六章 p64 的两道分隔护栏:精品站候选红离危险红 9.2967,不足 10.0,拒绝;报告蓝与牌卡紫相距 17.3691,通过。GK16 至 GK18 是本章 p72 的色阶:等色差弦长法变异系数 0.0167,通过;等明度法 0.1225,超过 0.05,K8;在 0.40 到 0.38 这么窄的明度里硬取 15 档,出现重复的颜色,K7 与 K8 同时触发。GK19 与 GK20 是不合法的输入,返回 E0。八个 K 码每一个都至少有一条样本触发。 (https://5dlibrary.com/answers/zh/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c07/) - Ch.8 Practice: Unifying the Palettes of Five Representative Hengyu Sites, with a Pre-Registered First-Impression Test — Chapter 8 takes five sites through the whole process: 17 roles out of tolerance and 746 of 1,655 declarations hard-coded; a theme function that fills four families with 15 colour values and derives a new token because gold reaches a contrast of only 2.69; a migration in which only 0.060 of hard-coded values can be replaced automatically and the group's colours shrink from 103 to 15 tokens plus 37 exceptions; and a pre-registered test of 90 raters, while detecting a 10% lift in sales would take 26.9 weeks. Each hypothesis states what would refute it. (https://5dlibrary.com/answers/en/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c08/) 第8章 实战:衡宇五个代表站点的色板统一与预先登记的第一印象测试 — 第八章让五个站点走完全程:17 个角色越界,1,655 条声明里 746 条写死;主题函数用 15 个颜色值填满四个家族,并因金色的对比度只有 2.69 而派生出新令牌;迁移时写死的值只有 0.060 能自动替换,集团的颜色从 103 个收拢到 15 个令牌加 37 个例外;预先登记的测试用 90 位评分人,而在成交上检出 10% 的提升要 26.9 周。每条假说都写明什么结果会推翻它。 (https://5dlibrary.com/answers/zh/design-token-algebra-thirty-sites-perceptually-uniform-colour-spaces/c08/) ## 5DL-B-141 · Narrative Transportation, Cognitive-Dissonance Disarming and Dopaminergic Scroll Rhythms in High-Conversion Long-Form Copy 中文书名:叙事沉浸传输、认知失调破防与高转化长文滑屏多巴胺节律 DOI: pending Cite: Fan, Y. (2026). Narrative Transportation, Cognitive-Dissonance Disarming and Dopaminergic Scroll Rhythms in High-Conversion Long-Form Copy (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not individual legal, advertising-compliance or financial advice, and it teaches no deception. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Copywriting circles pass around an explanation that sounds scientific: stories immerse the reader, admitting a fault breaks the reader's defences, suspense leaves the brain unable to let go, and small rewards between paragraphs release dopamine as short videos do. This book writes each of these claims as a runnable model and asks when it holds, when it reverses and what an overdose costs. Its parameters are illustrative; all calculations are generated by its programs; it is not individual legal, advertising-compliance or financial advice, and it teaches no deception. Chapter 1 shows that a two-question survey sees only part of the true effect of narrative transportation, and that two hard-sell breaks inside a story leave 0.681 of the unbroken acceptance. Chapter 2 uses Festinger's dissonance ratio and a four-exit choice model to show that the dissonance of a hard attack is mostly absorbed by disparaging the source, while a falsify-first opening is accepted 9.62 times as often. Chapter 3 gives the Bayesian account of admitting a flaw: an admission carries a likelihood ratio of 6.0, yet admitting without answering is worse than silence, and pressure wording backfires beyond 4.58 markers per 1000 characters. Chapter 4 examines the replication debate over the Zeigarnik effect, shows that recording only significant results logs a zero effect as 1.249, and finds the best pace of revelation at 0.063 of the remaining answer per section. Chapter 5 treats scroll depth as a survival time: under negative aging the first two screens lose 0.347 of readers, while average scroll depth records a true willingness of 8.55 screens as 5.43. Chapter 6 tests the dopamine metaphor, showing that sales peak when hooks and payoffs are about one to one, that writing for depth cuts buyers per 1000 readers from 8.64 to 5.93, and that 0.794 of the excess net value of pressure copy comes from readers in financial distress. Chapter 7 examines persuasions of the Zhanguo ce and Aristotle's Rhetoric: six speeches coded by the author are soft before hard, an order that random shuffling yields 1 time in 4320, and at credibility 0.5 an argument can grow the odds at most 2.0 times. Chapter 8 gives a persuasive-copy checker of eight machine rules, limits and quiet hours for follow-up messages, and a pre-registered plan with at least 6,455 readers per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:文案圈流传着一套听起来很科学的说法:故事让人沉浸,先认错让人破防,悬念让大脑放不下,段落间的小奖励让读者像刷短视频一样分泌多巴胺。本书把这些说法逐一写成可以运行的模型,看它们在什么条件下成立、在什么条件下反转、过量时付出什么代价。书中参数均为示意;全部计算由书中程序生成;本书不是法律、广告合规或理财方面的个人化建议,也不教任何欺骗手法。 第一章说明,两题问卷只能看到叙事传输真实效应的一部分,而故事中途插入两次硬广告,接受度只剩不中断时的 0.681。第二章用费斯廷格的失调比率与四条出路的选择模型说明,硬攻击的失调大多被读者对来源的贬低吸收,先证伪开头的接受概率是硬攻击的 9.62 倍。第三章给出承认缺点的贝叶斯账:承认的似然比是 6.0,但只承认不回应比沉默更糟;压迫性措辞超过每 1000 字 4.58 个就会反噬。第四章考辨蔡加尼克效应的复制争议,说明只记录显著结果会把零效应记成 1.249 倍,并算出每节揭晓剩余答案 0.063 的最佳节奏。 第五章把滑屏深度当作生存时间:负老化时前两屏流失 0.347,而平均滑屏深度把 8.55 屏的真实意愿记成 5.43 屏。第六章检验多巴胺比喻,说明钩子与兑现约一比一时成交最多,以深度为目标会让每 1000 名读者的买家从 8.64 降到 5.93,施压文案的超额净值有 0.794 来自财务困境读者。第七章考辨《战国策》游说辞与亚里士多德《修辞学》:作者编码的六篇说辞先软后硬,随机得到这种次序的机会是 4320 分之一;可信度 0.5 时论证最多把赔率放大 2.0 倍。第八章给出八条机检规则的攻心文案检查器、私讯跟进上限与静默规则,以及每组至少 6,455 名读者的预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Narrative Transportation: Scale Reliability, Attenuation and the Cost of Breaking a Story — Chapter 1 shows that a two-question survey sees only part of the true effect of narrative transportation, and that two hard-sell breaks inside a story leave 0.681 of the unbroken acceptance. (https://5dlibrary.com/answers/en/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c01/) 第1章 叙事传输:量表信度、效应衰减与故事中断的代价 — 第一章说明,两题问卷只能看到叙事传输真实效应的一部分,而故事中途插入两次硬广告,接受度只剩不中断时的 0.681。 (https://5dlibrary.com/answers/zh/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c01/) - Ch.2 Cognitive Dissonance: the Dissonance Ratio, Four Exits and What "Breaking Defences" Really Means — Chapter 2 uses Festinger's dissonance ratio and a four-exit choice model to show that the dissonance of a hard attack is mostly absorbed by disparaging the source, while a falsify-first opening is accepted 9.62 times as often. (https://5dlibrary.com/answers/en/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c02/) 第2章 认知失调:失调比率、四条出路与「破防」的真实含义 — 第二章用费斯廷格的失调比率与四条出路的选择模型说明,硬攻击的失调大多被读者对来源的贬低吸收,先证伪开头的接受概率是硬攻击的 9.62 倍。 (https://5dlibrary.com/answers/zh/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c02/) - Ch.3 Falsify First: Two-Sided Messages, the Size of a Concession and the Persuasion-Knowledge Backfire — Chapter 3 gives the Bayesian account of admitting a flaw: an admission carries a likelihood ratio of 6.0, yet admitting without answering is worse than silence, and pressure wording backfires beyond 4.58 markers per 1000 characters. (https://5dlibrary.com/answers/en/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c03/) 第3章 先证伪卸防备:双面信息、让步尺度与说服知识反噬 — 第三章给出承认缺点的贝叶斯账:承认的似然比是 6.0,但只承认不回应比沉默更糟;压迫性措辞超过每 1000 字 4.58 个就会反噬。 (https://5dlibrary.com/answers/zh/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c03/) - Ch.4 Open Loops: the Zeigarnik Replication Debate and the Pace of Closing an Information Gap — Chapter 4 examines the replication debate over the Zeigarnik effect, shows that recording only significant results logs a zero effect as 1.249, and finds the best pace of revelation at 0.063 of the remaining answer per section. (https://5dlibrary.com/answers/en/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c04/) 第4章 未完悬念:蔡加尼克效应的复制争议与信息缺口的揭晓节奏 — 第四章考辨蔡加尼克效应的复制争议,说明只记录显著结果会把零效应记成 1.249 倍,并算出每节揭晓剩余答案 0.063 的最佳节奏。 (https://5dlibrary.com/answers/zh/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c04/) - Ch.5 Scroll Depth as Survival: Weibull Hazards, Censoring and Cliff Sections — Chapter 5 treats scroll depth as a survival time: under negative aging the first two screens lose 0.347 of readers, while average scroll depth records a true willingness of 8.55 screens as 5.43. (https://5dlibrary.com/answers/en/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c05/) 第5章 滑屏深度生存曲线:威布尔风险率、删失与断崖段 — 第五章把滑屏深度当作生存时间:负老化时前两屏流失 0.347,而平均滑屏深度把 8.55 屏的真实意愿记成 5.43 屏。 (https://5dlibrary.com/answers/zh/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c05/) - Ch.6 The "Dopamine" Rhythm: Reward-Prediction Errors, Hook Debt and the True Ledger of Pressure — Chapter 6 tests the dopamine metaphor, showing that sales peak when hooks and payoffs are about one to one, that writing for depth cuts buyers per 1000 readers from 8.64 to 5.93, and that 0.794 of the excess net value of pressure copy comes from readers in financial distress. (https://5dlibrary.com/answers/en/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c06/) 第6章 「多巴胺」节律:奖励预测误差、钩子负债与施压的真实账本 — 第六章检验多巴胺比喻,说明钩子与兑现约一比一时成交最多,以深度为目标会让每 1000 名读者的买家从 8.64 降到 5.93,施压文案的超额净值有 0.794 来自财务困境读者。 (https://5dlibrary.com/answers/zh/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c06/) - Ch.7 The Classics Examined: Persuasions of the Zhanguo ce and Aristotle's Rhetoric — Chapter 7 examines persuasions of the Zhanguo ce and Aristotle's Rhetoric: six speeches coded by the author are soft before hard, an order that random shuffling yields 1 time in 4320, and at credibility 0.5 an argument can grow the odds at most 2.0 times. (https://5dlibrary.com/answers/en/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c07/) 第7章 原典考辨:《战国策》游说辞与亚里士多德《修辞学》 — 第七章考辨《战国策》游说辞与亚里士多德《修辞学》:作者编码的六篇说辞先软后硬,随机得到这种次序的机会是 4320 分之一;可信度 0.5 时论证最多把赔率放大 2.0 倍。 (https://5dlibrary.com/answers/zh/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c07/) - Ch.8 Practice and Falsification: the Persuasive-Copy Checker, Rules for Closing in Private Messages and a Pre-Registered Test — Chapter 8 gives a persuasive-copy checker of eight machine rules, limits and quiet hours for follow-up messages, and a pre-registered plan with at least 6,455 readers per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c08/) 第8章 实践与证伪:攻心文案检查器、私讯收口规则与预注册方案 — 第八章给出八条机检规则的攻心文案检查器、私讯跟进上限与静默规则,以及每组至少 6,455 名读者的预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/narrative-transportation-cognitive-dissonance-disarming-dopaminergic-scroll-rhythms-high/c08/) ## 5DL-B-142 · Self-Fulfilling Identity Labels, In-Group Belonging Calls and the Rhetoric of Resistance-Free High-Ticket Self-Persuasion 中文书名:身份标签自我实现、群体归属召唤与高客单零抗拒自我说服修辞学 DOI: pending Cite: Fan, Y. (2026). Self-Fulfilling Identity Labels, In-Group Belonging Calls and the Rhetoric of Resistance-Free High-Ticket Self-Persuasion (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not legal advice and not personalised financial or business advice. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: A method circulates among sellers of expensive services: lead the reader through small commitments, use circle jargon so that he feels "this was written for me", hand affluent readers an identity label so that they take the seat named for them, and let the reader persuade himself. This book does not teach anyone to use these devices harder. It writes each of them as a transparent model and weighs the size of the effect, the shape of the cost and the ethical boundary. Its parameters are illustrative; all calculations are generated by its programs; it is not legal advice and not personalised financial or business advice. Chapter 1 uses a Bayesian self-perception model to show that identity labels move only readers with middling priors, with a maximum gain from noun labels of only 0.0468; magnitude errors in small studies can reach 18.64 times, which grounds the central counter-intuitive proposition that under large-sample field conditions the gain from noun identity labels is close to zero. Chapter 2 writes the foot-in-the-door technique as an absorbing Markov chain; the best ladder has two steps and converts 0.2258, and maximum closing pressure yields only 0.679 of the best net income. Chapter 3 shows that jargon pays only when insiders exceed 0.4098 of the audience. Chapter 4 uses a Tajfel-style matrix and a threshold cascade to show that exposed plants raise the critical seed from 0.1203 to 0.2320. Chapter 5 covers insufficient justification, the number of reasons and an affordability stop line: selling stops when the price exceeds a quarter of the reader's liquid funds, and the 0.2648 of prospects turned away would have produced 0.6430 of refunds and regrets. Chapter 6 computes the backfire of excess: beyond 5.07 identity cues per three hundred characters copy does worse than plain copy, and a standard screen removes 0.6475 of the harm. Chapter 7 examines the Xueji from the Book of Rites and Aristotle's Rhetoric. Chapter 8 gives an owner's checklist, a persuasive-copy checker and a pre-registered test; detecting a one-point uplift needs 14,751 readers per arm. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:高客单销售圈流传着一套方法:用一致性承诺一步步引导,用圈层黑话让读者觉得「这是写给我的」,用身份标签让高净值读者主动对号入座,最后由读者自己说服自己。本书不教人更狠地使用这些手法,而是把它们逐一写成透明的模型,称出效果的大小、成本的形状与伦理的边界。书中参数为示意;全部计算由书中程序生成;本书不是法律意见,也不是个人化的财务或经营建议。 第一章用贝叶斯自我知觉模型说明身份标签只移动先验处于中段的读者,名词标签的最大增益只有 0.0468;小样本研究的量级错误可达 18.64 倍,由此提出核心反常识命题:名词身份标签在大样本现场条件下的提升接近于零。第二章把登门槛写成吸收马尔可夫链,最优阶梯只有两步,成交 0.2258;最大收口压力的净收入只有最优值的 0.679。第三章算出圈内人占比超过 0.4098 时黑话才划算。第四章用泰弗尔式矩阵与阈值级联说明托被识破后临界种子从 0.1203 升到 0.2320。 第五章讨论不足理由、理由数量与可负担性停止线,价格超过读者流动资金四分之一时停止成交,被挡下的 0.2648 意向客户本会贡献 0.6430 的退款与后悔。第六章计算过量的反噬:每三百字超过 5.07 处身份提示即不如平实文案,标准筛查去掉 0.6475 的伤害。第七章考辨《礼记·学记》与亚里士多德《修辞学》。第八章给出经营者清单、攻心文案检查器与预注册检验方案,检出一个百分点的提升每组需要 14,751 人。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Self-Fulfilling Identity Labels: Bayesian Self-Perception and the Shrinking Noun Effect — Chapter 1 uses a Bayesian self-perception model to show that identity labels move only readers with middling priors, with a maximum gain from noun labels of only 0.0468; magnitude errors in small studies can reach 18.64 times, which grounds the central counter-intuitive proposition that under large-sample field conditions the gain from noun identity labels is close to zero. (https://5dlibrary.com/answers/en/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c01/) 第1章 身份标签的自我实现:贝叶斯自我知觉与名词效应的缩水 — 第一章用贝叶斯自我知觉模型说明身份标签只移动先验处于中段的读者,名词标签的最大增益只有 0.0468;小样本研究的量级错误可达 18.64 倍,由此提出核心反常识命题:名词身份标签在大样本现场条件下的提升接近于零。 (https://5dlibrary.com/answers/zh/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c01/) - Ch.2 The Commitment Ladder: Foot-in-the-Door as an Absorbing Markov Chain, and the Refund Payback — Chapter 2 writes the foot-in-the-door technique as an absorbing Markov chain; the best ladder has two steps and converts 0.2258, and maximum closing pressure yields only 0.679 of the best net income. (https://5dlibrary.com/answers/en/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c02/) 第2章 一致性承诺阶梯:登门槛的吸收马尔可夫链与退款回吐 — 第二章把登门槛写成吸收马尔可夫链,最优阶梯只有两步,成交 0.2258;最大收口压力的净收入只有最优值的 0.679。 (https://5dlibrary.com/answers/zh/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c02/) - Ch.3 Circle Jargon: Optimal Distinctiveness and the Half-Life of Signal Dilution — Chapter 3 shows that jargon pays only when insiders exceed 0.4098 of the audience. (https://5dlibrary.com/answers/en/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c03/) 第3章 圈层黑话:最优区分度与信号稀释的半衰期 — 第三章算出圈内人占比超过 0.4098 时黑话才划算。 (https://5dlibrary.com/answers/zh/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c03/) - Ch.4 In-Group Belonging Calls: Minimal-Group Matrices and Threshold Cascades — Chapter 4 uses a Tajfel-style matrix and a threshold cascade to show that exposed plants raise the critical seed from 0.1203 to 0.2320. (https://5dlibrary.com/answers/en/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c04/) 第4章 群体归属召唤:最小群体矩阵与阈值级联 — 第四章用泰弗尔式矩阵与阈值级联说明托被识破后临界种子从 0.1203 升到 0.2320。 (https://5dlibrary.com/answers/zh/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c04/) - Ch.5 High-Ticket Self-Persuasion: Insufficient Justification, the Number of Reasons and an Affordability Stop Line — Chapter 5 covers insufficient justification, the number of reasons and an affordability stop line: selling stops when the price exceeds a quarter of the reader's liquid funds, and the 0.2648 of prospects turned away would have produced 0.6430 of refunds and regrets. (https://5dlibrary.com/answers/en/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c05/) 第5章 高客单自我说服:不足理由、理由数量与可负担性停止线 — 第五章讨论不足理由、理由数量与可负担性停止线,价格超过读者流动资金四分之一时停止成交,被挡下的 0.2648 意向客户本会贡献 0.6430 的退款与后悔。 (https://5dlibrary.com/answers/zh/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c05/) - Ch.6 Dose and Backfire: Persuasion Knowledge, Screening for Vulnerable Readers and a Cap on Follow-Ups — Chapter 6 computes the backfire of excess: beyond 5.07 identity cues per three hundred characters copy does worse than plain copy, and a standard screen removes 0.6475 of the harm. (https://5dlibrary.com/answers/en/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c06/) 第6章 剂量与反噬:说服知识、弱势读者筛查与跟进上限 — 第六章计算过量的反噬:每三百字超过 5.07 处身份提示即不如平实文案,标准筛查去掉 0.6475 的伤害。 (https://5dlibrary.com/answers/zh/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c06/) - Ch.7 Reading the Classics: the Art of Guiding in the Xueji and Aristotle's Enthymeme — An offer on day 3 converts only 0.0128 at once, yet closes the door on 0.3343 of readers. As the first offer is delayed, the share of readers who are ready reaches 0.632 by day 14, and the combined conversion of both offers peaks at 0.2461 with a first offer on day 21. When only one offer is allowed, the best day is 20, with conversion 0.1683. A first offer delayed to day 42 yields only 0.1052 from both offers together; that is "laborious and hard to complete". (https://5dlibrary.com/answers/en/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c07/) 第7章 原典考辨:《礼记·学记》的善喻之道与亚里士多德《修辞学》的省略推理 — 第 3 天就提出邀约,首次成交只有 0.0128,却有 0.3343 的读者因此关上了门。随着首次邀约推迟,到第 14 天,已经准备好的读者占 0.632;两次邀约合计的成交在第 21 天首邀时达到最大值 0.2461。只能邀约一次时,最优是第 20 天,成交 0.1683。若拖到第 42 天才首邀,两次合计只剩 0.1052,这就是「时过然后学,则勤苦而难成」。 (https://5dlibrary.com/answers/zh/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c07/) - Ch.8 Practice and a Plan for Refutation: the Persuasive-Copy Checker and a Pre-Registered Test — Chapter 8 gives an owner's checklist, a persuasive-copy checker and a pre-registered test; detecting a one-point uplift needs 14,751 readers per arm. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c08/) 第8章 实践与证伪方案:攻心文案检查器与预注册检验 — 第八章给出经营者清单、攻心文案检查器与预注册检验方案,检出一个百分点的提升每组需要 14,751 人。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/self-fulfilling-identity-labels-group-belonging-calls-rhetoric-resistance/c08/) ## 5DL-B-143 · The Inverted-U of Fear Arousal, Concrete Pain-Point Imagery and the Compliance Boundary of High-Conversion Copy 中文书名:恐惧唤醒倒U型曲线、具体化痛点成像与合规免责高转化文案边界 DOI: pending Cite: Fan, Y. (2026). The Inverted-U of Fear Arousal, Concrete Pain-Point Imagery and the Compliance Boundary of High-Conversion Copy (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All parameters are illustrative; all calculations are generated by the book's programs; the book is not legal advice nor personalised business or financial advice, and makes no judgement about the efficacy of any product. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: Three maxims circulate about crisis copy: the deeper the pain the stronger the sale, a frightened reader runs, and pain and antidote must be mixed in a golden ratio. This book writes them as runnable models, computes what each predicts, and states what data would overturn them. All parameters are illustrative; all calculations are generated by the book's programs; the book is not legal advice nor personalised business or financial advice, and makes no judgement about the efficacy of any product. Chapter 1 uses a two-step model of reception and yielding to show that the peak of the inverted U is set jointly by the rates of attention and defence, at dose 3.901 under the illustrative parameters, and notes that the three data points of the dental hygiene study lie on a straight line. Chapter 2 writes the extended parallel process model as an exact three-way split and proves that the threshold of defensive exit equals the audience's median efficacy, a critical dose of 1.980 for a mixed audience. Chapter 3 allocates a fixed budget between threat and antidote sentences: the best threat share is 0.446 for a mixed audience and moves between 0.220 and 0.697 with audience and length, so no fixed golden ratio exists; all-threat copy with no antidote reaches adoption of only 0.0003. Chapter 4 shows that a concrete pain-point picture amplifies dose and, without fewer threat sentences, pushes copy past the peak at a loss of 0.0868. Chapter 5 faces the counterintuitive meta-analytic verdict that within the range of existing studies stronger fear generally works better and evidence for an inverted U is weak; it explains how both can hold through stacked individual peaks and statistical power (with 200 people per level the chance of detecting a real bend is only 0.719), and states the conditions for refutation. Chapter 6 sets out the compliance red lines, overload from repetition, stopping conditions for readers in financial hardship or emotional crisis (harm per sale reaches a weight of 203.92 for the latter in the model), and the ledger in which plain disclosure nets 241.2 more once complaint costs are counted. Chapter 7 examines the Mencian maxim on sorrow and calamity beside Aristotle's account of fear in the Rhetoric. Chapter 8 gives an owner's checklist, a persuasion-copy checker, a cap on follow-up messages and a pre-registered testing plan (2,013 people per cell for the main comparison). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:危机感文案流传着三种说法:越痛越买、吓过头就跑、痛点与解药要按黄金比例配。本书把它们写成可以运行的模型,算出各自的预言,并写明什么数据能推翻它们。书中参数均为示意;全部计算由书中程序生成;本书不是法律意见,也不构成个人化的商业或理财建议,对任何产品的功效不作判断。 第一章用「先接收、再顺从」的两步模型说明倒 U 型的顶点由注意与防御的速率共同决定,示意参数下顶点在剂量 3.901,并指出牙齿卫生实验的三个数据点恰在一条直线上。第二章把扩展平行过程模型写成精确的三分法,证明防御退出的临界点等于受众效能的中位数,混合受众的临界剂量为 1.980。第三章在固定篇幅里配比威胁句与解药句:最优威胁占比在混合受众中为 0.446,随受众与篇幅在 0.220 到 0.697 之间移动,固定的黄金比例并不存在;没有解药的全威胁文案采纳只有 0.0003。第四章说明具体化的痛点画面是剂量放大器,不减少威胁句就会越过顶点,损失 0.0868。 第五章面对反常识的元分析结论:在已有研究范围内恐惧越强大体越有效,倒 U 型证据很弱;本书用个体顶点的叠加与检出力(每档 200 人时检出真实弯曲的机会只有 0.719)解释二者如何同时为真,并写明推翻条件。第六章写出合规红线、反复推送的过载、对财务困境与情绪危机读者的停止条件(模型中后者每单伤害权重达 203.92),以及清楚披露条款在投诉成本计入后净收益多出 241.2。第七章考辨《孟子》「生于忧患」与亚里士多德《修辞学》论恐惧。第八章给出经营者清单、攻心文案检查器、私讯跟进上限与预注册检验方案(主要比较每格 2,013 人)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Where the Inverted U Comes From: Reception Times Yielding and the Three Points of the Dental Hygiene Study — Chapter 1 uses a two-step model of reception and yielding to show that the peak of the inverted U is set jointly by the rates of attention and defence, at dose 3.901 under the illustrative parameters, and notes that the three data points of the dental hygiene study lie on a straight line. (https://5dlibrary.com/answers/en/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c01/) 第1章 倒 U 型从哪里来:接收乘以顺从的两步模型与牙齿卫生实验的三个点 — 第一章用「先接收、再顺从」的两步模型说明倒 U 型的顶点由注意与防御的速率共同决定,示意参数下顶点在剂量 3.901,并指出牙齿卫生实验的三个数据点恰在一条直线上。 (https://5dlibrary.com/answers/zh/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c01/) - Ch.2 The Threshold of Defensive Exit: an Exact Three-Way Split of the Extended Parallel Process Model — Chapter 2 writes the extended parallel process model as an exact three-way split and proves that the threshold of defensive exit equals the audience's median efficacy, a critical dose of 1.980 for a mixed audience. (https://5dlibrary.com/answers/en/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c02/) 第2章 防御退出的临界点:扩展平行过程模型的精确三分法 — 第二章把扩展平行过程模型写成精确的三分法,证明防御退出的临界点等于受众效能的中位数,混合受众的临界剂量为 1.980。 (https://5dlibrary.com/answers/zh/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c02/) - Ch.3 Pairing the Antidote: the Best Mix of Threat and Efficacy Sentences in a Fixed Budget, and the Bankruptcy of the Golden Ratio — Chapter 3 allocates a fixed budget between threat and antidote sentences: the best threat share is 0.446 for a mixed audience and moves between 0.220 and 0.697 with audience and length, so no fixed golden ratio exists; all-threat copy with no antidote reaches adoption of only 0.0003. (https://5dlibrary.com/answers/en/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c03/) 第3章 解药配对:固定篇幅里威胁句与效能句的最优配比与「黄金比例」的破产 — 第三章在固定篇幅里配比威胁句与解药句:最优威胁占比在混合受众中为 0.446,随受众与篇幅在 0.220 到 0.697 之间移动,固定的黄金比例并不存在;没有解药的全威胁文案采纳只有 0.0003。 (https://5dlibrary.com/answers/zh/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c03/) - Ch.4 Concrete Pain-Point Imagery: an Identifiable Picture Amplifies Dose, and Its Real Gain Is the Space It Frees — Chapter 4 shows that a concrete pain-point picture amplifies dose and, without fewer threat sentences, pushes copy past the peak at a loss of 0.0868. (https://5dlibrary.com/answers/en/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c04/) 第4章 具体化痛点成像:可识别的画面是剂量放大器,真正的收益是腾出篇幅 — 第四章说明具体化的痛点画面是剂量放大器,不减少威胁句就会越过顶点,损失 0.0868。 (https://5dlibrary.com/answers/zh/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c04/) - Ch.5 Does the Inverted U Exist? The Monotone Verdict of Meta-Analysis, Stacked Individual Peaks and the Power of Three-Arm Studies — Chapter 5 faces the counterintuitive meta-analytic verdict that within the range of existing studies stronger fear generally works better and evidence for an inverted U is weak; it explains how both can hold through stacked individual peaks and statistical power (with 200 people per level the chance of detecting a real bend is only 0.719), and states the conditions for refutation. (https://5dlibrary.com/answers/en/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c05/) 第5章 倒 U 型真的存在吗:元分析的单调结论、个体顶点的叠加与三档研究的检出力 — 第五章面对反常识的元分析结论:在已有研究范围内恐惧越强大体越有效,倒 U 型证据很弱;本书用个体顶点的叠加与检出力(每档 200 人时检出真实弯曲的机会只有 0.719)解释二者如何同时为真,并写明推翻条件。 (https://5dlibrary.com/answers/zh/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c05/) - Ch.6 The Compliance Boundary: Overload from Repetition, Stopping Conditions for Readers in Hardship, and the Ledger of Hidden Terms — Chapter 6 sets out the compliance red lines, overload from repetition, stopping conditions for readers in financial hardship or emotional crisis (harm per sale reaches a weight of 203.92 for the latter in the model), and the ledger in which plain disclosure nets 241.2 more once complaint costs are counted. (https://5dlibrary.com/answers/en/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c06/) 第6章 合规边界:反复推送的过载、处境艰难读者的停止条件与隐藏条款的账 — 第六章写出合规红线、反复推送的过载、对财务困境与情绪危机读者的停止条件(模型中后者每单伤害权重达 203.92),以及清楚披露条款在投诉成本计入后净收益多出 241.2。 (https://5dlibrary.com/answers/zh/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c06/) - Ch.7 The Classics Examined: Mencius on Being Born of Sorrow and Calamity, and Aristotle on Fear in the Rhetoric — The four regimes end very differently. In complete ease, capacity sinks from 0.400 to 0.272 by round 20. Under a fixed moderate challenge of 0.6, capacity reaches 0.579 by round 10 and then approaches 0.598 and stays there, since the challenge no longer exceeds capacity. When the challenge always sits a little above capacity and rises with it, capacity reaches 0.700 by round 10 and 0.951 by round 20. But facing an overwhelming challenge of 0.95 from the start, the gap is 0.55, far outside the growth band, and capacity slides every round to 0.000 by round 20. (https://5dlibrary.com/answers/en/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c07/) 第7章 原典考辨:《孟子》「生于忧患」与亚里士多德《修辞学》论恐惧 — 四种情形的结局截然不同。完全安乐时,能力从 0.400 一路降到第 20 轮的 0.272。固定在 0.6 的适度挑战下,能力在第 10 轮升到 0.579,之后逼近 0.598 并停在那里,因为挑战不再高于能力。挑战始终比能力高出一小截、随能力上调时,能力在第 10 轮达到 0.700,第 20 轮达到 0.951。而一开始就面对 0.95 的压倒性挑战时,差距是 0.55,远超成长带,能力每轮下滑,到第 20 轮降为 0.000。 (https://5dlibrary.com/answers/zh/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c07/) - Ch.8 Practice and a Plan for Refutation: an Owner's Checklist, a Persuasion-Copy Checker and a Pre-Registered Test — Chapter 8 gives an owner's checklist, a persuasion-copy checker, a cap on follow-up messages and a pre-registered testing plan (2,013 people per cell for the main comparison). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c08/) 第8章 实践与证伪方案:经营者清单、攻心文案检查器与预注册检验 — 第八章给出经营者清单、攻心文案检查器、私讯跟进上限与预注册检验方案(主要比较每格 2,013 人)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/inverted-u-fear-arousal-concrete-pain-point-imagery-compliance/c08/) ## 5DL-B-144 · Native Cross-Context Prosody, De-Translationese Syntax-Tree Rewriting and a Conversion Corpus of Malaysian, Singaporean and Taiwanese Chinese 中文书名:跨语境母语修辞韵律、去翻译腔句法树重构与马新台华语转化语料学 DOI: pending Cite: Fan, Y. (2026). Native Cross-Context Prosody, De-Translationese Syntax-Tree Rewriting and a Conversion Corpus of Malaysian, Singaporean and Taiwanese Chinese (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its example sentences and parameters are illustrative; all calculations are generated by its programs; it is not individual legal, advertising-compliance or marketing advice, and it discusses only persuasion that keeps the rules. Chapter 2 proves that the dependency length of stacked modifiers grows roughly with the square of the layers, so that at four layers the illustrative probability of keeping every relation is only 0.098. Abstract: Translationese in Chinese copy is usually treated as a matter of feel that can only be sensed. This book turns it into claims that can be computed, checked and refuted by data: sentences are annotated as dependency trees, the distance between related words is counted, and combinatorial counts, Poisson models, dynamic programming and simple probability models are applied to chains of the particle de, translationese markers, breath groups and regional vocabulary. Its example sentences and parameters are illustrative; all calculations are generated by its programs; it is not individual legal, advertising-compliance or marketing advice, and it discusses only persuasion that keeps the rules. Chapter 1 shows on illustrative hand-annotated trees that Europeanised drafts have a mean dependency distance of 3.429 and native rewrites 1.515; against random orders that keep the tree fixed, the drafts score 0.950, practically random, and the rewrites 0.640. Chapter 2 proves that the dependency length of stacked modifiers grows roughly with the square of the layers, so that at four layers the illustrative probability of keeping every relation is only 0.098. Chapter 3 writes translationese as ten machine-checkable markers, and a Poisson likelihood-ratio rule reaches an accuracy of 0.9614 at 300 hanzi. Chapter 4 gives three syntax-tree operations, relative-clause extraction bringing an example sentence down to 0.721 of its mean dependency distance, and proves that ordering modifiers long-outside is optimal, cutting total length at seven dependents to 0.725 of a random order. Chapter 5 places commas and breathing points in long sentences by dynamic programming. Chapter 6 builds a small lexicon of confident pairs only for Malaysian, Singaporean, Taiwanese and Hong Kong Chinese; a mainland draft read in Taiwan has an illustrative probability of 0.968 of being noticed as written elsewhere. Chapter 7 examines the Wenxin Diaolong on sentences and pruning and Orwell's six rules of writing: groups of three to six characters make up 0.842 of Liu Xie's own prose, while mechanical cutting closes only 0.154 of the gap between literal drafts and native rewrites. Chapter 8 gives the ten rules of a persuasive-copy checker, a ceiling on private follow-ups (two are best in the illustrative model, worth 0.0648 per enquiring customer), red lines and stopping conditions for vulnerable readers, and a pre-registered conversion trial of 6,745 customers per arm. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:中文文案的翻译腔常被当作一种只能意会的语感。本书把它改写成可以计算、可以复核、可以被数据推翻的命题:把句子标成依存树,数出相互依存的词之间隔了多远,再用组合计数、泊松模型、动态规划与简单的概率模型处理「的」字链、翻译腔标记、气口与跨地区用词。书中的例句与参数均为示意;全部计算由书中程序生成;本书不构成个别化的法律、广告合规或营销建议,只讨论守规矩的说服。 第一章用手工标注的示意依存树说明,西化稿的平均依存距离是 3.429,母语改写是 1.515;与保持树结构的随机语序相比,西化稿的比值为 0.950,几乎等于随机,改写稿为 0.640。第二章证明定语堆叠的依存长度近似平方增长,四层时全部关系保持住的示意概率只剩 0.098。第三章把翻译腔写成十种可机检标记,泊松似然比判别在 300 字时准确率达到 0.9614。第四章给出三种句法树操作,定语从句外置使示意句的平均依存距离降到原来的 0.721,并证明修饰语「长者在外」的排序最优,七个成分时把总长度降到随机次序的 0.725。 第五章用动态规划为长句安排逗号与气口。第六章建立只收确信对照的马来西亚、新加坡、台湾与香港华语小词表,大陆草稿给台湾读者看时被察觉为外地写的示意概率为 0.968。第七章考辨《文心雕龙》的章句、镕裁与奥威尔的六条写作规则:刘勰自己的文字中三到六字的段占 0.842,而机械删字只消除了直译稿与母语改写之间 0.154 的差距。第八章给出攻心文案检查器的十条规则、私讯跟进上限(示意模型中 2 次最优,每位询价顾客期望净值 0.0648)、红线与弱势读者的停止条件,以及每组 6,745 位顾客的预注册转化率检验。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Dependency Distance: Why Europeanised Long Sentences Are Hard to Read — Chapter 1 shows on illustrative hand-annotated trees that Europeanised drafts have a mean dependency distance of 3.429 and native rewrites 1.515; against random orders that keep the tree fixed, the drafts score 0.950, practically random, and the rewrites 0.640. (https://5dlibrary.com/answers/en/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c01/) 第1章 依存距离:西化长句为什么难读 — 第一章用手工标注的示意依存树说明,西化稿的平均依存距离是 3.429,母语改写是 1.515;与保持树结构的随机语序相比,西化稿的比值为 0.950,几乎等于随机,改写稿为 0.640。 (https://5dlibrary.com/answers/zh/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c01/) - Ch.2 Chains of De and Stacked Modifiers: an Exact Account of the Left-Branching Burden — Chapter 2 proves that the dependency length of stacked modifiers grows roughly with the square of the layers, so that at four layers the illustrative probability of keeping every relation is only 0.098. (https://5dlibrary.com/answers/en/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c02/) 第2章 「的」字链与定语堆叠:左分支负担的精确计算 — 第二章证明定语堆叠的依存长度近似平方增长,四层时全部关系保持住的示意概率只剩 0.098。 (https://5dlibrary.com/answers/zh/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c02/) - Ch.3 Fingerprints of Translationese: Ten Machine-Checkable Markers and a Poisson Likelihood-Ratio Test — Chapter 3 writes translationese as ten machine-checkable markers, and a Poisson likelihood-ratio rule reaches an accuracy of 0.9614 at 300 hanzi. (https://5dlibrary.com/answers/en/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c03/) 第3章 翻译腔指纹:十种可机检标记与泊松似然比判别 — 第三章把翻译腔写成十种可机检标记,泊松似然比判别在 300 字时准确率达到 0.9614。 (https://5dlibrary.com/answers/zh/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c03/) - Ch.4 Rewriting the Tree: Three Operations and the Long-Outside Ordering Theorem — Chapter 4 gives three syntax-tree operations, relative-clause extraction bringing an example sentence down to 0.721 of its mean dependency distance, and proves that ordering modifiers long-outside is optimal, cutting total length at seven dependents to 0.725 of a random order. (https://5dlibrary.com/answers/en/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c04/) 第4章 句法树重构:三种树操作与「长者在外」的排序定理 — 第四章给出三种句法树操作,定语从句外置使示意句的平均依存距离降到原来的 0.721,并证明修饰语「长者在外」的排序最优,七个成分时把总长度降到随机次序的 0.725。 (https://5dlibrary.com/answers/zh/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c04/) - Ch.5 Prosody and Breath: Segment Lengths, Dynamic Programming for Commas, and the Rhythm of Fours and Sixes — Breath groups can be counted. The illustrative drafts of this chapter have mean group lengths between 21.00 and 24.29 hanzi, the rewrites between 5.10 and 6.67, with groups of four to seven in the majority. Given the places where a sentence may be broken, dynamic programming finds the commas that come closest to a target length; with a target of 6, a sentence of 53 hanzi falls into 8 groups at a cost of 29, with every sentence boundary kept. The next chapter turns from the inside of the sentence to the place where it is read: does the same sentence read the same way in Taipei, Singapore, Kuala Lumpur and Hong Kong? (https://5dlibrary.com/answers/en/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c05/) 第5章 韵律与气口:分段长度、逗号的动态规划与四六节奏 — 气口段是可以数的。本章的示意直译稿平均段长在 21.00 到 24.29 字之间,改写稿在 5.10 到 6.67 字之间,四到七字的段占多数。给定可断的位置,动态规划能找到最接近目标段长的逗号位置;目标取 6 时,53 个字的长句被分成 8 段,代价 29,语义边界全部保留。下一章从句子的内部转向它所处的地区:同一句话,在台北、新加坡、吉隆坡与香港读起来是否一样。 (https://5dlibrary.com/answers/zh/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c05/) - Ch.6 Converting Between Malaysian, Singaporean, Taiwanese and Hong Kong Chinese: a Lexicon, an Affinity Matrix and a Library of Closing Sentences — Chapter 6 builds a small lexicon of confident pairs only for Malaysian, Singaporean, Taiwanese and Hong Kong Chinese; a mainland draft read in Taiwan has an illustrative probability of 0.968 of being noticed as written elsewhere. (https://5dlibrary.com/answers/en/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c06/) 第6章 马新台港华语转化:词汇对照、亲和矩阵与成交句式库 — 第六章建立只收确信对照的马来西亚、新加坡、台湾与香港华语小词表,大陆草稿给台湾读者看时被察觉为外地写的示意概率为 0.968。 (https://5dlibrary.com/answers/zh/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c06/) - Ch.7 The Classics Examined: Sentences and Pruning in the Wenxin Diaolong, and Orwell's Six Rules — Chapter 7 examines the Wenxin Diaolong on sentences and pruning and Orwell's six rules of writing: groups of three to six characters make up 0.842 of Liu Xie's own prose, while mechanical cutting closes only 0.154 of the gap between literal drafts and native rewrites. (https://5dlibrary.com/answers/en/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c07/) 第7章 原典考辨:《文心雕龙》的章句镕裁与奥威尔的六条规则 — 第七章考辨《文心雕龙》的章句、镕裁与奥威尔的六条写作规则:刘勰自己的文字中三到六字的段占 0.842,而机械删字只消除了直译稿与母语改写之间 0.154 的差距。 (https://5dlibrary.com/answers/zh/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c07/) - Ch.8 Practice and Refutation: the Persuasive-Copy Checker, a Ceiling on Private Follow-Ups, and a Pre-Registered Test — Chapter 8 gives the ten rules of a persuasive-copy checker, a ceiling on private follow-ups (two are best in the illustrative model, worth 0.0648 per enquiring customer), red lines and stopping conditions for vulnerable readers, and a pre-registered conversion trial of 6,745 customers per arm. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c08/) 第8章 实践与证伪:攻心文案检查器、私讯跟进上限与预注册检验 — 第八章给出攻心文案检查器的十条规则、私讯跟进上限(示意模型中 2 次最优,每位询价顾客期望净值 0.0648)、红线与弱势读者的停止条件,以及每组 6,745 位顾客的预注册转化率检验。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/native-cross-context-prosody-de-translationese-syntax-tree-rewriting/c08/) ## 5DL-B-145 · Instant-Messaging Bubble Rhythm, Read-No-Reply Icebreaking Games and High-Ticket Closing in Private Messages 中文书名:即时通讯气泡节奏学、已读不回破冰博弈与高客单私讯逼单母典 DOI: pending Cite: Fan, Y. (2026). Instant-Messaging Bubble Rhythm, Read-No-Reply Icebreaking Games and High-Ticket Closing in Private Messages (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not individual legal, compliance or business advice, and it discusses only closing that keeps the rules. Chapter 1 splits an illustrative 184-character quote by dynamic programming into 5 bubbles at a cost of 46.52, against 142.00 for one sentence per bubble. Abstract: More and more high-ticket business is concluded inside instant-messaging windows, yet what the trade says about the rhythm of private messages, that messages must be short, that read-but-silent customers must be chased at once, that payment reminders should spell out consequences, mostly counts one side only, gives no thresholds and argues from survivors, so it cannot be refuted. This book rewrites such claims as models with parameters: dynamic programming, reading-completion probabilities, mixtures of exponential decays, Erlang C queues, Bayesian updating, a two-player game, mixed logit, quasi-hyperbolic discounting and exhaustive enumeration of reminder schedules, and every conclusion comes with the conditions under which data would overturn it. Its parameters are illustrative; all calculations are generated by its programs; it is not individual legal, compliance or business advice, and it discusses only closing that keeps the rules. Chapter 1 splits an illustrative 184-character quote by dynamic programming into 5 bubbles at a cost of 46.52, against 142.00 for one sentence per bubble. Chapter 2 shows interest in an inquiry halving in about 63 minutes, with 6 agents needed for 40 inquiries an hour to keep waits over 5 minutes below 0.10. Chapter 3 shows that after about 5.1 hours of read-but-silent the probability of interest drops below 0.30, that the best permitted moment for an icebreaker is the next morning, and that a public commitment never to chase on read raises the seller's payoff in an illustrative game from 0.680 to 0.800. Chapter 4 shows a two-track quote earning 1.774 times the expected revenue per inquiry of the basic track alone, with a real three-track menu reaching 2,209, above a foil track. Chapter 5 finds the best share for a refundable deposit at about 0.102 and gives a break-even table for handing chats to a person. Chapter 6 enumerates every reminder schedule within 72 hours: two gentle reminders bring the probability of payment to 0.581, a third does not pay its way, and a threatening tone loses 47,736 over a year. Chapter 7 examines the chapter on assessing feelings in the Guiguzi and Bacon's Of Negotiating: several cues combined reach an AUC of 0.756, and a ladder of commitments is best at 4 steps, closing at 0.165. Chapter 8 gives a seven-scene closing skill pack for customer service, a stop-word replay with a recall of 1.000 and a pre-registered testing plan, in which a trial of first-reply delay needs 8,158 inquiries per arm. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:高客单生意越来越多在即时通讯的对话框里谈成,而行业里关于私讯节奏的说法,例如消息要短、已读不回要立刻追、催款要讲后果,大多只算一边、不给阈值、靠幸存者讲道理,因而无法证伪。本书把这些说法改写成带参数的模型:动态规划、阅读完成概率、指数衰减混合、爱尔朗 C 排队、贝叶斯更新、两人博弈、混合 logit、准双曲贴现与提醒排程的穷举枚举,每一个结论都给出可以被数据推翻的条件。书中参数为示意;全部计算由书中程序生成;本书不构成个别化的法律、合规或经营建议,只讨论守规矩的成交。 第一章用动态规划切分一条 184 字的示意报价,最优为 5 个气泡,代价 46.52,一句一条的代价是 142.00。第二章显示询价热度约 63 分钟减半,每小时 40 条询价需要 6 位客服才能把超过 5 分钟的等待压在 0.10 以下。第三章证明已读沉默约 5.1 小时后仍有兴趣的概率跌破 0.30,破冰消息的最佳可发时刻是次日早上,而公开承诺「已读不追」使商家在示意博弈中的收益从 0.680 升到 0.800。第四章显示双轨报价相对只报基础档,每条询价期望收入为 1.774 倍,真实的三档报价达到 2,209,高于陪衬档。第五章求出可退定金的最优比例约为 0.102,并给出转人工的盈亏平衡表。 第六章枚举 72 小时内全部提醒排程,两次温和提醒使付款概率达到 0.581,第三次得不偿失,威胁语气一年净值少 47,736。第七章考辨《鬼谷子·揣篇》与培根《论谈判》:多条线索合并的 AUC 为 0.756,承诺阶梯以 4 步为最优,成交率 0.165。第八章给出七个场景的客服促成成交技能包、召回率 1.000 的停止词回放与预注册检验方案,首次回复时延试验每组需要 8,158 条询价。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Bubble Splitting: Working-Memory Capacity, Interruption Cost and the Optimal Length of a Private Message by Dynamic Programming — Chapter 1 splits an illustrative 184-character quote by dynamic programming into 5 bubbles at a cost of 46.52, against 142.00 for one sentence per bubble. (https://5dlibrary.com/answers/en/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c01/) 第1章 气泡切分:工作记忆容量、打断成本与单条私讯最优字数的动态规划 — 第一章用动态规划切分一条 184 字的示意报价,最优为 5 个气泡,代价 46.52,一句一条的代价是 142.00。 (https://5dlibrary.com/answers/zh/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c01/) - Ch.2 First-Reply Latency: the Cooling Curve of an Inquiry, Night-Time Auto-Acknowledgements and Erlang C Staffing — Chapter 2 shows interest in an inquiry halving in about 63 minutes, with 6 agents needed for 40 inquiries an hour to keep waits over 5 minutes below 0.10. (https://5dlibrary.com/answers/en/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c02/) 第2章 首响时延:询价冷却曲线、夜间自动回执与爱尔朗 C 排班 — 第二章显示询价热度约 63 分钟减半,每小时 40 条询价需要 6 位客服才能把超过 5 分钟的等待压在 0.10 以下。 (https://5dlibrary.com/answers/zh/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c02/) - Ch.3 Read but No Reply: Bayesian Updating on Receipts, the Timing of an Icebreaker and a Game of Commitment — Chapter 3 shows that after about 5.1 hours of read-but-silent the probability of interest drops below 0.30, that the best permitted moment for an icebreaker is the next morning, and that a public commitment never to chase on read raises the seller's payoff in an illustrative game from 0.680 to 0.800. (https://5dlibrary.com/answers/en/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c03/) 第3章 已读不回:回执信号的贝叶斯更新、破冰时机与承诺博弈 — 第三章证明已读沉默约 5.1 小时后仍有兴趣的概率跌破 0.30,破冰消息的最佳可发时刻是次日早上,而公开承诺「已读不追」使商家在示意博弈中的收益从 0.680 升到 0.800。 (https://5dlibrary.com/answers/zh/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c03/) - Ch.4 Laying the Ground for a Two-Track Quote: Mixed Logit, the Compromise Effect and Anchoring on the Higher Track — Chapter 4 shows a two-track quote earning 1.774 times the expected revenue per inquiry of the basic track alone, with a real three-track menu reaching 2,209, above a foil track. (https://5dlibrary.com/answers/en/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c04/) 第4章 双轨报价铺垫:混合 logit、折中效应与先报高档的锚定 — 第四章显示双轨报价相对只报基础档,每条询价期望收入为 1.774 倍,真实的三档报价达到 2,209,高于陪衬档。 (https://5dlibrary.com/answers/zh/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c04/) - Ch.5 Closing a High-Ticket Sale: Refundable Deposits, Default Options and the Threshold for Handing Over to a Human — Chapter 5 finds the best share for a refundable deposit at about 0.102 and gives a break-even table for handing chats to a person. (https://5dlibrary.com/answers/en/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c05/) 第5章 高客单收口:可退定金、默认选项与转人工阈值 — 第五章求出可退定金的最优比例约为 0.102,并给出转人工的盈亏平衡表。 (https://5dlibrary.com/answers/zh/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c05/) - Ch.6 Placed but Not Paid: Present Bias, Gentle Reminder Pulses and a Year's Account of a Threatening Tone — Chapter 6 enumerates every reminder schedule within 72 hours: two gentle reminders bring the probability of payment to 0.581, a third does not pay its way, and a threatening tone loses 47,736 over a year. (https://5dlibrary.com/answers/en/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c06/) 第6章 超时未付:现时偏差、温和催款脉冲与威胁语气的年度账 — 第六章枚举 72 小时内全部提醒排程,两次温和提醒使付款概率达到 0.581,第三次得不偿失,威胁语气一年净值少 47,736。 (https://5dlibrary.com/answers/zh/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c06/) - Ch.7 The Classics Examined: the Chapter on Assessing Feelings in the Guiguzi and Bacon's Of Negotiating — Chapter 7 examines the chapter on assessing feelings in the Guiguzi and Bacon's Of Negotiating: several cues combined reach an AUC of 0.756, and a ladder of commitments is best at 4 steps, closing at 0.165. (https://5dlibrary.com/answers/en/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c07/) 第7章 原典考辨:《鬼谷子·揣篇》与培根《论谈判》 — 第七章考辨《鬼谷子·揣篇》与培根《论谈判》:多条线索合并的 AUC 为 0.756,承诺阶梯以 4 步为最优,成交率 0.165。 (https://5dlibrary.com/answers/zh/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c07/) - Ch.8 Practice and Refutation: a Closing Skill Pack for Customer Service, a Rule Replay and Pre-Registered Tests — Chapter 8 gives a seven-scene closing skill pack for customer service, a stop-word replay with a recall of 1.000 and a pre-registered testing plan, in which a trial of first-reply delay needs 8,158 inquiries per arm. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c08/) 第8章 实践与证伪:客服促成成交技能包、规则回放与预注册检验 — 第八章给出七个场景的客服促成成交技能包、召回率 1.000 的停止词回放与预注册检验方案,首次回复时延试验每组需要 8,158 条询价。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/instant-messaging-bubble-rhythm-read-no-reply-icebreaking-games/c08/) ## 5DL-B-146 · Consultative Objection Judo, a Bayesian Decomposition of "Too Expensive" and Deposit-Lock Stopping-Time Protocols 中文书名:顾问式销售异议柔道、价格嫌贵贝叶斯拆解与定金锁单停时协议 DOI: pending Cite: Fan, Y. (2026). Consultative Objection Judo, a Bayesian Decomposition of "Too Expensive" and Deposit-Lock Stopping-Time Protocols (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its customer populations, probabilities and amounts are illustrative; all calculations are generated by its programs; it is not individual legal, financial or investment advice, and it discusses only persuasion that keeps the rules. Chapter 1's objection tree shows that, under illustrative priors, 0.734 of those who say "too expensive" are not short of money. Abstract: The three objections most often heard in consultative selling are "too expensive", "let me think" and "I need to ask my partner". Trade slogans treat them as obstacles to break down and fit every outcome. This book treats the three sentences as data: behind each lie hidden causes such as budget, value, rival quotes, timing, joint decisions and polite refusal, and Bayes' rule, information gain, expected utility, survival models, simulations of choice conflict, a unanimity model, prospect theory, a screening model and optimal stopping turn "how to reply" and "when to stop" into claims that can be computed and refuted by data. Its customer populations, probabilities and amounts are illustrative; all calculations are generated by its programs; it is not individual legal, financial or investment advice, and it discusses only persuasion that keeps the rules. Chapter 1's objection tree shows that, under illustrative priors, 0.734 of those who say "too expensive" are not short of money. Chapter 2 takes the objection apart on a simulated population: at the market price an unclear value accounts for 0.570 and lack of budget for only 0.136; one honest walk-through dissolves 0.334 of objections against 0.241 for a ten per cent cut, and a reflexive cut earns 0.391 less margin than asking first; a twelve-instalment plan with a low-looking monthly fee carries a nominal annual rate of 0.1303. Chapter 3 reads silence: seven days after "let me think" the chance of a return within the next week is only 0.091, and going from two plans to six lowers the chance of a sale from 0.635 to 0.492. Chapter 4 uses a unanimity model to show that a one-page brief for both decision-makers earns 2.044 times the margin of pressing one partner to sign, while the posterior that the sentence is true rises from 0.529 in the low price band to 0.973 in the high one. Chapter 5 computes the three faces of a deposit: a forfeitable deposit borrows a grip of about 2.5131 times its amount, and the Chinese Civil Code sets a deposit cap and a penalty rule; the book's protocol refunds in full without reason for seven days. Chapter 6 writes closing as a stopping problem: a single closing request succeeds with a probability approaching 0.3679, follow-ups are capped at two, and a three-day refundable lock window raises the final purchase probability from 0.225 to 0.488. Chapter 7 examines Han Feizi's "Difficulties of Persuasion" and Franklin's moral algebra: four days of weighing agree with the right decision 0.8275 of the time. Chapter 8 gives a closing skill pack for service teams with red lines and hand-over conditions; in the illustrative simulation its net margin is 1.809 times that of a pressure playbook, and a pre-registered trial needs 1,251 leads per group. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:顾问式销售里最常听到的三句异议是「太贵了」「我再想想」「我问问爱人」。行业口诀把它们当作要攻破的障碍,而且无论结果如何都说得通。本书把三句话当作观测数据:每一句背后都有预算、价值、比价、时机、共同决策与婉拒等隐藏原因,用贝叶斯规则、信息增益、期望效用、生存模型、选项冲突模拟、一致同意模型、前景理论、筛选模型与最优停止理论,把「怎样回应」「什么时候停」改写成可以计算、可以被数据推翻的命题。书中的顾客群、概率与金额均为示意;全部计算由书中程序生成;本书不构成个人化的法律、财务或投资建议,只讨论守规矩的说服。 第一章的判别树说明,在示意先验下,说「太贵了」的人有 0.734 并不缺钱。第二章用模拟顾客群拆解嫌贵:市场价处没看懂价值占 0.570,缺预算只占 0.136;一次诚实讲解化解 0.334 的异议,降价一成只化解 0.241,反射式降价比先问后答少赚 0.391 的毛利;十二期分期的月费率看似很低,名义年化却是 0.1303。第三章读解沉默:说完「再想想」后第七天,七天内回来的概率只剩 0.091;方案从两个增加到六个,成交概率从 0.635 降到 0.492。第四章用一致同意模型说明,为两位决策者准备一页说明,毛利是催一方当场签约的 2.044 倍,而这句话为真的后验从低价段的 0.529 升到高价段的 0.973。 第五章计算定金的三重身份:不可退定金借来的锁定力约为定金的 2.5131 倍,并按《民法典》说明定金上限与罚则,给出七天内无理由全额退的定金协议。第六章把收口写成停时问题:只提一次的成交请求成功率趋近 0.3679,跟进以两次为上限,三天的可退锁单时窗把最终购买概率从 0.225 提到 0.488。第七章考辨《韩非子·说难》与富兰克林的道德代数:称四天的决定一致率达到 0.8275。第八章给出客服收口技能包、红线与转人工条件,示意模拟中技能包的净毛利是施压剧本的 1.809 倍,并给出每组 1,251 条线索的预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Objection Tree: Six Hidden Causes behind Three Sentences — Chapter 1's objection tree shows that, under illustrative priors, 0.734 of those who say "too expensive" are not short of money. (https://5dlibrary.com/answers/en/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c01/) 第1章 异议判别树:三句话背后的六个隐因 — 第一章的判别树说明,在示意先验下,说「太贵了」的人有 0.734 并不缺钱。 (https://5dlibrary.com/answers/zh/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c01/) - Ch.2 A Bayesian Decomposition of "Too Expensive": Budget, Value, Rival Quotes and Polite Refusal — Chapter 2 takes the objection apart on a simulated population: at the market price an unclear value accounts for 0.570 and lack of budget for only 0.136; one honest walk-through dissolves 0.334 of objections against 0.241 for a ten per cent cut, and a reflexive cut earns 0.391 less margin than asking first; a twelve-instalment plan with a low-looking monthly fee carries a nominal annual rate of 0.1303. (https://5dlibrary.com/answers/en/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c02/) 第2章 「太贵了」的贝叶斯拆解:预算、价值、比价与婉拒 — 第二章用模拟顾客群拆解嫌贵:市场价处没看懂价值占 0.570,缺预算只占 0.136;一次诚实讲解化解 0.334 的异议,降价一成只化解 0.241,反射式降价比先问后答少赚 0.391 的毛利;十二期分期的月费率看似很低,名义年化却是 0.1303。 (https://5dlibrary.com/answers/zh/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c02/) - Ch.3 "Let Me Think": the Posterior of Silence and Choice Conflict — Chapter 3 reads silence: seven days after "let me think" the chance of a return within the next week is only 0.091, and going from two plans to six lowers the chance of a sale from 0.635 to 0.492. (https://5dlibrary.com/answers/en/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c03/) 第3章 「我再想想」:沉默的后验与选项冲突 — 第三章读解沉默:说完「再想想」后第七天,七天内回来的概率只剩 0.091;方案从两个增加到六个,成交概率从 0.635 降到 0.492。 (https://5dlibrary.com/answers/zh/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c03/) - Ch.4 "I Need to Ask My Partner": a Unanimity Model of Joint Decisions — Chapter 4 uses a unanimity model to show that a one-page brief for both decision-makers earns 2.044 times the margin of pressing one partner to sign, while the posterior that the sentence is true rises from 0.529 in the low price band to 0.973 in the high one. (https://5dlibrary.com/answers/en/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c04/) 第4章 「我问问爱人」:共同决策的一致同意模型 — 第四章用一致同意模型说明,为两位决策者准备一页说明,毛利是催一方当场签约的 2.044 倍,而这句话为真的后验从低价段的 0.529 升到高价段的 0.973。 (https://5dlibrary.com/answers/zh/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c04/) - Ch.5 Three Faces of a Deposit: Loss Aversion, Screening and the Civil Code — Chapter 5 computes the three faces of a deposit: a forfeitable deposit borrows a grip of about 2.5131 times its amount, and the Chinese Civil Code sets a deposit cap and a penalty rule; the book's protocol refunds in full without reason for seven days. (https://5dlibrary.com/answers/en/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c05/) 第5章 定金的三重身份:损失厌恶、筛选信号与民法典规则 — 第五章计算定金的三重身份:不可退定金借来的锁定力约为定金的 2.5131 倍,并按《民法典》说明定金上限与罚则,给出七天内无理由全额退的定金协议。 (https://5dlibrary.com/answers/zh/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c05/) - Ch.6 Stopping-Time Protocols: the Secretary Problem, Follow-up Caps and Lock Windows — Chapter 6 writes closing as a stopping problem: a single closing request succeeds with a probability approaching 0.3679, follow-ups are capped at two, and a three-day refundable lock window raises the final purchase probability from 0.225 to 0.488. (https://5dlibrary.com/answers/en/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c06/) 第6章 停时协议:秘书问题、跟进上限与锁单时窗 — 第六章把收口写成停时问题:只提一次的成交请求成功率趋近 0.3679,跟进以两次为上限,三天的可退锁单时窗把最终购买概率从 0.225 提到 0.488。 (https://5dlibrary.com/answers/zh/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c06/) - Ch.7 Classical Sources: Han Feizi's "Difficulties of Persuasion" and Franklin's Moral Algebra — Chapter 7 examines Han Feizi's "Difficulties of Persuasion" and Franklin's moral algebra: four days of weighing agree with the right decision 0.8275 of the time. (https://5dlibrary.com/answers/en/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c07/) 第7章 原典考辨:《韩非子·说难》与富兰克林的道德代数 — 第七章考辨《韩非子·说难》与富兰克林的道德代数:称四天的决定一致率达到 0.8275。 (https://5dlibrary.com/answers/zh/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c07/) - Ch.8 Practice and Refutation: a Closing Skill Pack and a Pre-registered Trial — Chapter 8 gives a closing skill pack for service teams with red lines and hand-over conditions; in the illustrative simulation its net margin is 1.809 times that of a pressure playbook, and a pre-registered trial needs 1,251 leads per group. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c08/) 第8章 实践与证伪方案:客服收口技能包与预注册检验 — 第八章给出客服收口技能包、红线与转人工条件,示意模拟中技能包的净毛利是施压剧本的 1.809 倍,并给出每组 1,251 条线索的预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/consultative-objection-judo-bayesian-decomposition-too-expensive-deposit-lock/c08/) ## 5DL-B-147 · Seven-Syllable Auditory Hooks in Short Video and Livestreams, Comment-Section Herding Ignition and Pulse-Closing Scripts 中文书名:短视频与直播前七字听觉劫持、评论区从众引爆与脉冲逼单剧本学 DOI: pending Cite: Fan, Y. (2026). Seven-Syllable Auditory Hooks in Short Video and Livestreams, Comment-Section Herding Ignition and Pulse-Closing Scripts (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not individual legal, advertising-compliance or marketing advice. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Selling through short video and live streams runs on slogans: the first seven characters decide everything, a hook every three seconds, steer the comments first, shout out the last few orders. This book rewrites them as claims that can be computed, checked and refuted by data, and it discusses only persuasion that keeps the rules: announce only real back-end stock, no fake countdowns, no paid comments, no planted shills. Its parameters are illustrative; all calculations are generated by its programs; it is not individual legal, advertising-compliance or marketing advice. Chapter 1 computes the Bayesian surprise of an opening's first seven syllables with a Gamma-Poisson model; staying is an inverted-U function of surprise, with a best value of 1.79 bits and a band of 1.13 to 2.62 bits keeping nine tenths of the gain in the illustrative model, and the seven-character rule turns out to depend on speech rate. Chapter 2 recasts the retention curve as a censored survival function; the same proportional edit in the first three seconds gains 3.65 times as much mean watch time as one later on. Chapter 3 shows that hooks habituate, the last hook at a three-second spacing keeping only 0.102 of its effect, and that net value is an inverted U once negative feedback is counted. Chapter 4 solves information cascades exactly as a Markov chain; a single fake recommendation makes a wrong cascade 2.63 times as likely. Chapter 5 uses Hawkes processes and the branching ratio to explain how a comment section is genuinely set alight. Chapter 6 gives a geometric ladder for announcing real stock and the best frequency of closing pulses, a pulse every minute earning only 0.863 of the best; fake scarcity stops paying once the chance of exposure per show reaches 0.0123. Chapter 7 examines the Yueji on how things move the heart and Helmholtz's theory of beats, roughness peaking at about 26.0 beats per second. Chapter 8 gives a rule-abiding closing kit, a ceiling and quiet hours for private follow-ups, rules for stopping and handing over to a person, and a pre-registered conversion trial of 6,745 viewers per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:短视频与直播卖货流传着许多口诀:前七个字定生死,每三秒一个钩子,评论区要先带节奏,最后几单要喊出来。本书把这些口诀改写成可以计算、可以复核、可以被数据推翻的命题,并且只讨论守规矩的说服:只报后台真实库存,不搞虚假倒计时,不雇人刷评论,不安排托。书中参数均为示意;全部计算由书中程序生成;本书不构成个别化的法律、广告合规或营销建议。 第一章用伽马-泊松模型计算开场七个音节的贝叶斯惊奇度,留存与惊奇度呈倒 U 关系,示意模型的最优值为 1.79 比特,九成增益区间为 1.13 到 2.62 比特,「七个字」取决于语速。第二章把留存曲线还原为有删失的生存函数,前三秒的同幅改动对平均观看时长的增益是后段的 3.65 倍。第三章说明钩子会习惯化,每三秒一个钩子时最后一个钩子只剩 0.102 的效果,计入负面反馈后净价值呈倒 U 形。第四章用精确马尔可夫链求解信息级联,一条假推荐就把错误级联的概率放大到 2.63 倍。 第五章用霍克斯过程与分支比解释评论区如何被真实地点燃。第六章给出真实库存的几何报数阶梯与收口脉冲的最优频率,每分钟一次脉冲只得到最优的 0.863;虚假稀缺只要每场被识破的概率达到 0.0123 就不再划算。第七章考辨《乐记》的感物动心之说与亥姆霍兹的拍音理论,粗糙度在每秒约 26.0 次拍时最强。第八章给出合规收口技能包、私讯跟进上限与静默时段、停止与转人工规则,以及每组 6,745 位观众的预注册转化率检验。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Acoustic Surprise in the First Seven Syllables: Bayesian Surprise, an Inverted-U Threshold and Speech Rate — Chapter 1 computes the Bayesian surprise of an opening's first seven syllables with a Gamma-Poisson model; staying is an inverted-U function of surprise, with a best value of 1.79 bits and a band of 1.13 to 2.62 bits keeping nine tenths of the gain in the illustrative model, and the seven-character rule turns out to depend on speech rate. (https://5dlibrary.com/answers/en/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c01/) 第1章 前七字的声学惊奇度:贝叶斯惊奇、倒 U 阈值与语速 — 第一章用伽马-泊松模型计算开场七个音节的贝叶斯惊奇度,留存与惊奇度呈倒 U 关系,示意模型的最优值为 1.79 比特,九成增益区间为 1.13 到 2.62 比特,「七个字」取决于语速。 (https://5dlibrary.com/answers/zh/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c01/) - Ch.2 Attention Survival Curves: Censoring, the Hazard Function and the Leverage of the First Three Seconds — Chapter 2 recasts the retention curve as a censored survival function; the same proportional edit in the first three seconds gains 3.65 times as much mean watch time as one later on. (https://5dlibrary.com/answers/en/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c02/) 第2章 注意力生存曲线:删失、风险函数与前三秒的杠杆 — 第二章把留存曲线还原为有删失的生存函数,前三秒的同幅改动对平均观看时长的增益是后段的 3.65 倍。 (https://5dlibrary.com/answers/zh/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c02/) - Ch.3 The Density of Anti-Swipe Hooks: Habituation, Diminishing Returns and the Backlash of Excess — Chapter 3 shows that hooks habituate, the last hook at a three-second spacing keeping only 0.102 of its effect, and that net value is an inverted U once negative feedback is counted. (https://5dlibrary.com/answers/en/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c03/) 第3章 防划走钩子密度:习惯化、边际递减与过量的反噬 — 第三章说明钩子会习惯化,每三秒一个钩子时最后一个钩子只剩 0.102 的效果,计入负面反馈后净价值呈倒 U 形。 (https://5dlibrary.com/answers/zh/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c03/) - Ch.4 Herding in the Comment Section: Asch, Information Cascades and Cumulative Advantage — Chapter 4 solves information cascades exactly as a Markov chain; a single fake recommendation makes a wrong cascade 2.63 times as likely. (https://5dlibrary.com/answers/en/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c04/) 第4章 评论区从众:阿希实验、信息级联与累积优势 — 第四章用精确马尔可夫链求解信息级联,一条假推荐就把错误级联的概率放大到 2.63 倍。 (https://5dlibrary.com/answers/zh/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c04/) - Ch.5 Igniting the Comment Section: Branching Ratio, Hawkes Processes and Genuine First Comments — The key quantity in igniting a comment section is the branching ratio. In the illustrative model, ten genuine questions grow into 20.0 comments on average at a ratio of 0.5 and into 100.0 at 0.9; the nearer to criticality, the more each improvement is amplified. In the first five minutes, answering by name doubles the expected comments, while hired commenters mostly buy fake comments and add little from real viewers. Near criticality a comment section amplifies complaints too, so the hotter the room, the more it needs complaints handled in real time. The next chapter turns to another core step of live selling: the cadence of announcing real stock, and closing in pulses. (https://5dlibrary.com/answers/en/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c05/) 第5章 评论区引爆:分支比、霍克斯过程与真实首评 — 引爆评论区的关键量是分支比。示意模型中,从 10 条真实提问出发,分支比 0.5 平均只得 20.0 条,0.9 平均得 100.0 条;越接近临界,同样的改进放大越多。前五分钟里,认真点名回答使期望评论数翻倍,雇人刷评论买来的主要是假评论,真实评论只多了一点。临界附近的评论区同样会放大投诉,因此越热的直播间越需要实时处理投诉。下一章转向直播间的另一个核心环节:真实库存的报数节奏与脉冲收口。 (https://5dlibrary.com/answers/zh/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c05/) - Ch.6 The Cadence of Scarcity: Geometric Announcements of Real Stock and Closing in Pulses — Chapter 6 gives a geometric ladder for announcing real stock and the best frequency of closing pulses, a pulse every minute earning only 0.863 of the best; fake scarcity stops paying once the chance of exposure per show reaches 0.0123. (https://5dlibrary.com/answers/en/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c06/) 第6章 稀缺报数节奏:真实库存的几何报数与脉冲收口 — 第六章给出真实库存的几何报数阶梯与收口脉冲的最优频率,每分钟一次脉冲只得到最优的 0.863;虚假稀缺只要每场被识破的概率达到 0.0123 就不再划算。 (https://5dlibrary.com/answers/zh/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c06/) - Ch.7 The Classics Examined: the Yueji on Being Moved by Things, and Helmholtz on Beats — Chapter 7 examines the Yueji on how things move the heart and Helmholtz's theory of beats, roughness peaking at about 26.0 beats per second. (https://5dlibrary.com/answers/en/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c07/) 第7章 原典考辨:《乐记》的感物动心与亥姆霍兹的拍音 — 第七章考辨《乐记》的感物动心之说与亥姆霍兹的拍音理论,粗糙度在每秒约 26.0 次拍时最强。 (https://5dlibrary.com/answers/zh/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c07/) - Ch.8 Practice and Refutation: a Rule-Abiding Closing Procedure for the Live Room and Pre-Registered Tests — Chapter 8 gives a rule-abiding closing kit, a ceiling and quiet hours for private follow-ups, rules for stopping and handing over to a person, and a pre-registered conversion trial of 6,745 viewers per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c08/) 第8章 实践与证伪:直播间合规收口流程与预注册检验 — 第八章给出合规收口技能包、私讯跟进上限与静默时段、停止与转人工规则,以及每组 6,745 位观众的预注册转化率检验。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/seven-syllable-auditory-hijacking-short-video-livestreams-comment-section/c08/) ## 5DL-B-148 · Identity Ratchets in High-Ticket Communities, Variable-Reward Quest Loops and First-Week Epiphany Retention Engineering 中文书名:高客单社群身份棘轮、可变奖励闯关上瘾与首周顿悟留存工程 DOI: pending Cite: Fan, Y. (2026). Identity Ratchets in High-Ticket Communities, Variable-Reward Quest Loops and First-Week Epiphany Retention Engineering (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not legal advice and not personalised financial, investment or business advice. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Owners of high-ticket paid communities pass around a set of retention practices: first-week icebreaking and a moment of epiphany, quest loops with variable rewards, milestone unlocks, identity levels that only rise, and a renewal loop meant to hold people before they renew. This book does not teach anyone to hold members tighter. It writes each mechanism as a transparent model and weighs the size of its effect, the shape of its cost, the backfire of excess and the point where it must stop. Its parameters are illustrative; all calculations are generated by its programs; it is not legal advice and not personalised financial, investment or business advice. Chapter 1 estimates retention with the product-limit method; members with and without a first-week epiphany differ in six-month survival by 0.3069. It also sets out the central counter-intuitive proposition: in a heterogeneity model where nobody changes, the cohort's weekly churn hazard falls by 0.5052 within eight weeks, so a flattening curve does not mean members became loyal. Chapter 2 writes the first week as an absorbing Markov chain; the icebreak route reaches 3.50 times the day-seven epiphany rate of dumping everything at once, and a queueing formula sizes the hosts behind the first human reply. Chapter 3 shows that variable ratios make it harder to stop once rewards end, and that manufacturing surprise costs 2.0500 times the flat plan. Chapter 4 gives a front-loaded milestone schedule and shows that the median member covers only 0.2722 of the habit plateau in the first week. Chapter 5 shows that levels which only rise inflate, with one-year retention of 0.2905 against 0.3371 for levels earned by contribution, and weighs renewals held by identity. Chapter 6 computes the inverted U of push dosage and stop conditions for vulnerable members (a standard screen keeps 0.9282 of kept sales and removes 0.6002 of harm), and shows that "zero refunds" can only come from delivery: the refund channel stays open, and obstructing refunds nets only 801.9 against 962.5 with an open channel. Chapter 7 examines rewards and the warning against draining the pond in the Lüshi Chunqiu, and William James on habit; a single chance miss lowers automaticity by only 0.0151. Chapter 8 gives an owner's checklist, a customer-service closing kit and a pre-registered plan whose ninety-day retention test needs 1,016 members per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:高客单付费社群的经营者流传着一套留存做法:首周破冰与顿悟时刻,可变奖励的闯关任务,前后排布的里程碑解锁,只升不降的身份等级,以及在续费前把人留住的续约闭环。本书不教人把会员困得更紧,而是把这些机制逐一写成透明的模型,称出效果的大小、成本的形状、过量的反噬与必须停下的地方。书中参数为示意;全部计算由书中程序生成;本书不是法律意见,也不是个人化的财务、投资或经营建议。 第一章用乘积极限法估计留存,首周顿悟组与非顿悟组的半年生存率相差 0.3069;同时提出核心反常识命题:在无人改变的异质性模型中,群组周流失风险八周内下降 0.5052,变平的曲线不等于会员变忠诚。第二章把首周写成吸收马尔可夫链,破冰路线的第七天顿悟率是一次性交付的 3.50 倍,并用排队公式安排第一次真人回复的人手。第三章说明可变比率让人在奖励停止后更难停下,持续制造惊喜的账单是固定方案的 2.0500 倍。第四章给出前置的里程碑排程,并指出中位数会员在第一周只走完习惯平台的 0.2722。第五章显示只升不降的等级会通胀,一年留存 0.2905,凭贡献升级为 0.3371,并称量被身份困住的续费。 第六章计算推送剂量的倒 U、弱势会员的停止条件(标准筛查保留 0.9282 的有效成交,去掉 0.6002 的伤害),并说明「零退费」只能靠交付实现:退款渠道始终敞开,阻挠退款的净值只有 801.9,低于敞开渠道的 962.5。第七章考辨《吕氏春秋》的赏罚与竭泽之戒和威廉·詹姆斯论习惯,一次偶然错过只让自动化程度下降 0.0151。第八章给出经营者清单、客服促成成交技能包与预注册方案,九十天留存检验每组需要 1,016 人。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The First-Week Fork: Product-Limit Survival Curves, the Ruse of Heterogeneity and the Price of an Epiphany Rate — Chapter 1 estimates retention with the product-limit method; members with and without a first-week epiphany differ in six-month survival by 0.3069. It also sets out the central counter-intuitive proposition: in a heterogeneity model where nobody changes, the cohort's weekly churn hazard falls by 0.5052 within eight weeks, so a flattening curve does not mean members became loyal. (https://5dlibrary.com/answers/en/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c01/) 第1章 首周分叉:乘积极限生存曲线、异质性的假象与顿悟率的价格 — 第一章用乘积极限法估计留存,首周顿悟组与非顿悟组的半年生存率相差 0.3069;同时提出核心反常识命题:在无人改变的异质性模型中,群组周流失风险八周内下降 0.5052,变平的曲线不等于会员变忠诚。 (https://5dlibrary.com/answers/zh/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c01/) - Ch.2 The First-Week Icebreaking Route: Absorbing Markov Chains, Bottleneck Sensitivity and a Queueing Formula for the First Reply — Chapter 2 writes the first week as an absorbing Markov chain; the icebreak route reaches 3.50 times the day-seven epiphany rate of dumping everything at once, and a queueing formula sizes the hosts behind the first human reply. (https://5dlibrary.com/answers/en/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c02/) 第2章 首周破冰动线:吸收马尔可夫链、瓶颈灵敏度与第一次回复的排队公式 — 第二章把首周写成吸收马尔可夫链,破冰路线的第七天顿悟率是一次性交付的 3.50 倍,并用排队公式安排第一次真人回复的人手。 (https://5dlibrary.com/answers/zh/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c02/) - Ch.3 Reward Schedules for Quests: the Extinction Resistance of Variable Ratios, the Migration of the Reward-Prediction Error and the Escalating Bill for Surprise — Chapter 3 shows that variable ratios make it harder to stop once rewards end, and that manufacturing surprise costs 2.0500 times the flat plan. (https://5dlibrary.com/answers/en/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c03/) 第3章 闯关的奖励程式:变比率的消退抗性、奖赏预测误差的迁移与惊喜的升级账单 — 第三章说明可变比率让人在奖励停止后更难停下,持续制造惊喜的账单是固定方案的 2.0500 倍。 (https://5dlibrary.com/answers/zh/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c03/) - Ch.4 Scheduling Milestone Unlocks: the Goal Gradient, an Exhaustive Optimum that Front-Loads, and How Far an Epiphany Is from a Habit — Chapter 4 gives a front-loaded milestone schedule and shows that the median member covers only 0.2722 of the habit plateau in the first week. (https://5dlibrary.com/answers/en/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c04/) 第4章 里程碑解锁排程:目标梯度、前置的穷举最优与顿悟离习惯有多远 — 第四章给出前置的里程碑排程,并指出中位数会员在第一周只走完习惯平台的 0.2722。 (https://5dlibrary.com/answers/zh/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c04/) - Ch.5 The Identity Ratchet: Level Inflation, the Completion Condition of the IKEA Effect and Renewals Held by Identity — Chapter 5 shows that levels which only rise inflate, with one-year retention of 0.2905 against 0.3371 for levels earned by contribution, and weighs renewals held by identity. (https://5dlibrary.com/answers/en/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c05/) 第5章 身份棘轮:等级通胀、宜家效应的完成条件与被身份困住的续费 — 第五章显示只升不降的等级会通胀,一年留存 0.2905,凭贡献升级为 0.3371,并称量被身份困住的续费。 (https://5dlibrary.com/answers/zh/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c05/) - Ch.6 Excess, Stop Conditions and the Renewal Loop: the Inverted U of Push Dosage, Screening Vulnerable Members, and Low Refunds by Delivery rather than Obstruction — Chapter 6 computes the inverted U of push dosage and stop conditions for vulnerable members (a standard screen keeps 0.9282 of kept sales and removes 0.6002 of harm), and shows that "zero refunds" can only come from delivery: the refund channel stays open, and obstructing refunds nets only 801.9 against 962.5 with an open channel. (https://5dlibrary.com/answers/en/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c06/) 第6章 过量、停止条件与续约闭环:推送剂量的倒 U、弱势会员筛查与靠交付而非阻挠的低退款 — 第六章计算推送剂量的倒 U、弱势会员的停止条件(标准筛查保留 0.9282 的有效成交,去掉 0.6002 的伤害),并说明「零退费」只能靠交付实现:退款渠道始终敞开,阻挠退款的净值只有 801.9,低于敞开渠道的 962.5。 (https://5dlibrary.com/answers/zh/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c06/) - Ch.7 Sources Examined: the Handle of Reward and the Warning against Draining the Pond in the Lüshi Chunqiu, and the Fly-Wheel of Habit in William James's Principles of Psychology — Chapter 7 examines rewards and the warning against draining the pond in the Lüshi Chunqiu, and William James on habit; a single chance miss lowers automaticity by only 0.0151. (https://5dlibrary.com/answers/en/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c07/) 第7章 原典考辨:《吕氏春秋》的赏罚之柄与竭泽之戒,威廉·詹姆斯《心理学原理》的习惯飞轮 — 第七章考辨《吕氏春秋》的赏罚与竭泽之戒和威廉·詹姆斯论习惯,一次偶然错过只让自动化程度下降 0.0151。 (https://5dlibrary.com/answers/zh/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c07/) - Ch.8 Practice and Refutation: an Owner's Checklist, a Customer-Service Closing Kit and a Pre-Registered Test — Chapter 8 gives an owner's checklist, a customer-service closing kit and a pre-registered plan whose ninety-day retention test needs 1,016 members per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c08/) 第8章 实践与证伪:经营者清单、客服促成成交技能包与预注册检验方案 — 第八章给出经营者清单、客服促成成交技能包与预注册方案,九十天留存检验每组需要 1,016 人。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/identity-ratchets-high-ticket-communities-variable-reward-quest-loops/c08/) ## 5DL-B-149 · Information-Theoretic Rhetoric of Bestselling Monographs, Concrete Metaphor Mapping of Hard Formulas and the Chapter-End Suspense Engine 中文书名:畅销专著修辞信息论、硬核公式具象隐喻映射与章末悬念母机 DOI: pending Cite: Fan, Y. (2026). Information-Theoretic Rhetoric of Bestselling Monographs, Concrete Metaphor Mapping of Hard Formulas and the Chapter-End Suspense Engine (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not legal advice and not personalised business or financial advice. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: Can a scholarly monograph be read the way a novel is read, without a word of exaggeration? Publishing passes around many sayings: fewer equations sell better, readers want stories, every chapter should end on a hook, the more vivid the metaphor the better, a title must grab the eye. This book writes each piece of craft as a transparent model and uses programs that rely only on the Python standard library to weigh the size of its effect, the backfire of excess and the red lines that may not be crossed. Its parameters are illustrative; all calculations are generated by its programs; it is not legal advice and not personalised business or financial advice. Chapter 1 weighs three drafts with Shannon entropy: the story draft sits at 0.930 of the range from boilerplate to jargon, so it is easy because it is cheap to decode, not because it says less, and repeating a claim verbatim three times leaves 5.22 times the misses of restating it from fresh angles. Chapter 2 treats sentence length as a waveform; in a periodic paragraph the strongest period holds 0.829 of the power, and readability formulas give chopped-up text the top score of 111.8. Chapter 3 scores metaphors by structure mapping: a highway convoy reaches a consistency of 1.000 with the noisy-channel theorem, the telephone game only 0.143. Chapter 4 checks that a concrete sentence follows each formula, and two metaphors per formula give the best net effect, 0.7552. Chapter 5 is the chapter-end suspense engine: the pull of an information gap peaks with 4 candidates left (0.7358), and early, spaced hooks bring the expected chapters read to 4.6318. The central counter-intuitive proposition is that inflated hooks add 0.4926 chapters read while raising the share of readers who meet two broken promises from 0.0156 to 0.6517. Chapter 6 discusses sharing in light of evidence on valence and arousal; wording inflated to "changes everything" opens 2.50 times as often as accurate wording yet is worth -0.16491 per impression. Chapter 7 examines the fable, weighty word and goblet word of the Zhuangzi and reversal in Aristotle's Poetics; the product of surprise and coherence peaks with six planted clues (1.6917). Chapter 8 gives a monograph copy checker, an editorial checklist, limits on how authors contact readers and a pre-registered trial whose completion test needs 1,094 readers per group. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:学术专著能不能像小说一样让人读下去,同时一个字也不夸大?出版界流传着许多说法:公式越少越好卖,读者爱听故事,每章结尾要留钩子,比喻越生动越好,标题要抓眼球。本书把这些手艺逐一写成透明的模型,用只依赖 Python 标准库的程序称出效果的大小、过量的反效果与不可逾越的红线。书中参数为示意;全部计算由书中程序生成;本书不是法律意见,也不是个人化的商业或财务建议。 第一章用香农熵称量三份草稿,故事草稿的词熵落在套话到术语区间的 0.930 处,好读来自解码成本低而不是信息少;原句照抄三遍留下的漏读是换角度重述的 5.22 倍。第二章把句长当作波形,有周期的段落最强周期独占 0.829 的能量,并说明可读性公式会给切碎的文字打出 111.8 的最高分。第三章用结构映射给比喻打分,公路车队对噪声信道定理的一致度是 1.000,传话游戏只有 0.143。第四章检查公式之后有没有具象句,每个公式配两个比喻时净效果 0.7552 最高。第五章是章末悬念母机:信息缺口的拉力在剩下 4 个候选时最高(0.7358),早而分散的钩子让期望读完章数达到 4.6318;核心反常识命题是,夸大的钩子让读完章数多 0.4926 章,却让经历两次落空承诺的读者从 0.0156 升到 0.6517。 第六章依据效价与唤醒的证据讨论传播,夸大到「改变一切」的措辞打开率是准确措辞的 2.50 倍,每次展示的价值却是 -0.16491。第七章考辨《庄子》的寓言、重言、卮言与亚里士多德《诗学》的突转,惊讶与连贯的乘积在埋下六条线索时最高(1.6917)。第八章给出专著文案检查器、编辑部清单、作者与读者沟通的边界与预注册试验方案,完读率检验每组需要 1,094 名读者。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Chapter Entropy: the Surprise Budget, Restatement from Fresh Angles and a Grammar of Rhetorical Moves — Chapter 1 weighs three drafts with Shannon entropy: the story draft sits at 0.930 of the range from boilerplate to jargon, so it is easy because it is cheap to decode, not because it says less, and repeating a claim verbatim three times leaves 5.22 times the misses of restating it from fresh angles. (https://5dlibrary.com/answers/en/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c01/) 第1章 章节信息熵:惊讶度预算、换角度重述与修辞动作的语法 — 第一章用香农熵称量三份草稿,故事草稿的词熵落在套话到术语区间的 0.930 处,好读来自解码成本低而不是信息少;原句照抄三遍留下的漏读是换角度重述的 5.22 倍。 (https://5dlibrary.com/answers/zh/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c01/) - Ch.2 The Waveform of Sentence Length: Coefficient of Variation, Spectral Period and the Blind Spot of Readability Formulas — Chapter 2 treats sentence length as a waveform; in a periodic paragraph the strongest period holds 0.829 of the power, and readability formulas give chopped-up text the top score of 111.8. (https://5dlibrary.com/answers/en/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c02/) 第2章 句长节奏波形:变异系数、频谱周期与可读性公式的盲区 — 第二章把句长当作波形,有周期的段落最强周期独占 0.829 的能量,并说明可读性公式会给切碎的文字打出 111.8 的最高分。 (https://5dlibrary.com/answers/zh/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c02/) - Ch.3 The Structure Mapping of Metaphor: Relational Consistency, Systematicity and a Permutation Test for Lucky Fit — Chapter 3 scores metaphors by structure mapping: a highway convoy reaches a consistency of 1.000 with the noisy-channel theorem, the telephone game only 0.143. (https://5dlibrary.com/answers/en/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c03/) 第3章 隐喻的结构映射:关系一致度、系统性与偶然拟合的置换检验 — 第三章用结构映射给比喻打分,公路车队对噪声信道定理的一致度是 1.000,传话游戏只有 0.143。 (https://5dlibrary.com/answers/zh/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c03/) - Ch.4 Landing the Formula: the Share of Concrete Sentences, the Dose of Metaphor and the Diversity of Source Domains — Chapter 4 checks that a concrete sentence follows each formula, and two metaphors per formula give the best net effect, 0.7552. (https://5dlibrary.com/answers/en/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c04/) 第4章 公式落地:具象句比例、隐喻剂量与源域多样性 — 第四章检查公式之后有没有具象句,每个公式配两个比喻时净效果 0.7552 最高。 (https://5dlibrary.com/answers/zh/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c04/) - Ch.5 The Chapter-End Suspense Engine: the Inverted U of the Information Gap, the Placement of Hooks and the Beta Update of Trust — Chapter 5 is the chapter-end suspense engine: the pull of an information gap peaks with 4 candidates left (0.7358), and early, spaced hooks bring the expected chapters read to 4.6318. The central counter-intuitive proposition is that inflated hooks add 0.4926 chapters read while raising the share of readers who meet two broken promises from 0.0156 to 0.6517. (https://5dlibrary.com/answers/en/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c05/) 第5章 章末悬念母机:信息缺口的倒 U、钩子排布与信任的贝塔更新 — 第五章是章末悬念母机:信息缺口的拉力在剩下 4 个候选时最高(0.7358),早而分散的钩子让期望读完章数达到 4.6318;核心反常识命题是,夸大的钩子让读完章数多 0.4926 章,却让经历两次落空承诺的读者从 0.0156 升到 0.6517。 (https://5dlibrary.com/answers/zh/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c05/) - Ch.6 Arousal and Sharing: Emotional Valence, Branching Processes and the Backfire of Overclaiming — Chapter 6 discusses sharing in light of evidence on valence and arousal; wording inflated to "changes everything" opens 2.50 times as often as accurate wording yet is worth -0.16491 per impression. (https://5dlibrary.com/answers/en/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c06/) 第6章 唤醒与传播:情绪效价、分支过程与夸大措辞的反噬 — 第六章依据效价与唤醒的证据讨论传播,夸大到「改变一切」的措辞打开率是准确措辞的 2.50 倍,每次展示的价值却是 -0.16491。 (https://5dlibrary.com/answers/zh/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c06/) - Ch.7 Sources Examined: Fable, Weighty Word and Goblet Word in the Zhuangzi, and Reversal and Recognition in Aristotle's Poetics — Chapter 7 examines the fable, weighty word and goblet word of the Zhuangzi and reversal in Aristotle's Poetics; the product of surprise and coherence peaks with six planted clues (1.6917). (https://5dlibrary.com/answers/en/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c07/) 第7章 原典考辨:《庄子》的寓言、重言与卮言,亚里士多德《诗学》的突转与发现 — 第七章考辨《庄子》的寓言、重言、卮言与亚里士多德《诗学》的突转,惊讶与连贯的乘积在埋下六条线索时最高(1.6917)。 (https://5dlibrary.com/answers/zh/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c07/) - Ch.8 Practice and a Plan for Falsification: the Monograph Copy Checker, an Editorial Workflow and a Pre-Registered Reader Trial — Chapter 8 gives a monograph copy checker, an editorial checklist, limits on how authors contact readers and a pre-registered trial whose completion test needs 1,094 readers per group. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c08/) 第8章 实践与证伪方案:专著文案检查器、编辑部流程与预注册读者试验 — 第八章给出专著文案检查器、编辑部清单、作者与读者沟通的边界与预注册试验方案,完读率检验每组需要 1,094 名读者。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/information-theoretic-rhetoric-bestselling-monographs-concrete-metaphor-mapping-hard/c08/) ## 5DL-B-150 · Lossless Distillation of Monographs into Agent Skill Packs, Executable Rule Compilation and Self-Evolving Systems 中文书名:专著向智能体技能包无损蒸馏、规则可执行化编译与系统自进化工程 DOI: pending Cite: Fan, Y. (2026). Lossless Distillation of Monographs into Agent Skill Packs, Executable Rule Compilation and Self-Evolving Systems (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All samples are invented; pricing and closing rules are written in general terms and are not legal or compliance advice. Abstract: What an agent can actually execute from a monograph is its decision rules, thresholds, prohibitions and reproducible calculations. This volume defines one protocol for distilling them into a skill pack: an instruction file, validator scripts, a decision table and golden test samples, plus a manifest that pins every rule to its page in the source and to a content fingerprint. "Lossless" is defined as four countable ratios: rule coverage, executable rate, golden-sample regression pass rate and refusal completeness; the release gate requires the last two to equal 1 and effective retention to be no lower than 0.90. Every claim is graded by evidence, the Xunzi chapter "An Exhortation to Learning" is examined on tools and accumulation, every calculation is produced by the book's programs, and the author's judgements are placed in hypothesis boxes that state what would refute them. All samples are invented; pricing and closing rules are written in general terms and are not legal or compliance advice. Chapter 1 establishes two counter-intuitive results: pasting a whole book into the prompt gives coverage 1.000 but effective retention 0.000; with twelve samples all passing, the 95% upper bound on the failure rate is still 0.2209, and bringing it below one percent takes 299; under stated assumptions, a twelve-rule pack without samples loses at least one rule within ten edits with probability 0.9115. Chapters 2 to 7 cover extraction, compilation, guardrails and validators, the uniform specification of the four kits, self-evolution and governance. Chapter 8 distils an invented twelve-rule pricing book into a pricing calculator: a signal-word scan reaches precision 0.714 and recall 0.833 and misses exactly the definition of margin and one refusal condition; one casual edit drops the pass rate to 0.788 and refusal completeness to 0.833 while coverage and executable rate stay at 1; the pre-registered three-arm comparison is set at 228 questions per arm, 684 answers in all, with an upper bound of 0.0131 if no violation is seen. 摘要:一本专著里真正能被智能体执行的,是它的决策规则、阈值、禁忌与可复现计算。本卷定义一份统一协议,把这些内容蒸馏成技能包:说明文件、校验脚本、决策表与黄金测试样本,外加一份把每条规则钉回原书页码与内容指纹的清单。「无损」被定义为四个可以数出来的比率:规则覆盖率、可执行率、黄金样本回归通过率与拒绝条件完整率;发布门槛是后两者等于 1、有效保留率不低于 0.90。全书按证据分级标注每一条说法,以《荀子·劝学》考辨工具与积累,所有计算由书中程序生成,作者的判断一律写进假说框并写明什么结果会推翻它。书中样本全部虚构,定价与成交规则只写一般框架,不是法律或合规意见。 第一章证明两个反常识结论:把整本书贴进提示词,覆盖率是 1.000,有效保留率却是 0.000;十二个样本全部通过时,失败率的 95% 上限仍有 0.2209,要压到百分之一以下需要 299 个;按写明的假设,没有样本的十二条规则包十轮修改之内至少坏一条的概率是 0.9115。第二至第七章依次写抽取、编译、护栏与校验、四件套统一规格、自进化与治理。第八章把一本虚构的十二条规则定价书蒸馏成定价试算器:信号词扫描的精确率 0.714、召回率 0.833,漏掉的正是毛利的定义与一条拒绝条件;一次顺手的修改让回归通过率掉到 0.788、拒绝条件完整率掉到 0.833,覆盖率与可执行率却仍是 1;预先登记的三臂比较定为每臂 228 题、共 684 个回答,零违规时的上限是 0.0131。 Chapter answers: - Ch.1 Method: Skill Packs, Four Retention Indicators, Evidence Grades and the Four Steps of Distillation — Chapter 1 establishes two counter-intuitive results: pasting a whole book into the prompt gives coverage 1.000 but effective retention 0.000; with twelve samples all passing, the 95% upper bound on the failure rate is still 0.2209, and bringing it below one percent takes 299; under stated assumptions, a twelve-rule pack without samples loses at least one rule within ten edits with probability 0.9115. (https://5dlibrary.com/answers/en/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c01/) 第1章 方法:技能包、四个保留指标、证据分级与蒸馏四步 — 第一章证明两个反常识结论:把整本书贴进提示词,覆盖率是 1.000,有效保留率却是 0.000;十二个样本全部通过时,失败率的 95% 上限仍有 0.2209,要压到百分之一以下需要 299 个;按写明的假设,没有样本的十二条规则包十轮修改之内至少坏一条的概率是 0.9115。第二至第七章依次写抽取、编译、护栏与校验、四件套统一规格、自进化与治理。 (https://5dlibrary.com/answers/zh/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c01/) - Ch.2 Extraction: Picking the Executable Sentences out of a Manuscript One by One — Rule sentences come in eight patterns: condition–action, threshold, prohibition, ordering, exception, formula, refusal and definition; the annotation box is a ninth signal. On the 30-sentence fictional quotation handbook, the box-aware cue extractor reaches precision 0.8966 and recall 0.8387 in Chinese, precision 0.8333 and recall 0.9677 in English; box signals remove false positives at no cost in recall, and the bilingual union misses only one implicit conditional that carries no cue at all. Chinese quantity words must be converted first, and those that cannot be converted must be refused. Conclusions about recall are bounded by the number of gold rules: a target of 0.90 with no misses needs at least 35. Kappa between two annotators is driven by the share of rule sentences, so compare that share before comparing manuscripts. The product of extraction is an extraction sheet with every field filled, and Chapter 3 starts from that sheet to compile rule sentences into executable predicates. (https://5dlibrary.com/answers/en/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c02/) 第2章 抽取:把书稿里能执行的句子一句一句拣出来 — 规则句有八种句型:条件—动作、阈值、禁忌、次序、例外、公式、拒绝、定义,标注框是第九种信号。在 30 句的虚构报价手册上,识框的提示词抽取器中文精确率 0.8966、召回率 0.8387,英文精确率 0.8333、召回率 0.9677;标注框信号不花召回率就拿掉了误报,中英并集只漏掉一句不带任何提示词的隐性条件。中文数量词要先换算,换算不了的要拒绝。召回率的结论受金标准条数限制:目标 0.90 而一条不漏,至少要 35 条金标准。两位标注者的 kappa 受规则句比例左右,比较不同书稿时要先看比例。抽取的产物是一张逐栏填满的抽取单,第三章从这张单子出发,把规则句编译成可以执行的谓词。 (https://5dlibrary.com/answers/zh/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c02/) - Ch.3 Compilation: Turning Rule Sentences into Predicates, Decision Tables and Decision Trees — Compilation turns the extraction sheet into three runnable forms: atomic predicates, decision tables and decision trees. Comparison words decide open and closed; three writers' wordings of one deposit rule come out complete, conflicting and gapped at the single point of 8000 yuan, so this book accepts only left-closed, right-open bands and gives every threshold three boundary samples. Units and floating point must be pinned down first: read 0.3 as 30 and the rush fee jumps from 2040 yuan to 204000 yuan; with a margin floor of 0.2 the wrong formula rejects good orders, while a floor of 0.35 happens to hide the defect. The handbook's five rules expanded over 32 cells show 12 conflicts and 1 gap; a first-match precedence order plus a refusal branch settles them all, and the report states how many cells each rule loses to shadowing. Information gain is fooled into a perfect score by an identifier field and gain ratio does not stop it, so identifiers are excluded by schema before any tree is grown; a decision tree reproduces rules from cases but copies one rule into several branches, so rules stated in a book are always compiled into a decision table. The compile report and the boundary golden samples pass to Chapter 4, where they become the validation gates before release. (https://5dlibrary.com/answers/en/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c03/) 第3章 编译:把规则句变成谓词、决策表与决策树 — 编译把抽取单变成三种可运行的形式:原子谓词、决策表、决策树。比较词决定开闭,三位写手对同一条定金规则的三种写法,在 8000 元这一个点上分别是完备、冲突与缺口;本书只接受左闭右开的切法,并为每个阈值配三个边界样本。单位与浮点要先钉死:0.3 读成 30,加急费就从 2040 元变成 204000 元;底线 0.2 的毛利率判断,用错一种写法就会误拒,而底线 0.35 恰好看不出来。手册五条规则展开到 32 格上,有 12 格冲突、1 格缺口,首条命中的优先次序加一个拒绝分支把它们全部收住,同时报告每条规则被遮蔽了多少格。信息增益会被标识类字段骗到满分,增益比也拦不住,只能在建树之前按模式排除;决策树能从案例里复现规则,却会把一条规则复制到多个分支,所以书里写明的规则一律编译成决策表。编译报告与边界黄金样本交给第四章,变成发布前的校验门槛。 (https://5dlibrary.com/answers/zh/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c03/) - Ch.4 Guardrails and Validators: Banned Phrases, Required Sentences, Refusal Conditions, Lint and Golden Samples — Guardrails come in three sentence types, banned phrases, required sentences and refusal conditions, each paired with replacement wording, trigger predicates and fixed replies; outside text is always data. Lint has a word half and a structural half. On 80 drafts the best word matcher reaches precision 0.897 and recall 0.963, while the structural check is exact when fields are complete, so any rule that can be expressed as fields is not checked with a word list. The unit costs of false alarms and misses set the threshold, which equals $C_{FP}/(C_{FP}+C_{FN})$ and must be recomputed whenever the base rate moves; otherwise every thousand drafts pay the gap between 6231.4 and 4404.2. Golden samples belong on the boundaries: 18 boundary samples reach a mutation score of 1.000 where 200 random samples reach only 0.767. A clean run only gives a bound, and only 299 clean samples justify saying the error rate is below 0.01. The next chapter builds these checks into the unified specification of the four kits. (https://5dlibrary.com/answers/en/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c04/) 第4章 护栏与校验器:禁词、合规句、拒答条件、lint 与黄金样本 — 护栏有三种句型:禁词、合规句、拒答条件,各配替代说法、触发谓词与固定回复;外来文本一律当数据。lint 分词表与结构两半,词表在 80 份草稿上最好的一档精确率 0.897、召回率 0.963,结构检查在字段齐全时是精确的,所以能写成字段的规则就不用词表。误报与漏报的单价决定门槛,门槛等于 $C_{FP}/(C_{FP}+C_{FN})$,基础率一变就要重算,否则每千份草稿多付 6231.4 与 4404.2 之间的差额。黄金样本要取在边界上:18 个边界样本的变异得分是 1.000,200 个随机样本只有 0.767;全部通过只给出上界,299 个样本零失败才能说错误率低于 0.01。下一章把这些检查装进四件套的统一规格里。 (https://5dlibrary.com/answers/zh/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c04/) - Ch.5 One Specification for the Four Kits: Pricing Calculator, Front-End Style Library, Persuasive-Copy Checker and Customer-Service Closing Pack — The four kits share a directory tree of 10 paths and an instruction-file header of 15 fields, and release follows the gate of Chapter 1: effective retention of at least 0.90, a regression pass rate and refusal completeness of exactly 1, and a golden sample that triggers every refusal condition. The pricing calculator solves for price in exact fractions, rounds up to the lattice and refuses with four codes, E1 to E4. The front-end style library guards semantic tokens with the contrast and tap-target sizes of WCAG 2.1, and does not pass even a near miss such as the alert red at 4.43:1. The persuasive-copy checker asks only whether each sentence can be checked against data, and returns invented scarcity and invented deadlines. The customer-service closing pack follows up at 24, 72 and 168 hours, holds messages through quiet hours, sends at most 3, and stops at no. There are 5 edges among the four kits, each with a contract; version numbers are computed from contract and golden-sample diffs, and a consumer pinned to a major version reruns its own samples whenever that major version moves. Chapter 6 goes on to show how these skill packs keep from decaying through round after round of edits. (https://5dlibrary.com/answers/en/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c05/) 第5章 四件套统一规格:定价试算器、前端样式库、攻心文案检查器与客服逼单技能包 — 四件套共用一棵 10 个路径的目录树与一个 15 个字段的说明文件头部,放行沿用第一章的门槛:有效保留率不低于 0.90,回归通过率与拒绝条件完整率都是 1,且每个拒绝条件都要有触发它的黄金样本。定价试算器用精确分数解出价格、向上取格点,以 E1 到 E4 四个代码拒绝;前端样式库用语义令牌和 WCAG 2.1 的对比度与点按尺寸把关,警示红 4.43:1 这样差一点的颜色也不放行;攻心文案检查器只问每一句能不能被数据核实,虚构稀缺与虚构期限都被退回;客服逼单技能包按 24、72、168 小时跟进、静默时段顺延、最多 3 条、听到不就停。四件之间有 5 条边,每条边一份契约,版本号由契约差异与黄金样本差异算出,钉住主版本的消费方在主版本跳动时先重跑自己的样本。第六章接着写这些技能包在一轮轮修改之下怎样保持不退化。 (https://5dlibrary.com/answers/zh/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c05/) - Ch.6 Self-Evolution: Feedback Loops, Regression Gates, Drift and the Accumulation of a Skill Library — A skill pack with no golden samples decays quietly under frequent editing: in the p60 model only 0.7080 of its rules are intact after 30 rounds, and repair by complaint can hold it only at a floor of 0.6757, while a pack with nine rules in ten under sample guard keeps 0.9708 at the same point. The regression gate ties version numbers, sample results and coverage together and blocked four of ten candidate releases. Drift has to be watched per day and per conversation, and the cumulative sum raises the alarm about five days earlier than a weekly mean. Making the visible pass rate the target produces a pack with a perfect score that has got worse; a skill library becomes a mountain through an admission gate, not by heaping. Self-critique pulls a pack toward its own fixed point, and only external samples can raise that point. The next chapter, on governance, answers the other half of the question: who may change a pack, how a change is traced, and when a rule should be retired. (https://5dlibrary.com/answers/en/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c06/) 第6章 自进化:反馈回路、回归门、漂移与技能库的累积 — 一份没有黄金样本的技能包,在频繁修改下会悄悄退化:p60 的模型里,30 轮之后只剩 0.7080 的规则完好,靠投诉来修只能把它托在 0.6757 的地板上;九成规则有样本守着的包,同一时刻保留 0.9708。回归门把版本号、样本通过与覆盖率三件事绑在一起,十次候选发布挡下了四次。漂移要按天、按对话来监测,累积和比周平均早约五天报警。把可见样本通过率当成目标,会得到一个满分却变坏的包;技能库要靠入库门积成山,而不是堆成堆。自我批评会把技能包拉向它自己的固定点,只有外部样本能把这个固定点抬高。下一章讨论的治理,要回答的是另一半问题:谁有权改、改了以后怎样追溯,以及什么时候该让一条规则退场。 (https://5dlibrary.com/answers/zh/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c06/) - Ch.7 Governance: Least Privilege, Machine-Checked Red Lines, Disclaimers, Provenance and Rule Retirement — R04, the reply template for price objections, fired 96 times and was overruled 31 times; its interval runs from 0.2378 to 0.4217 and the lower bound is above 0.20, so the template is probably wrong and is retired for a rewrite. R05 fired 18 times with 3 overrides; the rate of 0.1667 looks modest, but the upper bound reaches 0.3922, so it cannot be cleared yet and goes to review. R07 fired only 12 times with 1 override but drew a red-line complaint, and is suspended at once. R03 and R11 do not have traced anchors and are suspended. R08, the guard on stock lines, did not fire once in ninety days and is kept. S13, a birthday-offer rule from an earlier version, did not fire in ninety days and is retired to the archive. (https://5dlibrary.com/answers/en/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c07/) 第7章 治理:最小权限、红线机检、免责、来源追溯与规则下架 — R04,即价格异议的回复模板,触发 96 次被推翻 31 次,区间是 0.2378 到 0.4217,下界已经超过 0.20:这条模板多半是错的,下架重写。R05 触发 18 次推翻 3 次,比率 0.1667 看上去不高,区间上界却到 0.3922,还不能放心,进入复核。R07 只有 12 次触发、1 次推翻,但有一次红线投诉,直接暂停。R03 与 R11 的锚点不是已追溯,暂停。R08 是库存提示的护栏,九十天一次没触发,保留。S13 是旧版本里的生日优惠规则,九十天没触发,下架归档。 (https://5dlibrary.com/answers/zh/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c07/) - Ch.8 Practice: Distilling a Twelve-Rule Pricing Book into a Pricing Calculator — Chapter 8 distils an invented twelve-rule pricing book into a pricing calculator: a signal-word scan reaches precision 0.714 and recall 0.833 and misses exactly the definition of margin and one refusal condition; one casual edit drops the pass rate to 0.788 and refusal completeness to 0.833 while coverage and executable rate stay at 1; the pre-registered three-arm comparison is set at 228 questions per arm, 684 answers in all, with an upper bound of 0.0131 if no violation is seen. (https://5dlibrary.com/answers/en/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c08/) 第8章 实战:把一本十二条规则的定价小书蒸馏成定价试算器 — 第八章把一本虚构的十二条规则定价书蒸馏成定价试算器:信号词扫描的精确率 0.714、召回率 0.833,漏掉的正是毛利的定义与一条拒绝条件;一次顺手的修改让回归通过率掉到 0.788、拒绝条件完整率掉到 0.833,覆盖率与可执行率却仍是 1;预先登记的三臂比较定为每臂 228 题、共 684 个回答,零违规时的上限是 0.0131。 (https://5dlibrary.com/answers/zh/lossless-distillation-monographs-agent-skill-packs-executable-rule-compilation/c08/) ## 5DL-B-151 · Content Fission Algebra: Compiling Hard-Core Monographs into Two-Host Podcasts, Study Slides and Quiz Banks 中文书名:一书多吃内容裂变代数:从硬核专著自动编译双人播客、导学投影片与通关题库 DOI: pending Cite: Fan, Y. (2026). Content Fission Algebra: Compiling Hard-Core Monographs into Two-Host Podcasts, Study Slides and Quiz Banks (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not legal advice and not personalised business or investment advice. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: Publishing and knowledge services have embraced the practice of turning one hard-core monograph into many products: a two-host podcast, study slides and knowledge cards, and a pass-or-retry quiz bank. Every adaptation is a compression. This book does not teach anyone to chop books finer. It writes each compression as a transparent model, computes what readers, listeners and learners actually carry away, and states the backfire of excess and the red lines that must not be crossed. Its parameters are illustrative; all calculations are generated by its programs; it is not legal advice and not personalised business or investment advice. Chapter 1 writes sentence selection as submodular coverage maximisation, where greedy selection carries a worst-case guarantee of 0.6321 and salience ranking piles its places onto one topic. Chapter 2 shows that a thirty-minute podcast carrying about 10 arguments keeps 0.4820, while stating 20 arguments once each keeps only 0.3117, and that the best uninterrupted turn of one host lasts about 41.5 seconds. Chapter 3 finds that 3 points per slide keep 4.0909 points a minute and that an evenly spread card takes away most at 40 characters, 0.7113. Chapter 4 gives quiz items a check-up, assembles tests around the pass line, and states the core counter-intuitive proposition: through the testing effect, a tested group without feedback still leads after a week with 0.3004 against 0.0253 for restudy. Chapter 5 shows that the best review gap grows with the retention interval and that pausing new items during a backlog lets more cards graduate. Chapter 6 finds the best mix of podcast, slides and quizzes within 40 production hours, worth 2.0685 against 1.4164 without quizzes; fact checks belong upstream; 2 reminders a week give the most expected study sessions, 6.2792, while excess reminders harm, with a lower threshold for fragile learners; only books one has the right to adapt may be adapted, and synthetic content must be labelled. Chapter 7 examines Zhu Xi on following the order and reading thoroughly and Bacon's Of Studies: only 7 of 5040 orders of presentation respect the dependencies, and close reading of about 8 passes per passage yields the most restatable passages, 2.3396. Chapter 8 gives an owner's checklist, the distilled front-end style library (body-text contrast 15.80) and a pre-registered plan. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:出版与知识服务行业流行「一书多吃」:把一本硬核专著改编成双人播客、导学投影片与知识卡片、通关题库。每一次改编都是一次压缩。本书不教人把书拆得更碎,而是把每一种压缩写成透明的模型,计算读者、听众与学习者真正带走的信息,并给出过量的反效果与不可逾越的红线。书中参数为示意;全部计算由书中程序生成;本书不是法律意见,也不是个人化的经营或投资建议。 第一章把选句写成子模覆盖的最大化,贪心算法有 0.6321 的最坏保证,而显著度排序会把名额集中到同一个话题。第二章显示三十分钟的播客只讲约 10 个论点时保留 0.4820,把 20 个论点各讲一遍只保留 0.3117,单人连续讲述的最优时长约 41.5 秒。第三章给出每页 3 条要点时每分钟留住 4.0909 条,平铺排版的卡片在 40 字时带走最多,为 0.7113。第四章给题目做体检、围绕通关线组卷,并提出核心反常识命题:测试效应使不给反馈的测验组一周后仍以 0.3004 领先反复学习组的 0.0253。第五章说明最佳复习间隔随保持期增长,复习积压时暂停新题能让更多卡片毕业。 第六章在 40 个制作工时下求出播客、投影片与题库的最优组合,价值 2.0685,不做题库只有 1.4164;事实核对应放在上游;每周 2 次提醒使期望学习次数最高,为 6.2792,过量提醒有害,对脆弱学习者阈值更低;只改编有权改编的书,合成内容必须标识。第七章考辨朱熹「循序渐进、熟读精思」与培根《论读书》,在 5040 种呈现顺序中只有 7 种遵守依赖关系,每篇约 8 遍的精读得到最多能复述的篇目 2.3396。第八章给出经营者清单、前端样式库蒸馏件(正文对比度 15.80)与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Submodular Coverage: the Greedy Algebra of Picking Sentences and the 1−1/e Guarantee — Chapter 1 writes sentence selection as submodular coverage maximisation, where greedy selection carries a worst-case guarantee of 0.6321 and salience ranking piles its places onto one topic. (https://5dlibrary.com/answers/en/content-fission-algebra-compiling-hard-core-monographs-two-host/c01/) 第1章 子模覆盖:从专著里选句的贪心代数与 1−1/e 的保证 — 第一章把选句写成子模覆盖的最大化,贪心算法有 0.6321 的最坏保证,而显著度排序会把名额集中到同一个话题。 (https://5dlibrary.com/answers/zh/content-fission-algebra-compiling-hard-core-monographs-two-host/c01/) - Ch.2 Spoken Compression: Breadth against Repetition, Turn Length and Shannon Naming in a Two-Host Podcast — Chapter 2 shows that a thirty-minute podcast carrying about 10 arguments keeps 0.4820, while stating 20 arguments once each keeps only 0.3117, and that the best uninterrupted turn of one host lasts about 41.5 seconds. (https://5dlibrary.com/answers/en/content-fission-algebra-compiling-hard-core-monographs-two-host/c02/) 第2章 口语化压缩:双人播客的广度与重复、换手时长与香农命名 — 第二章显示三十分钟的播客只讲约 10 个论点时保留 0.4820,把 20 个论点各讲一遍只保留 0.3117,单人连续讲述的最优时长约 41.5 秒。 (https://5dlibrary.com/answers/zh/content-fission-algebra-compiling-hard-core-monographs-two-host/c02/) - Ch.3 Visual Card Compression: the Item Ceiling of a Slide, Dynamic Programming over Dependencies, and the Length of a Card — Chapter 3 finds that 3 points per slide keep 4.0909 points a minute and that an evenly spread card takes away most at 40 characters, 0.7113. (https://5dlibrary.com/answers/en/content-fission-algebra-compiling-hard-core-monographs-two-host/c03/) 第3章 视觉卡片压缩:投影片每页信息量上限、按依赖切页的动态规划与卡片字数 — 第三章给出每页 3 条要点时每分钟留住 4.0909 条,平铺排版的卡片在 40 字时带走最多,为 0.7113。 (https://5dlibrary.com/answers/zh/content-fission-algebra-compiling-hard-core-monographs-two-host/c03/) - Ch.4 The Pass-or-Retry Quiz Bank: Difficulty and Discrimination, Test Information at the Cut, and the Counter-Intuitive Testing Effect — Chapter 4 gives quiz items a check-up, assembles tests around the pass line, and states the core counter-intuitive proposition: through the testing effect, a tested group without feedback still leads after a week with 0.3004 against 0.0253 for restudy. (https://5dlibrary.com/answers/en/content-fission-algebra-compiling-hard-core-monographs-two-host/c04/) 第4章 通关题库:难度与区分度、围绕通关线的测验信息与测试效应的反直觉 — 第四章给题目做体检、围绕通关线组卷,并提出核心反常识命题:测试效应使不给反馈的测验组一周后仍以 0.3004 领先反复学习组的 0.0253。 (https://5dlibrary.com/answers/zh/content-fission-algebra-compiling-hard-core-monographs-two-host/c04/) - Ch.5 Spaced Review Scheduling: Optimal Gaps that Grow with the Retention Interval, the Leitner Boxes as an Absorbing Chain, and a Daily Review Cap — The timid schedule has the highest mean pass rate, 0.94, and needs only 5.7856 reviews per card, or 115.7 a day; but at graduation it has shown only that a card can be recalled after 5 days, so its reviews per proven day are 1.1571, the least efficient. The doubling schedule passes 0.90 on average, with 6.9351 reviews per card, 138.7 a day and 0.4334 per proven day. The bold schedule's pass rate falls to 0.82, with 9.5008 reviews per card, 190.0 a day and 0.2969 per proven day. The reckless schedule passes only 0.69, needing 17.1856 reviews per card, 12.1856 of them wrong answers, and 343.7 a day. (https://5dlibrary.com/answers/en/content-fission-algebra-compiling-hard-core-monographs-two-host/c05/) 第5章 间隔复习排程:最佳间隔随保持期增长、莱特纳盒子的吸收链与每日复习上限 — 保守排程的平均答对率最高,0.94,每张卡片只需 5.7856 次复习,每天 115.7 次;但它毕业时只证明了卡片能在 5 天后被回忆,按「每证明一天的复习次数」算是 1.1571,效率最低。加倍排程的平均答对率 0.90,每张卡 6.9351 次复习,每天 138.7 次,每证明一天 0.4334 次。激进排程答对率降到 0.82,每张卡 9.5008 次,每天 190.0 次,每证明一天 0.2969 次。鲁莽排程答对率只有 0.69,每张卡 17.1856 次复习,其中 12.1856 次是答错,每天 343.7 次。 (https://5dlibrary.com/answers/zh/content-fission-algebra-compiling-hard-core-monographs-two-host/c05/) - Ch.6 The Portfolio Algebra of One Book, Many Products: Production Hours, Fidelity Audits, the Backfire of Excess and the Copyright Red Line — Chapter 6 finds the best mix of podcast, slides and quizzes within 40 production hours, worth 2.0685 against 1.4164 without quizzes; fact checks belong upstream; 2 reminders a week give the most expected study sessions, 6.2792, while excess reminders harm, with a lower threshold for fragile learners; only books one has the right to adapt may be adapted, and synthetic content must be labelled. (https://5dlibrary.com/answers/en/content-fission-algebra-compiling-hard-core-monographs-two-host/c06/) 第6章 一书多吃的组合代数:制作工时分配、事实保真审计、过量的反效果与版权红线 — 第六章在 40 个制作工时下求出播客、投影片与题库的最优组合,价值 2.0685,不做题库只有 1.4164;事实核对应放在上游;每周 2 次提醒使期望学习次数最高,为 6.2792,过量提醒有害,对脆弱学习者阈值更低;只改编有权改编的书,合成内容必须标识。 (https://5dlibrary.com/answers/zh/content-fission-algebra-compiling-hard-core-monographs-two-host/c06/) - Ch.7 The Classics Examined: Zhu Xi on Following the Order and Reading Thoroughly, and Bacon on Books to be Tasted, Swallowed and Chewed — Chapter 7 examines Zhu Xi on following the order and reading thoroughly and Bacon's Of Studies: only 7 of 5040 orders of presentation respect the dependencies, and close reading of about 8 passes per passage yields the most restatable passages, 2.3396. (https://5dlibrary.com/answers/en/content-fission-algebra-compiling-hard-core-monographs-two-host/c07/) 第7章 原典考辨:朱熹「循序渐进、熟读精思」的读书法与培根《论读书》的「浅尝、吞咽与咀嚼」 — 第七章考辨朱熹「循序渐进、熟读精思」与培根《论读书》,在 5040 种呈现顺序中只有 7 种遵守依赖关系,每篇约 8 遍的精读得到最多能复述的篇目 2.3396。 (https://5dlibrary.com/answers/zh/content-fission-algebra-compiling-hard-core-monographs-two-host/c07/) - Ch.8 Practice and Falsification: an Owner's Checklist, a Front-End Style Library and a Pre-Registered Testing Plan — Chapter 8 gives an owner's checklist, the distilled front-end style library (body-text contrast 15.80) and a pre-registered plan. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/content-fission-algebra-compiling-hard-core-monographs-two-host/c08/) 第8章 实践与证伪:经营者清单、前端样式库蒸馏件与预注册检验方案 — 第八章给出经营者清单、前端样式库蒸馏件(正文对比度 15.80)与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/content-fission-algebra-compiling-hard-core-monographs-two-host/c08/) ## 5DL-B-152 · Drift Control for Parallel Multi-Window Book Writing, Cross-Volume Terminology Graph Audits and Zero-Hallucination Citation Verifiers 中文书名:多窗口并行写书防漂移控制、术语跨卷一致性图审计与零幻觉文献核验器 DOI: pending Cite: Fan, Y. (2026). Drift Control for Parallel Multi-Window Book Writing, Cross-Volume Terminology Graph Audits and Zero-Hallucination Citation Verifiers (Version 0.1.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Every error rate without a source is an assumed parameter or a measurement on synthetic data; the book is not legal advice and recommends no writing tool or model. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it. Abstract: This volume is written for publishers and studio principals who produce manuscripts with several AI writing windows at once. Its claim is that when several windows write one series, the errors are not random: they gather in five failure modes, terminology drift, number mismatch, duplicate sentences, citation hallucination and style degradation. Each can be written as a detector that runs; the detectors are grouped by Chen Yuan's four methods of collation into parallel, internal, external and rational gates, and their detection rates and misses can be measured by injecting known errors. The machine checks what has a copy to check against, Duan Yucai's "right and wrong of the base text", and people check what has none, "the right and wrong of the argument". Every error rate without a source is an assumed parameter or a measurement on synthetic data; the book is not legal advice and recommends no writing tool or model. Chapter 1 is the method. When the windows' errors are correlated, adding windows does not average them away: at an intraclass correlation of 0.02, sixteen windows are worth only 48.35 independent checks. Four windows sharing one stock of boilerplate should expect 18.22 duplicate pairs in a nine-chapter book, against 4.55 with separate stocks; a "check, then create" lock ends in double ownership in 2160 of the 2520 interleavings of four windows, and an exclusive create brings that to zero; at the error rates a published study found for GPT-4 citations, a chapter with 12 citations has a probability of only 0.003432 that all are real and correct. Chapters 2 to 7 cover concurrency locks and wave scheduling, the cross-volume term graph, bilingual number reconciliation and duplicate detection, zero-hallucination citation checking, protection against style degradation and the automation of the four methods of collation, ending with the distillation spec for the four kits. Chapter 8 has four windows write a synthetic nine-chapter book with 119 injected errors in five classes: the four gates stop 0.723, 28 of the 33 escapes belong to techniques the machine cannot see by construction, and without them the detection rate is 0.945; on the clean book the first term run raises 1672 false alarms, which fall to zero after a person reviews 48 words; a serious injection audit needs at least 75 errors per class. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it. 摘要:本卷写给用多个人工智能写作窗口并发生产书稿的出版者与工作室主理人。它的命题是:几个窗口同时写同一套书时,错误不是随机的,而是集中在术语漂移、数字不一致、重复句、引文幻觉与风格退化五种失效模式上;每一种都能写成一个可运行的检测器,检测器按陈垣的校勘四法归入对校、本校、他校、理校四道闸门,检出率与漏检可以用注入已知错误的办法测出来。机器查有本可对的「底本之是非」,人查无本可对的「立说之是非」,这是本书借段玉裁校勘学划出的分工。书中未标出处的错误率都是假设参数或合成数据上的测量值;本书不是法律意见,也不推荐任何写作工具或模型。 第一章是方法:窗口之间的错误相关时,加窗口平均不掉错误,组内相关系数 0.02 时十六个窗口只等效于 48.35 处独立核对;四个窗口共用一个套话库,九章书期望 18.22 对重复句,各用各的只有 4.55 对;「先检查再创建」的锁在四个窗口的 2520 种交错里有 2160 种双重拥有,独占新建把它降为零;按一项已发表研究里 GPT-4 引文的出错比例,一章 12 条引文全部真实且无误的概率只有 0.003432。第二至第七章依次写并发锁与分波次调度、跨卷术语图、中英数字对账与重复句检测、零幻觉文献核验、风格防退化与校勘四法的自动化,并以四件套蒸馏规格收尾。第八章让四个窗口并发写一本九章合成书,注入五类共 119 个错误:四道闸门拦住 0.723,漏网的 33 个里有 28 个属于机器在结构上看不见的手法,扣除后检出率为 0.945;干净书上术语检测首轮误报 1672 处,人审 48 个词后归零;一份严肃的注入审计每类至少要 75 个错误。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它。 Chapter answers: - Ch.1 Method: Five Failure Modes, Exclusive Locks and the Four Methods of Collation — Chapter 1 is the method. When the windows' errors are correlated, adding windows does not average them away: at an intraclass correlation of 0.02, sixteen windows are worth only 48.35 independent checks. Four windows sharing one stock of boilerplate should expect 18.22 duplicate pairs in a nine-chapter book, against 4.55 with separate stocks; a "check, then create" lock ends in double ownership in 2160 of the 2520 interleavings of four windows, and an exclusive create brings that to zero; at the error rates a published study found for GPT-4 citations, a chapter with 12 citations has a probability of only 0.003432 that all are real and correct. (https://5dlibrary.com/answers/en/drift-control-parallel-multi-window-book-writing-cross-volume/c01/) 第1章 方法:五种失效模式、独占锁与校勘四法 — 第一章是方法:窗口之间的错误相关时,加窗口平均不掉错误,组内相关系数 0.02 时十六个窗口只等效于 48.35 处独立核对;四个窗口共用一个套话库,九章书期望 18.22 对重复句,各用各的只有 4.55 对;「先检查再创建」的锁在四个窗口的 2520 种交错里有 2160 种双重拥有,独占新建把它降为零;按一项已发表研究里 GPT-4 引文的出错比例,一章 12 条引文全部真实且无误的概率只有 0.003432。第二至第七章依次写并发锁与分波次调度、跨卷术语图、中英数字对账与重复句检测、零幻觉文献核验、风格防退化与校勘四法的自动化,并以四件套蒸馏规格收尾。 (https://5dlibrary.com/answers/zh/drift-control-parallel-multi-window-book-writing-cross-volume/c01/) - Ch.2 Concurrency Control: Exclusive-Create Locks, Leases, Version Numbers and Wave Scheduling — Testing and creating must be a single atomic action; otherwise, with four windows starting together, double ownership occurs in 0.8571 of all interleavings. Conflicts over chapters rise steeply with the number of windows: four windows picking among nine chapters already collide with probability 0.5391. Locks drive duplicate drafts to zero, the writer table makes responsibility traceable, and the two are used together. Leases reclaim the locks of silent holders, and their length is set by the heartbeat loss rate and the clock skew; reclamation may only rename and then create, and acquisition numbers stop late writes. Shared files compare version numbers and are sharded when retries pile up; locks everyone needs go through tickets; multiple locks follow one global order. Dependencies between chapters yield a wave zero, cutting the critical path from 22 to 17. Reviews sort by logical clocks and never assign blame by wall clocks. The next chapter turns to the term lock table these mechanisms protect. (https://5dlibrary.com/answers/en/drift-control-parallel-multi-window-book-writing-cross-volume/c02/) 第2章 并发控制:独占新建锁、租约、版本号与分波次调度 — 检查与新建必须是一个原子动作,否则四个窗口同时起跑时,双重占有在所有交错里占 0.8571。抢章节的冲突随窗口数急剧上升,九章四窗口同时挑,撞车概率已是 0.5391;加锁让重复草稿归零,写手表让责任可追,两者叠加使用。租约用来回收失联者的锁,长短由心跳丢失率和时钟偏差决定;回收只准改名再新建,并用取锁序号挡住迟到的写入。共享文件比对版本号,重试过多就分片;人人都要的锁一律排号;多把锁按全局次序拿。章节之间的依赖拆出第零波,关键路径因此从 22 缩到 17。复盘用逻辑时钟排序,不用墙上时钟追责。下一章转向这些机制所保护的那张术语锁表本身。 (https://5dlibrary.com/answers/zh/drift-control-parallel-multi-window-book-writing-cross-volume/c02/) - Ch.3 The Terminology Graph: Lock Tables, Synonym Drift and Cross-Volume Consistency Audits — Terminology drift has three sources: literal variants, synonym substitution and translation drift. String rulers catch only the first; the second needs aliases registered in the lock table; the third needs the term graph built from Chinese–English pairings. Edit distance weighs heavily on short Chinese terms, where one character's difference in a two-character term leaves a similarity of only 0.500; bigrams must be padded, and English compact keys compared first. On the constructed sample, the two rulers reach only about six in ten precision when recall is near perfect, so they only produce review lists, and automatic verdicts come from the lock table alone. The cross-volume term graph takes components along hard edges only, and soft edges never merge anything automatically; the five audit rules put collisions and translation drift first, and return definition conflicts to the volume that owns the definition. The blast radius of every lock-table change is computed exactly as "uses it and is pinned before the change". The next chapter turns to something else that drifts between windows just as terms do: numbers. (https://5dlibrary.com/answers/en/drift-control-parallel-multi-window-book-writing-cross-volume/c03/) 第3章 术语图:锁表、同义漂移与跨卷一致性审计 — 术语漂移有字面变体、同义替换与对译漂移三种来源,字符串尺子只能抓第一种,第二种靠锁表里登记的别名,第三种靠中英配对构成的术语图。编辑距离在短中文术语上一字千钧,两字术语差一个字相似度就只剩 0.500;字二元组要补边界,英文要先比紧凑键。构造样本上,两把尺子在召回率接近满分时精确率只有六成左右,所以它们只负责产生复核名单,自动判定一律来自锁表。跨卷术语图只沿硬边求分量,软边从不自动合并;五条审计规则把碰撞与对译漂移放在最前,定义冲突交还定义权所在的那一卷。锁表每次改动的波及面,按「用了且版本钉早于修改」精确计算。下一章转向与术语同样会在多窗口之间漂移的另一样东西:数字。 (https://5dlibrary.com/answers/zh/drift-control-parallel-multi-window-book-writing-cross-volume/c03/) - Ch.4 Number Parity and Duplicate Sentences: Bilingual Agreement, Program Reconciliation and Near-Duplicate Detection — Number reconciliation checks two things: the sets of checkable numbers in the Chinese and English drafts are equal, and every checkable number in the prose can be found in a program's output or source. Its blind spots come from its own definition: digits glued to Chinese characters are invisible, comma-less integers in the year band are never checked, 386.5 and 386.50 count as two numbers, and 1,032 with a comma cannot be found in an output that prints it without one. The patches are conventions: spaces around numerals, thousands commas on every non-year large integer, and programs printing exactly the format the prose will copy. Duplicate detection has two layers. Verbatim repeats of 20 characters or more use a dictionary or 64-bit fingerprints, with normalisation limited to NFKC, whitespace removal and dropping the closing mark. Reworded near-duplicates use 2- to 3-character shingles, MinHash signatures and banded candidates, closed by an exact Jaccard check. Splits and thresholds are not copied from elsewhere but set from the cost ratio of misses to false alarms; under this chapter's assumptions, 64×2 with a threshold of 0.35 has the lowest expected cost per book. (https://5dlibrary.com/answers/en/drift-control-parallel-multi-window-book-writing-cross-volume/c04/) 第4章 数字对账与重复句:中英一致、程序对账与近似重复检测 — 数字对账看的是两件事:中英两份稿的可检数字集合相等,正文的每个可检数字都能在程序输出或源码里找到。它的盲区来自它的定义:紧贴汉字的数字看不见,年份带里的整数不检查,386.5 与 386.50 被当作两个数,带逗号的 1,032 在不带逗号的输出里找不到。补法是格式约定:数字两侧留空格,非年份的大整数一律写千分位,程序按正文要抄的格式打印。重复句检测分两层:20 字以上的逐字重复用字典或 64 位指纹,规范化只做 NFKC、删空白与去句末标点;改写过的近似重复用 2 到 3 字的字符片段、MinHash 签名与分带候选,再用精确 Jaccard 收口。切法与阈值不靠抄,要按漏报与误报的代价比来定;在本章的假设下,64×2 加 0.35 的阈值每本书期望成本最低。 (https://5dlibrary.com/answers/zh/drift-control-parallel-multi-window-book-writing-cross-volume/c04/) - Ch.5 A Zero-Hallucination Citation Verifier: Fields, Identifiers, Ledger Matching and Item-by-Item Verdicts — Zero-hallucination verification has three gates. The format gate confirms complete fields, a well-formed DOI, a correct ISBN check digit and a sensible year and volume; it stops copying errors but not a well-made fabrication, because a fabricator can compute the check digit first and assemble a DOI to pattern. The ledger gate matches author, year, title and venue against a local ledger checked record by record against originals; only a title similarity of 0.90 or more can pass, and anything from 0.60 to 0.90 goes to a person. In weighted scoring the "title differs, rest agree" pattern still scores +11.973, so the title must be a gate and not a bonus. Verdicts fall into pass, fix, hold and reject, and only the first two may exist in a print-ready manuscript. In this chapter's fictional sample, F6 with its borrowed DOI and the chimera F5 are the two entries the format checks let straight through and only the ledger stops; and the higher the ledger's coverage, the shorter the hold list and the more suspicious each hold becomes. (https://5dlibrary.com/answers/en/drift-control-parallel-multi-window-book-writing-cross-volume/c05/) 第5章 零幻觉文献核验器:字段、标识符、台账比对与逐条判决 — 零幻觉核验分三道。格式检查确认字段齐全、DOI 合语法、书号校验位正确、年份卷期说得通;它拦得住抄写错误,拦不住像样的编造,因为编造者可以先算好校验位、照格式拼出 DOI。台账比对拿作者、年份、标题与出处去对一份逐条核过原件的本地台账,标题相似度不低于 0.90 才可能通过,0.60 到 0.90 之间一律交人判断;加权打分里「只有标题不同」的组合仍有 +11.973 的高分,所以标题必须是闸门而不是加分项。判决分通过、改正、暂扣、拒收四档,交印稿只允许前两档存在。在本章的虚构样本里,借用别人 DOI 的 F6 与嵌合标题的 F5 是格式检查完全放过、只有台账才拦得住的两条;而台账覆盖率越高,暂扣清单越短,每一条暂扣也越值得怀疑。 (https://5dlibrary.com/answers/zh/drift-control-parallel-multi-window-book-writing-cross-volume/c05/) - Ch.6 Style Drift: Gates for Sentence Length, Lexical Diversity, Stock Phrases and Paragraph Openers — The two passages are of similar size, 271 Chinese characters for A and 233 for B, yet they separate on almost every line. The coefficient of variation of sentence length is 0.670 for A and only 0.371 for B; the share of sentences within three characters of the median is 0.267 for A and 0.455 for B. A mixes long and short: a sentence of eight characters about how quickly a number can be corrected sits right next to a forty-character explanation. B's sentences nearly all crowd around twenty characters. The type-token ratio is 0.590 for A and 0.322 for B; MTLD is 128.1 for A and falls to 44.5 for B. Stock-phrase hits run at zero per thousand characters in A and 128.8 in B. Four of B's six paragraphs share one four-character opener, and the share of character 8-grams that occur more than once is 0.000 in A against 0.199 in B. (https://5dlibrary.com/answers/en/drift-control-parallel-multi-window-book-writing-cross-volume/c06/) 第6章 风格防退化:句长、词汇多样性、套话与段首的门槛 — 两段的篇幅相近,A 有 271 个汉字,B 有 233 个,可是几乎每一项都分开了。句长的变异系数,A 是 0.670,B 只有 0.371;落在中位数上下三个字以内的句子,A 占 0.267,B 占 0.455。A 的句子有长有短,「改一个数只要十秒」只有八个字,紧挨着一句四十字的解释;B 的句子几乎全挤在二十字上下。类符形符比 A 是 0.590,B 是 0.322;按 MTLD 算,A 是 128.1,B 掉到 44.5。套话命中,A 每千字为零,B 每千字 128.8 次。六个段落里,B 有四段共用同一个四字开头;字符八元组里重复出现的比例,A 为 0.000,B 为 0.199。 (https://5dlibrary.com/answers/zh/drift-control-parallel-multi-window-book-writing-cross-volume/c06/) - Ch.8 Practice: Four Windows Write a Nine-Chapter Book, Five Classes of Injected Error and the Detection Ledger of Four Gates — Chapter 8 has four windows write a synthetic nine-chapter book with 119 injected errors in five classes: the four gates stop 0.723, 28 of the 33 escapes belong to techniques the machine cannot see by construction, and without them the detection rate is 0.945; on the clean book the first term run raises 1672 false alarms, which fall to zero after a person reviews 48 words; a serious injection audit needs at least 75 errors per class. Marked boxes separate published evidence, runnable calculations, sources and the author's hypotheses, each stating what would refute it. (https://5dlibrary.com/answers/en/drift-control-parallel-multi-window-book-writing-cross-volume/c08/) 第8章 实战:四个窗口并发写一本九章书,五类注入错误与四道闸门的检出账 — 第八章让四个窗口并发写一本九章合成书,注入五类共 119 个错误:四道闸门拦住 0.723,漏网的 33 个里有 28 个属于机器在结构上看不见的手法,扣除后检出率为 0.945;干净书上术语检测首轮误报 1672 处,人审 48 个词后归零;一份严肃的注入审计每类至少要 75 个错误。书中以标注框区分已发表的证据、可运行的计算、原典与作者的假说,每条假说都写明什么结果会推翻它。 (https://5dlibrary.com/answers/zh/drift-control-parallel-multi-window-book-writing-cross-volume/c08/) ## 5DL-B-153 · Da Liu Ren Heaven-Earth Plate Permutations, Four-Lesson Three-Transmission Digraphs and the Depth Algebra of the Nine Methods 中文书名:大六壬天地盘置换、四课三传有向图与九宗门涉害深度代数 DOI: pending Cite: Fan, Y. (2026). Da Liu Ren Heaven-Earth Plate Permutations, Four-Lesson Three-Transmission Digraphs and the Depth Algebra of the Nine Methods (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters about users, archives and prices are illustrative; all calculations are generated by its programs; it analyses the structure and probabilities of a system, makes no personal reading, and is not legal, financial or business advice. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: Da Liu Ren has long been ranked first among the three cosmic boards, yet its rules have rarely been computed as a deterministic machine. This book answers only a structural question: what distribution do the plates, the four lessons and the nine methods produce over all situations? Sixty days, twelve monthly generals and twelve hours give 8,640 situations, and the book enumerates every one. Its parameters about users, archives and prices are illustrative; all calculations are generated by its programs; it analyses the structure and probabilities of a system, makes no personal reading, and is not legal, financial or business advice. Chapter 1 writes the heaven plate as the cyclic group of order twelve acting on twelve palaces: general and hour act only through their difference, fu-yin is the identity and fan-yin the only element of order two, together 0.1667 of all situations; trines, meng-zhong-ji and clashes are cosets of subgroups. Chapter 2 derives the four lessons from the lodging palaces and shows that the lodging table avoids the four zhong palaces. Chapter 3 writes the nine methods as a decision tree and enumerates it exactly: zei-ke takes 0.4681, she-hai only 0.1056, fu-yin and fan-yin 0.0833 each, and the method entropy is only 2.419 bits. Chapter 4 proves that she-hai depth is a count summed over an arc of the ground plate that saturates quickly, so depth decides only 0.585 of she-hai boards; four common readings together change only 0.0431 of the classes. Chapter 5 treats the three transmissions as paths on rotation orbits whose shape is fixed by the order of the step; regular paths take 0.8611 and closed trines 0.1653. Chapter 6 tests the uniform weights on the real calendar of 2026 and finds that the step accounts for a share of 0.434 of the method entropy. Chapter 7 examines the Rengui, the case records of Shao Yanhe and Ptolemy's Tetrabiblos, showing that the 720 lessons reduce to 141 sets of transmissions and that Ptolemy's aspects and triplicities are cosets on the same cyclic group. Chapter 8 gives an operator's checklist, design tokens for a front-end style library and a pre-registered archive test with a power of 0.895 at 200 cases. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:大六壬向居三式之首,却很少有人把它的规则当作一台确定的机器来计算。本书只回答结构问题:天地盘、四课与九宗门在全部局面上产出什么样的分布。六十日、十二月将、十二时辰一共 8,640 个局面,本书对它们逐一穷举。书中关于用户、档案与价格的参数均为示意;全部计算由书中程序生成;本书分析体系的结构与概率,不做任何个人占断,也不是法律、财务或经营建议。 第一章把天盘写成十二阶循环群在十二宫上的作用:月将与占时只通过差值起作用,伏吟是单位元,返吟是唯一的二阶元,两者合计占 0.1667;三合、孟仲季、六冲都是子群的陪集。第二章由日干寄宫推出四课,说明寄宫表完全避开四仲。第三章把九宗门写成决策树并精确枚举:贼克门占 0.4681,涉害只占 0.1056,伏吟与返吟各 0.0833,法门熵只有 2.419 比特。第四章证明涉害深度是地盘一段弧上的计数和,会很快饱和,深浅只决定 0.585 的涉害局面;四种常见写法合计只改变 0.0431 的局面。 第五章把三传看作旋转轨道上的路径,路径形状由步长的阶决定,规则路径占 0.8611,三传成三合局占 0.1653。第六章用 2026 年的真实历法检验均匀权重,并算出法门熵中由步长决定的份额为 0.434。第七章考辨《壬归》、邵彦和断案与托勒密《四书》,说明七百二十课在三传层面只有 141 种,托勒密的相位与三分宫是同一循环群上的陪集。第八章给出经营者清单、前端样式库的设计令牌与一份预注册档案检验,至少 200 个案例时功效为 0.895。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Plates: a Cyclic Permutation Group on Twelve Palaces — Chapter 1 writes the heaven plate as the cyclic group of order twelve acting on twelve palaces: general and hour act only through their difference, fu-yin is the identity and fan-yin the only element of order two, together 0.1667 of all situations; trines, meng-zhong-ji and clashes are cosets of subgroups. (https://5dlibrary.com/answers/en/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c01/) 第1章 天地盘:十二宫上的循环置换群 — 第一章把天盘写成十二阶循环群在十二宫上的作用:月将与占时只通过差值起作用,伏吟是单位元,返吟是唯一的二阶元,两者合计占 0.1667;三合、孟仲季、六冲都是子群的陪集。 (https://5dlibrary.com/answers/zh/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c01/) - Ch.2 The Four Lessons: Lodging Palaces, the Sixty-Day Cycle and Lesson Counts — The sixty-day cycle is the cyclic subgroup of order 60 in the direct product, and stem and branch always share parity. The lodging table maps ten stems onto eight palaces and avoids the zhong palaces entirely. The four lessons are fixed by palace, branch and step; the distance from palace to branch is uniform on 0 to 11, five days each. Of the 720 classes, 500 have complete lessons, 100 have three, 115 have two and 5 have one. The share of zei and ke runs from 0.025 to 0.392 across the steps, and that swing shapes the distribution of methods in the next chapter. (https://5dlibrary.com/answers/en/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c02/) 第2章 四课:日干寄宫、六十甲子与课体计数 — 六十甲子是直积中阶为 60 的循环子群,干支奇偶永远一致。寄宫表把十个天干映到八个宫位,完全避开四仲。四课由寄宫、日支与步长三者决定;寄宫与日支的距离在 0 到 11 上均匀分布,每个距离 5 日。720 类局面中,四课齐全 500 类,不备 100 类,两课 115 类,一课 5 类。贼克的比例随步长从 0.025 变到 0.392,这个起伏决定了下一章各法门的分布。 (https://5dlibrary.com/answers/zh/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c02/) - Ch.3 The Nine Methods as a Decision Tree: Exact Enumeration of 8,640 Situations — Chapter 3 writes the nine methods as a decision tree and enumerates it exactly: zei-ke takes 0.4681, she-hai only 0.1056, fu-yin and fan-yin 0.0833 each, and the method entropy is only 2.419 bits. (https://5dlibrary.com/answers/en/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c03/) 第3章 九宗门决策树:8,640 个局面的精确枚举 — 第三章把九宗门写成决策树并精确枚举:贼克门占 0.4681,涉害只占 0.1056,伏吟与返吟各 0.0833,法门熵只有 2.419 比特。 (https://5dlibrary.com/answers/zh/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c03/) - Ch.4 The Depth Algebra of She-hai: Arc Sums, Saturation and Disagreement Between Schools — Chapter 4 proves that she-hai depth is a count summed over an arc of the ground plate that saturates quickly, so depth decides only 0.585 of she-hai boards; four common readings together change only 0.0431 of the classes. (https://5dlibrary.com/answers/en/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c04/) 第4章 涉害深度代数:弧和、饱和与流派分歧 — 第四章证明涉害深度是地盘一段弧上的计数和,会很快饱和,深浅只决定 0.585 的涉害局面;四种常见写法合计只改变 0.0431 的局面。 (https://5dlibrary.com/answers/zh/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c04/) - Ch.5 The Three Transmissions as a Digraph: Rotation Orbits and Path Shapes — Chapter 5 treats the three transmissions as paths on rotation orbits whose shape is fixed by the order of the step; regular paths take 0.8611 and closed trines 0.1653. (https://5dlibrary.com/answers/en/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c05/) 第5章 三传有向图:旋转的轨道与路径形状 — 第五章把三传看作旋转轨道上的路径,路径形状由步长的阶决定,规则路径占 0.8611,三传成三合局占 0.1653。 (https://5dlibrary.com/answers/zh/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c05/) - Ch.6 Branch Probabilities and Information: Calendar Sampling, Mutual Information and Uniform Weights — Chapter 6 tests the uniform weights on the real calendar of 2026 and finds that the step accounts for a share of 0.434 of the method entropy. (https://5dlibrary.com/answers/en/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c06/) 第6章 分支概率与信息:历法抽样、互信息与均匀权重 — 第六章用 2026 年的真实历法检验均匀权重,并算出法门熵中由步长决定的份额为 0.434。 (https://5dlibrary.com/answers/zh/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c06/) - Ch.7 The Classics Examined: the Rengui, the Case Records of Shao Yanhe and Ptolemy's Tetrabiblos — Chapter 7 examines the Rengui, the case records of Shao Yanhe and Ptolemy's Tetrabiblos, showing that the 720 lessons reduce to 141 sets of transmissions and that Ptolemy's aspects and triplicities are cosets on the same cyclic group. (https://5dlibrary.com/answers/en/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c07/) 第7章 原典考辨:《壬归》、邵彦和断案与托勒密《四书》 — 第七章考辨《壬归》、邵彦和断案与托勒密《四书》,说明七百二十课在三传层面只有 141 种,托勒密的相位与三分宫是同一循环群上的陪集。 (https://5dlibrary.com/answers/zh/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c07/) - Ch.8 Practice and Falsification: a Style Library, a Pre-registered Test and the Conditions for Refutation — Chapter 8 gives an operator's checklist, design tokens for a front-end style library and a pre-registered archive test with a power of 0.895 at 200 cases. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c08/) 第8章 实践与证伪:样式库、预注册检验与推翻条件 — 第八章给出经营者清单、前端样式库的设计令牌与一份预注册档案检验,至少 200 个案例时功效为 0.895。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/da-liu-ren-heaven-earth-plate-permutations-four-lesson/c08/) ## 5DL-B-154 · Taiyi Shenshu: Sixteen-Palace Walks, Accumulated-Year Congruence Cycles and the Game Bounds of Host and Guest Counting Rods 中文书名:太乙神数十六宫游走、积年同余周期与主客算筹博弈边界 DOI: pending Cite: Fan, Y. (2026). Taiyi Shenshu: Sixteen-Palace Walks, Accumulated-Year Congruence Cycles and the Game Bounds of Host and Guest Counting Rods (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; its chronicle of calamities is synthetic data generated with a fixed seed. Chapter 6 runs random-permutation tests on a synthetic chronicle: the most "significant" configuration has a family-wise p of 0.501, and once the choice of convention and tolerance is counted, the chance that some path is nominally significant is 0.867. Abstract: Every Taiyi Shenshu reading starts from a very large accumulated year, derives the positions of the gods by division and remainder, and computes host and guest counts from those positions. This book writes the rules of the Taiyi Jinjing Shijing as runnable programs and asks three questions: what is the mathematical structure of the algorithm, is the comparison of long and short counts even-handed, and can later writings that set Taiyi configurations against calamities and campaigns serve as evidence. Its parameters are illustrative; all calculations are generated by its programs; its chronicle of calamities is synthetic data generated with a fixed seed. The book analyses structure and probability, offers no personal reading of any kind and makes no judgement about any year, event or person. Chapter 1 shows that the accumulated year acts only through residues: three documented conventions agree exactly modulo 720, and no god can read more than 12.81 bits. Chapter 2 enumerates the ring of sixteen gods and finds that six common configurations together fill 0.681 of the seventy-two years. Chapter 3 finds a joint period of 7200 years for Wufu and Dayou, on which all disagreement between the conventions lies. Chapter 4 gives the exact distribution of the counts over the seventy-two ju: the host count is longer in 33 and the guest count in 38, with a strong lean by palace, and eight readings favour the host in up to 0.708 of years. Chapter 5 derives the game bounds of the readings: on records in which the host side holds seven in ten, an uninformed reading still beats a coin by 0.083. Chapter 6 runs random-permutation tests on a synthetic chronicle: the most "significant" configuration has a family-wise p of 0.501, and once the choice of convention and tolerance is counted, the chance that some path is nominally significant is 0.867. Chapter 7 reads the Taiyi Jinjing Shijing alongside Kepler's De Stella Nova and finds the arithmetic of the accumulated year consistent but two of its ties to dates wrong. Chapter 8 gives a pre-registration plan (eleven hundred years of records can detect, per palace, only a calamity share of 0.289 or more), a checklist for owners and a style specification for a sixteen-palace walk component. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:太乙神数的全部推演,从一个巨大的积年出发,经过除法与取余得出诸神位置,再由位置算出主算与客算。本书把《太乙金镜式经》的规则写成可以运行的程序,只问三个问题:这套算法的数学结构是什么,主客算的「长短」比较是否对等,后世把太乙格局与灾异、兵事相对照的文字能否作为证据。书中参数为示意;全部计算由书中程序生成;书中灾异年表为固定种子生成的合成数据。本书分析结构与概率,不做任何个人占断,不对任何年份、事件或个人作出判断。 第一章说明积年只以余数起作用:三种有据可查的积年在模 720 上完全一致,任何一个神能读到的信息不超过 12.81 比特。第二章对十六神环做完全枚举,六种常见格局合起来占七十二年中的 0.681。第三章算出五福与大游的联合周期为 7200 年,三种积年的全部分歧都在这两位长周期神上。第四章给出主客算在七十二局中的精确分布:主算较长 33 局,客算较长 38 局,各宫位的偏向悬殊;八条读法偏向主方的比例最高达 0.708。 第五章推导主客读法的博弈边界:在主方占七成的记载上,毫无信息的读法也能比掷硬币高出 0.083。第六章在合成年表上做随机置换检验,最「显著」的格局族系 p 值为 0.501,计入积年与容差的选择后,至少一条路径名义显著的机会是 0.867。第七章对读《太乙金镜式经》与开普勒的《论新星》,判定积年算术自洽而两处年代挂钩有误。第八章给出预注册方案(一千一百年史料每宫只能检出 0.289 以上的灾异比例)、经营者清单与十六宫游走组件的样式规格。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Accumulated Year Is a Residue: Thirteen Readable Bits in a Seven-Digit Number — Chapter 1 shows that the accumulated year acts only through residues: three documented conventions agree exactly modulo 720, and no god can read more than 12.81 bits. (https://5dlibrary.com/answers/en/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c01/) 第1章 积年只是余数:七位数里能被读到的十三比特 — 第一章说明积年只以余数起作用:三种有据可查的积年在模 720 上完全一致,任何一个神能读到的信息不超过 12.81 比特。 (https://5dlibrary.com/answers/zh/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c01/) - Ch.2 A Deterministic Walk on the Ring of Sixteen Gods: Occupancy and the True Base Rate of "Rare" Configurations — Chapter 2 enumerates the ring of sixteen gods and finds that six common configurations together fill 0.681 of the seventy-two years. (https://5dlibrary.com/answers/en/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c02/) 第2章 十六神环上的确定性游走:占位频率与「罕见格局」的真实基率 — 第二章对十六神环做完全枚举,六种常见格局合起来占七十二年中的 0.681。 (https://5dlibrary.com/answers/zh/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c02/) - Ch.3 The Great Orbits of Wufu and Dayou: Least Common Periods, Irregular Revisits and the Whole Extent of the Schools' Disagreement — Chapter 3 finds a joint period of 7200 years for Wufu and Dayou, on which all disagreement between the conventions lies. (https://5dlibrary.com/answers/en/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c03/) 第3章 五福与大游的大轨道:最小公倍周期、不规则重访与流派分歧的全部所在 — 第三章算出五福与大游的联合周期为 7200 年,三种积年的全部分歧都在这两位长周期神上。 (https://5dlibrary.com/answers/zh/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c03/) - Ch.4 Host and Guest Counts: the Exact Distribution over Seventy-Two Ju and the Structural Asymmetry of "Long and Short" — Chapter 4 gives the exact distribution of the counts over the seventy-two ju: the host count is longer in 33 and the guest count in 38, with a strong lean by palace, and eight readings favour the host in up to 0.708 of years. (https://5dlibrary.com/answers/en/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c04/) 第4章 主算与客算:七十二局的精确分布与「长短」比较的结构性不对称 — 第四章给出主客算在七十二局中的精确分布:主算较长 33 局,客算较长 38 局,各宫位的偏向悬殊;八条读法偏向主方的比例最高达 0.708。 (https://5dlibrary.com/answers/zh/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c04/) - Ch.5 The Game Bounds of the Counting Rods: Base-Rate Leakage and the Ceiling on Choosing the Best of Many Readings — Chapter 5 derives the game bounds of the readings: on records in which the host side holds seven in ten, an uninformed reading still beats a coin by 0.083. (https://5dlibrary.com/answers/en/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c05/) 第5章 主客算筹的博弈边界:基率泄漏与「多条读法择优」的上限 — 第五章推导主客读法的博弈边界:在主方占七成的记载上,毫无信息的读法也能比掷硬币高出 0.083。 (https://5dlibrary.com/answers/zh/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c05/) - Ch.6 Random-Permutation Tests on a Chronicle of Calamities: Why Fitting after the Fact Almost Always Succeeds — Chapter 6 runs random-permutation tests on a synthetic chronicle: the most "significant" configuration has a family-wise p of 0.501, and once the choice of convention and tolerance is counted, the chance that some path is nominally significant is 0.867. (https://5dlibrary.com/answers/en/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c06/) 第6章 灾异年表的随机置换检验:事后拟合为何几乎总能成功 — 第六章在合成年表上做随机置换检验,最「显著」的格局族系 p 值为 0.501,计入积年与容差的选择后,至少一条路径名义显著的机会是 0.867。 (https://5dlibrary.com/answers/zh/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c06/) - Ch.7 The Sources Examined: the Self-Consistency of the Golden Mirror's Accumulated Years and Kepler's Cycle of Great Conjunctions — The Golden Mirror's algorithm for the accumulated year is strictly consistent at the level of residues, but two of its ties to dates can be judged wrong: Kaiyuan 15 was not a jiazi year, and the thirty-seventh year of King Li of Zhou lies 5 years from the computed jiazi start. Under today's mean motions Kepler's great-conjunction cycle is about 882 years, longer than the eight hundred he used; Jupiter and Saturn are only approximately commensurable, whereas the Taiyi gods share an exact common period of 7200 years. (https://5dlibrary.com/answers/en/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c07/) 第7章 原典考辨:《太乙金镜式经》的积年自洽性与开普勒的大会合周期 — 《金镜》的积年算法在余数层面严格自洽,但有两处年代挂钩可以判错:开元十五年不是甲子年,周厉王三十七年与推算的甲子起点差 5 年。开普勒的大会合在今天的平均运动下周期约 882 年,比他采用的约八百年偏长;木星与土星只有近似公度,而太乙诸神有严格的 7200 年公倍周期。 (https://5dlibrary.com/answers/zh/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c07/) - Ch.8 Practice and a Plan for Falsification: Pre-Registered Tests, a Sixteen-Palace Walk Component and a Checklist for Owners — Chapter 8 gives a pre-registration plan (eleven hundred years of records can detect, per palace, only a calamity share of 0.289 or more), a checklist for owners and a style specification for a sixteen-palace walk component. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c08/) 第8章 实践与证伪方案:预注册检验、十六宫游走组件与经营者清单 — 第八章给出预注册方案(一千一百年史料每宫只能检出 0.289 以上的灾异比例)、经营者清单与十六宫游走组件的样式规格。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/taiyi-shenshu-sixteen-palace-walks-accumulated-year-congruence-cycles/c08/) ## 5DL-B-155 · Seven Governors and Four Remainders: Ecliptic Lunar-Mansion Perturbations, Shadow-Node Orbits and Error Propagation in Dongwei Limits 中文书名:七政四余黄道宿度摄动、四余暗曜交点轨道与洞微限度误差传播 DOI: pending Cite: Fan, Y. (2026). Seven Governors and Four Remainders: Ecliptic Lunar-Mansion Perturbations, Shadow-Node Orbits and Error Propagation in Dongwei Limits (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative or sourced; all calculations are generated by its programs; it analyses only the structure and probabilities of the system, makes no personal reading of any kind, and is not advice of any kind. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: The Seven Governors and Four Remainders is the branch of traditional Chinese star lore closest to astronomy: Rahu and Ketu are the nodes of the lunar orbit, Yuebei is the lunar apogee, and the twenty-eight mansions are defined by particular determinative stars. This book writes the popular claims of the system as runnable computations and asks how precise each position, palace boundary and period really is, and where the disagreements between programs and almanacs come from. Its parameters are illustrative or sourced; all calculations are generated by its programs; it analyses only the structure and probabilities of the system, makes no personal reading of any kind, and is not advice of any kind. Chapters 1 and 2 compute the nodes and Yuebei from the polynomials of modern lunar theory: the nodal period is 18.6130 years against the equinox and 18.5995 years against the stars; the swing of the true node about the mean node changes the palace on 0.0306 of all days; and "Yuebei goes round in nine years" is wrong by 68.99 degrees a century, half a palace off within about twenty years. Chapter 3 shows that Ziqi is a purely calendrical point whose two readings give different mansions for 0.6027 of the time, and that any two of eight programs built from three conventions agree completely in only 0.3479 of cases. Chapter 4 reconstructs the reversal of the mansions Zi and Shen by rigorous precession; equatorial mansion widths stretch by up to 5.780 du over two thousand years. Chapter 5 computes the disagreement between palaces by degree and palaces by mansion: with frames aligned in 1645, a single body disagrees today with probability 0.1773, and at least one of eleven bodies with probability 0.8831. Chapter 6 shows that the double-hour and rising-sign rules for the Ming palace agree only 0.5908 of the time, and that a fifteen-minute birth-time error shifts the age of the first limit by about 1.88 years. Chapter 7 reads the Guolao xingzong, the Qintang zhijin ge and Ptolemy's Almagest side by side and finds Ptolemy's longitude of Spica 1.264 degrees too small at its own epoch. Chapter 8 gives a product checklist, a chart-component specification, a pre-registered protocol and the conditions that would overturn the book. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:七政四余是中国传统星命体系中最接近天文学的一支:罗睺、计都是月球轨道的交点,月孛是月球远地点,二十八宿由具体的距星定义。本书把这一体系中流行的说法逐条写成可以运行的计算,问每一个位置、每一条宫界、每一个年限究竟有多精确,不同软件与历书之间的分歧来自哪里。书中参数为示意或注明出处;全部计算由书中程序生成;本书只分析体系的结构与概率,不做任何个人占断,也不是任何形式的建议。 第一、二章用现代月球理论的多项式计算罗计与月孛:交点周期相对春分点为 18.6130 年,相对恒星为 18.5995 年;真交点围绕平交点摆动,使 0.0306 的日子宫位改变;「月孛九年一周」每百年误差 68.99 度,二十余年即差半宫。第三章说明紫气是纯历法虚点,两种读法在 0.6027 的时间里给出不同的宿,由三项约定组成的八部软件任意两部完全一致的比例只有 0.3479。第四章用严格的岁差旋转重现觜参两宿的换序,赤道宿度在两千余年间伸缩达 5.780 古度。 第五章计算宫度分宫与宿度分宫的分歧率:若两套框架在 1645 年对齐,今天单个星体的分歧率为 0.1773,十一个星体中至少一个不同的概率为 0.8831。第六章说明命宫的时辰法与东升法只有 0.5908 的机会一致,十五分钟的出生时刻误差使命宫限度的年龄移动约 1.88 年。第七章对读《果老星宗》《琴堂指金歌》与托勒密《至大论》,发现托勒密星表的角宿一黄经在其历元上偏小 1.264 度。第八章给出产品清单、星盘组件规格、预注册检验方案与推翻本书的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Rahu and Ketu: Mean Motion, True Motion and the Eclipse Year of the Lunar Nodes — Chapters 1 and 2 compute the nodes and Yuebei from the polynomials of modern lunar theory: the nodal period is 18.6130 years against the equinox and 18.5995 years against the stars; the swing of the true node about the mean node changes the palace on 0.0306 of all days; and "Yuebei goes round in nine years" is wrong by 68.99 degrees a century, half a palace off within about twenty years. (https://5dlibrary.com/answers/en/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c01/) 第1章 罗睺与计都:黄白交点的平运动、真运动与食年 — 第一、二章用现代月球理论的多项式计算罗计与月孛:交点周期相对春分点为 18.6130 年,相对恒星为 18.5995 年;真交点围绕平交点摆动,使 0.0306 的日子宫位改变;「月孛九年一周」每百年误差 68.99 度,二十余年即差半宫。 (https://5dlibrary.com/answers/zh/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c01/) - Ch.2 Yuebei: the Advance of the Lunar Apogee and the Price of a Rounded Period — Yuebei corresponds to the Moon's mean apogee. It can be written as a polynomial, with a circuit of 8.8474 years against the equinox and 3232.61 days against the stars. Its higher terms and the swing of its true position are far larger than those of the node. The price of a rounded period can be computed exactly: "nine years a circuit" is half a palace off after 22 years and a full palace after 43, whereas rounding to 3232 days takes 2446 years to reach half a palace. A sidereal period combined with palaces counted from the equinox makes an error equal to precession. (https://5dlibrary.com/answers/en/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c02/) 第2章 月孛:月球远地点的进动与取整周期的代价 — 月孛对应月球的平远地点,可以写成多项式,相对春分点 8.8474 年一周,相对恒星 3232.61 日一周。它的高次项与真位置的摆动都比交点大得多。取整周期的代价可以精确计算:「月孛九年一周」22 年就差半宫,43 年差一整宫;而 3232 日的取整要 2446 年才差半宫。用恒星周期配春分点起量的宫,其误差正好等于岁差。 (https://5dlibrary.com/answers/zh/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c02/) - Ch.3 Ziqi and the Lattice of Conventions: Can a Purely Calendrical Point Be Identified? — Chapter 3 shows that Ziqi is a purely calendrical point whose two readings give different mansions for 0.6027 of the time, and that any two of eight programs built from three conventions agree completely in only 0.3479 of cases. (https://5dlibrary.com/answers/en/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c03/) 第3章 紫气与四余约定格:一个纯历法虚点的可辨识性 — 第三章说明紫气是纯历法虚点,两种读法在 0.6027 的时间里给出不同的宿,由三项约定组成的八部软件任意两部完全一致的比例只有 0.3479。 (https://5dlibrary.com/answers/zh/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c03/) - Ch.4 Precession and the Determinative Stars of the Twenty-Eight Mansions: the Zi–Shen Reversal and Mansion Widths that Stretch — Chapter 4 reconstructs the reversal of the mansions Zi and Shen by rigorous precession; equatorial mansion widths stretch by up to 5.780 du over two thousand years. (https://5dlibrary.com/answers/en/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c04/) 第4章 二十八宿距星的岁差漂移:觜参换序与宿度伸缩 — 第四章用严格的岁差旋转重现觜参两宿的换序,赤道宿度在两千余年间伸缩达 5.780 古度。 (https://5dlibrary.com/answers/zh/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c04/) - Ch.5 Palaces by Degree and Palaces by Mansion: the Disagreement Rate of Two Frames — Chapter 5 computes the disagreement between palaces by degree and palaces by mansion: with frames aligned in 1645, a single body disagrees today with probability 0.1773, and at least one of eleven bodies with probability 0.8831. (https://5dlibrary.com/answers/en/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c05/) 第5章 宫度分宫与宿度分宫:两套框架的分歧率 — 第五章计算宫度分宫与宿度分宫的分歧率:若两套框架在 1645 年对齐,今天单个星体的分歧率为 0.1773,十一个星体中至少一个不同的概率为 0.8831。 (https://5dlibrary.com/answers/zh/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c05/) - Ch.6 Dongwei Limits: Two Rules for the Ming Palace and the Propagation of Birth-Time Error — Chapter 6 shows that the double-hour and rising-sign rules for the Ming palace agree only 0.5908 of the time, and that a fifteen-minute birth-time error shifts the age of the first limit by about 1.88 years. (https://5dlibrary.com/answers/en/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c06/) 第6章 洞微限度:命宫两法与出生时刻误差的传播 — 第六章说明命宫的时辰法与东升法只有 0.5908 的机会一致,十五分钟的出生时刻误差使命宫限度的年龄移动约 1.88 年。 (https://5dlibrary.com/answers/zh/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c06/) - Ch.7 The Classics Examined: the Guolao xingzong, the Qintang zhijin ge and the Almagest — Chapter 7 reads the Guolao xingzong, the Qintang zhijin ge and Ptolemy's Almagest side by side and finds Ptolemy's longitude of Spica 1.264 degrees too small at its own epoch. (https://5dlibrary.com/answers/en/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c07/) 第7章 原典考辨:《果老星宗》《琴堂指金歌》与《至大论》 — 第七章对读《果老星宗》《琴堂指金歌》与托勒密《至大论》,发现托勒密星表的角宿一黄经在其历元上偏小 1.264 度。 (https://5dlibrary.com/answers/zh/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c07/) - Ch.8 Practice and Falsification: Chart Components, Pre-Registered Tests and the Conditions for Overturning — Chapter 8 gives a product checklist, a chart-component specification, a pre-registered protocol and the conditions that would overturn the book. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c08/) 第8章 实践与证伪方案:星盘组件、预注册检验与推翻条件 — 第八章给出产品清单、星盘组件规格、预注册检验方案与推翻本书的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/seven-governors-four-remainders-ecliptic-lunar-mansion-perturbations-shadow/c08/) ## 5DL-B-156 · Plum-Blossom Numerology as a Body-Use Automaton, Hash Collisions in Date-Time Casting and Base Rates of Attention-Captured Omens 中文书名:梅花易数体用生克自动机、年月日时哈希碰撞与外应注意捕获基率 DOI: pending Cite: Fan, Y. (2026). Plum-Blossom Numerology as a Body-Use Automaton, Hash Collisions in Date-Time Casting and Base Rates of Attention-Captured Omens (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Population shares, attention rates and memory rates are illustrative; all calculations are generated by the book's programs; the book analyses structure and probability, makes no personal reading, and is not business, legal or financial advice. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: The Meihua Yishu, attributed to Shao Yong, casts a hexagram by adding the numbers of year, month, day and hour and keeping remainders, judges it by the five-phase relation of body and use, and confirms it with external omens seen by chance. This book leaves aside the metaphysics of the figures and writes the procedure as runnable arithmetic: casting is a hash function, the body-use rule a verdict machine, and the omen a base-rate problem. Population shares, attention rates and memory rates are illustrative; all calculations are generated by the book's programs; the book analyses structure and probability, makes no personal reading, and is not business, legal or financial advice. Chapter 1 proves that the 51,840 combinations of year, month, day and hour reach only 192 of 384 designed slots, because the lower trigram and the moving line come from one total divided by 8 and by 6 and so share a parity; the output entropy is only 7.499 bits, and a day has only 12 results. Chapter 2 computes collisions: two people casting at random on the same day share a figure with probability 0.0833, and a crowd in one double-hour shares it entirely; across a year the actual rule makes pairs collide 2.232 times as often as an ideal hash, and 16 people make a collision more likely than not. Chapter 3 places the five phases on a ring modulo 5, turning the body-use rule into one subtraction; the verdict class carries 2.318 bits, nearly a fair five-sided die, and 0.7005 of figures carry signals of both classes. Chapter 4 adds seasonal strength and the nuclear and changed hexagrams; with five signals, 0.8609 of figures support both conclusions. Chapter 5 shows that the associations of the Shuo Gua are partitions, so a reading covers more than half the trigrams, a chance match along one dimension has probability 0.5108, even the strictest standard named in advance gives a match within 30 minutes with probability 0.9994, and attention capture inflates matches further. Chapter 6 prices fits found after the fact by their likelihood ratios and simulates how selective memory turns even odds into a remembered fit rate of 0.795. Chapter 7 recomputes the plum-viewing case, finds the same result on 382 days, and reads Wang Chong's chapters on divination and ghosts beside Bacon's votive tablets. Chapter 8 gives owners a self-audit list, a persuasion-copy checker with eight rules and a pre-registered testing plan (the blind match needs 325 participants). Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:托名邵雍的《梅花易数》用年、月、日、时的数字相加取余起卦,再按体用生克判定,并以随手所见的外应相印证。本书不讨论卦象的形而上意义,只把这套流程写成可以运行的算术:起卦是一个哈希函数,体用是一台判定机,外应是一个基率问题。书中的人群分布、注意率与记忆率均为示意;全部计算由书中程序生成;本书分析结构与概率,不做任何个人占断,也不构成商业、法律或理财建议。 第一章证明,51,840 个年月日时组合只打到 384 个设计槽位中的 192 个,因为下卦与动爻由同一个总和分别除以 8 与 6 得出,奇偶被锁在一起;输出熵只有 7.499 比特,一天之内只有 12 种结果。第二章计算碰撞:同一天随机起卦的两人撞卦概率是 0.0833,同一时辰的人群全部撞卦;全年范围内,实际规则的两人撞卦概率是理想哈希的 2.232 倍,16 个人就足以让撞卦更可能发生。第三章把五行放在模 5 的环上,体用生克变成一次减法;判定类别的熵是 2.318 比特,几乎是一个均匀的五面骰子,而 0.7005 的卦同时含有正反两类信号。第四章加入月令旺衰与互卦、变卦,五个信号中两种结论都有依据的卦占 0.8609。 第五章说明《说卦》的类象是一组分划,一次解读平均覆盖一半以上的卦,一个维度上随手对上的概率是 0.5108,按最严格的事前指定标准,30 分钟内对上的概率也有 0.9994,注意捕获还会进一步放大匹配。第六章用似然比为「事后对上」定价,并模拟选择性记忆如何把一半对一半的运气变成 0.795 的回忆相符率。第七章复算观梅占,发现同一结果出现在 382 天上,并把王充〈卜筮〉〈订鬼〉与培根的还愿画板对读。第八章给出经营者的自查清单、八条规则的攻心文案检查器与预注册检验方案(盲法匹配需要 325 人)。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Date-Time Casting as a Hash: Only Half of 384 Slots Are Ever Hit — Chapter 1 proves that the 51,840 combinations of year, month, day and hour reach only 192 of 384 designed slots, because the lower trigram and the moving line come from one total divided by 8 and by 6 and so share a parity; the output entropy is only 7.499 bits, and a day has only 12 results. (https://5dlibrary.com/answers/en/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c01/) 第1章 年月日时是一个哈希:384 个槽位只命中一半 — 第一章证明,51,840 个年月日时组合只打到 384 个设计槽位中的 192 个,因为下卦与动爻由同一个总和分别除以 8 与 6 得出,奇偶被锁在一起;输出熵只有 7.499 比特,一天之内只有 12 种结果。 (https://5dlibrary.com/answers/zh/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c01/) - Ch.2 Same Double-Hour, Same Hexagram: the Birthday Problem and Crowd Degeneracy — Chapter 2 computes collisions: two people casting at random on the same day share a figure with probability 0.0833, and a crowd in one double-hour shares it entirely; across a year the actual rule makes pairs collide 2.232 times as often as an ideal hash, and 16 people make a collision more likely than not. (https://5dlibrary.com/answers/en/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c02/) 第2章 同一时辰,同一卦:生日问题与人群退化 — 第二章计算碰撞:同一天随机起卦的两人撞卦概率是 0.0833,同一时辰的人群全部撞卦;全年范围内,实际规则的两人撞卦概率是理想哈希的 2.232 倍,16 个人就足以让撞卦更可能发生。 (https://5dlibrary.com/answers/zh/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c02/) - Ch.3 The Body-Use Rule Is Subtraction Modulo 5: a Five-State Automaton and Its Verdicts — Chapter 3 places the five phases on a ring modulo 5, turning the body-use rule into one subtraction; the verdict class carries 2.318 bits, nearly a fair five-sided die, and 0.7005 of figures carry signals of both classes. (https://5dlibrary.com/answers/en/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c03/) 第3章 体用生克是模 5 减法:五状态自动机与判定分布 — 第三章把五行放在模 5 的环上,体用生克变成一次减法;判定类别的熵是 2.318 比特,几乎是一个均匀的五面骰子,而 0.7005 的卦同时含有正反两类信号。 (https://5dlibrary.com/answers/zh/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c03/) - Ch.4 Forking Paths through Season, Nucleus and Change: How Often One Figure Supports Both Verdicts — Chapter 4 adds seasonal strength and the nuclear and changed hexagrams; with five signals, 0.8609 of figures support both conclusions. (https://5dlibrary.com/answers/en/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c04/) 第4章 旺衰与互变的分叉:一卦能讲出两种结论的比例 — 第四章加入月令旺衰与互卦、变卦,五个信号中两种结论都有依据的卦占 0.8609。 (https://5dlibrary.com/answers/zh/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c04/) - Ch.5 Base Rates of the External Omen: Partitions of Symbols, a Poisson Clock and Attention Capture — Chapter 5 shows that the associations of the Shuo Gua are partitions, so a reading covers more than half the trigrams, a chance match along one dimension has probability 0.5108, even the strictest standard named in advance gives a match within 30 minutes with probability 0.9994, and attention capture inflates matches further. (https://5dlibrary.com/answers/en/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c05/) 第5章 外应的基率:类象分划、泊松时钟与注意捕获 — 第五章说明《说卦》的类象是一组分划,一次解读平均覆盖一半以上的卦,一个维度上随手对上的概率是 0.5108,按最严格的事前指定标准,30 分钟内对上的概率也有 0.9994,注意捕获还会进一步放大匹配。 (https://5dlibrary.com/answers/zh/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c05/) - Ch.6 What a Fit Found After the Fact Is Worth: Likelihood Ratios, Selective Memory and Barnum Sentences — Chapter 6 prices fits found after the fact by their likelihood ratios and simulates how selective memory turns even odds into a remembered fit rate of 0.795. (https://5dlibrary.com/answers/en/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c06/) 第6章 事后对上值多少证据:似然比、选择性记忆与巴纳姆句式 — 第六章用似然比为「事后对上」定价,并模拟选择性记忆如何把一半对一半的运气变成 0.795 的回忆相符率。 (https://5dlibrary.com/answers/zh/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c06/) - Ch.7 The Sources Examined: the Plum-Viewing Case, Wang Chong on Divination and Ghosts, and Bacon's Votive Tablets — Chapter 7 recomputes the plum-viewing case, finds the same result on 382 days, and reads Wang Chong's chapters on divination and ghosts beside Bacon's votive tablets. (https://5dlibrary.com/answers/en/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c07/) 第7章 原典考辨:观梅占、王充〈卜筮〉〈订鬼〉与培根的还愿画板 — 第七章复算观梅占,发现同一结果出现在 382 天上,并把王充〈卜筮〉〈订鬼〉与培根的还愿画板对读。 (https://5dlibrary.com/answers/zh/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c07/) - Ch.8 Practice and Refutation: a Copy Checker, Pre-Registered Tests and the Conditions That Would Overturn This Book — Chapter 8 gives owners a self-audit list, a persuasion-copy checker with eight rules and a pre-registered testing plan (the blind match needs 325 participants). Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c08/) 第8章 实践与证伪方案:文案检查器、预注册检验与推翻条件 — 第八章给出经营者的自查清单、八条规则的攻心文案检查器与预注册检验方案(盲法匹配需要 325 人)。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/plum-blossom-numerology-body-use-automaton-hash-collisions-date/c08/) ## 5DL-B-157 · Tieban Shenshu Ke-Fen Decision Trees, Affine Coding of the Verse Bank and Entropy Lower Bounds on Elicitation 中文书名:铁板神数考刻分决策树、条文库仿射编码与信息熵套取下界 DOI: pending Cite: Fan, Y. (2026). Tieban Shenshu Ke-Fen Decision Trees, Affine Coding of the Verse Bank and Entropy Lower Bounds on Elicitation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it analyses structure and probability, makes no reading of any individual, and is not legal or financial advice. Chapter 5 shows that a verse carries at most 0.5307 bits of expected information and that at an illustrative scale 19.01 people are born each minute. Abstract: Tieban Shenshu is famed for precision: it is said to cut the double-hour into ke and fen and to retrieve, from a bank of tens of thousands of verses, the verses naming a client's parents' birth animals and number of brothers. Using Shannon's information theory and elementary number theory, this book asks a question that can be calculated: how many bits does it take to locate one verse, and where do those bits come from? Lineages disagree about the bank size and the subdivision of the clock; the book adopts 8 ke to the double-hour and 15 fen to the ke, and takes a bank of 120,000 verses as its main case. Its parameters are illustrative; all calculations are generated by its programs; it analyses structure and probability, makes no reading of any individual, and is not legal or financial advice. Chapter 1 finds that locating one of 120,000 verses takes 16.8727 bits, that 13.2877 bits are still missing once the double-hour is known, and that a four-question checking questionnaire already has a raw capacity of 11.340 bits. Chapter 2 writes kaoke as a decision tree: twelve-way questions fix one of 120 fen in 1.9833 questions on average with a Huffman tree. Chapter 3 studies affine coding of verse numbers: with modulus 12,000 the invertible keys are worth only 25.195 bits, an average of 3.288 known pairs breaks the table, and no one-to-one code adds information. Chapter 4 is the core: searching the 120 fen of a double-hour, the parents' animals match exactly with probability 0.6975, and a match that looks like a rare event of 6.451 bits carries only 0.520 bits of real evidence. Chapter 5 shows that a verse carries at most 0.5307 bits of expected information and that at an illustrative scale 19.01 people are born each minute. Chapter 6 computes year-boundary errors in birth animals, with about 0.0911 of people misreporting by Gregorian year, and gives lower bounds from the liar game and Fano's inequality. Chapter 7 examines the 17,024 syllable positions of the Sound-and-Tone chart in Shao Yong's Huangji jingshi and Leibniz's binary. Chapter 8 gives a pre-registered blind test, needing 65 clients if the true hit rate is 0.10, and distils the results into a script skeleton and red lines for service-desk intake. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:铁板神数以「精确」著称:据说它把时辰切成刻与分,再从一部数以万计的条文库中检索出写有父母属相、兄弟人数的条文。本书用香农信息论与初等数论追问一个可以计算的问题:定位一条条文需要多少比特,这些比特从哪里来。条文总数与刻分制度各家说法不一,本书采用一时辰 8 刻、一刻 15 分之说,条文库取 120,000 条为主算例。书中参数为示意;全部计算由书中程序生成;本书分析结构与概率,不对任何个人作任何占断,也不构成法律或理财建议。 第一章算出定位 120,000 条需要 16.8727 比特,已知时辰后还差 13.2877 比特,而一份四问的考刻问卷原始容量已有 11.340 比特。第二章把考刻写成决策树:十二选一的问题确定 120 分,哈夫曼树平均只要 1.9833 问。第三章研究条文编号的仿射映射:模数 12,000 时可逆钥匙只值 25.195 比特,平均 3.288 对已知样本即可破解,任何一一对应的编码都不增加信息。第四章是全书核心:在一个时辰的 120 个分中搜索,父母属相完全对上的概率为 0.6975,表面上 6.451 比特的罕事,真实证据只有 0.520 比特。 第五章说明一句条文的期望信息量至多 0.5307 比特,按示意规模每分钟有 19.01 人出生。第六章计算属相的年界误差,约 0.0911 的人按公历自报属相会出错,并用说谎者游戏与费诺不等式给出下界。第七章考辨邵雍《皇极经世》声音唱和图的 17,024 个音节位置与莱布尼茨的二进制。第八章给出预注册的盲法检验,真实命中率为 0.10 时需 65 名求测者,并把结论蒸馏为一套客服问诊的话术骨架与红线。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Bit Budget of a Verse Bank: What the Double-Hour, the Ke, the Fen and the Bank Are Worth — Chapter 1 finds that locating one of 120,000 verses takes 16.8727 bits, that 13.2877 bits are still missing once the double-hour is known, and that a four-question checking questionnaire already has a raw capacity of 11.340 bits. (https://5dlibrary.com/answers/en/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c01/) 第1章 条文库的比特账:时辰、刻、分与条文总数各值几比特 — 第一章算出定位 120,000 条需要 16.8727 比特,已知时辰后还差 13.2877 比特,而一份四问的考刻问卷原始容量已有 11.340 比特。 (https://5dlibrary.com/answers/zh/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c01/) - Ch.2 The Kaoke Decision Tree: Entropy Bounds on the Mean Number of Questions and Huffman Optimality — Chapter 2 writes kaoke as a decision tree: twelve-way questions fix one of 120 fen in 1.9833 questions on average with a Huffman tree. (https://5dlibrary.com/answers/en/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c02/) 第2章 考刻决策树:平均问答次数的熵下界与哈夫曼最优 — 第二章把考刻写成决策树:十二选一的问题确定 120 分,哈夫曼树平均只要 1.9833 问。 (https://5dlibrary.com/answers/zh/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c02/) - Ch.3 An Affine Cipher for Verse Numbers: Invertibility, Key Space and Known-Pair Attacks — Chapter 3 studies affine coding of verse numbers: with modulus 12,000 the invertible keys are worth only 25.195 bits, an average of 3.288 known pairs breaks the table, and no one-to-one code adds information. (https://5dlibrary.com/answers/en/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c03/) 第3章 条文编号的仿射密码:可逆条件、密钥空间与已知样本破解 — 第三章研究条文编号的仿射映射:模数 12,000 时可逆钥匙只值 25.195 比特,平均 3.288 对已知样本即可破解,任何一一对应的编码都不增加信息。 (https://5dlibrary.com/answers/zh/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c03/) - Ch.4 Questions Are Leaks: The Cold Reader's Entropy Budget and Why a Match Is Nearly Certain — Chapter 4 is the core: searching the 120 fen of a double-hour, the parents' animals match exactly with probability 0.6975, and a match that looks like a rare event of 6.451 bits carries only 0.520 bits of real evidence. (https://5dlibrary.com/answers/en/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c04/) 第4章 问答即泄露:冷读者的熵预算与「对上」为何几乎必然 — 第四章是全书核心:在一个时辰的 120 个分中搜索,父母属相完全对上的概率为 0.6975,表面上 6.451 比特的罕事,真实证据只有 0.520 比特。 (https://5dlibrary.com/answers/zh/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c04/) - Ch.5 Verse Specificity and Clock Resolution: Barnum Statements, Same-Minute Births and Twins — Chapter 5 shows that a verse carries at most 0.5307 bits of expected information and that at an illustrative scale 19.01 people are born each minute. (https://5dlibrary.com/answers/en/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c05/) 第5章 条文的特异度与刻分的分辨率:巴纳姆句、同分钟出生者与双胞胎 — 第五章说明一句条文的期望信息量至多 0.5307 比特,按示意规模每分钟有 19.01 人出生。 (https://5dlibrary.com/answers/zh/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c05/) - Ch.6 Noisy Answers: The Liar Game, Year-Boundary Errors in Birth Animals, and Fano's Inequality — Chapter 6 computes year-boundary errors in birth animals, with about 0.0911 of people misreporting by Gregorian year, and gives lower bounds from the liar game and Fano's inequality. (https://5dlibrary.com/answers/en/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c06/) 第6章 有噪问答:说谎者游戏、属相的年界误差与费诺不等式 — 第六章计算属相的年界误差,约 0.0911 的人按公历自报属相会出错,并用说谎者游戏与费诺不等式给出下界。 (https://5dlibrary.com/answers/zh/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c06/) - Ch.7 The Classical Sources: The Combinatorics of the Sound-and-Tone Chart and Leibniz's Binary — Chapter 7 examines the 17,024 syllable positions of the Sound-and-Tone chart in Shao Yong's Huangji jingshi and Leibniz's binary. (https://5dlibrary.com/answers/en/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c07/) 第7章 原典考辨:声音唱和图的组合数与莱布尼茨的二进制 — 第七章考辨邵雍《皇极经世》声音唱和图的 17,024 个音节位置与莱布尼茨的二进制。 (https://5dlibrary.com/answers/zh/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c07/) - Ch.8 Practice and Refutation: A Bit Budget for Intake, Red Lines for Scripts, and a Pre-Registered Test — Chapter 8 gives a pre-registered blind test, needing 65 clients if the true hit rate is 0.10, and distils the results into a script skeleton and red lines for service-desk intake. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c08/) 第8章 实践与证伪方案:问诊的比特预算、话术红线与预注册检验 — 第八章给出预注册的盲法检验,真实命中率为 0.10 时需 65 名求测者,并把结论蒸馏为一套客服问诊的话术骨架与红线。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/tieban-shenshu-ke-fen-decision-trees-affine-coding-verse/c08/) ## 5DL-B-158 · Turing Reaction-Diffusion Patterns in Palm Dermatoglyphics, Flexion-Crease Topology and Embryonic Stress Inversion 中文书名:手掌皮纹图灵反应扩散斑图、屈肌折痕拓扑分类与胚胎期应力反演 DOI: pending Cite: Fan, Y. (2026). Turing Reaction-Diffusion Patterns in Palm Dermatoglyphics, Flexion-Crease Topology and Embryonic Stress Inversion (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: It analyses structure and probability, makes no personal judgment from anyone's hand, and is not medical, legal or business advice. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: The hand carries two kinds of line: the fine epidermal ridges of fingertips and palm, and the thick flexion creases of the palm. Physiognomy reads them as a finished document; this book asks three questions that can be computed. How do the lines grow? How are they distributed in a population, and are they independent? Can an adult fingerprint tell us what happened in the embryo? Its parameters are illustrative, its tabulated counts synthetic, and all calculations are generated by its programs. It analyses structure and probability, makes no personal judgment from anyone's hand, and is not medical, legal or business advice. Chapter 1 finds the Turing instability of the Schnakenberg system by linear stability analysis: the critical diffusion ratio is 8.5676 and the fastest-growing wavelength 12.910, and under one law and one length the number of ridges still varies with the initial noise. In Chapter 2 two-dimensional finite differences show that directionality in one coefficient turns spots into parallel ridges, while minutiae are set by noise. Chapter 3 uses the Poincaré index to prove that a loop has one core and one delta and a whorl two deltas; of random direction fields only 0.462 meet the fingertip boundary condition. Chapter 4 analyses the single transverse crease against pattern type with chi-square and exact tests and shows that two traits independent within strata can show a pooled odds ratio of 1.393. Chapter 5 writes the timing of pad regression and pattern type as an inverse problem whose information matrix has an eigenvalue of exactly zero: pattern type pins down only pad height at ridge onset. Chapter 6 shows that ridge directions give only the shape of the deviatoric stress, and that running growth backwards needs regularisation that does not come from the data. Chapter 7 reads the Mayi Shenxiang beside Galton's Finger Prints: correlation between fingers makes ten whorls 11176 times likelier, and Galton's chance of accidental agreement reaches even odds at 308,651 fingers. Chapter 8 gives a reaction-diffusion texture style library, three business rules and a pre-registered protocol whose main test needs 3,690 people. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:手上有两套纹:指尖与掌面细密的表皮嵴,以及掌心粗深的屈肌折痕。相学把它们当作写好的文书来读,本书则问三个可以计算的问题:这些纹怎样长出来?它们在人群中如何分布、彼此是否独立?从一枚成人指纹能否倒推胚胎期发生了什么?书中参数为示意,表格计数为合成数据,全部计算由书中程序生成;本书分析结构与概率,不对任何人的手作个人判断,也不是医疗、法律或经营建议。 第一章用线性稳定性分析求施纳肯贝格系统的图灵失稳:临界扩散比为 8.5676,最快增长波长为 12.910,同一定律、同一长度下嵴数仍随初始噪声变化。第二章的二维有限差分模拟显示,一个系数的方向性就能把斑点变成平行嵴,细节点则由噪声决定。第三章用庞加莱指数证明箕纹一核一三角、斗纹两个三角,随机方向场中只有 0.462 满足指尖边界条件。第四章用卡方与精确检验分析通贯掌与纹型的关联,并说明分层内独立的两种特征在合并后可以出现 1.393 的比值比。 第五章把指垫消退时机与纹型写成反演问题,信息矩阵有一个精确为零的特征值,纹型只钉住成嵴时的指垫高度。第六章说明嵴线走向只给出应力场的偏应力形状,倒放生长需要不来自数据的正则化。第七章对读《麻衣神相》与高尔顿的《指纹》:指间相关使十斗的机会增加 11176 倍,高尔顿的偶然一致概率在 308,651 个手指时已是一半机会。第八章给出反应扩散纹理样式库、三条经营规则与预注册方案,主检验需要 3,690 人。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Turing Instability: Linear Stability Sets the Scale of Ridge Spacing — Chapter 1 finds the Turing instability of the Schnakenberg system by linear stability analysis: the critical diffusion ratio is 8.5676 and the fastest-growing wavelength 12.910, and under one law and one length the number of ridges still varies with the initial noise. In Chapter 2 two-dimensional finite differences show that directionality in one coefficient turns spots into parallel ridges, while minutiae are set by noise. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c01/) 第1章 图灵失稳:线性稳定性分析给出嵴距的尺度 — 第一章用线性稳定性分析求施纳肯贝格系统的图灵失稳:临界扩散比为 8.5676,最快增长波长为 12.910,同一定律、同一长度下嵴数仍随初始噪声变化。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c01/) - Ch.2 From Spots to Parallel Ridges: Two-Dimensional Finite Differences, Anisotropy and Minutiae as Defects — The default two-dimensional Turing pattern is a field of spots, and directionality in a single coefficient is enough to turn it into parallel ridges that run along the direction of slower spreading. Ridge spacing is set by the law and, under the whole-number constraint, jumps around the fastest wavelength; minutiae are set by noise, ranging from 0 to 8 in six runs under one law. Palmistry treats every break and join in the lines as information. This chapter shows that the breaks are first of all the product of a random process. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c02/) 第2章 从斑点到平行嵴:二维有限差分、各向异性与细节点缺陷 — 二维图灵斑图默认是斑点,一个系数的方向性就能把它变成平行嵴,嵴沿扩散较慢的方向延伸。嵴距由定律决定,在整数约束下围绕最快波长跳动;细节点由噪声决定,同一定律下的六次模拟从 0 到 8 个不等。相学把纹线的每一处断续当成信息,而本章说明,断续首先是一个随机过程的产物。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c02/) - Ch.3 Singularities of the Ridge Field and the Poincaré Index: One Core and One Delta for a Loop, Two Deltas for a Whorl — Chapter 3 uses the Poincaré index to prove that a loop has one core and one delta and a whorl two deltas; of random direction fields only 0.462 meet the fingertip boundary condition. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c03/) 第3章 方向场奇点与庞加莱指数:箕纹一核一三角,斗纹两个三角 — 第三章用庞加莱指数证明箕纹一核一三角、斗纹两个三角,随机方向场中只有 0.462 满足指尖边界条件。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c03/) - Ch.4 Flexion Creases: Population Frequency of the Single Transverse Crease, Chi-Square Independence and Common Causes — Chapter 4 analyses the single transverse crease against pattern type with chi-square and exact tests and shows that two traits independent within strata can show a pooled odds ratio of 1.393. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c04/) 第4章 屈肌折痕:通贯掌的人群频率、卡方独立性与共同原因 — 第四章用卡方与精确检验分析通贯掌与纹型的关联,并说明分层内独立的两种特征在合并后可以出现 1.393 的比值比。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c04/) - Ch.5 Timing of Volar Pad Regression and Pattern Type: an Inverse Problem without a Unique Solution — The forward map squeezes three parameters, peak height, regression rate and onset week, into one combined quantity, and pattern type reads only that quantity. The Bayesian inversion pins pad height near 0.658, but the onset week may lie anywhere from 10.05 to 13.25 depending on the assumed rate of regression. The information matrix of ten fingers still has an eigenvalue of exactly zero. A fingerprint is a record of the embryo, but what it records is a product, not a timetable. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c05/) 第5章 指垫消退时机与纹型:一个不能唯一求解的反演问题 — 正向映射把(峰值高度、消退速率、成嵴周数)三个参数压成一个组合量,纹型只读这个组合量。贝叶斯反演把指垫高度钉在 0.658 附近,成嵴周数却可以落在 10.05 到 13.25 之间,取决于假定的消退速率。十个手指的信息矩阵仍有一个精确为零的特征值。指纹是胚胎期的记录,但它记录的是一个乘积,而不是一份时间表。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c05/) - Ch.6 Inverting for Embryonic Stress: the Direction Field Gives Only Half the Information — Ridge angles give homogeneous linear equations in the stress components. For a linear stress field with 9 parameters the null space has 4 dimensions: three pressures and one scale with its sign. The shape of the deviatoric stress can be recovered, with a mean error of 2.05 degrees, but the data cannot say whether ridges run perpendicular or parallel to the greatest compression. Cores and deltas correspond to isotropic points where the deviatoric stress vanishes, the most robust part of the inversion. Running later growth backwards is unstable: the condition number is 3.622e+05, the naive inversion has an error of 9.179e+02, and the best regularised error, 0.0548, depends on a choice of smoothness that does not come from the data. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c06/) 第6章 胚胎期应力反演:方向场只给出一半信息 — 嵴线角度对应力分量给出线性齐次方程。在一个 9 参数的线性应力场中,零空间是 4 维的:三个压强、一个尺度与符号。偏应力的形状可以恢复,平均误差 2.05 度,但嵴线垂直还是平行于最大压应力,数据无法判断。核心与三角对应偏应力为零的各向同性点,是反演中最稳健的部分。把后来的生长倒放回去是不稳定的:条件数 3.622e+05,朴素反演误差 9.179e+02,正则化后最好的误差 0.0548 依赖一个不来自数据的光滑性选择。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c06/) - Ch.7 The Classics Examined: the Mayi Shenxiang and Galton's Finger Prints Read Side by Side — Chapter 7 reads the Mayi Shenxiang beside Galton's Finger Prints: correlation between fingers makes ten whorls 11176 times likelier, and Galton's chance of accidental agreement reaches even odds at 308,651 fingers. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c07/) 第7章 原典考辨:《麻衣神相》与高尔顿《指纹》对读 — 第七章对读《麻衣神相》与高尔顿的《指纹》:指间相关使十斗的机会增加 11176 倍,高尔顿的偶然一致概率在 308,651 个手指时已是一半机会。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c07/) - Ch.8 Practice and Falsification: a Texture Style Library, a Pre-Registered Test and the Conditions of Refutation — Chapter 8 gives a reaction-diffusion texture style library, three business rules and a pre-registered protocol whose main test needs 3,690 people. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c08/) 第8章 实践与证伪方案:纹理样式库、预注册检验与推翻条件 — 第八章给出反应扩散纹理样式库、三条经营规则与预注册方案,主检验需要 3,690 人。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-patterns-palm-dermatoglyphics-flexion-crease-topology/c08/) ## 5DL-B-159 · The Unclosed Circle of Pythagorean Tuning, Zhu Zaiyu's Root Approximation of Equal Temperament and the Beat Roughness of Just-Intonation Overtones 中文书名:三分损益律旋宫不归、朱载堉十二平均律开方逼近与纯律泛音拍频粗糙度 DOI: pending Cite: Fan, Y. (2026). The Unclosed Circle of Pythagorean Tuning, Zhu Zaiyu's Root Approximation of Equal Temperament and the Beat Roughness of Just-Intonation Overtones (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its listeners, samples and copy are illustrative, and all calculations are generated by its programs; the book is not medical advice, claims no effect of any frequency, scale or instrument on body or mind, and is not purchasing advice. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: A chain of pure fifths generated by adding and cutting a third does not return to huangzhong after twelve steps; it is the oldest problem in tuning theory. This book writes it, and several popular claims around it, as runnable arithmetic: how large the remainder is and why it can never be zero; why Jing Fang's sixty pipes and Qian Lezhi's three hundred and sixty sometimes helped and sometimes did not; whether Zhu Zaiyu's twenty-five-digit close ratio was right and whether it was needed; whether the major thirds of three tunings become beats or roughness in the ear; and what moving standard pitch to 432 Hz really changes. Its listeners, samples and copy are illustrative, and all calculations are generated by its programs; the book is not medical advice, claims no effect of any frequency, scale or instrument on body or mind, and is not purchasing advice. Chapter 1 computes in exact fractions the ratio of huangzhong to the returning pipe, 3^12/219 or 23.46 cents, and shows that the Guanzi's start of eighty-one is the smallest that keeps whole numbers, twelve pipes needing 177,147. Chapter 2 uses continued fractions to show that records of closure fall only at 5, 12, 41, 53 and 306 steps, that Jing Fang's seyu misses by only 3.6150 cents, and that by the three-gap theorem a tuning only ever has two or three step sizes. Chapter 3 replays Zhu Zaiyu's three root extractions digit by digit and confirms that all twenty-five digits of the close ratio usually quoted, 1.059463094359295264561825, are correct, while an octave error below one cent needs only six digits. Chapter 4 compares the three tunings: the equal-tempered major third is 13.686 cents wider than just, and the schisma third eight steps down the chain of fifths is only 1.954 cents off. Chapter 5 computes dissonance with the Plomp-Levelt curve and the Sethares model, finds the equal-tempered major third on middle C 1.151 times as rough as the just one, and cites experiments showing that preference for consonance follows harmonicity more than roughness. Chapter 6 shows that the same mistuning is a slow beat in the low register and roughness in the high, and that 432 Hz merely lowers everything by 31.77 cents, with modelled dissonance rising slightly rather than falling, by no more than 3.08 %, far less than the 15.15 % made by the choice of tuning. Chapter 7 examines the Diyuan chapter of the Guanzi, Zhu Zaiyu's Lvlv jingyi and the blacksmith legend in Boethius, showing that by the law of strings hammer weights of twelve to six give only 600.00 cents. Chapter 8 gives a checklist, copy-checking rules and a pre-registered test with a sample of 169. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:三分损益的律链走完十二步回不到黄钟,这是律学最古老的难题。本书把它和相关的几个流行说法写成可以运行的算术:那一点余数有多大,为什么不可能为零;京房六十律、钱乐之三百六十律为什么时好时坏;朱载堉的二十五位密率对不对、要不要;三种律制的大三度在耳朵里变成拍还是粗糙;把标准音改为 432 Hz 又改变了什么。书中的听者、样本与文案均为示意,全部计算由书中程序生成;本书不是医疗建议,不宣称任何频率、音阶或乐器有身心功效,也不构成购买建议。 第一章用精确分数算出返生之律与黄钟之比 3^12/219,即 23.46 音分,并说明《管子》以八十一起算是保持整数的最小选择,十二律需要 177,147。第二章用连分数说明闭合纪录只落在 5、12、41、53、306 步上,京房的色育只差 3.6150 音分,而三间隙定理使律制内部永远只有两三种步长。第三章逐位重演朱载堉的三次开方,确认通行文献所载密率 1.059463094359295264561825 的二十五位全部正确,而让八度误差小于一音分只需六位数字。第四章比较三种律制,平均律大三度比纯律宽 13.686 音分,五度链向下八步的施斯玛三度只差 1.954 音分。 第五章以 Plomp–Levelt 曲线与 Sethares 模型计算不协和度,平均律大三度在中央 C 上是纯律的 1.151 倍,并引用实验指出协和偏好主要随谐和性而非粗糙度变化。第六章说明同一失谐在低音区是慢拍、在高音区是粗糙,改用 432 Hz 只是整体降低 31.77 音分,模型中的不协和度略升不降,变化不超过 3.08 %,远小于律制选择造成的 15.15 %。第七章考辨《管子·地员》、朱载堉《律吕精义》与波爱修斯的铁匠传说,指出按弦的规律十二比六的锤重只得 600.00 音分。第八章给出执行清单、文案检查规则与样本量为 169 的预先登记检验。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Twelve Steps of Adding and Cutting a Third: the Exact Arithmetic of the Pythagorean Comma — Chapter 1 computes in exact fractions the ratio of huangzhong to the returning pipe, 3^12/219 or 23.46 cents, and shows that the Guanzi's start of eighty-one is the smallest that keeps whole numbers, twelve pipes needing 177,147. (https://5dlibrary.com/answers/en/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c01/) 第1章 三分损益十二步:毕达哥拉斯音差的精确算术 — 第一章用精确分数算出返生之律与黄钟之比 3^12/219,即 23.46 音分,并说明《管子》以八十一起算是保持整数的最小选择,十二律需要 177,147。 (https://5dlibrary.com/answers/zh/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c01/) - Ch.2 Jing Fang's Sixty Pipes and Qian Lezhi's Three Hundred and Sixty: Why Continued Fractions Favour 12, 41 and 53 — Chapter 2 uses continued fractions to show that records of closure fall only at 5, 12, 41, 53 and 306 steps, that Jing Fang's seyu misses by only 3.6150 cents, and that by the three-gap theorem a tuning only ever has two or three step sizes. (https://5dlibrary.com/answers/en/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c02/) 第2章 京房六十律与钱乐之三百六十律:连分数为何偏爱 12、41、53 — 第二章用连分数说明闭合纪录只落在 5、12、41、53、306 步上,京房的色育只差 3.6150 音分,而三间隙定理使律制内部永远只有两三种步长。 (https://5dlibrary.com/answers/zh/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c02/) - Ch.3 Zhu Zaiyu's Close Ratio: Digit-by-Digit Roots, a Twenty-Five-Digit Recomputation and the Marginal Value of Precision — Chapter 3 replays Zhu Zaiyu's three root extractions digit by digit and confirms that all twenty-five digits of the close ratio usually quoted, 1.059463094359295264561825, are correct, while an octave error below one cent needs only six digits. (https://5dlibrary.com/answers/en/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c03/) 第3章 朱载堉的密率:逐位开方、二十五位复算与精度的边际价值 — 第三章逐位重演朱载堉的三次开方,确认通行文献所载密率 1.059463094359295264561825 的二十五位全部正确,而让八度误差小于一音分只需六位数字。 (https://5dlibrary.com/answers/zh/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c03/) - Ch.4 Three Ledgers for the Major Third: Just 5/4, Pythagorean 81/64 and Equal Temperament — Chapter 4 compares the three tunings: the equal-tempered major third is 13.686 cents wider than just, and the schisma third eight steps down the chain of fifths is only 1.954 cents off. (https://5dlibrary.com/answers/en/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c04/) 第4章 大三度的三本账:纯律 5/4、五度律 81/64 与平均律 — 第四章比较三种律制,平均律大三度比纯律宽 13.686 音分,五度链向下八步的施斯玛三度只差 1.954 音分。 (https://5dlibrary.com/answers/zh/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c04/) - Ch.5 The Roughness Curve: the Plomp and Levelt Data and the Sethares Dissonance Model — Chapter 5 computes dissonance with the Plomp-Levelt curve and the Sethares model, finds the equal-tempered major third on middle C 1.151 times as rough as the just one, and cites experiments showing that preference for consonance follows harmonicity more than roughness. (https://5dlibrary.com/answers/en/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c05/) 第5章 粗糙度曲线:Plomp–Levelt 与 Sethares 不协和度模型 — 第五章以 Plomp–Levelt 曲线与 Sethares 模型计算不协和度,平均律大三度在中央 C 上是纯律的 1.151 倍,并引用实验指出协和偏好主要随谐和性而非粗糙度变化。 (https://5dlibrary.com/answers/zh/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c05/) - Ch.6 Beats and High-Register Roughness: the Same Mistuning Beats Low and Grates High — Chapter 6 shows that the same mistuning is a slow beat in the low register and roughness in the high, and that 432 Hz merely lowers everything by 31.77 cents, with modelled dissonance rising slightly rather than falling, by no more than 3.08 %, far less than the 15.15 % made by the choice of tuning. (https://5dlibrary.com/answers/en/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c06/) 第6章 拍频与高频粗糙:同一失谐在低音区拍、在高音区糙 — 第六章说明同一失谐在低音区是慢拍、在高音区是粗糙,改用 432 Hz 只是整体降低 31.77 音分,模型中的不协和度略升不降,变化不超过 3.08 %,远小于律制选择造成的 15.15 %。 (https://5dlibrary.com/answers/zh/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c06/) - Ch.7 The Sources Examined: the Diyuan Chapter of the Guanzi, Zhu Zaiyu's Lvlv jingyi and the Blacksmith Legend in Boethius — Chapter 7 examines the Diyuan chapter of the Guanzi, Zhu Zaiyu's Lvlv jingyi and the blacksmith legend in Boethius, showing that by the law of strings hammer weights of twelve to six give only 600.00 cents. (https://5dlibrary.com/answers/en/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c07/) 第7章 原典考辨:《管子·地员》、朱载堉《律吕精义》与波爱修斯的铁匠传说 — 第七章考辨《管子·地员》、朱载堉《律吕精义》与波爱修斯的铁匠传说,指出按弦的规律十二比六的锤重只得 600.00 音分。 (https://5dlibrary.com/answers/zh/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c07/) - Ch.8 Practice and Falsification: Fitting Marketing Copy with a Detector for Pseudo-Precise Tuning Claims — Chapter 8 gives a checklist, copy-checking rules and a pre-registered test with a sample of 169. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c08/) 第8章 实践与证伪:给营销文案装一个音律伪精确检测器 — 第八章给出执行清单、文案检查规则与样本量为 169 的预先登记检验。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/unclosed-circle-pythagorean-tuning-zhu-zaiyus-root-approximation-equal/c08/) ## 5DL-B-160 · Two-Cavity Vocal-Tract Coupling in Overtone Singing, Airflow Impedance Spectra of the Liuzijue Six Sounds and Oscillatory Gain in Exhaled Nitric Oxide 中文书名:双声泛音唱法声道双腔耦合、六字诀气流阻抗谱与呼出气一氧化氮震荡增益 DOI: pending Cite: Fan, Y. (2026). Two-Cavity Vocal-Tract Coupling in Overtone Singing, Airflow Impedance Spectra of the Liuzijue Six Sounds and Oscillatory Gain in Exhaled Nitric Oxide (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Most anatomical sizes, pressures and concentrations are illustrative; all calculations are generated by the book's programs; the book is not medical advice and claims no bodily benefit of any breathing method, singing style or vocal exercise. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: This book places overtone singing, the six exhalations of the Liuzijue and nasal nitric oxide during humming inside one physical frame: the vocal tract is a set of coupled tubes and cavities, a mouth shape is an impedance in the airflow, and a sinus is a resonator ventilated through a small hole. Its central question is how the positive-pressure resistance of the slit between tongue tip and teeth changes, at the same time, the focus of an overtone and the flux of nitric oxide out of the sinuses. Most anatomical sizes, pressures and concentrations are illustrative; all calculations are generated by the book's programs; the book is not medical advice and claims no bodily benefit of any breathing method, singing style or vocal exercise. Chapter 1 derives the coupling split of formants in a two-tube tract and shows that the second and third formants come within one harmonic spacing only when the ratio of the narrowest cross-section to the pharynx falls below about 0.0140, whereas two lumped Helmholtz cavities cannot merge however they are tuned. Chapter 2 uses the source-filter model to show that merged formants can lift the twelfth harmonic 30.5 decibels above its neighbours. Chapter 3 computes Bernoulli and Poiseuille resistance and impedance spectra for the six Liuzijue mouth shapes: at the same driving pressure one breath on si lasts 13.8 times as long as on he. Chapter 4 shows that the oral positive pressure of a tongue-teeth slit is shared according to squared areas, and that voicing stops once the combined oral and nasal exit falls below 7.75 square millimetres. Chapter 5 uses a two-compartment model of sinus and nose to interpret the published finding that humming raises nasal nitric oxide about fifteen-fold: it is the instantaneous peak of a store being washed out, the second hum peaks at only 0.398 of the first, and recovery to nine tenths takes 13.7 minutes. The joint model of Chapter 6 finds that overtone focus and sinus flux conflict at one tongue gap: the unconstrained optimum is 0.755 millimetres with a gain of 4.48, while requiring focus moves it to 0.296 millimetres and lowers the gain to 3.26; Monte Carlo puts the median share of excess gain given up for focus at 0.373. Chapter 7 reads Tao Hongjing's six exhalations in the Yangxing yanming lu beside the resonators of Helmholtz's On the Sensations of Tone. Chapter 8 gives an operator's checklist, a persuasion-copy checker and a pre-registered test plan starting from 22 participants. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:本书把泛音唱法、六字诀与哼鸣时的鼻腔一氧化氮放进同一个物理框架:声道是彼此耦合的管与腔,口型是气流通路上的阻抗,鼻窦是靠一个小孔通气的共鸣器。核心问题是:舌尖与齿之间那道窄缝造成的正压阻力,怎样同时改变泛音的聚焦与鼻窦一氧化氮的通量。书中的解剖尺寸、压力与浓度多为示意参数;全部计算由书中程序生成;本书不是医疗建议,不宣称任何呼吸法、唱法或发声练习的身体功效。 第一章用两段管求出共振峰的耦合分裂,说明只有当最窄处与咽腔的截面比低于约 0.0140 时,第 2 与第 3 共振峰才能靠近到一个谐波间距以内,而两个集总亥姆霍兹腔无论怎样调谐都无法合并。第二章用源-滤波器模型说明,合并的共振峰可以使第 12 个谐波高出邻居 30.5 分贝。第三章计算六字诀六种口型的伯努利与泊肃叶流阻和阻抗谱:同样的驱动压力下,呬的一口气是呵的 13.8 倍长。第四章说明舌尖齿缝的口内正压按面积平方分配,口鼻出口总面积低于 7.75 平方毫米时发声停止。 第五章用鼻窦与鼻腔的两室模型解释已发表的「哼鸣使鼻腔一氧化氮升高约十五倍」:这是一个被冲刷的储存量的瞬时峰值,第二次哼鸣的峰值只有第一次的 0.398,恢复九成需要 13.7 分钟。第六章的联合模型发现泛音聚焦与鼻窦通量在同一间隙上冲突:无约束最优间隙 0.755 毫米、增益 4.48,要求聚焦后移到 0.296 毫米、增益降到 3.26;蒙特卡洛显示为聚焦放弃的超额增益中位数为 0.373。第七章对读陶弘景《养性延命录》的吐气六字与亥姆霍兹《论音觉》的共鸣器。第八章给出经营者清单、攻心文案检查器与样本量为 22 人起的预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Two Tubes and Two Helmholtz Cavities: Vocal-Tract Resonances and Coupling Splits — Chapter 1 derives the coupling split of formants in a two-tube tract and shows that the second and third formants come within one harmonic spacing only when the ratio of the narrowest cross-section to the pharynx falls below about 0.0140, whereas two lumped Helmholtz cavities cannot merge however they are tuned. (https://5dlibrary.com/answers/en/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c01/) 第1章 两段管与两个亥姆霍兹腔:声道共振与耦合分裂 — 第一章用两段管求出共振峰的耦合分裂,说明只有当最窄处与咽腔的截面比低于约 0.0140 时,第 2 与第 3 共振峰才能靠近到一个谐波间距以内,而两个集总亥姆霍兹腔无论怎样调谐都无法合并。 (https://5dlibrary.com/answers/zh/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c01/) - Ch.2 Source, Filter and Overtone Focus: the Gain of Every Harmonic — Chapter 2 uses the source-filter model to show that merged formants can lift the twelfth harmonic 30.5 decibels above its neighbours. (https://5dlibrary.com/answers/en/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c02/) 第2章 源-滤波器与泛音聚焦:每个谐波的增益 — 第二章用源-滤波器模型说明,合并的共振峰可以使第 12 个谐波高出邻居 30.5 分贝。 (https://5dlibrary.com/answers/zh/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c02/) - Ch.3 Airflow Impedance Spectra of the Six Liuzijue Mouth Shapes: Bernoulli, Poiseuille and Exhalation Time — Chapter 3 computes Bernoulli and Poiseuille resistance and impedance spectra for the six Liuzijue mouth shapes: at the same driving pressure one breath on si lasts 13.8 times as long as on he. (https://5dlibrary.com/answers/en/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c03/) 第3章 六字诀口型的气流阻抗谱:伯努利、泊肃叶与呼气时长 — 第三章计算六字诀六种口型的伯努利与泊肃叶流阻和阻抗谱:同样的驱动压力下,呬的一口气是呵的 13.8 倍长。 (https://5dlibrary.com/answers/zh/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c03/) - Ch.4 Positive Pressure behind the Tongue Tip: Pressure Division, Transglottal Pressure and Nasal Diversion — Chapter 4 shows that the oral positive pressure of a tongue-teeth slit is shared according to squared areas, and that voicing stops once the combined oral and nasal exit falls below 7.75 square millimetres. (https://5dlibrary.com/answers/en/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c04/) 第4章 舌尖齿缝的正压:分压、声门跨压与鼻腔分流 — 第四章说明舌尖齿缝的口内正压按面积平方分配,口鼻出口总面积低于 7.75 平方毫米时发声停止。 (https://5dlibrary.com/answers/zh/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c04/) - Ch.5 A Two-Compartment Model of Sinus Nitric Oxide: Peak, Decay and Recovery of the Humming Gain — Chapter 5 uses a two-compartment model of sinus and nose to interpret the published finding that humming raises nasal nitric oxide about fifteen-fold: it is the instantaneous peak of a store being washed out, the second hum peaks at only 0.398 of the first, and recovery to nine tenths takes 13.7 minutes. The joint model of Chapter 6 finds that overtone focus and sinus flux conflict at one tongue gap: the unconstrained optimum is 0.755 millimetres with a gain of 4.48, while requiring focus moves it to 0.296 millimetres and lowers the gain to 3.26; Monte Carlo puts the median share of excess gain given up for focus at 0.373. (https://5dlibrary.com/answers/en/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c05/) 第5章 鼻窦一氧化氮的两室模型:哼鸣振荡增益的峰值、衰减与恢复 — 第五章用鼻窦与鼻腔的两室模型解释已发表的「哼鸣使鼻腔一氧化氮升高约十五倍」:这是一个被冲刷的储存量的瞬时峰值,第二次哼鸣的峰值只有第一次的 0.398,恢复九成需要 13.7 分钟。 (https://5dlibrary.com/answers/zh/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c05/) - Ch.6 One Gap, Three Consequences: a Joint Model of Overtone Focus and Sinus Flux — In the joint model, the gap of the tongue-teeth slit fixes overtone focus, source strength and sinus flux at once. Focus needs a very small gap, and a small gap weakens the source through oral positive pressure, lowering nasal sound pressure and nitric oxide gain. With the example parameters the unconstrained optimum gap is 0.755 millimetres; requiring focus moves it to 0.296 millimetres and lowers the gain from 4.48 to 3.26. Across 2000 random parameter sets the conflict is everywhere, and its size is set mainly by how far the velum opens and by the glottal area. (https://5dlibrary.com/answers/en/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c06/) 第6章 一个间隙,三个后果:泛音聚焦与鼻窦通量的联合模型 — 在联合模型中,舌尖齿缝的间隙同时决定泛音聚焦、声源强度与鼻窦通量。聚焦需要很小的间隙,而小间隙通过口内正压削弱声源,使鼻腔声压与一氧化氮增益下降。示例参数下,增益的无约束最优在 0.755 毫米,要求聚焦后最优移到 0.296 毫米,增益从 4.48 降到 3.26。2000 组随机参数中,这一冲突普遍存在,其大小主要由软腭开放程度与声门面积决定。 (https://5dlibrary.com/answers/zh/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c06/) - Ch.7 The Classics Examined: the Six Exhalations of the Yangxing yanming lu and Helmholtz's Resonators — He, the most open shape, makes a jet about 19.5 millimetres across that is still inside its potential core at 10 centimetres; the air arrives at 33.7 degrees and only 0.83 metres per second, and the cooling index is negative (-0.7), so the hand feels warmth, which is why people blow on their hands this way to warm them. Chui makes a jet about 6.2 millimetres across that has entrained a great deal of room air by the time it reaches the hand: 25.7 degrees, 3.09 metres per second and a cooling index of 12.8. Hu makes a jet about 8.7 millimetres across, at 27.2 degrees with a cooling index of 8.5. Chui and hu both feel cool on the hand, chui cooler than hu; the warm one is he. (https://5dlibrary.com/answers/en/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c07/) 第7章 原典考辨:《养性延命录》的吐气六字与亥姆霍兹的共鸣器 — 开口最大的呵,射流直径约 19.5 毫米,在 10 厘米处仍处于射流的势核之内,温度 33.7 摄氏度,流速只有每秒 0.83 米,冷却指数为负(-0.7),手感到的是暖气,这就是人们对着手「哈气」取暖的物理原因。吹的射流直径约 6.2 毫米,到达手时已经卷入大量室内空气,温度降到 25.7 摄氏度,流速每秒 3.09 米,冷却指数 12.8;呼的射流直径约 8.7 毫米,温度 27.2 摄氏度,冷却指数 8.5。也就是说,吹与呼吹到手上都是凉的,吹比呼更凉;真正暖的是呵。 (https://5dlibrary.com/answers/zh/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c07/) - Ch.8 Practice and Falsification: an Operator's Checklist, a Copy Checker and a Pre-Registered Test — Chapter 8 gives an operator's checklist, a persuasion-copy checker and a pre-registered test plan starting from 22 participants. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c08/) 第8章 实践与证伪方案:经营者清单、文案检查器与预注册检验 — 第八章给出经营者清单、攻心文案检查器与样本量为 22 人起的预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/two-cavity-vocal-tract-coupling-overtone-singing-airflow-impedance/c08/) ## 5DL-B-161 · Rayleigh Thin-Shell Modes of Bronze Singing Bowls, Nonlinear Strike Beat Envelopes and Coupling to Somatic Mechanoreceptors 中文书名:青铜颂钵薄壳瑞利振动模态、非线性敲击差拍包络与低频躯体机械感受器耦合 DOI: pending Cite: Fan, Y. (2026). Rayleigh Thin-Shell Modes of Bronze Singing Bowls, Nonlinear Strike Beat Envelopes and Coupling to Somatic Mechanoreceptors (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 1 derives the ring ratio n(n² − 1)/√(n² + 1) from the Rayleigh quotient: (3,0) lies 2.828 times above (2,0), exactly an octave and a tritone, and the (2,0) mode of the illustrative bowl is at 207.6 Hz. Chapter 2 shows that only the 2n-th thickness harmonic splits mode n at first order, that the beat rate equals the split (illustratively 1.25 Hz), and that the depth of the beat depends on where the bowl is struck relative to the mode axes. Abstract: This book treats the bronze singing bowl as a computable physical object: a bronze shell vibrating in bending. Which partials does it have, why does one partial split into two and beat, why does rubbing always sing the lowest note, what do the mallet, water and the palm change, and which components can the mechanoreceptors of the palm actually feel? Its material constants, dimensions, quality factors and threshold curves are illustrative; all calculations are generated by its programs; the book is not medical advice and claims no bodily benefit. Chapter 1 derives the ring ratio n(n² − 1)/√(n² + 1) from the Rayleigh quotient: (3,0) lies 2.828 times above (2,0), exactly an octave and a tritone, and the (2,0) mode of the illustrative bowl is at 207.6 Hz. Chapter 2 shows that only the 2n-th thickness harmonic splits mode n at first order, that the beat rate equals the split (illustratively 1.25 Hz), and that the depth of the beat depends on where the bowl is struck relative to the mode axes. Chapter 3 shows that rubbing works through negative damping from velocity-weakening friction, that (2,0) needs 5.66 times less pressure to start than (3,0), and that the swell of a rubbed bowl is set by the hand. Chapter 4 uses Hertzian contact to show how mallet hardness filters the partials; with Q at 4000 the T60 of (2,0) is 42.37 s, while geometric nonlinearity in a 4 mm wall amounts to only 1.373 cents. Chapter 5 shows that the pitch drop with water is concentrated in the last stretch as the water nears the rim, 603 cents when full, and that the ripples are millimetre capillary waves. Chapter 6 shows that (2,0) falls in the most sensitive band of the Pacinian channel and is still 34.6 dB above threshold at a palm under the base; the component at the beat frequency is only 0.0008 microns of numerical leakage, the swell felt is the envelope of the carrier, and one strike is felt for about 22.7 s. Chapter 7 uses these models to test the Kaogongji passage on bell-founding and Rayleigh's Theory of Sound line by line. Chapter 8 gives a shop measurement checklist, a copy checker and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:本书把青铜颂钵当作一个可计算的物理对象:一圈弯曲振动的青铜薄壳。它有哪些分音,为什么一个分音会裂成两个并产生拍,磨钵为什么总是唱出最低的音,锤头、注水与手掌各改变了什么,手掌皮肤里的机械感受器能感到其中哪些成分?书中的材料常数、尺寸、品质因数与阈值曲线均为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何身体功效。 第一章由瑞利商推出圆环弯曲模态的频率比 n(n² − 1)/√(n² + 1):(3,0) 是 (2,0) 的 2.828 倍,恰为一个八度加三全音;示意钵的 (2,0) 在 207.6 赫兹。第二章说明只有第 2n 个厚度谐波在一阶上分裂第 n 阶模态,拍频等于分裂量,示意分裂 1.25 赫兹,拍的深浅由敲击点相对模态轴的方位决定。第三章说明磨钵靠速度弱化摩擦的负阻尼,(2,0) 起振所需压力比 (3,0) 小 5.66 倍,磨钵时的起伏由手速决定。第四章由赫兹接触算出锤头软硬对分音的筛选,Q 为 4000 时 (2,0) 的 T60 为 42.37 秒,而 4 毫米壁厚的几何非线性只有 1.373 音分。 第五章说明注水的音高下移集中在水面接近口沿的最后一段,注满降 603 音分,水纹是毫米级的毛细波。第六章说明 (2,0) 落在帕西尼通道最敏感的频段,托底手掌处仍高出阈值 34.6 分贝;拍频处的振动成分只有 0.0008 微米的数值泄漏,手感到的起伏是载波的包络,一次敲击约可感到 22.7 秒。第七章用这些模型逐句检验《考工记·凫氏为钟》的厚薄、侈弇之说与瑞利《声学理论》。第八章给出门店测量清单、攻心文案检查器与一份预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Rings and Thin Shells: Deriving and Computing Rayleigh's Inextensional Mode Ratios — Chapter 1 derives the ring ratio n(n² − 1)/√(n² + 1) from the Rayleigh quotient: (3,0) lies 2.828 times above (2,0), exactly an octave and a tritone, and the (2,0) mode of the illustrative bowl is at 207.6 Hz. (https://5dlibrary.com/answers/en/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c01/) 第1章 圆环与薄壳:瑞利不可伸长模态的频率比推导与计算 — 第一章由瑞利商推出圆环弯曲模态的频率比 n(n² − 1)/√(n² + 1):(3,0) 是 (2,0) 的 2.828 倍,恰为一个八度加三全音;示意钵的 (2,0) 在 207.6 赫兹。 (https://5dlibrary.com/answers/zh/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c01/) - Ch.2 Degenerate Splitting: Thickness Harmonics Set the Beat, Superposition Sets the Envelope — Chapter 2 shows that only the 2n-th thickness harmonic splits mode n at first order, that the beat rate equals the split (illustratively 1.25 Hz), and that the depth of the beat depends on where the bowl is struck relative to the mode axes. (https://5dlibrary.com/answers/en/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c02/) 第2章 简并分裂:厚度谐波决定拍频,叠加模型算包络 — 第二章说明只有第 2n 个厚度谐波在一阶上分裂第 n 阶模态,拍频等于分裂量,示意分裂 1.25 赫兹,拍的深浅由敲击点相对模态轴的方位决定。 (https://5dlibrary.com/answers/zh/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c02/) - Ch.3 Rubbing: Velocity-Weakening Friction, the Negative-Damping Threshold and a Mode Shape That Follows the Stick — Chapter 3 shows that rubbing works through negative damping from velocity-weakening friction, that (2,0) needs 5.66 times less pressure to start than (3,0), and that the swell of a rubbed bowl is set by the hand. (https://5dlibrary.com/answers/en/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c03/) 第3章 磨钵:速度弱化摩擦、负阻尼阈值与随摩擦点转动的振型 — 第三章说明磨钵靠速度弱化摩擦的负阻尼,(2,0) 起振所需压力比 (3,0) 小 5.66 倍,磨钵时的起伏由手速决定。 (https://5dlibrary.com/answers/zh/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c03/) - Ch.4 Striking: Hertzian Contact Time, Quality Factors and the Size of Geometric Nonlinearity — Chapter 4 uses Hertzian contact to show how mallet hardness filters the partials; with Q at 4000 the T60 of (2,0) is 42.37 s, while geometric nonlinearity in a 4 mm wall amounts to only 1.373 cents. (https://5dlibrary.com/answers/en/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c04/) 第4章 敲击:赫兹接触时间、品质因数与几何非线性的量级 — 第四章由赫兹接触算出锤头软硬对分音的筛选,Q 为 4000 时 (2,0) 的 T60 为 42.37 秒,而 4 毫米壁厚的几何非线性只有 1.373 音分。 (https://5dlibrary.com/answers/zh/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c04/) - Ch.5 Water in the Bowl: Added-Mass Frequency Drop and Faraday Waves — Chapter 5 shows that the pitch drop with water is concentrated in the last stretch as the water nears the rim, 603 cents when full, and that the ripples are millimetre capillary waves. (https://5dlibrary.com/answers/en/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c05/) 第5章 注水:附加质量的频率下移与法拉第波 — 第五章说明注水的音高下移集中在水面接近口沿的最后一段,注满降 603 音分,水纹是毫米级的毛细波。 (https://5dlibrary.com/answers/zh/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c05/) - Ch.6 The Palm: the Pacinian U-Shaped Threshold, the Meissner Channel and the Spectral Truth About Beats — Chapter 6 shows that (2,0) falls in the most sensitive band of the Pacinian channel and is still 34.6 dB above threshold at a palm under the base; the component at the beat frequency is only 0.0008 microns of numerical leakage, the swell felt is the envelope of the carrier, and one strike is felt for about 22.7 s. (https://5dlibrary.com/answers/en/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c06/) 第6章 手掌:帕西尼小体的 U 形阈值、梅斯纳通道与拍频的频谱真相 — 第六章说明 (2,0) 落在帕西尼通道最敏感的频段,托底手掌处仍高出阈值 34.6 分贝;拍频处的振动成分只有 0.0008 微米的数值泄漏,手感到的起伏是载波的包络,一次敲击约可感到 22.7 秒。 (https://5dlibrary.com/answers/zh/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c06/) - Ch.7 The Classical Sources: the Kaogongji on Bell-Founding Read Against Rayleigh's Theory of Sound — With p = 4 the flared bowl has ka of 0.456, its radiation Q falls to 2914, the share of vibrational energy that becomes sound is 0.579 and the tone lasts 17.86 s; the closed bowl has ka of 0.266, a radiation Q of 8249, a sound share of only 0.327 and a tone of 28.53 s; the reference bowl lies between, with a share of 0.443 and a tone of 23.61 s. With p = 6 the values change a great deal, the reference share dropping to 0.086, but the order is exactly the same: the flared bowl largest at 0.228, the closed bowl smallest at 0.033. (https://5dlibrary.com/answers/en/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c07/) 第7章 原典考辨:《考工记·凫氏为钟》与瑞利《声学理论》对读 — 取 p = 4 时,侈口钵的 ka 是 0.456,辐射 Q 降到 2914,振动能量中变成声音的份额是 0.579,余音 17.86 秒;弇口钵的 ka 是 0.266,辐射 Q 升到 8249,变成声音的份额只有 0.327,余音 28.53 秒;参考钵居中,份额 0.443,余音 23.61 秒。取 p = 6 时数值变化很大,参考钵的份额降到 0.086,但排序完全一样:侈口钵 0.228 最大,弇口钵 0.033 最小。 (https://5dlibrary.com/answers/zh/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c07/) - Ch.8 Practice and Refutation: a Shop Measurement Checklist, a Copy Checker and a Pre-Registered Test — Chapter 8 gives a shop measurement checklist, a copy checker and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c08/) 第8章 实践与证伪:门店测量清单、文案检查器与预注册检验 — 第八章给出门店测量清单、攻心文案检查器与一份预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/rayleigh-thin-shell-modes-bronze-singing-bowls-nonlinear-strike/c08/) ## 5DL-B-162 · Ear-Canal Pressure and Vagal Afference, Exhale-Prolonged Phase Locking and Gain Bounds of the Slow-Wave Baroreflex 中文书名:外耳道声压迷走传入、呼气延长相位锁定与慢波压力反射增益边界 DOI: pending Cite: Fan, Y. (2026). Ear-Canal Pressure and Vagal Afference, Exhale-Prolonged Phase Locking and Gain Bounds of the Slow-Wave Baroreflex (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Loop parameters, skin moduli and population ranges are illustrative; all calculations are generated by the book's programs; the book is not medical advice and claims no bodily benefit for ear pressing, ear stimulation, breathing exercises or sound products. Chapter 4 finds that four in and eight out gives 0.818 of the heart-rate swing of five in and five out in the illustrative loop, and separates a linear from a nonlinear reading of "phase locking". Abstract: This book quantifies the real gain ceilings behind two popular claims: that pressing the cymba conchae or playing sound into the ear canal "activates the vagus nerve", and that breathing in for four seconds and out for eight pushes heart rate variability to its maximum for everyone. It writes the anatomy of the auricular branch of the vagus, the low-pass path from breathing to heart rate, the delayed baroreflex loop and the arithmetic of heart rate variability indices as runnable models, derives their gain curves and upper bounds, and states each bound as a proposition that data could overturn. Loop parameters, skin moduli and population ranges are illustrative; all calculations are generated by the book's programs; the book is not medical advice and claims no bodily benefit for ear pressing, ear stimulation, breathing exercises or sound products. Chapter 1 shows that "every ear has vagal supply in the cymba conchae" rests on 14 ears, with a 95 % lower bound of 0.768; of the skin a finger or earbud actually touches, the expected vagally supplied share is about 0.448; and moving the skin 10 micrometres would take a sound pressure of 138.4 decibels. Chapter 2 writes respiratory sinus arrhythmia as a delayed low-pass output and shows how the heartbeat samples its own modulation. Chapter 3 places the resonance of the baroreflex loop near 6 breaths a minute, bounds resonant amplification at about 2.39, and shows that higher gain means a narrower band; of the 10.43-fold heart-rate oscillation of slow deep breathing over quiet breathing only 2.94 comes from the loop, and in a simulated population a fixed 6 a minute gives a median of 0.926 of each person's peak gain. Chapter 4 finds that four in and eight out gives 0.818 of the heart-rate swing of five in and five out in the illustrative loop, and separates a linear from a nonlinear reading of "phase locking". Chapter 5 shows that with the autonomic state unchanged, RMSSD can change 2.33-fold and LF/HF several hundredfold with breathing rate alone. Chapter 6 cites a sham-controlled meta-analysis that finds the acute effect of auricular electrical stimulation on vagally mediated heart rate variability close to zero, and computes that the winner's curse can inflate small-study effects 6.2 times and that regression to the mean can raise RMSSD by 19.0 % without any stimulation. Chapter 7 reads the yawning, sobbing, sighing and ringing ears of the Lingshu beside Ludwig's breathing-linked pulse waves of 1847 and Hales's blood column of 1733: in the illustrative model one sigh moves the RR interval by 110.4 milliseconds, and Hales's column equals 26.0 kilopascals. Chapter 8 gives an operator's checklist, a persuasion-copy checker and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:本书量化两类流行说法的真实增益上限:按压耳甲艇或向外耳道播放声音能「激活迷走神经」,以及吸四秒、呼八秒的呼吸对所有人都能把心率变异度推到最高。书中把迷走神经耳支的解剖分布、呼吸到心率的低通通路、压力反射的时延回路与心率变异度指标的计算方式写成可以运行的模型,求出各自的增益曲线与上界,并把每一个上界写成可以被数据推翻的命题。书中的回路参数、皮肤模量与人群分布均为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何按耳、耳部刺激、呼吸练习或声音产品的身体功效。 第一章说明「耳甲艇每一只耳朵都由迷走神经耳支支配」的依据是 14 只耳朵,95 % 置信下界为 0.768;手指或耳塞实际接触的皮肤中,带迷走分布的期望份额约为 0.448;要让皮肤位移达到 10 微米,声压需要 138.4 分贝。第二章把呼吸性窦性心律不齐写成带时延的低通输出,并说明心跳对自身调制的采样。第三章求出压力反射回路的共振在每分钟约 6 次,共振放大的上界约为 2.39,增益越高共振带越窄;慢深呼吸相对平静呼吸 10.43 倍的心率振荡幅度中,只有 2.94 倍来自回路;模拟人群中固定每分钟 6 次的增益中位数为各自峰值的 0.926。第四章说明四秒吸八秒呼在示意回路中的心率摆幅是五秒吸五秒呼的 0.818,并区分「相位锁定」的线性与非线性两种读法。 第五章说明在自主神经状态不变时,RMSSD 可随呼吸频率改变 2.33 倍,LF/HF 可改变数百倍。第六章引用假刺激对照的元分析,其结论是急性耳部电刺激对迷走介导心率变异度的效应接近零,并计算赢家诅咒可把小样本研究的效应夸大 6.2 倍、回归均值可在无刺激时使 RMSSD 上升 19.0 %。第七章对读《灵枢·口问》的欠、唏、太息与耳鸣、路德维希一八四七年的呼吸性脉搏波动与黑尔斯一七三三年的血柱测量:一次叹息在示意模型中使 RR 间期偏离 110.4 毫秒,黑尔斯的血柱合 26.0 千帕。第八章给出经营者清单、攻心文案检查器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A Nerve Map of the Cymba Conchae: Fourteen Ears, Contact Patches and Sound Pressure — Chapter 1 shows that "every ear has vagal supply in the cymba conchae" rests on 14 ears, with a 95 % lower bound of 0.768; of the skin a finger or earbud actually touches, the expected vagally supplied share is about 0.448; and moving the skin 10 micrometres would take a sound pressure of 138.4 decibels. (https://5dlibrary.com/answers/en/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c01/) 第1章 耳甲艇的神经地图:十四只耳朵、接触斑与声压量级 — 第一章说明「耳甲艇每一只耳朵都由迷走神经耳支支配」的依据是 14 只耳朵,95 % 置信下界为 0.768;手指或耳塞实际接触的皮肤中,带迷走分布的期望份额约为 0.448;要让皮肤位移达到 10 微米,声压需要 138.4 分贝。 (https://5dlibrary.com/answers/zh/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c01/) - Ch.2 Respiratory Sinus Arrhythmia: Heart Rate as a Low-Pass Output of the Breathing Drive — The path from breathing to heart rate behaves like a low-pass filter with delay. Under illustrative parameters the relative gain is 0.768 at 6 breaths a minute and 0.433 at 15, so at equal tidal volume the swings stand in a ratio of 1.78, and heart rate lags lung volume by about 1.61 seconds. The heartbeat is itself a sampler: with 4.40 beats per breath the swing we see is 0.842 of the continuous value, and with 2.20 beats only 0.410. Slow breathing enlarges RSA first because of filtering and sampling, and only then, perhaps, because regulation itself changes. (https://5dlibrary.com/answers/en/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c02/) 第2章 呼吸性窦性心律不齐:心率作为呼吸驱动的低通输出 — 呼吸到心率的通路像一个带时延的低通滤波器。在示意参数下,每分钟 6 次呼吸的相对增益是 0.768,每分钟 15 次是 0.433,同样潮气量下两者的峰谷差之比是 1.78;心率的变化比肺容积滞后 1.61 秒左右。心跳本身又是一个采样器:每次呼吸只有 4.40 次心跳时,看到的峰谷差只有连续值的 0.842,只有 2.20 次时只剩 0.410。慢呼吸使 RSA 变大,首先是因为滤波与采样,其次才可能涉及调节本身的改变。 (https://5dlibrary.com/answers/zh/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c02/) - Ch.3 The Delayed Baroreflex Loop: Resonance at Six a Minute and a Ceiling on Gain — Chapter 3 places the resonance of the baroreflex loop near 6 breaths a minute, bounds resonant amplification at about 2.39, and shows that higher gain means a narrower band; of the 10.43-fold heart-rate oscillation of slow deep breathing over quiet breathing only 2.94 comes from the loop, and in a simulated population a fixed 6 a minute gives a median of 0.926 of each person's peak gain. (https://5dlibrary.com/answers/en/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c03/) 第3章 压力反射时延回路:每分钟六次的共振与增益上界 — 第三章求出压力反射回路的共振在每分钟约 6 次,共振放大的上界约为 2.39,增益越高共振带越窄;慢深呼吸相对平静呼吸 10.43 倍的心率振荡幅度中,只有 2.94 倍来自回路;模拟人群中固定每分钟 6 次的增益中位数为各自峰值的 0.926。 (https://5dlibrary.com/answers/zh/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c03/) - Ch.4 Four In, Eight Out: Detuning, Harmonics and Phase Locking at Five a Minute — Chapter 4 finds that four in and eight out gives 0.818 of the heart-rate swing of five in and five out in the illustrative loop, and separates a linear from a nonlinear reading of "phase locking". (https://5dlibrary.com/answers/en/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c04/) 第4章 四秒吸八秒呼:每分钟五次的失谐、谐波与相位锁定 — 第四章说明四秒吸八秒呼在示意回路中的心率摆幅是五秒吸五秒呼的 0.818,并区分「相位锁定」的线性与非线性两种读法。 (https://5dlibrary.com/answers/zh/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c04/) - Ch.5 Artefacts of the Indices: How RMSSD and LF/HF Follow the Breathing Rate — Chapter 5 shows that with the autonomic state unchanged, RMSSD can change 2.33-fold and LF/HF several hundredfold with breathing rate alone. (https://5dlibrary.com/answers/en/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c05/) 第5章 指标的伪影:RMSSD 与 LF/HF 怎样跟着呼吸频率走 — 第五章说明在自主神经状态不变时,RMSSD 可随呼吸频率改变 2.33 倍,LF/HF 可改变数百倍。 (https://5dlibrary.com/answers/zh/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c05/) - Ch.6 Effect Sizes of Ear Stimulation: Sham Control, the Winner's Curse and Regression to the Mean — Chapter 6 cites a sham-controlled meta-analysis that finds the acute effect of auricular electrical stimulation on vagally mediated heart rate variability close to zero, and computes that the winner's curse can inflate small-study effects 6.2 times and that regression to the mean can raise RMSSD by 19.0 % without any stimulation. (https://5dlibrary.com/answers/en/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c06/) 第6章 耳部刺激的效应量:假刺激对照、赢家诅咒与回归均值 — 第六章引用假刺激对照的元分析,其结论是急性耳部电刺激对迷走介导心率变异度的效应接近零,并计算赢家诅咒可把小样本研究的效应夸大 6.2 倍、回归均值可在无刺激时使 RMSSD 上升 19.0 %。 (https://5dlibrary.com/answers/zh/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c06/) - Ch.7 The Classics Examined: Yawning, Sobbing, Sighing and Ringing Ears in the Lingshu, Read with Ludwig and Hales — Chapter 7 reads the yawning, sobbing, sighing and ringing ears of the Lingshu beside Ludwig's breathing-linked pulse waves of 1847 and Hales's blood column of 1733: in the illustrative model one sigh moves the RR interval by 110.4 milliseconds, and Hales's column equals 26.0 kilopascals. (https://5dlibrary.com/answers/en/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c07/) 第7章 原典考辨:《灵枢·口问》的欠、唏、太息与耳鸣,对读路德维希与黑尔斯 — 第七章对读《灵枢·口问》的欠、唏、太息与耳鸣、路德维希一八四七年的呼吸性脉搏波动与黑尔斯一七三三年的血柱测量:一次叹息在示意模型中使 RR 间期偏离 110.4 毫秒,黑尔斯的血柱合 26.0 千帕。 (https://5dlibrary.com/answers/zh/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c07/) - Ch.8 Practice and Refutation: an Operator's Checklist, a Copy Checker and Pre-Registered Tests — Chapter 8 gives an operator's checklist, a persuasion-copy checker and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c08/) 第8章 实践与证伪方案:经营者清单、文案检查器与预注册检验 — 第八章给出经营者清单、攻心文案检查器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/ear-canal-pressure-vagal-afference-exhale-prolonged-phase-locking/c08/) ## 5DL-B-163 · Harmonic-Expectation Violation Entropy in Orchestral Scoring, the Shepard-Tone Illusion and the Dynamics of Musical Chills 中文书名:管弦配乐和声期待违背熵、谢帕德无限音阶错觉与音乐战栗立毛反射动力学 DOI: pending Cite: Fan, Y. (2026). Harmonic-Expectation Violation Entropy in Orchestral Scoring, the Shepard-Tone Illusion and the Dynamics of Musical Chills (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not medical advice, claims no health effect, and is not marketing-compliance or legal advice. Chapter 1 writes harmony as a Markov chain: under an illustrative matrix a deceptive cadence carries 3.06 bits of surprisal and an authentic cadence only 0.51. Abstract: Why can an orchestral cue give a listener goosebumps? Why does a scale that "keeps rising" never reach the top? How can long copy rise and fall well without invented urgency? This book argues that the three questions share one core: the listener carries a probability model of what will happen next, and at every moment sound or text honours or violates it. The book rewrites popular claims as arithmetic that can be run and refuted. Its parameters are illustrative; all calculations are generated by its programs; it is not medical advice, claims no health effect, and is not marketing-compliance or legal advice. Chapter 1 writes harmony as a Markov chain: under an illustrative matrix a deceptive cadence carries 3.06 bits of surprisal and an authentic cadence only 0.51. Chapter 2 uses Huron's ITPRA theory to explain why violation moves us and expresses the interaction of uncertainty and surprise as a saddle: the same vi gives 8.5 times the pleasure in a sharper context. Chapter 3 builds Shepard tones and Risset's accelerando and shows that the spectral overlaps of the two directions of a tritone are exactly equal, both 2.5858, so direction can only come from the listener. Chapter 4 combines a skin-conductance delay of 2.18 seconds with a leaky-integrator threshold model: only when violation and crescendo coincide does a chill fire, at second 102.00; of four hundred simulated listeners 68.8 % have chills, but only 21.2 % owe them to the violation. Chapter 5 keeps the loudness ledger of the crescendo: doubling the players adds only 23.2 % to loudness. Chapter 6 carries the harmonic model over to long copy, with an illustrative genre baseline of 3.35 violations per ten sentences, and draws a machine-checkable red line against false urgency. Chapter 7 reads the three hearings of the Xianchi music in the Zhuangzi, Ji Kang's essay on sound without sorrow or joy, and Hanslick's On the Musically Beautiful. Chapter 8 gives checklists, the persuasive-copy checker and a pre-registered protocol: the listening test needs 151 participants and the copy test 10,317 visitors per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:一段管弦配乐为什么能让人起鸡皮疙瘩?一条「永远上升」的音阶为什么永远到不了顶?一份长文案怎样才能起伏有致,而不靠编造的紧迫感?本书认为三个问题共用一个核心:听者心中有一个关于「接下来会发生什么」的概率模型,声音或文字在每一刻兑现或违背它。本书把流行说法改写成可以运行、也可以被推翻的算术。书中参数为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何健康效果,也不是营销合规或法律意见。 第一章把和声写成马尔可夫链:在示意矩阵中,阻碍终止带来 3.06 比特的惊奇,正格终止只有 0.51 比特。第二章用休伦的 ITPRA 理论解释违背为什么动人,并用一个马鞍面表达不确定与惊奇的交互:同一个 vi 在更尖锐的语境中带来 8.5 倍的愉悦。第三章构造谢帕德音与里塞渐快,说明三全音两个方向的频谱重合度完全相同,都是 2.5858,方向只能来自听者。第四章把 2.18 秒的皮肤电延迟与漏积分阈值模型结合:只有违背与渐强同时出现时,战栗才在第 102.00 秒触发;四百位模拟听众中 68.8 % 会战栗,但只有 21.2 % 由违背额外带来。 第五章给渐强记响度账:演奏者人数加倍,响度只增加 23.2 %。第六章把和声模型移植到长文案,示意文体基线为每十句 3.35 次违背,并以可机检规则划出虚假紧迫的红线。第七章读《庄子·天运》的咸池三闻、嵇康的《声无哀乐论》与汉斯立克的《论音乐的美》。第八章给出清单、攻心文案检查器与预注册方案:听众实验需要 151 位被试,文案对照每组需要 10,317 位访客。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Harmony as a Markov Chain: an Illustrative Transition Matrix, Surprisal and Entropy Rate — Chapter 1 writes harmony as a Markov chain: under an illustrative matrix a deceptive cadence carries 3.06 bits of surprisal and an authentic cadence only 0.51. (https://5dlibrary.com/answers/en/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c01/) 第1章 和声的马尔可夫链:示意转移矩阵、惊奇度与熵率 — 第一章把和声写成马尔可夫链:在示意矩阵中,阻碍终止带来 3.06 比特的惊奇,正格终止只有 0.51 比特。 (https://5dlibrary.com/answers/zh/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c01/) - Ch.2 Peaks of Violated Expectation: Deceptive Cadences, Chromatic Borrowing and the Five Responses of ITPRA — Chapter 2 uses Huron's ITPRA theory to explain why violation moves us and expresses the interaction of uncertainty and surprise as a saddle: the same vi gives 8.5 times the pleasure in a sharper context. (https://5dlibrary.com/answers/en/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c02/) 第2章 期待违背的峰:阻碍终止、半音化借用与 ITPRA 的五种反应 — 第二章用休伦的 ITPRA 理论解释违背为什么动人,并用一个马鞍面表达不确定与惊奇的交互:同一个 vi 在更尖锐的语境中带来 8.5 倍的愉悦。 (https://5dlibrary.com/answers/zh/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c02/) - Ch.3 Shepard Tones and the Endless Accelerando: Octave Partials, a Bell-Shaped Envelope and the Tritone Paradox — Chapter 3 builds Shepard tones and Risset's accelerando and shows that the spectral overlaps of the two directions of a tritone are exactly equal, both 2.5858, so direction can only come from the listener. (https://5dlibrary.com/answers/en/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c03/) 第3章 谢帕德音与无限渐快:八度分量、钟形包络与三全音悖论 — 第三章构造谢帕德音与里塞渐快,说明三全音两个方向的频谱重合度完全相同,都是 2.5858,方向只能来自听者。 (https://5dlibrary.com/answers/zh/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c03/) - Ch.4 The Dynamics of Chills: Skin-Conductance Delay, a Leaky Integrator with a Threshold, and Differences between Listeners — Chapter 4 combines a skin-conductance delay of 2.18 seconds with a leaky-integrator threshold model: only when violation and crescendo coincide does a chill fire, at second 102.00; of four hundred simulated listeners 68.8 % have chills, but only 21.2 % owe them to the violation. (https://5dlibrary.com/answers/en/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c04/) 第4章 战栗的动力学:皮肤电延迟、漏积分阈值与听众差异 — 第四章把 2.18 秒的皮肤电延迟与漏积分阈值模型结合:只有违背与渐强同时出现时,战栗才在第 102.00 秒触发;四百位模拟听众中 68.8 % 会战栗,但只有 21.2 % 由违背额外带来。 (https://5dlibrary.com/answers/zh/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c04/) - Ch.5 The Loudness Ledger of a Crescendo: Phon, Sone and the Logarithmic Cost of Players — Chapter 5 keeps the loudness ledger of the crescendo: doubling the players adds only 23.2 % to loudness. (https://5dlibrary.com/answers/en/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c05/) 第5章 渐强的响度账:方、宋与乐队人数的对数代价 — 第五章给渐强记响度账:演奏者人数加倍,响度只增加 23.2 %。 (https://5dlibrary.com/answers/zh/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c05/) - Ch.6 The Surprisal of Persuasive Copy: Carrying the Harmonic Model over to Long Copy, and Drawing a Red Line against False Urgency — Chapter 6 carries the harmonic model over to long copy, with an illustrative genre baseline of 3.35 violations per ten sentences, and draws a machine-checkable red line against false urgency. (https://5dlibrary.com/answers/en/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c06/) 第6章 攻心文案的惊奇度:把和声模型移植到长文案,并划出虚假紧迫的红线 — 第六章把和声模型移植到长文案,示意文体基线为每十句 3.35 次违背,并以可机检规则划出虚假紧迫的红线。 (https://5dlibrary.com/answers/zh/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c06/) - Ch.7 The Classics Examined: Three Hearings of the Xianchi Music, Sound without Sorrow or Joy, and Hanslick's "Forms Moved in Sound" — The fear, slackness and bewilderment of the three hearings of the Xianchi music correspond, in a learning listener, to an initial surprisal of 2.315 bits; a fall to 2.042 bits with excess surprise of only 0.086 once the music is familiar; and a rise to 3.167 bits with excess surprise of 1.085 once the source changes. Ji Kang's "the heart and the sound are plainly two things" holds in information theory: the surprisal profiles of the same music for two listeners correlate at only -0.10; but loudness is 32 sones for both, and arousal really does lie in the sound. Hanslick's restriction of musical content to forms moved in sound is strikingly close to Ji Kang's "restless or calm", and both underrated the force of expectation itself. (https://5dlibrary.com/answers/en/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c07/) 第7章 原典考辨:咸池三闻、声无哀乐与汉斯立克的「乐音运动的形式」 — 咸池三闻的「惧、怠、惑」,在学习型听众中对应为:起初惊奇度 2.315 比特;熟悉后降到 2.042 比特、超额惊奇只有 0.086;信源改变后升到 3.167 比特、超额惊奇 1.085。嵇康的「心之與聲,明為二物」在信息论中成立:同一段音乐对两位听众的惊奇度相关系数只有 −0.10;但响度对两人都是 32 宋,唤醒确实在声。汉斯立克把音乐的内容限定为乐音运动的形式,与嵇康的「躁静」惊人地相近,而两人都低估了期待本身的力量。 (https://5dlibrary.com/answers/zh/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c07/) - Ch.8 Practice and Falsification: the Persuasive-Copy Checker, a Pre-Registered Protocol and the Conditions of Refutation — Chapter 8 gives checklists, the persuasive-copy checker and a pre-registered protocol: the listening test needs 151 participants and the copy test 10,317 visitors per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c08/) 第8章 实践与证伪:攻心文案检查器、预注册方案与推翻条件 — 第八章给出清单、攻心文案检查器与预注册方案:听众实验需要 151 位被试,文案对照每组需要 10,317 位访客。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/harmonic-expectation-violation-entropy-orchestral-scoring-shepard-tone-illusion/c08/) ## 5DL-B-164 · LMS Adaptive Filtering in In-Ear Active Noise Cancellation, Occlusion-Effect Bone-Conduction Gain and SNR-Driven Cognitive Recovery 中文书名:耳内主动降噪最小均方自适应滤波、堵耳效应低频骨导增益与信噪比认知恢复 DOI: pending Cite: Fan, Y. (2026). LMS Adaptive Filtering in In-Ear Active Noise Cancellation, Occlusion-Effect Bone-Conduction Gain and SNR-Driven Cognitive Recovery (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 3 checks Bode's sensitivity integral numerically: the quiet side and the loud side of an illustrative feedback loop are each 32.9 dB·kHz. Chapter 5 prices cancellation with the Speech Intelligibility Index: one decibel at 250 Hz is worth only 0.00206 and one at 2000 Hz 0.00883, and in the illustrative scene cancellation lowers the playback level needed for equal intelligibility by 5.1 dB. Abstract: In which bands, and by how much, can an in-ear active noise canceller physically lower noise; what does it pay for doing so; and how much speech intelligibility and cognitive slack do those decibels buy? This book turns the common claims of noise-cancelling advertising into runnable arithmetic and asks when they hold. Its ear-canal dimensions, noise spectra, loop gains and cognitive parameters are illustrative; all calculations are generated by its programs; the book is not medical advice and claims no benefit to hearing, to ringing in the ears or to health from any product or way of listening. Chapter 1 runs least-mean-squares and filtered-x least-mean-squares on synthetic noise and shows that the textbook step bound 2/(L·P) is too loose for Gaussian input: for 16 taps the exact bound is 0.1111, and the algorithm diverges once the phase error of its secondary-path model passes 90 degrees. Chapter 2 derives the reduction −20log10|1−e^{−jφ}| under a delay τ, so that 10 dB requires fτ of at most 0.05054. Chapter 3 checks Bode's sensitivity integral numerically: the quiet side and the loud side of an illustrative feedback loop are each 32.9 dB·kHz. Chapter 4 computes the occlusion effect with a lumped model, 13.0 dB at 250 Hz for a shallow fit with a foam leak, and shows that the band in which a vent relieves boom is the band in which it gives up low-frequency isolation. Chapter 5 prices cancellation with the Speech Intelligibility Index: one decibel at 250 Hz is worth only 0.00206 and one at 2000 Hz 0.00883, and in the illustrative scene cancellation lowers the playback level needed for equal intelligibility by 5.1 dB. Chapter 6 uses Kahneman's model of effort and pupillometry to argue for an inverted U between signal-to-noise ratio and effort: the same 5 dB improvement can raise effort in the toy model by 0.568. Chapter 7 reads the Laozi line on "the great sound" against Lueg's 1936 patent on active silencing, finds the noise-cancelling philosophy a modern projection, and computes that a 10-degree temperature error cuts the 10 dB edge of a 0.5-metre duct to 1933 Hz. Chapter 8 first sets the advertised cancellation figures against the overall change users hear, then gives a persuasive-copy checker, an owner's checklist and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:一只入耳式主动降噪耳机,在物理上最多能在哪些频段把噪声压低多少,它为此付出什么代价,这些分贝又能换来多少言语可懂度与多少认知余裕?本书把降噪宣传中的常见说法写成可以运行的算术,再看它们在什么条件下成立。书中的耳道尺寸、噪声谱、环路增益与认知参数均为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何产品或聆听方式对听力、耳鸣或健康有任何益处。 第一章在合成噪声上运行最小均方与滤波 x 最小均方算法,说明教科书步长界 2/(L·P) 对高斯输入偏松:16 抽头时精确界是 0.1111,而次级通路模型的相位误差越过 90 度时算法发散。第二章推导时延 τ 下的降噪量 −20log10|1−e^{−jφ}|,10 dB 降噪要求 fτ 不超过 0.05054。第三章数值验证伯德灵敏度积分:示意反馈环路安静一侧与吵闹一侧各为 32.9 dB·kHz。第四章用集总参数模型计算堵耳效应,泡棉泄漏下浅插在 250 Hz 为 13.0 dB,并说明开孔缓解闷感的频段正是放弃低频隔声的频段。 第五章用言语可懂度指数给降噪定价:250 Hz 的 1 dB 只值 0.00206,2000 Hz 的 1 dB 值 0.00883,示意场景中主动降噪让同样可懂时的音量低 5.1 dB。第六章以卡尼曼的努力模型与瞳孔文献说明信噪比与听力努力之间的倒 U:同样 5 dB 的改善可以让玩具模型中的努力增加 0.568。第七章对读《老子》「大音希声」与吕格 1936 年的主动消声专利,指出降噪哲学之说是今人附会,并算出 10 摄氏度温差把 0.5 米管道的 10 dB 上限压到 1933 Hz。第八章先对照降噪的宣传数字与用户整体听到的变化,再给出攻心文案检查器、经营者执行清单与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 LMS and Filtered-x LMS: Where the Step-Size Bound 2/(L·P) Really Sits — Chapter 1 runs least-mean-squares and filtered-x least-mean-squares on synthetic noise and shows that the textbook step bound 2/(L·P) is too loose for Gaussian input: for 16 taps the exact bound is 0.1111, and the algorithm diverges once the phase error of its secondary-path model passes 90 degrees. (https://5dlibrary.com/answers/en/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c01/) 第1章 最小均方与滤波 x 最小均方:步长稳定界 2/(L·P) 的真实位置 — 第一章在合成噪声上运行最小均方与滤波 x 最小均方算法,说明教科书步长界 2/(L·P) 对高斯输入偏松:16 抽头时精确界是 0.1111,而次级通路模型的相位误差越过 90 度时算法发散。 (https://5dlibrary.com/answers/zh/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c01/) - Ch.2 Secondary-Path Delay: Phase Error 2πfτ and the Band Edge of 10 dB Cancellation — Chapter 2 derives the reduction −20log10|1−e^{−jφ}| under a delay τ, so that 10 dB requires fτ of at most 0.05054. (https://5dlibrary.com/answers/en/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c02/) 第2章 次级通路时延:相位误差 2πfτ 与 10 dB 降噪的带宽上限 — 第二章推导时延 τ 下的降噪量 −20log10|1−e^{−jφ}|,10 dB 降噪要求 fτ 不超过 0.05054。 (https://5dlibrary.com/answers/zh/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c02/) - Ch.3 The Waterbed of Feedback Cancellation: a Numerical Ledger of Bode's Sensitivity Integral — Chapter 3 checks Bode's sensitivity integral numerically: the quiet side and the loud side of an illustrative feedback loop are each 32.9 dB·kHz. (https://5dlibrary.com/answers/en/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c03/) 第3章 反馈式降噪的水床:伯德灵敏度积分的数值账本 — 第三章数值验证伯德灵敏度积分:示意反馈环路安静一侧与吵闹一侧各为 32.9 dB·kHz。 (https://5dlibrary.com/answers/zh/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c03/) - Ch.4 The Occlusion Effect: Lumped Equations of a Closed Ear Canal and the Price of a Vent — Chapter 4 computes the occlusion effect with a lumped model, 13.0 dB at 250 Hz for a shallow fit with a foam leak, and shows that the band in which a vent relieves boom is the band in which it gives up low-frequency isolation. (https://5dlibrary.com/answers/en/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c04/) 第4章 堵耳效应:闭塞耳道的集总参数方程与开孔的代价 — 第四章用集总参数模型计算堵耳效应,泡棉泄漏下浅插在 250 Hz 为 13.0 dB,并说明开孔缓解闷感的频段正是放弃低频隔声的频段。 (https://5dlibrary.com/answers/zh/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c04/) - Ch.5 An Exchange Rate for Signal-to-Noise: Pricing Cancellation with the Speech Intelligibility Index — Chapter 5 prices cancellation with the Speech Intelligibility Index: one decibel at 250 Hz is worth only 0.00206 and one at 2000 Hz 0.00883, and in the illustrative scene cancellation lowers the playback level needed for equal intelligibility by 5.1 dB. (https://5dlibrary.com/answers/en/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c05/) 第5章 信噪比的汇率:用言语可懂度指数给降噪定价 — 第五章用言语可懂度指数给降噪定价:250 Hz 的 1 dB 只值 0.00206,2000 Hz 的 1 dB 值 0.00883,示意场景中主动降噪让同样可懂时的音量低 5.1 dB。 (https://5dlibrary.com/answers/zh/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c05/) - Ch.6 Listening Effort: Kahneman's Capacity Model and the Inverted U of the Pupil — Chapter 6 uses Kahneman's model of effort and pupillometry to argue for an inverted U between signal-to-noise ratio and effort: the same 5 dB improvement can raise effort in the toy model by 0.568. (https://5dlibrary.com/answers/en/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c06/) 第6章 听力努力:卡尼曼的容量模型与瞳孔的倒 U — 第六章以卡尼曼的努力模型与瞳孔文献说明信噪比与听力努力之间的倒 U:同样 5 dB 的改善可以让玩具模型中的努力增加 0.568。 (https://5dlibrary.com/answers/zh/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c06/) - Ch.7 Sources Examined: "The Great Sound Is Faint" and Lueg's Duct — Chapter 7 reads the Laozi line on "the great sound" against Lueg's 1936 patent on active silencing, finds the noise-cancelling philosophy a modern projection, and computes that a 10-degree temperature error cuts the 10 dB edge of a 0.5-metre duct to 1933 Hz. (https://5dlibrary.com/answers/en/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c07/) 第7章 原典考辨:「大音希声」与吕格的管道 — 第七章对读《老子》「大音希声」与吕格 1936 年的主动消声专利,指出降噪哲学之说是今人附会,并算出 10 摄氏度温差把 0.5 米管道的 10 dB 上限压到 1933 Hz。 (https://5dlibrary.com/answers/zh/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c07/) - Ch.8 Practice and Refutation: a Persuasive-Copy Checker and Pre-Registered Tests — Chapter 8 first sets the advertised cancellation figures against the overall change users hear, then gives a persuasive-copy checker, an owner's checklist and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c08/) 第8章 实践与证伪方案:攻心文案检查器与预注册检验 — 第八章先对照降噪的宣传数字与用户整体听到的变化,再给出攻心文案检查器、经营者执行清单与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/lms-adaptive-filtering-ear-active-noise-cancellation-occlusion-effect/c08/) ## 5DL-B-165 · Solubility-Product Nucleation of Nano-Hydroxyapatite, Enamel Micropore Diffusion and the Oral Commensal Nitrate-Reduction Pathway 中文书名:纳米羟基磷灰石溶度积成核、釉质微孔扩散修复与口腔共生菌硝酸盐还原通路 DOI: pending Cite: Fan, Y. (2026). Solubility-Product Nucleation of Nano-Hydroxyapatite, Enamel Micropore Diffusion and the Oral Commensal Nitrate-Reduction Pathway (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not medical or dental advice and claims no health effect for any product or practice. Chapter 2 shows that the critical pH is not a constant but a function of calcium and phosphate, running from 5.12 to 6.07 across an illustrative grid, with a slope of −0.722 against log calcium. Abstract: Oral-care products often claim to "repair cavities", "rebuild enamel" or be "healthier because they kill bacteria". This book rewrites such claims as propositions that can be computed, measured and refuted: the solubility product and saturation ratio of enamel hydroxyapatite, the critical pH set by the three dissociations of phosphoric acid, the barrier of classical nucleation theory, diffusion in enamel micropores, the solubility product of fluorapatite, and the entero-salivary circulation in which oral commensals reduce nitrate to nitrite. Its parameters are illustrative; all calculations are generated by its programs; the book is not medical or dental advice and claims no health effect for any product or practice. Chapter 1 uses ΔG = −(RT/n) ln S to show that resting saliva at pH 6.8 has a log S of about 8.06, or −5.32 kJ per mole of ions, and that supersaturation falls off a cliff with pH. Chapter 2 shows that the critical pH is not a constant but a function of calcium and phosphate, running from 5.12 to 6.07 across an illustrative grid, with a slope of −0.722 against log calcium. Chapter 3 uses classical nucleation theory to show that each tenfold rise in supersaturation changes the nucleation rate by 10^7.13, that the switch for homogeneous nucleation lies near log S of 11.70, and that nanoparticles act as seeds and ion reservoirs. Chapter 4 gives the square-root law of diffusion (half-concentration depth 0.9539√(Dt)), shows that surface sealing stops mineral gain at 11.22, and finds that a 20 nm particle, even in wide pores, has only 0.01617 of the overall mobility of calcium. Chapter 5 computes the critical pH of fluorapatite (4.51 at 1 ppm fluoride, against 5.46 for hydroxyapatite) and shows that a non-inferiority trial would need 410 per arm to rule out a disadvantage of 10 percentage points. Chapter 6 uses a two-compartment model to infer that the oral pathway supplies about 0.278 of plasma nitrite, that a 90 % cut in oral reduction lowers it by 0.240, that the flora's recovery time constant is at least 55.9 hours, and that rinsing keeps only 0.101 of the peak nitrite rise after a nitrate load. Chapter 7 examines the tooth worms, rinsing, "jade spring" and "cannot grow again" of the Zhubing Yuanhou Lun sentence by sentence against Miller's chemico-parasitic theory, with a neighbour-pair share of 0.289 under a worm model against 0.079 under shared acid. Chapter 8 gives a persuasive-copy checker, a listing checklist, customer-service red lines and a pre-registered testing plan needing 1091 people. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:口腔护理产品常说「修复蛀牙」「重建釉质」「杀灭细菌更健康」。本书把这些说法改写成可以计算、可以测量、可以被推翻的命题:釉质羟基磷灰石的溶度积与过饱和比,磷酸三级电离决定的临界 pH,经典成核理论的能垒,釉质微孔中的扩散,氟磷灰石的溶度积,以及口腔共生菌把硝酸盐还原为亚硝酸盐的肠唾液循环。书中参数为示意;全部计算由书中程序生成;本书不是医疗或牙科建议,不宣称任何产品或做法的健康功效。 第一章用 ΔG = −(RT/n) ln S 说明,静息唾液在 pH 6.8 时 log S 约为 8.06,每摩尔离子 −5.32 kJ,而过饱和随 pH 呈悬崖式下降。第二章说明临界 pH 不是常数,而是钙与磷的函数,在示意网格中从 5.12 变到 6.07,对钙浓度对数的斜率为 −0.722。第三章用经典成核理论说明,过饱和每增加十倍,成核速率变化 10^7.13 倍,均相成核的开关约在 log S 为 11.70,纳米颗粒的角色是晶种与离子库。第四章给出扩散的平方根律(半浓度深度 0.9539√(Dt)),说明表层封闭会使矿物增量停在 11.22,直径 20 nm 的颗粒在大孔中的综合迁移能力也只有钙离子的 0.01617。 第五章计算氟磷灰石的临界 pH(1 ppm 氟时 4.51,对比羟基磷灰石的 5.46),并说明一项非劣效试验若要排除 10 个百分点的劣势,每组需要 410 人。第六章用两室模型推出口腔通路约占血浆亚硝酸盐的 0.278,削减 90 % 口腔还原使其下降 0.240,菌群恢复时间常数至少 55.9 小时,一次硝酸盐负荷后含漱只保留峰值上升的 0.101。第七章逐句考辨《诸病源候论》的齿虫、漱口、玉泉与「不能复生」,并与米勒的化学寄生学说对读:齿虫传播模型的相邻牙对比例为 0.289,共享酸环境模型为 0.079。第八章给出攻心文案检查器、上架清单、客服红线与共需 1091 人的预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Solubility Product and Supersaturation: the Gibbs-Energy Threshold of Remineralisation — Chapter 1 uses ΔG = −(RT/n) ln S to show that resting saliva at pH 6.8 has a log S of about 8.06, or −5.32 kJ per mole of ions, and that supersaturation falls off a cliff with pH. (https://5dlibrary.com/answers/en/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c01/) 第1章 溶度积与过饱和度:再矿化的吉布斯自由能阈值 — 第一章用 ΔG = −(RT/n) ln S 说明,静息唾液在 pH 6.8 时 log S 约为 8.06,每摩尔离子 −5.32 kJ,而过饱和随 pH 呈悬崖式下降。 (https://5dlibrary.com/answers/zh/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c01/) - Ch.2 The Critical pH: a Line Drawn by the Three Dissociations of Phosphoric Acid — Chapter 2 shows that the critical pH is not a constant but a function of calcium and phosphate, running from 5.12 to 6.07 across an illustrative grid, with a slope of −0.722 against log calcium. (https://5dlibrary.com/answers/en/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c02/) 第2章 临界 pH:由磷酸三级电离推出的那条线 — 第二章说明临界 pH 不是常数,而是钙与磷的函数,在示意网格中从 5.12 变到 6.07,对钙浓度对数的斜率为 −0.722。 (https://5dlibrary.com/answers/zh/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c02/) - Ch.3 Classical Nucleation Theory: the Critical Radius and the Extreme Sensitivity of the Nucleation Rate — Chapter 3 uses classical nucleation theory to show that each tenfold rise in supersaturation changes the nucleation rate by 10^7.13, that the switch for homogeneous nucleation lies near log S of 11.70, and that nanoparticles act as seeds and ion reservoirs. (https://5dlibrary.com/answers/en/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c03/) 第3章 经典成核理论:临界核半径与成核速率的极端敏感 — 第三章用经典成核理论说明,过饱和每增加十倍,成核速率变化 10^7.13 倍,均相成核的开关约在 log S 为 11.70,纳米颗粒的角色是晶种与离子库。 (https://5dlibrary.com/answers/zh/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c03/) - Ch.4 Diffusion in Enamel Micropores: the Square-Root Law, Pore Clogging and Molecular Sieving — Chapter 4 gives the square-root law of diffusion (half-concentration depth 0.9539√(Dt)), shows that surface sealing stops mineral gain at 11.22, and finds that a 20 nm particle, even in wide pores, has only 0.01617 of the overall mobility of calcium. (https://5dlibrary.com/answers/en/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c04/) 第4章 釉质微孔中的扩散:平方根律、孔隙封闭与分子筛 — 第四章给出扩散的平方根律(半浓度深度 0.9539√(Dt)),说明表层封闭会使矿物增量停在 11.22,直径 20 nm 的颗粒在大孔中的综合迁移能力也只有钙离子的 0.01617。 (https://5dlibrary.com/answers/zh/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c04/) - Ch.5 Fluorapatite: a Lower Solubility Product, a Window, and How to Read a Non-Inferiority Trial — Chapter 5 computes the critical pH of fluorapatite (4.51 at 1 ppm fluoride, against 5.46 for hydroxyapatite) and shows that a non-inferiority trial would need 410 per arm to rule out a disadvantage of 10 percentage points. (https://5dlibrary.com/answers/en/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c05/) 第5章 氟磷灰石:更低的溶度积、一扇窗口与非劣效试验的读法 — 第五章计算氟磷灰石的临界 pH(1 ppm 氟时 4.51,对比羟基磷灰石的 5.46),并说明一项非劣效试验若要排除 10 个百分点的劣势,每组需要 410 人。 (https://5dlibrary.com/answers/zh/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c05/) - Ch.6 Nitrate, Nitrite, Nitric Oxide: a Two-Compartment Ledger for Antiseptic Mouthwash — Chapter 6 uses a two-compartment model to infer that the oral pathway supplies about 0.278 of plasma nitrite, that a 90 % cut in oral reduction lowers it by 0.240, that the flora's recovery time constant is at least 55.9 hours, and that rinsing keeps only 0.101 of the peak nitrite rise after a nitrate load. (https://5dlibrary.com/answers/en/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c06/) 第6章 硝酸盐–亚硝酸盐–一氧化氮:杀菌漱口水的两室账本 — 第六章用两室模型推出口腔通路约占血浆亚硝酸盐的 0.278,削减 90 % 口腔还原使其下降 0.240,菌群恢复时间常数至少 55.9 小时,一次硝酸盐负荷后含漱只保留峰值上升的 0.101。 (https://5dlibrary.com/answers/zh/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c06/) - Ch.7 The Classics Examined: Tooth Worms and the "Jade Spring" in the Zhubing Yuanhou Lun, and Miller's Chemico-Parasitic Theory — Chapter 7 examines the tooth worms, rinsing, "jade spring" and "cannot grow again" of the Zhubing Yuanhou Lun sentence by sentence against Miller's chemico-parasitic theory, with a neighbour-pair share of 0.289 under a worm model against 0.079 under shared acid. (https://5dlibrary.com/answers/en/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c07/) 第7章 原典考辨:《诸病源候论》的齿虫与玉泉,米勒的化学寄生学说 — 第七章逐句考辨《诸病源候论》的齿虫、漱口、玉泉与「不能复生」,并与米勒的化学寄生学说对读:齿虫传播模型的相邻牙对比例为 0.289,共享酸环境模型为 0.079。 (https://5dlibrary.com/answers/zh/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c07/) - Ch.8 Practice and Falsification: a Copy Checker, a Listing Checklist and Pre-Registered Tests — Chapter 8 gives a persuasive-copy checker, a listing checklist, customer-service red lines and a pre-registered testing plan needing 1091 people. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c08/) 第8章 实践与证伪方案:文案检查器、上架清单与预注册检验 — 第八章给出攻心文案检查器、上架清单、客服红线与共需 1091 人的预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/solubility-product-nucleation-nano-hydroxyapatite-enamel-micropore-diffusion-oral/c08/) ## 5DL-B-166 · Portal Transport of Caprylic Triglycerides, Rapid Hepatic Oxidation and Monocarboxylate-Transporter Kinetics of Brain Ketone Uptake 中文书名:辛酸甘油三酯门静脉跨膜转运、肝脏快速氧化与单羧酸转运体脑酮摄取动力学 DOI: pending Cite: Fan, Y. (2026). Portal Transport of Caprylic Triglycerides, Rapid Hepatic Oxidation and Monocarboxylate-Transporter Kinetics of Brain Ketone Uptake (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not medical advice and makes no claim of any health or cognitive benefit. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: After a spoonful of medium-chain triglyceride oil, why does blood beta-hydroxybutyrate rise, by how much, and when does it peak? Why do caprylic acid, with eight carbons, and lauric acid, with twelve, differ by an order of magnitude? This book rewrites slogans such as "medium chains go straight to the liver", "no carnitine needed" and "coconut oil is natural MCT" as rate equations that can be computed, and asks when they hold. Its parameters are illustrative; all calculations are generated by its programs; it is not medical advice and makes no claim of any health or cognitive benefit. Chapter 1 uses a logistic switch along chain length to show that 0.988 of caprylic acid takes the portal vein against only 0.349 of lauric acid. Chapter 2 keeps a counter-intuitive carbon account: per gram, lauric acid's ketone ceiling is 1.105 times caprylic acid's, so caprylic acid's advantage is kinetic. Chapter 3 writes the closed-form solution of a first-order chain of four compartments (lumen, portal vein, liver, blood ketone): 20 g of tricaprylin gives a peak increment of about 0.573 mmol/L, 11.8 times that of trilaurin and 33.0 minutes earlier, while the elasticity of the peak with respect to portal transport is only 0.007. Chapter 4 shows that doubling the dose gives 1.866 times the peak, and that reading every 15 minutes and taking the highest value inflates the peak to 1.126 times its true size. Chapter 5 explains the near-linearity of brain ketone uptake in blood ketone by Michaelis-Menten kinetics: departure from the line reaches a tenth only at 1.39 mmol/L, and data sampled at low concentrations cannot exclude even a Michaelis constant of 1.96. Chapter 6 refutes "coconut oil equals MCT oil": 20 g of coconut oil peaks at only 0.143 of C8 oil, with 0.690 of that peak coming from its small eight- and ten-carbon fraction; matching C8 oil would take 140 g at once, and per unit of ketone exposure it costs 1.05 times as much. Chapter 7 reads the fat entries of Kou Zongshi's Bencao yanyi alongside Krebs on ketone bodies, and finds milk fat at 0.106 of C8 oil. Chapter 8 gives an owner's checklist, an advertising-copy checker and a pre-registered crossover plan for 32 people. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:一勺中链甘油三酯油吃下去之后,血液里的 β-羟丁酸为什么升高、升高多少、何时到顶?八碳的辛酸与十二碳的月桂酸为什么相差一个数量级?本书把「中链直接进入肝脏」「不需要肉碱」「椰子油是天然 MCT」这类说法改写成可以计算的速率方程,再看它们在什么条件下成立。书中参数为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何健康或认知方面的效益。 第一章用一条沿碳链长度变化的逻辑斯蒂开关说明,辛酸有 0.988 走门静脉,月桂酸只有 0.349。第二章做一笔反直觉的碳账:按每克计,月桂酸的酮体理论上限是辛酸的 1.105 倍,辛酸的优势是动力学的。第三章写出肠腔、门静脉、肝脏与血酮四室一阶链的闭式解:20 克三辛酸甘油酯的血酮增量峰值约 0.573 毫摩尔每升,是三月桂酸甘油酯的 11.8 倍,早 33.0 分钟达峰;门静脉运输速度对峰值的弹性只有 0.007。第四章说明剂量加倍只给出 1.866 倍的峰值,每 15 分钟测一次并取最高读数会把峰值高估到 1.126 倍。 第五章用米氏动力学解释脑酮摄取与血酮的近线性,偏离直线一成要到 1.39 毫摩尔每升,而只在低浓度取样的数据连 1.96 的米氏常数也排除不了。第六章证伪「椰子油等于 MCT 油」:20 克椰子油的峰值只有八碳油的 0.143,其中 0.690 来自它所含的少量八碳与十碳,要追平需一次吃 140 克,按单位酮体暴露计反而贵 1.05 倍。第七章对读寇宗奭《本草衍义》的油脂条目与克雷布斯的酮体论述,并算出乳脂的生酮能力为八碳油的 0.106。第八章给出经营者清单、攻心文案检查器与一份 32 人的预注册交叉试验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Two Roads: the Chain-Length Switch between Portal Blood and Chylomicrons — Chapter 1 uses a logistic switch along chain length to show that 0.988 of caprylic acid takes the portal vein against only 0.349 of lauric acid. (https://5dlibrary.com/answers/en/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c01/) 第1章 两条路:门静脉与乳糜微粒的链长开关 — 第一章用一条沿碳链长度变化的逻辑斯蒂开关说明,辛酸有 0.988 走门静脉,月桂酸只有 0.349。 (https://5dlibrary.com/answers/zh/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c01/) - Ch.2 The Side Door of the Mitochondrion: Carnitine-Independent Entry and Carbon Bookkeeping — Chapter 2 keeps a counter-intuitive carbon account: per gram, lauric acid's ketone ceiling is 1.105 times caprylic acid's, so caprylic acid's advantage is kinetic. (https://5dlibrary.com/answers/en/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c02/) 第2章 线粒体的侧门:肉碱非依赖进入与碳计量 — 第二章做一笔反直觉的碳账:按每克计,月桂酸的酮体理论上限是辛酸的 1.105 倍,辛酸的优势是动力学的。 (https://5dlibrary.com/answers/zh/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c02/) - Ch.3 A First-Order Chain of Four Compartments: the Bateman Solution, Peak and Time to Peak — Chapter 3 writes the closed-form solution of a first-order chain of four compartments (lumen, portal vein, liver, blood ketone): 20 g of tricaprylin gives a peak increment of about 0.573 mmol/L, 11.8 times that of trilaurin and 33.0 minutes earlier, while the elasticity of the peak with respect to portal transport is only 0.007. (https://5dlibrary.com/answers/en/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c03/) 第3章 四室一阶链:贝特曼解、峰值与达峰时间 — 第三章写出肠腔、门静脉、肝脏与血酮四室一阶链的闭式解:20 克三辛酸甘油酯的血酮增量峰值约 0.573 毫摩尔每升,是三月桂酸甘油酯的 11.8 倍,早 33.0 分钟达峰;门静脉运输速度对峰值的弹性只有 0.007。 (https://5dlibrary.com/answers/zh/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c03/) - Ch.4 Dose, Sampling and Individual Differences: How a Peak Gets Measured — Chapter 4 shows that doubling the dose gives 1.866 times the peak, and that reading every 15 minutes and taking the highest value inflates the peak to 1.126 times its true size. (https://5dlibrary.com/answers/en/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c04/) 第4章 剂量、采样与个体差异:峰值是怎样被测出来的 — 第四章说明剂量加倍只给出 1.866 倍的峰值,每 15 分钟测一次并取最高读数会把峰值高估到 1.126 倍。 (https://5dlibrary.com/answers/zh/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c04/) - Ch.5 Monocarboxylate Transporters: Michaelis-Menten Kinetics and the Near-Linearity of Brain Ketone Uptake — Chapter 5 explains the near-linearity of brain ketone uptake in blood ketone by Michaelis-Menten kinetics: departure from the line reaches a tenth only at 1.39 mmol/L, and data sampled at low concentrations cannot exclude even a Michaelis constant of 1.96. (https://5dlibrary.com/answers/en/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c05/) 第5章 单羧酸转运体:米氏动力学与脑酮摄取的近线性 — 第五章用米氏动力学解释脑酮摄取与血酮的近线性,偏离直线一成要到 1.39 毫摩尔每升,而只在低浓度取样的数据连 1.96 的米氏常数也排除不了。 (https://5dlibrary.com/answers/zh/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c05/) - Ch.6 Coconut Oil Is Not MCT Oil: Composition Weighting, the Tolerance Ceiling and Unit Cost — Chapter 6 refutes "coconut oil equals MCT oil": 20 g of coconut oil peaks at only 0.143 of C8 oil, with 0.690 of that peak coming from its small eight- and ten-carbon fraction; matching C8 oil would take 140 g at once, and per unit of ketone exposure it costs 1.05 times as much. (https://5dlibrary.com/answers/en/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c06/) 第6章 椰子油不是 MCT 油:组成加权、耐受上限与单位成本 — 第六章证伪「椰子油等于 MCT 油」:20 克椰子油的峰值只有八碳油的 0.143,其中 0.690 来自它所含的少量八碳与十碳,要追平需一次吃 140 克,按单位酮体暴露计反而贵 1.05 倍。 (https://5dlibrary.com/answers/zh/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c06/) - Ch.7 The Classics Examined: the Oils of the Bencao yanyi and Krebs's Ketone Bodies — Chapter 7 reads the fat entries of Kou Zongshi's Bencao yanyi alongside Krebs on ketone bodies, and finds milk fat at 0.106 of C8 oil. (https://5dlibrary.com/answers/en/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c07/) 第7章 原典考辨:《本草衍义》的油与克雷布斯的酮体 — 第七章对读寇宗奭《本草衍义》的油脂条目与克雷布斯的酮体论述,并算出乳脂的生酮能力为八碳油的 0.106。 (https://5dlibrary.com/answers/zh/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c07/) - Ch.8 Practice and a Plan for Falsification — Chapter 8 gives an owner's checklist, an advertising-copy checker and a pre-registered crossover plan for 32 people. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c08/) 第8章 实践与证伪方案 — 第八章给出经营者清单、攻心文案检查器与一份 32 人的预注册交叉试验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/portal-transport-caprylic-triglycerides-rapid-hepatic-oxidation-monocarboxylate-transporter/c08/) ## 5DL-B-167 · Fulvic-Acid Coordination Chemistry of Alpine Humus, Trace Boron and SHBG Dissociation Constants, and Heavy-Metal Leaching Bounds 中文书名:高山腐殖质富里酸配位化学、痕量硼对性激素结合球蛋白解离常数及重金属溶出边界 DOI: pending Cite: Fan, Y. (2026). Fulvic-Acid Coordination Chemistry of Alpine Humus, Trace Boron and SHBG Dissociation Constants, and Heavy-Metal Leaching Bounds (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its site capacities, stability constants, lumen parameters and lot contents are illustrative; all calculations are generated by its programs; the book is not medical advice, claims no health benefit for any ingredient, and is not legal or financial advice. The book contains 18 standard-library Python programs and separates evidence, calculation, classical sources and the author's hypotheses in marked boxes. Abstract: Alpine humus extracts, fulvic drops, Himalayan resins and boron blends are sold in one language: natural, mineral, chelated, balanced. This book rewrites those claims as equations of coordination chemistry and binding equilibria and asks under what conditions they hold. Its site capacities, stability constants, lumen parameters and lot contents are illustrative; all calculations are generated by its programs; the book is not medical advice, claims no health benefit for any ingredient, and is not legal or financial advice. Chapter 1 writes fulvic acid as a two-site polyprotic acid: carboxyl-type sites are 0.426 dissociated at pH 2.0 and 0.974 at pH 6.5, while a chelating motif gains 5.36 log units in conditional constant from stomach to gut and still sits 4.50 below its intrinsic value. Chapter 2 solves the mass-action balance: the usual approximation underestimates free metal when metal and ligand are comparable; with calcium and magnesium holding sites, free fractions of trace metals rise by factors of about 1.31 to 1.39, and 0.997 of the lead is carried in solution. Chapter 3 locates the window in which chelation helps: the absorbed share is 0.124 without a ligand, 0.188 at the best point, and 0.028 past the upper edge. Chapter 4 computes an open-loop elasticity of free testosterone with respect to SHBG of -0.653: a tenth's fall in SHBG adds only 19.0 picomoles per litre, and an eight-person uncontrolled design reports a significant rise with probability 0.276 when there is no effect at all. Chapter 5 multiplies daily intake by content: at 0.5 grams a day the lead red line is 7.56 ppm and the cadmium line 4.29 ppm; an illustrative lot has a lead margin of exposure of 30.2 at the label dose and 5.0 under heavy use; detecting a hot share of 0.05 takes 59 sachets tested one by one. Chapter 6 shows that dissolved fulvic acid carries lead into solution rather than locking it away, raising dissolved lead 32 times at pH 6.5, and that ten passing lots push the upper bound on the failing share only down to 0.259. Chapter 7 reads Ge Hong's Baopuzi Neipian beside the dose of Paracelsus: a nominal 0.1 gram dose of cinnabar carries 86220 micrograms of mercury, 2515 times a modern reference intake. Chapter 8 delivers a shelf checklist, a copy checker and a pre-registered protocol (279 per arm for an effect of 0.066 at a power of 0.90). The book contains 18 standard-library Python programs and separates evidence, calculation, classical sources and the author's hypotheses in marked boxes. 摘要:高山腐殖质提取物、富里酸滴剂、喜马拉雅树脂与含硼配方用同一套语言销售:天然、矿物、螯合、平衡。本书把这些说法改写成配位化学与结合平衡的方程,再看它们在什么条件下成立。书中的位点容量、稳定常数、肠腔参数与批次含量均为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何成分的健康功效,也不是法律或理财建议。 第一章把富里酸写成两类位点的多元酸:羧基型位点在 pH 2.0 时已解离 0.426,在 pH 6.5 时解离 0.974,而螯合模体的条件常数从胃到肠上升 5.36 个对数单位,仍比本征值低 4.50。第二章解质量作用平衡:常用近似在金属与配体同量级时低估游离金属,钙镁占据位点后微量金属的游离分数上升约 1.31 到 1.39 倍,铅有 0.997 被带在溶液中。第三章给出「螯合有利」的窗口:无配体时吸收份额 0.124,最佳点 0.188,越过上界后跌到 0.028。第四章计算游离睾酮对 SHBG 的开环弹性为 -0.653:SHBG 下降一成,游离睾酮只增加 19.0 皮摩尔每升;一项八人无对照设计在毫无效应时仍有 0.276 的机会报告显著上升。 第五章把每日摄入量乘以含量:每天 0.5 克时铅的红线是 7.56 ppm、镉 4.29 ppm,示意批次的铅暴露边际在标签剂量下为 30.2、重度使用时为 5.0,发现 0.05 的热袋份额需逐袋检测 59 袋。第六章说明溶解态富里酸把铅带进溶液而非锁住:pH 6.5 时可使溶解铅升至 32 倍;10 批全部合格只能把超标比例上限压到 0.259。第七章对读葛洪《抱朴子·内篇》与帕拉塞尔苏斯的剂量之说:一份 0.1 克丹砂含汞 86220 微克,为现代参考摄入量的 2515 倍。第八章给出上架清单、文案检查器与预注册方案(效应 0.066、检验力 0.90 时每臂 279 人)。全书含 18 段纯标准库 Python 程序,并以标注框分列证据、计算、原典与作者假说。 Chapter answers: - Ch.1 Titrating Fulvic Acid: Carboxyl Sites, Phenolic Sites and Conditional Stability Constants — Chapter 1 writes fulvic acid as a two-site polyprotic acid: carboxyl-type sites are 0.426 dissociated at pH 2.0 and 0.974 at pH 6.5, while a chelating motif gains 5.36 log units in conditional constant from stomach to gut and still sits 4.50 below its intrinsic value. (https://5dlibrary.com/answers/en/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c01/) 第1章 富里酸的多元酸滴定:羧基、酚羟基与条件稳定常数 — 第一章把富里酸写成两类位点的多元酸:羧基型位点在 pH 2.0 时已解离 0.426,在 pH 6.5 时解离 0.974,而螯合模体的条件常数从胃到肠上升 5.36 个对数单位,仍比本征值低 4.50。 (https://5dlibrary.com/answers/zh/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c01/) - Ch.2 Mass Action and Competitive Binding: Exact Free-Metal Fractions — Chapter 2 solves the mass-action balance: the usual approximation underestimates free metal when metal and ligand are comparable; with calcium and magnesium holding sites, free fractions of trace metals rise by factors of about 1.31 to 1.39, and 0.997 of the lead is carried in solution. (https://5dlibrary.com/answers/en/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c02/) 第2章 质量作用与竞争配位:游离金属离子分数的精确解 — 第二章解质量作用平衡:常用近似在金属与配体同量级时低估游离金属,钙镁占据位点后微量金属的游离分数上升约 1.31 到 1.39 倍,铅有 0.997 被带在溶液中。 (https://5dlibrary.com/answers/zh/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c02/) - Ch.3 When "Chelated Is Better Absorbed" Holds: a Window of Stability Constants — Chapter 3 locates the window in which chelation helps: the absorbed share is 0.124 without a ligand, 0.188 at the best point, and 0.028 past the upper edge. (https://5dlibrary.com/answers/en/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c03/) 第3章 「螯合吸收更好」的成立区间:稳定常数窗口 — 第三章给出「螯合有利」的窗口:无配体时吸收份额 0.124,最佳点 0.188,越过上界后跌到 0.028。 (https://5dlibrary.com/answers/zh/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c03/) - Ch.4 Boron, SHBG and Free Testosterone: the Sensitivity of a Binding Equilibrium — Chapter 4 computes an open-loop elasticity of free testosterone with respect to SHBG of -0.653: a tenth's fall in SHBG adds only 19.0 picomoles per litre, and an eight-person uncontrolled design reports a significant rise with probability 0.276 when there is no effect at all. (https://5dlibrary.com/answers/en/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c04/) 第4章 硼、SHBG 与游离睾酮:结合平衡的敏感度 — 第四章计算游离睾酮对 SHBG 的开环弹性为 -0.653:SHBG 下降一成,游离睾酮只增加 19.0 皮摩尔每升;一项八人无对照设计在毫无效应时仍有 0.276 的机会报告显著上升。 (https://5dlibrary.com/answers/zh/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c04/) - Ch.5 Heavy-Metal Red Lines for Shilajit-Type Products: Daily Intake Times Content — Chapter 5 multiplies daily intake by content: at 0.5 grams a day the lead red line is 7.56 ppm and the cadmium line 4.29 ppm; an illustrative lot has a lead margin of exposure of 30.2 at the label dose and 5.0 under heavy use; detecting a hot share of 0.05 takes 59 sachets tested one by one. (https://5dlibrary.com/answers/en/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c05/) 第5章 希拉吉特类产品的重金属红线:每日摄入量乘以含量 — 第五章把每日摄入量乘以含量:每天 0.5 克时铅的红线是 7.56 ppm、镉 4.29 ppm,示意批次的铅暴露边际在标签剂量下为 30.2、重度使用时为 5.0,发现 0.05 的热袋份额需逐袋检测 59 袋。 (https://5dlibrary.com/answers/zh/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c05/) - Ch.6 Leaching Bounds: Does Fulvic Acid Lock Lead In or Carry It Into Solution? — Chapter 6 shows that dissolved fulvic acid carries lead into solution rather than locking it away, raising dissolved lead 32 times at pH 6.5, and that ten passing lots push the upper bound on the failing share only down to 0.259. (https://5dlibrary.com/answers/en/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c06/) 第6章 溶出边界:富里酸是锁住铅,还是把铅带进溶液 — 第六章说明溶解态富里酸把铅带进溶液而非锁住:pH 6.5 时可使溶解铅升至 32 倍;10 批全部合格只能把超标比例上限压到 0.259。 (https://5dlibrary.com/answers/zh/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c06/) - Ch.7 Reading the Sources: Minerals in the Baopuzi and the Dose of Paracelsus — Chapter 7 reads Ge Hong's Baopuzi Neipian beside the dose of Paracelsus: a nominal 0.1 gram dose of cinnabar carries 86220 micrograms of mercury, 2515 times a modern reference intake. (https://5dlibrary.com/answers/en/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c07/) 第7章 原典考辨:《抱朴子》的矿物服食与帕拉塞尔苏斯的剂量 — 第七章对读葛洪《抱朴子·内篇》与帕拉塞尔苏斯的剂量之说:一份 0.1 克丹砂含汞 86220 微克,为现代参考摄入量的 2515 倍。 (https://5dlibrary.com/answers/zh/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c07/) - Ch.8 Practice, the Distillation Kit and a Pre-Registered Protocol — Chapter 8 delivers a shelf checklist, a copy checker and a pre-registered protocol (279 per arm for an effect of 0.066 at a power of 0.90). The book contains 18 standard-library Python programs and separates evidence, calculation, classical sources and the author's hypotheses in marked boxes. (https://5dlibrary.com/answers/en/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c08/) 第8章 实践、蒸馏件与预注册检验方案 — 第八章给出上架清单、文案检查器与预注册方案(效应 0.066、检验力 0.90 时每臂 279 人)。全书含 18 段纯标准库 Python 程序,并以标注框分列证据、计算、原典与作者假说。 (https://5dlibrary.com/answers/zh/fulvic-acid-coordination-chemistry-alpine-humus-trace-boron-shbg/c08/) ## 5DL-B-168 · Glow-Discharge Ionisation in Glass Electrodes, Micro-Gap Ozone Diffusion and Safety Thresholds for Stratum-Corneum Lipid Peroxidation 中文书名:玻璃电极辉光放电电离层、微间隙臭氧菲克扩散与角质层脂质过氧化安全阈值 DOI: pending Cite: Fan, Y. (2026). Glow-Discharge Ionisation in Glass Electrodes, Micro-Gap Ozone Diffusion and Safety Thresholds for Stratum-Corneum Lipid Peroxidation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not medical advice and claims no health or cosmetic benefit. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: This book studies one class of consumer device: a hand-held high-frequency discharge unit whose tip is a glass tube filled with a rare gas at low pressure, sold as a "violet wand" or "glass electrode". It asks only about physics and chemistry: at what voltage the air gap outside the tube breaks down, how much electrical energy the discharge turns into ozone and nitric oxide, how these gases spread to the breathing zone, and how they react with squalene on the skin surface. Its parameters are illustrative; all calculations are generated by its programs; the book is not medical advice and claims no health or cosmetic benefit. Chapter 1 uses Paschen's law to place the breakdown minimum of air at 1.115 pascal-metres and 305.3 volts and gives the voltages needed for gaps of a tenth of a millimetre to one millimetre. Chapter 2 treats the glass wall as a capacitor in series with the gap, finds that the easiest gap to ignite is about 0.099 millimetres, and uses Manley's relation and a circuit simulation to obtain mean gap powers of a fraction of a watt (at most 0.584 watts). Chapter 3 converts power into an ozone flux, 2.50 milligrams per hour in the reference case, and uses an afterglow model to show that nitric oxide from sparks titrates ozone into nitrogen dioxide: with as much nitric oxide as oxygen atoms, only 0.157 remains after a hundred milliseconds. Chapter 4 uses Fickian diffusion and a near-field and far-field model to obtain 48.4 micrograms per cubic metre in the near field. Chapter 5 computes the local concentration in the micro-gap; in the reference case 0.5256 of the ozone made in the gap lands directly on the skin, and the local dose is 234 times that of the breathing route. Chapter 6 combines three published limits into a table of stay limits, with a source ceiling of 4.04 milligrams per hour, and finds 0.0454 of forty thousand draws above the device ceiling. Chapter 7 reads Zhang Hua's light from combing in the Bowuzhi with Franklin, Schönbein and Paschen. Chapter 8 gives a shop checklist, a copy checker and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:本书研究一类消费器械:手持高频放电器,前端是充有低压惰性气体的玻璃管,俗称「紫光棒」或「玻璃电极」。它只问物理与化学:管外的空气间隙在什么电压下击穿,放电把多少电能变成臭氧与一氧化氮,这些气体怎样扩散到呼吸带,又怎样与皮肤表面的角鲨烯反应。书中参数为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何健康或美容功效。 第一章用帕邢定律算出空气的击穿最低点在 1.115 帕米、305.3 伏,并说明零点一到一毫米间隙所需的电压。第二章把玻璃壁当作与间隙串联的电容,算出最容易点燃的间隙约为 0.099 毫米,用曼利关系与电路模拟给出零点几瓦的间隙平均功率(最高 0.584 瓦)。第三章把功率换成臭氧通量,参考情形为每小时 2.50 毫克,并用余辉模型说明火花产生的一氧化氮会把臭氧滴定成二氧化氮:一氧化氮与氧原子等量时,一百毫秒后只剩 0.157。第四章用菲克扩散与近场远场模型求得近场每立方米 48.4 微克。 第五章计算微间隙内的局部浓度,参考情形下间隙生成的臭氧有 0.5256 直接落到皮肤上,局部剂量是呼吸途径的 234 倍。第六章把三条公开限值合成停留上限表,源强上限为每小时 4.04 毫克,四万次抽样中有 0.0454 超过器械上限。第七章对读张华《博物志》的梳头之光、富兰克林、舍恩拜因与帕邢。第八章给出门店清单、文案检查器与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Paschen Curves: the Voltages at which Air, Neon and Argon Break Down — Chapter 1 uses Paschen's law to place the breakdown minimum of air at 1.115 pascal-metres and 305.3 volts and gives the voltages needed for gaps of a tenth of a millimetre to one millimetre. (https://5dlibrary.com/answers/en/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c01/) 第1章 帕邢曲线:空气、氖与氩在什么电压下击穿 — 第一章用帕邢定律算出空气的击穿最低点在 1.115 帕米、305.3 伏,并说明零点一到一毫米间隙所需的电压。 (https://5dlibrary.com/answers/zh/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c01/) - Ch.2 The Glass Is a Capacitor: Dielectric Barriers, Manley's Power and the Townsend Avalanche — Chapter 2 treats the glass wall as a capacitor in series with the gap, finds that the easiest gap to ignite is about 0.099 millimetres, and uses Manley's relation and a circuit simulation to obtain mean gap powers of a fraction of a watt (at most 0.584 watts). (https://5dlibrary.com/answers/en/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c02/) 第2章 玻璃即电容:介质阻挡、曼利功率与汤森雪崩 — 第二章把玻璃壁当作与间隙串联的电容,算出最容易点燃的间隙约为 0.099 毫米,用曼利关系与电路模拟给出零点几瓦的间隙平均功率(最高 0.584 瓦)。 (https://5dlibrary.com/answers/zh/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c02/) - Ch.3 Grams per Kilowatt-Hour: Ozone Yield, the Fate of Oxygen Atoms and the Nitric Oxide in a Spark — Chapter 3 converts power into an ozone flux, 2.50 milligrams per hour in the reference case, and uses an afterglow model to show that nitric oxide from sparks titrates ozone into nitrogen dioxide: with as much nitric oxide as oxygen atoms, only 0.157 remains after a hundred milliseconds. (https://5dlibrary.com/answers/en/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c03/) 第3章 每千瓦时几克:臭氧产率、氧原子的去向与火花里的一氧化氮 — 第三章把功率换成臭氧通量,参考情形为每小时 2.50 毫克,并用余辉模型说明火花产生的一氧化氮会把臭氧滴定成二氧化氮:一氧化氮与氧原子等量时,一百毫秒后只剩 0.157。 (https://5dlibrary.com/answers/zh/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c03/) - Ch.4 From Electrode Tip to Nose Tip: Fickian Diffusion, Near and Far Fields, and the Time Course of One Session — Chapter 4 uses Fickian diffusion and a near-field and far-field model to obtain 48.4 micrograms per cubic metre in the near field. (https://5dlibrary.com/answers/en/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c04/) 第4章 从电极尖到鼻尖:菲克扩散、近场远场与一次使用的时间曲线 — 第四章用菲克扩散与近场远场模型求得近场每立方米 48.4 微克。 (https://5dlibrary.com/answers/zh/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c04/) - Ch.5 Six Double Bonds of Squalene: Skin Uptake, Micro-Gap Concentration and the Share Peroxidised — Chapter 5 computes the local concentration in the micro-gap; in the reference case 0.5256 of the ozone made in the gap lands directly on the skin, and the local dose is 234 times that of the breathing route. (https://5dlibrary.com/answers/en/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c05/) 第5章 角鲨烯的六个双键:皮肤摄取、微间隙浓度与过氧化份额 — 第五章计算微间隙内的局部浓度,参考情形下间隙生成的臭氧有 0.5256 直接落到皮肤上,局部剂量是呼吸途径的 234 倍。 (https://5dlibrary.com/answers/zh/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c05/) - Ch.6 Stay Limits: Three Published Levels, One Table and Forty Thousand Draws — Chapter 6 combines three published limits into a table of stay limits, with a source ceiling of 4.04 milligrams per hour, and finds 0.0454 of forty thousand draws above the device ceiling. (https://5dlibrary.com/answers/en/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c06/) 第6章 停留上限:三条限值、一张表与四万次抽样 — 第六章把三条公开限值合成停留上限表,源强上限为每小时 4.04 毫克,四万次抽样中有 0.0454 超过器械上限。 (https://5dlibrary.com/answers/zh/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c06/) - Ch.7 Reading the Sources: Light from Combing in the Bowuzhi and Paschen's Similarity Law — The Bowuzhi records accurately the breakdown of air after triboelectric charging: light that follows the action, with a crackling sound. At the breakdown limit on surface charge, a comb can make a spark of 1.77 kilovolts that jumps 0.268 millimetres with 23.5 microjoules, and too little ozone to smell. Franklin's observation on points corresponds to field enhancement, Schönbein turned a smell into a measurable substance, and Paschen's similarity law supplied the scaling, which breaks down in gaps of a few micrometres through field emission. (https://5dlibrary.com/answers/en/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c07/) 第7章 原典考辨:《博物志》的梳头之光与帕邢的相似律 — 《博物志》准确记录了摩擦起电后的空气击穿:光随动作出现,伴有爆裂声。按表面电荷的击穿上限,一把梳子能产生 1.77 千伏、跳过 0.268 毫米、能量 23.5 微焦的火花,臭氧量不足以被闻到。富兰克林的尖端观察对应于场增强,舍恩拜因把气味变成了可测的物质,帕邢的相似律给出了标度,但在几微米以下的间隙中被场致发射打破。 (https://5dlibrary.com/answers/zh/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c07/) - Ch.8 Practice and Falsification: a Copy Checker, a Shop Checklist and a Pre-Registered Protocol — Chapter 8 gives a shop checklist, a copy checker and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c08/) 第8章 实践与证伪方案:文案检查、门店清单与预注册 — 第八章给出门店清单、文案检查器与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/glow-discharge-ionisation-glass-electrodes-micro-gap-ozone-fickian/c08/) ## 5DL-B-169 · Ostwald Ripening in Liquid-Nitrogen-Frozen High-Fat Fruit Pulp, Volatile Thiol Release Kinetics and Triangle-Test Sensory Statistics 中文书名:高脂果肉液氮速冻奥斯特瓦尔德熟化、挥发性硫醇释放动力学与三点法感官盲测统计 DOI: pending Cite: Fan, Y. (2026). Ostwald Ripening in Liquid-Nitrogen-Frozen High-Fat Fruit Pulp, Volatile Thiol Release Kinetics and Triangle-Test Sensory Statistics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not medical or nutritional advice, claims no health benefit for any food, and its pricing formula is not personalised financial or business advice. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: This book studies frozen high-fat tropical fruit pulp, such as durian flesh and avocado purée, through four gates between freezing and tasting: what ice crystals form at freezing, how they grow in storage, how sulfurous aroma is released on thawing, and whether customers in a blind test can tell the product from fresh pulp. It breaks slogans such as "locked fresh with liquid nitrogen", "cold chain all the way" and "nobody can tell the difference" into propositions that can be calculated and refuted. Its parameters are illustrative; all calculations are generated by its programs; the book is not medical or nutritional advice, claims no health benefit for any food, and its pricing formula is not personalised financial or business advice. Chapter 1 uses classical nucleation theory and Plank's equation to show why liquid nitrogen makes small ice crystals and how the vapour film and internal conduction limit its advantage. Chapter 2 uses LSW ripening theory to show that the fine crystals of fast freezing are evened out in storage: after thirty days at minus eighteen degrees the radius ratio of blast-frozen to nitrogen-frozen crystals shrinks from 5.41 to 1.13, and nitrogen's lead is worth only 13.8 days. Chapter 3 shows that temperature fluctuation amplifies ripening through the Arrhenius relation, by 1.89 times for a daily swing of eight kelvin amplitude, and simulates three cold-chain grades with a Markov chain. Chapter 4 uses Henry's law and oil-water partitioning to show that fat tilts the headspace towards thiols, amplifying sulfur relative to fruit 10.60 times at a fat volume fraction of 0.15, and computes release on thawing and oxidation in storage. Chapter 5 gives exact critical values for the triangle test and the Thurstonian model: at a sensory distance d′ of one the proportion correct is only 0.4180, and detecting it with eighty per cent power takes 202 assessors. Chapter 6 covers overdispersion in repeated tasting, the similarity test of at least 39 people needed to claim "you can't tell", and false-positive rates as high as 0.3302 under unbalanced serving orders. Chapter 7 examines Ji Han's Nanfang Caomu Zhuang, Bai Juyi's Preface to a Lychee Painting and Bacon's snow-stuffed hen. Chapter 8 gives a buyer's checklist, a pricing calculator and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:本书研究冷冻高脂热带果肉,例如榴莲肉与牛油果泥,从冻结到品尝的四道关口:冻结时形成什么样的冰晶,储存时冰晶怎样长大,解冻时含硫香气怎样释放,顾客能否在盲测中把它与新鲜果肉区分开。它把「液氮锁鲜」「全程冷链」「吃不出区别」等宣传语拆成可以计算、可以被推翻的命题。书中参数为示意;全部计算由书中程序生成;本书不是医疗或营养建议,不宣称任何食品的健康功效,书中的定价公式也不构成个人化的财务或经营建议。 第一章用经典成核理论与普朗克公式说明液氮为什么能把冰晶做小,以及蒸气膜与内部导热如何限制它的优势。第二章用 LSW 熟化理论说明,冻结时的细小冰晶会在储存中被抹平:零下十八度储存三十天后,风冷与液氮冰晶的半径之比从 5.41 缩到 1.13,液氮的领先只值 13.8 天。第三章说明温度波动按阿伦尼乌斯关系放大熟化,日波动振幅八开尔文时放大 1.89 倍,并以马尔可夫链模拟三个冷链等级。第四章用亨利定律与油水分配说明脂肪使顶空偏向硫醇,脂肪体积分数 0.15 时硫味相对果香放大 10.60 倍,并计算解冻释放与储存氧化。 第五章给出三点检验的精确临界值与瑟斯顿模型:感官距离 d′ 为一时正确率只有 0.4180,以百分之八十的功效检出它需要 202 位评价员。第六章说明重复品尝的过度离散、宣称「吃不出区别」至少需要 39 人的相似性检验,以及呈送顺序不均衡时假阳性率可达 0.3302。第七章考辨嵇含《南方草木状》、白居易《荔枝图序》与培根的雪中之鸡。第八章给出买手清单、定价试算器与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Supercooling and Nucleation: Why Liquid-Nitrogen Immersion Makes Small Ice Crystals — Under the illustrative power law noted in the program, crystal radius grows with the square root of the characteristic freezing time, and the five methods leave radii of 72.5, 30.3, 18.6, 8.6 and 5.6 micrometres. The ratio of radii between still air and immersion is 12.98, and the ratio of single-crystal volumes is 2185. The square-root form is only a common empirical shape and the exponent varies between studies, but whatever value it takes, crystal size increases monotonically with freezing time, and that much is robust. (https://5dlibrary.com/answers/en/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c01/) 第1章 过冷与成核:液氮浸渍为什么能把冰晶做小 — 按程序注释里的示意幂律,冰晶半径随特征冻结时间的平方根增长,五种方式留下的半径依次是 72.5、30.3、18.6、8.6 与 5.6 微米。静止空气与液氮浸渍的半径之比是 12.98,单个冰晶的体积之比是 2185。这里的平方根关系只是一个常见的经验形式,指数在不同研究中有差别,但无论指数取多少,冰晶尺寸随冻结时间单调增长这一点是稳固的。 (https://5dlibrary.com/answers/zh/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c01/) - Ch.2 Ostwald Ripening: How the Cube Law Erases the Head Start of Fast Freezing — Chapter 2 uses LSW ripening theory to show that the fine crystals of fast freezing are evened out in storage: after thirty days at minus eighteen degrees the radius ratio of blast-frozen to nitrogen-frozen crystals shrinks from 5.41 to 1.13, and nitrogen's lead is worth only 13.8 days. (https://5dlibrary.com/answers/en/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c02/) 第2章 奥斯特瓦尔德熟化:立方律如何抹平速冻的先发优势 — 第二章用 LSW 熟化理论说明,冻结时的细小冰晶会在储存中被抹平:零下十八度储存三十天后,风冷与液氮冰晶的半径之比从 5.41 缩到 1.13,液氮的领先只值 13.8 天。 (https://5dlibrary.com/answers/zh/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c02/) - Ch.3 Cold-Chain Breaks and the Glassy State: Arrhenius Amplification and the Standstill below Tg′ — Chapter 3 shows that temperature fluctuation amplifies ripening through the Arrhenius relation, by 1.89 times for a daily swing of eight kelvin amplitude, and simulates three cold-chain grades with a Markov chain. (https://5dlibrary.com/answers/en/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c03/) 第3章 冷链断链与玻璃化:阿伦尼乌斯放大与 Tg′ 以下的停摆 — 第三章说明温度波动按阿伦尼乌斯关系放大熟化,日波动振幅八开尔文时放大 1.89 倍,并以马尔可夫链模拟三个冷链等级。 (https://5dlibrary.com/answers/zh/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c03/) - Ch.4 Where the Thiols Go: Henry Partitioning, Fat Retention and First-Order Release on Thawing — Chapter 4 uses Henry's law and oil-water partitioning to show that fat tilts the headspace towards thiols, amplifying sulfur relative to fruit 10.60 times at a fat volume fraction of 0.15, and computes release on thawing and oxidation in storage. (https://5dlibrary.com/answers/en/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c04/) 第4章 硫醇的去向:亨利分配、脂肪截留与解冻释放的一阶动力学 — 第四章用亨利定律与油水分配说明脂肪使顶空偏向硫醇,脂肪体积分数 0.15 时硫味相对果香放大 10.60 倍,并计算解冻释放与储存氧化。 (https://5dlibrary.com/answers/zh/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c04/) - Ch.5 The Arithmetic of the Triangle Test: Critical Numbers, d′ and Minimum Sample Size — Chapter 5 gives exact critical values for the triangle test and the Thurstonian model: at a sensory distance d′ of one the proportion correct is only 0.4180, and detecting it with eighty per cent power takes 202 assessors. (https://5dlibrary.com/answers/en/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c05/) 第5章 三点检验的算术:临界正确数、d′ 与最小样本量 — 第五章给出三点检验的精确临界值与瑟斯顿模型:感官距离 d′ 为一时正确率只有 0.4180,以百分之八十的功效检出它需要 202 位评价员。 (https://5dlibrary.com/answers/zh/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c05/) - Ch.6 Designing Blind Tests of Frozen against Fresh: Replication, Position Bias and the Burden of Proof for "Tastes the Same" — Chapter 6 covers overdispersion in repeated tasting, the similarity test of at least 39 people needed to claim "you can't tell", and false-positive rates as high as 0.3302 under unbalanced serving orders. (https://5dlibrary.com/answers/en/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c06/) 第6章 冷冻对新鲜的盲测设计:重复、位置偏好与「吃不出差别」的举证负担 — 第六章说明重复品尝的过度离散、宣称「吃不出区别」至少需要 39 人的相似性检验,以及呈送顺序不均衡时假阳性率可达 0.3302。 (https://5dlibrary.com/answers/zh/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c06/) - Ch.7 Reading the Sources: Ji Han's Nanfang Caomu Zhuang, Bai Juyi's Preface to a Lychee Painting and Bacon's Snow-Stuffed Hen — The Nanfang Caomu Zhuang describes the form and mouthfeel of lychee and coconut largely accurately; "white as fat" is a simile, while "very rich" matches the genuinely high fat of coconut meat, and "drinking it makes one drunk" does not fit fresh coconut water. Bai Juyi's order of decay in colour, aroma and taste is right, but "all gone in four or five days" fits first-order decay only for colour, implying that aroma and taste accelerate later. Bacon's intuition of cold in place of salt was right, but snow at zero cannot freeze meat, and even at minus three it takes 85.8 hours; the essence of fast freezing is the rate of heat transfer, not low temperature alone. (https://5dlibrary.com/answers/en/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c07/) 第7章 原典考辨:嵇含《南方草木状》、白居易《荔枝图序》与培根的雪中之鸡 — 《南方草木状》对荔枝与椰子的形态与口感描写大体准确,「白如肪」是比喻,「极肥美」对应椰肉真实的高脂;「饮之得醉」不符合新鲜椰子水。白居易给出的色、香、味衰减顺序正确,但「四五日尽去」与一阶衰减只对颜色相容,暗示香、味在后期加速。培根以冷代盐的直觉正确,但零度的雪不能冻结鸡肉,零下三度也要 85.8 小时;速冻的本质是传热速率,而不只是低温。 (https://5dlibrary.com/answers/zh/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c07/) - Ch.8 Practice and Falsification: a Pricing Calculator, a Buyer's Checklist and a Pre-Registered Protocol — Chapter 8 gives a buyer's checklist, a pricing calculator and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c08/) 第8章 实践与证伪:定价试算器、买手清单与预注册方案 — 第八章给出买手清单、定价试算器与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/ostwald-ripening-liquid-nitrogen-frozen-high-fat-fruit-pulp/c08/) ## 5DL-B-170 · Nernst Potentials in Soil Electrochemistry, Nanowatt Harvest Limits of Passive Copper Antennas and Field Block Designs for Electroculture Yield Claims 中文书名:土壤电化学能斯特电势、铜棒无源天线纳瓦级收集极限与电耕增产声称的田间区组设计 DOI: pending Cite: Fan, Y. (2026). Nernst Potentials in Soil Electrochemistry, Nanowatt Harvest Limits of Passive Copper Antennas and Field Block Designs for Electroculture Yield Claims (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not agronomic or investment advice and judges no brand. Chapter 1 uses the Nernst equation to find an open-circuit voltage of about 1.0734 volts for a copper-zinc couple in an illustrative soil solution and shows that its energy comes from corroding zinc. Abstract: This book studies a family of farm-input products and practices: spiral copper rods pushed into a field, or galvanised stakes wrapped in copper wire, claimed to collect atmospheric energy and raise crop yields, a practice known as electroculture. It asks only about physics, electrochemistry and the statistics of experiments: where the voltage between two metals in soil comes from, how much power a copper rod can obtain at most from the fair-weather field, the geomagnetic field and radio waves, what share of sunlight that power is, and how to design field trials that can answer the yield question. Its parameters are illustrative; all calculations are generated by its programs; the book is not agronomic or investment advice and judges no brand. Chapter 1 uses the Nernst equation to find an open-circuit voltage of about 1.0734 volts for a copper-zinc couple in an illustrative soil solution and shows that its energy comes from corroding zinc. Chapter 2 uses earthing resistance and the oxygen diffusion limit to bound the power of an electrode pair in loam at about 320.6 microwatts. Chapter 3 starts from Gauss's law and Ohm's law in air and uses the method of moments to find, for a two-metre copper rod, a short-circuit current of 4.882 picoamperes and a maximum power of 0.114 nanowatts; a spiral outline brings the power at most to 1.40 times that of a straight rod. Chapter 4 uses Faraday's law and Poynting's theorem to show that a still coil in the geomagnetic field has zero electromotive force and that radio waves carry far less energy than sunlight. Chapter 5 compares every power with a mean photosynthetically active radiation of 99.5 watts per square metre; spiral rods supply only $1.48\times10^{-12}$ of it, while shorted copper-zinc stakes can release 3.221 kilograms of zinc per hectare in a season, 10.7 times the crop's removal. Chapter 6 treats randomised blocks, split plots and power analysis: testing the main-plot factor against the subplot error raises the false-positive rate to 0.3160, and a rod's influence on the ground-level field falls below one per cent beyond 3.43 metres. Chapter 7 reads Chen Fu's Nongshu on terrain and manuring beside the electroculture writings of Lemström and Blackman, and simulation shows that a treatment with no effect, compared without replication, shows a gain of a tenth or more in about 0.2354 of seasons. Chapter 8 gives an operator's checklist, a copy checker and a pre-registered split-plot protocol: detecting a gain of a tenth needs about 5320 square metres. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:本书研究一类农资产品与做法:插在田里的螺旋铜杆,或缠着铜线的镀锌桩,宣称能收集大气能量、使作物增产,俗称「电耕」。它只问物理、电化学与试验统计:土壤里两种金属的电压从哪里来,一根铜杆从晴天大气电场、地磁场与无线电波中最多能得到多少功率,这些功率占阳光的多少,以及怎样设计能回答增产问题的田间试验。书中参数为示意;全部计算由书中程序生成;本书不是农艺或投资建议,不评价任何品牌。 第一章用能斯特方程算出铜–锌电偶在示意土壤溶液中的开路电压约 1.0734 伏,并说明其能量来自锌的腐蚀。第二章用接地电阻与氧扩散极限求出壤土中一对电极的功率上限约 320.6 微瓦。第三章从高斯定律与空气中的欧姆定律出发,用矩量法算出两米铜杆的短路电流 4.882 皮安、最大功率 0.114 纳瓦,螺旋外形最多使功率达到直杆的 1.40 倍。第四章用法拉第定律与坡印廷定理说明静止线圈在地磁场里的电动势为零,无线电波的能流远小于阳光。 第五章把各种功率与每平方米 99.5 瓦的平均光合有效辐射相比,螺旋杆只占其 $1.48\times10^{-12}$;而短路的铜–锌桩每公顷一季可溶出 3.221 千克锌,是作物带走量的 10.7 倍。第六章讨论随机区组、裂区设计与功效分析:错用副区误差检验主区因子,假阳性率升到 0.3160;铜杆对地面电场的影响在 3.43 米外降到百分之一以下。第七章对读陈旉《农书》的「地势之宜」「粪田之宜」与伦斯特伦、布莱克曼的电耕文献,并用模拟说明零效应处理在无重复对比中约有 0.2354 的季节显示一成以上增产。第八章给出经营者清单、文案检查器与预注册裂区方案:检测一成效应约需 5320 平方米。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Nernst Equation: Open-Circuit Voltages of Copper-Zinc and Copper-Iron Couples in Soil Solution — Chapter 1 uses the Nernst equation to find an open-circuit voltage of about 1.0734 volts for a copper-zinc couple in an illustrative soil solution and shows that its energy comes from corroding zinc. (https://5dlibrary.com/answers/en/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c01/) 第1章 能斯特方程:铜–锌与铜–铁电偶在土壤溶液里的开路电压 — 第一章用能斯特方程算出铜–锌电偶在示意土壤溶液中的开路电压约 1.0734 伏,并说明其能量来自锌的腐蚀。 (https://5dlibrary.com/answers/zh/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c01/) - Ch.2 Internal Resistance, Oxygen Diffusion and the Maximum Power $V^2/4R$: What a Soil Couple Can Deliver — Chapter 2 uses earthing resistance and the oxygen diffusion limit to bound the power of an electrode pair in loam at about 320.6 microwatts. (https://5dlibrary.com/answers/en/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c02/) 第2章 内阻、氧扩散与最大功率 $V^2/4R$:土壤电偶究竟能输出多少 — 第二章用接地电阻与氧扩散极限求出壤土中一对电极的功率上限约 320.6 微瓦。 (https://5dlibrary.com/answers/zh/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c02/) - Ch.3 Gauss's Law and Air Conductivity: How Much Current a Copper Rod Collects from the Fair-Weather Field — Chapter 3 starts from Gauss's law and Ohm's law in air and uses the method of moments to find, for a two-metre copper rod, a short-circuit current of 4.882 picoamperes and a maximum power of 0.114 nanowatts; a spiral outline brings the power at most to 1.40 times that of a straight rod. (https://5dlibrary.com/answers/en/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c03/) 第3章 高斯定律与空气电导:一根铜杆在晴天大气电场里能收集多少电流 — 第三章从高斯定律与空气中的欧姆定律出发,用矩量法算出两米铜杆的短路电流 4.882 皮安、最大功率 0.114 纳瓦,螺旋外形最多使功率达到直杆的 1.40 倍。 (https://5dlibrary.com/answers/zh/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c03/) - Ch.4 Faraday's Law and Poynting's Theorem: What a Still Coil, the Geomagnetic Field and Radio Waves Can Give a Field — Faraday's law says that only a changing flux induces an electromotive force; a still coil in the geomagnetic field has exactly zero. Rocked by wind, a coil of twenty turns yields a third of a microvolt and less than a picowatt, and that energy comes from the wind. The geomagnetic field's own daily variation, storms and micropulsations give powers below $10^{-22}$ watts. Poynting's theorem sets the ceiling on the electromagnetic power any device can obtain: a typical medium-wave signal carries $2.65\times10{-10}$ of the energy flux of sunshine, and an untuned two-metre rod takes only 0.315 nanowatts from it. The next chapter sets these numbers beside the photosynthetically active radiation on which a crop actually depends every day. (https://5dlibrary.com/answers/en/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c04/) 第4章 法拉第定律与坡印廷定理:静止线圈、地磁场与无线电波能给田地多少 — 法拉第定律说,只有磁通量变化才有感应电动势;静止线圈在地磁场里的电动势严格为零。让线圈随风摆动,二十匝的线圈也只能得到零点三微伏、不到一皮瓦的功率,而且这点能量来自风。地磁场自身的日变化、磁暴与微脉动,给出的功率在 $10^{-22}$ 瓦以下。坡印廷定理规定了任何装置能得到的电磁功率的上限:典型中波信号的能流是阳光的 $2.65\times10{-10}$,不加调谐的两米铜杆从中只得到 0.315 纳瓦。下一章把这些数与作物每天真正依赖的光合有效辐射放在一起。 (https://5dlibrary.com/answers/zh/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c04/) - Ch.5 The Energy Ledger: Electroculture Power as a Share of Photosynthetically Active Radiation, and Zinc as a Confounder — Chapter 5 compares every power with a mean photosynthetically active radiation of 99.5 watts per square metre; spiral rods supply only $1.48\times10^{-12}$ of it, while shorted copper-zinc stakes can release 3.221 kilograms of zinc per hectare in a season, 10.7 times the crop's removal. (https://5dlibrary.com/answers/en/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c05/) 第5章 能量账:电耕功率在光合有效辐射里占多少,以及锌这个混杂因素 — 第五章把各种功率与每平方米 99.5 瓦的平均光合有效辐射相比,螺旋杆只占其 $1.48\times10^{-12}$;而短路的铜–锌桩每公顷一季可溶出 3.221 千克锌,是作物带走量的 10.7 倍。 (https://5dlibrary.com/answers/zh/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c05/) - Ch.6 Testing in the Field: Randomised Blocks, Split Plots, Power Analysis and Guard Strips — Chapter 6 treats randomised blocks, split plots and power analysis: testing the main-plot factor against the subplot error raises the false-positive rate to 0.3160, and a rod's influence on the ground-level field falls below one per cent beyond 3.43 metres. (https://5dlibrary.com/answers/en/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c06/) 第6章 田间检验:随机区组、裂区设计、功效分析与保护行 — 第六章讨论随机区组、裂区设计与功效分析:错用副区误差检验主区因子,假阳性率升到 0.3160;铜杆对地面电场的影响在 3.43 米外降到百分之一以下。 (https://5dlibrary.com/answers/zh/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c06/) - Ch.7 Reading the Sources: Chen Fu's "Fitness of Terrain" and the Electroculture Trials of Lemström and Blackman — Chapter 7 reads Chen Fu's Nongshu on terrain and manuring beside the electroculture writings of Lemström and Blackman, and simulation shows that a treatment with no effect, compared without replication, shows a gain of a tenth or more in about 0.2354 of seasons. (https://5dlibrary.com/answers/en/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c07/) 第7章 原典考辨:陈旉《农书》的「地势之宜」与伦斯特伦、布莱克曼的电耕试验 — 第七章对读陈旉《农书》的「地势之宜」「粪田之宜」与伦斯特伦、布莱克曼的电耕文献,并用模拟说明零效应处理在无重复对比中约有 0.2354 的季节显示一成以上增产。 (https://5dlibrary.com/answers/zh/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c07/) - Ch.8 Practice and Falsification: A Farm Owner's Checklist, a Copy Checker and a Pre-Registered Split-Plot Trial — Chapter 8 gives an operator's checklist, a copy checker and a pre-registered split-plot protocol: detecting a gain of a tenth needs about 5320 square metres. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c08/) 第8章 实践与证伪方案:农场主清单、文案检查器与预注册裂区试验 — 第八章给出经营者清单、文案检查器与预注册裂区方案:检测一成效应约需 5320 平方米。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/nernst-potentials-soil-electrochemistry-nanowatt-harvest-limits-passive-copper/c08/) ## 5DL-B-171 · Multi-Feature Bayesian Authentication of Pre-Owned Luxury, Akerlof Lemon-Discount Equilibria and the Resale Half-Life of Limited Bags 中文书名:二手精品多特征贝叶斯鉴定、阿克洛夫柠檬折价均衡与限量包款二级市场折旧半衰期 DOI: pending Cite: Fan, Y. (2026). Multi-Feature Bayesian Authentication of Pre-Owned Luxury, Akerlof Lemon-Discount Equilibria and the Resale Half-Life of Limited Bags (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not a pricing opinion on any actual transaction and not personalised legal or investment advice. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: This book studies three computable questions in the pre-owned luxury trade: what an authentication actually delivers, what happens to a market when buyers cannot tell genuine from fake, and how long the secondary premium of a limited model lasts. It uses abstract feature names only, refers to no brand, and gives no method for authenticating the goods of any brand. Its parameters are illustrative; all calculations are generated by its programs; it is not a pricing opinion on any actual transaction and not personalised legal or investment advice. Chapter 1 writes authentication as Bayesian updating and shows the decisive role of the channel's base rate: the same two-of-four rule has a positive predictive value of only 0.223 in a channel where one item in a hundred is fake. Chapter 2 shows that when features share a cause, plain multiplication counts evidence twice: in a model whose marginal parameters are exactly right, one pattern has a naive posterior of 0.88503 and an exact posterior of only 0.34611, and a tempering weight fitted on confirmed records is 0.75. Chapter 3 discusses the two paradoxes of agreement between appraisers and shows that when one item in five is a copy that fools appraisers together, a panel of nine is right only 0.8757 of the time. Chapter 4 carries Akerlof's lemons model over to pre-owned bags; once the share of fakes passes 0.1622, owners of genuine bags leave the market as a body. Chapter 5 finds the fee threshold of third-party authentication, above 4.74 fakes are no longer submitted, and gives the conditions under which a refund guarantee deters. Chapter 6 estimates the premium half-life by log-linear regression at 8.35 months, shows that contamination by fakes shortens the estimate to 7.03 months, and gives the condition under which the implied price of a bundle quota makes the premium collapse. Chapter 7 reads the market officer of the Rites of Zhou, the man of Zheng in the Han Feizi and Akerlof's paper of 1970 side by side, and finds a likelihood ratio of 1.37374 for a full set of accessories. Chapter 8 gives a checklist for taking in stock, a buy-in quote calculator, machine rules for copy and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:本书研究二手精品生意里的三个可计算问题:鉴定给出的到底是什么,买家分不清真假时市场会怎样,限量款的二级溢价能撑多久。全书只用抽象的特征名,不针对任何品牌,也不提供鉴定某一品牌真伪的方法。书中参数为示意;全部计算由书中程序生成;本书不是对任何具体交易的定价意见,也不构成个人化的法律或理财建议。 第一章把鉴定写成贝叶斯更新,说明渠道基率的决定作用:同一条「四中二」规则,在百分之一假货的渠道里阳性预测值只有 0.223。第二章说明特征有共同原因时朴素连乘会重复计数证据:在边际参数完全正确的模型里,一组观察的朴素后验是 0.88503,精确后验只有 0.34611;用已确认记录拟合的回火权重为 0.75。第三章讨论鉴定师一致性的两个悖论,并说明当五分之一的货是共同受骗的高仿时,九人投票的准确率只有 0.8757。第四章把阿克洛夫的柠檬模型搬到二手包上,假货比例越过 0.1622 时真货卖家整体退出。 第五章算出第三方鉴定的费用门槛,费用超过 4.74 时假货不再送检,并给出退款担保的威慑条件。第六章用对数线性回归估计溢价半衰期为 8.35 个月,说明假货污染会把它压短到 7.03 个月,并给出配货隐含价格的崩塌条件。第七章对读《周礼·地官·质人》、《韩非子》郑人买履与阿克洛夫 1970 年的论文,算出全套附件的似然比为 1.37374。第八章给出收货清单、收购报价试算器、文案机检规则与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Authentication as Bayesian Updating: Likelihood Ratios of Four Features, Channel Base Rates and Sequential Reading — Chapter 1 writes authentication as Bayesian updating and shows the decisive role of the channel's base rate: the same two-of-four rule has a positive predictive value of only 0.223 in a channel where one item in a hundred is fake. (https://5dlibrary.com/answers/en/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c01/) 第1章 鉴定即贝叶斯更新:四个特征的似然比、渠道基率与序贯读法 — 第一章把鉴定写成贝叶斯更新,说明渠道基率的决定作用:同一条「四中二」规则,在百分之一假货的渠道里阳性预测值只有 0.223。 (https://5dlibrary.com/answers/zh/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c01/) - Ch.2 Double Counting: How Naive Bayes Counts One Piece of Evidence Several Times When Features Share a Cause — Chapter 2 shows that when features share a cause, plain multiplication counts evidence twice: in a model whose marginal parameters are exactly right, one pattern has a naive posterior of 0.88503 and an exact posterior of only 0.34611, and a tempering weight fitted on confirmed records is 0.75. (https://5dlibrary.com/answers/en/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c02/) 第2章 串联高估:特征共因时朴素贝叶斯把同一份证据数了几遍 — 第二章说明特征有共同原因时朴素连乘会重复计数证据:在边际参数完全正确的模型里,一组观察的朴素后验是 0.88503,精确后验只有 0.34611;用已确认记录拟合的回火权重为 0.75。 (https://5dlibrary.com/answers/zh/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c02/) - Ch.3 Agreement Between Appraisers: Cohen's Kappa, the Two Paradoxes of High Agreement and Low Kappa, and the Ceiling on Majority Votes — Chapter 3 discusses the two paradoxes of agreement between appraisers and shows that when one item in five is a copy that fools appraisers together, a panel of nine is right only 0.8757 of the time. (https://5dlibrary.com/answers/en/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c03/) 第3章 鉴定师一致性:Cohen κ、高一致低 κ 的两个悖论与多人投票的天花板 — 第三章讨论鉴定师一致性的两个悖论,并说明当五分之一的货是共同受骗的高仿时,九人投票的准确率只有 0.8757。 (https://5dlibrary.com/answers/zh/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c03/) - Ch.4 The Lemon Discount: Why Owners of Genuine Bags Leave All at Once When the Share of Fakes Crosses a Tipping Point — Chapter 4 carries Akerlof's lemons model over to pre-owned bags; once the share of fakes passes 0.1622, owners of genuine bags leave the market as a body. (https://5dlibrary.com/answers/en/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c04/) 第4章 柠檬折价:假货比例越过临界点时,真品卖家为何整体退出 — 第四章把阿克洛夫的柠檬模型搬到二手包上,假货比例越过 0.1622 时真货卖家整体退出。 (https://5dlibrary.com/answers/zh/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c04/) - Ch.5 Third-Party Authentication and Refund Guarantees: How Low a Fee and How High a Refund Multiple Restore Equilibrium — Chapter 5 finds the fee threshold of third-party authentication, above 4.74 fakes are no longer submitted, and gives the conditions under which a refund guarantee deters. (https://5dlibrary.com/answers/en/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c05/) 第5章 第三方鉴定与退款担保:恢复均衡需要多低的费用与多高的赔付倍数 — 第五章算出第三方鉴定的费用门槛,费用超过 4.74 时假货不再送检,并给出退款担保的威慑条件。 (https://5dlibrary.com/answers/zh/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c05/) - Ch.6 The Half-Life of a Limited Bag's Premium: Log-Linear Regression, Contamination by Fakes, and When a Bundle Quota Makes the Premium Collapse — Chapter 6 estimates the premium half-life by log-linear regression at 8.35 months, shows that contamination by fakes shortens the estimate to 7.03 months, and gives the condition under which the implied price of a bundle quota makes the premium collapse. (https://5dlibrary.com/answers/en/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c06/) 第6章 限量款溢价半衰期:对数线性回归、假货污染与配货隐含价格的崩塌条件 — 第六章用对数线性回归估计溢价半衰期为 8.35 个月,说明假货污染会把它压短到 7.03 个月,并给出配货隐含价格的崩塌条件。 (https://5dlibrary.com/answers/zh/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c06/) - Ch.7 Reading the Sources: the Market Officer's Contracts and Claim Periods, the Man of Zheng Who Trusted His Measure, and Akerlof 1970 — Chapter 7 reads the market officer of the Rites of Zhou, the man of Zheng in the Han Feizi and Akerlof's paper of 1970 side by side, and finds a likelihood ratio of 1.37374 for a full set of accessories. (https://5dlibrary.com/answers/en/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c07/) 第7章 原典考辨:质人的书契与期限、郑人买履的「宁信度」与阿克洛夫 1970 — 第七章对读《周礼·地官·质人》、《韩非子》郑人买履与阿克洛夫 1970 年的论文,算出全套附件的似然比为 1.37374。 (https://5dlibrary.com/answers/zh/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c07/) - Ch.8 Practice and Falsification: a Buy-In Quote Calculator, Machine Checks on Copy, and a Pre-Registered Test — Chapter 8 gives a checklist for taking in stock, a buy-in quote calculator, machine rules for copy and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c08/) 第8章 实践与证伪方案:收购报价试算器、文案机检与预注册检验 — 第八章给出收货清单、收购报价试算器、文案机检规则与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/multi-feature-bayesian-authentication-pre-owned-luxury-akerlof-lemon/c08/) ## 5DL-B-172 · Survivor Power Laws in Influencer-Agency Incubation, Sampling-Audit Games against Off-Platform Deals and Option Pricing of Contract Buyouts 中文书名:网红经纪机构孵化幸存者幂律、达人飞单抽样审计博弈与违约买断期权定价 DOI: pending Cite: Fan, Y. (2026). Survivor Power Laws in Influencer-Agency Incubation, Sampling-Audit Games against Off-Platform Deals and Option Pricing of Contract Buyouts (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not individual legal or financial advice. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: This book studies three critical moments in the business of an influencer agency: how to estimate the true incubation success rate before signing, how to use audits to curb deals that a rising creator takes privately around the agency, and how to design the revenue split and the contract buyout so that the agency does not lose a creator in her most valuable years. It writes them as problems of statistics, games and pricing. Its parameters are illustrative; all calculations are generated by its programs; the book is not individual legal or financial advice. Chapter 1 uses multiplicative growth with a drop-out floor to show that counting only active creators inflates the head rate 1.74 times and counting only visible creators inflates it 10.70 times, and that the floor itself turns a lognormal into a power law. Chapter 2 fits tail indices by the method of Clauset, Shalizi and Newman and shows that an incubation portfolio with a positive expected return still loses money with probability 0.637 when it holds only 10 signings. Chapter 3 writes off-platform dealing as an inspection game and derives the minimum effective audit rate, 0.4167 for a penalty of 30000 yuan under the reference parameters; when the penalty cannot be collected, raising it reduces audits but not diversion. Chapter 4 uses the Nash bargaining solution and the outside-option principle to show that a fixed split of 0.60 breaks at about 204192 followers, and that a ladder split raises the present value of commission on a simulated roster from 830.56 million yuan to 1253.26 million. Chapter 5 treats the buyout clause as an American call worth 162475 yuan in the reference case, with an early-exercise premium of 0.428 of that value. Chapter 6 combines the ladder split, the buyout and the audit rate into a three-in-one pricing calculator that sets the buyout at 1.712 times the remaining commission. Chapter 7 examines the keeping of retainers, the cock-crow and dog-thief episode and Feng Huan's burning of the bonds in the "Biography of Lord Mengchang" of the Records of the Grand Historian, read against Machiavelli and Adam Smith: when seating is unrelated to crisis value the most valuable retainer sits in the lowest rank with probability 0.323, and at a correlation of 0.3 only 0.007. Chapter 8 gives a principal's checklist, rules for checking copy and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:本书研究网红经纪机构的三个关键时刻:签约前如何估计真实的孵化成功率,达人起势后如何用抽查遏制绕过机构的私下接单(飞单),达人羽翼丰满后如何设计分成与违约买断条款,使机构不至于在她最值钱的几年失去她。书中把它们写成统计、博弈与定价问题。书中参数为示意;全部计算由书中程序生成;本书不是个人化的法律或理财建议。 第一章用带淘汰下限的乘法增长模型说明,只看仍活跃的达人会把头部率放大 1.74 倍,只看被外界看见的达人会放大 10.70 倍,而下限本身会把对数正态变成幂律。第二章用克劳塞特、沙利齐与纽曼的方法拟合尾指数,并说明期望收益为正的孵化组合在只签 10 人时仍有 0.637 的概率亏损。第三章把飞单写成检查博弈,求出最低有效抽查率:参考参数下违约金 30000 元对应 0.4167;违约金收不回时,提高它只会减少抽查而不会减少飞单。第四章用纳什讨价还价解与外部选项原理说明,固定 0.60 的分成在粉丝数约 204192 时破裂,阶梯分成把模拟名册的佣金现值从 830.56 百万元提高到 1253.26 百万元。 第五章把买断条款视为美式看涨期权,参考情形下值 162475 元,其中提前行权溢价占 0.428。第六章把阶梯分成、买断金与抽查率合成三联定价试算器,买断金取剩余佣金的 1.712 倍。第七章考辨《史记·孟尝君列传》的养士、鸡鸣狗盗与冯驩焚券,与马基雅维利、亚当·斯密对读:座次与危机价值无关时,价值最高的门客坐在最下等的概率是 0.323,相关系数升到 0.3 时只剩 0.007。第八章给出主事者清单、文案检查规则与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Survivorship in the Incubation Pool: Multiplicative Growth, Drop-Out Floors and the Heads Outsiders See — Chapter 1 uses multiplicative growth with a drop-out floor to show that counting only active creators inflates the head rate 1.74 times and counting only visible creators inflates it 10.70 times, and that the floor itself turns a lognormal into a power law. (https://5dlibrary.com/answers/en/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c01/) 第1章 孵化池的幸存者偏差:乘法增长、淘汰下限与被看见的头部 — 第一章用带淘汰下限的乘法增长模型说明,只看仍活跃的达人会把头部率放大 1.74 倍,只看被外界看见的达人会放大 10.70 倍,而下限本身会把对数正态变成幂律。 (https://5dlibrary.com/answers/zh/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c01/) - Ch.2 Tail Indices and the Failure of Sample Means: Clauset Fits, Heavy-Tailed Portfolios and the Rule of the Heads — Chapter 2 fits tail indices by the method of Clauset, Shalizi and Newman and shows that an incubation portfolio with a positive expected return still loses money with probability 0.637 when it holds only 10 signings. (https://5dlibrary.com/answers/en/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c02/) 第2章 尾指数与样本均值的失灵:克劳塞特拟合、厚尾组合与头部决定论 — 第二章用克劳塞特、沙利齐与纽曼的方法拟合尾指数,并说明期望收益为正的孵化组合在只签 10 人时仍有 0.637 的概率亏损。 (https://5dlibrary.com/answers/zh/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c02/) - Ch.3 Off-Platform Deals as an Inspection Game: Audit Rates, Penalties and the Mixed Equilibrium of Private Gains — Chapter 3 writes off-platform dealing as an inspection game and derives the minimum effective audit rate, 0.4167 for a penalty of 30000 yuan under the reference parameters; when the penalty cannot be collected, raising it reduces audits but not diversion. (https://5dlibrary.com/answers/en/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c03/) 第3章 飞单的检查博弈:抽查率、罚则与私单收益的混合均衡 — 第三章把飞单写成检查博弈,求出最低有效抽查率:参考参数下违约金 30000 元对应 0.4167;违约金收不回时,提高它只会减少抽查而不会减少飞单。 (https://5dlibrary.com/answers/zh/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c03/) - Ch.4 When a Creator Comes into Her Own: Outside Options, Nash Splits and Ladder Contracts — Chapter 4 uses the Nash bargaining solution and the outside-option principle to show that a fixed split of 0.60 breaks at about 204192 followers, and that a ladder split raises the present value of commission on a simulated roster from 830.56 million yuan to 1253.26 million. (https://5dlibrary.com/answers/en/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c04/) 第4章 羽翼丰满之后:外部选项、纳什分成与阶梯合约 — 第四章用纳什讨价还价解与外部选项原理说明,固定 0.60 的分成在粉丝数约 204192 时破裂,阶梯分成把模拟名册的佣金现值从 830.56 百万元提高到 1253.26 百万元。 (https://5dlibrary.com/answers/zh/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c04/) - Ch.5 The Contract Buyout as an Option: Black-Scholes, Binomial Trees and the Exit Boundary — Chapter 5 treats the buyout clause as an American call worth 162475 yuan in the reference case, with an early-exercise premium of 0.428 of that value. (https://5dlibrary.com/answers/en/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c05/) 第5章 违约买断即期权:布莱克-斯科尔斯、二叉树与退出边界 — 第五章把买断条款视为美式看涨期权,参考情形下值 162475 元,其中提前行权溢价占 0.428。 (https://5dlibrary.com/answers/zh/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c05/) - Ch.6 A Three-in-One Pricing Calculator: Designing the Ladder Split, the Buyout and the Audit Rate Together — Chapter 6 combines the ladder split, the buyout and the audit rate into a three-in-one pricing calculator that sets the buyout at 1.712 times the remaining commission. (https://5dlibrary.com/answers/en/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c06/) 第6章 三联定价试算器:阶梯分成、买断金与抽查率的联合设计 — 第六章把阶梯分成、买断金与抽查率合成三联定价试算器,买断金取剩余佣金的 1.712 倍。 (https://5dlibrary.com/answers/zh/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c06/) - Ch.7 Reading the Classics: Lord Mengchang's Three Thousand Retainers and Machiavelli's Mercenaries — Chapter 7 examines the keeping of retainers, the cock-crow and dog-thief episode and Feng Huan's burning of the bonds in the "Biography of Lord Mengchang" of the Records of the Grand Historian, read against Machiavelli and Adam Smith: when seating is unrelated to crisis value the most valuable retainer sits in the lowest rank with probability 0.323, and at a correlation of 0.3 only 0.007. (https://5dlibrary.com/answers/en/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c07/) 第7章 原典考辨:孟尝君的三千门客与马基雅维利的雇佣兵 — 第七章考辨《史记·孟尝君列传》的养士、鸡鸣狗盗与冯驩焚券,与马基雅维利、亚当·斯密对读:座次与危机价值无关时,价值最高的门客坐在最下等的概率是 0.323,相关系数升到 0.3 时只剩 0.007。 (https://5dlibrary.com/answers/zh/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c07/) - Ch.8 Practice and Falsification: A Principal's Checklist, the Calculator and a Pre-Registered Protocol — Chapter 8 gives a principal's checklist, rules for checking copy and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c08/) 第8章 实践与证伪:主事者清单、试算器说明与预注册方案 — 第八章给出主事者清单、文案检查规则与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/survivor-power-laws-influencer-agency-incubation-sampling-audit-games/c08/) ## 5DL-B-173 · One-Dimensional Time-Slot Packing for Multi-Format Studio Sets, Dynamic Cancellation-Option Ladders and Accelerated Depreciation of High-Wear Equipment 中文书名:多规格实景空间时段一维装箱、动态退改期权阶梯与高损耗器材加速折旧运筹 DOI: pending Cite: Fan, Y. (2026). One-Dimensional Time-Slot Packing for Multi-Format Studio Sets, Dynamic Cancellation-Option Ladders and Accelerated Depreciation of High-Wear Equipment (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not individual tax, legal or investment advice. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: This book studies three kinds of decision in a photography venue with sets of several formats: how to fit orders of unequal length into a limited number of sets and hours so that fragmented time is minimised; how to price and handle cancellations and no-shows; and how to depreciate and replace high-wear equipment and design deposits and damage waivers so that clients treat the space with care. Its parameters are illustrative; all calculations are generated by its programs; the book is not individual tax, legal or investment advice. Chapter 1 uses one-dimensional bin packing to show that first fit decreasing reaches a ratio of 1.2222 on an adversarial instance yet uses very few extra set-days on random order books, and that minimum-vacancy stitching raises the sold share by 0.0457 without any change in price. Chapter 2 shows that placement rules which spread the load evenly do worst, and that holding 4 blocks for higher-paying orders by Littlewood's rule earns 1581 yuan more. Chapter 3 prices each step of a refund ladder by the chance of resale; net revenue per booking under the ladder is 1.1035, above 1.0261 with no refund and 1.0113 with full refunds. Chapter 4 writes overbooking as a newsvendor problem: one extra booking earns 679 yuan more, but once the intra-day correlation of no-shows reaches 0.15 the venue should not overbook. Chapter 5 shows that the present-value gain of the tax shield from double declining balance over straight line is only 0.0139 of the cost, and that the faster second-hand prices fall, the longer equipment should be kept. Chapter 6 finds the incentive boundary of deposits: deposits with sample inspections beat no deposit only while an inspection costs less than 280.8 yuan, and a damage waiver needs a deductible of at least 2800 yuan; it then presents a three-in-one quotation calculator for slot price, refund ladder and deposit. Chapter 7 examines daily rent and rent remission at the Song rental office for state houses, and the standard of care and remission for force majeure in the Roman contract of hire: with reletting taking 20 days, remission pays once the exit share passes 0.1304, and a fault rule needs a judgement accuracy above 0.6085. Chapter 8 gives a principal's checklist to run daily, weekly, monthly and quarterly, describes the distilled kit built around the three-in-one quotation calculator together with its colour tokens and a checker for promotional copy, and sets out a pre-registered test protocol. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:本书研究多规格摄影场馆的三类经营决策:长短不一的订单怎样拼进有限的棚与钟点,使碎片档期最少;退改与爽约怎样定价与应对;高损耗器材怎样折旧与更换,押金与免损服务怎样设计才能让客户愿意小心使用场地。书中参数为示意;全部计算由书中程序生成;本书不是个人化的税务、法律或投资建议。 第一章用一维装箱说明,首次适配降序在对抗实例上达到 1.2222 的比值,在随机订单簿上平均只比最优多用很少的棚日,而最小空置拼接在不改价格的前提下使售出份额提高 0.0457。第二章说明平均分配负荷的放置规则最差,按利特尔伍德规则为高价订单保留 4 档可多赚 1581 元。第三章用再售概率给退改阶梯每一级定价,阶梯下每单净收入为 1.1035,高于不退款的 1.0261 与全额退款的 1.0113。第四章把超订写成报童问题,多接一单多赚 679 元,但爽约的日内相关达到 0.15 时就不应超订。 第五章说明双倍余额递减相对直线法的税盾现值增益只有原价的 0.0139,而二手价跌得越快,器材越应多用一年。第六章求出押金的激励边界:验场成本低于 280.8 元时押金加抽查才优于不收押金,免损服务的免赔额不得低于 2800 元;并给出时段价、退改阶梯与押金的三联报价试算器。第七章考辨宋代楼店务的按日计租与放免房钱,以及罗马法租赁契约中的注意标准与不可抗力减租:再租需 20 天时,放免的临界离开比例是 0.1304,过失责任需要 0.6085 以上的判定准确率。第八章给出按日、周、月、季执行的主事者清单,说明以三联报价试算器为核心、附配色令牌与宣传文案检查器的蒸馏件,并给出预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Time-Slot Packing: the 11/9 Bound of First Fit Decreasing, Exact Dynamic Programming and the Idle Share of Fragmented Calendars — Chapter 1 uses one-dimensional bin packing to show that first fit decreasing reaches a ratio of 1.2222 on an adversarial instance yet uses very few extra set-days on random order books, and that minimum-vacancy stitching raises the sold share by 0.0457 without any change in price. (https://5dlibrary.com/answers/en/one-dimensional-time-slot-bin-packing-multi-format-studio/c01/) 第1章 时段装箱:首次适配降序的 11/9 界、精确动态规划与碎片档期空置率 — 第一章用一维装箱说明,首次适配降序在对抗实例上达到 1.2222 的比值,在随机订单簿上平均只比最优多用很少的棚日,而最小空置拼接在不改价格的前提下使售出份额提高 0.0457。 (https://5dlibrary.com/answers/zh/one-dimensional-time-slot-bin-packing-multi-format-studio/c01/) - Ch.2 Online Arrivals: Greedy Placement, Littlewood Protection Levels and the Minimum Booking Length — Chapter 2 shows that placement rules which spread the load evenly do worst, and that holding 4 blocks for higher-paying orders by Littlewood's rule earns 1581 yuan more. (https://5dlibrary.com/answers/en/one-dimensional-time-slot-bin-packing-multi-format-studio/c02/) 第2章 在线到达:贪心放置、利特尔伍德保留量与最短起订时长 — 第二章说明平均分配负荷的放置规则最差,按利特尔伍德规则为高价订单保留 4 档可多赚 1581 元。 (https://5dlibrary.com/answers/zh/one-dimensional-time-slot-bin-packing-multi-format-studio/c02/) - Ch.3 The Cancellation-Option Ladder: Pricing Each Step by the Chance of Resale, Service Escape and Seasonal Dynamic Ladders — Chapter 3 prices each step of a refund ladder by the chance of resale; net revenue per booking under the ladder is 1.1035, above 1.0261 with no refund and 1.0113 with full refunds. (https://5dlibrary.com/answers/en/one-dimensional-time-slot-bin-packing-multi-format-studio/c03/) 第3章 退改期权阶梯:用再售概率给每一级定价、服务逃逸与淡旺季动态阶梯 — 第三章用再售概率给退改阶梯每一级定价,阶梯下每单净收入为 1.1035,高于不退款的 1.0261 与全额退款的 1.0113。 (https://5dlibrary.com/answers/zh/one-dimensional-time-slot-bin-packing-multi-format-studio/c03/) - Ch.4 Overbooking as a Newsvendor Problem: the Marginal-Booking Rule, the Cost of Compensation and the Trap of Correlated No-Shows — Chapter 4 writes overbooking as a newsvendor problem: one extra booking earns 679 yuan more, but once the intra-day correlation of no-shows reaches 0.15 the venue should not overbook. (https://5dlibrary.com/answers/en/one-dimensional-time-slot-bin-packing-multi-format-studio/c04/) 第4章 超订的报童模型:边际订单法则、补偿成本与成团爽约的相关性陷阱 — 第四章把超订写成报童问题,多接一单多赚 679 元,但爽约的日内相关达到 0.15 时就不应超订。 (https://5dlibrary.com/answers/zh/one-dimensional-time-slot-bin-packing-multi-format-studio/c04/) - Ch.5 Accelerated Depreciation of High-Wear Equipment: Double Declining Balance, Units of Production, the Present Value of the Tax Shield and Economic Life — Chapter 5 shows that the present-value gain of the tax shield from double declining balance over straight line is only 0.0139 of the cost, and that the faster second-hand prices fall, the longer equipment should be kept. (https://5dlibrary.com/answers/en/one-dimensional-time-slot-bin-packing-multi-format-studio/c05/) 第5章 高损耗器材的加速折旧:双倍余额递减、作业量法、税盾现值与经济寿命 — 第五章说明双倍余额递减相对直线法的税盾现值增益只有原价的 0.0139,而二手价跌得越快,器材越应多用一年。 (https://5dlibrary.com/answers/zh/one-dimensional-time-slot-bin-packing-multi-format-studio/c05/) - Ch.6 Deposits and Damage Waivers: the Incentive-Compatibility Boundary under Moral Hazard and a Three-in-One Quotation Calculator — Chapter 6 finds the incentive boundary of deposits: deposits with sample inspections beat no deposit only while an inspection costs less than 280.8 yuan, and a damage waiver needs a deductible of at least 2800 yuan; it then presents a three-in-one quotation calculator for slot price, refund ladder and deposit. (https://5dlibrary.com/answers/en/one-dimensional-time-slot-bin-packing-multi-format-studio/c06/) 第6章 押金与免损:道德风险下的激励相容边界与三联报价试算器 — 第六章求出押金的激励边界:验场成本低于 280.8 元时押金加抽查才优于不收押金,免损服务的免赔额不得低于 2800 元;并给出时段价、退改阶梯与押金的三联报价试算器。 (https://5dlibrary.com/answers/zh/one-dimensional-time-slot-bin-packing-multi-format-studio/c06/) - Ch.7 Classical Sources: the Song Rental Office for State Houses and the Roman Contract of Hire — Chapter 7 examines daily rent and rent remission at the Song rental office for state houses, and the standard of care and remission for force majeure in the Roman contract of hire: with reletting taking 20 days, remission pays once the exit share passes 0.1304, and a fault rule needs a judgement accuracy above 0.6085. (https://5dlibrary.com/answers/en/one-dimensional-time-slot-bin-packing-multi-format-studio/c07/) 第7章 原典考辨:宋代楼店务的官房租赁与罗马法的租赁契约 — 第七章考辨宋代楼店务的按日计租与放免房钱,以及罗马法租赁契约中的注意标准与不可抗力减租:再租需 20 天时,放免的临界离开比例是 0.1304,过失责任需要 0.6085 以上的判定准确率。 (https://5dlibrary.com/answers/zh/one-dimensional-time-slot-bin-packing-multi-format-studio/c07/) - Ch.8 Practice and Falsification: a Checklist for the Principal of a Venue, the Distilled Kit and a Pre-Registered Test — Chapter 8 gives a principal's checklist to run daily, weekly, monthly and quarterly, describes the distilled kit built around the three-in-one quotation calculator together with its colour tokens and a checker for promotional copy, and sets out a pre-registered test protocol. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/one-dimensional-time-slot-bin-packing-multi-format-studio/c08/) 第8章 实践与证伪方案:场馆主事者清单、蒸馏件与预注册检验 — 第八章给出按日、周、月、季执行的主事者清单,说明以三联报价试算器为核心、附配色令牌与宣传文案检查器的蒸馏件,并给出预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/one-dimensional-time-slot-bin-packing-multi-format-studio/c08/) ## 5DL-B-174 · Two-Dimensional Volume-Weight Knapsacks for Cross-Border Buyers, Port Risk-Profiling Inspection Games and the Scale Reversal Point to Formal Customs Clearance 中文书名:跨国买手体积重量双维背包优化、口岸风险画像抽检博弈与正规报关规模反转点 DOI: pending Cite: Fan, Y. (2026). Two-Dimensional Volume-Weight Knapsacks for Cross-Border Buyers, Port Risk-Profiling Inspection Games and the Scale Reversal Point to Formal Customs Clearance (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative and laws differ from place to place; all calculations are generated by its programs; the book is not individual legal or tax advice. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: This book studies one operating decision of a cross-border buying shop: at what size it should stop carrying goods in personal luggage and move to formal customs clearance. The two answers current in the trade, that small volumes go unnoticed and that formal clearance is taxed too heavily, both compare only visible costs. The book exposes the real cost of informal carrying and gives decision rules for moving to compliant declaration; it provides no method for evading customs control, avoiding inspection or escaping duties. Its parameters are illustrative and laws differ from place to place; all calculations are generated by its programs; the book is not individual legal or tax advice. Chapter 1 writes a checked bag of 23 kg and 100 litres as a two-dimensional 0-1 knapsack; dynamic programming finds a best margin of 10460 yuan, while greedy packing by margin per kilogram reaches only 0.9254 of it. Chapter 2 examines billing on the larger of actual and volumetric weight; with a divisor of 6000 the break-even density is 166.67 kg per cubic metre. Chapter 3 treats border inspection as an inspection game; under illustrative parameters the break-even inspection rate for not declaring is 0.1852, and risk profiling makes the inspection rate jump from 0.0380 to 0.5035 between two and four crossings a quarter. Chapter 4 itemises the bill after a finding and shows that tracing and thresholds make the expected hidden cost accelerate with the number of trips: at 12 trips a year the probability of a finding is 0.8889 and the expected hidden cost 144001 yuan. Chapter 5 compares the landed cost of carrying, cross-border e-commerce and general trade: the visible cost of carrying is always 690.0 yuan a unit, yet its expected cost already reaches 930.8 at 40 units a month, and the scale reversal point is pinned at about ten trips a year, insensitive to unit value and tax rate. Chapter 6 simulates a year of cash flow by Monte Carlo with a fixed seed, finds a conditional value at risk of 146183 yuan for carrying at 20 units a month, and presents the landed-cost calculator. Chapter 7 examines the salt policy of the Guanzi "Hai Wang" and Adam Smith on smugglers: the official markup on imported salt is 6.667 times, and Swift's arithmetic holds at a semi-elasticity of evasion of 3, where doubling the rate multiplies revenue by 0.8131. Chapter 8 gives checklists for buyers and owners, copy-checking rules and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:本书研究跨国买手店的一个经营决策:生意做到多大,就应当从人肉带货转向正规报关。坊间的两种说法,「量小没人管」与「正规税太重」,都只比较了看得见的成本。本书揭示非正规带货的真实成本并给出转向合规报关的决策准则,不提供任何规避海关监管、逃避查验或偷逃税款的方法。书中参数为示意,各地法规不同;全部计算由书中程序生成;本书不是个人化的法律或税务建议。 第一章把一只 23 公斤、100 升的托运箱写成二维 0-1 背包,动态规划的最优毛利是 10460 元,按每公斤毛利贪心只得到 0.9254。第二章讨论体积重与实重取大的计费规则,系数为 6000 时的盈亏平衡密度是每立方米 166.67 公斤。第三章把口岸查验写成检查博弈,示意参数下不申报的盈亏平衡查验率是 0.1852,而风险画像使查验率在每季度两到四趟之间从 0.0380 跳到 0.5035。第四章拆开被查之后的账单,说明追溯与门槛使期望隐性成本随行程数加速增长:一年 12 趟时全年被查获的概率是 0.8889,期望隐性成本 144001 元。 第五章比较带货、跨境电商与一般贸易的到岸成本:带货看得见的成本永远是每件 690.0 元,但每月 40 件时期望成本已达 930.8 元,规模反转点大约钉在一年十趟行程上,对货值与税率都不敏感。第六章用固定种子的蒙特卡洛模拟一年现金流,每月 20 件时带货的条件风险价值是 146183 元,并给出到岸成本试算器。第七章考辨《管子·海王》的盐策与亚当·斯密论走私者,管子笔下进口盐的官府加价是 6.667 倍,斯威夫特的算术在逃税半弹性为 3 时成立:税率翻倍,收入乘以 0.8131。第八章给出买手与店主清单、文案检查规则与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 One Suitcase as a Two-Dimensional Knapsack: Weight, Volume and the Price of Greedy Density — Chapter 1 writes a checked bag of 23 kg and 100 litres as a two-dimensional 0-1 knapsack; dynamic programming finds a best margin of 10460 yuan, while greedy packing by margin per kilogram reaches only 0.9254 of it. (https://5dlibrary.com/answers/en/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c01/) 第1章 一只箱子的二维背包:重量、体积与价值密度贪心的代价 — 第一章把一只 23 公斤、100 升的托运箱写成二维 0-1 背包,动态规划的最优毛利是 10460 元,按每公斤毛利贪心只得到 0.9254。 (https://5dlibrary.com/answers/zh/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c01/) - Ch.2 The Larger of Actual and Volumetric Weight: Break-Even Density, Mixed Loads and Vacuum Packing — Chapter 2 examines billing on the larger of actual and volumetric weight; with a divisor of 6000 the break-even density is 166.67 kg per cubic metre. (https://5dlibrary.com/answers/en/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c02/) 第2章 体积重与实重取大:计费密度、轻重拼装与真空压缩 — 第二章讨论体积重与实重取大的计费规则,系数为 6000 时的盈亏平衡密度是每立方米 166.67 公斤。 (https://5dlibrary.com/answers/zh/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c02/) - Ch.3 The Border as an Inspection Game: Mixed Equilibria of Declaring and Risk Profiling — Chapter 3 treats border inspection as an inspection game; under illustrative parameters the break-even inspection rate for not declaring is 0.1852, and risk profiling makes the inspection rate jump from 0.0380 to 0.5035 between two and four crossings a quarter. (https://5dlibrary.com/answers/en/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c03/) 第3章 口岸检查博弈:申报与不申报的混合均衡和风险画像 — 第三章把口岸查验写成检查博弈,示意参数下不申报的盈亏平衡查验率是 0.1852,而风险画像使查验率在每季度两到四趟之间从 0.0380 跳到 0.5035。 (https://5dlibrary.com/answers/zh/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c03/) - Ch.4 The Nonlinearity of Hidden Penalties: Back Taxes, Confiscation, Tracing and Credit — Chapter 4 itemises the bill after a finding and shows that tracing and thresholds make the expected hidden cost accelerate with the number of trips: at 12 trips a year the probability of a finding is 0.8889 and the expected hidden cost 144001 yuan. (https://5dlibrary.com/answers/en/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c04/) 第4章 隐性罚没成本的非线性:补税、没收、追溯与信用 — 第四章拆开被查之后的账单,说明追溯与门槛使期望隐性成本随行程数加速增长:一年 12 趟时全年被查获的概率是 0.8889,期望隐性成本 144001 元。 (https://5dlibrary.com/answers/zh/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c04/) - Ch.5 The Scale Reversal Point: Landed Cost of Carrying, Cross-Border E-Commerce and General Trade — Chapter 5 compares the landed cost of carrying, cross-border e-commerce and general trade: the visible cost of carrying is always 690.0 yuan a unit, yet its expected cost already reaches 930.8 at 40 units a month, and the scale reversal point is pinned at about ten trips a year, insensitive to unit value and tax rate. (https://5dlibrary.com/answers/en/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c05/) 第5章 规模反转点:带货、跨境电商与一般贸易的到岸成本 — 第五章比较带货、跨境电商与一般贸易的到岸成本:带货看得见的成本永远是每件 690.0 元,但每月 40 件时期望成本已达 930.8 元,规模反转点大约钉在一年十趟行程上,对货值与税率都不敏感。 (https://5dlibrary.com/answers/zh/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c05/) - Ch.6 Tail Risk in a Year of Cash Flow and the Landed-Cost Calculator — Chapter 6 simulates a year of cash flow by Monte Carlo with a fixed seed, finds a conditional value at risk of 146183 yuan for carrying at 20 units a month, and presents the landed-cost calculator. (https://5dlibrary.com/answers/en/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c06/) 第6章 一年现金流的尾部风险与到岸成本试算器 — 第六章用固定种子的蒙特卡洛模拟一年现金流,每月 20 件时带货的条件风险价值是 146183 元,并给出到岸成本试算器。 (https://5dlibrary.com/answers/zh/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c06/) - Ch.7 Classical Sources: the Salt Policy of the Guanzi "Hai Wang" and Smith's Smuggler — Chapter 7 examines the salt policy of the Guanzi "Hai Wang" and Adam Smith on smugglers: the official markup on imported salt is 6.667 times, and Swift's arithmetic holds at a semi-elasticity of evasion of 3, where doubling the rate multiplies revenue by 0.8131. (https://5dlibrary.com/answers/en/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c07/) 第7章 原典考辨:《管子·海王》的盐策与斯密笔下的走私者 — 第七章考辨《管子·海王》的盐策与亚当·斯密论走私者,管子笔下进口盐的官府加价是 6.667 倍,斯威夫特的算术在逃税半弹性为 3 时成立:税率翻倍,收入乘以 0.8131。 (https://5dlibrary.com/answers/zh/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c07/) - Ch.8 Practice and Falsification: Checklists for Buyers and Owners, the Calculator and a Pre-Registered Protocol — Chapter 8 gives checklists for buyers and owners, copy-checking rules and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c08/) 第8章 实践与证伪:买手与店主清单、试算器说明与预注册方案 — 第八章给出买手与店主清单、文案检查规则与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/two-dimensional-volume-weight-knapsacks-cross-border-buyers-port/c08/) ## 5DL-B-175 · Patience-Decay Distributions in Private-Message Queues, Red-Amber-Green Human-AI Handoff Thresholds and Markov Blocking of Complaint Escalation 中文书名:私域对话队列耐心衰减分布、人机协同红黄绿灯移交门限与客诉升级马尔可夫阻断 DOI: pending Cite: Fan, Y. (2026). Patience-Decay Distributions in Private-Message Queues, Red-Amber-Green Human-AI Handoff Thresholds and Markov Blocking of Complaint Escalation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not personalised legal, financial or business advice. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: While a high-intent buyer waits in a private-message channel, the owner must decide how fast to reply, after how many seconds the AI should hand the conversation to a person, which signals show that a conversation has turned dangerous, where to intervene when the buyer is dissatisfied, and how to make the buyer remember being looked after. This book writes these decisions as runnable mathematics and states what data would refute each conclusion. Its parameters are illustrative; all calculations are generated by its programs; it is not personalised legal, financial or business advice. Chapter 1 shows that with the same mean patience the hazard can rise, fall, or rise and then fall; under exponential patience a buyer who has waited any length of time leaves within the next 30 seconds with probability 0.221; a falling population hazard may be nothing but the impatient leaving first, and censoring makes simple averages of chat logs wrong. Chapter 2 uses Erlang-A to show that for the same six-person team Erlang C reports a mean wait of 103.6 seconds while, with abandonment counted, it is only 13.1 seconds, at the price of 0.109 of high-intent buyers leaving while they wait. Chapter 3 finds that a human handoff can pay only above a margin of 16.0 yuan, and that once the cutoff is applied to every buyer, the best cutoff for a five-person team raises the expected value per buyer from 202.6 to 270.2 yuan. Chapter 4 shows that red-amber-green thresholds must move with margin: a threshold fixed at the midpoint costs 1232.9 yuan on expensive goods where the optimum costs 550.2. Chapter 5 writes complaint escalation as an absorbing Markov chain: an inquiry ends in public exposure with probability 0.01469, and blocking upstream with a complete first answer saves 24.18 yuan per yuan spent, though above a unit price of 2.57 yuan intervention at dissatisfaction is better. Chapter 6 treats Maister's eight propositions on waiting, the peak-end rule and the fairness cost of a fast lane. Chapter 7 examines the nine postures of the Guiguzi's "Fu Yan" and Erlang's telephone-exchange papers of 1909 and 1917, finding that push notices raise the variance-to-mean ratio of message arrivals to 1.992 and break the Poisson premise. Chapter 8 gives an owner's checklist, six red lines for scripts and a pre-registered testing plan whose largest test needs 3470 conversations per arm. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:高意向买手在私讯里等待时,店主要决定多快回复、AI 接待多少秒后转给真人、什么信号说明对话已经危险、买手不满时在哪一步介入,以及怎样让买手记住的是被照顾。本书把这些决定写成可以运行的数学,并写明每一条结论在什么数据下会被推翻。书中参数为示意;全部计算由书中程序生成;本书不是法律、财务或经营上的个人化建议。 第一章说明,同样的平均耐心下,风险率可以上升、下降或先升后降,指数耐心下等过任何时刻的买手在下 30 秒离开的概率都是 0.221;下降的总体风险率可能只是急性子先走光,聊天记录的删失又会让直接平均出错。第二章用 Erlang-A 说明,同一支六人团队,Erlang C 给出 103.6 秒的平均等待,计入放弃后只有 13.1 秒,代价是 0.109 的高意向买手在等待中离开。第三章算出转人工只有在毛利高于 16.0 元时才可能划算,并说明把截断作用于所有买手后,五人团队的最优截断使每位买手的期望价值从 202.6 元升到 270.2 元。第四章说明红黄绿灯门限必须随毛利移动,固定在中点的门限对高价商品的代价是 1232.9 元,而最优只有 550.2 元。 第五章把客诉升级写成吸收马尔可夫链,一次咨询最终走向公开曝光的概率是 0.01469,上游的首答阻断每花一元省下 24.18 元,但其单价超过 2.57 元时应改为不满介入。第六章讨论梅斯特的等待八命题、峰终规则与快速通道的公平代价。第七章考辨《鬼谷子·符言》的九条主术与埃尔朗 1909 年和 1917 年的电话局论文,发现推送使私讯到达的方差均值比升到 1.992,打破泊松前提。第八章给出店主清单、六条话术红线与预注册检验方案,最大的一项每组需要 3470 段对话。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Shape of Patience: Exponential, Weibull and Lognormal Hazards, and Estimating Patience under Censoring — Chapter 1 shows that with the same mean patience the hazard can rise, fall, or rise and then fall; under exponential patience a buyer who has waited any length of time leaves within the next 30 seconds with probability 0.221; a falling population hazard may be nothing but the impatient leaving first, and censoring makes simple averages of chat logs wrong. (https://5dlibrary.com/answers/en/patience-decay-distributions-private-message-queues-red-amber-green/c01/) 第1章 耐心的形状:指数、威布尔与对数正态的风险率,以及删失下的耐心估计 — 第一章说明,同样的平均耐心下,风险率可以上升、下降或先升后降,指数耐心下等过任何时刻的买手在下 30 秒离开的概率都是 0.221;下降的总体风险率可能只是急性子先走光,聊天记录的删失又会让直接平均出错。 (https://5dlibrary.com/answers/zh/patience-decay-distributions-private-message-queues-red-amber-green/c01/) - Ch.2 Queues that Give Up: Erlang C, Erlang-A and Square-Root Staffing — Chapter 2 uses Erlang-A to show that for the same six-person team Erlang C reports a mean wait of 103.6 seconds while, with abandonment counted, it is only 13.1 seconds, at the price of 0.109 of high-intent buyers leaving while they wait. (https://5dlibrary.com/answers/en/patience-decay-distributions-private-message-queues-red-amber-green/c02/) 第2章 会放弃的队列:Erlang C、Erlang-A 与平方根配员 — 第二章用 Erlang-A 说明,同一支六人团队,Erlang C 给出 103.6 秒的平均等待,计入放弃后只有 13.1 秒,代价是 0.109 的高意向买手在等待中离开。 (https://5dlibrary.com/answers/zh/patience-decay-distributions-private-message-queues-red-amber-green/c02/) - Ch.3 The Handoff Cutoff in Seconds: Expected Lost Value, Human Cost and Queue Feedback — Chapter 3 finds that a human handoff can pay only above a margin of 16.0 yuan, and that once the cutoff is applied to every buyer, the best cutoff for a five-person team raises the expected value per buyer from 202.6 to 270.2 yuan. (https://5dlibrary.com/answers/en/patience-decay-distributions-private-message-queues-red-amber-green/c03/) 第3章 转人工截断秒数:期望流失价值、人工成本与队列反馈 — 第三章算出转人工只有在毛利高于 16.0 元时才可能划算,并说明把截断作用于所有买手后,五人团队的最优截断使每位买手的期望价值从 202.6 元升到 270.2 元。 (https://5dlibrary.com/answers/zh/patience-decay-distributions-private-message-queues-red-amber-green/c03/) - Ch.4 Red, Amber, Green: Cost-Sensitive Handoff Thresholds on the ROC Curve — Chapter 4 shows that red-amber-green thresholds must move with margin: a threshold fixed at the midpoint costs 1232.9 yuan on expensive goods where the optimum costs 550.2. (https://5dlibrary.com/answers/en/patience-decay-distributions-private-message-queues-red-amber-green/c04/) 第4章 红黄绿灯:ROC 曲线上的代价敏感移交门限 — 第四章说明红黄绿灯门限必须随毛利移动,固定在中点的门限对高价商品的代价是 1232.9 元,而最优只有 550.2 元。 (https://5dlibrary.com/answers/zh/patience-decay-distributions-private-message-queues-red-amber-green/c04/) - Ch.5 A Markov Chain of Complaint Escalation: Absorption Probabilities and the Cheapest Point of Blocking — Chapter 5 writes complaint escalation as an absorbing Markov chain: an inquiry ends in public exposure with probability 0.01469, and blocking upstream with a complete first answer saves 24.18 yuan per yuan spent, though above a unit price of 2.57 yuan intervention at dissatisfaction is better. (https://5dlibrary.com/answers/en/patience-decay-distributions-private-message-queues-red-amber-green/c05/) 第5章 客诉升级马尔可夫链:吸收概率与最划算的阻断点 — 第五章把客诉升级写成吸收马尔可夫链,一次咨询最终走向公开曝光的概率是 0.01469,上游的首答阻断每花一元省下 24.18 元,但其单价超过 2.57 元时应改为不满介入。 (https://5dlibrary.com/answers/zh/patience-decay-distributions-private-message-queues-red-amber-green/c05/) - Ch.6 The Mental Accounting of Waiting: the Peak-End Rule, Eight Propositions on Waiting, and the Fairness Cost of Priority — By the stopwatch, B is 30 seconds longer than A and total discomfort rises from 47 to 52; yet under the peak-end rule A scores 7.0 in memory and B only 5.0. To check that the ranking does not rest on one memory rule alone, the program also uses a blend that weights mean, peak and end by 0.3, 0.35 and 0.35; A scores 6.66 and B 5.06, and the ranking holds. Progress notes act on the peak: C's peak falls from 8 to 6 and its peak-end score from 7.0 to 6.0. D, which combines both, has a peak-end score of only 4.0 and a blended score of 4.06, the best of the four scripts. (https://5dlibrary.com/answers/en/patience-decay-distributions-private-message-queues-red-amber-green/c06/) 第6章 等待的心理账:峰终规则、等待八原则与优先级的公平代价 — 按秒表算,B 比 A 多 30 秒,不适的总量从 47 增加到 52;可是按峰终规则,A 的回顾评分是 7.0,B 只有 5.0。为了确认这个排序不依赖于单一的记忆规则,程序还用一个把均值、峰值与终值按 0.3、0.35、0.35 加权的混合规则复核,A 是 6.66,B 是 5.06,排序不变。进度提示的作用体现在峰值上:C 的峰值从 8 降到 6,峰终评分从 7.0 降到 6.0。两者叠加的 D,峰终评分只有 4.0,混合规则 4.06,是四个脚本中最好的。 (https://5dlibrary.com/answers/zh/patience-decay-distributions-private-message-queues-red-amber-green/c06/) - Ch.7 The Classics Examined: the Nine Postures of the Guiguzi's "Fu Yan" and Erlang's Telephone Exchange — Chapter 7 examines the nine postures of the Guiguzi's "Fu Yan" and Erlang's telephone-exchange papers of 1909 and 1917, finding that push notices raise the variance-to-mean ratio of message arrivals to 1.992 and break the Poisson premise. (https://5dlibrary.com/answers/en/patience-decay-distributions-private-message-queues-red-amber-green/c07/) 第7章 原典考辨:《鬼谷子·符言》的九条主术与埃尔朗的电话局 — 第七章考辨《鬼谷子·符言》的九条主术与埃尔朗 1909 年和 1917 年的电话局论文,发现推送使私讯到达的方差均值比升到 1.992,打破泊松前提。 (https://5dlibrary.com/answers/zh/patience-decay-distributions-private-message-queues-red-amber-green/c07/) - Ch.8 Practice and Falsification: a Closing-Skills Kit for Private-Message Service, Pre-Registered Tests and Refutation Conditions — Chapter 8 gives an owner's checklist, six red lines for scripts and a pre-registered testing plan whose largest test needs 3470 conversations per arm. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/patience-decay-distributions-private-message-queues-red-amber-green/c08/) 第8章 实践与证伪:私域客服逼单技能包、预注册检验与推翻条件 — 第八章给出店主清单、六条话术红线与预注册检验方案,最大的一项每组需要 3470 段对话。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/patience-decay-distributions-private-message-queues-red-amber-green/c08/) ## 5DL-B-176 · Hyperlink Authority Propagation in Machine-Readable Knowledge Bases, Retrieval-Truncation Games in Generative Answering and High-SNR Snippet Extraction 中文书名:机器可读知识库超链接权威传播、生成式问答检索截断博弈与高信噪比摘要提取 DOI: pending Cite: Fan, Y. (2026). Hyperlink Authority Propagation in Machine-Readable Knowledge Bases, Retrieval-Truncation Games in Generative Answering and High-SNR Snippet Extraction (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not personalised legal, financial or business advice and makes no promise about citation rankings in any system. Chapter 1 computes PageRank and HITS with power iteration and a random surfer: in an illustrative knowledge base with damping factor 0.85, the mother page receives 0.3240, or 2.59 times the uniform share. Abstract: When readers put their questions directly to a generative answering system, will the pages of a knowledge base be found, placed in the context and cited? Popular advice abounds, yet almost none of it says under what conditions it holds, how large its effect is, or what result would show it wrong. This book models, layer by layer, the path a piece of text travels before it is cited: authority on the link graph, lexical retrieval scores, truncation of the context, competition among content providers, and the reranker's choice. Its parameters are illustrative; all calculations are generated by its programs; the book is not personalised legal, financial or business advice and makes no promise about citation rankings in any system. Chapter 1 computes PageRank and HITS with power iteration and a random surfer: in an illustrative knowledge base with damping factor 0.85, the mother page receives 0.3240, or 2.59 times the uniform share. Chapter 2 proves that under star-shaped internal linking the mother page's share has a ceiling of $d/(1+d)$, or 0.4595, independent of the size of the base. Chapter 3 implements BM25 with the standard library and shows that stuffing soon peaks while padding is discounted by length: a twelve-word direct answer padded to two hundred and forty words keeps only 0.302 of its score. Chapter 4 computes inclusion probabilities under top-k truncation: a ninety percent chance of the top three needs a score advantage of 2.46 standard deviations. Chapter 5 borrows position auctions and ratio contests to show that ten competitors burn 0.900 of a position's value, and that systems which reward duplication homogenise content. Chapter 6, the core of the book, derives the selection gain of a direct-answer snippet in a softmax reranker: with twenty candidates at temperature one the gain is 4.65; structured elements lift the probability of selection to 0.5866 while rivals rarely adopt them, and to only 0.1204 once rivals adopt them widely. Chapter 7 reads Zheng Qiao's Treatise on Collation beside Vannevar Bush's As We May Think, and shows with a program that classification helps retrieval only when its accuracy exceeds 0.667. Chapter 8 gives a publisher's checklist, a copy checker and a pre-registered test; detecting a moderate lift needs 1,081 runs per group. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:当读者直接向生成式问答系统提问时,一个知识库的页面能否被找到、放进上下文并被引用?流行的建议很多,却几乎都不说明在什么条件下成立、效果多大、什么结果会证明它错。本书沿着一段文字被引用之前要经过的路径逐层建模:链接图上的权威、词法检索的打分、上下文的截断、内容方之间的竞争、重排器的选择。书中参数为示意;全部计算由书中程序生成;本书不是法律、理财或经营方面的个人化建议,也不对任何系统中的引用排名作出承诺。 第一章用幂迭代与随机冲浪者计算 PageRank 与 HITS:在阻尼系数 0.85 的示意知识库里,母典页拿到 0.3240,是均匀份额的 2.59 倍。第二章证明星形内链下母典页的份额有一个与库的大小无关的上限 $d/(1+d)$,即 0.4595。第三章用标准库实现 BM25,说明堆砌很快到顶、填充按长度打折:十二个词的直答句填充到二百四十个词只剩原分数的 0.302。第四章计算 top-k 截断下的入选概率:想有九成把握进入前三名,需要 2.46 个标准差的分数优势。第五章借用位置拍卖与比例竞赛,说明十个竞争者会烧掉位置价值的 0.900,而奖励重复的系统使内容同质化。 第六章是全书核心,推导直答片段在 softmax 重排器中的获选概率增益:在二十个候选、温度为一的示意设定下增益为 4.65;结构化元素在竞争者很少采用时把获选概率提高到 0.5866,在竞争者普遍采用后只剩 0.1204。第七章对读郑樵《通志·校雠略》与布什《诚如我们所思》,用程序说明分类只有在准确率高于 0.667 时才帮助检索。第八章给出出版者清单、文案检查器与预注册检验方案,检测一个中等提升每组需要 1,081 次运行。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Power Iteration and the Damping Factor: How Hyperlink Authority Flows Through a Knowledge Base — Chapter 1 computes PageRank and HITS with power iteration and a random surfer: in an illustrative knowledge base with damping factor 0.85, the mother page receives 0.3240, or 2.59 times the uniform share. (https://5dlibrary.com/answers/en/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c01/) 第1章 幂迭代与阻尼系数:超链接权威在知识库里怎样流动 — 第一章用幂迭代与随机冲浪者计算 PageRank 与 HITS:在阻尼系数 0.85 的示意知识库里,母典页拿到 0.3240,是均匀份额的 2.59 倍。 (https://5dlibrary.com/answers/zh/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c01/) - Ch.2 The Mother Page: How Internal-Link Topology Concentrates Authority — Chapter 2 proves that under star-shaped internal linking the mother page's share has a ceiling of $d/(1+d)$, or 0.4595, independent of the size of the base. (https://5dlibrary.com/answers/en/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c02/) 第2章 母典页:内链拓扑如何把权威集中到一处 — 第二章证明星形内链下母典页的份额有一个与库的大小无关的上限 $d/(1+d)$,即 0.4595。 (https://5dlibrary.com/answers/zh/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c02/) - Ch.3 BM25 and Length Normalisation: the Term-Frequency Ledger of Stuffing, Padding and Direct Answers — Chapter 3 implements BM25 with the standard library and shows that stuffing soon peaks while padding is discounted by length: a twelve-word direct answer padded to two hundred and forty words keeps only 0.302 of its score. (https://5dlibrary.com/answers/en/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c03/) 第3章 BM25 与长度归一化:堆砌、填充与直答句的词频账 — 第三章用标准库实现 BM25,说明堆砌很快到顶、填充按长度打折:十二个词的直答句填充到二百四十个词只剩原分数的 0.302。 (https://5dlibrary.com/answers/zh/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c03/) - Ch.4 Top-k Truncation: the Probability that a Passage Enters the Context — Chapter 4 computes inclusion probabilities under top-k truncation: a ninety percent chance of the top three needs a score advantage of 2.46 standard deviations. (https://5dlibrary.com/answers/en/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c04/) 第4章 top-k 截断:一个文段被放进上下文的概率 — 第四章计算 top-k 截断下的入选概率:想有九成把握进入前三名,需要 2.46 个标准差的分数优势。 (https://5dlibrary.com/answers/zh/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c04/) - Ch.5 Competing for k Positions: Position Auctions, Homogenisation and the Commons — Chapter 5 borrows position auctions and ratio contests to show that ten competitors burn 0.900 of a position's value, and that systems which reward duplication homogenise content. (https://5dlibrary.com/answers/en/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c05/) 第5章 争夺 k 个位置:位置拍卖、同质化与公地 — 第五章借用位置拍卖与比例竞赛,说明十个竞争者会烧掉位置价值的 0.900,而奖励重复的系统使内容同质化。 (https://5dlibrary.com/answers/zh/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c05/) - Ch.6 Signal-to-Noise and Softmax Reranking: the Selection Gain of Direct-Answer Snippets — Chapter 6, the core of the book, derives the selection gain of a direct-answer snippet in a softmax reranker: with twenty candidates at temperature one the gain is 4.65; structured elements lift the probability of selection to 0.5866 while rivals rarely adopt them, and to only 0.1204 once rivals adopt them widely. (https://5dlibrary.com/answers/en/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c06/) 第6章 信噪比与 softmax 重排:直答片段的获选概率增益 — 第六章是全书核心,推导直答片段在 softmax 重排器中的获选概率增益:在二十个候选、温度为一的示意设定下增益为 4.65;结构化元素在竞争者很少采用时把获选概率提高到 0.5866,在竞争者普遍采用后只剩 0.1204。 (https://5dlibrary.com/answers/zh/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c06/) - Ch.7 The Classics Examined: Zheng Qiao's Classification and Bush's Associative Trails — Chapter 7 reads Zheng Qiao's Treatise on Collation beside Vannevar Bush's As We May Think, and shows with a program that classification helps retrieval only when its accuracy exceeds 0.667. (https://5dlibrary.com/answers/en/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c07/) 第7章 原典考辨:郑樵《校雠略》的类例与布什的联想轨迹 — 第七章对读郑樵《通志·校雠略》与布什《诚如我们所思》,用程序说明分类只有在准确率高于 0.667 时才帮助检索。 (https://5dlibrary.com/answers/zh/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c07/) - Ch.8 Practice and Falsification: the Copy Checker, a Publisher's Checklist and a Pre-Registered Test — Chapter 8 gives a publisher's checklist, a copy checker and a pre-registered test; detecting a moderate lift needs 1,081 runs per group. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c08/) 第8章 实践与证伪方案:文案检查器、出版清单与预注册检验 — 第八章给出出版者清单、文案检查器与预注册检验方案,检测一个中等提升每组需要 1,081 次运行。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/hyperlink-authority-propagation-machine-readable-knowledge-bases-retrieval-truncation/c08/) ## 5DL-B-177 · Liability Percolation across Subsidiary Graphs, Tail-Ruin Amplification from Personal Guarantees and Risk-Bucket Partitioning of Operating Assets 中文书名:多子公司责任隔离图渗流、个人连带担保尾部破产放大与经营资产风险仓切分 DOI: pending Cite: Fan, Y. (2026). Liability Percolation across Subsidiary Graphs, Tail-Ruin Amplification from Personal Guarantees and Risk-Bucket Partitioning of Operating Assets (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. Abstract: Owners who run several companies often take limited liability for a firewall. This book draws a business group as a liability graph, whose nodes are the entities and the owner and whose edges are guarantees, mutual guarantees, cash pools and intercompany balances, and computes how much of the firewall is left after a single cross-guarantee, a personal joint guarantee or a court's consolidation order. Its balance sheets, loss laws and legal parameters are illustrative; all calculations are generated by its programs; it is not personalised legal, tax or financial advice, rules differ between jurisdictions, and it offers no method for avoiding lawful debts or escaping enforcement. Chapter 1 uses the Eisenberg–Noe clearing vector to show that once a parent guarantees one subsidiary, the same shock topples four entities instead of one and the loss landing outside the shocked entity rises from 12.50 to 41.10; in Merton's structural model the guarantee multiplies the parent's chance of default by between 1.47 and 3.57. Chapter 2 uses threshold contagion and an exact enumeration of all 1,024 mutual-guarantee patterns of a small group to show that guarantee payments do not thin out as partners multiply. Chapter 3 locates the percolation threshold: at 1.2 guarantee ties per entity the giant component already holds 0.3137, and a ring-fenced entity with one liability edge is reached by the risky entity with probability 0.1129 instead of zero. Chapter 4 shows that a personal guarantee multiplies the owner's ten-year chance of ruin by 7.0 under light tails and 48.6 under heavy tails, ruin being driven by one large year. Chapter 5 shows that the effect of risk buckets depends on the tail: eight buckets raise the chance of ruin from 0.0102 to 0.0355 under light tails and cut it from 0.0489 to 0.0027 under heavy tails, while a chain of seven guarantees between buckets raises it to 0.0943. Chapter 6 does the arithmetic of two isolation layers: with clean, independent layers the chance of getting through is 0.0050, but the multiplication fails once a layer signs a guarantee or was set up while insolvent, so timing and solvency come first. Chapter 7 reads Fan Li's "three times a thousand in gold, twice dispersed" in the Records of the Grand Historian (three tenfold rounds need growth to 1.4384 times a year) against Salomon in the House of Lords. Chapter 8 gives a pricing calculator (default yearly chance of ruin 0.00711, correction factor 1.25), an owner's checklist and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. 摘要:开设多家公司的经营者常以为有限责任是一道防火墙。本书把企业集团画成一张责任图:节点是各个实体与经营者本人,边是担保、互保、资金池与往来款,然后计算单独一笔交叉担保、一份个人连带保证或一次法院合并会让这道防火墙剩下多少。书中的资产负债表、损失分布与法律参数均为示意;全部计算由书中程序生成;本书不是个人化的法律、税务或财务建议,各法域规则不同,书中也不提供任何规避合法债务或逃避执行的方法。 第一章用 Eisenberg–Noe 清算向量说明,母公司为一家子公司提供一笔担保后,同一次冲击使违约实体从一个增加到四个,落在出事实体以外的损失从 12.50 增加到 41.10;在 Merton 结构模型中,这笔担保使母公司违约概率变为原来的 1.47 倍到 3.57 倍。第二章用阈值传染与对全部 1,024 种互保图样的精确枚举说明,互保的代偿不随伙伴增多而摊薄。第三章给出渗流阈值:平均担保关系为 1.2 时巨连通分量已占 0.3137;一个只有一条责任边的隔离实体,被风险实体波及的概率从零跳到 0.1129。第四章说明个人连带担保把经营者十年内破产的概率放大 7.0 倍(轻尾)到 48.6 倍(重尾),破产由单独一年的大损失主导。 第五章说明风险仓切分的效果取决于尾部:切成八个仓,轻尾时破产概率从 0.0102 升到 0.0355,重尾时从 0.0489 降到 0.0027;七条仓间担保串成的链使它升到 0.0943。第六章做双层隔离算术:两层干净独立时穿透概率为 0.0050,一层签了担保或在资不抵债时设立,这个乘法就失效,设立时点与偿债能力是一切分层的前提。第七章对读《史记·货殖列传》陶朱公「三致千金,再分散」(三轮十倍需每年增长到 1.4384 倍)与英国上议院 Salomon 案。第八章给出定价试算器(默认年度破产概率 0.00711,校正因子 1.25)、经营者清单与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Ledger of the Firewall: Clearing Vectors and One Cross-Guarantee — Chapter 1 uses the Eisenberg–Noe clearing vector to show that once a parent guarantees one subsidiary, the same shock topples four entities instead of one and the loss landing outside the shocked entity rises from 12.50 to 41.10; in Merton's structural model the guarantee multiplies the parent's chance of default by between 1.47 and 3.57. (https://5dlibrary.com/answers/en/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c01/) 第1章 防火墙的账本:清算向量与一笔交叉担保 — 第一章用 Eisenberg–Noe 清算向量说明,母公司为一家子公司提供一笔担保后,同一次冲击使违约实体从一个增加到四个,落在出事实体以外的损失从 12.50 增加到 41.10;在 Merton 结构模型中,这笔担保使母公司违约概率变为原来的 1.47 倍到 3.57 倍。 (https://5dlibrary.com/answers/zh/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c01/) - Ch.2 Threshold Contagion: Default Avalanches on Guarantee Graphs and an Exact Enumeration — Chapter 2 uses threshold contagion and an exact enumeration of all 1,024 mutual-guarantee patterns of a small group to show that guarantee payments do not thin out as partners multiply. (https://5dlibrary.com/answers/en/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c02/) 第2章 阈值传染:担保图上的违约雪崩与精确枚举 — 第二章用阈值传染与对全部 1,024 种互保图样的精确枚举说明,互保的代偿不随伙伴增多而摊薄。 (https://5dlibrary.com/answers/zh/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c02/) - Ch.3 The Percolation Threshold: Guarantee Density, the Giant Component and the Failure of Ring-Fencing — Chapter 3 locates the percolation threshold: at 1.2 guarantee ties per entity the giant component already holds 0.3137, and a ring-fenced entity with one liability edge is reached by the risky entity with probability 0.1129 instead of zero. (https://5dlibrary.com/answers/en/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c03/) 第3章 渗流阈值:交叉担保密度、巨连通分量与隔离失效 — 第三章给出渗流阈值:平均担保关系为 1.2 时巨连通分量已占 0.3137;一个只有一条责任边的隔离实体,被风险实体波及的概率从零跳到 0.1129。 (https://5dlibrary.com/answers/zh/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c03/) - Ch.4 Personal Guarantees: Ruin Amplification under Heavy-Tailed Losses and the Single Big Jump — Chapter 4 shows that a personal guarantee multiplies the owner's ten-year chance of ruin by 7.0 under light tails and 48.6 under heavy tails, ruin being driven by one large year. (https://5dlibrary.com/answers/en/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c04/) 第4章 个人连带担保:重尾损失下的破产放大与单次大跳 — 第四章说明个人连带担保把经营者十年内破产的概率放大 7.0 倍(轻尾)到 48.6 倍(重尾),破产由单独一年的大损失主导。 (https://5dlibrary.com/answers/zh/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c04/) - Ch.5 Partitioning into Risk Buckets: Number of Buckets, Tail Index and Piercing the Veil — Chapter 5 shows that the effect of risk buckets depends on the tail: eight buckets raise the chance of ruin from 0.0102 to 0.0355 under light tails and cut it from 0.0489 to 0.0027 under heavy tails, while a chain of seven guarantees between buckets raises it to 0.0943. (https://5dlibrary.com/answers/en/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c05/) 第5章 风险仓切分:切分数、尾部指数与刺破公司面纱 — 第五章说明风险仓切分的效果取决于尾部:切成八个仓,轻尾时破产概率从 0.0102 升到 0.0355,重尾时从 0.0489 降到 0.0027;七条仓间担保串成的链使它升到 0.0943。 (https://5dlibrary.com/answers/zh/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c05/) - Ch.6 The Arithmetic of Two Isolation Layers: Timing, Solvency and the Bypass of a Single Guarantee — Chapter 6 does the arithmetic of two isolation layers: with clean, independent layers the chance of getting through is 0.0050, but the multiplication fails once a layer signs a guarantee or was set up while insolvent, so timing and solvency come first. (https://5dlibrary.com/answers/en/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c06/) 第6章 双层隔离算术:设立时点、偿债能力与一笔担保的旁路 — 第六章做双层隔离算术:两层干净独立时穿透概率为 0.0050,一层签了担保或在资不抵债时设立,这个乘法就失效,设立时点与偿债能力是一切分层的前提。 (https://5dlibrary.com/answers/zh/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c06/) - Ch.7 Reading the Sources: Lord Zhu of Tao's "Twice Dispersed" and Separate Personality in Salomon — Chapter 7 reads Fan Li's "three times a thousand in gold, twice dispersed" in the Records of the Grand Historian (three tenfold rounds need growth to 1.4384 times a year) against Salomon in the House of Lords. (https://5dlibrary.com/answers/en/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c07/) 第7章 原典考辨:陶朱公「再分散」与 Salomon 案的独立人格 — 第七章对读《史记·货殖列传》陶朱公「三致千金,再分散」(三轮十倍需每年增长到 1.4384 倍)与英国上议院 Salomon 案。 (https://5dlibrary.com/answers/zh/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c07/) - Ch.8 Practice and Refutation: A Pricing Calculator, an Owner's Checklist and a Pre-Registered Test — Chapter 8 gives a pricing calculator (default yearly chance of ruin 0.00711, correction factor 1.25), an owner's checklist and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c08/) 第8章 实践与证伪:定价试算器、经营者清单与预注册检验 — 第八章给出定价试算器(默认年度破产概率 0.00711,校正因子 1.25)、经营者清单与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/liability-percolation-subsidiary-graphs-tail-ruin-amplification-personal-guarantees/c08/) ## 5DL-B-178 · Causal-DAG Validation of Long-Form Narrative, Sawtooth Scheduling of Chapter Suspense Entropy and Forgetting-Aware Foreshadowing Recovery 中文书名:长篇叙事因果有向无环图校验、章节悬念信息熵锯齿调度与伏笔遗忘衰减回收 DOI: pending Cite: Fan, Y. (2026). Causal-DAG Validation of Long-Form Narrative, Sawtooth Scheduling of Chapter Suspense Entropy and Forgetting-Aware Foreshadowing Recovery (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its serials, readers and parameters are illustrative models; all calculations are generated by its programs; it is not legal, financial or investment advice, and it keeps one red line: no false suspense to win clicks. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: Anyone who publishes a novel that runs to hundreds of chapters answers three questions every week: does the world contradict itself, how many chapters should separate the hooks, and how long should a planted clue wait for its payoff? Writing courses have slogans for all three, none of which data could refute. This book writes them as directed graphs, as entropy with a Markov retention chain, and as forgetting curves with survival analysis, and asks for which parameters each slogan holds. Its serials, readers and parameters are illustrative models; all calculations are generated by its programs; it is not legal, financial or investment advice, and it keeps one red line: no false suspense to win clicks. Chapters 1 and 2 use Kahn's topological sort and a version check grouped by key to find time paradoxes and attributes written with two values, and show that cross-volume retcons hide in implied order: one edge in volume six of an illustrative serial contradicts a chain of 10 edges written between volume 2 and volume 6. Chapter 3 writes suspense as entropy, shows that a hook is contrary evidence that raises entropy, and shows that the suspense budget is conserved. Chapter 4 solves for the best hook spacing under sawtooth entropy, gap-dependent churn and habituation: under illustrative parameters a hook every three chapters keeps 0.234 of readers at chapter 100 against 0.098 for a hook in every chapter, and under other reader traits a blanket spacing keeps up to 0.140 less. Chapter 5 gives a closed form for the best payoff time; one well-placed light reminder raises satisfaction from 0.018 to 0.678, and a climax leaning on 6 clues is fully understood with probability only 0.531. Chapter 6 shows that reader heterogeneity makes late hooks look spectacular in naive comparisons (a ratio of 0.194 for a true effect of zero), and that a naive ratio puts retention at chapter 50 at 0.0918 where the product-limit estimate gives 0.1656. Chapter 7 examines Jin Shengtan's grass-snake ash line, clouds across the mountain and lead-in, read against Aristotle's Poetics. Chapter 8 gives a checklist, an eight-rule persuasive-copy checker and a pre-registered protocol: comparing hook schedules needs 3,044 readers per arm, and evaluating a single hook needs 583,531 starting readers per arm. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:一部连载几百章的长篇,作者与出版方每周都在回答三个问题:前后设定有没有打架,章末钩子该隔几章留一个,伏笔埋多久再回收。写作课的口诀对这三个问题都有答案,却都无法被数据推翻。本书把它们分别写成有向图、信息熵与马尔可夫留存链、遗忘曲线与生存分析,问参数取什么值时口诀成立。书中的连载、读者与参数均为示意模型;全部计算由书中程序生成;本书不是法律、财务或投资建议,并坚持一条红线:不得以虚假悬念骗取点击。 第一、二章用卡恩拓扑排序与按键分组的版本检查找出时间悖论与同一属性的两个值,并说明跨卷吃书藏在推得出的先后里:示意连载第六卷的一条边与写在第 2 卷到第 6 卷之间的 10 条边相冲突。第三章把悬念写成熵,说明钩子是让熵上升的反向证据,而悬念预算是守恒的。第四章在锯齿形熵、随缺口变化的流失与钩子习惯化之下求最优间距:示意参数下每三章一个钩子第 100 章留住 0.234,章章留钩只留住 0.098,换一组读者特征,一刀切的间距最多少留住 0.140。 第五章给出伏笔最优回收时点的解析式,一次位置合适的轻提醒把满足感从 0.018 提到 0.678,同时依赖 6 条伏笔的高潮被完整理解的概率只有 0.531。第六章说明读者异质性让晚出现的钩子在朴素比较中显得效果惊人(比值 0.194,真实效应为零),而朴素比例把第 50 章留存算成 0.0918,乘积极限估计给出 0.1656。第七章考辨金圣叹《读第五才子书法》的草蛇灰线、横云断山与弄引,并对读亚里士多德《诗学》。第八章给出检查清单、八条规则的攻心文案检查器与预注册方案:比较钩子排期每组需要 3,044 名读者,评价单个钩子每组需要 583,531 名起始读者。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Causal Graph: Kahn's Topological Sort and Version Conflicts in a Setting — Chapters 1 and 2 use Kahn's topological sort and a version check grouped by key to find time paradoxes and attributes written with two values, and show that cross-volume retcons hide in implied order: one edge in volume six of an illustrative serial contradicts a chain of 10 edges written between volume 2 and volume 6. (https://5dlibrary.com/answers/en/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c01/) 第1章 因果有向图:卡恩拓扑排序与设定版本冲突 — 第一、二章用卡恩拓扑排序与按键分组的版本检查找出时间悖论与同一属性的两个值,并说明跨卷吃书藏在推得出的先后里:示意连载第六卷的一条边与写在第 2 卷到第 6 卷之间的 10 条边相冲突。 (https://5dlibrary.com/answers/zh/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c01/) - Ch.2 Cross-Volume Retcons: Incremental Cycle Detection, Transitive Closure and the Author's Own Forgetting — Cross-volume retcons have two sources. One is implied order: in the illustrative serial a harmless-looking edge in volume six closes a cycle, contradicting a chain of 10 stated edges written between volume 2 and volume 6, and incremental cycle detection raises the alarm the moment that edge is written. The other is the author's own forgetting: under the illustrative parameters, writing from memory leaves 2.06 contradictions on average by volume 4, a setting bible brings that down to 0.51, and yet by volume 10 a share of 0.813 of the runs still contains at least one. The bible, the version check and cycle detection each guard one stretch; only together do they approach zero retcons. (https://5dlibrary.com/answers/en/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c02/) 第2章 跨卷吃书:增量环检测、传递闭包与作者自己的遗忘 — 跨卷吃书有两个来源。一个是推得出的先后:示意连载在第六卷由一条看似无害的边闭合了一个环,与它冲突的是一条由 10 条明说的边组成、写在第 2 卷到第 6 卷之间的链;增量环检测在这条边写下的那一刻就能报警。另一个是作者自己的遗忘:在示意参数下,凭记忆写到第 4 卷平均已有 2.06 处矛盾,查阅设定集可以把它降到 0.51 处,但写到第 10 卷,仍有 0.813 的模拟至少出现一处。设定集、版本检查与环检测各管一段,三者合起来,才接近「零吃书」。 (https://5dlibrary.com/answers/zh/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c02/) - Ch.3 The Entropy of Suspense: Bayesian Readers, Contrary Evidence and the Suspense Budget — Once suspense is written as entropy, a hook acquires a precise mathematical identity: it is evidence for an ending the reader did not favour, so it can raise entropy, even though evidence lowers entropy on average. The illustrative story's entropy follows a sawtooth for six chapters and then falls to 0.060 bits at the reveal. The total budget of suspense is conserved, 0.2500 in the illustrative question; saving everything for the finale yields a sum of square roots of 0.500, and giving each chapter equal suspense yields 1.500. The spoiler studies remind us that this is only one part of what drives reading. The next chapter links the sawtooth to retention and solves for the best spacing of hooks. (https://5dlibrary.com/answers/en/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c03/) 第3章 悬念的信息熵:贝叶斯读者、反向证据与悬念预算 — 把悬念写成熵,钩子就有了明确的数学身份:它是支持读者原本不看好的结局的证据,所以能让熵上升,而平均来看证据只会让熵下降。示意故事的熵沿着锯齿走了六章,最后由揭晓降到 0.060 比特。悬念的总预算是守恒的,示意问题中为 0.2500;捂到结局只换来 0.500 的平方根之和,每章分到同样多的悬念则得到 1.500。剧透研究提醒我们,这只是阅读动力的一部分。下一章把锯齿与留存连起来,求钩子的最优间距。 (https://5dlibrary.com/answers/zh/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c03/) - Ch.4 Sawtooth Scheduling: Hook Spacing and a Markov Model of Serial Retention — Chapter 4 solves for the best hook spacing under sawtooth entropy, gap-dependent churn and habituation: under illustrative parameters a hook every three chapters keeps 0.234 of readers at chapter 100 against 0.098 for a hook in every chapter, and under other reader traits a blanket spacing keeps up to 0.140 less. (https://5dlibrary.com/answers/en/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c04/) 第4章 锯齿调度:钩子间距与追更留存的马尔可夫模型 — 第四章在锯齿形熵、随缺口变化的流失与钩子习惯化之下求最优间距:示意参数下每三章一个钩子第 100 章留住 0.234,章章留钩只留住 0.098,换一组读者特征,一刀切的间距最多少留住 0.140。 (https://5dlibrary.com/answers/zh/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c04/) - Ch.5 The Forgetting of Foreshadowing: Best Payoff Time, Mid-Way Reminders and a Ceiling on Open Threads — Chapter 5 gives a closed form for the best payoff time; one well-placed light reminder raises satisfaction from 0.018 to 0.678, and a climax leaning on 6 clues is fully understood with probability only 0.531. (https://5dlibrary.com/answers/en/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c05/) 第5章 伏笔的遗忘衰减:最优回收时点、中途提醒与并行线索上限 — 第五章给出伏笔最优回收时点的解析式,一次位置合适的轻提醒把满足感从 0.018 提到 0.678,同时依赖 6 条伏笔的高潮被完整理解的概率只有 0.531。 (https://5dlibrary.com/answers/zh/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c05/) - Ch.6 The Survivors Who Keep Reading: Reader Heterogeneity, Survivor Bias and Censoring — Chapter 6 shows that reader heterogeneity makes late hooks look spectacular in naive comparisons (a ratio of 0.194 for a true effect of zero), and that a naive ratio puts retention at chapter 50 at 0.0918 where the product-limit estimate gives 0.1656. (https://5dlibrary.com/answers/en/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c06/) 第6章 追更的幸存者:读者异质性、幸存者偏差与删失 — 第六章说明读者异质性让晚出现的钩子在朴素比较中显得效果惊人(比值 0.194,真实效应为零),而朴素比例把第 50 章留存算成 0.0918,乘积极限估计给出 0.1656。 (https://5dlibrary.com/answers/zh/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c06/) - Ch.7 The Classics Examined: Jin Shengtan's "How to Read the Fifth Book of Genius" against Aristotle's Poetics — Jin Shengtan's grass-snake ash line, clouds across the mountain and lead-in correspond to this book's spread light reminders, release of overload and short-range foreshadowing. In the illustrative models six light mentions give a satisfaction of 0.820 against 0.574 for one loud reminder, one interlude chapter raises retention over a twelve-chapter siege from 0.677 to 0.737, and the best distance for a lead-in is about 4.48 chapters. Aristotle's standard of unity removes 5 nodes from the illustrative graph, and his demand for the unexpected that follows from causes can be scored as earned surprise. The disagreement between the two classics reminds us to distinguish structural unity from sustained reading, and to hand both to data. (https://5dlibrary.com/answers/en/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c07/) 第7章 原典考辨:金圣叹《读第五才子书法》对读亚里士多德《诗学》 — 金圣叹的草蛇灰线、横云断山与弄引,分别对应本书的分散轻提醒、过载泄压与短距离伏笔;示意模型中,六次轻提的满足感 0.820 高于一次大声提醒的 0.574,插入一章间隔让十二章围城的留存从 0.677 升到 0.737,引子的最优距离约为 4.48 章。亚里士多德的整一标准在示意图上删去了 5 个节点,他对「出人意料而因果相承」的要求,可以量成挣来的惊奇。两部原典的分歧,提醒我们区分结构的整一与读者的持续阅读,并把两者都交给数据。 (https://5dlibrary.com/answers/zh/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c07/) - Ch.8 Practice and Falsification: A Publisher's Checklist, a Persuasive-Copy Checker and a Pre-Registered Test — Chapter 8 gives a checklist, an eight-rule persuasive-copy checker and a pre-registered protocol: comparing hook schedules needs 3,044 readers per arm, and evaluating a single hook needs 583,531 starting readers per arm. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c08/) 第8章 实践与证伪方案:连载主事者的检查清单、攻心文案检查器与预注册检验 — 第八章给出检查清单、八条规则的攻心文案检查器与预注册方案:比较钩子排期每组需要 3,044 名读者,评价单个钩子每组需要 583,531 名起始读者。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/causal-dag-validation-long-form-narrative-sawtooth-scheduling-chapter/c08/) ## 5DL-B-179 · Cost-Benefit Regression for Asymmetric Strategy Cards, Hypergeometric Brick-Rate Bounds in Deck Building and Cyclic Counters to Dominant Strategies 中文书名:非对称策略卡牌费用收益回归、牌库构筑超几何卡手率边界与优势策略循环克制 DOI: pending Cite: Fan, Y. (2026). Cost-Benefit Regression for Asymmetric Strategy Cards, Hypergeometric Brick-Rate Bounds in Deck Building and Cyclic Counters to Dominant Strategies (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its card pool, win rates and prices are illustrative; all calculations are generated by its programs; the book is not legal, financial or investment advice and recommends no purchase. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: Before a physical strategy card game goes to print, its designers and publisher must answer three kinds of question: is each card's cost right, how many resources should a deck hold so that it seldom stalls, and will a set of decks that counter one another be displaced by a deck that loses to nobody? This book turns these questions into runnable arithmetic using Hengshu, an invented, illustrative card game. Its card pool, win rates and prices are illustrative; all calculations are generated by its programs; the book is not legal, financial or investment advice and recommends no purchase. Chapter 1 uses least squares to find under-costed and over-costed cards: each point of cost buys 1.858 stat points on average, and a multiple regression shows that health is dearer than attack and that keywords are worth far less than assumed. Chapter 2 computes brick rates exactly with the hypergeometric distribution: a first player with seventeen resources bricks on turn four with probability 0.293, and if neither bricks nor floods may exceed a quarter the feasible band is 18 to 21 resources. Chapter 3 computes expected available mana with a Markov chain and simulates how a low-curve deck spends less because its hand runs dry. Chapter 4 solves the mixed equilibrium of counter cycles: strengthening aggro raises the share of control, its counter, by 0.121. Chapter 5 implements counterfactual regret minimisation, approaches the theoretical value of −0.0556 in a three-card bluffing duel, and within a thousand iterations hands 0.808 of the share to a broken deck built on an under-costed card. Chapter 6 computes the games needed to detect such a deck and the prediction interval of a patch; a one-point cost increase still has a 0.407 chance of falling short. Chapter 7 reads Ban Gu's Essay on the Purport of Go against von Neumann's paper of 1928: brick probability raises the games needed to rank players from 41 to 119. Chapter 8 gives a publisher's checklist, a pricing calculator (a named super rare appears in a box with probability only 0.182) and a pre-registered test protocol. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:一款实体策略卡牌游戏在印刷之前,设计者与发行商要回答三类问题:单张卡的费用定得对不对,一副牌放多少资源才不容易卡手,以及几种互相克制的卡组会不会被一副对谁都不吃亏的卡组取代。本书用一款自拟的示意卡牌游戏「衡枢」把这些问题写成可以运行的算术。书中卡池、胜率与价格参数均为示意;全部计算由书中程序生成;本书不是法律、财务或投资建议,也不推荐任何购买。 第一章用最小二乘找出被低估与被高估的卡:每一费平均买 1.858 点身材,多元回归还揭示生命比攻击贵、关键词远不如假设的值钱。第二章用超几何分布精确计算卡手率:资源 17 张的先手第 4 回合卡手概率是 0.293,若要求卡手与溢出都不超过四分之一,可行区间是 18 到 21 张。第三章用马尔可夫链计算期望可用费用,并用模拟说明低曲线卡组因手牌耗尽而花得更少。第四章求解循环克制的混合均衡:加强快攻,涨份额的是克制它的控制,增幅 0.121。 第五章实现反事实遗憾最小化,在三牌暗押上逼近理论值 −0.0556,并在一千轮内把 0.808 的份额推给建立在被低估卡上的坏牌组。第六章计算检出坏牌组所需的对局数与补丁效果的预测区间,改一费仍有 0.407 的概率不够。第七章对读班固《弈旨》与冯·诺依曼 1928 年的论文:卡手概率让排定等级所需的局数从 41 局升到 119 局。第八章给出发行商清单、定价试算器(指定超稀有一盒内出现的概率只有 0.182)与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Cost-Benefit Regression: Least Squares Finds the Under-Costed and the Over-Costed — Chapter 1 uses least squares to find under-costed and over-costed cards: each point of cost buys 1.858 stat points on average, and a multiple regression shows that health is dearer than attack and that keywords are worth far less than assumed. (https://5dlibrary.com/answers/en/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c01/) 第1章 费用–收益回归:用最小二乘找出被低估与被高估的卡 — 第一章用最小二乘找出被低估与被高估的卡:每一费平均买 1.858 点身材,多元回归还揭示生命比攻击贵、关键词远不如假设的值钱。 (https://5dlibrary.com/answers/zh/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c01/) - Ch.2 Hypergeometric Brick Rates: the Seven-Card Hand, Turn-by-Turn Bricks and a Feasible Band of Resources — Chapter 2 computes brick rates exactly with the hypergeometric distribution: a first player with seventeen resources bricks on turn four with probability 0.293, and if neither bricks nor floods may exceed a quarter the feasible band is 18 to 21 resources. (https://5dlibrary.com/answers/en/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c02/) 第2章 超几何卡手率:起手七张、逐回合卡手与资源牌的可行区间 — 第二章用超几何分布精确计算卡手率:资源 17 张的先手第 4 回合卡手概率是 0.293,若要求卡手与溢出都不超过四分之一,可行区间是 18 到 21 张。 (https://5dlibrary.com/answers/zh/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c02/) - Ch.3 The Mana Curve: a Markov Chain for Expected Available Mana and a Simulation of Mana Spent — Expected available mana is not the smaller of turn number and resources seen; it is a Markov chain in which a missed drop is never recovered. A first player with seventeen resources expects only 23.89 mana-turns over the first seven turns and misses at least one drop with probability 0.748. Available is not spent: the low-curve deck spends less because its hand runs dry and does best with sixteen resources, while the high-curve deck does best with nineteen. The next chapter leaves the single deck and turns to the counter relations between decks. (https://5dlibrary.com/answers/en/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c03/) 第3章 法力曲线:期望可用费用的马尔可夫链与实际花费的模拟 — 期望可用费用不是「回合数与资源数取小」,而是一条错过就无法补回的马尔可夫链。资源 17 张的先手,前 7 回合累计只有 23.89 点可用费用,且有 0.748 的概率至少错过一次放置。可用不等于花掉:低曲线卡组因手牌耗尽而花得更少,最佳资源数在 16 张;高曲线卡组的最佳资源数在 19 张。下一章离开单副卡组,进入卡组之间的克制关系。 (https://5dlibrary.com/answers/zh/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c03/) - Ch.4 Cyclic Counters: the Mixed Equilibrium of a Symmetric Zero-Sum Matrix Game and a Counter-Intuitive Patch Effect — Chapter 4 solves the mixed equilibrium of counter cycles: strengthening aggro raises the share of control, its counter, by 0.121. (https://5dlibrary.com/answers/en/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c04/) 第4章 循环克制:对称零和矩阵博弈的混合均衡与反直觉的补丁效应 — 第四章求解循环克制的混合均衡:加强快攻,涨份额的是克制它的控制,增幅 0.121。 (https://5dlibrary.com/answers/zh/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c04/) - Ch.5 Counterfactual Regret Minimisation: from a Three-Card Bluffing Duel to a Must-Pick Broken Deck — Chapter 5 implements counterfactual regret minimisation, approaches the theoretical value of −0.0556 in a three-card bluffing duel, and within a thousand iterations hands 0.808 of the share to a broken deck built on an under-costed card. (https://5dlibrary.com/answers/en/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c05/) 第5章 反事实遗憾最小化:从三牌暗押到卡组选择中的必胜坏牌组 — 第五章实现反事实遗憾最小化,在三牌暗押上逼近理论值 −0.0556,并在一千轮内把 0.808 的份额推给建立在被低估卡上的坏牌组。 (https://5dlibrary.com/answers/zh/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c05/) - Ch.6 Testing for a Must-Pick Broken Deck and Regression Checks on Balance Patches — Chapter 6 computes the games needed to detect such a deck and the prediction interval of a patch; a one-point cost increase still has a 0.407 chance of falling short. (https://5dlibrary.com/answers/en/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c06/) 第6章 必胜坏牌组的统计检验与平衡补丁的回归检查 — 第六章计算检出坏牌组所需的对局数与补丁效果的预测区间,改一费仍有 0.407 的概率不够。 (https://5dlibrary.com/answers/zh/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c06/) - Ch.7 The Classics Examined: Ban Gu's Essay on the Purport of Go and von Neumann's Theory of Parlour Games — Chapter 7 reads Ban Gu's Essay on the Purport of Go against von Neumann's paper of 1928: brick probability raises the games needed to rank players from 41 to 119. (https://5dlibrary.com/answers/en/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c07/) 第7章 原典考辨:班固《弈旨》与冯·诺依曼《论室内游戏理论》 — 第七章对读班固《弈旨》与冯·诺依曼 1928 年的论文:卡手概率让排定等级所需的局数从 41 局升到 119 局。 (https://5dlibrary.com/answers/zh/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c07/) - Ch.8 Practice and Falsification: a Publisher's Checklist, a Pricing Calculator and Pre-Registered Tests — Chapter 8 gives a publisher's checklist, a pricing calculator (a named super rare appears in a box with probability only 0.182) and a pre-registered test protocol. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c08/) 第8章 实践与证伪方案:发行商清单、定价试算器与预注册检验 — 第八章给出发行商清单、定价试算器(指定超稀有一盒内出现的概率只有 0.182)与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cost-benefit-regression-asymmetric-strategy-cards-hypergeometric-brick-rate/c08/) ## 5DL-B-180 · Envy-Free Auctions of Indivisible Assets among Multiple Heirs, the Altruistic-Parent Sloth Trap and Voting-Right Decoupling in Family Charters 中文书名:多继承人不可分资产无嫉妒拍卖、慈父利他主义懒惰陷阱与家族宪章投票权解耦 DOI: pending Cite: Fan, Y. (2026). Envy-Free Auctions of Indivisible Assets among Multiple Heirs, the Altruistic-Parent Sloth Trap and Voting-Right Decoupling in Family Charters (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 2 computes Knaster's sealed-bid procedure exactly (the illustrative surplus is 393.33), shows that it can leave envy of 30.00, and maps all envy-free cash schemes as a region with 5 vertices. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: This book addresses three entangled problems that family business owners face in succession: how several heirs can fairly divide indivisible assets such as an old house, a factory and a collection, with cash to balance the shares; why altruistic parents weaken their children's effort by catching them after the fact; and how much power each member really holds once a family charter separates votes from dividends. Its valuations, wealth levels, altruism weights and voting weights are illustrative parameters; all calculations are generated by its programs; the book is not legal, tax or financial advice for any individual or family, inheritance, trust and company law differ between jurisdictions, and specific arrangements require local licensed professionals. Chapter 1 enumerates the 27 ways three heirs can split three indivisible assets and finds none proportional; under random valuations with three assets only 0.2150 of profiles admit an envy-free division. Chapter 2 computes Knaster's sealed-bid procedure exactly (the illustrative surplus is 393.33), shows that it can leave envy of 30.00, and maps all envy-free cash schemes as a region with 5 vertices. Chapter 3 shows that under complete information a unilateral misreport can take 206.22 more, and that under incomplete information shrinking bids toward the middle misallocates at least one asset with probability 0.2657. Chapter 4 writes Buchanan's Samaritan's dilemma as a two-stage game: an ex-post safety net lowers the child's effort from 0.8828 to 0.3252, prior commitment restores it to 0.5975 and leaves the parent better off too, and the rotten kid theorem fails when the child can push herself into a corner where no transfer is needed. Chapter 5 compares fixed distributions, milestone trusts and matching trusts, showing that the value of commitment comes mainly from payments that do not float with outcomes; as the probability that the enforcer makes exceptions rises to 1, effort falls back to 0.325. Chapter 6 uses exact enumeration of Shapley values and Banzhaf indices to show that equity is not power: a member holding forty per cent has power 0.000 under a voting trust, and when the controller's cash-flow share is 0.10 the other members bear 1.053 units of loss for each 1 unit of the controller's private benefit. Chapter 7 reads Yan Zhitui's Family Instructions against Carnegie's "Wealth". Chapter 8 gives a checklist for the head of family, a communication skill pack for family clients and a pre-registered test protocol (388 families per group). Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:本书讨论家族企业主在传承时遇到的三个相互缠绕的问题:几位继承人怎样公平地分配老宅、工厂、藏品这类不可分资产并以现金找补;利他的父母为什么会因为事后兜底而削弱子女的努力;家族宪章把表决权与分红权分开之后,每个成员实际拥有多大的权力。书中的估价、财富、利他系数与表决权重均为示意参数;全部计算由书中程序生成;本书不是针对任何个人或家庭的法律、税务或理财建议,各法域的继承法、信托法与公司法不同,具体安排须咨询当地持牌专业人士。 第一章穷举三位继承人分三件不可分资产的 27 种分法,没有一种比例公平;随机估值下三件资产时只有 0.2150 的估值组合存在无嫉妒分法。第二章精确计算 Knaster 密封出价程序(示意例的盈余为 393.33),说明它可能留下 30.00 的嫉妒,而全部无嫉妒现金方案构成一个有 5 个顶点的区域。第三章说明完全信息下单方虚报可多拿 206.22,信息不完全时向中收缩使至少一件资产错配的概率达到 0.2657。第四章把布坎南的撒玛利亚人困境写成两阶段博弈:事后兜底使子女努力从 0.8828 降到 0.3252,事先承诺使其回到 0.5975,且父母自身的福利也更高;坏孩子定理在子女能把自己推到不需要转移的角点时失效。 第五章比较固定分配、里程碑信托与配比信托,说明承诺的价值主要来自给付不随结果浮动,执行者破例的概率升到 1 时努力回落到 0.325。第六章用夏普利值与 Banzhaf 指数的精确枚举说明股权不等于权力:持股四成的成员在表决权信托下权力为 0.000;控制者现金流权为 0.10 时,其他成员每承担 1.053 单位损失才换来控制者 1 单位私人收益。第七章对读颜之推《颜氏家训》与卡内基《财富》。第八章给出主事者清单、家族客户沟通技能包与一份预注册检验方案(每组 388 个家族)。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Three Yardsticks: How Proportionality, Envy-Freeness and Pareto Efficiency Nest over Indivisible Assets — Chapter 1 enumerates the 27 ways three heirs can split three indivisible assets and finds none proportional; under random valuations with three assets only 0.2150 of profiles admit an envy-free division. (https://5dlibrary.com/answers/en/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c01/) 第1章 三把尺子:比例公平、无嫉妒与帕累托效率在不可分资产上的包含关系 — 第一章穷举三位继承人分三件不可分资产的 27 种分法,没有一种比例公平;随机估值下三件资产时只有 0.2150 的估值组合存在无嫉妒分法。 (https://5dlibrary.com/answers/zh/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c01/) - Ch.2 Sealed Bids and Cash Balancing: Knaster's Procedure, the Envy-Free Cash Polytope and Internal Auctions — Chapter 2 computes Knaster's sealed-bid procedure exactly (the illustrative surplus is 393.33), shows that it can leave envy of 30.00, and maps all envy-free cash schemes as a region with 5 vertices. (https://5dlibrary.com/answers/en/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c02/) 第2章 密封出价与现金找补:Knaster 程序、无嫉妒现金多面体与内部拍卖 — 第二章精确计算 Knaster 密封出价程序(示意例的盈余为 393.33),说明它可能留下 30.00 的嫉妒,而全部无嫉妒现金方案构成一个有 5 个顶点的区域。 (https://5dlibrary.com/answers/zh/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c02/) - Ch.3 The Incentive to Misreport: Best Deviations under Complete Information and Shrinking toward the Middle under Uncertainty — Chapter 3 shows that under complete information a unilateral misreport can take 206.22 more, and that under incomplete information shrinking bids toward the middle misallocates at least one asset with probability 0.2657. (https://5dlibrary.com/answers/en/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c03/) 第3章 虚报的激励:完全信息下的最优偏离与不确定之下的向中收缩 — 第三章说明完全信息下单方虚报可多拿 206.22,信息不完全时向中收缩使至少一件资产错配的概率达到 0.2657。 (https://5dlibrary.com/answers/zh/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c03/) - Ch.4 The Samaritan's Dilemma: the Subgame-Perfect Equilibrium of an Ex-Post Safety Net and Where the Rotten Kid Theorem Fails — Chapter 4 writes Buchanan's Samaritan's dilemma as a two-stage game: an ex-post safety net lowers the child's effort from 0.8828 to 0.3252, prior commitment restores it to 0.5975 and leaves the parent better off too, and the rotten kid theorem fails when the child can push herself into a corner where no transfer is needed. (https://5dlibrary.com/answers/en/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c04/) 第4章 撒玛利亚人困境:事后兜底的子博弈精炼均衡与坏孩子定理的失效边界 — 第四章把布坎南的撒玛利亚人困境写成两阶段博弈:事后兜底使子女努力从 0.8828 降到 0.3252,事先承诺使其回到 0.5975,且父母自身的福利也更高;坏孩子定理在子女能把自己推到不需要转移的角点时失效。 (https://5dlibrary.com/answers/zh/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c04/) - Ch.5 Commitment, Not Conditions: Fixed Distributions, Milestone Trusts, Matching Trusts and the Softening of Trustees — Chapter 5 compares fixed distributions, milestone trusts and matching trusts, showing that the value of commitment comes mainly from payments that do not float with outcomes; as the probability that the enforcer makes exceptions rises to 1, effort falls back to 0.325. (https://5dlibrary.com/answers/en/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c05/) 第5章 承诺而非条件:固定分配、里程碑信托、配比信托与执行者的软化 — 第五章比较固定分配、里程碑信托与配比信托,说明承诺的价值主要来自给付不随结果浮动,执行者破例的概率升到 1 时努力回落到 0.325。 (https://5dlibrary.com/answers/zh/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c05/) - Ch.6 Equity Is Not Power: Shapley Values, Banzhaf Indices and the Externality of Decoupled Voting Rights — Chapter 6 uses exact enumeration of Shapley values and Banzhaf indices to show that equity is not power: a member holding forty per cent has power 0.000 under a voting trust, and when the controller's cash-flow share is 0.10 the other members bear 1.053 units of loss for each 1 unit of the controller's private benefit. (https://5dlibrary.com/answers/en/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c06/) 第6章 股权不等于权力:夏普利值、Banzhaf 指数与表决权解耦的外部性 — 第六章用夏普利值与 Banzhaf 指数的精确枚举说明股权不等于权力:持股四成的成员在表决权信托下权力为 0.000;控制者现金流权为 0.10 时,其他成员每承担 1.053 单位损失才换来控制者 1 单位私人收益。 (https://5dlibrary.com/answers/zh/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c06/) - Ch.7 Classical Sources Examined: Yan Zhitui's Family Instructions and Carnegie's "Wealth" — The second part is more direct. With no bequest the child's effort is 0.8828 and her own earnings make up 0.7793 of her total resources. With a bequest of 100 her effort falls to 0.3252, exactly the effort under the ex-post safety net of Chapter 4, since in both cases parental wealth plus base income comes to 110, and her own earnings are only 0.1057 of the total. With a bequest of 400 her effort is just 0.0966. The value of a modest skill lies not only in the fact that skills cannot be divided among heirs but also in the effort it takes to acquire one, and a large bequest weakens precisely that effort. (https://5dlibrary.com/answers/en/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c07/) 第7章 原典考辨:《颜氏家训》与卡内基《财富》 — 第二部分更直接:遗赠为零时,子女努力 0.8828,自己挣的收入占总资源的 0.7793;遗赠 100 时,努力降到 0.3252,与第四章事后兜底的努力完全相同(因为两者都让「父母财富加基础收入」等于 110),自己挣的比例只剩 0.1057;遗赠 400 时,努力只有 0.0966。「薄伎在身」的价值,不仅在于技能不会被分割,还在于获得技能需要努力,而大笔遗赠恰恰削弱了这份努力。 (https://5dlibrary.com/answers/zh/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c07/) - Ch.8 Practice and Falsification: a Communication Skill Pack for Family Offices and a Pre-Registered Test — Chapter 8 gives a checklist for the head of family, a communication skill pack for family clients and a pre-registered test protocol (388 families per group). Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c08/) 第8章 实践与证伪:家族办公室沟通技能包与预注册检验 — 第八章给出主事者清单、家族客户沟通技能包与一份预注册检验方案(每组 388 个家族)。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/envy-free-auctions-indivisible-assets-multiple-heirs-altruistic-parent/c08/) ## 5DL-B-181 · Costly-Signal Screening in Early Intimacy, Intermittent-Reinforcement Dopamine Traps and Reciprocal-Floor Stopping Games 中文书名:亲密关系早期高代价信号甄别、间歇性强化多巴胺成瘾陷阱与对等底线停时博弈 DOI: pending Cite: Fan, Y. (2026). Costly-Signal Screening in Early Intimacy, Intermittent-Reinforcement Dopamine Traps and Reciprocal-Floor Stopping Games (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not psychological counselling, claims no effect of any kind, and teaches no technique of manipulation. Chapter 2 enters each interaction in a likelihood-ratio ledger; with illustrative parameters one missed plan weighs as much as 2.52 kept ones, and Wald's sequential test saves 0.459 of the interactions a fixed observation period needs. Abstract: This book addresses three questions about the early stage of intimate relationships: how, seeing only behaviour, to tell someone who intends to invest for the long run from someone who wants short-term gain; why hot-and-cold treatment is harder to let go of than steady coolness; and what a stop-loss rule, written down in advance and followed when the time comes, should look like. Its parameters are illustrative; all calculations are generated by its programs; it is not psychological counselling, claims no effect of any kind, and teaches no technique of manipulation. Anyone facing threats, stalking, control or harm should contact local emergency services, a helpline or a qualified professional. Chapter 1 uses the signalling theory of Spence, Zahavi and Grafen to show that a signal separates types only if it is expensive to different degrees for people with different intentions: a money bar sorts wealth, a time bar sorts intention. Chapter 2 enters each interaction in a likelihood-ratio ledger; with illustrative parameters one missed plan weighs as much as 2.52 kept ones, and Wald's sequential test saves 0.459 of the interactions a fixed observation period needs. Chapter 3 uses the partial reinforcement extinction effect and a discrimination model to show that a variable schedule answering one attempt in four needs 21 times as many misses before quitting as steady responding, and that the mean size of the prediction error peaks at 0.5000 when responses come half the time. The small temporal-difference model of Chapter 4 shows that ranking partners by intensity of feeling almost reverses ranking by mean warmth, and that a partner who floods and then withdraws can produce a single rush of 0.887. Chapter 5 shows that the limiting success rate of the secretary problem is only 0.3679 and proposes the reciprocal floor: leave when the investment ratio stays below 0.5 for 3 weeks in a row; in the illustrative model it wounds a genuine investor within a year with probability 0.0258 and identifies an under-investor in 6.99 weeks on average. Chapter 6 writes the love-bombing and withdrawal cycle as computable features of shape; a combined rule catches a soft cycle at 0.7760 and raises false alarms of 0.0020 on genuine investors with irregular lives. Chapter 7 reads the six rites of the Book of Rites chapter on marriage against Zahavi's handicap principle; staging cuts the sincere suitor's sunk-cost risk from 0.848 to 0.342. Chapter 8 gives a personal interaction log, a stop-loss communication kit for client relationships and a pre-registered test protocol (163 per group). Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:本书讨论亲密关系早期的三个问题:只能看到对方行为时,怎样把打算长期投入的人与只想短期获利的人分开;为什么时冷时热的对待比稳定的冷淡更难放下;以及一条事先写下、到点就走的止损规则应当长什么样。书中参数均为示意值;全部计算由书中程序生成;本书不是心理咨询,不宣称任何效果,也不教授任何操控手段。遇到威胁、跟踪、控制或伤害,请联系当地紧急求助电话、援助热线或专业人士。 第一章用斯彭斯、扎哈维与格拉芬的信号理论说明,信号只有在对不同意图的人贵得不一样时才能分开类型,金钱门槛筛的是财力,时间门槛筛的才是意图。第二章把每次互动记入似然比账本,示意参数下一次失约的分量相当于 2.52 次兑现,沃尔德序贯检验平均比固定观察期节省 0.459 的互动。第三章用部分强化消退效应与辨别模型说明,四次一回应的变比率时间表需要的落空次数是稳定回应的 21 倍,预测误差的平均大小在回应概率为一半时达到 0.5000。第四章的时序差分小模型显示,按感受强烈程度排序几乎与按平均温暖排序相反,先轰炸后冷落的对象单次冲劲可达 0.887。 第五章说明秘书问题的极限成功率只有 0.3679,并提出对等底线:比值连续 3 周低于 0.5 即退出,示意模型中一年内误伤真投入者的概率为 0.0258,识别投入不足者平均需要 6.99 周。第六章把爱情轰炸与冷落循环写成可计算的形状特征,组合规则对弱循环命中 0.7760,对生活不规律的真投入者误报 0.0020。第七章对读《礼记·昏义》的六礼与扎哈维的累赘原理,分阶段安排使真心者的沉没风险从 0.848 降到 0.342。第八章给出个人互动日志、客户关系止损沟通技能包与一份预注册检验方案(每组 163 人)。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 When Costly Signals Separate: Single Crossing, the Handicap Principle and the Cost Structure of Honest Signals — The left half of the table is the honest equilibrium. A person of quality 0.4 sends 0.0600, one of quality 0.8 sends 0.3000, one of quality 1.0 sends 0.4800, and the receiver's reading equals true quality every time. The right half replaces the cost with a flat s that does not depend on quality: every type now chooses the same strength, 0.4800. Once everyone sends the same signal, nothing can be read from it. The column of 1.000 is only the illusion of a receiver who keeps the old reading, and in equilibrium that reading must be abandoned. (https://5dlibrary.com/answers/en/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c01/) 第1章 昂贵信号的分离条件:单交叉、累赘原理与诚实信号的成本结构 — 左半张表是诚实均衡:质量 0.4 的人发 0.0600,质量 0.8 的人发 0.3000,质量 1.0 的人发 0.4800,接收者的读数与真实质量完全一致。右半张表把代价换成与质量无关的 s:每一种质量的人都选择同一个强度 0.4800。一旦所有人发同样的信号,接收者就不可能从中读出任何东西;表里那一列 1.000 只是接收者沿用旧读法时的错觉,在均衡里这种读法必然被放弃。 (https://5dlibrary.com/answers/zh/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c01/) - Ch.2 Sequential Bayesian Screening: a Likelihood-Ratio Ledger and Wald's Boundaries for Going On or Stepping Back — Chapter 2 enters each interaction in a likelihood-ratio ledger; with illustrative parameters one missed plan weighs as much as 2.52 kept ones, and Wald's sequential test saves 0.459 of the interactions a fixed observation period needs. (https://5dlibrary.com/answers/en/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c02/) 第2章 序贯贝叶斯甄别:似然比账本与沃尔德检验的继续—止损边界 — 第二章把每次互动记入似然比账本,示意参数下一次失约的分量相当于 2.52 次兑现,沃尔德序贯检验平均比固定观察期节省 0.459 的互动。 (https://5dlibrary.com/answers/zh/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c02/) - Ch.3 Variable-Ratio Reinforcement and Resistance to Extinction: Why Responses That Come and Go Are the Hardest to Give Up — Chapter 3 uses the partial reinforcement extinction effect and a discrimination model to show that a variable schedule answering one attempt in four needs 21 times as many misses before quitting as steady responding, and that the mean size of the prediction error peaks at 0.5000 when responses come half the time. The small temporal-difference model of Chapter 4 shows that ranking partners by intensity of feeling almost reverses ranking by mean warmth, and that a partner who floods and then withdraws can produce a single rush of 0.887. (https://5dlibrary.com/answers/en/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c03/) 第3章 变比率强化与消退抗性:为什么时有时无的回应最难放下 — 第三章用部分强化消退效应与辨别模型说明,四次一回应的变比率时间表需要的落空次数是稳定回应的 21 倍,预测误差的平均大小在回应概率为一半时达到 0.5000。 (https://5dlibrary.com/answers/zh/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c03/) - Ch.4 Reward Prediction Errors: a Small Temporal-Difference Model of the Hot-and-Cold Relationship — By mean warmth the order is S1, S2, S3, S4: 0.800, 0.400, 0.372 and 0.244. By rush the order is almost exactly reversed. S1 and S2 both score 0.000, S3 scores 0.245 and S4 0.193, and the largest single rushes belong to S3 (0.917) and S4 (0.887). A learner who judged by how intense things feel would pick the two partners who give least; one who judged by mean warmth would pick S1. (https://5dlibrary.com/answers/en/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c04/) 第4章 奖赏预测误差:时冷时热关系的时序差分学习小模型 — 平均温暖的排序是 S1、S2、S3、S4:0.800、0.400、0.372、0.244。冲劲的排序几乎完全相反:S1 与 S2 都是 0.000,S3 是 0.245,S4 是 0.193;最大的单次冲劲出现在 S3(0.917)与 S4(0.887)。如果学习者用「感觉有多强烈」来判断谁值得留下,他会选中给得最少的两位;如果用平均温暖来判断,他会选中 S1。 (https://5dlibrary.com/answers/zh/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c04/) - Ch.5 Optimal Stopping: the 1/e of the Secretary Problem, Thresholds on a Known Scale and the Reciprocal-Floor Rule — Chapter 5 shows that the limiting success rate of the secretary problem is only 0.3679 and proposes the reciprocal floor: leave when the investment ratio stays below 0.5 for 3 weeks in a row; in the illustrative model it wounds a genuine investor within a year with probability 0.0258 and identifies an under-investor in 6.99 weeks on average. (https://5dlibrary.com/answers/en/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c05/) 第5章 最优停时:秘书问题的 1/e、已知尺度下的阈值与对等底线规则 — 第五章说明秘书问题的极限成功率只有 0.3679,并提出对等底线:比值连续 3 周低于 0.5 即退出,示意模型中一年内误伤真投入者的概率为 0.0258,识别投入不足者平均需要 6.99 周。 (https://5dlibrary.com/answers/zh/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c05/) - Ch.6 Detectable Fingerprints of Manipulation: Features and False Alarm Rates for the Love-Bombing and Withdrawal Cycle — Chapter 6 writes the love-bombing and withdrawal cycle as computable features of shape; a combined rule catches a soft cycle at 0.7760 and raises false alarms of 0.0020 on genuine investors with irregular lives. (https://5dlibrary.com/answers/en/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c06/) 第6章 操控模式的可检测指纹:爱情轰炸—冷落循环的特征与误报率 — 第六章把爱情轰炸与冷落循环写成可计算的形状特征,组合规则对弱循环命中 0.7760,对生活不规律的真投入者误报 0.0020。 (https://5dlibrary.com/answers/zh/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c06/) - Ch.7 Reading the Sources: the Six Rites of the Book of Rites and Zahavi's Handicap Principle — Chapter 7 reads the six rites of the Book of Rites chapter on marriage against Zahavi's handicap principle; staging cuts the sincere suitor's sunk-cost risk from 0.848 to 0.342. (https://5dlibrary.com/answers/en/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c07/) 第7章 原典考辨:《礼记·昏义》六礼与扎哈维累赘原理 — 第七章对读《礼记·昏义》的六礼与扎哈维的累赘原理,分阶段安排使真心者的沉没风险从 0.848 降到 0.342。 (https://5dlibrary.com/answers/zh/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c07/) - Ch.8 Practice and Refutation: a Personal Interaction Log, a Stop-Loss Communication Kit and a Pre-Registered Test — Chapter 8 gives a personal interaction log, a stop-loss communication kit for client relationships and a pre-registered test protocol (163 per group). Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c08/) 第8章 实践与证伪方案:个人互动日志、止损沟通技能包与预注册检验 — 第八章给出个人互动日志、客户关系止损沟通技能包与一份预注册检验方案(每组 163 人)。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/costly-signal-screening-early-intimacy-intermittent-reinforcement-dopamine-traps/c08/) ## 5DL-B-182 · Cross-Volume Theorem-Conflict Detection in Large Bilingual Monograph Libraries, Anti-Decay of Code Reproducibility and Erasure-Coded Cold Storage 中文书名:超大规模双语专著库跨卷定理冲突检测、代码可复现性衰减防腐与冷存储纠删码 DOI: pending Cite: Fan, Y. (2026). Cross-Volume Theorem-Conflict Detection in Large Bilingual Monograph Libraries, Anti-Decay of Code Reproducibility and Erasure-Coded Cold Storage (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Library sizes, error rates, hazards and failure rates are illustrative; all calculations are generated by the programs in the book; the book is not legal, financial, procurement or storage-service advice. Chapter 1 draws the library as a knowledge graph and shows that the more volumes there are, the fewer places an error can hide: at 200 volumes, 0.9821 of deviant definitions in the illustrative library sit where another volume can contradict them. Abstract: A library of two hundred bilingual monographs, each with runnable programs, accumulates three kinds of trouble over time: one term is defined or valued differently in different volumes, the programs behind the printed numbers stop reproducing them as dependencies drift, and the media that keep the manuscripts fail or rot silently. This book turns the sayings "every volume was checked, so the library is consistent", "the code is public, so it can be tested for ever" and "keep a few more copies and rest easy" into computable arithmetic and asks under which parameters they hold. Library sizes, error rates, hazards and failure rates are illustrative; all calculations are generated by the programs in the book; the book is not legal, financial, procurement or storage-service advice. Chapter 1 draws the library as a knowledge graph and shows that the more volumes there are, the fewer places an error can hide: at 200 volumes, 0.9821 of deviant definitions in the illustrative library sit where another volume can contradict them. Chapter 2 compares four comparators: one that only normalizes terms has a precision of 0.073, while the interval-overlap rule has no false alarms. Chapter 3 gives the rule of three for clean audits: showing a conflict rate below 0.01 with 95 % confidence takes 299 clean checks, almost regardless of library size. Chapter 4 uses a hazard model to show that standard-library programs have a half-life of 34.7 years against 2.2 years with five dependencies, and that a yearly byte-exact output comparison cuts the unknown failure-years of a decade to 9.21 per hundred programs. Chapter 5 implements a systematic Reed-Solomon erasure code over GF(2^8) in standard-library Python; an overhead of 1.50 survives three lost blocks, and all 1,001 survivor sets of RS(10+4) decode. Chapter 6 computes yearly data-loss probabilities with a binomial window model and Markov chains, showing that a twofold correlated acceleration of failures shrinks the MTTDL by 15.9 times and that, with a scrub every five years and no per-block checksums, a read returns wrong bytes with probability 0.11820. Chapter 7 examines the collation records of the Siku quanshu and the fate of its seven pavilion copies alongside Lachmann's stemmatics: with four kindred copies and one independent copy, voting by manuscripts leaves 20.300 wrong characters per thousand and comparing families only 0.100. Chapter 8 offers checklists for publishers, a cold-storage cost calculator, an errata register format and a pre-registered test protocol. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:一个拥有两百卷双语专著、每卷附带可运行程序的书库,会随时间累积三种麻烦:同一术语在不同卷中的定义或数值彼此矛盾,产生书中数字的程序因依赖漂移而不再复现,保存书稿的介质发生故障与静默腐烂。本书把「每卷都校过所以全库一致」「代码公开所以永久可检验」「多存几份就能放心」这些说法写成可以计算的算术,再看它们在什么参数下成立。书中的书库规模、错误率、风险率与故障率均为示意参数;全部计算由书中程序生成;本书不是法律、财务、采购或存储服务建议。 第一章把书库画成知识图谱,说明卷数越多错误越难藏身:示意书库到 200 卷时,0.9821 的偏差定义处在可以被另一卷反驳的位置。第二章比较四种比对器:只规范化术语的比对器精确率只有 0.073,改用区间相交规则后误报为零。第三章给出零失败审计的三法则,要以 95 % 的置信度证明冲突率低于 0.01,需要 299 条全部合格的抽查,且几乎与书库大小无关。第四章用风险率模型说明,纯标准库程序的半衰期为 34.7 年,带五个依赖的程序只有 2.2 年,每年逐字比对输出能把每百段程序十年中无人知晓的失效年压到 9.21。 第五章在 GF(2^8) 上用纯标准库实现系统形式的里德–所罗门擦除码,开销 1.50 倍即可容忍三块丢失,RS(10+4) 的 1,001 种存活组合全部可解。第六章用二项窗口模型与马尔可夫链计算年数据丢失概率,说明两倍的相关故障加速会让 MTTDL 缩小 15.9 倍,五年一次巡检又不做逐块校验时,读取有 0.11820 的概率拿到错误字节。第七章考辨四库全书的校勘记录与七阁存亡,并与拉赫曼谱系法对读:四份同源抄本加一份独立抄本,按抄本数投票每千字错 20.300 个,按家族比较只剩 0.100 个。第八章给出出版方清单、冷存储成本试算、勘误登记格式与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Library as a Graph: Volumes, Chapters, Terms and Numeric Claims as Nodes — Chapter 1 draws the library as a knowledge graph and shows that the more volumes there are, the fewer places an error can hide: at 200 volumes, 0.9821 of deviant definitions in the illustrative library sit where another volume can contradict them. (https://5dlibrary.com/answers/en/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c01/) 第1章 书库即图谱:卷、章、术语与数值命题的节点模型 — 第一章把书库画成知识图谱,说明卷数越多错误越难藏身:示意书库到 200 卷时,0.9821 的偏差定义处在可以被另一卷反驳的位置。 (https://5dlibrary.com/answers/zh/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c01/) - Ch.2 Normalization and Hash Indexing: False Alarms and Misses in Cross-Volume Conflict Detection — Chapter 2 compares four comparators: one that only normalizes terms has a precision of 0.073, while the interval-overlap rule has no false alarms. (https://5dlibrary.com/answers/en/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c02/) 第2章 规范化与哈希索引:跨卷定理冲突检测的误报与漏报 — 第二章比较四种比对器:只规范化术语的比对器精确率只有 0.073,改用区间相交规则后误报为零。 (https://5dlibrary.com/answers/zh/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c02/) - Ch.3 The Sampling Theory of Regression Audits: the Rule of Three and the Finite-Population Correction — Chapter 3 gives the rule of three for clean audits: showing a conflict rate below 0.01 with 95 % confidence takes 299 clean checks, almost regardless of library size. (https://5dlibrary.com/answers/en/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c03/) 第3章 回归审计的抽样理论:零失败时的三法则与有限总体修正 — 第三章给出零失败审计的三法则,要以 95 % 的置信度证明冲突率低于 0.01,需要 299 条全部合格的抽查,且几乎与书库大小无关。 (https://5dlibrary.com/answers/zh/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c03/) - Ch.4 The Decay of Code Reproducibility: Hazard Rates of Dependency Drift and Byte-Exact Output Regression — Chapter 4 uses a hazard model to show that standard-library programs have a half-life of 34.7 years against 2.2 years with five dependencies, and that a yearly byte-exact output comparison cuts the unknown failure-years of a decade to 9.21 per hundred programs. (https://5dlibrary.com/answers/en/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c04/) 第4章 代码可复现性衰减:依赖漂移的风险率与逐字输出回归 — 第四章用风险率模型说明,纯标准库程序的半衰期为 34.7 年,带五个依赖的程序只有 2.2 年,每年逐字比对输出能把每百段程序十年中无人知晓的失效年压到 9.21。 (https://5dlibrary.com/answers/zh/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c04/) - Ch.5 Reed-Solomon over GF(2^8): Erasure Coding and Recovery in the Standard Library — Chapter 5 implements a systematic Reed-Solomon erasure code over GF(2^8) in standard-library Python; an overhead of 1.50 survives three lost blocks, and all 1,001 survivor sets of RS(10+4) decode. (https://5dlibrary.com/answers/en/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c05/) 第5章 GF(2^8) 上的里德–所罗门:纯标准库编码与擦除恢复 — 第五章在 GF(2^8) 上用纯标准库实现系统形式的里德–所罗门擦除码,开销 1.50 倍即可容忍三块丢失,RS(10+4) 的 1,001 种存活组合全部可解。 (https://5dlibrary.com/answers/zh/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c05/) - Ch.6 The Reliability of Replicas and Erasure Codes: Binomial Windows, Markov Chains and Scrubbing for Bit Rot — Chapter 6 computes yearly data-loss probabilities with a binomial window model and Markov chains, showing that a twofold correlated acceleration of failures shrinks the MTTDL by 15.9 times and that, with a scrub every five years and no per-block checksums, a read returns wrong bytes with probability 0.11820. (https://5dlibrary.com/answers/en/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c06/) 第6章 副本与纠删码的可靠性:二项模型、马尔可夫链与位腐烂巡检 — 第六章用二项窗口模型与马尔可夫链计算年数据丢失概率,说明两倍的相关故障加速会让 MTTDL 缩小 15.9 倍,五年一次巡检又不做逐块校验时,读取有 0.11820 的概率拿到错误字节。 (https://5dlibrary.com/answers/zh/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c06/) - Ch.7 Classical Sources: Siku Collation, the Fate of the Seven Pavilions and Lachmann's Stemmatics — Chapter 7 examines the collation records of the Siku quanshu and the fate of its seven pavilion copies alongside Lachmann's stemmatics: with four kindred copies and one independent copy, voting by manuscripts leaves 20.300 wrong characters per thousand and comparing families only 0.100. (https://5dlibrary.com/answers/en/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c07/) 第7章 原典考辨:四库校勘、七阁存亡与拉赫曼谱系法 — 第七章考辨四库全书的校勘记录与七阁存亡,并与拉赫曼谱系法对读:四份同源抄本加一份独立抄本,按抄本数投票每千字错 20.300 个,按家族比较只剩 0.100 个。 (https://5dlibrary.com/answers/zh/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c07/) - Ch.8 Practice and Falsification: a Publisher's Checklists for Consistency, Reproducibility and Cold Storage — Chapter 8 offers checklists for publishers, a cold-storage cost calculator, an errata register format and a pre-registered test protocol. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c08/) 第8章 实践与证伪方案:出版方的一致性、可复现与冷存储清单 — 第八章给出出版方清单、冷存储成本试算、勘误登记格式与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cross-volume-theorem-conflict-detection-large-bilingual-monograph-libraries/c08/) ## 5DL-B-183 · Cold-Immersion Heat-Flow Equations, Brown-Fat Non-Shivering Thermogenesis and the Post-Training Cryotherapy Hypertrophy Penalty 中文书名:冷水浸浴热流方程、棕色脂肪非颤抖产热与练后冷疗肌肥大抵消母典 DOI: pending Cite: Fan, Y. (2026). Cold-Immersion Heat-Flow Equations, Brown-Fat Non-Shivering Thermogenesis and the Post-Training Cryotherapy Hypertrophy Penalty (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not medical advice, claims no health benefit, and asks readers to consult a licensed physician before any self-experiment. Chapter 5 derives a rule for the gap between training and cold water: with illustrative parameters a 6-hour gap cuts the interference to 0.135 of an immediate plunge, for a predicted growth ratio of 0.910. Abstract: Two claims surround ice baths and winter swimming: that an ice bath straight after strength training makes muscle grow more, and that cold water activates brown fat and burns fat. This book writes both claims as runnable calculations and asks when they hold. Its parameters are illustrative; all calculations are generated by its programs; it is not medical advice, claims no health benefit, and asks readers to consult a licensed physician before any self-experiment. Chapter 1 converts the raw data of the 2015 trial by Roberts and colleagues: the cold group gained quadriceps lean mass at 0.333 of the control rate, a Hedges g of 2.75; but pooled analysis puts the hypertrophy effect at only −0.22, with a credible interval crossing zero, so the single trial is 12.5 times the pooled value. Chapter 2 calibrates the rise in metabolism on an immersion experiment: five minutes at 10 °C cost about 48.0 kcal of excess energy, three plunges a week for a year come to 0.97 kg, and eating it back shrinks that. Chapter 3 solves the heat-flow equation of a skin-fat-core cylinder: ten minutes of cold water draw some 210 kcal of heat, mostly stored heat that the surroundings can repay; the convective coefficients of water and air differ 25-fold, steady heat loss only 1.90-fold. Chapter 4 shows that brown fat offsets at most 0.0164 of the heat lost in cold water and lowers the critical water temperature by only 0.71 °C. Chapter 5 derives a rule for the gap between training and cold water: with illustrative parameters a 6-hour gap cuts the interference to 0.135 of an immediate plunge, for a predicted growth ratio of 0.910. Chapter 6 compares the cooling dose to muscle and the core safety margin under published protocols. Chapter 7 examines the Suwen's "leave the cold, approach warmth; do not let the skin leak" and Seneca's New Year plunge. Chapter 8 gives a cold-and-heat bath scheduler, an A-B-A personal log and a pre-registered protocol: testing the pooled hypertrophy effect needs 325 people per arm. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:冰浴与冬泳流行着两条说法:力量训练后立刻泡冰水能长得更多;冷水能激活棕色脂肪、帮人燃脂。本书把两条说法都写成可以运行的计算,再看它们在什么条件下成立。书中参数为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何健康功效,自我试验前须先咨询执业医师。 第一章换算 Roberts 等人 2015 年试验的原始数据:冷水组股四头肌瘦体量的增加只有对照组的 0.333,Hedges g 为 2.75;但合并分析给出的增肌效应只有 −0.22,可信区间跨过零,单项试验是合并值的 12.5 倍。第二章用浸浴实验校准代谢上升:在 10 °C 水中泡 5 分钟额外耗能约 48.0 千卡,每周三次一年折合 0.97 千克,而且会被进食补偿削弱。第三章解皮肤—脂肪—核心三层圆柱的热流方程:十分钟冷浴散热 210 千卡上下,大部分是可由外界偿还的储热;水与空气的换热系数差 25 倍,稳态散热只差 1.90 倍。第四章说明棕色脂肪最多抵消冷水散热的 0.0164,只把临界水温降低 0.71 °C。 第五章推出训练与冷浴的间隔规则:在示意参数下,间隔 6 小时把干扰压到立刻冷浴时的 0.135,预测增肌比为 0.910。第六章比较文献方案对肌肉的冷却剂量与核心安全边界。第七章考辨《素问》「去寒就温,无泄皮肤」与塞涅卡的元旦冬泳。第八章给出冷热浴排程器、A-B-A 个人日志与预注册方案:检验合并增肌效应每组需 325 人。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Post-Training Ice-Bath Penalty: One Trial and Two Pooled Analyses — Chapter 1 converts the raw data of the 2015 trial by Roberts and colleagues: the cold group gained quadriceps lean mass at 0.333 of the control rate, a Hedges g of 2.75; but pooled analysis puts the hypertrophy effect at only −0.22, with a credible interval crossing zero, so the single trial is 12.5 times the pooled value. (https://5dlibrary.com/answers/en/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c01/) 第1章 练后冰浴的增肌罚款:一项试验与两份合并分析 — 第一章换算 Roberts 等人 2015 年试验的原始数据:冷水组股四头肌瘦体量的增加只有对照组的 0.333,Hedges g 为 2.75;但合并分析给出的增肌效应只有 −0.22,可信区间跨过零,单项试验是合并值的 12.5 倍。 (https://5dlibrary.com/answers/zh/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c01/) - Ch.2 The Fat-Burning Ledger: One Cold Plunge Is Worth a Few Dozen Kilocalories — Chapter 2 calibrates the rise in metabolism on an immersion experiment: five minutes at 10 °C cost about 48.0 kcal of excess energy, three plunges a week for a year come to 0.97 kg, and eating it back shrinks that. (https://5dlibrary.com/answers/en/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c02/) 第2章 冷水燃脂的账本:一次冷浴值几十千卡 — 第二章用浸浴实验校准代谢上升:在 10 °C 水中泡 5 分钟额外耗能约 48.0 千卡,每周三次一年折合 0.97 千克,而且会被进食补偿削弱。 (https://5dlibrary.com/answers/zh/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c02/) - Ch.3 A Skin-Fat-Core Cylinder: Finite Differences and Newton Cooling — Chapter 3 solves the heat-flow equation of a skin-fat-core cylinder: ten minutes of cold water draw some 210 kcal of heat, mostly stored heat that the surroundings can repay; the convective coefficients of water and air differ 25-fold, steady heat loss only 1.90-fold. (https://5dlibrary.com/answers/en/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c03/) 第3章 皮肤—脂肪—核心三层圆柱:有限差分与牛顿冷却 — 第三章解皮肤—脂肪—核心三层圆柱的热流方程:十分钟冷浴散热 210 千卡上下,大部分是可由外界偿还的储热;水与空气的换热系数差 25 倍,稳态散热只差 1.90 倍。 (https://5dlibrary.com/answers/zh/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c03/) - Ch.4 Brown-Fat Non-Shivering Thermogenesis: a Power Ceiling and a Critical Water Temperature — Chapter 4 shows that brown fat offsets at most 0.0164 of the heat lost in cold water and lowers the critical water temperature by only 0.71 °C. (https://5dlibrary.com/answers/en/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c04/) 第4章 棕色脂肪的非颤抖产热:功率上限与临界水温 — 第四章说明棕色脂肪最多抵消冷水散热的 0.0164,只把临界水温降低 0.71 °C。 (https://5dlibrary.com/answers/zh/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c04/) - Ch.5 A Suppression Function for the Training-to-Cold Gap, with a Sensitivity Scan — Chapter 5 derives a rule for the gap between training and cold water: with illustrative parameters a 6-hour gap cuts the interference to 0.135 of an immediate plunge, for a predicted growth ratio of 0.910. (https://5dlibrary.com/answers/en/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c05/) 第5章 训练—冷浴间隔的抑制函数与敏感度扫描 — 第五章推出训练与冷浴的间隔规则:在示意参数下,间隔 6 小时把干扰压到立刻冷浴时的 0.135,预测增肌比为 0.910。 (https://5dlibrary.com/answers/zh/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c05/) - Ch.6 Cooling Dose, Residual Muscle Cooling and the Core Safety Margin — In the equivalent-cylinder model, a 3-minute plunge at 3 °C cools the top of the muscle only to 34.6 °C, ten minutes at 10 °C take it to 31.7 °C, and protocols of 20 minutes or an hour bring it to about 29 °C; after exit the deeper muscle keeps cooling. A plunge of a few minutes barely touches core temperature, whereas an hour of immersion can take a lean core to 33.44 °C or below, followed by an afterdrop after leaving the water. The cooling dose should be measured in degree-minutes of muscle temperature, not in water temperature. The next chapter leaves the laboratory to read what two classical texts said about cold. (https://5dlibrary.com/answers/en/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c06/) 第6章 冷却剂量、肌温余冷与核心安全边界 — 在等效圆柱模型中,3 °C、3 分钟的冷浴让肌肉表层只降到 34.6 °C,10 °C、10 分钟降到 31.7 °C,20 分钟或一小时级的方案降到 29 °C 上下;出水后深处肌肉还会继续变冷。几分钟的冷浴对核心温度影响很小,一小时级浸浴则可以让瘦人核心降到 33.44 °C 以下,并伴随出水后的后降。冷却剂量应当用肌肉温度的度·分钟来衡量,而不是用水温。下一章离开实验室,去读两部古典文献对寒冷的看法。 (https://5dlibrary.com/answers/zh/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c06/) - Ch.7 Classical Sources East and West: Winter Storing in the Suwen and Seneca's New Year Plunge — The Suwen's "leave the cold, approach warmth; do not let the skin leak" can be written as a relation between clothing insulation and critical air temperature: in 3 clo, resting heat alone holds the body in air at 5.9 °C. Seneca's retreat can be written as the rise in critical water temperature once the ceiling of heat production falls, from 12.1 °C to 22.4 °C. Neither source treats cold water as a means of burning fat. The next chapter gathers the book's calculations into a loop that can be executed: scheduling, recording and testing. (https://5dlibrary.com/answers/en/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c07/) 第7章 东西方原典考辨:《素问》的冬藏与塞涅卡的元旦冬泳 — 《素问》的「去寒就温,无泄皮肤」,可以写成衣着热阻与临界气温的关系:穿 3 克罗时,只靠静息产热就能在 5.9 °C 的空气中维持体温。塞涅卡的退却,可以写成产热上限下降后临界水温的上升:从 12.1 °C 升到 22.4 °C。两部原典都没有把冷水当作燃脂手段。下一章把全书的计算收拢成一个可以执行的闭环:排程、记录、检验。 (https://5dlibrary.com/answers/zh/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c07/) - Ch.8 Closing the Loop: the Cold-and-Heat Bath Scheduler and a Pre-Registered A-B-A Test — Chapter 8 gives a cold-and-heat bath scheduler, an A-B-A personal log and a pre-registered protocol: testing the pooled hypertrophy effect needs 325 people per arm. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c08/) 第8章 实证闭环:冷热浴排程器与 A-B-A 预注册检验 — 第八章给出冷热浴排程器、A-B-A 个人日志与预注册方案:检验合并增肌效应每组需 325 人。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cold-immersion-heat-flow-brown-fat-cryotherapy-hypertrophy-penalty/c08/) ## 5DL-B-184 · Sitting Disc-Pressure Levers, the 135-Degree Recline and the Standing-Desk Energy Myth 中文书名:久坐椎间盘内压杠杆、一百三十五度后仰最低负荷与站立办公能耗神话母典 DOI: pending Cite: Fan, Y. (2026). Sitting Disc-Pressure Levers, the 135-Degree Recline and the Standing-Desk Energy Myth (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs. The book is not medical advice and claims no health effect for any intervention; readers with existing back problems should consult a licensed physician before changing a workstation or starting a self-experiment. Abstract: This book tests two popular pieces of bodily common sense: that "sitting up straight" is the sitting posture with the lowest lumbar load, and that working on one's feet burns a worthwhile amount of extra energy. It writes both claims as runnable arithmetic, using lumbar lever statics, Poisson posture switching, first-order creep and statistical power, and then asks when they hold. Its parameters are illustrative; all calculations are generated by its programs. The book is not medical advice and claims no health effect for any intervention; readers with existing back problems should consult a licensed physician before changing a workstation or starting a self-experiment. Chapter 1 reads Wilke's in vivo pressures of 1999: actively straightened sitting gives 0.55 MPa and reclining down into an armchair 0.27 MPa, a ratio of 2.037, set beside Nachemson's older data and the model estimates of Callaghan and McGill. Chapter 2 shows that the 135 degrees measure the trunk-thigh angle; a statics model puts compression at that angle at only 0.441 of actively straightened sitting, and the supporting studies had 1, 4 and 22 subjects. Chapter 3 writes down a lever with an arm of about 5 centimetres, on which a forward bend of about twenty degrees doubles the pressure. Chapter 4 audits the energy of standing work: about 9.0 kcal an hour, and once fading novelty and compensatory eating are included the median static upper bound for a year is only 0.208 kg. Chapter 5 uses a Poisson process and a threshold dose to show that regular switching beats random switching, and that the returns of the disc pump diminish as switching grows more frequent. Chapter 6 gives seat height, footrest, standing desk height and the screen position after reclining. Chapter 7 examines "if his mat was not straight, he did not sit" and "not like a corpse in bed, no formal bearing at home" from the Analects, and the bound sapling of Andry's Orthopaedia. Chapter 8 provides a workstation parameter card, a thirty-minute reminder and a two-week diary in randomised order, and simulates how a naive before-and-after design reports "significant" results in 0.588 of studies when there is no effect. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:本书检验两条流行的身体常识:「坐直」是腰椎负荷最低的坐姿;站着办公能多消耗可观的能量。它把两条说法写成可以运行的算术:腰椎杠杆静力学、泊松姿势切换、一阶蠕变与统计功效,再看它们在什么条件下成立。书中参数为示意;全部计算由书中程序生成。本书不是医疗建议,不宣称任何干预的健康效果;已有腰背问题的读者调整工位或开展自我试验前,应先咨询执业医师。 第一章读维尔克 1999 年的活体测压:主动挺直坐为 0.55 MPa,靠在扶手椅里后仰下滑为 0.27 MPa,两者之比为 2.037,并与纳赫姆森的旧数据和卡拉汉、麦吉尔的模型估计并陈。第二章说明 135 度量的是躯干-大腿角,静力学模型给出该角度的压缩力只有主动挺直坐的 0.441,并指出相关研究只有 1 人、4 人与 22 人。第三章写出约 5 厘米力臂的杠杆,前屈约二十度即可使压力翻倍。第四章核算站立办公的能耗:每小时约 9.0 千卡,把新鲜感衰退与进食补偿算进来,全年静态上限的中位数只有 0.208 kg。 第五章用泊松过程与阈值剂量说明规律切换胜过随机切换,椎间盘泵的收益随切换加密而递减。第六章给出座高、脚踏、站立桌面与后仰后的屏幕位置。第七章考辨《论语·乡党》「席不正,不坐」「寝不尸,居不容」与安德里《矫形术》中被绑直的小树。第八章给出工位参数卡、三十分钟提醒与随机顺序的两周日志,并模拟出朴素前后对照在无效应时有 0.588 的研究报告「显著」。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Pressure in Vivo: Why the Upright Sitting Posture Is Not the Lowest Load — Chapter 1 reads Wilke's in vivo pressures of 1999: actively straightened sitting gives 0.55 MPa and reclining down into an armchair 0.27 MPa, a ratio of 2.037, set beside Nachemson's older data and the model estimates of Callaghan and McGill. (https://5dlibrary.com/answers/en/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c01/) 第1章 活体测压:挺直的坐姿为什么不是腰椎负荷最低的坐姿 — 第一章读维尔克 1999 年的活体测压:主动挺直坐为 0.55 MPa,靠在扶手椅里后仰下滑为 0.27 MPa,两者之比为 2.037,并与纳赫姆森的旧数据和卡拉汉、麦吉尔的模型估计并陈。 (https://5dlibrary.com/answers/zh/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c01/) - Ch.2 One Hundred and Thirty-Five Degrees: the Backrest Takes the Load, the Hip Frees the Pelvis — Chapter 2 shows that the 135 degrees measure the trunk-thigh angle; a statics model puts compression at that angle at only 0.441 of actively straightened sitting, and the supporting studies had 1, 4 and 22 subjects. (https://5dlibrary.com/answers/en/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c02/) 第2章 一百三十五度:靠背承重、髋部放开骨盆 — 第二章说明 135 度量的是躯干-大腿角,静力学模型给出该角度的压缩力只有主动挺直坐的 0.441,并指出相关研究只有 1 人、4 人与 22 人。 (https://5dlibrary.com/answers/zh/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c02/) - Ch.3 A Five-Centimetre Lever Arm: How Forward Bending Multiplies Lumbar Compression — Chapter 3 writes down a lever with an arm of about 5 centimetres, on which a forward bend of about twenty degrees doubles the pressure. (https://5dlibrary.com/answers/en/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c03/) 第3章 五厘米力臂:前屈角如何放大腰椎压缩力 — 第三章写出约 5 厘米力臂的杠杆,前屈约二十度即可使压力翻倍。 (https://5dlibrary.com/answers/zh/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c03/) - Ch.4 Nine Kilocalories an Hour: the Standing-Desk Energy Myth and the Real Ledger — Chapter 4 audits the energy of standing work: about 9.0 kcal an hour, and once fading novelty and compensatory eating are included the median static upper bound for a year is only 0.208 kg. (https://5dlibrary.com/answers/en/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c04/) 第4章 每小时九千卡:站立办公的能耗神话与真实账本 — 第四章核算站立办公的能耗:每小时约 9.0 千卡,把新鲜感衰退与进食补偿算进来,全年静态上限的中位数只有 0.208 kg。 (https://5dlibrary.com/answers/zh/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c04/) - Ch.5 Poisson Switching: Why the Cumulative Load Integral Cannot See the Cost of Staying Still — The cumulative load integral is completely insensitive to switching frequency, so measures such as "minutes standing per day" cannot see the cost of postural monotony. Even at an average of one switch every 15 minutes, random switching leaves nearly an even chance of an hour-long static bout in the day; at the same mean interval, regular switching yields far less static dose than random switching, and a fall in obedience from nine to seven in ten multiplies the dose about fourfold. The disc-pump argument is close to its ceiling at one switch every half hour, and more frequent switching brings diminishing returns. The next chapter turns these conclusions into concrete workstation geometry: screen, chair angle and footrest. (https://5dlibrary.com/answers/en/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c05/) 第5章 泊松切换:累积负荷积分为什么看不出静态的代价 — 累积负荷积分对切换频率完全不敏感,所以「每天站多少分钟」这类指标看不出姿势单调的代价。泊松切换即使平均每 15 分钟一次,一天里仍有近一半的机会出现超过一小时的静态段;同样平均间隔下,规律切换的静态剂量远低于随机切换,执行率从九成降到七成会让剂量增加约四倍。椎间盘泵论证在半小时一换时已接近上限,更频繁的切换收益递减。下一章把这些结论落到具体的工位几何上:屏幕、椅角与脚踏。 (https://5dlibrary.com/answers/zh/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c05/) - Ch.6 Workstation Geometry: Seat Height, Footrest, Standing Desk and the Screen After Reclining — Seat height is set by popliteal height and typing height by elbow height; when the two disagree, short people need a footrest and tall people a higher desk, and with a fixed desk of 0.73 metres everyone shorter than 1.821 metres needs a footrest. Standing desk height varies with stature by more than twenty centimetres, so no fixed height fits a whole team. Reclining to 135 degrees leaves screen height almost unchanged but requires bringing the screen more than half a metre closer. The next chapter leaves the modern workstation for two old texts: one says "if the mat was not straight, he would not sit," the other pictures a young tree bound straight to a stake. (https://5dlibrary.com/answers/en/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c06/) 第6章 工位几何:座高、脚踏、站立桌面与后仰后的屏幕 — 座面高度由腘窝高决定,打字高度由肘高决定,两者对不上时,矮个子需要脚踏,高个子需要更高的桌面;以 0.73 米的固定桌面计,身高低于 1.821 米的人都需要脚踏。站立桌面高度随身高变化二十多厘米,固定高度无法适合整个团队。后仰到 135 度时,屏幕高度几乎不变,却要拉近半米以上。下一章离开现代工位,回到两部古老的文献:一部讲「席不正,不坐」,一部画了一棵被绑直的小树。 (https://5dlibrary.com/answers/zh/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c06/) - Ch.7 The Crooked Mat and the Bound Sapling: Classical Sources East and West Examined — The program sets five 8-hour schedules side by side. Andry's schedule of "upright all day," two bouts of 240 minutes, has a mean compression of 660 newtons, a load integral of 5280 newton-hours and a static dose of 44100. All-day relaxed sitting and all-day standing have exactly the same static dose, also 44100, with mean pressures lowered to 552 and 600 newtons respectively. Two bouts each of sitting and standing at 120 minutes bring the dose down to 16200. Rotating every 30 minutes among reclining, standing and relaxed sitting gives a mean compression of only 469 newtons, a load integral of 3753, zero static dose and a longest bout of 30 minutes. (https://5dlibrary.com/answers/en/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c07/) 第7章 席不正不坐与绑直的小树:东西方原典考辨 — 程序把五种 8 小时日程放在一起。安德里式的「全天挺直」日程,两段各 240 分钟,平均压缩力 660 牛,负荷积分 5280 牛·小时,静态剂量 44100。全天放松坐和全天站立的静态剂量完全相同,也是 44100,只是平均压力分别降到 552 和 600 牛。坐站各两段、每段 120 分钟,剂量降到 16200。而每 30 分钟在后仰、站立、放松坐三种姿势之间轮换,平均压缩力只有 469 牛,负荷积分 3753,静态剂量为零,最长一段 30 分钟。 (https://5dlibrary.com/answers/zh/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c07/) - Ch.8 Closing the Loop: the Workstation Parameter Card, the Thirty-Minute Reminder and the Two-Week Diary — Chapter 8 provides a workstation parameter card, a thirty-minute reminder and a two-week diary in randomised order, and simulates how a naive before-and-after design reports "significant" results in 0.588 of studies when there is no effect. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c08/) 第8章 实证闭环:工位参数卡、三十分钟提醒与两周日志 — 第八章给出工位参数卡、三十分钟提醒与随机顺序的两周日志,并模拟出朴素前后对照在无效应时有 0.588 的研究报告「显著」。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/sitting-disc-pressure-levers-135-degree-recline-standing-desk-myth/c08/) ## 5DL-B-185 · Lotka-Volterra Competition in Fermented-Food Microbiota, Probiotic-Capsule Colonisation Failure and Dietary-Diversity Inflammation Markers 中文书名:发酵食品菌群洛特卡—沃尔泰拉竞争、益生菌胶囊定植失败与膳食多样性炎症指标母典 DOI: pending Cite: Fan, Y. (2026). Lotka-Volterra Competition in Fermented-Food Microbiota, Probiotic-Capsule Colonisation Failure and Dietary-Diversity Inflammation Markers (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Under the book's illustrative parameters that number is -0.2900 a day, so a capsule strain is merely carried, passively, while dosing continues, at a standing carriage of 0.0690 and with a half-life of 2.39 days once it stops. Parameters are illustrative; every calculation is generated by the book's programs; the book is not medical advice, claims no effect, and any self-experiment should be preceded by consulting a licensed physician. Abstract: Can a microbial community whose space is already taken be rewritten by swallowing a capsule? This book writes that question as generalised Lotka-Volterra competition, solves the resident community to equilibrium, and compresses an incoming strain's prospects into one number, the invasion growth rate. Under the book's illustrative parameters that number is -0.2900 a day, so a capsule strain is merely carried, passively, while dosing continues, at a standing carriage of 0.0690 and with a half-life of 2.39 days once it stops. Flipping the sign requires pressing the residents below 0.6548 of baseline, which is what antibiotics do. Parameters are illustrative; every calculation is generated by the book's programs; the book is not medical advice, claims no effect, and any self-experiment should be preceded by consulting a licensed physician. Chapter 1 gives the full arithmetic of the colonisation threshold and of passive carriage. Chapter 2 states the counterintuitive claim head-on: capsules are mostly transient, and taking them after antibiotics postpones the return of a person's own community, the model ordering being 1.92 days for a stored community, 9.24 days spontaneously and 31.64 days with capsules. It also separates failing to settle from having no effect, where one pooled preventive use carries a risk ratio of 0.40. Chapter 3 asks which order of diversity rose, with one community changing by 0.1667 at order zero and only 0.0239 at order two. Chapter 4 splits fermented food and dietary fibre into an arrival stream and a substrate stream, the first taking the steady taxon count from 120.0 to 140.0 and the second moving only flux. Chapter 5 handles the statistics of an inflammation panel: a set of proteins falling together is extremely rare under the null, while eighteen people can see only 0.6603 of a standard deviation. Chapter 6 runs a grid scan and elasticities, with twenty of twenty-five cells supporting the central conclusion. Chapter 7 examines the sauce and bean-ferment methods of the Qimin Yaoshu and Metchnikoff's soured-milk argument of 1907, the former's serial-transfer model reaching a purity of 0.9821 in three transfers while the latter's displacement mechanism fails in the competition equations. Chapter 8 closes the loop with a product-description template free of body-effect wording, a four-week diversity scorecard, and a pre-registered protocol aimed at 13.1865. The book contains 18 runnable programs. 摘要:一个已经占满的微生物群落,能不能靠吞下一粒胶囊改写?本书把这个问题写成广义洛特卡—沃尔泰拉竞争方程,解出原住群落的平衡态,再把外来菌株的命运压缩成一个数:入侵增长率。在书中的示意参数下这个数是 -0.2900 每天,于是胶囊菌株只能在持续服用期间被动携带,携带量 0.0690,停服后半衰期 2.39 天。要让这个符号翻正,原住群落必须被压到基线的 0.6548 以下——而这正是抗生素做的事。书中参数为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何效果,自我试验须先咨询执业医师。 第一章给出定植阈值与被动携带的全部算术。第二章正面陈述本书的反常识命题:胶囊多为一过性,抗生素之后服用会推迟原生菌群恢复——模型给出的顺序是自体回输 1.92 天、自然恢复 9.24 天、胶囊 31.64 天;同时指出「不定植」不等于「无效应」,某项预防用途的汇总风险比是 0.40。第三章追问「多样性提高」到底是哪一阶,同一群落按零阶变化 0.1667,按二阶只有 0.0239。第四章把发酵食品与膳食纤维分成到达流与底物流两根杠杆:前者把稳态类群数从 120.0 推到 140.0,后者只推通量。 第五章处理炎症面板的统计:一组蛋白同向下降在零假设下极其罕见,而十八个人只能看见 0.6603 个标准差的效应。第六章做网格扫描与弹性分析,二十五格中二十格支持核心结论。第七章考辨《齐民要术》的作酱作豉之法与梅契尼科夫一九零七年的酸乳论,前者的传代模型三次即达 0.9821 的纯度,后者的「取代」机制在竞争方程里不成立。第八章给出实证闭环:零身体效果字样的选品说明模板、四周饮食多样性计分卡,以及以 13.1865 为靶的预注册方案。全书含 18 段可运行程序。 Chapter answers: - Ch.1 The Colonisation Threshold: Invasion Growth Rates in a Competition Model — Once topping up is written as equations the question becomes clean. Whether a strain stays is decided by the sign of its invasion growth rate, which under this chapter's illustrative parameters is -0.2900 a day. Under a negative sign a dose buys the carriage of 0.0690 that lasts while it is taken and nothing beyond; to flip the sign, the resident community must be pressed below 0.6548 of baseline. (https://5dlibrary.com/answers/en/fermented-food-lotka-volterra-probiotic-colonisation-failure/c01/) 第1章 定植阈值:竞争方程里的入侵增长率 — 把「补菌」翻译成方程以后,问题变得干净:一个菌株能不能留下,由入侵增长率的符号决定;在本章的示意参数下这个数是 -0.2900 每天。负号之下,剂量只能买到服用期内的携带量 0.0690,买不到停服后的存续;要让符号翻正,需要把原住群落压到基线的 0.6548 以下。 (https://5dlibrary.com/answers/zh/fermented-food-lotka-volterra-probiotic-colonisation-failure/c01/) - Ch.2 Transient by Design: The Order of Post-Antibiotic Recovery, and Why Failing to Settle Is Not Failing to Act — Chapter 2 states the counterintuitive claim head-on: capsules are mostly transient, and taking them after antibiotics postpones the return of a person's own community, the model ordering being 1.92 days for a stored community, 9.24 days spontaneously and 31.64 days with capsules. It also separates failing to settle from having no effect, where one pooled preventive use carries a risk ratio of 0.40. (https://5dlibrary.com/answers/en/fermented-food-lotka-volterra-probiotic-colonisation-failure/c02/) 第2章 一过性:抗生素之后的恢复顺序,与「不定植」不等于「无效应」 — 第二章正面陈述本书的反常识命题:胶囊多为一过性,抗生素之后服用会推迟原生菌群恢复——模型给出的顺序是自体回输 1.92 天、自然恢复 9.24 天、胶囊 31.64 天;同时指出「不定植」不等于「无效应」,某项预防用途的汇总风险比是 0.40。 (https://5dlibrary.com/answers/zh/fermented-food-lotka-volterra-probiotic-colonisation-failure/c02/) - Ch.3 Orders of Diversity: Trajectories, the Shannon Index and Hill Numbers — Chapter 3 asks which order of diversity rose, with one community changing by 0.1667 at order zero and only 0.0239 at order two. (https://5dlibrary.com/answers/en/fermented-food-lotka-volterra-probiotic-colonisation-failure/c03/) 第3章 多样性的阶数:轨迹、香农指数与希尔数 — 第三章追问「多样性提高」到底是哪一阶,同一群落按零阶变化 0.1667,按二阶只有 0.0239。 (https://5dlibrary.com/answers/zh/fermented-food-lotka-volterra-probiotic-colonisation-failure/c03/) - Ch.4 Two Levers: The Arrival Stream and the Substrate Stream — Chapter 4 splits fermented food and dietary fibre into an arrival stream and a substrate stream, the first taking the steady taxon count from 120.0 to 140.0 and the second moving only flux. (https://5dlibrary.com/answers/en/fermented-food-lotka-volterra-probiotic-colonisation-failure/c04/) 第4章 两根杠杆:到达流与底物流 — 第四章把发酵食品与膳食纤维分成到达流与底物流两根杠杆:前者把稳态类群数从 120.0 推到 140.0,后者只推通量。 (https://5dlibrary.com/answers/zh/fermented-food-lotka-volterra-probiotic-colonisation-failure/c04/) - Ch.5 The Statistics of a Panel: What a Set of Proteins Falling Together Means — Chapter 5 handles the statistics of an inflammation panel: a set of proteins falling together is extremely rare under the null, while eighteen people can see only 0.6603 of a standard deviation. (https://5dlibrary.com/answers/en/fermented-food-lotka-volterra-probiotic-colonisation-failure/c05/) 第5章 面板的统计学:一组蛋白同时下降意味着什么 — 第五章处理炎症面板的统计:一组蛋白同向下降在零假设下极其罕见,而十八个人只能看见 0.6603 个标准差的效应。 (https://5dlibrary.com/answers/zh/fermented-food-lotka-volterra-probiotic-colonisation-failure/c05/) - Ch.6 Parameter Scans and Elasticities: How Far the Conclusions Hold — Placed on a grid of twenty-five cells, the first conclusion holds in twenty, with the critical share spanning 0.2500 to 1.0000; the failing cells cluster in the corner of fast growth and weak competition, in agreement with experience. The elasticity analysis gives carriage an elasticity of 1.000 to the dose, -2.246 to the resident total, and 0.000 for the dose's effect on clearance speed, while the gain in taxon count responds only to the pool (1.000), the arrival coefficient per serving (0.917), the local loss rate (-0.350) and the arrival coefficient at zero servings (-0.547). The two mechanisms are orthogonal in the model, and that orthogonality is itself a prediction a two-by-two design could destroy. The next chapter leaves contemporary data for two older sources, one on making sauces and bean ferments, one on soured milk and long life. (https://5dlibrary.com/answers/en/fermented-food-lotka-volterra-probiotic-colonisation-failure/c06/) 第6章 参数扫描与弹性:结论在多大范围内成立 — 把第一个结论放到二十五格的网格上,有二十格成立,临界份额跨度从 0.2500 到 1.0000;失效的格子集中在增长快而竞争弱的角落,与经验一致。弹性分析显示,携带量对剂量的弹性是 1.000,对原住群落总量是 -2.246,对清除速度剂量的弹性是 0.000;类群数增量只对类群池(1.000)、每份到达系数(0.917)、本地消失率(-0.350)与零份到达系数(-0.547)敏感。两套机制在模型里正交,而这正交性本身是一个可以被二乘二设计推翻的预测。一个模型值不值得读,不看它的输出有多漂亮,而看它把自己放在多少条可以被切断的绳子上;本章的任务就是把这些绳子一根根数清楚。扫描与弹性这两件工具并不高深,它们只要求作者在发表之前先对自己做一次攻击;多数模型之所以经不起追问,不是因为算错了,而是因为从来没有人替它做过这一步。把攻击写进正文,读者就不必自己去猜哪里薄弱;而作者也会在写作过程中被迫承认,哪几条结论其实是靠参数撑住的,哪几条是靠结构撑住的。下一章离开当代数据,去看两部原典:一部讲怎么做酱做豉,一部讲酸乳与长寿。 (https://5dlibrary.com/answers/zh/fermented-food-lotka-volterra-probiotic-colonisation-failure/c06/) - Ch.7 Two Classical Sources: Sauce and Bean Ferments in the Qimin Yaoshu, and Metchnikoff on Soured Milk — Chapter 7 examines the sauce and bean-ferment methods of the Qimin Yaoshu and Metchnikoff's soured-milk argument of 1907, the former's serial-transfer model reaching a purity of 0.9821 in three transfers while the latter's displacement mechanism fails in the competition equations. (https://5dlibrary.com/answers/en/fermented-food-lotka-volterra-probiotic-colonisation-failure/c07/) 第7章 东西方原典考辨:《齐民要术》的作酱作豉与梅契尼科夫的酸乳论 — 第七章考辨《齐民要术》的作酱作豉之法与梅契尼科夫一九零七年的酸乳论,前者的传代模型三次即达 0.9821 的纯度,后者的「取代」机制在竞争方程里不成立。 (https://5dlibrary.com/answers/zh/fermented-food-lotka-volterra-probiotic-colonisation-failure/c07/) - Ch.8 Closing the Loop: A Scorecard, a Compliant Template and a Pre-Registered Protocol — Chapter 8 closes the loop with a product-description template free of body-effect wording, a four-week diversity scorecard, and a pre-registered protocol aimed at 13.1865. The book contains 18 runnable programs. (https://5dlibrary.com/answers/en/fermented-food-lotka-volterra-probiotic-colonisation-failure/c08/) 第8章 实证闭环:计分卡、合规模板与预注册方案 — 第八章给出实证闭环:零身体效果字样的选品说明模板、四周饮食多样性计分卡,以及以 13.1865 为靶的预注册方案。全书含 18 段可运行程序。 (https://5dlibrary.com/answers/zh/fermented-food-lotka-volterra-probiotic-colonisation-failure/c08/) ## 5DL-B-186 · Caffeine-Theanine Pharmacokinetics, Adenosine-Receptor Half-Lives and the Sleep Debt of Afternoon Tea 中文书名:咖啡因茶氨酸协同药代动力学、腺苷受体半衰期与下午茶睡眠负债母典 DOI: pending Cite: Fan, Y. (2026). Caffeine-Theanine Pharmacokinetics, Adenosine-Receptor Half-Lives and the Sleep Debt of Afternoon Tea (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Except where a source is named, its parameters are illustrative; all calculations are generated by its programs. The book is not medical advice and claims no health effect; anyone planning a self-experiment should first consult a licensed physician. Abstract: "Tea does not disturb sleep, because theanine cancels the caffeine" is a widely repeated claim. This book takes it apart and tests the pieces: how much theanine a cup of tea holds and how much the studies used; how much sleep caffeine still takes when drunk hours before bed; and by what time an afternoon tea should be finished. Except where a source is named, its parameters are illustrative; all calculations are generated by its programs. The book is not medical advice and claims no health effect; anyone planning a self-experiment should first consult a licensed physician. Chapter 1 measures the dose gap: a cup of tea holds about 7.9 to 29.9 mg of theanine, reaching the research dose of 200 mg takes 8.26 cups of black tea and brings 310.7 mg of caffeine, and the much-quoted "six hours before bed still costs an hour" comes from a twelve-person trial of 400 mg with an objective loss of 1.1 to 1.2 hours. Chapter 2 uses the Bateman curve of one-compartment oral absorption and calibrates a night-exposure threshold of 0.6054 mg per litre on the coffee cut-off of the Gardiner meta-analysis; it predicts a pre-workout cut-off of 13.92 hours (13.2 in the paper), and the two published cut-offs imply a half-life of 4.30 hours. Chapter 3 treats fast and slow CYP1A2 metabolism and stacked doses: in a lognormal population about 4.7 % cross the line after one black tea at four, against 0.694 after one coffee. Chapter 4 converts concentration into adenosine-receptor occupancy and explains circadian gating and saturation. Chapter 5 solves for the last cup by bisection: at a half-life of 5 hours a single cup below 34.5 mg needs no cut-off, and moving coffee from covering half of customers to nine in ten brings its cut-off 6.90 hours earlier. Chapter 6 brackets sleep debt with three loss models. Chapter 7 examines the bowl regime of Lu Yu's Classic of Tea (4.51 to 32.23 mg per guest) and Balzac's escalation and tolerance in his Treatise on Modern Stimulants. Chapter 8 gives caffeine labels for a shop, a fourteen-night personal trial (one person can detect only differences above about 22.5 minutes, with a power of just 0.205 for 9.1 minutes), a pre-registered protocol, the conditions that would refute the book, and the distilled tool, the Caffeine Cutoff Calculator. The book contains 22 programs. 摘要:「喝茶不影响睡眠,因为茶氨酸抵消了咖啡因」是一句流传极广的说法。本书把它拆开检验:一杯茶里有多少茶氨酸,研究用了多少;睡前几小时的咖啡因还会偷走多少睡眠;一杯下午茶最晚几点喝。书中参数除标明出处者外均为示意;全部计算由书中程序生成。本书不是医疗建议,不宣称任何健康效果;自我试验须先咨询执业医师。 第一章量出剂量鸿沟:一杯茶的茶氨酸约 7.9 至 29.9 毫克,要喝到研究用的 200 毫克需要 8.26 杯红茶,同时摄入 310.7 毫克咖啡因;常被引用的「睡前 6 小时仍少睡一小时」来自 400 毫克、12 人的试验,客观损失为 1.1 至 1.2 小时。第二章用一室口服吸收的贝特曼曲线,以 Gardiner 元分析的咖啡截止点标定夜间暴露阈值每升 0.6054 毫克,它预测补剂截止点 13.92 小时(论文 13.2 小时),两个截止点隐含的半衰期为 4.30 小时。第三章处理 CYP1A2 快慢代谢与多剂量叠加:在对数正态人群中,16 点一杯红茶越线的约 4.7 %,一杯咖啡为 0.694。第四章把浓度换算成腺苷受体占有率,并说明昼夜门控与效应饱和。 第五章用二分法反解最晚一杯:半衰期 5 小时时低于 34.5 毫克的单杯无需截止点,咖啡从覆盖一半顾客到覆盖九成,截止点要再提前 6.90 小时。第六章用三种损失模型夹出睡眠负债的区间。第七章考辨陆羽《茶经·六之饮》的碗数制度(每位客人 4.51 至 32.23 毫克)与巴尔扎克《论现代兴奋剂》的加码与耐受。第八章给出门店咖啡因标签、十四夜个人试验(单人只能检出约 22.5 分钟以上的差异,对 9.1 分钟的检验力仅 0.205)、预注册方案与推翻本书的条件,并附蒸馏小工具「咖啡因截止时刻计算器」。书中共含 22 段程序。 Chapter answers: - Ch.1 The Dose Gap: the Theanine in One Cup and Drake's Four Hundred Milligrams — Chapter 1 measures the dose gap: a cup of tea holds about 7.9 to 29.9 mg of theanine, reaching the research dose of 200 mg takes 8.26 cups of black tea and brings 310.7 mg of caffeine, and the much-quoted "six hours before bed still costs an hour" comes from a twelve-person trial of 400 mg with an objective loss of 1.1 to 1.2 hours. (https://5dlibrary.com/answers/en/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c01/) 第1章 剂量鸿沟:一杯茶的茶氨酸与 Drake 的四百毫克 — 第一章量出剂量鸿沟:一杯茶的茶氨酸约 7.9 至 29.9 毫克,要喝到研究用的 200 毫克需要 8.26 杯红茶,同时摄入 310.7 毫克咖啡因;常被引用的「睡前 6 小时仍少睡一小时」来自 400 毫克、12 人的试验,客观损失为 1.1 至 1.2 小时。 (https://5dlibrary.com/answers/zh/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c01/) - Ch.2 The Bateman Curve: One-Compartment Oral Absorption and One Calibrated Threshold — Chapter 2 uses the Bateman curve of one-compartment oral absorption and calibrates a night-exposure threshold of 0.6054 mg per litre on the coffee cut-off of the Gardiner meta-analysis; it predicts a pre-workout cut-off of 13.92 hours (13.2 in the paper), and the two published cut-offs imply a half-life of 4.30 hours. (https://5dlibrary.com/answers/en/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c02/) 第2章 贝特曼曲线:一室口服吸收与一个标定阈值 — 第二章用一室口服吸收的贝特曼曲线,以 Gardiner 元分析的咖啡截止点标定夜间暴露阈值每升 0.6054 毫克,它预测补剂截止点 13.92 小时(论文 13.2 小时),两个截止点隐含的半衰期为 4.30 小时。 (https://5dlibrary.com/answers/zh/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c02/) - Ch.3 Fast and Slow: the CYP1A2 Half-Life Distribution and Stacked Doses — Chapter 3 treats fast and slow CYP1A2 metabolism and stacked doses: in a lognormal population about 4.7 % cross the line after one black tea at four, against 0.694 after one coffee. (https://5dlibrary.com/answers/en/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c03/) 第3章 快慢代谢:CYP1A2 半衰期分布与多剂量叠加 — 第三章处理 CYP1A2 快慢代谢与多剂量叠加:在对数正态人群中,16 点一杯红茶越线的约 4.7 %,一杯咖啡为 0.694。 (https://5dlibrary.com/answers/zh/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c03/) - Ch.4 Adenosine Receptors: Occupancy, Masked Pressure and Saturation — With the human A1 measurement as anchor, a black tea peaks at about 6.90% occupancy and still holds 3.30% at eleven at night when drunk at four. A homeostat-only model inflates that occupancy into more than an hour of delayed sleep; with a circadian gate the delay falls to between 9.8 and 19.6 minutes. Drake's three timings lose the same sleep, and only a saturating model with a half-maximal concentration of 0.25 mg per litre fits all three, yet that model predicts that one black tea at four costs 42.8 minutes a night, in conflict with the meta-analysis. The conflict itself is the question this book hands to experiment. (https://5dlibrary.com/answers/en/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c04/) 第4章 腺苷受体:占有率、压力遮蔽与效应饱和 — 在 A1 受体的人体实测锚点下,一杯红茶的峰值占有率约 6.90%,下午四点喝到晚上十一点仍占 3.30%。只有稳态过程的模型会把这点占有率夸大成一个多小时的入睡延迟;加入昼夜门控后,延迟落在 9.8 至 19.6 分钟。Drake 三个时点损失相同,只有饱和模型(半数效应浓度每升 0.25 毫克)能同时吻合,而这个模型又预测一杯 16 点的红茶每晚损失 42.8 分钟,与元分析的结论冲突。冲突本身就是本书要交给实验的问题。 (https://5dlibrary.com/answers/zh/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c04/) - Ch.5 Solving for the Last Cup: Threshold, Quantiles and the Day's Budget — Chapter 5 solves for the last cup by bisection: at a half-life of 5 hours a single cup below 34.5 mg needs no cut-off, and moving coffee from covering half of customers to nine in ten brings its cut-off 6.90 hours earlier. (https://5dlibrary.com/answers/en/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c05/) 第5章 反解最晚一杯:阈值、分位数与当日预算 — 第五章用二分法反解最晚一杯:半衰期 5 小时时低于 34.5 毫克的单杯无需截止点,咖啡从覆盖一半顾客到覆盖九成,截止点要再提前 6.90 小时。 (https://5dlibrary.com/answers/zh/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c05/) - Ch.6 A Sleep-Debt Ledger: an Interval Bracketed by Three Loss Models — One black tea at four costs 0.0, 22.8 and 42.8 minutes a night under the three loss models, and the fourteen-day net debt runs from zero to 141.7 minutes. With three tenths repaid each night, the debt reaches nine tenths of steady state by night 7. Under the middle model, switching from coffee at four to black tea at two saves 145 minutes a week, and a coffee at ten in the morning costs almost the same each week as a black tea at four. The width of the interval is precisely the reason the trial of Chapter 8 is needed. (https://5dlibrary.com/answers/en/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c06/) 第6章 睡眠负债账本:三种损失模型夹出的区间 — 一杯下午四点的红茶,在三种损失模型下的每晚损失是 0.0、22.8 与 42.8 分钟,十四天净负债从零到 141.7 分钟。负债按每晚偿还三成计算,第 7 晚即达稳态的九成。按中位模型,从下午四点的咖啡改为下午两点的红茶,每周省下 145 分钟;上午十点的一杯咖啡与下午四点的一杯红茶,每周损失几乎相同。区间之宽,正是需要第八章那种试验的理由。 (https://5dlibrary.com/answers/zh/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c06/) - Ch.7 Classics East and West: Lu Yu's Chapter on Drinking and Balzac's Treatise on Modern Stimulants — Chapter 7 examines the bowl regime of Lu Yu's Classic of Tea (4.51 to 32.23 mg per guest) and Balzac's escalation and tolerance in his Treatise on Modern Stimulants. (https://5dlibrary.com/answers/en/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c07/) 第7章 东西方原典考辨:陆羽《茶经·六之饮》与巴尔扎克《论现代兴奋剂》 — 第七章考辨陆羽《茶经·六之饮》的碗数制度(每位客人 4.51 至 32.23 毫克)与巴尔扎克《论现代兴奋剂》的加码与耐受。 (https://5dlibrary.com/answers/zh/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c07/) - Ch.8 Closing the Loop: Menu Labels, a Fourteen-Night A/B Test and Pre-Registered Refutation — Chapter 8 gives caffeine labels for a shop, a fourteen-night personal trial (one person can detect only differences above about 22.5 minutes, with a power of just 0.205 for 9.1 minutes), a pre-registered protocol, the conditions that would refute the book, and the distilled tool, the Caffeine Cutoff Calculator. The book contains 22 programs. (https://5dlibrary.com/answers/en/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c08/) 第8章 实证闭环:菜单标签、十四夜 A/B 与预注册证伪 — 第八章给出门店咖啡因标签、十四夜个人试验(单人只能检出约 22.5 分钟以上的差异,对 9.1 分钟的检验力仅 0.205)、预注册方案与推翻本书的条件,并附蒸馏小工具「咖啡因截止时刻计算器」。书中共含 22 段程序。 (https://5dlibrary.com/answers/zh/caffeine-theanine-pharmacokinetics-afternoon-tea-sleep-debt/c08/) ## 5DL-B-187 · The Mortality Dose-Response of VO2max, Zone-2 Lactate Thresholds and the Metabolic Equivalents of the Five-Animal Frolics 中文书名:最大摄氧量死亡率剂量响应、二区有氧乳酸阈与五禽戏间歇代谢当量母典 DOI: pending Cite: Fan, Y. (2026). The Mortality Dose-Response of VO2max, Zone-2 Lactate Thresholds and the Metabolic Equivalents of the Five-Animal Frolics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 6 assigns illustrative metabolic equivalents to the Five-Animal Frolics, about 43.75 MET-minutes a round, and schedules a weekly dose. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: This book tests two popular pieces of common sense: that poor fitness is merely being out of shape and not a risk on a par with smoking or diabetes, and that the goal of exercise is ten thousand steps a day, the more the better. It writes each claim as a model the data are allowed to reject and tests it against published studies. Parameters given without a source are illustrative; all calculations are generated by the book's programs; the book is not medical advice, claims no effect of any activity on any disease, and anyone planning a self-experiment should first consult a licensed physician. Chapter 1 rereads the Cleveland Clinic treadmill cohort: the least fit group carries a hazard ratio of 5.04 against the elite and smoking 1.41, but the former is a contrast of extreme groups; per standard deviation, fitness carries 1.559 and smoking 1.131, so fitness still leads, by far less than the slogan claims. Chapter 2 traces the origin of ten thousand steps, places the step plateau near 6500 with a fall-then-flat spline, and rejects the purely linear claim of no plateau. Chapter 3 fits a piecewise log-linear curve with two knots to pooled data on two thirds of a million people, with a hazard ratio of 0.9431 per MET-hour on the first segment. Chapter 4 shows that the edge of zone 2 is a cut drawn by people on a smooth lactate curve. Chapter 5 shows that single tests inflate the share of training non-responders from 0.107 to 0.196. Chapter 6 assigns illustrative metabolic equivalents to the Five-Animal Frolics, about 43.75 MET-minutes a round, and schedules a weekly dose. Chapter 7 examines Hua Tuo's frolic of five animals in the Hou Han shu and Galen's On Exercise with the Small Ball, and shows that among 100 practitioners, even with no effect, the chance of one ninety-year-old is 0.4917. Chapter 8 gives a fitness plan card for one-person companies, an eight-week curve that admits a single watch estimate error of about 5.02, a distilled tool called the Fitness Grade Calculator and a pre-registered plan for falsification. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:本书检验两条流行的常识:体能差只是「状态不好」,算不上与吸烟、糖尿病同级的风险;运动的目标是每天一万步,越多越好。书中把每一条说法写成可以被数据拒绝的模型,再用已发表的研究检验。书中参数凡未注明出处者均为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何活动对任何疾病有作用,自我试验须先咨询执业医师。 第一章重读克利夫兰诊所的跑台队列:最低体能组对精英组的风险比为 5.04,吸烟为 1.41,但前者是极值组对比;换到每标准差尺度,体能为 1.559,吸烟为 1.131,体能仍然领先,只是远没有口号说的那么多。第二章追溯一万步的出身,用先降后平的样条把步数平台放在约 6500 步,并拒绝「没有平台」的纯线性说法。第三章用六十六万人的合并数据拟合两个拐点的分段对数线性曲线,第一段每 MET 小时风险比为 0.9431。第四章说明二区边界是在光滑乳酸曲线上的人为切口。 第五章说明单次测试会把训练无反应者的比例从 0.107 夸大到 0.196。第六章为五禽戏估出示意的代谢当量,一轮约 43.75 MET 分钟,并排出周剂量。第七章考辨《后汉书》所记华佗五禽之戏与盖伦《论小球运动》,并说明在 100 位练习者中,即使毫无效应,出现一位九十岁者的概率也有 0.4917。第八章给出一人公司体能计划卡、手表单次估值误差约 5.02 的八周曲线、名为「体能分级计算器」的蒸馏小工具与预注册的证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Five Grades on the Treadmill: Why the Least Fit Group Outranks Smoking — Chapter 1 rereads the Cleveland Clinic treadmill cohort: the least fit group carries a hazard ratio of 5.04 against the elite and smoking 1.41, but the former is a contrast of extreme groups; per standard deviation, fitness carries 1.559 and smoking 1.131, so fitness still leads, by far less than the slogan claims. (https://5dlibrary.com/answers/en/vo2max-mortality-dose-response-zone2-five-animal-frolics/c01/) 第1章 跑台上的五档:最低体能组的死亡风险为何压过吸烟 — 第一章重读克利夫兰诊所的跑台队列:最低体能组对精英组的风险比为 5.04,吸烟为 1.41,但前者是极值组对比;换到每标准差尺度,体能为 1.559,吸烟为 1.131,体能仍然领先,只是远没有口号说的那么多。 (https://5dlibrary.com/answers/zh/vo2max-mortality-dose-response-zone2-five-animal-frolics/c01/) - Ch.2 Where Ten Thousand Steps Came From: a Step-Risk Spline and Age-Specific Plateaus — Chapter 2 traces the origin of ten thousand steps, places the step plateau near 6500 with a fall-then-flat spline, and rejects the purely linear claim of no plateau. (https://5dlibrary.com/answers/en/vo2max-mortality-dose-response-zone2-five-animal-frolics/c02/) 第2章 一万步的出身:步数—风险样条与年龄平台 — 第二章追溯一万步的出身,用先降后平的样条把步数平台放在约 6500 步,并拒绝「没有平台」的纯线性说法。 (https://5dlibrary.com/answers/zh/vo2max-mortality-dose-response-zone2-five-animal-frolics/c02/) - Ch.3 Piecewise Log-Linear MET-Minutes: Knots and Ceilings in the Dose-Response — Chapter 3 fits a piecewise log-linear curve with two knots to pooled data on two thirds of a million people, with a hazard ratio of 0.9431 per MET-hour on the first segment. (https://5dlibrary.com/answers/en/vo2max-mortality-dose-response-zone2-five-animal-frolics/c03/) 第3章 MET 分钟的分段对数线性:剂量—反应的拐点与天花板 — 第三章用六十六万人的合并数据拟合两个拐点的分段对数线性曲线,第一段每 MET 小时风险比为 0.9431。 (https://5dlibrary.com/answers/zh/vo2max-mortality-dose-response-zone2-five-animal-frolics/c03/) - Ch.4 Zone 2 and the Lactate Threshold: One Curve, Five Ways to Cut It — For the less fit cyclist, lactate at the first stage of 50 W is 1.16. The three first-rise rules give 102.0 W, 126.5 W and 120.0 W; the lowest and highest differ by 24.5 W, or 14.4 % of the power at 4 mmol/L. For the trained cyclist, lactate at the first stage of 100 W is 0.90, and the three rules give 207.9, 244.9 and 250.0 W, a spread of 42.1 W, or 13.6 % of the 4-mmol power. Dmax falls between the two families: 143.1 W for the less fit rider (lactate 2.68) and 262.8 W for the trained one (lactate 2.28). (https://5dlibrary.com/answers/en/vo2max-mortality-dose-response-zone2-five-animal-frolics/c04/) 第4章 二区与乳酸阈:同一条曲线,五种切法 — 对体能较低的骑行者,第一级负荷 50 瓦时乳酸为 1.16。三种对应「第一次上升」的规则分别给出 102.0 瓦、126.5 瓦与 120.0 瓦,最低与最高相差 24.5 瓦,相当于 4 毫摩尔功率的 14.4 %。对训练有素者,第一级负荷 100 瓦时乳酸为 0.90,三种规则给出 207.9、244.9 与 250.0 瓦,相差 42.1 瓦,即 4 毫摩尔功率的 13.6 %。Dmax 法则落在两类之间:体能较低者 143.1 瓦(乳酸 2.68),训练有素者 262.8 瓦(乳酸 2.28)。 (https://5dlibrary.com/answers/zh/vo2max-mortality-dose-response-zone2-five-animal-frolics/c04/) - Ch.5 Trainability and the Slope of Ageing: Individual Differences and Measurement Noise in VO2max — Chapter 5 shows that single tests inflate the share of training non-responders from 0.107 to 0.196. (https://5dlibrary.com/answers/en/vo2max-mortality-dose-response-zone2-five-animal-frolics/c05/) 第5章 可训练性与衰老斜率:最大摄氧量的个体差异与测量噪声 — 第五章说明单次测试会把训练无反应者的比例从 0.107 夸大到 0.196。 (https://5dlibrary.com/answers/zh/vo2max-mortality-dose-response-zone2-five-animal-frolics/c05/) - Ch.6 The Metabolic Equivalents of the Five-Animal Frolics: Interval Load Integrals and Weekly Dose Scheduling — Chapter 6 assigns illustrative metabolic equivalents to the Five-Animal Frolics, about 43.75 MET-minutes a round, and schedules a weekly dose. (https://5dlibrary.com/answers/en/vo2max-mortality-dose-response-zone2-five-animal-frolics/c06/) 第6章 五禽戏的代谢当量:间歇负荷积分与周剂量排程 — 第六章为五禽戏估出示意的代谢当量,一轮约 43.75 MET 分钟,并排出周剂量。 (https://5dlibrary.com/answers/zh/vo2max-mortality-dose-response-zone2-five-animal-frolics/c06/) - Ch.7 Hua Tuo and Galen: the Five-Animal Frolics and the Small Ball Examined — Chapter 7 examines Hua Tuo's frolic of five animals in the Hou Han shu and Galen's On Exercise with the Small Ball, and shows that among 100 practitioners, even with no effect, the chance of one ninety-year-old is 0.4917. (https://5dlibrary.com/answers/en/vo2max-mortality-dose-response-zone2-five-animal-frolics/c07/) 第7章 华佗与盖伦:五禽之戏与小球运动的原典考辨 — 第七章考辨《后汉书》所记华佗五禽之戏与盖伦《论小球运动》,并说明在 100 位练习者中,即使毫无效应,出现一位九十岁者的概率也有 0.4917。 (https://5dlibrary.com/answers/zh/vo2max-mortality-dose-response-zone2-five-animal-frolics/c07/) - Ch.8 Closing the Loop: a Fitness Plan Card for One-Person Companies, an Eight-Week Wearable Curve and a Plan for Falsification — Chapter 8 gives a fitness plan card for one-person companies, an eight-week curve that admits a single watch estimate error of about 5.02, a distilled tool called the Fitness Grade Calculator and a pre-registered plan for falsification. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/vo2max-mortality-dose-response-zone2-five-animal-frolics/c08/) 第8章 实证闭环:一人公司体能计划卡、八周可穿戴曲线与证伪方案 — 第八章给出一人公司体能计划卡、手表单次估值误差约 5.02 的八周曲线、名为「体能分级计算器」的蒸馏小工具与预注册的证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/vo2max-mortality-dose-response-zone2-five-animal-frolics/c08/) ## 5DL-B-188 · Leucine Thresholds for Muscle Protein Synthesis, the Per-Meal Protein Cap Myth and the Resistance-Training Dose against Sarcopenia 中文书名:肌蛋白合成亮氨酸阈值、单餐蛋白吸收上限神话与肌少症抗阻剂量母典 DOI: pending Cite: Fan, Y. (2026). Leucine Thresholds for Muscle Protein Synthesis, the Per-Meal Protein Cap Myth and the Resistance-Training Dose against Sarcopenia (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not medical advice and promises nothing about any disease or individual health outcome. Chapter 6 sets strength reserve against the grip thresholds of EWGSOP2: in a simulated population 0.658 fall below the threshold by 85 with no training and 0.301 with full adherence, and the daily-total plateau lies near 1.62 g per kilogram. Abstract: "One meal can only absorb 30 g of protein" is one of the most widely repeated claims in fitness and nutrition. This book rewrites it as a model that can be computed and sets it against published experiments: the 2023 study of Trommelen and colleagues showed that after exercise a single 100 g dose of milk protein gave a larger and longer myofibrillar synthetic response than 25 g. The book also tests a second claim, that training twice a week with a few sets close to failure collects more than half the gain on the dose-response curve. Its parameters are illustrative; all calculations are generated by its programs; it is not medical advice and promises nothing about any disease or individual health outcome. Chapter 1 calibrates a Bateman curve on the paper's figures and lets the cap make its own prediction: the open model gives a 12-hour synthesis ratio of 1.322, the capped model only 1.057, outside the interval of 1.095 to 1.624 derived from the paper's approximate figures. Chapter 2 shows that absorption speed changes shape but not area: for the same 25 g, four sources share a total area of 29.17 g·h, and 53.0 g of a 100 g dose enter the circulation within 12 hours. Chapter 3 shows how a four-hour window manufactures a "20 g plateau": 20 g reaches 0.775 of the 40 g response, while over 12 hours 100 g reaches 1.331; the older half-saturation constant is about 1.47 times the young one, and two papers are shown to be in tension. Chapter 4 computes a day's anabolic area: four meals of 25 g beat one of 100 g by a ratio of 1.090. Chapter 5 uses a logarithmic response to show that 6 weekly sets collect 0.511 to 0.639 of the gain of 20, and compares two rules for training close to failure. Chapter 6 sets strength reserve against the grip thresholds of EWGSOP2: in a simulated population 0.658 fall below the threshold by 85 with no training and 0.301 with full adherence, and the daily-total plateau lies near 1.62 g per kilogram. Chapter 7 examines the advice of the Yinshan zhengyao to eat "often and little" and Quintilian's story of Milo and the calf. Chapter 8 gives a twelve-week closed loop for a gym's strength camp, the distilled protein distribution calculator, a pre-registered protocol and the conditions that would overturn the book. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:「一餐只能吸收 30 克蛋白」是健身与营养领域流传最广的说法之一。本书把它重写成可以计算的模型,并与已发表的实验对照:特罗梅伦等人 2023 年的实验显示,训练后一次 100 克乳蛋白比 25 克带来更大、更久的肌原纤维合成反应。本书同时检验第二条主张:每周两次、每次少组数、接近力竭的训练能拿到剂量反应曲线上过半的增益。书中参数为示意;全部计算由书中程序生成;本书不是医疗建议,不对任何疾病或个人健康结果作出承诺。 第一章以 Bateman 曲线校准原文数字,让「上限」说法自己给出预测:开放模型的 12 小时合成比值为 1.322,封顶模型只有 1.057,后者落在由原文近似值推出的区间 1.095 至 1.624 之外。第二章说明吸收速度只改变形状、不改变面积:同样 25 克,四种来源的总面积都是 29.17 克·小时;100 克在 12 小时内进入循环 53.0 克。第三章说明四小时窗口会制造「20 克平台」:20 克达到 40 克反应的 0.775,而 12 小时窗口里 100 克达到 1.331;老年人的半饱和常数约为年轻人的 1.47 倍,并揭示两份文献之间的张力。第四章计算一天的合成面积:100 克分四餐比一次吃完的面积比为 1.090。 第五章用对数响应说明每周 6 组拿到 20 组增益的 0.511 至 0.639,并比较两种接近力竭规则。第六章把力量储备放到 EWGSOP2 的握力阈值上:模拟人群中无人训练时 85 岁有 0.658 跌破阈值,全部坚持训练时为 0.301;每日总量的平台约在每千克 1.62 克。第七章考辨《饮膳正要》「食欲数而少,不欲顿而多」与昆体良笔下米洛扛牛犊的故事。第八章给出门店力量训练营的十二周闭环、蒸馏小工具「蛋白分配计算器」、预注册方案与推翻本书的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(共 18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Thirty-Gram Cap: a Number that Confuses Absorption with Use — Chapter 1 calibrates a Bateman curve on the paper's figures and lets the cap make its own prediction: the open model gives a 12-hour synthesis ratio of 1.322, the capped model only 1.057, outside the interval of 1.095 to 1.624 derived from the paper's approximate figures. (https://5dlibrary.com/answers/en/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c01/) 第1章 三十克上限:一个把吸收与利用混为一谈的数字 — 第一章以 Bateman 曲线校准原文数字,让「上限」说法自己给出预测:开放模型的 12 小时合成比值为 1.322,封顶模型只有 1.057,后者落在由原文近似值推出的区间 1.095 至 1.624 之外。 (https://5dlibrary.com/answers/zh/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c01/) - Ch.2 The Bateman Curve: Absorption Speed Changes Shape, Not Area — Chapter 2 shows that absorption speed changes shape but not area: for the same 25 g, four sources share a total area of 29.17 g·h, and 53.0 g of a 100 g dose enter the circulation within 12 hours. (https://5dlibrary.com/answers/en/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c02/) 第2章 Bateman 曲线:吸收速度改变形状,不改变面积 — 第二章说明吸收速度只改变形状、不改变面积:同样 25 克,四种来源的总面积都是 29.17 克·小时;100 克在 12 小时内进入循环 53.0 克。 (https://5dlibrary.com/answers/zh/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c02/) - Ch.3 Saturation and Threshold: Michaelis-Menten, Hill and the Higher Leucine Threshold of Older Muscle — Chapter 3 shows how a four-hour window manufactures a "20 g plateau": 20 g reaches 0.775 of the 40 g response, while over 12 hours 100 g reaches 1.331; the older half-saturation constant is about 1.47 times the young one, and two papers are shown to be in tension. (https://5dlibrary.com/answers/en/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c03/) 第3章 饱和与阈值:米氏曲线、Hill 曲线与老年人更高的亮氨酸门槛 — 第三章说明四小时窗口会制造「20 克平台」:20 克达到 40 克反应的 0.775,而 12 小时窗口里 100 克达到 1.331;老年人的半饱和常数约为年轻人的 1.47 倍,并揭示两份文献之间的张力。 (https://5dlibrary.com/answers/zh/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c03/) - Ch.4 The Duration Integral: Large Doses Stretch the Area, Splitting Gains a Few Per Cent — Chapter 4 computes a day's anabolic area: four meals of 25 g beat one of 100 g by a ratio of 1.090. (https://5dlibrary.com/answers/en/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c04/) 第4章 持续时间积分:大剂量拉长面积,分餐只差几个百分点 — 第四章计算一天的合成面积:100 克分四餐比一次吃完的面积比为 1.090。 (https://5dlibrary.com/answers/zh/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c04/) - Ch.5 A Logarithmic Response to Training Volume: Twice a Week, Few Sets, Close to Failure — Chapter 5 uses a logarithmic response to show that 6 weekly sets collect 0.511 to 0.639 of the gain of 20, and compares two rules for training close to failure. (https://5dlibrary.com/answers/en/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c05/) 第5章 训练量的对数响应:每周两次、少组数、接近力竭 — 第五章用对数响应说明每周 6 组拿到 20 组增益的 0.511 至 0.639,并比较两种接近力竭规则。 (https://5dlibrary.com/answers/zh/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c05/) - Ch.6 The Reserve Arithmetic of Sarcopenia: Grip Thresholds, Rates of Loss and the Daily Total — Chapter 6 sets strength reserve against the grip thresholds of EWGSOP2: in a simulated population 0.658 fall below the threshold by 85 with no training and 0.301 with full adherence, and the daily-total plateau lies near 1.62 g per kilogram. (https://5dlibrary.com/answers/en/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c06/) 第6章 肌少症的储备算术:握力阈值、流失速率与每日总量 — 第六章把力量储备放到 EWGSOP2 的握力阈值上:模拟人群中无人训练时 85 岁有 0.658 跌破阈值,全部坚持训练时为 0.301;每日总量的平台约在每千克 1.62 克。 (https://5dlibrary.com/answers/zh/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c06/) - Ch.7 Classical Sources East and West: "Often and Little" in the Yinshan zhengyao and Milo's Calf — Hu Sihui's "eat often and little, not all at once and much" points the same way as modern experiments on protein distribution, but its benefit is a few per cent to about a tenth, and "little" must not fall below the leucine threshold. The mutton-heavy recipes of the Yinshan zhengyao suggest that a single meal at court carried far more than "thirty grams" of protein. Quintilian's story of Milo gets the principle of progressive loading right and gets wrong the idea that load can rise linearly without limit: with illustrative parameters, the calf outweighs the carrier's capacity by day 233. The next chapter turns the models of this book into a testing scheme that a gym can carry out, and that could overturn the book. (https://5dlibrary.com/answers/en/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c07/) 第7章 东西方原典考辨:《饮膳正要》的「数而少」与米洛的牛犊 — 忽思慧「食欲数而少,不欲顿而多」与现代关于蛋白分配的实验方向一致,但收益只有百分之几到一成,并且「少」不能低于亮氨酸门槛;《饮膳正要》以羊为主料的菜谱暗示宫廷单餐蛋白远超「三十克」。昆体良的米洛故事说对了渐进负荷的原则,说错了负荷可以无限线性增加:按示意参数,牛犊在第 233 天就重于扛牛者的能力。下一章把全书的模型变成一套门店可以执行、也可以被推翻的检验方案。 (https://5dlibrary.com/answers/zh/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c07/) - Ch.8 Closing the Loop: a Strength Camp, a Twelve-Week Log and a Pre-Registered Test — Chapter 8 gives a twelve-week closed loop for a gym's strength camp, the distilled protein distribution calculator, a pre-registered protocol and the conditions that would overturn the book. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c08/) 第8章 实证闭环:力量训练营、十二周日志与预注册检验 — 第八章给出门店力量训练营的十二周闭环、蒸馏小工具「蛋白分配计算器」、预注册方案与推翻本书的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(共 18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/leucine-threshold-protein-cap-myth-sarcopenia-resistance-dose/c08/) ## 5DL-B-189 · Outdoor Illuminance Dose, Accommodative Lag and Peripheral-Defocus Optics in Refractive Development 中文书名:户外照度剂量近视抑制、调节滞后与周边离焦光学的屈光发育母典 DOI: pending Cite: Fan, Y. (2026). Outdoor Illuminance Dose, Accommodative Lag and Peripheral-Defocus Optics in Refractive Development (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable computations (20 programs), classical sources and the author's hypotheses. Abstract: This book tests a popular claim: children become short-sighted because they read too much, and looking far away relaxes the eyes. It states and quantifies a falsifiable counter-claim: as an explanation of who becomes myopic, near work is weaker than a shortfall of time outdoors, and the outdoors works mainly through a dose of illuminance multiplied by time, not through looking far. Its children, timetables and luminaire parameters are illustrative; all calculations are generated by its programs; the book is not medical advice, promises no effect for any individual, and any self-experiment should first be discussed with a licensed physician. Chapter 1 recomputes the Guangzhou cluster randomised trial from its raw counts: three-year new myopia of 0.3036 against 0.3953, a risk ratio of 0.768; if the outdoor class worked only by cutting reading, it would have to remove 81.0 minutes of reading a day, 2.02 times its own length. Chapter 2 uses lux-minutes as its scale: a full day at a classroom desk gives 0.180 of the benchmark, while 40 minutes in tree shade supply 0.333; in the Taiwanese trial, raising the threshold from 1000 to 3000 lux raised the benefit only 1.14 times. Chapter 3 writes axial growth as a linear mixed model; in simulation the longitudinal estimate detects the light effect every time, the cross-sectional one with a chance of only 0.110. Chapter 4 shows that accommodative lag rises only after myopia begins, and that add lenses saved just 0.135 of progression over three years. Chapter 5 computes peripheral defocus with thin lenses and the reduced eye; explaining defocus spectacles through average peripheral hyperopia would need a change of 9.2 D. Chapter 6 shows that observational data cannot identify the mechanism whereas a four-arm trial can, and that population arithmetic is most sensitive to dose coverage. Chapter 7 weighs Fu Renyu's Shenshi Yaohan and Kepler's Optical Part of Astronomy; reading by candlelight gives only 11.1 lux. Chapter 8 offers light logs, a lux planning sheet for shops, the distilled Illuminance Dose Logger and a pre-registered falsification protocol. Marked boxes separate published evidence, runnable computations (20 programs), classical sources and the author's hypotheses. 摘要:本书检验一个流行说法:孩子近视是因为书看得太多,看远处能让眼睛放松。本书提出并量化一个可以被证伪的反命题:在解释近视发生上,近距离用眼的解释力弱于户外时间不足,而户外起作用的主要是照度与时间相乘的剂量,不是远眺。书中儿童、日程与灯具参数为示意;全部计算由书中程序生成;本书不是医疗建议,不对任何个人作效果承诺,自我试验须先咨询执业医师。 第一章用广州整群随机试验的原始计数重算:三年新发近视 0.3036 对 0.3953,风险比 0.768;若户外课只靠减少看书起作用,它每天须减掉 81.0 分钟阅读,是课时的 2.02 倍。第二章以勒克斯·分钟为尺度,课桌前坐满一天只得基准的 0.180,树荫下 40 分钟补上 0.333;台湾试验中门槛由 1000 lux 提到 3000 lux,收益只增 1.14 倍。第三章写出眼轴生长的线性混合模型,模拟中纵向估计每次都能检出照度效应,横断面估计只有 0.110 的机会。第四章显示调节滞后在近视发生之后才升高,加光镜片三年只节省进展的 0.135。 第五章用薄透镜与简化眼计算周边离焦,若以平均周边远视解释离焦镜片的效应,需要 9.2 D 的改变。第六章说明观察数据无法辨识机制,而四臂试验可以;人群账对剂量覆盖率最敏感。第七章考辨《审视瑶函》与开普勒《天文光学》,烛下读书只有 11.1 lux。第八章给出照度日志、门店 lux 规划表、蒸馏小工具「照度剂量记录器」与预先登记的证伪方案。书中以标注框区分已发表证据、可运行计算(20 段程序)、原典与作者假说。 Chapter answers: - Ch.1 Forty Minutes of Evidence: the Guangzhou Cluster Trial and the Weak Reach of Near Work — Chapter 1 recomputes the Guangzhou cluster randomised trial from its raw counts: three-year new myopia of 0.3036 against 0.3953, a risk ratio of 0.768; if the outdoor class worked only by cutting reading, it would have to remove 81.0 minutes of reading a day, 2.02 times its own length. (https://5dlibrary.com/answers/en/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c01/) 第1章 四十分钟的证据:广州整群试验与近距离用眼的解释力 — 第一章用广州整群随机试验的原始计数重算:三年新发近视 0.3036 对 0.3953,风险比 0.768;若户外课只靠减少看书起作用,它每天须减掉 81.0 分钟阅读,是课时的 2.02 倍。 (https://5dlibrary.com/answers/zh/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c01/) - Ch.2 Illuminance, Not Distance: Lux-Minute Dose and Where Two Models Part — Chapter 2 uses lux-minutes as its scale: a full day at a classroom desk gives 0.180 of the benchmark, while 40 minutes in tree shade supply 0.333; in the Taiwanese trial, raising the threshold from 1000 to 3000 lux raised the benefit only 1.14 times. (https://5dlibrary.com/answers/en/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c02/) 第2章 照度而非远眺:勒克斯·分钟剂量与两种模型的分歧 — 第二章以勒克斯·分钟为尺度,课桌前坐满一天只得基准的 0.180,树荫下 40 分钟补上 0.333;台湾试验中门槛由 1000 lux 提到 3000 lux,收益只增 1.14 倍。 (https://5dlibrary.com/answers/zh/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c02/) - Ch.3 Axial Length in a Linear Mixed Model: How Dose Enters the Growth Slope — Chapter 3 writes axial growth as a linear mixed model; in simulation the longitudinal estimate detects the light effect every time, the cross-sectional one with a chance of only 0.110. (https://5dlibrary.com/answers/en/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c03/) 第3章 眼轴的线性混合模型:剂量如何进入生长斜率 — 第三章写出眼轴生长的线性混合模型,模拟中纵向估计每次都能检出照度效应,横断面估计只有 0.110 的机会。 (https://5dlibrary.com/answers/zh/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c03/) - Ch.4 Accommodative Lag: A Cause Overturned by the Order of Events — Chapter 4 shows that accommodative lag rises only after myopia begins, and that add lenses saved just 0.135 of progression over three years. (https://5dlibrary.com/answers/en/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c04/) 第4章 调节滞后:一个被时间顺序推翻的原因 — 第四章显示调节滞后在近视发生之后才升高,加光镜片三年只节省进展的 0.135。 (https://5dlibrary.com/answers/zh/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c04/) - Ch.5 The Thin-Lens Optics of Peripheral Defocus — Chapter 5 computes peripheral defocus with thin lenses and the reduced eye; explaining defocus spectacles through average peripheral hyperopia would need a change of 9.2 D. (https://5dlibrary.com/answers/en/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c05/) 第5章 周边离焦的薄透镜光学 — 第五章用薄透镜与简化眼计算周边离焦,若以平均周边远视解释离焦镜片的效应,需要 9.2 D 的改变。 (https://5dlibrary.com/answers/zh/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c05/) - Ch.6 Three Mechanisms in Competition: Identifiability, Sensitivity and Population Arithmetic — When bright, far and open always come together, observational data can confirm that the outdoors helps but cannot say which part helps; the simulation shows that in the observed world the wrong predictor still explains a respectable share of variance, whereas the four-arm design leaves fingerprints that can hardly be confused. The population arithmetic is most sensitive to coverage: at the Guangzhou effect, between 20.8 and 91.7 cases are averted per thousand children over three years, depending on how much of the dose is really delivered, and a single intervention cannot push incidence in a high-prevalence region below 0.30. The next chapter returns to two classical texts to see what the ancients got right about the eye and light. (https://5dlibrary.com/answers/en/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c06/) 第6章 三种机制的竞争:可辨识性、敏感度与人群账 — 当亮、远、开阔总是同时出现时,观察数据只能确认户外有用,不能指出哪一部分有用;模拟显示,在观察世界里错误的预测量也能解释可观的方差,而在四臂设计中,两种真相留下几乎不可混淆的指纹。人群账对覆盖率最敏感:按广州效应,每千名儿童三年少发 20.8 到 91.7 例,取决于实际剂量的覆盖;单一干预无法把高发地区的发病率压到 0.30 以下。下一章回到两部原典,看古人对眼与光说对了什么。 (https://5dlibrary.com/answers/zh/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c06/) - Ch.7 Fu Renyu and Kepler: Testable Sentences in Two Classics — Chapter 7 weighs Fu Renyu's Shenshi Yaohan and Kepler's Optical Part of Astronomy; reading by candlelight gives only 11.1 lux. (https://5dlibrary.com/answers/en/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c07/) 第7章 《审视瑶函》与开普勒:两部原典的可检验句 — 第七章考辨《审视瑶函》与开普勒《天文光学》,烛下读书只有 11.1 lux。 (https://5dlibrary.com/answers/zh/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c07/) - Ch.8 Light Logs, Shop Lux Plans and a Falsification Protocol — Chapter 8 offers light logs, a lux planning sheet for shops, the distilled Illuminance Dose Logger and a pre-registered falsification protocol. Marked boxes separate published evidence, runnable computations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c08/) 第8章 照度日志、门店 lux 规划与证伪方案 — 第八章给出照度日志、门店 lux 规划表、蒸馏小工具「照度剂量记录器」与预先登记的证伪方案。书中以标注框区分已发表证据、可运行计算(20 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/outdoor-illuminance-myopia-accommodative-lag-peripheral-defocus/c08/) ## 5DL-B-190 · Exponential Decay of SPF under Thin Application, UVA Photo-Ageing and the Vitamin D Synthesis Paradox 中文书名:防晒系数涂抹量指数衰减、长波紫外光老化与维生素D合成悖论母典 DOI: pending Cite: Fan, Y. (2026). Exponential Decay of SPF under Thin Application, UVA Photo-Ageing and the Vitamin D Synthesis Paradox (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its spectra, film thicknesses, minimal erythema doses and coverage shares are illustrative; all calculations are generated by its programs; it is not medical advice and promises no outcome for skin or health. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: The SPF on a bottle is measured at 2 milligrams per square centimetre, while people usually apply about a quarter of that. This book tests a counterintuitive claim: at that common amount an SPF50 product may deliver an SPF in single digits, and the vitamin D of regular sunscreen users usually does not fall much precisely because, in real life, the film is always too thin. The book writes the claim as runnable arithmetic and corrects the wording of its original plan against the primary literature. Its spectra, film thicknesses, minimal erythema doses and coverage shares are illustrative; all calculations are generated by its programs; it is not medical advice and promises no outcome for skin or health. Chapter 1 uses the Beer-Lambert power law to show that SPF50 at a quarter of the test amount keeps only 2.66, while the linear law reported by a multicentre study gives 12.50. Chapter 2 uses a spectral model to show when each law holds; a formulation with a UVA gap gives 7.86, and the model average is a single digit whenever the power law's weight is at least 0.144. Chapter 3 builds a dose ledger from UV index, time and minimal erythema dose; a clear mid-July day in Taipei delivers about 71.2 standard erythema doses. Chapter 4 shows that UVA carries 0.979 of ground-level ultraviolet energy yet barely enters the SPF, and that the Nambour photo-ageing trial had a power of only 0.56. Chapter 5 resolves the vitamin D paradox: the vitamin D effective dose of the field median pattern is 25.0 times that of the laboratory pattern, and the 2025 Sun-D trial found a between-group difference of 5.2 nmol/L for routine high-SPF use, a fall that is real but limited. Chapter 6 breaks the paradox into formulation, grams and coverage; at the same SPF15, a high-UVA formulation leaves vitamin D 4.63 times the share left by a low-UVA one. Chapter 7 examines the face preparations in Sun Simiao's Beiji qianjin yaofang and Niels Finsen's work on light. Chapter 8 gives a shelf usage card, a Taipei and Kuala Lumpur travel dose card, the distilled Sunscreen Dose Calculator and a pre-registered 40-person plan that refutes the claim with probability 0.931 if the linear law is true. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:瓶子上的 SPF 是按每平方厘米 2 毫克测出来的,而人们平时只涂它的四分之一左右。本书检验一个反常识命题:按这种常见用量涂抹 SPF50,实际防护可能只剩个位数;规律使用防晒者的维生素 D 之所以通常没有大幅下降,正是因为现实里总是涂不够。书中把命题写成可以运行的算术,并按原始文献修正了立项时的说法。书中的光谱、膜厚、最小红斑量与覆盖率均为示意参数;全部计算由书中程序生成;本书不是医疗建议,不对任何皮肤或健康结果作出承诺。 第一章用比尔—朗伯幂律说明 SPF50 在四分之一用量下只剩 2.66,而多中心研究报告的线性律给出 12.50。第二章用光谱模型说明两条律各自成立的条件,长波有漏洞的配方给出 7.86;只要给幂律的权重不低于 0.144,模型平均就是个位数。第三章建立紫外指数、时间与最小红斑量的剂量账本,台北七月中旬晴天一日约 71.2 个标准红斑量。第四章说明长波紫外占地面紫外能量的 0.979 却几乎不进入 SPF,南布尔光老化试验的把握度只有 0.56。 第五章解开维生素 D 悖论:现场中位情形的维生素 D 有效剂量是实验室情形的 25.0 倍;2025 年的 Sun-D 试验显示常规使用高 SPF 产品的组间差为 5.2 nmol/L,下降真实而有限。第六章把悖论拆成配方、克数与覆盖率,同为 SPF15 时高长波防护配方留给维生素 D 的份额是低长波防护配方的 4.63 倍。第七章考辨孙思邈《备急千金要方》的面药与芬森的光照研究。第八章给出货架用量说明卡、台北与吉隆坡行程剂量卡、蒸馏小工具“防晒剂量计算器”和 40 人的预注册方案,在线性律为真时该方案推翻命题的概率为 0.931。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Power Law of Application: Why a Quarter Dose Leaves a Single-Digit SPF — Chapter 1 uses the Beer-Lambert power law to show that SPF50 at a quarter of the test amount keeps only 2.66, while the linear law reported by a multicentre study gives 12.50. (https://5dlibrary.com/answers/en/spf-under-application-uva-photoageing-vitamin-d-paradox/c01/) 第1章 涂抹量的幂律:四分之一用量为何只剩个位数 SPF — 第一章用比尔—朗伯幂律说明 SPF50 在四分之一用量下只剩 2.66,而多中心研究报告的线性律给出 12.50。 (https://5dlibrary.com/answers/zh/spf-under-application-uva-photoageing-vitamin-d-paradox/c01/) - Ch.2 Exponential versus Linear: Spectra, Film Thickness and an Effect-Size Interval — Chapter 2 uses a spectral model to show when each law holds; a formulation with a UVA gap gives 7.86, and the model average is a single digit whenever the power law's weight is at least 0.144. (https://5dlibrary.com/answers/en/spf-under-application-uva-photoageing-vitamin-d-paradox/c02/) 第2章 指数律与线性律之争:光谱、膜厚与效应量区间 — 第二章用光谱模型说明两条律各自成立的条件,长波有漏洞的配方给出 7.86;只要给幂律的权重不低于 0.144,模型平均就是个位数。 (https://5dlibrary.com/answers/zh/spf-under-application-uva-photoageing-vitamin-d-paradox/c02/) - Ch.3 UV Index Times Time: a Dose Ledger in Minimal Erythema Doses — Chapter 3 builds a dose ledger from UV index, time and minimal erythema dose; a clear mid-July day in Taipei delivers about 71.2 standard erythema doses. (https://5dlibrary.com/answers/en/spf-under-application-uva-photoageing-vitamin-d-paradox/c03/) 第3章 紫外指数乘以时间:最小红斑量的剂量账本 — 第三章建立紫外指数、时间与最小红斑量的剂量账本,台北七月中旬晴天一日约 71.2 个标准红斑量。 (https://5dlibrary.com/answers/zh/spf-under-application-uva-photoageing-vitamin-d-paradox/c03/) - Ch.4 UVA and Photo-Ageing: the Band the SPF Does Not See — Chapter 4 shows that UVA carries 0.979 of ground-level ultraviolet energy yet barely enters the SPF, and that the Nambour photo-ageing trial had a power of only 0.56. (https://5dlibrary.com/answers/en/spf-under-application-uva-photoageing-vitamin-d-paradox/c04/) 第4章 长波紫外与光老化:SPF 看不见的那一段 — 第四章说明长波紫外占地面紫外能量的 0.979 却几乎不进入 SPF,南布尔光老化试验的把握度只有 0.56。 (https://5dlibrary.com/answers/zh/spf-under-application-uva-photoageing-vitamin-d-paradox/c04/) - Ch.5 The Vitamin D Paradox: Why Applying Too Little Preserves Synthesis — Chapter 5 resolves the vitamin D paradox: the vitamin D effective dose of the field median pattern is 25.0 times that of the laboratory pattern, and the 2025 Sun-D trial found a between-group difference of 5.2 nmol/L for routine high-SPF use, a fall that is real but limited. (https://5dlibrary.com/answers/en/spf-under-application-uva-photoageing-vitamin-d-paradox/c05/) 第5章 维生素 D 悖论:为什么涂不够反而保住了合成 — 第五章解开维生素 D 悖论:现场中位情形的维生素 D 有效剂量是实验室情形的 25.0 倍;2025 年的 Sun-D 试验显示常规使用高 SPF 产品的组间差为 5.2 nmol/L,下降真实而有限。 (https://5dlibrary.com/answers/zh/spf-under-application-uva-photoageing-vitamin-d-paradox/c05/) - Ch.6 Taking the Paradox Apart: Grams, Coverage and High-UVA Formulations — Chapter 6 breaks the paradox into formulation, grams and coverage; at the same SPF15, a high-UVA formulation leaves vitamin D 4.63 times the share left by a low-UVA one. (https://5dlibrary.com/answers/en/spf-under-application-uva-photoageing-vitamin-d-paradox/c06/) 第6章 拆开悖论:克数、覆盖率与高 UVA 防护配方 — 第六章把悖论拆成配方、克数与覆盖率,同为 SPF15 时高长波防护配方留给维生素 D 的份额是低长波防护配方的 4.63 倍。 (https://5dlibrary.com/answers/zh/spf-under-application-uva-photoageing-vitamin-d-paradox/c06/) - Ch.7 Classical Sources East and West: Sun Simiao's Face Preparations and Finsen's Light — With these illustrative parameters, violet light in normal skin falls to one per cent at a depth of only 0.26 millimetres; with pressure the depth is 0.51 millimetres, a gain of 2.00 times. UVA goes from 0.24 to 0.40 millimetres. Red light already penetrates deeply, from 2.71 to 2.97 millimetres, and pressure brings a gain of only 1.10 times. At a depth of one millimetre, pressed skin passes 8103.1 times as much violet light, and at 0.5 millimetres 90.0 times as much. The numbers are illustrative, but their structure shows why Finsen's intuition was sound: in bands where blood absorbs strongly, removing blood has an exponential effect. (https://5dlibrary.com/answers/en/spf-under-application-uva-photoageing-vitamin-d-paradox/c07/) 第7章 东西方原典考辨:孙思邈的面药与芬森的光 — 在示意参数下,紫光在正常皮肤中衰减到百分之一的深度只有 0.26 毫米,压迫后是 0.51 毫米,加深了 2.00 倍;长波紫外从 0.24 毫米加深到 0.40 毫米。红光本来就穿得很深,从 2.71 毫米到 2.97 毫米,压迫只带来 1.10 倍的增益。在 1 毫米深处,压迫后的皮肤让紫光多透过 8103.1 倍;在 0.5 毫米处也有 90.0 倍。这些数字是示意,但它们的结构说明了芬森的直觉为什么对:在血液吸收强的波段,减少血液的效果是指数级的。 (https://5dlibrary.com/answers/zh/spf-under-application-uva-photoageing-vitamin-d-paradox/c07/) - Ch.8 Closing the Loop and Planning Refutation: a Usage Card, a Travel Dose Card and Pre-Registration — Chapter 8 gives a shelf usage card, a Taipei and Kuala Lumpur travel dose card, the distilled Sunscreen Dose Calculator and a pre-registered 40-person plan that refutes the claim with probability 0.931 if the linear law is true. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/spf-under-application-uva-photoageing-vitamin-d-paradox/c08/) 第8章 实证闭环与证伪方案:用量说明卡、行程剂量卡与预注册 — 第八章给出货架用量说明卡、台北与吉隆坡行程剂量卡、蒸馏小工具“防晒剂量计算器”和 40 人的预注册方案,在线性律为真时该方案推翻命题的概率为 0.931。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/spf-under-application-uva-photoageing-vitamin-d-paradox/c08/) ## 5DL-B-191 · ALDH2 Flushing Genetics, the Tolerance-Carcinogenesis Paradox and Blood-Alcohol Curves at Business Banquets 中文书名:乙醛脱氢酶二型红脸基因、酒量耐受致癌悖论与商务宴请血醇曲线母典 DOI: pending Cite: Fan, Y. (2026). ALDH2 Flushing Genetics, the Tolerance-Carcinogenesis Paradox and Blood-Alcohol Curves at Business Banquets (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Every step is written as checkable arithmetic in programs using only the Python standard library; parameters without a cited source are illustrative, and all numbers are generated by those programs. This book is not medical advice and supplies no safe drinking quantity in any form; its blood-alcohol curves are for arranging transport and accommodation and must never be used to decide whether anyone may drive, because after drinking no one drives. Abstract: A sentence circulates at East Asian banquet tables: people who flush can train themselves to drink. This book splits that sentence into an observation and an inference and shows the inference to be wrong. The observation holds: the subjective flushing response fades with the number of occasions. The inference says risk falls with it, and the evidence points the other way. Every step is written as checkable arithmetic in programs using only the Python standard library; parameters without a cited source are illustrative, and all numbers are generated by those programs. This book is not medical advice and supplies no safe drinking quantity in any form; its blood-alcohol curves are for arranging transport and accommodation and must never be used to decide whether anyone may drive, because after drinking no one drives. Chapter 1 shows the blush to be an outward expression of acetaldehyde rather than a gauge of metabolism: the two enzyme forms differ in catalytic efficiency by a factor of 833.33, and habituation raises the acetaldehyde peak at a fixed subjective mark by a factor of 6.00. Chapter 2 states the received idea directly, gives the monotone gradient pooled by drinking stratum in which ten standard drinks a week multiply the odds ratio by 1.891, and reports an E-value of 13.72 for the heavy-stratum estimate of 7.12, while presenting the contrary evidence that carriers who do not drink show no detectable excess. Chapter 3 builds the blood-alcohol curve and the cup conversions. Chapter 4 derives the plateau formula, where a carrier's 24.00 micromolar stands six times above the common genotype's 4.00. Chapter 5 converts banquets into additive aldehyde hours: a carrier's ordinary year is 5299 micromolar hours, falling to 0.2613 if two fewer cups are poured and the nights are halved. Chapter 6 treats the flushing questionnaire as a classifier with a positive predictive value of 0.8084. Chapter 7 examines Zhu Hong's Beishan wine canon and Plato's Laws: the same cup holds 4.08 times the ethanol once the drink is distilled, and Plato's age ladder removes 0.3987 of lifetime exposure. Chapter 8 gives a stop-point table, a zero-alcohol default plan that takes a table from 618.2 to 309.1 micromolar hours at an acceptance share of 0.50, and a pre-registered protocol. The book contains 20 runnable programs. 摘要:东亚的酒桌上流传着一句话:红脸的人练练就能喝。本书把这句话拆成一个观察与一个推论,并证明推论是错的。观察是真的:主观的红脸反应会随出席次数衰减。推论说风险随之下降,而现有证据的方向相反。本书用纯标准库的程序把每一步写成可以核对的算术:参数凡未注明出处者均为示意,所有数字由书中程序生成。本书不是医疗建议,不提供任何形式的安全饮用量;书中的血醇曲线用于安排用车与住宿,绝不可用于判断能否驾驶——饮酒之后不驾车。 第一章说明红脸是乙醛的外部表现而非代谢能力的仪表:两种酶形式的催化效率相差 833.33 倍,习惯化让同一主观刻度下的乙醛峰值升到 6.00 倍。第二章正面陈述本书要推翻的常识,给出按饮酒层合并的单调梯度,每周每 10 个标准杯把比值比乘以 1.891,重度层的 7.12 对应的 E 值为 13.72;同时并陈反方证据:不饮酒的携带者没有可检出的超额风险。第三章给出血醇曲线与杯型换算。第四章导出乙醛平台公式,携带者的平台 24.00 微摩尔每升是常型 4.00 的六倍。 第五章把宴请折算成可加的乙醛小时,一个携带者的普通年份是 5299 微摩尔小时,每场少斟两杯且场次减半降到 0.2613。第六章把红脸问卷当作分类器,阳性预测值 0.8084。第七章考辨朱肱《北山酒经》与柏拉图《法律篇》:同一只杯子从发酵酒换成蒸馏酒,乙醇多出 4.08 倍,而柏拉图的年龄阶梯能移走 0.3987 的终生暴露。第八章给出停杯点表、无酒精默认值方案(接受率 0.50 时一桌从 618.2 降到 309.1 微摩尔小时)与预注册检验方案。全书含 20 段可运行程序。 Chapter answers: - Ch.1 The Blush Is Not a Broken Alarm: the Kinetics of One Amino Acid — Chapter 1 shows the blush to be an outward expression of acetaldehyde rather than a gauge of metabolism: the two enzyme forms differ in catalytic efficiency by a factor of 833.33, and habituation raises the acetaldehyde peak at a fixed subjective mark by a factor of 6.00. (https://5dlibrary.com/answers/en/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c01/) 第1章 红脸不是坏掉的警报器:一个氨基酸的酶学 — 第一章说明红脸是乙醛的外部表现而非代谢能力的仪表:两种酶形式的催化效率相差 833.33 倍,习惯化让同一主观刻度下的乙醛峰值升到 6.00 倍。 (https://5dlibrary.com/answers/zh/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c01/) - Ch.2 The Tolerance-Carcinogenesis Paradox: a Monotone Dose Curve — Chapter 2 states the received idea directly, gives the monotone gradient pooled by drinking stratum in which ten standard drinks a week multiply the odds ratio by 1.891, and reports an E-value of 13.72 for the heavy-stratum estimate of 7.12, while presenting the contrary evidence that carriers who do not drink show no detectable excess. (https://5dlibrary.com/answers/en/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c02/) 第2章 耐受—致癌悖论:一条单调上升的剂量曲线 — 第二章正面陈述本书要推翻的常识,给出按饮酒层合并的单调梯度,每周每 10 个标准杯把比值比乘以 1.891,重度层的 7.12 对应的 E 值为 13.72;同时并陈反方证据:不饮酒的携带者没有可检出的超额风险。 (https://5dlibrary.com/answers/zh/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c02/) - Ch.3 The Blood-Alcohol Curve: Widmark, Body Water and the Cup — The blood-alcohol curve is not an academic abstraction but a schedule that can be drawn before the first toast. A body-water regression gives an individualised distribution factor, and the illustrative guest's 0.7228 together with 0.1976 g per litre for one standard drink are two constants used repeatedly later in the book. Four guests at one table taking four identical cups can differ in peak by more than a third; a 50 mL cup of spirit is 2.05 standard drinks; slowing the pace lowers the area to 0.3909, while cutting six cups to three lowers it to 0.2514. The next chapter joins this ethanol curve to acetaldehyde and asks what the same cups become in different genotypes. (https://5dlibrary.com/answers/en/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c03/) 第3章 血醇曲线:Widmark 公式、体水回归与杯型 — 血醇曲线不是一条抽象的学术产物,它是一张可以在开席前画好的排程表。体水回归给出个体化的分布系数,示意客人的 0.7228 与一个标准杯 0.1976 克每升是本书后文反复用到的两个常数。同一张桌上的四位客人喝同样四杯,峰值可以相差三分之一以上;一只 50 毫升的白酒杯等于 2.05 个标准杯;放慢节奏把面积降到 0.3909,而把六杯减到三杯降到 0.2514。下一章把这条乙醇曲线接到乙醛上,看看同样的杯数在不同基因型身上变成了什么。 (https://5dlibrary.com/answers/zh/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c03/) - Ch.4 Two Tanks in Series: the Aldehyde Plateau and a Few Points of Margin — Chapter 4 derives the plateau formula, where a carrier's 24.00 micromolar stands six times above the common genotype's 4.00. (https://5dlibrary.com/answers/en/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c04/) 第4章 两室串联:乙醛平台与那几个百分点的裕量 — 第四章导出乙醛平台公式,携带者的平台 24.00 微摩尔每升是常型 4.00 的六倍。 (https://5dlibrary.com/answers/zh/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c04/) - Ch.5 The Annual Ledger: Banquets Converted into Aldehyde Hours — Chapter 5 converts banquets into additive aldehyde hours: a carrier's ordinary year is 5299 micromolar hours, falling to 0.2613 if two fewer cups are poured and the nights are halved. (https://5dlibrary.com/answers/en/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c05/) 第5章 一年的账:把宴请折算成乙醛小时 — 第五章把宴请折算成可加的乙醛小时,一个携带者的普通年份是 5299 微摩尔小时,每场少斟两杯且场次减半降到 0.2613。 (https://5dlibrary.com/answers/zh/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c05/) - Ch.6 Two Questions: the Flushing History as a Classifier — Chapter 6 treats the flushing questionnaire as a classifier with a positive predictive value of 0.8084. (https://5dlibrary.com/answers/en/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c06/) 第6章 两句问话:把红脸史当作一个分类器 — 第六章把红脸问卷当作分类器,阳性预测值 0.8084。 (https://5dlibrary.com/answers/zh/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c06/) - Ch.7 Two Classical Sources: Zhu Hong's Beishan Wine Canon and Plato's Laws — Chapter 7 examines Zhu Hong's Beishan wine canon and Plato's Laws: the same cup holds 4.08 times the ethanol once the drink is distilled, and Plato's age ladder removes 0.3987 of lifetime exposure. (https://5dlibrary.com/answers/en/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c07/) 第7章 东西方原典考辨:朱肱《北山酒经》与柏拉图《法律篇》 — 第七章考辨朱肱《北山酒经》与柏拉图《法律篇》:同一只杯子从发酵酒换成蒸馏酒,乙醇多出 4.08 倍,而柏拉图的年龄阶梯能移走 0.3987 的终生暴露。 (https://5dlibrary.com/answers/zh/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c07/) - Ch.8 The Empirical Loop: Stop Points, Zero-Alcohol Gifts and a Pre-Registered Protocol — Chapter 8 gives a stop-point table, a zero-alcohol default plan that takes a table from 618.2 to 309.1 micromolar hours at an acceptance share of 0.50, and a pre-registered protocol. The book contains 20 runnable programs. (https://5dlibrary.com/answers/en/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c08/) 第8章 实证闭环:停杯点、无酒精礼单与预注册方案 — 第八章给出停杯点表、无酒精默认值方案(接受率 0.50 时一桌从 618.2 降到 309.1 微摩尔小时)与预注册检验方案。全书含 20 段可运行程序。 (https://5dlibrary.com/answers/zh/aldh2-flushing-tolerance-paradox-business-banquet-blood-alcohol/c08/) ## 5DL-B-192 · Alkaline Water against Gastric Acid, the TDS Myth and Electrolyte Dosing in Drinking Water 中文书名:碱性水胃酸缓冲化学证伪、总溶解固体迷思与饮水电解质剂量母典 DOI: pending Cite: Fan, Y. (2026). Alkaline Water against Gastric Acid, the TDS Myth and Electrolyte Dosing in Drinking Water (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs. The book is not medical advice and claims no effect of any water or drink on any disease. Abstract: This book tests two sayings of the drinking-water market: that alkaline water above pH 9 neutralises acid in the body and changes an "acidic constitution", and that the lower the total dissolved solids (TDS) of a water, the healthier it is. It writes them as runnable arithmetic: charge balance and bisection, carbonate equilibria and buffer capacity, a dynamic titration model of the stomach, the Henderson-Hasselbalch equation, a ledger of potential renal acid load, and linear programming. Its parameters are illustrative; all calculations are generated by its programs. The book is not medical advice and claims no effect of any water or drink on any disease. Chapter 1 uses a strong-acid, strong-base charge balance to show that a glass of pH 9 water entering a fasting stomach at pH 1.7 contributes only 0.0011 pH unit through its alkali, and that bringing 50 mL of gastric juice to neutrality would take 99.8 litres. Chapter 2 shows that neutralising power is set by total alkalinity, not label pH: across six waters the rank correlation between the two is −0.883, and a bicarbonate mineral water labelled pH 6.6 carries 1126 times the alkali of ionised water at pH 9.5; water rich enough in bicarbonate does change urine pH, a correction the book makes to its commissioning claim. Chapter 3 writes the stomach as a titrator that secretes and empties continuously. Chapter 4 shows that a litre of ionised water, even if wholly absorbed, raises plasma pH by only 0.000041, and that a rise of 0.05 would take 6134 litres. Chapter 5 shows that TDS is a total, not a health score, and that the pooled relative risk of 0.89 for drinking-water magnesium and coronary death has an E-value of only 1.496. Chapter 6 uses linear programming to find a least-cost electrolyte recipe at 0.4613 yuan per litre, and shows that drinking plain water faster than 1.12 litres an hour, or a 30 mmol sodium drink faster than 1.44 litres an hour, drives plasma sodium below the threshold after four hours. Chapter 7 examines the Waters section of the Bencao gangmu and the Hippocratic Airs, Waters, Places: boiling a limestone well water drops 279.7 mg of scale per litre, in line with the old practice of boiling and waiting for the alkali to settle. Chapter 8 gives a product compliance checklist, the drinking-water pH calculator and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:本书检验饮用水市场上两句流行的话:喝 pH 9 以上的碱性水能中和体内的酸、改变「酸性体质」;水的总溶解固体(TDS)越低越健康。本书把它们写成可以运行的算术:电荷平衡与二分法、碳酸盐平衡与缓冲容量、胃的动态滴定模型、亨德森—哈塞尔巴尔赫方程、潜在肾酸负荷账本与线性规划。书中参数为示意;全部计算由书中程序生成。本书不是医疗建议,不宣称任何饮用水或饮品对任何疾病有效。 第一章用强酸—强碱电荷平衡说明,一杯 pH 9 的水进入 pH 1.7 的空腹胃,碱的贡献只有 0.0011 个 pH 单位,把 50 mL 胃液抬到中性需要 99.8 升。第二章说明决定中和能力的是总碱度而非标签 pH:六种水的标签 pH 与碱度的秩相关为 −0.883,标 pH 6.6 的碳酸氢盐矿泉水的碱度是 pH 9.5 离子水的 1126 倍;碳酸氢根足够多的水会改变尿液 pH,这是本书对立项命题的修正。第三章把胃写成持续泌酸、持续排空的滴定器。第四章说明一升离子水即使全部吸收,也只让血浆 pH 升高 0.000041,升高 0.05 需要 6134 升。 第五章说明 TDS 是总量而非健康分,饮水镁与冠心病死亡的合并相对风险 0.89 的 E 值只有 1.496。第六章用线性规划得出每升 0.4613 元的最低成本电解质配方,并说明白水饮用速度超过每小时 1.12 升、每升 30 mmol 钠的饮品超过 1.44 升时,四小时后血钠会跌破阈值。第七章考辨《本草纲目·水部》与希波克拉底《空气、水与地方》,石灰岩井水煮沸后每升析出 279.7 mg 水垢,与古人「煎滚、候碱澄」的做法吻合。第八章给出选品合规清单、饮水 pH 计算器与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A Glass of pH 9 Water Enters the Stomach: an Exact Strong-Acid, Strong-Base Charge Balance — Chapter 1 uses a strong-acid, strong-base charge balance to show that a glass of pH 9 water entering a fasting stomach at pH 1.7 contributes only 0.0011 pH unit through its alkali, and that bringing 50 mL of gastric juice to neutrality would take 99.8 litres. (https://5dlibrary.com/answers/en/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c01/) 第1章 一杯 pH 9 的水进胃:强酸—强碱电荷平衡的精确解 — 第一章用强酸—强碱电荷平衡说明,一杯 pH 9 的水进入 pH 1.7 的空腹胃,碱的贡献只有 0.0011 个 pH 单位,把 50 mL 胃液抬到中性需要 99.8 升。 (https://5dlibrary.com/answers/zh/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c01/) - Ch.2 Label pH Is Not Alkalinity: the Carbonate System and the Real Alkali Load — Chapter 2 shows that neutralising power is set by total alkalinity, not label pH: across six waters the rank correlation between the two is −0.883, and a bicarbonate mineral water labelled pH 6.6 carries 1126 times the alkali of ionised water at pH 9.5; water rich enough in bicarbonate does change urine pH, a correction the book makes to its commissioning claim. (https://5dlibrary.com/answers/en/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c02/) 第2章 标签 pH 不是碱度:碳酸盐体系与真正的碱负荷 — 第二章说明决定中和能力的是总碱度而非标签 pH:六种水的标签 pH 与碱度的秩相关为 −0.883,标 pH 6.6 的碳酸氢盐矿泉水的碱度是 pH 9.5 离子水的 1126 倍;碳酸氢根足够多的水会改变尿液 pH,这是本书对立项命题的修正。 (https://5dlibrary.com/answers/zh/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c02/) - Ch.3 The Stomach as a Constant-Acid Titrator: Secretion, Emptying and Buffer Capacity — Once the stomach is written as a pool with continuous inflow and emptying, ionised water at pH 9.5 and pure water draw the same curve: both fall back below pH 2.5 at about minute 19, some ten seconds apart. Hard tap water keeps the pH above 3 for a total of 29.7 minutes and bicarbonate mineral water for 88.7 minutes, the same order as the buffering of a meal. Buffer capacity explains the difference: 0.002 mmol per litre at pH 6 for ionised water against 27.340 mmol for the mineral water. The verdict on ionised water survives the whole parameter scan; the effect of bicarbonate water depends on the person and the moment. (https://5dlibrary.com/answers/en/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c03/) 第3章 胃是一台恒酸滴定器:分泌、排空与缓冲容量 — 把胃写成持续进料、持续排空的混合池以后,pH 9.5 离子水与纯水给出同一条曲线:两者都在第 19 分钟左右回落到 2.5 以下,差约十秒。硬自来水让 pH 高于 3 的时间累计 29.7 分钟,富碳酸氢盐矿泉水为 88.7 分钟,与一顿饭的缓冲同一量级。缓冲容量这个导数解释了差别:离子水在 pH 6 时的缓冲为每升 0.002 mmol,矿泉水为 27.340 mmol。离子水无效的结论在整个参数扫描中都成立,碳酸氢盐水的效应则取决于人与时刻。 (https://5dlibrary.com/answers/zh/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c03/) - Ch.4 Why Blood pH Does Not Move: the Henderson-Hasselbalch Equation and a PRAL Ledger — Chapter 4 shows that a litre of ionised water, even if wholly absorbed, raises plasma pH by only 0.000041, and that a rise of 0.05 would take 6134 litres. (https://5dlibrary.com/answers/en/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c04/) 第4章 血液 pH 为什么不动:亨德森—哈塞尔巴尔赫方程与 PRAL 账本 — 第四章说明一升离子水即使全部吸收,也只让血浆 pH 升高 0.000041,升高 0.05 需要 6134 升。 (https://5dlibrary.com/answers/zh/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c04/) - Ch.5 TDS Is Not a Health Score: Conductivity, Ionic Composition and the Strength of the Magnesium Evidence — Chapter 5 shows that TDS is a total, not a health score, and that the pooled relative risk of 0.89 for drinking-water magnesium and coronary death has an E-value of only 1.496. (https://5dlibrary.com/answers/en/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c05/) 第5章 TDS 不是健康分:电导率、离子构成与镁的证据强度 — 第五章说明 TDS 是总量而非健康分,饮水镁与冠心病死亡的合并相对风险 0.89 的 E 值只有 1.496。 (https://5dlibrary.com/answers/zh/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c05/) - Ch.6 Dosing Electrolyte Drinks: Linear Programming, Sodium Dilution and Differences between People — Chapter 6 uses linear programming to find a least-cost electrolyte recipe at 0.4613 yuan per litre, and shows that drinking plain water faster than 1.12 litres an hour, or a 30 mmol sodium drink faster than 1.44 litres an hour, drives plasma sodium below the threshold after four hours. (https://5dlibrary.com/answers/en/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c06/) 第6章 电解质饮品配比:线性规划、钠稀释与群体差异 — 第六章用线性规划得出每升 0.4613 元的最低成本电解质配方,并说明白水饮用速度超过每小时 1.12 升、每升 30 mmol 钠的饮品超过 1.44 升时,四小时后血钠会跌破阈值。 (https://5dlibrary.com/answers/zh/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c06/) - Ch.7 Eastern and Western Classics Examined: the Waters Section of the Bencao gangmu and Airs, Waters, Places — Chapter 7 examines the Waters section of the Bencao gangmu and the Hippocratic Airs, Waters, Places: boiling a limestone well water drops 279.7 mg of scale per litre, in line with the old practice of boiling and waiting for the alkali to settle. (https://5dlibrary.com/answers/en/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c07/) 第7章 东西方原典考辨:《本草纲目·水部》与《空气、水与地方》 — 第七章考辨《本草纲目·水部》与希波克拉底《空气、水与地方》,石灰岩井水煮沸后每升析出 279.7 mg 水垢,与古人「煎滚、候碱澄」的做法吻合。 (https://5dlibrary.com/answers/zh/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c07/) - Ch.8 Closing the Loop: a Product Compliance Checklist, the Drinking-Water pH Calculator and a Pre-Registered Protocol — Chapter 8 gives a product compliance checklist, the drinking-water pH calculator and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c08/) 第8章 实证闭环:选品合规清单、饮水 pH 计算器与预注册方案 — 第八章给出选品合规清单、饮水 pH 计算器与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/alkaline-water-gastric-buffering-tds-myth-electrolyte-dosing/c08/) ## 5DL-B-193 · Gompertz Mortality Doubling, Age Misreporting among Supercentenarians and Data Audits of Longevity Blue Zones 中文书名:冈珀茨死亡率倍增定律、超级百岁老人年龄误报与长寿蓝区数据审计母典 DOI: pending Cite: Fan, Y. (2026). Gompertz Mortality Doubling, Age Misreporting among Supercentenarians and Data Audits of Longevity Blue Zones (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not medical advice, claims no effect for any product or method, and any self-experiment should first be discussed with a licensed physician. Marked boxes separate published evidence, runnable computations (22 programs), classical sources and the author's hypotheses. Abstract: This book tests two popular beliefs: that ageing is slow and even, and that the world has longevity "blue zones" where centenarians are everywhere. It states and quantifies a falsifiable counter-claim: the adult force of mortality doubles roughly every eight years (a slope of 0.0866), so ageing compounds; and some clusters of extreme longevity may partly reflect missing birth registration, misreported ages and even pension fraud. The second point is contested, and the book sets out the evidence of both supporters and critics. Its parameters are illustrative; all calculations are generated by its programs; the book is not medical advice, claims no effect for any product or method, and any self-experiment should first be discussed with a licensed physician. Chapter 1 uses least squares and window-by-window fits to show how sensitive the doubling time is to the fitting window and to background mortality. Chapter 2 uses a base-rate model to show that the fall in US supercentenarian records after birth certificates needs only 1.437 to 2.941 age exaggerations per ten thousand records, while Japan's register "ghosts" never entered the life tables. Chapter 3 compares fitting methods in small villages and reads the tails of survival curves. Chapter 4 injects errors to show that large exaggerations can produce a plateau at very low error rates, whereas small ones cannot. Chapter 5 shows that real and artefactual plateaus are hard to tell apart in present samples. Chapter 6 audits Sardinia, Okinawa, Nicoya and Ikaria: Nicoya's "seven times the chance of reaching 100" requires a hazard ratio of 0.456, inconsistent with its death-rate ratio, and its advantage shrinks by about 1.43 years for every decade of later birth. Chapter 7 weighs the Five Blessings of the Hongfan against Gompertz's original paper of 1825. Chapter 8 offers a falsification checklist for longevity copy, a non-medical healthy-lifespan planning questionnaire (showing why hazard ratios must not be multiplied: three habits with no causal effect multiply to 0.437), the distilled Longevity Claim Checker and a pre-registered test protocol. Marked boxes separate published evidence, runnable computations (22 programs), classical sources and the author's hypotheses. 摘要:本书检验两个流行说法:衰老是缓慢均匀的过程,以及世界上存在百岁老人随处可见的长寿「蓝区」。本书提出并量化一个可证伪的反命题:成人死亡力大约每八年翻一倍(斜率 0.0866),衰老按复利进行;部分超高龄聚集区可能与出生登记缺失、年龄误报乃至冒领养老金有关。后一点属于学界仍有争议的命题,全书并陈支持与反对双方的证据。书中参数为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何产品或方法的效应,自我试验须先咨询执业医师。 第一章用最小二乘与分窗口拟合说明倍增时间对窗口与背景死亡的敏感;第二章用基础比率模型显示,解释美国出生证明前后超级百岁记录的下降,只需每万条记录 1.437 到 2.941 条年龄夸大,而日本户籍「幽灵」并未进入生命表;第三章比较小村庄中的拟合方法并读出生存曲线的尾部;第四章用误差注入说明,大幅度的年龄夸大能以极低错误率压出平台,小幅度则不能;第五章说明真平台与错误平台在现有样本中难以区分。 第六章审计撒丁岛、冲绳、尼科亚与伊卡里亚:尼科亚「百岁机会七倍」需要的风险比为 0.456,与其死亡率比不一致,其优势每晚生十年约丧失 1.43 岁。第七章考辨《尚书·洪范》的五福与冈珀茨 1825 年原论文。第八章给出长寿文案证伪清单、非医疗的健康寿命规划问卷(说明为何不能把多项风险比相乘:无因果作用的三种习惯相乘得 0.437)、蒸馏小工具「长寿声称核查表」与预注册检验方案。书中以标注框区分已发表证据、可运行计算(22 段程序)、原典与作者假说。 Chapter answers: - Ch.1 Doubling Every Eight Years: Gompertz Slopes, Fitting Windows and the Interval Around a Doubling Time — The Gompertz law writes the adult hazard as an exponential function of age. A slope of about 0.0866 corresponds to a doubling time of about eight years; published slopes range from 0.070 to 0.110, doubling times correspondingly from 6.30 to 9.90 years. Fits are sensitive to the window: background mortality flattens the young end, heterogeneity flattens the old end, and a bootstrap interval can be narrow while missing the truth. The next chapter turns to the second claim of the book: at the very top of the age range, how much of the flattening we see comes from the records themselves. (https://5dlibrary.com/answers/en/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c01/) 第1章 每八年翻一倍:冈珀茨斜率、拟合窗口与倍增时间的区间 — 冈珀茨定律把成年死亡力写成年龄的指数函数,斜率 B 约 0.0866,对应约 8 年的倍增时间;文献中的斜率在 0.070 到 0.110 之间,倍增时间相应在 6.30 到 9.90 年之间。拟合结果对窗口很敏感:背景死亡让低龄段变平,人群异质性让高龄段变平;自助法区间可以很窄却不包含真值。下一章转向本书的第二个主张:在最高龄的那一端,我们看到的变平,有多少来自记录本身。 (https://5dlibrary.com/answers/zh/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c01/) - Ch.2 When Birth Certificates Arrive, Supercentenarians Thin Out: Newman's Audit and the Case Against It — Chapter 2 uses a base-rate model to show that the fall in US supercentenarian records after birth certificates needs only 1.437 to 2.941 age exaggerations per ten thousand records, while Japan's register "ghosts" never entered the life tables. (https://5dlibrary.com/answers/en/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c02/) 第2章 出生证一到,超级百岁记录就少了:纽曼审计与反方证据 — 第二章用基础比率模型显示,解释美国出生证明前后超级百岁记录的下降,只需每万条记录 1.437 到 2.941 条年龄夸大,而日本户籍「幽灵」并未进入生命表; (https://5dlibrary.com/answers/zh/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c02/) - Ch.3 Least Squares and Survival Curves: Village Noise, Tails and the Age of 122 — Gompertz parameters translate directly into survival curves: modes and medians are stable, tails are extremely sensitive to the slope. In small samples Poisson likelihood beats least squares that drops zero-death ages; in a village of 300, chance alone can move the doubling time between seven and nine years. Calment's 122 years is all but impossible under a pure Gompertz extrapolation and merely rare under a plateau model; statistical prominence cannot replace archival validation. The next chapter tests error injection directly: how much age exaggeration does it take to press a Gompertz curve into a plateau? (https://5dlibrary.com/answers/en/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c03/) 第3章 最小二乘与生存曲线:小村庄的噪声、尾部与一百二十二岁 — 冈珀茨参数可以直接换成生存曲线:众数与中位数稳定,尾部却对斜率极端敏感。小样本中,泊松似然优于丢弃零死亡年龄的最小二乘;300 人的村庄单凭偶然就能让倍增时间在 7 年到 9 年之间摆动。卡尔芒的 122 岁在纯冈珀茨外推下几乎不可能出现,在平台模型下则只是罕见;统计上的突出不能代替档案核验。下一章正式检验误差注入:多少年龄夸大,能把一条冈珀茨曲线压成平台。 (https://5dlibrary.com/answers/zh/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c03/) - Ch.4 Error Injection: How a Few Exaggerated Ages in Ten Thousand Press the Curve Flat — Validated Italian and international data show a roughly flat hazard after 105 or 110, with an annual probability of death of about one half. Error-injection models show that very few exaggerated ages in ten thousand records can press a Gompertz curve into an approximate plateau, and that cohort blending can create deceleration below 100. Yet how many errors it takes to produce a plateau of 0.645 depends heavily on the size of the exaggeration: small shifts would require implausibly many errors, large shifts such as identity substitution very few. The next chapter turns to the other possibility: that the plateau is no error but real heterogeneity among people. (https://5dlibrary.com/answers/en/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c04/) 第4章 误差注入:万分之几的年龄夸大如何把曲线压平 — 经过核验的意大利与国际数据显示,105 岁或 110 岁以后的死亡力大致平坦,年死亡概率约为一半。误差注入模型显示,每万条记录中极少的年龄夸大就能把一条冈珀茨曲线压成近似平台,队列混合则能制造百岁以下的减速。但制造一个 0.645 的平台需要多少错误,强烈依赖夸大幅度:小幅夸大所需的错误率高得难以置信,大幅夸大(例如身份替换)只需极少。下一章转向另一种可能:平台不是错误,而是真实的人群异质性。 (https://5dlibrary.com/answers/zh/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c04/) - Ch.5 Real Plateau or Artefact: Heterogeneity Models, Cohort Trends and Identifiability — Gamma frailty can create a real plateau, but its simplest form only approaches the ceiling asymptotically and struggles to reproduce a curve that is flat from 105. The plateau falling across cohorts runs opposite to the simple prediction that errors decline over time, which counts against a pure error explanation. Slope tests show that the strongest claim, an unchanged Gompertz law up to 110, conflicts with validated data, while small slopes between 0.0200 and 0.0300 cannot be distinguished from zero in present samples. The next chapter takes these tools to the blue zones themselves. (https://5dlibrary.com/answers/en/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c05/) 第5章 真平台还是假平台:异质性模型、队列趋势与可辨识性 — 伽马脆弱度模型能产生真实的平台,但最简单的版本只能渐近地接近上限,难以复制「105 岁起即平直」的形状;平台随队列降低的观察,与「错误随时间减少」的简单预测方向相反,对纯错误解释不利。斜率检验显示,冈珀茨原样延续到 110 岁的最强说法与核验数据不符,而 0.0200 到 0.0300 的小斜率在现有样本中无法与零区分。下一章把这些工具带到蓝区本身。 (https://5dlibrary.com/answers/zh/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c05/) - Ch.6 Auditing the Blue Zones: Sardinia, Okinawa, Nicoya and Ikaria — Chapter 6 audits Sardinia, Okinawa, Nicoya and Ikaria: Nicoya's "seven times the chance of reaching 100" requires a hazard ratio of 0.456, inconsistent with its death-rate ratio, and its advantage shrinks by about 1.43 years for every decade of later birth. (https://5dlibrary.com/answers/en/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c06/) 第6章 蓝区数据审计:撒丁岛、冲绳、尼科亚与伊卡里亚 — 第六章审计撒丁岛、冲绳、尼科亚与伊卡里亚:尼科亚「百岁机会七倍」需要的风险比为 0.456,与其死亡率比不一致,其优势每晚生十年约丧失 1.43 岁。 (https://5dlibrary.com/answers/zh/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c06/) - Ch.7 Classical Sources East and West: the Five Blessings of the Hongfan and Gompertz in 1825 — Chapter 7 weighs the Five Blessings of the Hongfan against Gompertz's original paper of 1825. (https://5dlibrary.com/answers/en/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c07/) 第7章 东西方原典考辨:《洪范》五福与冈珀茨一八二五 — 第七章考辨《尚书·洪范》的五福与冈珀茨 1825 年原论文。 (https://5dlibrary.com/answers/zh/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c07/) - Ch.8 Closing the Loop: a Falsification Checklist for Longevity Copy, a Healthy-Lifespan Planning Questionnaire and a Pre-registered Protocol — Chapter 8 offers a falsification checklist for longevity copy, a non-medical healthy-lifespan planning questionnaire (showing why hazard ratios must not be multiplied: three habits with no causal effect multiply to 0.437), the distilled Longevity Claim Checker and a pre-registered test protocol. Marked boxes separate published evidence, runnable computations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c08/) 第8章 实证闭环:长寿文案证伪清单、健康寿命规划问卷与预注册方案 — 第八章给出长寿文案证伪清单、非医疗的健康寿命规划问卷(说明为何不能把多项风险比相乘:无因果作用的三种习惯相乘得 0.437)、蒸馏小工具「长寿声称核查表」与预注册检验方案。书中以标注框区分已发表证据、可运行计算(22 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/gompertz-mortality-doubling-supercentenarian-age-audit-blue-zones/c08/) ## 5DL-B-194 · Indoor CO2 and Cognitive Decline, Bernoulli Cross-Ventilation and an Engineering Re-reading of 'Storing Wind, Gathering Qi' 中文书名:室内二氧化碳认知衰减、穿堂风伯努利换气与藏风聚气的通风工程重读母典 DOI: pending Cite: Fan, Y. (2026). Indoor CO2 and Cognitive Decline, Bernoulli Cross-Ventilation and an Engineering Re-reading of 'Storing Wind, Gathering Qi' (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: This book tests a counterintuitive claim: the roughly 1400 ppm of carbon dioxide common in meeting rooms has gone with a marked fall in decision-test scores in controlled exposure experiments, while pot plants do almost nothing for indoor air, so that matching the air exchange of an ordinary room would take something like a hundred pots per square metre. The book writes the claim as runnable arithmetic and corrects the wording of its original plan against the primary literature: the "half" of Allen and colleagues in 2016 is measured against the two best days, the high carbon dioxide day and the conventional day differ by a ratio of 0.99, and the effect of pure carbon dioxide remains contested. Rooms, occupancy, exhaled rates and wind pressure coefficients are illustrative; all calculations are generated by the book's programs; it is not medical advice, promises no health outcome, and offers business owners general, non-personalised operating advice only. Chapter 1 recomputes the table of Allen and colleagues, confirms the fall of 0.503, and shows that giving the null half the weight shrinks the predicted fall at 1400 ppm from 0.446 to 0.223. Chapter 2 uses the stirred-tank equation to show that an eight-person meeting room with 0.5 air changes per hour of infiltration crosses 1000 ppm after 21.3 minutes and 1400 ppm after 38.6. Chapter 3 converts plant-chamber data into clean air delivery rates: a room with one air change per hour needs 117.4 pots per square metre. Chapter 4 uses Bernoulli's orifice formula to show that a cross draught of 694 cubic metres per hour is driven by only 1.92 pascals. Chapter 5 finds the lowest-cost window schedule by dynamic programming: a two-hour meeting held below 1000 ppm needs at least 35 open minutes and loses 2.34 kilowatt-hours of heat in winter, against 7.01 with the window always open. Chapter 6 shows that field data cannot separate carbon dioxide from bioeffluents, and reads 1400 ppm as a rebreathed fraction of 0.0258. Chapter 7 examines "qi scatters on the wind" in the Zangshu traditionally attributed to Guo Pu and the eight winds of Vitruvius; a fully sheltered room keeps only 0.317 of the open-site flow. Chapter 8 gives the empirical loop for studios, live-stream rooms and reception rooms, the distilled Ventilation Scheduler (the demonstration live-stream room is open for 0.243 of the time) and a pre-registered 32-person paired test with power 0.801 against a medium effect; a closing-rate test would need 2754 meetings and so serves for monitoring only. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:本书检验一个反常识命题:会议室里常见的约 1400 ppm 二氧化碳,在受控暴露实验中伴随着决策测验得分的显著下降;而室内盆栽几乎不能改善空气,要赶上普通房间的换气,每平方米得摆上百盆。书中把命题写成可以运行的算术,并按原始文献修正了立项时的说法:艾伦等人 2016 年的“低一半”是相对最好的两天而言,高二氧化碳日与常规日的得分比值为 0.99,纯二氧化碳的效应在学界仍有分歧。书中的房间、人数、呼出量与风压系数均为示意参数;全部计算由书中程序生成;本书不是医疗建议,不对任何健康结果作出承诺,对经营者的建议为非个人化的经营建议。 第一章重算艾伦等人的表格,复核 0.503 的降幅,并说明若给零效应一半权重,1400 ppm 处的预计下降从 0.446 缩到 0.223。第二章用连续搅拌釜方程算出一间八人会议室在每小时 0.5 次渗透下 21.3 分钟越过 1000 ppm、38.6 分钟越过 1400 ppm。第三章把盆栽实验舱数据换算成洁净空气输送率,每小时换气一次的房间每平方米需要 117.4 盆。第四章用伯努利孔口流公式说明,驱动每小时 694 立方米穿堂风的压差只有 1.92 帕。 第五章用动态规划求最省的开窗排程,两小时会议守住 1000 ppm 至少开窗 35 分钟,冬季热损 2.34 千瓦时,一直开窗则为 7.01。第六章说明现场数据分不清二氧化碳与生物排放物,并把 1400 ppm 读作 0.0258 的再呼吸份额。第七章考辨旧题郭璞所撰《葬书》的“气乘风则散”与维特鲁威的八风,完全遮蔽的房间只剩开阔地 0.317 的流量。第八章给出工作室、直播间与会客室的实证闭环、蒸馏小工具“换气排程器”(示范直播间开窗时间占 0.243)和 32 人的预注册配对检验,中等效应下把握度为 0.801;成交率检验需要 2754 场会谈,故只作监测。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 1400 ppm in the Meeting Room: Three Controlled Exposure Experiments and Decision Scores — Chapter 1 recomputes the table of Allen and colleagues, confirms the fall of 0.503, and shows that giving the null half the weight shrinks the predicted fall at 1400 ppm from 0.446 to 0.223. (https://5dlibrary.com/answers/en/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c01/) 第1章 会议室里的 1400 ppm:三项受控暴露实验与决策测验分数 — 第一章重算艾伦等人的表格,复核 0.503 的降幅,并说明若给零效应一半权重,1400 ppm 处的预计下降从 0.446 缩到 0.223。 (https://5dlibrary.com/answers/zh/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c01/) - Ch.2 The Stirred Tank: How an Eight-Person Meeting Room Climbs to 1400 ppm — Chapter 2 uses the stirred-tank equation to show that an eight-person meeting room with 0.5 air changes per hour of infiltration crosses 1000 ppm after 21.3 minutes and 1400 ppm after 38.6. (https://5dlibrary.com/answers/en/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c02/) 第2章 连续搅拌釜:一间八人会议室怎样升到 1400 ppm — 第二章用连续搅拌釜方程算出一间八人会议室在每小时 0.5 次渗透下 21.3 分钟越过 1000 ppm、38.6 分钟越过 1400 ppm。 (https://5dlibrary.com/answers/zh/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c02/) - Ch.3 The Clean Air Delivery Rate of a Pot Plant: the Arithmetic of a Hundred Pots per Square Metre — Chapter 3 converts plant-chamber data into clean air delivery rates: a room with one air change per hour needs 117.4 pots per square metre. (https://5dlibrary.com/answers/en/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c03/) 第3章 盆栽的洁净空气输送率:每平方米上百盆的算术 — 第三章把盆栽实验舱数据换算成洁净空气输送率,每小时换气一次的房间每平方米需要 117.4 盆。 (https://5dlibrary.com/answers/zh/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c03/) - Ch.4 Bernoulli's Orifice and the Cross Draught: Wind Pressure, Stack Pressure and Openings in Series — Chapter 4 uses Bernoulli's orifice formula to show that a cross draught of 694 cubic metres per hour is driven by only 1.92 pascals. (https://5dlibrary.com/answers/en/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c04/) 第4章 穿堂风的伯努利孔口:风压、热压与开口串联 — 第四章用伯努利孔口流公式说明,驱动每小时 694 立方米穿堂风的压差只有 1.92 帕。 (https://5dlibrary.com/answers/zh/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c04/) - Ch.5 Pulse Ventilation and Dynamic Programming: a Window Schedule inside a Sawtooth — Chapter 5 finds the lowest-cost window schedule by dynamic programming: a two-hour meeting held below 1000 ppm needs at least 35 open minutes and loses 2.34 kilowatt-hours of heat in winter, against 7.01 with the window always open. (https://5dlibrary.com/answers/en/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c05/) 第5章 脉冲开窗与动态规划:锯齿曲线里的换气排程 — 第五章用动态规划求最省的开窗排程,两小时会议守住 1000 ppm 至少开窗 35 分钟,冬季热损 2.34 千瓦时,一直开窗则为 7.01。 (https://5dlibrary.com/answers/zh/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c05/) - Ch.6 A Concentration Is Not a Dose: Confounding, Rebreathed Air and Sensor Error — Chapter 6 shows that field data cannot separate carbon dioxide from bioeffluents, and reads 1400 ppm as a rebreathed fraction of 0.0258. (https://5dlibrary.com/answers/en/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c06/) 第6章 浓度不是剂量:混杂、再呼吸份额与传感器误差 — 第六章说明现场数据分不清二氧化碳与生物排放物,并把 1400 ppm 读作 0.0258 的再呼吸份额。 (https://5dlibrary.com/answers/zh/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c06/) - Ch.7 Classical Sources East and West: Storing Wind and Gathering Qi in the Zangshu, and the Eight Winds of Vitruvius — Chapter 7 examines "qi scatters on the wind" in the Zangshu traditionally attributed to Guo Pu and the eight winds of Vitruvius; a fully sheltered room keeps only 0.317 of the open-site flow. (https://5dlibrary.com/answers/en/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c07/) 第7章 东西方原典考辨:《葬书》的藏风聚气与维特鲁威的八风 — 第七章考辨旧题郭璞所撰《葬书》的“气乘风则散”与维特鲁威的八风,完全遮蔽的房间只剩开阔地 0.317 的流量。 (https://5dlibrary.com/answers/zh/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c07/) - Ch.8 The Empirical Loop: Studio Ventilation Schedules, Reception-Room Closing Rates and a Plan for Refutation — Chapter 8 gives the empirical loop for studios, live-stream rooms and reception rooms, the distilled Ventilation Scheduler (the demonstration live-stream room is open for 0.243 of the time) and a pre-registered 32-person paired test with power 0.801 against a medium effect; a closing-rate test would need 2754 meetings and so serves for monitoring only. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c08/) 第8章 实证闭环:工作室换气排程、会客室成交率与证伪方案 — 第八章给出工作室、直播间与会客室的实证闭环、蒸馏小工具“换气排程器”(示范直播间开窗时间占 0.243)和 32 人的预注册配对检验,中等效应下把握度为 0.801;成交率检验需要 2754 场会谈,故只作监测。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/indoor-co2-cognition-cross-ventilation-storing-wind-gathering-qi/c08/) ## 5DL-B-195 · Arrhenius Kinetics of the Maillard Reaction, Wok-Hei Combustion Aerosols and the Myth That Searing Seals in the Juices 中文书名:美拉德反应阿伦尼乌斯动力学、镬气燃烧气溶胶与煎封锁汁神话证伪母典 DOI: pending Cite: Fan, Y. (2026). Arrhenius Kinetics of the Maillard Reaction, Wok-Hei Combustion Aerosols and the Myth That Searing Seals in the Juices (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not medical advice, claims no health effect for any diet or cooking method, and is not food-safety guidance. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: Two sayings circulate in kitchens almost unquestioned: sear a steak hard first and the juices are sealed inside; and wok hei is just a hot enough wok. This book writes both as runnable arithmetic and asks when they hold. Its parameters are illustrative; all calculations are generated by its programs; it is not medical advice, claims no health effect for any diet or cooking method, and is not food-safety guidance. Chapter 1 sets out three published weighing experiments: none shows searing first reducing cooking loss, and the 1974 study gives an increase of 3.3 percentage points; an energy balance shows that during searing the pan's heat goes mainly into boiling water off. Chapter 2 uses the Arrhenius formula to show that at the 100 °C of a wet surface browning runs at only 0.0104 of its rate at 150 °C, and gives a window between browning and charring that narrows as contact time shortens. Chapter 3 uses the erfc series to compute temperature profiles and brown layers, and shows that at the same core temperature searing first for 90 seconds adds 0.50 percentage points of water loss and 150 seconds adds 2.47. Chapter 4 shows that the crust is the dry zone left by a drying front, which advances about 1.317 millimetres in 60 seconds at a dry face of 200 °C, while a water film of 0.1 millimetres delays browning by 24.7 seconds. Chapter 5 writes the combustion channel of wok hei as a model: oil droplets burn by the d-squared law, and a restaurant burner returns 8.0 times the partly burnt oil mass of a domestic ring; the channel has not been measured directly, and the book lists it as a hypothesis awaiting test. Chapter 6 shows that what matters in tossing is contact share rather than frequency, and that batch size and power can substitute for each other. Chapter 7 examines Yuan Mei's notes on heat and Brillat-Savarin's theory of frying: above a surface temperature of 211.2 °C the skin burns before the inside is cooked; after 3 minutes of frying the upper bound on the dried share of thickness is 0.510, while solutes do not dissolve in oil. Chapter 8 gives a process card keyed to core temperature (a 15-millimetre cut seared first loses 3.38 more percentage points), the standalone Heat-and-Time Calculator, and a paired probe trial with a plan to refute the book. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:厨房里有两句几乎无人怀疑的话:牛排先用大火煎焦,就能把肉汁锁在里面;镬气就是锅够热。本书把这两句话写成可以运行的算术,再看它们在什么条件下成立。书中参数为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何饮食或烹调方法的健康效应,也不构成食品安全指引。 第一章摆出三组已发表的称重实验:没有一项显示先煎降低烹调损失,1974 年的研究给出 3.3 个百分点的增加;能量平衡说明,煎的那几十秒里锅的热量主要用于把水烧成蒸汽。第二章用阿伦尼乌斯公式说明,湿表面所在的 100 °C 下褐变速率只有 150 °C 的 0.0104,并给出褐变与炭化之间随接触时间变窄的窗口。第三章用 erfc 级数计算温度剖面与褐变层,并显示同一芯温下预煎 90 秒多失水 0.50 个百分点、150 秒多 2.47 个百分点。第四章说明焦壳是干燥前沿走过的干区,干面 200 °C、60 秒约推进 1.317 毫米,而 0.1 毫米的水膜会把褐变推迟 24.7 秒。 第五章把镬气的燃烧通道写成模型:油滴按 d 平方定律燃烧,餐馆炒炉回落的部分燃烧油质量是家用灶的 8.0 倍;这一通道尚无直接测量,本书列为待检验的假说。第六章说明抛锅的关键是接触份额而非频率,批量与火力可以互相替代。第七章考辨袁枚《随园食单》的火候须知与布里亚-萨瓦兰的油炸理论:表面高于 211.2 °C 时皮先焦而里未熟;油炸 3 分钟被烧干的厚度份额上界是 0.510,溶质却不溶于油。第八章给出以芯温为准的工艺卡(15 毫米薄切预煎多失水 3.38 个百分点)、独立运行的小工具「火候时间计算器」,以及配对探针实验与证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Searing Seals in the Juices: from Liebig's Conjecture to Three Weighing Experiments — Chapter 1 sets out three published weighing experiments: none shows searing first reducing cooking loss, and the 1974 study gives an increase of 3.3 percentage points; an energy balance shows that during searing the pan's heat goes mainly into boiling water off. (https://5dlibrary.com/answers/en/maillard-arrhenius-wok-hei-aerosols-searing-myth/c01/) 第1章 煎封锁汁:从李比希的设想到三组称重实验 — 第一章摆出三组已发表的称重实验:没有一项显示先煎降低烹调损失,1974 年的研究给出 3.3 个百分点的增加;能量平衡说明,煎的那几十秒里锅的热量主要用于把水烧成蒸汽。 (https://5dlibrary.com/answers/zh/maillard-arrhenius-wok-hei-aerosols-searing-myth/c01/) - Ch.2 Arrhenius: the Temperature Ladder of Browning Rates and the Feasible Window — Chapter 2 uses the Arrhenius formula to show that at the 100 °C of a wet surface browning runs at only 0.0104 of its rate at 150 °C, and gives a window between browning and charring that narrows as contact time shortens. (https://5dlibrary.com/answers/en/maillard-arrhenius-wok-hei-aerosols-searing-myth/c02/) 第2章 阿伦尼乌斯:褐变速率的温度阶梯与可行窗口 — 第二章用阿伦尼乌斯公式说明,湿表面所在的 100 °C 下褐变速率只有 150 °C 的 0.0104,并给出褐变与炭化之间随接触时间变窄的窗口。 (https://5dlibrary.com/answers/zh/maillard-arrhenius-wok-hei-aerosols-searing-myth/c02/) - Ch.3 The erfc Series: Semi-Infinite Conduction, Brown-Layer Depth and Water Loss at the Same Core Temperature — Chapter 3 uses the erfc series to compute temperature profiles and brown layers, and shows that at the same core temperature searing first for 90 seconds adds 0.50 percentage points of water loss and 150 seconds adds 2.47. (https://5dlibrary.com/answers/en/maillard-arrhenius-wok-hei-aerosols-searing-myth/c03/) 第3章 erfc 级数:半无限体导热、褐变层厚度与同芯温下的失水 — 第三章用 erfc 级数计算温度剖面与褐变层,并显示同一芯温下预煎 90 秒多失水 0.50 个百分点、150 秒多 2.47 个百分点。 (https://5dlibrary.com/answers/zh/maillard-arrhenius-wok-hei-aerosols-searing-myth/c03/) - Ch.4 Pinned at One Hundred Degrees: the Wet Surface, the Drying Front and a Layered Steak — Chapter 4 shows that the crust is the dry zone left by a drying front, which advances about 1.317 millimetres in 60 seconds at a dry face of 200 °C, while a water film of 0.1 millimetres delays browning by 24.7 seconds. (https://5dlibrary.com/answers/en/maillard-arrhenius-wok-hei-aerosols-searing-myth/c04/) 第4章 钉在一百度:湿表面、干燥前沿与外干内湿的分层 — 第四章说明焦壳是干燥前沿走过的干区,干面 200 °C、60 秒约推进 1.317 毫米,而 0.1 毫米的水膜会把褐变推迟 24.7 秒。 (https://5dlibrary.com/answers/zh/maillard-arrhenius-wok-hei-aerosols-searing-myth/c04/) - Ch.5 The Combustion Channel of Wok Hei: the d-Squared Law and Fall-Back Aerosol — Chapter 5 writes the combustion channel of wok hei as a model: oil droplets burn by the d-squared law, and a restaurant burner returns 8.0 times the partly burnt oil mass of a domestic ring; the channel has not been measured directly, and the book lists it as a hypothesis awaiting test. (https://5dlibrary.com/answers/en/maillard-arrhenius-wok-hei-aerosols-searing-myth/c05/) 第5章 镬气的燃烧通道:油滴 d 平方定律与回落气溶胶 — 第五章把镬气的燃烧通道写成模型:油滴按 d 平方定律燃烧,餐馆炒炉回落的部分燃烧油质量是家用灶的 8.0 倍;这一通道尚无直接测量,本书列为待检验的假说。 (https://5dlibrary.com/answers/zh/maillard-arrhenius-wok-hei-aerosols-searing-myth/c05/) - Ch.6 Tossing and Firepower: Contact Share, Batch Size and Wall Recovery — Tossing works not through frequency as such but through mixing, which brings each grain's share of wall contact close to the average. In the model, with a contact share of 0.15, tossing at 1.0 to 4.0 times a second gives nearly identical results, with charring indices no higher than 0.393; doubling the share to 0.30 lifts the charring index to 5.838. Wall temperature is set jointly by power and batch size: on a 3-kilowatt stove an 800-gram batch drags the wall down to 129.6 °C, a 200-gram batch only to 182.8 °C. The first step for a home kitchen short of power is to put in less at a time and toss more evenly. On a low-powered stove, halving the batch does more than a heavier wok; wok, burner and batch should be chosen together and accepted by measured temperatures rather than by feel. (https://5dlibrary.com/answers/en/maillard-arrhenius-wok-hei-aerosols-searing-myth/c06/) 第6章 抛锅与火力:接触份额、批量与锅壁回温 — 抛锅的作用不在于频率本身,而在于混合:让每一粒米的贴壁份额接近平均值。模型中接触份额 0.15 时,每秒 1.0 到 4.0 次的抛锅结果几乎相同,炭化指数都在 0.393 以下;份额加倍到 0.30,炭化指数升到 5.838。锅壁温度由火力与批量共同决定:3 千瓦的灶下锅 800 克,锅壁最低只有 129.6 °C;下锅 200 克则是 182.8 °C。家庭厨房弥补火力不足的第一步,是每次少放、翻得更匀。在小火力的灶上,减半批量的效果胜过换一口更重的锅;而锅、灶与批量三者应当配套选择,并用测温来验收,而不是凭手感。 (https://5dlibrary.com/answers/zh/maillard-arrhenius-wok-hei-aerosols-searing-myth/c06/) - Ch.7 Classics East and West: Yuan Mei on Heat and Brillat-Savarin's Theory of Frying — Chapter 7 examines Yuan Mei's notes on heat and Brillat-Savarin's theory of frying: above a surface temperature of 211.2 °C the skin burns before the inside is cooked; after 3 minutes of frying the upper bound on the dried share of thickness is 0.510, while solutes do not dissolve in oil. (https://5dlibrary.com/answers/en/maillard-arrhenius-wok-hei-aerosols-searing-myth/c07/) 第7章 东西方原典考辨:袁枚的火候须知与布里亚-萨瓦兰的油炸理论 — 第七章考辨袁枚《随园食单》的火候须知与布里亚-萨瓦兰的油炸理论:表面高于 211.2 °C 时皮先焦而里未熟;油炸 3 分钟被烧干的厚度份额上界是 0.510,溶质却不溶于油。 (https://5dlibrary.com/answers/zh/maillard-arrhenius-wok-hei-aerosols-searing-myth/c07/) - Ch.8 Closing the Loop: Process Cards, Paired Probe Trials and a Plan to Refute — Chapter 8 gives a process card keyed to core temperature (a 15-millimetre cut seared first loses 3.38 more percentage points), the standalone Heat-and-Time Calculator, and a paired probe trial with a plan to refute the book. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/maillard-arrhenius-wok-hei-aerosols-searing-myth/c08/) 第8章 实证闭环:门店工艺卡、探针对照实验与证伪方案 — 第八章给出以芯温为准的工艺卡(15 毫米薄切预煎多失水 3.38 个百分点)、独立运行的小工具「火候时间计算器」,以及配对探针实验与证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/maillard-arrhenius-wok-hei-aerosols-searing-myth/c08/) ## 5DL-B-196 · The Failed Replication of Ego Depletion, the Ordering Artefact behind the Hungry Judges and Decision-Fatigue Scheduling 中文书名:自我损耗大规模复制失败、法官饥饿效应排序伪影与决策疲劳日程母典 DOI: pending Cite: Fan, Y. (2026). The Failed Replication of Ego Depletion, the Ordering Artefact behind the Hungry Judges and Decision-Fatigue Scheduling (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not psychological or medical advice, and its procedures and tool are non-personalised business guidance. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: "Willpower is like a muscle", "hungry judges rule more harshly", "take important decisions in the morning": behind these sayings stand two famous pieces of evidence, a meta-analysis that pooled nearly two hundred tests of ego depletion to 0.62, and the saw-tooth curve of Israeli parole boards whose grant rates collapse before meals. This book sets both against preregistered replications, tests of bias in meta-analysis and simulations of docket order, and asks how much survives. Its parameters are illustrative; all calculations are generated by its programs; the book is not psychological or medical advice, and its procedures and tool are non-personalised business guidance. Chapter 1 recomputes two preregistered multi-laboratory tests: 23 laboratories in 2016 found 0.04, 36 laboratories in 2021 found 0.06, and a rough pool gives 0.0547. Chapter 2 simulates a literature whose true effect is zero but which passes through selective publication; a random-effects meta-analysis still returns 0.286, and Egger's test, PET and PEESE detect and correct it. Chapter 3 takes the hungry judges apart into three mechanisms: lawyers-first ordering, the slope left after controlling for representation, and Glöckner's time management; an ordering artefact can have a slope 4.68 times that of genuine fatigue. Chapter 4 shows how random effects amplify the bias of small studies and measures the replication evidence with Bayes factors. Chapter 5 sets out four explanations, resource, process, belief and ordering, and the designs that can tell them apart. Chapter 6 uses the rearrangement inequality and dynamic programming to derive rules for ordering and breaks: in a small-effect world, big matters first saves only 0.0100 of total stakes. Chapter 7 examines Xunzi's "empty, unified and still" and William James's "The Energies of Men". Chapter 8 gives an owner's decision-log procedure, the distilled tool Decision Sequencer and a preregistered plan: logging alone for 120 working days detects gaps no smaller than 0.072. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:「意志力像肌肉」「法官饿了判得更严」「重要的决定放在早上」,这些说法背后站着两项著名证据:一份把近两百次检验合并出 0.62 的自我损耗元分析,以及以色列假释委员会「饭前通过率骤降」的锯齿曲线。本书把两项证据放到预注册复制、元分析偏差检验与案件排序模拟面前,看它们还剩多少。书中参数为示意;全部计算由书中程序生成;本书不是心理或医疗建议,书中流程与工具为非个人化经营建议。 第一章复算两次预注册多实验室检验:2016 年的 23 个实验室得到 0.04,2021 年的 36 个实验室得到 0.06,粗略合并为 0.0547。第二章模拟一个真实为零、经选择性发表筛选的文献,随机效应元分析仍得到 0.286,并用 Egger 检验、PET 与 PEESE 检测与校正。第三章把「饥饿法官」拆成律师优先排序、控制代理后的剩余斜率与格勒克纳的时间管理三种机制;排序伪影的斜率可达真实疲劳的 4.68 倍。第四章说明随机效应模型怎样放大小研究偏差,并以贝叶斯因子度量复制证据。 第五章列出资源、过程、信念与排序四种解释,找出能区分它们的设计。第六章用重排不等式与动态规划计算排序与休息规则:在小效应世界,「大事先做」只节省总金额的 0.0100。第七章考辨荀子《解蔽》「虚壹而静」与詹姆斯《人的能量》。第八章给出老板的决策日志流程、蒸馏小工具「决策排序器」与预注册方案:独自记录 120 个工作日能发现的最小差距是 0.072。书中以标注框区分已发表证据、可运行的 Python 计算(18 段程序)、原典与作者假说。 Chapter answers: - Ch.1 From 0.62 to 0.04: The Preregistered Replications of Ego Depletion — Chapter 1 recomputes two preregistered multi-laboratory tests: 23 laboratories in 2016 found 0.04, 36 laboratories in 2021 found 0.06, and a rough pool gives 0.0547. (https://5dlibrary.com/answers/en/ego-depletion-replication-hungry-judges-ordering-artefact/c01/) 第1章 从 0.62 到 0.04:自我损耗的预注册复制 — 第一章复算两次预注册多实验室检验:2016 年的 23 个实验室得到 0.04,2021 年的 36 个实验室得到 0.06,粗略合并为 0.0547。 (https://5dlibrary.com/answers/zh/ego-depletion-replication-hungry-judges-ordering-artefact/c01/) - Ch.2 Air in the Funnel: Selective Publication, Egger's Test and PET-PEESE — Chapter 2 simulates a literature whose true effect is zero but which passes through selective publication; a random-effects meta-analysis still returns 0.286, and Egger's test, PET and PEESE detect and correct it. (https://5dlibrary.com/answers/en/ego-depletion-replication-hungry-judges-ordering-artefact/c02/) 第2章 漏斗里的空气:选择性发表、Egger 检验与 PET-PEESE — 第二章模拟一个真实为零、经选择性发表筛选的文献,随机效应元分析仍得到 0.286,并用 Egger 检验、PET 与 PEESE 检测与校正。 (https://5dlibrary.com/answers/zh/ego-depletion-replication-hungry-judges-ordering-artefact/c02/) - Ch.3 Hungry Judges or a Sorted Docket: Three Ways to Make an Ordering Artefact — Chapter 3 takes the hungry judges apart into three mechanisms: lawyers-first ordering, the slope left after controlling for representation, and Glöckner's time management; an ordering artefact can have a slope 4.68 times that of genuine fatigue. (https://5dlibrary.com/answers/en/ego-depletion-replication-hungry-judges-ordering-artefact/c03/) 第3章 饥饿的法官还是排好的案卷:排序伪影的三种生成方式 — 第三章把「饥饿法官」拆成律师优先排序、控制代理后的剩余斜率与格勒克纳的时间管理三种机制;排序伪影的斜率可达真实疲劳的 4.68 倍。 (https://5dlibrary.com/answers/zh/ego-depletion-replication-hungry-judges-ordering-artefact/c03/) - Ch.4 The Price of Random Effects: Heterogeneity, Prediction Intervals and Bayes Factors — The random-effects model acknowledges differences between studies, but when small studies are systematically too large it magnifies the bias: in the illustrative data the fixed-effect pool is 0.140, the random-effects pool 0.383, and the weight of the large test falls from 0.694 to 0.157. The prediction interval from −0.078 to 0.844 is far more honest than the confidence interval. Bayes factors show the two replications moderately supporting the null, yet with strong prior belief in the effect the posterior probability remains 0.5565. The next chapter leaves statistical method and turns to four competing explanations of why people might tire, looking for experimental designs that can tell them apart. (https://5dlibrary.com/answers/en/ego-depletion-replication-hungry-judges-ordering-artefact/c04/) 第4章 随机效应的代价:异质性、预测区间与贝叶斯因子 — 随机效应模型承认研究之间的差异,但当小研究系统性偏大时,它会把偏差放大:示意数据中,固定效应为 0.140,随机效应为 0.383,大检验的权重从 0.694 降到 0.157。预测区间从 −0.078 到 0.844,比置信区间诚实得多。贝叶斯因子显示两次复制中等程度地支持零假设,但若事先强烈相信效应,后验概率仍有 0.5565。下一章离开统计方法,讨论四种关于「为什么会疲劳」的竞争解释,并找出能区分它们的实验设计。 (https://5dlibrary.com/answers/zh/ego-depletion-replication-hungry-judges-ordering-artefact/c04/) - Ch.5 Four Fingerprints: Resource, Process, Belief and Ordering — Resource, process, belief and ordering can all produce the same within-day decline, and on an ordinary day the spread between them is 0.000. Shuffling the order is the most discriminating design, separating the ordering artefact from the other three; an incentive design separates resource from process. Randomising the timing of breaks is an easier substitute: in the simulations of p52 it correctly estimates the pure-fatigue slope on decisions since rest at −0.0092, where the naive analysis sees only −0.0029. The next chapter assumes that a decline exists with unknown size and works out how an owner should order a day's decisions and its breaks. (https://5dlibrary.com/answers/en/ego-depletion-replication-hungry-judges-ordering-artefact/c05/) 第5章 四种解释的指纹:资源、过程、信念与排序 — 资源、过程、信念与排序四个故事可以产生完全相同的日内衰退曲线,在一个普通的日子里,它们之间的差距为 0.000。打乱顺序的设计区分力最强,能把排序伪影与其余三者分开;激励设计能区分资源与过程。随机安排休息时点是一个更易执行的替代方案:在 p52 的模拟中,它把纯疲劳世界的休息距离斜率正确估计为 −0.0092,而朴素分析只看到 −0.0029。下一章假定衰退效应存在且大小未知,计算一位老板应当怎样安排每天的决策顺序与休息。 (https://5dlibrary.com/answers/zh/ego-depletion-replication-hungry-judges-ordering-artefact/c05/) - Ch.6 A Decision-Fatigue Schedule: The Rearrangement Inequality and a Dynamic Programme for Breaks — Chapter 6 uses the rearrangement inequality and dynamic programming to derive rules for ordering and breaks: in a small-effect world, big matters first saves only 0.0100 of total stakes. (https://5dlibrary.com/answers/en/ego-depletion-replication-hungry-judges-ordering-artefact/c06/) 第6章 决策疲劳日程:重排不等式与休息的动态规划 — 第六章用重排不等式与动态规划计算排序与休息规则:在小效应世界,「大事先做」只节省总金额的 0.0100。 (https://5dlibrary.com/answers/zh/ego-depletion-replication-hungry-judges-ordering-artefact/c06/) - Ch.7 Classical Sources, East and West: Xunzi's "Empty, Unified and Still" and James's "The Energies of Men" — In the ideal state accuracy is 0.7734. The depletion effect lowers it to 0.7673, a loss of 0.0061. Carry-over of 0.25 costs 0.0070, about the same as depletion; carry-over of 0.50 costs 0.0268. A parallel task adding noise variance of 0.25 costs 0.0245, and one adding 0.50 costs 0.0435. Mood noise of variance 0.10 costs 0.0106, and of 0.40 costs 0.0365. With all three at a low level the loss is 0.0391. (https://5dlibrary.com/answers/en/ego-depletion-replication-hungry-judges-ordering-artefact/c07/) 第7章 东西方原典考辨:荀子「虚壹而静」与詹姆斯《人的能量》 — 理想状态下的正确率为 0.7734。损耗效应让正确率降到 0.7673,损失 0.0061。拖带幅度为 0.25 时损失 0.0070,与损耗相当;拖带幅度为 0.50 时损失 0.0268。一项并行任务带来方差 0.25 的噪声时,损失 0.0245;方差为 0.50 时损失 0.0435。情绪噪声方差为 0.10 时损失 0.0106,为 0.40 时损失 0.0365。三者同时处于较低水平时,损失为 0.0391。 (https://5dlibrary.com/answers/zh/ego-depletion-replication-hungry-judges-ordering-artefact/c07/) - Ch.8 Closing the Loop: An Owner's Decision Log and a Preregistered Plan for Falsification — Chapter 8 gives an owner's decision-log procedure, the distilled tool Decision Sequencer and a preregistered plan: logging alone for 120 working days detects gaps no smaller than 0.072. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/ego-depletion-replication-hungry-judges-ordering-artefact/c08/) 第8章 实证闭环:老板的决策日志与预注册证伪方案 — 第八章给出老板的决策日志流程、蒸馏小工具「决策排序器」与预注册方案:独自记录 120 个工作日能发现的最小差距是 0.072。书中以标注框区分已发表证据、可运行的 Python 计算(18 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/ego-depletion-replication-hungry-judges-ordering-artefact/c08/) ## 5DL-B-197 · Asymptotic Habit-Formation Curves, Debunking the 21-Day Myth and a Cue-Based Automaticity Index 中文书名:习惯形成渐近曲线、二十一天神话证伪与情境线索自动化指数母典 DOI: pending Cite: Fan, Y. (2026). Asymptotic Habit-Formation Curves, Debunking the 21-Day Myth and a Cue-Based Automaticity Index (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its population parameters, forgetting rates and cue indices are illustrative; all calculations are generated by its programs; it is not medical or psychological advice, nor personalised business advice. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: "Stick with it for 21 days and it becomes a habit" is one of the most widely repeated claims in self-improvement. This book defines "becoming a habit" as T95, the day an automaticity curve A(t) = a − b·e^(−ct) reaches 95 % of its plateau, and asks when the claim holds. Its population parameters, forgetting rates and cue indices are illustrative; all calculations are generated by its programs; it is not medical or psychological advice, nor personalised business advice. Chapter 1 traces the 21 days to Maltz's clinical impression of how plastic-surgery patients adjusted their mental image, which he stated as a minimum; the 2010 tracking study of Lally and colleagues found a median of 66 days, with individuals ranging from 18 to 254. In a lognormal rebuilt from its summary, only 0.027 of people arrive within 21 days, and only 0.228 even at the largest spread a later review allows. Chapter 2 derives T95 in closed form and shows that "21 days" assumes the typical person learns 3.14 times faster. Chapter 3 writes a fitter from scratch with grid search and Gauss-Newton iteration and shows that when the true T95 is 250 days, the extrapolation band is 2.16 times as wide as the truth. Chapter 4 shows that one missed day delays T95 by only two or three days, while a rule of 66 days in a row implies an average wait of 10,461.1 days at a compliance of nine in ten. Chapter 5 proposes a cue-based automaticity index and replaces days with repetitions. Chapter 6 shows that the effect of one missed day is only 0.121 of a scale's minimal detectable change, and that the average curve moves more slowly than the median person. Chapter 7 examines "by nature near, by practice far" from the Analects and Book II of Aristotle's Nicomachean Ethics. Chapter 8 sets out a flexible 66-day cycle for paid communities (cumulative coverage 0.850 after four extension blocks), the distilled tool Habit Curve Fitter and a pre-registered falsification plan (28 curves passing the fit screen, about 59 people enrolled). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:「坚持 21 天就能养成一个习惯」是流传最广的自我提升说法之一。本书把「养成」定义为自动化曲线 A(t) = a − b·e^(−ct) 到达平台 95 % 的日子 T95,再看这个说法在什么条件下成立。书中的人群参数、遗忘速率与线索指数均为示意;全部计算由书中程序生成;本书不是医疗或心理建议,也不是个人化的经营建议。 第一章追溯 21 天的来历:它源自 Maltz 对整形病人心理意象的临床印象,原话是「至少」;Lally 等人 2010 年的追踪研究给出的中位数是 66 天,个体从 18 天到 254 天。按其摘要重建的对数正态分布中,21 天内到达的人只占 0.027;把离散度放到综述允许的最大值,也只有 0.228。第二章推导 T95 的闭式解,说明「21 天」等于假设典型的人学得快 3.14 倍。第三章用网格搜索与高斯—牛顿迭代从零写出拟合器,并说明真实 T95 为 250 天时,外推区间宽达真值的 2.16 倍。第四章说明单日缺席只让 T95 推迟两到三天,而「连续 66 天」规则在九成依从度下平均要等 10,461.1 天。 第五章提出情境线索自动化指数,并把天数换成重复次数。第六章说明单日缺席的效应只有量表最小可检测变化的 0.121,平均曲线比中位人走得更慢。第七章考辨《论语·阳货》「性相近也,习相远也」与亚里士多德《尼各马可伦理学》卷二。第八章给出付费社群的 66 天弹性周期(四个延长段后累计覆盖 0.850)、蒸馏小工具「习惯曲线拟合器」与预注册证伪方案(28 条通过筛选的曲线,约 59 人招募)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Where Twenty-One Days Came From: Maltz's Minimum and Lally's Sixty-Six — Chapter 1 traces the 21 days to Maltz's clinical impression of how plastic-surgery patients adjusted their mental image, which he stated as a minimum; the 2010 tracking study of Lally and colleagues found a median of 66 days, with individuals ranging from 18 to 254. In a lognormal rebuilt from its summary, only 0.027 of people arrive within 21 days, and only 0.228 even at the largest spread a later review allows. (https://5dlibrary.com/answers/en/habit-formation-asymptote-21-day-myth-automaticity-index/c01/) 第1章 二十一天从哪里来:Maltz 的「至少」与 Lally 的六十六天 — 第一章追溯 21 天的来历:它源自 Maltz 对整形病人心理意象的临床印象,原话是「至少」;Lally 等人 2010 年的追踪研究给出的中位数是 66 天,个体从 18 天到 254 天。按其摘要重建的对数正态分布中,21 天内到达的人只占 0.027;把离散度放到综述允许的最大值,也只有 0.228。 (https://5dlibrary.com/answers/zh/habit-formation-asymptote-21-day-myth-automaticity-index/c01/) - Ch.2 The Asymptotic Curve: Plateau, Gap, Rate and T95 in Closed Form — Chapter 2 derives T95 in closed form and shows that "21 days" assumes the typical person learns 3.14 times faster. (https://5dlibrary.com/answers/en/habit-formation-asymptote-21-day-myth-automaticity-index/c02/) 第2章 渐近曲线:平台、差距、速率与 T95 的闭式解 — 第二章推导 T95 的闭式解,说明「21 天」等于假设典型的人学得快 3.14 倍。 (https://5dlibrary.com/answers/zh/habit-formation-asymptote-21-day-myth-automaticity-index/c02/) - Ch.3 A Standard-Library Fitter: Grid Search, Gauss-Newton, Screening Bias and Extrapolation Error — Chapter 3 writes a fitter from scratch with grid search and Gauss-Newton iteration and shows that when the true T95 is 250 days, the extrapolation band is 2.16 times as wide as the truth. (https://5dlibrary.com/answers/en/habit-formation-asymptote-21-day-myth-automaticity-index/c03/) 第3章 纯标准库拟合:网格搜索、高斯—牛顿、筛选偏差与外推误差 — 第三章用网格搜索与高斯—牛顿迭代从零写出拟合器,并说明真实 T95 为 250 天时,外推区间宽达真值的 2.16 倍。 (https://5dlibrary.com/answers/zh/habit-formation-asymptote-21-day-myth-automaticity-index/c03/) - Ch.4 The Missed Day: Absence Shocks, the Dose of Compliance and the Cost of Streak Resets — Chapter 4 shows that one missed day delays T95 by only two or three days, while a rule of 66 days in a row implies an average wait of 10,461.1 days at a compliance of nine in ten. (https://5dlibrary.com/answers/en/habit-formation-asymptote-21-day-myth-automaticity-index/c04/) 第4章 漏做一天:缺席冲击、依从剂量与「断签清零」规则的代价 — 第四章说明单日缺席只让 T95 推迟两到三天,而「连续 66 天」规则在九成依从度下平均要等 10,461.1 天。 (https://5dlibrary.com/answers/zh/habit-formation-asymptote-21-day-myth-automaticity-index/c04/) - Ch.5 A Cue-Based Automaticity Index: Cue Entropy, Time of Day and a Repetition Clock — A habit is a bond between context and response, and the more stable the context, the faster the bond forms. This chapter defined a cue-based automaticity index from the Shannon entropy of preceding events and time slots: it is 1.000 with a single cue and falls to 0.145 when cues are random, which under the illustrative conversion stretches T95 from 59.0 to 121.2 days. In empirical work, the gap between morning and evening performance was 105.95 days against 154.01. Switching the axis from days to repetitions, an act done 6 times a day reaches its plateau in 11 days and one done twice a week takes 184. For operators, the lowest-cost way to speed things up is not more polished content but helping each member pick one fixed preceding event on the first day, and scheduling the cycle in repetitions rather than calendar days. The next chapter turns to measurement itself: how small a change a scale can see, and why the average curve misleads. (https://5dlibrary.com/answers/en/habit-formation-asymptote-21-day-myth-automaticity-index/c05/) 第5章 情境线索自动化指数:线索熵、时段与重复次数时钟 — 习惯是情境与反应之间的联结,情境越稳定,联结形成越快。本章用前置事件与时段的香农熵定义了情境线索自动化指数:只有一个线索时指数为 1.000,完全随机时降到 0.145,在示意换算下 T95 从 59.0 天延长到 121.2 天。早上与晚上执行的差距,在实证研究中是 105.95 天对 154.01 天。把横轴从天数换成重复次数,一天 6 次的行为 11 天就到达平台,一周 2 次的行为要 184 天。对经营者而言,成本最低的加速手段不是延长打磨内容,而是在开营第一天帮每个成员选定一个固定的前置事件,并按重复次数而不是日历天数安排周期。下一章讨论测量本身:量表能看见多小的变化,以及为什么平均曲线会误导。 (https://5dlibrary.com/answers/zh/habit-formation-asymptote-21-day-myth-automaticity-index/c05/) - Ch.6 Scales, Noise and the Average Curve: SRHI, SRBAI and the Trap of Right Skew — Chapter 6 shows that the effect of one missed day is only 0.121 of a scale's minimal detectable change, and that the average curve moves more slowly than the median person. (https://5dlibrary.com/answers/en/habit-formation-asymptote-21-day-myth-automaticity-index/c06/) 第6章 量表、噪声与平均曲线:SRHI、SRBAI 与右偏分布的陷阱 — 第六章说明单日缺席的效应只有量表最小可检测变化的 0.121,平均曲线比中位人走得更慢。 (https://5dlibrary.com/answers/zh/habit-formation-asymptote-21-day-myth-automaticity-index/c06/) - Ch.7 East and West in the Classics: "By Nature Near, by Practice Far" and Aristotle on Hexis — The Analects says that by nature people are near and by practice far apart. The model shows that on the same act, differences in rate pull people furthest apart around day 21, with a standard deviation of 5.68 points, narrowing to 0.16 after a year, while the difference between doing and not doing widens to 14.99. Aristotle says habit perfects a capacity given by nature, which matches the distinction between plateau and rate; his stone will not rise even if thrown ten thousand times, just as a curve with a zero plateau stays at zero. "Habits from youth make all the difference" is capped in the model: once an old habit has been kept for more than half a year, a new habit overtakes it around day 40. The final chapter turns the conclusions of the book into an empirical loop that can be carried out and can also be refuted. (https://5dlibrary.com/answers/en/habit-formation-asymptote-21-day-myth-automaticity-index/c07/) 第7章 东西方原典考辨:「性相近,习相远」与亚里士多德的习性 — 《论语》说「性相近也,习相远也」,模型显示:在同一件事上,速率差别让人在第 21 天前后拉开最大距离,标准差 5.68 分,一年后收窄到 0.16 分;做与不做的差别则持续扩大到 14.99 分。亚里士多德说习惯完成自然所赋予的能力,这与平台和速率的区分一致;他的石头即使抛一万次也不会向上,正如平台为零的曲线永远是零。「从小养成关系到一切」在模型中有上限:旧习惯保持半年以上,新习惯都在第 40 天前后超过它。最后一章把全书的结论变成一个可以执行、也可以被推翻的实证闭环。 (https://5dlibrary.com/answers/zh/habit-formation-asymptote-21-day-myth-automaticity-index/c07/) - Ch.8 Closing the Loop: a Flexible Sixty-Six-Day Cycle, the Habit Curve Fitter and a Pre-Registered Plan — Chapter 8 sets out a flexible 66-day cycle for paid communities (cumulative coverage 0.850 after four extension blocks), the distilled tool Habit Curve Fitter and a pre-registered falsification plan (28 curves passing the fit screen, about 59 people enrolled). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/habit-formation-asymptote-21-day-myth-automaticity-index/c08/) 第8章 实证闭环:六十六天弹性周期、习惯曲线拟合器与预注册方案 — 第八章给出付费社群的 66 天弹性周期(四个延长段后累计覆盖 0.850)、蒸馏小工具「习惯曲线拟合器」与预注册证伪方案(28 条通过筛选的曲线,约 59 人招募)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/habit-formation-asymptote-21-day-myth-automaticity-index/c08/) ## 5DL-B-198 · The Income-Wellbeing Satiation Debate, Hedonic Adaptation Rates and the Experiential Depreciation of Luxury Consumption 中文书名:收入幸福曲线饱和点争议、享乐适应速率与奢侈消费体验折旧母典 DOI: pending Cite: Fan, Y. (2026). The Income-Wellbeing Satiation Debate, Hedonic Adaptation Rates and the Experiential Depreciation of Luxury Consumption (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs. It is not personalised business advice, nor financial or investment advice. Abstract: "Past an income of 75,000 dollars a year, money buys no more happiness" is a widely repeated slogan. Checking the original studies, this book finds that most people's experienced happiness keeps rising with log income, that a plateau appears only among the least happy fifth or so of people and only above 100,000 dollars, and that even this residual plateau has been challenged. The book also examines a second slogan, "buy experiences, not things", by writing hedonic adaptation as a half-life that can be estimated and refuted. Its parameters are illustrative; all calculations are generated by its programs. It is not personalised business advice, nor financial or investment advice. Chapter 1 shows that under the log-linear reading 0.5356 of the happiness gained between 15,000 and 480,000 dollars accrues above the supposed plateau, and that a yes-or-no question can create a plateau through its ceiling. Chapter 2 explains quantile regression: one quantile plot can hide a world in which the initially least happy change by 0.837 on average, and when the upper tail is thin a plateau that does not exist appears in 0.117 of simulations. Chapter 3 prices happiness in doublings: the same ten thousand dollars weighs 14.0 times as much for a low earner as for a high earner. Chapter 4 builds an exponential baseline for hedonic adaptation and gives a closed formula, with its range of use, for the half-life from three visits on days 0, 30 and 90. Chapter 5 sets the evidence and counter-evidence on buying experiences side by side: good and bad outcomes, frequency and intensity, and customers' resources. Chapter 6 books satisfaction as a depreciating asset; under illustrative parameters a trip buys 1.86 times as many satisfaction point-days per unit of money as a handbag, and a follow-up ritual every 30 days nets 121.2 point-days. Chapter 7 examines the Zhuangzi's "Utmost Joy" and Epicurus's Letter to Menoeceus: the Marquis of Lu's seabird becomes a mismatch model in which 0.599 of customers are less satisfied after a "spectacle upgrade", and occasional plenty brings a pleasure of 1.052. Chapter 8 gives an experiential redesign, tracking at 30 and 90 days, the distilled tool called the satisfaction half-life tracker and a pre-registered protocol: with 300 enrolled per group a difference of 0.434 points can be detected, and the simulated pilot shows a difference in falls of -0.478. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:「年收入过了 75,000 美元,钱就买不来更多快乐」是一句流传很广的口号。本书核对原始文献后指出:多数人的经历幸福随对数收入持续上升,平台只出现在最不快乐的约两成人身上、并且在 100,000 美元以上;而这一残余的平台本身也受到质疑。本书同时检视另一句口号「买体验,不买东西」,把享乐适应写成可以估计、可以被否定的半衰期。书中参数为示意;全部计算由书中程序生成。本书不是个人化的经营建议,也不是理财或投资建议。 第一章说明,在对数线性读法下,从 15,000 美元到 480,000 美元的幸福增量中有 0.5356 发生在所谓的平台以上,而一道是否题就能凭天花板制造平台。第二章讲分位数回归:同一张分位数图,背后可以是起初最不快乐者平均变化 0.837 的世界;高收入端稀薄时,一个并不存在的平台会在 0.117 的模拟中出现。第三章按翻倍计价:同样一万美元,对低收入者的分量是高收入者的 14.0 倍。第四章建立享乐适应的指数基线,并给出由第 0、30、90 天三次回访解出半衰期的封闭公式及其适用范围。 第五章并陈「买体验」的证据与反证:结果好坏、频率与强度、顾客资源。第六章把满足当作会折旧的资产来记账,示意参数下旅行每元钱买到的满意度点日是手袋的 1.86 倍,每 30 天一次的回访仪式净得 121.2 个点日。第七章考辨《庄子·至乐》与伊壁鸠鲁《致美诺寇书》:鲁侯养鸟写成错配模型,0.599 的顾客在「排场升级」后更不满意;偶尔的丰盛带来 1.052 的快乐。第八章给出体验化改造、30 / 90 天追踪、蒸馏小工具「满意度半衰期追踪表」与预注册方案:每组入组 300 人时可发现 0.434 分的差别,模拟试点的下降量之差为 -0.478。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The 75,000-Dollar Plateau: Two Measures and an Adversarial Collaboration — Chapter 1 shows that under the log-linear reading 0.5356 of the happiness gained between 15,000 and 480,000 dollars accrues above the supposed plateau, and that a yes-or-no question can create a plateau through its ceiling. (https://5dlibrary.com/answers/en/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c01/) 第1章 七万五千美元的平台:两种测量与一场对抗性合作 — 第一章说明,在对数线性读法下,从 15,000 美元到 480,000 美元的幸福增量中有 0.5356 发生在所谓的平台以上,而一道是否题就能凭天花板制造平台。 (https://5dlibrary.com/answers/zh/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c01/) - Ch.2 Quantile Regression: Who Levels Off, Who Accelerates, Who Changes Places — Chapter 2 explains quantile regression: one quantile plot can hide a world in which the initially least happy change by 0.837 on average, and when the upper tail is thin a plateau that does not exist appears in 0.117 of simulations. (https://5dlibrary.com/answers/en/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c02/) 第2章 分位数回归:谁走平,谁加速,谁换了名次 — 第二章讲分位数回归:同一张分位数图,背后可以是起初最不快乐者平均变化 0.837 的世界;高收入端稀薄时,一个并不存在的平台会在 0.117 的模拟中出现。 (https://5dlibrary.com/answers/zh/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c02/) - Ch.3 Pricing in Doublings: Log Income, Relative Income and the Easterlin Paradox — Chapter 3 prices happiness in doublings: the same ten thousand dollars weighs 14.0 times as much for a low earner as for a high earner. (https://5dlibrary.com/answers/en/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c03/) 第3章 按翻倍计价:对数收入、相对收入与伊斯特林悖论 — 第三章按翻倍计价:同样一万美元,对低收入者的分量是高收入者的 14.0 倍。 (https://5dlibrary.com/answers/zh/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c03/) - Ch.4 An Exponential Baseline for Hedonic Adaptation: Estimating Half-Lives and Their Identifiability — Chapter 4 builds an exponential baseline for hedonic adaptation and gives a closed formula, with its range of use, for the half-life from three visits on days 0, 30 and 90. (https://5dlibrary.com/answers/en/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c04/) 第4章 享乐适应的指数基线:半衰期的估计与可识别性 — 第四章建立享乐适应的指数基线,并给出由第 0、30、90 天三次回访解出半衰期的封闭公式及其适用范围。 (https://5dlibrary.com/answers/zh/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c04/) - Ch.5 Goods and Experiences: Frequency, Intensity and the Constraint of Resources — In the first fourteen days the good gives 42.0 happy moments and the experience only 18.6, but the experience yields 30.3 units against the good's 21.3, since each experiential moment has an intensity of 1.629 against 0.507. By day 42 the good leads in moments by 126.0 to 35.4 and has overtaken in units as well, by 47.2 to 43.7; the program puts the crossing on day 31. Over a year the good accumulates 1095.0 moments and 155.7 units and the experience 229.2 moments and 86.9 units, while each experiential moment is still more than twice as intense as a material one, 0.379 against 0.142. (https://5dlibrary.com/answers/en/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c05/) 第5章 实物与体验:频率、强度与资源约束 — 前十四天里,实物带来 42.0 个快乐时刻,体验只有 18.6 个,但体验的快乐单位是 30.3,高于实物的 21.3:每个体验时刻的强度是 1.629,实物只有 0.507。到第 42 天,实物的时刻数已是 126.0 对 35.4,快乐单位也以 47.2 对 43.7 反超。程序找出的交叉日是第 31 天。一年下来,实物累计 1095.0 个时刻、155.7 个快乐单位,体验是 229.2 个时刻、86.9 个快乐单位;而每个体验时刻的强度仍是实物的两倍多,0.379 对 0.142。 (https://5dlibrary.com/answers/zh/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c05/) - Ch.6 Hedonic Depreciation Accounting: Point-Days, Rituals and the Peak-End Rule — Chapter 6 books satisfaction as a depreciating asset; under illustrative parameters a trip buys 1.86 times as many satisfaction point-days per unit of money as a handbag, and a follow-up ritual every 30 days nets 121.2 point-days. (https://5dlibrary.com/answers/en/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c06/) 第6章 享乐折旧会计:点日、仪式与峰终 — 第六章把满足当作会折旧的资产来记账,示意参数下旅行每元钱买到的满意度点日是手袋的 1.86 倍,每 30 天一次的回访仪式净得 121.2 个点日。 (https://5dlibrary.com/answers/zh/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c06/) - Ch.7 Sources East and West: the Zhuangzi's "Utmost Joy" and Epicurus's Letter to Menoeceus — Chapter 7 examines the Zhuangzi's "Utmost Joy" and Epicurus's Letter to Menoeceus: the Marquis of Lu's seabird becomes a mismatch model in which 0.599 of customers are less satisfied after a "spectacle upgrade", and occasional plenty brings a pleasure of 1.052. (https://5dlibrary.com/answers/en/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c07/) 第7章 东西方原典考辨:《庄子·至乐》与伊壁鸠鲁《致美诺寇书》 — 第七章考辨《庄子·至乐》与伊壁鸠鲁《致美诺寇书》:鲁侯养鸟写成错配模型,0.599 的顾客在「排场升级」后更不满意;偶尔的丰盛带来 1.052 的快乐。 (https://5dlibrary.com/answers/zh/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c07/) - Ch.8 Closing the Loop: Experiential Redesign and 30- and 90-Day Tracking — Chapter 8 gives an experiential redesign, tracking at 30 and 90 days, the distilled tool called the satisfaction half-life tracker and a pre-registered protocol: with 300 enrolled per group a difference of 0.434 points can be detected, and the simulated pilot shows a difference in falls of -0.478. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c08/) 第8章 实证闭环:体验化改造与 30 / 90 天追踪 — 第八章给出体验化改造、30 / 90 天追踪、蒸馏小工具「满意度半衰期追踪表」与预注册方案:每组入组 300 人时可发现 0.434 分的差别,模拟试点的下降量之差为 -0.478。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/income-wellbeing-satiation-hedonic-adaptation-luxury-depreciation/c08/) ## 5DL-B-199 · Hyperbolic-Discounting Preference Reversals, Deadline Commitment Devices and the Temporal Motivation Equation 中文书名:双曲贴现偏好反转、截止期承诺装置与时间动机方程母典 DOI: pending Cite: Fan, Y. (2026). Hyperbolic-Discounting Preference Reversals, Deadline Commitment Devices and the Temporal Motivation Equation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not individual business, legal or financial advice. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: Is "I will organise it myself" a systematic overestimate? Do imposed, evenly spaced deadlines always beat deadlines people set for themselves? This book was commissioned on a famous proofreading experiment from 2002, but checking the source before writing showed that the paper was retracted by its journal in September 2026 and that a preregistered replication found no deadline effect. The book therefore first falsifies its own commissioning claim and then uses hyperbolic discounting, the temporal motivation equation and the beta-delta model of present bias to ask again when a deadline is worth its price. Its parameters are illustrative; all calculations are generated by its programs; it is not individual business, legal or financial advice. Chapter 1 puts the original and the replication on one scale: the replication's ANOVA statistic for errors found is only 0.181, and the largest gap between groups compatible with it is 0.132 standard deviations; even a model that favours deadlines never pushes the score ratio of the single-deadline arm to the evenly spaced arm below 0.885, against 0.522 in the original paper. Chapter 2 derives the reversal point and shows that in a population of varied impatience, 0.4676 of those who take the smaller-sooner reward today would choose the opposite when planning a month ahead. Chapter 3 works through the temporal motivation equation and shows that interim checkpoints without stakes reduce rather than increase the number of motivated days. Chapter 4 shows that self-set deadlines fall short mainly through naivete about surprises rather than about procrastination. Chapter 5 finds almost the same overestimate in theses and gym contracts, 1.637 and 1.628, and shows that 80 % confidence of finishing on time requires multiplying one's own estimate by 2.690. Chapter 6 prices a deadline, through lost option value and a floor under late fees, and shows that hard interim deadlines lower completion in the face of surprises. Chapter 7 weighs Qian Fu's "Song of Tomorrow" and Seneca's On the Shortness of Life. Chapter 8 sets out milestone scheduling for a writing window, due dates for client instalments, a preregistered test and the distilled tool, the Deadline Scheduler. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:「我自己会安排好」是不是一种系统性高估?外部强制的等距截止期是否总比自设截止期更好?本书立项时依据的是 2002 年一项著名的校对实验,可是落笔前的核实发现,这篇论文已于 2026 年 9 月被期刊撤回,其预注册重复实验没有找到截止期效应。本书因此先证伪自己的立项命题,再用双曲贴现、时间动机方程与带现时偏好的 beta-delta 模型,重新回答截止期在什么条件下值得付出代价。书中参数为示意;全部计算由书中程序生成;本书不是个人化的经营、法律或理财建议。 第一章把原实验与重复实验放在同一把尺子上:重复实验找错数的方差分析统计量只有 0.181,与之相容的最大组间差距是 0.132 个标准差;即使模型偏向截止期,单一期限组与等距组的得分比最低也只到 0.885,而原论文报告的是 0.522。第二章推导偏好反转点,并说明在一个急躁程度各异的人群里,今天选择较小较早奖赏的人中有 0.4676 会在提前一个月计划时作出相反选择。第三章展开时间动机方程,说明没有赌注的中间节点会减少而不是增加有动机的天数。第四章说明自设截止期吃亏主要源于对意外的天真,而不是对拖延的天真。 第五章用毕业论文与健身房合同两组数据求出几乎相同的高估幅度 1.637 与 1.628,要有八成把握按期完成需要把自己的估计乘以 2.690。第六章给截止期定价:选择权的损失与迟交费的下限,并说明硬性中间截止期在意外面前会降低完成率。第七章考辨钱福《明日歌》与塞涅卡《论生命之短暂》。第八章给出写书窗口的里程碑排程、客户分期截止设计、预注册检验方案与蒸馏小工具「截止期排程器」。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A Retracted Deadline Experiment: the Evidence For and Against Evenly Spaced Deadlines — Chapter 1 puts the original and the replication on one scale: the replication's ANOVA statistic for errors found is only 0.181, and the largest gap between groups compatible with it is 0.132 standard deviations; even a model that favours deadlines never pushes the score ratio of the single-deadline arm to the evenly spaced arm below 0.885, against 0.522 in the original paper. (https://5dlibrary.com/answers/en/hyperbolic-discounting-deadline-commitment-temporal-motivation/c01/) 第1章 一项被撤回的截止期实验:等距截止期优势的证据与反证 — 第一章把原实验与重复实验放在同一把尺子上:重复实验找错数的方差分析统计量只有 0.181,与之相容的最大组间差距是 0.132 个标准差;即使模型偏向截止期,单一期限组与等距组的得分比最低也只到 0.885,而原论文报告的是 0.522。 (https://5dlibrary.com/answers/zh/hyperbolic-discounting-deadline-commitment-temporal-motivation/c01/) - Ch.2 Hyperbolic Discounting and Preference Reversal: Where 1/(1+kD) Crosses Over — Chapter 2 derives the reversal point and shows that in a population of varied impatience, 0.4676 of those who take the smaller-sooner reward today would choose the opposite when planning a month ahead. (https://5dlibrary.com/answers/en/hyperbolic-discounting-deadline-commitment-temporal-motivation/c02/) 第2章 双曲贴现与偏好反转:1/(1+kD) 的交叉点 — 第二章推导偏好反转点,并说明在一个急躁程度各异的人群里,今天选择较小较早奖赏的人中有 0.4676 会在提前一个月计划时作出相反选择。 (https://5dlibrary.com/answers/zh/hyperbolic-discounting-deadline-commitment-temporal-motivation/c02/) - Ch.3 The Temporal Motivation Equation: Four Levers of Expectancy, Value, Impulsiveness and Delay — The motivation curve already tells the story. Twenty-one days before the deadline motivation is only 0.364; at 14 days it is 0.533; at 7 days it is 1.000, exactly level with leisure; at 3 days 2.000; on the last day 8.000. The curve rises steeply at the end. This is the last-minute rush as the equation sees it. (https://5dlibrary.com/answers/en/hyperbolic-discounting-deadline-commitment-temporal-motivation/c03/) 第3章 时间动机方程:期望、价值、冲动与延迟的四根杠杆 — 动机曲线本身先说明了问题。距截止期 21 天时动机值只有 0.364,14 天时 0.533,7 天时 1.000 恰好与休闲持平,3 天时 2.000,最后一天 8.000。曲线在末端陡峭上升,这就是「截止日前的冲刺」在方程里的样子。 (https://5dlibrary.com/answers/zh/hyperbolic-discounting-deadline-commitment-temporal-motivation/c03/) - Ch.4 Naive, Partly Naive and Sophisticated: Who Sets Deadlines for Themselves — In the cost case, people with beta of 1.0 or 0.9 finish on day 1 at a cost of 4.0. At beta 0.7 the three types part ways. The naive and the partly naive believe each day that they will act tomorrow and drag the task to day 4, at a cost of 9.0; the sophisticate foresees that his day-3 self would put it off once more and finishes on day 2 at a cost of 5.0. At beta 0.5 the gap widens: the naif still waits until day 4, while the sophisticate finishes on day 1 at 4.0, because he knows that once today is missed every later self will keep postponing. (https://5dlibrary.com/answers/en/hyperbolic-discounting-deadline-commitment-temporal-motivation/c04/) 第4章 天真、半天真与老练:谁会给自己定截止期 — 在成本情形下,beta 为 1.0 或 0.9 的人都在第 1 天以成本 4.0 完成。beta 降到 0.7 时,三种人分道扬镳:天真与半天真的人每天都以为明天会做,一直拖到第 4 天以成本 9.0 完成;老练的人预见到第 3 天的自己会再拖一天,于是在第 2 天以成本 5.0 完成。beta 为 0.5 时差距更大:天真的人仍拖到第 4 天,老练的人则在第 1 天就以 4.0 完成,因为他知道一旦错过今天,接下来的每一个自己都会继续推。 (https://5dlibrary.com/answers/zh/hyperbolic-discounting-deadline-commitment-temporal-motivation/c04/) - Ch.5 Sizing the Overestimate in "I Will Organise It Myself": the Planning Fallacy and the Gym Contract — Chapter 5 finds almost the same overestimate in theses and gym contracts, 1.637 and 1.628, and shows that 80 % confidence of finishing on time requires multiplying one's own estimate by 2.690. (https://5dlibrary.com/answers/en/hyperbolic-discounting-deadline-commitment-temporal-motivation/c05/) 第5章 「我自己会安排好」的高估幅度:规划谬误与健身房合同 — 第五章用毕业论文与健身房合同两组数据求出几乎相同的高估幅度 1.637 与 1.628,要有八成把握按期完成需要把自己的估计乘以 2.690。 (https://5dlibrary.com/answers/zh/hyperbolic-discounting-deadline-commitment-temporal-motivation/c05/) - Ch.6 Two Prices of a Deadline: Lost Option Value and the Minimum Late Fee — A deadline has two prices. The first is lost option value, whose curve is steep at first and then flat: keeping two or three spare days preserves most of the value of waiting, and this price falls mainly on time-consistent people. The second is the late fee needed to make a soft deadline bite, with a floor of $c(1-\beta)/\beta$ that rises steeply as beta falls. Hard interim deadlines lower procrastinators' completion whenever surprises occur, and the more surprises, the larger the loss. Designing deadlines means trading these two prices against each other. (https://5dlibrary.com/answers/en/hyperbolic-discounting-deadline-commitment-temporal-motivation/c06/) 第6章 截止期的两种价格:选择权的损失与迟交费的下限 — 截止期有两种价格。第一种是选择权的损失,它的曲线前陡后平:保留两三个空余天就能留住大部分等待价值,而这种价格主要由时间一致的人承担。第二种是让软截止期生效的迟交费,下限为 $c(1-\beta)/\beta$,随 beta 下降急剧上升。硬性的中间截止期在意外存在时会降低拖延者的完成率,意外越多降得越多。设计截止期,就是在这两种价格之间作取舍。 (https://5dlibrary.com/answers/zh/hyperbolic-discounting-deadline-commitment-temporal-motivation/c06/) - Ch.7 Classical Sources East and West: the Endless Recursion of the "Song of Tomorrow" and Seneca's Ledger of Time — The first six lines of the "Song of Tomorrow" can be written as a theorem: a naif keeps postponing as long as the daily loss from delay is below $c(1-\beta)/\beta$, whatever the value of the matter. The poem is right about the recursion and the form of the cost but overstates "every matter" and fails to separate reasonable waiting from unreasonable. Seneca's ledger of time is right about the asymmetry between time and money and about the role of expectation, but it ignores the option value of waiting and the time a person does not control. Both sources stop at exhortation; a deadline turns exhortation into a loss that can be seen every day. (https://5dlibrary.com/answers/en/hyperbolic-discounting-deadline-commitment-temporal-motivation/c07/) 第7章 东西方原典考辨:《明日歌》的无穷递推与塞涅卡的时间账簿 — 《明日歌》的前六句可以写成一条定理:天真的人只要每天的拖延损失低于 $c(1-\beta)/\beta$ 就会一直推迟,与事情的价值无关;它说对了递推与代价的形式,却把「万事」说过了头,也没有区分合理的等待。塞涅卡的时间账簿说对了时间与金钱的不对称和对「期待」的判断,却忽视了等待的选择权与不由自己支配的时间。两部原典都止于劝诫,而截止期的作用,是把劝诫变成每天可见的损失。 (https://5dlibrary.com/answers/zh/hyperbolic-discounting-deadline-commitment-temporal-motivation/c07/) - Ch.8 Closing the Loop: Writing Milestones, Client Instalment Deadlines and a Preregistered Test — Chapter 8 sets out milestone scheduling for a writing window, due dates for client instalments, a preregistered test and the distilled tool, the Deadline Scheduler. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/hyperbolic-discounting-deadline-commitment-temporal-motivation/c08/) 第8章 实证闭环:写书里程碑、客户分期截止与预注册检验 — 第八章给出写书窗口的里程碑排程、客户分期截止设计、预注册检验方案与蒸馏小工具「截止期排程器」。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/hyperbolic-discounting-deadline-commitment-temporal-motivation/c08/) ## 5DL-B-200 · Falsifying the Catharsis Hypothesis: Two-Factor Arousal Decay and the Optimal Cooling-Off Period in Negotiation 中文书名:愤怒宣泄假说证伪、情绪唤醒双因素衰减与谈判冷却期最优长度母典 DOI: pending Cite: Fan, Y. (2026). Falsifying the Catharsis Hypothesis: Two-Factor Arousal Decay and the Optimal Cooling-Off Period in Negotiation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not psychological, medical or legal advice, nor personalised business advice. Chapter 4 computes the excitation-transfer window from Zillmann's two clocks, peaking at 10.75 minutes under illustrative parameters, and reports the controversy over two-factor theory. Abstract: "Let it out or it will eat you up" is among the most widespread beliefs about emotion. This book writes it as a prediction that can lose: anger and aggression after venting should be lower than without venting. Randomized experiments give the opposite answer, and the book then uses three kinds of mathematics, first-order decay, excitation transfer and optimal stopping, to show that what works is bringing arousal down and waiting, and how long to wait. Its parameters are illustrative; all calculations are generated by its programs; the book is not psychological, medical or legal advice, nor personalised business advice. Chapter 1 recomputes Table 1 of Bushman (2002): hitting a bag while dwelling on the provoker has an aggression effect of 0.298, with the whole interval above zero, and the meta-analysis of Kjærvik and Bushman (2024) finds an effect of −0.63 for arousal-lowering activities. Chapter 2 compares a hydraulic and a rehearsal model by isotonic regression, with a likelihood ratio of 164.5 on anger, and sets out the conditions under which venting might seem to work, with the counter-evidence. Chapter 3 uses first-order decay to compute the wait until a threshold: 16.9 minutes of plain waiting for someone provoked to 7 with a 20-minute time constant, and treats repeated provocations with Campbell's theorem. Chapter 4 computes the excitation-transfer window from Zillmann's two clocks, peaking at 10.75 minutes under illustrative parameters, and reports the controversy over two-factor theory. Chapter 5 calibrates acceptance to delay experiments in the ultimatum game; at moderate withdrawal risk the best pause is 14.7 minutes, worth 3.09 times deciding at once. Chapter 6 reports the value of anger as a signal, argues for selective display after calming down, and prices impatience: doubling it costs 0.1665 of the share. Chapter 7 examines the Zhongyong's "hitting the measure" and Seneca's delay in On Anger; expressing at once after a provocation hits the measure with probability 0.011, rising to 0.752 after 20 minutes. Chapter 8 gives complaint cooling rules (acknowledge first, with a best substantive delay of about 35.5 minutes), an owner's emotion-decision log, the Cooling-Off Calculator and a pre-registered test (176 events per group). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:「气要撒出来才不憋坏」是流传最广的情绪常识之一。本书把它写成一个会输的预测:宣泄之后的愤怒与攻击应当低于不宣泄。随机对照实验给出相反的答案,本书再用一阶衰减、兴奋转移与最优停时三类数学,说明真正有效的是把唤醒降下来、等一等,以及该等多久。书中参数为示意;全部计算由书中程序生成;本书不是心理、医疗或法律意见,也不是个人化经营建议。 第一章重算 Bushman 2002 的表 1:边打沙袋边想着冒犯者,攻击效应量为 0.298,整段区间在零之上;Kjærvik 与 Bushman 2024 的元分析显示降唤醒活动的效应为 −0.63。第二章以保序回归比较液压模型与复演模型,愤怒上的似然比为 164.5,并给出宣泄可能看起来有效的条件与反方证据。第三章用一阶衰减计算等到阈值的时间,被激怒到 7 分、时间常数 20 分钟时只是等需 16.9 分钟,并用坎贝尔定理处理接连不断的刺激。第四章用 Zillmann 的两只时钟计算兴奋转移窗口,示意峰值在 10.75 分钟,并并陈双因素理论的争议。 第五章用最后通牒博弈的延迟实验校准接受率,中等撤回风险下最优冷却 14.7 分钟,期望值是立刻决定的 3.09 倍。第六章并陈愤怒作为信号的价值,说明应在冷静之后有选择地表露,并算出急躁的价格:急躁倍数加倍损失 0.1665 的份额。第七章考辨《中庸》的「发而皆中节」与塞涅卡《论愤怒》的拖延,被激怒后立刻表达的中节概率 0.011,等 20 分钟升到 0.752。第八章给出客诉冷却规则(先确认,实质回复最优延迟约 35.5 分钟)、店主情绪—决策日志、冷却期计算器与预注册方案(每组 176 个事件)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Punching-Bag Ledger: Recomputing Bushman (2002) and the Arousal Meta-Analysis — Chapter 1 recomputes Table 1 of Bushman (2002): hitting a bag while dwelling on the provoker has an aggression effect of 0.298, with the whole interval above zero, and the meta-analysis of Kjærvik and Bushman (2024) finds an effect of −0.63 for arousal-lowering activities. (https://5dlibrary.com/answers/en/catharsis-falsified-arousal-decay-negotiation-cooling-off/c01/) 第1章 沙袋账本:重算 Bushman 2002 与降唤醒元分析 — 第一章重算 Bushman 2002 的表 1:边打沙袋边想着冒犯者,攻击效应量为 0.298,整段区间在零之上;Kjærvik 与 Bushman 2024 的元分析显示降唤醒活动的效应为 −0.63。 (https://5dlibrary.com/answers/zh/catharsis-falsified-arousal-decay-negotiation-cooling-off/c01/) - Ch.2 Hydraulics versus Rehearsal: an Order-Restricted Test of Two Venting Models — Chapter 2 compares a hydraulic and a rehearsal model by isotonic regression, with a likelihood ratio of 164.5 on anger, and sets out the conditions under which venting might seem to work, with the counter-evidence. (https://5dlibrary.com/answers/en/catharsis-falsified-arousal-decay-negotiation-cooling-off/c02/) 第2章 液压与复演:两种宣泄模型的保序检验 — 第二章以保序回归比较液压模型与复演模型,愤怒上的似然比为 164.5,并给出宣泄可能看起来有效的条件与反方证据。 (https://5dlibrary.com/answers/zh/catharsis-falsified-arousal-decay-negotiation-cooling-off/c02/) - Ch.3 First-Order Decay: Time Constants, Thresholds and Shot Noise — Chapter 3 uses first-order decay to compute the wait until a threshold: 16.9 minutes of plain waiting for someone provoked to 7 with a 20-minute time constant, and treats repeated provocations with Campbell's theorem. (https://5dlibrary.com/answers/en/catharsis-falsified-arousal-decay-negotiation-cooling-off/c03/) 第3章 一阶衰减:唤醒的时间常数、阈值与散粒噪声 — 第三章用一阶衰减计算等到阈值的时间,被激怒到 7 分、时间常数 20 分钟时只是等需 16.9 分钟,并用坎贝尔定理处理接连不断的刺激。 (https://5dlibrary.com/answers/zh/catharsis-falsified-arousal-decay-negotiation-cooling-off/c03/) - Ch.4 Two Clocks: the Excitation-Transfer Window and the Two-Factor Label — Chapter 4 computes the excitation-transfer window from Zillmann's two clocks, peaking at 10.75 minutes under illustrative parameters, and reports the controversy over two-factor theory. (https://5dlibrary.com/answers/en/catharsis-falsified-arousal-decay-negotiation-cooling-off/c04/) 第4章 两只时钟:兴奋转移窗口与双因素标签 — 第四章用 Zillmann 的两只时钟计算兴奋转移窗口,示意峰值在 10.75 分钟,并并陈双因素理论的争议。 (https://5dlibrary.com/answers/zh/catharsis-falsified-arousal-decay-negotiation-cooling-off/c04/) - Ch.5 Optimal Stopping: How Long a Cooling-Off Period Should Be — Chapter 5 calibrates acceptance to delay experiments in the ultimatum game; at moderate withdrawal risk the best pause is 14.7 minutes, worth 3.09 times deciding at once. (https://5dlibrary.com/answers/en/catharsis-falsified-arousal-decay-negotiation-cooling-off/c05/) 第5章 最优停时:谈判冷却期该多长 — 第五章用最后通牒博弈的延迟实验校准接受率,中等撤回风险下最优冷却 14.7 分钟,期望值是立刻决定的 3.09 倍。 (https://5dlibrary.com/answers/zh/catharsis-falsified-arousal-decay-negotiation-cooling-off/c05/) - Ch.6 Displayed Is Not Felt: Anger as a Signal, Power and the Price of Impatience — Chapter 6 reports the value of anger as a signal, argues for selective display after calming down, and prices impatience: doubling it costs 0.1665 of the share. (https://5dlibrary.com/answers/en/catharsis-falsified-arousal-decay-negotiation-cooling-off/c06/) 第6章 表露不等于感到:愤怒信号、权力与耐心的价格 — 第六章并陈愤怒作为信号的价值,说明应在冷静之后有选择地表露,并算出急躁的价格:急躁倍数加倍损失 0.1665 的份额。 (https://5dlibrary.com/answers/zh/catharsis-falsified-arousal-decay-negotiation-cooling-off/c06/) - Ch.7 Classics East and West: the Zhongyong's Measure and Seneca's Delay — Chapter 7 examines the Zhongyong's "hitting the measure" and Seneca's delay in On Anger; expressing at once after a provocation hits the measure with probability 0.011, rising to 0.752 after 20 minutes. (https://5dlibrary.com/answers/en/catharsis-falsified-arousal-decay-negotiation-cooling-off/c07/) 第7章 东西方原典考辨:《中庸》的中节与塞涅卡的拖延 — 第七章考辨《中庸》的「发而皆中节」与塞涅卡《论愤怒》的拖延,被激怒后立刻表达的中节概率 0.011,等 20 分钟升到 0.752。 (https://5dlibrary.com/answers/zh/catharsis-falsified-arousal-decay-negotiation-cooling-off/c07/) - Ch.8 Closing the Loop: Complaint Cooling Rules, an Owner's Emotion-Decision Log and Pre-Registration — Chapter 8 gives complaint cooling rules (acknowledge first, with a best substantive delay of about 35.5 minutes), an owner's emotion-decision log, the Cooling-Off Calculator and a pre-registered test (176 events per group). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/catharsis-falsified-arousal-decay-negotiation-cooling-off/c08/) 第8章 实证闭环:客诉冷却规则、店主情绪—决策日志与预注册 — 第八章给出客诉冷却规则(先确认,实质回复最优延迟约 35.5 分钟)、店主情绪—决策日志、冷却期计算器与预注册方案(每组 176 个事件)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/catharsis-falsified-arousal-decay-negotiation-cooling-off/c08/) ## 5DL-B-201 · Open-Label Placebos and Predictive Coding, Price-Tag Expectancy Gains and the Nocebo of the Bow in the Cup 中文书名:开放标签安慰剂预测编码、价格标签期望增益与杯弓蛇影反安慰剂母典 DOI: pending Cite: Fan, Y. (2026). Open-Label Placebos and Predictive Coding, Price-Tag Expectancy Gains and the Nocebo of the Bow in the Cup (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not medical advice, claims no effect of any product or practice on the body or on health, and its procedures and tool are non-personalised business guidance. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: "A placebo only works if people are kept in the dark" and "price does not change what people feel": this book tests both pieces of common sense against published randomised trials and puts them into one precision-weighted Bayesian model, in which perceived intensity is the precision-weighted average of expectation and bodily signal, price and explanation move the mean of expectation, ritual and relationship raise its precision, and nocebo is the same formula with a minus sign. Its parameters are illustrative; all calculations are generated by its programs; it is not medical advice, claims no effect of any product or practice on the body or on health, and its procedures and tool are non-personalised business guidance. Chapter 1 recomputes three tests of open-label placebo: a day-21 effect size of 0.788 in the 2010 pilot; 0.434 for open label against no pill in the three-arm trial of 2021, barely different from double-blind placebo; and 0.39 for self-reported against only 0.09 for objective outcomes in the meta-analysis of 2025. Chapter 2 recomputes the price-tag study: 0.854 of those told the regular price reported less pain against 0.610 of those told the discount price, and exact enumeration shows that with 41 per arm and a small true gap, significant results overstate it by 2.58 times on average. Chapters 3 and 4 build the precision-weighted model and ask how the precision of expectation is formed by experience, explanation and ritual. Chapter 5 recomputes the statin n-of-1 trial, with a nocebo ratio of 0.892 and only 0.055 of the symptom burden in statin months due to the drug itself, and builds a misattribution model of warning labels. Chapter 6 tests the earlier chapters with three knives: regression to the mean, reporting bias and the "powerless placebo". Chapter 7 examines the bow in the cup from Ying Shao's Fengsu Tongyi and the 1784 report of the French royal commission on animal magnetism. Chapter 8 gives a closed loop of expectation management for high-priced experiences, the distilled tool "Expectation Management Checklist" and a preregistered plan: raising the satisfied share from 0.70 to 0.80 needs 294 per arm. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:「安慰剂必须瞒着人才有用」「价格不改变感受」,本书用已发表的随机试验检验这两条常识,并把它们放进同一个精度加权的贝叶斯模型:被感知到的强度等于期望与身体信号按各自精度加权的平均,价格与说明移动期望的均值,仪式与关系提高期望的精度,反安慰剂是同一条公式的负号。书中参数为示意;全部计算由书中程序生成;本书不是医疗建议,不宣称任何产品或做法对身体或健康有任何作用,书中流程与工具为非个人化经营建议。 第一章复算开放标签安慰剂的三次检验:2010 年试点第 21 天的效应量为 0.788,2021 年三臂试验中开放标签相对不服药为 0.434、与双盲安慰剂相差无几,2025 年元分析中自评效应为 0.39、客观效应只有 0.09。第二章复算价格标签研究:被告知常规价者报告疼痛减轻的比例为 0.854,折扣价者为 0.610;精确枚举显示,每组 41 人的研究在真实差距较小时,显著结果平均会夸大 2.58 倍。第三、四章建立精度加权模型,讨论期望精度如何由经验、解释与仪式形成。 第五章复算他汀的 n-of-1 试验,反安慰剂比为 0.892,他汀月份的症状负担中只有 0.055 来自药物本身,并建立警示标签的误归因模型。第六章用回归均值、报告偏倚与「安慰剂无力」三把刀检验前文。第七章考辨应劭《风俗通义》中的杯弓蛇影与 1784 年法国皇家委员会的动物磁力报告。第八章给出高价体验产品的期望管理闭环、蒸馏小工具「期望管理检查表」与预注册方案:从 0.70 提到 0.80 的满意比例需要每组 294 人。书中以标注框区分已发表证据、可运行的 Python 计算(21 段程序)、原典与作者假说。 Chapter answers: - Ch.1 Telling People It Is a Sugar Pill: Three Tests of Open-Label Placebo — Chapter 1 recomputes three tests of open-label placebo: a day-21 effect size of 0.788 in the 2010 pilot; 0.434 for open label against no pill in the three-arm trial of 2021, barely different from double-blind placebo; and 0.39 for self-reported against only 0.09 for objective outcomes in the meta-analysis of 2025. (https://5dlibrary.com/answers/en/open-label-placebo-predictive-coding-price-expectancy-nocebo/c01/) 第1章 明说是糖丸:开放标签安慰剂的三次检验 — 第一章复算开放标签安慰剂的三次检验:2010 年试点第 21 天的效应量为 0.788,2021 年三臂试验中开放标签相对不服药为 0.434、与双盲安慰剂相差无几,2025 年元分析中自评效应为 0.39、客观效应只有 0.09。 (https://5dlibrary.com/answers/zh/open-label-placebo-predictive-coding-price-expectancy-nocebo/c01/) - Ch.2 Two Dollars Fifty or Ten Cents: the Expectancy Gain of a Price Tag and its Winner's Curse — Chapter 2 recomputes the price-tag study: 0.854 of those told the regular price reported less pain against 0.610 of those told the discount price, and exact enumeration shows that with 41 per arm and a small true gap, significant results overstate it by 2.58 times on average. (https://5dlibrary.com/answers/en/open-label-placebo-predictive-coding-price-expectancy-nocebo/c02/) 第2章 两块五与一毛钱:价格标签的期望增益及其赢家诅咒 — 第二章复算价格标签研究:被告知常规价者报告疼痛减轻的比例为 0.854,折扣价者为 0.610;精确枚举显示,每组 41 人的研究在真实差距较小时,显著结果平均会夸大 2.58 倍。 (https://5dlibrary.com/answers/zh/open-label-placebo-predictive-coding-price-expectancy-nocebo/c02/) - Ch.3 The Posterior Is a Weighted Mean: Precision-Weighted Perception and Price as a Shift of the Prior — The posterior is the precision-weighted mean of prior and signal. Program p31 shows that the ceiling of expectation's influence is set by the weight, which is 0.500 when expectation and signal are equally precise; p32 shows that reproducing Waber's gap requires a product of shift and weight of 7.74 points, a shift of 15.5 points at a weight of 0.500; p33 shows that the most precise third of participants report relief at 0.797 and the least precise third at only 0.640. Price moves the mean, ritual raises the precision, and their product decides the response. (https://5dlibrary.com/answers/en/open-label-placebo-predictive-coding-price-expectancy-nocebo/c03/) 第3章 后验等于加权平均:精度加权感知与价格作为先验位移 — 后验等于先验与信号按精度加权的平均。p31 说明,期望的作用上限由权重决定,期望与信号同样精确时权重为 0.500;p32 说明,要复现瓦伯的差距,期望位移与权重的乘积需要达到 7.74 分,权重为 0.500 时位移需要 15.5 分;p33 说明,精度最高的三分之一受试者报告缓解的比例为 0.797,精度最低的三分之一只有 0.640。价格移动均值,仪式提高精度,二者的乘积决定反应。 (https://5dlibrary.com/answers/zh/open-label-placebo-predictive-coding-price-expectancy-nocebo/c03/) - Ch.4 Where the Weight Comes From: Conditioned Learning, Rationale and the Additivity of Ritual — Program p41 shows that the weight of a prior can be built by experience: after four sessions of real relief, an openly declared placebo still brings 11.82 points of felt relief, falling only to 9.97 after six sessions. The studies of Locher and Kam-Hansen show that explanation and labels shift the mean of expectation. Program p42 shows that ritual and relationship each add about 0.432 and 0.456 on the probit scale, approximately additively, while the loss from open labelling relative to double blinding is only −0.09 of the ritual step. The next chapter turns the same machinery round: what happens when expectation predicts harm. (https://5dlibrary.com/answers/en/open-label-placebo-predictive-coding-price-expectancy-nocebo/c04/) 第4章 权重从哪里来:条件化学习、解释与仪式的可加性 — p41 说明,先验的权重可以由经验建立:四次真实缓解之后,明说是安慰剂的药丸仍带来 11.82 分的感受缓解,六次之后才降到 9.97。洛赫尔与卡姆-汉森的研究说明,解释与标签移动的是期望的均值。p42 说明,仪式与关系在概率单位上各高约 0.432 与 0.456,近似可加,而开放标签相对双盲的损失只有仪式台阶的 −0.09。下一章把同一套机制翻过来:期望伤害时会怎样。 (https://5dlibrary.com/answers/zh/open-label-placebo-predictive-coding-price-expectancy-nocebo/c04/) - Ch.5 Nocebo: the Same Formula with a Minus Sign — Chapter 5 recomputes the statin n-of-1 trial, with a nocebo ratio of 0.892 and only 0.055 of the symptom burden in statin months due to the drug itself, and builds a misattribution model of warning labels. (https://5dlibrary.com/answers/en/open-label-placebo-predictive-coding-price-expectancy-nocebo/c05/) 第5章 反安慰剂:同一条公式的负号 — 第五章复算他汀的 n-of-1 试验,反安慰剂比为 0.892,他汀月份的症状负担中只有 0.055 来自药物本身,并建立警示标签的误归因模型。 (https://5dlibrary.com/answers/zh/open-label-placebo-predictive-coding-price-expectancy-nocebo/c05/) - Ch.6 The Sceptic's Three Knives: Regression to the Mean, Reporting Bias and the "Powerless Placebo" — Program p61 shows that, with illustrative parameters, regression to the mean explains only part of the no-pill arm's improvement; producing a 52.3-point fall by itself would need a measurement error of 115.6 points. Program p62 shows that if objective outcomes are fully reliable, reporting bias could explain 0.769 of the self-reported effect, and after subtracting it the open-label effect in Lembo's trial is only 0.134. Program p63 shows that the "powerless placebo" analysis of 2001 and the open-label meta-analysis of 2025 point to the same boundary: a small to moderate effect on subjective continuous scales, almost none on objective hard outcomes. The next chapter turns to two classical texts that touched the same boundary some eighteen centuries and more than two centuries ago. (https://5dlibrary.com/answers/en/open-label-placebo-predictive-coding-price-expectancy-nocebo/c06/) 第6章 怀疑者的三把刀:回归均值、报告偏倚与「安慰剂无力」 — p61 说明,回归均值在示意参数下只解释不服药组改善的一部分,单凭它产生 52.3 分的回落需要 115.6 分的测量误差。p62 说明,若客观指标完全可信,报告偏倚可以解释自评效应的 0.769,扣除后伦博试验的开放标签效应只剩 0.134。p63 说明,2001 年的「安慰剂无力」分析与 2025 年的开放标签元分析指向同一个边界:主观连续量表上有小到中等的作用,客观硬结局上几乎没有。下一章转向两部古典文献:它们分别在一千八百多年前与两百多年前,已经触碰到了同一个边界。 (https://5dlibrary.com/answers/zh/open-label-placebo-predictive-coding-price-expectancy-nocebo/c06/) - Ch.7 Classical Sources East and West: Du Xuan's Snake in the Cup and Franklin's Blindfolded Boy — Chapter 7 examines the bow in the cup from Ying Shao's Fengsu Tongyi and the 1784 report of the French royal commission on animal magnetism. (https://5dlibrary.com/answers/en/open-label-placebo-predictive-coding-price-expectancy-nocebo/c07/) 第7章 东西方原典考辨:杜宣的杯中蛇影与富兰克林的被蒙眼的少年 — 第七章考辨应劭《风俗通义》中的杯弓蛇影与 1784 年法国皇家委员会的动物磁力报告。 (https://5dlibrary.com/answers/zh/open-label-placebo-predictive-coding-price-expectancy-nocebo/c07/) - Ch.8 Closing the Loop: Expectation Management, Ritualised Delivery and Blinded Satisfaction Comparisons for High-Priced Experiences — Chapter 8 gives a closed loop of expectation management for high-priced experiences, the distilled tool "Expectation Management Checklist" and a preregistered plan: raising the satisfied share from 0.70 to 0.80 needs 294 per arm. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/open-label-placebo-predictive-coding-price-expectancy-nocebo/c08/) 第8章 实证闭环:高价体验的期望管理、仪式化交付与盲法满意度对照 — 第八章给出高价体验产品的期望管理闭环、蒸馏小工具「期望管理检查表」与预注册方案:从 0.70 提到 0.80 的满意比例需要每组 294 人。书中以标注框区分已发表证据、可运行的 Python 计算(21 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/open-label-placebo-predictive-coding-price-expectancy-nocebo/c08/) ## 5DL-B-202 · Regression to the Mean and the Cover Jinx: Posterior Calibration of Fortune Reports and Serial Independence in the Tale of the Lost Horse 中文书名:均值回归与封面诅咒、运程报告后验校准与塞翁失马序列独立性母典 DOI: pending Cite: Fan, Y. (2026). Regression to the Mean and the Cover Jinx: Posterior Calibration of Fortune Reports and Serial Independence in the Tale of the Lost Horse (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: Owners widely believe that when the best store or creator does worse the next year it is because the award made them complacent, and that when the worst store improves after criticism the criticism worked. This book writes that common sense as runnable arithmetic and tests when it fails. The core model is the regression line of the bivariate normal: given this year's score, next year's conditional expectation is the correlation times this year's score. Its stores, creators, customers and reports are illustrative; all calculations are generated by its programs; the book is general business analysis, not business, legal or financial advice, and where it touches fortune reports and classical texts it analyses only structure and probability and makes no reading for any individual. Chapter 1 shows that with a year-to-year correlation of 0.5 a share of 0.8380 of top-decile stores score lower next year although nobody relaxed, and rebuilds Secrist's "triumph of mediocrity" with Hotelling's rebuttal. Chapter 2 replays the flight instructor's season: with feedback that does nothing, the instructor still sees criticism outperform praise by 2.529 units, and only a randomised comparison corrects it. Chapter 3 splits skill from luck with analysis of variance and the Spearman-Brown formula. Chapter 4 explains shrinkage with Efron and Morris's eighteen batters: in a simplified version the total squared error of the raw averages is 3.53 times that of the shrunk estimates. Chapter 5 shows that the cover jinx is a product of selecting at the peak: a winner chosen from a hundred gets worse in the next period with probability 0.918 even when the cover does nothing, and offers a residual test that can detect a real side effect. Chapter 6 treats purchases at the low point, Barnum statements and the Brier score, and shows that calibration is not information. Chapter 7 reads the Huainanzi's tale of the lost horse alongside Galton's 1886 paper and computes the bias of the "blessing after blessing" share in short records, 0.4048 for four entries. Chapter 8 gives an annual review by shrinkage ranking (the first-placed store in the example has only 0.634 probability of a top-three finish), a copy audit, a pre-registered plan and the distilled Shrinkage Ranking Calculator. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:经营者普遍相信:业绩最好的门店或达人第二年变差,是因为拿了奖就松懈;业绩最差的门店挨了批评之后变好,说明批评管用。本书把这条常识写成可以运行的算术,并检验它在什么条件下站不住。核心模型是双变量正态的回归直线:给定今年的成绩,明年的条件期望等于相关系数乘以今年的成绩。书中门店、达人、顾客与报告的参数均为示意;全部计算由书中程序生成;本书是非个人化的经营分析,不构成经营、法律或理财建议,涉及运程报告与原典之处只分析结构与概率,不作任何个人占断。 第一章说明,在年间相关为 0.5 时,前十分之一门店中有 0.8380 来年成绩低于今年,没有任何人松懈;并复现塞克里斯特「平庸的胜利」与霍特林的反驳。第二章重演飞行教官的训练季:反馈毫无作用时,教官仍会看到批评比表扬多带来 2.529 个单位的进步,只有随机对照能纠正。第三章用方差分析与斯皮尔曼—布朗公式分解技能与运气。第四章用埃夫隆与莫里斯的十八名击球手说明收缩估计:在简化版本中,原始打击率的总平方误差是收缩估计的 3.53 倍。 第五章说明封面诅咒是在峰值处挑选的产物:从一百人中选出的第一名,即使封面毫无作用,也有 0.918 在下一期变差,并给出分辨真实副作用的残差检验。第六章讨论低谷购买、巴纳姆陈述与布赖尔分数,说明校准不等于有信息。第七章对读《淮南子》塞翁失马与高尔顿 1886 年的论文,并算出短记录中「福后又福」比例的偏差,四次记录时为 0.4048。第八章给出收缩排名的年度评估(示例中排名第一的门店进入前三的概率只有 0.634)、文案审计、预注册方案与蒸馏小工具「收缩排名计算器」。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A Conditional Expectation with Slope Below One: Why Top Stores Slip the Next Year — Chapter 1 shows that with a year-to-year correlation of 0.5 a share of 0.8380 of top-decile stores score lower next year although nobody relaxed, and rebuilds Secrist's "triumph of mediocrity" with Hotelling's rebuttal. (https://5dlibrary.com/answers/en/regression-to-mean-cover-jinx-fortune-report-calibration/c01/) 第1章 条件期望是一条斜率小于一的直线:头部门店为什么来年下滑 — 第一章说明,在年间相关为 0.5 时,前十分之一门店中有 0.8380 来年成绩低于今年,没有任何人松懈;并复现塞克里斯特「平庸的胜利」与霍特林的反驳。 (https://5dlibrary.com/answers/zh/regression-to-mean-cover-jinx-fortune-report-calibration/c01/) - Ch.2 The Flight Instructor's Illusion: Why Criticism Looks More Effective than Praise — Chapter 2 replays the flight instructor's season: with feedback that does nothing, the instructor still sees criticism outperform praise by 2.529 units, and only a randomised comparison corrects it. (https://5dlibrary.com/answers/en/regression-to-mean-cover-jinx-fortune-report-calibration/c02/) 第2章 飞行教官的错觉:批评为何看起来比表扬有效 — 第二章重演飞行教官的训练季:反馈毫无作用时,教官仍会看到批评比表扬多带来 2.529 个单位的进步,只有随机对照能纠正。 (https://5dlibrary.com/answers/zh/regression-to-mean-cover-jinx-fortune-report-calibration/c02/) - Ch.3 Splitting Variance into Skill and Luck: Repeated Measures, Intraclass Correlation and Reliability — The correlation between two years is the share of the total variance that is skill, and it can be estimated from repeated measurements by a one-way random-effects analysis of variance; but the estimate is not precise, and 60 stores with 4 years of records still give a wide interval. The Spearman-Brown formula tells us how many periods are needed before a ranking means something, and for measures with low single-period reliability the answer is often dozens. Conversion rates under binomial sampling need hundreds or thousands of orders before skill and luck can be told apart. Skill exists, and sometimes it lasts; the use of a variance decomposition is to turn that sentence into a number that can be computed and tested. (https://5dlibrary.com/answers/en/regression-to-mean-cover-jinx-fortune-report-calibration/c03/) 第3章 技能与运气的方差分解:重复测量、组内相关与信度 — 两年之间的相关系数就是技能方差在总方差中所占的份额,可以用单因素随机效应方差分析从重复测量中估出来;但它的估计本身并不精确,60 家门店、4 年数据给出的区间仍然很宽。斯皮尔曼—布朗公式告诉我们需要多少期数据才能让排名有意义,对单期信度低的指标,答案往往是几十期。二项抽样下的转化率需要数百乃至上千个订单才能分开技能与运气。技能确实存在,有时还很持久;方差分解的用处是把这句话变成一个可以算、可以检验的数。 (https://5dlibrary.com/answers/zh/regression-to-mean-cover-jinx-fortune-report-calibration/c03/) - Ch.4 Shrinkage Estimation: James and Stein, Efron and Morris's Eighteen Batters, and Shrinkage Ranking — Chapter 4 explains shrinkage with Efron and Morris's eighteen batters: in a simplified version the total squared error of the raw averages is 3.53 times that of the shrunk estimates. (https://5dlibrary.com/answers/en/regression-to-mean-cover-jinx-fortune-report-calibration/c04/) 第4章 收缩估计:詹姆斯—斯坦、埃夫隆—莫里斯的十八名击球手与收缩排名 — 第四章用埃夫隆与莫里斯的十八名击球手说明收缩估计:在简化版本中,原始打击率的总平方误差是收缩估计的 3.53 倍。 (https://5dlibrary.com/answers/zh/regression-to-mean-cover-jinx-fortune-report-calibration/c04/) - Ch.5 The Cover Jinx: Select at the Peak and You Will See a Fall — Chapter 5 shows that the cover jinx is a product of selecting at the peak: a winner chosen from a hundred gets worse in the next period with probability 0.918 even when the cover does nothing, and offers a residual test that can detect a real side effect. (https://5dlibrary.com/answers/en/regression-to-mean-cover-jinx-fortune-report-calibration/c05/) 第5章 封面诅咒:在峰值处挑选,必然看到下滑 — 第五章说明封面诅咒是在峰值处挑选的产物:从一百人中选出的第一名,即使封面毫无作用,也有 0.918 在下一期变差,并给出分辨真实副作用的残差检验。 (https://5dlibrary.com/answers/zh/regression-to-mean-cover-jinx-fortune-report-calibration/c05/) - Ch.6 Posterior Calibration of Fortune Reports: Buying at the Low Point, Barnum Statements and the Brier Score — For an arbitrary month the share of next months that are better is 0.500, just as symmetry requires. But with $\rho$ at 0.3 and purchases only in the worst 0.05 of months, the next month is better in a share of 0.924, with an average rise of 1.426 standard deviations; with $\rho$ at 0.6 and the worst 0.15 it is 0.770. Even when the purchase threshold is relaxed to the worst 0.30 of months, the share is still between 0.716 and 0.786. A report that says nothing but "it will get better" achieves that hit rate among these buyers without any knowledge at all. (https://5dlibrary.com/answers/en/regression-to-mean-cover-jinx-fortune-report-calibration/c06/) 第6章 运程报告的后验校准:低谷购买、巴纳姆陈述与布赖尔分数 — 随便挑一个月,下个月更好的比例是 0.500,这正是对称性给出的值。可是在 $\rho$ 为 0.3、只在最差的 0.05 份额月份购买时,下个月更好的比例是 0.924,平均回升 1.426 个标准差;在 $\rho$ 为 0.6、最差 0.15 份额时是 0.770。即使购买门槛放宽到最差的 0.30 份额,比例也还有 0.716 到 0.786。一份只写「会好起来」的报告,在这些购买者身上的命中率就是这么高,它不需要任何知识。 (https://5dlibrary.com/answers/zh/regression-to-mean-cover-jinx-fortune-report-calibration/c06/) - Ch.7 Sources East and West: the Huainanzi's Old Man Who Lost His Horse and Galton's 1886 "Regression towards Mediocrity" — Chapter 7 reads the Huainanzi's tale of the lost horse alongside Galton's 1886 paper and computes the bias of the "blessing after blessing" share in short records, 0.4048 for four entries. (https://5dlibrary.com/answers/en/regression-to-mean-cover-jinx-fortune-report-calibration/c07/) 第7章 东西方原典考辨:《淮南子》塞翁失马与高尔顿 1886 年的「向平庸回归」 — 第七章对读《淮南子》塞翁失马与高尔顿 1886 年的论文,并算出短记录中「福后又福」比例的偏差,四次记录时为 0.4048。 (https://5dlibrary.com/answers/zh/regression-to-mean-cover-jinx-fortune-report-calibration/c07/) - Ch.8 Closing the Loop: Annual Reviews by Shrinkage Ranking, Copy without Causal Overstatement, and a Pre-registered Plan — Chapter 8 gives an annual review by shrinkage ranking (the first-placed store in the example has only 0.634 probability of a top-three finish), a copy audit, a pre-registered plan and the distilled Shrinkage Ranking Calculator. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/regression-to-mean-cover-jinx-fortune-report-calibration/c08/) 第8章 实证闭环:收缩排名的年度评估、去因果夸大的文案与预注册方案 — 第八章给出收缩排名的年度评估(示例中排名第一的门店进入前三的概率只有 0.634)、文案审计、预注册方案与蒸馏小工具「收缩排名计算器」。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/regression-to-mean-cover-jinx-fortune-report-calibration/c08/) ## 5DL-B-203 · The Shrinking Marshmallow Effect, Family-Resource Confounding and the Limits of Child Self-Control Interventions 中文书名:延迟满足棉花糖实验复制收缩、家庭资源混杂与儿童自控干预边界母典 DOI: pending Cite: Fan, Y. (2026). The Shrinking Marshmallow Effect, Family-Resource Confounding and the Limits of Child Self-Control Interventions (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not a psychological assessment or medical or parenting advice, and its suggestions for shops and communities are non-personalised business advice. Marked boxes separate published evidence, the author's hypotheses, classical sources and runnable Python calculations (18 programs). Abstract: "Whether a four-year-old can resist one marshmallow foretells the success or failure of his whole life" has been parenting common sense for decades. This book turns the claim into runnable arithmetic and checks it against the original papers: once family background is controlled, the predictive power of a single waiting test at four and a half for achievement at fifteen shrinks by about two thirds, and a child who does not wait may well be responding rationally to an unreliable environment. Its parameters are illustrative; all calculations are generated by its programs; the book is not a psychological assessment or medical or parenting advice, and its suggestions for shops and communities are non-personalised business advice. Chapter 1 recomputes the replication by Watts, Duncan and Quan: among children whose mothers had not completed college, the standardised coefficient falls from 0.236 to 0.081 with early controls, a shrinkage of 0.657. Chapter 2 recomputes the "rational snacking" experiment of Kidd and colleagues, where mean waits differ by a ratio of 3.98, and shows that in a world where waiting has no causal effect, rational not-waiting alone produces an uncontrolled slope of 0.208. Chapter 3 treats waiting as optimal stopping: the longer one waits without result, the further away the reward looks. Chapter 4 splits the shrinkage after controls into residual confounding, measurement error and bad controls, and quantifies the rebuttal by Falk and colleagues. Chapter 5 shows that the controlled category coefficients look more like a twenty-second threshold than a slope, and examines single-trial reliability and the seven-minute ceiling. Chapter 6 gives an optimistic upper bound for training children to wait: even treating the coefficient as causal with no fadeout, the academic gain ten years later is only 1.77 points on a scale with mean 100 and standard deviation 15, and only 0.61 points with the controlled coefficient. Chapter 7 examines "to rear without teaching is the father's fault" in the Three Character Classic and Locke's teaching on self-denial. Chapter 8 offers a toll-booth scenario capsule, a family promise ledger (telling a home that keeps nine promises in ten from one that keeps six needs at least 14 records) and a pre-registered testing plan. Marked boxes separate published evidence, the author's hypotheses, classical sources and runnable Python calculations (18 programs). 摘要:「四岁时能不能忍住一块棉花糖,预示着孩子一生的成败」是流传几十年的育儿常识。本书把它拆成可以运行的算术,并核对原始文献:控制家庭背景之后,四岁半时一次等待测验对十五岁学业成就的预测力缩小约三分之二;孩子不等,很可能是对不可靠环境的理性反应。书中参数为示意;全部计算由书中程序生成;本书不是心理评估、医疗或育儿意见,书中关于门店与社群的建议是非个人化经营建议。 第一章重算 Watts、Duncan 与 Quan 的复制研究:母亲未完成大学的子样本中,标准化系数从 0.236 降到加入早期控制后的 0.081,缩水 0.657。第二章重算 Kidd 等人的「理性零食」实验,两组平均等待之比为 3.98,并演示在等待毫无因果效应的世界里,理性的「不等」就能制造 0.208 的未控制斜率。第三章把等待写成最优停时问题,等得越久却等不到,回报看起来越远。第四章把控制后的缩水拆成残留混杂、测量误差与坏控制三个方向,并量化 Falk 等人的反驳。 第五章说明控制后的类别系数更像一个二十秒门槛而不是斜率,并检查单次信度与七分钟封顶。第六章给出训练等待的乐观上界:即使把系数当作因果、效果永不消退,十年后的学业收益也只有 1.77 分(均值 100、标准差 15 的量表),用控制后的系数只有 0.61 分。第七章考辨《三字经》「养不教,父之过」与洛克《教育漫话》的克己之说。第八章给出收费站情境胶囊、家庭承诺记录卡(区分九成与六成兑现的家庭至少要记 14 次)与预注册检验方案。书中以标注框区分已发表的证据、作者的假说、原典与可运行的 Python 计算(18 段程序)。 Chapter answers: - Ch.1 Two Thirds Gone: the Marshmallow Coefficient Meets Family Background — Chapter 1 recomputes the replication by Watts, Duncan and Quan: among children whose mothers had not completed college, the standardised coefficient falls from 0.236 to 0.081 with early controls, a shrinkage of 0.657. (https://5dlibrary.com/answers/en/marshmallow-replication-family-confounding-self-control-limits/c01/) 第1章 三分之二消失:棉花糖系数遇上家庭背景 — 第一章重算 Watts、Duncan 与 Quan 的复制研究:母亲未完成大学的子样本中,标准化系数从 0.236 降到加入早期控制后的 0.081,缩水 0.657。 (https://5dlibrary.com/answers/zh/marshmallow-replication-family-confounding-self-control-limits/c01/) - Ch.2 Rational Snacking: Not Waiting under a Prior on Environmental Reliability — Chapter 2 recomputes the "rational snacking" experiment of Kidd and colleagues, where mean waits differ by a ratio of 3.98, and shows that in a world where waiting has no causal effect, rational not-waiting alone produces an uncontrolled slope of 0.208. (https://5dlibrary.com/answers/en/marshmallow-replication-family-confounding-self-control-limits/c02/) 第2章 理性零食:环境可信度先验下的「不等」 — 第二章重算 Kidd 等人的「理性零食」实验,两组平均等待之比为 3.98,并演示在等待毫无因果效应的世界里,理性的「不等」就能制造 0.208 的未控制斜率。 (https://5dlibrary.com/answers/zh/marshmallow-replication-family-confounding-self-control-limits/c02/) - Ch.3 How Long Is It Rational to Persist? Waiting-Time Distributions and the Optimal Give-Up Point — The values of q in the table are posterior means taken from p22 in Chapter 2. A child with a very low cost per minute (c of 0.02) waits the full 15 minutes whenever q is at least 0.250, and even with q of only 0.143 waits 12.0 minutes. At a moderate cost (c of 0.05) a clear gradient appears: with q of 0.143 the child eats at once, with 0.250 waits 2.2 minutes, with 0.429 waits 6.6, with 0.571 waits 9.8, with 0.750 waits 14.2, and with 0.857 waits it out. At a higher cost (c of 0.08) the whole gradient shifts down: 1.9 minutes at q of 0.429, 5.0 at 0.571, 9.5 at 0.750 and 13.3 at 0.857. With q of 0.750 the hazard at the start is 0.1542 per minute and its long-run floor is 0.0167 per minute; a child whose cost lies below the floor never gives up voluntarily in this model. (https://5dlibrary.com/answers/en/marshmallow-replication-family-confounding-self-control-limits/c03/) 第3章 坚持多久才算理性:等待时间分布与最优放弃点 — 表中的 q 取自第二章 p22 的几种后验均值。对一个每分钟代价很低(c 为 0.02)的孩子,只要 q 不低于 0.250,就会等满 15 分钟;即使 q 只有 0.143,也会等 12.0 分钟。代价中等(c 为 0.05)时,表格出现了一条清晰的梯度:q 为 0.143 时立刻吃,0.250 时等 2.2 分钟,0.429 时 6.6 分钟,0.571 时 9.8 分钟,0.750 时 14.2 分钟,0.857 时等满。代价较高(c 为 0.08)时,梯度整体下移:q 为 0.429 时只等 1.9 分钟,0.571 时 5.0 分钟,0.750 时 9.5 分钟,0.857 时 13.3 分钟。q 为 0.750 时开始的风险率是每分钟 0.1542,长期下限是每分钟 0.0167;代价低于下限的孩子,在这个模型里永远不会主动放弃。 (https://5dlibrary.com/answers/zh/marshmallow-replication-family-confounding-self-control-limits/c03/) - Ch.4 Why the Coefficient Shrinks: Omitted Variables, Measurement Error and Bad Controls — Shrinkage after controls can come from three directions: confounding correctly removed, confounding left behind by imperfect measurement, and true effect wrongly deleted by a bad control. The omitted-variable formula shows that an unmeasured factor correlating 0.3 with waiting and moderately affecting achievement is enough to erase the 0.081; lowering the reliability of family resources from 1 to 0.6 lifts the controlled coefficient from near the truth to more than double it; and controlling for a concurrent variable that is itself an outcome cuts the total effect down to the direct effect. All three disputes take place within a range that is already small. The next chapter turns to measurement itself: what can a test given once, lasting seven minutes and involving one sweet, actually measure? (https://5dlibrary.com/answers/en/marshmallow-replication-family-confounding-self-control-limits/c04/) 第4章 系数为什么缩水:遗漏变量、测量误差与坏控制 — 控制后系数的缩水可以来自三个方向:被正确移除的混杂、因测量不准而残留的混杂,以及因坏控制而被误删的真实效应。遗漏变量偏误公式显示,一个与等待相关 0.3、对成就有中等影响的未测因素,就足以把 0.081 抹平;家庭资源的信度从 1 降到 0.6,控制后的系数就会从接近真值抬到两倍多;而控制一个本身是结果的同期变量,会把总效应砍到直接效应。三个方向的争论都发生在一个已经很小的区间内。下一章从测量本身入手:一个只做一次、只有七分钟、只有一块糖的测验,究竟能测出什么。 (https://5dlibrary.com/answers/zh/marshmallow-replication-family-confounding-self-control-limits/c04/) - Ch.5 The Twenty-Second Threshold: Category Coefficients, Single-Trial Reliability and the Seven-Minute Ceiling — The marshmallow test is a single, short, capped measurement. After controls, the replication's category coefficients look more like a 20-second threshold than a slope, and in a pure-threshold world a linear model still reports a plausible-looking "effect per minute". Single-trial reliability cuts correlations by about a fifth, and the ceiling mainly depresses associations among children who wait long, though by far less than the gap between the original study and the replication. The next chapter pulls these numbers together and does the arithmetic of intervention: if someone promises to raise achievement by training children to wait, what is the most they could deliver? (https://5dlibrary.com/answers/en/marshmallow-replication-family-confounding-self-control-limits/c05/) 第5章 二十秒门槛:类别系数、单次信度与七分钟封顶 — 棉花糖测验是一个单次、短时、封顶的测量。复制研究的类别系数在控制后更像一个 20 秒的门槛而不是一条斜率;在纯门槛的世界里,线性模型依然会报告一个看似可信的「每分钟效应」。单次测验的信度把相关打了约八折,上限主要在等待较长的孩子身上压低关联,但量级远小于原研究与复制研究之间的落差。下一章把这些数字收拢起来,算一笔干预的账:如果有人承诺通过训练孩子等待来提高成绩,最乐观能提高多少。 (https://5dlibrary.com/answers/zh/marshmallow-replication-family-confounding-self-control-limits/c05/) - Ch.6 The Ceiling on Intervention: Fadeout, an Optimistic Upper Bound and Two Levers — Chapter 6 gives an optimistic upper bound for training children to wait: even treating the coefficient as causal with no fadeout, the academic gain ten years later is only 1.77 points on a scale with mean 100 and standard deviation 15, and only 0.61 points with the controlled coefficient. (https://5dlibrary.com/answers/en/marshmallow-replication-family-confounding-self-control-limits/c06/) 第6章 干预的天花板:消退、乐观上界与两个杠杆 — 第六章给出训练等待的乐观上界:即使把系数当作因果、效果永不消退,十年后的学业收益也只有 1.77 分(均值 100、标准差 15 的量表),用控制后的系数只有 0.61 分。 (https://5dlibrary.com/answers/zh/marshmallow-replication-family-confounding-self-control-limits/c06/) - Ch.7 Classics East and West: the Father's Fault in the Three Character Classic and Locke on Self-Denial — With perfectly consistent parents (k of 1.0), five experiences give an expected posterior mean of 0.857 and a probability of waiting of 1.000; after sixty, the posterior mean is 0.984. With parents consistent nine times in ten (k of 0.9), five experiences give 0.786 and a probability of waiting of 0.991; from twenty onwards the probability is already 1.000. With seven in ten, five experiences give 0.643 and a probability of 0.837, twenty give 0.952 and sixty 0.999. But with parents whose word holds only half the time (k of 0.5), however many experiences pass, the child's posterior mean stays at 0.500, and the probability of waiting is 0.500 after five, 0.412 after twenty and 0.449 after sixty, swinging around one half for ever. (https://5dlibrary.com/answers/en/marshmallow-replication-family-confounding-self-control-limits/c07/) 第7章 东西方原典考辨:《三字经》的「父之过」与洛克的克己 — 完全一致的父母(k 为 1.0),孩子只要经历 5 次,后验均值的期望就是 0.857,会等的概率是 1.000;经历 60 次后,后验均值是 0.984。九成一致(k 为 0.9)时,5 次后是 0.786,会等的概率 0.991;20 次以后,会等的概率已经是 1.000。七成一致时,5 次后是 0.643,会等的概率 0.837;20 次后 0.952;60 次后 0.999。可是一半时候算数、一半时候不算数的父母(k 为 0.5),无论经历多少次,孩子的后验均值都停在 0.500,会等的概率在 5 次后是 0.500,20 次后是 0.412,60 次后是 0.449,始终在一半上下摆动。 (https://5dlibrary.com/answers/zh/marshmallow-replication-family-confounding-self-control-limits/c07/) - Ch.8 Closing the Loop: the Toll-Booth Scenario Capsule, the Family Promise Ledger and Pre-Registration — Chapter 8 offers a toll-booth scenario capsule, a family promise ledger (telling a home that keeps nine promises in ten from one that keeps six needs at least 14 records) and a pre-registered testing plan. Marked boxes separate published evidence, the author's hypotheses, classical sources and runnable Python calculations (18 programs). (https://5dlibrary.com/answers/en/marshmallow-replication-family-confounding-self-control-limits/c08/) 第8章 实证闭环:收费站情境胶囊、家庭承诺记录卡与预注册 — 第八章给出收费站情境胶囊、家庭承诺记录卡(区分九成与六成兑现的家庭至少要记 14 次)与预注册检验方案。书中以标注框区分已发表的证据、作者的假说、原典与可运行的 Python 计算(18 段程序)。 (https://5dlibrary.com/answers/zh/marshmallow-replication-family-confounding-self-control-limits/c08/) ## 5DL-B-204 · The Commuting Paradox, Compensating-Income Equations, Remote-Work Location Choice and the Foolish Old Man's Relocation Decision 中文书名:通勤悖论补偿收入方程、远程工作选址与愚公移山迁居决策母典 DOI: pending Cite: Fan, Y. (2026). The Commuting Paradox, Compensating-Income Equations, Remote-Work Location Choice and the Foolish Old Man's Relocation Decision (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its teams, sites, rents, pay and probabilities are illustrative; all calculations are generated by its programs; it is not individualised business, legal or financial advice. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: Because people choose freely where to live and work, a longer commute must already be compensated by income or housing: this book tests that belief (it fails on the original panel evidence, though contrary evidence exists) and corrects its most popular version along the way. Its teams, sites, rents, pay and probabilities are illustrative; all calculations are generated by its programs; it is not individualised business, legal or financial advice. Chapter 1 returns to the paper by Stutzer and Frey: its 35.4 % compensation belongs to the mean one-way commute of 22 minutes, not to an hour; extending its coefficients, an hour corresponds to 0.6896 of labour income or 0.3266 of household income, with a 90 % band running from 0.406 to 0.973 and beyond. Chapter 2 sets out the opposing evidence and shows that detecting an effect of 0.20 points per hour needs 13,440 person-waves of within variation. Chapter 3 uses a compensating-variation equation to show that pure time cost is only 0.0458 of daily income, the rest coming from a stress term and mispredicted adaptation. Chapter 4 computes the net present value of a studio's site: with time at full rate, a central site at 1006.2 thousand beats the lowest-rent site at 1078.2 thousand. Chapter 5 works out the marginal arithmetic of office days. Chapter 6 uses real options to show that at a volatility of 0.20 the optimal threshold for moving is 2.175 times the net-present-value threshold. Chapter 7 examines the Foolish Old Man of the Liezi and Pascal's Pensées. Chapter 8 gives a team commuting audit, the Site Time-Cost Calculator and a pre-registered test, and shows that a team of 40 can detect only differences above 0.532 points. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:人们可以自由选择住处与工作地点,所以更长的通勤一定已被收入或住房补偿:本书检验这个常识(在原文的面板证据下它不成立,但有反方证据),同时修正它最流行的说法。书中的团队、地点、租金、薪酬与概率均为示意参数;全部计算由书中程序生成;本书不是个人化的经营、法律或理财建议。 第一章回到斯图策与弗雷的原文:35.4 % 的补偿对应的是 22 分钟的平均单程通勤,而不是一小时;用原文系数延伸,一小时对应劳动收入的 0.6896 或家庭收入的 0.3266,且其 90 % 区间从 0.406 延伸到 0.973 以上。第二章并陈反方证据,说明检测每小时 0.20 分的效应需要 13,440 人次的组内变化。第三章用补偿变差方程说明,纯时间成本只占日收入的 0.0458,其余来自压力项与错估的适应。 第四章计算工作室选址的净现值:按全价计入时间,交通居中的地点以 1006.2 千胜过租金最低地点的 1078.2 千。第五章计算到岗日数的边际账。第六章用实物期权说明,波动率 0.20 时搬家的最优门槛是净现值门槛的 2.175 倍。第七章考辨《列子》愚公移山与帕斯卡《思想录》。第八章给出团队通勤盘点、选址时间成本器与预注册检验,40 人的团队只能检测 0.532 分以上的差别。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Commuting Paradox: Where the Thirty-Five Per Cent Comes From — Chapter 1 returns to the paper by Stutzer and Frey: its 35.4 % compensation belongs to the mean one-way commute of 22 minutes, not to an hour; extending its coefficients, an hour corresponds to 0.6896 of labour income or 0.3266 of household income, with a 90 % band running from 0.406 to 0.973 and beyond. (https://5dlibrary.com/answers/en/commuting-paradox-compensating-income-relocation-decision/c01/) 第1章 通勤悖论:三十五个百分点从哪里来 — 第一章回到斯图策与弗雷的原文:35.4 % 的补偿对应的是 22 分钟的平均单程通勤,而不是一小时;用原文系数延伸,一小时对应劳动收入的 0.6896 或家庭收入的 0.3266,且其 90 % 区间从 0.406 延伸到 0.973 以上。 (https://5dlibrary.com/answers/zh/commuting-paradox-compensating-income-relocation-decision/c01/) - Ch.2 The Case Against: Estimators, Sample Sizes and an Effect That Comes and Goes — Chapter 2 sets out the opposing evidence and shows that detecting an effect of 0.20 points per hour needs 13,440 person-waves of within variation. (https://5dlibrary.com/answers/en/commuting-paradox-compensating-income-relocation-decision/c02/) 第2章 反方证据:估计量、样本量与时隐时现的效应 — 第二章并陈反方证据,说明检测每小时 0.20 分的效应需要 13,440 人次的组内变化。 (https://5dlibrary.com/answers/zh/commuting-paradox-compensating-income-relocation-decision/c02/) - Ch.3 The Compensating-Variation Equation: the Price of Time, a Stress Term and Mispredicted Adaptation — Chapter 3 uses a compensating-variation equation to show that pure time cost is only 0.0458 of daily income, the rest coming from a stress term and mispredicted adaptation. (https://5dlibrary.com/answers/en/commuting-paradox-compensating-income-relocation-decision/c03/) 第3章 补偿变差方程:时间价、压力项与错估的适应 — 第三章用补偿变差方程说明,纯时间成本只占日收入的 0.0458,其余来自压力项与错估的适应。 (https://5dlibrary.com/answers/zh/commuting-paradox-compensating-income-relocation-decision/c03/) - Ch.4 Siting a Studio: the Net Present Value of Rent and the Team's Commute — Chapter 4 computes the net present value of a studio's site: with time at full rate, a central site at 1006.2 thousand beats the lowest-rent site at 1078.2 thousand. (https://5dlibrary.com/answers/en/commuting-paradox-compensating-income-relocation-decision/c04/) 第4章 工作室选址:租金与团队通勤的净现值 — 第四章计算工作室选址的净现值:按全价计入时间,交通居中的地点以 1006.2 千胜过租金最低地点的 1078.2 千。 (https://5dlibrary.com/answers/zh/commuting-paradox-compensating-income-relocation-decision/c04/) - Ch.5 Remote and Hybrid Work: the Marginal Arithmetic of Office Days — Reducing office days is another route to shorter commuting. Experimental evidence shows that people would on average give up 8 % of pay for the option of working from home, and that for largely independent jobs home working need not hurt output, though it may carry a hidden cost in promotion. The marginal arithmetic shows that the benefit of collaboration is concentrated in the first one or two office days and that the best number of days falls as the commute lengthens. Several prices for the same 22-minute commute differ several-fold, each answering a different question. The next chapter returns to relocation itself: when the place of work may change again, should one move or keep commuting, and is waiting worth anything? (https://5dlibrary.com/answers/en/commuting-paradox-compensating-income-relocation-decision/c05/) 第5章 远程与混合:到岗日数的边际账 — 减少到岗日子是缩短通勤的另一条路。实验证据显示,人们平均愿意为在家工作的选项让出 8 % 的薪酬,对独立性强的岗位,在家工作不一定损害产出,但可能带来提拔上的隐性代价。边际账显示,协作收益集中在头一两个到岗日,最优到岗天数随通勤变长而减少。同一段 22 分钟通勤的几种价格相差数倍,各自回答不同的问题。下一章回到搬迁本身:当工作地点可能再变时,搬家还是继续通勤,等待有没有价值。 (https://5dlibrary.com/answers/zh/commuting-paradox-compensating-income-relocation-decision/c05/) - Ch.6 Move or Commute: Real Options and the Value of Waiting — Chapter 6 uses real options to show that at a volatility of 0.20 the optimal threshold for moving is 2.175 times the net-present-value threshold. (https://5dlibrary.com/answers/en/commuting-paradox-compensating-income-relocation-decision/c06/) 第6章 搬家还是通勤:实物期权与等待的价值 — 第六章用实物期权说明,波动率 0.20 时搬家的最优门槛是净现值门槛的 2.175 倍。 (https://5dlibrary.com/answers/zh/commuting-paradox-compensating-income-relocation-decision/c06/) - Ch.7 The Foolish Old Man and Pascal: Moving Mountains, Moving House and Sitting in a Room — The share among men is 0.667, with a 95 % interval from 0.437 to 0.837; among women it is 0.250, from 0.120 to 0.449; overall it is 0.429, from 0.291 to 0.578. The samples are small and the intervals wide; the direction of the difference between men and women is clear, but its size is estimated very imprecisely. As far as our search reached, we found no large independent replication of the shock result, and after publication some commentators offered other readings of the claim that people dislike being alone with their thoughts. It can serve as an intriguing illustration of Pascal's claim, but not yet as a settled finding. (https://5dlibrary.com/answers/en/commuting-paradox-compensating-income-relocation-decision/c07/) 第7章 愚公与帕斯卡:移山、迁居与安坐一室 — 男性的比例是 0.667,95 % 区间是 0.437 到 0.837;女性是 0.250,区间是 0.120 到 0.449;合计 0.429,区间是 0.291 到 0.578。样本很小,区间很宽;男女之间差别的方向清楚,但幅度的估计很不精确。就我们检索所及,没有看到对这一电击结果的大规模独立重复;该研究发表后,也有评论者对「人们厌恶独处思考」的解读提出了不同看法。所以,它可以作为帕斯卡命题的一个有趣例证,但还不足以作为定论。 (https://5dlibrary.com/answers/zh/commuting-paradox-compensating-income-relocation-decision/c07/) - Ch.8 Closing the Loop: a Team Commuting Audit, a Siting Tool and a Pre-Registered Test — Chapter 8 gives a team commuting audit, the Site Time-Cost Calculator and a pre-registered test, and shows that a team of 40 can detect only differences above 0.532 points. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/commuting-paradox-compensating-income-relocation-decision/c08/) 第8章 实证闭环:团队通勤盘点、选址小工具与预注册检验 — 第八章给出团队通勤盘点、选址时间成本器与预注册检验,40 人的团队只能检测 0.532 分以上的差别。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/commuting-paradox-compensating-income-relocation-decision/c08/) ## 5DL-B-205 · Agarwood Sesquiterpene Volatilisation, Combinatorial Olfactory-Receptor Coding and the Time Constant of Olfactory Adaptation 中文书名:沉香倍半萜挥发动力学、嗅觉受体组合编码与嗅觉适应时间常数母典 DOI: pending Cite: Fan, Y. (2026). Agarwood Sesquiterpene Volatilisation, Combinatorial Olfactory-Receptor Coding and the Time Constant of Olfactory Adaptation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its rooms, wood, heaters and adaptation parameters are illustrative; all calculations are generated by its programs; it is general rather than individualised business guidance, does not discuss health effects and is not medical advice. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: This book recalculates two sayings of the agarwood and fragrance trade: that humans can discriminate a trillion smells, and that incense is better the stronger and longer it burns. Its rooms, wood, heaters and adaptation parameters are illustrative; all calculations are generated by its programs; it is general rather than individualised business guidance, does not discuss health effects and is not medical advice. Chapter 1 reproduces digit for digit the 2014 estimate of 1.72 × 10^12 by Bushdid and colleagues and sets the rebuttals of Gerkin and Castro and of Meister beside the original authors' case: the same formula can give anything from 4.56 × 103 to 1.54 × 10^29; holding one nose fixed and adding tests lifts the estimate from 10 to the power 15.27 to the power 26.15; a one-dimensional model that separates only 9 odours is credited with 10 to the power 18.19. The extrapolation fails and the number remains unknown. Chapter 2 shows that receptor code space is not the bottleneck, that reaching 10 to the power 12 needs 12.6 to 17.2 perceptual dimensions, and demonstrates olfactory white. Chapter 3 models release from a heated chip with first-order kinetics and the Clausius-Clapeyron relation, with a half-life of 11.6 minutes for the main fraction at 150 degrees Celsius. Chapter 4 shows that concentration peaks after lighting, at minute 38.2 in a closed reception room, and that a sorptive sink of soft furnishings lowers the peak to 13.2. Chapter 5 writes olfactory adaptation as a process with a time constant: with illustrative parameters the host keeps only 0.250 of a fresh nose's sensitivity at minute 15, and chasing it with more incense takes 12.4 times the concentration; whether the host can "hardly smell it" depends on the depth of adaptation and must be measured. Chapter 6 gives a scheduling rule (a lead of 17.7 minutes in a closed reception room) and a day's ledger: lighting on cue uses 0.375 of the wood of continuous burning for 0.671 of the entry sensation. Chapter 7 examines the water-sinking grades and burning method of the Chen Shi Xiang Pu (a chip lasts 45.2 times longer at 130 degrees than at 220) and Theophrastus' On Odours. Chapter 8 gives a shop-level loop, the distilled tool Scent Scheduler, a pre-registered protocol (459 trading days if purchases are the outcome) and the logic of authenticity checks (sinking in water has a likelihood ratio of only 1.75). Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:本书重算两句在沉香与香氛生意里流传的常识:人能分辨一万亿种气味;香点得越浓越久越好。书中的房间、香材、香炉与适应参数均为示意;全部计算由书中程序生成;本书是非个人化经营建议,不讨论健康影响,不是医疗建议。 第一章逐位复现 Bushdid 等人 2014 年的估计 1.72 × 10^12,再并陈 Gerkin 与 Castro、Meister 的反驳与原作者一方的理由:同一公式可以给出从 4.56 × 103 到 1.54 × 10^29 的任何数;固定一个鼻子、只增加测试次数,估计从 10 的 15.27 次方爬到 26.15 次方;只能区分 9 种气味的一维模型也被说成 10 的 18.19 次方。结论是推算不成立、数目仍未知。第二章说明受体的代码空间不是瓶颈,要达到 10 的 12 次方需要 12.6 到 17.2 个感知维度,并演示嗅觉白。第三章用一级动力学与克劳修斯—克拉珀龙关系描述香片释放,150 摄氏度下主体组分半衰期 11.6 分钟。第四章说明浓度峰值在点香之后,紧闭会客室在第 38.2 分钟,软装吸附汇把峰值压到 13.2。 第五章把嗅觉适应写成带时间常数的过程:示意参数下,点香后第 15 分钟主人只剩新鲜鼻子 0.250 的灵敏度,用加量追赶需要 12.4 倍浓度;「主人几乎闻不到」依赖适应深度,须实测。第六章给出排程规则(紧闭会客室提前 17.7 分钟)与整日账本:按客到点香用持续点香 0.375 的料换来 0.671 的进门感受。第七章考辨《陈氏香谱》的沉水分级与焚香法(130 摄氏度下一片香可用时间是 220 摄氏度的 45.2 倍)及泰奥弗拉斯托斯《论气味》。第八章给出门店闭环、蒸馏小工具「香氛排程器」、预注册方案(以成交率为结局需 459 个营业日)与真伪判据(入水下沉似然比仅 1.75)。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A Trillion Smells? Re-deriving the Bushdid Estimator and Its Two Rebuttals — Chapter 1 reproduces digit for digit the 2014 estimate of 1.72 × 10^12 by Bushdid and colleagues and sets the rebuttals of Gerkin and Castro and of Meister beside the original authors' case: the same formula can give anything from 4.56 × 103 to 1.54 × 10^29; holding one nose fixed and adding tests lifts the estimate from 10 to the power 15.27 to the power 26.15; a one-dimensional model that separates only 9 odours is credited with 10 to the power 18.19. The extrapolation fails and the number remains unknown. (https://5dlibrary.com/answers/en/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c01/) 第1章 一万亿种气味?重算 Bushdid 估计量及其两份反驳 — 第一章逐位复现 Bushdid 等人 2014 年的估计 1.72 × 10^12,再并陈 Gerkin 与 Castro、Meister 的反驳与原作者一方的理由:同一公式可以给出从 4.56 × 103 到 1.54 × 10^29 的任何数;固定一个鼻子、只增加测试次数,估计从 10 的 15.27 次方爬到 26.15 次方;只能区分 9 种气味的一维模型也被说成 10 的 18.19 次方。结论是推算不成立、数目仍未知。 (https://5dlibrary.com/answers/zh/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c01/) - Ch.2 Combinatorial Receptor Codes and the Dimensionality of Odour Space: Why Four Hundred Receptors Do Not Buy an Astronomical Number of Smells — Chapter 2 shows that receptor code space is not the bottleneck, that reaching 10 to the power 12 needs 12.6 to 17.2 perceptual dimensions, and demonstrates olfactory white. (https://5dlibrary.com/answers/en/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c02/) 第2章 受体组合编码与气味空间的维数:四百种受体为何换不来天文数字的气味 — 第二章说明受体的代码空间不是瓶颈,要达到 10 的 12 次方需要 12.6 到 17.2 个感知维度,并演示嗅觉白。 (https://5dlibrary.com/answers/zh/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c02/) - Ch.3 Off the Ember: First-Order Release Kinetics of Agarwood Volatiles and a Temperature Scale — Chapter 3 models release from a heated chip with first-order kinetics and the Clausius-Clapeyron relation, with a half-life of 11.6 minutes for the main fraction at 150 degrees Celsius. (https://5dlibrary.com/answers/en/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c03/) 第3章 隔火之下:沉香挥发组分的一级释放动力学与温度标尺 — 第三章用一级动力学与克劳修斯—克拉珀龙关系描述香片释放,150 摄氏度下主体组分半衰期 11.6 分钟。 (https://5dlibrary.com/answers/zh/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c03/) - Ch.4 A Room Is a Reactor: Well-Mixed Ventilation and the Surface Sink — Chapter 4 shows that concentration peaks after lighting, at minute 38.2 in a closed reception room, and that a sorptive sink of soft furnishings lowers the peak to 13.2. (https://5dlibrary.com/answers/en/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c04/) 第4章 一间屋子就是一个反应器:全混流换气稀释与墙面吸附汇 — 第四章说明浓度峰值在点香之后,紧闭会客室在第 38.2 分钟,软装吸附汇把峰值压到 13.2。 (https://5dlibrary.com/answers/zh/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c04/) - Ch.5 The Time Constant of Adaptation: Why the Host Cannot Smell His Own Incense — Chapter 5 writes olfactory adaptation as a process with a time constant: with illustrative parameters the host keeps only 0.250 of a fresh nose's sensitivity at minute 15, and chasing it with more incense takes 12.4 times the concentration; whether the host can "hardly smell it" depends on the depth of adaptation and must be measured. (https://5dlibrary.com/answers/en/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c05/) 第5章 适应时间常数:主人为什么闻不到自己点的香 — 第五章把嗅觉适应写成带时间常数的过程:示意参数下,点香后第 15 分钟主人只剩新鲜鼻子 0.250 的灵敏度,用加量追赶需要 12.4 倍浓度;「主人几乎闻不到」依赖适应深度,须实测。 (https://5dlibrary.com/answers/zh/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c05/) - Ch.6 Light It Before the Guest Arrives: a Lead-Time Rule, a Rule Map and a Day's Ledger — Chapter 6 gives a scheduling rule (a lead of 17.7 minutes in a closed reception room) and a day's ledger: lighting on cue uses 0.375 of the wood of continuous burning for 0.671 of the entry sensation. (https://5dlibrary.com/answers/en/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c06/) 第6章 客到前几分钟点香:提前量规则、规则地图与整日账本 — 第六章给出排程规则(紧闭会客室提前 17.7 分钟)与整日账本:按客到点香用持续点香 0.375 的料换来 0.671 的进门感受。 (https://5dlibrary.com/answers/zh/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c06/) - Ch.7 Classical Sources East and West: Chen Jing's Chen Shi Xiang Pu and Theophrastus' On Odours — Chapter 7 examines the water-sinking grades and burning method of the Chen Shi Xiang Pu (a chip lasts 45.2 times longer at 130 degrees than at 220) and Theophrastus' On Odours. (https://5dlibrary.com/answers/en/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c07/) 第7章 东西方原典考辨:《陈氏香谱》与泰奥弗拉斯托斯《论气味》 — 第七章考辨《陈氏香谱》的沉水分级与焚香法(130 摄氏度下一片香可用时间是 220 摄氏度的 45.2 倍)及泰奥弗拉斯托斯《论气味》。 (https://5dlibrary.com/answers/zh/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c07/) - Ch.8 Closing the Loop: a Shop Scheduling Log, a Pre-Registered Test and Authenticity Checks — Chapter 8 gives a shop-level loop, the distilled tool Scent Scheduler, a pre-registered protocol (459 trading days if purchases are the outcome) and the logic of authenticity checks (sinking in water has a likelihood ratio of only 1.75). Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c08/) 第8章 实证闭环:门店排程日志、预注册检验与真伪判据 — 第八章给出门店闭环、蒸馏小工具「香氛排程器」、预注册方案(以成交率为结局需 459 个营业日)与真伪判据(入水下沉似然比仅 1.75)。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/agarwood-sesquiterpene-olfactory-coding-adaptation-time-constant/c08/) ## 5DL-B-206 · Hydrostatic Specific Gravity and Bayesian Jadeite Identification, Window-Cut Signalling Games and the Winner's Curse in Rough-Jade Tenders 中文书名:翡翠静水比重贝叶斯鉴别、赌石开窗信号博弈与原石拍卖赢家诅咒母典 DOI: pending Cite: Fan, Y. (2026). Hydrostatic Specific Gravity and Bayesian Jadeite Identification, Window-Cut Signalling Games and the Winner's Curse in Rough-Jade Tenders (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not personalised legal or financial advice, nor an identification report on any stone. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: This book overturns a popular belief: that gambling on jadeite rough, "poor or rich with one cut", roughly breaks even over time. Its proposition is that, taken as a whole, buying windowed rough has a negative expected return from three sources: the window is a sample chosen by the seller, the decision to sell as rough is itself a selection, and a sealed tender is won by the most optimistic estimate. Published research supplies the general laws: in the jar-of-coins experiment the mean winning bid was 1.251 times the true value; in the laboratory larger groups lost more, yet a field study of oil leases found that professional bidders as a group did not lose. Its parameters are illustrative; all calculations are generated by its programs; it is not personalised legal or financial advice, nor an identification report on any stone. Chapter 1 decomposes returns with a model: window selection alone gives a lone buyer a portfolio return of −0.225, while 0.313 of buyers who purchase only five stones see a profit in their ledgers. Chapter 2 covers hydrostatic weighing: a 2-gram stone on a 0.01-gram balance gives a specific-gravity reading with standard deviation 0.0393; specific gravity separates jadeite from nephrite but not from B jade, with an error rate still of 0.2833 for a 10-gram stone. Chapter 3 chains specific gravity, refractive index and fluorescence into likelihood ratios: the same readings give a jadeite posterior of 0.8622 in a tourist market and 0.9657 at a reputable dealer. Chapter 4 computes the window bias (a seller who examines four spots makes an uncorrecting buyer overpay 0.311), a separating equilibrium in window size (when faking is cheap a naive buyer overpays 42.0 per stone) and the unravelling of unwindowed rough. Chapter 5 simulates N-bidder common-value tenders: under the naive rule the winner of a twelve-bidder tender loses 19.46 per auction on average, while bidding as if one had already won restores the reserved profit. Chapter 6 folds the three corrections into a bid cap: 0.309 of the window valuation with a chosen window, six bidders and a reliable certificate, and 0.274 with counter tests only; in a tender dominated by naive buyers one should leave; and the Kelly fraction under the illustrative edge is 0.090. Chapter 7 examines Song Yingxing's jade workers who "judge the rough and set the price" in the Tiangong Kaiwu and Pliny's weight test in Book 37 of the Natural History: telling jadeite from B jade by hand has a discriminating power of only 0.5446. Chapter 8 gives the buyer's flow card (a false-pass rate of 0.0068 once infrared is added), a compliant script for declining to buy rough for clients, the distilled Jade Likelihood Identifier and a pre-registered testing plan (test one needs 223 stones at a standard deviation of 0.9). Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:本书要推翻一个流行的信念:「一刀穷一刀富」的翡翠原石赌局长期大致扯平,抢的料没有差的。本书的命题是,作为一个整体,买入开窗原石的期望回报为负,来源有三:窗口是卖方挑过的样本,以原石出售本身经过挑选,暗标中标的总是最乐观的估计。已发表的研究给出一般规律:罐装硬币实验中中标价平均是真值的 1.251 倍;实验里人数越多亏得越多,但石油租让的现场研究发现专业投标企业整体并未亏损,本书把两方面证据并列。书中参数为示意;全部计算由书中程序生成;本书不是个人化的法律或理财建议,也不是任何一件石头的鉴定意见。 第一章用模型拆解回报:只靠挑窗一项,单一买家的组合回报就是 −0.225,而只买五块的买手中有 0.313 的人账本盈利。第二章讲静水称重:0.01 克天平称 2 克石头,比重读数标准差 0.0393;比重分得开翡翠与软玉,分不开翡翠与 B 货,10 克石头的误判率仍有 0.2833。第三章把比重、折射率与荧光串成似然比:同一组读数在游客市场给出 0.8622 的翡翠后验,在信誉商家给出 0.9657。第四章计算挑窗偏差(卖方看四个点,买方不修正多付 0.311)、开窗大小的分离均衡(造假便宜时天真买家每块多付 42.0)与全赌料的拆解。第五章模拟 N 人共同价值暗标:天真规则下十二人竞价的中标者平均每场亏 19.46,按「假如我已中标」修正后利润回到预留水平。 第六章把三项修正折成出价上限:挑窗、六人竞价、证书可靠时为窗口估值的 0.309,只做门店检验时为 0.274;在天真买家占多数的公盘里应当离场;凯利比例在示意优势下为 0.090。第七章考辨宋应星《天工开物·珠玉》的「玉工辨璞高下定价」与普林尼《博物志》卷三十七的重量检验:凭手掂区分翡翠与 B 货的判别力只有 0.5446。第八章给出买手流程卡(加上红外后误放率 0.0068)、拒绝代购赌石的合规话术、蒸馏小工具「玉石鉴别似然器」与预注册检验方案(检验一在标准差 0.9 时需要 223 块)。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Ledger of Rough: Adverse Selection, Chosen Windows and the Winner's Curse — Chapter 1 decomposes returns with a model: window selection alone gives a lone buyer a portfolio return of −0.225, while 0.313 of buyers who purchase only five stones see a profit in their ledgers. (https://5dlibrary.com/answers/en/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c01/) 第1章 赌石的账本:逆向选择、挑窗与赢家诅咒的三重负期望 — 第一章用模型拆解回报:只靠挑窗一项,单一买家的组合回报就是 −0.225,而只买五块的买手中有 0.313 的人账本盈利。 (https://5dlibrary.com/answers/zh/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c01/) - Ch.2 Hydrostatic Weighing: Specific-Gravity Error and the Limits of a Single Reading — Chapter 2 covers hydrostatic weighing: a 2-gram stone on a 0.01-gram balance gives a specific-gravity reading with standard deviation 0.0393; specific gravity separates jadeite from nephrite but not from B jade, with an error rate still of 0.2833 for a 10-gram stone. (https://5dlibrary.com/answers/en/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c02/) 第2章 静水称重:比重测量误差与单一读数的鉴别边界 — 第二章讲静水称重:0.01 克天平称 2 克石头,比重读数标准差 0.0393;比重分得开翡翠与软玉,分不开翡翠与 B 货,10 克石头的误判率仍有 0.2833。 (https://5dlibrary.com/answers/zh/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c02/) - Ch.3 Chained Likelihood Ratios: Bayesian Identification by Specific Gravity, Refractive Index and Fluorescence — Chapter 3 chains specific gravity, refractive index and fluorescence into likelihood ratios: the same readings give a jadeite posterior of 0.8622 in a tourist market and 0.9657 at a reputable dealer. (https://5dlibrary.com/answers/en/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c03/) 第3章 似然比串联:比重、折射率与荧光的贝叶斯鉴别 — 第三章把比重、折射率与荧光串成似然比:同一组读数在游客市场给出 0.8622 的翡翠后验,在信誉商家给出 0.9657。 (https://5dlibrary.com/answers/zh/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c03/) - Ch.4 Window Signals: Best-Spot Bias, a Separating Equilibrium and the Unravelling of Unwindowed Rough — Chapter 4 computes the window bias (a seller who examines four spots makes an uncorrecting buyer overpay 0.311), a separating equilibrium in window size (when faking is cheap a naive buyer overpays 42.0 per stone) and the unravelling of unwindowed rough. (https://5dlibrary.com/answers/en/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c04/) 第4章 开窗信号:最优点位偏差、分离均衡与全赌料的拆解 — 第四章计算挑窗偏差(卖方看四个点,买方不修正多付 0.311)、开窗大小的分离均衡(造假便宜时天真买家每块多付 42.0)与全赌料的拆解。 (https://5dlibrary.com/answers/zh/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c04/) - Ch.5 Sealed Tenders: a Monte Carlo of the Winner's Curse in N-Bidder Common-Value Auctions — Chapter 5 simulates N-bidder common-value tenders: under the naive rule the winner of a twelve-bidder tender loses 19.46 per auction on average, while bidding as if one had already won restores the reserved profit. (https://5dlibrary.com/answers/en/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c05/) 第5章 公盘暗标:N 人共同价值拍卖的赢家诅咒蒙特卡洛 — 第五章模拟 N 人共同价值暗标:天真规则下十二人竞价的中标者平均每场亏 19.46,按「假如我已中标」修正后利润回到预留水平。 (https://5dlibrary.com/answers/zh/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c05/) - Ch.6 The Bid Cap: Folding Identification, Window and Curse into One Number, with Stake Size and Ruin — Chapter 6 folds the three corrections into a bid cap: 0.309 of the window valuation with a chosen window, six bidders and a reliable certificate, and 0.274 with counter tests only; in a tender dominated by naive buyers one should leave; and the Kelly fraction under the illustrative edge is 0.090. (https://5dlibrary.com/answers/en/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c06/) 第6章 出价上限:把鉴别、挑窗与诅咒折进一个数,以及仓位与出局 — 第六章把三项修正折成出价上限:挑窗、六人竞价、证书可靠时为窗口估值的 0.309,只做门店检验时为 0.274;在天真买家占多数的公盘里应当离场;凯利比例在示意优势下为 0.090。 (https://5dlibrary.com/answers/zh/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c06/) - Ch.7 Classical Sources East and West: the Tiangong Kaiwu on Jade and Book 37 of Pliny's Natural History — Chapter 7 examines Song Yingxing's jade workers who "judge the rough and set the price" in the Tiangong Kaiwu and Pliny's weight test in Book 37 of the Natural History: telling jadeite from B jade by hand has a discriminating power of only 0.5446. (https://5dlibrary.com/answers/en/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c07/) 第7章 东西方原典考辨:《天工开物·珠玉》与普林尼《博物志》卷三十七 — 第七章考辨宋应星《天工开物·珠玉》的「玉工辨璞高下定价」与普林尼《博物志》卷三十七的重量检验:凭手掂区分翡翠与 B 货的判别力只有 0.5446。 (https://5dlibrary.com/answers/zh/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c07/) - Ch.8 Closing the Loop: the Buyer's Flow Card, the Bid Cap, a Compliant Script and Pre-Registered Tests — Chapter 8 gives the buyer's flow card (a false-pass rate of 0.0068 once infrared is added), a compliant script for declining to buy rough for clients, the distilled Jade Likelihood Identifier and a pre-registered testing plan (test one needs 223 stones at a standard deviation of 0.9). Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c08/) 第8章 实证闭环:买手流程卡、出价上限、合规话术与预注册检验 — 第八章给出买手流程卡(加上红外后误放率 0.0068)、拒绝代购赌石的合规话术、蒸馏小工具「玉石鉴别似然器」与预注册检验方案(检验一在标准差 0.9 时需要 223 块)。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/jadeite-specific-gravity-bayes-window-signalling-winners-curse/c08/) ## 5DL-B-207 · Escapement Isochronism Errors, Hairspring Thermal Coefficients and the Half-Life of Scarcity Premiums in the Secondary Watch Market 中文书名:机械擒纵等时性误差、游丝温度系数与名表二级市场稀缺溢价半衰期母典 DOI: pending Cite: Fan, Y. (2026). Escapement Isochronism Errors, Hairspring Thermal Coefficients and the Half-Life of Scarcity Premiums in the Secondary Watch Market (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: This book tests two popular sentences of the watch trade: chronometer-certified mechanical watches are "precise", and prestige watches "hold their value". It writes both as formulas with parameters and argues the opposite: on tolerance, a mechanical chronometer is still an order of magnitude worse than a cheap quartz watch; and holding value is concentrated in a few references with rationed allocation, so that once allocation loosens the secondary premium collapses on a time scale of months. Its parameters are illustrative; all calculations are generated by its programs; the book is a non-personalised analysis of business and finance, not personalised legal or financial advice, and anything regulatory is subject to manual review before publication. Chapter 1 sets out two ledgers: the mechanical chronometer tolerance is 12.0 times wider than a quartz watch erring by 15 seconds a month and 85.7 times wider than the quartz chronometer standard; collectible watches in auction samples show positive returns (a published study reports 5.5 % a year in real terms), while in an illustrative model 0.816 of all gains sit in a tenth of the references. Chapters 2 to 4 take the daily rate apart: the tolerance budget of the balance-spring oscillator, a steel spring at −10.886 seconds a day per degree against a chronometer limit of 0.6, the middle-temperature error that linear thermal criteria cannot see, Airy's escapement error, and the poise error changing sign at 219.5 degrees. Chapter 5 simulates a chronometer test and computes Allan deviations, finding a typical wrist error only 4.6 times that of a 15-second-a-month quartz watch and 13.8 times that of a 5-second one, so "an order of magnitude" depends on the benchmark and the statistic. Chapter 6 shows that fitting a premium half-life needs the long-run level first, or 5.69 months becomes 8.61, and uses a repeat-sales index and a supply-and-demand model to separate the end point set by looser allocation from the speed of collapse set by market structure. Chapter 7 examines the water-driven tianheng escapement of Su Song's Xin Yixiang Fayao and the cycloidal isochronism of Huygens's Horologium Oscillatorium. Chapter 8 gives a buy-side quote sheet for pre-owned watches, the Pre-owned Watch Buy-side Quoter, a compliance red line for "holds its value" (a buyer at the secondary price has only a 0.044 probability of no loss after three years) and a pre-registered testing protocol. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:本书检验名表行业两句流行的话:天文台认证的机械表「精准」,名表「保值」。它把两句话写成带参数的算式,并给出相反的主张:按容差比较,天文台机械表仍比廉价石英表差一个数量级;保值集中在少数受配货限制的型号上,配货放松后二级溢价会以月为单位塌缩。书中参数为示意;全部计算由书中程序生成;本书是非个人化的经营与理财分析,不构成个人化法律或理财建议,涉及法规之处以上架前人工复核为准。 第一章并排摆出两本账:天文台机械表的容差比月差 15 秒的石英表宽 12.0 倍,比天文台石英标准宽 85.7 倍;拍卖样本里的收藏表有正收益(已发表研究报告实际年化 5.5 %),而示意模型中全部增值的 0.816 集中在一成型号上。第二至四章拆开日差的物理来源:摆轮—游丝谐振器的公差预算,钢游丝每度 −10.886 秒每日而天文台判据只许 0.6,线性温度判据看不见的中温误差,艾里擒纵误差,以及偏重误差在 219.5 度变号。第五章模拟天文台检验并计算阿伦偏差,发现腕上典型误差对月差 15 秒石英表只差 4.6 倍、对月差 5 秒石英表差 13.8 倍,因此「一个数量级」取决于参照物与口径。 第六章说明拟合溢价半衰期必须先估长期水平,否则会把 5.69 个月误估成 8.61 个月,并用重复交易指数与供求模型区分配货放松决定的溢价终点与市场结构决定的塌缩速度。第七章考辨苏颂《新仪象法要》的天衡水力擒纵与惠更斯《摆钟论》的摆线等时性。第八章给出二手表收购报价表、名表收购报价器、「保值」话术的合规红线(以二级价买入者三年不亏的概率只有 0.044)与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Two Ledgers: the Order-of-Magnitude Gap between Chronometer Certificates and Quartz, and the Concentration of "Holding Value" — Chapter 1 sets out two ledgers: the mechanical chronometer tolerance is 12.0 times wider than a quartz watch erring by 15 seconds a month and 85.7 times wider than the quartz chronometer standard; collectible watches in auction samples show positive returns (a published study reports 5.5 % a year in real terms), while in an illustrative model 0.816 of all gains sit in a tenth of the references. (https://5dlibrary.com/answers/en/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c01/) 第1章 两本账:天文台证书与石英表的数量级差距,以及「保值」的集中度 — 第一章并排摆出两本账:天文台机械表的容差比月差 15 秒的石英表宽 12.0 倍,比天文台石英标准宽 85.7 倍;拍卖样本里的收藏表有正收益(已发表研究报告实际年化 5.5 %),而示意模型中全部增值的 0.816 集中在一成型号上。 (https://5dlibrary.com/answers/zh/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c01/) - Ch.2 The Balance-Spring Oscillator: a Tolerance Budget for T = 2π√(I/k) — Chapters 2 to 4 take the daily rate apart: the tolerance budget of the balance-spring oscillator, a steel spring at −10.886 seconds a day per degree against a chronometer limit of 0.6, the middle-temperature error that linear thermal criteria cannot see, Airy's escapement error, and the poise error changing sign at 219.5 degrees. (https://5dlibrary.com/answers/en/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c02/) 第2章 摆轮—游丝谐振器:周期公式 T = 2π√(I/k) 的公差预算 — 第二至四章拆开日差的物理来源:摆轮—游丝谐振器的公差预算,钢游丝每度 −10.886 秒每日而天文台判据只许 0.6,线性温度判据看不见的中温误差,艾里擒纵误差,以及偏重误差在 219.5 度变号。 (https://5dlibrary.com/answers/zh/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c02/) - Ch.3 Hairspring Thermal Coefficients: Elastic Modulus, Expansion and the Middle-Temperature Error — The thermal coefficient of rate is 86400 times half the algebraic sum of the modulus coefficient, three times the spring's expansion and minus twice the balance's expansion. A steel spring with a brass balance gives −10.886 seconds a day per degree and loses 163.3 seconds a day over 15 degrees; a tuned iron-nickel spring gives 0.518 and an oxide-tuned silicon spring −0.009. Over an illustrative day the steel spring accumulates 1632.9 seconds a month, iron-nickel 77.7, silicon −1.3 and uncompensated quartz 3.6. A COSC-style linear criterion cannot see the middle-temperature error, which with a large quadratic coefficient costs 1.944 seconds a day at 23 degrees. (https://5dlibrary.com/answers/en/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c03/) 第3章 游丝温度系数:弹性模量、线胀与中温误差 — 日差温度系数等于 86400 乘以弹性模量温度系数、三倍游丝线胀与两倍摆轮线胀之代数和的一半。钢游丝配黄铜摆轮是 −10.886 秒每日每度,温度变化 15 度时每天慢 163.3 秒;调好的铁镍合金为 0.518,氧化层调好的硅游丝为 −0.009。在示意的一天温度历程里,钢游丝每月累计 1632.9 秒,铁镍合金 77.7 秒,硅游丝 −1.3 秒,未补偿石英 3.6 秒。COSC 式的线性判据看不见中温误差,二次系数大时 23 度可慢 1.944 秒每日。 (https://5dlibrary.com/answers/zh/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c03/) - Ch.4 Isochronism Errors: Airy's Escapement Error, the 220-Degree Amplitude and the Power Reserve — Airy's conclusion can be written as a relative rate error per swing of $-\tan\psi/(2Q)$: a one-degree lead gains 0.617 seconds a day at 280 degrees and 1.080 at 160, while a symmetric impulse gives 0.000. The poise error follows the Bessel function and changes sign at 219.5 degrees, which is where "220 degrees" comes from. With all three terms, an illustrative watch drifts from 0.144 to 1.033 seconds a day over a 42 hour power reserve. A buying inspection should record amplitude together with rate, and an amplitude below 220 degrees is itself a signal to service. (https://5dlibrary.com/answers/en/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c04/) 第4章 等时性误差:艾里擒纵误差、220 度振幅与动力储存 — 艾里的结论可以写成每次摆动的相对日差 $-\tan\psi/(2Q)$:冲量偏前 1 度,振幅 280 度时快 0.617 秒,160 度时快 1.080 秒;对称冲量的误差为 0.000。偏重误差按一阶贝塞尔函数变化,在 219.5 度变号,这就是「220 度」的来源。三项合计,一只示意表在 42 小时动力储存里日差从 0.144 漂到 1.033 秒每日。收购检测应把振幅与日差同时记录,低于 220 度的振幅本身就是送修信号。 (https://5dlibrary.com/answers/zh/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c04/) - Ch.5 The Statistics of Daily Rates: a Simulated Chronometer Test and the Allan Deviation — Chapter 5 simulates a chronometer test and computes Allan deviations, finding a typical wrist error only 4.6 times that of a 15-second-a-month quartz watch and 13.8 times that of a 5-second one, so "an order of magnitude" depends on the benchmark and the statistic. (https://5dlibrary.com/answers/en/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c05/) 第5章 日差的统计:模拟天文台检验与阿伦偏差 — 第五章模拟天文台检验并计算阿伦偏差,发现腕上典型误差对月差 15 秒石英表只差 4.6 倍、对月差 5 秒石英表差 13.8 倍,因此「一个数量级」取决于参照物与口径。 (https://5dlibrary.com/answers/zh/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c05/) - Ch.6 Half-Lives of Scarcity Premiums: Exponential Fits, Repeat-Sales Indices and an Allocation Model — Chapter 6 shows that fitting a premium half-life needs the long-run level first, or 5.69 months becomes 8.61, and uses a repeat-sales index and a supply-and-demand model to separate the end point set by looser allocation from the speed of collapse set by market structure. (https://5dlibrary.com/answers/en/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c06/) 第6章 稀缺溢价半衰期:指数衰减拟合、重复交易指数与配货模型 — 第六章说明拟合溢价半衰期必须先估长期水平,否则会把 5.69 个月误估成 8.61 个月,并用重复交易指数与供求模型区分配货放松决定的溢价终点与市场结构决定的塌缩速度。 (https://5dlibrary.com/answers/zh/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c06/) - Ch.7 Eastern and Western Sources Examined: Su Song's "Heavenly Balance" and Huygens's Horologium Oscillatorium — Su Song's tianheng is a "fill, then release" escapement with no natural frequency, whose beat is set by flow: one per cent more drive makes it gain 864.0 seconds a day, while a balance changes by −0.021; compensating vessels in series reduce the error from head offsets stage by stage. Huygens proved that the circular pendulum is not isochronous and the cycloidal pendulum is; the exact formula gives a loss of 164.78 seconds a day at 10 degrees. His geometry was entirely right, though in practice cycloidal cheeks gave way to small amplitudes. The two sources correspond to the two halves of today's mechanical watch: the escapement and the resonator. (https://5dlibrary.com/answers/en/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c07/) 第7章 东西方原典考辨:苏颂的「天衡」与惠更斯的《摆钟论》 — 苏颂的天衡是一种没有固有频率的「积满再放」式擒纵,节拍由流量决定:动力增加百分之一,它每天快 864.0 秒,而摆轮只变 −0.021 秒;多级补偿壶把水头偏移造成的日差逐级压缩。惠更斯证明了圆弧摆不等时、摆线摆等时,精确公式显示摆幅 10 度时每天慢 164.78 秒;他的几何结论完全正确,摆线夹板却在工程上被小摆幅取代。两部原典分别对应今天机械表的两半:擒纵机构与谐振器。 (https://5dlibrary.com/answers/zh/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c07/) - Ch.8 Closing the Loop: a Buy-Side Quote Sheet for Pre-Owned Watches, the Compliance Boundary of "Holds Its Value", and Pre-Registered Tests — Chapter 8 gives a buy-side quote sheet for pre-owned watches, the Pre-owned Watch Buy-side Quoter, a compliance red line for "holds its value" (a buyer at the secondary price has only a 0.044 probability of no loss after three years) and a pre-registered testing protocol. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c08/) 第8章 实证闭环:二手表收购报价表、「保值」话术的合规边界与预注册检验 — 第八章给出二手表收购报价表、名表收购报价器、「保值」话术的合规红线(以二级价买入者三年不亏的概率只有 0.044)与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/escapement-isochronism-hairspring-thermal-watch-premium-half-life/c08/) ## 5DL-B-208 · Repeat-Sales Selection Bias in Collectibles, Forgery Priors and the Illiquidity Discount of Alternative Assets 中文书名:收藏品重复成交指数选择偏差、赝品先验与另类资产流动性折价母典 DOI: pending Cite: Fan, Y. (2026). Repeat-Sales Selection Bias in Collectibles, Forgery Priors and the Illiquidity Discount of Alternative Assets (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not personalised legal or financial advice, and fee schedules and rules are subject to manual review before release. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: The claim that art returns more than equities over the long run rests mostly on repeat-sales indices, and whether a work is sold again depends on whether it has risen or fallen. This book tests the claim with runnable arithmetic. A study of 32,928 repeat-sold paintings, once corrected for selection, moves the average annual return of a broad painting portfolio from 8.7% to 6.3% and its Sharpe ratio from 0.27 to 0.11; the book then deducts forgery, fees, holding costs and illiquidity one by one. Its parameters are illustrative; all calculations are generated by its programs; it is not personalised legal or financial advice, and fee schedules and rules are subject to manual review before release. Chapter 1 sets out the original evidence and uses the smallest model to show that the more the sale decision responds to gains, the more the index overstates. Chapter 2 writes the Bailey-Muth-Nourse regression by hand and discusses Case and Shiller's weighting and the arithmetic and geometric conventions. Chapter 3 separates resale timing, permanent exit and censoring in a twenty-year dynamic market: timing alone biases the index by about 0.0113 a year, permanent exit lifts this to 0.0263 or more, and correction needs a model of who sells. Chapter 4 treats forgery priors and the value of authentication with Bayes' rule. Chapter 5 shows that at the corrected return nominal break-even takes only 4.7 years, but with a holding cost of 0.01 a year catching a 0.05 benchmark takes 127.6 years; a single piece has a 0.389 chance of being below its total cost after ten years. Chapter 6 gives a perfect-timing bound of 0.2081 on the illiquidity discount and shows that a seller who must raise cash within three months loses 0.1715 more. Chapter 7 examines Mi Fu's distinction between dilettante and connoisseur in his History of Painting, his forgery prior of two genuine Li Cheng paintings against three hundred fakes, and Vasari's sleeping Cupid and copied papal portrait in the Lives. Chapter 8 demonstrates an allocation cap on alternative assets for wealthy clients: the promotional view allocates 0.251, the full-cost view zero; it also gives a pre-purchase due-diligence checklist, the distilled tool Collectibles Return Debiaser and a pre-registered falsification plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:「艺术品的长期回报跑赢股票」这句话,多半建立在重复成交指数之上,而一件作品会不会被再次卖出,本身取决于它涨了还是跌了。本书用可运行的算术检验这句话:一项覆盖 32,928 幅重复成交画作的研究在修正选择偏差之后,把宽泛画作组合的年均回报从 8.7% 调到 6.3%,夏普比率从 0.27 调到 0.11;本书再把赝品、佣金、持有成本与流动性逐项扣除。书中参数为示意;全部计算由书中程序生成;本书非个人化法律或理财建议,费率与法规以上架前人工复核为准。 第一章陈述原始证据,并用最小模型展示卖出概率对涨幅越敏感、指数高估越多。第二章手写 Bailey-Muth-Nourse 回归,讨论 Case 与 Shiller 的加权以及算术与几何口径。第三章在二十年的动态市场中区分转售时机、永久退出与截尾:单纯挑时机的偏差每年约 0.0113,加入永久退出后升到 0.0263 以上;纠正需要一个关于谁会卖的模型。第四章以贝叶斯定理处理赝品先验与鉴定的信息价值。第五章说明按修正后的回报名义回本只需 4.7 年,但在每年 0.01 的持有成本下追上 0.05 的基准需要 127.6 年;单件作品十年后低于总支出的概率是 0.389。第六章给出完美择时的流动性折扣上界 0.2081,并说明三个月内必须变现的卖家要多损失 0.1715。 第七章考辨米芾《画史》的好事者与赏鉴家之分、「李成真见两本,伪见三百本」的赝品先验,以及瓦萨里《名人传》中的沉睡丘比特与教皇像复制品。第八章演示高净值客户另类资产配置上限:宣传口径的模型配置为 0.251,扣除各项成本后为零;并给出藏品入手前的尽调清单、蒸馏小工具「收藏回报去偏器」与一份预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 After the Correction: Selection Bias in Repeat-Sales Samples — A repeat-sales index sees only the works that sell again, and whether a work sells again depends on how much it has gained. A large study that corrects for this selection moves the average annual return of paintings from 8.7% to 6.3% and the Sharpe ratio from 0.27 to 0.11; other studies that handle their samples carefully also report long-run returns below those of equities. The smallest model shows that the more sensitive the sale decision is to the gain, the more the index overstates, and the bias can be as large as the true return. The next chapter returns to the method itself: how the regression of Bailey, Muth and Nourse turns pairs of prices into an index, and when it is unbiased. (https://5dlibrary.com/answers/en/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c01/) 第1章 修正之后:重复成交样本里的选择偏差 — 重复成交指数只看得见被再次卖出的作品,而一件作品会不会被再卖,取决于它涨了多少。大样本研究在修正这种选择之后,把画作的年均回报从 8.7% 调到 6.3%,夏普比率从 0.27 调到 0.11;其他认真处理过样本的研究,也给出了低于股票的长期回报。最小模型表明,卖出概率对涨幅越敏感,指数就越高估,偏差可以和真实回报一样大。下一章回到方法本身:Bailey、Muth 与 Nourse 的回归是怎样从一对对成交价里算出指数的,它在什么条件下是无偏的。 (https://5dlibrary.com/answers/zh/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c01/) - Ch.2 The Bailey-Muth-Nourse Regression: from Pairs of Sales to a Price Index — When a work's own volatility is only 0.10 the two conventions differ by 0.0053; after ten years the median work is worth 1.84 times its price and the mean work 1.94 times, and the chance of a loss on a single work is only 0.027. At an own volatility of 0.30 the arithmetic rate is 0.1119, which is 0.0489 above the geometric one; after ten years the mean work is worth 2.89 times its price, yet the median work is still worth only 1.84 times, and the chance that a single work is below its purchase price is 0.260. At a volatility of 0.50 the arithmetic rate is 0.2045 while the geometric rate stays at 0.0630, and the chance of a loss on one work is 0.350. (https://5dlibrary.com/answers/en/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c02/) 第2章 Bailey-Muth-Nourse 回归:从成交配对到价格指数 — 当单件波动只有 0.10 时,两种口径只差 0.0053,十年后中位数作品的价值是买入价的 1.84 倍,平均作品是 1.94 倍,单件亏损的概率只有 0.027。单件波动升到 0.30,算术年率是 0.1119,比几何年率高出 0.0489;十年后平均作品是 2.89 倍,但中位数作品仍然只有 1.84 倍,而单件作品低于买入价的概率是 0.260。波动到 0.50 时,算术年率是 0.2045,几何年率仍是 0.0630,单件亏损的概率是 0.350。 (https://5dlibrary.com/answers/zh/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c02/) - Ch.3 Who Sells When: Resale Timing, Exit and Censoring — Chapter 3 separates resale timing, permanent exit and censoring in a twenty-year dynamic market: timing alone biases the index by about 0.0113 a year, permanent exit lifts this to 0.0263 or more, and correction needs a model of who sells. (https://5dlibrary.com/answers/en/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c03/) 第3章 谁在什么时候卖:转售时机、退出与截尾 — 第三章在二十年的动态市场中区分转售时机、永久退出与截尾:单纯挑时机的偏差每年约 0.0113,加入永久退出后升到 0.0263 以上;纠正需要一个关于谁会卖的模型。 (https://5dlibrary.com/answers/zh/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c03/) - Ch.4 Forgery Priors: Likelihood Ratios of Authentication Evidence and the Value of Information — Forgery is an invisible deduction from collectible returns. It does not appear in the index, because exposed fakes drop straight out of the data. Under Bayes' rule the same evidence weighs very differently under different priors, and in a low-prior category the posterior may still be only about three in five after four kinds of evidence have passed. A residual forgery probability translates into a yearly drag on returns, comparable at 0.20 to the selection correction. The value of a test is an inverted U, highest only where the decision is in the balance. The next chapter turns to a deduction that always happens: fees. (https://5dlibrary.com/answers/en/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c04/) 第4章 赝品先验:鉴定证据的似然比与信息价值 — 赝品是收藏品回报里看不见的一项扣减:它不出现在指数里,因为被揭穿的赝品直接从数据中消失。用贝叶斯定理看,同一条证据在不同先验下分量完全不同;在先验极低的品类里,即使四条证据全部通过,后验仍可能只有六成多。残余的赝品概率可以折算成每年的回报拖累,在 0.20 时与选择偏差的修正幅度相当。检测的价值呈倒 U 形,只在决定悬而未决时最高。下一章转向另一项确定发生的扣减:佣金。 (https://5dlibrary.com/answers/zh/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c04/) - Ch.5 The Round Trip: Break-Even Years and the Chance of Loss on a Single Piece — Chapter 5 shows that at the corrected return nominal break-even takes only 4.7 years, but with a holding cost of 0.01 a year catching a 0.05 benchmark takes 127.6 years; a single piece has a 0.389 chance of being below its total cost after ten years. (https://5dlibrary.com/answers/en/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c05/) 第5章 往返佣金:盈亏平衡年数与单件藏品的亏损概率 — 第五章说明按修正后的回报名义回本只需 4.7 年,但在每年 0.01 的持有成本下追上 0.05 的基准需要 127.6 年;单件作品十年后低于总支出的概率是 0.389。 (https://5dlibrary.com/answers/zh/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c05/) - Ch.6 The Illiquidity Discount: a Perfect-Timing Bound and Sales under a Deadline — Chapter 6 gives a perfect-timing bound of 0.2081 on the illiquidity discount and shows that a seller who must raise cash within three months loses 0.1715 more. (https://5dlibrary.com/answers/en/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c06/) 第6章 流动性折价:完美择时的上界与期限压力下的成交 — 第六章给出完美择时的流动性折扣上界 0.2081,并说明三个月内必须变现的卖家要多损失 0.1715。 (https://5dlibrary.com/answers/zh/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c06/) - Ch.7 Classics East and West: Mi Fu's History of Painting and Vasari's Lives — The papal portrait table: when copies pass with probability 0.10, stylistic judgement has a likelihood ratio of 9.50 and the posterior probability of the original after a pass is 0.905. When copy quality rises to 0.60, the ratio is only 1.58 and the posterior 0.613. At 0.90, which is Giulio Romano's situation, the ratio is 1.06 and the posterior 0.514, almost as if he had not looked at all. The insider's mark, by contrast, has a likelihood ratio of 95 and a posterior of 0.990. If, after stylistic judgement has passed the panel, the mark is found to be missing, the probability of the original falls from 0.514 to 0.0506. Vasari's pointing finger was worth far more than Giulio's two eyes. (https://5dlibrary.com/answers/en/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c07/) 第7章 东西方原典考辨:米芾《画史》与瓦萨里《名人传》 — 教皇像一表:复制品通过率为 0.10 时,风格判断的似然比是 9.50,通过后的原作概率是 0.905;复制水平升到 0.60,似然比只剩 1.58,后验 0.613;到了 0.90,也就是朱利奥·罗马诺遇到的情形,似然比 1.06,后验 0.514,几乎等于没有看。相反,知情人的记号似然比是 95,后验 0.990。如果风格判断通过之后,再发现画上没有那个记号,原作的概率就从 0.514 跌到 0.0506。瓦萨里那一指,比朱利奥的一双眼睛有用得多。 (https://5dlibrary.com/answers/zh/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c07/) - Ch.8 Closing the Loop: an Allocation Cap, a Due-Diligence Checklist and a Falsification Plan — Chapter 8 demonstrates an allocation cap on alternative assets for wealthy clients: the promotional view allocates 0.251, the full-cost view zero; it also gives a pre-purchase due-diligence checklist, the distilled tool Collectibles Return Debiaser and a pre-registered falsification plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c08/) 第8章 实证闭环:配置上限、尽调清单与证伪方案 — 第八章演示高净值客户另类资产配置上限:宣传口径的模型配置为 0.251,扣除各项成本后为零;并给出藏品入手前的尽调清单、蒸馏小工具「收藏回报去偏器」与一份预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/collectibles-repeat-sales-selection-bias-forgery-illiquidity/c08/) ## 5DL-B-209 · Diamond Magic-Size Price Jumps, the Lab-Grown Learning-Curve Slide and Supply-Cartel Signalling 中文书名:钻石克拉魔术尺寸价格跳跃、培育钻学习曲线崩价与供给卡特尔信号母典 DOI: pending Cite: Fan, Y. (2026). Diamond Magic-Size Price Jumps, the Lab-Grown Learning-Curve Slide and Supply-Cartel Signalling (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 3 uses a model of the cutter to show how a step leaves a hole and a pile in the weight distribution, with 0.650 of the 1.00 ct class pushed up in the illustrative model, and compares left-digit attention with focal-size conformity. Marked boxes separate published evidence, the author's hypotheses and classical sources, and the book contains 20 programs. Abstract: Two beliefs rule the jewellery counter: one carat is one carat and a fraction short will not do, and diamonds hold their value whether they come from a mine or a laboratory. This book turns both beliefs into quantities that can be computed and asks where published data place them. Apart from those with a stated source, its price, share and cost parameters are illustrative; all calculations are generated by its programs; it is not personalised legal, financial or business advice, and labelling and advertising rules are subject to the law in force as checked by a qualified person before listing. Chapter 1 shows that a 0.99 and a 1.00 ct stone differ in diameter by only 0.0214 mm, below what the eye resolves at close range, while the published one-carat premium, net of the rise owed to a smooth elasticity, equals 2.50 to 4.89 ordinary hundredths. Chapter 2 shows that a smooth fit ignoring the steps overstates the elasticity by 0.227 and that a blind scan for steps must raise its bar to 3.47. Chapter 3 uses a model of the cutter to show how a step leaves a hole and a pile in the weight distribution, with 0.650 of the 1.00 ct class pushed up in the illustrative model, and compares left-digit attention with focal-size conformity. Chapter 4 derives from published quotations a yearly change of −0.1734 in the average online lab-grown price, with its ratio to mined prices falling to 0.317, and discusses Wright's law and a cost floor. Chapter 5 separates replacement from realisable value: after five years a lab-grown stone's replacement value is about 0.386 of its price, and by user cost it is a consumer good. Chapter 6 uses the models of Cournot, Stigler and Green and Porter to explain the long stability of mined prices and how lab-grown stones weaken it from inside and outside. Chapter 7 examines Cao Zhao's Gegu yaolun and Tavernier's Six Voyages: the square rule overstates a 100 ct stone by a factor of 1.318 against the modern elasticity, and when half the market is lab-grown a stone that passes the old test is natural with probability only 0.494. Chapter 8 offers a jewellery quoting tool, twin-track copy rules and a pre-registered test plan whose shop trial needs 356 customers per arm. Marked boxes separate published evidence, the author's hypotheses and classical sources, and the book contains 20 programs. 摘要:珠宝柜台上有两条常识:一克拉就是一克拉,差一点也不行;钻石保值,不论来自矿山还是实验室。本书把这两条常识写成可以计算的量,再看已发表的数据把它们放在什么位置。书中的价格、比例与成本参数,除注明出处者外均为示意;全部计算由书中程序生成;本书不是个人化的法律、理财或经营建议,涉及的标注与广告规则以上架前人工复核的现行法条为准。 第一章说明 0.99 与 1.00 克拉的直径只差 0.0214 毫米,低于近距离肉眼的分辨力,而已发表研究报告的一克拉溢价在扣除平滑弹性应得的涨幅之后,相当于普通一分涨幅的 2.50 到 4.89 倍。第二章说明忽略台阶的平滑拟合会把弹性高估 0.227,盲扫台阶须把门槛提高到 3.47。第三章用切磨者模型说明台阶怎样在重量分布上留下缺口与堆积,示意模型中 1.00 克拉一档有 0.650 的石头是被推上关口的;并比较左位数注意力与焦点从众两种解释。第四章用公开报价算出培育钻在线均价每年变化 −0.1734,与开采钻的价格比降到 0.317,并讨论莱特定律与成本地板。 第五章区分重置价值与变现价值,培育钻五年后的重置价值约为买价的 0.386,按使用者成本它是一件消费品。第六章用古诺、施蒂格勒与格林、波特的模型解释天然钻价格的长期平稳,以及培育钻怎样从内外两个方向削弱它。第七章考辨曹昭《格古要论》与塔韦尼耶《六次旅行》:平方定价规则在 100 克拉处比现代弹性高估 1.318 倍;培育钻占一半时,古法检验通过的石头是天然钻的概率只剩 0.494。第八章给出钻饰报价器、双轨文案规则与预注册检验方案,门店对照试验每组需要 356 位顾客。书中以标注框区分已发表的证据、作者的假说与原典,共含 20 段程序。 Chapter answers: - Ch.1 Magic Sizes: an Invisible Hundredth of a Carat and a Price Step — Chapter 1 shows that a 0.99 and a 1.00 ct stone differ in diameter by only 0.0214 mm, below what the eye resolves at close range, while the published one-carat premium, net of the rise owed to a smooth elasticity, equals 2.50 to 4.89 ordinary hundredths. (https://5dlibrary.com/answers/en/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c01/) 第1章 魔术尺寸:肉眼分不出的一分与价格台阶 — 第一章说明 0.99 与 1.00 克拉的直径只差 0.0214 毫米,低于近距离肉眼的分辨力,而已发表研究报告的一克拉溢价在扣除平滑弹性应得的涨幅之后,相当于普通一分涨幅的 2.50 到 4.89 倍。 (https://5dlibrary.com/answers/zh/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c01/) - Ch.2 Piecewise Power Laws and Breakpoint Scans: Finding the Steps in a Price List — Chapter 2 shows that a smooth fit ignoring the steps overstates the elasticity by 0.227 and that a blind scan for steps must raise its bar to 3.47. (https://5dlibrary.com/answers/en/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c02/) 第2章 分段幂律与断点扫描:价格表里的台阶怎样被找出来 — 第二章说明忽略台阶的平滑拟合会把弹性高估 0.227,盲扫台阶须把门槛提高到 3.47。 (https://5dlibrary.com/answers/zh/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c02/) - Ch.3 The Cutter's Arithmetic: Bunching, Holes and Left-Digit Attention — Chapter 3 uses a model of the cutter to show how a step leaves a hole and a pile in the weight distribution, with 0.650 of the 1.00 ct class pushed up in the illustrative model, and compares left-digit attention with focal-size conformity. (https://5dlibrary.com/answers/en/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c03/) 第3章 切磨者的算盘:聚集、缺口与左位数注意力 — 第三章用切磨者模型说明台阶怎样在重量分布上留下缺口与堆积,示意模型中 1.00 克拉一档有 0.650 的石头是被推上关口的;并比较左位数注意力与焦点从众两种解释。 (https://5dlibrary.com/answers/zh/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c03/) - Ch.4 The Lab-Grown Learning Curve: How Fast Prices Fall and Where the Floor Is — Chapter 4 derives from published quotations a yearly change of −0.1734 in the average online lab-grown price, with its ratio to mined prices falling to 0.317, and discusses Wright's law and a cost floor. (https://5dlibrary.com/answers/en/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c04/) 第4章 培育钻的莱特曲线:价格每年掉多少,地板在哪里 — 第四章用公开报价算出培育钻在线均价每年变化 −0.1734,与开采钻的价格比降到 0.317,并讨论莱特定律与成本地板。 (https://5dlibrary.com/answers/zh/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c04/) - Ch.5 The Value Ledger: Consumer Good or Asset? — Chapter 5 separates replacement from realisable value: after five years a lab-grown stone's replacement value is about 0.386 of its price, and by user cost it is a consumer good. (https://5dlibrary.com/answers/en/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c05/) 第5章 保值账:消费品还是资产 — 第五章区分重置价值与变现价值,培育钻五年后的重置价值约为买价的 0.386,按使用者成本它是一件消费品。 (https://5dlibrary.com/answers/zh/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c05/) - Ch.6 Cartels, Cournot and Noisy Price Signals — The Cournot model shows why a cartel is worth organising: with two competing firms joint profit is only 0.889 of the monopoly level, with five only 0.556. Stigler and Green and Porter show why a cartel is hard to sustain: members cannot see one another's output and must infer cheating from noisy prices, and once the noise passes a critical level, about 7.5 to 8.5 under this chapter's illustrative parameters, collusion can no longer enforce itself. Lab-grown stones weaken this structure from two directions, pushing down the natural price ceiling from outside and making price signals noisier from inside. The next chapter goes back several centuries to see how two earlier writers described the pricing of gems, and which of their statements can be tested against today's data. (https://5dlibrary.com/answers/en/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c06/) 第6章 卡特尔、古诺与带噪声的价格信号 — 古诺模型说明了卡特尔为什么值得组织:两家竞争时联合利润只剩垄断的 0.889,五家时只剩 0.556。施蒂格勒与格林、波特说明了卡特尔为什么难以维持:成员看不到彼此的产量,只能从嘈杂的价格里推断背叛;噪声超过一个临界值,在本章的示意参数下约为 7.5 到 8.5,合谋就无法自我维持。培育钻从两个方向削弱了这种结构:从外部压低天然钻的价格上限,从内部让价格信号更嘈杂。下一章回到几百年前,看两位古人怎样描述宝石的定价,以及他们的哪些说法可以用今天的数据来检验。 (https://5dlibrary.com/answers/zh/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c06/) - Ch.7 Sources East and West: Cao Zhao's Gegu yaolun and Tavernier's Six Voyages — Chapter 7 examines Cao Zhao's Gegu yaolun and Tavernier's Six Voyages: the square rule overstates a 100 ct stone by a factor of 1.318 against the modern elasticity, and when half the market is lab-grown a stone that passes the old test is natural with probability only 0.494. (https://5dlibrary.com/answers/en/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c07/) 第7章 东西方原典考辨:曹昭《格古要论》与塔韦尼耶《六次旅行》 — 第七章考辨曹昭《格古要论》与塔韦尼耶《六次旅行》:平方定价规则在 100 克拉处比现代弹性高估 1.318 倍;培育钻占一半时,古法检验通过的石头是天然钻的概率只剩 0.494。 (https://5dlibrary.com/answers/zh/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c07/) - Ch.8 Closing the Loop: a Jewellery Quoting Tool, Twin-Track Copy and a Pre-Registered Test — Chapter 8 offers a jewellery quoting tool, twin-track copy rules and a pre-registered test plan whose shop trial needs 356 customers per arm. Marked boxes separate published evidence, the author's hypotheses and classical sources, and the book contains 20 programs. (https://5dlibrary.com/answers/en/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c08/) 第8章 实证闭环:钻饰报价器、双轨文案与预注册检验 — 第八章给出钻饰报价器、双轨文案规则与预注册检验方案,门店对照试验每组需要 356 位顾客。书中以标注框区分已发表的证据、作者的假说与原典,共含 20 段程序。 (https://5dlibrary.com/answers/zh/diamond-magic-size-jumps-lab-grown-learning-curve-cartel/c08/) ## 5DL-B-210 · Blind-Box Coupon Collecting, Geometric Waits for Hidden Variants and Odds Disclosure for Lucky Bags 中文书名:盲盒收集券问题、隐藏款几何等待时间与福袋概率公示合规母典 DOI: pending Cite: Fan, Y. (2026). Blind-Box Coupon Collecting, Geometric Waits for Hidden Variants and Odds Disclosure for Lucky Bags (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is general information, not individualised legal or financial advice, and every statute must be checked by a person before listing. Chapter 5 computes a lucky bag's mean, spread and chance of being worth less than its price (0.8280 for the illustrative bag) and the weighting of small chances. Abstract: This book recalculates two pieces of counter-side common sense: "a set of 12, buy a few more and you will have them all" and "the hidden variant is 1/144, so 144 boxes should give one". If every design is equally likely and every box independent, a set of 12 takes 37.24 boxes on average, with a median of 35 and a ninetieth percentile of 55; a 1/144 hidden variant appears at least once in 144 boxes with probability only 0.6334, and a nine-in-ten chance needs 331 boxes. Its parameters are illustrative; all calculations are generated by its programs; it is general information, not individualised legal or financial advice, and every statute must be checked by a person before listing. Chapter 1 completes equally likely sets with harmonic numbers, the exact distribution and seeded simulation. Chapter 2 treats the hidden variant as a geometric wait, sets out the contrary evidence on judgements of compound events, and shows that a guarantee changes the wait (91.21 draws on average with a guarantee at 144) but not the stated rate. Chapter 3 handles unequal odds by inclusion-exclusion and a Poissonisation integral: 12 regulars plus a 1/144 hidden variant take 149.07 boxes. Chapter 4 uses the double Dixie cup problem for pooling, duplicates and when to stop opening. Chapter 5 computes a lucky bag's mean, spread and chance of being worth less than its price (0.8280 for the illustrative bag) and the weighting of small chances. Chapter 6 takes the seller's side, comparing case packing, set-completion prizes and two allocation rules. Chapter 7 examines the guanpu stake games in Meng Yuanlao's Dongjing Meng Hua Lu and de Moivre's De Mensura Sortis. Chapter 8 gives an odds-disclosure generator for lucky-bag pages, the distilled tool Lucky-Bag Odds Discloser, a checklist against the rules of Taiwan and Malaysia, a pre-registered audit (2808 boxes to test 1/144 against 1/288) and the conditions that would overturn the central claim. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. 摘要:本书重算盲盒柜台前两句常识:「一套 12 款,多买几盒就齐」与「隐藏款 1/144,买满 144 盒总该出一个」。在每款等概率、每盒独立的前提下,集齐 12 款的期望是 37.24 盒,中位数 35,九成分位 55;1/144 的隐藏款买满 144 盒,至少一个的概率只有 0.6334,九成把握需要 331 盒。书中参数为示意;全部计算由书中程序生成;本书是非个人化法律或理财建议,法条以上架前人工复核为准。 第一章用调和数、精确分布与固定种子模拟算等概率集齐;第二章用几何分布算隐藏款等待,并陈复合事件判断偏差的相反证据,说明保底只改变等待(保底 144 次时平均 91.21 次)而不改变标称概率;第三章用容斥公式与泊松化积分算不等概率,12 款加 1/144 隐藏款的全套期望是 149.07 盒;第四章用双杯问题算合买与重复款,以及开盒止损;第五章算福袋的期望、方差与低于售价的概率(示意福袋为 0.8280)以及小概率加权。 第六章站在卖家一侧,比较整箱装配、集齐有奖与两种投放规则;第七章考辨孟元老《东京梦华录》所记关扑与 de Moivre《De Mensura Sortis》;第八章给出福袋页概率公示生成器、蒸馏小工具「福袋概率公示器」、台湾与马来西亚规范核对表、预注册抽检方案(标称 1/144 对 1/288 需 2808 盒)与推翻本书核心命题的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 How Many Boxes for Twelve Designs: Harmonic Numbers, Variance and the Full Distribution — For 12 equally likely designs, the expected number of boxes for a full set is 37.2385, with a standard deviation of 13.72, a median of 35 and a ninetieth percentile of 55. Of the expectation, 0.790 is spent on the last six designs. One hundred thousand seeded simulations agree with the formulas. These are mathematical facts; whether buyers generally underestimate them remains a hypothesis awaiting a survey. The next chapter asks a sharper question: with a hidden variant stated at 1/144, how sure can a buyer of 144 boxes be? (https://5dlibrary.com/answers/en/blind-box-coupon-collector-hidden-variant-odds-disclosure/c01/) 第1章 集齐十二款要多少盒:调和数、方差与完整分布 — 等概率的 12 款,集齐的期望盒数是 37.2385,标准差 13.72,中位数 35,九成分位 55。全部期望中有 0.790 花在后 6 款上。十万次固定种子的模拟与公式相符。这些都是数学事实;买家是否普遍低估它们,则是一条等待调查的假说。下一章的问题更尖锐:一个标称 1/144 的隐藏款,买满 144 盒,到底有多大把握? (https://5dlibrary.com/answers/zh/blind-box-coupon-collector-hidden-variant-odds-disclosure/c01/) - Ch.2 The Hidden Variant at 1/144: Geometric Waits, Quantiles and Guarantees — Chapter 2 treats the hidden variant as a geometric wait, sets out the contrary evidence on judgements of compound events, and shows that a guarantee changes the wait (91.21 draws on average with a guarantee at 144) but not the stated rate. (https://5dlibrary.com/answers/en/blind-box-coupon-collector-hidden-variant-odds-disclosure/c02/) 第2章 隐藏款 1/144:几何等待、分位数与保底机制 — 第二章用几何分布算隐藏款等待,并陈复合事件判断偏差的相反证据,说明保底只改变等待(保底 144 次时平均 91.21 次)而不改变标称概率; (https://5dlibrary.com/answers/zh/blind-box-coupon-collector-hidden-variant-odds-disclosure/c02/) - Ch.3 Collecting under Unequal Odds: Inclusion–Exclusion, a Poissonisation Integral and Tilted Designs — Chapter 3 handles unequal odds by inclusion-exclusion and a Poissonisation integral: 12 regulars plus a 1/144 hidden variant take 149.07 boxes. (https://5dlibrary.com/answers/en/blind-box-coupon-collector-hidden-variant-odds-disclosure/c03/) 第3章 不等概率的收集:容斥公式、泊松化积分与款式倾斜 — 第三章用容斥公式与泊松化积分算不等概率,12 款加 1/144 隐藏款的全套期望是 149.07 盒; (https://5dlibrary.com/answers/zh/blind-box-coupon-collector-hidden-variant-odds-disclosure/c03/) - Ch.4 Duplicates, Pooling and Buying the Gaps: the Double Dixie Cup and When to Stop Opening — A buyer who completes 12 designs holds 25.24 duplicates on average, and after box 24 each further box has less than a one-in-eight chance of being new. The double Dixie cup formula of Newman and Shepp shows that pooling cuts each member's share from 37.24 boxes to 29.02 for two people and 18.91 for ten, but never down to the number of designs. With single designs at 139 on the resale market and no sale of duplicates, the best plan is to stop opening when about 5 designs are missing, at an expected 1373.9, a little over half the 2569.5 of opening to the end. That closes the mathematics of blind boxes. Another product that depends just as much on chance, but is sold by value rather than by design, is the lucky bag. How should its expected value and spread be computed, and how should they be disclosed? Chapter 5 answers. (https://5dlibrary.com/answers/en/blind-box-coupon-collector-hidden-variant-odds-disclosure/c04/) 第4章 重复款、合买与补款:双杯问题与开盒止损 — 集齐 12 款时平均会多出 25.24 个重复款,拆到第 24 盒以后,每盒是新款的机会已不到八分之一。Newman 与 Shepp 的双杯公式说明,合买可以把每人分摊的盒数从 37.24 降到两人时的 29.02、十人时的 18.91,但永远降不到款数本身。若二手单款价为 139 且重复款无法回收,最优策略是拆到只缺 5 款左右就停手,期望花费 1373.9,约为一直拆到底的 2569.5 的一半多一点。到这里,盲盒的数学告一段落。另一种同样依赖概率、却按「价值」而不是「款式」出售的商品是福袋:它的期望值与方差该怎样算,又该怎样公示?第五章回答。 (https://5dlibrary.com/answers/zh/blind-box-coupon-collector-hidden-variant-odds-disclosure/c04/) - Ch.5 Expected Value and Spread of a Lucky Bag: Tiers, Convolution and the Weighting of Small Chances — Chapter 5 computes a lucky bag's mean, spread and chance of being worth less than its price (0.8280 for the illustrative bag) and the weighting of small chances. (https://5dlibrary.com/answers/en/blind-box-coupon-collector-hidden-variant-odds-disclosure/c05/) 第5章 福袋的期望价值与方差:分档、卷积与小概率加权 — 第五章算福袋的期望、方差与低于售价的概率(示意福袋为 0.8280)以及小概率加权。 (https://5dlibrary.com/answers/zh/blind-box-coupon-collector-hidden-variant-odds-disclosure/c05/) - Ch.6 The Seller's Design Choices: Case Packing, Set-Completion Prizes and Two Allocation Rules — Packing cases without repeats raises the chance of completing the regulars from 4.941e-05 to 0.9167, a ratio of 18551, while the "1/144" on the package is unchanged. Set-completion prizes turn the collecting multiplier into a cost multiplier: six rare items at 1 % take 245.0 draws, 2.45 times a single one, and Japan held such schemes to be prohibited in 2012. Under the same stated 1/144, the number of hidden variants in a run of 7,200 independently filled boxes has a standard deviation of 7.05, and only fixed allocation can guarantee "50 in this run". The seller's design choices are now covered. Where did these rules and odds first appear? What testable claims did the market streets of the East and the studies of the West leave behind? Chapter 7 returns to the classical sources. (https://5dlibrary.com/answers/en/blind-box-coupon-collector-hidden-variant-odds-disclosure/c06/) 第6章 卖家的设计变量:整箱装配、集齐有奖与两种投放规则 — 整箱不重复的装箱使集齐常规款的概率从 4.941e-05 升到 0.9167,相差 18551 倍,而包装上的「1/144」不变。「集齐有奖」把收集乘数变成成本乘数:集齐 6 个 1 % 道具要 245.0 次抽取,是单个的 2.45 倍,日本已于 2012 年将此类机制认定为被禁止的方式。同样标称 1/144,独立抽取的一批 7,200 盒中隐藏款数量的标准差是 7.05,只有固定投放才能保证「本批 50 个」。卖家的设计变量讲完了。这些规则和概率最早从哪里来?东方的市井与西方的书斋,各自留下了什么可以检验的说法?第七章回到原典。 (https://5dlibrary.com/answers/zh/blind-box-coupon-collector-hidden-variant-odds-disclosure/c06/) - Ch.7 Classical Sources East and West: Guanpu in the Dongjing Meng Hua Lu and de Moivre's De Mensura Sortis — The Dongjing Meng Hua Lu records that Kaifeng "released guanpu for three days" at festivals, that coins were thrown into earthenware basins, that some staked "one hu for thirty", and that prices were agreed beforehand, but it records no rule for winning; if the rule was five coins on a named face, odds of one to thirty left the stall 0.0312 of each stake. De Mensura Sortis wrote games of chance as numbered problems of calculation, and later literature traces an early form of the collector's problem to it, though this book could not verify the specific problem number. A six-sided die needs 13 throws before all faces are more likely than not, and a twelve-sided die thrown 24 times shows every face with the same probability as 24 boxes completing 12 designs, 0.1502. The institutions and the calculations of the past are in place. What remains is the question of how a shop today can turn them into a disclosure page that can be listed, checked and refuted. Chapter 8 closes the loop. (https://5dlibrary.com/answers/en/blind-box-coupon-collector-hidden-variant-odds-disclosure/c07/) 第7章 东西方原典考辨:《东京梦华录》的关扑与 de Moivre 的《De Mensura Sortis》 — 《东京梦华录》记录了北宋开封在节日「放关扑三日」、以瓦盆掷头钱、「以一笏扑三十笏」以及「约以价而扑之」,却没有记录输赢规则;若规则是五枚钱全为指定一面,一比三十的赔率让摊主每份抽取 0.0312。《De Mensura Sortis》把机会游戏写成编号的计算问题,后世文献把收集问题的早期形式追溯到它,具体编号本书未能核实。六面骰掷 13 次才有过半的机会六面齐现,12 面骰掷 24 次全部出现的概率与 24 盒集齐 12 款相同,都是 0.1502。古人的制度与计算都已就位,剩下的问题是:一家今天的门店,怎样把它们变成一页可以上架、可以核对、可以被推翻的公示?第八章给出闭环。 (https://5dlibrary.com/answers/zh/blind-box-coupon-collector-hidden-variant-odds-disclosure/c07/) - Ch.8 Closing the Loop: an Odds-Disclosure Generator for Lucky-Bag Pages, a Taiwan and Malaysia Checklist and a Pre-Registered Audit — Chapter 8 gives an odds-disclosure generator for lucky-bag pages, the distilled tool Lucky-Bag Odds Discloser, a checklist against the rules of Taiwan and Malaysia, a pre-registered audit (2808 boxes to test 1/144 against 1/288) and the conditions that would overturn the central claim. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/blind-box-coupon-collector-hidden-variant-odds-disclosure/c08/) 第8章 实证闭环:福袋页概率公示生成器、台马规范核对与预注册抽检 — 第八章给出福袋页概率公示生成器、蒸馏小工具「福袋概率公示器」、台湾与马来西亚规范核对表、预注册抽检方案(标称 1/144 对 1/288 需 2808 盒)与推翻本书核心命题的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/blind-box-coupon-collector-hidden-variant-odds-disclosure/c08/) ## 5DL-B-211 · Newsvendor Critical Fractiles for Pre-Season Buying, Dirichlet Size-Curve Priors and Dynamic-Programming End-of-Season Markdowns 中文书名:买手季前订货报童临界分位、尺码曲线狄利克雷先验与季末降价动态规划母典 DOI: pending Cite: Fan, Y. (2026). Newsvendor Critical Fractiles for Pre-Season Buying, Dirichlet Size-Curve Priors and Dynamic-Programming End-of-Season Markdowns (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its goods, stores and demand parameters are illustrative; all calculations are generated by its programs; the book is not individual business advice. Chapter 1 derives the critical fractile from marginal arithmetic; under illustrative parameters the optimal order equals the mean only at a margin of exactly 0.3750, and ordering the mean for an eight-tenths-margin item loses 6.96 % of expected profit. Abstract: The buyer's most common rule of thumb is to order the expected average. This book refutes it. The order that maximises expected profit is the critical fractile of the newsvendor model, set by the ratio of the cost of one unit short to the cost of one unit left over; ordering the average systematically under-buys high-margin goods and over-buys low-margin goods. Its goods, stores and demand parameters are illustrative; all calculations are generated by its programs; the book is not individual business advice. Chapter 1 derives the critical fractile from marginal arithmetic; under illustrative parameters the optimal order equals the mean only at a margin of exactly 0.3750, and ordering the mean for an eight-tenths-margin item loses 6.96 % of expected profit. Reconstructed from published data of the Schweitzer and Cachon experiment, participants pulled their orders 0.644 (high profit) and 0.787 (low profit) of the way toward the mean. Chapter 2 shows how skewed demand, salvage value and lost goodwill move the fractile, and that the mean order looks better after the fact in 55.90 % of seasons, so that 51.7 seasons are needed to see the difference in profit. Chapter 3 uses a Dirichlet-multinomial prior to weigh the chain's size curve against a store's own sample. Chapter 4 shows that per-size orders for a high-margin style add up to 17.9 units more than the style order. Chapter 5 solves for an end-of-season markdown table by finite-horizon dynamic programming; a weeks-of-cover rule earns at least 0.9878 of optimal revenue, while calendar markdowns do worse than holding the price when stock is thin. Chapter 6 folds the markdown option and in-season reorders back into the pre-season buy; with a correlation of 0.9 between early and season sales, reorders raise profit by 4.47 %. Chapter 7 examines Bai Gui's "what others cast off I take up" in the Records of the Grand Historian and Cantillon's account of the undertaker's risk in his Essai; the model shows that when free entry stops, the marginal shop loses money in a season with probability 0.4385. Chapter 8 gives an order quantity calculator, a post-season bias review and a pre-registered test; if the pull is as strong as in the laboratory, 33 style-seasons suffice to detect it. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:买手最常用的经验是按预计平均销量订货。本书证伪这条经验:能让期望利润最大的订量是报童模型的临界分位,由缺货一件与滞销一件的成本之比决定;按平均数订货,高毛利商品系统性订少,低毛利商品系统性订多。书中的商品、门店与需求参数均为示意;全部计算由书中程序生成;本书是非个人化经营建议。 第一章用边际算术推出临界分位,示意参数下只有毛利率恰为 0.3750 时最优订量才等于平均数,毛利八成的商品按平均数订会少赚 6.96 %;据施韦策与卡雄实验的已发表数据重建,受试者把订量向平均数拉了 0.644(高利润)与 0.787(低利润)的距离。第二章说明偏态需求、残值与商誉损失如何移动临界分位,并显示按平均数订货在 55.90 % 的季节里事后看反而更赚,要从利润中看出差别需要 51.7 个季度。第三章用狄利克雷—多项式先验在总部尺码曲线与门店样本之间加权。第四章说明高毛利款分尺码订量之和比整款订量多出 17.9 件。 第五章用有限期动态规划求季末降价表,库存周数规则可得最优收入的 0.9878 以上,而按日历降价在库存偏少时不如不降价。第六章把降价期权与季中追单折回季前订量,早期销售与全季相关系数为 0.9 时,追单使利润提高 4.47 %。第七章考辨《史记·货殖列传》中白圭「人弃我取,人取我与」与坎蒂隆《商业性质概论》对经营者风险的论述,模型显示自由进入停止时边际店铺每季亏损的概率为 0.4385。第八章给出订货量计算器、季后偏差复盘与预注册检验方案,若拉动强度与实验室相当,33 个款式季即可检出偏差。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Critical Fractile: Cost Ratios, Not Averages, Set the Buy — Chapter 1 derives the critical fractile from marginal arithmetic; under illustrative parameters the optimal order equals the mean only at a margin of exactly 0.3750, and ordering the mean for an eight-tenths-margin item loses 6.96 % of expected profit. Reconstructed from published data of the Schweitzer and Cachon experiment, participants pulled their orders 0.644 (high profit) and 0.787 (low profit) of the way toward the mean. (https://5dlibrary.com/answers/en/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c01/) 第1章 临界分位:订量由成本比决定,而不是由平均数决定 — 第一章用边际算术推出临界分位,示意参数下只有毛利率恰为 0.3750 时最优订量才等于平均数,毛利八成的商品按平均数订会少赚 6.96 %;据施韦策与卡雄实验的已发表数据重建,受试者把订量向平均数拉了 0.644(高利润)与 0.787(低利润)的距离。 (https://5dlibrary.com/answers/zh/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c01/) - Ch.2 Skewed Demand, Salvage and Goodwill: Three Forces on the Fractile, and Why Experience Fails to Correct the Bias — Chapter 2 shows how skewed demand, salvage value and lost goodwill move the fractile, and that the mean order looks better after the fact in 55.90 % of seasons, so that 51.7 seasons are needed to see the difference in profit. (https://5dlibrary.com/answers/en/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c02/) 第2章 偏态需求、残值与商誉:临界分位的三个推手,以及经验为何纠不正偏差 — 第二章说明偏态需求、残值与商誉损失如何移动临界分位,并显示按平均数订货在 55.90 % 的季节里事后看反而更赚,要从利润中看出差别需要 51.7 个季度。 (https://5dlibrary.com/answers/zh/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c02/) - Ch.3 Dirichlet-Multinomial Priors for the Size Curve: Weighing Head-Office Experience Against the Store's Own Sample — The result is a sobering reminder. With 12 stores and 720 units in all, the estimate is 139.0, the band starts at 49.6 and has no upper end: the data can rule out a weak prior but cannot say how strong the prior is. The log-likelihood differences tell the same story. A strength of 10 is 16.922 log units worse than 40 and can be excluded, while a strength of 160 is actually 3.036 better than 40 and cannot be told apart. With 48 stores the estimate is 56.1 and the band runs from 39.0 to 85.9; only with 192 stores does the band, from 37.1 to 52.1 around an estimate of 43.9, pin the true value down firmly. (https://5dlibrary.com/answers/en/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c03/) 第3章 尺码曲线的狄利克雷—多项式先验:在总部经验与门店样本之间加权 — 结果给经营者一个清醒的提醒。只有 12 家门店、共 720 件销售时,估计值是 139.0,区间下端 49.6,上端不封闭:数据只能排除「先验很弱」,却说不出先验有多强。对数似然的差也显示这一点:先验强度 10 比 40 差了 16.922 个对数单位,可以排除;先验强度 160 反而比 40 好 3.036,无法区分。门店增加到 48 家,估计值 56.1,区间 39.0 到 85.9;到 192 家,估计值 43.9,区间 37.1 到 52.1,才把真实值稳稳框住。 (https://5dlibrary.com/answers/zh/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c03/) - Ch.4 From the Total to the Sizes: Per-Size Newsvendors, the Fragmentation Premium and Broken Assortments — Chapter 4 shows that per-size orders for a high-margin style add up to 17.9 units more than the style order. (https://5dlibrary.com/answers/en/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c04/) 第4章 从总量到尺码:分尺码报童、碎片溢价与断码 — 第四章说明高毛利款分尺码订量之和比整款订量多出 17.9 件。 (https://5dlibrary.com/answers/zh/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c04/) - Ch.5 A Finite-Horizon Dynamic Programme for End-of-Season Markdowns: When to Cut and by How Much — Chapter 5 solves for an end-of-season markdown table by finite-horizon dynamic programming; a weeks-of-cover rule earns at least 0.9878 of optimal revenue, while calendar markdowns do worse than holding the price when stock is thin. (https://5dlibrary.com/answers/en/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c05/) 第5章 季末降价的有限期动态规划:什么时候降、降多少 — 第五章用有限期动态规划求季末降价表,库存周数规则可得最优收入的 0.9878 以上,而按日历降价在库存偏少时不如不降价。 (https://5dlibrary.com/answers/zh/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c05/) - Ch.6 Folding Markdowns and Reorders Back into the Pre-Season Buy: Effective Salvage and Quick Response — Chapter 6 folds the markdown option and in-season reorders back into the pre-season buy; with a correlation of 0.9 between early and season sales, reorders raise profit by 4.47 %. (https://5dlibrary.com/answers/en/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c06/) 第6章 把降价与追单折回季前订量:有效残值与快速反应 — 第六章把降价期权与季中追单折回季前订量,早期销售与全季相关系数为 0.9 时,追单使利润提高 4.47 %。 (https://5dlibrary.com/answers/zh/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c06/) - Ch.7 Classical Sources Examined: Bai Gui's "Take What Others Discard" and Cantillon's "Buy at a Certain Price, Sell at an Uncertain One" — Chapter 7 examines Bai Gui's "what others cast off I take up" in the Records of the Grand Historian and Cantillon's account of the undertaker's risk in his Essai; the model shows that when free entry stops, the marginal shop loses money in a season with probability 0.4385. (https://5dlibrary.com/answers/en/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c07/) 第7章 东西方原典考辨:白圭的「人弃我取」与坎蒂隆的「以定价买、以不定价卖」 — 第七章考辨《史记·货殖列传》中白圭「人弃我取,人取我与」与坎蒂隆《商业性质概论》对经营者风险的论述,模型显示自由进入停止时边际店铺每季亏损的概率为 0.4385。 (https://5dlibrary.com/answers/zh/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c07/) - Ch.8 Closing the Loop: the Order Quantity Calculator, the Post-Season Bias Review and a Pre-Registered Test — Chapter 8 gives an order quantity calculator, a post-season bias review and a pre-registered test; if the pull is as strong as in the laboratory, 33 style-seasons suffice to detect it. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c08/) 第8章 实证闭环:订货量计算器、季后偏差复盘与预注册检验方案 — 第八章给出订货量计算器、季后偏差复盘与预注册检验方案,若拉动强度与实验室相当,33 个款式季即可检出偏差。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/newsvendor-preseason-buying-dirichlet-size-curve-markdown-dp/c08/) ## 5DL-B-212 · J-Shaped Self-Selection in Star Ratings, Bayesian-Average Ranking and Rating Inflation from the Nine-Rank System to Online Marketplaces 中文书名:评论星级J型自选择偏差、贝叶斯平均排序与九品中正评级通胀母典 DOI: pending Cite: Fan, Y. (2026). J-Shaped Self-Selection in Star Ratings, Bayesian-Average Ranking and Rating Inflation from the Nine-Rank System to Online Marketplaces (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 1 explains the J-shaped star distribution with two filters, a purchasing bias and an under-reporting bias; with illustrative parameters the purchasing bias lifts the mean by +0.532 stars. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: Almost every online shop owner believes that higher star ratings sell better and that a star rating is what customers think. This book writes both beliefs as runnable arithmetic and asks when they hold. Its customers, products and parameters are illustrative models; all calculations are generated by its programs; it is general business reference, not personalised business advice for any shop, and not legal advice, and any regulation it mentions must be checked by a professional before going live. It opposes fake orders, bought reviews and gated review requests. Chapter 1 explains the J-shaped star distribution with two filters, a purchasing bias and an under-reporting bias; with illustrative parameters the purchasing bias lifts the mean by +0.532 stars. Chapter 2 returns to the 2017 report of the Spiegel Research Center: the report states the peak in three ways, whose intersection with its source article is 4.2 to 4.5 stars; it prints no curve values or confidence intervals and does not test the proposed psychological mechanism; and an illustrative catalogue shows that neither small-sample noise nor moderate boosting explains the whole inverted U. Chapter 3 sorts by the Wilson lower bound: 0.4385 for three five-star reviews against 0.8859 for two hundred at 4.6, while the textbook Wald interval can cover the truth as rarely as 0.049 at small samples and high shares. Chapter 4 uses the shop's star mix as a Dirichlet prior, giving Bayesian averages of 4.415 and 4.583, and discusses how to estimate the prior strength. Chapter 5 shows that sorting is also a policy for collecting reviews, comparing the exploration costs of cautious sorting and Thompson sampling and the value of the first five reviews. Chapter 6 models rating inflation as norm chasing: with honest feelings unchanged, the mean rises from 3.485 to 4.090 in nine years and more reviews are needed to tell sellers apart; it also shows how the friction of asking decides who speaks. Chapter 7 examines Liu Yi's critique of the Nine-Rank system in the Book of Jin, "no humble families above, no powerful clans below" (with family weighted about three times talent, humble men hold 0.072 of the upper grades in the model), and Galton's "Vox Populi" of 1907 (the crowd mean rebuilt as a two-piece normal is 1197.5 lb). Chapter 8 gives a shop procedure: one invitation to every buyer, the red line that inviting only satisfied buyers inflates the mean by +0.583, sorting by the posterior lower bound, the distilled Review Sorter, and pre-registered tests needing 53,211 visits per arm, with the conditions that would overturn the book. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:几乎每一位网店店主都相信星级越高越好卖,也相信星级就是顾客的评价。本书把这两条常识写成可以运行的算术,再看它们在什么条件下成立。书中的顾客、商品与参数均为示意模型;全部计算由书中程序生成;本书是非个人化的经营参考,不是针对具体店铺的经营建议,也不是法律意见,书中提到的法规以上架前人工复核为准;本书反对一切刷单、购买评论与筛选式求评。 第一章用「购买偏差加少报偏差」的两段筛子解释评论星级的 J 型分布,示意参数下购买偏差使均值上移 +0.532 星。第二章回到 Spiegel 研究中心 2017 年报告的原文:同一份报告对峰值区间有三种说法,与其所依据论文的交集是 4.2 至 4.5 星;报告没有印出曲线数值与置信区间,下降的心理机制未经检验;示意目录表明,小样本噪声与适度刷单都解释不了整条倒 U 曲线。第三章用威尔逊下界给出排序:3 条五星为 0.4385,200 条 4.6 星为 0.8859;教科书 Wald 区间在小样本高好评率处覆盖率可低到 0.049。第四章用全店星级构成作狄利克雷先验,两者的贝叶斯平均分别为 4.415 与 4.583,并讨论先验强度的估计。 第五章说明排序同时是求评策略,比较保守排序与汤普森抽样的探索代价,以及前五条评论的价值。第六章用规范追逐模型说明评分通胀:诚实感受不变,均分九年间从 3.485 升到 4.090,分辨两个卖家所需的评论随之增加;并讨论求评摩擦如何决定谁开口。第七章考辨《晋书·刘毅传》对九品中正制「上品无寒门,下品无势族」的批评(示意模型中门第权重约为才能的三倍时,上品寒门比例为 0.072)与高尔顿 1907 年的《民意》(两段正态重建的众人平均为 1197.5 磅)。第八章给出店铺流程:向全部买家发同一封邀请、只邀满意买家会使均分虚高 +0.583 的红线、后验下界排序、蒸馏小工具「评论排序器」,以及每组 53,211 次访问的预先登记检验和推翻本书的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Where the J Comes From: a U-Shaped Urge to Post and Two Self-Selection Biases — Chapter 1 explains the J-shaped star distribution with two filters, a purchasing bias and an under-reporting bias; with illustrative parameters the purchasing bias lifts the mean by +0.532 stars. (https://5dlibrary.com/answers/en/star-rating-j-shape-bayesian-average-nine-rank-inflation/c01/) 第1章 J 型从哪里来:U 形发帖意愿与两种自选择偏差 — 第一章用「购买偏差加少报偏差」的两段筛子解释评论星级的 J 型分布,示意参数下购买偏差使均值上移 +0.532 星。 (https://5dlibrary.com/answers/zh/star-rating-j-shape-bayesian-average-nine-rank-inflation/c01/) - Ch.2 Five Stars Do Not Sell Best: the Spiegel Curve at Source and a Small-Sample Reading — Chapter 2 returns to the 2017 report of the Spiegel Research Center: the report states the peak in three ways, whose intersection with its source article is 4.2 to 4.5 stars; it prints no curve values or confidence intervals and does not test the proposed psychological mechanism; and an illustrative catalogue shows that neither small-sample noise nor moderate boosting explains the whole inverted U. (https://5dlibrary.com/answers/en/star-rating-j-shape-bayesian-average-nine-rank-inflation/c02/) 第2章 满分不是最好卖:Spiegel 曲线的原始口径与小样本解读 — 第二章回到 Spiegel 研究中心 2017 年报告的原文:同一份报告对峰值区间有三种说法,与其所依据论文的交集是 4.2 至 4.5 星;报告没有印出曲线数值与置信区间,下降的心理机制未经检验;示意目录表明,小样本噪声与适度刷单都解释不了整条倒 U 曲线。 (https://5dlibrary.com/answers/zh/star-rating-j-shape-bayesian-average-nine-rank-inflation/c02/) - Ch.3 The Wilson Lower Bound: Why Three Five-Star Reviews Rank Below Two Hundred at 4.6 — Chapter 3 sorts by the Wilson lower bound: 0.4385 for three five-star reviews against 0.8859 for two hundred at 4.6, while the textbook Wald interval can cover the truth as rarely as 0.049 at small samples and high shares. (https://5dlibrary.com/answers/en/star-rating-j-shape-bayesian-average-nine-rank-inflation/c03/) 第3章 威尔逊下界:3 条五星为什么排不过 200 条 4.6 星 — 第三章用威尔逊下界给出排序:3 条五星为 0.4385,200 条 4.6 星为 0.8859;教科书 Wald 区间在小样本高好评率处覆盖率可低到 0.049。 (https://5dlibrary.com/answers/zh/star-rating-j-shape-bayesian-average-nine-rank-inflation/c03/) - Ch.4 The Dirichlet-Prior Bayesian Average: Pseudo-Reviews Borrowed from the Shop — Chapter 4 uses the shop's star mix as a Dirichlet prior, giving Bayesian averages of 4.415 and 4.583, and discusses how to estimate the prior strength. (https://5dlibrary.com/answers/en/star-rating-j-shape-bayesian-average-nine-rank-inflation/c04/) 第4章 狄利克雷先验贝叶斯平均:向全店目录借来的伪评论 — 第四章用全店星级构成作狄利克雷先验,两者的贝叶斯平均分别为 4.415 与 4.583,并讨论先验强度的估计。 (https://5dlibrary.com/answers/zh/star-rating-j-shape-bayesian-average-nine-rank-inflation/c04/) - Ch.5 Sorting Is Soliciting: the Shelf Feedback Loop, the Cost of Exploring and the First Five Reviews — A sort key is also a policy for collecting reviews. On the illustrative shelf, the beta posterior mean loses 85.8 orders and tests 6.12 products; the Wilson bound, the most cautious, tests only 3.00 and loses 118.8; Thompson sampling tests 18.38 and loses 131.6 in the short run. The first 5 reviews raise the weight of a product's own data to 0.333, while reviews 6 to 50 add only 0.500 more, so effort to solicit reviews should go first to products with fewer than 5. Do ratings, like rankings, drift as a whole because raters look at one another? Chapter 6 takes up rating inflation. (https://5dlibrary.com/answers/en/star-rating-j-shape-bayesian-average-nine-rank-inflation/c05/) 第5章 排序即求评:货架反馈环、探索代价与前五条评论 — 排序键也是求评策略。在示意货架上,贝塔后验均值损失 85.8 单、检验 6.12 件商品;威尔逊下界最保守,只检验 3.00 件,损失 118.8 单;汤普森抽样检验 18.38 件,短期损失 131.6 单。前 5 条评论把自身数据权重提高到 0.333,第 6 至 50 条只再加 0.500,所以求评资源应优先给评论不足 5 条的商品。评分会不会像排序一样,因为互相参照而整体漂移?第六章讨论评分通胀。 (https://5dlibrary.com/answers/zh/star-rating-j-shape-bayesian-average-nine-rank-inflation/c05/) - Ch.6 Rating Inflation: Norm Chasing, Information Decay and the Friction of Asking — Chapter 6 models rating inflation as norm chasing: with honest feelings unchanged, the mean rises from 3.485 to 4.090 in nine years and more reviews are needed to tell sellers apart; it also shows how the friction of asking decides who speaks. (https://5dlibrary.com/answers/en/star-rating-j-shape-bayesian-average-nine-rank-inflation/c06/) 第6章 评分通胀:规范追逐、信息衰减与求评摩擦 — 第六章用规范追逐模型说明评分通胀:诚实感受不变,均分九年间从 3.485 升到 4.090,分辨两个卖家所需的评论随之增加;并讨论求评摩擦如何决定谁开口。 (https://5dlibrary.com/answers/zh/star-rating-j-shape-bayesian-average-nine-rank-inflation/c06/) - Ch.7 Classic Texts East and West: Liu Yi on the Nine-Rank System and Galton's "Vox Populi" — Chapter 7 examines Liu Yi's critique of the Nine-Rank system in the Book of Jin, "no humble families above, no powerful clans below" (with family weighted about three times talent, humble men hold 0.072 of the upper grades in the model), and Galton's "Vox Populi" of 1907 (the crowd mean rebuilt as a two-piece normal is 1197.5 lb). (https://5dlibrary.com/answers/en/star-rating-j-shape-bayesian-average-nine-rank-inflation/c07/) 第7章 东西方原典考辨:《晋书·刘毅传》的九品之弊与高尔顿的《民意》 — 第七章考辨《晋书·刘毅传》对九品中正制「上品无寒门,下品无势族」的批评(示意模型中门第权重约为才能的三倍时,上品寒门比例为 0.072)与高尔顿 1907 年的《民意》(两段正态重建的众人平均为 1197.5 磅)。 (https://5dlibrary.com/answers/zh/star-rating-j-shape-bayesian-average-nine-rank-inflation/c07/) - Ch.8 Closing the Loop: Shop Sorting, the Timing of Review Requests and a Falsification Plan — Chapter 8 gives a shop procedure: one invitation to every buyer, the red line that inviting only satisfied buyers inflates the mean by +0.583, sorting by the posterior lower bound, the distilled Review Sorter, and pre-registered tests needing 53,211 visits per arm, with the conditions that would overturn the book. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/star-rating-j-shape-bayesian-average-nine-rank-inflation/c08/) 第8章 实证闭环:店铺排序算法、求评时机与证伪方案 — 第八章给出店铺流程:向全部买家发同一封邀请、只邀满意买家会使均分虚高 +0.583 的红线、后验下界排序、蒸馏小工具「评论排序器」,以及每组 53,211 次访问的预先登记检验和推翻本书的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/star-rating-j-shape-bayesian-average-nine-rank-inflation/c08/) ## 5DL-B-213 · Nonlinear Commission Ladders, End-of-Period Timing Manipulation and Quota-Induced Game Distortion 中文书名:销售提成阶梯非线性、期末冲量操纵与业绩配额博弈扭曲母典 DOI: pending Cite: Fan, Y. (2026). Nonlinear Commission Ladders, End-of-Period Timing Manipulation and Quota-Induced Game Distortion (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: Owners widely believe that commissions need steps: the more you sell, the more you earn, with a target bonus at some number. This book tests a statement pointing the other way: a stepped target bonus makes salespeople push deals into the next quarter when the target is out of reach or has just been met, and buy next quarter's deals in with discounts when one or two deals short; a linear commission often leaves the owner more profit than a steep ladder. Its salespeople, shops, rates and discounts are illustrative; all calculations are generated by its programs; the book is non-personalised business guidance, not legal, tax or financial advice, and its statements about regulations are subject to human review before publication. Chapter 1 writes linear, step and accelerator plans as tables, defines the kink ratio, and uses Jensen's inequality to show that convex plans reward bunching while plans with a jump reward trimming. Chapter 2 checks Oyer's study of fiscal-year-end seasonality and Larkin's finding that discounted pull-ins cost 6% to 8% of revenue, and sets beside them the evidence of Steenburgh and of Chung and co-authors that bonuses mainly buy effort; in the book's model each extra deal of effort comes with 0.78 deals moved. Chapter 3 finds the salesperson's timing policy by dynamic programming: under the accelerator the last two weeks run at 1.32 times the average week, while the step plan's quarter end slumps to 0.67 times. Chapter 4 compares the owner's profit by Monte Carlo: the linear plan earns 0.2266 a quarter more than the step plan, or 0.089; but at equal pay and one elasticity of effort the step plan wins narrowly, by 0.017, so the claim can only say "often". Chapter 5 shows that when the quota follows results upward, long-run profit falls from 2.5333 to 2.2633. Chapter 6 designs a test on weekly shares and a quarter-end discount monitor, which catches the accelerator with probability 0.540 from only 4 people over 1 quarter. Chapter 7 examines the "unified rewards" of the Book of Lord Shang and Taylor's differential piece rate: with the same bonus split into eight steps, the gain from moving a deal is only 0.0249 of expected pay. Chapter 8 gives a commission plan simulator, monitoring red lines and a pre-registered test protocol; softening the step plan raises the model's profit from 2.5347 to 2.7005. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:经营者普遍相信提成要有梯度:卖得越多拿得越多,到了某个数再给一笔达标奖。本书检验一个相反的说法:阶梯达标奖会让销售员在达标无望或刚刚达标之后把单子压到下个季度,在差一两单时用折扣把下个季度的单子抢进来;线性提成往往比高阶梯给老板留下更多利润。书中的销售员、门店、费率与折扣均为示意参数;全部计算由书中程序生成;本书是非个人化的经营建议,不构成法律、税务或理财建议,涉及法规的表述以上架前人工复核为准。 第一章把线性、台阶与加速三种方案写成表,定义折点比,并用詹森不等式说明凸的方案奖励堆单、带跳台的方案奖励削平。第二章核对 Oyer 关于财年末季节性的研究与 Larkin 关于折扣抢单造成收入 6% 至 8% 损失的研究,并陈 Steenburgh 与 Chung 等人认为奖金主要带来努力的反方证据;在本书模型中,每多一单努力伴随 0.78 单挪动。第三章用动态规划求出销售员的成交时点策略:加速方案季末两周是平均周的 1.32 倍,台阶方案的季末反而塌陷到 0.67 倍。第四章用蒙特卡洛模拟比较老板利润,线性方案比台阶方案每季度多 0.2266,相当于 0.089;但在同等薪酬与某一努力弹性下,台阶方案反而小胜 0.017,所以命题只能说「往往」。 第五章说明配额随实绩上调时,长期利润从 2.5333 降到 2.2633。第六章设计周度份额检验与期末折扣率监控,后者只看 4 个人 1 个季度就能以 0.540 的概率发现加速方案。第七章考辨《商君书·赏刑》的「壹赏」与泰勒的差别计件:同一笔奖金拆成八个台阶时,挪单收益只占期望薪酬的 0.0249。第八章给出提成方案模拟器、监控红线与预注册检验方案;把台阶方案软化之后,模型中的利润从 2.5347 升到 2.7005。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Geometry of a Commission Ladder: Kinks, Convexity and Marginal Rates — The shape of a commission plan matters more than its average level. The linear plan has a kink ratio of 1.00 and is wholly neutral about timing. The step plan raises the marginal pay of the target deal to 0.280, a kink ratio of 5.09, and rewards trimming each quarter to just reach the line. The accelerator, with a kink ratio of 2.67, is convex and rewards piling deals into one quarter. As long as a salesperson can decide which quarter a deal falls in, these shapes turn into behaviour. (https://5dlibrary.com/answers/en/commission-ladder-end-of-period-timing-quota-distortion/c01/) 第1章 提成阶梯的几何:折点、凸性与边际提成率 — 提成方案的形状比它的平均水平更重要。线性方案的折点比为 1.00,对成交时点完全中立;台阶方案在达标那一单上把边际提成抬到 0.280,折点比 5.09,奖励把每个季度削平到刚好达标;加速方案折点比 2.67,是凸的,奖励把单子堆进同一个季度。只要销售员能决定一单落在哪个季度,这些形状就会转化为行为。 (https://5dlibrary.com/answers/zh/commission-ladder-end-of-period-timing-quota-distortion/c01/) - Ch.2 The Evidence for the Claim: Fiscal-Year Seasonality, Discounted Pull-Ins and the Case Against — Chapter 2 checks Oyer's study of fiscal-year-end seasonality and Larkin's finding that discounted pull-ins cost 6% to 8% of revenue, and sets beside them the evidence of Steenburgh and of Chung and co-authors that bonuses mainly buy effort; in the book's model each extra deal of effort comes with 0.78 deals moved. (https://5dlibrary.com/answers/en/commission-ladder-end-of-period-timing-quota-distortion/c02/) 第2章 反常识命题的证据:财年末季节性、折扣抢单与反方数据 — 第二章核对 Oyer 关于财年末季节性的研究与 Larkin 关于折扣抢单造成收入 6% 至 8% 损失的研究,并陈 Steenburgh 与 Chung 等人认为奖金主要带来努力的反方证据;在本书模型中,每多一单努力伴随 0.78 单挪动。 (https://5dlibrary.com/answers/zh/commission-ladder-end-of-period-timing-quota-distortion/c02/) - Ch.3 Dynamic Programming of Close Timing: Holding, Dragging and Quarter-End Pull-Ins — Chapter 3 finds the salesperson's timing policy by dynamic programming: under the accelerator the last two weeks run at 1.32 times the average week, while the step plan's quarter end slumps to 0.67 times. (https://5dlibrary.com/answers/en/commission-ladder-end-of-period-timing-quota-distortion/c03/) 第3章 成交时点的动态规划:压单、拖单与期末抢单 — 第三章用动态规划求出销售员的成交时点策略:加速方案季末两周是平均周的 1.32 倍,台阶方案的季末反而塌陷到 0.67 倍。 (https://5dlibrary.com/answers/zh/commission-ladder-end-of-period-timing-quota-distortion/c03/) - Ch.4 The Owner's Profit: a Monte Carlo Contest of Linear and Ladder Plans — Chapter 4 compares the owner's profit by Monte Carlo: the linear plan earns 0.2266 a quarter more than the step plan, or 0.089; but at equal pay and one elasticity of effort the step plan wins narrowly, by 0.017, so the claim can only say "often". (https://5dlibrary.com/answers/en/commission-ladder-end-of-period-timing-quota-distortion/c04/) 第4章 老板的利润:线性与阶梯方案的蒙特卡洛对决 — 第四章用蒙特卡洛模拟比较老板利润,线性方案比台阶方案每季度多 0.2266,相当于 0.089;但在同等薪酬与某一努力弹性下,台阶方案反而小胜 0.017,所以命题只能说「往往」。 (https://5dlibrary.com/answers/zh/commission-ladder-end-of-period-timing-quota-distortion/c04/) - Ch.5 The Quota Ratchet and Where to Set the Quota: Holding Back and the "One or Two Short" Band — Chapter 5 shows that when the quota follows results upward, long-run profit falls from 2.5333 to 2.2633. (https://5dlibrary.com/answers/en/commission-ladder-end-of-period-timing-quota-distortion/c05/) 第5章 配额棘轮与配额定位:藏力与「差一两单」区间 — 第五章说明配额随实绩上调时,长期利润从 2.5333 降到 2.2633。 (https://5dlibrary.com/answers/zh/commission-ladder-end-of-period-timing-quota-distortion/c05/) - Ch.6 Detecting the Quarter-End Push: a Test on Weekly Shares and a Discount Monitor — Chapter 6 designs a test on weekly shares and a quarter-end discount monitor, which catches the accelerator with probability 0.540 from only 4 people over 1 quarter. (https://5dlibrary.com/answers/en/commission-ladder-end-of-period-timing-quota-distortion/c06/) 第6章 期末冲量的检测:周度份额检验与折扣率监控 — 第六章设计周度份额检验与期末折扣率监控,后者只看 4 个人 1 个季度就能以 0.540 的概率发现加速方案。 (https://5dlibrary.com/answers/zh/commission-ladder-end-of-period-timing-quota-distortion/c06/) - Ch.7 Classical Sources East and West: "Rewards and Punishments" in the Book of Lord Shang and Taylor's "A Piece-Rate System" — Chapter 7 examines the "unified rewards" of the Book of Lord Shang and Taylor's differential piece rate: with the same bonus split into eight steps, the gain from moving a deal is only 0.0249 of expected pay. (https://5dlibrary.com/answers/en/commission-ladder-end-of-period-timing-quota-distortion/c07/) 第7章 东西方原典考辨:《商君书·赏刑》与泰勒《计件报酬制》 — 第七章考辨《商君书·赏刑》的「壹赏」与泰勒的差别计件:同一笔奖金拆成八个台阶时,挪单收益只占期望薪酬的 0.0249。 (https://5dlibrary.com/answers/zh/commission-ladder-end-of-period-timing-quota-distortion/c07/) - Ch.8 Closing the Loop: a Commission Plan Simulator, a Monitoring Sheet and a Pre-Registered Test — Chapter 8 gives a commission plan simulator, monitoring red lines and a pre-registered test protocol; softening the step plan raises the model's profit from 2.5347 to 2.7005. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/commission-ladder-end-of-period-timing-quota-distortion/c08/) 第8章 实证闭环:提成方案模拟器、监控表与预注册检验 — 第八章给出提成方案模拟器、监控红线与预注册检验方案;把台阶方案软化之后,模型中的利润从 2.5347 升到 2.7005。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/commission-ladder-end-of-period-timing-quota-distortion/c08/) ## 5DL-B-214 · Huff Gravity Siting for Pop-Up Stores, Poisson Thinning of Foot Traffic and Central-Place Rent Gradients 中文书名:快闪店赫夫引力选址、步行客流泊松稀释与中心地等级租金梯度母典 DOI: pending Cite: Fan, Y. (2026). Huff Gravity Siting for Pop-Up Stores, Poisson Thinning of Foot Traffic and Central-Place Rent Gradients (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses, and the book comes with a Pop-up Site Scorer. Abstract: The popular rule for siting a pop-up store is that footfall is king. This book breaks it into models that can be computed, tested and overturned: sales come from target density times dwell time, not from total footfall, and a site at twice the rent pays its way only above a conversion threshold that can be solved for. The streets, sites, footfall, rents and costs in the book are illustrative; all calculations are generated by its programs; it offers non-personalised business guidance, not legal or financial advice, and lease terms are subject to manual review before publication. Chapter 1 uses three street types to show that a crossing with 4.00 times the passers of a lane earns only 0.212 of the lane's daily margin, and cites store research from an apparel chain: sales show diminishing returns to traffic, and conversion falls as traffic rises. Chapter 2 solves for the break-even conversion rate: a base threshold of 0.1070 rises to 0.1783 when rent doubles, only 1.667 times as high, because fixed costs do not move with rent; if a landlord discounts a tenant who brings traffic, the threshold can fall to 0.0756. Chapter 3 uses Poisson thinning to show why a daily report should divide by entrants. Chapter 4 compares three candidate sites with the Huff model: the site next to the target segment captures 4922 trips a week against 2464 at the central crossing, and the two exponents must be calibrated on a local survey. Chapter 5 finds the best season length under fading novelty (a turning point at 3.94 weeks) and shows that the pair of pop-ups chosen by footfall earns only 10358 a week against 259142 for the pair chosen by profit. Chapter 6 uses bid-rent curves and Christaller's threshold to show that central rent is 3.86 times the rent three kilometres out while the conversion required is only 1.27 times, and that high-threshold goods survive only through rotating pop-ups. Chapter 7 examines the shops and night markets of Wu Zimu's Mengliang lu and Christaller's hexagons. Chapter 8 gives a site scorecard, a daily funnel report and a pre-registered test for seasons in Taipei and Kuala Lumpur. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses, and the book comes with a Pop-up Site Scorer. 摘要:快闪店选址的流行说法是「人流为王」。本书把它拆成可以计算、可以检验、也可以被推翻的模型:成交来自目标客层密度乘停留时间,而不是总人流;租金贵一倍的铺位,回本需要的转化率是一个可以反解的阈值。书中的街道、铺位、人流、租金与成本均为示意;全部计算由书中程序生成;本书是非个人化经营建议,不构成法律或理财建议,租约条款以上架前人工复核为准。 第一章用三种街型说明,路人是巷弄 4.00 倍的大路口,日毛利只有巷弄的 0.212,并引用服装连锁的门店研究:销售对人流报酬递减,转化率随人流下降。第二章反解盈亏平衡转化率:基准门槛 0.1070,租金翻倍时升到 0.1783,只是原来的 1.667 倍,因为固定成本不随租金变化;若业主给带来人流的租户打折,门槛可降到 0.0756。第三章用泊松稀释说明日报为何应以进店数为分母。第四章用赫夫模型比较三个候选铺位:靠近目标客层的铺位每周拿到 4922 次出行,市中心路口只有 2464 次,并说明两个指数须用本地问卷标定。 第五章求出新鲜感衰减下的最优档期(拐点 3.94 周),并说明按人流挑出的两家快闪店周利润只有 10358,按利润挑出的组合为 259142。第六章用竞租曲线与克里斯塔勒的门槛说明:中心租金是三公里外的 3.86 倍,所需转化率只是 1.27 倍,而高门槛商品只能靠轮转的快闪生存。第七章考辨吴自牧《梦粱录》的铺席夜市与克里斯塔勒的六边形。第八章给出台北与吉隆坡档期的选址打分表、漏斗日报与预登记检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说,并附快闪选址打分器一件。 Chapter answers: - Ch.1 The Crossroads Trap: Total Footfall, Target Density and Dwell Time — Chapter 1 uses three street types to show that a crossing with 4.00 times the passers of a lane earns only 0.212 of the lane's daily margin, and cites store research from an apparel chain: sales show diminishing returns to traffic, and conversion falls as traffic rises. (https://5dlibrary.com/answers/en/popup-store-huff-gravity-poisson-thinning-central-place-rent/c01/) 第1章 最大路口的陷阱:总人流、目标客层密度与停留时间 — 第一章用三种街型说明,路人是巷弄 4.00 倍的大路口,日毛利只有巷弄的 0.212,并引用服装连锁的门店研究:销售对人流报酬递减,转化率随人流下降。 (https://5dlibrary.com/answers/zh/popup-store-huff-gravity-poisson-thinning-central-place-rent/c01/) - Ch.2 When Rent Doubles: Solving for the Break-Even Conversion Rate — Chapter 2 solves for the break-even conversion rate: a base threshold of 0.1070 rises to 0.1783 when rent doubles, only 1.667 times as high, because fixed costs do not move with rent; if a landlord discounts a tenant who brings traffic, the threshold can fall to 0.0756. (https://5dlibrary.com/answers/en/popup-store-huff-gravity-poisson-thinning-central-place-rent/c02/) 第2章 租金翻倍的回本阈值:盈亏平衡转化率反解 — 第二章反解盈亏平衡转化率:基准门槛 0.1070,租金翻倍时升到 0.1783,只是原来的 1.667 倍,因为固定成本不随租金变化;若业主给带来人流的租户打折,门槛可降到 0.0756。 (https://5dlibrary.com/answers/zh/popup-store-huff-gravity-poisson-thinning-central-place-rent/c02/) - Ch.3 Poisson Thinning: the Random Laws Linking Passers, Entrants and Buyers — In the simulation, passers average 249.767 with a variance-to-mean ratio of 0.983; entrants average 14.991 against a theoretical 15.000; buyers average 4.503 against 4.500. All three counts have a variance-to-mean ratio close to one, the hallmark of the Poisson law. The correlation between buyers and browsers is -0.0120, indistinguishable from zero within the sampling error of 6000 hours. (https://5dlibrary.com/answers/en/popup-store-huff-gravity-poisson-thinning-central-place-rent/c03/) 第3章 泊松稀释:路过、进店、成交三级计数的随机规律 — 模拟中,路人数的均值为 249.767,方差与均值之比为 0.983;进店数均值 14.991,理论值 15.000;成交数均值 4.503,理论值 4.500。三个计数的方差与均值之比都接近一,正是泊松分布的特征。成交者与只逛不买者的相关系数为 -0.0120,在 6000 小时的抽样误差范围内与零无异。 (https://5dlibrary.com/answers/zh/popup-store-huff-gravity-poisson-thinning-central-place-rent/c03/) - Ch.4 Huff Gravity: Trade-Area Probabilities, Distance Decay and Calibration — Chapter 4 compares three candidate sites with the Huff model: the site next to the target segment captures 4922 trips a week against 2464 at the central crossing, and the two exponents must be calibrated on a local survey. (https://5dlibrary.com/answers/en/popup-store-huff-gravity-poisson-thinning-central-place-rent/c04/) 第4章 赫夫引力:商圈概率、距离衰减与参数标定 — 第四章用赫夫模型比较三个候选铺位:靠近目标客层的铺位每周拿到 4922 次出行,市中心路口只有 2464 次,并说明两个指数须用本地问卷标定。 (https://5dlibrary.com/answers/zh/popup-store-huff-gravity-poisson-thinning-central-place-rent/c04/) - Ch.5 Seasons and Portfolios: Fading Novelty, the Best Number of Weeks and Two Shops Eating Each Other — Chapter 5 finds the best season length under fading novelty (a turning point at 3.94 weeks) and shows that the pair of pop-ups chosen by footfall earns only 10358 a week against 259142 for the pair chosen by profit. (https://5dlibrary.com/answers/en/popup-store-huff-gravity-poisson-thinning-central-place-rent/c05/) 第5章 档期与组合:新鲜感衰减、最优周数与两店互食 — 第五章求出新鲜感衰减下的最优档期(拐点 3.94 周),并说明按人流挑出的两家快闪店周利润只有 10358,按利润挑出的组合为 259142。 (https://5dlibrary.com/answers/zh/popup-store-huff-gravity-poisson-thinning-central-place-rent/c05/) - Ch.6 Central-Place Hierarchy and Rent Gradients: Thresholds, Ranges and Rotating Pop-Ups — Chapter 6 uses bid-rent curves and Christaller's threshold to show that central rent is 3.86 times the rent three kilometres out while the conversion required is only 1.27 times, and that high-threshold goods survive only through rotating pop-ups. (https://5dlibrary.com/answers/en/popup-store-huff-gravity-poisson-thinning-central-place-rent/c06/) 第6章 中心地等级与租金梯度:门槛、销售范围与轮转快闪 — 第六章用竞租曲线与克里斯塔勒的门槛说明:中心租金是三公里外的 3.86 倍,所需转化率只是 1.27 倍,而高门槛商品只能靠轮转的快闪生存。 (https://5dlibrary.com/answers/zh/popup-store-huff-gravity-poisson-thinning-central-place-rent/c06/) - Ch.7 Sources East and West: the Shops and Night Markets of the Mengliang lu and Christaller's Hexagons — The admissible numbers up to thirty are 1, 3, 4, 7, 9, 12, 13, 16, 19, 21, 25, 27 and 28, and Christaller's three principles take the three smallest non-trivial values. Under the marketing principle the number of new centres at each level is 1, 2, 6, 18, 54, 162 and 486, multiplying by three at each step; under the traffic principle it is 1, 3, 12, 48, 192, 768 and 3072; the administrative principle grows fastest, reaching 100842 by the seventh level. The spacing between centres grows by $\sqrt{K}$ per level: 1.732, 2.000 and 2.646 for the three principles. If the lowest level's spacing is set at 7.0 kilometres (an illustrative value, not Christaller's), the seven levels under the marketing principle are spaced 7.0, 12.1, 21.0, 36.4, 63.0, 109.1 and 189.0 kilometres apart, and one top-level market area contains 729 lowest-level areas. (https://5dlibrary.com/answers/en/popup-store-huff-gravity-poisson-thinning-central-place-rent/c07/) 第7章 东西方原典考辨:《梦粱录》的铺席夜市与克里斯塔勒的六边形 — 三十以内满足条件的数是 1、3、4、7、9、12、13、16、19、21、25、27、28;克里斯塔勒的三条原则对应其中最小的三个非平凡值。按供给原则,各级新增的中心地数目是 1、2、6、18、54、162、486,层层乘以三;按交通原则是 1、3、12、48、192、768、3072;按行政原则增长最快,第七级已有 100842 个。中心地之间的间距每升一级乘以 $\sqrt{K}$,三条原则分别为 1.732、2.000、2.646。若把最低一级的间距设为 7.0 公里(示意值,不是克里斯塔勒的数),供给原则下七级的间距依次为 7.0、12.1、21.0、36.4、63.0、109.1、189.0 公里,一个最高级市场区里包含 729 个最低级市场区。 (https://5dlibrary.com/answers/zh/popup-store-huff-gravity-poisson-thinning-central-place-rent/c07/) - Ch.8 Closing the Loop: Site Scorecards, Daily Funnel Reports and a Pre-Registered Test for Taipei and Kuala Lumpur Seasons — Chapter 8 gives a site scorecard, a daily funnel report and a pre-registered test for seasons in Taipei and Kuala Lumpur. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses, and the book comes with a Pop-up Site Scorer. (https://5dlibrary.com/answers/en/popup-store-huff-gravity-poisson-thinning-central-place-rent/c08/) 第8章 实证闭环与证伪方案:台北与吉隆坡档期的打分表、漏斗日报与预登记检验 — 第八章给出台北与吉隆坡档期的选址打分表、漏斗日报与预登记检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说,并附快闪选址打分器一件。 (https://5dlibrary.com/answers/zh/popup-store-huff-gravity-poisson-thinning-central-place-rent/c08/) ## 5DL-B-215 · Anthropometric Principal-Component Sizing, Vanity-Size Drift and Cross-Border Return-Rate Minimisation 中文书名:人体测量主成分尺码体系、虚荣尺码漂移与跨境退货率最小化母典 DOI: pending Cite: Fan, Y. (2026). Anthropometric Principal-Component Sizing, Vanity-Size Drift and Cross-Border Return-Rate Minimisation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its brands, shops, prices and body dimensions are illustrative; all calculations are generated by its programs; the book is not individual business advice, deals only with the measurement of garment sizes, and makes no health judgement. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: The most common response of online apparel shops to returns for size is to make the size chart finer. This book refutes the belief behind that practice: garments marked with the same M can differ by several centimetres in finished chest from brand to brand, and against that dispersion a finer size chart reduces returns far less than inferring brand residuals from the items a customer has already kept and converting them into a recommendation in the new brand. Its brands, shops, prices and body dimensions are illustrative; all calculations are generated by its programs; the book is not individual business advice, deals only with the measurement of garment sizes, and makes no health judgement. Chapter 1 writes brand offsets as a random effect: with an offset standard deviation of 2 cm, two brands' M differ by more than a full grade with probability 0.1573, and one customer needs 2.147 different labels across five brands. Chapter 2 splits one size choice into five errors: a body size chart has a return probability of 0.2170, a finer chart 0.1760 and even a perfect chart 0.1636, whereas the brand-residual method gives 0.0302 in closed form and 0.0499 once the selection of kept items is simulated honestly. Chapter 3 shows that the first principal component of height and chest carries only 0.654 of the variance, so sixteen one-axis sizes cover only 0.6022 while 11 sizes on a two-dimensional grid cover 0.9. Chapter 4 uses empirical Bayes shrinkage to cut the mean squared error of brand offsets to 0.808 of the raw means, and shows that with one garment measured per brand the "largest-running" brand on a page is overstated by 1.3922 cm on average. Chapter 5 tracks vanity-size drift with a Kalman filter and shows that under drift a 12-season average, with error 0.6015, does worse than a 4-season average at 0.4640. Chapter 6 converts errors into cross-border costs: the residual method saves 15.08 money units per order over a finer chart. Chapter 7 examines the tailors and the office of women's work in the Zhou li (made-to-measure misfits with probability only 0.00362) and Quetelet's average man (only 0.3750 of the soldiers lie within one inch of the mean chest). Chapter 8 gives a buyer's direct-message size recommender (fit probability 0.9441 in the worked example), a return-reason code table and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:服装电商应对尺码退货,最常见的做法是把尺码表做细。本书证伪这个做法背后的信念:同样标 M 的衣服,不同品牌的成衣胸围可以差出好几厘米;在品牌离散面前,更细的尺码表减少退货的效果,远不如利用顾客已经留下的单品推算品牌残差、再换算到新品牌去推荐。书中的品牌、门店、价格与身体尺寸参数均为示意;全部计算由书中程序生成;本书是非个人化经营建议,只涉及服装尺寸的测量,不作任何健康判断。 第一章把品牌偏移写成随机效应:偏移标准差为 2 厘米时,两个品牌的 M 相差超过一整档的概率是 0.1573,一位顾客在五个品牌里平均需要 2.147 个不同的标签。第二章把一次选码的误差拆成五块:身体尺码表的退货概率是 0.2170,细尺码表 0.1760,完美尺码表也只有 0.1636,而品牌残差法的闭式解是 0.0302,如实模拟留存单品的筛选后是 0.0499。第三章说明身高与胸围的第一主成分只承载 0.654 的方差,一维尺码用十六档也只覆盖 0.6022,两维网格用 11 档即覆盖 0.9。第四章用经验贝叶斯收缩把品牌偏移的均方误差降到原始均值的 0.808,并说明只量一件衣服时页面上「最偏大」的品牌平均被高估 1.3922 厘米。 第五章用卡尔曼滤波跟踪虚荣尺码漂移,说明漂移存在时 12 季平均的误差 0.6015 反而不如 4 季平均的 0.4640。第六章换算跨境退货成本:残差法相对细尺码表每单节省 15.08 货币单位。第七章考辨《周礼》缝人与典妇功(量体裁衣的不合身机会只有 0.00362)以及凯特勒的平均人(平均胸围上下一英寸内只有 0.3750 的士兵)。第八章给出买手私讯尺码推荐器(示范推荐的合身概率 0.9441)、退货原因编码表与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A Label Is Not a Measurement: Brand Dispersion of the Same "M" as a Random Effect — Chapter 1 writes brand offsets as a random effect: with an offset standard deviation of 2 cm, two brands' M differ by more than a full grade with probability 0.1573, and one customer needs 2.147 different labels across five brands. (https://5dlibrary.com/answers/en/anthropometric-pca-sizing-vanity-drift-return-minimisation/c01/) 第1章 标签不是尺寸:同标 M 的品牌离散与随机效应 — 第一章把品牌偏移写成随机效应:偏移标准差为 2 厘米时,两个品牌的 M 相差超过一整档的概率是 0.1573,一位顾客在五个品牌里平均需要 2.147 个不同的标签。 (https://5dlibrary.com/answers/zh/anthropometric-pca-sizing-vanity-drift-return-minimisation/c01/) - Ch.2 A Finer Size Chart or a Brand Residual: the Error Budget of Five Sizing Policies — Chapter 2 splits one size choice into five errors: a body size chart has a return probability of 0.2170, a finer chart 0.1760 and even a perfect chart 0.1636, whereas the brand-residual method gives 0.0302 in closed form and 0.0499 once the selection of kept items is simulated honestly. (https://5dlibrary.com/answers/en/anthropometric-pca-sizing-vanity-drift-return-minimisation/c02/) 第2章 细尺码表还是品牌残差:五种选码策略的误差预算 — 第二章把一次选码的误差拆成五块:身体尺码表的退货概率是 0.2170,细尺码表 0.1760,完美尺码表也只有 0.1636,而品牌残差法的闭式解是 0.0302,如实模拟留存单品的筛选后是 0.0499。 (https://5dlibrary.com/answers/zh/anthropometric-pca-sizing-vanity-drift-return-minimisation/c02/) - Ch.3 A Two-Dimensional Principal-Component Grid of Height and Girth: What a One-Axis Label Throws Away — Chapter 3 shows that the first principal component of height and chest carries only 0.654 of the variance, so sixteen one-axis sizes cover only 0.6022 while 11 sizes on a two-dimensional grid cover 0.9. (https://5dlibrary.com/answers/en/anthropometric-pca-sizing-vanity-drift-return-minimisation/c03/) 第3章 身高—围度两维主成分网格:一根轴的标签丢掉了什么 — 第三章说明身高与胸围的第一主成分只承载 0.654 的方差,一维尺码用十六档也只覆盖 0.6022,两维网格用 11 档即覆盖 0.9。 (https://5dlibrary.com/answers/zh/anthropometric-pca-sizing-vanity-drift-return-minimisation/c03/) - Ch.4 Shrinkage Estimates of Brand Offsets: How Far to Trust Two Measured Garments — Chapter 4 uses empirical Bayes shrinkage to cut the mean squared error of brand offsets to 0.808 of the raw means, and shows that with one garment measured per brand the "largest-running" brand on a page is overstated by 1.3922 cm on average. (https://5dlibrary.com/answers/en/anthropometric-pca-sizing-vanity-drift-return-minimisation/c04/) 第4章 品牌偏移的收缩估计:只量了两件衣服时该信几分 — 第四章用经验贝叶斯收缩把品牌偏移的均方误差降到原始均值的 0.808,并说明只量一件衣服时页面上「最偏大」的品牌平均被高估 1.3922 厘米。 (https://5dlibrary.com/answers/zh/anthropometric-pca-sizing-vanity-drift-return-minimisation/c04/) - Ch.5 Vanity-Size Drift: Tracking a Moving Brand Offset with a Kalman Filter — Chapter 5 tracks vanity-size drift with a Kalman filter and shows that under drift a 12-season average, with error 0.6015, does worse than a 4-season average at 0.4640. (https://5dlibrary.com/answers/en/anthropometric-pca-sizing-vanity-drift-return-minimisation/c05/) 第5章 虚荣尺码漂移:用卡尔曼滤波跟踪会走动的品牌偏移 — 第五章用卡尔曼滤波跟踪虚荣尺码漂移,说明漂移存在时 12 季平均的误差 0.6015 反而不如 4 季平均的 0.4640。 (https://5dlibrary.com/answers/zh/anthropometric-pca-sizing-vanity-drift-return-minimisation/c05/) - Ch.6 The Expected Cost of Returns: Sizing Up, Cross-Border Freight and Ordering Several Sizes — Chapter 6 converts errors into cross-border costs: the residual method saves 15.08 money units per order over a finer chart. (https://5dlibrary.com/answers/en/anthropometric-pca-sizing-vanity-drift-return-minimisation/c06/) 第6章 退货期望成本:向上取码、跨境运费与多码试穿 — 第六章换算跨境退货成本:残差法相对细尺码表每单节省 15.08 货币单位。 (https://5dlibrary.com/answers/zh/anthropometric-pca-sizing-vanity-drift-return-minimisation/c06/) - Ch.7 Sources East and West: the Palace Tailors of the Zhou li and Quetelet's Average Man — Chapter 7 examines the tailors and the office of women's work in the Zhou li (made-to-measure misfits with probability only 0.00362) and Quetelet's average man (only 0.3750 of the soldiers lie within one inch of the mean chest). (https://5dlibrary.com/answers/en/anthropometric-pca-sizing-vanity-drift-return-minimisation/c07/) 第7章 东西方原典考辨:《周礼》的缝人与凯特勒的平均人 — 第七章考辨《周礼》缝人与典妇功(量体裁衣的不合身机会只有 0.00362)以及凯特勒的平均人(平均胸围上下一英寸内只有 0.3750 的士兵)。 (https://5dlibrary.com/answers/zh/anthropometric-pca-sizing-vanity-drift-return-minimisation/c07/) - Ch.8 Closing the Loop: a Buyer's Direct-Message Size Recommender, a Return-Reason Code Table and a Pre-Registered Test — Chapter 8 gives a buyer's direct-message size recommender (fit probability 0.9441 in the worked example), a return-reason code table and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/anthropometric-pca-sizing-vanity-drift-return-minimisation/c08/) 第8章 实证闭环:买手私讯尺码推荐器、退货原因编码表与预注册检验 — 第八章给出买手私讯尺码推荐器(示范推荐的合身概率 0.9441)、退货原因编码表与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/anthropometric-pca-sizing-vanity-drift-return-minimisation/c08/) ## 5DL-B-216 · Benford's First-Digit Law, Ledger Fraud Detection and Chi-Square Audits of Buying-Agent Cash Flows and Expense Claims 中文书名:本福特首位数定律、账目舞弊检测与代购流水报销卡方审计母典 DOI: pending Cite: Fan, Y. (2026). Benford's First-Digit Law, Ledger Fraud Detection and Chi-Square Audits of Buying-Agent Cash Flows and Expense Claims (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not legal advice and not personalised business advice, and test results are only signals for review. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Popular audit lore says that in real books amounts beginning with 1 are the most common, that invented numbers use the nine first digits equally, and that counting first digits therefore catches false books. This book splits the lore into two halves and tests each with published research and programs that run. The first half is only partly right: when people invent numbers, their first digits fall between an even spread and Benford's curve, and the steadier tells are a refusal to repeat amounts and an avoidance of endings in 0 and 5. The second half is dangerously wrong: honest ledgers dominated by a fixed price, capped by an allowance or squeezed into one decade are born outside the law, and auditing them against it wrongs the innocent. Its parameters are illustrative; all calculations are generated by its programs; the book is not legal advice and not personalised business advice, and test results are only signals for review. Chapter 1 explains the law through uniform mantissae, scale invariance and multiplicative chains; the expected share of a leading 1 is 0.30103. Chapter 2 states the counterintuitive claim head on: a naive inventor's first-digit MAD is more than three times that of an honest ledger (0.0415 against 0.0133), honest fixed-price, capped and narrow-range ledgers are all flagged by chi-square at a rate of 1.000, and even small broad honest ledgers are misjudged by the MAD standard at a rate of 0.583. Chapter 3 shows that nine per-digit z tests give a thousand honest records a 0.3242 chance of being marked, and that data almost following the law are rejected by chi-square at ten thousand rows with probability 0.999. Chapter 4 gives the noise floor of the MAD, 0.2348 divided by the square root of the row count, and catches piling below an approval limit with the first-two-digit test at a detection rate of 0.794. Chapter 5 turns the precondition into a measurable gate: a worst-case MAD below 0.015 needs a span of 1.57 decades, and no single amount may take a tenth of the rows. Chapter 6 covers last digits, repeats and the limit window: honest rounding lifts the share of 0 and 5 endings to 0.5616 while an inventor presses it to 0.0911, and an honest ledger of 100 rows has only a 0.0025 chance of no repeat. Chapter 7 weighs the four-pillar settlement, Newcomb's note of 1881 and Pacioli's double entry of 1494. Chapter 8 gives a monthly self-audit, a three-light review, the distilled tool (the Ledger First-Digit Auditor) and a pre-registration plan: the gate cuts honest shops' false-alarm rate from 0.4280 to 0.0280 and, with invention at five per cent, lifts the positive predictive value from 0.110 to 0.653. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:流行的查账口诀说:真实账目以 1 开头的金额最多,编造的数字九个首位数用得一样多,所以数一数首位数就能抓假账。本书把这条口诀拆成两半,用已发表的研究与可运行的程序分别检验。前一半只对了一部分:人凭空写数时,首位数介于均匀分布与本福特分布之间,更稳定的破绽是不肯写重复金额、回避 0 和 5 结尾。后一半错得危险:主力商品固定价、有报销上限或挤在一个数量级内的诚实账本,天生就不服从定律,拿定律去审,会冤枉人。书中参数为示意;全部计算由书中程序生成;本书不是法律意见,也不是个人化经营建议,检验结果只是复核信号。 第一章从尾数均匀、尺度不变与乘法链解释定律,首位数为 1 的期望是 0.30103。第二章正面陈述反常识命题:朴素编造者的首位数 MAD 是诚实账本的三倍多(0.0415 对 0.0133),而固定价、有上限与窄区间的诚实账本被卡方判为偏离的比例都是 1.000,宽分布的诚实小账本被 MAD 标准误判的比例也有 0.583。第三章说明九次逐个数字的 z 检验让一千笔诚实记录有 0.3242 的机会被标红,几乎服从定律的数据在一万笔上被卡方拒绝的概率是 0.999。第四章给出 MAD 的噪声底线 0.2348 除以笔数的平方根,并用前两位数检验抓住审批限额下的堆积,检出率 0.794。 第五章把适用前提写成可测量的闸门:最坏情况 MAD 低于 0.015 需要跨越 1.57 个数量级,单一金额不能占到一成。第六章讲末位、重复与限额窗口:诚实凑整把 0、5 结尾推到 0.5616,编造者压到 0.0911;100 笔诚实账一笔不重复的概率只有 0.0025。第七章考辨四柱结算法、纽科姆 1881 年的短文与帕乔利 1494 年的复式记账。第八章给出月度自检流程、三色灯复核、蒸馏小工具「流水首位数审计器」与预注册方案:闸门把诚实店铺的误报率从 0.4280 降到 0.0280,编造占百分之五时阳性预测值从 0.110 升到 0.653。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Where the First-Digit Law Comes From: Uniform Mantissae, Scale Invariance and Multiplicative Chains — Chapter 1 explains the law through uniform mantissae, scale invariance and multiplicative chains; the expected share of a leading 1 is 0.30103. (https://5dlibrary.com/answers/en/benford-first-digit-ledger-fraud-chi-square-expense-audit/c01/) 第1章 首位数定律从哪里来:对数均匀、尺度不变与乘法链 — 第一章从尾数均匀、尺度不变与乘法链解释定律,首位数为 1 的期望是 0.30103。 (https://5dlibrary.com/answers/zh/benford-first-digit-ledger-fraud-chi-square-expense-audit/c01/) - Ch.2 The Counterintuitive Claim: Invented Numbers Are Too Clean, and the Law Wrongs Honest Fixed-Price Shops — Chapter 2 states the counterintuitive claim head on: a naive inventor's first-digit MAD is more than three times that of an honest ledger (0.0415 against 0.0133), honest fixed-price, capped and narrow-range ledgers are all flagged by chi-square at a rate of 1.000, and even small broad honest ledgers are misjudged by the MAD standard at a rate of 0.583. (https://5dlibrary.com/answers/en/benford-first-digit-ledger-fraud-chi-square-expense-audit/c02/) 第2章 反常识命题:编的数太「干净」,而定律会冤枉诚实的固定价店铺 — 第二章正面陈述反常识命题:朴素编造者的首位数 MAD 是诚实账本的三倍多(0.0415 对 0.0133),而固定价、有上限与窄区间的诚实账本被卡方判为偏离的比例都是 1.000,宽分布的诚实小账本被 MAD 标准误判的比例也有 0.583。 (https://5dlibrary.com/answers/zh/benford-first-digit-ledger-fraud-chi-square-expense-audit/c02/) - Ch.3 The Chi-Square Test: the Critical Value 15.507, the False Alarms of Nine z Tests and Certain Rejection in Large Samples — Chapter 3 shows that nine per-digit z tests give a thousand honest records a 0.3242 chance of being marked, and that data almost following the law are rejected by chi-square at ten thousand rows with probability 0.999. (https://5dlibrary.com/answers/en/benford-first-digit-ledger-fraud-chi-square-expense-audit/c03/) 第3章 卡方检验:临界值 15.507、九次 z 检验的虚警与大样本下的必然拒绝 — 第三章说明九次逐个数字的 z 检验让一千笔诚实记录有 0.3242 的机会被标红,几乎服从定律的数据在一万笔上被卡方拒绝的概率是 0.999。 (https://5dlibrary.com/answers/zh/benford-first-digit-ledger-fraud-chi-square-expense-audit/c03/) - Ch.4 The Mean Absolute Deviation: the Noise Floor, Nigrini's Bands and the First-Two-Digit Test — Chapter 4 gives the noise floor of the MAD, 0.2348 divided by the square root of the row count, and catches piling below an approval limit with the first-two-digit test at a detection rate of 0.794. (https://5dlibrary.com/answers/en/benford-first-digit-ledger-fraud-chi-square-expense-audit/c04/) 第4章 平均绝对偏差:噪声底线、尼格里尼区间与前两位数检验 — 第四章给出 MAD 的噪声底线 0.2348 除以笔数的平方根,并用前两位数检验抓住审批限额下的堆积,检出率 0.794。 (https://5dlibrary.com/answers/zh/benford-first-digit-ledger-fraud-chi-square-expense-audit/c04/) - Ch.5 The Precondition: How Many Decades, Fixed Prices and the Mixing of Price Points — Chapter 5 turns the precondition into a measurable gate: a worst-case MAD below 0.015 needs a span of 1.57 decades, and no single amount may take a tenth of the rows. (https://5dlibrary.com/answers/en/benford-first-digit-ledger-fraud-chi-square-expense-audit/c05/) 第5章 适用前提:跨越几个数量级、固定定价与价格点的混合 — 第五章把适用前提写成可测量的闸门:最坏情况 MAD 低于 0.015 需要跨越 1.57 个数量级,单一金额不能占到一成。 (https://5dlibrary.com/answers/zh/benford-first-digit-ledger-fraud-chi-square-expense-audit/c05/) - Ch.6 Beyond the First Digit: Last Digits, Repeated Amounts and Piling Up Below an Approval Limit — Chapter 6 covers last digits, repeats and the limit window: honest rounding lifts the share of 0 and 5 endings to 0.5616 while an inventor presses it to 0.0911, and an honest ledger of 100 rows has only a 0.0025 chance of no repeat. (https://5dlibrary.com/answers/en/benford-first-digit-ledger-fraud-chi-square-expense-audit/c06/) 第6章 首位数之外:末位数、重复金额与审批限额下的堆积 — 第六章讲末位、重复与限额窗口:诚实凑整把 0、5 结尾推到 0.5616,编造者压到 0.0911;100 笔诚实账一笔不重复的概率只有 0.0025。 (https://5dlibrary.com/answers/zh/benford-first-digit-ledger-fraud-chi-square-expense-audit/c06/) - Ch.7 Chinese and European Sources Weighed: the Four-Pillar Settlement, Newcomb's Logarithm Tables and Pacioli's Trial Balance — Chapter 7 weighs the four-pillar settlement, Newcomb's note of 1881 and Pacioli's double entry of 1494. (https://5dlibrary.com/answers/en/benford-first-digit-ledger-fraud-chi-square-expense-audit/c07/) 第7章 东西方原典考辨:四柱结算法、纽科姆的对数表与帕乔利的试算平衡 — 第七章考辨四柱结算法、纽科姆 1881 年的短文与帕乔利 1494 年的复式记账。 (https://5dlibrary.com/answers/zh/benford-first-digit-ledger-fraud-chi-square-expense-audit/c07/) - Ch.8 Closing the Loop: Self-Audit of Buying-Agent Flows and Expense Claims, Red-Light Review and a Pre-Registered Plan — Chapter 8 gives a monthly self-audit, a three-light review, the distilled tool (the Ledger First-Digit Auditor) and a pre-registration plan: the gate cuts honest shops' false-alarm rate from 0.4280 to 0.0280 and, with invention at five per cent, lifts the positive predictive value from 0.110 to 0.653. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/benford-first-digit-ledger-fraud-chi-square-expense-audit/c08/) 第8章 实证闭环:代购流水与报销自检、红灯复核与预注册方案 — 第八章给出月度自检流程、三色灯复核、蒸馏小工具「流水首位数审计器」与预注册方案:闸门把诚实店铺的误报率从 0.4280 降到 0.0280,编造占百分之五时阳性预测值从 0.110 升到 0.653。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/benford-first-digit-ledger-fraud-chi-square-expense-audit/c08/) ## 5DL-B-217 · Peeking-Inflated False Positives in A/B Tests, Sequential Probability Ratio Tests and Minimum Samples for Low-Traffic Sites 中文书名:A/B测试偷看早停假阳性膨胀、序贯概率比检验与小流量站点最小样本母典 DOI: pending Cite: Fan, Y. (2026). Peeking-Inflated False Positives in A/B Tests, Sequential Probability Ratio Tests and Minimum Samples for Low-Traffic Sites (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its sites, traffic, conversion rates and lifts are illustrative parameters; all calculations are generated by its programs; the book is not individualized business, legal or financial advice. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: What does it cost when an owner opens the dashboard every day and stops a test, launching the new page to everyone, as soon as it reads "significant"? How large an improvement can a site with two hundred visitors a day actually see with an A/B test? This book writes both questions as calculations that rely only on the Python standard library and can be reproduced character for character, and states which data would overturn its conclusions. Its sites, traffic, conversion rates and lifts are illustrative parameters; all calculations are generated by its programs; the book is not individualized business, legal or financial advice. Chapter 1 computes exactly, by grid recursion, that looking once a day and stopping when significant gives a false-positive rate of 0.2197 after two weeks and 0.2802 after a month, correcting the founding claim of "a fifth to nearly a third". Chapter 2 shows that at a 3% base a 5% relative lift needs 207,938 visitors per arm, 5.7 years at 200 visitors a day, and that a one-month test has power of 0.0609. Chapter 3 explains the inflation with random walks and the law of the iterated logarithm and finds that a peeking small site reports on average 18.3 times the true lift. Chapter 4 derives Wald's sequential probability ratio test, which at the specified effect needs on average 0.64 of the fixed sample. Chapter 5 gives group-sequential bars: with five looks, 2.4133 for Pocock and a final 2.0402 for O'Brien-Fleming. Chapter 6 shows that a mixture sequential test watched daily for a year keeps its false-alarm rate at 0.0280; that a Bayesian "chance to beat control" under a default prior is true only 0.2158 of the time in a world where most changes do nothing; and that pre-experiment data would need a correlation of 0.9932 for a small site to see a 5% lift in a month. Chapter 7 examines the standard and the three tests of the Mozi chapter "Against Fate, Part One" and Fisher's lady tasting tea (all four cups right by pure guessing: 0.0143); written as a Bayes formula, the three tests show that only 0.133 of the changes a peeking small site launches are real. Chapter 8 offers a stopping-rule dashboard for thirty sites, the use of the A/B Stopping-Rule Calculator and a pre-registered plan for refuting the book. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:店主每天打开后台、看到「显著」就停止测试并全量上线,这个习惯有多大代价?日均两百访客的小站,用 A/B 测试究竟能看见多大的改进?本书把这两个问题写成只依赖 Python 标准库、可以逐字复现的计算,并给出什么样的数据会推翻本书的结论。书中的站点、流量、转化率与提升幅度均为示意参数;全部计算由书中程序生成;本书不是个人化的经营、法律或理财建议。 第一章用格点递推精确计算:每天看一次、显著即停,两周后的假阳性率是 0.2197,一个月后是 0.2802,并修正了立项时「两到三成」的说法。第二章说明基线 3%、检出 5% 的相对提升每组需要 207,938 个访客,日均 200 访客要 5.7 年,一个月测试的功效只有 0.0609。第三章用随机游走与重对数律解释膨胀的来源,并算出偷看的小站报告的提升平均是真实值的 18.3 倍。第四章推导 Wald 的序贯概率比检验,它在指定效应上平均只需固定样本的 0.64。第五章给出成组序贯门槛,五次查看时 Pocock 为 2.4133、O'Brien-Fleming 最后一次为 2.0402。 第六章说明混合序贯检验每天看一年误报率只有 0.0280;默认先验下的贝叶斯「胜出概率」在多数改动无效的世界里只有 0.2158 为真;事前数据要让小站一个月看见 5% 的提升,需要 0.9932 的相关系数。第七章考辨《墨子·非命上》的立仪与三表,以及 Fisher 的女士品茶(纯猜四杯全对的概率 0.0143);三表写成贝叶斯公式后,偷看的小站上线的改动只有 0.133 是真的。第八章给出三十站停止规则仪表、A/B 停止规则器的用法与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Price of Peeking: an Exact Account of "Look Daily, Stop When Significant" — Chapter 1 computes exactly, by grid recursion, that looking once a day and stopping when significant gives a false-positive rate of 0.2197 after two weeks and 0.2802 after a month, correcting the founding claim of "a fifth to nearly a third". (https://5dlibrary.com/answers/en/ab-test-peeking-false-positives-sprt-low-traffic-samples/c01/) 第1章 偷看的价格:每天一看、显著即停的假阳性精确账 — 第一章用格点递推精确计算:每天看一次、显著即停,两周后的假阳性率是 0.2197,一个月后是 0.2802,并修正了立项时「两到三成」的说法。 (https://5dlibrary.com/answers/zh/ab-test-peeking-false-positives-sprt-low-traffic-samples/c01/) - Ch.2 The Ceiling of a Small Site: Power, Minimum Detectable Effects and a Five-and-a-Half-Year Wait — Chapter 2 shows that at a 3% base a 5% relative lift needs 207,938 visitors per arm, 5.7 years at 200 visitors a day, and that a one-month test has power of 0.0609. (https://5dlibrary.com/answers/en/ab-test-peeking-false-positives-sprt-low-traffic-samples/c02/) 第2章 小站的天花板:功效、最小可检测效应与五年半的等待 — 第二章说明基线 3%、检出 5% 的相对提升每组需要 207,938 个访客,日均 200 访客要 5.7 年,一个月测试的功效只有 0.0609。 (https://5dlibrary.com/answers/zh/ab-test-peeking-false-positives-sprt-low-traffic-samples/c02/) - Ch.3 Where the Inflation Comes From: Random Walks, the Iterated Logarithm and Exaggerated Winners — Chapter 3 explains the inflation with random walks and the law of the iterated logarithm and finds that a peeking small site reports on average 18.3 times the true lift. (https://5dlibrary.com/answers/en/ab-test-peeking-false-positives-sprt-low-traffic-samples/c03/) 第3章 膨胀从哪里来:随机游走、重对数律与被夸大的赢家 — 第三章用随机游走与重对数律解释膨胀的来源,并算出偷看的小站报告的提升平均是真实值的 18.3 倍。 (https://5dlibrary.com/answers/zh/ab-test-peeking-false-positives-sprt-low-traffic-samples/c03/) - Ch.4 Wald's Sequential Probability Ratio Test: a Look at Every Single Conversion — Chapter 4 derives Wald's sequential probability ratio test, which at the specified effect needs on average 0.64 of the fixed sample. (https://5dlibrary.com/answers/en/ab-test-peeking-false-positives-sprt-low-traffic-samples/c04/) 第4章 Wald 的序贯概率比检验:允许每一笔转化都看一眼 — 第四章推导 Wald 的序贯概率比检验,它在指定效应上平均只需固定样本的 0.64。 (https://5dlibrary.com/answers/zh/ab-test-peeking-false-positives-sprt-low-traffic-samples/c04/) - Ch.5 Looking Once a Week: Group-Sequential Designs and Alpha-Spending Functions — Chapter 5 gives group-sequential bars: with five looks, 2.4133 for Pocock and a final 2.0402 for O'Brien-Fleming. (https://5dlibrary.com/answers/en/ab-test-peeking-false-positives-sprt-low-traffic-samples/c05/) 第5章 每周看一次:成组序贯设计与 α 消耗函数 — 第五章给出成组序贯门槛,五次查看时 Pocock 为 2.4133、O'Brien-Fleming 最后一次为 2.0402。 (https://5dlibrary.com/answers/zh/ab-test-peeking-false-positives-sprt-low-traffic-samples/c05/) - Ch.6 Always-Valid Inference, the Bayesian "Chance to Beat Control" and the Limits of Pre-Experiment Data — Chapter 6 shows that a mixture sequential test watched daily for a year keeps its false-alarm rate at 0.0280; that a Bayesian "chance to beat control" under a default prior is true only 0.2158 of the time in a world where most changes do nothing; and that pre-experiment data would need a correlation of 0.9932 for a small site to see a 5% lift in a month. (https://5dlibrary.com/answers/en/ab-test-peeking-false-positives-sprt-low-traffic-samples/c06/) 第6章 随时可看的推断、贝叶斯「胜出概率」与事前数据的极限 — 第六章说明混合序贯检验每天看一年误报率只有 0.0280;默认先验下的贝叶斯「胜出概率」在多数改动无效的世界里只有 0.2158 为真;事前数据要让小站一个月看见 5% 的提升,需要 0.9932 的相关系数。 (https://5dlibrary.com/answers/zh/ab-test-peeking-false-positives-sprt-low-traffic-samples/c06/) - Ch.7 Eastern and Western Sources Examined: Mozi's Three Tests and Fisher's Cup of Tea — Chapter 7 examines the standard and the three tests of the Mozi chapter "Against Fate, Part One" and Fisher's lady tasting tea (all four cups right by pure guessing: 0.0143); written as a Bayes formula, the three tests show that only 0.133 of the changes a peeking small site launches are real. (https://5dlibrary.com/answers/en/ab-test-peeking-false-positives-sprt-low-traffic-samples/c07/) 第7章 东西方原典考辨:墨子的三表与 Fisher 的一杯茶 — 第七章考辨《墨子·非命上》的立仪与三表,以及 Fisher 的女士品茶(纯猜四杯全对的概率 0.0143);三表写成贝叶斯公式后,偷看的小站上线的改动只有 0.133 是真的。 (https://5dlibrary.com/answers/zh/ab-test-peeking-false-positives-sprt-low-traffic-samples/c07/) - Ch.8 Closing the Loop: a Stopping-Rule Dashboard for Thirty Sites, a Minimum-Sample Calculator and a Plan to Refute This Book — Chapter 8 offers a stopping-rule dashboard for thirty sites, the use of the A/B Stopping-Rule Calculator and a pre-registered plan for refuting the book. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/ab-test-peeking-false-positives-sprt-low-traffic-samples/c08/) 第8章 实证闭环:三十站停止规则仪表、最小样本计算器与证伪方案 — 第八章给出三十站停止规则仪表、A/B 停止规则器的用法与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/ab-test-peeking-false-positives-sprt-low-traffic-samples/c08/) ## 5DL-B-218 · Password Entropy and Credential Stuffing, the Forced-Rotation Paradox and Least-Privilege Accounts for One-Person Firms 中文书名:口令熵与撞库攻击、强制改密反安全悖论与一人公司账号最小权限母典 DOI: pending Cite: Fan, Y. (2026). Password Entropy and Credential Stuffing, the Forced-Rotation Paradox and Least-Privilege Accounts for One-Person Firms (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its firms, accounts and probabilities are illustrative, and all calculations are generated by its programs; it is not personalised business, legal, financial or compliance advice, and its audit routine and tool never touch a real password. Chapter 6 uses minimum role covers and the recovery graph to show that the illustrative firm expects to lose 0.4972 accounts a year, and 0.1899 after hardening the main mailbox alone. Abstract: This book tests the sentence most common in small firms' security policies: passwords must change every 90 days and must contain four classes of character. It splits the sentence into two claims that data can overturn: forcing changes by the calendar, with no sign of compromise, brings no measurable security benefit and produces predictable variants; and randomly generated long passphrases beat complex short passwords that people assemble themselves. Its firms, accounts and probabilities are illustrative, and all calculations are generated by its programs; it is not personalised business, legal, financial or compliance advice, and its audit routine and tool never touch a real password. Chapter 1 shows that an attacker who needs one day of access still succeeds with probability 0.991 under a 90-day forced change, corrects the blueprint's "no longer recommends" to the "shall not" of Revision 4, and sets out the counter-evidence of a study of self-reported behaviour. Chapter 2 computes entropy: eight random printable characters carry 52.44 bits, four random words 51.70 and five random words 64.62, and human-chosen phrases cannot be scored by the formula. Chapter 3 shows in a Zipf population that a ten-guess success rate of 0.0477 is 817 times that of a uniform distribution with the same entropy; if refused users switch to predictable mutations, the hundred-guess rate under a hundred-entry blocklist, 0.1336, exceeds the 0.1087 with no blocklist. Chapter 4 computes the stuffing funnel: ten accounts sharing one password expose 4.37 times as many accounts as unique passwords, and going from half to full second-factor coverage cuts residual risk 28.3 times. Chapter 5 computes the ceilings and leaks of throttling, blocklists and hash-prefix queries. Chapter 6 uses minimum role covers and the recovery graph to show that the illustrative firm expects to lose 0.4972 accounts a year, and 0.1899 after hardening the main mailbox alone. Chapter 7 examines the secret tallies and split letters of the Six Secret Teachings and Caesar's shift as recorded by Suetonius. Chapter 8 gives the permission-matrix audit sheet and the tool of the same name: the illustrative firm's weighted residual risk index falls from 11.640 to 1.017; it also gives a pre-registered testing plan and the results that would overturn the claims. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:本书检验小公司安全制度里最常见的一句话:口令每 90 天必须更换,并且必须包含四类字符。全书把它拆成两个可以被数据推翻的命题:没有泄露迹象时按日历强制改密不带来可测的安全收益,还会逼出可预测的变体;随机生成的长口令短语胜过人自己凑出的复杂短密码。书中的公司、账户与概率参数均为示意,全部计算由书中程序生成;本书不是个性化的经营、法律、财务或合规建议,审计流程与小工具不接触任何真实口令。 第一章说明,对只需要一天访问的攻击者,每 90 天强制改密后其成功概率仍有 0.991,并依据原文把蓝图里的「已不建议」修正为修订第 4 版的「不得」,同时并陈自述行为研究的反方证据。第二章计算熵:8 位全字符随机口令 52.44 比特,四个随机词 51.70 比特,五个随机词 64.62 比特,并说明人选短语不能按熵公式计算。第三章在 Zipf 人群中说明,猜 10 次的成功率 0.0477 是同熵均匀分布的 817 倍;被拒用户若改用可预测变形,百条黑名单下猜 100 次的成功率 0.1336 反而高于不设黑名单的 0.1087。第四章计算撞库漏斗,十个账户共用一个口令时期望暴露数是独立口令的 4.37 倍,第二因素从半覆盖到全覆盖使剩余风险下降 28.3 倍。 第五章计算限速、黑名单与哈希前缀查询的天花板与泄露。第六章用最小角色覆盖与找回关系图说明,示意公司期望每年失守 0.4972 个账户,只加固主邮箱就降到 0.1899。第七章考辨《六韬》的阴符、阴书与苏维托尼乌斯记下的凯撒移位。第八章给出权限矩阵审计表与同名小工具:示意公司的加权剩余风险指数从 11.640 降到 1.017,并给出预注册检验方案与推翻条件。书中以标注框区分已发表证据、可运行的 Python 计算(19 段程序)、原典与作者假说。 Chapter answers: - Ch.1 The Forced-Rotation Paradox: Predictable Variants and an Overrated Attack Window — Chapter 1 shows that an attacker who needs one day of access still succeeds with probability 0.991 under a 90-day forced change, corrects the blueprint's "no longer recommends" to the "shall not" of Revision 4, and sets out the counter-evidence of a study of self-reported behaviour. (https://5dlibrary.com/answers/en/password-entropy-credential-stuffing-forced-rotation-least-privilege/c01/) 第1章 强制改密悖论:可预测的变体与被高估的攻击窗口 — 第一章说明,对只需要一天访问的攻击者,每 90 天强制改密后其成功概率仍有 0.991,并依据原文把蓝图里的「已不建议」修正为修订第 4 版的「不得」,同时并陈自述行为研究的反方证据。 (https://5dlibrary.com/answers/zh/password-entropy-credential-stuffing-forced-rotation-least-privilege/c01/) - Ch.2 The Entropy Ledger: log2(N^L), Random Long Passphrases and Complex Short Passwords — Chapter 2 computes entropy: eight random printable characters carry 52.44 bits, four random words 51.70 and five random words 64.62, and human-chosen phrases cannot be scored by the formula. (https://5dlibrary.com/answers/en/password-entropy-credential-stuffing-forced-rotation-least-privilege/c02/) 第2章 熵的账本:log2(N^L)、随机长短语与复杂短密码 — 第二章计算熵:8 位全字符随机口令 52.44 比特,四个随机词 51.70 比特,五个随机词 64.62 比特,并说明人选短语不能按熵公式计算。 (https://5dlibrary.com/answers/zh/password-entropy-credential-stuffing-forced-rotation-least-privilege/c02/) - Ch.3 Guessing under a Zipf Law: Why Shannon Entropy Overrates Human-Chosen Passwords — Chapter 3 shows in a Zipf population that a ten-guess success rate of 0.0477 is 817 times that of a uniform distribution with the same entropy; if refused users switch to predictable mutations, the hundred-guess rate under a hundred-entry blocklist, 0.1336, exceeds the 0.1087 with no blocklist. (https://5dlibrary.com/answers/en/password-entropy-credential-stuffing-forced-rotation-least-privilege/c03/) 第3章 Zipf 分布下的猜测:香农熵为什么高估人选口令 — 第三章在 Zipf 人群中说明,猜 10 次的成功率 0.0477 是同熵均匀分布的 817 倍;被拒用户若改用可预测变形,百条黑名单下猜 100 次的成功率 0.1336 反而高于不设黑名单的 0.1087。 (https://5dlibrary.com/answers/zh/password-entropy-credential-stuffing-forced-rotation-least-privilege/c03/) - Ch.4 The Credential-Stuffing Funnel: Reuse Rates, Breach Corpora and Second Factors — Chapter 4 computes the stuffing funnel: ten accounts sharing one password expose 4.37 times as many accounts as unique passwords, and going from half to full second-factor coverage cuts residual risk 28.3 times. (https://5dlibrary.com/answers/en/password-entropy-credential-stuffing-forced-rotation-least-privilege/c04/) 第4章 撞库漏斗:复用率、泄露库与第二因素 — 第四章计算撞库漏斗,十个账户共用一个口令时期望暴露数是独立口令的 4.37 倍,第二因素从半覆盖到全覆盖使剩余风险下降 28.3 倍。 (https://5dlibrary.com/answers/zh/password-entropy-credential-stuffing-forced-rotation-least-privilege/c04/) - Ch.5 Throttling, Blocklists and Hash Prefixes: the Ceiling on Online Guessing — Without the prefix, a single guess succeeds with the top candidate's probability, 0.0314. Knowing a one-character prefix raises it to 0.1240, 3.9 times; two characters give 0.3041, 9.7 times; three give 0.6214, 19.8 times; four give 0.9463, 30.1 times. The column of mean bucket sizes explains why. With 20,000 candidates in 4,096 buckets, each bucket holds 4.9 on average; with 65,536 buckets, 0.3 on average, so a bucket nearly always holds a single candidate and the prefix is the answer. (https://5dlibrary.com/answers/en/password-entropy-credential-stuffing-forced-rotation-least-privilege/c05/) 第5章 限速、黑名单与哈希前缀:在线猜测的天花板 — 不知道前缀时,攻击者只猜一次的成功率是第一名候选的概率 0.0314。知道 1 个十六进制字符的前缀后,成功率升到 0.1240,是原来的 3.9 倍;2 个字符时 0.3041,9.7 倍;3 个字符时 0.6214,19.8 倍;4 个字符时 0.9463,30.1 倍。原因在「平均桶大小」那一列:20,000 个候选分到 4,096 个桶里,每个桶平均只有 4.9 个;分到 65,536 个桶里,平均每桶 0.3 个,桶里几乎只剩一个候选,前缀就等于答案。 (https://5dlibrary.com/answers/zh/password-entropy-credential-stuffing-forced-rotation-least-privilege/c05/) - Ch.6 The Permission Graph and its Minimum Cover: a One-Person Firm's Blast Radius — Chapter 6 uses minimum role covers and the recovery graph to show that the illustrative firm expects to lose 0.4972 accounts a year, and 0.1899 after hardening the main mailbox alone. (https://5dlibrary.com/answers/en/password-entropy-credential-stuffing-forced-rotation-least-privilege/c06/) 第6章 权限图与最小覆盖:一人公司的爆炸半径 — 第六章用最小角色覆盖与找回关系图说明,示意公司期望每年失守 0.4972 个账户,只加固主邮箱就降到 0.1899。 (https://5dlibrary.com/answers/zh/password-entropy-credential-stuffing-forced-rotation-least-privilege/c06/) - Ch.8 The Permission-Matrix Audit Sheet: an Empirical Loop and a Falsification Plan — Chapter 8 gives the permission-matrix audit sheet and the tool of the same name: the illustrative firm's weighted residual risk index falls from 11.640 to 1.017; it also gives a pre-registered testing plan and the results that would overturn the claims. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/password-entropy-credential-stuffing-forced-rotation-least-privilege/c08/) 第8章 权限矩阵审计表:实证闭环与证伪方案 — 第八章给出权限矩阵审计表与同名小工具:示意公司的加权剩余风险指数从 11.640 降到 1.017,并给出预注册检验方案与推翻条件。书中以标注框区分已发表证据、可运行的 Python 计算(19 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/password-entropy-credential-stuffing-forced-rotation-least-privilege/c08/) ## 5DL-B-219 · Price-to-Rent Ratios and Holding-Cost NPV, Non-Ergodic Leveraged Home Buying and the Redemption Option in Chinese Dian Tenure 中文书名:租售比与持有成本净现值、杠杆购房非遍历性与典权回赎期权母典 DOI: pending Cite: Fan, Y. (2026). Price-to-Rent Ratios and Holding-Cost NPV, Non-Ergodic Leveraged Home Buying and the Redemption Option in Chinese Dian Tenure (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 1 uses user cost to show that in an illustrative city with a rent yield of 0.016 the rent does not even cover the landlord's interest and taxes, and that the verdict is most sensitive to the opportunity cost of funds. Chapter 8 presents the distilled Rent-or-Buy Comparison Calculator, under whose Kuala-Lumpur-like illustrative parameters the renter leads with probability of only 0.2116, together with a pre-registered test and the conditions that would overturn the book. Abstract: "Renting pays your landlord's mortgage; buying is saving for yourself" is close to the first rule of housing in Chinese-speaking societies. This book writes that rule as runnable arithmetic and tests when it fails. Its core model is a pair of ledgers on one budget: the buyer pays the down payment, the mortgage and holding costs, while the renter puts the same money, and every yearly saving, into an index fund, and both are liquidated at the end. Its parameters are illustrative; all calculations are generated by its programs; the book is general analysis, not personalised legal or financial advice, and where it touches historical legal provisions it gives only their general sense, subject to human review before publication. Chapter 1 uses user cost to show that in an illustrative city with a rent yield of 0.016 the rent does not even cover the landlord's interest and taxes, and that the verdict is most sensitive to the opportunity cost of funds. The ledger of Chapter 2 finds that house prices must rise by 0.0359 a year for buying to break even, and sets an American study of rent-or-buy decisions beside long-run evidence on asset returns that points the other way. Chapter 3 runs a fixed-seed Monte Carlo in which renting and investing finishes ahead on 0.7035 of the paths, with more dispersed outcomes and half of its edge owed to investment discipline. Chapter 4 contrasts ensemble and time averages: the average household with an eighty per cent loan grows its wealth by 0.0601 a year while the typical household's time-average growth rate is -0.0174, and the Kelly formula gives the growth-optimal loan-to-value ratio. Chapter 5 treats floating rates and the land share. Chapter 6 writes dian tenure as a redemption call held by the debtor and uses a binomial tree to show that a dian without a lock-up never pays for the holder, while with a three-year lock-up the fair dian price is about 0.8216 of the property's value. Chapter 7 examines the Song Penal Code's provisions on dian sales, Qing dian deeds and Ricardo's theory of rent. Chapter 8 presents the distilled Rent-or-Buy Comparison Calculator, under whose Kuala-Lumpur-like illustrative parameters the renter leads with probability of only 0.2116, together with a pre-registered test and the conditions that would overturn the book. Marked boxes separate published evidence, runnable Python calculations (20 programs), summaries of classical sources and the author's hypotheses. 摘要:「租房是在帮房东还贷,买房才是给自己存钱」是华人社会关于住房的第一常识。本书把它写成可以运行的算术,并检验它在什么条件下站不住。核心模型是同预算的两本账:买房户付首付、按揭与持有成本,租房户把同一笔钱与每年省下的差额投入指数基金,期末全部变现比较。书中参数为示意;全部计算由书中程序生成;本书是非个人化的分析,不构成法律或理财建议,涉及古代法条之处只述大意,以上架前人工复核为准。 第一章用用户成本说明,在租金回报率为 0.016 的示意城市里,房租连房东的利息与税费都付不清,结论最敏感的参数是资金的机会成本。第二章的账本求出买租打平所需的房价年涨幅为 0.0359,并并陈美国租买比较研究与长期资产回报研究的不同结论。第三章用固定种子的蒙特卡洛显示租房定投在 0.7035 的路径中期末领先,但其结局更分散,且一半优势来自投资纪律。第四章对比集合平均与时间平均:八成贷款的平均家庭财富年增 0.0601,典型家庭的时间平均增长率却是 -0.0174,并用凯利公式求出增长最优的贷款成数。第五章讨论浮动利率与土地份额。 第六章把典权写成出典人持有的回赎看涨期权,用二叉树说明没有锁定期的典对典权人永远不划算,三年锁定期下公平典价约为房价的 0.8216。第七章考辨《宋刑统》典卖规定、清代典契与李嘉图地租论。第八章给出蒸馏小工具「租买对比计算器」,在吉隆坡式示意参数下租房领先的概率只有 0.2116,并写下预注册检验方案与推翻本书的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典摘述与作者的假说。 Chapter answers: - Ch.1 Rent Does Not Pay the Landlord's Mortgage: Holding Costs, User Cost and the Price-to-Rent Ratio — Chapter 1 uses user cost to show that in an illustrative city with a rent yield of 0.016 the rent does not even cover the landlord's interest and taxes, and that the verdict is most sensitive to the opportunity cost of funds. The ledger of Chapter 2 finds that house prices must rise by 0.0359 a year for buying to break even, and sets an American study of rent-or-buy decisions beside long-run evidence on asset returns that points the other way. (https://5dlibrary.com/answers/en/price-to-rent-npv-leveraged-housing-dian-redemption-option/c01/) 第1章 租金并不替房东还贷:持有成本、用户成本与租售比 — 第一章用用户成本说明,在租金回报率为 0.016 的示意城市里,房租连房东的利息与税费都付不清,结论最敏感的参数是资金的机会成本。 (https://5dlibrary.com/answers/zh/price-to-rent-npv-leveraged-housing-dian-redemption-option/c01/) - Ch.2 Two Ledgers on One Budget: NPV and Break-Even Growth for Renting-and-Investing against Mortgaged Buying — Extending the one-year comparison into a multi-year ledger on equal budgets, the renter-investor leads by 58.07 after 20 years under the illustrative parameters, and the lead widens with time. The break-even growth $g^*$ is a more useful indicator than the price-to-rent ratio because it states plainly the bet that a purchase implies. The sensitivity analysis shows that the outcome is driven mainly by price growth, the fund return and the rent yield, with interest rates and transaction costs behind them. The deterministic ledger still assumes that every year's return equals its average; the next chapter lets the house and the fund move along random paths and asks how large the "large share of scenarios" really is. (https://5dlibrary.com/answers/en/price-to-rent-npv-leveraged-housing-dian-redemption-option/c02/) 第2章 同一份预算的两本账:租房定投与贷款买房的净现值与打平涨幅 — 把一年的比较扩展成同预算的多年账本以后,在示意参数下租房定投户在第 20 年领先 58.07,且领先随时间扩大。打平所需的房价涨幅 $g^*$ 是一个比租售比更有用的指标,它把买房的隐含押注说得一清二楚。敏感度分析显示,结局主要由房价涨幅、基金回报与租金回报率决定,利率与交易费用居次。确定性账本仍然假设每年的回报都等于平均值,下一章让房价与基金都按随机路径运动,看看「大量情景」究竟有多大。 (https://5dlibrary.com/answers/zh/price-to-rent-npv-leveraged-housing-dian-redemption-option/c02/) - Ch.3 How Large Is "a Large Share of Scenarios"? A Monte Carlo of Linked House and Fund Paths — Chapter 3 runs a fixed-seed Monte Carlo in which renting and investing finishes ahead on 0.7035 of the paths, with more dispersed outcomes and half of its edge owed to investment discipline. (https://5dlibrary.com/answers/en/price-to-rent-npv-leveraged-housing-dian-redemption-option/c03/) 第3章 「大量情景」有多大:房价与基金联动路径的蒙特卡洛 — 第三章用固定种子的蒙特卡洛显示租房定投在 0.7035 的路径中期末领先,但其结局更分散,且一半优势来自投资纪律。 (https://5dlibrary.com/answers/zh/price-to-rent-npv-leveraged-housing-dian-redemption-option/c03/) - Ch.4 The Non-Ergodicity of Leveraged Home Buying: Ensemble Averages, Time Averages and an Absorbing Barrier — Chapter 4 contrasts ensemble and time averages: the average household with an eighty per cent loan grows its wealth by 0.0601 a year while the typical household's time-average growth rate is -0.0174, and the Kelly formula gives the growth-optimal loan-to-value ratio. (https://5dlibrary.com/answers/en/price-to-rent-npv-leveraged-housing-dian-redemption-option/c04/) 第4章 杠杆购房的非遍历性:集合平均、时间平均与吸收壁 — 第四章对比集合平均与时间平均:八成贷款的平均家庭财富年增 0.0601,典型家庭的时间平均增长率却是 -0.0174,并用凯利公式求出增长最优的贷款成数。 (https://5dlibrary.com/answers/zh/price-to-rent-npv-leveraged-housing-dian-redemption-option/c04/) - Ch.5 The Fine Print of Holding Costs: Floating Rates, Building Depreciation and the Land Share — Once holding costs are itemised, the buyer's accounts look a little worse. Under a floating rate, the average interest over 20 years in the illustrative set rises from 26.05 to 27.18, and on 0.2493 of the paths the annual payment at some point exceeds 1.10 times the first one. Buildings depreciate and only land appreciates; the lower the land share, the less price growth a given rise in land values buys, and the faster land must rise to reach the break-even growth of p22. The next chapter leaves the modern mortgage for an ancient property arrangement, dian tenure, whose economic structure happens to be an option that can be priced on a binomial tree. (https://5dlibrary.com/answers/en/price-to-rent-npv-leveraged-housing-dian-redemption-option/c05/) 第5章 持有成本的细账:浮动利率、建物折旧与土地份额 — 把持有成本拆细以后,买房的账更难看了一些。浮动利率下,示意参数中 20 年的平均利息从 26.05 升到 27.18,且有 0.2493 的路径年供曾超过首年的 1.10 倍。建筑会折旧,只有土地能增值;土地份额越低,同样的地价上涨换来的房价上涨越少,要达到 p22 的打平涨幅所需的地价涨幅也越高。下一章离开现代的按揭,转向一种古老的产权安排:典权。它的经济结构,恰好是一份可以用二叉树定价的期权。 (https://5dlibrary.com/answers/zh/price-to-rent-npv-leveraged-housing-dian-redemption-option/c05/) - Ch.6 Dian Tenure as a Redemption Call: Binomial Pricing and the Fair Dian Price — Chapter 6 writes dian tenure as a redemption call held by the debtor and uses a binomial tree to show that a dian without a lock-up never pays for the holder, while with a three-year lock-up the fair dian price is about 0.8216 of the property's value. (https://5dlibrary.com/answers/en/price-to-rent-npv-leveraged-housing-dian-redemption-option/c06/) 第6章 典权是一份回赎看涨期权:二叉树定价与公平典价 — 第六章把典权写成出典人持有的回赎看涨期权,用二叉树说明没有锁定期的典对典权人永远不划算,三年锁定期下公平典价约为房价的 0.8216。 (https://5dlibrary.com/answers/zh/price-to-rent-npv-leveraged-housing-dian-redemption-option/c06/) - Ch.7 Sources East and West: the Song Penal Code on Dian Sales, Qing Dian Deeds and Ricardo's Theory of Rent — Program p71 turns the argument into a table. Five grades of land, each receiving the same outlay of 60.0, produce 100, 90, 80, 70 and 60 units of corn. With the margin of cultivation at grade B, corn costs 0.6667, the money rent of grade A is 6.667, and rent is 0.0526 of the value of output. When the margin extends to grade C, corn rises to 0.7500, grade A's rent to 15.000 and the rent share to 0.1111; at grade D corn costs 0.8571 and the rent share is 0.1765; at grade E corn costs 1.0000, grade A pays a rent of 40.000, all land together pays 100.000, and the rent share is 0.2500. (https://5dlibrary.com/answers/en/price-to-rent-npv-leveraged-housing-dian-redemption-option/c07/) 第7章 东西方原典考辨:《宋刑统》典卖规定、清代典契与李嘉图地租论 — 程序 p71 把这段论证写成一张表。五个等级的土地,同样投入 60.0 的成本,分别产出 100、90、80、70、60 单位谷物。耕作的边际在 B 级时,谷价是 0.6667,A 级的货币地租 6.667,地租占总产出价值的 0.0526。边际扩展到 C 级,谷价升到 0.7500,A 级地租 15.000,地租份额 0.1111;扩展到 D 级时谷价 0.8571,地租份额 0.1765;扩展到 E 级时谷价 1.0000,A 级地租 40.000,全部土地地租合计 100.000,地租份额 0.2500。 (https://5dlibrary.com/answers/zh/price-to-rent-npv-leveraged-housing-dian-redemption-option/c07/) - Ch.8 Closing the Loop: a Rent-or-Buy Calculator for a Taipei-like and a Kuala-Lumpur-like City, a Pre-Registered Test and Conditions for Refutation — Chapter 8 presents the distilled Rent-or-Buy Comparison Calculator, under whose Kuala-Lumpur-like illustrative parameters the renter leads with probability of only 0.2116, together with a pre-registered test and the conditions that would overturn the book. Marked boxes separate published evidence, runnable Python calculations (20 programs), summaries of classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/price-to-rent-npv-leveraged-housing-dian-redemption-option/c08/) 第8章 实证闭环:台北式与吉隆坡式租买计算器、预注册方案与证伪条件 — 第八章给出蒸馏小工具「租买对比计算器」,在吉隆坡式示意参数下租房领先的概率只有 0.2116,并写下预注册检验方案与推翻本书的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典摘述与作者的假说。 (https://5dlibrary.com/answers/zh/price-to-rent-npv-leveraged-housing-dian-redemption-option/c08/) ## 5DL-B-220 · Decomposing the IRR of Savings Insurance: Halley's Life Table and the Limits of Buying Term and Investing the Difference 中文书名:储蓄型保单内部收益率拆解、哈雷生命表精算与买定期投差额边界母典 DOI: pending Cite: Fan, Y. (2026). Decomposing the IRR of Savings Insurance: Halley's Life Table and the Limits of Buying Term and Investing the Difference (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 1 shows that the illustrative policy held for thirty years returns only 0.02462, its cash value just 0.9906 of a bond ladder fed with the same premiums. Chapter 4 follows ten thousand illustrative policies: 0.3285 of holders surrender at a loss, the pooled return falls to 0.02129, and the cost of surrender and policy loans under a sudden need for cash is computed. Abstract: Two opposite sayings about savings insurance are widely repeated: that it is steady, guaranteed and doubles over the long run, and that one should buy only term cover and invest the difference, winning for certain in the long run. This book puts both into one runnable model and shows that each states a comparison that depends on parameters as if it were settled. Its core tools are the internal rate of return of a cash-value table (roots by Newton's method and bisection, every present value summed exactly), annuity values from Halley's Breslau life table of 1693, and sensitivity analysis of the behaviour gap. Its policies, premiums, cash values, charges and mortality rates are illustrative and name or evaluate no insurer or product; all calculations are generated by its programs; it is not personalised financial advice and not legal or tax advice, and any statute or tax question is subject to human review before publication. Chapter 1 shows that the illustrative policy held for thirty years returns only 0.02462, its cash value just 0.9906 of a bond ladder fed with the same premiums. Chapter 2 finds the break-even behaviour gap for buying term and investing the difference: over thirty years a gap of 0.0148 a year erases the paper advantage, and even with perfect discipline 0.3532 of simulated paths lose to the policy. Chapter 3 splits the return shortfall into spread, expenses, cost of insurance and surrender charges, removing order effects with the Shapley value. Chapter 4 follows ten thousand illustrative policies: 0.3285 of holders surrender at a loss, the pooled return falls to 0.02129, and the cost of surrender and policy loans under a sudden need for cash is computed. Chapter 5 recomputes Halley's printed annuity values age by age; at age 10 and 6 % the value of 13.44 years' purchase matches his digit for digit, and at the flat price of seven years a buyer aged 10 could expect 0.1303. Chapter 6 compares historical and modern mortality with the Gompertz line and computes the returns of annuities and policies under different lifetimes. Chapter 7 reads the granary officer of the Rites of Zhou, the Song charity granaries and Halley's own text; in an illustrative model a one-tenth levy with low losses meets 0.9445 of famine relief needs. Chapter 8 offers the distilled Policy IRR Decomposer, a personal log, a disclosure procedure for advice shops and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:关于储蓄型保单流行着两句相反的话:一句说它稳健保证、长期翻倍,一句说只该买定期寿险、把差额拿去投资,长期一定赢。本书把两句话放进同一个可运行的模型,说明二者都把一个依赖参数的比较说成了定论。核心工具是现金价值表的内部收益率(牛顿迭代与二分法求根,现值一律精确求和)、哈雷 1693 年布雷斯劳生命表的年金现值,以及行为落差的敏感度分析。书中保单、保费、现金价值、费率与死亡率均为示意,不点名、不评价任何保险公司或产品;全部计算由书中程序生成;本书属于非个人化理财建议,不构成法律或税务意见,法条与税务问题以上架前人工复核为准。 第一章显示,示意保单持有三十年的收益率只有 0.02462,现金价值仅为同样交费买国债阶梯终值的 0.9906。第二章求出「买定期、投差额」的盈亏平衡行为落差,三十年尺度上每年 0.0148 就足以抹平纸面优势;即使完全守纪律,仍有 0.3532 的模拟路径输给保单。第三章把收益率缺口拆成利差、费用、保障成本与退保费用,并用沙普利值消除顺序影响。第四章追踪一万张示意保单,0.3285 的投保人亏本退保,整群合并收益率降到 0.02129,并计算急用钱时退保与保单贷款的代价。 第五章逐龄复算哈雷印出的年金价值,10 岁按 6 % 为 13.44 年购买价,与原文逐位相同;按七年购买价出售时,10 岁买家的期望收益率高达 0.1303。第六章用冈珀茨直线对比历史与现代死亡率,计算年金与保单在不同寿命下的收益率。第七章对读《周礼》遗人委积、宋代义仓与哈雷原文,示意模型中按十分之一附加征收、年损耗低时,义仓能满足 0.9445 的赈济需要。第八章给出蒸馏小工具「保单 IRR 拆解器」、个人日志、门店披露流程与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The IRR of a Cash-Value Table: Newton's Method, Bisection and the Government-Bond Benchmark — Chapter 1 shows that the illustrative policy held for thirty years returns only 0.02462, its cash value just 0.9906 of a bond ladder fed with the same premiums. (https://5dlibrary.com/answers/en/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c01/) 第1章 现金价值表的内部收益率:牛顿迭代、二分法与国债基准 — 第一章显示,示意保单持有三十年的收益率只有 0.02462,现金价值仅为同样交费买国债阶梯终值的 0.9906。 (https://5dlibrary.com/answers/zh/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c01/) - Ch.2 Where "Buy Term and Invest the Difference" Breaks Even: Behaviour Gaps, Stopping Risk and Dollar-Weighted Returns — Chapter 2 finds the break-even behaviour gap for buying term and investing the difference: over thirty years a gap of 0.0148 a year erases the paper advantage, and even with perfect discipline 0.3532 of simulated paths lose to the policy. (https://5dlibrary.com/answers/en/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c02/) 第2章 买定期投差额的盈亏边界:行为落差、停投风险与资金加权收益 — 第二章求出「买定期、投差额」的盈亏平衡行为落差,三十年尺度上每年 0.0148 就足以抹平纸面优势;即使完全守纪律,仍有 0.3532 的模拟路径输给保单。 (https://5dlibrary.com/answers/zh/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c02/) - Ch.3 Four Parts of the Return Shortfall: Spread, Expenses, Cost of Insurance and Surrender Charges — The total shortfall in year ten is 0.03506. Expenses receive a Shapley value of 0.02421, with marginal drops between 0.02379 and 0.02463 across orders; the spread receives 0.00975, ranging from 0.00943 to 0.01005; the surrender charge 0.00090; the cost of insurance 0.00020. In year thirty the total is 0.01538, of which the spread takes 0.00992, expenses 0.00540, the cost of insurance only 0.00005 and the surrender charge nothing. The four Shapley values add up to the total digit for digit. (https://5dlibrary.com/answers/en/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c03/) 第3章 收益率缺口的四段拆解:利差、费用、保障成本与退保费用 — 第十年的总缺口是 0.03506。按沙普利值,费用分到 0.02421,在不同顺序下的边际降幅介于 0.02379 与 0.02463 之间;利差分到 0.00975,范围是 0.00943 到 0.01005;退保费用 0.00090;保障成本 0.00020。第三十年的总缺口是 0.01538,利差分到 0.00992,费用分到 0.00540,保障成本只有 0.00005,退保费用为零。四个沙普利值之和与总缺口逐位相同。 (https://5dlibrary.com/answers/zh/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c03/) - Ch.4 Surrender and Liquidity: What Ten Thousand Policies Actually Earn and the Cost of Needing Cash — Chapter 4 follows ten thousand illustrative policies: 0.3285 of holders surrender at a loss, the pooled return falls to 0.02129, and the cost of surrender and policy loans under a sudden need for cash is computed. (https://5dlibrary.com/answers/en/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c04/) 第4章 退保与流动性:一万张保单的实现收益率与急用钱的代价 — 第四章追踪一万张示意保单,0.3285 的投保人亏本退保,整群合并收益率降到 0.02129,并计算急用钱时退保与保单贷款的代价。 (https://5dlibrary.com/answers/zh/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c04/) - Ch.5 Halley, 1693: the Breslau Life Table and the Present Value of a Life Annuity — Chapter 5 recomputes Halley's printed annuity values age by age; at age 10 and 6 % the value of 13.44 years' purchase matches his digit for digit, and at the flat price of seven years a buyer aged 10 could expect 0.1303. (https://5dlibrary.com/answers/en/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c05/) 第5章 哈雷 1693:布雷斯劳生命表与终身年金的现值 — 第五章逐龄复算哈雷印出的年金价值,10 岁按 6 % 为 13.44 年购买价,与原文逐位相同;按七年购买价出售时,10 岁买家的期望收益率高达 0.1303。 (https://5dlibrary.com/answers/zh/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c05/) - Ch.6 Mortality Curves and Longevity Risk: the Gompertz Line, the Annuity as a Wager on Lifetime and the Return on Early Death — The Gompertz line puts Halley's table and the book's illustrative law on the same axes: Halley's world had high mortality rising slowly, today's has low mortality rising fast. The annuity buyer's return is a function of lifetime; the illustrative annuity crosses zero at 80 and the bond yield at 84, so long lives win and short lives lose. The return of a savings policy on death runs the other way, high for early death and identical to the surrender return for long lives. The two products carry lifetime risks in opposite directions, and any comparison that reports a single average return erases the difference. The next chapter returns to the classics, to see what the granary institutions of ancient China and Halley's criticism of annuity prices each got right. (https://5dlibrary.com/answers/en/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c06/) 第6章 死亡率曲线与长寿风险:冈珀茨直线、年金的寿命赌注与身故收益率 — 冈珀茨直线把哈雷表与本书示意死亡律放进同一坐标:哈雷的世界死亡率高、增长慢,今天的世界死亡率低、增长快。年金买家的收益率是寿命的函数,示意年金在 80 岁越过零、84 岁越过国债收益率,长寿者赢,早逝者输;储蓄型保单的身故收益率则相反,早逝时高,长寿时与退保收益率重合。两种产品承担的是方向相反的寿命风险,任何只报一个平均收益率的比较,都把这一点抹掉了。下一章回到原典,看看中国古代的仓储制度与哈雷的年金批评各自说对了什么。 (https://5dlibrary.com/answers/zh/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c06/) - Ch.7 Classics East and West: the Granary Officer of the Rites of Zhou, the Song Charity Granaries and Halley's Seven Years' Purchase — Chapter 7 reads the granary officer of the Rites of Zhou, the Song charity granaries and Halley's own text; in an illustrative model a one-tenth levy with low losses meets 0.9445 of famine relief needs. (https://5dlibrary.com/answers/en/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c07/) 第7章 东西方原典考辨:《周礼》遗人委积、宋代义仓与哈雷的七年购买价 — 第七章对读《周礼》遗人委积、宋代义仓与哈雷原文,示意模型中按十分之一附加征收、年损耗低时,义仓能满足 0.9445 的赈济需要。 (https://5dlibrary.com/answers/zh/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c07/) - Ch.8 Closing the Loop: the Policy IRR Decomposer, a Pre-Registered Plan and the Conditions of Refutation — Chapter 8 offers the distilled Policy IRR Decomposer, a personal log, a disclosure procedure for advice shops and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c08/) 第8章 实证闭环:保单 IRR 拆解器、预注册方案与证伪条件 — 第八章给出蒸馏小工具「保单 IRR 拆解器」、个人日志、门店披露流程与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/savings-insurance-irr-halley-life-table-buy-term-invest-difference/c08/) ## 5DL-B-221 · Jiaozi Over-Issue and Gresham's Law, Thiers' Reversal under Floating Exchange and Stablecoin Run Thresholds 中文书名:交子超发与格雷欣定律、浮动兑换下的提尔斯逆转与稳定币挤兑临界母典 DOI: pending Cite: Fan, Y. (2026). Jiaozi Over-Issue and Gresham's Law, Thiers' Reversal under Floating Exchange and Stablecoin Run Thresholds (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not personalised financial advice, it recommends no currency, stablecoin or platform, and regulatory matters are subject to manual review before publication. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: "Bad money drives out good" is often treated as a universal law, and "under floating rates good money drives out bad, so receive good money and pay in bad" as its clever reverse. This book writes these claims as runnable arithmetic and asks when they hold. Its parameters are illustrative; all calculations are generated by its programs; it is not personalised financial advice, it recommends no currency, stablecoin or platform, and regulatory matters are subject to manual review before publication. Chapter 1 shows that Gresham's law holds only when the legal ratio is enforced and using good money at a premium is costly: with no enforcement, the bad coin's share of payments falls below one half only at monthly inflation of 0.084, and with enforcement at one half it never does. Chapter 2 shows the order of the Thiers reversal: a business that reprices monthly changes its unit of account at monthly inflation of 0.0203, and dollarization, once established, does not reverse when inflation falls. Chapter 3 explains jiaozi over-issue with the quantity theory and Cagan's model: doubling per issue period gives a steady-state ratio of prices to notes of 2.828, and seigniorage peaks at growth of 1.667 per period. Chapter 4 shows that liquidity insurance and run fragility share one source; with noisy signals the optimal deposit contract has a run probability of 0.6248. Chapter 5 shows that one-for-one reserves do not prevent stablecoin runs: with a cash share of 0.2 and a fire-sale discount of 0.10 a coin needs backing of 1.085; on a weekend in March 2023 the worst-case backing of USDC was still 0.9216, yet its price fell to near 0.88. Chapter 6 shows that "receive good, pay bad" is itself Gresham behaviour and fails once quote pass-through exceeds 0.367; under floating quotes receipts and payments converge on one low-volatility currency. Chapter 7 examines the jiaozi passage of the Song shi and Copernicus on coinage. Chapter 8 gives settlement rules, the distilled Settlement Currency Selector and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:「劣币驱逐良币」常被当作万能定律,「浮动汇率下良币驱逐劣币,所以跨境收款收良币、付款付劣币」又常被当作它的反面妙招。本书把这些说法写成可以运行的算术,看它们在什么条件下成立。书中参数为示意;全部计算由书中程序生成;本书不是个人化的理财建议,不推荐任何币种、稳定币或平台,法规事项以上架前人工复核为准。 第一章说明格雷欣定律只在法定比价被执行、且以溢价使用良币成本较高时成立:完全不执行比价时,劣币的支付份额要到月通胀 0.084 才跌破一半,而执行概率达到一半时永远不会跌破。第二章说明提尔斯逆转的次序:按月改价的经营者在月通胀 0.0203 时就改换计价单位,美元化一旦形成,通胀回落也不回头。第三章用数量论与卡甘模型解释交子超发:每界翻倍时稳态物价与纸币之比为 2.828,使铸币税最大的发行速度是每界 1.667 倍。第四章说明流动性保险与挤兑脆弱性同源,最优存款合同在噪声信号下的挤兑概率为 0.6248。 第五章说明一比一储备不足以防止稳定币挤兑:现金占比 0.2、变卖折价 0.10 时需要 1.085 的背书;2023 年 3 月的周末,USDC 最坏情形背书仍有 0.9216,市价却跌到 0.88 附近。第六章说明「收良付劣」本身是格雷欣式行为,报价消化率超过 0.367 即失效,浮动报价下收付合流于同一种低波动货币。第七章考辨《宋史·食货志》交子条与哥白尼《铸币论》。第八章给出收付规则、蒸馏小工具「收付币种选择器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Limits of Gresham's Law: Legal Ratios, Enforcement and Three Fates of the Good Coin — Chapter 1 shows that Gresham's law holds only when the legal ratio is enforced and using good money at a premium is costly: with no enforcement, the bad coin's share of payments falls below one half only at monthly inflation of 0.084, and with enforcement at one half it never does. (https://5dlibrary.com/answers/en/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c01/) 第1章 格雷欣定律的边界:法定比价、执行概率与良币的三种去向 — 第一章说明格雷欣定律只在法定比价被执行、且以溢价使用良币成本较高时成立:完全不执行比价时,劣币的支付份额要到月通胀 0.084 才跌破一半,而执行概率达到一半时永远不会跌破。 (https://5dlibrary.com/answers/zh/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c01/) - Ch.2 The Thiers Reversal: the Unit of Account Leaves First, the Medium of Exchange Later, and Dollarization Stays — Chapter 2 shows the order of the Thiers reversal: a business that reprices monthly changes its unit of account at monthly inflation of 0.0203, and dollarization, once established, does not reverse when inflation falls. (https://5dlibrary.com/answers/en/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c02/) 第2章 提尔斯逆转:计价单位先走,交易媒介后走,美元化不回头 — 第二章说明提尔斯逆转的次序:按月改价的经营者在月通胀 0.0203 时就改换计价单位,美元化一旦形成,通胀回落也不回头。 (https://5dlibrary.com/answers/zh/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c02/) - Ch.3 Jiaozi Over-Issue: a Quantity-Theory Recursion, the Seigniorage Laffer Curve and the Dilution of Backing — Chapter 3 explains jiaozi over-issue with the quantity theory and Cagan's model: doubling per issue period gives a steady-state ratio of prices to notes of 2.828, and seigniorage peaks at growth of 1.667 per period. (https://5dlibrary.com/answers/en/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c03/) 第3章 交子超发:数量论递推、铸币税拉弗曲线与背书稀释 — 第三章用数量论与卡甘模型解释交子超发:每界翻倍时稳态物价与纸币之比为 2.828,使铸币税最大的发行速度是每界 1.667 倍。 (https://5dlibrary.com/answers/zh/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c03/) - Ch.4 Reserve Runs: the Diamond-Dybvig Contract and the Global-Game Threshold — Chapter 4 shows that liquidity insurance and run fragility share one source; with noisy signals the optimal deposit contract has a run probability of 0.6248. (https://5dlibrary.com/answers/en/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c04/) 第4章 准备金挤兑:Diamond–Dybvig 合同与全局博弈阈值 — 第四章说明流动性保险与挤兑脆弱性同源,最优存款合同在噪声信号下的挤兑概率为 0.6248。 (https://5dlibrary.com/answers/zh/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c04/) - Ch.5 The Stablecoin Run Threshold: Cash Shares, Fire-Sale Discounts and a Weekend in March 2023 — Chapter 5 shows that one-for-one reserves do not prevent stablecoin runs: with a cash share of 0.2 and a fire-sale discount of 0.10 a coin needs backing of 1.085; on a weekend in March 2023 the worst-case backing of USDC was still 0.9216, yet its price fell to near 0.88. (https://5dlibrary.com/answers/en/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c05/) 第5章 稳定币挤兑临界:现金占比、变卖折价与 2023 年 3 月的周末 — 第五章说明一比一储备不足以防止稳定币挤兑:现金占比 0.2、变卖折价 0.10 时需要 1.085 的背书;2023 年 3 月的周末,USDC 最坏情形背书仍有 0.9216,市价却跌到 0.88 附近。 (https://5dlibrary.com/answers/zh/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c05/) - Ch.6 Choosing Currencies for Cross-Border Receipts and Payments: Quote Pass-Through Decides Between a Split and a Convergence — Chapter 6 shows that "receive good, pay bad" is itself Gresham behaviour and fails once quote pass-through exceeds 0.367; under floating quotes receipts and payments converge on one low-volatility currency. (https://5dlibrary.com/answers/en/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c06/) 第6章 跨境收付的币种选择:报价消化率决定分流还是合流 — 第六章说明「收良付劣」本身是格雷欣式行为,报价消化率超过 0.367 即失效,浮动报价下收付合流于同一种低波动货币。 (https://5dlibrary.com/answers/zh/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c06/) - Ch.7 Classical Sources East and West: the Jiaozi Passage of the Song shi and Copernicus on Coinage — The jiaozi passage of the Song shi records why paper money arose, how private issue failed, the official quantity rule, and the first way the rule was broken: two jie in circulation raised the ceiling on notes to 1.995 times the cap, enough under the quantity theory to bring the note's price down to 0.501. Copernicus's treatise was right that debasement raises prices and drives good coin out of circulation, but it stated both proportionality and necessity too strongly. The next chapter turns the book's models into rules a business owner can follow and states what result would overturn the book. (https://5dlibrary.com/answers/en/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c07/) 第7章 东西方原典考辨:《宋史·食货志》交子条与哥白尼《铸币论》 — 《宋史·食货志》交子条记下了纸币的起因、私人发行的失败、官方的数量规则,以及规则第一次被突破的方式:两界并行使流通上限变为界额的 1.995 倍,按数量论足以使纸币价格降到 0.501。哥白尼的《铸币论》说对了减色抬高物价与良币离开流通,但把比例与必然说得过满。下一章把全书的模型变成经营者可以照做的规则,并说明什么样的结果会推翻本书。 (https://5dlibrary.com/answers/zh/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c07/) - Ch.8 Closing the Loop: Multi-Currency Settlement Rules, a Pre-Registered Test and Conditions for Refutation — Chapter 8 gives settlement rules, the distilled Settlement Currency Selector and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c08/) 第8章 实证闭环:多币种收付规则、预注册检验与证伪条件 — 第八章给出收付规则、蒸馏小工具「收付币种选择器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/jiaozi-gresham-thiers-reversal-stablecoin-run-threshold/c08/) ## 5DL-B-222 · Multilateral Punishment in Merchant Networks: Maghribi Agency Games and Trust Networks in Buyer Communities 中文书名:商帮多边惩罚声誉机制、马格里布商人代理博弈与买手社群信任网络母典 DOI: pending Cite: Fan, Y. (2026). Multilateral Punishment in Merchant Networks: Maghribi Agency Games and Trust Networks in Buyer Communities (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its agents, orders and network parameters are illustrative; all calculations are generated by its programs; the book is general business analysis, not business, legal or financial advice. Chapter 4 finds the critical group size: tens to over a hundred members in chain-like networks of acquaintances under illustrative parameters, but 33742 in a well-mixed network. Abstract: Cross-border traders widely believe that multilateral punishment inside a small circle, blacklisting together, disciplines overseas agents better than contracts and lawsuits, and that once the group grows too large news stops spreading and the blacklist fails. This book writes that belief as runnable arithmetic and tests when it holds. The core model is the efficiency-wage condition of a repeated game: an agent stays honest when the present value of future rent exceeds a one-shot theft; an SI model of information diffusion then yields the critical group size at which reputation fails. The project blueprint said multilateral punishment constrains better than contracts and litigation; after checking Greif's 1993 article and Edwards and Ogilvie's 2012 critique, the book restates this as a conditional claim and sets out both sides. Its agents, orders and network parameters are illustrative; all calculations are generated by its programs; the book is general business analysis, not business, legal or financial advice. Chapter 1 derives the discount threshold G/(G+r) under multilateral punishment, 0.8333 when theft is worth five periods of rent; if the cheat has a 0.15 chance per period of being rehired elsewhere it rises to 0.9804. Chapter 2 sets Greif against Edwards and Ogilvie and finds courts and multilateral punishment to be complements: a court that wins with probability 0.4 and no blacklist gives a threshold of 0.8497, close to that of a blacklist alone. Chapter 3 compares the speed of news in well-mixed, ring and small-world networks and shows that a rumour which stops leaves 0.2032 of a large population unreached. Chapter 4 finds the critical group size: tens to over a hundred members in chain-like networks of acquaintances under illustrative parameters, but 33742 in a well-mixed network. Chapter 5 prices a buyer alliance's shared blacklist: at a coverage of 0.95 the largest order that can be entrusted is 7.163 periods of rent, falling to 3.673 with a reporting lag of one period; false reports and deposits are quantified too. Chapter 6 treats last periods, exits and new identities. Chapter 7 examines the Huizhou merchant manuals represented by Cheng Chunyu's Shishang leiyao and the Cairo Geniza business letters. Chapter 8 gives eight clauses for sharing a blacklist, a pre-registered test and the distilled Agent Credit Scorecard; in simulation a verified shared list cuts the share of agents with a motive to default to 0.245, while an unverified one lets it climb back to 0.417. The scorecard uses transaction conduct only and must not be used to discriminate. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:做跨境生意的人普遍相信:小圈子里「一起拉进黑名单」的多边惩罚,比合同和诉讼更能约束海外代理人,但群一旦大了,消息传不开,黑名单就不灵了。本书把这条常识写成可以运行的算术,并检验它在什么条件下成立。核心模型是重复博弈中的效率薪酬条件:未来租金的现值大于一次性侵吞的所得,代理人就守信;再加上 SI 信息扩散模型,求出声誉机制失效的临界群规模。立项蓝图称多边惩罚「比合同和诉讼更能约束」,核对格雷夫 1993 年的原文与爱德华兹、奥格尔维 2012 年的质疑后,本书把它修正为一个条件命题,并把正反两方并陈。书中代理人、订单与网络参数均为示意;全部计算由书中程序生成;本书是非个人化的经营分析,不构成经营、法律或理财建议。 第一章推导多边惩罚下的折现因子门槛 G/(G+r),侵吞所得为五期租金时为 0.8333;骗子每期只要有 0.15 的机会被别人再雇用,门槛就升到 0.9804。第二章对读格雷夫与爱德华兹—奥格尔维,计算表明法庭与多边惩罚是补充关系:只有胜诉概率 0.4 的法庭时门槛为 0.8497,与只有黑名单时相差无几。第三章比较均匀混合、环状与小世界通信网中的消息速度,并说明会停下的谣言在大群体中留下 0.2032 的人永远听不到。第四章求临界群规模:链状熟人网络在示意参数下只有几十到一百多人,均匀混合网络要到 33742 人。 第五章计算买手联盟的共享黑名单:覆盖率 0.95 时可托付的订单额上限为 7.163 期租金,上报滞后一个周期就降到 3.673;误报与保证金的作用也逐一量化。第六章讨论终局、退出与换马甲。第七章考辨明代程春宇编《士商类要》所代表的徽商商书与开罗 Geniza 商业书信。第八章给出八条黑名单共享条款、预注册检验方案与蒸馏小工具「代理人信用评分卡」;模拟中,核实式共享名单把有违约动机的代理人比例降到 0.245,不核实的名单则回升到 0.417。评分卡只用交易行为数据,不得用于歧视性用途。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 One-Shot Theft against Future Rent: the Discount Threshold of Multilateral Punishment — Chapter 1 derives the discount threshold G/(G+r) under multilateral punishment, 0.8333 when theft is worth five periods of rent; if the cheat has a 0.15 chance per period of being rehired elsewhere it rises to 0.9804. (https://5dlibrary.com/answers/en/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c01/) 第1章 一次性侵吞与未来租金:多边惩罚的折现因子门槛 — 第一章推导多边惩罚下的折现因子门槛 G/(G+r),侵吞所得为五期租金时为 0.8333;骗子每期只要有 0.15 的机会被别人再雇用,门槛就升到 0.9804。 (https://5dlibrary.com/answers/zh/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c01/) - Ch.2 The Maghribi Dispute: Greif's Coalition and Edwards and Ogilvie's Courts — Chapter 2 sets Greif against Edwards and Ogilvie and finds courts and multilateral punishment to be complements: a court that wins with probability 0.4 and no blacklist gives a threshold of 0.8497, close to that of a blacklist alone. (https://5dlibrary.com/answers/en/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c02/) 第2章 马格里布商人之争:格雷夫的联盟与爱德华兹—奥格尔维的法庭 — 第二章对读格雷夫与爱德华兹—奥格尔维,计算表明法庭与多边惩罚是补充关系:只有胜诉概率 0.4 的法庭时门槛为 0.8497,与只有黑名单时相差无几。 (https://5dlibrary.com/answers/zh/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c02/) - Ch.3 How Bad News Travels: SI Diffusion, Rings of Correspondents and Rumours that Stop — Chapter 3 compares the speed of news in well-mixed, ring and small-world networks and shows that a rumour which stops leaves 0.2032 of a large population unreached. (https://5dlibrary.com/answers/en/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c03/) 第3章 坏消息怎样走:SI 扩散、环状通信网与会停下的谣言 — 第三章比较均匀混合、环状与小世界通信网中的消息速度,并说明会停下的谣言在大群体中留下 0.2032 的人永远听不到。 (https://5dlibrary.com/answers/zh/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c03/) - Ch.4 The Critical Group Size: How Large a Network Can Reputation Police — Chapter 4 finds the critical group size: tens to over a hundred members in chain-like networks of acquaintances under illustrative parameters, but 33742 in a well-mixed network. (https://5dlibrary.com/answers/en/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c04/) 第4章 临界群规模:网络多大时声誉机制失效 — 第四章求临界群规模:链状熟人网络在示意参数下只有几十到一百多人,均匀混合网络要到 33742 人。 (https://5dlibrary.com/answers/zh/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c04/) - Ch.5 Blacklists in Buyer Alliances: Coverage, False Reports and Deposits — Chapter 5 prices a buyer alliance's shared blacklist: at a coverage of 0.95 the largest order that can be entrusted is 7.163 periods of rent, falling to 3.673 with a reporting lag of one period; false reports and deposits are quantified too. (https://5dlibrary.com/answers/en/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c05/) 第5章 买手联盟的黑名单:共享覆盖率、误报与保证金 — 第五章计算买手联盟的共享黑名单:覆盖率 0.95 时可托付的订单额上限为 7.163 期租金,上报滞后一个周期就降到 3.673;误报与保证金的作用也逐一量化。 (https://5dlibrary.com/answers/zh/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c05/) - Ch.6 Where Reputation Ends: Last Periods, Exits and New Names — Reputation's collateral is the future, so whatever weakens the future weakens reputation. Once an agent's chance of leaving reaches 0.05 per period, the rent at stake under illustrative parameters no longer outweighs a single theft; with a known end, large sums entrusted within 13 periods of departure need escrow; and when new names are free, every honest newcomer pays 8 periods of probation for the cheat's ability to disguise himself. The common rule of repair is this: where the future is not worth enough, give the agent something to lose now. With that, the modelling part of the book is complete. How did people in earlier centuries think about these problems? The next chapter returns to two sources, one from Ming-dynasty Huizhou and one from that storeroom in Cairo. (https://5dlibrary.com/answers/en/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c06/) 第6章 声誉的边界:终局、退出与换马甲 — 声誉的抵押品是未来,所以凡是削弱未来的东西都会削弱声誉。代理人每期离开的概率达到 0.05,示意参数下租金现值就不够抵消一次侵吞;已知终局时,离开前 13 期以内的大额托付都需要托管款补位;换马甲免费时,所有诚实的新人都要为骗子的伪装能力交 8 期考察期的学费。补救的共同原则是:在未来不够值钱的地方,让代理人现在就有东西可以失去。到这里,模型部分已经完整。古人对这些问题是怎么想的?下一章回到两部原典,一部来自明代徽州,一部来自开罗的那间储藏室。 (https://5dlibrary.com/answers/zh/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c06/) - Ch.7 Sources East and West: the Shishang leiyao and the Cairo Geniza Letters — The Ming merchant manuals represented by the Shishang leiyao and the letters of the Maghribi traders in the Cairo Geniza answer the same question from two directions. The manuals' precepts tell a merchant how to protect himself where news does not flow, and p71 shows that native-place networks, social sanctions and guarantees, layered on one another, can multiply the order value that can be entrusted more than twentyfold. The Geniza letters show a group of merchants exchanging reputational information by correspondence, and p72 shows that this information was bound to the seasons of shipping, with winter delays long enough to loosen the hold of reputation for months. Neither source tells us where the limits of these mechanisms lie; that is a question today's data can answer. The last chapter turns the book's models into a protocol a buyer alliance can follow: how to build the list, how to take deposits, how to compute the scorecard, and what results would overturn this book? (https://5dlibrary.com/answers/en/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c07/) 第7章 东西方原典考辨:《士商类要》与开罗 Geniza 书信 — 《士商类要》代表的明代商书,与开罗 Geniza 中的马格里布商人书信,从两个方向回答了同一个问题。商书的规训告诉一位商人在消息不通的地方如何自保,p71 说明同乡网络、社会惩罚与担保层层叠加,可以把可托付的订单额放大二十多倍。Geniza 书信展示了一群商人如何通过通信交换声誉信息,p72 说明这种信息受季节航运的束缚,冬季的延迟足以让声誉的约束在几个月里松动。两部原典都没有告诉我们这些机制的边界在哪里,那是今天的数据可以回答的问题。最后一章要把全书的模型变成一份买手联盟可以照着执行的协议:名单怎么建、保证金怎么收、评分卡怎么算、什么样的结果会推翻本书? (https://5dlibrary.com/answers/zh/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c07/) - Ch.8 Closing the Loop: a Blacklist-Sharing Protocol, Deposits and the Agent Credit Scorecard — Chapter 8 gives eight clauses for sharing a blacklist, a pre-registered test and the distilled Agent Credit Scorecard; in simulation a verified shared list cuts the share of agents with a motive to default to 0.245, while an unverified one lets it climb back to 0.417. The scorecard uses transaction conduct only and must not be used to discriminate. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c08/) 第8章 实证闭环:黑名单共享协议、保证金与代理人信用评分卡 — 第八章给出八条黑名单共享条款、预注册检验方案与蒸馏小工具「代理人信用评分卡」;模拟中,核实式共享名单把有违约动机的代理人比例降到 0.245,不核实的名单则回升到 0.417。评分卡只用交易行为数据,不得用于歧视性用途。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/merchant-multilateral-punishment-maghribi-agency-buyer-trust/c08/) ## 5DL-B-223 · The Deadweight Loss of Gift-Giving, Reciprocity-Ledger Decay and the Signalling Value of Corporate Gifts 中文书名:送礼无谓损失、人情账本互惠衰减与企业节礼信号价值母典 DOI: pending Cite: Fan, Y. (2026). The Deadweight Loss of Gift-Giving, Reciprocity-Ledger Decay and the Signalling Value of Corporate Gifts (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its clients, gifts and parameters are illustrative; all calculations are generated by its programs; the book is general business analysis, not business, legal, tax or financial advice, and statutes must be checked by hand before publication. Chapter 5 derives the break-even yearly gift: for the same lift a loyal client justifies 2.67 times the gift, and an illustrative allocation by value nets 356.63 against 90.98 for an equal split. Abstract: Business owners commonly believe two contradictory things: that giving is wasteful, since recipients value gifts below their price and cash would be cheaper, and that the more important a relationship, the more it calls for expensive, impractical gifts, with cash doing the most damage. This book writes both beliefs as runnable arithmetic and tests when they hold. The project blueprint said recipients value gifts a tenth to a third below price and that cash hurts relationships most; after checking Waldfogel's 1993 article and the counter-evidence of Solnick and Hemenway and of List and Shogren, the book restates the first half as a claim conditional on the method of measurement, and the second as a claim that varies with relationship and culture and is, in corporate giving, first a matter of compliance, setting out both sides. Its clients, gifts and parameters are illustrative; all calculations are generated by its programs; the book is general business analysis, not business, legal, tax or financial advice, and statutes must be checked by hand before publication. Chapter 1 models yields as lognormal: at a mean of two thirds of the price, 0.853 of gifts fall below price; the same gifts show a mean yield of 0.866 on a questionnaire that excludes sentiment and 1.269 in an auction that counts sentiment and the endowment effect. Chapter 2 traces the loss to missing information about preferences: a business contact achieves a mean yield of only 0.607, and distant givers switch to cash more often. Chapter 3 gives the separating equilibrium of costly signals and shows that for the same 100, cash leaves partners at only 0.462 of those who turn up while a personalised gift reaches 0.774. Chapter 4 writes the return-gift recursion of a favour ledger: with a tenth fading each year the return ratio must reach 1.111 to hold; as the chance of exchange falls from 0.95 to 0.50, a relationship's mean lifetime falls from 160.1 years to 9.1. Chapter 5 derives the break-even yearly gift: for the same lift a loyal client justifies 2.67 times the gift, and an illustrative allocation by value nets 356.63 against 90.98 for an equal split. Chapter 6 draws the red lines against commercial bribery, with a screening rule stricter than the law and a rolling-total check. Chapter 7 examines the passage on reciprocity in the Book of Rites and Mauss's The Gift and shows how a gift list anchors returns to prices. Chapter 8 gives the distilled Holiday Gift Budgeter, rules for choose-one gift boxes (a net yield of 0.753 against 0.607 for a fixed box) and a pre-registered test (detecting a renewal lift of 0.05 needs 1094 clients per arm). Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:经营者普遍相信两件互相矛盾的事:送礼是一种浪费,收礼人对礼物的估值低于售价,直接给钱最省;可是越重要的关系越要送昂贵而不实用的礼物,送现金反而最伤关系。本书把这两种说法写成可以运行的算术,并检验它们在什么条件下成立。立项蓝图称收礼人估值平均比售价低一成到三成、现金最伤关系;核对沃尔德福格尔 1993 年的原文及索尔尼克与海明威、李斯特与肖格伦的反证之后,本书把前半句修正为取决于测量方法的条件命题,把后半句改写为因关系与文化而异、在企业节礼中首先是合规问题的命题,并把正反两方并陈。书中客户、礼品与参数均为示意;全部计算由书中程序生成;本书是非个人化的经营分析,不构成经营、法律、税务或理财建议,法条以上架前人工复核为准。 第一章用对数正态分布刻画估值比:均值为售价的三分之二时,有 0.853 的礼物低于售价;同一批礼物在不计情感的问卷下平均估值比为 0.866,计入情感与禀赋效应的拍卖下为 1.269。第二章说明估值损失来自偏好信息不足,生意联系人送礼的平均估值比只有 0.607,远亲更常改送现金。第三章给出昂贵信号的分离均衡,并说明同样花 100,现金让出面者中伙伴只占 0.462,定制礼物可达 0.774。第四章写人情账本的回礼递推:每年淡去一成时,回礼比率须达 1.111 才能维持;往来概率从 0.95 降到 0.50,关系的平均寿命从 160.1 年降到 9.1 年。 第五章推导盈亏平衡的年礼,同样的提升下忠诚客户值得 2.67 倍的礼,按价值分配的示意预算净收益为 356.63,平均分配只有 90.98。第六章划出反商业贿赂红线,给出比法律更严的筛查规则与滚动累计检查。第七章考辨《礼记·曲礼上》「礼尚往来」一段与莫斯的《礼物》,说明礼单把回礼锚定在价格上。第八章给出蒸馏小工具「节礼预算器」、多选一礼盒的选品规则(净估值比 0.753,固定礼盒为 0.607)与预注册检验方案(检测 0.05 的续约提升每组需 1094 位客户)。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Distribution of Valuation Losses: Where a Tenth to a Third Goes — Chapter 1 models yields as lognormal: at a mean of two thirds of the price, 0.853 of gifts fall below price; the same gifts show a mean yield of 0.866 on a questionnaire that excludes sentiment and 1.269 in an auction that counts sentiment and the endowment effect. (https://5dlibrary.com/answers/en/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c01/) 第1章 估值损失分布:一成到三成的无谓损失从哪里来 — 第一章用对数正态分布刻画估值比:均值为售价的三分之二时,有 0.853 的礼物低于售价;同一批礼物在不计情感的问卷下平均估值比为 0.866,计入情感与禀赋效应的拍卖下为 1.269。 (https://5dlibrary.com/answers/zh/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c01/) - Ch.2 Who Gives Worst: Preference Information, Social Distance and the Stigma of Cash — Chapter 2 traces the loss to missing information about preferences: a business contact achieves a mean yield of only 0.607, and distant givers switch to cash more often. (https://5dlibrary.com/answers/en/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c02/) 第2章 谁送得最差:偏好信息、关系远近与现金的污名 — 第二章说明估值损失来自偏好信息不足,生意联系人送礼的平均估值比只有 0.607,远亲更常改送现金。 (https://5dlibrary.com/answers/zh/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c02/) - Ch.3 The Separating Equilibrium of Costly Signals: Why "Wasteful" Gifts Are Credible — Chapter 3 gives the separating equilibrium of costly signals and shows that for the same 100, cash leaves partners at only 0.462 of those who turn up while a personalised gift reaches 0.774. (https://5dlibrary.com/answers/en/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c03/) 第3章 昂贵信号的分离均衡:为什么「浪费」的礼物反而可信 — 第三章给出昂贵信号的分离均衡,并说明同样花 100,现金让出面者中伙伴只占 0.462,定制礼物可达 0.774。 (https://5dlibrary.com/answers/zh/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c03/) - Ch.4 The Return-Gift Recursion and the Decay of Reciprocity in a Favour Ledger — Chapter 4 writes the return-gift recursion of a favour ledger: with a tenth fading each year the return ratio must reach 1.111 to hold; as the chance of exchange falls from 0.95 to 0.50, a relationship's mean lifetime falls from 160.1 years to 9.1. (https://5dlibrary.com/answers/en/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c04/) 第4章 人情账本的回礼递推与互惠衰减 — 第四章写人情账本的回礼递推:每年淡去一成时,回礼比率须达 1.111 才能维持;往来概率从 0.95 降到 0.50,关系的平均寿命从 160.1 年降到 9.1 年。 (https://5dlibrary.com/answers/zh/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c04/) - Ch.5 Holiday Budgets for High-Value Clients: Breaking Even on Retention — Chapter 5 derives the break-even yearly gift: for the same lift a loyal client justifies 2.67 times the gift, and an illustrative allocation by value nets 356.63 against 90.98 for an equal split. (https://5dlibrary.com/answers/en/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c05/) 第5章 高客单客户的节礼预算:留存提升的盈亏平衡 — 第五章推导盈亏平衡的年礼,同样的提升下忠诚客户值得 2.67 倍的礼,按价值分配的示意预算净收益为 356.63,平均分配只有 90.98。 (https://5dlibrary.com/answers/zh/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c05/) - Ch.6 Red Lines against Commercial Bribery: Where a Gift Ends and a Bribe Begins — Recipient R-A receives two gifts of 120 a year; at every point the rolling total is 240 and always compliant. Recipient R-B is different. Each gift is only 70, far below the per-gift cap, but six were scheduled in 2026. At the fifth the rolling total reaches 350 and the count 5, breaching both limits, so that gift and the one right after it are refused. Refused gifts are not given and so do not enter the total; by January of the following year the earliest gifts have left the window, and a new gift passes again. (https://5dlibrary.com/answers/en/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c06/) 第6章 反商业贿赂红线:礼与贿的分界 — 收礼人 R-A 每年收到两份价值 120 的礼物,任何时点上滚动累计都是 240,始终合规。收礼人 R-B 的情形不同:每份礼物只有 70,远低于单件上限,但在 2026 年里一共安排了六次。到第五次时,滚动累计达到 350,次数达到 5 次,两项同时超标,这次和紧随其后的一次都被拒绝。被拒绝的礼物没有送出,因此不计入累计;到次年一月,最早的几笔滑出窗口,新的礼物又可以通过。 (https://5dlibrary.com/answers/zh/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c06/) - Ch.7 Eastern and Western Sources Examined: the Quli Chapter of the Book of Rites and Mauss's The Gift — The result is clear. Without a gift list and with a mean perceived yield of 0.87, within ten rounds almost every pair has one side feeling slighted, a share of 0.998; 0.800 of the exchanges lapse, and the median gift size ends at only 0.265. The cause is the valuation loss of Chapter 1: recipients feel the gift is worth less than its price, return according to that perception, and so returns shrink systematically. Even with a mean yield of one, valuation noise alone leaves 0.958 of the pairs with at least one slight and 0.283 of exchanges lapsed. Adding a tenth to every return, without a gift list, only brings the lapse share down to 0.446. (https://5dlibrary.com/answers/en/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c07/) 第7章 东西方原典考辨:《礼记·曲礼上》与莫斯《礼物》 — 结果很清楚。没有礼单、感知估值比均值为 0.87 时,十轮之内几乎每一对都会有一方感到被怠慢,比例是 0.998,有 0.800 的往来中断,礼物大小的中位数最终只剩 0.265。原因是第一章的估值损失:收礼人觉得礼物不如售价值钱,按感知价值回礼,回礼就系统性地变小。即使估值比的均值为一,单纯的估值噪声也让 0.958 的配对至少有一次感到被怠慢,0.283 的往来中断。回礼时多添一成,在没有礼单的情况下也只能把中断比例降到 0.446。 (https://5dlibrary.com/answers/zh/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c07/) - Ch.8 Closing the Loop: the Holiday Gift Budgeter, Gift-Box SKUs and a Pre-Registered Test — Chapter 8 gives the distilled Holiday Gift Budgeter, rules for choose-one gift boxes (a net yield of 0.753 against 0.607 for a fixed box) and a pre-registered test (detecting a renewal lift of 0.05 needs 1094 clients per arm). Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c08/) 第8章 实证闭环:节礼预算器、企业礼盒 SKU 与预注册检验 — 第八章给出蒸馏小工具「节礼预算器」、多选一礼盒的选品规则(净估值比 0.753,固定礼盒为 0.607)与预注册检验方案(检测 0.05 的续约提升每组需 1094 位客户)。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/gift-deadweight-loss-reciprocity-ledger-corporate-gift-signalling/c08/) ## 5DL-B-224 · Exact Enumeration of Baccarat House Edges, the Futility of Card Counting and Gambler's Ruin Probabilities 中文书名:百家乐庄闲边际精确枚举、算牌无效性与赌徒破产概率母典 DOI: pending Cite: Fan, Y. (2026). Exact Enumeration of Baccarat House Edges, the Futility of Card Counting and Gambler's Ruin Probabilities (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative, all calculations are generated by its programs, and nothing in it is personalised financial advice. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: It is often said that baccarat, like blackjack, can be beaten by keeping track of the cards, and that the scoreboard of past results beside the table hides a pattern. This book tests both beliefs with exact calculation. It analyses probabilities and ruin only; it does not encourage any gambling, teaches no gambling technique and recommends no venue or platform. Its parameters are illustrative, all calculations are generated by its programs, and nothing in it is personalised financial advice. Chapter 1 enumerates every card sequence of an eight-deck shoe in rational arithmetic: Banker wins 0.458597, Player 0.446247 and a tie 0.095156, with house edges of 0.010579 on Banker and 0.012351 on Player, agreeing digit for digit with the values Thorp and Walden published in 1966. Chapter 2 computes effects of removal, simulates whole shoes and repeats the small-pack experiment: Banker has a positive expectation on only 0.00286 of coups, and an observer with perfect information gains only 0.0014 units per hundred coups. Chapter 3 uses the gambler's ruin formula to show that doubling 20 units on Player succeeds with probability only 0.3668. Chapter 4 proves that doubling and other staking systems merely rearrange the same price: the expected loss per unit staked is always 0.013650. Chapter 5 tests the road on two thousand really dealt shoes; the runs test rejects in a share of 0.0480, as expected without any signal, and the within-scorecard share of Banker after three Bankers averages only 0.4502, a small-sample bias. Chapter 6 shows that a hard stop line makes the chance of losing more than the budget zero, while one buy-back raises it to 0.1484 within 300 coups. Chapter 7 examines the biography of Tao Kan in the Book of Jin and Cardano's Book on Games of Chance; by equality of conditions the fair commission would be only 0.026932. Chapter 8 gives a command-line tool, the Entertainment Stop-Loss Calculator, a risk-notice procedure and pre-registered tests. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:人们常说,百家乐和 21 点一样可以靠记牌找到下注时机,牌桌边的路单也藏着规律。本书用精确计算检验这两种说法。本书只做概率与破产分析,不鼓励任何赌博,不教任何博彩技巧,不推荐任何场所或平台;书中参数为示意,全部计算由书中程序生成;书中内容是非个人化理财建议。 第一章用有理数精确枚举八副牌的全部发牌序列:庄赢 0.458597,闲赢 0.446247,和局 0.095156,押庄的庄家优势 0.010579,押闲 0.012351,与 Thorp 和 Walden 在 1966 年发表的数值逐位一致。第二章计算移牌效应、模拟整副牌靴并重做小牌堆实验:押庄为正期望的局只占 0.00286,完美信息观察者每一百局只得 0.0014 个单位。第三章用赌徒破产公式说明,带 20 个单位押闲想翻倍,成功概率只有 0.3668。第四章证明倍投等注码法只重新排列同一个价格,每押一个单位的期望损失恒为 0.013650。 第五章在两千副真实发出的牌靴中检验路单,游程检验的拒绝比例为 0.0480,与无信号时的预期一致;单看一张路单得出的「三庄之后开庄」比例平均只有 0.4502,这是小样本偏差。第六章说明硬止损线让输掉超过预算的概率为零,而追损一次在 300 局内会把它升到 0.1484。第七章考辨《晋书·陶侃传》与卡尔达诺的《论赌博游戏》,按条件相等原则,公平的佣金只应是 0.026932。第八章给出命令行小工具「娱乐止损计算器」、风险告知流程与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Exact Enumeration of the Eight-Deck Shoe: Where Banker 1.06 % and Player 1.24 % Come From — Chapter 1 enumerates every card sequence of an eight-deck shoe in rational arithmetic: Banker wins 0.458597, Player 0.446247 and a tie 0.095156, with house edges of 0.010579 on Banker and 0.012351 on Player, agreeing digit for digit with the values Thorp and Walden published in 1966. (https://5dlibrary.com/answers/en/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c01/) 第1章 八副牌的精确枚举:庄 1.06 %、闲 1.24 % 从哪里来 — 第一章用有理数精确枚举八副牌的全部发牌序列:庄赢 0.458597,闲赢 0.446247,和局 0.095156,押庄的庄家优势 0.010579,押闲 0.012351,与 Thorp 和 Walden 在 1966 年发表的数值逐位一致。 (https://5dlibrary.com/answers/zh/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c01/) - Ch.2 Why Counting Barely Works: Effects of Removal, Simulated Shoes and the Fundamental Theorem — Chapter 2 computes effects of removal, simulates whole shoes and repeats the small-pack experiment: Banker has a positive expectation on only 0.00286 of coups, and an observer with perfect information gains only 0.0014 units per hundred coups. (https://5dlibrary.com/answers/en/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c02/) 第2章 算牌为什么几乎无效:移牌效应、牌靴模拟与计牌基本定理 — 第二章计算移牌效应、模拟整副牌靴并重做小牌堆实验:押庄为正期望的局只占 0.00286,完美信息观察者每一百局只得 0.0014 个单位。 (https://5dlibrary.com/answers/zh/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c02/) - Ch.3 Gambler's Ruin: a Finite Bankroll Against a House Without a Ceiling — Chapter 3 uses the gambler's ruin formula to show that doubling 20 units on Player succeeds with probability only 0.3668. (https://5dlibrary.com/answers/en/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c03/) 第3章 赌徒破产:有限本金对没有上限的庄家 — 第三章用赌徒破产公式说明,带 20 个单位押闲想翻倍,成功概率只有 0.3668。 (https://5dlibrary.com/answers/zh/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c03/) - Ch.4 Staking Systems Cannot Change the Expectation: Doubling, d'Alembert and the Optional Stopping Theorem — Chapter 4 proves that doubling and other staking systems merely rearrange the same price: the expected loss per unit staked is always 0.013650. (https://5dlibrary.com/answers/en/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c04/) 第4章 注码法改不了期望:倍投、达朗贝尔与可选停止定理 — 第四章证明倍投等注码法只重新排列同一个价格,每押一个单位的期望损失恒为 0.013650。 (https://5dlibrary.com/answers/zh/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c04/) - Ch.5 The Road Has No Predictive Power: Streak Lengths, Runs Tests and Small-Sample Bias — Chapter 5 tests the road on two thousand really dealt shoes; the runs test rejects in a share of 0.0480, as expected without any signal, and the within-scorecard share of Banker after three Bankers averages only 0.4502, a small-sample bias. (https://5dlibrary.com/answers/en/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c05/) 第5章 路单没有预测力:连串长度、游程检验与小样本偏差 — 第五章在两千副真实发出的牌靴中检验路单,游程检验的拒绝比例为 0.0480,与无信号时的预期一致;单看一张路单得出的「三庄之后开庄」比例平均只有 0.4502,这是小样本偏差。 (https://5dlibrary.com/answers/zh/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c05/) - Ch.6 The Price of a Session: Loss Distributions, a Hard Stop Line and Chasing — Chapter 6 shows that a hard stop line makes the chance of losing more than the budget zero, while one buy-back raises it to 0.1484 within 300 coups. (https://5dlibrary.com/answers/en/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c06/) 第6章 一场会话的价格:损失分布、硬止损线与追损 — 第六章说明硬止损线让输掉超过预算的概率为零,而追损一次在 300 局内会把它升到 0.1484。 (https://5dlibrary.com/answers/zh/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c06/) - Ch.7 Sources East and West: Tao Kan Throws the Gaming Pieces into the River, and Cardano's Book on Games of Chance — Chapter 7 examines the biography of Tao Kan in the Book of Jin and Cardano's Book on Games of Chance; by equality of conditions the fair commission would be only 0.026932. (https://5dlibrary.com/answers/en/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c07/) 第7章 东西方原典考辨:陶侃投蒱博之具于江与卡尔达诺《论赌博游戏》 — 第七章考辨《晋书·陶侃传》与卡尔达诺的《论赌博游戏》,按条件相等原则,公平的佣金只应是 0.026932。 (https://5dlibrary.com/answers/zh/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c07/) - Ch.8 Closing the Loop: the Entertainment Stop-Loss Calculator, Risk Notices and Pre-Registered Tests — Chapter 8 gives a command-line tool, the Entertainment Stop-Loss Calculator, a risk-notice procedure and pre-registered tests. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c08/) 第8章 实证闭环:娱乐止损计算器、风险告知文案与预注册证伪方案 — 第八章给出命令行小工具「娱乐止损计算器」、风险告知流程与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/baccarat-house-edge-enumeration-card-counting-futility-gamblers-ruin/c08/) ## 5DL-B-225 · Rigid-Body Landing Probabilities of Moon Blocks, the Likelihood of Three Consecutive Holy Casts and Barnum Effects in Temple Oracle Verses 中文书名:掷筊杯刚体落地概率、连续三圣杯似然与签诗文本巴纳姆效应母典 DOI: pending Cite: Fan, Y. (2026). Rigid-Body Landing Probabilities of Moon Blocks, the Likelihood of Three Consecutive Holy Casts and Barnum Effects in Temple Oracle Verses (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs. Chapter 7 examines the entry on divination cups in Cheng Dachang's Yanfanlu and Cicero's De Divinatione, setting the illustrative Venus throw of four knucklebones, 0.0384, beside three holy casts. Abstract: In temples of southern Fujian, Taiwan and Southeast Asia, worshippers consult the deity by casting a pair of moon blocks, each with one flat and one convex face; a weighty matter often needs three holy casts in a row, and a drawn lot is followed by its verse. Popular belief holds that the blocks are as fair as coins, that three holy casts are too rare to be chance, and that a verse feels accurate because it names the asker's situation. This book takes the computable parts of the rite and tests them with rigid-body resting stability, Bernoulli trials and runs, Markov chains, likelihood ratios and a score for the universality of text. It analyses only structure, probability and text, gives no personal reading of anyone's question, and passes no judgement on any temple, deity or worshipper. Its parameters are illustrative; all calculations are generated by its programs. Chapter 1 applies a quasi-static rule to find the chance that one block lands flat face up; an illustrative half-disc model gives 0.6278, which lowers the holy cast of a pair to 0.4673, and the chapter proves that "a third each" is impossible for an identical pair. Chapter 2 corrects the project blueprint: for an identical, independent pair three holy casts come at most one time in eight, and only 0.1021 under the half-disc model; they carry at most 3.000 bits of evidence, and the more matters one asks about, the likelier a chance confirmation becomes. Chapter 3 gives estimates and an independence test for a logbook of three hundred casts, shows that a pair with bias 0.6 is identified within 97 casts, and shows that a mismatched pair can raise the run of three to 0.1951, breaking the one-eighth ceiling. Chapter 4 checks Forer's classroom experiment of 1949: its mean rating of 4.26 corresponds in a toy model to a universality of 0.852 per sentence; two verse-style lines carry only 1.380 bits, while one specific statement carries 6.508. Chapter 5 shows how ambiguous clauses, open deadlines and selective memory multiply, so that 0.5330 of simulated readers sincerely remember that most lots came true. Chapter 6 reads the rite as a protocol, shows that the final share of confirmations means nothing when restarts are allowed, and finds that a worn lot tube is caught by an audit of 2400 confirmations with chance only 0.650. Chapter 7 examines the entry on divination cups in Cheng Dachang's Yanfanlu and Cicero's De Divinatione, setting the illustrative Venus throw of four knucklebones, 0.0384, beside three holy casts. Chapter 8 gives a randomness statement, a controlled experiment on satisfaction and universality, a pre-registered protocol and the Moon Block Probability Simulator. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:在闽台与东南亚的宫庙里,信众掷一对一平一凸的筊杯请示神明,大事常要连得三次圣杯,抽签之后再读签诗。流行的说法是:筊杯像硬币一样公平,三圣杯难得到不可能出于偶然,签诗之所以准是因为说中了求签者的处境。本书把仪式中可以计算的部分拿出来,用刚体静止稳定性、伯努利与游程、马尔可夫链、似然比与文本普适度计分逐一检验。本书只分析结构、概率与文本,不做任何个人占断,也不评价任何宫庙、神明或信众。书中参数为示意;全部计算由书中程序生成。 第一章以准静态规则求单片筊杯平面朝上的机会,半圆截面的示意模型给出 0.6278,一对的圣杯随之降到 0.4673,并证明「三种结果各占三分之一」对相同的一对不可能成立。第二章修正立项蓝图:对相同且独立的一对,三圣杯至多八分之一,半圆截面下只有 0.1021;它最多提供 3.000 比特的证据,问的事情越多,单凭偶然被确认的机会越高。第三章给出三百次投掷记录本的估计与独立性检验,指出偏置为 0.6 的一对只需 97 次投掷即可识别,而失配的一对能把三连抬到 0.1951,突破八分之一的上限。 第四章核对福勒 1949 年的课堂实验,平均分 4.26 在玩具模型中相当于每句话的普适度 0.852;两句签诗式句子只携带 1.380 比特,一句具体陈述却有 6.508 比特。第五章说明多义条文、无期限与选择性记忆的乘积,使 0.5330 的模拟读者真诚地记得「大多对上了」。第六章把仪轨读成协议,指出允许重来时最终确认比例不说明任何问题,并说明一个磨损的签筒在 2400 次确认时也只有 0.650 的机会被审计发现。第七章考辨程大昌《演繁露》杯珓条与西塞罗《论占卜》,以距骨维纳斯一掷 0.0384 的示意值对照三圣杯。第八章给出随机性声明、满意度与普适性对照实验、预注册方案与小工具「筊杯概率模拟器」。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Flat Face and Convex Face: the Rigid-Body Resting Stability of a Single Moon Block — Chapter 1 applies a quasi-static rule to find the chance that one block lands flat face up; an illustrative half-disc model gives 0.6278, which lowers the holy cast of a pair to 0.4673, and the chapter proves that "a third each" is impossible for an identical pair. (https://5dlibrary.com/answers/en/moon-block-landing-probability-three-holy-casts-barnum-verses/c01/) 第1章 平面与凸面:筊杯单片落地的刚体静止稳定性 — 第一章以准静态规则求单片筊杯平面朝上的机会,半圆截面的示意模型给出 0.6278,一对的圣杯随之降到 0.4673,并证明「三种结果各占三分之一」对相同的一对不可能成立。 (https://5dlibrary.com/answers/zh/moon-block-landing-probability-three-holy-casts-barnum-verses/c01/) - Ch.2 Three Holy Casts in a Row: One Eighth Is a Ceiling, Not a Miracle — Chapter 2 corrects the project blueprint: for an identical, independent pair three holy casts come at most one time in eight, and only 0.1021 under the half-disc model; they carry at most 3.000 bits of evidence, and the more matters one asks about, the likelier a chance confirmation becomes. (https://5dlibrary.com/answers/en/moon-block-landing-probability-three-holy-casts-barnum-verses/c02/) 第2章 连续三圣杯:八分之一是上限而不是神迹 — 第二章修正立项蓝图:对相同且独立的一对,三圣杯至多八分之一,半圆截面下只有 0.1021;它最多提供 3.000 比特的证据,问的事情越多,单凭偶然被确认的机会越高。 (https://5dlibrary.com/answers/zh/moon-block-landing-probability-three-holy-casts-barnum-verses/c02/) - Ch.3 A Logbook of Three Hundred Casts: Three-Way Counts, a Test of Independence and the Mismatched Pair — Chapter 3 gives estimates and an independence test for a logbook of three hundred casts, shows that a pair with bias 0.6 is identified within 97 casts, and shows that a mismatched pair can raise the run of three to 0.1951, breaking the one-eighth ceiling. (https://5dlibrary.com/answers/en/moon-block-landing-probability-three-holy-casts-barnum-verses/c03/) 第3章 三百次投掷的记录本:三态计数、独立性检验与失配的一对 — 第三章给出三百次投掷记录本的估计与独立性检验,指出偏置为 0.6 的一对只需 97 次投掷即可识别,而失配的一对能把三连抬到 0.1951,突破八分之一的上限。 (https://5dlibrary.com/answers/zh/moon-block-landing-probability-three-holy-casts-barnum-verses/c03/) - Ch.4 How Universal a Temple Verse Is: Scoring Barnum Statements and a Model of Acceptance — Chapter 4 checks Forer's classroom experiment of 1949: its mean rating of 4.26 corresponds in a toy model to a universality of 0.852 per sentence; two verse-style lines carry only 1.380 bits, while one specific statement carries 6.508. (https://5dlibrary.com/answers/en/moon-block-landing-probability-three-holy-casts-barnum-verses/c04/) 第4章 签诗的普适度:巴纳姆陈述的计分与接受模型 — 第四章核对福勒 1949 年的课堂实验,平均分 4.26 在玩具模型中相当于每句话的普适度 0.852;两句签诗式句子只携带 1.380 比特,一句具体陈述却有 6.508 比特。 (https://5dlibrary.com/answers/zh/moon-block-landing-probability-three-holy-casts-barnum-verses/c04/) - Ch.5 Matching After the Fact: Ambiguous Clauses, Time Windows and Selective Memory — Chapter 5 shows how ambiguous clauses, open deadlines and selective memory multiply, so that 0.5330 of simulated readers sincerely remember that most lots came true. (https://5dlibrary.com/answers/en/moon-block-landing-probability-three-holy-casts-barnum-verses/c05/) 第5章 事后对上:多义条文、时间窗与选择性记忆 — 第五章说明多义条文、无期限与选择性记忆的乘积,使 0.5330 的模拟读者真诚地记得「大多对上了」。 (https://5dlibrary.com/answers/zh/moon-block-landing-probability-three-holy-casts-barnum-verses/c05/) - Ch.6 The Rite as a Protocol: Restart Rules, Stopping Rules and the Evenness of the Lot Tube — Chapter 6 reads the rite as a protocol, shows that the final share of confirmations means nothing when restarts are allowed, and finds that a worn lot tube is caught by an audit of 2400 confirmations with chance only 0.650. (https://5dlibrary.com/answers/en/moon-block-landing-probability-three-holy-casts-barnum-verses/c06/) 第6章 仪轨即协议:重掷规则、停止规则与签筒均匀性 — 第六章把仪轨读成协议,指出允许重来时最终确认比例不说明任何问题,并说明一个磨损的签筒在 2400 次确认时也只有 0.650 的机会被审计发现。 (https://5dlibrary.com/answers/zh/moon-block-landing-probability-three-holy-casts-barnum-verses/c06/) - Ch.7 Classical Sources East and West: the Entry on Divination Cups in the Yanfanlu and Cicero's De Divinatione — Chapter 7 examines the entry on divination cups in Cheng Dachang's Yanfanlu and Cicero's De Divinatione, setting the illustrative Venus throw of four knucklebones, 0.0384, beside three holy casts. (https://5dlibrary.com/answers/en/moon-block-landing-probability-three-holy-casts-barnum-verses/c07/) 第7章 东西方原典考辨:《演繁露》杯珓条与西塞罗《论占卜》 — 第七章考辨程大昌《演繁露》杯珓条与西塞罗《论占卜》,以距骨维纳斯一掷 0.0384 的示意值对照三圣杯。 (https://5dlibrary.com/answers/zh/moon-block-landing-probability-three-holy-casts-barnum-verses/c07/) - Ch.8 Closing the Loop: Randomness Statements, a Controlled Experiment on Satisfaction and a Pre-Registered Protocol — Chapter 8 gives a randomness statement, a controlled experiment on satisfaction and universality, a pre-registered protocol and the Moon Block Probability Simulator. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/moon-block-landing-probability-three-holy-casts-barnum-verses/c08/) 第8章 实证闭环:随机性声明、满意度对照实验与预注册方案 — 第八章给出随机性声明、满意度与普适性对照实验、预注册方案与小工具「筊杯概率模拟器」。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/moon-block-landing-probability-three-holy-casts-barnum-verses/c08/) ## 5DL-B-226 · Ideomotor Micro-Movements in Spirit Writing, Random-Walk Stylus Trajectories and Stylometric Attribution of Revealed Texts 中文书名:扶乩意动效应微肌电、乩笔轨迹随机游走与乩文语言模型溯源母典 DOI: pending Cite: Fan, Y. (2026). Ideomotor Micro-Movements in Spirit Writing, Random-Walk Stylus Trajectories and Stylometric Attribution of Revealed Texts (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative and its revealed texts are synthetic demonstrations written for it; all calculations are generated by its programs. Chapter 5 demonstrates three attribution methods on a synthetic corpus, character-bigram cosine, Burrows's Delta and naive Bayes; all six revealed texts are attributed correctly, while the naive-Bayes log odds of 8.92 to 21.37 are systematically too large. Abstract: People who take part in spirit writing, table-turning and the Ouija board widely and sincerely believe that the instrument is not moved by them and that the words it writes come from beyond them. This book tests the step from "I do not feel myself pushing" to "the force does not come from me" and splits its counterintuitive claim into three layers: the movement of the spirit-writing stylus comes from the holders' small unnoticed muscular activity; legibility falls markedly once visual feedback is lost; and the characters of revealed texts resemble the holders' own style. The project blueprint stated all three layers as facts; after checking the original literature, the book restates the last two as modelled hypotheses awaiting tests and notes that Faraday tested table-turning in 1853, not the stylus. The book analyses only structure, probability and text, makes no personal reading of anyone's fate and respects the religious and folk practices concerned. Its parameters are illustrative and its revealed texts are synthetic demonstrations written for it; all calculations are generated by its programs. Chapter 1, after Carpenter and Jacobson, shows that an ideomotor burst only 0.05 above the noise floor is detected 0.925 of the time in a 2.0-second window. Chapter 2 states Faraday's card and indicator experiments directly: written as a feedback gain, an indicator gain of 0.75 cuts the share of sittings in which the table moves within a minute to 0.087; the 65 % against 50 % reported by Gauchou and colleagues in 2012 is a Cohen's h of 0.3047, and a replication needs at least 85 questions per condition. Chapter 3 writes shared holding as a biased random walk: four hands reach the target within the time limit with probability 0.853, one hand with only 0.120. Chapter 4 derives the effect of a blindfold from closed-loop and open-loop control: the legibility of a nine-stroke character falls from 0.9936 to 0.1490. Chapter 5 demonstrates three attribution methods on a synthetic corpus, character-bigram cosine, Burrows's Delta and naive Bayes; all six revealed texts are attributed correctly, while the naive-Bayes log odds of 8.92 to 21.37 are systematically too large. Chapter 6 tests short texts and open candidate sets: at 300 characters and a threshold of 0.45, writers missing from the list are wrongly named only 0.043 of the time while 0.940 of in-list texts are still named correctly. Chapter 7 examines Xu Dishan's 1941 study of spirit writing and Faraday's reports and turns them into an out-of-vocabulary test and a sealed-question test. Chapter 8 gives an integrity statement for spiritual products, exercises for teaching the ideomotor effect, the distilled Style Provenance Comparator and three pre-registered tests: a blindfold legibility test needs 15 to 44 sessions per condition, and with three candidates 13 revealed texts suffice to test attribution. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:扶乩、转桌与通灵板的参与者普遍真诚地相信:器具的运动不是自己推动的,写出的文字来自自己之外。本书检验从「感觉不到自己用力」到「力不来自自己」的这一步推断,并把反常识命题分为三层:乩笔的运动来自扶笔者不自觉的微小肌肉活动;失去视觉反馈后乩文的可读性显著下降;乩文的用字与扶笔者本人的文风相近。立项蓝图把三层都写成既定事实;核对原始文献后,本书把后两层修正为有模型、待检验的假说,并说明法拉第 1853 年检验的是转桌而不是乩笔。本书只分析结构、概率与文本,不做任何个人占断,尊重相关的宗教与民俗实践。书中参数为示意,乩文语料全部为自拟合成示范;全部计算由书中程序生成。 第一章从卡朋特的意动概念与雅各布森的肌电记录出发,说明幅度只高出底噪 0.05 的意动爆发在 2.0 秒窗口里的检出率可达 0.925。第二章正面陈述法拉第的纸板与指示器实验:把指示器写成反馈增益,增益为 0.75 时一分钟内桌子会动的比例降到 0.087;高修等人 2012 年报告的 65 % 对 50 % 相当于科恩 h 0.3047,重复研究每种条件至少需要 85 道题。第三章把多手扶笔写成有偏随机游走,四只手在限时内到达目标的概率是 0.853,单手只有 0.120。第四章用闭环与开环控制推导蒙眼的影响,九笔字的可读率从 0.9936 降到 0.1490。 第五章在自拟语料上演示字二元组余弦、伯罗斯 Delta 与朴素贝叶斯三种文风溯源方法,六篇乩文全部归属正确,而朴素贝叶斯的对数优势比 8.92 到 21.37 系统性偏大。第六章检验短文本与开放集:300 个字、阈值 0.45 时,名单外作者被错扣的比例降到 0.043,名单内仍有 0.940 被正确指认。第七章考辨许地山 1941 年的《扶箕迷信底研究》与法拉第的报告,并把二者写成超出词表率检验与密封问题检验。第八章给出灵性产品诚信声明、意动体验教学、蒸馏小工具「文风溯源比对器」与三项预注册检验:蒙眼可读性检验每种条件需要 15 到 44 场,三位候选人时 13 篇乩文即可检验文风归属。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Ideomotor Action and Micro-EMG: Carpenter's Term and the Invisible Command — Chapter 1, after Carpenter and Jacobson, shows that an ideomotor burst only 0.05 above the noise floor is detected 0.925 of the time in a 2.0-second window. (https://5dlibrary.com/answers/en/spirit-writing-ideomotor-random-walk-text-attribution/c01/) 第1章 意动效应与微肌电:卡朋特的命名与看不见的指令 — 第一章从卡朋特的意动概念与雅各布森的肌电记录出发,说明幅度只高出底噪 0.05 的意动爆发在 2.0 秒窗口里的检出率可达 0.925。 (https://5dlibrary.com/answers/zh/spirit-writing-ideomotor-random-walk-text-attribution/c01/) - Ch.2 Faraday's Indicator: the Original Evidence and Its Effect Sizes — Chapter 2 states Faraday's card and indicator experiments directly: written as a feedback gain, an indicator gain of 0.75 cuts the share of sittings in which the table moves within a minute to 0.087; the 65 % against 50 % reported by Gauchou and colleagues in 2012 is a Cohen's h of 0.3047, and a replication needs at least 85 questions per condition. (https://5dlibrary.com/answers/en/spirit-writing-ideomotor-random-walk-text-attribution/c02/) 第2章 法拉第的指示器:反常识命题的原始证据与效应量 — 第二章正面陈述法拉第的纸板与指示器实验:把指示器写成反馈增益,增益为 0.75 时一分钟内桌子会动的比例降到 0.087;高修等人 2012 年报告的 65 % 对 50 % 相当于科恩 h 0.3047,重复研究每种条件至少需要 85 道题。 (https://5dlibrary.com/answers/zh/spirit-writing-ideomotor-random-walk-text-attribution/c02/) - Ch.3 The Biased Random Walk of Many Small Forces: Drift, Noise and the Leading Hand — Chapter 3 writes shared holding as a biased random walk: four hands reach the target within the time limit with probability 0.853, one hand with only 0.120. (https://5dlibrary.com/answers/en/spirit-writing-ideomotor-random-walk-text-attribution/c03/) 第3章 多人微力的有偏随机游走:漂移、噪声与主笔 — 第三章把多手扶笔写成有偏随机游走,四只手在限时内到达目标的概率是 0.853,单手只有 0.120。 (https://5dlibrary.com/answers/zh/spirit-writing-ideomotor-random-walk-text-attribution/c03/) - Ch.4 Closed Loop and Open Loop: Why Revealed Texts Become Hard to Read under a Blindfold — Chapter 4 derives the effect of a blindfold from closed-loop and open-loop control: the legibility of a nine-stroke character falls from 0.9936 to 0.1490. (https://5dlibrary.com/answers/en/spirit-writing-ideomotor-random-walk-text-attribution/c04/) 第4章 闭环与开环:蒙眼之后乩文为何难认 — 第四章用闭环与开环控制推导蒙眼的影响,九笔字的可读率从 0.9936 降到 0.1490。 (https://5dlibrary.com/answers/zh/spirit-writing-ideomotor-random-walk-text-attribution/c04/) - Ch.5 Stylometric Attribution: Character-Bigram Cosine, Burrows's Delta and Naive Bayes — Chapter 5 demonstrates three attribution methods on a synthetic corpus, character-bigram cosine, Burrows's Delta and naive Bayes; all six revealed texts are attributed correctly, while the naive-Bayes log odds of 8.92 to 21.37 are systematically too large. (https://5dlibrary.com/answers/en/spirit-writing-ideomotor-random-walk-text-attribution/c05/) 第5章 文风溯源:字二元组余弦、Burrows Delta 与朴素贝叶斯 — 第五章在自拟语料上演示字二元组余弦、伯罗斯 Delta 与朴素贝叶斯三种文风溯源方法,六篇乩文全部归属正确,而朴素贝叶斯的对数优势比 8.92 到 21.37 系统性偏大。 (https://5dlibrary.com/answers/zh/spirit-writing-ideomotor-random-walk-text-attribution/c05/) - Ch.6 The Limits of Attribution: Text Length, Open Sets and Register Confounds — Chapter 6 tests short texts and open candidate sets: at 300 characters and a threshold of 0.45, writers missing from the list are wrongly named only 0.043 of the time while 0.940 of in-list texts are still named correctly. (https://5dlibrary.com/answers/en/spirit-writing-ideomotor-random-walk-text-attribution/c06/) 第6章 溯源的边界:文本长度、开放集与语体混淆 — 第六章检验短文本与开放集:300 个字、阈值 0.45 时,名单外作者被错扣的比例降到 0.043,名单内仍有 0.940 被正确指认。 (https://5dlibrary.com/answers/zh/spirit-writing-ideomotor-random-walk-text-attribution/c06/) - Ch.7 Classics East and West: Xu Dishan's Study of Spirit Writing and Faraday's Table-Turning Experiments — Chapter 7 examines Xu Dishan's 1941 study of spirit writing and Faraday's reports and turns them into an out-of-vocabulary test and a sealed-question test. (https://5dlibrary.com/answers/en/spirit-writing-ideomotor-random-walk-text-attribution/c07/) 第7章 东西方原典考辨:许地山的扶箕研究与法拉第的转桌实验 — 第七章考辨许地山 1941 年的《扶箕迷信底研究》与法拉第的报告,并把二者写成超出词表率检验与密封问题检验。 (https://5dlibrary.com/answers/zh/spirit-writing-ideomotor-random-walk-text-attribution/c07/) - Ch.8 Closing the Loop: Integrity Statements for Spiritual Products, Teaching the Ideomotor Effect and Pre-Registered Tests — Chapter 8 gives an integrity statement for spiritual products, exercises for teaching the ideomotor effect, the distilled Style Provenance Comparator and three pre-registered tests: a blindfold legibility test needs 15 to 44 sessions per condition, and with three candidates 13 revealed texts suffice to test attribution. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/spirit-writing-ideomotor-random-walk-text-attribution/c08/) 第8章 实证闭环:灵性产品诚信声明、意动体验教学与预注册方案 — 第八章给出灵性产品诚信声明、意动体验教学、蒸馏小工具「文风溯源比对器」与三项预注册检验:蒙眼可读性检验每种条件需要 15 到 44 场,三位候选人时 13 篇乩文即可检验文风归属。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/spirit-writing-ideomotor-random-walk-text-attribution/c08/) ## 5DL-B-227 · Spherical Geometry of Western House Systems, Polar-Circle Singularities and Non-Astrological Explanations of Birth-Season Effects 中文书名:西洋本命盘宫位制球面几何、极圈奇点退化与出生季节效应的非占星解释母典 DOI: pending Cite: Fan, Y. (2026). Spherical Geometry of Western House Systems, Polar-Circle Singularities and Non-Astrological Explanations of Birth-Season Effects (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: Two views of the Western natal chart are widespread: that date, time and place of birth fix one unique chart, and that season of birth bears on a person's life, as shown by the many athletes of certain signs. This book keeps the phenomena behind both views and overturns the explanations attached to them. First, a house system is only a coordinate convention: a Placidus cusp is the root of an equation that requires every degree of the ecliptic to have a semi-arc, a premise that fails north of latitude 66.5607 degrees. Second, the link between birth month and athletes' selection comes from the relative age effect produced by the cut-off dates of school entry and selection, and the peak month moves when the cut-off moves. After checking the original literature, the book corrects the blueprint's "Placidus is undefined inside the polar circle" to "undefined for part of the day inside the polar circle, a share that grows with latitude". The book analyses only structure, probability and text and makes no personal reading of anyone. Its parameters are illustrative; all calculations are generated by its programs. Chapters 1 to 4 form the geometric part. In Tromsø only 0.158 of the day yields a complete Placidus chart, and less than a tenth of a degree beyond the circle the share of complete charts falls to 0.892; at latitude 66.5 the ascendant moves at up to 93.948 degrees a minute; in London, switching between Placidus and equal houses changes a planet's house number in 0.417 of placements. Chapters 5 and 6 form the statistical part. Two Capricorns born on 1 January and 31 December differ by a factor of 2.020 in their chance of selection under a 1 January cut-off; in birth-month data on multiple sclerosis the ratio of the May and November observed-to-expected ratios is 1.193, and the seasonal account predicts a six-month shift between hemispheres; with the wrong baseline a sample of 20000 manufactures a spurious sign effect with probability 0.943; and the Gauquelin Mars effect vanishes in independent tests. Chapter 7 examines the Kaiyuan Zhanjing and Ptolemy's Tetrabiblos: an omen with two outcomes and a three-year window is matched afterwards with probability 0.882, and Ptolemy's seasonal qualities of the signs all fail at 35 degrees south, where the insolation of the two hemispheres correlates at −0.9958. Chapter 8 gives a high-latitude patch for chart services, a birth-season explanation card, the distilled High-Latitude House Checker and two pre-registered tests: a three-choice matching test needs 111 charts at a true success rate of 0.45, and a cut-off-shift test needs 512 selected players per federation. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:关于西洋本命盘,流行的看法有两条:出生的日期、时刻与地点决定一张唯一的星盘;出生季节与人的际遇有关,证据是某些星座的运动员特别多。本书保留这两条背后的现象,推翻附在现象上的解释。第一,宫位制只是一种坐标约定,Placidus 宫头是一个方程的根,这个方程要求每个黄道度数都有半弧,而在纬度 66.5607 度以北这一前提不再成立。第二,出生月份与运动员入选率的关联来自入学与选拔截止日造成的相对年龄效应,截止日不同,峰值月份随之平移。核对原始文献后,本书把立项蓝图中「Placidus 在极圈内无定义」修正为「在极圈内一天中的部分时刻无定义,比例随纬度上升」。本书只分析结构、概率与文本,不做任何个人占断。书中参数为示意;全部计算由书中程序生成。 第一至四章是几何部分:特罗姆瑟一天中只有 0.158 的时刻能排出完整的 Placidus 星盘,越过极圈不到十分之一度时完整星盘的比例降到 0.892;纬度 66.5 度时上升点最快每分钟移动 93.948 度;在伦敦,同一颗行星换用 Placidus 与等宫制,宫位编号不同的比例为 0.417。第五、六章是统计部分:同为摩羯座的 1 月 1 日与 12 月 31 日出生者,在 1 月 1 日截止的选拔中入选概率相差 2.020 倍;多发性硬化出生月份数据中 5 月与 11 月观察期望比之比为 1.193,季节解释预言南北半球相差六个月;用错基线时,20000 人的样本有 0.943 的概率造出虚假的星座效应;戈克兰火星效应在独立检验中消失。 第七章考辨《开元占经》与托勒密《占星四书》:两种结局、三年窗口的占辞被事后对上的概率为 0.882,托勒密的星座季节性质在南纬 35 度全部失效,两半球辐照度相关系数为 −0.9958。第八章给出排盘服务的高纬防错补丁、出生季节解释卡、蒸馏小工具「高纬宫位校验器」与两项预注册检验:三选一配对检验在真实正确率 0.45 时需要 111 张星盘,截止日平移检验每个协会需要 512 名入选者。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Two Counterintuitive Claims: the Polar Singularity and the Moving Cut-off — Chapters 1 to 4 form the geometric part. In Tromsø only 0.158 of the day yields a complete Placidus chart, and less than a tenth of a degree beyond the circle the share of complete charts falls to 0.892; at latitude 66.5 the ascendant moves at up to 93.948 degrees a minute; in London, switching between Placidus and equal houses changes a planet's house number in 0.417 of placements. (https://5dlibrary.com/answers/en/western-house-systems-polar-singularity-birth-season-effects/c01/) 第1章 两个反常识命题:极圈奇点与截止日平移 — 第一至四章是几何部分:特罗姆瑟一天中只有 0.158 的时刻能排出完整的 Placidus 星盘,越过极圈不到十分之一度时完整星盘的比例降到 0.892;纬度 66.5 度时上升点最快每分钟移动 93.948 度;在伦敦,同一颗行星换用 Placidus 与等宫制,宫位编号不同的比例为 0.417。 (https://5dlibrary.com/answers/zh/western-house-systems-polar-singularity-birth-season-effects/c01/) - Ch.2 Three Great Circles: the Equator, the Ecliptic, the Horizon and the Ascendant Formula — This chapter reduces the inputs of the house calculation to a latitude and a sidereal time, writes the three great circles of ecliptic, equator and horizon as two rotation formulas, and uses the three checks of p21 to confirm their meaning: a round-trip error of 1.1e-13 degrees and an ascendant altitude of 8.0e-14 degrees. At latitude 70, for 0.181 of the day the formula returns the western intersection. Program p22 shows that at latitude 66.5 the ascendant moves at up to 93.948 degrees a minute, and that beyond the polar circle it runs backward and jumps 2 times a day. The midheaven and the ascendant form the skeleton of every quadrant house system; the next chapter builds the Placidus trisection of semi-arcs on that skeleton and watches it lose its definition near the polar circle. (https://5dlibrary.com/answers/en/western-house-systems-polar-singularity-birth-season-effects/c02/) 第2章 三个大圆:赤道、黄道、地平与上升点公式 — 本章把宫位计算的输入还原为纬度与恒星时,把黄道、赤道、地平三个大圆写成两个旋转公式,并用 p21 的三项检查确认它们的含义:往返误差 1.1e-13 度,上升点高度 8.0e-14 度。纬度 70 度时,有 0.181 的时刻公式给出的是西方交点。p22 显示上升点在纬度 66.5 度最快每分钟走 93.948 度,越过极圈后会倒行并每天跳变 2 次。天顶点与上升点是一切象限宫位制的骨架;下一章在这副骨架上搭建 Placidus 的半弧三分,看它在极圈附近怎样失去定义。 (https://5dlibrary.com/answers/zh/western-house-systems-polar-singularity-birth-season-effects/c02/) - Ch.3 Trisecting the Placidus Semi-Arcs: Equations, Roots and Failure inside the Polar Circle — A Placidus cusp is the root of an equation whose unknown appears on both sides. Program p31 shows that in London scanning and iteration give the same four cusps, while in Tromsø iteration fails in its first round and the scan finds both missing cusps and surplus ones. Program p32 shows the share of complete charts at 1.000 at latitude 66.5, falling to 0.892 at 66.6, 0.147 at 70 and 0.000 at 80. Program p33 shows that south of the polar circle the cusp that runs out of control is the ascendant, at up to 532.29 degrees a minute at 66.55, while the two intermediate cusps still move at only 0.41 and 0.82. The next chapter applies the same tests to Koch, Porphyry, equal and whole-sign houses and asks how far house numbers differ between conventions. (https://5dlibrary.com/answers/en/western-house-systems-polar-singularity-birth-season-effects/c03/) 第3章 Placidus 半弧三分:方程、根与极圈内的失效 — Placidus 的宫头是一个未知数出现在两边的方程的根。p31 显示,在伦敦扫描与迭代给出相同的四个宫头;在特罗姆瑟,迭代第一轮即失败,扫描则找到缺失的宫头与多余的宫头。p32 显示完整星盘的比例在 66.5 度为 1.000,66.6 度降到 0.892,70 度为 0.147,80 度为 0.000。p33 显示极圈以南真正失控的是上升点,在 66.55 度最快每分钟 532.29 度,而两个中间宫头仍只有 0.41 度与 0.82 度。下一章把同样的检验施加到 Koch、Porphyry、等宫制与整宫制,看宫位编号在不同约定之间相差多少。 (https://5dlibrary.com/answers/zh/western-house-systems-polar-singularity-birth-season-effects/c03/) - Ch.4 Koch, Porphyry, Equal and Whole-Sign Houses: House Numbers Are Conventions, Not Observations — Program p41 shows that at one moment in London the eleventh cusp of five house systems ranges from 120 to 152 degrees. Koch starts failing at latitude 66.6 and is undefined for 0.366 of the day at 70 and 0.734 at 80, while north of the circle Porphyry produces reversed quadrants in which the arc from the midheaven to the ascendant exceeds 180 degrees. Program p42 shows that, in London, switching between Placidus and equal houses changes the house number of a planet in 0.417 of placements, and at latitude 65 in 0.550; between equal and whole-sign houses the share is about one half at every latitude. House numbers are conventions, not observations. The next chapter leaves the sphere and turns to the book's second claim, the link between season of birth and selection. (https://5dlibrary.com/answers/en/western-house-systems-polar-singularity-birth-season-effects/c04/) 第4章 Koch、Porphyry、等宫与整宫:宫位编号是约定,不是观测 — p41 显示,在伦敦同一时刻,五种宫位制的第十一宫宫头从 120 度到 152 度不等;Koch 在纬度 66.6 度开始失效,70 度时无定义的比例为 0.366,80 度为 0.734;Porphyry 在极圈以北会出现天顶点到上升点弧长超过 180 度的反向象限。p42 显示,同一颗行星在伦敦换用 Placidus 与等宫制,宫位编号不同的比例是 0.417,在纬度 65 度是 0.550;等宫制与整宫制之间在各纬度都约为一半。宫位编号是约定,不是观测。下一章离开球面,转向本书的第二个命题:出生季节与入选率的关系。 (https://5dlibrary.com/answers/zh/western-house-systems-polar-singularity-birth-season-effects/c04/) - Ch.5 The Relative Age Effect: Dose-Response, Displaced Signs and Sample Size — Chapters 5 and 6 form the statistical part. Two Capricorns born on 1 January and 31 December differ by a factor of 2.020 in their chance of selection under a 1 January cut-off; in birth-month data on multiple sclerosis the ratio of the May and November observed-to-expected ratios is 1.193, and the seasonal account predicts a six-month shift between hemispheres; with the wrong baseline a sample of 20000 manufactures a spurious sign effect with probability 0.943; and the Gauquelin Mars effect vanishes in independent tests. (https://5dlibrary.com/answers/en/western-house-systems-polar-singularity-birth-season-effects/c05/) 第5章 相对年龄效应:剂量反应、星座错位与样本量 — 第五、六章是统计部分:同为摩羯座的 1 月 1 日与 12 月 31 日出生者,在 1 月 1 日截止的选拔中入选概率相差 2.020 倍;多发性硬化出生月份数据中 5 月与 11 月观察期望比之比为 1.193,季节解释预言南北半球相差六个月;用错基线时,20000 人的样本有 0.943 的概率造出虚假的星座效应;戈克兰火星效应在独立检验中消失。 (https://5dlibrary.com/answers/zh/western-house-systems-polar-singularity-birth-season-effects/c05/) - Ch.6 Other Non-Astrological Mechanisms of Birth Season, and the Mars Effect Controversy — Birth-season effects have at least three classes of non-astrological mechanism: the seasonality of the environment, the seasonality of births, and institutional cut-off dates. Program p61 uses Willer's counts to obtain a ratio of 1.193 between the May and November ratios of observed to expected births, and uses an illustrative model to show that the seasonal account predicts troughs six months apart in the two hemispheres, whereas the signs keep the same dates in both. Program p62 shows that with the wrong baseline a sample of 20000 manufactures a spurious sign effect with probability 0.943. Program p63 shows that the Gauquelin effect, a deviation of 3.23 in the Comité Para sample, vanishes in the American and French tests, and that leaving out 0.181 of the champions outside the key sectors is enough to produce a share of 0.20. The next chapter returns to antiquity and examines which statements in the Kaiyuan Zhanjing and Ptolemy's Tetrabiblos can be tested against today's data. (https://5dlibrary.com/answers/en/western-house-systems-polar-singularity-birth-season-effects/c06/) 第6章 出生季节的其他非占星机制与火星效应之争 — 出生季节效应的非占星机制至少有三类:环境的季节性、人口的出生季节性与制度的截止日。p61 用维勒等人的计数算出 5 月与 11 月的观察期望比之比为 1.193,并用示意模型说明季节解释预言两个半球的低谷相差六个月,而星座在两个半球日期相同。p62 显示用错基线时,20000 人的样本有 0.943 的概率造出虚假的星座效应。p63 显示戈克兰效应在帕拉委员会样本中偏差 3.23,在美国与法国的检验中消失,而漏掉 0.181 的非关键区冠军就足以造出一个 0.20 的比例。下一章回到古代,考辨《开元占经》与托勒密《占星四书》中可以被今天的数据检验的说法。 (https://5dlibrary.com/answers/zh/western-house-systems-polar-singularity-birth-season-effects/c06/) - Ch.7 Eastern and Western Sources Examined: the Kaiyuan Zhanjing and Ptolemy's Tetrabiblos — Chapter 7 examines the Kaiyuan Zhanjing and Ptolemy's Tetrabiblos: an omen with two outcomes and a three-year window is matched afterwards with probability 0.882, and Ptolemy's seasonal qualities of the signs all fail at 35 degrees south, where the insolation of the two hemispheres correlates at −0.9958. (https://5dlibrary.com/answers/en/western-house-systems-polar-singularity-birth-season-effects/c07/) 第7章 东西方原典考辨:《开元占经》与托勒密《占星四书》 — 第七章考辨《开元占经》与托勒密《占星四书》:两种结局、三年窗口的占辞被事后对上的概率为 0.882,托勒密的星座季节性质在南纬 35 度全部失效,两半球辐照度相关系数为 −0.9958。 (https://5dlibrary.com/answers/zh/western-house-systems-polar-singularity-birth-season-effects/c07/) - Ch.8 Closing the Loop: a High-Latitude Patch for Chart Services, a Birth-Season Card and Pre-Registered Tests — Chapter 8 gives a high-latitude patch for chart services, a birth-season explanation card, the distilled High-Latitude House Checker and two pre-registered tests: a three-choice matching test needs 111 charts at a true success rate of 0.45, and a cut-off-shift test needs 512 selected players per federation. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/western-house-systems-polar-singularity-birth-season-effects/c08/) 第8章 实证闭环:排盘服务的高纬防错补丁、季节解释卡与预注册检验 — 第八章给出排盘服务的高纬防错补丁、出生季节解释卡、蒸馏小工具「高纬宫位校验器」与两项预注册检验:三选一配对检验在真实正确率 0.45 时需要 111 张星盘,截止日平移检验每个协会需要 512 名入选者。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/western-house-systems-polar-singularity-birth-season-effects/c08/) ## 5DL-B-228 · The 98% Lexical-Coverage Threshold, Zipfian Frequency Curves and Fast-Track Limits for Malay and Hokkien Learners 中文书名:二语词汇覆盖率九十八阈值、齐普夫频率曲线与马来语闽南语速成边界母典 DOI: pending Cite: Fan, Y. (2026). The 98% Lexical-Coverage Threshold, Zipfian Frequency Curves and Fast-Track Limits for Malay and Hokkien Learners (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative and all calculations are generated by its programs. It is general reference material for owners and teachers, not business advice for any person or firm, and it promises no fluency from any course within any period. Abstract: Language courses like to say that two thousand common words let you understand eighty per cent of any text, so that everyday communication is solved. This book tests that claim and states its counterintuitive thesis in a form that can be computed and refuted: the coverage of a word list is a count, while comprehension is an ability; according to Hu and Nation (2000), Nation (2006) and Laufer and Ravenhorst-Kalovski (2010), comfortable unassisted reading needs about 95 % to 98 % coverage, and getting from two thousand word families into that range multiplies the vocabulary by more than four. After checking the original papers the book corrects its project blueprint: 98 % was estimated by regression rather than tested directly, Schmitt and colleagues (2011) found a slope rather than a cliff, and the replication by Kremmel and colleagues (2023) could not fully reproduce the original results; what is expensive is not "the last 2 %" but every point beyond 95 %. Its parameters are illustrative and all calculations are generated by its programs. It is general reference material for owners and teachers, not business advice for any person or firm, and it promises no fluency from any course within any period. Chapter 1 translates coverage into the density of unknown words on a page and computes effect sizes separating a cliff from a slope; from 95 % to 98 % each coverage point costs about 1166.7 word families, 7.0 times the earlier price, and going from two thousand families to 98 % multiplies the vocabulary by 4.0 to 4.5. Chapter 2 derives the closed form for harmonic numbers and a ratio formula for power-law tails, and fits a Zipf-Mandelbrot reference curve through two published anchors, on which two thousand families cover about 0.8618. Chapter 3 counts forms, lemmas and families and demonstrates Malay affix rules. Chapter 4 shows that a list averaging 98 % still leaves 0.396 of pages below 98 %, and that a customer-service domain list of sixteen hundred items covers at most 0.7905. Chapter 5 schedules review with half-life regression: a "ninety-day fast track" reaches a coverage of only 0.8050, and half an hour a day takes 500 days to reach eight thousand families. Chapter 6 computes that Malay affixes and loanwords cut the time to 95 % by a factor of 3.78, while second readings alone add 47 days to a Taiwanese Hokkien listening goal. Chapter 7 weighs the Erya and the Orbis Pictus against frequency lists. Chapter 8 gives a procedure for building, measuring and re-measuring word lists for service teams in Malaysia and Taiwan, the distilled Word-List Coverage Calculator (in its demonstration the affix rules raise coverage from 0.6667 to 0.8667), three pre-registered tests and the results that would refute the book's central claim. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:语言课程常说「掌握两千个常用词就能读懂八成文字,日常交流从此无忧」。本书检验这句话,并把反常识命题写成可以计算、可以推翻的形式:词表的覆盖率是计数,理解是能力;按胡与内申 2000 年、内申 2006 年与劳弗等人 2010 年的研究,舒适的独立阅读约需 95 % 到 98 % 的覆盖率,从两千词族到这个区间,词汇量要乘以四倍以上。核对原始文献后,本书修正了立项蓝图:98 % 是回归推算的数而非直接测试的条件,施密特等人 2011 年看到的是斜坡而不是断崖,克雷梅尔等人 2023 年的重复研究未能完全重现原结果;昂贵的不是「最后 2 %」,而是 95 % 之后的每一个点。书中参数为示意,全部计算由书中程序生成;本书是非个人化的经营与教学参考,不是针对任何个人或企业的经营建议,也不承诺任何课程能在某个期限内带来流利。 第一章把覆盖率译成纸面上的生词密度,并计算断崖与斜坡两种剂量—反应曲线的效应量;从九成五到九成八,每个覆盖点约需 1166.7 个词族,是前一段价格的 7.0 倍,从两千词族到九成八覆盖,词汇量要乘以 4.0 到 4.5。第二章推导调和数的闭式近似与幂律尾部的比值公式,并用两个已发表锚点拟合齐普夫—曼德博参考曲线,两千词族约覆盖 0.8618。第三章讨论词形、词元与词族,演示一套只认词典词根的马来语词缀规则。第四章说明平均 98 % 的词表仍有 0.396 的页面低于 98 %,而一千六百个词的客服领域词表最多覆盖 0.7905。第五章用半衰期回归排复习表:「九十天速成」只能走到 0.8050 的覆盖率,每天约半小时要 500 天才到八千词族。 第六章计算马来语的词缀与借词能把达到 95 % 的时间缩短为 3.78 分之一,台语的文白异读一项就让听力目标多出 47 天。第七章考辨《尔雅》与夸美纽斯的《世界图解》,比较按主题与按频率编排的词表。第八章给出马来西亚与台湾客服团队的建表、测量与复测流程,蒸馏小工具「词表覆盖率计算器」(示范中词缀规则把覆盖率从 0.6667 提到 0.8667),以及三项预注册检验与推翻本书核心命题的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Ninety-Eight Per Cent: the Original Evidence for a Coverage Threshold and Its Limits — Chapter 1 translates coverage into the density of unknown words on a page and computes effect sizes separating a cliff from a slope; from 95 % to 98 % each coverage point costs about 1166.7 word families, 7.0 times the earlier price, and going from two thousand families to 98 % multiplies the vocabulary by 4.0 to 4.5. (https://5dlibrary.com/answers/en/lexical-coverage-98-threshold-zipf-malay-hokkien/c01/) 第1章 百分之九十八:覆盖率阈值的原始证据与它的边界 — 第一章把覆盖率译成纸面上的生词密度,并计算断崖与斜坡两种剂量—反应曲线的效应量;从九成五到九成八,每个覆盖点约需 1166.7 个词族,是前一段价格的 7.0 倍,从两千词族到九成八覆盖,词汇量要乘以 4.0 到 4.5。 (https://5dlibrary.com/answers/zh/lexical-coverage-98-threshold-zipf-malay-hokkien/c01/) - Ch.2 The Zipf Curve: Closed Forms for Cumulative Coverage and the Inverse Problem of List Size — Chapter 2 derives the closed form for harmonic numbers and a ratio formula for power-law tails, and fits a Zipf-Mandelbrot reference curve through two published anchors, on which two thousand families cover about 0.8618. (https://5dlibrary.com/answers/en/lexical-coverage-98-threshold-zipf-malay-hokkien/c02/) 第2章 齐普夫曲线:累积覆盖率的解析式与词表大小的反解 — 第二章推导调和数的闭式近似与幂律尾部的比值公式,并用两个已发表锚点拟合齐普夫—曼德博参考曲线,两千词族约覆盖 0.8618。 (https://5dlibrary.com/answers/zh/lexical-coverage-98-threshold-zipf-malay-hokkien/c02/) - Ch.3 How to Count a Word: Forms, Lemmas, Families and Malay Affixes — Between "two thousand words" and "eight thousand families" lies a unit of counting that is usually left out. p31 shows that on the same reference curve the forms needed for 98 % can be between 1.95 and 4.62 times the families; and a 98 % counted in families falls to between 0.7389 and 0.8167 when a learner has only an even chance of recognising derived forms. Malay is rich in affixes, and p32 uses a rule set that accepts only lexicon roots to assign thirty-five common forms to sixteen families; Hokkien has few affixes, the difficulty moves to segmentation, and the Tâi-lô hyphen is what makes counting repeatable. The next chapter leaves averages behind and looks at text page by page: what does an average coverage of 98 % look like on each single page? (https://5dlibrary.com/answers/en/lexical-coverage-98-threshold-zipf-malay-hokkien/c03/) 第3章 一个词怎么数:词形、词元、词族与马来语的词缀 — 「两千词」与「八千词族」之间隔着一个常被省略的计数单位。p31 表明,同一条参考曲线上,98 % 所需的词形数可以是词族数的 1.95 倍到 4.62 倍;而以词族计数的 98 %,在学习者认出派生形式的把握只有一半时,会降到 0.7389 至 0.8167。马来语词缀丰富,p32 用一个只认词典词根的规则集,把三十五个常见词形归入十六个词族;闽南语词缀少,难点转到切分,台罗的连字符是让计数可以重复的办法。下一章离开平均数,去看一页一页的文字:平均 98 % 的覆盖率,落到每一页上会是什么样子? (https://5dlibrary.com/answers/zh/lexical-coverage-98-threshold-zipf-malay-hokkien/c03/) - Ch.4 Page by Page: How Coverage Varies Across Pages and Where a Narrow Domain List Stops — Chapter 4 shows that a list averaging 98 % still leaves 0.396 of pages below 98 %, and that a customer-service domain list of sixteen hundred items covers at most 0.7905. (https://5dlibrary.com/answers/en/lexical-coverage-98-threshold-zipf-malay-hokkien/c04/) 第4章 一页一页地看:覆盖率的逐页波动与窄领域词表的边界 — 第四章说明平均 98 % 的词表仍有 0.396 的页面低于 98 %,而一千六百个词的客服领域词表最多覆盖 0.7905。 (https://5dlibrary.com/answers/zh/lexical-coverage-98-threshold-zipf-malay-hokkien/c04/) - Ch.5 Forgetting and Review: Half-Life Regression and an Honest Timetable — Chapter 5 schedules review with half-life regression: a "ninety-day fast track" reaches a coverage of only 0.8050, and half an hour a day takes 500 days to reach eight thousand families. (https://5dlibrary.com/answers/en/lexical-coverage-98-threshold-zipf-malay-hokkien/c05/) 第5章 遗忘与复习:半衰期回归与一份诚实的时间表 — 第五章用半衰期回归排复习表:「九十天速成」只能走到 0.8050 的覆盖率,每天约半小时要 500 天才到八千词族。 (https://5dlibrary.com/answers/zh/lexical-coverage-98-threshold-zipf-malay-hokkien/c05/) - Ch.6 Malay and Hokkien: the Structural Limits of Fast-Track Promises — Chapter 6 computes that Malay affixes and loanwords cut the time to 95 % by a factor of 3.78, while second readings alone add 47 days to a Taiwanese Hokkien listening goal. (https://5dlibrary.com/answers/en/lexical-coverage-98-threshold-zipf-malay-hokkien/c06/) 第6章 马来语与闽南语:速成承诺的结构边界 — 第六章计算马来语的词缀与借词能把达到 95 % 的时间缩短为 3.78 分之一,台语的文白异读一项就让听力目标多出 47 天。 (https://5dlibrary.com/answers/zh/lexical-coverage-98-threshold-zipf-malay-hokkien/c06/) - Ch.7 Classical Sources East and West: the Grouping of the Erya and Comenius's Orbis Pictus — The Erya glosses rare synonyms with common words, an ancient word family grouped by meaning; the Orbis Pictus shows things in pictures and organises lessons by theme, lowering the cost of acquiring meaning. p71 shows that on general text a list built from whole themes falls far short of a frequency list, and that a hybrid arrangement wins only when text is highly concentrated on a theme (a theme share of 0.5 or more); p72 shows that pictures and glossaries change the price of each unknown word, not their number. Both classics remind us that there is meaning and context beyond coverage; coverage reminds them that frequency decides what to learn first. The final chapter gathers the book's calculations into a procedure that can be carried out: how to build word lists and measure coverage for customer-service teams in Malaysia and Taiwan, and what pre-registered tests could refute this book itself. (https://5dlibrary.com/answers/en/lexical-coverage-98-threshold-zipf-malay-hokkien/c07/) 第7章 东西方原典考辨:《尔雅》的类聚与夸美纽斯的《世界图解》 — 《尔雅》以常用词训释罕见同义词,是按意义归并的古代词族;《世界图解》以图示物、按主题组织课程,降低了获得意义的成本。p71 表明,在一般文字上按主题整块收词的词表远不如频率词表,只有当文本高度集中于某个主题(主题比例达到 0.5 以上)时,混合编排才胜出;p72 表明图画与词典改变的是每个生词的价格,而不是生词的数量。两部原典都在提醒我们,覆盖率之外还有意义与情境;而覆盖率提醒它们,先学什么由频率决定。最后一章把全书的计算收拢成一套可以执行的流程:怎样为马来西亚与台湾的客服团队建词表、测覆盖率,又用什么样的预注册检验去推翻本书自己? (https://5dlibrary.com/answers/zh/lexical-coverage-98-threshold-zipf-malay-hokkien/c07/) - Ch.8 Closing the Loop: Customer-Service Word Lists, Team Learning Plans and How to Refute This Book — Chapter 8 gives a procedure for building, measuring and re-measuring word lists for service teams in Malaysia and Taiwan, the distilled Word-List Coverage Calculator (in its demonstration the affix rules raise coverage from 0.6667 to 0.8667), three pre-registered tests and the results that would refute the book's central claim. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/lexical-coverage-98-threshold-zipf-malay-hokkien/c08/) 第8章 实证闭环:客服话术词表、团队学习计划与推翻本书的方案 — 第八章给出马来西亚与台湾客服团队的建表、测量与复测流程,蒸馏小工具「词表覆盖率计算器」(示范中词缀规则把覆盖率从 0.6667 提到 0.8667),以及三项预注册检验与推翻本书核心命题的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/lexical-coverage-98-threshold-zipf-malay-hokkien/c08/) ## 5DL-B-229 · Minimum-Jerk Brush Trajectories, Capillary Ink Flow in Flying-White Strokes and the Fractal Dimension of Stroke Beauty 中文书名:书法运笔最小加加速度轨迹、飞白墨量毛细流与笔势美感分形维数母典 DOI: pending Cite: Fan, Y. (2026). Minimum-Jerk Brush Trajectories, Capillary Ink Flow in Flying-White Strokes and the Fractal Dimension of Stroke Beauty (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Trajectories, paper, ink and scoring parameters are illustrative; all calculations are generated by the book's programs; the book is general analysis, not business, design or medical advice. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: People who write about calligraphy often say that a master's strokes have a higher fractal dimension and that beauty lives in this self-similar complexity. This book writes the claim as runnable arithmetic and sets against it a claim that data could overturn: a master's script need not have a higher fractal dimension than a skilled learner's, and beauty depends more on a smooth speed profile (low jerk) and on the structure of pauses. The project blueprint stated this as settled; after checking the literature the book found no published study comparing the dimensions of masters' and learners' scripts, so it restates the claim as a hypothesis to be tested and draws its evidence only from the minimum-jerk model of Flash and Hogan and from both sides of the dispute over fractal analysis of Pollock's drip paintings. Trajectories, paper, ink and scoring parameters are illustrative; all calculations are generated by the book's programs; the book is general analysis, not business, design or medical advice. Chapter 1 derives the quintic stroke: peak speed is 1.875 times mean speed, the jerk cost is 720, a cycloidal profile costs 1.0823 times as much and one stop midway 16 times as much, and a dimensionless jerk is introduced that ignores size and tempo. Chapter 2 shows that box-counting dimension cannot see the moving brush: one arc written with three timings has dimension 1.2251 each time while smoothness falls from -7.653 to -11.140; in two illustrative groups the effect size for dimension is only -0.36 (95 % interval -0.98 to +0.27) and for smoothness 3.74; a random walk reaches 1.704, close to the 1.72 reported for drip paintings. Chapter 3 separates planned stops from hesitations: a stroke with a planned stop scores 100.0 once segmented, a four-piece hesitant stroke 25.0. Chapter 4 computes flying white with Washburn's equation: ink crosses unsized paper in 0.16 ms but needs 23.04 ms on sized paper, and a dry brush at 100 mm a second leaves half the paper white by 104 mm. Chapter 5 validates box counting and shows that widening the brush from 0.5 to 5.0 pixels pushes the dimension from 1.0635 to 1.5777, far more than the 0.015 between groups. Chapter 6 computes judge reliability, statistical power and forking paths: picking the best of 16 windows lifts the false-positive rate to 0.3725. Chapter 7 examines Sun Guoting's Treatise on Calligraphy and Hogarth's Analysis of Beauty. Chapter 8 gives a specification for handwritten brand marks, a closed loop for speed logs in writing practice, a pre-registered protocol and the distilled Stroke Smoothness Scorer. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:谈书法的人常说,名家笔迹的分形维数更高,美就藏在这种自相似的复杂里。本书把这条说法写成可以运行的算术,并提出一个可被数据推翻的对立命题:名家书迹的分形维数未必显著高于熟练的习书者,美更多取决于速度剖面的平滑(低加加速度)与停顿的结构。立项蓝图把这一命题写成定论;核对文献后,本书找不到直接比较名家与习书者书迹维数的已发表研究,因此把它改写为待检验的假说,证据只取自 Flash 与 Hogan 的最小加加速度模型和波洛克滴画分形分析的正反两方。书中轨迹、纸墨与评分参数均为示意;全部计算由书中程序生成;本书是非个人化的分析,不构成经营、设计或医疗建议。 第一章推导五次多项式笔迹:峰值速度是平均速度的 1.875 倍,加加速度代价为 720,摆线速度剖面贵 1.0823 倍,中途停一次贵 16 倍,并引入不随大小与快慢改变的无量纲加加速度。第二章证明盒计数维数看不见运笔:同一条弧线以三种时序写出,维数都是 1.2251,平滑度却从 -7.653 落到 -11.140;示意的两组笔迹中,维数的效应量只有 -0.36(95 % 区间 -0.98 到 +0.27),平滑度的效应量是 3.74;随机游走的维数 1.704 已接近滴画研究报告的 1.72。第三章区分有意的顿笔与迟疑:顿笔分段后得 100.0 分,四段迟疑的笔只得 25.0 分。第四章用 Washburn 方程计算飞白:未上胶的纸 0.16 毫秒就被墨穿透,上胶的纸要 23.04 毫秒,枯笔在每秒 100 毫米时写到 104 毫米处已有一半留白。 第五章校验盒计数法,并说明笔宽从 0.5 到 5.0 像素就把维数从 1.0635 推到 1.5777,远大于两组之间 0.015 的差。第六章计算评分者信度、统计功效与分叉路径:在 16 个窗口里挑最好的一个,假阳性率升到 0.3725。第七章考辨孙过庭《书谱》与荷加斯《美的分析》。第八章给出品牌手写体规范、书写练习速度日志的闭环、预注册方案与蒸馏小工具「笔势平滑度评分器」。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Minimum Jerk: the Quintic Stroke and a Dimensionless Jerk — Chapter 1 derives the quintic stroke: peak speed is 1.875 times mean speed, the jerk cost is 720, a cycloidal profile costs 1.0823 times as much and one stop midway 16 times as much, and a dimensionless jerk is introduced that ignores size and tempo. (https://5dlibrary.com/answers/en/calligraphy-minimum-jerk-flying-white-capillary-fractal/c01/) 第1章 最小加加速度:五次多项式笔迹与无量纲加加速度 — 第一章推导五次多项式笔迹:峰值速度是平均速度的 1.875 倍,加加速度代价为 720,摆线速度剖面贵 1.0823 倍,中途停一次贵 16 倍,并引入不随大小与快慢改变的无量纲加加速度。 (https://5dlibrary.com/answers/zh/calligraphy-minimum-jerk-flying-white-capillary-fractal/c01/) - Ch.2 The Fractal Dimension Cannot See the Moving Brush: the Counterintuitive Claim and the Pollock Dispute — Chapter 2 shows that box-counting dimension cannot see the moving brush: one arc written with three timings has dimension 1.2251 each time while smoothness falls from -7.653 to -11.140; in two illustrative groups the effect size for dimension is only -0.36 (95 % interval -0.98 to +0.27) and for smoothness 3.74; a random walk reaches 1.704, close to the 1.72 reported for drip paintings. (https://5dlibrary.com/answers/en/calligraphy-minimum-jerk-flying-white-capillary-fractal/c02/) 第2章 分形维数看不见运笔:反常识命题与波洛克之争 — 第二章证明盒计数维数看不见运笔:同一条弧线以三种时序写出,维数都是 1.2251,平滑度却从 -7.653 落到 -11.140;示意的两组笔迹中,维数的效应量只有 -0.36(95 % 区间 -0.98 到 +0.27),平滑度的效应量是 3.74;随机游走的维数 1.704 已接近滴画研究报告的 1.72。 (https://5dlibrary.com/answers/zh/calligraphy-minimum-jerk-flying-white-capillary-fractal/c02/) - Ch.3 The Structure of Pauses: Planned Stops, Hesitations and the Two-Thirds Power Law — Chapter 3 separates planned stops from hesitations: a stroke with a planned stop scores 100.0 once segmented, a four-piece hesitant stroke 25.0. (https://5dlibrary.com/answers/en/calligraphy-minimum-jerk-flying-white-capillary-fractal/c03/) 第3章 停顿的结构:顿笔、迟疑与三分之二幂律 — 第三章区分有意的顿笔与迟疑:顿笔分段后得 100.0 分,四段迟疑的笔只得 25.0 分。 (https://5dlibrary.com/answers/zh/calligraphy-minimum-jerk-flying-white-capillary-fractal/c03/) - Ch.4 The Capillary Physics of Flying White: Washburn's Equation and the Drying Brush — Chapter 4 computes flying white with Washburn's equation: ink crosses unsized paper in 0.16 ms but needs 23.04 ms on sized paper, and a dry brush at 100 mm a second leaves half the paper white by 104 mm. (https://5dlibrary.com/answers/en/calligraphy-minimum-jerk-flying-white-capillary-fractal/c04/) 第4章 飞白的毛细物理:Washburn 方程与笔毫枯润 — 第四章用 Washburn 方程计算飞白:未上胶的纸 0.16 毫秒就被墨穿透,上胶的纸要 23.04 毫秒,枯笔在每秒 100 毫米时写到 104 毫米处已有一半留白。 (https://5dlibrary.com/answers/zh/calligraphy-minimum-jerk-flying-white-capillary-fractal/c04/) - Ch.5 The Metrology of Box Counting: What a Dimension Actually Measures — Chapter 5 validates box counting and shows that widening the brush from 0.5 to 5.0 pixels pushes the dimension from 1.0635 to 1.5777, far more than the 0.015 between groups. (https://5dlibrary.com/answers/en/calligraphy-minimum-jerk-flying-white-capillary-fractal/c05/) 第5章 盒计数法的度量学:维数到底量了什么 — 第五章校验盒计数法,并说明笔宽从 0.5 到 5.0 像素就把维数从 1.0635 推到 1.5777,远大于两组之间 0.015 的差。 (https://5dlibrary.com/answers/zh/calligraphy-minimum-jerk-flying-white-capillary-fractal/c05/) - Ch.6 Testing Beauty: Judge Reliability, Statistical Power and Forking Paths — Chapter 6 computes judge reliability, statistical power and forking paths: picking the best of 16 windows lifts the false-positive rate to 0.3725. (https://5dlibrary.com/answers/en/calligraphy-minimum-jerk-flying-white-capillary-fractal/c06/) 第6章 美感如何检验:评分者信度、统计功效与分叉路径 — 第六章计算评分者信度、统计功效与分叉路径:在 16 个窗口里挑最好的一个,假阳性率升到 0.3725。 (https://5dlibrary.com/answers/zh/calligraphy-minimum-jerk-flying-white-capillary-fractal/c06/) - Ch.7 Classics East and West: Sun Guoting's Treatise on Calligraphy and Hogarth's Analysis of Beauty — Sun Guoting writes about the time course of the brush, Hogarth about the geometry of the line. The Shupu's "within a single stroke, rise and fall change at the tip of the brush" and "lingering is not always slow, sending is not always fast" both have counterparts in one bell-shaped sweep, which lingers for 0.415 of its moving time and sends for 0.352; a constant-speed stroke, with a coefficient of variation of 0.000, breaks the sentence outright. "Stopped, it rests like a mountain" supports scoring planned stops apart from hesitations, and "the brush never moves in vain" is a prediction that can be tested now. Hogarth's beauty of the middle has no counterpart in the measures of the line: the smoothness of tracing falls steadily from -6.575 to -13.870 and the wobble of the dimension is an edge effect, so it is a viewer's preference that judges must test. The next chapter puts the models of the whole book to work, with a specification for handwritten brand marks, speed logs for writing practice and a pre-registered plan for falsification. (https://5dlibrary.com/answers/en/calligraphy-minimum-jerk-flying-white-capillary-fractal/c07/) 第7章 东西方原典考辨:孙过庭《书谱》与荷加斯《美的分析》 — 孙过庭谈运笔的时间进程,荷加斯谈线条的几何形状。《书谱》的「一画之间,变起伏于锋杪」与「留不常迟,遣不恒疾」,在一次钟形扫过中都有对应:有 0.415 的时间在留,0.352 的时间在遣;匀速之笔的变异系数为 0.000,恰恰违反了这句话。「导之则泉注,顿之则山安」支持把顿笔与迟疑分开评分,「翰不虚动」则是一个今天就能检验的预测。荷加斯的居中之美在线条的度量中找不到对应:书写平滑度从 -6.575 单调降到 -13.870,维数的起伏只是边缘效应,所以它是观看者的偏好,要用评审来检验。下一章把全书的模型落到实处,给出品牌手写体规范、书写练习的速度日志与一份预注册的证伪方案。 (https://5dlibrary.com/answers/zh/calligraphy-minimum-jerk-flying-white-capillary-fractal/c07/) - Ch.8 Closing the Loop: Handwritten Brand Marks, Writing-Practice Speed Logs and a Falsification Plan — Chapter 8 gives a specification for handwritten brand marks, a closed loop for speed logs in writing practice, a pre-registered protocol and the distilled Stroke Smoothness Scorer. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/calligraphy-minimum-jerk-flying-white-capillary-fractal/c08/) 第8章 实证闭环:品牌手写体规范、书写练习速度日志与证伪方案 — 第八章给出品牌手写体规范、书写练习速度日志的闭环、预注册方案与蒸馏小工具「笔势平滑度评分器」。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/calligraphy-minimum-jerk-flying-white-capillary-fractal/c08/) ## 5DL-B-230 · Post-AI Leaps in Human Go Decision Quality, Novel-Move Rates and Strategic Entropy in the Thirteen-Chapter Classic of Go 中文书名:围棋人工智能后人类决策质量跃迁、新颖着法率与棋经十三篇策略熵母典 DOI: pending Cite: Fan, Y. (2026). Post-AI Leaps in Human Go Decision Quality, Novel-Move Rates and Strategic Entropy in the Thirteen-Chapter Classic of Go (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Apart from the paper's figures, all parameters are illustrative and all calculations are generated by the book's programs; the book is non-personalised business guidance and addresses no individual decision. Chapter 2 demonstrates regression discontinuity in time on simulated data: with a true jump of 0.30 a 12-month bandwidth detects it only 0.685 of the time, while a world with nothing but a ramp is declared significant with probability 0.258 at 36 months. Abstract: After AlphaGo, do people grow duller by leaning on AI, and does Go become uniform? This book starts from the 2023 study by Shin, Kim, van Opheusden and Griffiths in the Proceedings of the National Academy of Sciences, which used KataGo to score 5,857,513 decisions by professional players from 1950 to 2021 and found that both decision quality and novelty rose clearly after AI appeared. The book sets out the paper's numbers as published and then asks, item by item, under what conditions they hold. Apart from the paper's figures, all parameters are illustrative and all calculations are generated by the book's programs; the book is non-personalised business guidance and addresses no individual decision. Chapter 1 turns the paper's Table 1 into effect sizes: after AI an ordinary move's Decision Quality Index is 0.59754 points higher, a novel move's 1.11258 points, and the old penalty for novelty shrinks by 0.848; it also shows that median summaries hit a ceiling once more than half of moves match the AI. Chapter 2 demonstrates regression discontinuity in time on simulated data: with a true jump of 0.30 a 12-month bandwidth detects it only 0.685 of the time, while a world with nothing but a ramp is declared significant with probability 0.258 at 36 months. Chapter 3 shows that the Novelty Index drifts mechanically as the historical database grows and tests the inserted-AI-games check against memorisation: copying particular games is fully exposed, but learning the AI's opening distribution keeps 0.860 of the gain. Chapter 4 measures joseki concentration with Shannon entropy; the strategic entropy of an illustrative corner position rises from 1.608 to 2.360 bits, and a model of conformist learning shows an outside evaluator bringing forward the escape from a local optimum by about 12 years. Chapter 5 separates internalisation, memorisation and composition, showing that player fixed effects push a spurious interaction from 0.831 down to -0.053 but cannot touch memorisation. Chapter 6 tests the yardstick: evaluator noise averages out in large samples, and the same 3-point mistake costs 14.566 points of win rate in an even position and 1.431 when far ahead. Chapter 7 weighs Zhang Ni's Classic of Go in Thirteen Chapters and Franklin's The Morals of Chess, computing that the break-even point of "rather lose several stones than lose the initiative" falls from 7.00 stones in the opening to 0.79 in the endgame. Chapter 8 gives seven rules for team reviews with AI, an audit of decision quality in the publisher's AI-assisted writing, the distilled Review Quality Scorer and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. 摘要:AlphaGo 之后,人会不会因为依赖 AI 而变笨,围棋会不会变得千篇一律?本书以 Shin、Kim、van Opheusden 与 Griffiths 二〇二三年在《美国国家科学院院刊》上的研究为起点:他们用 KataGo 给一九五〇到二〇二一年职业棋手的 5,857,513 个决策逐一打分,发现 AI 出现后决策质量与新颖度都明显上升。本书把论文的数字原样摆出,再逐项追问它们在什么条件下成立。书中除论文数字外的参数均为示意,全部计算由书中程序生成;本书是非个人化经营建议,不针对任何个人决定。 第一章把论文表 1 换算成效应量:AI 后普通着法的决策质量指数高出 0.59754 个百分点,新着高出 1.11258 个点,新着原有的代价缩小了 0.848;并说明匹配率超过一半时中位数汇总会碰到上限。第二章在模拟数据上演示时间断点回归:真实跳变为 0.30 时,12 个月带宽的检出率只有 0.685,而一个只有斜坡的世界在 36 个月带宽下有 0.258 的机会被判为显著。第三章说明新颖指数随历史库膨胀而机械漂移,并检验插入 AI 对局的反背谱检验:照抄棋谱会被完全识破,学会 AI 的开局分布却仍保留 0.860 的增益。第四章用香农熵度量定式集中度,示意角部局面的策略熵从 1.608 升到 2.360 比特,从众学习模型显示外部评估者让群体提早约 12 年离开局部最优。 第五章区分内化、背谱与构成偏差,说明棋手固定效应能把虚假交互项从 0.831 压到 -0.053,却对背谱无效。第六章检验量尺:评估噪声在大样本中被平均掉,同一个 3 目错误在均势与大优局面中的胜率代价是 14.566 对 1.431。第七章考辨张拟《棋经十三篇》与富兰克林《下棋的道德》,算出「宁输数子,勿失一先」的盈亏平衡点从开局的 7.00 颗子降到终局的 0.79 颗。第八章给出团队 AI 复盘的七条规则、出版方 AI 写作的决策质量审计、蒸馏小工具「复盘质量评分器」与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Leap at the Break: a Decision Quality Index on a KataGo Yardstick — Chapter 1 turns the paper's Table 1 into effect sizes: after AI an ordinary move's Decision Quality Index is 0.59754 points higher, a novel move's 1.11258 points, and the old penalty for novelty shrinks by 0.848; it also shows that median summaries hit a ceiling once more than half of moves match the AI. (https://5dlibrary.com/answers/en/go-post-ai-human-decision-quality-novelty-strategic-entropy/c01/) 第1章 断点上的跃迁:KataGo 量尺下的决策质量指数 — 第一章把论文表 1 换算成效应量:AI 后普通着法的决策质量指数高出 0.59754 个百分点,新着高出 1.11258 个点,新着原有的代价缩小了 0.848;并说明匹配率超过一半时中位数汇总会碰到上限。 (https://5dlibrary.com/answers/zh/go-post-ai-human-decision-quality-novelty-strategic-entropy/c01/) - Ch.2 A Break in Time: Regression Discontinuity at March 2016 — Chapter 2 demonstrates regression discontinuity in time on simulated data: with a true jump of 0.30 a 12-month bandwidth detects it only 0.685 of the time, while a world with nothing but a ramp is declared significant with probability 0.258 at 36 months. (https://5dlibrary.com/answers/en/go-post-ai-human-decision-quality-novelty-strategic-entropy/c02/) 第2章 时间上的断点:二〇一六年三月的回归不连续设计 — 第二章在模拟数据上演示时间断点回归:真实跳变为 0.30 时,12 个月带宽的检出率只有 0.685,而一个只有斜坡的世界在 36 个月带宽下有 0.258 的机会被判为显著。 (https://5dlibrary.com/answers/zh/go-post-ai-human-decision-quality-novelty-strategic-entropy/c02/) - Ch.3 Novel-Move Rates: the First Historically New Move, a Growing Database and the Memorisation Test — Chapter 3 shows that the Novelty Index drifts mechanically as the historical database grows and tests the inserted-AI-games check against memorisation: copying particular games is fully exposed, but learning the AI's opening distribution keeps 0.860 of the gain. (https://5dlibrary.com/answers/en/go-post-ai-human-decision-quality-novelty-strategic-entropy/c03/) 第3章 新颖着法率:首个历史新着、数据库膨胀与背谱检验 — 第三章说明新颖指数随历史库膨胀而机械漂移,并检验插入 AI 对局的反背谱检验:照抄棋谱会被完全识破,学会 AI 的开局分布却仍保留 0.860 的增益。 (https://5dlibrary.com/answers/zh/go-post-ai-human-decision-quality-novelty-strategic-entropy/c03/) - Ch.4 Strategic Entropy: Joseki Concentration, Local Optima and Conformist Lock-In — Chapter 4 measures joseki concentration with Shannon entropy; the strategic entropy of an illustrative corner position rises from 1.608 to 2.360 bits, and a model of conformist learning shows an outside evaluator bringing forward the escape from a local optimum by about 12 years. (https://5dlibrary.com/answers/en/go-post-ai-human-decision-quality-novelty-strategic-entropy/c04/) 第4章 策略熵:定式集中度、局部最优与从众锁定 — 第四章用香农熵度量定式集中度,示意角部局面的策略熵从 1.608 升到 2.360 比特,从众学习模型显示外部评估者让群体提早约 12 年离开局部最优。 (https://5dlibrary.com/answers/zh/go-post-ai-human-decision-quality-novelty-strategic-entropy/c04/) - Ch.5 The Contest over Mechanism: Internalisation, Memorisation and Composition — Chapter 5 separates internalisation, memorisation and composition, showing that player fixed effects push a spurious interaction from 0.831 down to -0.053 but cannot touch memorisation. (https://5dlibrary.com/answers/en/go-post-ai-human-decision-quality-novelty-strategic-entropy/c05/) 第5章 机制之争:内化、背谱与构成偏差 — 第五章区分内化、背谱与构成偏差,说明棋手固定效应能把虚假交互项从 0.831 压到 -0.053,却对背谱无效。 (https://5dlibrary.com/answers/zh/go-post-ai-human-decision-quality-novelty-strategic-entropy/c05/) - Ch.6 The Limits of the Yardstick: Evaluator Noise, Engine Versions and the Win-Rate Ceiling — Chapter 6 tests the yardstick: evaluator noise averages out in large samples, and the same 3-point mistake costs 14.566 points of win rate in an even position and 1.431 when far ahead. (https://5dlibrary.com/answers/en/go-post-ai-human-decision-quality-novelty-strategic-entropy/c06/) 第6章 量尺的边界:评估噪声、引擎版本与胜率天花板 — 第六章检验量尺:评估噪声在大样本中被平均掉,同一个 3 目错误在均势与大优局面中的胜率代价是 14.566 对 1.431。 (https://5dlibrary.com/answers/zh/go-post-ai-human-decision-quality-novelty-strategic-entropy/c06/) - Ch.7 Classics East and West: the Classic of Go in Thirteen Chapters and Franklin's The Morals of Chess — Chapter 7 weighs Zhang Ni's Classic of Go in Thirteen Chapters and Franklin's The Morals of Chess, computing that the break-even point of "rather lose several stones than lose the initiative" falls from 7.00 stones in the opening to 0.79 in the endgame. (https://5dlibrary.com/answers/en/go-post-ai-human-decision-quality-novelty-strategic-entropy/c07/) 第7章 东西方原典考辨:《棋经十三篇》与富兰克林《下棋的道德》 — 第七章考辨张拟《棋经十三篇》与富兰克林《下棋的道德》,算出「宁输数子,勿失一先」的盈亏平衡点从开局的 7.00 颗子降到终局的 0.79 颗。 (https://5dlibrary.com/answers/zh/go-post-ai-human-decision-quality-novelty-strategic-entropy/c07/) - Ch.8 Closing the Loop: a Team Routine for AI Reviews and an Audit of Editorial Decision Quality — Chapter 8 gives seven rules for team reviews with AI, an audit of decision quality in the publisher's AI-assisted writing, the distilled Review Quality Scorer and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/go-post-ai-human-decision-quality-novelty-strategic-entropy/c08/) 第8章 实证闭环:团队 AI 复盘流程与出版局决策质量审计 — 第八章给出团队 AI 复盘的七条规则、出版方 AI 写作的决策质量审计、蒸馏小工具「复盘质量评分器」与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/go-post-ai-human-decision-quality-novelty-strategic-entropy/c08/) ## 5DL-B-231 · Dunbar's Layered Numbers, Relationship-Maintenance Time Budgets and the Decay Half-Life of Private-Channel Customers 中文书名:邓巴数分层、关系维护时间预算与私域客户衰减半衰期母典 DOI: pending Cite: Fan, Y. (2026). Dunbar's Layered Numbers, Relationship-Maintenance Time Budgets and the Decay Half-Life of Private-Channel Customers (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its layer floors, half-lives, weights and stop probabilities are illustrative; all calculations are generated by its programs; it is non-individualized business guidance, and its compliance rules yield to the law in force locally. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: Private-channel trade runs on a rule: one account or one adviser can maintain at most 150 customers. This book examines where the rule comes from and offers an alternative: layer the list by time budget instead of capping it by head count. Its layer floors, half-lives, weights and stop probabilities are illustrative; all calculations are generated by its programs; it is non-individualized business guidance, and its compliance rules yield to the law in force locally. Chapter 1 recomputes the 2021 reanalysis: the 95% intervals of eight models stretch from 2.1 to 520.0, and the preferred model gives 69.2 people with an interval of 3.8 to 292.0, holding 15, 50 and 150 alike. Chapter 2 recomputes the layer ratio from six mean group sizes, obtaining 3.258 to 3.379, and shows why a common multiplicative factor cancels in the division. Chapter 3 writes relationships as exponential decay and gives the longest contact interval that holds each layer; a customer with a 60-day half-life needs contact about every 10.2 days to stay in the VIP layer. Chapter 4 allocates weekly time with a Lagrange multiplier: the innermost 5 take 0.406 of it, and spreading it evenly over 150 people yields only 0.598 of the optimal value. Chapter 5 shows that differences between customers make a cohort's half-life grow with time, with survivors' conditional half-life reaching 1061.4 days, and uses the BG/NBD formula to judge whether a silent customer is still active. Chapter 6 writes quiet hours, stop means stop, at most one message per 7 days and data minimization into the schedule, and shows that the value peak of messaging makes 0.446 of customers stop within a year. Chapter 7 reads the Analects on helpful and harmful friends alongside Cicero's On Friendship. Chapter 8 gives the layering procedure for a buyer's private channel, a VIP care calendar, the tool "Private-Channel Layer Scheduler" and a pre-registered test protocol. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:私域经营流传一条规矩:一个号、一位顾问最多维护 150 人。本书检验这条规矩的来源,并给出替代方案:按时间预算分层,而不是设人头上限。书中的层级下限、半衰期、权重与停止概率均为示意参数;全部计算由书中程序生成;本书是非个人化经营建议,合规规则以各地现行法规为准。 第一章重算 2021 年的重估:八个模型的 95% 区间从 2.1 延伸到 520.0,偏好模型给出 69.2 人、区间 3.8 到 292.0,同时容得下 15、50 与 150。第二章由六个平均群体规模重算层比,得到 3.258 到 3.379,并说明共同乘数因子为何在相除时被约掉。第三章把关系写成指数衰减,给出守住每层的最长联系间隔,半衰期 60 天的客户守住贵宾层约需每 10.2 天联系一次。第四章用拉格朗日乘子分配每周时间,最内 5 人占 0.406,平均分给 150 人只得到最优价值的 0.598。 第五章说明客户之间的差异会让一批客户的半衰期随时间变长,幸存者的条件半衰期可达 1061.4 天,并用 BG/NBD 公式判断沉默客户是否仍活跃。第六章把静默时段、停止即停、每 7 天至多一条与数据最小化写进排程,并显示推送的价值峰值会让 0.446 的客户在一年内停止。第七章对读《论语》的益者三友、损者三友与西塞罗《论友谊》。第八章给出买手私域的分层流程、贵宾关怀日历、小工具「私域分层排程器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Confidence Interval of 150: from a Neocortex Regression to an Envelope of Eight Models — Chapter 1 recomputes the 2021 reanalysis: the 95% intervals of eight models stretch from 2.1 to 520.0, and the preferred model gives 69.2 people with an interval of 3.8 to 292.0, holding 15, 50 and 150 alike. (https://5dlibrary.com/answers/en/dunbar-layers-maintenance-time-budget-private-customer-half-life/c01/) 第1章 150 的置信区间:从新皮层回归到八个模型的包络 — 第一章重算 2021 年的重估:八个模型的 95% 区间从 2.1 延伸到 520.0,偏好模型给出 69.2 人、区间 3.8 到 292.0,同时容得下 15、50 与 150。 (https://5dlibrary.com/answers/zh/dunbar-layers-maintenance-time-budget-private-customer-half-life/c01/) - Ch.2 A Ratio of About Three: 5, 15, 50, 150 and the Robustness of Proportions — Chapter 2 recomputes the layer ratio from six mean group sizes, obtaining 3.258 to 3.379, and shows why a common multiplicative factor cancels in the division. (https://5dlibrary.com/answers/en/dunbar-layers-maintenance-time-budget-private-customer-half-life/c02/) 第2章 约三倍的层比:5、15、50、150 与比例的稳健性 — 第二章由六个平均群体规模重算层比,得到 3.258 到 3.379,并说明共同乘数因子为何在相除时被约掉。 (https://5dlibrary.com/answers/zh/dunbar-layers-maintenance-time-budget-private-customer-half-life/c02/) - Ch.3 Exponential Decay of Tie Strength: Half-Lives, Slide Times and Maintenance Intervals — Chapter 3 writes relationships as exponential decay and gives the longest contact interval that holds each layer; a customer with a 60-day half-life needs contact about every 10.2 days to stay in the VIP layer. (https://5dlibrary.com/answers/en/dunbar-layers-maintenance-time-budget-private-customer-half-life/c03/) 第3章 关系强度的指数衰减:半衰期、滑层时间与维护间隔 — 第三章把关系写成指数衰减,给出守住每层的最长联系间隔,半衰期 60 天的客户守住贵宾层约需每 10.2 天联系一次。 (https://5dlibrary.com/answers/zh/dunbar-layers-maintenance-time-budget-private-customer-half-life/c03/) - Ch.4 A Lagrangian Split of the Time Budget: Head Counts Are Computed, Not Decreed — Chapter 4 allocates weekly time with a Lagrange multiplier: the innermost 5 take 0.406 of it, and spreading it evenly over 150 people yields only 0.598 of the optimal value. (https://5dlibrary.com/answers/en/dunbar-layers-maintenance-time-budget-private-customer-half-life/c04/) 第4章 时间预算的拉格朗日分配:人数是算出来的,不是定出来的 — 第四章用拉格朗日乘子分配每周时间,最内 5 人占 0.406,平均分给 150 人只得到最优价值的 0.598。 (https://5dlibrary.com/answers/zh/dunbar-layers-maintenance-time-budget-private-customer-half-life/c04/) - Ch.5 The Decay Half-Life of Private-Channel Customers: Why a Cohort's Half-Life Grows — Chapter 5 shows that differences between customers make a cohort's half-life grow with time, with survivors' conditional half-life reaching 1061.4 days, and uses the BG/NBD formula to judge whether a silent customer is still active. (https://5dlibrary.com/answers/en/dunbar-layers-maintenance-time-budget-private-customer-half-life/c05/) 第5章 私域客户的衰减半衰期:群体的半衰期为什么会变长 — 第五章说明客户之间的差异会让一批客户的半衰期随时间变长,幸存者的条件半衰期可达 1061.4 天,并用 BG/NBD 公式判断沉默客户是否仍活跃。 (https://5dlibrary.com/answers/zh/dunbar-layers-maintenance-time-budget-private-customer-half-life/c05/) - Ch.6 A Compliant Contact Schedule: Quiet Hours, Stop Means Stop and a Frequency Cap — Chapter 6 writes quiet hours, stop means stop, at most one message per 7 days and data minimization into the schedule, and shows that the value peak of messaging makes 0.446 of customers stop within a year. (https://5dlibrary.com/answers/en/dunbar-layers-maintenance-time-budget-private-customer-half-life/c06/) 第6章 触达频率的合规排程:静默时段、停止即停与频率上限 — 第六章把静默时段、停止即停、每 7 天至多一条与数据最小化写进排程,并显示推送的价值峰值会让 0.446 的客户在一年内停止。 (https://5dlibrary.com/answers/zh/dunbar-layers-maintenance-time-budget-private-customer-half-life/c06/) - Ch.7 Classical Sources East and West: Confucius's Three Helpful Friends and Cicero's Measures of Salt — Confucius's three helpful and three harmful friends can be placed on a measurable quantity, the half-lives of customers recruited in different ways; under illustrative parameters the honest cohort overtakes in active count on day 40.7 and accumulates 1.875 times as much activity over the long run. "Stop when it does not work" and "too insistent brings estrangement" correspond to stop means stop and the frequency cap. Cicero's confinement of friendship to "a few" matches the innermost layer of the layered structure, and his "many measures of salt" converts to 1746 hours of shared meals, far more than the modern estimate of 200 hours yet pointing the same way. The final chapter turns the book's calculations into a plan that can be carried out and can also be refuted. (https://5dlibrary.com/answers/en/dunbar-layers-maintenance-time-budget-private-customer-half-life/c07/) 第7章 东西方原典考辨:孔子的益者三友与西塞罗的几斗盐 — 孔子的「益者三友,损者三友」可以落到一个可测的量上,即获客方式不同的两批客户的半衰期;示意参数下,诚实批次在第 40.7 天反超活跃人数,长期累计是奉承批次的 1.875 倍。「不可则止」与「朋友数,斯疏矣」对应停止即停与频率上限。西塞罗把友谊收束到「少数人之间」,与分层结构的最内层一致;他的「好几斗盐」换算下来是 1746 个共餐小时,远多于现代估计的 200 小时,但方向相同。最后一章把全书的计算变成一份可以执行、也可以被推翻的方案。 (https://5dlibrary.com/answers/zh/dunbar-layers-maintenance-time-budget-private-customer-half-life/c07/) - Ch.8 Closing the Loop: Layered Contact for a Buyer's Private Channel, a VIP Care Calendar and a Plan for Refutation — Chapter 8 gives the layering procedure for a buyer's private channel, a VIP care calendar, the tool "Private-Channel Layer Scheduler" and a pre-registered test protocol. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/dunbar-layers-maintenance-time-budget-private-customer-half-life/c08/) 第8章 实证闭环:买手私域的分层触达、贵宾关怀日历与证伪方案 — 第八章给出买手私域的分层流程、贵宾关怀日历、小工具「私域分层排程器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/dunbar-layers-maintenance-time-budget-private-customer-half-life/c08/) ## 5DL-B-232 · Temperature Phase Lag across the Twenty-Four Solar Terms, First-Order Thermal Inertia behind the Dog Days and Seasonal Demand Forecasting 中文书名:二十四节气气温滞后相位、三伏一阶热惯性与季节性消费需求预测母典 DOI: pending Cite: Fan, Y. (2026). Temperature Phase Lag across the Twenty-Four Solar Terms, First-Order Thermal Inertia behind the Dog Days and Seasonal Demand Forecasting (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters in this book are illustrative; all calculations are generated by the book's programs; the book is non-personalised business guidance, and its order dates are demonstrations that promise no sales to any merchant. Chapter 4 shows that "the hottest day" is a statistic dominated by weather noise: in illustrative data the single hottest day misses by 16.81 days, while the harmonic phase from three years of daily data misses by only 1.25 days. Abstract: The solstice has the most sunshine, so it must be the hottest time; the names of the solar terms are the seasons, so ordering by them must be right. This book turns that common belief into a proposition that can be computed and can be refuted by data. Surface temperature behaves like a vessel with thermal inertia: under sinusoidal sunshine with a period of one year, the steady solution of the first-order equation arrives arctan(ωτ)/ω days late with its amplitude reduced to cos(arctan(ωτ)), and however large the heat capacity, the lag never exceeds 91.31 days. Trenberth reported in the Bulletin of the American Meteorological Society in 1983 a Northern Hemisphere mid-latitude lag of 32.5 days; in 2026 the solstice and Major Heat are 31.45 days apart, so the heat peak falls around Major Heat. Parameters in this book are illustrative; all calculations are generated by the book's programs; the book is non-personalised business guidance, and its order dates are demonstrations that promise no sales to any merchant. Chapters 1 and 2 derive the phase lag of first-order thermal inertia and compute the twenty-four terms of 2026 from solar longitude, showing that the term names mark the season of sunshine. Chapter 3 lays out the three fu periods by the geng-day rule: their midpoint lies about eleven days after the Northern Hemisphere heat peak, but because the crest is flat they are on average only 0.245 degrees Celsius cooler. Chapter 4 shows that "the hottest day" is a statistic dominated by weather noise: in illustrative data the single hottest day misses by 16.81 days, while the harmonic phase from three years of daily data misses by only 1.25 days. Chapter 5 links demand to temperature: a plan that follows the term names passes halfway in week 10 while true demand does so in week 14, exactly a month early; only daylight-driven categories escape the lag. Chapter 6 prices the error with the newsvendor model: ordering by the term names earns only 0.7196 of the best profit, less than flat ordering that ignores the season (0.7530), while the phase drift from climate change costs only 0.00147. Chapter 7 weighs the Shi Xun chapter of the Yi Zhou Shu and Hesiod's Works and Days. Chapter 8 gives a buyer's order calendar, a store's term-based promotion schedule, the distilled Solar-Term Order Calendar and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:夏至日照最多,所以夏至最热;节气的名字就是季节,照着节气名订货就对了。本书把这个人人都信的常识写成一个可以计算、也可以被数据推翻的命题。地表气温是一个有热惯性的容器,对一年一周期的正弦日照,一阶方程的稳态解晚到 arctan(ωτ)/ω 天,振幅缩小到 cos(arctan(ωτ)) 倍,而且无论热容多大,滞后都不超过 91.31 天。Trenberth 一九八三年在《美国气象学会会刊》报告的北半球中纬度滞后为 32.5 天;二〇二六年夏至到大暑相隔 31.45 天,热峰因此落在大暑前后。书中参数为示意;全部计算由书中程序生成;本书是非个人化经营建议,订货日期只是演示,不对任何商家承诺销量。 第一、二章推导一阶热惯性的相位滞后,并用太阳黄经算出二〇二六年的二十四节气,说明节气名标的是日照的季节。第三章按庚日规则排出三伏:三伏的中点比北半球中纬度的热峰晚约十一天,但因为峰顶平坦,平均只凉 0.245 摄氏度。第四章说明「哪天最热」是一个被天气噪声主导的统计量:示意数据中单日最高温的定位误差为 16.81 天,而三年逐日数据的谐波相位误差只有 1.25 天。 第五章把需求接到气温上:照节气名字订货,计划在第 10 周累计过半,真实需求要到第 14 周,正好早一个月;只有由日照驱动的品类没有这个滞后。第六章用报童模型标价:照节气名字订货只拿到最优利润的 0.7196,比完全不考虑季节的平均订货(0.7530)还低,而气候变化造成的相位漂移只损失 0.00147。第七章考辨《逸周书·时训解》与赫西俄德《工作与时日》。第八章给出买手订货日历、门店节气排期、蒸馏小工具「节气订货日历」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Solstice Is Not the Hottest Day: First-Order Thermal Inertia and the Phase Lag arctan(ωτ) — Three points in the first half of the table are worth keeping. First, when τ is small the lag is almost equal to τ itself: with τ of 10 days the lag is 9.90 days and the amplitude kept is 0.986, because arctan(x) is close to x when x is small. Second, as τ grows the lag grows more slowly. With τ of 30 days, ωτ is 0.516, the lag is 27.70 days and the amplitude kept is 0.889; doubling τ to 60 days only brings the lag to 46.58 days; τ of 90 days gives 57.97 days, and doubling again to 180 days still gives only 73.15 days. Third, even with τ as long as a full year the lag is just 82.13 days, while the amplitude has fallen to 0.157 and the seasons are almost flattened out. The ceiling is 91.31 days, a quarter of a tropical year. (https://5dlibrary.com/answers/en/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c01/) 第1章 夏至不是最热:一阶热惯性与 arctan(ωτ) 相位滞后 — 表的前半部分有三点值得记住。其一,τ 小的时候滞后几乎等于 τ 本身:τ 为 10 天,滞后为 9.90 天,振幅保留 0.986。原因是 arctan(x) 在 x 很小时约等于 x。其二,τ 变大以后滞后增长变慢:τ 为 30 天时 ωτ 为 0.516,滞后为 27.70 天,振幅保留 0.889;τ 翻倍到 60 天,滞后只到 46.58 天;τ 为 90 天时滞后为 57.97 天;再翻倍到 180 天也只有 73.15 天。其三,就算 τ 长达一整年,滞后也只有 82.13 天,振幅却只剩 0.157,季节几乎被抹平。上限是 91.31 天,也就是四分之一个回归年。 (https://5dlibrary.com/answers/zh/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c01/) - Ch.2 Fifteen Degrees a Step: Solar Longitude, the Twenty-Four Terms and the Sinusoid of Sunshine — The left column is the first half of the year and the right column the second; the two terms on each row stand opposite each other on the ecliptic. In 2026 the summer solstice falls on 21 June at 16:24 Beijing time and Major Heat on 23 July at 03:17, an interval of 31.45 days. Put the Northern Hemisphere lag of 32½ days from Chapter 1 on top, and the model's heat peak falls about one day after Major Heat. The blueprint said that it is hottest around Major Heat; for the average Northern Hemisphere mid-latitude case that statement is off by only about a day. (https://5dlibrary.com/answers/en/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c02/) 第2章 十五度一格:太阳黄经、二十四节气与日照的正弦 — 再看日照。夏至时赤纬为 23.4 度,北纬 35 度的日照为 480 瓦每平方米;冬至时只有 192,不到夏至的一半。日照年变化的形状接近一条正弦,峰顶非常平:芒种的日照是夏至的 0.994,小暑是 0.990,大暑仍有 0.966,到立秋也还有 0.926。换句话说,名叫「立秋」的那一天,黄河中下游大气顶的日照仍然是一年最大值的九成以上。从日照看,立秋之后的夏天并没有结束,只是开始走下坡路;从气温看,热峰还没到,或刚刚过去。 (https://5dlibrary.com/answers/zh/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c02/) - Ch.3 The Arithmetic of the Dog Days: the Geng-Day Rule, the Heat Window and Thermal Mass — Chapter 3 lays out the three fu periods by the geng-day rule: their midpoint lies about eleven days after the Northern Hemisphere heat peak, but because the crest is flat they are on average only 0.245 degrees Celsius cooler. (https://5dlibrary.com/answers/en/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c03/) 第3章 三伏的算术:庚日规则、热峰窗口与热容差异 — 第三章按庚日规则排出三伏:三伏的中点比北半球中纬度的热峰晚约十一天,但因为峰顶平坦,平均只凉 0.245 摄氏度。 (https://5dlibrary.com/answers/zh/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c03/) - Ch.4 Harmonic Regression: Estimating Amplitude and Phase from Daily Readings — Chapter 4 shows that "the hottest day" is a statistic dominated by weather noise: in illustrative data the single hottest day misses by 16.81 days, while the harmonic phase from three years of daily data misses by only 1.25 days. (https://5dlibrary.com/answers/en/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c04/) 第4章 谐波回归:从逐日读数估计振幅与相位 — 第四章说明「哪天最热」是一个被天气噪声主导的统计量:示意数据中单日最高温的定位误差为 16.81 天,而三年逐日数据的谐波相位误差只有 1.25 天。 (https://5dlibrary.com/answers/zh/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c04/) - Ch.5 The Phase of Demand: Regressing Demand on Temperature, and the Misplaced Nominal Season — Chapter 5 links demand to temperature: a plan that follows the term names passes halfway in week 10 while true demand does so in week 14, exactly a month early; only daylight-driven categories escape the lag. (https://5dlibrary.com/answers/en/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c05/) 第5章 需求的相位:气温与需求的回归,以及名义季节的错位 — 第五章把需求接到气温上:照节气名字订货,计划在第 10 周累计过半,真实需求要到第 14 周,正好早一个月;只有由日照驱动的品类没有这个滞后。 (https://5dlibrary.com/answers/zh/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c05/) - Ch.6 The Price of a Phase Error: Order Timing under the Newsvendor Model — Chapter 6 prices the error with the newsvendor model: ordering by the term names earns only 0.7196 of the best profit, less than flat ordering that ignores the season (0.7530), while the phase drift from climate change costs only 0.00147. (https://5dlibrary.com/answers/en/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c06/) 第6章 相位误差的价钱:报童模型下的订货时点 — 第六章用报童模型标价:照节气名字订货只拿到最优利润的 0.7196,比完全不考虑季节的平均订货(0.7530)还低,而气候变化造成的相位漂移只损失 0.00147。 (https://5dlibrary.com/answers/zh/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c06/) - Ch.7 Classical Sources East and West: the Shi Xun Chapter of the Yi Zhou Shu and Hesiod's Works and Days — Look first at the term names. The table shows a clear division of labour. Terms named for heat or cold follow temperature. Major Heat lies 31.5 days after the solstice, where the temperature anomaly is +1.00, exactly at the heat peak; Major Cold has a temperature anomaly of −1.00, exactly at the cold trough; Minor Heat and Minor Cold stand at ±0.96, on the leading slopes of peak and trough. Terms named "Start" follow sunshine instead. At the Start of Autumn the sunshine anomaly is +0.69, just where sunshine is coming down from its crest, while the temperature anomaly is still +0.97. At the Start of Spring the sunshine anomaly is −0.71, while the temperature anomaly is −0.98, still near the cold trough. (https://5dlibrary.com/answers/en/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c07/) 第7章 东西方原典考辨:《逸周书·时训解》与赫西俄德《工作与时日》 — 先看「夏至后五十天,酷热结束」。在北半球中纬度的平均滞后 32.5 天下,夏至后第五十天的气温距平仍有 0.955,离热峰不远;距平要到第 58.7 天才降到 0.9 以下,到第 65.0 天才回到夏至当天的水平。滞后为 27.5 天的大陆性气候,第五十天的距平为 0.926,降到 0.9 以下是在第 53.7 天,离赫西俄德的五十天只差三四天。滞后为 44 天的海洋性气候,第五十天的距平高达 0.995,几乎还在峰顶。赫西俄德住在希腊中部的玻俄提亚,三面近海,按理说滞后不短。如果按气温理解「酷热结束」,他的五十天偏早;如果理解为「最难熬的那一段过去了」,在滞后较短的内陆,这个说法与模型相差不过几天。 (https://5dlibrary.com/answers/zh/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c07/) - Ch.8 Closing the Loop: the Solar-Term Order Calendar, Store Scheduling and a Pre-Registered Test — Chapter 8 gives a buyer's order calendar, a store's term-based promotion schedule, the distilled Solar-Term Order Calendar and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c08/) 第8章 实证闭环:节气订货日历、门店排期与预注册检验 — 第八章给出买手订货日历、门店节气排期、蒸馏小工具「节气订货日历」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/solar-terms-temperature-phase-lag-thermal-inertia-seasonal-demand/c08/) ## 5DL-B-233 · Optimal Transport and Kantorovich Duality: Minimum-Cost Rebalancing of Luxury Stock across Regions and Client-Migration Matrices 中文书名:最优传输康托罗维奇对偶、跨区域奢品调货最小搬运代价与客群迁移矩阵母典 DOI: pending Cite: Fan, Y. (2026). Optimal Transport and Kantorovich Duality: Minimum-Cost Rebalancing of Luxury Stock across Regions and Client-Migration Matrices (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 5 applies optimal transport to client migration: in an illustrative matrix, gross movement is 13.43 times what the net change requires, so goods follow net flows and people follow gross flows. Chapter 7 examines equitable transport in the Treatise on the Balanced Standard of the Shiji (lateral transfers save 0.721 of freight in an illustrative model) and Monge's cuttings and fillings. Abstract: When a luxury retailer with many stores transfers stock each week, the most common rule is to serve the store with the largest shortage first, each time from the nearest store with spare stock. This book refutes two beliefs behind that rule: that queuing transfers by gap costs little more than the optimum, and that the store with the largest gap or the largest loss per missing unit is where one more unit is worth most. Using the potential method of the transportation problem and Kantorovich duality, it gives exact transfer sheets and store shadow prices, every one of which can be checked by addition and subtraction. Its stores, city coordinates, rates, stock levels, margins and probabilities are illustrative; all calculations are generated by its programs; the book is not individual business advice, and every transfer proposed by its small tool must be released by a person before it is carried out. Chapter 1 measures the excess of shortage-first over three thousand random snapshots: the median is only 0.0338, but the 99th percentile reaches 0.320. Chapter 2 walks from a northwest-corner start of 2564 yuan to the optimum of 1281 yuan in six exchanges, and shows that Vogel's start is already optimal in 0.777 of snapshots. Chapter 3 reads potentials as location prices: with stock-out margins counted, the illustrative network's optimal total is 7322 yuan against 11003 for shortage-first; every store's shadow price of one more unit clusters around 1320 yuan, the loss at the marginal short store, the store with the largest loss per missing unit is worth only 1253, and since a shadow price is a local slope, 8 extra units at one store are worth only 5709. Chapter 4 shows that entropic regularisation with the Sinkhorn iteration costs only 0.0033 more at a temperature of 10 yuan, but its blurred sheets cannot be shipped as they are. Chapter 5 applies optimal transport to client migration: in an illustrative matrix, gross movement is 13.43 times what the net change requires, so goods follow net flows and people follow gross flows. Chapter 6 deals with whole units, fixed fees per shipment and moving rates: at a fee of 200 yuan the linear optimum exceeds the exact optimum by 0.2117, shortage-first then ties with the optimum, and pruning recovers it. Chapter 7 examines equitable transport in the Treatise on the Balanced Standard of the Shiji (lateral transfers save 0.721 of freight in an illustrative model) and Monge's cuttings and fillings. Chapter 8 gives the weekly transfer procedure, the Cross-Store Rebalancing Optimiser and a pre-registered test plan: the cost test needs 197 pairs of weeks, about 34 weeks across twelve regions. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:多店奢品零售商每周调货,最常见的规则是先补缺得最多的店,每次从最近的有货店调。本书证伪这条规则背后的两个信念:按缺口排队调货不会比最优解差多少;缺得最多或缺一件损失最大的店,就是多一件货最值钱的店。书中用运输问题的位势法与康托罗维奇对偶给出精确的调货单与门店影子价格,每一个数都可以用加减法核对。书中的门店、城市坐标、运价、库存、毛利与概率均为示意;全部计算由书中程序生成;本书是非个人化经营建议,书中小工具给出的每一条调拨都须由经营者人工放行。 第一章在三千张随机快照上量出「缺货店优先补」的超出:中位数只有 0.0338,第 99 百分位却达 0.320。第二章从西北角的 2564 元出发,经六次换基走到 1281 元的最优,并说明伏格尔法有 0.777 的快照起点即最优。第三章把位势读成地点价格:计入缺货毛利损失后,示意网络的最优总成本是 7322 元,「缺货店优先补」是 11003 元;每家店多一件货的影子价格都挤在边际缺货店的 1320 元附近,缺一件损失最大的店只值 1253 元,而影子价格是局部斜率,同一家店多 8 件只值 5709 元。第四章说明熵正则与 Sinkhorn 迭代在温度 10 元时成本只高出 0.0033,但模糊的单子不能直接发货。 第五章把最优传输用到会员迁移:示意矩阵中毛流动是净变化所需的 13.43 倍,货跟净流,人跟毛流。第六章处理整件、单票固定费与运价波动:单票费 200 元时线性最优比精确最优高 0.2117,此时「缺货店优先补」反而打平最优,剪枝法找回最优。第七章考辨《史记·平准书》的均输(示意模型中横向调拨省下 0.721 的运费)与蒙日的挖填土方。第八章给出周度调货流程、跨店调货最优搬运器与预注册检验方案:成本检验需要 197 对周,十二个区域约 34 周。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Price of Serving the Emptiest Shelf First: Greedy Transfers against the Transportation Optimum — Chapter 1 measures the excess of shortage-first over three thousand random snapshots: the median is only 0.0338, but the 99th percentile reaches 0.320. (https://5dlibrary.com/answers/en/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c01/) 第1章 缺货优先的代价:贪心调拨与运输问题最优解 — 第一章在三千张随机快照上量出「缺货店优先补」的超出:中位数只有 0.0338,第 99 百分位却达 0.320。 (https://5dlibrary.com/answers/zh/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c01/) - Ch.2 The Potential Method: Northwest-Corner Starts, Closed Loops and an Exact MODI Solution — Chapter 2 walks from a northwest-corner start of 2564 yuan to the optimum of 1281 yuan in six exchanges, and shows that Vogel's start is already optimal in 0.777 of snapshots. (https://5dlibrary.com/answers/en/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c02/) 第2章 位势法:西北角初解、闭回路与 MODI 精确求解 — 第二章从西北角的 2564 元出发,经六次换基走到 1281 元的最优,并说明伏格尔法有 0.777 的快照起点即最优。 (https://5dlibrary.com/answers/zh/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c02/) - Ch.3 Kantorovich Duality: Potentials Are the Shadow Prices of Stores — Chapter 3 reads potentials as location prices: with stock-out margins counted, the illustrative network's optimal total is 7322 yuan against 11003 for shortage-first; every store's shadow price of one more unit clusters around 1320 yuan, the loss at the marginal short store, the store with the largest loss per missing unit is worth only 1253, and since a shadow price is a local slope, 8 extra units at one store are worth only 5709. (https://5dlibrary.com/answers/en/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c03/) 第3章 康托罗维奇对偶:位势就是门店影子价格 — 第三章把位势读成地点价格:计入缺货毛利损失后,示意网络的最优总成本是 7322 元,「缺货店优先补」是 11003 元;每家店多一件货的影子价格都挤在边际缺货店的 1320 元附近,缺一件损失最大的店只值 1253 元,而影子价格是局部斜率,同一家店多 8 件只值 5709 元。 (https://5dlibrary.com/answers/zh/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c03/) - Ch.4 Entropic Regularisation and Sinkhorn: Faster, but Blurred — Chapter 4 shows that entropic regularisation with the Sinkhorn iteration costs only 0.0033 more at a temperature of 10 yuan, but its blurred sheets cannot be shipped as they are. (https://5dlibrary.com/answers/en/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c04/) 第4章 熵正则与 Sinkhorn:更快,但是模糊 — 第四章说明熵正则与 Sinkhorn 迭代在温度 10 元时成本只高出 0.0033,但模糊的单子不能直接发货。 (https://5dlibrary.com/answers/zh/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c04/) - Ch.5 Client-Migration Matrices: Measuring Real Movement by the Least Movement Required — Chapter 5 applies optimal transport to client migration: in an illustrative matrix, gross movement is 13.43 times what the net change requires, so goods follow net flows and people follow gross flows. (https://5dlibrary.com/answers/en/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c05/) 第5章 客群迁移矩阵:用最小搬运量量出会员的真实流动 — 第五章把最优传输用到会员迁移:示意矩阵中毛流动是净变化所需的 13.43 倍,货跟净流,人跟毛流。 (https://5dlibrary.com/answers/zh/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c05/) - Ch.6 Whole Units, Fees per Shipment and Moving Rates: From a Linear Optimum to an Executable Transfer Sheet — Chapter 6 deals with whole units, fixed fees per shipment and moving rates: at a fee of 200 yuan the linear optimum exceeds the exact optimum by 0.2117, shortage-first then ties with the optimum, and pruning recovers it. (https://5dlibrary.com/answers/en/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c06/) 第6章 整件、单票费与运价波动:从线性最优到可执行调货单 — 第六章处理整件、单票固定费与运价波动:单票费 200 元时线性最优比精确最优高 0.2117,此时「缺货店优先补」反而打平最优,剪枝法找回最优。 (https://5dlibrary.com/answers/zh/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c06/) - Ch.7 East and West in the Sources: Sang Hongyang's Equitable Transport and Monge's Cuttings and Fillings — Chapter 7 examines equitable transport in the Treatise on the Balanced Standard of the Shiji (lateral transfers save 0.721 of freight in an illustrative model) and Monge's cuttings and fillings. (https://5dlibrary.com/answers/en/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c07/) 第7章 东西方原典考辨:桑弘羊均输与蒙日的挖填土方 — 第七章考辨《史记·平准书》的均输(示意模型中横向调拨省下 0.721 的运费)与蒙日的挖填土方。 (https://5dlibrary.com/answers/zh/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c07/) - Ch.8 Closing the Loop: The Weekly Transfer Sheet, the Shadow Price Table and Human Release — Chapter 8 gives the weekly transfer procedure, the Cross-Store Rebalancing Optimiser and a pre-registered test plan: the cost test needs 197 pairs of weeks, about 34 weeks across twelve regions. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c08/) 第8章 实证闭环:周度调货单、影子价表与人工放行 — 第八章给出周度调货流程、跨店调货最优搬运器与预注册检验方案:成本检验需要 197 对周,十二个区域约 34 周。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/optimal-transport-kantorovich-duality-luxury-stock-rebalancing/c08/) ## 5DL-B-234 · Hawkes Self-Exciting Point Processes: Branching-Ratio Criticality in Comment Cascades and the Aftershock Half-Life of Livestream Herd Buying 中文书名:霍克斯自激点过程、评论裂变分支比临界值与直播跟风抢购余震半衰期母典 DOI: pending Cite: Fan, Y. (2026). Hawkes Self-Exciting Point Processes: Branching-Ratio Criticality in Comment Cascades and the Aftershock Half-Life of Livestream Herd Buying (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: Owners who run businesses on content and livestreams tend to credit hits to "heat", as if heat were a lasting state decided from outside. This book refutes that account: most hits are aftershock cascades with branching ratio n below 1, the expected number of comments that one outside comment eventually brings is 1/(1-n), and as n moves from 0.8 to 0.9 the multiplier doubles from 5.0 to 10.0. Its posts, accounts, traffic, costs and comment values are illustrative parameters; all calculations are generated by its programs; the book is not individual business advice, deals only with comments by genuine users and exposure bought through platforms' official tools, and covers no means of manufacturing comments or inventing scarcity. Chapter 1 uses the Borel distribution and cluster simulation to show why post sizes differ so much at the same branching ratio. Chapter 2 reproduces Crane and Sornette's decay classes under power-law memory, gives the half-life under exponential memory, stretched from the bare kernel's 6.93 minutes to 138.6 minutes at n of 0.95, and separates pushed peaks (median peak fraction 0.538) from grown ones (0.093). Chapter 3 simulates comment streams by Ogata's thinning and audits fits by time rescaling, finding that the gap test can hardly see clumping while the count-dispersion test rejects a flat Poisson model 0.646 of the time. Chapter 4 finds maximum likelihood estimates with Ozaki's recursion, the compensator identity and golden-section search; the illustrative post's 95 % interval for the branching ratio runs from 0.42 to 0.92, and two-hour estimates have a standard deviation of about 0.186. Chapter 5 computes the stock and half-life of aftershocks once outside traffic stops; in a backtest, shrinkage toward the account mean cuts the median forecast error from 0.589 to 0.418, and convexity lifts the expected multiplier to 3.40. Chapter 6 shows the apparent criticality of paid reach posing as word of mouth: a pushed post with a true branching ratio of 0.3 is estimated naively as 0.864 and as 0.276 once the push window enters the model, the naive fit overstating the push's yield 5.15-fold. Chapter 7 examines "Zeng Shen has killed a man" in the Zhan guo ce (at branching ratio 0.8, three reports share one source with probability 0.6400) and Omori's aftershock law (p of 1.4 under a gamma mixture). Chapter 8 gives the loop of estimating the branching ratio in the first two hours before deciding on a push, the Cascade Branching-Ratio Estimator, and a pre-registered push experiment (43 posts per group, 13 after covariate adjustment). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:做内容与直播的经营者习惯把爆款归因于「热度」,仿佛热度是一种由外部决定、持续存在的状态。本书证伪这个说法:大多数爆款是分支比 n 小于 1 的余震级联,每一条外来评论最终带来的评论总数期望为 1/(1-n),n 从 0.8 到 0.9,乘数从 5.0 翻到 10.0。书中的帖子、账号、流量、成本与评论价值均为示意参数;全部计算由书中程序生成;本书是非个人化经营建议,只讨论真实用户的评论与平台官方工具买来的曝光,不涉及任何制造评论或虚构稀缺的手段。 第一章用博雷尔分布与族群模拟说明同样的分支比下帖子规模为何悬殊。第二章在幂律记忆下复现克兰与索内特的衰减分类,在指数记忆下给出半衰期,n 为 0.95 时从裸核的 6.93 分钟拉长到 138.6 分钟,并用峰值占比区分推起来的峰(中位数 0.538)与长起来的峰(0.093)。第三章用尾形稀疏化模拟评论流,用时间重标审计拟合,发现间隔检验几乎看不见成团,计数离散度检验则拒绝平坦泊松模型 0.646 的次数。第四章用小崎递推、补偿子恒等式与黄金分割求最大似然估计,示意帖的分支比 95 % 区间从 0.42 到 0.92,两小时估计的标准差约 0.186。 第五章计算外部给量停止后的余震存量与半衰期,回测中向账号均值收缩把预报误差中位数从 0.589 降到 0.418,并说明凸性使期望乘数被抬高到 3.40。第六章说明推流冒充口碑的伪临界:真实分支比 0.3 的推流帖被朴素估计成 0.864,把推流窗口写进模型后为 0.276,朴素估计把推流产出高估 5.15 倍。第七章考辨《战国策》曾参杀人(分支比 0.8 时三次告知同源的概率 0.6400)与大森房吉的余震律(伽马混合下 p 为 1.4)。第八章给出前两小时估分支比再决定推流的闭环、小工具「裂变分支比估计器」与预注册推流实验(每组 43 条帖子,协变量调整后 13 条)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Hits Are Aftershocks: the Branching Ratio and the 1/(1-n) Multiplier — Once comments are seen as family trees, "heat" splits into an outside rate $\mu$ and a branching ratio $n$ that can be measured separately. The expected number of comments that one outside comment eventually brings is $1/(1-n)$; as $n$ moves from 0.8 to 0.9 the multiplier doubles from 5.0 to 10.0. At one and the same $n$, post sizes can differ several-fold, and the largest tenth of posts holds more than a quarter of all comments, differences that have nothing to do with the quality of the content. Published research on video views supports the self-exciting cascade framework and finds that most content never bursts, with the decay of bursts changing with the branching ratio. The next chapter turns to time: whether the same aftershocks were pushed up from outside at a stroke or grew bit by bit from inside is told by the decay rate and by the shape of the peak. (https://5dlibrary.com/answers/en/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c01/) 第1章 爆款是余震:分支比与 1/(1-n) 乘数 — 把评论看成族谱以后,「热度」被拆成外来速率 $\mu$ 与分支比 $n$ 两个可以分别测量的数。每一条外来评论最终带来的评论总数期望是 $1/(1-n)$;$n$ 从 0.8 到 0.9,乘数从 5.0 翻到 10.0。同样的 $n$ 之下,帖子规模的差异可以大到好几倍,最大的一成帖子装下全部评论的四分之一以上,这些差异与内容质量无关。已发表的视频观看研究支持自激级联的框架,并发现大多数内容根本不爆,爆发的衰减形态随分支比变化。下一章转向时间:同样的余震,是被外部一次推起来的,还是自己一点点长起来的,衰减速率与峰形会说出来。 (https://5dlibrary.com/answers/zh/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c01/) - Ch.2 Exogenous and Endogenous: Three Decay Classes under Power-Law Memory and the Half-Life of the Exponential Kernel — Chapter 2 reproduces Crane and Sornette's decay classes under power-law memory, gives the half-life under exponential memory, stretched from the bare kernel's 6.93 minutes to 138.6 minutes at n of 0.95, and separates pushed peaks (median peak fraction 0.538) from grown ones (0.093). (https://5dlibrary.com/answers/en/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c02/) 第2章 外生与内生:幂律记忆的三类衰减与指数核的半衰期 — 第二章在幂律记忆下复现克兰与索内特的衰减分类,在指数记忆下给出半衰期,n 为 0.95 时从裸核的 6.93 分钟拉长到 138.6 分钟,并用峰值占比区分推起来的峰(中位数 0.538)与长起来的峰(0.093)。 (https://5dlibrary.com/answers/zh/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c02/) - Ch.3 Thinning and Time Rescaling: Simulating a Comment Stream, Then Auditing It — Chapter 3 simulates comment streams by Ogata's thinning and audits fits by time rescaling, finding that the gap test can hardly see clumping while the count-dispersion test rejects a flat Poisson model 0.646 of the time. (https://5dlibrary.com/answers/en/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c03/) 第3章 稀疏化与时间重标:模拟一条评论流,再审计它 — 第三章用尾形稀疏化模拟评论流,用时间重标审计拟合,发现间隔检验几乎看不见成团,计数离散度检验则拒绝平坦泊松模型 0.646 的次数。 (https://5dlibrary.com/answers/zh/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c03/) - Ch.4 Log-Likelihood, Grid and Golden Section: How Precisely Two Hours Pin Down the Branching Ratio — Chapter 4 finds maximum likelihood estimates with Ozaki's recursion, the compensator identity and golden-section search; the illustrative post's 95 % interval for the branching ratio runs from 0.42 to 0.92, and two-hour estimates have a standard deviation of about 0.186. (https://5dlibrary.com/answers/en/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c04/) 第4章 对数似然、网格与黄金分割:两小时能把分支比估多准 — 第四章用小崎递推、补偿子恒等式与黄金分割求最大似然估计,示意帖的分支比 95 % 区间从 0.42 到 0.92,两小时估计的标准差约 0.186。 (https://5dlibrary.com/answers/zh/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c04/) - Ch.5 Half-Life and the Stock of Aftershocks: Forecasting the Tail of a Post — Chapter 5 computes the stock and half-life of aftershocks once outside traffic stops; in a backtest, shrinkage toward the account mean cuts the median forecast error from 0.589 to 0.418, and convexity lifts the expected multiplier to 3.40. (https://5dlibrary.com/answers/en/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c05/) 第5章 半衰期与余震总量:预报一条帖子的尾巴 — 第五章计算外部给量停止后的余震存量与半衰期,回测中向账号均值收缩把预报误差中位数从 0.589 降到 0.418,并说明凸性使期望乘数被抬高到 3.40。 (https://5dlibrary.com/answers/zh/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c05/) - Ch.6 Paid Reach Posing as Word of Mouth: Non-Stationary Outside Rates and Apparent Criticality — Chapter 6 shows the apparent criticality of paid reach posing as word of mouth: a pushed post with a true branching ratio of 0.3 is estimated naively as 0.864 and as 0.276 once the push window enters the model, the naive fit overstating the push's yield 5.15-fold. (https://5dlibrary.com/answers/en/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c06/) 第6章 推流冒充口碑:非平稳外生率与伪临界 — 第六章说明推流冒充口碑的伪临界:真实分支比 0.3 的推流帖被朴素估计成 0.864,把推流窗口写进模型后为 0.276,朴素估计把推流产出高估 5.15 倍。 (https://5dlibrary.com/answers/zh/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c06/) - Ch.7 Eastern and Western Sources Examined: "Zeng Shen Has Killed a Man" and Omori's Aftershock Law — Chapter 7 examines "Zeng Shen has killed a man" in the Zhan guo ce (at branching ratio 0.8, three reports share one source with probability 0.6400) and Omori's aftershock law (p of 1.4 under a gamma mixture). (https://5dlibrary.com/answers/en/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c07/) 第7章 东西方原典考辨:曾参杀人与大森余震律 — 第七章考辨《战国策》曾参杀人(分支比 0.8 时三次告知同源的概率 0.6400)与大森房吉的余震律(伽马混合下 p 为 1.4)。 (https://5dlibrary.com/answers/zh/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c07/) - Ch.8 Closing the Loop: Estimate the Branching Ratio in the First Two Hours, Then Decide Whether to Push — Chapter 8 gives the loop of estimating the branching ratio in the first two hours before deciding on a push, the Cascade Branching-Ratio Estimator, and a pre-registered push experiment (43 posts per group, 13 after covariate adjustment). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c08/) 第8章 实证闭环:前两小时估分支比,再决定是否追加推流 — 第八章给出前两小时估分支比再决定推流的闭环、小工具「裂变分支比估计器」与预注册推流实验(每组 43 条帖子,协变量调整后 13 条)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/hawkes-self-exciting-comment-cascade-branching-ratio-half-life/c08/) ## 5DL-B-235 · Fisher Information Matrices and the Cramér-Rao Bound: the Fewest Questions Needed to Estimate a Budget or a Birth Time 中文书名:费希尔信息矩阵、克拉美罗方差下界与预算及出生时刻探询最少题数母典 DOI: pending Cite: Fan, Y. (2026). Fisher Information Matrices and the Cramér-Rao Bound: the Fewest Questions Needed to Estimate a Budget or a Birth Time (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: Whether an owner is sounding out a budget before a sale or chart-casting software is rectifying a birth time, questionnaires tend to grow longer and longer. This book refutes the common belief that more questions mean more precision: when questions share a source of bias, information hits a ceiling, and every extra question loses some respondents. Its prices, budgets, question banks and discrimination parameters are illustrative; all calculations are generated by its programs; it is not individual business advice; and its treatment of birth times covers only estimation theory and questionnaire design, assesses the predictive validity of no system and makes no reading of any individual's birth time. Chapter 1 introduces Fisher information and the Cramér-Rao bound and shows that going below the bound is possible only through bias and prior. Chapter 2 carries the main argument: at a shared-bias correlation of 0.5 the sixth question adds only 0.0476 of the first question's information and five questions already reach 0.8333 of the ceiling; once dropout is counted, the best length is two to five questions. The project blueprint stated the "sixth question" as a fact; the book corrects it to a model consequence that depends on the correlation. Chapter 3 rereads a price ladder through a grid posterior and shows how anchored clusters inflate the confidence a system reports. In the simulations of Chapter 4, five adaptive questions reach an error of 0.3866, better than the 0.3969 of eight fixed ones, and stopping on precision needs only 4.33 questions on average; channel capacity sets a floor that no way of asking can beat. Chapter 5 uses the Fisher information matrix to handle a yes-saying response style, and a design with three forward and two reverse-keyed questions pushes the entanglement term down to 0.0344. Chapter 6 treats birth-time rectification as an estimation problem: a record rounded to the hour lands in the wrong block with probability 0.2500; with discrimination of 0.7 against 0.3 a sequential test still needs 9.35 questions on average between adjacent blocks; and when the questions do not discriminate, a standard of eight matches out of ten is passed by luck among twelve candidates with probability 0.4908. Chapter 7 examines "knocking at both ends" in the Analects (6.417 questions on average over twelve levels against 3.667 for bisection) and Fisher's 1925 "Theory of Statistical Estimation" (median efficiency 0.6366). Chapter 8 gives a five-question pre-sale budget questionnaire (the demonstration stops at the fourth question with a band of 3,173 to 8,905), a read-only question order, the distilled "fewest-question adaptive questionnaire" tool and a pre-registered test plan with 555 per group, 1,110 in all. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:成交前摸预算、排盘软件校准出生时刻,问卷都习惯越问越长。本书证伪「题数越多越准」这条常识:题目共享同一个偏差来源时,信息会撞上天花板,而每多问一题都有人中途离开。书中的价格、预算、题库与判别力参数均为示意;全部计算由书中程序生成;本书是非个人化经营建议;涉及出生时刻的部分只讨论估计理论与问卷设计,不评价任何体系的预测效度,本书不做任何个人占断。 第一章介绍费希尔信息与克拉美罗下界,说明低于下界只能靠偏差与先验。第二章是全书主轴:共享偏差的相关为 0.5 时,第六题的信息增量只有首题的 0.0476,五题已拿到天花板的 0.8333;把流失记进账本后,最优题数在二到五之间。立项蓝图把「第六题」写成定论,本书修正为依赖相关强度的模型推论。第三章用网格后验重读价格阶梯,并说明锚定簇会让系统报告的把握偏高。第四章的模拟中,自适应五题的误差 0.3866 优于固定八题的 0.3969,按精度停止平均只需 4.33 题;信道容量给出任何问法都绕不过的题数下界。第五章用费希尔信息矩阵处理「爱说是」的作答风格,三正两反的设计把纠缠项压到 0.0344。 第六章把出生时刻校准当作估计问题:取整到整点的记录有 0.2500 的机会落在错的时段;判别力为 0.7 对 0.3 时,序贯检验在相邻两时段间平均仍需 9.35 题;题目毫无判别力时,十题八中的标准在十二个候选中偶然达标的概率是 0.4908。第七章考辨《论语·子罕》的「叩其两端」(十二级上逐级叩问平均 6.417 题,二分只要 3.667 题)与费希尔 1925 年的《统计估计理论》(中位数效率 0.6366)。第八章给出成交前五题预算问卷(示范在第四题停止,区间 3,173 到 8,905)、只读题序、蒸馏小工具「最少题数自适应问卷器」与每组 555 人、共 1,110 人的预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Fisher Information and the Cramér-Rao Bound: How Precise Can a Questionnaire Be? — Fisher information is a ruler for how much a question tells you; the information of independent questions adds up, and the Cramér-Rao bound says the variance of an unbiased estimator cannot fall below its reciprocal. Program p11 shows that going below the bound is possible only through bias and prior, and p12 shows that a yes/no question concentrates its information near its threshold, so a five-step ladder is precise only in the middle. The next chapter removes the premise of independence and shows how correlated questions drive information into a ceiling. (https://5dlibrary.com/answers/en/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c01/) 第1章 费希尔信息与克拉美罗下界:一份问卷最多能有多准 — 费希尔信息是一把量「一道题告诉你多少」的尺子;独立题目的信息相加,克拉美罗下界说无偏估计的方差不低于信息的倒数。p11 说明低于下界只能靠偏差与先验,p12 说明是非题的信息集中在门槛附近,五级阶梯只在中段准。下一章拆掉「独立」这个前提,看相关题目如何让信息撞上天花板。 (https://5dlibrary.com/answers/zh/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c01/) - Ch.2 The Information Ceiling of Correlated Questions: Why the Sixth Question Is Nearly Wasted — Chapter 2 carries the main argument: at a shared-bias correlation of 0.5 the sixth question adds only 0.0476 of the first question's information and five questions already reach 0.8333 of the ceiling; once dropout is counted, the best length is two to five questions. The project blueprint stated the "sixth question" as a fact; the book corrects it to a model consequence that depends on the correlation. (https://5dlibrary.com/answers/en/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c02/) 第2章 相关题的信息天花板:第六题之后为什么几乎白问 — 第二章是全书主轴:共享偏差的相关为 0.5 时,第六题的信息增量只有首题的 0.0476,五题已拿到天花板的 0.8333;把流失记进账本后,最优题数在二到五之间。立项蓝图把「第六题」写成定论,本书修正为依赖相关强度的模型推论。 (https://5dlibrary.com/answers/zh/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c02/) - Ch.3 An Item Response Model for Budget Questions: How Tightly Do Five Ladder Answers Pin Down a Budget? — Five ladder answers yield a posterior distribution, not a number. With three "yes" and two "no" answers the posterior mean is 3,733, the interval runs from 2,214 to 6,313, and the posterior standard deviation is already close to the Bayesian information bound; customers at either end of the ladder get much wider intervals. A price ladder is a natural testlet, and a shared shift makes the true error exceed the reported one, by a factor of 1.857 when its standard deviation is 1.2, with coverage falling below one half while the system notices nothing. The next chapter takes another road: instead of fixing the five steps in advance, it lets the price of each question move with the answers before it. (https://5dlibrary.com/answers/en/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c03/) 第3章 预算题的项目反应模型:五个阶梯答案能把预算锁多紧 — 五个阶梯答案给出的是一个后验分布,而不是一个数。三个「是」两个「否」时,期望估计 3,733,区间从 2,214 到 6,313,后验标准差已接近贝叶斯信息下界;落在阶梯两端的客户,区间要宽得多。价格阶梯是天然的题组,共享偏移让真实误差高出报告值,标准差为 1.2 时高出 1.857 倍,覆盖率跌破一半,而系统自己毫无察觉。下一章讨论另一条路:不再固定问哪五档,而是让每一道题的价格跟着前面的回答移动。 (https://5dlibrary.com/answers/zh/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c03/) - Ch.4 Adaptive Selection: the Maximum-Information Rule and an Information-Theoretic Floor on the Number of Questions — The maximum-information rule makes the next question follow the current estimate. In the simulation five adaptive questions have an error of 0.3866, slightly better than the 0.3969 of eight fixed ones, and stopping on precision needs only 4.33 questions on average. Information theory gives a hard floor: each yes/no question carries at most $1 - H(e)$ bits, and with an answer error rate of 0.2 choosing one of twelve blocks takes at least 13 questions. The next chapter widens the view from one parameter to two: when the answers are also mixed with a "yes-saying" response style, the Fisher information matrix tells us how to write the questions. (https://5dlibrary.com/answers/en/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c04/) 第4章 自适应选题:最大信息准则与题数的信息论下界 — 最大信息准则让下一题跟着当前估计走。模拟中自适应五题的误差 0.3866,略优于固定八题的 0.3969;按精度停止平均只需 4.33 题。信息论给出硬性下界:每道是非题至多 $1 - H(e)$ 比特,回答错误率 0.2 时从十二个时段中选一至少要问 13 题。下一章把视野从一个参数扩展到两个:当回答里还混着「爱说是」的作答风格时,费希尔信息矩阵告诉我们该怎样出题。 (https://5dlibrary.com/answers/zh/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c04/) - Ch.5 The Fisher Information Matrix: Yes-Saying, Reverse-Keyed Questions and Optimal Design — Chapter 5 uses the Fisher information matrix to handle a yes-saying response style, and a design with three forward and two reverse-keyed questions pushes the entanglement term down to 0.0344. (https://5dlibrary.com/answers/en/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c05/) 第5章 费希尔信息矩阵:爱说「是」的作答风格、反向题与最优设计 — 第五章用费希尔信息矩阵处理「爱说是」的作答风格,三正两反的设计把纠缠项压到 0.0344。 (https://5dlibrary.com/answers/zh/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c05/) - Ch.6 Estimation Theory for a Birth Time: Rounding Error, Sequential Tests and "It Matched Afterwards" — Chapter 6 treats birth-time rectification as an estimation problem: a record rounded to the hour lands in the wrong block with probability 0.2500; with discrimination of 0.7 against 0.3 a sequential test still needs 9.35 questions on average between adjacent blocks; and when the questions do not discriminate, a standard of eight matches out of ten is passed by luck among twelve candidates with probability 0.4908. (https://5dlibrary.com/answers/en/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c06/) 第6章 出生时刻的估计理论:取整误差、序贯检验与「事后对上」 — 第六章把出生时刻校准当作估计问题:取整到整点的记录有 0.2500 的机会落在错的时段;判别力为 0.7 对 0.3 时,序贯检验在相邻两时段间平均仍需 9.35 题;题目毫无判别力时,十题八中的标准在十二个候选中偶然达标的概率是 0.4908。 (https://5dlibrary.com/answers/zh/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c06/) - Ch.7 Classical Sources East and West: "Knocking at Both Ends" and Fisher's Amount of Information — Chapter 7 examines "knocking at both ends" in the Analects (6.417 questions on average over twelve levels against 3.667 for bisection) and Fisher's 1925 "Theory of Statistical Estimation" (median efficiency 0.6366). (https://5dlibrary.com/answers/en/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c07/) 第7章 东西方原典考辨:「叩其两端」与费希尔的信息量 — 第七章考辨《论语·子罕》的「叩其两端」(十二级上逐级叩问平均 6.417 题,二分只要 3.667 题)与费希尔 1925 年的《统计估计理论》(中位数效率 0.6366)。 (https://5dlibrary.com/answers/zh/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c07/) - Ch.8 Closing the Loop: a Five-Question Pre-Sale Budget Questionnaire, a Read-Only Question Order and a Pre-Registered Test — Chapter 8 gives a five-question pre-sale budget questionnaire (the demonstration stops at the fourth question with a band of 3,173 to 8,905), a read-only question order, the distilled "fewest-question adaptive questionnaire" tool and a pre-registered test plan with 555 per group, 1,110 in all. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c08/) 第8章 实证闭环:成交前五题预算问卷、只读题序与预注册检验 — 第八章给出成交前五题预算问卷(示范在第四题停止,区间 3,173 到 8,905)、只读题序、蒸馏小工具「最少题数自适应问卷器」与每组 555 人、共 1,110 人的预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/fisher-information-cramer-rao-fewest-questions-budget-birth-time/c08/) ## 5DL-B-236 · Persistent Homology and Betti Numbers: Structural Voids in High-Net-Worth Private Communities and the Review of Fake Cliques 中文书名:持续同调拓扑贝蒂数、高净值私域社群结构空洞与虚假抱团水军识别母典 DOI: pending Cite: Fan, Y. (2026). Persistent Homology and Betti Numbers: Structural Voids in High-Net-Worth Private Communities and the Review of Fake Cliques (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Every community and member in the book is synthetic data with fixed seeds, and no data from any real community or person is used; all calculations are generated by its programs, and its parameters are illustrative. The book is not individual business advice, and its "detection of paid boosters" produces only a list of suspects for human review, never an automatic penalty for any member. Abstract: Organisers of private communities usually judge whether a circle is real by its activity and a cohesion index. This book refutes that intuition: a community with plenty of engagement and triangles can have no real circle at all, most of its triangles coming from a solid block of a few synchronised boosting accounts, while the mark of a real circle is a loop that survives across many scales, the one-dimensional barcode of persistent homology. Every community and member in the book is synthetic data with fixed seeds, and no data from any real community or person is used; all calculations are generated by its programs, and its parameters are illustrative. The book is not individual business advice, and its "detection of paid boosters" produces only a list of suspects for human review, never an automatic penalty for any member. Chapter 1 shows that a boosting block can raise transitivity to 0.996 (0.671 in a genuine community) while being topologically solid, with first Betti number zero. Chapter 2 computes the zero-dimensional barcode by union-find and the one-dimensional barcode by column reduction mod 2; across two hundred synthetic communities total likes have an AUC of only 0.427 while L3, the sum of the three longest one-dimensional bars, has 0.975, falling to 0.673 when the genuine core of the padded communities grows to 30, where short zero-dimensional bars must take over. Chapter 3 discusses the Jaccard distance and the stability of barcodes. Chapter 4 draws an alarm line of 0.1786 with a permutation null model, crossed by 0.775 of genuine communities, and sets a minimum community size. Chapter 5 turns short zero-dimensional bars into a review list: the catch rate is 0.996 and the false-alarm rate per genuine account 0.0127, yet when boosting accounts are rare 0.717 of the people listed are innocent; camouflage cuts the catch rate to 0.028 while the AUC of L3 stays at 0.900. Chapter 6 follows a twelve-month hollowing in which likes stay at 0.998 times the baseline while the synchrony and loop alarms sound in turn. Chapter 7 examines "broad and not partisan" in the Analects and Euler's polyhedron formula. Chapter 8 gives the monthly checkup procedure, the Community Topology Checkup tool and a pre-registered test plan (41 audited communities per group, 93 reviewed list entries). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:私域社群的主事者习惯用活跃度与抱团指数判断圈层真假。本书证伪这一直觉:互动多、三角多的社群可以没有任何真实的圈层结构,大量三角来自少数同步互刷账号组成的实心团块;真实圈层的标志是一条跨越多个尺度都不消失的环,即持续同调的一维条码。书中所有社群与成员都是固定种子的合成数据,不使用任何真实社群或个人的数据;全部计算由书中程序生成,参数为示意;本书是非个人化经营建议,「水军识别」只产出供人工复核的嫌疑名单,不自动处罚任何成员。 第一章说明互刷团块可以把传递性抬到 0.996(真实社群为 0.671),而它在拓扑上是实心的,一维贝蒂数为零。第二章用并查集求零维条码、用模 2 列约化求一维条码,在两百个合成社群上,点赞总数的 AUC 只有 0.427,一维条码前三长之和 L3 为 0.975;补量社群的真实核心增至 30 人时降到 0.673,此时须靠零维短条。第三章讨论 Jaccard 距离与条码稳定性。第四章用置换零模型画出警戒线 0.1786,真实社群中 0.775 越过它,并给出社群人数的下限。 第五章把零维短条做成嫌疑名单:检出率 0.996、真实账号误报率 0.0127,但互刷账号占比很低时名单上 0.717 的人是无辜的;伪装使检出率跌到 0.028,L3 的 AUC 仍为 0.900。第六章做十二个月的掏空序列,点赞维持在基线的 0.998 倍,同步报警与环的报警先后响起。第七章考辨《论语》「周而不比」与欧拉多面体公式。第八章给出社群月度拓扑体检流程、社群拓扑体检器与预注册检验方案(每组 41 个审计社群、复核 93 个名单账号)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Activity Is Not a Loop: Triangle Closure, Transitivity and Blocks of Mutual Boosting — Chapter 1 shows that a boosting block can raise transitivity to 0.996 (0.671 in a genuine community) while being topologically solid, with first Betti number zero. (https://5dlibrary.com/answers/en/persistent-homology-betti-community-voids-fake-cliques/c01/) 第1章 活跃不等于有环:三角闭合、传递性与互刷团块 — 第一章说明互刷团块可以把传递性抬到 0.996(真实社群为 0.671),而它在拓扑上是实心的,一维贝蒂数为零。 (https://5dlibrary.com/answers/zh/persistent-homology-betti-community-voids-fake-cliques/c01/) - Ch.2 Persistent Homology: the Rips Filtration, Union-Find for Dimension Zero and Reduction Mod 2 for Dimension One — Chapter 2 computes the zero-dimensional barcode by union-find and the one-dimensional barcode by column reduction mod 2; across two hundred synthetic communities total likes have an AUC of only 0.427 while L3, the sum of the three longest one-dimensional bars, has 0.975, falling to 0.673 when the genuine core of the padded communities grows to 30, where short zero-dimensional bars must take over. (https://5dlibrary.com/answers/en/persistent-homology-betti-community-voids-fake-cliques/c02/) 第2章 持续同调:Rips 过滤、并查集零维条码与模 2 约化一维条码 — 第二章用并查集求零维条码、用模 2 列约化求一维条码,在两百个合成社群上,点赞总数的 AUC 只有 0.427,一维条码前三长之和 L3 为 0.975;补量社群的真实核心增至 30 人时降到 0.673,此时须靠零维短条。 (https://5dlibrary.com/answers/zh/persistent-homology-betti-community-voids-fake-cliques/c02/) - Ch.3 Distance Comes Before Topology: Jaccard Distance, the Gap Between Classes and the Stability of Barcodes — Barcodes stand on a distance, and the distance must pass inspection first. This book uses only the table of which posts each member took part in, without counts, text or identities. The median Jaccard distance between boosting accounts is 0.152, the median distance from a genuine member to her nearest neighbour is 0.472, and between the two classes lies only the narrow gap from 0.333 to 0.344. The stability theorem guarantees that when each distance moves by ε, zero-dimensional bars move by at most ε and L3 by at most 6ε. Switching to the cosine distance changes the AUC of L3 from 0.983 to 0.980, hardly at all, but the zero-dimensional alarm line must be reset. The next question is how long a one-dimensional bar must be to count as more than noise. Without something to compare with, "long" means nothing; Chapter 4 supplies the comparison with a permutation null model. (https://5dlibrary.com/answers/en/persistent-homology-betti-community-voids-fake-cliques/c03/) 第3章 距离先于拓扑:Jaccard 距离、类间缝隙与条码稳定性 — 条码在距离之上,距离要先过关。本书只用「成员参与过哪些帖子」这一张表,不用次数、正文与身份。互刷账号之间的 Jaccard 距离中位数是 0.152,真实成员到最近邻的中位数是 0.472,两类之间只有 0.333 到 0.344 的一条窄缝。稳定性定理保证每个距离只动 ε 时零维条码最多动 ε,L3 最多动 6ε;换成余弦距离,L3 的 AUC 从 0.983 变成 0.980,几乎不变,零维报警线却必须重定。下一个问题是:一根一维条要多长,才不是噪声?没有比较对象,「长」字没有意义,第四章用置换零模型给出比较对象。 (https://5dlibrary.com/answers/zh/persistent-homology-betti-community-voids-fake-cliques/c03/) - Ch.4 How Long Is Long: a Permutation Null Model, an Alarm Line and the Dose-Response of Community Size — Chapter 4 draws an alarm line of 0.1786 with a permutation null model, crossed by 0.775 of genuine communities, and sets a minimum community size. (https://5dlibrary.com/answers/en/persistent-homology-betti-community-voids-fake-cliques/c04/) 第4章 多长才算长:置换零模型、警戒线与社群规模的剂量反应 — 第四章用置换零模型画出警戒线 0.1786,真实社群中 0.775 越过它,并给出社群人数的下限。 (https://5dlibrary.com/answers/zh/persistent-homology-betti-community-voids-fake-cliques/c04/) - Ch.5 The Review List: Short Zero-Dimensional Bars, Camouflage and Base Rates — Chapter 5 turns short zero-dimensional bars into a review list: the catch rate is 0.996 and the false-alarm rate per genuine account 0.0127, yet when boosting accounts are rare 0.717 of the people listed are innocent; camouflage cuts the catch rate to 0.028 while the AUC of L3 stays at 0.900. (https://5dlibrary.com/answers/en/persistent-homology-betti-community-voids-fake-cliques/c05/) 第5章 嫌疑名单:零维短条码、伪装与基础比率 — 第五章把零维短条做成嫌疑名单:检出率 0.996、真实账号误报率 0.0127,但互刷账号占比很低时名单上 0.717 的人是无辜的;伪装使检出率跌到 0.028,L3 的 AUC 仍为 0.900。 (https://5dlibrary.com/answers/zh/persistent-homology-betti-community-voids-fake-cliques/c05/) - Ch.6 The Monthly Checkup: a Barcode Time Series of Hollowing, and the Cost of Computation — Chapter 6 follows a twelve-month hollowing in which likes stay at 0.998 times the baseline while the synchrony and loop alarms sound in turn. (https://5dlibrary.com/answers/en/persistent-homology-betti-community-voids-fake-cliques/c06/) 第6章 月度体检:掏空过程的条码时间序列与计算规模 — 第六章做十二个月的掏空序列,点赞维持在基线的 0.998 倍,同步报警与环的报警先后响起。 (https://5dlibrary.com/answers/zh/persistent-homology-betti-community-voids-fake-cliques/c06/) - Ch.7 Eastern and Western Classics Examined: "Broad, Not Partisan" and Euler's Polyhedron Formula — The "broad and not partisan" of the Analects can be written as two computable indices: zhou as the one-dimensional L3, bi as short zero-dimensional bars. On synthetic data, 41 of 50 genuine ring communities are broad and not partisan and 48 of 50 padded communities partisan and not broad; but 49 of 50 genuine closed clubs are also partisan and not broad, indistinguishable from padded ones, and in communities with families close to half are both or neither. Euler's polyhedron formula holds for the five regular solids and fails for the seven-vertex torus, and the Betti numbers 1, 2, 1 explain why. Each classic draws a correct structural distinction, and each states it too broadly. The last chapter puts these structural quantities back into the daily running of a community: what should an organiser do each month, and what data would refute this book? (https://5dlibrary.com/answers/en/persistent-homology-betti-community-voids-fake-cliques/c07/) 第7章 东西方原典考辨:「周而不比」与欧拉多面体公式 — 《论语》的「周而不比」可以写成两个可算的指标:周对应一维条码 L3,比对应零维短条。在合成数据上,真实环形社群 50 个中有 41 个周而不比,补量社群 48 个比而不周;但真实的封闭俱乐部有 49 个比而不周,与补量社群无法区分,含家庭的社群里周与比兼有或皆无的接近一半。欧拉的多面体公式对五种正多面体成立,对七顶点环面不成立,贝蒂数 1、2、1 解释了原因。两部原典都给出了一个正确的结构区分,也都说得过宽。最后一章要把这些结构量放回社群的日常运营:一个主事者每个月该怎么做,什么样的数据会推翻本书? (https://5dlibrary.com/answers/zh/persistent-homology-betti-community-voids-fake-cliques/c07/) - Ch.8 Closing the Empirical Loop: the Community Topology Checkup, Human Review and a Pre-Registered Test — Chapter 8 gives the monthly checkup procedure, the Community Topology Checkup tool and a pre-registered test plan (41 audited communities per group, 93 reviewed list entries). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/persistent-homology-betti-community-voids-fake-cliques/c08/) 第8章 实证闭环:社群拓扑体检器、人工复核与预注册检验 — 第八章给出社群月度拓扑体检流程、社群拓扑体检器与预注册检验方案(每组 41 个审计社群、复核 93 个名单账号)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/persistent-homology-betti-community-voids-fake-cliques/c08/) ## 5DL-B-237 · Vickrey Second-Price Sealed Bids and Myerson Optimal Reserve Prices: Incentive-Compatible Auctions for One-of-a-Kind Luxury Pieces and Limited Seats 中文书名:维克里二价暗标、迈尔森最优保留价与孤品奢包限量席位激励相容拍卖母典 DOI: pending Cite: Fan, Y. (2026). Vickrey Second-Price Sealed Bids and Myerson Optimal Reserve Prices: Incentive-Compatible Auctions for One-of-a-Kind Luxury Pieces and Limited Seats (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: When a one-of-a-kind handbag or a set of limited seats is sold, the trade holds two intuitions: format sets the price, and a floor is a mind game, best omitted or set as high as possible. This book refutes their shared premise. With independent, symmetric, private values and risk-neutral buyers, the English, Dutch, first-price and second-price formats have the same expected revenue; what systematically raises expected revenue is a reserve set by virtual value, worth more the fewer the buyers; and attracting one more genuine, qualified buyer is often worth more than getting the reserve exactly right. Its parameters are illustrative and all calculations are generated by its programs; it is not individual business advice or a legal opinion, formal auctions follow local rules checked by a person, and urgency may come only from real stock and real deadlines, never from fictitious bidders or shill bids. Chapter 1 checks revenue equivalence by exact integration in fractions, and uses twenty thousand simulations on common random numbers to show the same mean with different risk: the standard deviation of the price is 16185 yuan for the second-price sealed bid and 9930 yuan for the first-price one. Chapter 2 introduces virtual values and the optimal reserve: under uniform values the relative gain from the reserve falls from 0.2500 with two buyers to 0.0078 with five, but with an exponential tail the gain rises from 5,666 yuan with one buyer to 6,843 yuan with two, a correction to the project blueprint; one more genuine buyer beats the optimal reserve in every case when the seller's own value is zero, and the result reverses when that value is high and values have a long tail. Chapter 3 draws up the bill for a reserve: with three buyers the chance of no sale rises to 0.2441, and buyers give up 1.868 yuan for every 1 yuan the owner gains. Chapter 4 covers truth-telling, overbidding and shill bids (a hidden shill takes 16,000 yuan from each winner) and locks bids and reserve with SHA-256 commitment and reveal. Chapter 5 discusses where revenue equivalence breaks: risk aversion, correlated values and bidding rings, with the optimal reserve absorbing 0.6333 of the loss from a ring of three. Chapter 6 extends the single piece to limited seats: with twice as many applicants as seats, the (k+1)-th price earns 1.0806 times the optimal posted price. Chapter 7 examines tax-farm bidding in the Song shi with Su Zhe's proposal to fix a middle price and ban outbidding (in an illustrative model, bidding one's own estimate gives a chance of default of 0.9691), and the Babylonian marriage auction recorded by Herodotus (receipts 3.373 times the dowries). Chapter 8 gives a six-step sealed-bid procedure, a public notice for second-price seats, the Sealed-Bid Reserve Price Calculator and a pre-registered test plan: with three buyers the no-sale test needs 33 pieces per group and the revenue test 164. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:出售孤品手袋或限量席位时,行内流传两种直觉:格式决定价格,公开喊价或一价暗标一定卖得更贵;底价只是心理游戏,要么不设,要么越高越好。本书证伪它们的共同前提。在独立、对称、私人价值且风险中性的标准情形下,英式、荷式、一价与二价四种格式的期望收入相同;真正能系统抬高期望收入的是按虚拟价值设定的保留价,而且买家越少它越值钱;与此同时,多招一位真实的合格买家常常比把保留价算准更值钱。书中的估值分布、保有价值、买家人数与价格均为示意;全部计算由书中程序生成;本书是非个人化经营建议,也不是法律意见,正式拍卖以上架前人工复核的当地法规为准;紧迫感只能来自真实库存与真实截止,不虚构竞价者,不托价。 第一章用分数精确积分核对收入等价,并用两万次共同随机数的模拟说明平均相同而风险不同:二价暗标的成交价标准差 16185 元,一价暗标 9930 元。第二章引入虚拟价值与最优保留价:均匀分布下保留价的相对增益从两位买家的 0.2500 降到五位的 0.0078,但在指数长尾下增益从一位买家的 5,666 元先升到两位的 6,843 元,这是对立项蓝图的修正;多一位真实买家在卖方自留价值为零时全部胜过最优保留价,在自留价值较高且有长尾时反转。第三章算保留价的账单:三位买家时流拍概率升到 0.2441,店主每多拿 1 元,买家付出 1.868 元。第四章讲二价暗标的说真话性质、过度出价与托价,暗中托价每笔从赢家身上拿走 16,000 元,并用 SHA-256 承诺与揭示锁定出价与保留价。 第五章讨论收入等价破裂的情形:风险厌恶、估值相关与围标,最优保留价能挡住三人围标损失的 0.6333。第六章把单件推广到限量席位:申请者为席位两倍时,第 k+1 价的利润是最优标价的 1.0806 倍。第七章考辨《宋史》的买扑与苏辙「立定中价,不许添刬」之议(示意模型中按自身估计出价的违约概率达 0.9691)以及希罗多德记载的巴比伦婚配拍卖(收入是嫁资的 3.373 倍)。第八章给出孤品暗标六步流程、席位二价公示、暗标保留价计算器与预注册检验方案:三位买家时流拍检验每组 33 件,所得检验每组 164 件。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 One Expectation for Four Formats: Exact Integration and Monte Carlo for Revenue Equivalence — Chapter 1 checks revenue equivalence by exact integration in fractions, and uses twenty thousand simulations on common random numbers to show the same mean with different risk: the standard deviation of the price is 16185 yuan for the second-price sealed bid and 9930 yuan for the first-price one. (https://5dlibrary.com/answers/en/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c01/) 第1章 四种格式的同一个期望:收入等价的精确积分与蒙特卡洛 — 第一章用分数精确积分核对收入等价,并用两万次共同随机数的模拟说明平均相同而风险不同:二价暗标的成交价标准差 16185 元,一价暗标 9930 元。 (https://5dlibrary.com/answers/zh/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c01/) - Ch.2 Virtual Values and Myerson's Reserve: the Fewer the Buyers, the More the Floor Is Worth — Chapter 2 introduces virtual values and the optimal reserve: under uniform values the relative gain from the reserve falls from 0.2500 with two buyers to 0.0078 with five, but with an exponential tail the gain rises from 5,666 yuan with one buyer to 6,843 yuan with two, a correction to the project blueprint; one more genuine buyer beats the optimal reserve in every case when the seller's own value is zero, and the result reverses when that value is high and values have a long tail. (https://5dlibrary.com/answers/en/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c02/) 第2章 虚拟价值与迈尔森保留价:买家越少,底价越值钱 — 第二章引入虚拟价值与最优保留价:均匀分布下保留价的相对增益从两位买家的 0.2500 降到五位的 0.0078,但在指数长尾下增益从一位买家的 5,666 元先升到两位的 6,843 元,这是对立项蓝图的修正;多一位真实买家在卖方自留价值为零时全部胜过最优保留价,在自留价值较高且有长尾时反转。 (https://5dlibrary.com/answers/zh/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c02/) - Ch.3 The Bill for a Reserve: No-Sales, Forgone Trades and the Wrong Distribution — Chapter 3 draws up the bill for a reserve: with three buyers the chance of no sale rises to 0.2441, and buyers give up 1.868 yuan for every 1 yuan the owner gains. (https://5dlibrary.com/answers/en/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c03/) 第3章 保留价的账单:流拍、被放弃的成交与看错分布 — 第三章算保留价的账单:三位买家时流拍概率升到 0.2441,店主每多拿 1 元,买家付出 1.868 元。 (https://5dlibrary.com/answers/zh/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c03/) - Ch.4 The Truthful Sealed Bid: Dominant Strategies, Overbidding, Shill Bids and Commit-Reveal — Chapter 4 covers truth-telling, overbidding and shill bids (a hidden shill takes 16,000 yuan from each winner) and locks bids and reserve with SHA-256 commitment and reveal. (https://5dlibrary.com/answers/en/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c04/) 第4章 说真话的暗标:占优策略、过度出价、托价与承诺揭示 — 第四章讲二价暗标的说真话性质、过度出价与托价,暗中托价每笔从赢家身上拿走 16,000 元,并用 SHA-256 承诺与揭示锁定出价与保留价。 (https://5dlibrary.com/answers/zh/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c04/) - Ch.5 When Equivalence Breaks: Risk Aversion, Bidding Rings and Field Experiments — Chapter 5 discusses where revenue equivalence breaks: risk aversion, correlated values and bidding rings, with the optimal reserve absorbing 0.6333 of the loss from a ring of three. (https://5dlibrary.com/answers/en/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c05/) 第5章 等价何时破裂:风险厌恶、围标与实地实验 — 第五章讨论收入等价破裂的情形:风险厌恶、估值相关与围标,最优保留价能挡住三人围标损失的 0.6333。 (https://5dlibrary.com/answers/zh/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c05/) - Ch.6 Limited Seats: the (k+1)-th Price for k Identical Seats, Compared with a Posted Price — Chapter 6 extends the single piece to limited seats: with twice as many applicants as seats, the (k+1)-th price earns 1.0806 times the optimal posted price. (https://5dlibrary.com/answers/en/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c06/) 第6章 限量席位:k 个相同席位的第 k+1 价与标价对照 — 第六章把单件推广到限量席位:申请者为席位两倍时,第 k+1 价的利润是最优标价的 1.0806 倍。 (https://5dlibrary.com/answers/zh/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c06/) - Ch.7 Classical Sources East and West: Song Tax-Farm Bidding and Herodotus's Babylonian Marriage Auction — Chapter 7 examines tax-farm bidding in the Song shi with Su Zhe's proposal to fix a middle price and ban outbidding (in an illustrative model, bidding one's own estimate gives a chance of default of 0.9691), and the Babylonian marriage auction recorded by Herodotus (receipts 3.373 times the dowries). (https://5dlibrary.com/answers/en/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c07/) 第7章 东西方原典考辨:宋代买扑与希罗多德的巴比伦婚配拍卖 — 第七章考辨《宋史》的买扑与苏辙「立定中价,不许添刬」之议(示意模型中按自身估计出价的违约概率达 0.9691)以及希罗多德记载的巴比伦婚配拍卖(收入是嫁资的 3.373 倍)。 (https://5dlibrary.com/answers/zh/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c07/) - Ch.8 Closing the Loop: a Sealed-Bid Procedure for Unique Pieces, a Public Second-Price Rule for Seats, and a Pre-Registered Test — Chapter 8 gives a six-step sealed-bid procedure, a public notice for second-price seats, the Sealed-Bid Reserve Price Calculator and a pre-registered test plan: with three buyers the no-sale test needs 33 pieces per group and the revenue test 164. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c08/) 第8章 实证闭环:孤品暗标流程、席位二价公示与预注册检验 — 第八章给出孤品暗标六步流程、席位二价公示、暗标保留价计算器与预注册检验方案:三位买家时流拍检验每组 33 件,所得检验每组 164 件。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/vickrey-myerson-reserve-price-one-of-a-kind-luxury-auctions/c08/) ## 5DL-B-238 · Koopman Operators and Dynamic Mode Decomposition: Linearising Nonlinear Cash-Flow Swings and Forecasting Turning Points in Livestream Audience Curves 中文书名:库普曼算子动态模态分解、现金流非线性波动线性化与直播在线曲线拐点预测母典 DOI: pending Cite: Fan, Y. (2026). Koopman Operators and Dynamic Mode Decomposition: Linearising Nonlinear Cash-Flow Swings and Forecasting Turning Points in Livestream Audience Curves (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: Livestream teams and shop operators widely believe two things: audience curves and daily takings are essentially a mess, and the safest way to read a turning point is a moving average. This book tests both beliefs with the Koopman operator and dynamic mode decomposition: curves that look chaotic are often sums of a few decaying or oscillating modes, and moving averages carry a built-in lag at turning points that comes from the structure of the filter. All computation uses the Python standard library alone, with singular value and eigen decompositions built from Jacobi rotations. The book's audience curves, shop takings and parameters are illustrative; all calculations are generated by its programs; the book is not individual business advice, its cash-flow forecasts are not advice on financing, lending or investment, and every warning given by its small tool is only an internal prompt on which the operator must judge. Chapter 1 shows that the first two modes of an illustrative audience curve hold 0.9285 of its energy, but a random walk also reaches 0.872, so explained variance is not evidence and out-of-sample forecasting is; a moving average that removes a 12-minute swing turns 5.89 minutes late. Chapter 2 explains how a nonlinear system is lifted into a linear one, how a forced fit in raw coordinates gives a distorted eigenvalue of 0.744427, and why chaotic orbits have no finite set of modes. Chapter 3 builds standard-library decompositions from Jacobi rotations and shows why small singular values must not be taken from the matrix times its transpose. Chapter 4 assembles dynamic mode decomposition and shows that noise makes a swing that never decays look like one with modulus 0.7744 and a half-life of 2.7 minutes, which a forward-backward correction repairs. Chapter 5 compares three rules over two hundred random illustrative sessions: the mode warning has a median lead of 3.45 minutes, comes on average 13.13 minutes before the moving average and is premature in 0.055 of sessions, against 0.115 for the raw-curve rule. Chapter 6 applies the method to a shop's daily cash flow, with a 28-day forecast error of 2301 yuan against 4665 for "the same day last week"; taking logarithms is a Koopman lift that cuts the rank-3 error from 0.1645 to 0.0257. Chapter 7 examines "knowing the incipient" in the Book of Changes (ten minutes before the turn the second difference has a signal-to-noise ratio of only 0.031) and Fourier's analysis of heat (a fixed Fourier basis holds only 0.603 of the energy of a decaying swing). Chapter 8 gives the five-minute fitting procedure, the Online-Curve Turning-Point Early Warner and a pre-registered test plan: a lead difference of 3 minutes needs 51 sessions and the false-alarm test needs 68. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:做直播的团队和开店的经营者普遍相信两件事:在线曲线与日流水本质上是乱的;读拐点最稳妥的办法是移动平均。本书用库普曼算子与动态模态分解检验这两个信念:看似混乱的曲线常由少数衰减或振荡的模态叠加而成,而移动平均对拐点有来自滤波器结构的固有滞后。全部计算只用 Python 标准库,奇异值分解与特征分解由 Jacobi 旋转实现。书中的直播曲线、店铺流水与参数均为示意;全部计算由书中程序生成;本书是非个人化经营建议,现金流预报不构成融资、借贷或投资建议,书中小工具给出的每一条预警都只是内部提示,须由经营者人工判断。 第一章显示示意直播曲线的前两个模态装下 0.9285 的能量,但随机游走也有 0.872,所以方差解释不是证据,样本外预报才是;去掉 12 分钟起伏的移动平均,拐点迟到 5.89 分钟。第二章说明非线性系统怎样抬升成线性,原始坐标的硬拟合会给出扭曲的特征值 0.744427,而混沌轨道没有有限的模态。第三章用 Jacobi 旋转搭建标准库的分解,并说明为什么不能从矩阵乘以转置求小奇异值。第四章组装动态模态分解,指出噪声会把永不衰减的起伏估成模长 0.7744、半衰期 2.7 分钟,前向后向修正可以纠正。 第五章在两百场随机示意直播上比较三条规则:模态预警的提前量中位数 3.45 分钟,平均比移动平均早 13.13 分钟,过早误报 0.055,原始曲线规则则有 0.115。第六章把方法用到店铺日现金流,28 天预报误差 2301 元,「上周同一天」为 4665 元;取对数是一次库普曼抬升,秩 3 误差从 0.1645 降到 0.0257。第七章考辨《周易·系辞下》的「知几」(拐点前十分钟二阶差分的信噪比只有 0.031)与傅里叶的热分析(固定的傅里叶基只装下衰减起伏 0.603 的能量)。第八章给出每五分钟一次的拟合流程、在线曲线拐点预警器与预注册检验方案:检验 3 分钟的提前量差需要 51 场,检验过早误报需要 68 场。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Two or Three Modes: the Low-Rank Structure of a Livestream Audience Curve and the Built-In Lag of Moving Averages — Chapter 1 shows that the first two modes of an illustrative audience curve hold 0.9285 of its energy, but a random walk also reaches 0.872, so explained variance is not evidence and out-of-sample forecasting is; a moving average that removes a 12-minute swing turns 5.89 minutes late. (https://5dlibrary.com/answers/en/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c01/) 第1章 两三个模态:直播在线曲线的低秩结构与移动平均的固有滞后 — 第一章显示示意直播曲线的前两个模态装下 0.9285 的能量,但随机游走也有 0.872,所以方差解释不是证据,样本外预报才是;去掉 12 分钟起伏的移动平均,拐点迟到 5.89 分钟。 (https://5dlibrary.com/answers/zh/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c01/) - Ch.2 The Koopman Operator: Lifting Nonlinear Dynamics into Linear Ones, and the Limits of the Lift — Chapter 2 explains how a nonlinear system is lifted into a linear one, how a forced fit in raw coordinates gives a distorted eigenvalue of 0.744427, and why chaotic orbits have no finite set of modes. (https://5dlibrary.com/answers/en/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c02/) 第2章 库普曼算子:把非线性抬升成线性,以及抬升的边界 — 第二章说明非线性系统怎样抬升成线性,原始坐标的硬拟合会给出扭曲的特征值 0.744427,而混沌轨道没有有限的模态。 (https://5dlibrary.com/answers/zh/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c02/) - Ch.3 Jacobi Rotations: Symmetric Eigendecomposition and the Singular Value Decomposition with the Standard Library Alone — A Jacobi rotation turns one off-diagonal element into zero, off-diagonal energy can only flow onto the diagonal, and cyclic sweeps converge quadratically in their later stage: in program p31 six sweeps take an 8 by 8 Gram matrix from 9.217e-01 to 1.022e-16, with orthogonality error 2.7e-15. Singular values must not be obtained by the shortcut of multiplying by the transpose and taking roots: in program p32 the one-billionth singular value comes out negative on that road, with a relative error of 8.6e+00, while one-sided Jacobi errs by 2.6e-08, right at the limit set by rounding in the input. Program p33 confirms the theorem of Eckart and Young: the truncation error equals the root of the sum of the discarded squared singular values. With these standard-library factorisations in hand, the next chapter builds dynamic mode decomposition proper, computing eigenvalues and modes, and shows how noise disguises a swing that never decays as a signal that is fading away. (https://5dlibrary.com/answers/en/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c03/) 第3章 Jacobi 旋转:只用标准库做对称特征分解与奇异值分解 — Jacobi 旋转每次把一个非对角元转成零,非对角能量只会流向对角线,循环扫描在后期二次收敛:程序 p31 在 8 乘 8 的 Gram 矩阵上六轮就从 9.217e-01 降到 1.022e-16,正交性误差 2.7e-15。求奇异值不能走「先乘转置再开方」的捷径:程序 p32 中,十亿分之一的奇异值在那条路上被算成负数,相对误差 8.6e+00,而单边 Jacobi 的误差是 2.6e-08,已经贴着输入舍入的极限。程序 p33 验证了 Eckart 与 Young 的定理:截断误差恰好等于尾部奇异值平方和的平方根。有了这套只用标准库的分解工具,下一章就可以正式搭建动态模态分解:求特征值与模态,并看清噪声怎样把一个永不衰减的起伏伪装成正在消失的信号。 (https://5dlibrary.com/answers/zh/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c03/) - Ch.4 Dynamic Mode Decomposition: Eigenvalues, Amplitudes and the False Decay Caused by Noise — Chapter 4 assembles dynamic mode decomposition and shows that noise makes a swing that never decays look like one with modulus 0.7744 and a half-life of 2.7 minutes, which a forward-backward correction repairs. (https://5dlibrary.com/answers/en/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c04/) 第4章 动态模态分解:特征值、振幅与噪声造成的假衰减 — 第四章组装动态模态分解,指出噪声会把永不衰减的起伏估成模长 0.7744、半衰期 2.7 分钟,前向后向修正可以纠正。 (https://5dlibrary.com/answers/zh/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c04/) - Ch.5 Lead Time at the Turning Point: a Mode-Based Warning Every Five Minutes against Moving Averages — Chapter 5 compares three rules over two hundred random illustrative sessions: the mode warning has a median lead of 3.45 minutes, comes on average 13.13 minutes before the moving average and is premature in 0.055 of sessions, against 0.115 for the raw-curve rule. (https://5dlibrary.com/answers/en/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c05/) 第5章 拐点提前量:每五分钟一次的模态预警与移动平均的比较 — 第五章在两百场随机示意直播上比较三条规则:模态预警的提前量中位数 3.45 分钟,平均比移动平均早 13.13 分钟,过早误报 0.055,原始曲线规则则有 0.115。 (https://5dlibrary.com/answers/zh/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c05/) - Ch.6 Cash Flow: Linearising Daily, Weekly and Monthly Swings with Modes, and Choosing the Observable — Chapter 6 applies the method to a shop's daily cash flow, with a 28-day forecast error of 2301 yuan against 4665 for "the same day last week"; taking logarithms is a Koopman lift that cuts the rank-3 error from 0.1645 to 0.0257. (https://5dlibrary.com/answers/en/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c06/) 第6章 现金流:日、周、月三重波动的模态线性化与观测量的选择 — 第六章把方法用到店铺日现金流,28 天预报误差 2301 元,「上周同一天」为 4665 元;取对数是一次库普曼抬升,秩 3 误差从 0.1645 降到 0.0257。 (https://5dlibrary.com/answers/zh/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c06/) - Ch.7 East and West in the Sources: "Knowing the Incipient" in the Book of Changes and Fourier's Analysis of Heat — Chapter 7 examines "knowing the incipient" in the Book of Changes (ten minutes before the turn the second difference has a signal-to-noise ratio of only 0.031) and Fourier's analysis of heat (a fixed Fourier basis holds only 0.603 of the energy of a decaying swing). (https://5dlibrary.com/answers/en/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c07/) 第7章 东西方原典考辨:《周易》的「知几」与傅里叶的热分析 — 第七章考辨《周易·系辞下》的「知几」(拐点前十分钟二阶差分的信噪比只有 0.031)与傅里叶的热分析(固定的傅里叶基只装下衰减起伏 0.603 的能量)。 (https://5dlibrary.com/answers/zh/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c07/) - Ch.8 Closing the Empirical Loop: a Mode Fit Every Five Minutes, a Cooling Warning and a Pre-Registered Test — Chapter 8 gives the five-minute fitting procedure, the Online-Curve Turning-Point Early Warner and a pre-registered test plan: a lead difference of 3 minutes needs 51 sessions and the false-alarm test needs 68. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c08/) 第8章 实证闭环:每五分钟一次的模态拟合、降温预警与预注册检验 — 第八章给出每五分钟一次的拟合流程、在线曲线拐点预警器与预注册检验方案:检验 3 分钟的提前量差需要 51 场,检验过早误报需要 68 场。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/koopman-dynamic-mode-decomposition-cashflow-audience-turning-points/c08/) ## 5DL-B-239 · Kuramoto Coupled-Oscillator Synchrony: the Collective-Ordering Phase Transition in Hundred-Viewer Livestreams and the Critical Coupling of Rhythmic Entrainment 中文书名:藏本耦合振子同步、百人直播间集体下单相变与声频节律夹带临界耦合母典 DOI: pending Cite: Fan, Y. (2026). Kuramoto Coupled-Oscillator Synchrony: the Collective-Ordering Phase Transition in Hundred-Viewer Livestreams and the Critical Coupling of Rhythmic Entrainment (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 8 gives the A/B loop for a unified countdown, the Livestream Synchrony Order-Parameter Calculator, and a pre-registered protocol (26 sessions per arm under illustrative parameters, simulated power 0.907). Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Owners who sell through livestreams tend to believe that the more often the call is repeated, the more the audience moves as one and the tighter orders cluster. This book tests that belief with Kuramoto's coupled-oscillator model: treat viewers' decision rhythms as oscillators and the sight of genuine sales as coupling, and synchrony has a critical coupling set by the spread of rhythms; below it there is almost no synchrony, and above it the order parameter rises steeply by a square-root law, reaching 0.2182 in a Lorentzian crowd five per cent past threshold. Its audiences, rhythms, couplings, call strengths, stock and orders are illustrative parameters; all calculations are generated by its programs; the book is not individual business advice, deals only with genuine orders, holds that urgency may come only from real stock and real deadlines, and treats rhythm as synchrony mathematics only, with no claim of any physical or psychological effect. Chapter 1 explains the critical coupling with the self-consistency equation and a fourth-order Runge-Kutta simulation of a hundred viewers that matches theory to two decimal places above threshold. Chapter 2 gives the noise floor of finite crowds (an expectation of 0.0886 for a hundred people) and shows that below threshold repetition with fatigue never crosses, while squeezing the spread crosses at step five. Chapter 3 compares the thresholds of three audience shapes (1.0000, 1.1829 and 1.2732 at equal interquartile range), shows that coupling and spread enter only as a ratio, and simulates two rhythms breaking each other up in one room. Chapter 4 folds order timestamps into phases, computes the order parameter, the Rayleigh test and a small-sample bias correction, and scans candidate periods. Chapter 5 writes the call as an external force: Adler locking, near-critical amplification in the forced model (the same call amplified 4.37 times at threshold) and phase correction in beat following. Chapter 6 derives the load of the busiest minute from the order parameter and simulates real stock selling out in a synchronized peak. Chapter 7 examines the Book of Changes on "like sounds answer each other, like breaths seek each other" and Huygens's two pendulum clocks. Chapter 8 gives the A/B loop for a unified countdown, the Livestream Synchrony Order-Parameter Calculator, and a pre-registered protocol (26 sessions per arm under illustrative parameters, simulated power 0.907). Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:做直播带货的经营者习惯相信,口令喊得越多,观众越齐,下单越集中。本书用藏本耦合振子模型检验这一说法:把观众的决策节奏看成振子、把看见真实成交看成耦合,同步就有一个由节奏离散度决定的临界耦合,临界以下几乎不同步,越过以后序参量以平方根律陡然升起;洛伦兹人群越过临界百分之五,序参量即达 0.2182。书中的观众、节奏、耦合、口令强度、库存与订单均为示意参数;全部计算由书中程序生成;本书是非个人化经营建议,只讨论真实订单,紧迫感只能来自真实库存与真实截止,节律部分只讲同步数学,不作任何身心功效断言。 第一章用自洽方程与一百人的四阶龙格库塔模拟说明临界耦合,模拟与理论在临界以上吻合到小数点后两位。第二章给出有限人群的噪声底(一百人时期望 0.0886),并说明在临界以下,带疲劳的重复永远越不过临界,而压缩离散度在第五步就越过。第三章比较三种观众形状的临界值(同样四分位距下为 1.0000、1.1829 与 1.2732),说明耦合与离散度只以比值进入结果,并模拟一个房间里两种节奏互相打散。第四章把下单时间戳折叠成相位,计算序参量、瑞利检验与小样本偏差校正,并扫描候选周期。 第五章把口令写成外力:阿德勒锁定、受迫模型的近临界放大(同样的口令在临界处被放大 4.37 倍)与节拍跟随的相位校正。第六章由序参量算出最忙一分钟的订单压力,并模拟真实库存在同步高峰里售罄。第七章考辨《周易》「同声相应,同气相求」与惠更斯的两台摆钟。第八章给出统一倒数的 A/B 闭环、小工具「直播间同步序参量计算器」与预注册方案(示意参数下每组 26 场,模拟检验力 0.907)。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Critical Coupling: Why Synchrony Is Not Gradual — Once viewers are treated as oscillators and the sight of genuine sales as coupling, the togetherness of a livestream room is described by an order parameter $r$ obeying a self-consistency equation with a critical point. The critical coupling $K_c=2/(\pi g(0))$ depends only on the central density of the distribution of audience rhythms; for a Lorentzian crowd of half-width 0.5 the threshold is 1.0, five per cent past it the order parameter reaches 0.2182, and below it even a doubling of the coupling buys no togetherness. A simulation of a hundred people agrees with theory to two decimal places above threshold and leaves a layer of chance togetherness below it. So in a room of a hundred, how large must the order parameter be before the room is really in step, and how much does "one more call" buy below threshold? The next chapter calculates both. (https://5dlibrary.com/answers/en/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c01/) 第1章 临界耦合:同步为什么不是渐进的 — 把观众看成振子、把看见真实成交看成耦合以后,直播间的整齐程度由一个序参量 $r$ 描述,它服从一个有临界点的自洽方程。临界耦合 $K_c=2/(\pi g(0))$ 只由观众节奏分布的中心密度决定;洛伦兹人群的半宽为 0.5 时,临界值为 1.0,越过百分之五序参量就到 0.2182,而临界以下即使耦合翻倍也换不来任何整齐。一百人的模拟在临界以上与理论吻合到小数点后两位,在临界以下则留下一层偶然的整齐。那么,一百人的房间里,序参量多大才算真的齐了,「再喊一遍」在临界以下究竟能换来多少?下一章逐一计算。 (https://5dlibrary.com/answers/zh/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c01/) - Ch.2 Finite Rooms: the Noise Floor, the Blurred Threshold and "One More Call" — Chapter 2 gives the noise floor of finite crowds (an expectation of 0.0886 for a hundred people) and shows that below threshold repetition with fatigue never crosses, while squeezing the spread crosses at step five. (https://5dlibrary.com/answers/en/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c02/) 第2章 有限的房间:噪声底、模糊的阈值与「再喊一遍」 — 第二章给出有限人群的噪声底(一百人时期望 0.0886),并说明在临界以下,带疲劳的重复永远越不过临界,而压缩离散度在第五步就越过。 (https://5dlibrary.com/answers/zh/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c02/) - Ch.3 The Spread Lever: Audience Shapes, the Ratio Rule and Two Rhythms in One Room — Chapter 3 compares the thresholds of three audience shapes (1.0000, 1.1829 and 1.2732 at equal interquartile range), shows that coupling and spread enter only as a ratio, and simulates two rhythms breaking each other up in one room. (https://5dlibrary.com/answers/en/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c03/) 第3章 离散度杠杆:观众形状、比值律与一个房间里的两种节奏 — 第三章比较三种观众形状的临界值(同样四分位距下为 1.0000、1.1829 与 1.2732),说明耦合与离散度只以比值进入结果,并模拟一个房间里两种节奏互相打散。 (https://5dlibrary.com/answers/zh/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c03/) - Ch.4 From Order Timestamps to an Order Parameter: Folding, the Rayleigh Test and Small-Sample Bias — Fold order times into phases using the call period, and the length of the mean vector is the order parameter in the data while its direction is the average lag of orders behind the call. In the illustrative sessions, session A with a rhythm has an order parameter of 0.3249, a lag of 48.3 seconds and a Rayleigh statistic of 19.00; session B without one has 0.0434 and a statistic of 0.34; the noise floor for one hundred and eighty orders is 0.0661. Small samples bias the order parameter upward, so that a session of twenty orders with no rhythm still averages 0.195, and the bias-corrected $R^2_{adj}$ brings it back to zero. A period scan must watch for harmonics and pay for seventeen candidates with a family-wise bar of 5.83. Is the togetherness measured in the data pushed by the call or carried by viewers among themselves? The next chapter writes the call as an external force and separates the two. (https://5dlibrary.com/answers/en/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c04/) 第4章 从下单时间戳到序参量:折叠、瑞利检验与小样本偏差 — 把下单时刻按口令周期折叠成相位,平均向量的长度就是数据里的序参量,方向就是订单相对口令的平均滞后。示意直播里,有节奏的甲场序参量 0.3249、滞后 48.3 秒,瑞利统计量 19.00;无节奏的乙场 0.0434,统计量 0.34;一百八十笔订单的噪声底是 0.0661。小样本会让序参量系统性偏高,二十笔订单时毫无节奏的直播平均也有 0.195,偏差校正 $R^2_{adj}$ 能把它扣回零。扫描周期时要警惕谐波,并为十七个候选付出族错误率门槛 5.83 的代价。数据里测到的整齐,到底是口令推出来的,还是观众彼此带出来的?下一章把口令写成外力,把两者分开。 (https://5dlibrary.com/answers/zh/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c04/) - Ch.5 The Call Is an External Force: Adler Locking, Near-Critical Amplification and Beat Following — Chapter 5 writes the call as an external force: Adler locking, near-critical amplification in the forced model (the same call amplified 4.37 times at threshold) and phase correction in beat following. (https://5dlibrary.com/answers/en/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c05/) 第5章 口令是外力:阿德勒锁定、近临界放大与节拍跟随 — 第五章把口令写成外力:阿德勒锁定、受迫模型的近临界放大(同样的口令在临界处被放大 4.37 倍)与节拍跟随的相位校正。 (https://5dlibrary.com/answers/zh/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c05/) - Ch.6 Real Stock and Real Deadlines: Which Minute Synchrony Loads — The higher the order parameter, the more orders crowd into a few minutes: in the illustrative session averaging 3 orders a minute, the busiest minute sees 7.00 orders at an order parameter of 0.5 and 12.04 at 0.8, and the busiest tenth of the cycle holds 0.383 of all orders. The Millennium Bridge shows that the cost of crowd synchrony falls on whatever carries the load. In the stock simulation, stronger synchrony brings earlier sell-out and more vain clicks; with a synchronized share of 0.9, an average of 3.22 people click in vain in the thirty seconds after sell-out. Urgency may come only from real stock and real deadlines, stock must be displayed in real time, and a deadline once stated must be kept. What did the ancients observe about "like answering like", and what did Huygens's two clocks get right? The next chapter examines the two classical sources. (https://5dlibrary.com/answers/en/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c06/) 第6章 真实库存与真实截止:同步把压力压到哪一分钟 — 序参量越高,订单越挤在少数几分钟里:示意直播每分钟平均 3 笔,序参量 0.5 时最忙一分钟 7.00 笔,0.8 时 12.04 笔,最忙的十分之一时间装下 0.383 的订单。千禧桥的事件说明,人群同步的代价落在承载者身上。库存模拟里,同步越强,售罄越早、扑空越多,同步比例 0.9 时售罄后三十秒内平均 3.22 人扑空。紧迫感只能来自真实库存与真实截止,库存显示必须实时,截止时间一旦说出就要守住。古人对「同频相应」有过怎样的观察,惠更斯的两台摆钟又说对了什么?下一章考辨两部原典。 (https://5dlibrary.com/answers/zh/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c06/) - Ch.7 Classical Sources East and West: "Like Sounds Answer Each Other" and Huygens's Two Clocks — The Wenyan's "like sounds answer each other" can be read as a testable proposition: those drawn into synchrony are those whose natural rhythms are close to the mainstream. Program p72 shows that just past threshold the locked viewers lie on average only 0.192 from the centre, against 0.777 for those drifting; yet "like" need not mean identical, and the locking band widens as the chorus grows. Huygens rightly attributed the "sympathy" of his two clocks to coupling through a shared support, and p71 supplies the boundary he could not have written down: with coupling 0.05 the clocks no longer synchronize once the rate difference exceeds 0.10, and near the boundary the time to recover stretches to 225.5. Repetitions of his experiment show that the anti-phase synchrony he saw depended on his apparatus happening to fall in the right range of parameters. How do these observations reach a shop's daily livestreams, and what data could overturn the book's claims? The final chapter gives the loop and the pre-registered protocol. (https://5dlibrary.com/answers/en/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c07/) 第7章 东西方原典考辨:同声相应与惠更斯的两台摆钟 — 《文言》的「同声相应」可以读成一个可检验的命题:被同步吸纳的是固有节奏接近主流的人。p72 显示,耦合刚过临界时,锁住的人偏离中心的平均值只有 0.192,漂移者是 0.777;但「同声」不必完全相同,锁定带宽随合唱的壮大而变宽。惠更斯正确地把两台钟的「共感」归于共同支架的耦合,p71 补上了他无从写出的边界:耦合 0.05 时速率差超过 0.10 就不再同步,接近边界时恢复所需的时间拉长到 225.5。重做实验的研究表明,他看到的反相同步依赖于装置恰好落在合适的参数范围。这些观察怎样落到一家店铺每天的直播里,什么样的数据能推翻全书的主张?最后一章给出闭环与预注册方案。 (https://5dlibrary.com/answers/zh/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c07/) - Ch.8 Closing the Loop: an A/B Test of the Unified Countdown, the Calculator and a Pre-Registered Protocol — Chapter 8 gives the A/B loop for a unified countdown, the Livestream Synchrony Order-Parameter Calculator, and a pre-registered protocol (26 sessions per arm under illustrative parameters, simulated power 0.907). Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c08/) 第8章 实证闭环:统一倒数的 A/B、计算器与预注册方案 — 第八章给出统一倒数的 A/B 闭环、小工具「直播间同步序参量计算器」与预注册方案(示意参数下每组 26 场,模拟检验力 0.907)。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/kuramoto-synchrony-livestream-collective-ordering-critical-coupling/c08/) ## 5DL-B-240 · Fokker-Planck Drift-Diffusion on Long Sales Pages: Belief-Drift Rates and First-Passage Times to the Payment Barrier 中文书名:福克普朗克漂移扩散、长销售页信念漂移速率与付款吸收壁首达时间母典 DOI: pending Cite: Fan, Y. (2026). Fokker-Planck Drift-Diffusion on Long Sales Pages: Belief-Drift Rates and First-Passage Times to the Payment Barrier (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 4 pushes a page forward section by section: deleting the feature table of an illustrative page gains 0.0081, and the best and worst orders of sections differ by 0.1764. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Most owners who build long sales pages believe that the longer the page, the more it persuades. This book uses the drift-diffusion model to refute that unconditional belief: a visitor's belief wanders at random between two absorbing barriers, "leave" and "pay", the noise produced by cluttered information raises the probability of a mistaken exit, and when clutter accumulates with the number of sections the best page length has an upper limit. The project blueprint attached the proposition to Ratcliff's 1978 paper; after checking, the book corrects it: that paper studies recognition memory, and the page-length ceiling is a model consequence derived here under explicit assumptions and set mainly by clutter rather than by the size of the drift. The parameters are illustrative; all calculations are generated by the book's programs; the book is not individual business advice; the model describes how visitors weigh information and is not a technique of persuasion or manipulation; and every page change uses true information only, with no invented countdowns or scarcity. Chapter 1 gives closed forms for first-passage probabilities and mean decision times between two absorbing barriers. Chapter 2 carries the main argument: with a drift of 0.30 and noise of 1.0 the probability of a mistaken exit is 0.3543, and with a clutter coefficient of 0.06 the best page has 6 sections, while without clutter the ceiling disappears. Chapter 3 solves the Fokker-Planck equation with a conservative finite-volume scheme and checks the first-passage density against a series. Chapter 4 pushes a page forward section by section: deleting the feature table of an illustrative page gains 0.0081, and the best and worst orders of sections differ by 0.1764. Chapter 5 estimates drift from dwell and payment data by EZ-diffusion and shows that the absolute size of noise cannot be identified. Chapter 6 shows that compressing the barriers raises unsuited buyers per hundred decided visitors from 13.89 to 25.53, that noise cuts kept sales to 0.7767, and writes the opposing evidence on argument quantity as a cue into the model. Chapter 7 examines "branching", "tilting" and "doubling" in the Xunzi chapter "Undoing Fixation" and Einstein's 1905 paper on Brownian motion. Chapter 8 sets out a loop of recording, estimating, changing one thing and reviewing, the distilled "Long-Page Belief-Drift Estimator" tool, and a pre-registered deletion test with 17943 visitors per arm together with conditions for overturning the book. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:做长销售页的经营者大多相信「页面越长,说服越多」。本书用漂移扩散模型证伪这条不加条件的常识:访客的信念在「离开」与「付款」两个吸收壁之间随机游动,信息杂乱带来的噪音提高误撞错误出口的概率,杂乱随段数累积时最优页长有上限。立项蓝图把命题挂在拉特克利夫 1978 年的论文名下;本书核对后修正为:该文研究记忆再认,页长上限是本书在明确假设下推出的模型结论,且主要由杂乱程度而非漂移大小决定。书中参数为示意;全部计算由书中程序生成;本书是非个人化经营建议,模型用于描述访客怎样权衡信息,不是说服或操纵的技术,所有页面改动只用真实信息,不虚构倒计时或稀缺。 第一章给出双吸收壁首达概率与平均决定时间的闭式解。第二章是主轴:漂移 0.30、噪音 1.0 时误撞离开壁的概率为 0.3543;杂乱系数 0.06 时最优页长为 6 段,没有杂乱时上限消失。第三章用守恒的有限体积格式解福克普朗克方程,并用级数核对首达时间密度。第四章逐段推进页面:示意页面删去功能表多出 0.0081,最好与最差段落次序相差 0.1764。第五章用 EZ 扩散从停留与付款数据估计漂移,并说明噪音的绝对大小无法识别。第六章说明压缩壁距使每百位已决定访客中不适合的买家从 13.89 位增加到 25.53 位,噪音使留存成交降到 0.7767,并把「论据数量作为线索」的反方证据写进模型。 第七章考辨《荀子·解蔽》的「心枝」「倾」「贰」与爱因斯坦 1905 年的布朗运动论文。第八章给出记录、估计、改一处、复盘的闭环,蒸馏小工具「长页信念漂移估算器」,以及每组 17943 位访客的预注册删段检验方案与推翻条件。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Two Absorbing Barriers and First Passage: a Visitor's Path between "Pay" and "Leave" — The drift-diffusion model writes a visitor's hesitation as a random path between a "leave" barrier and a "pay" barrier. Program p11 gives closed forms for the probability of paying and the mean decision time and checks them by simulation with a continuity correction; p12 shows that payers and leavers have equal conditional times only when the start is central, so dwell data must be read together with the starting point. The next chapter isolates the noise and shows how it makes visitors who would have paid hit the leave barrier by mistake. (https://5dlibrary.com/answers/en/fokker-planck-drift-diffusion-sales-page-first-passage/c01/) 第1章 双吸收壁与首达:访客在「付款」与「离开」之间走完一条信念路 — 漂移扩散模型把访客的犹豫写成一条在「离开」与「付款」两壁之间的随机路。p11 给出付款概率与平均决定时间的闭式解,并用带连续性校正的模拟核对;p12 说明付款者与离开者的条件时间只有在起点居中时才相等,停留数据必须结合起点来读。下一章把噪音单独拿出来,看它怎样让本来会付款的人误撞离开壁。 (https://5dlibrary.com/answers/zh/fokker-planck-drift-diffusion-sales-page-first-passage/c01/) - Ch.2 Noise and Mistaken Exits: Why a Longer Page Does Not Mean More Persuasion — Chapter 2 carries the main argument: with a drift of 0.30 and noise of 1.0 the probability of a mistaken exit is 0.3543, and with a clutter coefficient of 0.06 the best page has 6 sections, while without clutter the ceiling disappears. (https://5dlibrary.com/answers/en/fokker-planck-drift-diffusion-sales-page-first-passage/c02/) 第2章 噪音与误撞离开壁:页面加长为什么不等于说服更多 — 第二章是主轴:漂移 0.30、噪音 1.0 时误撞离开壁的概率为 0.3543;杂乱系数 0.06 时最优页长为 6 段,没有杂乱时上限消失。 (https://5dlibrary.com/answers/zh/fokker-planck-drift-diffusion-sales-page-first-passage/c02/) - Ch.3 The Fokker-Planck Equation by Finite Differences: a Crowd of Visitors as a Cloud of Probability — The Fokker-Planck equation describes how a crowd of undecided visitors flows, and the flux through each barrier is the moment-by-moment rate of payment and of exit. Program p31 reproduces the closed forms with a mass-conserving finite-volume scheme whose error converges at second order; p32 computes the right-skewed distribution of payment times by a series and by the scheme, showing that most eventual payers decide within the first few screens; p33 shows that the step of an explicit scheme must satisfy the stability condition or its numbers mean nothing. The next chapter applies these tools to real page structure, where every section has its own drift and noise, and where which section is cut and which comes first both change the outcome. (https://5dlibrary.com/answers/en/fokker-planck-drift-diffusion-sales-page-first-passage/c03/) 第3章 福克普朗克方程的有限差分:把一群访客当作一团概率云 — 福克普朗克方程描述一群尚未决定的访客怎样流动,壁上的通量就是每一刻的付款率与离开率。p31 用质量守恒的有限体积格式重现了闭式解,误差呈二阶收敛;p32 用级数与格式两种方法算出付款时刻的右偏分布,最终付款者大多在前几屏就做了决定;p33 说明显式格式的步长必须满足稳定性条件,否则数字没有意义。下一章用这套工具处理真正的页面:每一段的漂移与噪音都不同,删哪一段、先放哪一段,都会改变结局。 (https://5dlibrary.com/answers/zh/fokker-planck-drift-diffusion-sales-page-first-passage/c03/) - Ch.4 A Page Pushed Forward Section by Section: Deletion, Order and the Best Length — Chapter 4 pushes a page forward section by section: deleting the feature table of an illustrative page gains 0.0081, and the best and worst orders of sections differ by 0.1764. (https://5dlibrary.com/answers/en/fokker-planck-drift-diffusion-sales-page-first-passage/c04/) 第4章 逐段推进的页面:删段、排序与最优页长 — 第四章逐段推进页面:示意页面删去功能表多出 0.0081,最好与最差段落次序相差 0.1764。 (https://5dlibrary.com/answers/zh/fokker-planck-drift-diffusion-sales-page-first-passage/c04/) - Ch.5 Estimating Drift and Noise from Dwell and Payment: EZ-Diffusion and Identifiability — EZ-diffusion turns three numbers, the share of payers and the mean and variance of dwell, into three parameters, drift, barrier separation and non-decision time, in closed form. Program p51 shows that drift is the hardest parameter to estimate: with one hundred decided visitors per version its error exceeds three tenths, and only around sixteen hundred does it fall below a tenth. Program p52 shows that when drift, barrier separation and noise are scaled together nothing observable changes, so noise can be estimated only relative to an anchor; this book's anchor is that visitors' caution does not change with a redesign. The next chapter uses this estimation perspective to discuss the trade-off between speed and accuracy, and the evidence that points against this book. (https://5dlibrary.com/answers/en/fokker-planck-drift-diffusion-sales-page-first-passage/c05/) 第5章 从停留与付款估漂移与噪音:EZ 扩散与可识别性 — EZ 扩散把付款比例、停留均值与停留方差三个数,闭式地换成漂移、壁距与非决策时间三个参数。p51 显示漂移是最难估的参数,每版只有一百位已决定访客时误差在三成以上,到一千六百位时才降到一成以内。p52 显示漂移、壁距与噪音同时缩放时一切观测量不变,噪音只能相对于一个锚来估;本书的锚是「访客谨慎程度不因改版而变」。下一章用这套估计的视角,讨论速度与准确的权衡,以及与本书方向相反的证据。 (https://5dlibrary.com/answers/zh/fokker-planck-drift-diffusion-sales-page-first-passage/c05/) - Ch.6 Speed, Accuracy and the Opposing Evidence: Compressed Barriers, Mixed Audiences and the Cue of Argument Quantity — Chapter 6 shows that compressing the barriers raises unsuited buyers per hundred decided visitors from 13.89 to 25.53, that noise cuts kept sales to 0.7767, and writes the opposing evidence on argument quantity as a cue into the model. (https://5dlibrary.com/answers/en/fokker-planck-drift-diffusion-sales-page-first-passage/c06/) 第6章 速度、准确与反方证据:压缩壁距、混合人群与「论据数量」线索 — 第六章说明压缩壁距使每百位已决定访客中不适合的买家从 13.89 位增加到 25.53 位,噪音使留存成交降到 0.7767,并把「论据数量作为线索」的反方证据写进模型。 (https://5dlibrary.com/answers/zh/fokker-planck-drift-diffusion-sales-page-first-passage/c06/) - Ch.7 Eastern and Western Sources: Xunzi on the Divided, Tilted and Doubled Mind, and Einstein's 1905 Brownian Motion — Chapter 7 examines "branching", "tilting" and "doubling" in the Xunzi chapter "Undoing Fixation" and Einstein's 1905 paper on Brownian motion. (https://5dlibrary.com/answers/en/fokker-planck-drift-diffusion-sales-page-first-passage/c07/) 第7章 东西方原典考辨:《荀子·解蔽》的心枝与倾贰,爱因斯坦 1905 年的布朗运动 — 第七章考辨《荀子·解蔽》的「心枝」「倾」「贰」与爱因斯坦 1905 年的布朗运动论文。 (https://5dlibrary.com/answers/zh/fokker-planck-drift-diffusion-sales-page-first-passage/c07/) - Ch.8 Closing the Loop: Scroll Depth, Controlled Deletions and a Pre-Registered Test — Chapter 8 sets out a loop of recording, estimating, changing one thing and reviewing, the distilled "Long-Page Belief-Drift Estimator" tool, and a pre-registered deletion test with 17943 visitors per arm together with conditions for overturning the book. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/fokker-planck-drift-diffusion-sales-page-first-passage/c08/) 第8章 实证闭环:滚动深度、删段对照与预注册检验 — 第八章给出记录、估计、改一处、复盘的闭环,蒸馏小工具「长页信念漂移估算器」,以及每组 17943 位访客的预注册删段检验方案与推翻条件。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/fokker-planck-drift-diffusion-sales-page-first-passage/c08/) ## 5DL-B-241 · Pontryagin's Maximum Principle at Product Launch: Optimal Ad-Pulse Burn Rates and the Hand-Over to Word of Mouth 中文书名:庞特里亚金极大值原理、新品冷启动广告脉冲烧钱速率与口碑接棒最优控制母典 DOI: pending Cite: Fan, Y. (2026). Pontryagin's Maximum Principle at Product Launch: Optimal Ad-Pulse Burn Rates and the Hand-Over to Word of Mouth (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 8 gives the first-week pulse schedule, the switch-day judgement, the Launch Ad-Pulse Switch-Day Calculator and a pre-registered protocol (simulated power 0.840 with 20 units per arm). Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: When a new product launches, owners tend to spread the promotion budget evenly across the days. This book tests that habit with an optimal control model of Bass diffusion with advertising. When advertising acts additively and linearly on the innovation term under a daily cap, Pontryagin's maximum principle and Bellman's dynamic programming give the same answer: spend at the cap from the start, stop once on the day the switching function turns negative, and hand the baton to word of mouth; under illustrative parameters the same budget spent at the cap over the first six days earns 13.4 % more than spread over sixty days. After checking the originals, the book corrects its founding proposition: the shape of the optimal path is set by the way advertising enters the diffusion equation, gliding smoothly down under diminishing returns and moving later when advertising multiplies the whole adoption rate. Its markets, margins, diffusion coefficients, advertising effects, caps and capacities are illustrative parameters; all calculations are generated by its programs; the book is not individual business advice, and urgency in it comes only from real stock and real deadlines. Chapter 1 states and quantifies the proposition with the Bass curve and five ways of spending one budget. Chapter 2 writes down the discrete-time Hamiltonian, co-state and switching function, checks by needle variation that the switching function is the marginal profit of each yuan, and shows why stopping by daily return keeps spending too long. Chapter 3 assumes no bang-bang structure and solves again by Bellman's backward induction, a grid convergence check and enumeration under a hard budget. Chapter 4 compares linear, concave, S-shaped and multiplicative responses. Chapter 5 splits buyers by source: under illustrative parameters the pulse brings 7064 extra buyers, of whom advertising bought 2264 directly and word of mouth relayed the rest; under parameter uncertainty the planned pulse loses a median of 0.0131 and the best even level chosen after the fact 0.0775. Chapter 6 deals with daily caps, burn rate and real capacity. Chapter 7 examines the Sunzi's "his momentum is steep, his release is short" and Bernoulli's brachistochrone of 1696. Chapter 8 gives the first-week pulse schedule, the switch-day judgement, the Launch Ad-Pulse Switch-Day Calculator and a pre-registered protocol (simulated power 0.840 with 20 units per arm). Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:新品上市时,经营者习惯把推广预算平均摊到每一天。本书用巴斯扩散加广告控制的最优控制模型检验这一做法:广告以加法、线性方式作用于创新项并有每日上限时,庞特里亚金极大值原理与贝尔曼动态规划给出同一个答案,开头打满上限、在开关函数变负的那天一次停投,把接力棒交给口碑;示意参数下同额预算前六天打满比六十天均匀多赚 13.4 %。核对原文后,本书修正了立项命题:最优路径的形状由广告进入扩散方程的方式决定,反应边际递减时是平滑滑降,广告乘到整个采用率上时最优支出反而后移。书中的市场、毛利、扩散系数、广告效果、上限与产能均为示意参数;全部计算由书中程序生成;本书是非个人化经营建议,紧迫感只来自真实库存与真实截止。 第一章用巴斯曲线与同一笔预算的五种花法陈述并量化命题。第二章写出离散时间的哈密顿量、协态与开关函数,用针状变分检验开关函数就是每一元的边际利润,并说明按当天回报率停投为何会投得太久。第三章不假设开关结构,用贝尔曼逆推、网格收敛检查与硬预算下的穷举重新求解。第四章比较线性、凹形、S 形与乘法四种反应。 第五章拆分买家来源:示意参数下脉冲多带来 7064 位买家,广告直接买来 2264 位,其余由口碑接力;参数不确定时,计划好的脉冲损失中位数 0.0131,事后选出的最好均匀档为 0.0775。第六章讨论日上限、烧钱速率与真实产能。第七章考辨《孙子·势篇》的「其势险,其节短」与伯努利 1696 年的最速降线。第八章给出首周脉冲排程、切换日判定、小工具「冷启动投放切换点计算器」与预注册方案(每组 20 个单位时模拟检验力 0.840)。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Why Even Spending Loses: Equal Budgets on a Bass Curve — In the Bass equation buyers come from two directions, spontaneous trial and word of mouth, and word of mouth is the engine. When advertising acts linearly on the innovation term, the earlier the same money is spent, the larger its multiplier; six days at the cap earn 13.4 % more than sixty even days, and the same money in the last six days is almost wasted. But where does "six days" come from? Why six and not five or eight, and why should not a single yuan be spent after day six? Answering that needs a tool that prices every yuan day by day, and that tool is the Hamiltonian and switching function of the next chapter. (https://5dlibrary.com/answers/en/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c01/) 第1章 均匀撒钱为什么输:巴斯扩散上的同额对照 — 巴斯方程里,买家从自发尝鲜与口碑两个方向来,而口碑是引擎。广告线性作用于创新项时,同一笔钱越早花,乘数越大;前六天打满比六十天均匀多赚 13.4 %,放在最后六天几乎白花。可是「前六天」这个数从哪里来?为什么是六天而不是五天或八天,第六天之后为什么一分都不该再花?要回答它,需要一个能逐日给每一元钱定价的工具,这就是下一章的哈密顿量与开关函数。 (https://5dlibrary.com/answers/zh/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c01/) - Ch.2 Pricing Every Yuan: the Hamiltonian, the Co-State and the Switching Function — The two columns are almost identical: on the first day the needle gives +0.4627 and the switching function +0.4579; on day fifteen they are -0.4638 and -0.4633; on day thirty, -0.7314 and -0.7313. The small gaps arise because a hundred yuan is not infinitesimal, and nonlinearity makes a finite perturbation drift slightly from the derivative; the later the day, the smaller the gap. The switching function is therefore not an abstract multiplier. It is how much the whole window's profit changes, net, for one more yuan on that day, the true price of each yuan. (https://5dlibrary.com/answers/en/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c02/) 第2章 给每一元钱定价:哈密顿量、协态与开关函数 — 两列几乎相同:第一天针状检验得 +0.4627,开关函数 +0.4579;第十五天分别是 -0.4638 与 -0.4633;第三十天是 -0.7314 与 -0.7313。差异来自一百元并不是无穷小,非线性让有限扰动与导数略有偏离;越往后差异越小。所以开关函数不是抽象的乘子,它就是「这一天再多花一元,整个窗口的利润净变多少」,是每一元钱的真实价格。 (https://5dlibrary.com/answers/zh/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c02/) - Ch.3 Reasoning Back from the Last Day: Bellman's Induction, a Grid Check and a Hard Budget — Bellman's backward induction assumes no shape and, with intermediate amounts allowed, independently finds the same six-day pulse; the optimal value of 531904 and the path profit of 531909 differ only by interpolation error, and the conclusion is unchanged from a hundred and one grid points to two thousand and one, and from three levels to twenty-one. The decision table can also be read as a threshold rule that rises over time and copes with sales that depart from plan. Under a hard budget, enumerating every placement shows that the best is always a run at the cap from the first day. Everything so far rests on one key assumption: the effect of advertising is linear in spending. If the first thousand yuan work hardest and returns diminish after that, or if spending too little does nothing at all, is the optimal path still a clean cut? The next chapter answers. (https://5dlibrary.com/answers/en/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c03/) 第3章 从最后一天往回推:贝尔曼逆推、网格检查与硬预算 — 贝尔曼逆推不假设任何形状,在允许中间金额的条件下独立得到同一条六天脉冲,最优价值 531904 与路径利润 531909 只差插值误差;从一百零一个到两千零一个网格点、从三档到二十一档,结论不变。决策表还可以读成一条随时间上升的阈值规则,用来应对销售偏离计划。硬预算下穷举所有摆法,最优永远是从第一天起连续打满。到此为止的结论都建立在一个关键假设上:广告效果对花费是线性的。如果头一千元最有效、此后边际递减,或者花得太少根本不起作用,最优路径还是一刀切吗?下一章回答。 (https://5dlibrary.com/answers/zh/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c03/) - Ch.4 The Response Curve Sets the Shape: a Concave Glide, S-Shaped Pulses and a Multiplicative Counterexample — The response curve sets the shape of the optimal path. Linear response gives a switch: spend at the cap at the start and stop once. Concave response gives a slope: spend most at the start, a little less each day, and stop after twenty-one days, earning 540233, more than the pulse at the cap. S-shaped response makes even spending lose to alternate days. The multiplicative form is the most contrary: its optimal path starts low and rises, the opposite of the additive model. Concentrating early holds only when advertising substitutes for word of mouth. On the additive side, which parameters set the switch day? How many buyers does advertising buy, and how many does word of mouth bring? And when the parameters are poorly known, is the pulse still safe? The next chapter answers. (https://5dlibrary.com/answers/en/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c04/) 第4章 反应曲线决定形状:凹形滑降、S 形脉冲与乘法反例 — 反应曲线决定最优路径的形状。线性反应给出开关:开头打满,一次停掉。凹形反应给出滑坡:开头花得最多,每天减一点,二十一天后停投,利润 540233 高于打满上限的脉冲。S 形反应让均匀投放输给隔天投放。乘法形式最为反常:最优路径先低后高,与加法模型方向相反。前段集中的命题只在广告替代口碑时成立。在加法这一侧,切换日由哪些参数决定?广告买来的人与口碑带来的人各占多少?参数估不准时,脉冲还安全吗?下一章回答。 (https://5dlibrary.com/answers/zh/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c04/) - Ch.5 Word of Mouth Takes the Baton: Splitting Buyers by Source, the Sensitivity of the Switch Day and Robustness under Uncertainty — Chapter 5 splits buyers by source: under illustrative parameters the pulse brings 7064 extra buyers, of whom advertising bought 2264 directly and word of mouth relayed the rest; under parameter uncertainty the planned pulse loses a median of 0.0131 and the best even level chosen after the fact 0.0775. (https://5dlibrary.com/answers/en/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c05/) 第5章 口碑接棒:买家来源拆分、切换日的灵敏度与参数不确定下的稳健性 — 第五章拆分买家来源:示意参数下脉冲多带来 7064 位买家,广告直接买来 2264 位,其余由口碑接力;参数不确定时,计划好的脉冲损失中位数 0.0131,事后选出的最好均匀档为 0.0775。 (https://5dlibrary.com/answers/zh/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c05/) - Ch.6 Burn Rate and Real Capacity: Daily Caps, the Impulse Limit and Orders That Cannot Be Filled — The daily cap sets the burn rate, not the total. In the linear model, raising the cap from 5000 yuan to 20000 leaves the optimal total at 30000 yuan, while the profit gains shrink step by step toward the theoretical limit of a single impulse; real diminishing returns keep that limit out of reach. Capacity is another matter. When shipping capacity cannot absorb the demand advertising piles up, the optimal pulse becomes lower and longer, and spending at the cap loses more orders and earns less. Urgency may come only from real stock and real deadlines. With the models, the evidence and the constraints in place, two questions remain: what did the ancients say about momentum and about the fastest path, and how much of it did they get right? The next chapter examines the Sunzi and Bernoulli. (https://5dlibrary.com/answers/en/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c06/) 第6章 烧钱速率与真实产能:日上限、脉冲极限与接不住的订单 — 日上限决定的是烧钱速率,而不是总额:在线性模型里,上限从 5000 元提高到 20000 元,最优总花费始终是 30000 元,利润的增幅逐级变小,逼近一次性脉冲的理论极限;现实中的边际递减会让这个极限够不着。产能是另一回事:当发货能力接不住广告堆出来的需求时,最优脉冲变低、变长,打满上限反而多流失订单、少赚钱。紧迫感只能来自真实库存与真实截止。说完了模型、证据与约束,还剩下两个问题:古人对「势」与「最快的路」说过什么,他们说对了多少?下一章考辨孙子与伯努利。 (https://5dlibrary.com/answers/zh/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c06/) - Ch.7 Classical Sources East and West: the Sunzi on Momentum and Bernoulli's Brachistochrone — Chapter 7 examines the Sunzi's "his momentum is steep, his release is short" and Bernoulli's brachistochrone of 1696. (https://5dlibrary.com/answers/en/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c07/) 第7章 东西方原典考辨:《孙子·势篇》与伯努利的最速降线 — 第七章考辨《孙子·势篇》的「其势险,其节短」与伯努利 1696 年的最速降线。 (https://5dlibrary.com/answers/zh/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c07/) - Ch.8 The Empirical Loop: Scheduling the First-Week Pulse, Judging the Switch Day and a Pre-Registered Test — Chapter 8 gives the first-week pulse schedule, the switch-day judgement, the Launch Ad-Pulse Switch-Day Calculator and a pre-registered protocol (simulated power 0.840 with 20 units per arm). Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c08/) 第8章 实证闭环:首周脉冲排程、切换日判定与预注册检验 — 第八章给出首周脉冲排程、切换日判定、小工具「冷启动投放切换点计算器」与预注册方案(每组 20 个单位时模拟检验力 0.840)。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/pontryagin-maximum-principle-launch-ad-pulse-word-of-mouth/c08/) ## 5DL-B-242 · Compressed Sensing and L1 Minimisation: Sparse Mystery-Shopper Probes and the Exact Recovery of Competitors' Hidden-Discount Matrices 中文书名:压缩感知一范数极小化、神秘访客稀疏探针与竞品暗折矩阵精确重构母典 DOI: pending Cite: Fan, Y. (2026). Compressed Sensing and L1 Minimisation: Sparse Mystery-Shopper Probes and the Exact Recovery of Competitors' Hidden-Discount Matrices (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its items, amounts, costs and rivals are illustrative and all calculations are generated by its programs; it is a general business reference, not a legal opinion. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: The trade assumes that learning which of a rival's items carry a hidden discount at checkout, one not printed on the label, and how large each is, requires quoting every item in a full census, since sampling yields at best an average. This book uses compressed sensing to refute the premise of that view. As long as hidden discounts are sparse, putting items into random carts and reading only the checkout totals needs a number of quotes roughly proportional to the number of discounts times the logarithm of the ratio of items to discounts, far fewer than the items themselves, and L1 minimisation under nonnegativity returns, without noise, the exact table. Its items, amounts, costs and rivals are illustrative and all calculations are generated by its programs; it is a general business reference, not a legal opinion. Probes collect only the prices and conditions a rival shows the public: no impersonation, no inducing staff to reveal internal information, no scraping of protected data, no exchange of results with competitors, and compliance boundaries follow local rules checked by a person before publication. Chapter 1 checks and corrects the sources of the founding claim and uses an exact simplex method in fractions to recover 3 hidden discounts among 40 items from 12 basket quotes, while least squares smears them over the whole table with a worst cell error of 26.73 yuan; it measures the scheduling constant, with the quote share falling to 0.0825 of the item count. Chapter 2 writes L1 as a two-phase simplex, treats degeneracy and perturbation, issues certificates of uniqueness, and gives a design check for baskets: zero and identical columns pull the recovery share of three-cell supports from 0.969 to 0.704. Chapter 3 compares orthogonal matching pursuit: on one hundred items L1 reaches ninety percent success with 22 quotes, nonnegative OMP with 26 and plain OMP with 34, while worst-case coherence guarantees are nearly useless. Chapter 4 draws the phase diagram of zero-one baskets and prices the knowledge that there are discounts only. Chapter 5 handles rounding, anniversary coupons and threshold rules: with rounding to ten fen the equality model is mostly infeasible, while the robust model still finds every support with a worst error of 0.03 yuan, and only 0.0403 of wrong tables pass one hold-out quote. Chapter 6 sets several public checkout conditions side by side as a hidden-discount matrix, compares a greedy method for joint sparsity with column-wise L1, and shows that baskets that are too small cost more than a census. Chapter 7 examines Sunzi's "foreknowledge must be obtained from people" and Shannon's count of samples: in an illustrative account analogy loses 59400 yuan, a census costs 57600 yuan and sparse probes 12960 yuan, and each tenfold increase in items adds only 17.3 quotes. Chapter 8 gives a six-step monthly probe procedure, a seven-item compliance checklist, the Sparse Probe Reconstructor and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:行内默认的看法是:想知道竞品每件商品在结算时有没有价签上不写的暗折、各减多少,就只能逐件全盘抽样,抽样最多给出平均让利。本书用压缩感知证伪这一看法的前提。只要暗折是稀疏的,把商品随机组合成购物车、只读结算总额,所需报价次数大致与暗折数乘以商品数对暗折数之比的对数成正比,远少于商品数;在非负约束下极小化一范数,无噪声时得到的是精确的表。书中商品、金额、成本与竞品均为示意;全部计算由书中程序生成;本书是非个人化经营参考,不是法律意见。探针只采集竞品向公众公开展示的价格与条件,不冒充身份,不诱导员工泄露内部信息,不抓取受保护数据,不与竞争者交换结果;合规边界以上架前人工复核的当地法规为准。 第一章核对并修正立项命题的出处,用分数精确的单纯形法以 12 次篮子报价还原 40 件商品里的 3 笔暗折,而最小二乘把它们摊满全表,最坏一格偏离 26.73 元;并量出排程常数,报价占商品数的比例可降到 0.0825。第二章把一范数写成两阶段单纯形,讲退化与扰动,开出唯一性证书,并给出篮子设计体检:零列与同列把三格支撑的还原比例从 0.969 拉到 0.704。第三章比较正交匹配追踪:一百件商品上一范数 22 次、非负 OMP 26 次、普通 OMP 34 次报价达到九成成功,互相干度的最坏情形保证几乎无用。第四章画出 0 与 1 篮子的相变图,并量出「只有暗折」这一先验的价值。 第五章处理抹零、店庆券与门槛规则:取整到角时等式模型多半不可行,稳健模型仍找对全部支撑,最坏误差 0.03 元;错表通过一次留出检验的比例只有 0.0403。第六章把多个公开结算条件并排成暗折矩阵,比较联合稀疏的贪心算法与逐列一范数,并说明篮子太小反而比全盘抽样更费。第七章考辨《孙子·用间》「先知必取于人」与香农的采样计数:示意账目中类推损失 59400 元、全盘抽样 57600 元、稀疏探针 12960 元;商品每多十倍,稀疏计数只多 17.3 次。第八章给出每月六步探针流程、七条合规清单、稀疏探针重构器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 No Full Census Needed: Basket Quotes, Sparse Hidden Discounts and Exact L1 Recovery — Chapter 1 checks and corrects the sources of the founding claim and uses an exact simplex method in fractions to recover 3 hidden discounts among 40 items from 12 basket quotes, while least squares smears them over the whole table with a worst cell error of 26.73 yuan; it measures the scheduling constant, with the quote share falling to 0.0825 of the item count. (https://5dlibrary.com/answers/en/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c01/) 第1章 不必全盘抽样:篮子报价、稀疏暗折与一范数的精确重构 — 第一章核对并修正立项命题的出处,用分数精确的单纯形法以 12 次篮子报价还原 40 件商品里的 3 笔暗折,而最小二乘把它们摊满全表,最坏一格偏离 26.73 元;并量出排程常数,报价占商品数的比例可降到 0.0825。 (https://5dlibrary.com/answers/zh/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c01/) - Ch.2 L1 as a Linear Program: the Two-Phase Simplex, Dual Certificates and a Design Check for Probes — Chapter 2 writes L1 as a two-phase simplex, treats degeneracy and perturbation, issues certificates of uniqueness, and gives a design check for baskets: zero and identical columns pull the recovery share of three-cell supports from 0.969 to 0.704. (https://5dlibrary.com/answers/en/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c02/) 第2章 一范数写成线性规划:两阶段单纯形、对偶证书与探针设计体检 — 第二章把一范数写成两阶段单纯形,讲退化与扰动,开出唯一性证书,并给出篮子设计体检:零列与同列把三格支撑的还原比例从 0.969 拉到 0.704。 (https://5dlibrary.com/answers/zh/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c02/) - Ch.3 The Price of Greed: Orthogonal Matching Pursuit, Coherence and Worst-Case Guarantees — Chapter 3 compares orthogonal matching pursuit: on one hundred items L1 reaches ninety percent success with 22 quotes, nonnegative OMP with 26 and plain OMP with 34, while worst-case coherence guarantees are nearly useless. (https://5dlibrary.com/answers/en/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c03/) 第3章 贪心的代价:正交匹配追踪、相干度与最坏情形保证 — 第三章比较正交匹配追踪:一百件商品上一范数 22 次、非负 OMP 26 次、普通 OMP 34 次报价达到九成成功,互相干度的最坏情形保证几乎无用。 (https://5dlibrary.com/answers/zh/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c03/) - Ch.4 The Phase Diagram: Probe Share, Sparsity and the Value of Knowing the Sign — Chapter 4 draws the phase diagram of zero-one baskets and prices the knowledge that there are discounts only. (https://5dlibrary.com/answers/en/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c04/) 第4章 相变图:探针占比、稀疏度与已知符号的价值 — 第四章画出 0 与 1 篮子的相变图,并量出「只有暗折」这一先验的价值。 (https://5dlibrary.com/answers/zh/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c04/) - Ch.5 Rounding, Anniversary Coupons and Hold-Out Quotes: from Exact to Stable Recovery — Chapter 5 handles rounding, anniversary coupons and threshold rules: with rounding to ten fen the equality model is mostly infeasible, while the robust model still finds every support with a worst error of 0.03 yuan, and only 0.0403 of wrong tables pass one hold-out quote. (https://5dlibrary.com/answers/en/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c05/) 第5章 抹零、店庆券与留出报价:从精确重构到稳健重构 — 第五章处理抹零、店庆券与门槛规则:取整到角时等式模型多半不可行,稳健模型仍找对全部支撑,最坏误差 0.03 元;错表通过一次留出检验的比例只有 0.0403。 (https://5dlibrary.com/answers/zh/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c05/) - Ch.6 The Hidden-Discount Matrix: Joint Sparsity across Public Conditions, and Basket Size — Discounts under several public conditions form a matrix side by side, and the discounted items are largely the same across conditions; this is joint sparsity. Program p61 showed that joint information lets the greedy SOMP clearly beat column-wise OMP, but that on nonnegative zero-one baskets column-wise L1 is still the best: from 24 quotes on the whole matrix is always right, 96 quotes over 4 conditions against 320 for a full census. Program p62 showed that basket size carries costs at both ends: with 2 items per basket it takes 78 quotes, more than a full census; from 15 items upward 16 quotes suffice, but the items handled grow as baskets grow. With this the mathematical part is complete. The next chapter returns to two classical sources: Sunzi, who held that foreknowledge "must be obtained from people", and Shannon, who held that a signal is completely determined by its samples. What did they get right, and where do today's data rewrite them? (https://5dlibrary.com/answers/en/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c06/) 第6章 暗折矩阵:多个公开条件的联合稀疏与篮子大小 — 多个公开条件下的暗折并排成一张矩阵,带暗折的商品在各条件之间大体相同,这叫联合稀疏。程序 p61 显示,联合信息让贪心算法 SOMP 明显胜过逐列 OMP,但在非负的 0 与 1 篮子上,逐列一范数仍然最好:24 次报价起整张矩阵全对,4 个条件合计 96 次,而全盘抽样要 320 次。程序 p62 显示篮子大小有两头的代价:每篮 2 件时要 78 次报价,比全盘抽样还多;每篮 15 件以上时 16 次即可,但经手件数随篮子变大而上升。到这里,数学部分已经完整。下一章回到两部原典:《孙子》说先知「必取于人」,香农说信号由它的样本完全决定;它们说对了什么,又在哪里被今天的数据改写。 (https://5dlibrary.com/answers/zh/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c06/) - Ch.7 Classical Sources East and West: Foreknowledge in Sunzi's "Use of Spies" and Shannon's Count of Samples — Chapter 7 examines Sunzi's "foreknowledge must be obtained from people" and Shannon's count of samples: in an illustrative account analogy loses 59400 yuan, a census costs 57600 yuan and sparse probes 12960 yuan, and each tenfold increase in items adds only 17.3 quotes. (https://5dlibrary.com/answers/en/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c07/) 第7章 东西方原典考辨:《孙子·用间》的先知与香农的采样计数 — 第七章考辨《孙子·用间》「先知必取于人」与香农的采样计数:示意账目中类推损失 59400 元、全盘抽样 57600 元、稀疏探针 12960 元;商品每多十倍,稀疏计数只多 17.3 次。 (https://5dlibrary.com/answers/zh/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c07/) - Ch.8 Closing the Loop: a Monthly Mystery-Shopper Probe Schedule, a Compliance Checklist and a Pre-Registered Test — Chapter 8 gives a six-step monthly probe procedure, a seven-item compliance checklist, the Sparse Probe Reconstructor and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c08/) 第8章 实证闭环:每月神秘访客探针排程、合规清单与预注册检验 — 第八章给出每月六步探针流程、七条合规清单、稀疏探针重构器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/compressed-sensing-l1-mystery-shopper-hidden-discount-recovery/c08/) ## 5DL-B-243 · Graph-Laplacian Spectral Gaps and Cheeger Constants: Single-Point Bottleneck Cuts in Multi-Agent Workflows and Cross-Border Supply Chains 中文书名:图拉普拉斯谱间隙、奇格常数与多智能体工作流及跨国供应链单点瓶颈割集母典 DOI: pending Cite: Fan, Y. (2026). Graph-Laplacian Spectral Gaps and Cheeger Constants: Single-Point Bottleneck Cuts in Multi-Agent Workflows and Cross-Border Supply Chains (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: The parameters are illustrative; all calculations are generated by the book's programs; the book is not individual business advice, and any change to a production system or a real supply chain must be reviewed and released by hand by the person in charge. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: Most owners who build multi-agent workflows or run cross-border supply chains believe that more links mean more robustness. This book uses the spectrum of the graph Laplacian to refute that unconditional belief: whether a network breaks is decided by its narrowest neck, not by how many links an average node has. The project blueprint attached the claim to the papers of Fiedler in 1973 and Cheeger in 1970; after checking, the book corrects it: Fiedler proved that the algebraic connectivity does not exceed the vertex and edge connectivities, "a small algebraic connectivity guarantees a cheap cut" comes from later Cheeger-type inequalities on graphs and guarantees a relative cut ratio, Cheeger studied manifolds, the placement of new links was studied systematically by Ghosh and Boyd, and the counting form of the matrix-tree theorem was not written by Kirchhoff himself. The parameters are illustrative; all calculations are generated by the book's programs; the book is not individual business advice, and any change to a production system or a real supply chain must be reviewed and released by hand by the person in charge. Chapter 1 computes the algebraic connectivity by Jacobi rotation and shows that a one-bridge barbell with a higher mean degree has an algebraic connectivity of only 0.25834, against 2.00000 for a circulant, a factor of 7.74. Chapter 2 carries the main argument: the Cheeger constant is clamped from both sides by the spectrum, and in a dose-response scan, links added inside clusters raise the mean degree while the algebraic connectivity falls from 0.4384 to 0.2042. Chapter 3 tries every new link and finds that cross-team links gain on average 87.5 times as much as inside-team links. Chapter 4 treats consensus speed among agents, the tempo ceiling of an orchestrating hub and single points of hand-off. Chapter 5 compares three supply networks with the same number of lanes by effective resistance and maximum flow; the cross-woven design pushes the probability of a quarter delivering half or less down to 0.0203. Chapter 6 sets out the opposing evidence: the preferential-attachment graph with the highest algebraic connectivity keeps only 0.300 of its nodes connected after a fifth of its hubs are removed on purpose. Chapter 7 examines the Changshan snake of the Sunzi's "Nine Grounds" and Kirchhoff's circuit equations of 1847. Chapter 8 gives a quarterly loop of drawing, checking, adding and reviewing, in which three rounds of advice raise the algebraic connectivity of an illustrative workflow from 0.1824 to 1.0000; the distilled "Bottleneck Cut-Set Checker"; and a pre-registered test with 74 team-quarters per arm together with conditions for overturning the book. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:搭建多智能体工作流、管理跨国供应链的经营者大多相信「连接越多越稳」。本书用图拉普拉斯矩阵的谱证伪这条不加条件的常识:网络会不会断,由最窄的颈口决定,而不由平均每个节点连了几条边决定。立项蓝图把命题挂在菲德勒 1973 年与奇格 1970 年的论文名下;本书核对后修正为:菲德勒证明的是代数连通度不超过点连通度与边连通度,「代数连通度小则存在便宜的割」来自后来图上的奇格型不等式,且保证的是相对割比;奇格研究的是流形;补边位置的问题由戈什与博伊德系统研究;矩阵树定理的计数表述并非基尔霍夫本人写出。书中参数为示意;全部计算由书中程序生成;本书是非个人化经营建议,任何生产系统或真实供应链的改动须由主事者人工复核放行。 第一章用雅可比旋转求代数连通度,显示平均度数更高的一桥哑铃代数连通度只有 0.25834,而循环图为 2.00000,相差 7.74 倍。第二章是主轴:奇格常数被谱从两侧夹住;剂量与反应扫描中,团内加边使平均度数升高,代数连通度却从 0.4384 降到 0.2042。第三章逐条试算新边,跨队边的平均增量是队内边的 87.5 倍。第四章讨论多智能体共识速度、调度中枢的节拍上限与交接单点。第五章用等效电阻与最大流比较三种线路数相同的供应网,交织设计把单季送达不足一半的概率压到 0.0203。第六章陈列反方证据:代数连通度最高的优先连接图在定向删去两成枢纽后只剩 0.300 的节点连通。 第七章考辨《孙子·九地》的常山之蛇与基尔霍夫 1847 年的电路方程。第八章给出画图、体检、补边、复查的季度闭环,示意工作流三轮补边后代数连通度从 0.1824 升到 1.0000;蒸馏小工具「瓶颈割集体检器」;以及每组 74 个团队季度的预注册检验方案与推翻条件。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Algebraic Connectivity: Why a High Degree Is Not Robustness — Chapter 1 computes the algebraic connectivity by Jacobi rotation and shows that a one-bridge barbell with a higher mean degree has an algebraic connectivity of only 0.25834, against 2.00000 for a circulant, a factor of 7.74. (https://5dlibrary.com/answers/en/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c01/) 第1章 代数连通度:度数高为什么不等于稳固 — 第一章用雅可比旋转求代数连通度,显示平均度数更高的一桥哑铃代数连通度只有 0.25834,而循环图为 2.00000,相差 7.74 倍。 (https://5dlibrary.com/answers/zh/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c01/) - Ch.2 The Cheeger Constant: How Narrow the Neck Is, Clamped from Both Sides by the Spectrum — Chapter 2 carries the main argument: the Cheeger constant is clamped from both sides by the spectrum, and in a dose-response scan, links added inside clusters raise the mean degree while the algebraic connectivity falls from 0.4384 to 0.2042. (https://5dlibrary.com/answers/en/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c02/) 第2章 奇格常数:颈口有多窄,谱给出上下两道夹板 — 第二章是主轴:奇格常数被谱从两侧夹住;剂量与反应扫描中,团内加边使平均度数升高,代数连通度却从 0.4384 降到 0.2042。 (https://5dlibrary.com/answers/zh/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c02/) - Ch.3 Where to Add One Link: Greedy Addition, the Fiedler Score and the Illusion of Quantity — Chapter 3 tries every new link and finds that cross-team links gain on average 87.5 times as much as inside-team links. (https://5dlibrary.com/answers/en/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c03/) 第3章 补一条边的位置:贪心加边、菲德勒评分与数量的幻觉 — 第三章逐条试算新边,跨队边的平均增量是队内边的 87.5 倍。 (https://5dlibrary.com/answers/zh/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c03/) - Ch.4 Multi-Agent Workflows: Consensus Speed, the Orchestrating Hub and Single Points of Hand-off — Once a workflow is drawn as a graph of hand-off channels, the speed of consensus is set jointly by the algebraic connectivity and the maximum degree. Program p41 shows that the chain pipeline is slowest and that the orchestrating hub, despite a respectable algebraic connectivity, is slower than a ring because its tempo is capped by the hub's load; p42 finds two bridges, three cut vertices and the seam between thinking and making in a twelve-agent workflow, and shows that the link suggestion with the largest gain need not remove cut vertices. The next chapter carries the same tools into cross-border supply chains, where links carry not only messages but capacity. (https://5dlibrary.com/answers/en/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c04/) 第4章 多智能体工作流:共识速度、调度中枢与交接单点 — 把工作流写成交接通道构成的图之后,共识的快慢由代数连通度与最大度数共同决定。p41 显示链式流水线最慢,调度中枢虽然代数连通度不低,却因为节拍受中枢负荷所限而慢于环;p42 在一张十二个智能体的工作流上找出两座桥、三个割点与「想清楚」和「做出来」之间的接缝,并说明增量第一的补边建议未必能消除割点。下一章把同一套工具搬到跨国供应链,那里的边不只是通道,还有运力。 (https://5dlibrary.com/answers/zh/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c04/) - Ch.5 Cross-Border Supply Chains: Effective Resistance, Maximum Flow and the Weight of One Port — Chapter 5 compares three supply networks with the same number of lanes by effective resistance and maximum flow; the cross-woven design pushes the probability of a quarter delivering half or less down to 0.0203. (https://5dlibrary.com/answers/en/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c05/) 第5章 跨国供应链:等效电阻、最大流与一个港口的分量 — 第五章用等效电阻与最大流比较三种线路数相同的供应网,交织设计把单季送达不足一半的概率压到 0.0203。 (https://5dlibrary.com/answers/zh/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c05/) - Ch.6 Opposing Evidence and Boundaries: Where the Algebraic Connectivity Misleads — Chapter 6 sets out the opposing evidence: the preferential-attachment graph with the highest algebraic connectivity keeps only 0.300 of its nodes connected after a fifth of its hubs are removed on purpose. (https://5dlibrary.com/answers/en/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c06/) 第6章 反方证据与边界:代数连通度在哪里会误导 — 第六章陈列反方证据:代数连通度最高的优先连接图在定向删去两成枢纽后只剩 0.300 的节点连通。 (https://5dlibrary.com/answers/zh/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c06/) - Ch.7 Eastern and Western Sources: the Changshan Snake of Sunzi's "Nine Grounds" and Kirchhoff's Circuit Equations of 1847 — Chapter 7 examines the Changshan snake of the Sunzi's "Nine Grounds" and Kirchhoff's circuit equations of 1847. (https://5dlibrary.com/answers/en/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c07/) 第7章 东西方原典考辨:《孙子·九地》的常山之蛇,基尔霍夫 1847 年的电路方程 — 第七章考辨《孙子·九地》的常山之蛇与基尔霍夫 1847 年的电路方程。 (https://5dlibrary.com/answers/zh/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c07/) - Ch.8 The Empirical Loop: a Quarterly Checkup of the Workflow Graph, Link Advice and a Plan for Refutation — Chapter 8 gives a quarterly loop of drawing, checking, adding and reviewing, in which three rounds of advice raise the algebraic connectivity of an illustrative workflow from 0.1824 to 1.0000; the distilled "Bottleneck Cut-Set Checker"; and a pre-registered test with 74 team-quarters per arm together with conditions for overturning the book. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c08/) 第8章 实证闭环:工作流图的季度体检、补边建议与证伪方案 — 第八章给出画图、体检、补边、复查的季度闭环,示意工作流三轮补边后代数连通度从 0.1824 升到 1.0000;蒸馏小工具「瓶颈割集体检器」;以及每组 74 个团队季度的预注册检验方案与推翻条件。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/graph-laplacian-spectral-gap-cheeger-workflow-supply-bottlenecks/c08/) ## 5DL-B-244 · Extreme-Value Theory and Generalised Pareto Tails: Whale-Order Ceilings and Reserves for Extreme Chargebacks 中文书名:极值理论广义帕累托尾部、超级鲸鱼客单上限与极端退单准备金母典 DOI: pending Cite: Fan, Y. (2026). Extreme-Value Theory and Generalised Pareto Tails: Whale-Order Ceilings and Reserves for Extreme Chargebacks (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 5 chooses the threshold with the mean excess plot and 150 simulated shops; the log error of the ten-year level is smallest at the ninetieth percentile, at 0.194. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: Shopkeepers, bookkeepers and those who run server rooms commonly draw the line for "unusually large" at the mean plus three standard deviations: orders above it are reviewed, reserves are held at it and server capacity is estimated from it. This book tests that ruler with extreme-value theory. For heavy-tailed data such as order values, chargeback losses and traffic bursts it systematically underestimates extremes, by a multiple set by the shape parameter of the tail that grows the deeper one goes, while the standard deviation itself is held hostage by a few whale orders. The book's orders, chargebacks, request volumes and parameters are illustrative; all calculations are generated by its programs; its discussion of reserves and server headroom is general business and financial method, not financial, accounting, tax or investment advice, and actual amounts and production changes must be decided by the operator. Chapter 1 shows that on an illustrative order tail with shape about 0.4 the true one-in-a-thousand quantile is 2.57 times the normal estimate and the one-in-ten-thousand quantile 5.56 times; on samples the normal estimate has a median of only 0.363 of the truth, and the 20 largest orders hold 0.353 of the sum of squared deviations. The blueprint's "several times" is accordingly corrected to two to three times at one in a thousand and five or six times at one in ten thousand. Chapter 2 presents the peaks-over-threshold theorem and shows that a lognormal looks like a power-law tail with shape 0.2148 at finite thresholds. Chapter 3 derives probability-weighted moments, whose root mean square error with 50 excesses is 0.169 against 0.209 for maximum likelihood. Chapter 4 turns parameters into return levels: the exact once-in-ten-years whale order is 27214 yuan where the normal estimate is only 1459, and a once-in-ten-years event occurs within ten years with probability 0.632. Chapter 5 chooses the threshold with the mean excess plot and 150 simulated shops; the log error of the ten-year level is smallest at the ninetieth percentile, at 0.194. Chapter 6 writes return levels into chargeback reserves and server headroom: at once in a thousand months the true monthly bill is 1.77 times the normal rule, a reserve spliced and fitted with the shape at its upper end is 9798 yuan, and the residual method gives servers a median of 0.98 overloaded minutes a year. Chapter 7 examines "treading on frost, hard ice comes" from Kun in the Book of Changes (on a heavy tail the chance of doubling once past a threshold is about one in five) and the extreme-value laws of Fisher and Tippett of 1928 (the normal maximum is still 0.0109 from its limit with a hundred thousand draws). Chapter 8 gives a monthly tail check-up, the Tail Reserve Estimator and pre-registered tests: testing the normal one-in-a-thousand line needs only 280 orders, testing the peaks-over-threshold line 9100. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:开店、管账与管机房的人,普遍用均值加三倍标准差给「异常大」划线:超过的订单要复核,准备金按它留,服务器容量照它估。本书用极值理论检验这把尺子:对订单金额、退单损失与突发流量这类重尾数据,它系统性地低估极端值,低估的倍数由尾部形状参数决定,越往深处越大,而标准差本身被少数鲸鱼单绑架。全部计算只用 Python 标准库。书中的订单、退单、请求量与参数均为示意;全部计算由书中程序生成;准备金与服务器冗余的讨论属于非个人化经营与财务方法,不构成理财、会计、税务或投资建议,实际金额与生产变更须由经营者人工决定。 第一章显示,在形状约 0.4 的示意订单尾部上,真实千分之一分位是正态估计的 2.57 倍,万分之一分位是 5.56 倍;样本上的正态估计中位数只有真实值的 0.363,最大的 20 笔订单占去离差平方和的 0.353。立项蓝图的「大数倍」据此修正为千分之一层两到三倍、万分之一层五六倍。第二章讲皮坎兹与巴尔凯马—德哈恩的峰过阈值定理,说明对数正态在有限阈值上看起来像形状 0.2148 的幂律尾。第三章推导概率加权矩,50 个超额时其均方根误差 0.169,低于最大似然的 0.209。第四章把参数换成回报水平:十年一遇的鲸鱼单精确值 27214 元,正态估计只有 1459 元;十年一遇的事件十年内发生的概率是 0.632。 第五章用平均超额图与 150 家模拟店铺选阈值,第九十百分位上十年一遇水平的对数误差最小,为 0.194。第六章把回报水平写进退单准备金与服务器冗余:月度账单在千个月一遇时真实值是正态规则的 1.77 倍,按形状上端拼接拟合的准备金为 9798 元;去节律后的残差方法使服务器一年过载分钟数的中位数为 0.98。第七章考辨《周易·坤》初六的「履霜,坚冰至」(重尾上越过阈值后再翻倍的概率约五分之一)与费希尔—蒂皮特 1928 年的极值分布(正态最大值在十万个样本时与极限仍差 0.0109)。第八章给出每月一次的尾部体检流程、尾部准备金估算器与预注册检验:检验正态千分之一线只需 280 笔订单,检验峰过阈值线需要 9100 笔。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Blind Spot of Mean Plus Three Standard Deviations: the One-in-a-Thousand Quantile and Whale Orders under Heavy Tails — Chapter 1 shows that on an illustrative order tail with shape about 0.4 the true one-in-a-thousand quantile is 2.57 times the normal estimate and the one-in-ten-thousand quantile 5.56 times; on samples the normal estimate has a median of only 0.363 of the truth, and the 20 largest orders hold 0.353 of the sum of squared deviations. The blueprint's "several times" is accordingly corrected to two to three times at one in a thousand and five or six times at one in ten thousand. (https://5dlibrary.com/answers/en/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c01/) 第1章 均值加三倍标准差的盲区:重尾下的千分之一分位与鲸鱼客单 — 第一章显示,在形状约 0.4 的示意订单尾部上,真实千分之一分位是正态估计的 2.57 倍,万分之一分位是 5.56 倍;样本上的正态估计中位数只有真实值的 0.363,最大的 20 笔订单占去离差平方和的 0.353。立项蓝图的「大数倍」据此修正为千分之一层两到三倍、万分之一层五六倍。 (https://5dlibrary.com/answers/zh/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c01/) - Ch.2 Peaks over Thresholds: the Pickands-Balkema-de Haan Theorem and the Convergence of Excesses — Chapter 2 presents the peaks-over-threshold theorem and shows that a lognormal looks like a power-law tail with shape 0.2148 at finite thresholds. (https://5dlibrary.com/answers/en/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c02/) 第2章 峰过阈值:皮坎兹—巴尔凯马—德哈恩定理与超额分布的收敛 — 第二章讲皮坎兹与巴尔凯马—德哈恩的峰过阈值定理,说明对数正态在有限阈值上看起来像形状 0.2148 的幂律尾。 (https://5dlibrary.com/answers/zh/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c02/) - Ch.3 Probability-Weighted Moments: Estimating the Two Generalised Pareto Parameters with the Standard Library — Chapter 3 derives probability-weighted moments, whose root mean square error with 50 excesses is 0.169 against 0.209 for maximum likelihood. (https://5dlibrary.com/answers/en/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c03/) 第3章 概率加权矩:只用标准库估计广义帕累托的两个参数 — 第三章推导概率加权矩,50 个超额时其均方根误差 0.169,低于最大似然的 0.209。 (https://5dlibrary.com/answers/zh/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c03/) - Ch.4 Return Levels: How Large Is the Whale Order That Comes Once in So Many Months — Chapter 4 turns parameters into return levels: the exact once-in-ten-years whale order is 27214 yuan where the normal estimate is only 1459, and a once-in-ten-years event occurs within ten years with probability 0.632. (https://5dlibrary.com/answers/en/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c04/) 第4章 回报水平:多少个月一遇的鲸鱼单有多大 — 第四章把参数换成回报水平:十年一遇的鲸鱼单精确值 27214 元,正态估计只有 1459 元;十年一遇的事件十年内发生的概率是 0.632。 (https://5dlibrary.com/answers/zh/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c04/) - Ch.5 The Mean Excess Plot and the Choice of Threshold: Bias, Variance and the Lognormal Trap — Chapter 5 chooses the threshold with the mean excess plot and 150 simulated shops; the log error of the ten-year level is smallest at the ninetieth percentile, at 0.194. (https://5dlibrary.com/answers/en/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c05/) 第5章 平均超额图与阈值选择:偏差、方差与对数正态陷阱 — 第五章用平均超额图与 150 家模拟店铺选阈值,第九十百分位上十年一遇水平的对数误差最小,为 0.194。 (https://5dlibrary.com/answers/zh/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c05/) - Ch.6 Chargeback Reserves and Server Headroom: Writing Return Levels into the Ledger and the Server Room — Chapter 6 writes return levels into chargeback reserves and server headroom: at once in a thousand months the true monthly bill is 1.77 times the normal rule, a reserve spliced and fitted with the shape at its upper end is 9798 yuan, and the residual method gives servers a median of 0.98 overloaded minutes a year. (https://5dlibrary.com/answers/en/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c06/) 第6章 退单准备金与服务器冗余:把回报水平写进账本与机房 — 第六章把回报水平写进退单准备金与服务器冗余:月度账单在千个月一遇时真实值是正态规则的 1.77 倍,按形状上端拼接拟合的准备金为 9798 元;去节律后的残差方法使服务器一年过载分钟数的中位数为 0.98。 (https://5dlibrary.com/answers/zh/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c06/) - Ch.7 Classical Sources East and West: "Treading on Frost, Hard Ice Comes" and the Extreme-Value Laws of Fisher and Tippett — Chapter 7 examines "treading on frost, hard ice comes" from Kun in the Book of Changes (on a heavy tail the chance of doubling once past a threshold is about one in five) and the extreme-value laws of Fisher and Tippett of 1928 (the normal maximum is still 0.0109 from its limit with a hundred thousand draws). (https://5dlibrary.com/answers/en/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c07/) 第7章 东西方原典考辨:「履霜,坚冰至」与费希尔—蒂皮特的极值分布 — 第七章考辨《周易·坤》初六的「履霜,坚冰至」(重尾上越过阈值后再翻倍的概率约五分之一)与费希尔—蒂皮特 1928 年的极值分布(正态最大值在十万个样本时与极限仍差 0.0109)。 (https://5dlibrary.com/answers/zh/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c07/) - Ch.8 Closing the Empirical Loop: a Monthly Tail Check-Up, the Tail Reserve Estimator and Pre-Registered Tests — Chapter 8 gives a monthly tail check-up, the Tail Reserve Estimator and pre-registered tests: testing the normal one-in-a-thousand line needs only 280 orders, testing the peaks-over-threshold line 9100. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c08/) 第8章 实证闭环:每月一次的尾部体检、尾部准备金估算器与预注册检验 — 第八章给出每月一次的尾部体检流程、尾部准备金估算器与预注册检验:检验正态千分之一线只需 280 笔订单,检验峰过阈值线需要 9100 笔。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/extreme-value-generalised-pareto-whale-orders-chargeback-reserves/c08/) ## 5DL-B-245 · Kalman Filtering and State-Space Assimilation: Separating Observation Noise from True Signals in Wearable Readings and Advertising Reports 中文书名:卡尔曼滤波状态空间同化、可穿戴读数观测噪音与广告归因真实信号提纯母典 DOI: pending Cite: Fan, Y. (2026). Kalman Filtering and State-Space Assimilation: Separating Observation Noise from True Signals in Wearable Readings and Advertising Reports (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its noise variances, drifts, returns and budgets are illustrative parameters; all calculations are generated by its programs. On wearable readings it is not medical advice, claims no effect, judges no one's state of health, and asks readers to consult a licensed physician before any self-experiment; on advertising attribution it is not individual business advice. Abstract: A glance at the wristband or the advertising report each day, followed by a decision for the day, rests on the common sense that newer is more accurate. This book tests that common sense with local level and local trend state space models: a reading is the true level plus measurement error, and when the error variance far exceeds the variance of the true daily change, yes-or-no decisions made on raw readings overreact to noise; with illustrative parameters raw readings flip the call 23.6 times per hundred days and the Kalman estimate only 5.5 times. After checking the sources, the book corrects its founding proposition: Kalman's paper of 1960 is a methods paper with no wearable or advertising data; the benefit of filtering grows steadily with the noise-to-signal ratio and is almost nil near zero, and its price is a lag of a few days after a real jump. Its noise variances, drifts, returns and budgets are illustrative parameters; all calculations are generated by its programs. On wearable readings it is not medical advice, claims no effect, judges no one's state of health, and asks readers to consult a licensed physician before any self-experiment; on advertising attribution it is not individual business advice. Chapter 1 quantifies the proposition by simulation and sweeps the noise-to-signal ratio. Chapter 2 derives the Kalman gain, shows that the steady filter is exponential smoothing with the best weight, and compares it with the seven-day average. Chapter 3 shows that a slow trend takes weeks to see and that, at equal false-alarm rates, the filter alarms no later than raw readings. Chapter 4 covers backward smoothing, missing days and an innovation gate for artefact readings. Chapter 5 covers maximum likelihood, innovation tests and moments of differences: a gain estimated from four weeks of data lies between 0.031 and 0.481 and is only a starting point. Chapter 6 turns to advertising reports: switching the budget on the raw report changes level 33.8 times in ninety days and earns 0.714 of the ideal, switching on the filtered estimate changes only 5.6 times and earns 0.903, and with constant spend no method learns the effect of advertising. Chapter 7 examines the Mencius line "believing the book entirely is worse than having no book" and Gauss's least squares of 1809. Chapter 8 gives the closed loops for a four-week filtered log and for filtered advertising reports, the Noisy-Reading Filter tool and two pre-registered protocols (96 advertising accounts under conservative assumptions, 57 personal logs). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:每天看一眼手环读数或投放日报,再据此决定今天怎么做,背后的常识是「越新越准」。本书用局部水平与局部趋势的状态空间模型检验这一常识:一个读数等于真实水平加测量误差,当误差方差远大于真实日变化方差时,按原始读数做是或否的决定会对噪音过度反应;示意参数下原始读数每百天翻转 23.6 次,卡尔曼滤波估计只翻转 5.5 次。核对原文后,本书修正了立项命题:卡尔曼 1960 年的论文是方法论文,不含任何可穿戴或广告数据;滤波的好处随噪音信号比单调增大,比值接近零时几乎没有,代价是对真实突变迟钝几天。书中的噪音方差、漂移、回报率与预算均为示意参数;全部计算由书中程序生成。关于可穿戴读数:本书非医疗建议,不宣称任何功效,不判断任何人的健康状况,自我试验先咨询执业医师;关于广告归因:本书是非个人化经营建议。 第一章用模拟量化命题并扫描噪音信号比。第二章推出卡尔曼增益,说明稳态滤波就是权重最优的指数平滑,并与七日均线比较。第三章讨论慢趋势要几周才看得出,以及在误报相同时滤波并不比原始读数报警更慢。第四章讲回看平滑、缺测日与伪读数的新息门限。第五章讲极大似然、新息检验与差分矩估计:四周数据估出的增益在 0.031 到 0.481 之间,只能当作起点。 第六章转到广告日报:按原始日报切换预算九十天切换 33.8 次、利润只有理想的 0.714,按滤波估计只切换 5.6 次、达到 0.903;花费不变时任何方法都学不到广告效果。第七章考辨《孟子·尽心下》的「尽信书,则不如无书」与高斯 1809 年的最小二乘。第八章给出四周滤波日志与广告日报滤波的闭环、小工具「噪音读数滤波器」与两份预注册方案(广告账户保守需要 96 个,个人日志需要 57 份)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Why Raw Readings Overreact: the Noise-to-Signal Ratio and Decision Flips — A reading is the true level plus measurement error. When the variance of the error far exceeds the variance of the true daily change, decisions taken on raw readings flip often, and most of those flips respond to error; in the illustrative case with a noise-to-signal ratio of 16, raw readings flip 4.26 times as often as the filtered estimate. How much is gained depends entirely on the ratio: near zero the filter is useless, and the larger the ratio the more useful and the more sluggish it becomes. Where, then, does this gain that absorbs a fixed share of the error come from, and how does it relate to the familiar seven-day average and to exponential smoothing? That is the question of the next chapter. (https://5dlibrary.com/answers/en/kalman-state-space-wearable-noise-ad-attribution/c01/) 第1章 原始读数为什么过度反应:噪音信号比与决策翻转 — 一个读数等于真实水平加测量误差。当误差的方差远大于真实水平的日变化方差时,按原始读数做的决定会频繁翻转,翻转的绝大部分是在回应误差;在噪音信号比为 16 的示意情形下,原始读数的翻转次数是滤波估计的 4.26 倍。好处的大小完全取决于噪音信号比:比值接近零时滤波没有用,比值越大越有用,也越迟钝。那么,这个按比例吸收误差的增益究竟从何而来,它与人们熟悉的七日均线和指数平滑又是什么关系?这是下一章的问题。 (https://5dlibrary.com/answers/zh/kalman-state-space-wearable-noise-ad-attribution/c01/) - Ch.2 Where the Kalman Gain Comes From: One Prediction, One Correction and Exponential Smoothing — The Kalman filter for the local level model is four lines of arithmetic: the predicted variance grows by $q$, the gain is the predicted variance over the predicted variance plus $r$, the estimate absorbs the innovation in proportion to the gain, and the variance shrinks by the gain. The gain settles within days, and once settled the filter is exponential smoothing with weight equal to the gain; with the illustrative parameters it is 0.2207, with half the weight on the last three days. The seven-day average loses little on noisy readings but does worse than no smoothing on clean ones. Real metrics, however, do not always drift at random around one level. Scale readings may have a sustained trend, and orders may step up on a given day because of a real event. What happens to the filter when it meets a trend or a step? The next chapter answers. (https://5dlibrary.com/answers/en/kalman-state-space-wearable-noise-ad-attribution/c02/) 第2章 卡尔曼增益从哪里来:一步预测、一步修正与指数平滑 — 局部水平模型下的卡尔曼滤波只有四行算术:预测方差加 $q$,增益等于预测方差除以预测方差加 $r$,估计按增益吸收新息,方差按增益缩小。增益几天之内稳定下来,稳定之后就是平滑系数等于增益的指数平滑;在示意参数下它是 0.2207,一半的权重落在最近三天。七日均线在嘈杂读数上差得不多,在干净读数上却比不平滑还差。可是真实的指标并不总是围着一个水平随机漂移:体重秤读数可能有持续的趋势,订单可能在某一天因为一次真实的事件跳上一个台阶。滤波遇到趋势与台阶会怎样?下一章回答。 (https://5dlibrary.com/answers/zh/kalman-state-space-wearable-noise-ad-attribution/c02/) - Ch.3 Trends and Steps: Slow Drift Takes Weeks to Show, a Real Jump Costs Days of Lag — Adding a slope state turns the local level model into the local trend model. With the illustrative parameters, a single reading's error is twelve times the daily drift, and comparing two readings tells almost nothing about trend; the filter needs about eight weeks of continuous records to have a 0.882 chance of seeing it. For a real step the filter does need a few days, but at equal false-alarm rates it is no slower than alarming on raw readings. The first three chapters have looked at today in real time. If we are willing to look back and use today's readings to revise our judgement of last week, how much further can the error be cut? And what should be done with missing days and readings that are plainly absurd? The next chapter answers. (https://5dlibrary.com/answers/en/kalman-state-space-wearable-noise-ad-attribution/c03/) 第3章 趋势与台阶:慢漂移要几周才看得出,真突变要付几天迟钝 — 加一个斜率状态,局部水平模型就变成局部趋势模型。在示意参数下,单次读数误差是每日漂移的十二倍,用两次读数比较趋势几乎没有信息;滤波要连续记录约八周,才有 0.882 的机会看出趋势。对于真实的台阶,滤波确实需要几天,但在误报相同的条件下,它不比按原始读数报警更慢。前三章都在实时地看今天;如果我们愿意回头看,用今天的读数去修正对上周的判断,又能把误差压低多少?缺测的日子和明显荒唐的读数又该怎样处理?下一章回答。 (https://5dlibrary.com/answers/zh/kalman-state-space-wearable-noise-ad-attribution/c03/) - Ch.4 Looking Back and Filling Gaps: the RTS Smoother, Missing Days and Artefact Readings — In review, backward smoothing lowers the error of the middle days from 0.940 to 0.705, but it does nothing for the last day. Missing days need no imputation: the filter lets its uncertainty grow honestly and speeds up its catch-up when readings resume. Artefact readings should be stopped by an innovation gate before they enter the estimate and passed to a person for review. Up to this point the book has assumed that we know the two noise variances $q$ and $r$. In reality nobody tells us these numbers. Can they be estimated from the readings alone, and are four weeks of data enough? The next chapter answers. (https://5dlibrary.com/answers/en/kalman-state-space-wearable-noise-ad-attribution/c04/) 第4章 回看与补洞:RTS 平滑、缺测日与伪读数 — 回看平滑在复盘时把中间日子的误差从 0.940 降到 0.705,但对最后一天毫无帮助。缺测日不需要插补,滤波器会让不确定性如实增大,读数恢复后自动加快追赶。伪读数应当用新息门限在进入估计之前拦下并交人复核。到这里,本书一直假设我们知道两种噪音的方差 $q$ 与 $r$。现实中没有人告诉我们这两个数。只凭读数本身,能把它们估出来吗?四周的数据够不够?下一章回答。 (https://5dlibrary.com/answers/zh/kalman-state-space-wearable-noise-ad-attribution/c04/) - Ch.5 How Large Is the Noise? Maximum Likelihood, Innovation Tests and Moments of Differences — Chapter 5 covers maximum likelihood, innovation tests and moments of differences: a gain estimated from four weeks of data lies between 0.031 and 0.481 and is only a starting point. (https://5dlibrary.com/answers/en/kalman-state-space-wearable-noise-ad-attribution/c05/) 第5章 噪音有多大:极大似然、新息检验与差分矩估计 — 第五章讲极大似然、新息检验与差分矩估计:四周数据估出的增益在 0.031 到 0.481 之间,只能当作起点。 (https://5dlibrary.com/answers/zh/kalman-state-space-wearable-noise-ad-attribution/c05/) - Ch.6 The Advertising Report: Attribution Noise, On/Off Flighting and Budget Switching — Chapter 6 turns to advertising reports: switching the budget on the raw report changes level 33.8 times in ninety days and earns 0.714 of the ideal, switching on the filtered estimate changes only 5.6 times and earns 0.903, and with constant spend no method learns the effect of advertising. (https://5dlibrary.com/answers/en/kalman-state-space-wearable-noise-ad-attribution/c06/) 第6章 广告日报:归因噪音、开关试验与预算切换 — 第六章转到广告日报:按原始日报切换预算九十天切换 33.8 次、利润只有理想的 0.714,按滤波估计只切换 5.6 次、达到 0.903;花费不变时任何方法都学不到广告效果。 (https://5dlibrary.com/answers/zh/kalman-state-space-wearable-noise-ad-attribution/c06/) - Ch.7 Classical Sources East and West: the Mencius and Gauss's Least Squares — Chapter 7 examines the Mencius line "believing the book entirely is worse than having no book" and Gauss's least squares of 1809. (https://5dlibrary.com/answers/en/kalman-state-space-wearable-noise-ad-attribution/c07/) 第7章 东西方原典考辨:《孟子·尽心下》与高斯的最小二乘 — 第七章考辨《孟子·尽心下》的「尽信书,则不如无书」与高斯 1809 年的最小二乘。 (https://5dlibrary.com/answers/zh/kalman-state-space-wearable-noise-ad-attribution/c07/) - Ch.8 Closing the Loop: a Four-Week Filtered Log, Filtered Advertising Reports and Pre-Registered Tests — Chapter 8 gives the closed loops for a four-week filtered log and for filtered advertising reports, the Noisy-Reading Filter tool and two pre-registered protocols (96 advertising accounts under conservative assumptions, 57 personal logs). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/kalman-state-space-wearable-noise-ad-attribution/c08/) 第8章 实证闭环:四周滤波日志、广告日报滤波与预注册检验 — 第八章给出四周滤波日志与广告日报滤波的闭环、小工具「噪音读数滤波器」与两份预注册方案(广告账户保守需要 96 个,个人日志需要 57 份)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/kalman-state-space-wearable-noise-ad-attribution/c08/) ## 5DL-B-246 · Braess's Paradox and Wardrop Equilibrium: When One More Purchase Entry or One More Coordination Layer Slows Everyone Down 中文书名:布雷斯悖论瓦德罗普均衡、购买入口加法反效与协同层级的无秩序代价母典 DOI: pending Cite: Fan, Y. (2026). Braess's Paradox and Wardrop Equilibrium: When One More Purchase Entry or One More Coordination Layer Slows Everyone Down (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: Most owners who sell through live rooms, private chat and mini-programs, and most owners who run a few small delivery teams, believe that "one more way to order means one more sale, and one more coordination layer means less friction". This book tests that intuition, that adding cannot hurt, with Wardrop equilibrium and Braess's paradox: when a new entry or a coordination layer sends people through two bottlenecks staffed by people at once, and everyone takes the route quickest for himself, everyone can end up slower. The commissioning blueprint attached the claim to Braess in 1968 and to Roughgarden and Tardos in 2002; after checking the sources the book narrows it: Braess wrote about roads, and entries and layers are this book's analogy; the paradox appears only inside a window of demand; the four-thirds ceiling holds only for linear waits; and going from "slower" to "fewer orders" needs an assumption of elastic demand. The book divides the work with Volume 32, which treats the road prototype and Pigouvian taxes; this volume concentrates on purchase entries and coordination layers and implements the equilibrium algorithm of Frank and Wolfe from scratch in the standard library. The parameters are illustrative; all calculations are generated by the book's programs; the book is not individual business advice, and any production change that removes an entry or a coordination layer must be reviewed and released by hand by the person in charge. Chapter 1 explains Wardrop's two principles with two examples in fractions. Chapter 2 recomputes Braess's original network, finds the paradox's demand window from 2.581 to 8.889, and shows that fourth-power queue costs push the ceiling on losses to 2.1505. Chapter 3 implements the method of Frank and Wolfe and finds the system optimum. Chapter 4 shows on a funnel with three entries that a new entry raises the average time to order from 12.200 to 16.000 minutes and, with a patience scale of 40 minutes, cuts orders by 8.5 %, and gives an audit that closes entries one at a time. Chapter 5 shows that a relay-only coordination layer lengthens the cycle of requests from 8.623 to 10.957 days and gets worse the more efficient it becomes, while a layer with its own handling capacity is the best design. Chapter 6 sets out the opposing evidence: in random networks the share in which a new link hurts is only 0.020 to 0.084, vaguer information among buyers shrinks the harm, and a new entry that brings about 10.0 % new customers comes out ahead. Chapter 7 examines "the great road is very smooth, yet the people love byways" from chapter 53 of the Laozi, Pigou's two roads, and Knight's criticism. Chapter 8 shows that a user-level split reports an entry that adds 3.800 minutes as one that saves 2.616 minutes, and therefore uses whole-block switchbacks instead; it gives the distilled "Entry-Deletion Equilibrium Calculator" and a pre-registered protocol, with 49 days per arm for the entry test and 99 requests per arm for the layer audit. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:做直播、私域与小程序的店主,带着几个小团队交付项目的经营者,大多相信「多一个下单入口就多一份成交,多一层协同就少一分内耗」。本书用瓦德罗普均衡与布雷斯悖论检验这条「加法不会有害」的直觉:当一个新入口或一层协同同时把人引向两处靠人的瓶颈,而每个人都选对自己最快的路线时,所有人都可能更慢。立项蓝图把命题挂在布雷斯 1968 年与拉夫加登、塔尔多什 2002 年的论文名下;本书核对原文后收窄为:布雷斯讲的是道路,入口与协同层是本书的类比;悖论只在一个需求窗口里出现;三分之四的上界只对线性等待成立;从「更慢」推到「更少成交」还需要弹性需求的假设。本书与第三十二卷分工:后者讲道路原型与庇古税,本卷聚焦购买入口与协同层,并用标准库从头实现弗兰克与沃尔夫的均衡算法。书中参数为示意;全部计算由书中程序生成;本书是非个人化经营建议,任何删入口、撤协同层的生产变更须由主事者人工复核放行。 第一章用两个分数例子讲清瓦德罗普的两条原则。第二章复算布雷斯的原始网络,求出悖论的需求窗口 2.581 到 8.889,并显示四次方排队成本把损失上界推到 2.1505。第三章实现弗兰克与沃尔夫的方法并求系统最优。第四章在三入口漏斗上显示新入口让平均下单时长从 12.200 分钟升到 16.000 分钟,耐心尺度 40 分钟时成交减少 8.5 %,并给出逐条关闭入口的体检。第五章显示只做转交的协同层把请求周期从 8.623 天拉长到 10.957 天,而且越高效越糟,自带处理能力的协同层却最好。第六章陈列反方证据:随机网络里新链路有害的比例只有 0.020 到 0.084,买家信息越模糊害处越小,新入口带来约 10.0 % 新客即可扭亏。 第七章考辨《老子》第五十三章「大道甚夷,而民好径」与庇古的两条路及奈特的批评。第八章说明按用户分流的实验会把增加 3.800 分钟的入口报告成缩短 2.616 分钟,因而改用整时段切换;给出蒸馏小工具「入口删减均衡计算器」与预注册方案,删入口实验每组 49 天,层级审计每组 99 个请求。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Two Principles: Wardrop Equilibrium, and How a "One-Tap" Entry Makes Every Buyer Wait Four Minutes Longer — Wardrop's first principle describes the result of buyers choosing their routes for themselves, the second the result with the least average wait. Program p11 shows that a crowded fast lane is used, at the user equilibrium, until it is as slow as the standard checkout, an average of 10.000 minutes, whereas the optimal split needs only 7.500, a ratio of 1.3333. Program p12 shows that a zero-cost "one-tap" entry takes every buyer from 12 minutes to 16, and that nobody can get quicker by switching alone. The next chapter returns to Braess's original network and asks a sharper question: at what levels of demand does a new entry actually hurt? (https://5dlibrary.com/answers/en/braess-wardrop-equilibrium-extra-entry-coordination-layer/c01/) 第1章 两条原则:瓦德罗普均衡与「一键直达」让每位买家多等四分钟 — 瓦德罗普第一原则描述买家各自选路的结果,第二原则描述平均等待最短的结果。程序 p11 显示,一条拥挤的快车道在用户均衡下被用到与标准结账一样慢,平均 10.000 分钟,而最优分流只需 7.500,比值 1.3333。程序 p12 显示,一个零成本的「一键直达」入口让每位买家从 12 分钟变成 16 分钟,没有人能单独换路变快。下一章回到布雷斯的原始网络,追问一个更尖锐的问题:在什么样的需求下,新入口才会伤人? (https://5dlibrary.com/answers/zh/braess-wardrop-equilibrium-extra-entry-coordination-layer/c01/) - Ch.2 The Main Argument: the Demand Window of Braess's Paradox and the Limits of the Four-Thirds Ceiling — Chapter 2 recomputes Braess's original network, finds the paradox's demand window from 2.581 to 8.889, and shows that fourth-power queue costs push the ceiling on losses to 2.1505. (https://5dlibrary.com/answers/en/braess-wardrop-equilibrium-extra-entry-coordination-layer/c02/) 第2章 主轴:布雷斯悖论的需求窗口与三分之四上界的边界 — 第二章复算布雷斯的原始网络,求出悖论的需求窗口 2.581 到 8.889,并显示四次方排队成本把损失上界推到 2.1505。 (https://5dlibrary.com/answers/zh/braess-wardrop-equilibrium-extra-entry-coordination-layer/c02/) - Ch.3 Frank and Wolfe in the Standard Library: Equilibrium, Optimum and the Relative Gap — Program p31 reproduces 83.0000 and 92.0000 on Braess's network, with link flows equal to the exact fractions. Program p32 shows that Frank and Wolfe with a line search brings the gap to 1.588e-03 by the tenth step, while successive averages is still at 3.952e-04 after three hundred. Program p33 uses the same code to find the system optimum: the new entry lowers the optimal average from 12.030 to 11.397 but raises the user equilibrium from 12.200 to 16.000, a ratio of 1.4039. The next chapter hands these tools to the shop owner: from minutes to orders, and then closing entries one at a time to see which one is the Braess link. (https://5dlibrary.com/answers/en/braess-wardrop-equilibrium-extra-entry-coordination-layer/c03/) 第3章 标准库里的弗兰克与沃尔夫算法:均衡、最优与相对间隙 — 程序 p31 在布雷斯网络上复现了 83.0000 与 92.0000,路段流量与分数解一致。程序 p32 显示,带一维搜索的弗兰克与沃尔夫的方法在第十步就把间隙降到 1.588e-03,逐次平均法第三百步还停在 3.952e-04。程序 p33 用同一套代码求出系统最优:新入口让最优平均从 12.030 降到 11.397,却让用户均衡从 12.200 升到 16.000,比值 1.4039。下一章把这套工具交给店主:从分钟推到订单,再逐个关闭入口,看看哪一个才是那条布雷斯链路。 (https://5dlibrary.com/answers/zh/braess-wardrop-equilibrium-extra-entry-coordination-layer/c03/) - Ch.4 Purchase Entries: a Funnel Network, Elastic Orders and an Audit that Closes Entries One at a Time — Chapter 4 shows on a funnel with three entries that a new entry raises the average time to order from 12.200 to 16.000 minutes and, with a patience scale of 40 minutes, cuts orders by 8.5 %, and gives an audit that closes entries one at a time. (https://5dlibrary.com/answers/en/braess-wardrop-equilibrium-extra-entry-coordination-layer/c04/) 第4章 购买入口:漏斗网络、弹性订单与逐条关闭入口的体检 — 第四章在三入口漏斗上显示新入口让平均下单时长从 12.200 分钟升到 16.000 分钟,耐心尺度 40 分钟时成交减少 8.5 %,并给出逐条关闭入口的体检。 (https://5dlibrary.com/answers/zh/braess-wardrop-equilibrium-extra-entry-coordination-layer/c04/) - Ch.5 Coordination Layers: the More Efficient a Pure Relay, the Slower Every Request — Chapter 5 shows that a relay-only coordination layer lengthens the cycle of requests from 8.623 to 10.957 days and gets worse the more efficient it becomes, while a layer with its own handling capacity is the best design. (https://5dlibrary.com/answers/en/braess-wardrop-equilibrium-extra-entry-coordination-layer/c05/) 第5章 协同层:只做转交的中转层越高效,所有请求越慢 — 第五章显示只做转交的协同层把请求周期从 8.623 天拉长到 10.957 天,而且越高效越糟,自带处理能力的协同层却最好。 (https://5dlibrary.com/answers/zh/braess-wardrop-equilibrium-extra-entry-coordination-layer/c05/) - Ch.6 The Opposing Evidence: How Common the Paradox Is, How Much Buyers Know, and How Many New Customers a New Entry Brings — Chapter 6 sets out the opposing evidence: in random networks the share in which a new link hurts is only 0.020 to 0.084, vaguer information among buyers shrinks the harm, and a new entry that brings about 10.0 % new customers comes out ahead. (https://5dlibrary.com/answers/en/braess-wardrop-equilibrium-extra-entry-coordination-layer/c06/) 第6章 反方证据:悖论有多常见、买家知道多少、新入口带来多少新客 — 第六章陈列反方证据:随机网络里新链路有害的比例只有 0.020 到 0.084,买家信息越模糊害处越小,新入口带来约 10.0 % 新客即可扭亏。 (https://5dlibrary.com/answers/zh/braess-wardrop-equilibrium-extra-entry-coordination-layer/c06/) - Ch.7 Classical Sources East and West: "The Great Road Is Very Smooth, yet People Love the Byways" and Pigou's Two Roads — Program p71 writes "the great road is very smooth, yet the people love byways" as Pigou's two roads: as long as anyone uses the main road, the byway is squeezed until it is exactly as slow, and those who take it gain nothing; the optimal split could lower the average from 10.000 to 9.000. The ancient description of behaviour is right; its moral attribution of the cause can be refuted by the model. Program p72 tests Knight's criticism of Pigou: when the main road is uncongested, the owner's toll of 4.000 is exactly the optimal toll; when the main road congests too, the owner raises his toll to 6.500 and the average, 11.440, stays above the optimal 11.143. The last chapter turns these models into a loop an owner can execute: how can one run, without harming customers, an experiment that could actually overturn this book? (https://5dlibrary.com/answers/en/braess-wardrop-equilibrium-extra-entry-coordination-layer/c07/) 第7章 东西方原典考辨:「大道甚夷,而民好径」与庇古的两条路 — 程序 p71 把「大道甚夷,而民好径」写成庇古形状的两条路:只要还有人走大路,小路就被挤到与大路一样慢,走小路的人一无所获;最优分流能把平均从 10.000 降到 9.000。古人对行为的描述是对的,对原因的道德归因可以被模型反驳。程序 p72 检验了奈特对庇古的批评:大路不拥挤时,所有者的通行费 4.000 恰好是最优通行费;大路也拥挤时,所有者把通行费抬到 6.500,平均 11.440,仍高于最优的 11.143。最后一章要把这些模型变成一个经营者可以执行的闭环:怎样在不伤害顾客的前提下,做一次真正能推翻本书的实验? (https://5dlibrary.com/answers/zh/braess-wardrop-equilibrium-extra-entry-coordination-layer/c07/) - Ch.8 Closing the Loop: Whole-Block Switchback Tests for Entries, an Audit of Organisational Layers, and a Pre-Registered Protocol — Chapter 8 shows that a user-level split reports an entry that adds 3.800 minutes as one that saves 2.616 minutes, and therefore uses whole-block switchbacks instead; it gives the distilled "Entry-Deletion Equilibrium Calculator" and a pre-registered protocol, with 49 days per arm for the entry test and 99 requests per arm for the layer audit. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/braess-wardrop-equilibrium-extra-entry-coordination-layer/c08/) 第8章 实证闭环:删入口的整时段切换实验、组织层级审计与预注册方案 — 第八章说明按用户分流的实验会把增加 3.800 分钟的入口报告成缩短 2.616 分钟,因而改用整时段切换;给出蒸馏小工具「入口删减均衡计算器」与预注册方案,删入口实验每组 49 天,层级审计每组 99 个请求。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/braess-wardrop-equilibrium-extra-entry-coordination-layer/c08/) ## 5DL-B-247 · Turing Reaction-Diffusion Patterns: How the Outposts of a Niche Word-of-Mouth Brand Emerge in a New City, and the Diffusion-Ratio Condition for Regional Dominance 中文书名:图灵反应扩散斑图、小众口碑品牌陌生城市据点涌现与区域垄断扩散比条件母典 DOI: pending Cite: Fan, Y. (2026). Turing Reaction-Diffusion Patterns: How the Outposts of a Niche Word-of-Mouth Brand Emerge in a New City, and the Diffusion-Ratio Condition for Regional Dominance (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: The book's reaches, decay rates, city lengths and data are illustrative; every calculation is generated by its programs; it offers general business method, not investment, legal or site-selection advice. Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. Abstract: When a niche word-of-mouth brand enters an unfamiliar city, the two usual approaches are to spread outposts evenly or to win one spot first; both assume that the operator can simply draw the layout. This book tests that assumption with Turing's diffusion-driven instability of 1952 and the activator-inhibitor equations of Gierer and Meinhardt of 1972. Reading word of mouth as short-range self-amplification, and saturation and cannibalisation as long-range inhibition, even activity breaks by itself into a row of equally spaced peaks once the ratio of the cannibalisation reach to the word-of-mouth reach passes a threshold. All calculations use only the Python standard library. The book's reaches, decay rates, city lengths and data are illustrative; every calculation is generated by its programs; it offers general business method, not investment, legal or site-selection advice. "Regional dominance" means only the concentration of one brand's own activity and share, never coordination or territorial agreements with other firms. Chapter 1 derives the threshold: 11.657 as a ratio of diffusion coefficients and 2.414 as a ratio of reaches, the latter independent of decay rates; the blueprint's "certain multiple" is corrected accordingly, and the finding of Marcon and colleagues that unequal diffusion is not generally necessary is set beside it. Chapter 2 gives an implicit-explicit finite-difference scheme with grid checks. Chapter 3 treats the dispersion relation: with default parameters the fastest wavelength is 7.13 km and the coefficient of variation of gaps over twenty starts is only 0.073, a layout that is regular but not unique. Chapter 4 estimates a word-of-mouth reach of 0.551 km and a cannibalisation reach of 2.343 km from referral and opening records, with a one-sided lower bound of 3.863 for their ratio, and gives the data needed for a call. Chapter 5 shows that past the threshold the hottest fifth of the corridor holds 0.493 of activity, while existing category demand pushes the threshold up to 6.464 and eventually rules out any pattern. Chapter 6 places shops in the model as fixed seeds: the layout collapses when the gap shrinks to 4.44 km, and gaps that are too wide do not fill themselves without ambient demand. Chapter 7 examines Sunzi's "we concentrate into one, the enemy divides into ten" (a peak is only 2.72 times the even level, far from tenfold) and von Thunen's Isolated State of 1826 (catchment spacing moves with price, Turing spacing does not). Chapter 8 gives an eight-step planning procedure, the Outpost Spacing Wavelength Calculator and a pre-registered test: the regression of spacing on reaches has power 0.989 with 15 cities. Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. 摘要:口碑型小众品牌进入陌生城市时,常见的两种做法是均匀撒点与先做透一个点,两者都默认据点格局可以由经营者随意画出。本书用图灵 1952 年的扩散驱动失稳与吉尔、迈恩哈特 1972 年的激活子—抑制子方程检验这一默认:把口碑读成短程自我放大,把饱和与同店蚕食读成长程抑制,当蚕食半径与口碑半径之比越过门槛时,均匀热度会自己碎成一串等距高点。全部计算只用 Python 标准库。书中半径、衰减率、城市长度与数据均为示意;全部计算由书中程序生成;本书属非个人化经营方法,不构成投资、法律或选址决策建议。「区域垄断」只指品牌自身热度与份额的集中,不涉及与任何企业的协同或分区约定。 第一章推出门槛:以扩散系数之比计是 11.657,以作用半径之比计是 2.414,后者与衰减率无关;立项蓝图的「一定倍数」据此修正,并并陈马尔孔等人关于差速扩散并非普遍必要的反方结论。第二章给出隐显式有限差分与网格校核。第三章讲色散关系:默认参数下最快波长 7.13 公里,二十个起点的间距变异系数只有 0.073,格局规整却不唯一。第四章从转介距离与开店记录估出口碑半径 0.551 公里、蚕食半径 2.343 公里,比值单侧下界 3.863,并给出判断所需的数据量。 第五章说明过门槛后最热的五分之一走廊占 0.493 的热度,而品类既有需求会把门槛推高到 6.464 乃至无法成斑。第六章把门店作为固定种子放进模型:间距缩到 4.44 公里时格局塌缩,过疏的空档在没有环境需求时不会自补。第七章考辨《孙子·虚实》的「我专为一,敌分为十」(高点只有均匀水平的 2.72 倍,达不到十倍)与冯·杜能 1826 年的《孤立国》(间距随价格变,图灵间距不随价格变)。第八章给出八步规划流程、据点间距波长计算器与预注册检验:间距对半径的回归斜率检验在 15 座城市时功效 0.989。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Why Evenness Breaks: the Turing Condition and the Threshold on the Diffusion Ratio — Chapter 1 derives the threshold: 11.657 as a ratio of diffusion coefficients and 2.414 as a ratio of reaches, the latter independent of decay rates; the blueprint's "certain multiple" is corrected accordingly, and the finding of Marcon and colleagues that unequal diffusion is not generally necessary is set beside it. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-niche-brand-outposts-new-city/c01/) 第1章 均匀为什么会失稳:图灵条件与扩散比门槛 — 第一章推出门槛:以扩散系数之比计是 11.657,以作用半径之比计是 2.414,后者与衰减率无关;立项蓝图的「一定倍数」据此修正,并并陈马尔孔等人关于差速扩散并非普遍必要的反方结论。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-niche-brand-outposts-new-city/c01/) - Ch.2 One-Dimensional Finite Differences: How Even Activity Breaks into Equally Spaced Peaks — The implicit-explicit finite-difference scheme solves diffusion and decay at the next time level, keeps production at the current one and solves the tridiagonal systems with the Thomas algorithm, so that pure Python integrates a 40 km corridor in a second. Program p21 shows that, starting from even activity, the layout passes through linear amplification, nonlinear sharpening and the merging of peaks, and with the default parameters comes to rest at 6 peaks with a mean gap of 7.20 km, while total activity falls from 80.00 to 63.67. Program p22 shows that when cell width and time step vary within a sensible range the number of peaks stays fixed and the spacing moves by less than one per cent, whereas a different draw of initial noise can move the spacing by a tenth. What, then, sets the spacing? And why can the same equations come to rest at different spacings? The next chapter returns to the dispersion relation to answer both questions. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-niche-brand-outposts-new-city/c02/) 第2章 一维有限差分:均匀热度怎样碎成等距高点 — 隐显式有限差分把扩散与衰减放到下一时刻去解,把产生项留在当前时刻,用托马斯算法解三对角方程组,让纯 Python 在一秒内算完一条 40 公里的走廊。程序 p21 显示,从均匀热度出发,格局经历线性放大、非线性变尖、高点合并三个阶段,最终在默认参数下停在 6 个高点,平均间距 7.20 公里;总热度从 80.00 降到 63.67。程序 p22 显示,格宽与步长在合理范围内变化时,高点个数不变、间距变化不到百分之一;而换一组初始噪声,间距可以变一成。那么,究竟是什么决定了间距?为什么同一组方程可以停在不同的间距上?下一章回到色散关系,把这两个问题算清楚。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-niche-brand-outposts-new-city/c02/) - Ch.3 The Dispersion Relation and the Characteristic Wavelength: Spacing Set Jointly by Two Reaches — Chapter 3 treats the dispersion relation: with default parameters the fastest wavelength is 7.13 km and the coefficient of variation of gaps over twenty starts is only 0.073, a layout that is regular but not unique. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-niche-brand-outposts-new-city/c03/) 第3章 色散关系与特征波长:间距由两个半径共同定下 — 第三章讲色散关系:默认参数下最快波长 7.13 公里,二十个起点的间距变异系数只有 0.073,格局规整却不唯一。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-niche-brand-outposts-new-city/c03/) - Ch.4 Estimating the Two Reaches from a Shop's Own Records: Referral Distances and Cannibalisation — Chapter 4 estimates a word-of-mouth reach of 0.551 km and a cannibalisation reach of 2.343 km from referral and opening records, with a one-sided lower bound of 3.863 for their ratio, and gives the data needed for a call. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-niche-brand-outposts-new-city/c04/) 第4章 从门店账本估计两种半径:转介距离与同店蚕食 — 第四章从转介距离与开店记录估出口碑半径 0.551 公里、蚕食半径 2.343 公里,比值单侧下界 3.863,并给出判断所需的数据量。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-niche-brand-outposts-new-city/c04/) - Ch.5 The Diffusion-Ratio Condition for Regional Dominance: Concentrated Share, Ambient Demand and a Drifting Threshold — Chapter 5 shows that past the threshold the hottest fifth of the corridor holds 0.493 of activity, while existing category demand pushes the threshold up to 6.464 and eventually rules out any pattern. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-niche-brand-outposts-new-city/c05/) 第5章 区域垄断的扩散比条件:份额集中、环境需求与门槛漂移 — 第五章说明过门槛后最热的五分之一走廊占 0.493 的热度,而品类既有需求会把门槛推高到 6.464 乃至无法成斑。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-niche-brand-outposts-new-city/c05/) - Ch.6 Seeding the First Outposts: Too Dense Collapses, Too Sparse Does Not Fill Itself — Chapter 6 places shops in the model as fixed seeds: the layout collapses when the gap shrinks to 4.44 km, and gaps that are too wide do not fill themselves without ambient demand. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-niche-brand-outposts-new-city/c06/) 第6章 首批据点的播种:过密会塌缩,过疏不会自补 — 第六章把门店作为固定种子放进模型:间距缩到 4.44 公里时格局塌缩,过疏的空档在没有环境需求时不会自补。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-niche-brand-outposts-new-city/c06/) - Ch.7 Classical Sources East and West: Sunzi on Concentrating and Dividing, and von Thunen's Isolated State — Chapter 7 examines Sunzi's "we concentrate into one, the enemy divides into ten" (a peak is only 2.72 times the even level, far from tenfold) and von Thunen's Isolated State of 1826 (catchment spacing moves with price, Turing spacing does not). (https://5dlibrary.com/answers/en/turing-reaction-diffusion-niche-brand-outposts-new-city/c07/) 第7章 东西方原典考辨:《孙子·虚实》的「我专敌分」与杜能的孤立国 — 第七章考辨《孙子·虚实》的「我专为一,敌分为十」(高点只有均匀水平的 2.72 倍,达不到十倍)与冯·杜能 1826 年的《孤立国》(间距随价格变,图灵间距不随价格变)。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-niche-brand-outposts-new-city/c07/) - Ch.8 Closing the Empirical Loop: Planning the Spacing of First Outposts in a New City, and a Pre-Registered Test — Chapter 8 gives an eight-step planning procedure, the Outpost Spacing Wavelength Calculator and a pre-registered test: the regression of spacing on reaches has power 0.989 with 15 cities. Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/turing-reaction-diffusion-niche-brand-outposts-new-city/c08/) 第8章 实证闭环:新城市首批据点间距规划与预注册检验 — 第八章给出八步规划流程、据点间距波长计算器与预注册检验:间距对半径的回归斜率检验在 15 座城市时功效 0.989。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/turing-reaction-diffusion-niche-brand-outposts-new-city/c08/) ## 5DL-B-248 · Renormalisation Group and Cross-Scale Scaling Laws: Which Conversion Elasticities Survive from a One-Person Workshop to a Hundred-Site Matrix 中文书名:重整化群跨尺度标度律、一人作坊到百站矩阵的转化弹性守恒与经验法则崩塌母典 DOI: pending Cite: Fan, Y. (2026). Renormalisation Group and Cross-Scale Scaling Laws: Which Conversion Elasticities Survive from a One-Person Workshop to a Hundred-Site Matrix (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: The trade assumes that a rule of thumb verified on one site, multiplied by the number of sites, is the law of a hundred-site matrix. This book tests that view with the block transform of the renormalisation group and finds that it holds for one class of metrics only. A metric that follows a power law in the number of sites keeps its conversion elasticity under scale-up; a metric with fixed overhead has an elasticity that rises with scale, so small-scale experience becomes irrelevant; a metric with cross-coupling, such as several sites sharing one keyword pool, has an elasticity that collapses, so single-site test results extrapolated to a hundred sites are often too high. The sources of the founding claim are corrected after checking: the urban-scaling paper gives only three classes of exponents for city indicators, and the shrinkage of trials at scale is cited instead to work on the scalability of experiments and on interference between treatment and control groups. Its site counts, traffic, conversions, costs and owners are illustrative, and all calculations are generated by its programs; it is a general business reference, not advice on any specific business decision, and it covers only compliant matrices of distinct sites run under a real identity. Chapter 1 defines conversion elasticity; in illustration the elasticity of a cost with fixed overhead rises from 0.158 to 0.949, that of coupled orders falls from 0.980 to 0.329, and in a saturated pool a split-test estimate is 25.0 times the effect of a full rollout. Chapter 2 covers intervals, a curvature test and the bias from dropped zeros, which pushes a true exponent of 1 down by 0.135. Chapter 3 performs block transforms on a two-dimensional Ising lattice and a one-dimensional chain and classifies the terms of a business metric as relevant or irrelevant. Chapter 4 shows that revenue elasticity along a funnel is conserved at 0.92 while profit elasticity has no value that can be extrapolated, and that a picked winning trial realises only 0.239 after rollout. Chapter 5 uses finite-size scaling for the threshold of a referral network between sites; after rescaling, the largest difference among sizes falls from 3.949 times to 1.297 times. Chapter 6 covers stratified regression across tiers of a matrix and the gap between a cross-sectional exponent of 1.150 among peers and an own-path exponent of 0.892. Chapter 7 examines the line in the Zhuangzi that from the small one does not see the large whole and from the large one does not see the small clearly, together with Galileo's square-cube law. Chapter 8 gives a six-step quarterly procedure, compliance boundaries, the Scaling Elasticity Tester and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:行内默认的看法是:一个站上验证过的经营口诀,按站数乘上去就是百站矩阵的规律。本书用重整化群的块变换检验这一看法,结论是它只对一类指标成立。随站数呈幂律的指标,转化弹性在放大中守恒;含固定开销的指标,弹性随规模上升,小规模经验逐渐无关;含交叉耦合的指标,例如多个站共用一个关键词池,弹性随规模塌陷,单站试验的结论外推到百站常被高估。立项命题的出处经核对后修正:城市标度论文只给出城市指标的三类指数,「试验放大后缩水」改引关于实验可放大性与处理组对照组干扰的文献。书中站数、流量、转化、成本与经营者均为示意;全部计算由书中程序生成;本书是非个人化经营参考,不构成对任何具体经营决策的建议,只讨论以真实主体运营、内容彼此不同的合规站群。 第一章定义转化弹性,示意中含固定开销的成本弹性从 0.158 升到 0.949,含交叉耦合的订单弹性从 0.980 塌到 0.329,饱和池里拆分试验的估计是真实铺开效果的 25.0 倍。第二章讲指数的区间、曲率检验与丢零偏差,对数直线把真实指数 1 压低 0.135。第三章在二维伊辛格点与一维链上做块变换,并把经营指标各项分为相关项与无关项。第四章说明漏斗里营收弹性守恒在 0.92 而利润弹性没有可外推的值,挑出来的试验赢家铺开后只兑现 0.239。 第五章用有限尺寸标度处理站间互推网络的门槛,重新标度后不同规模间的最大差异从 3.949 倍缩到 1.297 倍。第六章讲站群分层回归,并说明同行截面指数 1.150 与自身路径指数 0.892 的差别。第七章考辨《庄子·秋水》「自细视大者不尽,自大视细者不明」与伽利略的平方立方律。第八章给出季度六步流程、合规边界、规模标度弹性检验器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Frog in the Well: Conversion Elasticity and the Three Fates of Scale-Up — Chapter 1 defines conversion elasticity; in illustration the elasticity of a cost with fixed overhead rises from 0.158 to 0.949, that of coupled orders falls from 0.980 to 0.329, and in a saturated pool a split-test estimate is 25.0 times the effect of a full rollout. (https://5dlibrary.com/answers/en/renormalisation-group-scaling-one-person-to-hundred-sites/c01/) 第1章 井蛙之误:转化弹性的定义与放大的三种命运 — 第一章定义转化弹性,示意中含固定开销的成本弹性从 0.158 升到 0.949,含交叉耦合的订单弹性从 0.980 塌到 0.329,饱和池里拆分试验的估计是真实铺开效果的 25.0 倍。 (https://5dlibrary.com/answers/zh/renormalisation-group-scaling-one-person-to-hundred-sites/c01/) - Ch.2 Estimating a Scaling Exponent from Scattered Points: Intervals, a Curvature Test and the Zeros that Get Dropped — Chapter 2 covers intervals, a curvature test and the bias from dropped zeros, which pushes a true exponent of 1 down by 0.135. (https://5dlibrary.com/answers/en/renormalisation-group-scaling-one-person-to-hundred-sites/c02/) 第2章 从散点估计标度指数:区间、曲率检验与被丢掉的零 — 第二章讲指数的区间、曲率检验与丢零偏差,对数直线把真实指数 1 压低 0.135。 (https://5dlibrary.com/answers/zh/renormalisation-group-scaling-one-person-to-hundred-sites/c02/) - Ch.3 Kadanoff's Block Transform: Which Terms Grow and Which Vanish When Sites Are Merged — The block transform turns scaling up into an operation that can be applied again and again. Program p31 shows on a two-dimensional Ising lattice that two temperatures with similar correlations at the finest scale flow apart after coarse-graining, one to order and one to disorder, with the flip between 2.269 and 2.35. Program p32 shows exactly, on a one-dimensional chain, that an interaction travelling along a line must dissipate at large scale: a neighbour correlation of 0.90 leaves only 0.0012 after six decimations. Program p33 classifies the terms of a business metric by their growth under the block transform: fixed overhead is irrelevant, cross-coupling is relevant, and in the illustrative shop the two cross near 35 sites. The next chapter applies this frame to the most common objects of business, the conversion funnel, profit and the shared keyword pool, and answers the question owners care about most: how much of a lift measured in a test remains after rolling it out to a hundred sites? (https://5dlibrary.com/answers/en/renormalisation-group-scaling-one-person-to-hundred-sites/c03/) 第3章 卡丹诺夫块变换:合并站点时哪些项长大、哪些项消失 — 块变换把「放大」写成一种可以反复施加的运算。程序 p31 在二维伊辛格点上显示,同样在细尺度上相关程度相近的两个温度,粗粒化以后一个流向有序、一个流向无序,流向在 2.269 与 2.35 之间翻转。程序 p32 在一维链上精确地说明,只沿一条线传递的相互作用必然在大尺度上消散,相关 0.90 六次抽取后只剩 0.0012。程序 p33 把经营指标的各项按块变换下的增长率分类:固定开销是无关项,交叉耦合是相关项,示意门店的两者在 35 个站附近交叉。下一章把这个框架用到最常见的经营对象上:转化漏斗、利润和共享的关键词池,并回答一个经营者最关心的问题:试验测到的提升,铺到百站以后还剩多少? (https://5dlibrary.com/answers/zh/renormalisation-group-scaling-one-person-to-hundred-sites/c03/) - Ch.4 Conservation along the Funnel and the Collapse of Profit: Saturating Keyword Pools and Shrinking Pilot Lifts — Chapter 4 shows that revenue elasticity along a funnel is conserved at 0.92 while profit elasticity has no value that can be extrapolated, and that a picked winning trial realises only 0.239 after rollout. (https://5dlibrary.com/answers/en/renormalisation-group-scaling-one-person-to-hundred-sites/c04/) 第4章 漏斗守恒与利润崩塌:共享词池的饱和与试点提升的缩水 — 第四章说明漏斗里营收弹性守恒在 0.92 而利润弹性没有可外推的值,挑出来的试验赢家铺开后只兑现 0.239。 (https://5dlibrary.com/answers/zh/renormalisation-group-scaling-one-person-to-hundred-sites/c04/) - Ch.5 Finite-Size Scaling: How a Small Matrix Smears the Threshold of a Referral Network — Chapter 5 uses finite-size scaling for the threshold of a referral network between sites; after rescaling, the largest difference among sizes falls from 3.949 times to 1.297 times. (https://5dlibrary.com/answers/en/renormalisation-group-scaling-one-person-to-hundred-sites/c05/) 第5章 有限尺寸标度:互推网络的门槛在小站群里被抹平 — 第五章用有限尺寸标度处理站间互推网络的门槛,重新标度后不同规模间的最大差异从 3.949 倍缩到 1.297 倍。 (https://5dlibrary.com/answers/zh/renormalisation-group-scaling-one-person-to-hundred-sites/c05/) - Ch.6 Stratified Regression across Tiers: How One Straight Line Hides a Drifting Elasticity — Chapter 6 covers stratified regression across tiers of a matrix and the gap between a cross-sectional exponent of 1.150 among peers and an own-path exponent of 0.892. (https://5dlibrary.com/answers/en/renormalisation-group-scaling-one-person-to-hundred-sites/c06/) 第6章 站群分层回归:一条直线怎样藏起弹性漂移 — 第六章讲站群分层回归,并说明同行截面指数 1.150 与自身路径指数 0.892 的差别。 (https://5dlibrary.com/answers/zh/renormalisation-group-scaling-one-person-to-hundred-sites/c06/) - Ch.7 Classical Sources East and West: the Frog in the Well of the Zhuangzi and Galileo's Square-Cube Law — The line in "Autumn Floods" about looking at the large from the small and the small from the large splits the limits of scale into two directions, which match the relevant and irrelevant terms of the block transform. Program p72 shows that all four forward and backward extrapolations fail and that none of the prediction intervals covers the truth: extrapolating cost from the small window predicts 36494 yuan against a true 158000 yuan. Galileo's square-cube law is the earliest example of a relevant term: in program p71, enlarging length three times requires width to grow 5.196 times to keep stress unchanged. Both books were right that enlarging is not copying; Zhuangzi's relativism and Galileo's upper limit are revised by today's scaling calculations into the statement that the relation between scales can be computed. The last chapter turns the whole book into a procedure an owner can follow every quarter: how to collect the data, how to report the exponents, and which result would overturn this book? (https://5dlibrary.com/answers/en/renormalisation-group-scaling-one-person-to-hundred-sites/c07/) 第7章 东西方原典考辨:《庄子·秋水》的井蛙与伽利略的平方立方律 — 《秋水》的「自细视大者不尽,自大视细者不明」把尺度局限分成两个方向,恰好对应块变换中的相关项与无关项。程序 p72 显示,四种外推与倒推全部失手,而且预测区间都没有覆盖真值:小窗口外推成本预测 36494 元,真实 158000 元。伽利略的平方立方律是相关项的最早例子:程序 p71 中长度放大 3 倍,保持应力不变要求宽度放大 5.196 倍。两部书都说对了「放大不是复制」;庄子的相对主义与伽利略的上限论,则被今天的标度计算改写为「不同尺度之间的关系是可以计算的」。最后一章把全书变成一个经营者每个季度都能照做的流程:数据怎样收,指数怎样报,什么结果会推翻这本书? (https://5dlibrary.com/answers/zh/renormalisation-group-scaling-one-person-to-hundred-sites/c07/) - Ch.8 Closing the Loop: Stratified Regression by Matrix Size, a Quarterly Report on Scaling Exponents and a Falsification Plan — Chapter 8 gives a six-step quarterly procedure, compliance boundaries, the Scaling Elasticity Tester and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/renormalisation-group-scaling-one-person-to-hundred-sites/c08/) 第8章 实证闭环:站群规模分层回归、标度指数季报与证伪方案 — 第八章给出季度六步流程、合规边界、规模标度弹性检验器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/renormalisation-group-scaling-one-person-to-hundred-sites/c08/) ## 5DL-B-249 · Doob's Martingales and the Optional Stopping Theorem: the Doubling Fallacy and Rigorous Stop-Loss Bounds for Business Trials 中文书名:杜布鞅论可选停时定理、加倍追注谬误与商业试错见好就收停损界母典 DOI: pending Cite: Fan, Y. (2026). Doob's Martingales and the Optional Stopping Theorem: the Doubling Fallacy and Rigorous Stop-Loss Bounds for Business Trials (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs. It is not individual financial advice, does not encourage gambling of any kind, and analyses doubling only as a counter-example. Abstract: People in business often trust two rules of thumb: double after a loss, and quit while ahead. This book tests them with the mathematics of martingales and stopping times. The optional stopping theorem says that as long as the moment of stopping does not peek into the future and has an upper limit, a break-even venture has an expected result at stopping equal to its start; with illustrative parameters the rule of stopping the first time ahead wins with probability 0.824, yet its expectation is exactly zero. The doubling ladder ties many small wins to one large loss, and its probability of doubling capital is only 0.3681, below the one half of level stakes. After checking the sources the book corrects its founding proposition: the result is owed to Ville, Halmos and Doob together; the theorem holds only under conditions such as a bounded stopping time; and in business trials of unknown quality a stop line can raise the average result, but the gain comes from telling good from bad, not from stopping as such. Its parameters are illustrative; all calculations are generated by its programs. It is not individual financial advice, does not encourage gambling of any kind, and analyses doubling only as a counter-example. Chapter 1 shows exact conservation of the expectation under six stopping rules and marks the theorem's boundary. Chapter 2 prices the doubling ladder and shows that doubling after a loss on projects of unknown quality has expectation minus 6.805. Chapter 3 covers Wald's identity: a stop line sells time, and a log must be read by the pooled ratio. Chapter 4 shows that quitting while ahead changes only the shape, and that selling winners early while riding losers has expectation minus 0.270. Chapter 5 shows that a stop line is a line of belief whose value rises with the gap between good and bad, capturing at most 0.646 of the value of foresight. Chapter 6 shows that peeking a hundred times in an online experiment raises false alarms to 0.3726, and presents test martingales and sequential tests. Chapter 7 examines Laozi's Chapter 44 on knowing what is enough and when to stop, and Bernoulli's logarithmic utility of 1738. Chapter 8 gives the three-line register of a trial budget, the Trial Stop-Loss Bound Calculator and a pre-registered protocol (601 trials per group). Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:做生意的人常信两条经验:亏了就加倍,见好就收。本书用鞅与停时的数学检验它们。可选停时定理说,只要停止时刻不偷看未来且有上限,不赚不亏的事情在停下时的期望结果就等于起点;示意参数下,「第一次领先就停」的胜率是 0.824,期望却精确为零。加倍阶梯把许多次小赢与一次大亏捆在一起,靠它把资本翻倍的概率只有 0.3681,低于平注的一半。核对原典后,本书修正了立项命题:这一结论的出处是维尔、哈尔莫斯与杜布三方;定理只在停时有界等条件下成立;在质量未知的商业试错中,止损线能提高平均结果,但收益来自对好坏的区分,而不来自停止本身。书中参数均为示意;全部计算由书中程序生成。本书属非个人化理财建议,不鼓励任何赌博,加倍追注只作反例分析。 第一章精确展示六条停止规则下期望守恒,并说明定理的边界。第二章精算加倍阶梯,并显示在好坏未知的项目上「亏了加倍」的期望是负 6.805。第三章讲瓦尔德恒等式:止损线卖掉的是时间,台账要用合并比率读漂移。第四章讲「见好就收」只改形状,「早卖赢家、死扛输家」的期望是负 0.270。 第五章讲止损线就是信念线,其价值随好坏间距上升,最多可捕获先知价值的 0.646。第六章讲在线试验里偷看一百次时误报率升到 0.3726,以及检验鞅与序贯检验。第七章考辨《老子》第四十四章「知足不辱,知止不殆」与伯努利 1738 年的对数效用。第八章给出试错预算三线登记表、小工具「试错停损界计算器」与预注册方案(每组需 601 项试错)。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A Conservation Law for Fair Games: Martingales, Bounded Stopping Times and the Optional Stopping Theorem — A martingale is a cumulative process that neither gains nor loses; a stopping time is a rule that looks only at the past; the optional stopping theorem says that under a bounded stopping time the expectation of a martingale is conserved. Program p11 displays this conservation exactly on six rules, whose win rates range from 0.300 to 0.824 while every expectation is zero. Program p12 shows that the theorem fails only when the stopping time is unbounded, and that the price of the failure is a loss on a vanishing set of paths and an average wait that both grow without limit, which cannot be collected within finite capital and finite time. What, then, are the accounts of the most common and most tempting rule of all, doubling after a loss, when capital is finite? That is the question of the next chapter. (https://5dlibrary.com/answers/en/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c01/) 第1章 公平博弈的守恒律:鞅、有界停时与可选停时定理 — 鞅是不赚不亏的累计过程;停时是只看过去的停止规则;可选停时定理说,有界停时下鞅的期望守恒。程序 p11 在六条规则上精确展示了这一守恒,胜率从 0.300 到 0.824 不等,期望全部是零。程序 p12 显示,定理只在停时无界时失效,而失效的代价是极少数路径的亏损与平均等待同时趋于无穷,这在有限资本与有限时间里不可能兑现。那么,最常见、最有诱惑力的那条规则,也就是「亏了就加倍」,在有限资本下究竟是什么样的一笔账?这是下一章的问题。 (https://5dlibrary.com/answers/zh/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c01/) - Ch.2 The Exact Accounts of the Doubling Ladder: Finite Capital, Tail Losses and Doubling After a Loss — Chapter 2 prices the doubling ladder and shows that doubling after a loss on projects of unknown quality has expectation minus 6.805. (https://5dlibrary.com/answers/en/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c02/) 第2章 加倍阶梯的精确账:有限资本、尾部亏损与「亏了加倍」 — 第二章精算加倍阶梯,并显示在好坏未知的项目上「亏了加倍」的期望是负 6.805。 (https://5dlibrary.com/answers/zh/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c02/) - Ch.3 Wald's Identities: a Stop Line Sells Time — Wald's first identity says that the expected result at stopping equals the drift per period times the expected number of periods; programs p31 and p32 confirm it exactly for two lines and for three. Stop lines, profit lines and deadlines can move only time: for a project with success probability 0.45, moving from no stop line to a tight one shrinks the expected loss from 2.469 to 0.905; for a project with success probability 0.55, the same stop line shrinks the expected gain from 2.164 to 1.157. To read a drift from a log, use the pooled ratio rather than the per-trial average, or a losing class of projects will read as a winning one. If stopping rules change only time, what does quitting while ahead change, and why does it make people feel they win more than they lose? That is the question of the next chapter. (https://5dlibrary.com/answers/en/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c03/) 第3章 瓦尔德恒等式:止损线卖掉的是时间 — 瓦尔德第一恒等式说,停下时的期望结果等于每期漂移乘以期望期数;程序 p31 与 p32 在两条线与三条线上精确验证了它。止损线、止盈线与截止日期能动的只有时间:对成功概率 0.45 的项目,止损线从无到紧,把期望亏损从 2.469 缩到 0.905;对成功概率 0.55 的项目,同样的止损线把期望盈利从 2.164 缩到 1.157。从台账里读漂移,要用合并比率而不是逐项平均,否则一个亏损的项目类别会读成盈利。既然停止规则只改变时间,「见好就收」又改变了什么?它为什么让人觉得自己赢多输少?这是下一章的问题。 (https://5dlibrary.com/answers/zh/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c03/) - Ch.4 Quitting While Ahead Changes Only the Shape: the Win-Rate Illusion and Selling Winners While Holding Losers — Chapter 4 shows that quitting while ahead changes only the shape, and that selling winners early while riding losers has expectation minus 0.270. (https://5dlibrary.com/answers/en/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c04/) 第4章 见好就收只改形状:胜率幻觉与「早卖赢家、死扛输家」 — 第四章讲「见好就收」只改形状,「早卖赢家、死扛输家」的期望是负 0.270。 (https://5dlibrary.com/answers/zh/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c04/) - Ch.5 The Learning Value of a Stop Line: Lines of Belief, the Gap Between Good and Bad, and Escalating Sunk Costs — Chapter 5 shows that a stop line is a line of belief whose value rises with the gap between good and bad, capturing at most 0.646 of the value of foresight. (https://5dlibrary.com/answers/en/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c05/) 第5章 止损线的学习价值:信念线、好坏间距与沉没成本的加码 — 第五章讲止损线就是信念线,其价值随好坏间距上升,最多可捕获先知价值的 0.646。 (https://5dlibrary.com/answers/zh/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c05/) - Ch.6 Peeking and Early Stopping: Optional Stopping and Test Martingales in Online Experiments — Chapter 6 shows that peeking a hundred times in an online experiment raises false alarms to 0.3726, and presents test martingales and sequential tests. (https://5dlibrary.com/answers/en/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c06/) 第6章 偷看与早停:在线试验里的可选停止与检验鞅 — 第六章讲在线试验里偷看一百次时误报率升到 0.3726,以及检验鞅与序贯检验。 (https://5dlibrary.com/answers/zh/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c06/) - Ch.7 Eastern and Western Sources Examined: Laozi on Knowing Enough and Knowing When to Stop, and Bernoulli's Logarithm — Chapter 7 examines Laozi's Chapter 44 on knowing what is enough and when to stop, and Bernoulli's logarithmic utility of 1738. (https://5dlibrary.com/answers/en/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c07/) 第7章 东西方原典考辨:《老子》「知足不辱,知止不殆」与伯努利的对数 — 第七章考辨《老子》第四十四章「知足不辱,知止不殆」与伯努利 1738 年的对数效用。 (https://5dlibrary.com/answers/zh/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c07/) - Ch.8 Closing the Loop: the Three-Line Register of a Trial Budget, the Stop-Loss Bound Calculator and a Pre-Registered Protocol — Chapter 8 gives the three-line register of a trial budget, the Trial Stop-Loss Bound Calculator and a pre-registered protocol (601 trials per group). Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c08/) 第8章 实证闭环:试错预算三线登记表、停损界计算器与预注册方案 — 第八章给出试错预算三线登记表、小工具「试错停损界计算器」与预注册方案(每组需 601 项试错)。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/doob-martingale-optional-stopping-doubling-fallacy-stop-loss/c08/) ## 5DL-B-250 · The Information Bottleneck and Rate-Distortion Limits: How Far a Bilingual Monograph Can Be Compressed into Guardrails before Discriminative Power Falls Away 中文书名:信息瓶颈率失真极值、双语专著到全局护栏的极限压缩比与判别力保真母典 DOI: pending Cite: Fan, Y. (2026). The Information Bottleneck and Rate-Distortion Limits: How Far a Bilingual Monograph Can Be Compressed into Guardrails before Discriminative Power Falls Away (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 1 builds an illustrative source of eight kinds of draft and three verdicts and the counting bound: three rules keep 0.960 of the verdict information, two only 0.629. Chapter 4 shows that the critical exchange rate is the reciprocal of the squared maximal correlation, 0.6844 and 1.4610 on the illustrative source while information rises continuously as rules split. Abstract: When a bilingual monograph is finished, its publisher wants to press it into a few guardrail rules. The commissioning claim was that compression has no free lunch, that the number of guardrail rules has a rate-distortion lower bound, and that beyond a certain compression ratio discriminative power falls off a cliff. This book writes candidate rules as a discrete source and reviewers' verdicts as a relevance variable and tests the claim with rate-distortion theory and the information bottleneck, correcting the blueprint. The sources never speak of a cliff; the optimal curve is concave and continuous and grows steeper, with the cost of one bit removed rising from 0.012 to 0.592 on the illustrative source; cliffs appear only when a whole number of rules falls below the number of verdict groups, when the structure of the solution crosses a critical point, or when discrete scores are used; and when reviewers never hesitate the cost is exactly linear. Volume 150 covers the distillation workflow; this volume gives its limits. The parameters are illustrative; all calculations are generated by the book's programs; the book is not individual business advice or a compliance opinion, and any merger or release of a guardrail rule must be reviewed and released by hand by the person in charge. Chapter 1 builds an illustrative source of eight kinds of draft and three verdicts and the counting bound: three rules keep 0.960 of the verdict information, two only 0.629. Chapter 2 quantifies the cliff: the cliff in share, 0.331, is 41.7 times the earlier average fall. Chapter 3 implements Blahut and Arimoto in the standard library, matches the binary closed form digit by digit, and shows that draft-by-draft rules use 0.2620 bits more than the rate-distortion curve. Chapter 4 shows that the critical exchange rate is the reciprocal of the squared maximal correlation, 0.6844 and 1.4610 on the illustrative source while information rises continuously as rules split. Chapter 5 uses the agglomerative bottleneck to press forty-eight candidate rules to a knee of five, keeping 0.958; exact bilingual redundancy is the only free lunch, and translation drift pulls the share to 0.948. Chapter 6 sets out the opposing evidence: with 50 drafts, 0.920 of repeated runs choose the wrong knee, and six rules learnt from few records keep only 0.017 bits. Chapter 7 examines "writing does not exhaust speech, speech does not exhaust intent" from the Xici commentary, Wang Bi's "forget the image once the intent is grasped", and Shannon's decision to set meaning outside the engineering problem: with the same three rules, compressing to reproduce the draft keeps 0.6108 bits and compressing to keep the verdict 0.7234 bits. Chapter 8 gives the loop of one compression-recall curve per volume, the distilled "Rule-Compression Knee Calculator" and a paired pre-registered protocol needing 75 to 103 drafts per test when the gap in accuracy is large. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:一部双语专著写完,出版人总想把它压成几条护栏规则。流行的说法有两种:好书可以无限提炼而不丢东西,或者规则砍掉一半判别力也掉一半。立项命题说压缩没有免费午餐,护栏规则条数有率失真下界,越过某个压缩比判别力断崖式下降。本书把候选规则写成离散信源、把审稿裁决写成相关变量,用香农的率失真函数、布拉赫特与有本的迭代和提斯比等人的信息瓶颈逐条检验,并修正蓝图:原文没有「断崖」之说;最优曲线是凹的、连续的,越压越陡,示意信源上每删一比特的代价从 0.012 升到 0.592;断崖只在整数条数低于需要区分的裁决组数、解的结构跨过临界点、或用离散计分去量时出现;审稿人不犹豫时代价恰是直线。本卷给出压缩极限,蒸馏流程见第一百五十卷。书中参数为示意;全部计算由书中程序生成;本书是非个人化经营建议,不构成法律或内容合规意见,任何护栏规则的删并与上线须由主事者人工复核放行。 第一章建立八类草稿、三种裁决的示意信源与计数下界,三条规则保留 0.960 的裁决信息,两条只剩 0.629。第二章量化断崖:份额断崖 0.331 是此前平均降幅的 41.7 倍,审稿人越犹豫比值越大,准确率断崖却越小。第三章在标准库里实现布拉赫特与有本的算法,与二元信源解析解逐位一致,并显示逐稿规则比率失真曲线多用 0.2620 比特。第四章证明临界汇率等于最大相关平方的倒数,示意信源为 0.6844 与 1.4610,规则数阶梯式分裂而信息量连续。第五章用凝聚式信息瓶颈把四十八条候选规则压到拐点五条,保留 0.958;双语精确冗余是唯一的免费午餐,翻译漂移把它拉到 0.948。第六章陈列反方证据:50 份样本时 0.920 的重复实验选错拐点,从少量记录学出的六条规则只保留 0.017 比特。 第七章考辨《周易·系辞上》「书不尽言,言不尽意」、王弼「得意忘象」与香农把语义放在工程问题之外的立场:同样三条规则,按复现草稿压缩保留 0.6108 比特,按保留裁决压缩保留 0.7234 比特。第八章给出每卷一条压缩召回曲线的闭环、蒸馏小工具「规则压缩拐点计算器」与配对预注册方案,准确率差较大时每次检验 75 到 103 份稿件。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Guardrail Source: Eight Kinds of Draft, Three Verdicts and the Counting Bound on Rules — Chapter 1 builds an illustrative source of eight kinds of draft and three verdicts and the counting bound: three rules keep 0.960 of the verdict information, two only 0.629. (https://5dlibrary.com/answers/en/information-bottleneck-rate-distortion-monograph-to-guardrails/c01/) 第1章 护栏信源:八类草稿、三种裁决与规则条数的计数下界 — 第一章建立八类草稿、三种裁决的示意信源与计数下界,三条规则保留 0.960 的裁决信息,两条只剩 0.629。 (https://5dlibrary.com/answers/zh/information-bottleneck-rate-distortion-monograph-to-guardrails/c01/) - Ch.2 The Axis: Cliffs, Concave Curves and the Straight Line under Deterministic Labels — Chapter 2 quantifies the cliff: the cliff in share, 0.331, is 41.7 times the earlier average fall. (https://5dlibrary.com/answers/en/information-bottleneck-rate-distortion-monograph-to-guardrails/c02/) 第2章 主轴:断崖、凹曲线与确定性标签下的直线 — 第二章量化断崖:份额断崖 0.331 是此前平均降幅的 41.7 倍,审稿人越犹豫比值越大,准确率断崖却越小。 (https://5dlibrary.com/answers/zh/information-bottleneck-rate-distortion-monograph-to-guardrails/c02/) - Ch.3 Blahut and Arimoto: Computing the Rate-Distortion Function of a Guardrail in the Standard Library — Chapter 3 implements Blahut and Arimoto in the standard library, matches the binary closed form digit by digit, and shows that draft-by-draft rules use 0.2620 bits more than the rate-distortion curve. (https://5dlibrary.com/answers/en/information-bottleneck-rate-distortion-monograph-to-guardrails/c03/) 第3章 布拉赫特与有本:用标准库算护栏的率失真函数 — 第三章在标准库里实现布拉赫特与有本的算法,与二元信源解析解逐位一致,并显示逐稿规则比率失真曲线多用 0.2620 比特。 (https://5dlibrary.com/answers/zh/information-bottleneck-rate-distortion-monograph-to-guardrails/c03/) - Ch.4 Phase Transitions: Maximal Correlation, the Critical Exchange Rate and Deterministic Annealing — Chapter 4 shows that the critical exchange rate is the reciprocal of the squared maximal correlation, 0.6844 and 1.4610 on the illustrative source while information rises continuously as rules split. (https://5dlibrary.com/answers/en/information-bottleneck-rate-distortion-monograph-to-guardrails/c04/) 第4章 相变:最大相关、临界汇率与确定性退火 — 第四章证明临界汇率等于最大相关平方的倒数,示意信源为 0.6844 与 1.4610,规则数阶梯式分裂而信息量连续。 (https://5dlibrary.com/answers/zh/information-bottleneck-rate-distortion-monograph-to-guardrails/c04/) - Ch.5 From Monograph to Guardrails: Agglomerative Bottlenecks, Five Knee Criteria and Bilingual Redundancy — Chapter 5 uses the agglomerative bottleneck to press forty-eight candidate rules to a knee of five, keeping 0.958; exact bilingual redundancy is the only free lunch, and translation drift pulls the share to 0.948. (https://5dlibrary.com/answers/en/information-bottleneck-rate-distortion-monograph-to-guardrails/c05/) 第5章 从专著到护栏:凝聚式瓶颈、五种拐点判法与双语冗余 — 第五章用凝聚式信息瓶颈把四十八条候选规则压到拐点五条,保留 0.958;双语精确冗余是唯一的免费午餐,翻译漂移把它拉到 0.948。 (https://5dlibrary.com/answers/zh/information-bottleneck-rate-distortion-monograph-to-guardrails/c05/) - Ch.6 Opposing Evidence: Finite Samples, Learnt Compression and How Discriminative Power Is Scored — Chapter 6 sets out the opposing evidence: with 50 drafts, 0.920 of repeated runs choose the wrong knee, and six rules learnt from few records keep only 0.017 bits. (https://5dlibrary.com/answers/en/information-bottleneck-rate-distortion-monograph-to-guardrails/c06/) 第6章 反方证据:有限样本、学来的压缩与判别力的计分方式 — 第六章陈列反方证据:50 份样本时 0.920 的重复实验选错拐点,从少量记录学出的六条规则只保留 0.017 比特。 (https://5dlibrary.com/answers/zh/information-bottleneck-rate-distortion-monograph-to-guardrails/c06/) - Ch.7 Classical Sources East and West: "Writing Does Not Exhaust Speech, Speech Does Not Exhaust Intent" and Shannon's Suspension of Meaning — Chapter 7 examines "writing does not exhaust speech, speech does not exhaust intent" from the Xici commentary, Wang Bi's "forget the image once the intent is grasped", and Shannon's decision to set meaning outside the engineering problem: with the same three rules, compressing to reproduce the draft keeps 0.6108 bits and compressing to keep the verdict 0.7234 bits. (https://5dlibrary.com/answers/en/information-bottleneck-rate-distortion-monograph-to-guardrails/c07/) 第7章 东西方原典考辨:「书不尽言,言不尽意」与香农的语义悬置 — 第七章考辨《周易·系辞上》「书不尽言,言不尽意」、王弼「得意忘象」与香农把语义放在工程问题之外的立场:同样三条规则,按复现草稿压缩保留 0.6108 比特,按保留裁决压缩保留 0.7234 比特。 (https://5dlibrary.com/answers/zh/information-bottleneck-rate-distortion-monograph-to-guardrails/c07/) - Ch.8 Closing the Loop: One Compression-Recall Curve per Volume, a Knee Calculator and a Pre-Registered Protocol — Chapter 8 gives the loop of one compression-recall curve per volume, the distilled "Rule-Compression Knee Calculator" and a paired pre-registered protocol needing 75 to 103 drafts per test when the gap in accuracy is large. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/information-bottleneck-rate-distortion-monograph-to-guardrails/c08/) 第8章 实证闭环:每卷一条压缩召回曲线、拐点小工具与预注册方案 — 第八章给出每卷一条压缩召回曲线的闭环、蒸馏小工具「规则压缩拐点计算器」与配对预注册方案,准确率差较大时每次检验 75 到 103 份稿件。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/information-bottleneck-rate-distortion-monograph-to-guardrails/c08/) ## 5DL-B-251 · Lotka-Volterra Competition Phase Portraits: Cannibalisation within Multi-Site, Multi-Account Matrices versus Symbiotic Barriers 中文书名:洛特卡沃尔泰拉竞争相图、多子站多账号矩阵内耗与共生壁垒判定母典 DOI: pending Cite: Fan, Y. (2026). Lotka-Volterra Competition Phase Portraits: Cannibalisation within Multi-Site, Multi-Account Matrices versus Symbiotic Barriers (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Sites, accounts, visitors, costs and owners are illustrative, and all calculations are generated by the book's programs. It is not individual business advice and covers only compliant brand matrices run under a real identity with public links between sites, never sockpuppet accounts or bought traffic. Abstract: The trade assumes that one more site or one more brand account takes one more slice of the market, so that the larger the matrix, the larger its total traffic. This book tests that view with the competition equations of Lotka and Volterra, treating sites as species, audiences as food and the monthly visitors a site could hold alone as its carrying capacity. It finds that whether an extra site takes more of the market depends on the competition coefficients between sites and on the ratio of their capacities. The founding claim is corrected after checking: with coefficients above one, only an interior equilibrium propped up by the owner falls below a single site, while free competition ends with one site left; and a product of coefficients below one is merely necessary for coexistence, since on an illustrative grid a share of 0.627 of the cells that satisfy it still end with one site out. Sites, accounts, visitors, costs and owners are illustrative, and all calculations are generated by the book's programs. It is not individual business advice and covers only compliant brand matrices run under a real identity with public links between sites, never sockpuppet accounts or bought traffic. Chapter 1 draws the four outcomes for two sites and shows that with running costs at a quarter, a second site pays only when the coefficient is below 0.600. Chapter 2 estimates coefficients from audience profiles with the overlap formula of MacArthur and Levins and shows that a sub-site cut from the main site's core makes the main site feel a pressure of 1.230 and is still pushed out. Chapter 3 covers Jacobian eigenvalues, the advantage of an early start in the bistable case and critical slowing down; at a coefficient of 0.80 the split of traffic needs about 30.2 quarters to return to equilibrium. Chapter 4 extends the model to many sites; the best number of sites is roughly the reciprocal of the coefficient, and of five unequal sites free competition leaves two, with 13126 visitors between them. Chapter 5 covers symbiosis through disclosed cross-referral and the entry barrier it raises; here the product rule is exactly the condition for bounded traffic. Chapter 6 shows that passive watching can barely read the coefficients while a pause test can, and that pairwise outcomes predict three-site outcomes with an accuracy of 0.934. Chapter 7 examines the Zhuangzi's line on stranded fish that moisten each other, set against forgetting each other in rivers and lakes, together with the struggle for existence in the third chapter of Darwin's Origin of Species. Chapter 8 gives a quarterly audit of audience overlap, the Site Matrix Competition Portrait and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:行内默认的看法是:多开一个站、多注册一个品牌账号,就多吃一块市场,矩阵越大总流量越大。本书用洛特卡与沃尔泰拉的竞争方程检验这一看法,把站当作物种、受众当作食物、单站可留住的月访客当作承载量。结论是:多开站是否多吃市场,取决于站与站之间的竞争系数与承载量之比。立项命题经核对后修正:系数大于一时,只有被经营者撑住的内部平衡点低于单站,自由竞争的结局是只剩一个站;两个系数之积小于一只是共存的必要条件,示意网格中满足它的格子仍有 0.627 以一方出局告终。书中站、账号、访客、成本与经营者均为示意;全部计算由书中程序生成;本书是非个人化经营建议,只讨论以真实主体运营、彼此关系公开的合规品牌矩阵,不涉及马甲账号与刷量。 第一章画出两站的四种结局,并说明运营成本占四分之一时,第二个站只有在系数低于 0.600 时才划算。第二章用麦克阿瑟与莱文斯的重叠公式从受众画像估系数,说明切走主站核心人群的子站会让主站感受到 1.230 的压力而自己仍被挤出。第三章讲雅可比特征值、双稳情形的先发优势与临界放慢,系数为 0.80 时流量分配回到平衡约需 30.2 个季度。第四章推广到多站,最优站数大致是系数的倒数,五个不等的站在自由竞争下只留下两个,合计 13126 名访客。 第五章讲公开互导形成的共生与进入门槛,乘积条件在这里恰好是总流量有界的条件。第六章说明被动观察几乎读不出系数,暂停试验能读出,两两结局预测三站结局的准确率为 0.934。第七章考辨《庄子·大宗师》「相濡以沫,不如相忘于江湖」与达尔文《物种起源》第三章的生存斗争。第八章给出季度受众重叠审计、站群内耗相图判定器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Competition Portrait: Four Outcomes for Two Sites and a Corrected Founding Claim — Chapter 1 draws the four outcomes for two sites and shows that with running costs at a quarter, a second site pays only when the coefficient is below 0.600. (https://5dlibrary.com/answers/en/lotka-volterra-competition-multi-site-matrix-cannibalisation/c01/) 第1章 竞争相图:两个站的四种结局与立项命题的修正 — 第一章画出两站的四种结局,并说明运营成本占四分之一时,第二个站只有在系数低于 0.600 时才划算。 (https://5dlibrary.com/answers/zh/lotka-volterra-competition-multi-site-matrix-cannibalisation/c01/) - Ch.2 Competition Coefficients from Audience Overlap: the Overlap Formula, Sub-Sites Cut from the Core, and Quarterly Sample Sizes — Chapter 2 estimates coefficients from audience profiles with the overlap formula of MacArthur and Levins and shows that a sub-site cut from the main site's core makes the main site feel a pressure of 1.230 and is still pushed out. (https://5dlibrary.com/answers/en/lotka-volterra-competition-multi-site-matrix-cannibalisation/c02/) 第2章 受众重叠估竞争系数:重叠公式、切走核心人群的子站与季度样本量 — 第二章用麦克阿瑟与莱文斯的重叠公式从受众画像估系数,说明切走主站核心人群的子站会让主站感受到 1.230 的压力而自己仍被挤出。 (https://5dlibrary.com/answers/zh/lotka-volterra-competition-multi-site-matrix-cannibalisation/c02/) - Ch.3 Stability of Equilibria: Jacobian Eigenvalues, the Advantage of an Early Start, and Critical Slowing Down — Chapter 3 covers Jacobian eigenvalues, the advantage of an early start in the bistable case and critical slowing down; at a coefficient of 0.80 the split of traffic needs about 30.2 quarters to return to equilibrium. (https://5dlibrary.com/answers/en/lotka-volterra-competition-multi-site-matrix-cannibalisation/c03/) 第3章 平衡点的稳定性:雅可比特征值、先发优势与临界放慢 — 第三章讲雅可比特征值、双稳情形的先发优势与临界放慢,系数为 0.80 时流量分配回到平衡约需 30.2 个季度。 (https://5dlibrary.com/answers/zh/lotka-volterra-competition-multi-site-matrix-cannibalisation/c03/) - Ch.4 From Two Sites to Many: the Traffic Ceiling, the Best Number of Sites, and Pruning a Five-Site Matrix — Chapter 4 extends the model to many sites; the best number of sites is roughly the reciprocal of the coefficient, and of five unequal sites free competition leaves two, with 13126 visitors between them. (https://5dlibrary.com/answers/en/lotka-volterra-competition-multi-site-matrix-cannibalisation/c04/) 第4章 从两站到多站:流量天花板、最优站数与五站矩阵的裁剪 — 第四章推广到多站,最优站数大致是系数的倒数,五个不等的站在自由竞争下只留下两个,合计 13126 名访客。 (https://5dlibrary.com/answers/zh/lotka-volterra-competition-multi-site-matrix-cannibalisation/c04/) - Ch.5 Symbiotic Barriers: Disclosed Cross-Referral, Negative Coefficients and the Entry Threshold for Outside Rivals — Coefficients can be negative: disclosed cross-referral can make one site's growth lift another. Program p51 shows that in the symbiotic case a product of coefficients below one is exactly the condition for bounded traffic, so the product rule of the founding claim finds its proper place here; it also shows that the linear model exaggerates to absurdity once negative coefficients grow, reaching 20.000 times one site at a product of 0.81, so it should be used only when they are small. Program p52 shows that the entry threshold is proportional to the total traffic the matrix keeps, and that a symbiotic pair's threshold of 12500 exceeds the 10526 of four competing sites. Sign and size both have to be measured, yet the chapters so far have taken the coefficients as known. The next chapter faces the hardest question: with only a few years of monthly traffic, can the coefficients be read at all? (https://5dlibrary.com/answers/en/lotka-volterra-competition-multi-site-matrix-cannibalisation/c05/) 第5章 共生壁垒:公开互导、负系数与外来竞争者的进入门槛 — 系数可以为负:公开的互导使一个站的访客增长反而带动另一个站。程序 p51 说明,在共生情形里两个系数之积小于一正是总流量有界的条件,立项命题里的乘积条件在这里找到了准确的位置;同时线性模型在负系数稍大时会夸大到荒唐,乘积 0.81 时总流量已是 20.000 倍,所以只宜在负系数绝对值不大时使用。程序 p52 说明,进入门槛与矩阵留住的总流量成正比,两个共生站的门槛 12500 高于四个竞争站的 10526。系数的正负与大小都靠测量,而前面几章一直假定系数已知。下一章面对最难的问题:只有几年的月度流量,系数到底能不能读出来? (https://5dlibrary.com/answers/zh/lotka-volterra-competition-multi-site-matrix-cannibalisation/c05/) - Ch.6 Reading Coefficients from Traffic Series: Shared Swings, Pause Tests and the Accuracy of Pairwise Verdicts — Chapter 6 shows that passive watching can barely read the coefficients while a pause test can, and that pairwise outcomes predict three-site outcomes with an accuracy of 0.934. (https://5dlibrary.com/answers/en/lotka-volterra-competition-multi-site-matrix-cannibalisation/c06/) 第6章 从流量序列读系数:共同波动、暂停试验与两两判别的准确率 — 第六章说明被动观察几乎读不出系数,暂停试验能读出,两两结局预测三站结局的准确率为 0.934。 (https://5dlibrary.com/answers/zh/lotka-volterra-competition-multi-site-matrix-cannibalisation/c06/) - Ch.7 Classical Sources East and West: the Stranded Fish of the Zhuangzi and Darwin's Struggle for Existence — The fish in the dry spring of "The Great and Venerable Teacher" become, in program p71, a survival table of four arrangements: competition in the pool fails at a critical capacity of 5400, mutual referral holds out to 2400, and the open waters are unaffected. Zhuangzi was right about the value of leaving a shrinking pool, underrated mutual help when there is nowhere to go, and left out the cost of moving. Darwin's struggle for existence becomes, in program p72, a curve of the coefficient falling steadily with distance between positions. He was right that the similar struggle hardest and that each species is limited first by itself, but he did not reach the ratio of capacities: against a rival twice as strong, a site must sit at least 2.498 segments away to coexist. Both books point to the same business move: keep positions apart, and move away when the pool shrinks. The next chapter packs the first seven chapters into a quarterly routine: what to measure each quarter, how to decide, and what results would overturn this book? (https://5dlibrary.com/answers/en/lotka-volterra-competition-multi-site-matrix-cannibalisation/c07/) 第7章 东西方原典考辨:《庄子·大宗师》的涸泉之鱼与达尔文的生存斗争 — 《大宗师》的涸泉之鱼,在程序 p71 里成了一张四种安排的生存表:同池竞争在临界承载量 5400 时撑不住,互导可以撑到 2400,各奔江湖不受影响。庄子说对了离开萎缩池子的价值,低估了互助在无处可去时的作用,也省掉了迁移的成本。达尔文的生存斗争在程序 p72 里成了系数随定位距离单调下降的曲线;他说对了相似者斗争最烈,以及每个物种首先受自身限制,但没有说到承载量之比,对手强一倍时至少要偏离 2.498 个细分才能共存。两部书都指向同一个经营动作:定位要错开,萎缩时要迁走。下一章把前七章装进一个季度流程:每季量什么,怎么判,什么结果会推翻本书? (https://5dlibrary.com/answers/zh/lotka-volterra-competition-multi-site-matrix-cannibalisation/c07/) - Ch.8 Closing the Loop: a Quarterly Audit of Audience Overlap, Merge and Split Decisions, and a Falsification Protocol — Chapter 8 gives a quarterly audit of audience overlap, the Site Matrix Competition Portrait and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/lotka-volterra-competition-multi-site-matrix-cannibalisation/c08/) 第8章 实证闭环:季度受众重叠审计、合并与拆分判定与证伪方案 — 第八章给出季度受众重叠审计、站群内耗相图判定器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/lotka-volterra-competition-multi-site-matrix-cannibalisation/c08/) ## 5DL-B-252 · Boolean Satisfiability Phase Transitions: the Critical Constraint Ratio at Which Stacked Promotion, Compliance and Gate Rules Deadlock, and How to See It Coming 中文书名:布尔可满足性相变阈值、促销合规与门禁叠加的约束比临界与全盘死锁预警母典 DOI: pending Cite: Fan, Y. (2026). Boolean Satisfiability Phase Transitions: the Critical Constraint Ratio at Which Stacked Promotion, Compliance and Gate Rules Deadlock, and How to See It Coming (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: The rulebooks and data are illustrative; every calculation is generated by the book's programs; the book offers general business method, not legal, compliance or business-decision advice. Chapter 4 checks an illustrative rulebook of 24 switches: bare ratio 0.958, load 0.475, a median clause ratio at deadlock of only 1.292, and a minimal conflict core spanning three rule families. Abstract: A business team's rulebook only grows: coupons never stack, member prices exclude flash sales, the price floor must hold, a countdown needs a real deadline, contrast and touch targets must meet their minimums, messages need consent. The common belief is that more rules make the system tighter, and gradually so. This book tests that belief with the phase transition of Boolean satisfiability: with switches written as Boolean variables and rules as clauses, a deadlocked rulebook is an unsatisfiable formula, and the satisfiable share of random 3-SAT plunges from near one to near zero within a narrow band of the clause-to-variable ratio. All calculations use only the Python standard library. The rulebooks and data are illustrative; every calculation is generated by the book's programs; the book offers general business method, not legal, compliance or business-decision advice. "Gate" means only the stacking of constraints in a rule engine, with no judgement of fortune of any kind. Chapter 1 states and quantifies the claim and corrects the blueprint: 4.26 comes from the cavity-method value of 4.267 of 2002, not from the 1992 experiments (about 4.3) or the 1994 extrapolation (about 4.17) that the blueprint cites, and only the bounds 3.52 and 4.4898 have been proved; with 50 switches the half-way point is 4.33. Chapter 2 writes DPLL from scratch and reproduces "easy, hard, easy". Chapter 3 gives the first-moment bound of 5.191, the exact threshold of 2-SAT and the role of rule length. Chapter 4 checks an illustrative rulebook of 24 switches: bare ratio 0.958, load 0.475, a median clause ratio at deadlock of only 1.292, and a minimal conflict core spanning three rule families. Chapter 5 shows that short rules govern the cliff and that deadlock needs demands and prohibitions together, the most dangerous balance bringing the clause ratio at deadlock down to 2.638. Chapter 6 proves that the hazard of the next random rule depends only on the size of the backbone, and shows the backbone jumping from 6 to 23 while the campaign count falls smoothly. Chapter 7 examines the spear and shield of the Han Feizi and Boole's Laws of Thought of 1854. Chapter 8 gives an eight-step monthly scan, the Rule Constraint-Ratio Deadlock Early Warning and a pre-registered test: grading can be tested with power 0.851 using 28 snapshots per group. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:经营团队的规则册只增不减:券不能叠用、会员价与秒杀互斥、底价不得击穿、倒计时要有真实截止、对比度与点按尺寸要达标、私讯要先取得同意。常识以为规则越多系统越紧,而且是渐进地紧。本书用布尔可满足性的相变检验这一常识:把开关写成布尔变量、把规则写成子句,规则册的死锁就是公式不可满足;随机 3-SAT 的可满足比例在子句与变量之比的一个窄区间里从接近一骤降到接近零。全部计算只用 Python 标准库。书中规则册与数据均为示意;全部计算由书中程序生成;本书属非个人化经营方法,不构成法律、合规或经营决策意见。「门禁」只指规则引擎里约束的叠加,不作任何吉凶推断。 第一章正面陈述并量化命题,并修正立项蓝图:4.26 出自 2002 年空腔方法的 4.267,而非蓝图所引的 1992 年实验(约 4.3)或 1994 年外推(约 4.17),且至今只严格证明了 3.52 与 4.4898 两个界;开关数 50 时一半位置在 4.33。第二章从零写出 DPLL 并复现「易、难、易」。第三章给出一阶矩上界 5.191、2-SAT 的精确阈值与规则长度的作用。第四章给一本 24 个开关的示意规则册做体检:裸约束比 0.958,载荷 0.475,死锁时约束比中位数只有 1.292,最小冲突核横跨三个规则族。 第五章说明短规则主导悬崖,死锁需要「要求」与「禁止」同在,最危险的配比使死锁约束比低到 2.638。第六章证明下一条随机规则的危险率只取决于骨干大小,并显示方案数平滑下降时骨干从 6 跳到 23。第七章考辨《韩非子·难一》的矛与楯与布尔 1854 年的《思维规律》。第八章给出八步月度扫描、规则约束比死锁预警器与预注册检验:分级检验每组 28 个快照即有 0.851 的功效。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Cliff of Stacked Rules: the Satisfiability Phase Transition of Random 3-SAT — Chapter 1 states and quantifies the claim and corrects the blueprint: 4.26 comes from the cavity-method value of 4.267 of 2002, not from the 1992 experiments (about 4.3) or the 1994 extrapolation (about 4.17) that the blueprint cites, and only the bounds 3.52 and 4.4898 have been proved; with 50 switches the half-way point is 4.33. (https://5dlibrary.com/answers/en/boolean-sat-phase-transition-stacked-rules-deadlock/c01/) 第1章 规则叠加的悬崖:随机 3-SAT 的可满足性相变 — 第一章正面陈述并量化命题,并修正立项蓝图:4.26 出自 2002 年空腔方法的 4.267,而非蓝图所引的 1992 年实验(约 4.3)或 1994 年外推(约 4.17),且至今只严格证明了 3.52 与 4.4898 两个界;开关数 50 时一半位置在 4.33。 (https://5dlibrary.com/answers/zh/boolean-sat-phase-transition-stacked-rules-deadlock/c01/) - Ch.2 DPLL with the Standard Library: Unit Propagation, Branching and "Easy, Hard, Easy" — DPLL decides whether a rulebook has a solution by simplification, unit propagation and branching with backtracking, and the steps of unit propagation are themselves a readable account of a deadlock. p21 shows a rulebook of only eight switches whose ratio after one new rule in week 2 is just 1.375, yet seven steps of unit propagation prove deadlock, because single-switch hard requirements pin the other rules down. p22 reproduces "easy, hard, easy": the median number of dead ends climbs from 0.0 to 26.5 and falls back to 9.0. p23 shows that unit propagation cuts the work by two orders of magnitude. But where exactly is the cliff, and can a bound be computed without experiments? The next chapter answers with the expected number of solutions, the exact threshold of 2-SAT and the length of clauses. (https://5dlibrary.com/answers/en/boolean-sat-phase-transition-stacked-rules-deadlock/c02/) 第2章 标准库 DPLL:单元传播、分支与「易、难、易」 — DPLL 用赋值化简、单元传播与分支回溯判定规则册有没有解,单元传播推出的步骤本身就是一份可读的死锁说明。p21 显示,一本只有八个开关的规则册,第二周加一条规则后约束比只是 1.375,单元传播七步就证明了死锁,因为单开关硬规定把其余规则钉死了。p22 复现了「易、难、易」:死路数的中位数从 0.0 升到 26.5 再回落到 9.0。p23 显示单元传播把工作量压低两个数量级。可是,悬崖究竟在哪里,能不能不做实验就算出一个界?下一章用解数的期望、2-SAT 的精确阈值与子句长度给出答案。 (https://5dlibrary.com/answers/zh/boolean-sat-phase-transition-stacked-rules-deadlock/c02/) - Ch.3 Where the Cliff Lies: the First-Moment Bound, the Exact Threshold of 2-SAT and Rule Length — Chapter 3 gives the first-moment bound of 5.191, the exact threshold of 2-SAT and the role of rule length. (https://5dlibrary.com/answers/en/boolean-sat-phase-transition-stacked-rules-deadlock/c03/) 第3章 悬崖在哪里:一阶矩上界、2-SAT 的精确阈值与规则长度 — 第三章给出一阶矩上界 5.191、2-SAT 的精确阈值与规则长度的作用。 (https://5dlibrary.com/answers/zh/boolean-sat-phase-transition-stacked-rules-deadlock/c03/) - Ch.4 Measuring a Rulebook: Load, Forced Values, Minimal Conflict Cores and Headroom — Chapter 4 checks an illustrative rulebook of 24 switches: bare ratio 0.958, load 0.475, a median clause ratio at deadlock of only 1.292, and a minimal conflict core spanning three rule families. (https://5dlibrary.com/answers/en/boolean-sat-phase-transition-stacked-rules-deadlock/c04/) 第4章 给规则册量约束比:载荷、强制值、最小冲突核与余量 — 第四章给一本 24 个开关的示意规则册做体检:裸约束比 0.958,载荷 0.475,死锁时约束比中位数只有 1.292,最小冲突核横跨三个规则族。 (https://5dlibrary.com/answers/zh/boolean-sat-phase-transition-stacked-rules-deadlock/c04/) - Ch.5 Real Rules Are Not Random Rules: Mixed Lengths, the 2+p Model and the Balance of Demands and Prohibitions — Chapter 5 shows that short rules govern the cliff and that deadlock needs demands and prohibitions together, the most dangerous balance bringing the clause ratio at deadlock down to 2.638. (https://5dlibrary.com/answers/en/boolean-sat-phase-transition-stacked-rules-deadlock/c05/) 第5章 真规则不是随机规则:长短混合、2+p 模型与「要求」和「禁止」的配比 — 第五章说明短规则主导悬崖,死锁需要「要求」与「禁止」同在,最危险的配比使死锁约束比低到 2.638。 (https://5dlibrary.com/answers/zh/boolean-sat-phase-transition-stacked-rules-deadlock/c05/) - Ch.6 Signals on the Eve of Deadlock: Solution Counts, the Backbone and the Hazard of the Next Rule — Chapter 6 proves that the hazard of the next random rule depends only on the size of the backbone, and shows the backbone jumping from 6 to 23 while the campaign count falls smoothly. (https://5dlibrary.com/answers/en/boolean-sat-phase-transition-stacked-rules-deadlock/c06/) 第6章 死锁前夜的信号:解数、骨干与下一条规则的危险率 — 第六章证明下一条随机规则的危险率只取决于骨干大小,并显示方案数平滑下降时骨干从 6 跳到 23。 (https://5dlibrary.com/answers/zh/boolean-sat-phase-transition-stacked-rules-deadlock/c06/) - Ch.7 Classical Sources East and West: the Spear and Shield of the Han Feizi, and Boole's x² = x — Chapter 7 examines the spear and shield of the Han Feizi and Boole's Laws of Thought of 1854. (https://5dlibrary.com/answers/en/boolean-sat-phase-transition-stacked-rules-deadlock/c07/) 第7章 东西方原典考辨:《韩非子·难一》的矛与楯,布尔《思维规律》的 x² = x — 第七章考辨《韩非子·难一》的矛与楯与布尔 1854 年的《思维规律》。 (https://5dlibrary.com/answers/zh/boolean-sat-phase-transition-stacked-rules-deadlock/c07/) - Ch.8 Closing the Loop: a Monthly Constraint-Ratio Scan of the Five-Kit Rulebooks and a Pre-Registered Test — Chapter 8 gives an eight-step monthly scan, the Rule Constraint-Ratio Deadlock Early Warning and a pre-registered test: grading can be tested with power 0.851 using 28 snapshots per group. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/boolean-sat-phase-transition-stacked-rules-deadlock/c08/) 第8章 实证闭环:五件套规则库的月度约束比扫描与预注册检验 — 第八章给出八步月度扫描、规则约束比死锁预警器与预注册检验:分级检验每组 28 个快照即有 0.851 的功效。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/boolean-sat-phase-transition-stacked-rules-deadlock/c08/) ## 5DL-B-253 · Constitutional Engineering for a One-Commander Silicon Empire: Serial Decision Bandwidth and Breaking Amdahl's Bottleneck 中文书名:单兵硅基帝国宪章学、串行总司令决策带宽与阿姆达尔串行瓶颈破局母典 DOI: pending Cite: Fan, Y. (2026). Constitutional Engineering for a One-Commander Silicon Empire: Serial Decision Bandwidth and Breaking Amdahl's Bottleneck (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: With illustrative parameters, cutting the serial share from a fifth to a twentieth raises the speedup by 3.325, while raising the agents from ten to a hundred adds only 1.236. Its parameters are illustrative; all calculations are generated by its programs. Abstract: Now that agents are cheap, a popular belief holds that one person who opens more parallel windows multiplies output with the number of windows. This book tests that belief with Amdahl's law and queueing theory. As long as every item of output returns to the same commander for approval, the serial share he does by hand caps the speed of the whole. With illustrative parameters, cutting the serial share from a fifth to a twentieth raises the speedup by 3.325, while raising the agents from ten to a hundred adds only 1.236. After checking the sources the book corrects its founding blueprint: Amdahl's paper of 1967 writes no formula, and treating approval as the serial part is this book's modelling assumption; the line "the ruler holds the name, the minister renders the form" comes from the chapter "Yang Quan" of the Han Feizi, not "Zhu Dao". Its parameters are illustrative; all calculations are generated by its programs. It is not individual business advice. Chapter 1 computes Amdahl's ceiling and the founding claim exactly. Chapter 2 shows why Gustafson's law cannot get round per-item approval and reads the serial share back with the Karp and Flatt metric. Chapter 3 treats the commander as a single server; in the finite-source model lowering the serial share releases 9.667 items an hour, while adding agents only adds backlog. Chapter 4 uses Harris's square-root formula to find an optimal approval window of 58.55 minutes, lowering the serial share from 0.250 to 0.112. Chapter 5 covers automatic gates: reviewing about a fifth lifts the ceiling to 21.75, and coherency overhead between windows makes capacity peak at 21.21 windows. Chapter 6 treats a written constitution as a decision cache; at a nine-in-ten hit rate ten agents reach a speedup of 8.696. Chapter 7 examines Han Fei's tally of name and form and Smith's pin factory (of the team gain of 2,400, heads give only ten). Chapter 8 gives a weekly decision log (in an invented example the serial share falls from 0.2430 to 0.0841), the Serial-Share Speedup Ceiling Calculator and a pre-registered crossover design (simulated power 0.8262). Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:智能体便宜以后,流行的说法是:一个人只要多开并行窗口,产出就随窗口数成倍增长。本书用阿姆达尔定律与排队论检验这个说法。只要每份产出都要回到同一位主事者那里审批,他亲手做的那部分所占的串行份额就给整体速度设了上限。示意参数下,把串行份额从两成降到半成使加速比增加 3.325,把智能体从十个加到一百个只增加 1.236。核对原典后,本书修正了立项蓝图:阿姆达尔 1967 年的原文没有写出公式,把审批看成串行部分是本书的模型假设;「君操其名,臣效其形」出自《韩非子·扬权》而非《主道》。书中参数均为示意;全部计算由书中程序生成。本书属非个人化经营建议。 第一章精确计算阿姆达尔天花板与立项命题。第二章说明古斯塔夫森定律为何绕不开逐项审批,并用卡普—弗拉特指标反推串行份额。第三章把主事者看成单服务台,有限源模型中降低串行份额使每小时放行 9.667 份,加智能体则只增加积压。第四章用哈里斯平方根公式求出最优审批窗口 58.55 分钟,串行份额从 0.250 降到 0.112。 第五章讲自动闸门,送审约两成时天花板升到 21.75 倍,窗口间相干开销使容量在 21.21 个窗口处见顶。第六章把成文宪章看成决策缓存,命中九成时十个智能体的加速比为 8.696。第七章考辨《韩非子》形名参同与斯密制针工场(团队增益 2,400 倍中人头只占十倍)。第八章给出每周决策日志(虚构示例中串行份额从 0.2430 降到 0.0841)、小工具「串行份额加速上限计算器」与预注册交叉设计(模拟检验力 0.8262)。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Ceiling of the Serial Share: Amdahl's Law and the Exact Account of Twenty Percent against Five — The serial share is the proportion of the work that only the commander can do, in sequence; Amdahl's law says the parallel speedup is capped by its reciprocal. Program p11 shows that at a serial share of 0.20 even a thousand agents reach only 4.980; program p12 shows that cutting the serial share from 0.20 to 0.05 gains 3.325, nearly three times what going from ten agents to a hundred gains; program p13 shows that the curve has already reached half its ceiling at $(1-s)/s$ agents. Yet Amdahl assumed a fixed amount of work. If the work grows as the agents multiply, does the ceiling still stand? That is the question of the next chapter. (https://5dlibrary.com/answers/en/one-commander-silicon-empire-amdahl-serial-bottleneck/c01/) 第1章 串行份额的天花板:阿姆达尔定律与「两成对半成」的精确账 — 串行份额是只能由主事者本人按顺序完成的那部分工作所占的比例;阿姆达尔定律说,并行加速比的上限是它的倒数。程序 p11 显示,串行份额为 0.20 时一千个智能体也只到 4.980 倍;程序 p12 显示,把串行份额从 0.20 降到 0.05 带来的增量是 3.325,接近把智能体从十个加到一百个的三倍;程序 p13 显示,曲线在 $(1-s)/s$ 个智能体处就已经到了天花板的一半。可是,阿姆达尔假设工作总量固定不变。如果智能体多了,工作量也跟着长大,天花板还在吗?这是下一章的问题。 (https://5dlibrary.com/answers/zh/one-commander-silicon-empire-amdahl-serial-bottleneck/c01/) - Ch.2 Does the Problem Grow? Gustafson's Law, Per-Item Review and the Measured Serial Share — Gustafson's law says that when the problem grows with the number of processors, the share of the serial part is diluted and the speedup can approach proportional growth; program p21 shows that on 1,024 processors a serial share of 0.004 on the parallel machine corresponds to a scaled speedup of 1019.9. Program p22 shows that when every item takes 12 minutes of the commander's time, a growing problem does not help, and the speedup stops at 5.000 from four agents onwards. Program p23 shows that the Karp and Flatt metric separates a fixed serial share from overhead that grows with scale. Yet $\min(N/a,1/r)$ is only an optimistic bound: items from agents do not arrive evenly, and the commander's review times are sometimes short and sometimes long. How long a queue does randomness build at the commander's door? That is the question of the next chapter. (https://5dlibrary.com/answers/en/one-commander-silicon-empire-amdahl-serial-bottleneck/c02/) 第2章 问题会长大吗:古斯塔夫森定律、逐项审批与实测串行份额 — 古斯塔夫森定律说,问题随处理器数目长大时,串行部分的份额被摊薄,加速比可以接近线性;程序 p21 显示,1,024 个处理器上 0.004 的并行机串行份额对应 1019.9 倍的规模扩展加速比。程序 p22 显示,当每一份产出都要占用主事者 12 分钟时,问题长大也无济于事,加速比在四个智能体处就停在 5.000 倍。程序 p23 显示,卡普—弗拉特指标可以把固定的串行份额与随规模增长的开销区分开。可是 $\min(N/a,1/r)$ 只是一个乐观的上界:智能体交来的产出并不均匀,主事者的审查时间也有长有短。随机性会让主事者门前排起多长的队?这是下一章的问题。 (https://5dlibrary.com/answers/zh/one-commander-silicon-empire-amdahl-serial-bottleneck/c02/) - Ch.3 The Commander Is a Single Server: Queueing Theory, the Finite-Source Model and Little's Law — Chapter 3 treats the commander as a single server; in the finite-source model lowering the serial share releases 9.667 items an hour, while adding agents only adds backlog. (https://5dlibrary.com/answers/en/one-commander-silicon-empire-amdahl-serial-bottleneck/c03/) 第3章 主事者是一张单服务台:排队论、有限源模型与利特尔定律 — 第三章把主事者看成单服务台,有限源模型中降低串行份额使每小时放行 9.667 份,加智能体则只增加积压。 (https://5dlibrary.com/answers/zh/one-commander-silicon-empire-amdahl-serial-bottleneck/c03/) - Ch.4 Batch Approval Windows: Harris's Square-Root Formula and the Price of Waiting — Chapter 4 uses Harris's square-root formula to find an optimal approval window of 58.55 minutes, lowering the serial share from 0.250 to 0.112. (https://5dlibrary.com/answers/en/one-commander-silicon-empire-amdahl-serial-bottleneck/c04/) 第4章 批量审批窗口:哈里斯平方根公式与等待的代价 — 第四章用哈里斯平方根公式求出最优审批窗口 58.55 分钟,串行份额从 0.250 降到 0.112。 (https://5dlibrary.com/answers/zh/one-commander-silicon-empire-amdahl-serial-bottleneck/c04/) - Ch.5 Send Up Only the Exceptions: Automatic Gates, Random Audits and Coordination Overhead — Chapter 5 covers automatic gates: reviewing about a fifth lifts the ceiling to 21.75, and coherency overhead between windows makes capacity peak at 21.21 windows. (https://5dlibrary.com/answers/en/one-commander-silicon-empire-amdahl-serial-bottleneck/c05/) 第5章 只把例外送上来:自动闸门、随机抽检与协调开销 — 第五章讲自动闸门,送审约两成时天花板升到 21.75 倍,窗口间相干开销使容量在 21.21 个窗口处见顶。 (https://5dlibrary.com/answers/zh/one-commander-silicon-empire-amdahl-serial-bottleneck/c05/) - Ch.6 Decision Bandwidth: a Written Constitution as a Cache, a Faster Serial Part and the Span of Control — Chapter 6 treats a written constitution as a decision cache; at a nine-in-ten hit rate ten agents reach a speedup of 8.696. (https://5dlibrary.com/answers/en/one-commander-silicon-empire-amdahl-serial-bottleneck/c06/) 第6章 决策带宽:成文宪章作缓存、更快的串行部分与管理幅度 — 第六章把成文宪章看成决策缓存,命中九成时十个智能体的加速比为 8.696。 (https://5dlibrary.com/answers/zh/one-commander-silicon-empire-amdahl-serial-bottleneck/c06/) - Ch.7 Eastern and Western Sources Examined: Han Fei's Tally of Name and Form and Smith's Pin Factory — Chapter 7 examines Han Fei's tally of name and form and Smith's pin factory (of the team gain of 2,400, heads give only ten). (https://5dlibrary.com/answers/en/one-commander-silicon-empire-amdahl-serial-bottleneck/c07/) 第7章 东西方原典考辨:《韩非子》形名参同与斯密制针工场 — 第七章考辨《韩非子》形名参同与斯密制针工场(团队增益 2,400 倍中人头只占十倍)。 (https://5dlibrary.com/answers/zh/one-commander-silicon-empire-amdahl-serial-bottleneck/c07/) - Ch.8 Closing the Loop: a Weekly Decision Log, Batch Approval Windows and a Pre-Registered Test — Chapter 8 gives a weekly decision log (in an invented example the serial share falls from 0.2430 to 0.0841), the Serial-Share Speedup Ceiling Calculator and a pre-registered crossover design (simulated power 0.8262). Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/one-commander-silicon-empire-amdahl-serial-bottleneck/c08/) 第8章 实证闭环:每周决策日志、批量审批窗口与预注册证伪方案 — 第八章给出每周决策日志(虚构示例中串行份额从 0.2430 降到 0.0841)、小工具「串行份额加速上限计算器」与预注册交叉设计(模拟检验力 0.8262)。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/one-commander-silicon-empire-amdahl-serial-bottleneck/c08/) ## 5DL-B-254 · Atomic File-Lock Protocols for Multiple Windows: Self-Recovering Orphan Locks after Power Loss and Cross-Window Hand-Over State Machines 中文书名:多窗口文件原子锁并发协议、断电孤儿锁自愈与跨窗口交接状态机母典 DOI: pending Cite: Fan, Y. (2026). Atomic File-Lock Protocols for Multiple Windows: Self-Recovering Orphan Locks after Power Loss and Cross-Window Hand-Over State Machines (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: When several writing windows claim volume numbers at once, the commonest code checks whether a lock exists, creates it if not, and adds a timeout said to repair the lock after a power cut. The commissioning claim was that check-then-create must have a race window, that only atomic exclusive creation guarantees mutual exclusion, and that a self-repairing lease without a monotonic clock and fencing tokens produces two holders. This book writes every step of a protocol as an atomic operation, simulates the file system in memory, enumerates every interleaving and corrects the blueprint in four places. "Must" speaks of the space of schedules: two windows give a double claim in 12 of 16 schedules, while the frequency on the wall clock depends on whether starts cluster, 0.02966 per batch when four windows rush in within 2 seconds. Dijkstra proved in 1965 that reads and writes alone suffice for exclusion; the limit is that the roster must be closed. Fencing tokens come from Burrows's sequencers and Kleppmann's account, not from Gray and Cheriton. A monotonic clock cannot stop process pauses: with perfect clocks 0.000976 of writes still land late. The parameters are illustrative; all calculations are generated by the book's programs, which read and write no file; the book is non-individual business advice, and any change to a lock protocol in use and any ruling on an orphan lock must be reviewed and released by hand by the person in charge. Chapter 1 enumerates check-then-create and converts the race to the wall clock. Chapter 2 model-checks six entry protocols built from reads and writes, reproduces the shortest counterexample to Hyman's flawed solution of 1966 and shows that the roster is the weak point of read/write algorithms. Chapter 3 treats power loss: flushing cannot remove empty locks, writing a temporary file and linking it can, and "delete empty locks" kills the living. Chapter 4 treats leases, clocks and fencing tokens. Chapter 5 treats the judgement of orphans: an hour of silence on a laptop says little about death, and once automatic stealing is allowed even release gives double claims, 12 of 80 schedules, a share of 0.150. Chapter 6 treats hand-over between windows and a status that only moves forward. Chapter 7 examines "he keeps and issues the keys, to open and close the gates of the state" from the Office of the Gates in the Rites of Zhou and Dijkstra's four conditions of 1965. Chapter 8 gives a formal health check of the lock protocol this series uses: none of its schedules gives a double claim as written, while an automatic cleaner gives double claims in 45 of 48 final states. It presents the distilled "Lock-Protocol Race Enumerator" and a pre-registered test; showing a double-claim rate below one in a thousand with no failures takes 2995 batches. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:同时开多个写作窗口去领卷号,最常见的写法是先看锁在不在、不在就建,再给锁加一个超时,号称断电后能自愈。立项命题说:先检查后创建必有竞态窗口,只有原子独占创建才保证互斥;租约自愈若不配单调计时与围栏令牌,会出现双持锁。本书把协议的每一步写成原子操作,在内存里模拟文件系统,穷举全部交错调度,并修正蓝图四处:「必有」说的是调度空间,两窗口 16 种调度有 12 种双领,墙钟上的频率取决于出发是否扎堆,四个窗口 2 秒内一拥而上时每批 0.02966;迪杰斯特拉 1965 年恰恰证明只靠读写也能互斥,限制在于名册必须封闭;围栏令牌出自伯罗斯的序列号与克莱普曼的整理,不出自格雷与切里顿;单调计时挡不住进程停顿,时钟完美时仍有 0.000976 的写入迟到。书中参数为示意;全部计算由书中程序生成;程序不读写任何文件;本书是非个人化经营建议,任何锁协议的上线变更与孤儿锁裁决须由主事者人工复核放行。 第一章穷举先检查后创建并把竞态换算到墙钟。第二章用模型检验器检验六个只靠读写的入口协议,复现海曼 1966 年错误解法的最短反例,并说明名册是读写算法的命门。第三章讨论断电:刷盘消不掉空锁,先写临时再链接可以;「空锁即删」会误杀活人。第四章讨论租约、时钟与围栏令牌。第五章讨论孤儿锁判别:笔记本沉默一小时主人已死的后验很低,一旦允许自动抢锁,连释放都会双领,80 种调度有 12 种,份额 0.150。第六章讨论跨窗口交接与只进不退的状态。 第七章考辨《周礼·地官司徒·司门》「掌授管键,以启闭国门」与迪杰斯特拉 1965 年的四个条件。第八章对本百卷正在使用的锁协议做形式化体检:原样零双领,加自动清理者则 48 个终止状态有 45 个双领;给出蒸馏小工具「锁协议竞态穷举器」与预注册方案,零失败证明双领率低于千分之一需 2995 批。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Race Window: Exhaustive Interleavings of Check-then-Create and Their Width on the Wall Clock — Program p11 counted every schedule: with two windows using check-then-create, 12 of 16 schedules give a double claim; with four windows, 74880 of 74976. Atomic exclusive creation has no bad schedule at any window count. Program p12 converted the gap to the wall clock: collisions are rare when starts are spread out, and about 0.02966 per batch when windows rush in together. Yet this chapter has only answered which primitive to use. A more fundamental question remains: if the file system offered no exclusive creation and only atomic reads and writes, could several windows still exclude each other? Dijkstra answered yes in 1965. The next chapter checks his answer with a small model checker and explains why it cannot be used for many windows. (https://5dlibrary.com/answers/en/atomic-file-lock-multi-window-orphan-lock-handover/c01/) 第1章 竞态窗口:先检查后创建的交错穷举与墙钟宽度 — 程序 p11 数清了全部调度:两个窗口用先检查后创建,16 种调度里有 12 种双领;四个窗口时 74976 种里有 74880 种。原子独占创建在任何窗口数下都没有坏调度。程序 p12 把缝换算到墙钟上:分散出发时撞车罕见,一拥而上时每批约 0.02966。可是本章只回答了「用什么原语」,还没有回答一个更根本的问题:如果文件系统不提供独占创建,只给原子的读和写,多个窗口还能不能互斥?迪杰斯特拉在 1965 年给出了肯定的回答,下一章用一个小小的模型检验器来复核它,并说明这个回答为什么在多窗口场景里用不上。 (https://5dlibrary.com/answers/zh/atomic-file-lock-multi-window-orphan-lock-handover/c01/) - Ch.2 Mutual Exclusion from Reads and Writes Alone: Model-Checking Six Entry Protocols and the Roster Bound — Chapter 2 model-checks six entry protocols built from reads and writes, reproduces the shortest counterexample to Hyman's flawed solution of 1966 and shows that the roster is the weak point of read/write algorithms. (https://5dlibrary.com/answers/en/atomic-file-lock-multi-window-orphan-lock-handover/c02/) 第2章 只靠读写能否互斥:六个入口协议的模型检验与名册下界 — 第二章用模型检验器检验六个只靠读写的入口协议,复现海曼 1966 年错误解法的最短反例,并说明名册是读写算法的命门。 (https://5dlibrary.com/answers/zh/atomic-file-lock-multi-window-orphan-lock-handover/c02/) - Ch.3 After the Power Cut: Empty Locks, Missing Status and Write-Temporary-then-Link — Program p31 shows that when a claim is cut short by a power failure, the original method can leave 7 disk states in the lagging model, 3 of them empty locks; flushing reduces empty locks to 1 kind but cannot remove it; write-temporary-then-link brings them to zero. Program p32 shows that the cleaning rule "delete empty locks" gives a double claim in 10 of 25 schedules, and looking twice before deleting only lowers the share from 0.400 to 0.250. The best treatment of the half-finished remains of a power cut is to keep them from ever appearing in a living window. There is, however, a harder case: a window that has not lost power but has merely paused and will wake up later. Leases were designed for this case. Are they good enough for many windows? (https://5dlibrary.com/answers/en/atomic-file-lock-multi-window-orphan-lock-handover/c03/) 第3章 断电之后:空锁、缺状态与先写临时再链接 — 程序 p31 显示,一次领题被断电打断后,原样写法在滞后模型下可能留下 7 种磁盘状态,其中 3 种是空锁;刷盘把空锁减到 1 种,却消不掉;先写临时再链接把空锁减到零。程序 p32 显示,「空锁即删」的清理规则在 25 种调度里有 10 种造成双领,看两次再删也只把份额从 0.400 降到 0.250。断电留下的半成品,最好的处理是让它不出现在活人身上。可是还有一类更难的情形:窗口没有断电,只是停住了,过一会儿又醒来。租约正是为这类情形设计的,它在多窗口场景里够不够用? (https://5dlibrary.com/answers/zh/atomic-file-lock-multi-window-orphan-lock-handover/c03/) - Ch.4 Leases, Clocks and Fencing Tokens — Program p41 showed that steps of a single shared clock create no overlap, while mixing a monotonic clock with the wall clock does: a 5-second step gives 3.000 seconds of overlap, and a 3600-second step gives 58.000 seconds. Program p42 showed that with perfect clocks a 60-second lease with a 2-second margin still lets 0.000976 of writes land late, almost all because of sleep. Program p43 showed that the 380 late writes all overwrite new data under blind overwriting and never do when the storage checks tokens. A lease answers the question of what to do when the owner dies, but its answer about whether the owner is dead is a guess based on time. How costly is a wrong guess, and how long must a silence last before the owner can be declared dead? The next chapter does that reckoning. (https://5dlibrary.com/answers/en/atomic-file-lock-multi-window-orphan-lock-handover/c04/) 第4章 租约、时钟与围栏令牌 — 程序 p41 显示,同一个钟的跳变不会制造重叠,混用单调钟与墙钟才会:墙钟跳 5 秒造成 3.000 秒重叠,跳 3600 秒造成 58.000 秒重叠。程序 p42 显示,时钟完美时,60 秒租约留 2 秒余量仍有 0.000976 的写入迟到,几乎全部来自休眠。程序 p43 显示,迟到的 380 次写入在盲写下全部覆盖新数据,在存储端核对令牌时一次也不会。租约解决的是「主人死了怎么办」,可它对主人的死活作出的是一个基于时间的猜测。猜错的代价有多大?沉默多久才可以认定主人已死?下一章算这笔账。 (https://5dlibrary.com/answers/zh/atomic-file-lock-multi-window-orphan-lock-handover/c04/) - Ch.5 Judging Orphan Locks: How Long a Silence Means Death — Chapter 5 treats the judgement of orphans: an hour of silence on a laptop says little about death, and once automatic stealing is allowed even release gives double claims, 12 of 80 schedules, a share of 0.150. (https://5dlibrary.com/answers/en/atomic-file-lock-multi-window-orphan-lock-handover/c05/) 第5章 孤儿锁判别:沉默多久才算死 — 第五章讨论孤儿锁判别:笔记本沉默一小时主人已死的后验很低,一旦允许自动抢锁,连释放都会双领,80 种调度有 12 种,份额 0.150。 (https://5dlibrary.com/answers/zh/atomic-file-lock-multi-window-orphan-lock-handover/c05/) - Ch.6 State Machines for Hand-Over between Windows: Transfer, Tokens and a Status That Only Moves Forward — Program p61 showed that hand-over by delete-then-create is overtaken by a third party in 40 of 54 schedules; atomic replacement removes the queue-jumping, yet 0.250 of schedules are still damaged by the departing window's late write; adding a token brings both loss and damage to zero. Program p62 showed that a blindly written status file moves backwards in 4 of 6 schedules, that reading before writing still moves backwards in 0.333 of them, and that one exclusive mark per stage never does. We have now tested the lock protocol of many-window collaboration, from claiming, power cuts, leases and orphan locks to hand-over, with the mathematics of the second half of the twentieth century. Yet exclusion and hand-over are not problems that began with computers. How were the keys of a city gate kept more than two thousand years ago? And what exactly does Dijkstra's single page say, and which of its statements can still be tested today? The next chapter returns to the classics. (https://5dlibrary.com/answers/en/atomic-file-lock-multi-window-orphan-lock-handover/c06/) 第6章 跨窗口交接状态机:让渡、令牌与只进不退的状态 — 程序 p61 显示,先删后建的交接在 54 种调度里有 40 种被第三方插队,原子替换消除了插队,却仍有 0.250 的调度被交出方的迟到写入损坏;加上令牌之后丢失与损坏都是零。程序 p62 显示,盲写的状态文件在 6 种调度里有 4 种倒退,先读后写仍有 0.333 倒退,每个阶段一个独占标记则为零。至此,从领题、断电、租约、孤儿锁到交接,我们已经用二十世纪下半叶的数学把多窗口协作的锁协议检验了一遍。可是独占与交接并不是计算机时代才有的问题。两千多年前的城门怎样管钥匙?迪杰斯特拉那一页纸的原文究竟说了什么,哪些说法今天仍可检验?下一章回到原典。 (https://5dlibrary.com/answers/zh/atomic-file-lock-multi-window-orphan-lock-handover/c06/) - Ch.7 Classical Sources East and West: the Gatekeeper Who "Keeps and Issues the Keys" and Dijkstra in 1965 — Chapter 7 examines "he keeps and issues the keys, to open and close the gates of the state" from the Office of the Gates in the Rites of Zhou and Dijkstra's four conditions of 1965. (https://5dlibrary.com/answers/en/atomic-file-lock-multi-window-orphan-lock-handover/c07/) 第7章 东西方原典考辨:司门「掌授管键」与迪杰斯特拉 1965 — 第七章考辨《周礼·地官司徒·司门》「掌授管键,以启闭国门」与迪杰斯特拉 1965 年的四个条件。 (https://5dlibrary.com/answers/zh/atomic-file-lock-multi-window-orphan-lock-handover/c07/) - Ch.8 Closing the Loop: a Formal Health Check of This Series' Lock Protocol and a Pre-Registered Test — Chapter 8 gives a formal health check of the lock protocol this series uses: none of its schedules gives a double claim as written, while an automatic cleaner gives double claims in 45 of 48 final states. It presents the distilled "Lock-Protocol Race Enumerator" and a pre-registered test; showing a double-claim rate below one in a thousand with no failures takes 2995 batches. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/atomic-file-lock-multi-window-orphan-lock-handover/c08/) 第8章 实证闭环:本百卷锁协议的形式化体检与预注册方案 — 第八章对本百卷正在使用的锁协议做形式化体检:原样零双领,加自动清理者则 48 个终止状态有 45 个双领;给出蒸馏小工具「锁协议竞态穷举器」与预注册方案,零失败证明双领率低于千分之一需 2995 批。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/atomic-file-lock-multi-window-orphan-lock-handover/c08/) ## 5DL-B-255 · Mid-Context Forgetting in Long-Context Attention: Positional Bias and the Threshold for Lossless Segment-Summary Relays 中文书名:长上下文注意力中段遗忘曲线、位置偏置与零损耗分段摘要接力阈值母典 DOI: pending Cite: Fan, Y. (2026). Mid-Context Forgetting in Long-Context Attention: Positional Bias and the Threshold for Lossless Segment-Summary Relays (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Tasks, material and parameters are illustrative and the model constants come from one 2023 model; all calculations are generated by the book's programs, which call no model. It is not individual business advice and neither rates nor recommends any model or service. Abstract: Two views circulate. One says that whatever fits in the window gets read, so long material can simply be stuffed in; the other says that models reading long inputs forget the middle, so material should be cut into segments, summarised and relayed layer by layer. This book tests the founding claim of the second view: recall in the middle of a long context is markedly lower than at the ends, tracing a U (Liu et al., Transactions of the Association for Computational Linguistics, 2024), so a relay beats stuffing the window, while relay depth is bounded by the product of per-layer retention rates. Checking the original corrects the claim. The gap of 0.220 between beginning and middle is robust, but only two of the paper's four models draw a U, a middle below the closed-book score appears in one model only with a confidence interval spanning zero, and the paper never tested a relay, so the "therefore" is an inference. Tasks, material and parameters are illustrative and the model constants come from one 2023 model; all calculations are generated by the book's programs, which call no model. It is not individual business advice and neither rates nor recommends any model or service. Chapters 1 and 2 re-examine the original numbers, effect sizes and dose-response curve, and set out dissenting results on other models, extended windows, calibration and training, and single-needle tests. Chapter 3 builds a two-exponential recall model, a primacy bump and a recency bump over a floor that sinks with length, and extrapolates a capture curve. Chapter 4 computes the product of relay retention rates, shows that a running summary turns the U into a recency slope, and finds detailed facts keeping only 0.4305 after eight layers. Chapter 5 solves for the best segment length and the relay threshold: for the 2023 reader each summary call must keep key facts above about 0.911 and segments must be short, and once the reader's loss falls below 0.413 of the original curve a relay at fidelity 0.97 loses to the whole window. Chapter 6 shows that zero loss can come only from a key-fact restatement slot copied by template and verified by checksum, which lifts key-fact survival from 0.601 to above nine in ten but holds only about seven lines. Chapter 7 examines Mencius on one day of sun and ten days of cold and Ebbinghaus on memory. Chapter 8 gives a relay template for long tasks, the Segment Relay Threshold Calculator and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:流行的说法有两种:一种认为窗口装得下就读得到,长材料一次塞进去即可;另一种认为模型读长文会「中段遗忘」,所以应当分段写摘要、层层接力。本书检验后一种说法的立项命题:中段召回显著低于首尾、呈 U 形(Liu 等 2024 年刊于《计算语言学协会汇刊》),因此整窗不如接力,而接力层数受每层保留率连乘约束。核对原文后命题被修正:开头与中间之差 0.220 的效应稳固,但原文四个模型只有两个呈 U 形,「中段低于闭卷」只见于一个模型且置信区间跨零;原论文没有测试接力,「因此」只是推论。书中任务、材料与参数均为示意,模型常数来自 2023 年的一个模型;全部计算由书中程序生成,不调用任何模型接口;本书是非个人化经营建议,不评测、不推荐任何模型或服务。 第一、二章复核原始数字、效应量与剂量曲线,并陈不同模型、扩展窗口、校准与训练研究及单针测试的反方结果。第三章建立首因峰、近因峰加随长度下沉的地板的双指数召回模型,并外推出捕获曲线。第四章计算接力保留率连乘,说明滚动摘要把 U 形变成近因斜坡,细节类事实八层后只剩 0.4305。第五章求最优段长与接力阈值:对 2023 年的读者,每次摘要调用的保真度须高于约 0.911,段长须很短;读者损失低于原曲线的 0.413 倍时,保真度 0.97 的接力就输给整窗。 第六章说明零损耗只能来自模板复制加校验码的关键事实复述位,它把关键事实的存活从 0.601 提到九成以上,但只容得下约七行。第七章考辨《孟子·告子上》「一日暴之,十日寒之」与艾宾浩斯《论记忆》。第八章给出长任务接力摘要模板、分段接力阈值计算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Collapse in the Middle: Original Numbers, Effect Sizes and Dose for the U-Shaped Position Curve — Program p11 fits GPT-3.5-Turbo's scores at twenty documents with an upward-opening parabola 17.09 points deep. Program p12 gives the confidence interval of the 0.220 gap between beginning and middle, shows that "below closed-book" straddles zero, and splits the loss at the worst position into position and dilution. Program p13 shows that more documents push the worst position lower and the mean capture from 0.741 down to 0.626. All of these numbers come from one model on one task. Does another model still draw a U? And does a version with a larger window read any better? (https://5dlibrary.com/answers/en/long-context-mid-forgetting-positional-bias-summary-relay/c01/) 第1章 中段塌陷:U 形位置曲线的原始数字、效应量与剂量 — 程序 p11 把 GPT-3.5-Turbo 在二十篇文档上的成绩拟合成一条开口朝上的抛物线,深度 17.09 个百分点。程序 p12 给出开头与中间之差 0.220 的置信区间,并说明「中段低于闭卷」跨过零;它还把最坏位置的损失拆成位置与稀释两份。程序 p13 显示文档越多,最坏位置越低,平均捕获率从 0.741 降到 0.626。这些数字都来自同一个模型、同一种任务。那么换一个模型,曲线还是 U 形吗?窗口更大的版本,会不会读得更好? (https://5dlibrary.com/answers/zh/long-context-mid-forgetting-positional-bias-summary-relay/c01/) - Ch.2 Dissent and Diversity: Four Curves, Bigger Windows and the Single-Needle Test — Program p21 shows the four models of the paper drawing four shapes, with only one model's middle below its closed-book score and every model carrying a clear dilution loss. Program p22 shows that extended-window versions differ from their originals by no more than 0.3 points, far below sampling error. Program p23 shows that a single-needle recall of 0.997 leaves only a 0.740 chance of a clean sweep over a hundred facts. Description has gone far enough. To predict a length that has never been measured we need a formula that extrapolates. Can primacy, recency and dilution be written into one curve? (https://5dlibrary.com/answers/en/long-context-mid-forgetting-positional-bias-summary-relay/c02/) 第2章 反方与异质:四种曲线、更大的窗口与单针测试 — 程序 p21 显示原文四个模型画出四种形状,只有一个模型的中段低于闭卷成绩,而所有模型都有明显的稀释损失。程序 p22 显示扩展窗口版本与原版的差距不超过 0.3 个百分点,远小于抽样误差。程序 p23 说明单针召回 0.997 换到一百条事实上只剩 0.740 的全中机会。描述到此为止已经够了;要预测一个从未测过的长度,我们需要一个能外推的公式。首因、近因与稀释,能不能写进同一条曲线? (https://5dlibrary.com/answers/zh/long-context-mid-forgetting-positional-bias-summary-relay/c02/) - Ch.3 A Position-Recall Model: a Primacy Bump, a Recency Bump and a Floor That Sinks with Length — Program p31 fits the two-exponential model separately to thirty and twenty documents: the primacy bump is 0.249 in both, the decay lengths are 2.5 and 2.2 documents, and only the floor falls, from 0.609 to 0.578. Program p32 turns that structure into a joint model, fitting twelve cells with five parameters at an error of 0.0105, and extrapolates a capture curve: 0.908 at four documents, 0.612 at thirty-two and 0.477 at two hundred and fifty-six. Every number on this curve can be remeasured by an owner on his own reader, and Chapter 8 gives the method. With this curve in hand we can ask the relay question. If long material is cut into short segments, summarised and passed on, how much is lost at each hand-off, and after how many layers does relaying stop paying? (https://5dlibrary.com/answers/en/long-context-mid-forgetting-positional-bias-summary-relay/c03/) 第3章 位置召回模型:首因峰、近因峰与随长度下沉的地板 — 程序 p31 用双指数模型分别拟合三十篇与二十篇数据,首因峰高度都是 0.249,衰减长度 2.5 与 2.2 篇,只有地板从 0.609 降到 0.578。程序 p32 把这一结构写成联合模型,五个参数拟合十二个格子,误差 0.0105,并外推出捕获曲线:四篇 0.908,三十二篇 0.612,两百五十六篇 0.477。这条曲线的每一个数都可以由经营者用自己的读者重测,第八章会给出方法。有了这条曲线,我们就能问接力的问题了:把长材料切成短段、写摘要、再往下传,每传一层会丢多少,传几层就不划算了? (https://5dlibrary.com/answers/zh/long-context-mid-forgetting-positional-bias-summary-relay/c03/) - Ch.4 The Product of Retention Rates: a Ceiling on Layers, the Recency Slope of Running Summaries and the Casualties among Details — Chapter 4 computes the product of relay retention rates, shows that a running summary turns the U into a recency slope, and finds detailed facts keeping only 0.4305 after eight layers. (https://5dlibrary.com/answers/en/long-context-mid-forgetting-positional-bias-summary-relay/c04/) 第4章 接力保留率连乘:层数上限、滚动摘要的近因斜坡与细节的伤亡 — 第四章计算接力保留率连乘,说明滚动摘要把 U 形变成近因斜坡,细节类事实八层后只剩 0.4305。 (https://5dlibrary.com/answers/zh/long-context-mid-forgetting-positional-bias-summary-relay/c04/) - Ch.5 The Best Segment Length and the Relay Threshold: a Tug of War between Losses in Reading and in Writing — Chapter 5 solves for the best segment length and the relay threshold: for the 2023 reader each summary call must keep key facts above about 0.911 and segments must be short, and once the reader's loss falls below 0.413 of the original curve a relay at fidelity 0.97 loses to the whole window. (https://5dlibrary.com/answers/en/long-context-mid-forgetting-positional-bias-summary-relay/c05/) 第5章 最优段长与接力阈值:读的损失与写的损失之间的拉锯 — 第五章求最优段长与接力阈值:对 2023 年的读者,每次摘要调用的保真度须高于约 0.911,段长须很短;读者损失低于原曲线的 0.413 倍时,保真度 0.97 的接力就输给整窗。 (https://5dlibrary.com/answers/zh/long-context-mid-forgetting-positional-bias-summary-relay/c05/) - Ch.6 The Key-Fact Restatement Slot: Zero Loss Comes Only from Template Copying and Checksums — Chapter 6 shows that zero loss can come only from a key-fact restatement slot copied by template and verified by checksum, which lifts key-fact survival from 0.601 to above nine in ten but holds only about seven lines. (https://5dlibrary.com/answers/en/long-context-mid-forgetting-positional-bias-summary-relay/c06/) 第6章 关键事实复述位:零损耗只能来自模板复制与校验 — 第六章说明零损耗只能来自模板复制加校验码的关键事实复述位,它把关键事实的存活从 0.601 提到九成以上,但只容得下约七行。 (https://5dlibrary.com/answers/zh/long-context-mid-forgetting-positional-bias-summary-relay/c06/) - Ch.7 The Classics Examined: Mencius on One Day of Sun and Ten Days of Cold, and Ebbinghaus on Memory — Program p72 turns Mencius's image into parameters: with a five per cent cold-day loss, the warm day must add 0.670 to break even; in a relay, restating a key fact only every eleventh layer leaves a mean survival of 0.863 and a worst of 0.737. Program p71 shows that Ebbinghaus's original data look more like a power law (an error of 0.0206 against 0.0933) while the replication looks more like an exponential, so the form of forgetting is not settled. Both classics point to the same business conclusion: what is key must get sun every day, and strong and weak facts must be told apart and counted separately. The last chapter must say what workflow an owner can follow tomorrow morning, and what result would overturn this book. (https://5dlibrary.com/answers/en/long-context-mid-forgetting-positional-bias-summary-relay/c07/) 第7章 原典考辨:孟子的「一日暴之,十日寒之」与艾宾浩斯的《论记忆》 — 程序 p72 把孟子的比喻写成参数:冷天损失百分之五时,晴天须生长 0.670 才能持平;在接力里,每十一层才复述一次,关键事实平均只剩 0.863,最差 0.737。程序 p71 显示艾宾浩斯原始数据更像幂律(误差 0.0206 对 0.0933),但重复实验更像指数,遗忘的函数形式并未定论。两部原典都指向同一个经营上的结论:关键的东西要「日日暴之」,而且要分清强弱、分开计量。最后一章要回答的,是经营者明天早上就能照着做的流程长什么样,以及什么样的结果会推翻这本书? (https://5dlibrary.com/answers/zh/long-context-mid-forgetting-positional-bias-summary-relay/c07/) - Ch.8 Closing the Loop: a Relay Template for Long Tasks, the Key-Fact Slot and a Pre-Registered Test — Chapter 8 gives a relay template for long tasks, the Segment Relay Threshold Calculator and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/long-context-mid-forgetting-positional-bias-summary-relay/c08/) 第8章 实证闭环:长任务接力摘要模板、关键事实复述位与预注册检验 — 第八章给出长任务接力摘要模板、分段接力阈值计算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/long-context-mid-forgetting-positional-bias-summary-relay/c08/) ## 5DL-B-256 · Hard Gates across the Client Lifecycle: the Decay of Soft-Prompt Compliance over Turns and State Machines for Interception before Publication 中文书名:客户端生命周期硬门禁工程、软提示遵从率轮次衰减与保存前物理拦截状态机母典 DOI: pending Cite: Fan, Y. (2026). Hard Gates across the Client Lifecycle: the Decay of Soft-Prompt Compliance over Turns and State Machines for Interception before Publication (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Soft reminders work in proportion; when layers are independent the residual violation rate is the product of their miss rates, and with illustrative parameters four layers still leave 0.0048, not zero. Its parameters are illustrative; all calculations are generated by its programs. Abstract: People who work through AI agents often believe that if reminders are clear enough and layered enough, violations will approach zero. This book tests that belief with the mathematics of reliability. Soft reminders work in proportion; when layers are independent the residual violation rate is the product of their miss rates, and with illustrative parameters four layers still leave 0.0048, not zero. Only a deterministic code gate can reach zero, and only inside the pattern class it defines. After checking the sources the book corrects its founding blueprint: Reason's Human Error of 1990 writes no product formula, which holds only under independence; no reliable per-turn primary data exist for the decay of soft-prompt compliance, so the book states it as the author's hypothesis with conditions for refutation; and the author's own file gate actually sits after the write, so what guarantees the zero is a publication gate that refuses on error. Its parameters are illustrative; all calculations are generated by its programs. It is not individual business advice. Chapter 1 does the product account and splits the claim into an in-class and an out-of-class part. Chapter 2 treats common-cause failure: with a common-cause share of only 0.01, it supplies 0.933 of the residual after six soft layers. Chapter 3 gives an exponential decay model with a floor, with a half-life of 6.93 turns, under which a forty-turn session lets through 4.71 times the static estimate. Chapter 4 compares interception before the write, after it and before publication; the combined design E costs only 0.60 at a hook failure rate of 0.010. Chapter 5 proves by exhausting a state machine that a dirty publication is unreachable under the right design; with a green light not bound to content, 0.1964 of event sequences of length ten go wrong. Chapter 6 prices wrong blocks: a wide rule set of coverage 0.97 survives only 15.52 weeks a year and lets through 354.64 violations. Chapter 7 examines the chapter "You Du" of the Han Feizi and Reason's classification of error; behind the gate, 0.627 of the residual is knowledge-based mistakes. Chapter 8 gives a weekly interception report (in an invented example the block rate falls from 42.39 to 12.71), the Gate Residual Violation Rate Calculator and a pre-registered test (power 0.916). Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:用智能体干活的人常以为:提醒写得够清楚、够多层,违规就会趋近于零。本书用可靠性数学检验这个说法。软提醒按比例起作用,各层独立时残余违规率是漏过率之积,示意参数下四层之后仍是 0.0048,不是零;只有确定性的代码门禁能在它自己定义的模式类别之内做到零。核对原典后,本书修正了立项蓝图:里森 1990 年的《人为差错》没有写出乘积公式,乘积只在独立假设下成立;「软提示遵从率随轮次衰减」没有可靠的逐轮一手数据,本书把它写成带证伪条件的作者假说;作者自己的文件门禁其实挂在写入之后,真正保证零的是出错即不放行的出版闸门。书中参数均为示意;全部计算由书中程序生成。本书属非个人化经营建议。 第一章算乘积账,并把命题拆成类内与类外两部分。第二章讲共因失效:共同原因份额只有 0.01 时,六层软防护的残余里就有 0.933 来自它。第三章给出带地板的指数衰减模型,半衰期 6.93 轮,四十轮会话的漏网是静态估计的 4.71 倍。第四章比较写入前、写入后与出版前三种拦截位置,组合设计 E 在钩子故障率 0.010 时成本只有 0.60。 第五章用状态机穷举证明正确设计下带病出版不可达,绿灯不绑定内容时长度为十的事件序列有 0.1964 会出错。第六章讲误拦的代价:覆盖 0.97 的宽规则集一年只活 15.52 周,全年漏网 354.64 次。第七章考辨《韩非子·有度》与里森的差错分类,门禁之后的残余有 0.627 是知识型错误。第八章给出门禁拦截日志周报(虚构示例中阻断率从 42.39 降到 12.71)、小工具「门禁残余违规率计算器」与预注册检验(检验力 0.916)。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Product Account: Why Stacked Soft Layers Never Leave a Zero Residual — Soft reminders work in proportion; stacked, they leave a product, and with illustrative parameters four layers leave 0.004800, not zero. A hard gate sorts violations into an in-class part, which escapes only when the gate breaks and passes, and an out-of-class part, which it does not touch. Seeing zero yields only an upper bound; the guarantee of zero comes from structure. The product account, however, rests on a strong assumption: that the layers are independent. If every soft layer reads the same context and suffers the same lapse, will the holes open together, and where will the gain from stacking stop? That is the question of the next chapter. (https://5dlibrary.com/answers/en/client-lifecycle-hard-gates-soft-prompt-decay-interception/c01/) 第1章 乘积账:软层叠加的残余违规率为何永远不是零 — 软提醒是按比例起作用的,叠层之后残余率是乘积,示意参数下四层之后是 0.004800,不是零。硬门禁把违规分成类内与类外:类内只有在门禁故障且放行时才漏,类外完全不受影响。观察到零只给出一个上界,零的保证来自结构。可是,乘积账依赖一个很强的假设:各层彼此独立。假如所有软层读的是同一段上下文、犯的是同一种迷糊,漏洞会不会一起出现?叠层的收益又会在哪里停下?这是下一章的问题。 (https://5dlibrary.com/answers/zh/client-lifecycle-hard-gates-soft-prompt-decay-interception/c01/) - Ch.2 Holes That Line Up: the Floor That Common-Cause Failure Puts under Stacked Soft Layers — Chapter 2 treats common-cause failure: with a common-cause share of only 0.01, it supplies 0.933 of the residual after six soft layers. (https://5dlibrary.com/answers/en/client-lifecycle-hard-gates-soft-prompt-decay-interception/c02/) 第2章 漏洞对齐:共因失效给软层叠加设下的地板 — 第二章讲共因失效:共同原因份额只有 0.01 时,六层软防护的残余里就有 0.933 来自它。 (https://5dlibrary.com/answers/zh/client-lifecycle-hard-gates-soft-prompt-decay-interception/c02/) - Ch.3 Decay over Turns: an Exponential Model of Soft-Prompt Compliance, Re-Sending Periods and Reading the Log — Chapter 3 gives an exponential decay model with a floor, with a half-life of 6.93 turns, under which a forty-turn session lets through 4.71 times the static estimate. (https://5dlibrary.com/answers/en/client-lifecycle-hard-gates-soft-prompt-decay-interception/c03/) 第3章 轮次衰减:软提示遵从率的指数模型、重发周期与日志判别 — 第三章给出带地板的指数衰减模型,半衰期 6.93 轮,四十轮会话的漏网是静态估计的 4.71 倍。 (https://5dlibrary.com/answers/zh/client-lifecycle-hard-gates-soft-prompt-decay-interception/c03/) - Ch.4 Where to Intercept: before the Write, after It or before Publication, and Whether a Broken Gate Passes or Refuses — Chapter 4 compares interception before the write, after it and before publication; the combined design E costs only 0.60 at a hook failure rate of 0.010. (https://5dlibrary.com/answers/en/client-lifecycle-hard-gates-soft-prompt-decay-interception/c04/) 第4章 拦截位置:写入前、写入后与出版前,门禁坏了放行还是拒绝 — 第四章比较写入前、写入后与出版前三种拦截位置,组合设计 E 在钩子故障率 0.010 时成本只有 0.60。 (https://5dlibrary.com/answers/zh/client-lifecycle-hard-gates-soft-prompt-decay-interception/c04/) - Ch.5 Exhausting the State Machine: Proving a Dirty Publication Unreachable and Finding the Shortest Counterexamples of Three Faulty Designs — Chapter 5 proves by exhausting a state machine that a dirty publication is unreachable under the right design; with a green light not bound to content, 0.1964 of event sequences of length ten go wrong. (https://5dlibrary.com/answers/en/client-lifecycle-hard-gates-soft-prompt-decay-interception/c05/) 第5章 状态机穷举:证明带病出版不可达,并找出三种错误设计的最短反例 — 第五章用状态机穷举证明正确设计下带病出版不可达,绿灯不绑定内容时长度为十的事件序列有 0.1964 会出错。 (https://5dlibrary.com/answers/zh/client-lifecycle-hard-gates-soft-prompt-decay-interception/c05/) - Ch.6 The Price of Hardness: Pattern Width, Wrong Blocks and Gates That Get Switched Off — Chapter 6 prices wrong blocks: a wide rule set of coverage 0.97 survives only 15.52 weeks a year and lets through 354.64 violations. (https://5dlibrary.com/answers/en/client-lifecycle-hard-gates-soft-prompt-decay-interception/c06/) 第6章 硬的代价:模式宽度、误拦与被关掉的门禁 — 第六章讲误拦的代价:覆盖 0.97 的宽规则集一年只活 15.52 周,全年漏网 354.64 次。 (https://5dlibrary.com/answers/zh/client-lifecycle-hard-gates-soft-prompt-decay-interception/c06/) - Ch.7 Classical Sources East and West: Law and the Inked Line in the Han Feizi, and Reason's Classification of Error — Chapter 7 examines the chapter "You Du" of the Han Feizi and Reason's classification of error; behind the gate, 0.627 of the residual is knowledge-based mistakes. (https://5dlibrary.com/answers/en/client-lifecycle-hard-gates-soft-prompt-decay-interception/c07/) 第7章 东西方原典考辨:《韩非子·有度》的法与绳,里森《人为差错》的差错分类 — 第七章考辨《韩非子·有度》与里森的差错分类,门禁之后的残余有 0.627 是知识型错误。 (https://5dlibrary.com/answers/zh/client-lifecycle-hard-gates-soft-prompt-decay-interception/c07/) - Ch.8 Closing the Loop: the Weekly Interception Report, the Residual Calculator and a Pre-Registered Test — Chapter 8 gives a weekly interception report (in an invented example the block rate falls from 42.39 to 12.71), the Gate Residual Violation Rate Calculator and a pre-registered test (power 0.916). Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/client-lifecycle-hard-gates-soft-prompt-decay-interception/c08/) 第8章 实证闭环:门禁拦截日志周报、残余违规率计算器与预注册检验 — 第八章给出门禁拦截日志周报(虚构示例中阻断率从 42.39 降到 12.71)、小工具「门禁残余违规率计算器」与预注册检验(检验力 0.916)。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/client-lifecycle-hard-gates-soft-prompt-decay-interception/c08/) ## 5DL-B-257 · A Lazy-Loading Private Tool Hub: Near-Zero Opening Context for Hundreds of Checkers and On-Demand Routing Trees 中文书名:延迟加载私有工具中枢、数百检查器零初始上下文开销与按需路由树母典 DOI: pending Cite: Fan, Y. (2026). A Lazy-Loading Private Tool Hub: Near-Zero Opening Context for Hundreds of Checkers and On-Demand Routing Trees (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Selection accuracy lacks reliable peer-reviewed primary evidence and is demoted to an author's hypothesis. Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. Abstract: As checkers, calculators and writing gates pile up, the easiest wiring puts every tool's full description into context at the opening. The commissioning claim was that loading everything at once eats context linearly, and that a multi-level routing tree built from call frequencies brings the expected description cost close to the entropy lower bound. This book writes the claim as calculable models and corrects the blueprint in five places. The entropy bound comes from Shannon in 1948, not Huffman in 1952. Entropy bounds decisions, not tokens: each decision reads a menu, and the lowest-cost branching factor per lookup is about 2.72 with no hop cost and 5.09 with a hop cost of 80 tokens. A Huffman tree's inner nodes have no names; a navigable tree must keep a semantic order, which costs only 0.112 extra decisions in a random order. Lazy loading is not always cheaper: for three hundred tools a session costs 150000 tokens loaded flat against 8558.5 loaded lazily, a ratio of 17.53, but the order reverses above 1479 calls per session. Selection accuracy lacks reliable peer-reviewed primary evidence and is demoted to an author's hypothesis. The parameters are illustrative; all calculations are generated by the book's programs, which read and write no file; the book is non-individual business advice, and any change to a live tool catalogue must be reviewed and released by hand by the person in charge. Chapter 1 sets the session bills of the two ways of loading side by side with the entropy bound. Chapter 2 treats dummy leaves in many-way Huffman trees, exhaustive checking and the price of a menu. Chapter 3 treats the price of alphabetic order and three assumed laws of wrong turns. Chapter 4 treats depth against width and hanging on the root any tool called in a session with probability above 0.862. Chapter 5 treats cold starts with a queue with set-up times and the ski-rental rule, whose break-even warm period is never worse than 2.000 times the optimum under any distribution of gaps. Chapter 6 treats frequency drift: the cost of a mismatch is about the relative entropy, and rebuilding must be triggered by a held-out test. Chapter 7 examines "establishing offices and dividing duties" and "great matters go through the head, small matters are dealt with directly" in the Rites of Zhou, and Dewey's decimal scheme and subject index of 1876. Chapter 8 gives a monthly report on catalogue layering and call frequency, the distilled "Tool Routing-Tree Cost Calculator", a pre-registered test plan and the conditions that would overturn the book. Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. 摘要:检查器、计算器与写作闸门越积越多,最省事的接法是把每个工具的完整说明一开场就全部放进上下文。立项命题说:一次性加载会线性吃掉上下文,按调用频率构造的多级路由树能把期望说明开销降到接近信息熵下界。本书把命题写成可计算的模型,并修正蓝图五处:熵下界出自香农 1948 年而非霍夫曼 1952 年;熵限制的是决策次数而不是词元,每次决策要读一屏菜单,单次查找最省的分支数在跳开销为零时约 2.72、为 80 个词元时约 5.09;霍夫曼树的内部节点没有名字,可导航的树须保持语义次序,随机次序下只多 0.112 次决策;按需加载并非总更省,三百个工具时会话账单一次性加载 150000 个词元、按需加载 8558.5,比值 17.53,但每次会话超过 1479 次调用即反转;选择准确率缺乏可靠的同行评审一手证据,整体降为作者假说。书中参数为示意;全部计算由书中程序生成;程序不读写任何文件;本书是非个人化经营建议,任何线上工具目录的改动须由主事者人工复核放行。 第一章并排算出两种加载方式的会话账单与熵下界。第二章讲多叉霍夫曼树的哑叶补齐、穷举复核与菜单代价。第三章讲字母序约束的价格与逐层误选的三种假设律。第四章讲深度与宽度,以及把会话中被调用概率超过 0.862 的工具直挂在根上。第五章用带启动时间的排队模型与租雪橇规则讨论冷启动,盈亏平衡常驻在任何间隔分布下最坏不超过最优的 2.000 倍。第六章讲频率漂移:错配代价约等于相对熵,重建须用留出检验触发。 第七章考辨《周礼》「设官分职」与「大事则从其长,小事则专达」,以及杜威 1876 年的十分法与主题索引。第八章给出工具目录分层与调用频率的月报流程、蒸馏小工具「工具路由树成本计算器」、预先登记的检验方案与推翻本书的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Linear Bill: Context Cost of Loading Everything and the Entropy Bound — At exponent zero the entropy is 8.229 bits and the Huffman tree needs 8.293 decisions, a saving of only 0.079 of the balanced tree's cost: when all tools are equally used, frequency has nothing to sort. At exponent one the entropy is 6.372, the Huffman tree needs 6.406, a gap of 0.034, and the saving over the balanced tree is 0.288. At exponent two the entropy is only 2.324, the Huffman tree needs 2.379, and the saving is 0.736. In all five rows the gap between the Huffman tree and the entropy is at most 0.065, far below the one unit that Shannon's theorem allows. The last two columns concern an 8-ary tree: counting three binary choices as one eight-way choice, the entropy at exponent one equals 2.124 eight-way choices, while the 8-ary Huffman tree needs 2.207. (https://5dlibrary.com/answers/en/lazy-loading-tool-hub-zero-initial-context-routing-tree/c01/) 第1章 线性账单:一次性加载的上下文开销与熵下界 — 指数为零时,熵是 8.229 比特,霍夫曼树要 8.293 次决策,比平衡树只少 0.079 的份额:大家一样常用,按频率排也排不出什么。指数为一时,熵是 6.372,霍夫曼树是 6.406,差距 0.034,比平衡树少了 0.288。指数为二时,熵只有 2.324,霍夫曼树是 2.379,比平衡树少了 0.736。五行里霍夫曼树与熵的差距都不超过 0.065,远小于香农定理允许的一。最后两列是八叉树:以三次二选一折合一次八选一,指数为一时熵折合 2.124 次八选一,八叉霍夫曼树实际要 2.207 次。 (https://5dlibrary.com/answers/zh/lazy-loading-tool-hub-zero-initial-context-routing-tree/c01/) - Ch.2 Building Huffman Trees: Many-Way Merging, Dummy Leaves and the Price of a Menu — A many-way Huffman tree must be padded with dummy leaves first; otherwise a 10-ary tree needs more than a quarter more decisions. On small instances the Huffman tree matched the exhaustive optimum in all 720 cases. Decisions are not tokens: each decision reads a menu, and the best branching factor in tokens is about 2.72 with zero hop cost and about 5.09 with a hop cost of 80 tokens. Yet this chapter has kept avoiding one question. A Huffman tree lumps cold tools together by frequency into a branch, so the model sees on its menu a node without a name. How is it to know that the tool it wants lies down that branch? The next chapter answers this, and puts a price on the answer. (https://5dlibrary.com/answers/en/lazy-loading-tool-hub-zero-initial-context-routing-tree/c02/) 第2章 霍夫曼构树:多叉合并、哑叶补齐与菜单代价 — 多叉霍夫曼树必须先补哑叶,否则十叉树会多出四分之一以上的决策。在小规模上,霍夫曼树与穷举的最优解在 720 个实例中全部相等。决策次数不是词元:每次决策要读一屏菜单,词元意义上的最优分支数在跳开销为零时约为 2.72,跳开销为 80 个词元时约为 5.09。可是本章一直回避了一个问题:霍夫曼树把冷门工具按频率胡乱拼成一枝,模型在菜单上看到的是一个没有名字的节点,它凭什么知道要找的工具在这一枝里?下一章回答这个问题,并给出它的价格。 (https://5dlibrary.com/answers/zh/lazy-loading-tool-hub-zero-initial-context-routing-tree/c02/) - Ch.3 Navigable Trees: the Price of Alphabetic Order and the Accumulation of Wrong Turns — A navigable tree must keep a semantic order. In theory this costs at most about one decision more than a Huffman tree; under a Zipf distribution, program p31 finds the real cost to be only about a tenth of a decision, 0.112 in a random order. The real uncertainty lies in wrong turns: whether a routing tree raises accuracy depends on how errors change with menu length, and that law has not yet been reliably measured in public. The next chapter sets accuracy aside for the moment and returns to tokens and round trips to ask a more practical question: given three hundred tools, how many levels should there be, how wide should each be, and which tools deserve to hang directly on the root? (https://5dlibrary.com/answers/en/lazy-loading-tool-hub-zero-initial-context-routing-tree/c03/) 第3章 可导航的树:字母序约束的价格与逐层误选的累积 — 可导航的树必须保持语义次序。理论上这最多比霍夫曼树多花约一次决策;在齐普夫分布下,程序 p31 算出的实际代价只有十分之一次左右,随机次序下为 0.112。真正的不确定性在误选:路由树能否提高准确率,取决于误选随菜单长度变化的规律,而这个规律至今没有可靠的公开测量。下一章暂时把准确率放在一边,回到词元与往返,问一个更实际的问题:给定三百个工具,到底该分几层、每层多宽,哪些工具值得直接挂在根上? (https://5dlibrary.com/answers/zh/lazy-loading-tool-hub-zero-initial-context-routing-tree/c03/) - Ch.4 The Branching Factor: Depth against Width, and Hanging Popular Tools on the Root — Chapter 4 treats depth against width and hanging on the root any tool called in a session with probability above 0.862. (https://5dlibrary.com/answers/en/lazy-loading-tool-hub-zero-initial-context-routing-tree/c04/) 第4章 分支因子:深度与宽度的取舍,以及把热门工具直挂在根上 — 第四章讲深度与宽度,以及把会话中被调用概率超过 0.862 的工具直挂在根上。 (https://5dlibrary.com/answers/zh/lazy-loading-tool-hub-zero-initial-context-routing-tree/c04/) - Ch.5 Queueing with Cold Starts: the Delay Bill of Lazy Loading and How Long to Stay Warm — Chapter 5 treats cold starts with a queue with set-up times and the ski-rental rule, whose break-even warm period is never worse than 2.000 times the optimum under any distribution of gaps. (https://5dlibrary.com/answers/en/lazy-loading-tool-hub-zero-initial-context-routing-tree/c05/) 第5章 冷启动排队:按需加载的延迟账与常驻时长 — 第五章用带启动时间的排队模型与租雪橇规则讨论冷启动,盈亏平衡常驻在任何间隔分布下最坏不超过最优的 2.000 倍。 (https://5dlibrary.com/answers/zh/lazy-loading-tool-hub-zero-initial-context-routing-tree/c05/) - Ch.6 Frequency Drift: the Cost of Wrong Frequencies and When to Rebuild the Routing Tree — The cost of using wrong frequencies is roughly the relative entropy: small drift is almost harmless, drift 1.0 adds 0.678 decisions per lookup, and large drift makes a frequency tree worse than a balanced one. The trigger for rebuilding must not be the in-sample gap, which fires every month; it should be a held-out test, so that rebuilding happens only when frequencies really change. With this, the mathematics of routing trees is complete. The next chapter leaves the computer for a while and goes back to an administrative order more than two thousand years old and a library scheme from a century and a half ago, to see how earlier ages layered hundreds of offices and thousands of books, what they got right and what they missed. (https://5dlibrary.com/answers/en/lazy-loading-tool-hub-zero-initial-context-routing-tree/c06/) 第6章 频率漂移:用错频率的代价与重建路由树的时机 — 用错频率的代价约等于相对熵:小漂移几乎无害,漂移 1.0 时每次查找多做 0.678 次决策,大漂移会让频率树比平衡树还差。重建的触发信号不能用样本内的差距,它每个月都会触发;应当用留出检验,让重建只在频率真的变化时发生。到这里,我们已经把路由树的数学讲完了。下一章暂时离开计算机,回到两千多年前的官制与一百五十年前的图书馆,看古人怎样给成百上千的职事与书籍分层,他们说对了什么,又漏掉了什么。 (https://5dlibrary.com/answers/zh/lazy-loading-tool-hub-zero-initial-context-routing-tree/c06/) - Ch.7 Classical Sources East and West: Offices and Duties in the Rites of Zhou and Dewey's Decimal Scheme — Chapter 7 examines "establishing offices and dividing duties" and "great matters go through the head, small matters are dealt with directly" in the Rites of Zhou, and Dewey's decimal scheme and subject index of 1876. (https://5dlibrary.com/answers/en/lazy-loading-tool-hub-zero-initial-context-routing-tree/c07/) 第7章 东西方原典考辨:《周礼》设官分职与杜威十分法 — 第七章考辨《周礼》「设官分职」与「大事则从其长,小事则专达」,以及杜威 1876 年的十分法与主题索引。 (https://5dlibrary.com/answers/zh/lazy-loading-tool-hub-zero-initial-context-routing-tree/c07/) - Ch.8 Closing the Loop: a Monthly Report on Catalogue Layering and Call Frequency — Chapter 8 gives a monthly report on catalogue layering and call frequency, the distilled "Tool Routing-Tree Cost Calculator", a pre-registered test plan and the conditions that would overturn the book. Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/lazy-loading-tool-hub-zero-initial-context-routing-tree/c08/) 第8章 实证闭环:工具目录分层与调用频率月报 — 第八章给出工具目录分层与调用频率的月报流程、蒸馏小工具「工具路由树成本计算器」、预先登记的检验方案与推翻本书的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/lazy-loading-tool-hub-zero-initial-context-routing-tree/c08/) ## 5DL-B-258 · Cross-Volume Rule Resolution and Axiom Reduction: Detecting Threshold Conflicts among a Thousand Machine Rules and Finding a Minimal Primitive Set 中文书名:跨卷规则消解与公理归约、千条机器规则阈值冲突检测与最小原语集母典 DOI: pending Cite: Fan, Y. (2026). Cross-Volume Rule Resolution and Axiom Reduction: Detecting Threshold Conflicts among a Thousand Machine Rules and Finding a Minimal Primitive Set (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: The rule libraries and values are illustrative; every calculation is generated by the book's programs; the book offers general business method, not legal, compliance or business-decision advice, and adding or deleting a rule must be released by its owner. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: The rules a business team hands to machines will sooner or later number in the thousands: price floors and exclusions, contrast ratios and touch targets, forbidden claims and countdowns, follow-up caps and quiet hours. A scanner marks in red every pair of rules on one metric whose numbers differ, and the common belief concludes that the rulebook contradicts itself. This book tests that belief. It splits conflict into four relations, duplicate, subsumption, compatibility and true contradiction; writes threshold rules as intervals in exact fractions; and reduces a thousand rules to a minimal primitive set with transitive closure, prime implicants and greedy set cover. All calculations use only the Python standard library. The rule libraries and values are illustrative; every calculation is generated by the book's programs; the book offers general business method, not legal, compliance or business-decision advice, and adding or deleting a rule must be released by its owner. Chapter 1 states and quantifies the claim: in an illustrative library of a thousand rules, true contradictions are only 0.0392 of flagged pairs, 971 rules are redundant after the rulings and just 18 stay binding; even when one rule in ten crosses over, the contradiction share is only 0.2667, although every metric is by then infeasible. The preface corrects the blueprint: the greedy ratio Johnson proved in 1974 is of harmonic order, at most ln n plus one; Quine's paper of 1952 concerns prime implicants, not set cover; and "most conflicts are redundancy" is the author's hypothesis, not a published law. Chapter 2 shows one-dimensional thresholds decided in one sweep, constraints between two quantities needing negative-cycle detection, and floating point inventing false conflicts on 0.1151 of boundary pairs. Chapter 3 uses transitive closure to find contradictions in chains and dead rules. Chapter 4 merges 40 narrow prohibitions into 4 primitives. Chapter 5 gives the bounds and a gap certificate for greedy set cover, which on ordinary instances is off by a mean factor of only 1.0919. Chapter 6 shows text similarity recalling only 0.417 of true duplicates at the 0.85 cut-off and proposes canonical keys. Chapter 7 examines the Xunzi on correct names and Book I of Euclid's Elements. Chapter 8 gives an eight-step monthly routine for the pruner, the distilled tool, the Rule Conflict Reducer, and a pre-registered test: 60 blind-labelled pairs, a rejection bound of 23 and a test size of 0.0462. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:一个经营团队交给机器执行的规则迟早上千:定价的底价与互斥,界面的对比度与点按尺寸,文案的红线词与倒计时,私讯的跟进上限与静默时段。扫描器把同一指标上数值不同的规则全部标红,常识便以为规则册已经自相矛盾。本书检验这一常识:把冲突拆成重复、包含、可并存与真矛盾四种关系,用精确分数把阈值规则写成区间,用传递闭包、素蕴含项与贪心集合覆盖把上千条规则归约为最小原语集。全部计算只用 Python 标准库。书中规则库与数值均为示意;全部计算由书中程序生成;本书属非个人化经营方法,不构成法律、合规或经营决策意见,规则的增删须由主事者人工放行。 第一章正面陈述并量化命题:一千条示意规则里,标红配对中真矛盾只占 0.0392,裁决后 971 条是冗余,只剩 18 条约束性规则;越界规则占一成时真矛盾比例也只有 0.2667,但无解指标已遍及全部。序中修正立项蓝图:约翰逊 1974 年证明的贪心比值是调和数级,至多 ln n 加一倍;奎因 1952 年讨论的是素蕴含项而非集合覆盖;「冲突多半是冗余」是作者假说而非已发表定律。第二章说明一维阈值一次扫描即可判定,两量约束须做负环检测,浮点会在 0.1151 的边界配对上造出假冲突。第三章用传递闭包找出链条上的矛盾与死规则。第四章把 40 条窄禁令并成 4 个本原。 第五章给出贪心集合覆盖的界与差距证书,平常实例上平均只差 1.0919 倍。第六章显示文本相似度在 0.85 界线上的召回只有 0.417,并给出规范键。第七章考辨《荀子·正名》与欧几里得《几何原本》第一卷。第八章给出剪枝器接入的八步月度流程、蒸馏小工具「规则冲突归约器」与预注册检验:60 对盲判配对、拒绝界 23、检验水平 0.0462。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Most Conflicts among a Thousand Rules Are Redundancy — Chapter 1 states and quantifies the claim: in an illustrative library of a thousand rules, true contradictions are only 0.0392 of flagged pairs, 971 rules are redundant after the rulings and just 18 stay binding; even when one rule in ten crosses over, the contradiction share is only 0.2667, although every metric is by then infeasible. The preface corrects the blueprint: the greedy ratio Johnson proved in 1974 is of harmonic order, at most ln n plus one; Quine's paper of 1952 concerns prime implicants, not set cover; and "most conflicts are redundancy" is the author's hypothesis, not a published law. (https://5dlibrary.com/answers/en/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c01/) 第1章 千条规则里的冲突,多半是同义冗余 — 第一章正面陈述并量化命题:一千条示意规则里,标红配对中真矛盾只占 0.0392,裁决后 971 条是冗余,只剩 18 条约束性规则;越界规则占一成时真矛盾比例也只有 0.2667,但无解指标已遍及全部。序中修正立项蓝图:约翰逊 1974 年证明的贪心比值是调和数级,至多 ln n 加一倍;奎因 1952 年讨论的是素蕴含项而非集合覆盖;「冲突多半是冗余」是作者假说而非已发表定律。 (https://5dlibrary.com/answers/zh/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c01/) - Ch.2 Rules as Intervals: Deciding Threshold Conflicts Mechanically — Chapter 2 shows one-dimensional thresholds decided in one sweep, constraints between two quantities needing negative-cycle detection, and floating point inventing false conflicts on 0.1151 of boundary pairs. (https://5dlibrary.com/answers/en/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c02/) 第2章 规则即区间:阈值冲突的机械判定 — 第二章说明一维阈值一次扫描即可判定,两量约束须做负环检测,浮点会在 0.1151 的边界配对上造出假冲突。 (https://5dlibrary.com/answers/zh/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c02/) - Ch.3 Conflict Graphs and Transitive Closure: Contradictions Hidden in Chains — Switch rules form a conflict graph in which implication is transitive and exclusion spreads along implication. In the illustrative library of p31, 4 written exclusions become 6 after closure, and two dead rules emerge, the same-day launch and the evergreen page; the latter is the formal reason why a resetting countdown is always a violation. Random libraries show derived exclusions and dead rules multiplying a hundredfold once the mean out-degree passes one. p32 shows that restatements declared in both directions shrink 280 rules to 40 synonym classes and 617 links to a skeleton of 71, while each tenth of declarations left out adds about twenty singleton classes. Graph theory tells us which rules are equivalent and which redundant, but it works rule by rule. When dozens of prohibitions each forbid one narrow combination, and together amount to a few broad prohibitions, the graph cannot see it. How can they be merged into the fewest primitives? The next chapter calls on Quine and McCluskey. (https://5dlibrary.com/answers/en/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c03/) 第3章 冲突图与传递闭包:藏在链条里的矛盾 — 开关型规则构成冲突图,蕴含会传递,互斥沿蕴含传播。p31 的示意库里,4 条写下的互斥闭包后变成 6 对,并推出两条死规则:当天上线与常青页,后者正是会重置的倒计时为什么一律违规的形式化理由。随机库显示,蕴含的平均出度越过一,派生互斥与死规则成百倍增长。p32 显示,双向声明的重述能把 280 条规则缩成 40 个同义类,617 条牵连缩成 71 条骨架;每少声明一成,就多出约二十个单点类。图论告诉我们哪些规则等价、哪些冗余,但它处理的是一条一条的规则。如果几十条禁令各自只禁止一个很窄的组合,合起来其实等于几条宽的禁令,图上看不出来。怎样把它们合并成最少的本原?下一章请出 Quine 与 McCluskey。 (https://5dlibrary.com/answers/zh/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c03/) - Ch.4 Prime Implicants: Merging Forty Narrow Prohibitions into Four Primitives — Chapter 4 merges 40 narrow prohibitions into 4 primitives. (https://5dlibrary.com/answers/en/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c04/) 第4章 素蕴含项:把四十条窄禁令并成四个本原 — 第四章把 40 条窄禁令并成 4 个本原。 (https://5dlibrary.com/answers/zh/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c04/) - Ch.5 Greedy Set Cover and ln n: Provable Bounds on a Minimal Primitive Set — Chapter 5 gives the bounds and a gap certificate for greedy set cover, which on ordinary instances is off by a mean factor of only 1.0919. (https://5dlibrary.com/answers/en/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c05/) 第5章 贪心集合覆盖与 ln n:最小原语集的可证界 — 第五章给出贪心集合覆盖的界与差距证书,平常实例上平均只差 1.0919 倍。 (https://5dlibrary.com/answers/zh/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c05/) - Ch.6 Judging Synonyms: Two Errors of Text Similarity, and Canonical Keys — Chapter 6 shows text similarity recalling only 0.417 of true duplicates at the 0.85 cut-off and proposes canonical keys. (https://5dlibrary.com/answers/en/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c06/) 第6章 同义判定:文本相似度的两种错与规范键 — 第六章显示文本相似度在 0.85 界线上的召回只有 0.417,并给出规范键。 (https://5dlibrary.com/answers/zh/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c06/) - Ch.7 Classical Sources East and West: the Xunzi on Correct Names and Book I of Euclid's Elements — The Xunzi's "the same are named the same, the different are named differently" and "where a single name does not suffice, a compound name is used" correspond to merging, separating and adding scope in a rule library. p71 shows compound names removing all 3 false conflicts, while only grouping by true meaning finds all 53 duplicate pairs. Xunzi gave no detailed criterion for sameness, and in particular did not treat one meaning written with two numbers. Book I of the Elements built all its propositions on a few postulates and common notions and pioneered the minimal primitive set; the independence of the fifth postulate needed models to be settled, and Hilbert supplied the assumptions Euclid had left tacit. p72 shows 8 of 9 rules redundant one at a time and yet impossible to delete together, with irredundant bases of three sizes, 4, 5 and 6. The classics gave a direction and modern mathematics a method. What remains is practice: how to fit these methods into monthly rule governance, and how to test the central claim of this book? The last chapter gives the routine and a pre-registered protocol. (https://5dlibrary.com/answers/en/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c07/) 第7章 东西方原典考辨:《荀子·正名》与《几何原本》第一卷 — 《荀子·正名》的「同则同之,异则异之」与「单不足以喻则兼」,对应规则库里的合并、分开与加上适用范围;p71 显示兼名消除了全部 3 处虚假冲突,而只有按真实意思分组才能找全 53 对重复。荀子没有给出判定同异的细则,尤其没有处理同一个意思写了两个数值的情形。《几何原本》第一卷把全部命题建立在少数公设与公理上,开了最小原语集的先河;第五公设的独立性要靠模型才能确认,而希尔伯特补上了欧几里得默认的假定。p72 显示,9 条规则里 8 条单独看都冗余,却不能一起删;不可再删的基有 4、5、6 条三种大小。原典给了方向,现代数学给了方法。剩下的问题是落地:怎样把这些方法接进每个月的规则治理,又怎样检验本书的核心命题?最后一章给出流程与预注册方案。 (https://5dlibrary.com/answers/zh/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c07/) - Ch.8 Closing the Loop: Plugging in the Pruner, Monthly Reduction and a Pre-Registered Test — Chapter 8 gives an eight-step monthly routine for the pruner, the distilled tool, the Rule Conflict Reducer, and a pre-registered test: 60 blind-labelled pairs, a rejection bound of 23 and a test size of 0.0462. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c08/) 第8章 实证闭环:剪枝器接入、月度归约与预注册检验 — 第八章给出剪枝器接入的八步月度流程、蒸馏小工具「规则冲突归约器」与预注册检验:60 对盲判配对、拒绝界 23、检验水平 0.0462。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cross-volume-rule-resolution-axiom-reduction-minimal-primitives/c08/) ## 5DL-B-259 · Red-Blue Synthetic-Customer Sandboxes: Multi-Round Bargaining against Parameterised Demanding Buyers and Pre-Launch Leak Detection 中文书名:红蓝军合成客群对抗沙盘、参数化挑剔买家多轮砍价博弈与漏单破绽预检母典 DOI: pending Cite: Fan, Y. (2026). Red-Blue Synthetic-Customer Sandboxes: Multi-Round Bargaining against Parameterised Demanding Buyers and Pre-Launch Leak Detection (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations are generated by the book's programs, which call no model interface. This is not individual business advice. Abstract: Before a sales page or chat script goes live, owners often stress-test it on one imagined average buyer. This book tests the opposite founding claim: the average buyer misses leaks in the tail, a Latin hypercube crowd of synthetic buyers covers more persona combinations for the same number of tests (McKay, Beckman and Conover, Technometrics, 1979), and multi-round haggling has the discounted equilibrium Rubinstein gave in 1982. Checking the originals corrects the claim. The original promise of the Latin hypercube is a steadier estimate of the mean; with 24 buyers it gives full one-way coverage, two-way coverage of 0.6419 against 0.5797 for simple random sampling, and three-way coverage of 0.1743 against 0.1671, nearly a tie, so pairs belong to a covering array. A leak in this book means only a factual error, invented scarcity, undue pressure or a dodged question; the aim of stress testing is an honest, dependable sales page, and urgency may come only from real stock and real deadlines. Parameters are illustrative; all calculations are generated by the book's programs, which call no model interface. This is not individual business advice. Chapters 1 and 2 compare the average buyer with simple random and Latin hypercube crowds, and set out covering arrays from software testing, over-tidy synthetic personas and the average-buyer fallacy. Chapter 3 covers Rubinstein's alternating offers, the power of the last round and fairness thresholds seen in experiments. Chapter 4 uses 240 parameterised demanding buyers to compare one price, a concession ladder and real tiers: the ladder hands an average discount of 0.0512 to buyers who would have paid in full, and while the average buyer passes right on the line, 0.4112 of the crowd's deals break the margin floor. Chapter 5 draws the leak discovery curve and derives stopping rules from Good's estimate and Chao's lower bound; 24 clean buyers bound the leak rate no lower than 0.1173. Chapter 6 writes detection rules for four kinds of leak and a gate that puts red lines before conversion. Chapter 7 examines Sunzi on knowing the other side and oneself and Reisswitz's war game of 1824, showing that an informed umpire lets a script with no advantage win 0.5215 of comparisons. Chapter 8 gives an eight-step loop from the owner's seat, the Synthetic Customer Coverage Sampler and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:店主在销售页与私讯话术上线前,常用一位想象中的平均买家做压测。本书检验相反的立项命题:只用平均买家会漏掉尾部破绽,拉丁超立方合成客群在同样测试量下覆盖更多人格组合(McKay、Beckman 与 Conover 1979 年刊于《技术计量学》),多轮砍价有 Rubinstein 1982 年给出的贴现均衡。核对原文后命题被修正:拉丁超立方的原始承诺是估计均值更稳;24 位买家时它一维全覆盖,两维 0.6419 对简单随机的 0.5797,三维 0.1743 对 0.1671 几乎打平,两两组合应当交给覆盖数组。本书所说的破绽只指事实错误、虚构稀缺、不当施压与答非所问,压测的目的是让销售页诚实可靠,紧迫感只能来自真实库存与真实截止。书中参数为示意;全部计算由书中程序生成,不调用任何模型接口;本书是非个人化经营建议。 第一、二章比较平均买家、简单随机与拉丁超立方客群,并陈软件测试的覆盖数组、过于整齐的合成人格与平均买家谬误。第三章讲 Rubinstein 交替出价、最后一轮的优势与实验中的公平门槛。第四章用 240 位参数化挑剔买家比较一口价、让价阶梯与真实分档:阶梯把平均 0.0512 的折扣送给本来愿付全价的买家,平均买家压线通过时,整群 0.4112 的成交跌破毛利底线。 第五章画破绽发现曲线,用古德估计与赵氏下界给出停测规则,24 位买家零发现只能把破绽率上限压到 0.1173。第六章写出四类破绽的检测规则与先过红线再看转化的闸门。第七章考辨《孙子·谋攻》的「知彼知己」与莱斯维茨 1824 年的兵棋,知情裁判会让毫无优势的话术在 0.5215 的对比中胜出。第八章给出店主视角的八步闭环、合成客群覆盖抽样器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Average Buyer's Blind Spot: Latin Hypercube Crowds and Combination Coverage — Program p11 shows that the average buyer finds nothing among twelve illustrative flaws, while a synthetic crowd of 24 finds about 7. Program p12 shows that the Latin hypercube achieves full one-way coverage, wins narrowly in two dimensions and ties simple random sampling in three. Program p13 shows that its real strength lies in estimating smooth averages, with a variance ratio as high as 336.86, and that it does nothing for corner flaws. How many buyers, then, does it take to cover every two-way combination on purpose? And will synthetic buyers turn out tidier than real ones, wearing down the very tail we want to reach? (https://5dlibrary.com/answers/en/red-blue-synthetic-customer-sandbox-bargaining-leaks/c01/) 第1章 平均买家的盲区:拉丁超立方合成客群与组合覆盖率 — 程序 p11 显示,平均买家在十二种示意破绽上一无所获,24 人的合成客群平均找到约 7 种。程序 p12 显示,拉丁超立方在一维上做到全覆盖,在两维上小胜,在三维上与简单随机打平。程序 p13 说明它真正的长处在于估计平滑的平均量,方差比可达 336.86,而对角落破绽毫无帮助。那么,要系统地覆盖两维组合,最少需要多少位买家?合成出来的买家,又会不会比真实买家更整齐,把尾部本身磨平? (https://5dlibrary.com/answers/zh/red-blue-synthetic-customer-sandbox-bargaining-leaks/c01/) - Ch.2 Dissent and Limits: Pairwise Covering Designs, Over-Tidy Synthetic Personas and the Average-Buyer Fallacy — Program p21 shows that covering all 392 two-way combinations takes a greedy covering array 51 buyers and random crowds more than two hundred. Program p22 shows that a generator shrinking the standard deviation to 0.6 cuts one-trait tail buyers by a factor of 16.3 and joint two-trait tail buyers by a factor of 267.2. Program p23 shows that when the average buyer clears the bar, 0.4505 of the crowd may still sit below it. The coverage question is settled for now. The next question is this: when a demanding buyer sits down to haggle, how much does each side walk away with? (https://5dlibrary.com/answers/en/red-blue-synthetic-customer-sandbox-bargaining-leaks/c02/) 第2章 反方与边界:两两覆盖设计、过于整齐的合成人格与平均买家谬误 — 程序 p21 显示,覆盖全部 392 个两两组合,贪心覆盖数组只要 51 人,随机客群要两百多人。程序 p22 显示,生成器把标准差缩到 0.6,单维尾部买家就少了 16.3 倍,两维联合尾部少了 267.2 倍。程序 p23 显示,平均买家过线时,整群里仍可能有 0.4505 的买家低于门槛。人格覆盖的问题讲完了,下一个问题是:一位挑剔的买家坐下来砍价,双方各自能拿到多少? (https://5dlibrary.com/answers/zh/red-blue-synthetic-customer-sandbox-bargaining-leaks/c02/) - Ch.3 Alternating Offers: Rubinstein's Discounted Equilibrium, the Last Round and Fairness Thresholds — Program p31 gives the Rubinstein split: with symmetric patience of 0.90 the opener keeps 0.5263, and as the interval between offers tends to zero the share depends only on the ratio of patience. Program p32 shows that a finite number of rounds creates a huge last-round advantage, with rounds 2 and 3 differing by 0.7200 of the surplus. Program p33 shows that once fairness thresholds are added, the equilibrium offer may persuade nobody at all. So how much margin will a crowd of demanding buyers, each with their own patience, budget and sense of fairness, carry away from different concession policies? (https://5dlibrary.com/answers/en/red-blue-synthetic-customer-sandbox-bargaining-leaks/c03/) 第3章 交替出价:Rubinstein 贴现均衡、最后一轮与公平门槛 — 程序 p31 给出 Rubinstein 均衡的切分:对称耐心 0.90 时先开口的一方拿 0.5263,出价间隔趋于零时份额只取决于耐心之比。程序 p32 显示,有限轮数会制造巨大的最后一轮优势,第 2 轮与第 3 轮相差 0.7200 个剩余。程序 p33 显示,加上公平门槛之后,均衡报价可能一个买家也说不动。那么,一群各有耐心、预算和公平感的挑剔买家,面对不同的让价策略,会把多少毛利带走? (https://5dlibrary.com/answers/zh/red-blue-synthetic-customer-sandbox-bargaining-leaks/c03/) - Ch.4 Parameterised Demanding Buyers: a Multi-Round Haggling Sandbox, Concession Ladders and the Margin Floor — Chapter 4 uses 240 parameterised demanding buyers to compare one price, a concession ladder and real tiers: the ladder hands an average discount of 0.0512 to buyers who would have paid in full, and while the average buyer passes right on the line, 0.4112 of the crowd's deals break the margin floor. (https://5dlibrary.com/answers/en/red-blue-synthetic-customer-sandbox-bargaining-leaks/c04/) 第4章 参数化挑剔买家:多轮砍价沙盘、让价阶梯与毛利底线 — 第四章用 240 位参数化挑剔买家比较一口价、让价阶梯与真实分档:阶梯把平均 0.0512 的折扣送给本来愿付全价的买家,平均买家压线通过时,整群 0.4112 的成交跌破毛利底线。 (https://5dlibrary.com/answers/zh/red-blue-synthetic-customer-sandbox-bargaining-leaks/c04/) - Ch.5 The Leak Discovery Curve: Unseen Probability, Species Estimates and Stopping Rules — Chapter 5 draws the leak discovery curve and derives stopping rules from Good's estimate and Chao's lower bound; 24 clean buyers bound the leak rate no lower than 0.1173. (https://5dlibrary.com/answers/en/red-blue-synthetic-customer-sandbox-bargaining-leaks/c05/) 第5章 破绽发现曲线:未见概率、物种数估计与停测规则 — 第五章画破绽发现曲线,用古德估计与赵氏下界给出停测规则,24 位买家零发现只能把破绽率上限压到 0.1173。 (https://5dlibrary.com/answers/zh/red-blue-synthetic-customer-sandbox-bargaining-leaks/c05/) - Ch.6 Four Kinds of Leak: Factual Errors, Invented Scarcity, Undue Pressure and Dodged Questions — Program p61 uses one fact sheet and four rules to stop four of five illustrative scripts, and shows that keyword rules can be fooled by an invented countdown. Program p62 shows that a script answering only price passes the average-buyer test while dodging seven tenths of the objections. Program p63 shows that the script with a resetting countdown converts best and is still stopped at the gate. The idea behind these checks is not new: a military classic of more than two thousand years ago and a war game of two centuries ago were both concerned with seeing oneself and the other side clearly before battle. The next chapter returns to those sources. (https://5dlibrary.com/answers/en/red-blue-synthetic-customer-sandbox-bargaining-leaks/c06/) 第6章 四类漏单破绽:事实错误、虚构稀缺、不当施压与答非所问 — 程序 p61 用一份事实表和四条规则拦下了五份示意话术中的四份,同时暴露了关键词规则会被虚构倒计时骗过。程序 p62 显示,只回答价格的话术通过平均买家测试,却对七成异议答非所问。程序 p63 显示,用了重置倒计时的话术转化最高,仍被闸门拦下。这套检查的思想并不新:两千多年前的兵书和两百年前的兵棋,都在讨论怎样在开战之前看清自己和对手。下一章回到原典。 (https://5dlibrary.com/answers/zh/red-blue-synthetic-customer-sandbox-bargaining-leaks/c06/) - Ch.7 The Classics Examined: Sunzi on Knowing the Other Side and Oneself, and Reisswitz's War Game of 1824 — Chapter 7 examines Sunzi on knowing the other side and oneself and Reisswitz's war game of 1824, showing that an informed umpire lets a script with no advantage win 0.5215 of comparisons. (https://5dlibrary.com/answers/en/red-blue-synthetic-customer-sandbox-bargaining-leaks/c07/) 第7章 原典考辨:《孙子·谋攻》的「知彼知己」与莱斯维茨 1824 年的兵棋 — 第七章考辨《孙子·谋攻》的「知彼知己」与莱斯维茨 1824 年的兵棋,知情裁判会让毫无优势的话术在 0.5215 的对比中胜出。 (https://5dlibrary.com/answers/zh/red-blue-synthetic-customer-sandbox-bargaining-leaks/c07/) - Ch.8 Closing the Loop: Extended Persona Dimensions, the Synthetic Customer Coverage Sampler and a Pre-Registered Test — Chapter 8 gives an eight-step loop from the owner's seat, the Synthetic Customer Coverage Sampler and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/red-blue-synthetic-customer-sandbox-bargaining-leaks/c08/) 第8章 实证闭环:扩展人格维度、合成客群覆盖抽样器与预注册检验 — 第八章给出店主视角的八步闭环、合成客群覆盖抽样器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/red-blue-synthetic-customer-sandbox-bargaining-leaks/c08/) ## 5DL-B-260 · The Geometry of Citation Weight in Generative Search: Retrieval and Reranking Features and the Paragraph Topology of Highly Cited Bilingual Answer Pages 中文书名:生成式搜索引用权重几何学、问答引擎检索重排特征与高被引双语直答页段落拓扑母典 DOI: pending Cite: Fan, Y. (2026). The Geometry of Citation Weight in Generative Search: Retrieval and Reranking Features and the Paragraph Topology of Highly Cited Bilingual Answer Pages (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs. It is not individual business advice. Abstract: A common belief holds that repeating keywords will make answer engines cite a page more often. This book tests that belief with the mathematics of retrieval and ranking. After checking the GEO paper by Aggarwal and colleagues, published at KDD in 2024, it corrects its founding blueprint: on the paper's benchmark keyword stuffing changed position-adjusted word count by -0.0829 relative to the baseline, quotation addition raised it by 0.4093, but a fluency rewrite that adds no information gained almost as much; the later C-SEO Bench found most such rewrites largely ineffective on new tasks, with gains shrinking toward zero as adopters multiply. The book covers only true conclusions, clear sources and compliant structured data, never disguise, hidden text or any manipulation of rankings, and labels every inferred scoring feature as an inference. Its parameters are illustrative; all calculations are generated by its programs. It is not individual business advice. Chapter 1 rebuilds the paper's table of effects and uses a proportional-share model to show that a rewrite adopted alone buys only a 0.3027 gain in share and nothing once all adopt. Chapter 2 covers BM25 saturation, cosine direction and rank fusion: one new true matching term earns nearly three times what one repeated term does. Chapter 3 covers chunking and truncation: at a cut of 128 words a conclusion placed first is read 4.5000 times as often as one placed last, and paragraph length has an interior optimum of about 60 to 130 words. Chapter 4 covers position discounts and the sag in the middle of long contexts. Chapter 5 treats verifiability and shows by simulation how fragile it is to infer feature weights from outside: leaving out length overstates the weight on statistics by 0.3252. Chapter 6 gives the question-and-answer topology of a bilingual direct-answer page and a check that four surfaces carry the same numbers. Chapter 7 examines the three gauges of the Mozi and Garfield's citation index of 1955: cumulative advantage lowers the rank correlation between quality and citations from 0.8688 to 0.6773. Chapter 8 gives an A/B test of answer-paragraph templates, the Answer-Paragraph Citability Scorer and a pre-registered protocol (524 checks per group, simulated power 0.8085). Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:常见的说法是:把关键词多写几遍,答案引擎就会更常引用你的页面。本书用检索与排序的数学检验这一说法。核对阿加瓦尔等人 2024 年发表于 KDD 的生成式引擎优化论文以后,本书修正了立项蓝图:关键词堆砌在原文基准上使位置调整词数相对下降 0.0829;加引语使之提升 0.4093,但只改写措辞的流畅度优化也提升了几乎同样多;后续的 C-SEO Bench 发现多数此类改写在新任务上基本无效,且随采用者增多而趋零。本书只讲用真实结论、清楚出处与合规结构化数据提高可引用度,不写伪装、隐藏文本或任何操纵排名的手法;书中推断的打分特征均标明为推断。书中参数均为示意;全部计算由书中程序生成。本书属非个人化经营建议。 第一章还原原文效应表,并用比例分配模型说明单独采用的改写只换来 0.3027 的份额提升、全部采用时归零。第二章讲 BM25 饱和、余弦方向与秩融合:加入一个新的真实匹配词所得是重复一次旧词的近三倍。第三章讲切块与截断:截断在 128 个词时,段首结论被读到的比例是段尾的 4.5000 倍,段落长度的内点最优约为 60 到 130 个词。第四章讲位次折扣与长上下文的中段塌陷。 第五章讲可核查性,并用模拟说明从外部推断特征权重的脆弱:漏掉长度时统计数字的权重被高估 0.3252。第六章给出双语直答页的问答单元拓扑与四面同数检查。第七章考辨《墨子·非命上》三表法与加菲尔德 1955 年的引文索引:累积优势使质量与被引次数的秩相关从 0.8688 降到 0.6773。第八章给出直答段落模板的 A/B 检验、小工具「直答段落可引用度评分器」与预注册方案(每组 524 次检查,模拟检验力 0.8085)。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Measuring Visibility: Why Keyword Stuffing Fails and Why Sources, Figures and Quotations Help — Chapter 1 rebuilds the paper's table of effects and uses a proportional-share model to show that a rewrite adopted alone buys only a 0.3027 gain in share and nothing once all adopt. (https://5dlibrary.com/answers/en/generative-search-citation-weight-answer-page-topology/c01/) 第1章 可见度的度量:关键词堆砌为何失灵,出处、数字与引语为何有效 — 第一章还原原文效应表,并用比例分配模型说明单独采用的改写只换来 0.3027 的份额提升、全部采用时归零。 (https://5dlibrary.com/answers/zh/generative-search-citation-weight-answer-page-topology/c01/) - Ch.2 The Geometry of Hybrid Scoring: BM25 Saturation, Cosine Direction and Rank Fusion — Chapter 2 covers BM25 saturation, cosine direction and rank fusion: one new true matching term earns nearly three times what one repeated term does. (https://5dlibrary.com/answers/en/generative-search-citation-weight-answer-page-topology/c02/) 第2章 混合打分几何:BM25 饱和、余弦方向与秩融合 — 第二章讲 BM25 饱和、余弦方向与秩融合:加入一个新的真实匹配词所得是重复一次旧词的近三倍。 (https://5dlibrary.com/answers/zh/generative-search-citation-weight-answer-page-topology/c02/) - Ch.3 Paragraph Length and the Point of the Cut: Chunking, Truncation and an Interior Optimum — Chapter 3 covers chunking and truncation: at a cut of 128 words a conclusion placed first is read 4.5000 times as often as one placed last, and paragraph length has an interior optimum of about 60 to 130 words. (https://5dlibrary.com/answers/en/generative-search-citation-weight-answer-page-topology/c03/) 第3章 段落长度与截断位置:切块、截断与内点最优 — 第三章讲切块与截断:截断在 128 个词时,段首结论被读到的比例是段尾的 4.5000 倍,段落长度的内点最优约为 60 到 130 个词。 (https://5dlibrary.com/answers/zh/generative-search-citation-weight-answer-page-topology/c03/) - Ch.4 The Discount by Citation Position: Cumulated Gain, Position-Adjusted Word Count and the Sag in the Middle — With $\tau=1$ the five prior shares are 0.6364, 0.2341, 0.0861, 0.0317 and 0.0117. A rewrite of the first source gains only +0.1181, because it already holds the bulk; a rewrite of the fifth gains +0.4026, close to the ceiling of 0.4093. With $\tau=2$ attention is spread more evenly: the first source gains +0.1990 and the fifth +0.3766. The last column is what the first source loses when another source rewrites: when the second source rewrites, the first loses -0.0875 or -0.0962; when the fifth rewrites, it loses only -0.0047 or -0.0232. (https://5dlibrary.com/answers/en/generative-search-citation-weight-answer-page-topology/c04/) 第4章 引用位次折扣:累积增益、位置调整词数与中段塌陷 — $\tau=1$ 时,五个来源的先验份额是 0.6364、0.2341、0.0861、0.0317、0.0117。第一名改写只提升 +0.1181,因为它已经占了大头;第五名改写提升 +0.4026,接近上限 0.4093。$\tau=2$ 时注意力分得更平,第一名的提升是 +0.1990,第五名是 +0.3766。最后一列是第一名在其他来源改写时的损失:第二名改写时,第一名损失 -0.0875 或 -0.0962,第五名改写时只损失 -0.0047 或 -0.0232。 (https://5dlibrary.com/answers/zh/generative-search-citation-weight-answer-page-topology/c04/) - Ch.5 Verifiability: Why Sources Raise the Chance of Citation, and the Trap of Inferring Features from Outside — Chapter 5 treats verifiability and shows by simulation how fragile it is to infer feature weights from outside: leaving out length overstates the weight on statistics by 0.3252. (https://5dlibrary.com/answers/en/generative-search-citation-weight-answer-page-topology/c05/) 第5章 可核查性:出处为何提高被引概率,以及从外部推断特征的陷阱 — 第五章讲可核查性,并用模拟说明从外部推断特征权重的脆弱:漏掉长度时统计数字的权重被高估 0.3252。 (https://5dlibrary.com/answers/zh/generative-search-citation-weight-answer-page-topology/c05/) - Ch.6 The Paragraph Topology of a Bilingual Direct-Answer Page: Self-Contained Paragraphs, One Set of Numbers on Four Surfaces, and Compliant Structured Data — The narrative page has five paragraphs; only 0.2000 of them stand alone, none names a source, and a single lifted paragraph has an expected value of 0.4400. The inverted-pyramid page has four paragraphs; half stand alone, for an expected value of 0.6500. The page of question-and-answer pairs has three paragraphs; every one stands alone, carries a figure, names a source and opens with the answer, for an expected value of 1.0000. The lower table varies the illustrative value kept by a paragraph that cannot stand alone: even if it keeps 0.6, the narrative page reaches only 0.6800 and the inverted pyramid 0.8000, while the pairs stay at 1.0000. (https://5dlibrary.com/answers/en/generative-search-citation-weight-answer-page-topology/c06/) 第6章 双语直答页的段落拓扑:自足段落、四面同数与合规结构化数据 — 叙事式页面有五段,只有 0.2000 能单独成立,没有一段写明出处,单独拎走一段的期望价值是 0.4400。倒金字塔式页面有四段,一半能单独成立,期望价值 0.6500。问答对式页面有三段,每一段都能单独成立,都带数字、都写明出处、首句都是答案,期望价值 1.0000。下半张表变动「不能单独成立的段落还剩多少价值」这个示意参数:即使它还剩 0.6,叙事式页面也只有 0.6800,倒金字塔 0.8000,问答对式始终是 1.0000。 (https://5dlibrary.com/answers/zh/generative-search-citation-weight-answer-page-topology/c06/) - Ch.7 Classical Sources East and West: the Three Gauges of the Mozi and Garfield's Citation Index — Chapter 7 examines the three gauges of the Mozi and Garfield's citation index of 1955: cumulative advantage lowers the rank correlation between quality and citations from 0.8688 to 0.6773. (https://5dlibrary.com/answers/en/generative-search-citation-weight-answer-page-topology/c07/) 第7章 东西方原典考辨:《墨子·非命上》三表法与加菲尔德的引文索引 — 第七章考辨《墨子·非命上》三表法与加菲尔德 1955 年的引文索引:累积优势使质量与被引次数的秩相关从 0.8688 降到 0.6773。 (https://5dlibrary.com/answers/zh/generative-search-citation-weight-answer-page-topology/c07/) - Ch.8 Closing the Loop: an A/B Test of Answer-Paragraph Templates, the Citability Scorer and a Pre-Registered Test — Chapter 8 gives an A/B test of answer-paragraph templates, the Answer-Paragraph Citability Scorer and a pre-registered protocol (524 checks per group, simulated power 0.8085). Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/generative-search-citation-weight-answer-page-topology/c08/) 第8章 实证闭环:直答段落模板 A/B、可引用度评分器与预注册检验 — 第八章给出直答段落模板的 A/B 检验、小工具「直答段落可引用度评分器」与预注册方案(每组 524 次检查,模拟检验力 0.8085)。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/generative-search-citation-weight-answer-page-topology/c08/) ## 5DL-B-261 · Grid Search over Knowledge-Base Chunk Boundaries: Overlap Ratios and the Hybrid Weighting of Vector and Term-Frequency Retrieval 中文书名:知识库分块边界网格寻优、重叠率与向量词频混合检索配比母典 DOI: pending Cite: Fan, Y. (2026). Grid Search over Knowledge-Base Chunk Boundaries: Overlap Ratios and the Hybrid Weighting of Vector and Term-Frequency Retrieval (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Implementing BM25 and a concept-level bag-of-words cosine in standard-library programs on fixed-seed synthetic corpora, the book corrects the blueprint in three places. Parameters are illustrative and all numbers come from the book's programs, which read and write no file and connect to no live knowledge base. Abstract: Building a knowledge base means fixing three numbers: chunk length, overlap between neighbouring chunks, and the blend weight between vector and term-frequency retrieval. Against the folk wisdom that smaller is more precise and one blend suffices, the commissioning claim says recall traces an inverted U against chunk length, overlap has diminishing returns, and the best blend depends on the query type. Implementing BM25 and a concept-level bag-of-words cosine in standard-library programs on fixed-seed synthetic corpora, the book corrects the blueprint in three places. The cited 2009 survey by Robertson and Zaragoza sets out BM25 but does not test chunking; the direct evidence comes from passage retrieval and retrieval granularity. The inverted U appears only under a fixed reading budget: with a fixed Top K of 5 recall rises steadily to 0.809, while a budget of 192 tokens gives a peak of 0.689 at a chunk length of 48. Overlap pays until a saturation point and then stops: with chunks of 32 and answers of 24 tokens, the chance that one chunk holds eighty per cent of an answer is 0.531 without overlap and 1.000 at half overlap, with an index 2.00 times as large. Parameters are illustrative and all numbers come from the book's programs, which read and write no file and connect to no live knowledge base. This is non-individual business advice; changes to a live knowledge base need an offline evaluation and a manual release. Chapter 1 draws the inverted U and shows its peak following the answer span. Chapter 2 writes BM25 from scratch; its length normalisation idles under equal-length chunks. Chapter 3 treats cosine dilution in long chunks: term frequency recognises identifiers, vectors recognise paraphrase. Chapter 4 gives exact odds of a boundary cut and a closed formula for the saturation point of overlap. Chapter 5 sweeps the blend weight: the best value is 0.0 for exact-ID queries and 1.0 for paraphrase queries, and the illustrative 0.7 / 0.3 baseline regrets 0.098 on an all-paraphrase workload; routing by type at a classification error rate of 0.10 gives held-out recall of 0.928, already below the 0.938 of a tuned global weight. Chapter 6 runs the joint grid and finds an optimism of 0.028 when the champion is picked from only 30 labelled queries. Chapter 7 examines Cook Ding carving the ox in the Zhuangzi and Plato's division at the natural joints in the Phaedrus: cutting by paragraph gives zero overlap and zero breakage, fixed windows of 64 at an overlap of one half still break 0.401 of the answers, and misread joints quickly erase this advantage. Chapter 8 gives a quarterly tuning routine for knowledge-base chunking, the distilled "Chunk-and-Blend Grid Optimizer", a pre-registered test with sample sizes from Connor's formula (the exact McNemar test reaches an actual power of 0.762) and three conditions that would overturn the core claim. Boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. 摘要:知识库建库时要填三个数:块长、相邻块的重叠率、向量与词频的混合权重。流行说法是「块越小越准」与「混合检索给一个固定配比就行」。立项命题说:召回随块长呈倒 U,重叠率超过一定比例收益递减,最优混合权重依查询类型而变。本书在固定种子的合成语料上用纯标准库程序实现 BM25 与概念层面的词袋余弦,把命题写成可计算的模型,并修正蓝图三处:规格所引罗伯逊与萨拉戈萨 2009 年综述只阐述 BM25,不检验分块,直接证据来自段落检索与检索粒度研究;倒 U 只在固定阅读预算下出现,每次固定取 5 块时召回随块长单调升到 0.809,固定 192 个词元预算时在块长 48 处达到峰值 0.689;重叠的收益不是平滑递减而是到饱和点即止,块长 32、答案跨度 24 时,装下八成答案的概率在不重叠时为 0.531,重叠一半时为 1.000,索引为 2.00 倍。书中参数为示意;全部计算由书中程序生成;程序不读写任何文件,不连接任何线上知识库;本书是非个人化经营建议,任何线上知识库参数的改动须经离线评测并由主事者人工复核放行。 第一章在固定预算下画出倒 U 并说明峰位随答案跨度移动。第二章从头写 BM25,说明长度归一在等长分块里几乎闲置。第三章写词袋余弦与长块稀释,说明词频一路认编号、向量一路认改写。第四章精确枚举边界切断概率,给出重叠率饱和点的闭合公式。第五章扫描混合权重:编号型查询最优 0.0,改写型最优 1.0,示意基线 0.7 / 0.3 在全改写型时遗憾 0.098;按类型分流在判别出错率为 0.10 时留出召回 0.928,已低于调好的全局权重 0.938。第六章做联合网格,只用 30 条标注查询挑冠军时乐观偏差 0.028。 第七章考辨《庄子·养生主》庖丁解牛与柏拉图《斐德若篇》依自然关节切分:顺着段落切零重叠零折断,块长 64、重叠一半的固定窗口仍折断 0.401 的答案,而关节识别出错会迅速吃掉这一优势。第八章给出知识库分块参数的季度寻优流程、蒸馏小工具「分块配比网格寻优器」、按康纳公式取样本量的预注册检验(精确麦克尼马尔检验的实测把握度 0.762)与推翻本书核心命题的三个条件。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Inverted U: Chunk Length and Recall under a Reading Budget — With a span of 12 the best chunk length is 48, which is 4.00 times the span; with a span of 24 the best length is still 48, now 2.00 times the span; with a span of 48 it moves to 96, again 2.00 times. With a span of 48, recall at a chunk length of 8 is 0.000: no chunk can hold eighty per cent of the evidence, so no ranking, however good, can help. With a span of 12, small chunks fare much better, reaching 0.681 at a length of 8, but this is still below the 0.869 reached at 48. (https://5dlibrary.com/answers/en/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c01/) 第1章 倒 U:阅读预算下的块长与召回 — 事实跨度为 12 时,最优块长是 48,是跨度的 4.00 倍;跨度为 24 时,最优块长仍是 48,是跨度的 2.00 倍;跨度为 48 时,最优块长移到 96,同样是 2.00 倍。跨度为 48 时,块长 8 的召回是 0.000:任何一块都不可能装下八成证据,排得再准也没有用。跨度为 12 时,小块的处境好得多,块长 8 的召回有 0.681,但仍低于块长 48 的 0.869。 (https://5dlibrary.com/answers/zh/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c01/) - Ch.2 BM25 from Scratch: Inverse Document Frequency, Saturation and Length Normalisation — Chapter 2 writes BM25 from scratch; its length normalisation idles under equal-length chunks. (https://5dlibrary.com/answers/en/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c02/) 第2章 从头写 BM25:逆文档频率、词频饱和与长度归一 — 第二章从头写 BM25,说明长度归一在等长分块里几乎闲置。 (https://5dlibrary.com/answers/zh/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c02/) - Ch.3 Bag-of-Words Cosine and Dilution in Long Chunks: Which Retriever Recognises Which Query — The wrong chunks are more instructive. At a chunk length of 12, the strongest wrong chunk has a mean cosine of 0.584, above the 0.573 of the gold chunk, so the margin is negative. A gold chunk of 12 tokens holds only half the fact, the other half sits in the neighbouring chunk, and among thousands of small chunks in the base some always happen to gather a few key concepts. At a chunk length of 24 the gold chunk leads by the widest margin, 0.085; from 96 onwards the margin is back near zero. The share of queries in which the gold chunk beats every wrong chunk therefore also traces an inverted U: 0.528 at a length of 12, a peak of 0.787 at 24, and back down to 0.506 at 192. (https://5dlibrary.com/answers/en/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c03/) 第3章 词袋余弦与长块稀释:两路检索各认哪一类查询 — 更值得看的是错块。块长 12 时,最强错块的平均余弦 0.584 反而高于金块的 0.573,差值为负:12 个词元的金块只装得下事实的一半,另一半在隔壁块里,而全库成千上万个小块中,总有一些恰好凑齐了几个要点概念。块长 24 时金块领先最多,差值 0.085;块长 96 以后差值又回到零附近。金块击败所有错块的比例因此也呈倒 U:块长 12 时 0.528,块长 24 时达到 0.787,块长 192 时回落到 0.506。 (https://5dlibrary.com/answers/zh/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c03/) - Ch.4 Overlap: Exact Odds of a Boundary Cut and the Knee of Diminishing Returns — With chunks of 32, recall rises from 0.584 without overlap to 0.970 at an overlap of one half, in line with the saturation point predicted by p41. At an overlap of three quarters the number of chunks grows from 2400 to 8400, recall falls slightly to 0.964, and the duplicated share of returned text grows from 0.120 to 0.365. With chunks of 64, recall reaches 0.873 at an overlap of 0.375 and falls after it, to 0.851 at one half and 0.824 at three quarters, with a duplicated share of 0.331. (https://5dlibrary.com/answers/en/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c04/) 第4章 重叠率:边界切断的精确概率与收益递减的拐点 — 块长 32 时,召回从不重叠的 0.584 一路升到重叠一半的 0.970,与 p41 预测的饱和点吻合;重叠到四分之三,块数从 2400 涨到 8400,召回反而略降到 0.964,取回文字里重复的比例从 0.120 升到 0.365。块长 64 时,召回在重叠 0.375 处达到 0.873,之后下降,重叠一半时 0.851,四分之三时 0.824,重复比例 0.331。 (https://5dlibrary.com/answers/zh/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c04/) - Ch.5 The Blend: How the Best Weight Moves with the Query Mix — Chapter 5 sweeps the blend weight: the best value is 0.0 for exact-ID queries and 1.0 for paraphrase queries, and the illustrative 0.7 / 0.3 baseline regrets 0.098 on an all-paraphrase workload; routing by type at a classification error rate of 0.10 gives held-out recall of 0.928, already below the 0.938 of a tuned global weight. (https://5dlibrary.com/answers/en/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c05/) 第5章 混合配比:最优权重随查询构成移动 — 第五章扫描混合权重:编号型查询最优 0.0,改写型最优 1.0,示意基线 0.7 / 0.3 在全改写型时遗憾 0.098;按类型分流在判别出错率为 0.10 时留出召回 0.928,已低于调好的全局权重 0.938。 (https://5dlibrary.com/answers/zh/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c05/) - Ch.6 The Joint Grid and the Winner's Curse: Why the Held-Out Test Cannot Be Skipped — Chapter 6 runs the joint grid and finds an optimism of 0.028 when the champion is picked from only 30 labelled queries. (https://5dlibrary.com/answers/en/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c06/) 第6章 联合网格与赢家诅咒:留出检验为什么不可省 — 第六章做联合网格,只用 30 条标注查询挑冠军时乐观偏差 0.028。 (https://5dlibrary.com/answers/zh/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c06/) - Ch.7 Cook Ding and the Phaedrus: Two Ancient Arguments for Cutting at the Natural Joints — Chapter 7 examines Cook Ding carving the ox in the Zhuangzi and Plato's division at the natural joints in the Phaedrus: cutting by paragraph gives zero overlap and zero breakage, fixed windows of 64 at an overlap of one half still break 0.401 of the answers, and misread joints quickly erase this advantage. (https://5dlibrary.com/answers/en/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c07/) 第7章 庖丁与斐德若:依自然关节切分的两种古老主张 — 第七章考辨《庄子·养生主》庖丁解牛与柏拉图《斐德若篇》依自然关节切分:顺着段落切零重叠零折断,块长 64、重叠一半的固定窗口仍折断 0.401 的答案,而关节识别出错会迅速吃掉这一优势。 (https://5dlibrary.com/answers/zh/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c07/) - Ch.8 The Quarterly Tuning Loop and a Plan for Refutation — Chapter 8 gives a quarterly tuning routine for knowledge-base chunking, the distilled "Chunk-and-Blend Grid Optimizer", a pre-registered test with sample sizes from Connor's formula (the exact McNemar test reaches an actual power of 0.762) and three conditions that would overturn the core claim. Boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c08/) 第8章 季度寻优闭环与证伪方案 — 第八章给出知识库分块参数的季度寻优流程、蒸馏小工具「分块配比网格寻优器」、按康纳公式取样本量的预注册检验(精确麦克尼马尔检验的实测把握度 0.762)与推翻本书核心命题的三个条件。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/knowledge-base-chunking-overlap-hybrid-retrieval-grid-search/c08/) ## 5DL-B-262 · A Zero-Leak Socratic Examiner: Server-Side Isolation of Answers and Bayesian Mastery Probing Trees 中文书名:零泄题苏格拉底闯关考官、服务端答案物理隔离与掌握度贝叶斯追问树母典 DOI: pending Cite: Fan, Y. (2026). A Zero-Leak Socratic Examiner: Server-Side Isolation of Answers and Bayesian Mastery Probing Trees (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations are generated by the book's programs, which call no model interface, contain no keys or server addresses and include no concrete extraction prompts. This is not individual business advice. Abstract: When an online academy uses a language model as a gatekeeping examiner, it often writes the model answer into the prompt and instructs the model never to reveal it. This book tests the opposite founding claim: with the answer in context, the cumulative probability of extraction, 1-(1-p)^k, tends to 1 as attempts accumulate and an instruction changes only the speed; only when the answer stays out of the context is content leakage zero; and mastery is booked with the Bayesian knowledge tracing of Corbett and Anderson, published in User Modeling and User-Adapted Interaction in 1995. Checking the originals corrects the claim: the curve slows when attackers differ and speeds up when they learn; grading itself is an oracle that reveals at most one bit per verdict, the model may solve an item by itself, and learners share the items they answered correctly. Parameters are illustrative; all calculations are generated by the book's programs, which call no model interface, contain no keys or server addresses and include no concrete extraction prompts. This is not individual business advice. Chapters 1 and 2 quantify why instructions fail: in an illustrative course of 40 items with 3 attempts per item, a class of only 30 leaks at least one item with probability 0.9727 when p is 0.001. They then draw information-flow graphs of four examiner architectures and compute the bit bound of the grading oracle and the cost of attempt caps. Chapters 3 and 4 cover the four-parameter update of knowledge tracing, the mastery line and identifiability, and show that when a fifth of learners hold the answers, false certificates make up 0.2534 of all certificates, while a fit after the fact reads this as a stronger intake. Chapter 5 builds probing trees from information gain, whose benefit comes mainly from the stopping rule: 7.71 items on average against 9.59 for a fixed order, with better accuracy. Chapter 6 computes correlated failure in defence in depth, digest-only output checks and item exposure. Chapter 7 examines "no opening without striving" in the Analects and the slave-boy geometry lesson in Plato's Meno. Chapter 8 gives the owner's eight-step red-team loop, the distilled Mastery Probing Tree and a pre-registered test; vouching for a term's safety of a build that keeps the answer in context would take 1437950 clean red-team attempts. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:在线学堂用语言模型做闯关考官时,常把标准答案写进提示词,再叮嘱模型不许说出。本书检验相反的立项命题:答案在上下文里,套话的累计泄露概率 1-(1-p)^k 随尝试次数趋近 1,叮嘱只能改变速度;只有答案不进入上下文,内容泄露才为零;掌握度改用 Corbett 与 Anderson 1995 年刊于《用户建模与用户自适应交互》的贝叶斯知识追踪记账。核对原文后命题被修正:攻击者各不相同时曲线变慢、会学习时变快;判分本身是每次至多一比特的预言机,模型还可能自己把题解出来,学员也会分享自己答对的题。书中参数为示意;全部计算由书中程序生成,不调用任何模型接口,不含任何密钥或服务器地址,也不收录具体的套话提示词;本书是非个人化经营建议。 第一、二章量化叮嘱为何无效,在每题 3 次尝试、40 道题的示意课程里,p 等于 0.001 时一个 30 人的小班至少漏掉一道题的概率是 0.9727;再画出四种考官架构的信息流图,计算判分预言机的比特上限与尝试上限的代价。第三、四章讲知识追踪的四参数更新、掌握线与可识别性,并说明五分之一的学员拿到答案时,假证书占全部证书的 0.2534,而事后拟合把它解释成生源更好。 第五章用信息增益搭追问树,收益主要来自停问规则:平均 7.71 题对固定顺序的 9.59 题,判对率反而更高。第六章计算纵深防御的相关失效、只存摘要的输出检查与题库暴露。第七章考辨《论语·述而》的「不愤不启」与柏拉图《美诺篇》的童奴几何问答。第八章给出主事者视角的八步红队闭环、蒸馏小工具「掌握度追问树」与预注册检验;要为答案在上下文的构建担保一学期安全,需要 1437950 次干净的红队尝试。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Telling It Not to Tell: Answers in Context and the Cumulative Leak Probability — Chapters 1 and 2 quantify why instructions fail: in an illustrative course of 40 items with 3 attempts per item, a class of only 30 leaks at least one item with probability 0.9727 when p is 0.001. They then draw information-flow graphs of four examiner architectures and compute the bit bound of the grading oracle and the cost of attempt caps. (https://5dlibrary.com/answers/en/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c01/) 第1章 叮嘱无效:上下文里的答案与累计泄露概率 — 第一、二章量化叮嘱为何无效,在每题 3 次尝试、40 道题的示意课程里,p 等于 0.001 时一个 30 人的小班至少漏掉一道题的概率是 0.9727;再画出四种考官架构的信息流图,计算判分预言机的比特上限与尝试上限的代价。 (https://5dlibrary.com/answers/zh/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c01/) - Ch.2 Physical Isolation: Server-Side Grading, Information-Flow Graphs and the One-Bit Oracle — Program p21 draws four architectures as information-flow graphs. Only architecture C, with grading on the server and only bits for the model, pinches every path to the browser down to one bit, and a single hint writer that reads the answer is enough to turn it back into D. Program p22 shows that grading is itself an oracle that reveals at most one bit per verdict and offers almost no protection to small-space items such as four-option questions. Program p23 shows that caps and parallel items hold the oracle down, but that small-space items cannot guard against guessing and against false lockouts at the same time. With the content channel closed, the examiner faces one question: holding nothing but a string of rights and wrongs, on what grounds can it judge that a learner has mastered the skill? The next chapter introduces Bayesian knowledge tracing. (https://5dlibrary.com/answers/en/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c02/) 第2章 物理隔离:服务端判分、信息流图与判分预言机的一比特 — 程序 p21 把四种架构画成信息流图,只有服务器端判分、模型只拿比特的架构 C 把每一条到达浏览器的路径都卡在一个比特上,而一个读取答案的提示生成器就能把它退回到 D。程序 p22 说明判分本身是一台预言机,每次最多透露一个比特,对四选一这样的小空间题几乎没有保护。程序 p23 说明尝试上限与平行题能压住预言机,但小空间题无法同时照顾防猜与防误伤。内容通道关上之后,考官还剩下一个问题:手里只有一串「对」和「错」,它凭什么判断学员已经掌握?下一章引入贝叶斯知识追踪。 (https://5dlibrary.com/answers/zh/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c02/) - Ch.3 Bayesian Knowledge Tracing: Four Parameters, the Mastery Line and Identifiability — Chapters 3 and 4 cover the four-parameter update of knowledge tracing, the mastery line and identifiability, and show that when a fifth of learners hold the answers, false certificates make up 0.2534 of all certificates, while a fit after the fact reads this as a stronger intake. (https://5dlibrary.com/answers/en/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c03/) 第3章 贝叶斯知识追踪:四个参数、掌握线与可识别性 — 第三、四章讲知识追踪的四参数更新、掌握线与可识别性,并说明五分之一的学员拿到答案时,假证书占全部证书的 0.2534,而事后拟合把它解释成生源更好。 (https://5dlibrary.com/answers/zh/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c03/) - Ch.4 How Leaks Forge Mastery: Guess Rates, False Certificates and the Blind Spot of Fitting — Program p41 shows that a leak lets learners who hold the answers pass the examiner with certainty, and that a leak reaching a fifth of the class is enough to make one certificate in four false. Program p42 shows that a fit after the fact cannot see the leak: it explains it as a stronger intake with fewer slips, and the curve still fits. Detection therefore relies on structures designed when items are written, namely parallel variants, delayed transfer items and fixed guess rates, and all of them are repairs that come after prevention. The examiner's proper job is to judge mastery quickly and accurately without leaking anything; so, given a string of right-and-wrong bits, which question is the most valuable to ask next? The next chapter builds the probing tree. (https://5dlibrary.com/answers/en/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c04/) 第4章 泄题如何伪造掌握:猜对率、假证书与拟合的盲区 — 程序 p41 说明,泄题让拿到答案的人以百分之百的概率通过考官,百分之二十的泄题率就足以让四张证书里有一张是假的。程序 p42 说明,事后拟合看不出泄题:它把泄题解释成生源更好、失手更少,曲线依旧吻合。检测因此要靠平行变体、延迟迁移题与固定猜对率这些出题时就设计好的结构,而它们都只是预防之后的补救。考官的本职,是在不泄题的前提下尽快、尽准地判断掌握,那么在一串对错比特之间,下一道题该问什么才最有价值?下一章搭建追问树。 (https://5dlibrary.com/answers/zh/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c04/) - Ch.5 The Probing Tree: Information-Gain Item Selection, Greedy Trees and Stopping Rules — Chapter 5 builds probing trees from information gain, whose benefit comes mainly from the stopping rule: 7.71 items on average against 9.59 for a fixed order, with better accuracy. (https://5dlibrary.com/answers/en/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c05/) 第5章 追问树:信息增益选题、贪心树与停问规则 — 第五章用信息增益搭追问树,收益主要来自停问规则:平均 7.71 题对固定顺序的 9.59 题,判对率反而更高。 (https://5dlibrary.com/answers/zh/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c05/) - Ch.6 Defence in Depth against Prompt Extraction: Correlated Layer Failure, Digest-Only Output Checks and Item Exposure — Program p61 shows that the layers of defence in depth can be multiplied only when they fail independently, whereas the failure mechanisms revealed by current research make them fail together; only isolation moves the limit from 1 to 0. Program p62 demonstrates a digest-only output check: in the isolated architecture it guards against regressions and against the model computing the answer by itself, and it stops written variants but not rewording. Program p63 estimates the one channel isolation cannot close, learners sharing the items they know they answered correctly, and shows the role of variants and of refreshing the bank. With the account settled, an older question comes next: how did Socrates and Confucius, long before any server, handle asking without telling? The next chapter returns to the classical sources. (https://5dlibrary.com/answers/en/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c06/) 第6章 防套话的纵深架构:分层失效的相关性、只存摘要的输出检查与题库暴露 — 程序 p61 说明,纵深防御的各层只有在独立失效时才能连乘,而现有研究揭示的失效机制恰恰会让它们一起失效;只有隔离把极限从 1 变成 0。程序 p62 演示了只存摘要的输出检查:它在隔离架构里负责防倒退和防模型自己算出答案,能拦书写变体,拦不住改写。程序 p63 估算了隔离之后唯一关不上的通道,即学员对自己答对题目的分享,并说明变体与题库更新的作用。账算完了,接下来要回答一个更古老的问题:苏格拉底和孔子在没有服务器的年代,是怎样处理「问而不告」的?下一章回到原典。 (https://5dlibrary.com/answers/zh/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c06/) - Ch.7 Classical Sources: "No Opening without Striving" in the Analects and the Slave-Boy Geometry of Plato's Meno — Program p71 writes "no opening without striving" and "one corner shown, three returned" as rules that can be computed: prompting after the attempt preserves the evidence, three correct transfer items give a posterior of 0.9891, and a repeated worked example is worth nothing as evidence. Program p72 recomputes the doubled square of the Meno and shows that leading questions carry almost no weight as evidence of mastery, so Socrates's claim to ask without teaching does not stand in terms of information. The direction both teachers point in agrees with this book's architecture: the answer stays outside the student, and the evidence comes only from the student's own responses. The final chapter asks what red-team procedure an owner can carry out tomorrow, how large the samples must be, and what result would overturn the book's central claim. (https://5dlibrary.com/answers/en/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c07/) 第7章 原典考辨:《论语·述而》的「不愤不启」与柏拉图《美诺篇》的童奴几何问答 — 程序 p71 把「不愤不启」「举一隅不以三隅反」写成可算的规则:先尝试后点拨保住了证据,三道迁移题全对的后验 0.9891,而重复展示过的例题毫无证据价值。程序 p72 复算了《美诺篇》的倍积正方形,并说明引导性问句作为掌握证据几乎没有分量,苏格拉底「只问不教」的说法在信息意义上站不住。两位老师给出的方向,与本书的架构一致:答案留在学生之外,证据只来自学生自己的作答。最后一章要回答的是:主事者明天就能执行的红队测试流程是什么,样本量要多大,什么结果会推翻本书的核心命题? (https://5dlibrary.com/answers/zh/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c07/) - Ch.8 The Empirical Loop: Red-Teaming the Academy Examiner for Leaks, the Mastery Probing Tree and a Pre-Registered Test — Chapter 8 gives the owner's eight-step red-team loop, the distilled Mastery Probing Tree and a pre-registered test; vouching for a term's safety of a build that keeps the answer in context would take 1437950 clean red-team attempts. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c08/) 第8章 实证闭环:学堂考官泄题红队测试、掌握度追问树与预注册检验 — 第八章给出主事者视角的八步红队闭环、蒸馏小工具「掌握度追问树」与预注册检验;要为答案在上下文的构建担保一学期安全,需要 1437950 次干净的红队尝试。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/zero-leak-socratic-examiner-server-side-answers-mastery-tree/c08/) ## 5DL-B-263 · Empirical Retirement of Live Rules: Wiring the Sequential Probability Ratio Test into Real Conversion Streams to Promote, Demote and Retire Copy Rules 中文书名:线上规则实证淘汰、序贯概率比检验接入真实转化流与新文案规则转正降级母典 DOI: pending Cite: Fan, Y. (2026). Empirical Retirement of Live Rules: Wiring the Sequential Probability Ratio Test into Real Conversion Streams to Promote, Demote and Retire Copy Rules (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: A business team often has dozens of copy, button and private-message rules on test at once. Common practice is to glance at the dashboard every week and promote whatever is significant, or, conversely, to believe that reliability means waiting for a full fixed sample. This book tests both defaults and wires Wald's sequential probability ratio test into a lifecycle state machine for rules, so that every rule passes through drafting, queueing, trial, promotion, watch, demotion and retirement under one discipline. All calculations use only the Python standard library. Conversion rates, traffic and weekly counts are illustrative; every calculation is generated by the book's programs; the book offers general business method, not business, legal or investment advice; every promotion, demotion and retirement must be released by the rule's owner, and urgency may come only from real stock and real deadlines. Chapter 1 states and quantifies the claim: ten weekly two-sided looks give a false-alarm rate of 0.1934 and a year of looks 0.3235; in a queue of 40 rules peeking promotes 5.58 useless rules and the sequential test 0.73. The preface corrects the blueprint: one in five is passed only at the eleventh look; "about half" holds only near the design points, where the exact values are 0.464 and 0.632 of the fixed sample, rising to 0.748 for middling rules; and optimality was proved by Wald and Wolfowitz in 1948. Chapter 2 draws the boundaries as two parallel lines in the count plane. Chapter 3 gives exact error rates and the cost of a cap. Chapter 4 shows that weekly batches make the test more conservative, while re-estimating the baseline every week lifts false promotions to 0.1138. Chapter 5 checks an eight-state rule machine exhaustively, watches for wear with Page's cumulative sum, and uses an absorbing chain to show that a worn rule stays live 8.54 weeks on average. Chapter 6 shows that splitting alpha among 100 rules lengthens verdicts on useless rules by a ratio of only 1.12. Chapter 7 examines checking by trial in the Han Feizi and Wald's Sequential Analysis. Chapter 8 wires the weekly radar's verdicts into the state machine and gives the distilled tool, the Rule Promotion and Demotion Judge, and a pre-registered A/A calibration: 100 A/A rules, with 11 or more promotions overturning the book. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:一个经营团队同时在测的文案、按钮与私讯模板规则常有几十条。常识的做法是每周看一眼后台,显著就转正;或者反过来,认为要可靠就必须等够固定样本。本书检验这两个默认,并把 Wald 的序贯概率比检验接进一台规则生命周期状态机,让每条规则按同一套纪律经历起草、排队、试用、转正、观察、降级与退役。全部计算只用 Python 标准库。书中的转化率、流量与每周计数均为示意;全部计算由书中程序生成;本书属非个人化经营方法,不构成经营、法律或投资建议,规则的转正、降级与退役须由主事者人工放行,紧迫感只能来自真实库存与真实截止。 第一章正面陈述并量化命题:每周双侧检验看 10 次,误报率 0.1934,看满一年 0.3235;40 条规则的队列里,偷看误转正 5.58 条,序贯检验 0.73 条。序中修正立项蓝图:两成以上要看到第 11 次才越过;「约一半」只在设计点附近成立,精确值为固定样本的 0.464 与 0.632,平庸规则处升到 0.748;最优性由 Wald 与沃尔福威茨 1948 年证明。第二章把边界画成计数平面上的两条平行直线。第三章用精确递推给出真实错误率与截断代价。第四章说明周批次让检验更保守,而每周重估基线会把误转正率抬到 0.1138。 第五章用穷举检查八状态的规则状态机,用 Page 累积和监视磨损,以吸收链算出磨损规则平均还在线 8.54 周。第六章显示平分 α 给 100 条规则时,无效规则的判决周数只增长 1.12 倍。第七章考辨《韩非子·显学》的参验与 Wald 的《序贯分析》。第八章把每周雷达的判决接进状态机,给出蒸馏小工具「规则转正降级判定器」与预注册的 A/A 校准:100 条 A/A 规则,不少于 11 条转正即推翻本书。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A Weekly Glance: How a Mediocre Copy Rule Gets Promoted by Mistake — Chapter 1 states and quantifies the claim: ten weekly two-sided looks give a false-alarm rate of 0.1934 and a year of looks 0.3235; in a queue of 40 rules peeking promotes 5.58 useless rules and the sequential test 0.73. The preface corrects the blueprint: one in five is passed only at the eleventh look; "about half" holds only near the design points, where the exact values are 0.464 and 0.632 of the fixed sample, rising to 0.748 for middling rules; and optimality was proved by Wald and Wolfowitz in 1948. (https://5dlibrary.com/answers/en/live-rule-retirement-sprt-promote-demote-copy-rules/c01/) 第1章 每周看一眼:平庸文案怎样被误转正 — 第一章正面陈述并量化命题:每周双侧检验看 10 次,误报率 0.1934,看满一年 0.3235;40 条规则的队列里,偷看误转正 5.58 条,序贯检验 0.73 条。序中修正立项蓝图:两成以上要看到第 11 次才越过;「约一半」只在设计点附近成立,精确值为固定样本的 0.464 与 0.632,平庸规则处升到 0.748;最优性由 Wald 与沃尔福威茨 1948 年证明。 (https://5dlibrary.com/answers/zh/live-rule-retirement-sprt-promote-demote-copy-rules/c01/) - Ch.2 Wald's Two Straight Lines: Boundaries, the OC Function and the Mean Sample Size — Wald's sequential test writes evidence as a visitor-by-visitor sum, and its two boundaries are parallel lines in the count plane with slope 0.03291 and a gap of 22.97 conversions, which anyone can compare with the cumulative visitors and conversions in a weekly report. The formulas of p22 show that the saving of "about half" holds at the baseline, 0.460, reaches only 0.616 at the target, and that middling rules between the two push the mean sample size up to 0.728 of the fixed sample. All of these are approximations that neglect the overshoot. Computed exactly, visitor by visitor, what are the true error rates and the true mean sample size? How long is the tail of the sample-size distribution, and when should a pipeline call time? Chapter 3 answers with exact recursions. (https://5dlibrary.com/answers/en/live-rule-retirement-sprt-promote-demote-copy-rules/c02/) 第2章 Wald 的两条直线:边界、OC 函数与平均样本量 — Wald 的序贯检验把证据写成逐个访客的累加,两条边界在计数平面上是斜率 0.03291、间距 22.97 个转化的两条平行直线,任何人都能拿周报里的累计访客与转化数对照它们。p22 的公式显示,「约一半」的节省在基线处成立,0.460;在目标处只到 0.616;介于两者之间的平庸规则,平均样本量升到固定样本的 0.728。这些都是忽略越界量的近似。那么,按逐个访客精确计算,真实的错误率与平均样本量是多少?样本量的分布有多长的尾巴,一条流水线又该在什么时候喊停?第三章用精确递推回答。 (https://5dlibrary.com/answers/zh/live-rule-retirement-sprt-promote-demote-copy-rules/c02/) - Ch.3 Exact Recursions: the Long Tail of the Sample Size and the Cost of a Cap — Exact recursion adds three things to Wald's formulas. First, the true error rates are 0.0470 and 0.2000, both inside the bounds 0.0625 and 0.2105. Second, the mean sample size is 0.464 and 0.632 of the fixed sample at the two design points and rises to 0.748 for middling rules, nearly a quarter of which run longer than a fixed sample. Third, a cap has a price: removing all open rules at a nine-week cap leaves a power of only 0.6548, while settling them by sign lifts the type I error rate to 0.0876. All of this assumes a look after every visitor and an exactly known baseline. Yet the weekly radar looks once a week, and the baseline must be estimated from the control. Which way do batches and the source of the baseline push these numbers? Chapter 4 connects the test to real conversion streams. (https://5dlibrary.com/answers/en/live-rule-retirement-sprt-promote-demote-copy-rules/c03/) 第3章 精确递推:样本量的长尾与截断的代价 — 精确递推给出了 Wald 公式之外的三件事。第一,真实错误率是 0.0470 与 0.2000,都在上界 0.0625 与 0.2105 之内。第二,平均样本量在两个设计点上是固定样本的 0.464 与 0.632,平庸规则那里升到 0.748,而且有接近四分之一的平庸规则会比固定样本跑得更久。第三,截断有代价:到期一律淘汰,9 周上限的功效只有 0.6548;到期按符号判,第一类错误率升到 0.0876。这些都假定每来一个访客就看一次,基线也确切已知。可是每周雷达是一周看一次,基线也要从对照组估计。批次与基线来源,会把这些数字推向哪边?第四章把检验接进真实的转化流。 (https://5dlibrary.com/answers/zh/live-rule-retirement-sprt-promote-demote-copy-rules/c03/) - Ch.4 Connecting Real Conversion Streams: Weekly Batches and Where the Baseline Comes From — Chapter 4 shows that weekly batches make the test more conservative, while re-estimating the baseline every week lifts false promotions to 0.1138. (https://5dlibrary.com/answers/en/live-rule-retirement-sprt-promote-demote-copy-rules/c04/) 第4章 接入真实转化流:周批次与基线从哪里来 — 第四章说明周批次让检验更保守,而每周重估基线会把误转正率抬到 0.1138。 (https://5dlibrary.com/answers/zh/live-rule-retirement-sprt-promote-demote-copy-rules/c04/) - Ch.5 The Rule Lifecycle as a State Machine: Promotion, Demotion and Retirement — Chapter 5 checks an eight-state rule machine exhaustively, watches for wear with Page's cumulative sum, and uses an absorbing chain to show that a worn rule stays live 8.54 weeks on average. (https://5dlibrary.com/answers/en/live-rule-retirement-sprt-promote-demote-copy-rules/c05/) 第5章 规则生命周期状态机:转正、降级与淘汰 — 第五章用穷举检查八状态的规则状态机,用 Page 累积和监视磨损,以吸收链算出磨损规则平均还在线 8.54 周。 (https://5dlibrary.com/answers/zh/live-rule-retirement-sprt-promote-demote-copy-rules/c05/) - Ch.6 Many Rules in Parallel: Sharing Alpha, and How Many Promoted Rules Are Duds — Chapter 6 shows that splitting alpha among 100 rules lengthens verdicts on useless rules by a ratio of only 1.12. (https://5dlibrary.com/answers/en/live-rule-retirement-sprt-promote-demote-copy-rules/c06/) 第6章 多条规则并行:α 怎么分,转正的规则里几成是空的 — 第六章显示平分 α 给 100 条规则时,无效规则的判决周数只增长 1.12 倍。 (https://5dlibrary.com/answers/zh/live-rule-retirement-sprt-promote-demote-copy-rules/c06/) - Ch.7 Classical Sources East and West: Checking by Trial in the Han Feizi, and Wald's Sequential Analysis — "Certainty without checking is folly" from the Han Feizi becomes a concrete number in p71: promote without trial and 0.9000 of the promotion list is useless. Harnessing the horse and watching it to the end of the road is one sequential test, with a useless share of 0.2554; ministers rising from the provinces is two-stage selection, with a useless share of 0.0129 but power down to 0.7594, much like one single-stage test at alpha 0.0025. Wald's Sequential Analysis passes nine exact checks in p72: all three inequalities hold, the true beta' occasionally exceeds its target slightly, and the inequality on the sum always holds. The ancient author supplied an attitude, Wald supplied an algorithm, and the first six chapters supplied the pipeline. The remaining question is the most practical one: on Monday morning, when the owner opens the weekly radar's report, what should be done and in what order? How is the distilled tool used? And how should a test be pre-registered so that the claims of this book face the verdict of real data? Chapter 8 answers. (https://5dlibrary.com/answers/en/live-rule-retirement-sprt-promote-demote-copy-rules/c07/) 第7章 东西方原典考辨:《韩非子·显学》的参验与 Wald 的《序贯分析》 — 《韩非子·显学》的「无参验而必之者,愚也」,在 p71 里是一个具体的数:不试就转正,转正名单中 0.9000 是无效规则。「授车就驾,观其末涂」是一次序贯检验,无效份额 0.2554;「宰相必起于州部」是两级选拔,无效份额 0.0129,但功效降到 0.7594,与一次 α 为 0.0025 的单级检验相差不大。Wald 的《序贯分析》在 p72 里接受了九组精确核对,三条不等式全部成立,真实 β′ 偶尔略超目标,但和的不等式始终守住。古人给出了态度,Wald 给出了算法,前六章补上了流水线。剩下的问题是最实际的:下周一早上,经营者打开每周雷达的报告,应该按什么顺序做什么?蒸馏出来的小工具怎么用?又该怎样预先登记一次检验,让本书的主张接受真实数据的裁决?第八章回答。 (https://5dlibrary.com/answers/zh/live-rule-retirement-sprt-promote-demote-copy-rules/c07/) - Ch.8 Closing the Loop: Wiring the Weekly Radar's Verdicts into the Rule State Machine — Chapter 8 wires the weekly radar's verdicts into the state machine and gives the distilled tool, the Rule Promotion and Demotion Judge, and a pre-registered A/A calibration: 100 A/A rules, with 11 or more promotions overturning the book. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/live-rule-retirement-sprt-promote-demote-copy-rules/c08/) 第8章 实证闭环:每周雷达的判决接进规则状态机 — 第八章把每周雷达的判决接进状态机,给出蒸馏小工具「规则转正降级判定器」与预注册的 A/A 校准:100 条 A/A 规则,不少于 11 条转正即推翻本书。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/live-rule-retirement-sprt-promote-demote-copy-rules/c08/) ## 5DL-B-264 · Dynamic Allocation of Compute and Budget across Business Lines: Gittins-Index Exploration and Weekly Quotas for Tollbooth, Buying and Publishing Lines 中文书名:多业务线算力与预算动态裁决、吉廷斯指数探索与收费站代购出版线周度配额母典 DOI: pending Cite: Fan, Y. (2026). Dynamic Allocation of Compute and Budget across Business Lines: Gittins-Index Exploration and Weekly Quotas for Tollbooth, Buying and Publishing Lines (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative, and margins and costs are accounting points, not prices; all calculations are generated by the book's programs. It is not individual business advice, and changes to budgets or live systems need sign-off by the person in charge. Abstract: A common weekly quota practice gives the whole budget to whichever business line returned most last week. This book tests that practice with the mathematics of multi-armed bandits. After checking Gittins' paper of 1979 and later work, it corrects its founding blueprint: the optimality of the Gittins index holds only when one line is worked at a time, idle lines are frozen, switching is free and discounting is geometric, so split weekly budgets, switching costs and changing markets all lie outside the theorem; lock-in is a statement about probability, since betting everything on cumulative returns ends the second half-year wholly on the worse of two illustrative lines in 0.2133 of runs, yet at a short-horizon discount of 0.9 per unit greedy loses only 0.0050 to 0.0166 against the index. Parameters are illustrative, and margins and costs are accounting points, not prices; all calculations are generated by the book's programs. It is not individual business advice, and changes to budgets or live systems need sign-off by the person in charge. Chapter 1 quantifies lock-in and shows that its cost grows with the horizon: over 500 periods greedy's regret per period stays at 0.0563 while Thompson sampling's falls to 0.0216. Chapter 2 computes the Gittins index by calibration and by restart; a line never tried has an index of 0.7029 at a discount of 0.9. Chapter 3 compares Thompson sampling with the index: the first over-explores at short horizons and guarantees logarithmic regret over long ones. Chapter 4 compares five weekly quota rules; the quota rule, in which the leader takes the bulk and every line whose index beats the leader's mean receives one trial unit, earns 0.9042 of the ideal, against 0.8370 for the last-week rule. Chapter 5 deals with two frictions: switching costs call for a hysteresis band, and changing markets call for fading records. Chapter 6 writes the tollbooth, buying and publishing lines as arms with margins and unit costs and gives net indices, prior strength and a stop-funding table. Chapter 7 examines Jiran's "in drought, lay in boats; in flood, lay in carts" from the Records of the Grand Historian and Thompson's two-sample problem of 1933: preparing against the present pays only in a regular cycle with a lag beyond a quarter of it. Chapter 8 gives the weekly quota meeting table, the Business-Line Quota Indexer and a pre-registered protocol (250 cells per group, 162 after adjusting for the run-in). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:常见的周度配额做法是:哪条业务线上周回报最好,本周就把预算整份押给它。本书用多臂老虎机的数学检验这一做法。核对吉廷斯 1979 年的论文与后续文献以后,本书修正了立项蓝图:吉廷斯指数的最优性只在单线经营、闲置线不变、换线无成本、几何贴现时成立,周度拆分预算、换线成本与会变的市场都在定理之外;锁死是一个概率说法,两条示意线上按累计回报全押有 0.2133 的重复在下半年完全押错,但在每单位贴现 0.9 的短视界里贪心相对指数只损失 0.0050 到 0.0166。书中参数为示意,毛利与成本用记账点数而非价格;全部计算由书中程序生成。本书属非个人化经营建议,预算与线上改动须由主事者人工放行。 第一章量化锁死,并显示视界越长代价越大:视界 500 期时贪心每期后悔停在 0.0563,汤普森抽样降到 0.0216。第二章用校准法与重启法计算吉廷斯指数,从未试过的线在贴现 0.9 时指数为 0.7029。第三章比较汤普森抽样与指数:前者在短视界过度探索,在长视界保证后悔按对数增长。第四章比较五种周度配额规则,「领先线拿大头、指数高过领先线均值者各拿一个试投单位」的配额规则拿到理想值的 0.9042,上周规则只有 0.8370。 第五章处理两种摩擦:换线成本要求滞后带,会变的市场要求记录褪色。第六章把收费站、代购、出版写成带毛利与单位成本的臂,给出净指数、先验强度与停线表。第七章考辨《史记·货殖列传》所载计然「旱则资舟,水则资车」与汤普森 1933 年的两样本问题:逆着当下准备只在规律周期且时滞超过四分之一周期时有利。第八章给出周度配额会议表、小工具「业务线配额指数器」与预注册方案(每组 250 个单元,按磨合期调整后 162 个)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Lock-In: Why Betting Everything on Last Week's Winner Traps a Business on a Second-Best Line — Chapter 1 quantifies lock-in and shows that its cost grows with the horizon: over 500 periods greedy's regret per period stays at 0.0563 while Thompson sampling's falls to 0.0216. (https://5dlibrary.com/answers/en/gittins-index-weekly-quota-business-lines/c01/) 第1章 锁死:按上周回报全押为何会长期困在次优业务线 — 第一章量化锁死,并显示视界越长代价越大:视界 500 期时贪心每期后悔停在 0.0563,汤普森抽样降到 0.0216。 (https://5dlibrary.com/answers/zh/gittins-index-weekly-quota-business-lines/c01/) - Ch.2 Calibration: Setting a Business Line against a Permanent Safe Yield and Bisecting for the Gittins Index — Chapter 2 computes the Gittins index by calibration and by restart; a line never tried has an index of 0.7029 at a discount of 0.9. (https://5dlibrary.com/answers/en/gittins-index-weekly-quota-business-lines/c02/) 第2章 校准法:把一条业务线与永久安全收益相比,二分求出吉廷斯指数 — 第二章用校准法与重启法计算吉廷斯指数,从未试过的线在贴现 0.9 时指数为 0.7029。 (https://5dlibrary.com/answers/zh/gittins-index-weekly-quota-business-lines/c02/) - Ch.3 Thompson Sampling against the Gittins Index: Two Ways of Keeping a Chance Open on the Same Decision — The relative entropy between the two lines is 0.00803, and the constant c of the lower bound is 7.47. Over a horizon of 100, the greedy rule's cumulative regret is 2.53 and Thompson sampling's 2.17, hardly different. Over 1000 units the first is 23.04 and the second 12.08; over 5000 units the first is 113.04 and the second only 20.85. The greedy rule's regret is roughly proportional to the horizon: 0.0226 per unit, a little over a third of the regret per unit of a rule that is always wrong, which means that on average more than a third of the units went to the worse line. From 1000 to 5000 units, Thompson sampling's regret grows by less than nine, exactly the pattern of logarithmic growth. (https://5dlibrary.com/answers/en/gittins-index-weekly-quota-business-lines/c03/) 第3章 汤普森抽样对吉廷斯指数:同一个决定上两种留机会的方式 — 两条线的相对熵是 0.00803,下界常数 c 是 7.47。视界 100 时,贪心规则累计后悔 2.53,汤普森抽样 2.17,相差不大;视界 1000 时,前者 23.04,后者 12.08;视界 5000 时,前者 113.04,后者只有 20.85。贪心规则的后悔差不多与视界成正比:每单位后悔 0.0226,大约是始终押错时每单位后悔的三分之一多,也就是说,平均有三分之一以上的单位押在了较差的线上。汤普森抽样的后悔从 1000 到 5000 只增加了不到九,正是对数增长的样子。 (https://5dlibrary.com/answers/zh/gittins-index-weekly-quota-business-lines/c03/) - Ch.4 From Choosing a Line to Splitting a Budget: Five Weekly Quota Rules under Batched Feedback — Chapter 4 compares five weekly quota rules; the quota rule, in which the leader takes the bulk and every line whose index beats the leader's mean receives one trial unit, earns 0.9042 of the ideal, against 0.8370 for the last-week rule. (https://5dlibrary.com/answers/en/gittins-index-weekly-quota-business-lines/c04/) 第4章 从选一条线到分一份预算:批量反馈下的五种周度配额规则 — 第四章比较五种周度配额规则,「领先线拿大头、指数高过领先线均值者各拿一个试投单位」的配额规则拿到理想值的 0.9042,上周规则只有 0.8370。 (https://5dlibrary.com/answers/zh/gittins-index-weekly-quota-business-lines/c04/) - Ch.5 Two Frictions: Switching Costs Call for a Hysteresis Band, Changing Markets Call for Fading Records — With switching costs the Gittins index loses its optimality, and a hysteresis band is a simple, effective patch: the higher the cost, the wider the best band, about three to five tenths of the cost in the illustrative model. When markets change, old records must fade: with γ at 0.97 both Thompson sampling and the greedy rule rise from a little over seven tenths of the ideal to nearly nine tenths. So far, though, the "sale" in these rules is still an abstract unit. A sale on the tollbooth, the buying and the publishing lines differs greatly in value, and so does the cost of a unit; how should the index be converted into money? (https://5dlibrary.com/answers/en/gittins-index-weekly-quota-business-lines/c05/) 第5章 两种摩擦:换线成本要求滞后带,会变的市场要求记录褪色 — 有换线成本时,吉廷斯指数不再有最优性,滞后带是一个简单有效的补丁:成本越高,最好的滞后带越宽,在示意模型里约为成本的三到五成。市场会变时,旧记录必须褪色:γ 取 0.97 时,汤普森抽样与贪心规则都把收益从理想值的七成多提高到近九成。到这里,规则里的「一次成交」还只是一个抽象的单位。收费站、代购、出版三条线的一次成交,价值相差很大,单位成本也不同;指数该怎样换算成钱? (https://5dlibrary.com/answers/zh/gittins-index-weekly-quota-business-lines/c05/) - Ch.6 Writing the Tollbooth, Buying and Publishing Lines as Arms: Margins, Unit Costs, Prior Strength and a Stop-Funding Table — Ranked by conversion rate, buying comes first (0.5000), the tollbooth second (0.3095) and publishing last (0.2500). Ranked by net mean, the tollbooth comes first (5.2857), buying second (5.0000) and publishing last (2.0000): buying converts best but its margin per order is too thin. Ranked by net index, publishing comes first (7.1851), the tollbooth second (5.8142) and buying last (5.1241). (https://5dlibrary.com/answers/en/gittins-index-weekly-quota-business-lines/c06/) 第6章 把收费站、代购与出版写成臂:毛利、单位成本、先验强度与停线表 — 按转化率排,代购第一(0.5000),收费站第二(0.3095),出版最后(0.2500)。按净均值排,收费站第一(5.2857),代购第二(5.0000),出版最后(2.0000):代购转化率最高,但每单毛利太薄。按净指数排,出版第一(7.1851),收费站第二(5.8142),代购最后(5.1241)。 (https://5dlibrary.com/answers/zh/gittins-index-weekly-quota-business-lines/c06/) - Ch.7 Classical Sources East and West: Jiran's "In Drought, Lay In Boats; In Flood, Lay In Carts" and Thompson's Two-Sample Problem of 1933 — Chapter 7 examines Jiran's "in drought, lay in boats; in flood, lay in carts" from the Records of the Grand Historian and Thompson's two-sample problem of 1933: preparing against the present pays only in a regular cycle with a lag beyond a quarter of it. (https://5dlibrary.com/answers/en/gittins-index-weekly-quota-business-lines/c07/) 第7章 东西方原典考辨:计然「旱则资舟,水则资车」与汤普森 1933 年的两样本问题 — 第七章考辨《史记·货殖列传》所载计然「旱则资舟,水则资车」与汤普森 1933 年的两样本问题:逆着当下准备只在规律周期且时滞超过四分之一周期时有利。 (https://5dlibrary.com/answers/zh/gittins-index-weekly-quota-business-lines/c07/) - Ch.8 Closing the Loop: the Weekly Quota Meeting Table, the Business-Line Quota Indexer and a Pre-Registered Test — Chapter 8 gives the weekly quota meeting table, the Business-Line Quota Indexer and a pre-registered protocol (250 cells per group, 162 after adjusting for the run-in). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/gittins-index-weekly-quota-business-lines/c08/) 第8章 实证闭环:周度配额会议表、业务线配额指数器与预注册检验 — 第八章给出周度配额会议表、小工具「业务线配额指数器」与预注册方案(每组 250 个单元,按磨合期调整后 162 个)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/gittins-index-weekly-quota-business-lines/c08/) ## 5DL-B-265 · Anti-Collapse Engineering with Two Anchors of Truth: the Variance-Collapse Rate of Self-Generated Training Data and Immunity through Standard-Library Recomputation and Textual Collation 中文书名:抗模型崩溃双锚真源工程、自产语料回喂方差坍缩速率与标准库验算校勘免疫母典 DOI: pending Cite: Fan, Y. (2026). Anti-Collapse Engineering with Two Anchors of Truth: the Variance-Collapse Rate of Self-Generated Training Data and Immunity through Standard-Library Recomputation and Textual Collation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations are generated by the book's programs, which call no model interface and use no reader's or customer's data. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: When a model is retrained generation after generation on content it produced itself, the tails of the distribution disappear and its width collapses, which Shumailov and colleagues called model collapse in Nature in 2024. This book tests the founding claim that the variance collapses as sigma^2 (1-1/N)n and that mixing a fixed share of real sources into each generation prevents collapse, and corrects it after checking the originals: the formula is only the expectation under a maximum-likelihood refit, while the paper's theorem uses unbiased estimators, under which the expectation stays put and the variance still goes to zero almost surely; the fixed-share statement is not a conclusion of that paper, and how much real sources preserve depends on their absolute number per generation. Parameters are illustrative; all calculations are generated by the book's programs, which call no model interface and use no reader's or customer's data. This is a non-individual business and engineering reference. Chapters 1 and 2 derive the Gaussian feedback chain and the discrete absorbing chain: with 100 samples per generation the expected variance is down to 0.3660 at generation 100, and the typical value collapses about twice as fast; a category is eventually lost with probability one minus its initial share. Chapter 3 sets out the contrary evidence of Gerstgrasser and colleagues: when real data are accumulated rather than replaced, the error of the mean in the Gaussian twin is capped by pi^2/6, that is 1.644934. Chapter 4 computes the fixed point of real-source mixing: at the same real share of 0.01 per generation, raising N from 100 to 10000 lifts the expected variance from 0.6644 to 0.9950; tail coverage answers only to counts, and 1000 real items per generation cover only 0.0256 of the tail categories whatever the share. Chapter 5 shows that inward-looking detection metrics grow more optimistic during collapse, the Good–Turing missing mass falling from 0.0677 to 0.0009 while the true loss rises. Chapter 6 covers the first anchor: standard-library recomputation, the order independence of math.fsum and the correlated failure of two anchors, which together raise the probability of an error-free book of 300 numbers to 0.6769. Chapter 7 examines Duan Yucai's "difficulty of settling right and wrong" and Erasmus's New Testament of 1516. Chapter 8 gives an eight-step two-anchor procedure for every volume, the distilled Feedback Collapse Rate Calculator and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:用模型自产的内容逐代重训,分布的尾部会逐代消失、宽度会逐代塌缩,Shumailov 等人 2024 年在《自然》上称之为模型坍缩。本书检验立项命题「方差按 σ²(1-1/N)^n 坍缩,每代混入固定比例真源可阻止坍缩」,核对原文后作出修正:这条公式只是最大似然重估下的期望,论文的定理用无偏估计,期望不变而方差几乎必然归零;固定比例的说法不是该论文的结论,真源能保住多少取决于每代真源的绝对个数。书中参数为示意;全部计算由书中程序生成,不调用任何模型接口,不使用任何读者或客户的数据;本书是非个人化的经营与工程参考。 第一、二章推导高斯回喂链与离散吸收链:每代 100 个样本时第 100 代期望方差只剩 0.3660,典型值约以两倍速度坍缩;一个类别最终丢失的概率等于 1 减去初始份额。第三章并陈 Gerstgrasser 等人的反方证据,累积而非替换真数据时,高斯孪生版的均值误差被 π²/6 即 1.644934 封顶。第四章算出真源混合的不动点:同样每代混入 0.01 的真源,N 从 100 增到 10000,期望方差从 0.6644 升到 0.9950;尾部覆盖只认个数,每代 1000 个真样本时不论比例如何都只覆盖 0.0256 的尾部类别。 第五章说明向内看的检测指标在坍缩中越来越乐观,Good–Turing 缺失质量从 0.0677 降到 0.0009,而真实缺失在上升。第六章讨论第一锚:标准库验算、math.fsum 的顺序无关与双锚的相关失效,两锚合用使一本 300 个数字的书无错的概率升到 0.6769。第七章考辨段玉裁「定其是非之难」与伊拉斯谟 1516 年的《新约全书》。第八章给出每卷八步双锚流程、蒸馏小工具「回喂坍缩速率计算器」与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Variance Collapsing Generation by Generation: the Expectation Formula of a Gaussian Feedback Chain and Almost-Sure Extinction — Chapters 1 and 2 derive the Gaussian feedback chain and the discrete absorbing chain: with 100 samples per generation the expected variance is down to 0.3660 at generation 100, and the typical value collapses about twice as fast; a category is eventually lost with probability one minus its initial share. (https://5dlibrary.com/answers/en/anti-model-collapse-two-anchors-variance-collapse-rate/c01/) 第1章 方差逐代坍缩:高斯回喂链的期望公式与几乎必然归零 — 第一、二章推导高斯回喂链与离散吸收链:每代 100 个样本时第 100 代期望方差只剩 0.3660,典型值约以两倍速度坍缩;一个类别最终丢失的概率等于 1 减去初始份额。 (https://5dlibrary.com/answers/zh/anti-model-collapse-two-anchors-variance-collapse-rate/c01/) - Ch.2 The Tails Die First: Generational Extinction in a Discrete Law and the Absorbing States of Wright–Fisher — Program p21 shows that a single round of finite sampling deletes a large stretch of the tail: with 1000 draws per generation, the Zipf list of 1000 categories is expected to lose 666.3 of them, the tail loses 0.8254 of its own mass, and the head is practically untouched. Program p22 uses the exact transition matrix to show that a single category is eventually lost with probability one minus its initial share, and that the end point of the feedback chain is a point mass. Program p23 shows the two-phase decline of a whole list, from the tails dying first to the head categories swallowing each other. Since collapse comes from treating the output of the previous generation as the whole training set, a natural question follows: what if old data are not thrown away, and the output of every generation is accumulated instead? The next chapter tests this contrary evidence, put forward by Gerstgrasser and colleagues. (https://5dlibrary.com/answers/en/anti-model-collapse-two-anchors-variance-collapse-rate/c02/) 第2章 尾部先死:离散分布的逐代灭绝与 Wright–Fisher 吸收态 — 程序 p21 说明,一代有限抽样就会删掉一大片尾部:每代 1000 个样本时,1000 类的齐普夫名单期望消失 666.3 类,尾部自身的质量丢掉 0.8254,头部几乎无损。程序 p22 用精确的转移矩阵说明,单个类别最终丢失的概率等于 1 减去初始份额,回喂链的终点是单点分布。程序 p23 展示了整张名单从尾部先死到头部互吞的两阶段衰亡。既然坍缩源于每一代都把上一代的产出当作全部训练数据,一个自然的问题是:如果不丢旧数据,而是把每一代的产出累积起来呢?下一章检验 Gerstgrasser 等人提出的这个反方证据。 (https://5dlibrary.com/answers/zh/anti-model-collapse-two-anchors-variance-collapse-rate/c02/) - Ch.3 Replace or Accumulate: the Counter-Evidence of Gerstgrasser and the Bound of pi^2/6 — Chapter 3 sets out the contrary evidence of Gerstgrasser and colleagues: when real data are accumulated rather than replaced, the error of the mean in the Gaussian twin is capped by pi^2/6, that is 1.644934. (https://5dlibrary.com/answers/en/anti-model-collapse-two-anchors-variance-collapse-rate/c03/) 第3章 替换还是累积:Gerstgrasser 反证与 π²/6 上界 — 第三章并陈 Gerstgrasser 等人的反方证据,累积而非替换真数据时,高斯孪生版的均值误差被 π²/6 即 1.644934 封顶。 (https://5dlibrary.com/answers/zh/anti-model-collapse-two-anchors-variance-collapse-rate/c03/) - Ch.4 The Threshold of Mixing in Real Sources: How Much Variance a Fixed Share Preserves, and How Much of the Tail It Misses — Chapter 4 computes the fixed point of real-source mixing: at the same real share of 0.01 per generation, raising N from 100 to 10000 lifts the expected variance from 0.6644 to 0.9950; tail coverage answers only to counts, and 1000 real items per generation cover only 0.0256 of the tail categories whatever the share. (https://5dlibrary.com/answers/en/anti-model-collapse-two-anchors-variance-collapse-rate/c04/) 第4章 真源混合阈值:固定比例保住了多少方差,又漏掉了多少尾部 — 第四章算出真源混合的不动点:同样每代混入 0.01 的真源,N 从 100 增到 10000,期望方差从 0.6644 升到 0.9950;尾部覆盖只认个数,每代 1000 个真样本时不论比例如何都只覆盖 0.0256 的尾部类别。 (https://5dlibrary.com/answers/zh/anti-model-collapse-two-anchors-variance-collapse-rate/c04/) - Ch.5 Examining for Collapse: Why Inward-Looking Metrics Go Blind, and How Much Data an Outward-Looking Anchor Needs — Chapter 5 shows that inward-looking detection metrics grow more optimistic during collapse, the Good–Turing missing mass falling from 0.0677 to 0.0009 while the true loss rises. (https://5dlibrary.com/answers/en/anti-model-collapse-two-anchors-variance-collapse-rate/c05/) 第5章 坍缩体检:向内看的指标为何失明,向外看的锚点要多少数据 — 第五章说明向内看的检测指标在坍缩中越来越乐观,Good–Turing 缺失质量从 0.0677 降到 0.0009,而真实缺失在上升。 (https://5dlibrary.com/answers/zh/anti-model-collapse-two-anchors-variance-collapse-rate/c05/) - Ch.6 The First Anchor: Standard-Library Recomputation, Deterministic Fingerprints and the Correlated Failure of Two Anchors — Chapter 6 covers the first anchor: standard-library recomputation, the order independence of math.fsum and the correlated failure of two anchors, which together raise the probability of an error-free book of 300 numbers to 0.6769. (https://5dlibrary.com/answers/en/anti-model-collapse-two-anchors-variance-collapse-rate/c06/) 第6章 第一锚:标准库验算、确定性指纹与双锚的相关失效 — 第六章讨论第一锚:标准库验算、math.fsum 的顺序无关与双锚的相关失效,两锚合用使一本 300 个数字的书无错的概率升到 0.6769。 (https://5dlibrary.com/answers/zh/anti-model-collapse-two-anchors-variance-collapse-rate/c06/) - Ch.7 The Classics Examined: Duan Yucai on "the Difficulty of Settling Right and Wrong" and Erasmus's Return to the Original in 1516 — Chapter 7 examines Duan Yucai's "difficulty of settling right and wrong" and Erasmus's New Testament of 1516. (https://5dlibrary.com/answers/en/anti-model-collapse-two-anchors-variance-collapse-rate/c07/) 第7章 原典考辨:段玉裁「定其是非之难」与伊拉斯谟 1516 年的回到原文 — 第七章考辨段玉裁「定其是非之难」与伊拉斯谟 1516 年的《新约全书》。 (https://5dlibrary.com/answers/zh/anti-model-collapse-two-anchors-variance-collapse-rate/c07/) - Ch.8 The Empirical Loop: a Two-Anchor Procedure for Every Volume, the Feedback Collapse Rate Calculator and a Pre-Registered Test — Chapter 8 gives an eight-step two-anchor procedure for every volume, the distilled Feedback Collapse Rate Calculator and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/anti-model-collapse-two-anchors-variance-collapse-rate/c08/) 第8章 实证闭环:每卷双锚流程、回喂坍缩速率计算器与预注册检验 — 第八章给出每卷八步双锚流程、蒸馏小工具「回喂坍缩速率计算器」与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/anti-model-collapse-two-anchors-variance-collapse-rate/c08/) ## 5DL-B-266 · Three Lights for Handing Over between Bot and Human in Private Customer Service: Misclassification Loss Matrices, Chow's Reject Rule and Optimal Cut Points on Calibrated Confidence 中文书名:私域客服人机交接三级灯号精算、误判损失矩阵与最优置信度切分点母典 DOI: pending Cite: Fan, Y. (2026). Three Lights for Handing Over between Bot and Human in Private Customer Service: Misclassification Loss Matrices, Chow's Reject Rule and Optimal Cut Points on Calibrated Confidence (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Private customer service has a third action, a bot draft checked by a person, so it needs a six-cell loss matrix with two cut points; under illustrative parameters red lights below a chance of being right of 0.4000 and green from 0.9500. Parameters are illustrative, all conversations are fictional, and all numbers come from the book's programs, which read and write no file and connect to no customer-service system. Abstract: Once a bot joins a shop's private customer-service channel, the usual practice is to set a confidence threshold by accuracy and let the bot answer whenever it is high enough. The commissioning claim says that a threshold set by accuracy alone is wrong, because with a reject option the optimal threshold is decided by the ratio of misclassification costs (Chow 1970). Writing the claim as computable models in standard-library programs, the book corrects the blueprint in three places. Optimal recognition with a rejection channel already appears in Chow's 1957 paper, and the 1970 paper gives the general relation between error and rejection. The rule holds only for calibrated probabilities: cut points derived from the cost ratio and applied to the self-reported scores of an overconfident bot lose 265.4 yuan per hundred conversations, against 239.8 yuan after isotonic calibration. Private customer service has a third action, a bot draft checked by a person, so it needs a six-cell loss matrix with two cut points; under illustrative parameters red lights below a chance of being right of 0.4000 and green from 0.9500. When a wrong answer costs 16 times a handover, a threshold set by accuracy loses 1093.1 yuan per hundred conversations and one set by the cost ratio only 514.8 yuan. Parameters are illustrative, all conversations are fictional, and all numbers come from the book's programs, which read and write no file and connect to no customer-service system. This is non-individual business advice; any change to live lights needs a manual release by the person in charge. Chapter 1 compares Chow's rule with two accuracy yardsticks and checks the slope of the error-reject curve. Chapter 2 derives closed-form cut points and the condition for an amber band: the review cost must be below the catch rate times the takeover cost. Chapter 3 compares two bots on cost curves; the less accurate bot that rejects better is cheaper above a cost ratio of 4.14. Chapter 4 calibrates confidence with the pool-adjacent-violators algorithm and Platt scaling. Chapter 5 treats the prior shift of a big sale: when the true share of right answers falls to 0.600, the bot's mean self-reported chance is still 0.843. Chapter 6 makes "stop means stop" and "vulnerable customers go to staff" hard red lights outside the loss matrix and handles staffing caps with a shadow price. Chapter 7 examines the Analects' "hear much, set aside the doubtful, speak with care about the rest, and you will seldom err" and Bayes's paper of 1763: with 19 right out of 20 reviewed, the posterior mean is only 0.9091. Chapter 8 gives a monthly recalculation routine with an event rule for version changes, the distilled "Three-Light Cut-Point Calculator", a pre-registered test needing 1177 conversations per group, and the conditions that would overturn the book's three pillars. Boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:私域客服接入机器人之后,流行做法是按准确率定一个置信度门限,够高就让机器人自己答。立项命题说:只按准确率定阈值是错的,带拒判选项的最优阈值由误判成本比决定(Chow 1970)。本书用纯标准库程序把命题写成可计算的模型,并修正蓝图三处:带拒识通道的最优识别早见于 Chow 1957 年的论文,1970 年的论文给出误差与拒判的一般关系;规则只对校准过的概率成立,把成本比算出的切分点直接套在过度自信机器人的自报分数上,每百段对话损失 265.4 元,等渗回归校准后 239.8 元;私域客服有「机器人起草、真人过目」的第三种动作,需要六格损失矩阵与两条切分点,示意参数下答对概率低于 0.4000 亮红灯、不低于 0.9500 亮绿灯。答错代价是人工代价 16 倍时,按准确率定的门限每百段损失 1093.1 元,按成本比定的只有 514.8 元。书中参数为示意,对话全部虚构;全部计算由书中程序生成;程序不读写任何文件,不连接任何客服系统;本书是非个人化经营建议,任何线上灯号的改动须由主事者人工复核放行。 第一章对照 Chow 规则与两种准确率口径,并核对误差—拒判曲线的斜率。第二章推出两条切分点的封闭解与黄灯存在的条件:人审代价须低于拦截率乘接手代价。第三章用成本曲线比较两个机器人,准确率较低但更会拒判的那个在成本比高于 4.14 时更省钱。第四章用合并相邻违序者算法与普拉特缩放校准置信度。第五章讨论大促带来的先验迁移:真实答对比例降到 0.600 时,机器人自报的平均答对概率仍有 0.843。第六章把对方说停就停与弱势顾客转人工作为不进损失矩阵的硬红灯,并用影子价格处理人手上限。 第七章考辨《论语·为政》「多闻阙疑,慎言其余,则寡尤」与贝叶斯 1763 年的论文:复核 20 段答对 19 段时,后验均值只有 0.9091。第八章给出月度重算流程与换版本当周立即重算的事件规则、蒸馏小工具「三级灯号切分点计算器」、每组需 1177 段对话的预先登记检验,以及推翻本书三根支柱的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Blind Spot of Accuracy: Chow's Reject Rule and the Cost Ratio — Now read down. When a wrong answer costs 8 times as much as staff time, Chow's threshold moves to 0.8750 and the loss is 389.2 yuan; A1 stays put and loses 558.7 yuan; A2, with its threshold of 0.8170, is not far from the optimum and loses 402.8 yuan. At 16 times, Chow's threshold is 0.9375 with a loss of 514.8 yuan, A1 climbs to 1093.1 yuan and A2 to 666.2 yuan, so the worse of the two accuracy yardsticks pays 578.3 yuan more. At 32 times the best threshold is 0.9688, and A1 loses 2161.8 yuan, more than three times the optimum of 623.4 yuan. (https://5dlibrary.com/answers/en/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c01/) 第1章 准确率的盲区:Chow 拒判规则与成本比 — 再往下看。答错代价升到人工的 8 倍,Chow 门限移到 0.8750,损失 389.2 元;A1 口径不动,损失 558.7 元;A2 口径的门限 0.8170 离最优不远,损失 402.8 元。到 16 倍,Chow 门限 0.9375,损失 514.8 元,A1 口径升到 1093.1 元,A2 口径 666.2 元,两种准确率口径中较差的那个多付 578.3 元。到 32 倍,最优门限 0.9688,A1 口径的损失 2161.8 元,是最优 623.4 元的三倍多。 (https://5dlibrary.com/answers/zh/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c01/) - Ch.2 A Six-Cell Loss Matrix: Bayes Risk in Three Zones and Two Cut Points — A guess at 0.50 of the true value, 25.0 yuan, loses an extra 17.5 yuan per hundred conversations, 7.3 % of the optimal loss; a guess at 2.00 times, 100.0 yuan, loses an extra 23.0 yuan, or 9.6 %. A guess off by about a third costs roughly two per cent: 35.0 yuan adds 5.2 yuan and 70.0 yuan adds 5.4 yuan. Only a fourfold error makes the loss substantial: a guess of 12.5 yuan adds 73.4 yuan, or 30.7 %, and a guess of 200.0 yuan adds 89.9 yuan, or 37.6 %. (https://5dlibrary.com/answers/en/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c02/) 第2章 六格损失矩阵:三区决策的贝叶斯风险与两条切分点 — 估成真实值的 0.50 倍,即 25.0 元时,每百段多损失 17.5 元,相当于最优损失的 7.3 %;估成 2.00 倍,即 100.0 元时,多损失 23.0 元,即 9.6 %。估偏三成左右,代价只有两个百分点上下:估成 35.0 元多付 5.2 元,估成 70.0 元多付 5.4 元。只有偏到四倍,损失才变得显著:估成 12.5 元多付 73.4 元,即 30.7 %;估成 200.0 元多付 89.9 元,即 37.6 %。 (https://5dlibrary.com/answers/zh/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c02/) - Ch.3 Cost Curves: Why the More Accurate Bot Can Cost More — Chapter 3 compares two bots on cost curves; the less accurate bot that rejects better is cheaper above a cost ratio of 4.14. (https://5dlibrary.com/answers/en/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c03/) 第3章 成本曲线:准确率更高的机器人为什么反而更贵 — 第三章用成本曲线比较两个机器人,准确率较低但更会拒判的那个在成本比高于 4.14 时更省钱。 (https://5dlibrary.com/answers/zh/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c03/) - Ch.4 Confidence Is Not Probability: Isotonic Calibration and the Price of Few Labels — The optimal loss under the true probability is 240.0 yuan per hundred conversations. With only 50 labels, isotonic regression loses an extra 38.4 yuan on average and Platt scaling 12.9 yuan; with 100 labels, 22.5 and 6.6 yuan; with 300, 7.6 and 1.9; with 1,000, 3.6 and 0.8; with 3,000, 1.8 and 0.3. The mean gap from the true probability tells the same story: 0.0687 for isotonic and 0.0382 for Platt with 50 labels, falling to 0.0145 and 0.0051 with 3,000. (https://5dlibrary.com/answers/en/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c04/) 第4章 置信度不是概率:等渗回归校准与少样本的代价 — 真实概率下的最优损失是每百段 240.0 元。只有 50 段标注时,等渗回归平均多损失 38.4 元,普拉特缩放多损失 12.9 元;100 段时分别是 22.5 元与 6.6 元;300 段时 7.6 元与 1.9 元;1,000 段时 3.6 元与 0.8 元;3,000 段时 1.8 元与 0.3 元。校准值与真实概率的平均差同样如此:50 段时等渗回归 0.0687、普拉特 0.0382,3,000 段时降到 0.0145 与 0.0051。 (https://5dlibrary.com/answers/zh/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c04/) - Ch.5 Big Sales and Drift: Bayesian Correction for Prior Shift and the Sampling Error of Cut Points — Chapter 5 treats the prior shift of a big sale: when the true share of right answers falls to 0.600, the bot's mean self-reported chance is still 0.843. (https://5dlibrary.com/answers/en/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c05/) 第5章 大促与漂移:先验迁移的贝叶斯修正与切分点的抽样误差 — 第五章讨论大促带来的先验迁移:真实答对比例降到 0.600 时,机器人自报的平均答对概率仍有 0.843。 (https://5dlibrary.com/answers/zh/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c05/) - Ch.6 Red Lights Beyond Price: Stop Means Stop, Vulnerable Customers Go to Staff, and the Staffing Cap — Without a cap the optimal cut points are those of Chapter 2, 0.400 and 0.950, with a load of 22.1 units and a loss of 295.2 yuan per hundred conversations, which here includes the staff cost of the hard-rule conversations. With a cap of 20, the shadow price is 1.96 yuan, the green threshold falls from 0.950 to 0.938, the red threshold from 0.400 to 0.253, green rises to 0.572, and the loss grows only to 297.1 yuan. With a cap of 16, the shadow price is 11.05 yuan and the red threshold falls to zero, so that apart from the hard rules no conversation is handed over in full for low confidence; the green threshold falls to 0.881 and the loss is 320.0 yuan. With a cap of 12, the shadow price is 40.14 yuan, the green threshold 0.699, green covers 0.911, and the loss is 408.9 yuan. (https://5dlibrary.com/answers/en/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c06/) 第6章 不容定价的红灯:说停就停、弱势转人工与人手上限 — 不设上限时,最优切分点就是第二章的 0.400 与 0.950,负荷 22.1 个单位,每百段损失 295.2 元,这里包含了硬规则对话的人工代价。上限降到 20 时,影子价格 1.96 元,绿灯门限从 0.950 降到 0.938,红灯门限从 0.400 降到 0.253,绿灯比例升到 0.572,损失只增加到 297.1 元。上限降到 16 时,影子价格 11.05 元,红灯门限降到零,即除了硬规则,不再有任何对话因为置信度低而整段转人工,绿灯门限降到 0.881,损失 320.0 元。上限降到 12 时,影子价格 40.14 元,绿灯门限 0.699,绿灯占 0.911,损失 408.9 元。 (https://5dlibrary.com/answers/zh/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c06/) - Ch.7 Setting the Doubtful Aside, and Bayes: Two Classical Texts on Being Unsure — Chapter 7 examines the Analects' "hear much, set aside the doubtful, speak with care about the rest, and you will seldom err" and Bayes's paper of 1763: with 19 right out of 20 reviewed, the posterior mean is only 0.9091. (https://5dlibrary.com/answers/en/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c07/) 第7章 阙疑与贝叶斯:两份关于「把握不足」的古典文本 — 第七章考辨《论语·为政》「多闻阙疑,慎言其余,则寡尤」与贝叶斯 1763 年的论文:复核 20 段答对 19 段时,后验均值只有 0.9091。 (https://5dlibrary.com/answers/zh/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c07/) - Ch.8 The Monthly Recalculation Loop, the Distilled Tool and a Falsification Plan — Chapter 8 gives a monthly recalculation routine with an event rule for version changes, the distilled "Three-Light Cut-Point Calculator", a pre-registered test needing 1177 conversations per group, and the conditions that would overturn the book's three pillars. Boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c08/) 第8章 月度重算闭环、蒸馏小工具与证伪方案 — 第八章给出月度重算流程与换版本当周立即重算的事件规则、蒸馏小工具「三级灯号切分点计算器」、每组需 1177 段对话的预先登记检验,以及推翻本书三根支柱的条件。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/three-light-handover-loss-matrix-optimal-confidence-cutpoints/c08/) ## 5DL-B-267 · Fault-Tolerant Parsing of Payment Receipts and Spoken Orders: Last-Five-Digit Write-Off Automata and Blocking Duplicate-Screenshot Fraud 中文书名:多模态汇款凭证与口语订单容错解析、转账末五码核销自动机与重复截图欺诈阻断母典 DOI: pending Cite: Fan, Y. (2026). Fault-Tolerant Parsing of Payment Receipts and Spoken Orders: Last-Five-Digit Write-Off Automata and Blocking Duplicate-Screenshot Fraud (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Orders, account tails, amounts, screenshots and hashes are synthesised from fixed seeds, parameters are illustrative, and every calculation is generated by the book's programs. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Buying agents and small shops selling through private messages often reconcile payments the same way: the buyer sends a screenshot of a transfer, the owner finds a credit of the same amount on the statement and marks the order paid. This book tests two defaults behind that routine, that "prices all differ, so same-amount collisions are unlikely" and that "if the screenshot matches, the money has arrived", and quantifies the specification's three claims one by one: matching on amount alone must collide; with the last five digits and a time window collisions can become negligible; repeated screenshots can be detected by the Hamming distance of a perceptual hash. All calculations use only the Python standard library. Orders, account tails, amounts, screenshots and hashes are synthesised from fixed seeds, parameters are illustrative, and every calculation is generated by the book's programs. The book discusses only how the party doing the write-off detects repeated or altered receipts and gives no way of producing one; it offers general business method, not legal, financial or investment advice, and every order marked paid or shipped must be released by the owner. Chapter 1 quantifies collisions through the birthday problem: from 10 to 100 orders the expected number of colliding pairs grows 110.0 times while the collision probability grows only 41.3 times; an illustrative catalogue has only 19.3 effective amounts, giving even odds of a collision at 5.2 orders. Chapter 2 shows that at a thousand orders a day, amount plus last five digits plus a thirty-minute window collides once in 93.9 days on average, and at five thousand once in 3.8 days; repeat orders by the same buyer still tie on the key 1.220 times a day. The finite-state transducer of Chapter 3 fails nowhere in an exhaustive round trip of 509,498 texts. Chapter 4 proves by exhaustive search that a screenshot is a claim and a credit the fact; with amount alone about 0.17823 of credits are misassigned. Chapter 5 shows that a whole-image average hash at threshold 4 flags 0.9135 of same-layout genuine receipts as resends, while a hash of the fields alone recognises 0.8333 of resends at threshold 10 with no false positives. Chapter 6 reworks the birthday problem inside the fingerprint library and builds exact neighbour search from the pigeonhole principle. Chapter 7 examines the half-seal tallies of the Collected Statutes of the Ming and Galton's Finger Prints. Chapter 8 gives the reconciliation loop, the distilled tool, the Payment Receipt Reconciler, and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:做代购与私域小店的人常用同一套对账办法:买家发来转账截图,店主在流水里找一笔金额相同的到账,就把订单标为已付。本书检验这套办法背后的两个默认,即「价格各不相同,同额撞单是小概率」与「截图对得上,钱就到了」,并把规格书的三个命题逐条量化:只核金额必撞单;加末五码与时间窗后撞单可降到可忽略;重复截图可用感知哈希的汉明距离检出。全部计算只用 Python 标准库。书中订单、账号末码、金额、截图与哈希均由固定种子合成,参数为示意,全部计算由书中程序生成;本书只讨论核销方如何检出重复或改动过的凭证,不提供任何制作凭证的方法;属非个人化经营方法,不构成法律、财务或理财建议,任何置付与发货都须由主事者人工放行。 第一章用生日问题量化撞单:从 10 单到 100 单,期望撞单配对数增长 110.0 倍,撞单概率只增长 41.3 倍;示意目录的有效金额只有 19.3 种,5.2 单即有对半机会撞单。第二章显示,日单量一千单时金额加末五码加三十分钟时间窗平均 93.9 天撞一次,五千单时 3.8 天一次;同一买家的重复订单每天仍有 1.220 对撞键。第三章的有限状态转换器在 509,498 个文本的穷举往返中零失败。第四章穷举证明「截图是声明,入账才是事实」,只看金额时错配率约 0.17823。 第五章显示,整图平均哈希在门槛 4 下会把 0.9135 的同版式真凭证误判为重发,只对字段求哈希后在门槛 10 下认出 0.8333 的重发且零误判。第六章在指纹库里重算生日问题,并用鸽巢原理做精确近邻检索。第七章考辨《明会典》的勘合半印比对之制与高尔顿的《指纹》。第八章给出代购对账核销流程、蒸馏小工具「汇款凭证核销器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Matching on Amount Alone Must Collide: the Birthday Problem of Same-Day, Same-Amount Payments — Chapter 1 quantifies collisions through the birthday problem: from 10 to 100 orders the expected number of colliding pairs grows 110.0 times while the collision probability grows only 41.3 times; an illustrative catalogue has only 19.3 effective amounts, giving even odds of a collision at 5.2 orders. (https://5dlibrary.com/answers/en/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c01/) 第1章 只核金额必撞单:同日同额汇款的生日问题 — 第一章用生日问题量化撞单:从 10 单到 100 单,期望撞单配对数增长 110.0 倍,撞单概率只增长 41.3 倍;示意目录的有效金额只有 19.3 种,5.2 单即有对半机会撞单。 (https://5dlibrary.com/answers/zh/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c01/) - Ch.2 The Last Five Digits and a Time Window: How Far Collisions Fall, and What Will Not Fall — Chapter 2 shows that at a thousand orders a day, amount plus last five digits plus a thirty-minute window collides once in 93.9 days on average, and at five thousand once in 3.8 days; repeat orders by the same buyer still tie on the key 1.220 times a day. The finite-state transducer of Chapter 3 fails nowhere in an exhaustive round trip of 509,498 texts. (https://5dlibrary.com/answers/en/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c02/) 第2章 末五码与时间窗:碰撞能压多低,压不下去的又是什么 — 第二章显示,日单量一千单时金额加末五码加三十分钟时间窗平均 93.9 天撞一次,五千单时 3.8 天一次;同一买家的重复订单每天仍有 1.220 对撞键。 (https://5dlibrary.com/answers/zh/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c02/) - Ch.3 A Finite Automaton for Spoken and Written Chinese Amounts, Simplified and Traditional — Spoken amounts do not need an ever-growing pile of rules. A finite-state transducer with 15 states, 9 token classes and 45 transitions covers simplified and traditional script, banker's capitals, spoken cut-offs and mixed Arabic digits, and fails nowhere in an exhaustive round trip over 509,498 texts; whatever the table does not license is refused. Against misheard sound-alike digits, the automaton first removes illegal readings and the pending pool then removes amounts nobody owes, which takes misreadings in the dense pool from 0.0092 down to zero. Yet reading the amount correctly and matching the last five digits only shows that a screenshot appears to belong to an order. The screenshot itself is only a claim. The next chapter sets the floor of write-off: what evidence may make an order "paid", and how a write-off automaton guarantees that no sequence of events can get round that floor. (https://5dlibrary.com/answers/en/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c03/) 第3章 繁简口语金额的有限自动机 — 口语金额不需要靠越写越长的规则堆来解析。一台 15 个状态、9 类记号、45 条转移的有限状态转换器,覆盖了简体、繁体、银行大写、口语省略与阿拉伯数字混写,在 509,498 个文本的穷举往返中零失败;凡是表不许可的说法,一律拒绝。面对听错的近音数字,先让自动机删去不合法的读法,再用待核销池删去不存在的金额,在稠密池中把误读从 0.0092 压到零。可是,读对了金额、对上了末五码,也只说明这张截图「看起来」属于这一单。截图本身只是一个声明。下一章要建立核销的底线:什么证据才能让一单变成「已付」,以及一台核销自动机怎样保证这条底线不被任何事件序列绕过。 (https://5dlibrary.com/answers/zh/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c03/) - Ch.4 The Write-Off Automaton: a Screenshot Is a Claim, a Credit Is a Fact — Chapter 4 proves by exhaustive search that a screenshot is a claim and a credit the fact; with amount alone about 0.17823 of credits are misassigned. (https://5dlibrary.com/answers/en/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c04/) 第4章 核销自动机:截图是声明,入账才是事实 — 第四章穷举证明「截图是声明,入账才是事实」,只看金额时错配率约 0.17823。 (https://5dlibrary.com/answers/zh/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c04/) - Ch.5 Average Hash and Hamming Distance: Detecting Repeated Screenshots, and Where Detection Is Blind — Chapter 5 shows that a whole-image average hash at threshold 4 flags 0.9135 of same-layout genuine receipts as resends, while a hash of the fields alone recognises 0.8333 of resends at threshold 10 with no false positives. (https://5dlibrary.com/answers/en/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c05/) 第5章 平均哈希与汉明距离:重复截图的检出与它的盲区 — 第五章显示,整图平均哈希在门槛 4 下会把 0.9135 的同版式真凭证误判为重发,只对字段求哈希后在门槛 10 下认出 0.8333 的重发且零误判。 (https://5dlibrary.com/answers/zh/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c05/) - Ch.6 The Receipt Fingerprint Library: the Birthday Problem Among Hashes, and Pigeonhole Search — The hash library has its own birthday problem: chance neighbours of a single query grow in step with the library, and chance pairs within the library grow with its square. For an ideal 64-bit hash at threshold 10, a library of a million holds 4,991 chance pairs; at threshold 4, a library of 6,136,405 has better than even odds of one. A 390-bit field hash has almost no chance neighbours at the same size. The pigeonhole principle cuts a 64-bit hash into t+1 blocks; at threshold 4 a query checks about 10.56 candidates on average, and the answers agree with a full scan record by record. By now the three claims of the specification have been quantified, corrected and written into programs. The next chapter turns to the ancients. Ming officials guarded documents against misuse by matching half-seals against a register, and Galton argued the uniqueness of fingerprints with an estimate worth 36 bits. What did each of them get right, and where do they meet today's calculations? (https://5dlibrary.com/answers/en/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c06/) 第6章 凭证指纹库:哈希库里的生日问题与鸽巢检索 — 哈希库里同样有生日问题:单次查询的偶然近邻随库容线性增长,库内两两查重的偶然近邻对随库容平方增长。64 位理想哈希在门槛 10 下,一百万条记录里就有 4,991 对偶然近邻;门槛 4 下,库容到 6,136,405 时出现一对的机会过半。390 位字段哈希在同等库容下几乎没有偶然近邻。鸽巢原理把 64 位哈希切成 t+1 段,在门槛 4 时每次只需核对约 10.56 条候选,结果与全库扫描逐条一致。到这里,规格书的三个命题都已量化、修正并落到程序里。下一章回到古人:明代的勘合用半印与底簿比对防止文书被冒用,高尔顿用 36 位的估算论证指纹的独特性;他们各自说对了什么,又在哪里与今天的计算相遇? (https://5dlibrary.com/answers/zh/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c06/) - Ch.7 Classics East and West: Half-Seal Tallies in the Collected Statutes of the Ming and Galton's Finger Prints — The amount under the illustrative catalogue carries only 4.27 bits, with a birthday bound of 5.2 orders, as in Chapter 1. The last four digits carry 13.29 bits with a bound of 117.7; the last five, 16.61 bits with a bound of 372.3; a thirty-minute window, 4.60 bits. Amount, last five digits and window together carry 25.48 bits, with a birthday bound of 8,065.2: more than eight thousand orders in one window before even odds of a key clash. Galton's fingerprint estimate amounts to 36.00 bits with a bound of 308,651.0; an ideal 64-bit hash at threshold 4 amounts to 44.63 bits; a sixteen-digit transaction number carries 53.15 bits, with a bound above a hundred million. (https://5dlibrary.com/answers/en/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c07/) 第7章 东西方原典考辨:《明会典》勘合半印比对之制与高尔顿《指纹》 — 示意目录下的金额只有 4.27 比特,生日界 5.2 单,与第一章一致。末四位 13.29 比特、生日界 117.7;末五位 16.61 比特、生日界 372.3;三十分钟时间窗 4.60 比特。金额、末五码与时间窗合起来是 25.48 比特,生日界 8,065.2,即一个时间窗里要有八千多单,才有对半的机会撞键。高尔顿的指纹估算相当于 36.00 比特,生日界 308,651.0;64 位理想哈希在门槛 4 下相当于 44.63 比特;16 位数字的交易号有 53.15 比特,生日界一亿多。 (https://5dlibrary.com/answers/zh/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c07/) - Ch.8 The Buying Agent's Reconciliation Loop and a Protocol for Refutation — Chapter 8 gives the reconciliation loop, the distilled tool, the Payment Receipt Reconciler, and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c08/) 第8章 代购对账核销流程与证伪方案 — 第八章给出代购对账核销流程、蒸馏小工具「汇款凭证核销器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/payment-receipt-parsing-last-five-digits-duplicate-screenshot/c08/) ## 5DL-B-268 · Multimodal Derivative Compilation and Provenance Chains: A Structure-Preserving Pipeline from One Monograph to Podcasts, Slides and Quiz Banks 中文书名:母书多模态衍生编译与版权链、单部专著到播客投影片题库的结构同态流水线母典 DOI: pending Cite: Fan, Y. (2026). Multimodal Derivative Compilation and Provenance Chains: A Structure-Preserving Pipeline from One Monograph to Podcasts, Slides and Quiz Banks (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and costs are accounting points, not prices; all calculations are generated by the book's programs. It is not individual legal or business advice; its "copyright chain" means only provenance and permission records for the author's own works, and publishing any derivative needs sign-off by the person in charge. Abstract: A monograph is often adapted by relay into a summary, a podcast script, slides, a quiz bank and a short video, on the common assumption that if every step is done with care the final product stays close to the book. This book tests that assumption and corrects its founding blueprint against the original literature: Bartlett's serial reproduction of 1932 was a qualitative observation, and "fidelity p per layer leaves p to the power k after k layers" is a model of independent errors that are never repaired; at a fidelity of 0.95 per layer the fifth layer keeps only 0.7738, and as soon as a wrong fact has some chance of being put right a long chain settles at a positive level (0.2857 at a repair chance of 0.02 per layer). The claim that provenance hashes raise detection close to 1 holds only for anchored facts: with nine tenths anchored and a structure check that sees six tenths, detection is 0.960; with half anchored it is about 0.650. Parameters are illustrative and costs are accounting points, not prices; all calculations are generated by the book's programs. It is not individual legal or business advice; its "copyright chain" means only provenance and permission records for the author's own works, and publishing any derivative needs sign-off by the person in charge. Chapter 1 computes the compounding of drift and the effect of repair, and shows that directed drift such as rounding cannot be averaged away across versions. Chapter 2 writes facts as four-field atoms compared through canonical SHA-256 hashes, long enough prefixes and Merkle roots. Chapter 3 splits detection into anchor coverage and structural detection and shows that checks must go against the master, not the previous layer; on a set of sentence pairs the anchor checker has a precision of 1.000. Chapter 4 writes arguments as labelled directed graphs and requires each derivative to be a homomorphic image of the master's graph; the blind spots of structure checks and hashes barely overlap. Chapter 5 compares derivation topologies: without checks the five products of a chain carry 28.07 wrong facts between them and those of a star 10.00; with compilation and error costs in one expression the lowest-cost tree is usually a hub. Chapter 6 gives a derivation ledger with parent hashes, in which a later edit to one record is flagged together with everything downstream. Chapter 7 examines the lost fragment of Liu Xiang's Bielu, "one holds the copy, one reads the text, like enemies facing each other", and Bartlett's serial reproduction: two independent collators leave 0.12 of the errors. Chapter 8 gives the provenance hash manifest, the Derivative Drift Provenance Checker and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:一部专著常被接力改编成摘要、播客讲稿、投影片、题库与短视频,流行的默认是:每一步都认真,最后的产品就与原书差不多。本书检验这一默认,并按原始文献修正了立项蓝图:巴特利特 1932 年的串行复现是定性观察,「每层保真 p、k 层后剩 p 的 k 次方」是独立误差且永不修复时的模型,每层保真 0.95 时第五层只剩 0.7738;只要错了的事实有一点被改回来的机会,长链就停在正的水平上(每层 0.02 时为 0.2857)。「出处哈希可把检出率提到接近 1」只对被锚定的事实成立:锚定九成、结构检查认出六成时检出率是 0.960,锚定一半时约 0.650。书中参数为示意,代价用记账点数而非价格;全部计算由书中程序生成。本书属非个人化法律与经营建议,「版权链」只指作者自有作品的衍生溯源与授权记录,任何衍生品发布须由主事者人工放行。 第一章算漂移的复利与修复回流,并说明取整一类有方向的漂移无法靠多版取平均消除。第二章把事实写成四字段原子,用规范化 SHA-256 哈希、足够长的前缀与默克尔根做比对。第三章把检出率分解为锚定覆盖率与结构检出,并说明核对必须对母书而不对上一层;锚点比对在一组句对上准确率为 1.000。第四章把论证写成带标签的有向图,要求衍生品是母书论证图的同态像,结构检查与哈希的盲区几乎不重叠。 第五章比较衍生拓扑:不做检查时链形五件产品合计错 28.07 条事实,星形错 10.00 条;把编译成本与错误代价放进同一个式子,成本最低的通常是中枢形。第六章给出带父哈希的衍生账本,事后改动一条记录会连同下游一起被标出。第七章考辨刘向《别录》佚文「一人持本,一人读书,若怨家相对」与巴特利特的串行复现:两人独立对校剩下 0.12 的错误。第八章给出衍生品出处哈希清单、小工具「衍生漂移溯源核对器」与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Compounding of Drift: Serial Reproduction, p to the Power k, and Repair — Under independent errors that are never repaired, a relay with fidelity 0.95 per layer keeps only 0.7738 of its facts at the fifth layer; that figure is a warning, not a law. With any repair at all a long chain settles at a positive level, but repair itself requires going back to the book. Drift also tends to have a direction, which averaging cannot remove. The question left open is what exactly should be compared when going back to the book, and in what sense a piece of text should count as the same as the master. The next chapter answers with provenance hashes. (https://5dlibrary.com/answers/en/monograph-derivative-compilation-provenance-chain-homomorphism/c01/) 第1章 漂移的复利:串行复现、p 的 k 次方与修复回流 — 在独立误差、永不修复的假设下,每层保真 0.95 的接力在第五层只剩 0.7738;这个数字是一个警告,而不是定律。只要有一点修复,长链就停在正的水平上;但修复本身需要回到母书去对。走样又往往有方向,取平均救不了它。剩下的问题是:回到母书去对,具体对什么?一段文字和母书「相同」,应该在什么意义上相同?下一章用出处哈希回答。 (https://5dlibrary.com/answers/zh/monograph-derivative-compilation-provenance-chain-homomorphism/c01/) - Ch.2 Provenance Hashes: Pinning Facts Down as Atoms that Compare Byte for Byte — Chapter 2 writes facts as four-field atoms compared through canonical SHA-256 hashes, long enough prefixes and Merkle roots. (https://5dlibrary.com/answers/en/monograph-derivative-compilation-provenance-chain-homomorphism/c02/) 第2章 出处哈希:把事实钉成可以逐字节比对的原子 — 第二章把事实写成四字段原子,用规范化 SHA-256 哈希、足够长的前缀与默克尔根做比对。 (https://5dlibrary.com/answers/zh/monograph-derivative-compilation-provenance-chain-homomorphism/c02/) - Ch.3 What Detection Really Means: a Hash Sees Only What Was Anchored — Chapter 3 splits detection into anchor coverage and structural detection and shows that checks must go against the master, not the previous layer; on a set of sentence pairs the anchor checker has a precision of 1.000. (https://5dlibrary.com/answers/en/monograph-derivative-compilation-provenance-chain-homomorphism/c03/) 第3章 检出率的真相:哈希只看得见被锚定的东西 — 第三章把检出率分解为锚定覆盖率与结构检出,并说明核对必须对母书而不对上一层;锚点比对在一组句对上准确率为 1.000。 (https://5dlibrary.com/answers/zh/monograph-derivative-compilation-provenance-chain-homomorphism/c03/) - Ch.4 Structural Homomorphism: a Derivative Must Be a Homomorphic Image of the Book's Argument Graph — Reversals and flips are caught by the structure check every time and by the hash never, since the claim texts are unchanged. A randomly added edge is caught by the structure check in 0.9620 of trials; of the remainder, 0.0160 happened to add an edge that already exists in the master, which is not drift. When two claims swap places along with their ids, the structure check catches 0.9802; the rest are swaps of claims that lie on none of the edges the derivative draws, which changes nothing. A slipped digit is never caught by the structure check and always by the hash. Together the two checks catch every kind of edit except deletion at rates of 0.9620 or more. Deletion is reported by neither, because cutting is allowed and is reported through coverage. (https://5dlibrary.com/answers/en/monograph-derivative-compilation-provenance-chain-homomorphism/c04/) 第4章 结构同态:衍生品必须是母书论证图的同态像 — 倒置和翻转,结构检查每次都查得出,哈希一次也查不出,因为论断的文字没变。随机添加一条边,结构检查查出 0.9620;没查出的部分里,有 0.0160 是碰巧加上了一条母书里本来就有的边,这不算走样。两个论断连同编号互换位置,结构检查查出 0.9802,剩下的是那些互换的论断恰好不在衍生品画出的任何一条边上,等于什么也没改。数字被改动一位,结构检查一次也查不出,哈希每次都查出。两种检查合起来,除了删边以外,每一种改动的检出率都在 0.9620 以上。删边两者都不报,因为删边是允许的,它由覆盖率去报告。 (https://5dlibrary.com/answers/zh/monograph-derivative-compilation-provenance-chain-homomorphism/c04/) - Ch.5 Derivation Topology: Compile from the Master or from the Layer Before? — Chapter 5 compares derivation topologies: without checks the five products of a chain carry 28.07 wrong facts between them and those of a star 10.00; with compilation and error costs in one expression the lowest-cost tree is usually a hub. (https://5dlibrary.com/answers/en/monograph-derivative-compilation-provenance-chain-homomorphism/c05/) 第5章 衍生拓扑:从母书编译,还是从上一层编译 — 第五章比较衍生拓扑:不做检查时链形五件产品合计错 28.07 条事实,星形错 10.00 条;把编译成本与错误代价放进同一个式子,成本最低的通常是中枢形。 (https://5dlibrary.com/answers/zh/monograph-derivative-compilation-provenance-chain-homomorphism/c05/) - Ch.7 Classical Sources East and West: Liu Xiang's Collation and Bartlett's Serial Reproduction — Chapter 7 examines the lost fragment of Liu Xiang's Bielu, "one holds the copy, one reads the text, like enemies facing each other", and Bartlett's serial reproduction: two independent collators leave 0.12 of the errors. (https://5dlibrary.com/answers/en/monograph-derivative-compilation-provenance-chain-homomorphism/c07/) 第7章 东西方原典考辨:刘向校雠与巴特利特的串行复现 — 第七章考辨刘向《别录》佚文「一人持本,一人读书,若怨家相对」与巴特利特的串行复现:两人独立对校剩下 0.12 的错误。 (https://5dlibrary.com/answers/zh/monograph-derivative-compilation-provenance-chain-homomorphism/c07/) - Ch.8 Closing the Loop: the Provenance Hash Manifest and a Protocol for Refutation — Chapter 8 gives the provenance hash manifest, the Derivative Drift Provenance Checker and a pre-registered protocol. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/monograph-derivative-compilation-provenance-chain-homomorphism/c08/) 第8章 实证闭环:衍生品出处哈希清单与证伪方案 — 第八章给出衍生品出处哈希清单、小工具「衍生漂移溯源核对器」与预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/monograph-derivative-compilation-provenance-chain-homomorphism/c08/) ## 5DL-B-269 · Zero-Downtime Hot Migration and Rollback on Multi-Container Single Hosts: Lossless Schema Evolution behind a Shared Tunnel and a Ten-Second Stop-the-Bleeding Contract 中文书名:单机多容器零停机热迁与回滚、共享隧道架构生产库无损演进与十秒止血契约母典 DOI: pending Cite: Fan, Y. (2026). Zero-Downtime Hot Migration and Rollback on Multi-Container Single Hosts: Lossless Schema Evolution behind a Shared Tunnel and a Ten-Second Stop-the-Bleeding Contract (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Examples use fictional names, parameters are illustrative, and all calculations come from the book's programs, which connect to no server. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: A few containers on one host, a shared tunnel in front and a database behind make the most common production architecture of small teams, and the popular story is that a blue-green switch has zero downtime and rollback is instant. This book tests the founding claim that the downtime of a blue-green switch is not zero, that connection draining and the health-check interval set the worst interruption, and that expand-contract migration pushes irreversible steps later and markedly raises the rollback success rate. Checking the sources calls for a correction: Humble and Farley's Continuous Delivery sets out the practice but reports no comparative data on rollback success, so "markedly raises" is restated as a model claim, computed by the book's programs under explicit assumptions and handed over to pre-registered drills. Examples use fictional names, parameters are illustrative, and all calculations come from the book's programs, which connect to no server. This is a non-individual business and engineering reference; production changes must be released by a person under the operating protocol. Chapter 1 shows that a common probe configuration needs 35.0 seconds at worst to notice a dead backend, that a 10 second drain still cuts 9.5699 long connections in expectation, and that not one of four hundred fixed-seed switches failed zero requests. Chapter 2 draws migration as a rollbackability graph: with a 24 hour bake, expand-contract has a rollback failure rate 10.39 times lower than a one-shot migration, and with two coexisting containers only 8 of 560 interleavings are safe. Chapter 3 uses an availability Markov chain to show that the five-nines budget leaves only 12.1 seconds per bad release, and that expand-contract cuts yearly downtime from 2431.9 to 155.9 minutes. Chapter 4 shows that two colours on one host are capped by common-cause failure, saving only 262.1 minutes at a common-cause share of 0.5, and that a single tunnel connector with a short fixed retry recovers within ten seconds in a share of 0.9504, against 0.8518 for backoff with full jitter. Chapter 5 gives the backfill batch size, chunked-hash verification and the write freeze; a random sample of 1000 rows has only a 0.3017 chance of seeing one of 7 bad rows. Chapter 6 splits the ten-second contract into three terms: with an automatic decision and a warm old version the 95th percentile is 6.90 seconds, while a cold start meets the budget with probability only 0.0020; probing every second with four failures gives 0.0257 false triggers a month; and a rollback path drilled once a year works with probability only 0.2669. Chapter 7 examines "when a course runs out, change; changing, it opens a way; open, it endures" and the ship of Theseus, with expand-contract raising the best change frequency to 52.7 a year. Chapter 8 gives a ten-step local drill checklist, the distilled Hot-Migration Interruption Window Calculator and a pre-registered test, in which 59 consecutive drills without a breach are needed to say with 95 % confidence that the breach rate is below 0.05. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:一台主机上几个容器、前面一条共享隧道、后面一个数据库,是小团队最常见的生产架构,流行的说法是蓝绿切换停机为零、回滚瞬间完成。本书检验立项命题「蓝绿切换的停机并不为零,连接排空与健康检查间隔决定最坏中断;扩展收缩式迁移把不可逆步骤推后,回滚成功率显著提高」,核对原文后作出修正:Humble 与 Farley 的《持续交付》阐述了做法,但没有报告回滚成功率的对照数据,「显著提高」改写为模型命题,由书中程序在明确假设下算出,并交给预注册演练检验。书中只讲通用的数学模型,例子用虚构的 app-blue、app-green 与 db;参数为示意,全部计算由书中程序生成,程序不连接任何服务器;本书是非个人化的经营与工程参考,生产变更须按运维协议人工放行。 第一章说明常见探测配置最坏要 35.0 秒才发现后端已死,排空 10 秒仍期望切断 9.5699 个长连接,四百次固定种子切换中一次不失败的次数为零。第二章把迁移画成可回滚性图:观察期 24 小时时,扩展收缩式的回滚失败率比一步到位低 10.39 倍;两份容器并存时 560 种交错中只有 8 种安全。第三章用可用性马尔可夫链说明,五个 9 的预算摊到每次坏版本只有 12.1 秒,扩展收缩式把全年停机从 2431.9 分钟压到 155.9 分钟。第四章说明单机上双色副本受共因故障封顶,共因比例为 0.5 时只省 262.1 分钟;单个隧道连接器用固定短间隔重试,十秒内恢复的比例是 0.9504,带全抖动的退避只有 0.8518。 第五章给出回填批量、分块哈希核对与写入冻结的算法,随机抽 1000 行只有 0.3017 的把握看到 7 个坏行之一。第六章把十秒契约拆成三个条款:自动判定且旧版本保温时第 95 百分位 6.90 秒,冷启动时达成率只有 0.0020;每秒探一次、连续 4 次失败时每月误触发 0.0257 次;一年才演练一次的回滚路径能用概率只有 0.2669。第七章考辨「穷则变,变则通,通则久」与忒修斯之船,扩展收缩式把最优变更频率提到每年 52.7 次。第八章给出十步本地演练清单、蒸馏小工具「热迁中断窗口计算器」与预注册检验,连续 59 次零违约才能以 95 % 的把握说违约率低于 0.05。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Blue-Green Is Not Zero Downtime: the Interruption Window Set by Health Checks, Failure Thresholds and Connection Drain — Chapter 1 shows that a common probe configuration needs 35.0 seconds at worst to notice a dead backend, that a 10 second drain still cuts 9.5699 long connections in expectation, and that not one of four hundred fixed-seed switches failed zero requests. (https://5dlibrary.com/answers/en/single-host-zero-downtime-migration-rollback-ten-second-contract/c01/) 第1章 蓝绿切换并非零停机:健康检查、失败阈值与连接排空决定的中断时间分布 — 第一章说明常见探测配置最坏要 35.0 秒才发现后端已死,排空 10 秒仍期望切断 9.5699 个长连接,四百次固定种子切换中一次不失败的次数为零。 (https://5dlibrary.com/answers/zh/single-host-zero-downtime-migration-rollback-ten-second-contract/c01/) - Ch.2 Expand-Contract Migration: the Rollbackability Graph and Pushing Irreversible Steps to the End — Chapter 2 draws migration as a rollbackability graph: with a 24 hour bake, expand-contract has a rollback failure rate 10.39 times lower than a one-shot migration, and with two coexisting containers only 8 of 560 interleavings are safe. (https://5dlibrary.com/answers/en/single-host-zero-downtime-migration-rollback-ten-second-contract/c02/) 第2章 扩展收缩式迁移:可回滚性图与把不可逆步骤推到最后 — 第二章把迁移画成可回滚性图:观察期 24 小时时,扩展收缩式的回滚失败率比一步到位低 10.39 倍;两份容器并存时 560 种交错中只有 8 种安全。 (https://5dlibrary.com/answers/zh/single-host-zero-downtime-migration-rollback-ten-second-contract/c02/) - Ch.3 The Availability Markov Chain: How Repair Time Eats the Yearly Budget — Chapter 3 uses an availability Markov chain to show that the five-nines budget leaves only 12.1 seconds per bad release, and that expand-contract cuts yearly downtime from 2431.9 to 155.9 minutes. (https://5dlibrary.com/answers/en/single-host-zero-downtime-migration-rollback-ten-second-contract/c03/) 第3章 可用性马尔可夫链:修复时间如何吃掉全年预算 — 第三章用可用性马尔可夫链说明,五个 9 的预算摊到每次坏版本只有 12.1 秒,扩展收缩式把全年停机从 2431.9 分钟压到 155.9 分钟。 (https://5dlibrary.com/answers/zh/single-host-zero-downtime-migration-rollback-ten-second-contract/c03/) - Ch.4 Many Containers on One Host behind a Shared Tunnel: Common-Cause Failure, Reconnect Backoff and Memory Headroom for Dual Running — Chapter 4 shows that two colours on one host are capped by common-cause failure, saving only 262.1 minutes at a common-cause share of 0.5, and that a single tunnel connector with a short fixed retry recovers within ten seconds in a share of 0.9504, against 0.8518 for backoff with full jitter. (https://5dlibrary.com/answers/en/single-host-zero-downtime-migration-rollback-ten-second-contract/c04/) 第4章 单机多容器与共享隧道:共因故障、重连退避与双跑内存余量 — 第四章说明单机上双色副本受共因故障封顶,共因比例为 0.5 时只省 262.1 分钟;单个隧道连接器用固定短间隔重试,十秒内恢复的比例是 0.9504,带全抖动的退避只有 0.8518。 (https://5dlibrary.com/answers/zh/single-host-zero-downtime-migration-rollback-ten-second-contract/c04/) - Ch.5 Lossless Evolution of a Production Database: Backfill Pace, Dual-Write Verification and the Write-Freeze Window — Chapter 5 gives the backfill batch size, chunked-hash verification and the write freeze; a random sample of 1000 rows has only a 0.3017 chance of seeing one of 7 bad rows. (https://5dlibrary.com/answers/en/single-host-zero-downtime-migration-rollback-ten-second-contract/c05/) 第5章 生产库无损演进:回填节奏、双写核对与写入冻结窗口 — 第五章给出回填批量、分块哈希核对与写入冻结的算法,随机抽 1000 行只有 0.3017 的把握看到 7 个坏行之一。 (https://5dlibrary.com/answers/zh/single-host-zero-downtime-migration-rollback-ten-second-contract/c05/) - Ch.6 The Ten-Second Stop-the-Bleeding Contract: Convolving the Time Budget, False and Missed Alarms, and the Rot of Rollback Paths — Chapter 6 splits the ten-second contract into three terms: with an automatic decision and a warm old version the 95th percentile is 6.90 seconds, while a cold start meets the budget with probability only 0.0020; probing every second with four failures gives 0.0257 false triggers a month; and a rollback path drilled once a year works with probability only 0.2669. (https://5dlibrary.com/answers/en/single-host-zero-downtime-migration-rollback-ten-second-contract/c06/) 第6章 十秒止血契约:时间预算卷积、误报与漏报、回滚路径的锈蚀 — 第六章把十秒契约拆成三个条款:自动判定且旧版本保温时第 95 百分位 6.90 秒,冷启动时达成率只有 0.0020;每秒探一次、连续 4 次失败时每月误触发 0.0257 次;一年才演练一次的回滚路径能用概率只有 0.2669。 (https://5dlibrary.com/answers/zh/single-host-zero-downtime-migration-rollback-ten-second-contract/c06/) - Ch.7 The Classical Sources Examined: "When a Course Runs Out, Change" and the Ship of Theseus — Chapter 7 examines "when a course runs out, change; changing, it opens a way; open, it endures" and the ship of Theseus, with expand-contract raising the best change frequency to 52.7 a year. (https://5dlibrary.com/answers/en/single-host-zero-downtime-migration-rollback-ten-second-contract/c07/) 第7章 原典考辨:「穷则变,变则通,通则久」与忒修斯之船 — 第七章考辨「穷则变,变则通,通则久」与忒修斯之船,扩展收缩式把最优变更频率提到每年 52.7 次。 (https://5dlibrary.com/answers/zh/single-host-zero-downtime-migration-rollback-ten-second-contract/c07/) - Ch.8 The Empirical Loop: a Local Drill Checklist, the Hot-Migration Interruption Window Calculator and a Pre-Registered Test — Chapter 8 gives a ten-step local drill checklist, the distilled Hot-Migration Interruption Window Calculator and a pre-registered test, in which 59 consecutive drills without a breach are needed to say with 95 % confidence that the breach rate is below 0.05. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/single-host-zero-downtime-migration-rollback-ten-second-contract/c08/) 第8章 实证闭环:本地演练清单、热迁中断窗口计算器与预注册检验 — 第八章给出十步本地演练清单、蒸馏小工具「热迁中断窗口计算器」与预注册检验,连续 59 次零违约才能以 95 % 的把握说违约率低于 0.05。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/single-host-zero-downtime-migration-rollback-ten-second-contract/c08/) ## 5DL-B-270 · Peak-Shaving Periodic Quotas and Cross-Model Routing: Linear Programming against Reset Countdowns to Split Work between Heavy-Reasoning and Fast Models 中文书名:周期配额削峰填谷与跨模型路由、重置倒计时线性规划与重推理轻量模型算力分配母典 DOI: pending Cite: Fan, Y. (2026). Peak-Shaving Periodic Quotas and Cross-Model Routing: Linear Programming against Reset Countdowns to Split Work between Heavy-Reasoning and Fast Models (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Under illustrative parameters the linear programme gains 0.1668 to 0.2217 over arrival order, but only 0.41 % to 5.05 % over a fixed rule that sorts by complexity; most of the gain comes from sorting itself. Parameters are illustrative and contain no vendor's prices or quotas; all calculations are generated by the book's programs. Abstract: Teams that work with writing and reasoning tools often hold two tiers of service: a heavy-reasoning model with a small quota and a fast model with an ample quota that errs more on hard problems, both reset per cycle with leftovers expiring. The common practice is first come, first served. This book tests that practice and corrects its founding blueprint against the original literature: Dantzig's Linear Programming and Extensions of 1963 supplies theory and the simplex method, not measurements of quota routing; that a linear programme is no worse than fixed routing is a theorem inside the model, and what must be tested is the size of the gain. Under illustrative parameters the linear programme gains 0.1668 to 0.2217 over arrival order, but only 0.41 % to 5.05 % over a fixed rule that sorts by complexity; most of the gain comes from sorting itself. Parameters are illustrative and contain no vendor's prices or quotas; all calculations are generated by the book's programs. It is not individual business advice, covers only planning of quota already granted within the terms of each service, supplies no means of bypassing quotas or billing, and any change of routing in production needs sign-off by the person in charge. Chapter 1 computes the two failures of arrival order on busy and quiet days and reproduces the bound under which greedy captures only half in the worst order. Chapter 2 writes a simplex method in Fractions with Bland's rule and finds a shadow price of 9/40 for heavy quota. Chapter 3 uses dynamic programming to show that the unit value of quota falls as the reset nears, and that countdown re-solving needs a margin. Chapter 4 treats changing quotas, misjudged complexity and cascades; a cascade raises the optimum from 893.00 to 944.65 points. Chapter 5 treats peak shaving under short reset windows: deferral alone yields 863.29 points, pulling routine work forward raises it to 881.75, and a backlog turns expiring quota into output. Chapter 6 compares rules under random arrival: arrival order averages 0.8427 of the hindsight optimum, dynamic re-solving at geometric checkpoints 0.9875. Chapter 7 examines "years bring dearth and plenty, so grain is dear or cheap" from the Guanzi and Kantorovich's resolving multipliers of 1939. Chapter 8 gives the daily task routing table, the Quota Routing LP Planner and a pre-registered switchback protocol. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:用写作与推理工具做事的团队常有两档服务:重推理模型配额少,轻量模型配额宽裕但难题上容易出错,两档都按周期重置、过期作废。流行的做法是先到先用。本书检验这一做法,并按原始文献修正了立项蓝图:丹齐格 1963 年的《线性规划及其推广》给的是理论与单纯形法,不是配额路由的实测;「线性规划不劣于固定路由」是模型内的定理,要检验的是增益大小。示意参数下,线性规划相对先到先用多出 0.1668 到 0.2217,相对按复杂度分档的固定规则只多出 0.41 % 到 5.05 %,大头来自分档本身。书中参数为示意,不含任何厂商的价格或配额;全部计算由书中程序生成。本书属非个人化经营建议,只讨论在服务条款允许范围内对已获配额的规划,不提供任何绕过配额或计费的手法,生产路由变更须由主事者人工放行。 第一章算先到先用在忙日与闲日的两种失败,并复现最坏到达顺序下贪心只得一半的界。第二章用 Fraction 写出带布兰德规则的单纯形法,解出重推理配额的影子价格 9/40。第三章用动态规划说明配额的单位价值随重置临近而下跌,并说明倒计时重算须留余量。第四章讲配额变化、复杂度判别出错与级联升级,级联把最优产出从 893.00 点提到 944.65 点。 第五章讲短窗口重置下的削峰填谷:只延后时产出 863.29 点,再允许例行任务提前升到 881.75 点,积压池把作废的配额变成产出。第六章对照随机到达下的规则,先到先用平均得到事后最优的 0.8427,按几何检查点动态重算得到 0.9875。第七章考辨《管子·国蓄》「岁有凶穰,故谷有贵贱」与康托罗维奇 1939 年的解乘数。第八章给出每日任务路由表、小工具「配额路由线性规划器」与切换设计的预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Price of First Come, First Served: Exhaustion before the Reset and Waste at It — On the quiet day the picture reverses. Arrival order again exhausts H, this time by slot 8, because it spends H on a mass of routine work; only 5 of the 12 critical tasks get H, and output is 482.0 points. The fixed rule misses no critical task but leaves 48 H units to vanish at the reset. The third row is the instructive one: the plan built on the standing forecast leaves 128 units unused on the quiet day and yields 561.0 points, less than the fixed rule's 579.0. The fault is not the linear programme but the forecast it was fed. The plan reserves H for the usual number of complex tasks; on a quiet day fewer of them arrive, and the reserved units sit idle until the reset. Re-solved on the day's demand, the programme yields 581.4 points and leaves no H unit behind. (https://5dlibrary.com/answers/en/periodic-quota-peak-shaving-cross-model-routing-lp/c01/) 第1章 先到先用的代价:重置前耗尽与重置时作废 — 闲的那天,情形反过来。先到先用照样在第 8 个时段耗尽 H,因为它把 H 用在了大量例行任务上,12 个关键任务里只有 5 个用上 H,产出 482.0 点。固定路由一个关键任务都没落下,却在重置时剩下 48 个 H 单位作废。最有意思的是第三行:按常规预测解出的方案在闲日剩下 128 个单位作废,产出 561.0 点,反而不如固定路由的 579.0 点。原因不在线性规划,而在它用的需求预测:方案按常规日给复杂任务分配了 H 的份额,闲日的复杂任务没有那么多,多出来的 H 就躺在那里,一直躺到重置。用当天需求重新求解,产出 581.4 点,H 一个单位也不剩。 (https://5dlibrary.com/answers/zh/periodic-quota-peak-shaving-cross-model-routing-lp/c01/) - Ch.2 A Standard-Library Simplex: Exact Fractions, Bland's Rule and the Worst Case — Chapter 2 writes a simplex method in Fractions with Bland's rule and finds a shadow price of 9/40 for heavy quota. (https://5dlibrary.com/answers/en/periodic-quota-peak-shaving-cross-model-routing-lp/c02/) 第2章 标准库单纯形:分数精确运算、布兰德规则与最坏情形 — 第二章用 Fraction 写出带布兰德规则的单纯形法,解出重推理配额的影子价格 9/40。 (https://5dlibrary.com/answers/zh/periodic-quota-peak-shaving-cross-model-routing-lp/c02/) - Ch.3 Shadow Prices and the Reset Countdown: How the Marginal Value of Quota Falls with Time — Strong duality holds exactly on the problem of Chapter 1, and the shadow price of 9/40 for H is valid only between 400 and 900 units; with less quota it rises to 0.4375 and with more it falls to 0.0500. Tasks arriving online turn the plan into a dynamic programme: the nearer the reset and the larger the balance, the less one H unit is worth, down to nothing with ten or so tasks to come; with a tight balance it stays dear. Countdown re-solving approaches the optimum but needs a margin against fluctuation. The next chapter turns from time to the tasks themselves: complexity has to be judged, judgements err, and how much of the benefit of routing survives the errors? (https://5dlibrary.com/answers/en/periodic-quota-peak-shaving-cross-model-routing-lp/c03/) 第3章 影子价格与重置倒计时:配额的边际价值怎样随时间下跌 — 强对偶在第一章的问题上精确成立,H 的影子价格 9/40 只在 400 到 900 个单位之间有效,配额更少时升到 0.4375,更多时降到 0.0500。在线到达的任务把规划变成动态规划:离重置越近、余额越宽裕,一个 H 单位越不值钱,还剩十来个任务时可以跌到零;余额紧张时它仍然昂贵。倒计时重算能逼近最优,但要留出应付波动的余量。下一章把目光从时间转到任务本身:任务的复杂度要靠判别,判别会出错,出错以后路由还剩多少好处? (https://5dlibrary.com/answers/zh/periodic-quota-peak-shaving-cross-model-routing-lp/c03/) - Ch.4 Routing across Models: Complexity Classes, Label Errors and Cascades — Chapter 4 treats changing quotas, misjudged complexity and cascades; a cascade raises the optimum from 893.00 to 944.65 points. (https://5dlibrary.com/answers/en/periodic-quota-peak-shaving-cross-model-routing-lp/c04/) 第4章 跨模型路由:复杂度分档、标签误差与级联升级 — 第四章讲配额变化、复杂度判别出错与级联升级,级联把最优产出从 893.00 点提到 944.65 点。 (https://5dlibrary.com/answers/zh/periodic-quota-peak-shaving-cross-model-routing-lp/c04/) - Ch.5 Peak Shaving and Valley Filling: Deferral, Pulling Forward and a Backlog under Short Reset Windows — Chapter 5 treats peak shaving under short reset windows: deferral alone yields 863.29 points, pulling routine work forward raises it to 881.75, and a backlog turns expiring quota into output. (https://5dlibrary.com/answers/en/periodic-quota-peak-shaving-cross-model-routing-lp/c05/) 第5章 削峰填谷:短窗口重置下的延后、提前与积压池 — 第五章讲短窗口重置下的削峰填谷:只延后时产出 863.29 点,再允许例行任务提前升到 881.75 点,积压池把作废的配额变成产出。 (https://5dlibrary.com/answers/zh/periodic-quota-peak-shaving-cross-model-routing-lp/c05/) - Ch.6 Greedy against Linear Programming: Random Arrival, Prices Learned from a Sample, and Drift — Chapter 6 compares rules under random arrival: arrival order averages 0.8427 of the hindsight optimum, dynamic re-solving at geometric checkpoints 0.9875. (https://5dlibrary.com/answers/en/periodic-quota-peak-shaving-cross-model-routing-lp/c06/) 第6章 贪心与线性规划对照:随机到达、样本定价与分布漂移 — 第六章对照随机到达下的规则,先到先用平均得到事后最优的 0.8427,按几何检查点动态重算得到 0.9875。 (https://5dlibrary.com/answers/zh/periodic-quota-peak-shaving-cross-model-routing-lp/c06/) - Ch.7 Eastern and Western Classics: Gathering and Releasing in the Guanzi, and Kantorovich's Resolving Multipliers — Chapter 7 examines "years bring dearth and plenty, so grain is dear or cheap" from the Guanzi and Kantorovich's resolving multipliers of 1939. (https://5dlibrary.com/answers/en/periodic-quota-peak-shaving-cross-model-routing-lp/c07/) 第7章 东西方原典考辨:《管子·国蓄》的敛散与康托罗维奇的解乘数 — 第七章考辨《管子·国蓄》「岁有凶穰,故谷有贵贱」与康托罗维奇 1939 年的解乘数。 (https://5dlibrary.com/answers/zh/periodic-quota-peak-shaving-cross-model-routing-lp/c07/) - Ch.8 A Daily Task Routing Table: the Empirical Loop, the Distilled Tool and a Pre-registered Test — Chapter 8 gives the daily task routing table, the Quota Routing LP Planner and a pre-registered switchback protocol. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/periodic-quota-peak-shaving-cross-model-routing-lp/c08/) 第8章 每日任务路由表:实证闭环、蒸馏小工具与预注册检验 — 第八章给出每日任务路由表、小工具「配额路由线性规划器」与切换设计的预注册方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/periodic-quota-peak-shaving-cross-model-routing-lp/c08/) ## 5DL-B-271 · Rate Control for Private Messaging Channels without Account Bans: Token-Bucket Shaping, Platform-Published Limits and Consent-Based Reach Throughput 中文书名:私域通道防封号速率控制论、令牌桶限流与平台公示上限内的同意制触达吞吐母典 DOI: pending Cite: Fan, Y. (2026). Rate Control for Private Messaging Channels without Account Bans: Token-Bucket Shaping, Platform-Published Limits and Consent-Based Reach Throughput (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: An illustrative plan averaging 10.0 messages a minute still has 420 refused when sent in batches, and none once a token bucket of depth 20 shapes it. Parameters are illustrative, platform figures carry the document and the date read, and all numbers come from the book's programs, which read and write no file and use no network. Abstract: In this book's title, "without account bans" has one meaning only: staying unbanned by keeping the rules. The book studies compliant shaping alone: a token bucket keeps bursts under the limit a platform publishes, messages go only to people who agreed, stop means stop, quiet hours are kept, and sending slows when readers object. It does not study or describe any way of evading platform risk controls, anti-spam detection or rate limits. Checking the commissioning claim led to three corrections. The working idea of a "Poisson jitter fingerprint" became compliant shaping. Platforms do not publish a quality formula, saying only that ratings rest on usage, customer feedback and engagement and that significant negative feedback leads to limits, so the claim narrows to "quality is driven by recipients' feedback and engagement". And the per-recipient limit a platform publishes is itself written as a token bucket, which strengthens the claim. An illustrative plan averaging 10.0 messages a minute still has 420 refused when sent in batches, and none once a token bucket of depth 20 shapes it. Parameters are illustrative, platform figures carry the document and the date read, and all numbers come from the book's programs, which read and write no file and use no network. This is non-individual business advice, not legal or compliance advice; any change to a live sending rhythm needs a manual release by the person in charge. Chapter 1 gives the token bucket's bound of b plus r times W. Chapter 2 writes a Markov chain of four account tiers: as consent falls from 0.90 to 0.80, the expected daily limit drops from 99054 to 24692, while changing send times alone moves expected negative feedback by 0.000000. Chapter 3 solves Poisson arrivals exactly and shows that depth only shifts the tail of the leaky-bucket queue. Chapter 4 puts per-minute shaping and a moving 24-hour limit on people into one plan. Chapter 5 slows down on feedback with additive increase, multiplicative decrease and Wald's sequential test. Chapter 6 makes stop-means-stop, quiet hours, vulnerable readers to a person and no pretending into red lines: checking consent only at queueing a week ahead lifts objections to 2.4851 per thousand. Chapter 7 examines Mencius's parable of pulling up shoots and Erlang's Poisson paper of 1909; in the illustrative model the frequency with the most useful contacts is 2.25 messages per person a week. Chapter 8 gives a six-step weekly loop, the distilled "Compliant Reach Rate Shaper" and a pre-registered test needing 78390 messages per arm. Boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:书名里的「防封号」只有一个意思:通过遵守规则不被封。本书只研究合规整形:用令牌桶让突发不超过平台公示上限,只给同意者发,说停就停,静默时段不发,按读者反馈降速;不研究、不描述任何规避平台风控、反垃圾检测或速率限制的手法。立项命题核实后有三处修正:原题「泊松抖动指纹」改为合规整形;平台没有公开质量评级公式,只写明依据使用、用户反馈与互动,负面反馈多会被限制,所以命题收窄为「质量由收件人的反馈与互动驱动」;平台自己公布的单收件人限额就是令牌桶的写法,命题因此加强。示意计划每分钟平均 10.0 条,攒批猛发仍被拒收 420 条,接上桶深 20 的令牌桶后为零。书中参数为示意,平台数字注明文档与查阅日期;全部计算由书中程序生成,程序不读写文件、不联网;本书是非个人化经营建议,不是法律或合规意见,任何线上发送节奏的改动须由主事者人工复核放行。 第一章给出令牌桶的 b 加 r 乘 W 界。第二章写出四级账号等级的马尔可夫链:同意率从 0.90 降到 0.80,平均每日上限从 99054 跌到 24692;只改发送时间,期望负面反馈的变化是 0.000000。第三章在泊松到达下精确求解,证明桶深只平移漏桶队列的尾巴。第四章把每分钟整形与滑动 24 小时人数上限排进同一份计划。第五章用加性增、乘性减与瓦尔德序贯检验按反馈降速。第六章把说停就停、静默时段、弱势转人工与不伪装列为红线:只在排队时核对同意、提前一周排队,每千条负面反馈升到 2.4851 次。 第七章考辨《孟子》揠苗助长与埃尔朗 1909 年的泊松论文,示意模型下有用触达最多的频率是每人每周 2.25 条。第八章给出每周六步的闭环、蒸馏小工具「合规触达速率整形器」,以及每组需 78390 条消息的预先登记检验。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Bursts That Are Compliant on Average: Sliding-Window Caps and the b + rW Bound of the Token Bucket — The table lays out how the burst allowance relates to the long-run rate. In a window of 6 seconds one recipient could receive at most 46 messages, equal to 460.000 a minute, which looks alarming. Stretch the window to 600 seconds and the equivalent is only 14.500 a minute; stretch it to a whole day of 86400 seconds and the most is 14445, or 10.031 a minute, close to the long-run 10.000. The burst allowance is a share of 0.0031 of the day's bound. After a burst, the wait owed is 270 seconds. In other words, the burst allowance is a one-off credit, not extra capacity; once used, it is repaid at the normal rate. (https://5dlibrary.com/answers/en/private-channel-rate-control-token-bucket-consent-throughput/c01/) 第1章 平均合规的突发:滑动窗口上限与令牌桶的 b + rW 界 — 这张表把「突发额度」与「长期速率」的关系摊开了。在 6 秒的窗口里,一个收件人最多可能收到 46 条,折合每分钟 460.000 条,看起来很惊人;可窗口拉长到 600 秒,折合每分钟只剩 14.500 条;拉长到一天 86400 秒,最多 14445 条,折合每分钟 10.031 条,几乎就是长期速率 10.000。突发额度在一天的总量里只占 0.0031。突发之后要补的等待是 270 秒。换句话说,突发额度是一次性的信用,不是额外的产能;用掉了,就得按正常速率还回来。 (https://5dlibrary.com/answers/zh/private-channel-rate-control-token-bucket-consent-throughput/c01/) - Ch.2 Who Decides Account Quality: Consent Rates, Negative Feedback and a Markov Chain of Tiers — Chapter 2 writes a Markov chain of four account tiers: as consent falls from 0.90 to 0.80, the expected daily limit drops from 99054 to 24692, while changing send times alone moves expected negative feedback by 0.000000. (https://5dlibrary.com/answers/en/private-channel-rate-control-token-bucket-consent-throughput/c02/) 第2章 账号质量由谁决定:同意率、负面反馈与等级马尔可夫链 — 第二章写出四级账号等级的马尔可夫链:同意率从 0.90 降到 0.80,平均每日上限从 99054 跌到 24692;只改发送时间,期望负面反馈的变化是 0.000000。 (https://5dlibrary.com/answers/zh/private-channel-rate-control-token-bucket-consent-throughput/c02/) - Ch.3 Overflow under Poisson Arrivals: the Bucket Depth Only Shifts the Tail of the Leaky-Bucket Queue — Look at the three rows at a load of 0.60. With a pure leaky bucket, the probability that some message is still queued at the end of a minute is 0.048635, the mean queue 0.0921 messages and the mean extra wait 0.01535 minutes; at depth 10 the queueing probability falls to 0.000008, and at depth 20 it vanishes from six decimals. At a load of 0.80, the leaky bucket queues with probability 0.281302 and the mean extra wait is 0.10983 minutes; depth 10 brings this to 0.004549 and depth 20 to 0.000061. At a load of 0.90, the leaky bucket queues with probability 0.553500 and holds 3.1012 messages on average; depth 10 gives 0.074575, and depth 20 gives 0.009397 with a mean extra wait of 0.00558 minutes, under half a second. (https://5dlibrary.com/answers/en/private-channel-rate-control-token-bucket-consent-throughput/c03/) 第3章 泊松到达下的溢出:桶深只平移漏桶队列的尾巴 — 看负载 0.60 的三行。纯漏桶时,每分钟结束仍有消息排队的概率是 0.048635,平均排队 0.0921 条,平均多等 0.01535 分钟;桶深 10 时排队概率降到 0.000008;桶深 20 时,六位小数里已看不到。负载 0.80 时,漏桶的排队概率是 0.281302,平均多等 0.10983 分钟;桶深 10 降到 0.004549,桶深 20 降到 0.000061。负载 0.90 时,漏桶的排队概率是 0.553500,平均排队 3.1012 条;桶深 10 时 0.074575;桶深 20 时 0.009397,平均多等 0.00558 分钟,不到半秒。 (https://5dlibrary.com/answers/zh/private-channel-rate-control-token-bucket-consent-throughput/c03/) - Ch.4 Leaky Buckets, Token Buckets and Two Limits at Once: Per-Minute Shaping and Distinct Recipients in 24 Hours — Chapter 4 puts per-minute shaping and a moving 24-hour limit on people into one plan. (https://5dlibrary.com/answers/en/private-channel-rate-control-token-bucket-consent-throughput/c04/) 第4章 漏桶、令牌桶与双重上限:每分钟整形与 24 小时独立收件人 — 第四章把每分钟整形与滑动 24 小时人数上限排进同一份计划。 (https://5dlibrary.com/answers/zh/private-channel-rate-control-token-bucket-consent-throughput/c04/) - Ch.5 Slowing Down on Feedback: Additive Increase, Multiplicative Decrease and Wald's Sequential Test — A small sender with 300 messages a day, whose true rate holds firmly at 0.10%, sees a single day read above 0.30% with probability 0.2593, about one false alarm every four days; pooling a week brings this down to 0.005834. At 1000 a day the daily false-alarm chance is 0.0802 and the weekly one 0.000014. At 5000 a day the daily chance is only 0.0002. The last column gives the reason: the 0.30% line lies only 1.10 standard errors above the true value at 300 a day and 8.95 at 20000 a day. (https://5dlibrary.com/answers/en/private-channel-rate-control-token-bucket-consent-throughput/c05/) 第5章 按反馈降速:加性增、乘性减与瓦尔德序贯检验 — 一天只发 300 封的小发件人,真实投诉率明明守在 0.10%,单日读数越过 0.30% 的概率却有 0.2593,大约每四天就有一天误报一次;把一周合并起来读,概率降到 0.005834。一天 1000 封时,单日误报概率 0.0802,一周合并后 0.000014。到一天 5000 封,单日误报概率只剩 0.0002。最后一列说明原因:0.30% 这条线离真实值有几个标准误,一天 300 封时只有 1.10 个,一天 20000 封时有 8.95 个。 (https://5dlibrary.com/answers/zh/private-channel-rate-control-token-bucket-consent-throughput/c05/) - Ch.6 Red Lines outside the Model: Stop Means Stop, Quiet Hours, Vulnerable Readers to a Person, and No Pretending — Chapter 6 makes stop-means-stop, quiet hours, vulnerable readers to a person and no pretending into red lines: checking consent only at queueing a week ahead lifts objections to 2.4851 per thousand. (https://5dlibrary.com/answers/en/private-channel-rate-control-token-bucket-consent-throughput/c06/) 第6章 红线不进模型:退订即停、静默时段、弱势转人工与不伪装 — 第六章把说停就停、静默时段、弱势转人工与不伪装列为红线:只在排队时核对同意、提前一周排队,每千条负面反馈升到 2.4851 次。 (https://5dlibrary.com/answers/zh/private-channel-rate-control-token-bucket-consent-throughput/c06/) - Ch.7 Pulling Up Shoots and Telephone Traffic: Mencius's Parable and Erlang's Paper of 1909 — Chapter 7 examines Mencius's parable of pulling up shoots and Erlang's Poisson paper of 1909; in the illustrative model the frequency with the most useful contacts is 2.25 messages per person a week. (https://5dlibrary.com/answers/en/private-channel-rate-control-token-bucket-consent-throughput/c07/) 第7章 揠苗与话务:《孟子》的寓言与埃尔朗 1909 年的论文 — 第七章考辨《孟子》揠苗助长与埃尔朗 1909 年的泊松论文,示意模型下有用触达最多的频率是每人每周 2.25 条。 (https://5dlibrary.com/answers/zh/private-channel-rate-control-token-bucket-consent-throughput/c07/) - Ch.8 Closing the Loop: a Weekly Review, the Rate Shaper and a Pre-Registered Test — Chapter 8 gives a six-step weekly loop, the distilled "Compliant Reach Rate Shaper" and a pre-registered test needing 78390 messages per arm. Boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/private-channel-rate-control-token-bucket-consent-throughput/c08/) 第8章 实证闭环:周度复盘、整形器用法与预注册检验 — 第八章给出每周六步的闭环、蒸馏小工具「合规触达速率整形器」,以及每组需 78390 条消息的预先登记检验。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/private-channel-rate-control-token-bucket-consent-throughput/c08/) ## 5DL-B-272 · Disaster Recovery for Digital Assets: Shamir Secret-Sharing Thresholds, Off-Site Incremental Snapshots and Self-Rescue from Single Points of Loss 中文书名:数字资产灾难恢复与密钥分片、沙米尔秘密共享门限与异地增量快照单点失联自救母典 DOI: pending Cite: Fan, Y. (2026). Disaster Recovery for Digital Assets: Shamir Secret-Sharing Thresholds, Off-Site Incremental Snapshots and Self-Rescue from Single Points of Loss (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 6 computes self-rescue after one share goes missing: losing the fastest holder is like raising the threshold by one, and the illustrative rescue window is 263 hours. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: The advice most often heard about safeguarding digital assets is to copy the secret needed for recovery several times and keep the copies apart. This book splits that advice into two ledgers, loss and leakage, recomputes it with the binomial distribution, finite-field interpolation, Markov chains and order statistics, and corrects the project's claim: under independent losses three copies in three places are the best protection against loss, a threshold scheme trades loss for leakage relative to three copies, and winning on both ledgers at once needs more pieces. Every secret in the book is a toy integer, all parameters are illustrative, and all calculations are generated by its programs; the book is a mathematical model and a drill sheet, contacts and describes no real server, requires human release for every production backup or key operation, and is not individual legal or financial advice. With illustrative yearly rates of 0.05 for loss and 0.02 for leakage per piece, Chapter 1 finds a loss probability of 0.000125 and a leak probability of 0.058808 for three copies, 0.007250 and 0.001184 for two-of-three, and two-of-five as the smallest scheme that beats three copies on both. Chapter 2 splits and joins in the field modulo 2147483647, proves by exhaustive counting that fewer pieces than the threshold reveal nothing, and shows that locating one crafted bad share costs two units of redundancy. Chapter 3 handles unequal holders, shared buildings and collusion among acquaintances: three-of-five split three and two over two places has a loss probability of 0.012547, worse than one place. Chapter 4 computes off-site incremental snapshots: hourly incrementals push the worst recovery time to 51.11 hours, an hourly chain of 168 links is intact with probability only 0.8453 without read-back checks, and two copies last 30.0 years on average without repair and 10430.0 years with weekly repair. Chapter 5 shows that leakage accumulates: three-of-five left unrefreshed for twenty years leaks with probability 0.208484, and yearly refresh brings that to 0.001551. Chapter 6 computes self-rescue after one share goes missing: losing the fastest holder is like raising the threshold by one, and the illustrative rescue window is 263 hours. Chapter 7 examines the tiger tally of the Shiji and the split tally of the Dialogue of the Exchequer. Chapter 8 gives the drill procedure, the Threshold Share Risk Calculator and a pre-registered test of 30 drills with at least 28 successes. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:关于数字资产保管,最常听到的建议是把恢复所需的秘密多抄几份、分开存放。本书把这条建议拆成丢失与泄露两本账,用二项分布、有限域插值、马尔可夫链与顺序统计量重新计算,并修正立项命题:三份副本分放三处在独立丢失下反而最防丢,门限分片相对三份副本是用丢失换泄露,要在两本账上同时胜出需要更多份数。书中所有秘密都是玩具整数,参数均为示意,全部计算由书中程序生成;本书是数学模型与演练清单,不连接也不描述任何真实服务器,生产环境的备份与密钥操作须人工放行,不构成个人化法律或理财建议。 第一章在每份每年丢失 0.05、泄露 0.02 的示意参数下,算出三份副本的年丢失概率 0.000125、泄露概率 0.058808,三取二分别为 0.007250 与 0.001184,五取二是两项都胜过三份副本的最小方案。第二章在模 2147483647 的域上分与合,并穷举证明不足门限的份不泄露信息;定位一份精心构造的坏份需要两份冗余。第三章处理异质持有人、同址共灾与熟人串通:五取三按三加二分在两处,丢失概率 0.012547,比放在一处还差。第四章计算异地增量快照:每小时增量把最坏恢复时间推到 51.11 小时,168 环的小时链未经回读校验时只有 0.8453 的概率完好;两份副本从不修复平均 30.0 年丢失,一周内修复则是 10430.0 年。 第五章说明泄露会累积:五取三二十年不刷新的泄露概率 0.208484,每年刷新降到 0.001551。第六章计算单点失联后的自救:失去最快的持有人相当于把门限提高一级,示意自救窗口 263 小时。第七章考辨《史记·魏公子列传》的虎符与《财政署对话录》的劈木记账符。第八章给出演练流程、门限分片风险计算器与预注册检验:30 次演练至少 28 次成功。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Illusion of Three Copies: a Binomial Ledger of Loss and Leak — Three copies in three places are, under independent losses, the practice that best protects against loss: with the illustrative parameters the yearly loss probability is only 0.000125, but the leak probability rises to 0.058808, which is 2.940 times that of a single copy. Two-of-three improves both ledgers against one copy, and against three copies it trades loss for leakage. To beat three copies on both ledgers at once, at least 5 pieces are needed under these parameters, and two-of-five is the smallest such scheme. Shamir's paper supports the mathematics of threshold schemes and the dilemma of key management; it does not support the claim that a threshold scheme dominates under every set of parameters. The next chapter opens a share and looks inside: finite fields, polynomials and interpolation, why $k-1$ pieces reveal nothing, and whether a share that has been quietly corrupted can be detected. (https://5dlibrary.com/answers/en/digital-asset-disaster-recovery-shamir-threshold-snapshots/c01/) 第1章 三份副本的错觉:丢失与泄露的二项分布账 — 抄三份分放三处,在独立丢失的假定下是最防丢的做法:示意参数下的年丢失概率只有 0.000125,但泄露概率升到 0.058808,是一份副本的 2.940 倍。三取二相对一份副本两项同时变好,相对三份副本则是用丢失换泄露。要在两本账上都赢过三份副本,示意参数下至少要 5 份,五取二是最小的方案。沙米尔的论文支撑门限方案的数学性质与钥匙管理的两难,不支撑「门限在任何参数下都同时占优」。下一章打开分片的内部:有限域、多项式与插值,看 $k-1$ 份为什么不泄露任何信息,以及一份被悄悄改坏的分片能不能被发现。 (https://5dlibrary.com/answers/zh/digital-asset-disaster-recovery-shamir-threshold-snapshots/c01/) - Ch.2 Lagrange Interpolation over a Finite Field: Making, Rebuilding and Checking Shares — Chapter 2 splits and joins in the field modulo 2147483647, proves by exhaustive counting that fewer pieces than the threshold reveal nothing, and shows that locating one crafted bad share costs two units of redundancy. (https://5dlibrary.com/answers/en/digital-asset-disaster-recovery-shamir-threshold-snapshots/c02/) 第2章 有限域上的拉格朗日插值:分片如何生成、还原与验错 — 第二章在模 2147483647 的域上分与合,并穷举证明不足门限的份不泄露信息;定位一份精心构造的坏份需要两份冗余。 (https://5dlibrary.com/answers/zh/digital-asset-disaster-recovery-shamir-threshold-snapshots/c02/) - Ch.3 Loss and Collusion: Unequal Holders, Shared Buildings and Acquaintance Graphs — Chapter 3 handles unequal holders, shared buildings and collusion among acquaintances: three-of-five split three and two over two places has a loss probability of 0.012547, worse than one place. (https://5dlibrary.com/answers/en/digital-asset-disaster-recovery-shamir-threshold-snapshots/c03/) 第3章 失联与共谋:异质持有人、同址共灾与熟人网络 — 第三章处理异质持有人、同址共灾与熟人串通:五取三按三加二分在两处,丢失概率 0.012547,比放在一处还差。 (https://5dlibrary.com/answers/zh/digital-asset-disaster-recovery-shamir-threshold-snapshots/c03/) - Ch.4 Off-Site Incremental Snapshots: Recovery Point, Recovery Time and Mean Time to Data Loss — Chapter 4 computes off-site incremental snapshots: hourly incrementals push the worst recovery time to 51.11 hours, an hourly chain of 168 links is intact with probability only 0.8453 without read-back checks, and two copies last 30.0 years on average without repair and 10430.0 years with weekly repair. (https://5dlibrary.com/answers/en/digital-asset-disaster-recovery-shamir-threshold-snapshots/c04/) 第4章 异地增量快照:恢复点、恢复时间与数据丢失的平均间隔 — 第四章计算异地增量快照:每小时增量把最坏恢复时间推到 51.11 小时,168 环的小时链未经回读校验时只有 0.8453 的概率完好;两份副本从不修复平均 30.0 年丢失,一周内修复则是 10430.0 年。 (https://5dlibrary.com/answers/zh/digital-asset-disaster-recovery-shamir-threshold-snapshots/c04/) - Ch.5 Leaks Accumulate: Proactive Share Refresh — Chapter 5 shows that leakage accumulates: three-of-five left unrefreshed for twenty years leaks with probability 0.208484, and yearly refresh brings that to 0.001551. (https://5dlibrary.com/answers/en/digital-asset-disaster-recovery-shamir-threshold-snapshots/c05/) 第5章 泄露会累积:主动刷新分片 — 第五章说明泄露会累积:五取三二十年不刷新的泄露概率 0.208484,每年刷新降到 0.001551。 (https://5dlibrary.com/answers/zh/digital-asset-disaster-recovery-shamir-threshold-snapshots/c05/) - Ch.6 Self-Rescue after a Single Point of Loss: Time to Quorum and the Exposed Window — Chapter 6 computes self-rescue after one share goes missing: losing the fastest holder is like raising the threshold by one, and the illustrative rescue window is 263 hours. (https://5dlibrary.com/answers/en/digital-asset-disaster-recovery-shamir-threshold-snapshots/c06/) 第6章 单点失联的自救:凑齐门限的时间与暴露窗口 — 第六章计算单点失联后的自救:失去最快的持有人相当于把门限提高一级,示意自救窗口 263 小时。 (https://5dlibrary.com/answers/zh/digital-asset-disaster-recovery-shamir-threshold-snapshots/c06/) - Ch.7 Classics East and West: the Tiger Tally of Shiji 77 and the Split Tally of the Exchequer — The tiger tally is a two-of-two scheme in which the verifier holds the other half, so stealing the ruler's half is enough to forge an order. Jin Bi's doubt was anomaly detection, Zhu Hai's mallet was coercion of the verifier, and the step Hou Ying feared, asking the king again, was out-of-band confirmation. In the illustrative model a forged order succeeds with probability 0.0500 on the tally alone and 0.0025 with a call back to the capital. The split tally encoded amounts in notches, authenticated by grain and achieved off-site keeping by giving each party a half. A forged half matches all of 20 features with probability only 0.0000009537, keeping both halves in one building makes their loss five times as likely as keeping them apart, and the fire of 1834 shows that destroying everything in one place is just as dangerous. The final chapter gathers the book's models into a procedure that can be drilled, a small tool and a pre-registered test plan. (https://5dlibrary.com/answers/en/digital-asset-disaster-recovery-shamir-threshold-snapshots/c07/) 第7章 东西方原典考辨:窃符救赵的虎符与财政署的劈木记账符 — 虎符是一个二取二方案,验证者自己持有另一半,于是窃得国君那一半就足以伪造调令;晋鄙的怀疑是一道异常检测,朱亥的铁椎是对验证者的胁迫,侯生最怕的「复请」是带外核实。示意模型中,只靠虎符时伪造得逞的概率是 0.0500,加上回都城核实降到 0.0025。记账符用缺口编码金额,用木纹验真,用两方各持一半实现异地保管;伪造的一半在 20 处特征上全部吻合的概率只有 0.0000009537,两半同楼保管的丢失概率是分开保管的五倍,而 1834 年的大火说明集中销毁同样危险。下一章把全书的模型收拢成一份可演练的流程、一个小工具与一份预注册检验方案。 (https://5dlibrary.com/answers/zh/digital-asset-disaster-recovery-shamir-threshold-snapshots/c07/) - Ch.8 Closing the Loop: Threshold Drills, the Risk Calculator and a Pre-Registered Test — Chapter 8 gives the drill procedure, the Threshold Share Risk Calculator and a pre-registered test of 30 drills with at least 28 successes. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/digital-asset-disaster-recovery-shamir-threshold-snapshots/c08/) 第8章 实证闭环:门限方案演练、风险计算器与预注册检验 — 第八章给出演练流程、门限分片风险计算器与预注册检验:30 次演练至少 28 次成功。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/digital-asset-disaster-recovery-shamir-threshold-snapshots/c08/) ## 5DL-B-273 · Cold-Start Dynamics of Short-Video Traffic Pools: Step Thresholds in Completion, Engagement and Replay and a Zero-Ad-Spend Breakout Test Matrix 中文书名:全域短视频赛马池冷启动动力学、完播互动复播阶跃阈值与零投流破圈测试矩阵母典 DOI: pending Cite: Fan, Y. (2026). Cold-Start Dynamics of Short-Video Traffic Pools: Step Thresholds in Completion, Engagement and Replay and a Zero-Ad-Spend Breakout Test Matrix (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 3 shows that a hard threshold is a ramp in the data, whose width narrows from 0.124 to 0.023 as the batch grows from 100 to 3,000 viewers; the illustrative base video breaks out with probability 0.03585, with the bottleneck at the fourth gate. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: Among short-video operators, one piece of advice has almost the standing of common sense: the cold start is won by volume, so post daily and post often and a hit will come. This book refutes the assumption behind it, that every video has roughly the same fixed chance of breaking out. It writes promotion through traffic pools as a product of several threshold gates: a small gain in quality is amplified once at every gate and pays nonlinearly, while posting more at fixed quality adds chances only linearly. All thresholds, pool sizes and weights in the book are illustrative, and each platform's own published rules prevail; all calculations are generated by the book's programs; the book is not individual business advice, deals only with real content quality and compliant testing, and covers no way of inflating numbers or working around platform rules. The project blueprint attributed the claim to the 2014 cascade-prediction paper by Cheng and colleagues; on checking the original, the book recasts it as a model inference, since that paper actually found that early temporal features predict whether a cascade doubles with an accuracy of 0.795, while content features are weak. Chapter 1 shows that in the middle of the quality range, 0.1 standard deviation of quality multiplies the breakout probability by 1.89, while doubling the posting rate multiplies it by 2. Chapter 2 uses branching processes to show that raising transmission strength by a tenth can multiply the chance of a large cascade by 4.1. Chapter 3 shows that a hard threshold is a ramp in the data, whose width narrows from 0.124 to 0.023 as the batch grows from 100 to 3,000 viewers; the illustrative base video breaks out with probability 0.03585, with the bottleneck at the fourth gate. Chapter 4 shows that a 7-second video completes 2.89 times as often as a 60-second one purely because of length, and that telling an interaction rate of 0.03 from 0.04 takes 5,301 views. Chapter 5 finds an interior optimum for expected breakouts under a budget of production time (1.189 and 1.334 a week in the illustrations); the polished plan goes a quarter without a breakout with probability 0.069, the high-volume plan with 0.687. Chapter 6 presents a zero-ad-spend Plackett-Burman matrix of twelve videos a week and three factors, whose main-effect variance is 0.333 of a single video's; detecting a completion difference of 0.04 takes 11 weeks. Chapter 7 examines Sunzi's "first make yourself unbeatable" (a video of quality 1.0 breaks out with probability only 0.0645) and the fixed point 0.522588 that Watson missed. Chapter 8 gives the weekly procedure, the Traffic-Pool Promotion Probability Calculator and a pre-registered refutation plan: the product model predicts a top-third rate of 0.3449 at the third level against 0.1451 under an unchanged ratio, testable with 72 videos per third over 18 weeks. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:短视频经营者之间流传一条近乎常识的建议:冷启动靠数量,日更多发,碰上爆款的机会自然更大。本书证伪它背后的假设,即每条视频破圈的机会大致固定。书中把流量池晋级写成多道阈值门的连乘:质量的小幅提升在每一道门上各被放大一次,回报是非线性的;质量不变时多发只线性增加机会。书中所有门槛、池子规模与权重均为示意,平台规则以其官方公示为准;全部计算由书中程序生成;本书是非个人化经营建议,只讨论真实内容质量与合规测试,不涉及任何刷量或规避平台规则的做法。立项蓝图把命题归在程等人 2014 年的级联预测论文名下,核对原文后本书改为模型推论:该文实际发现的是早期时间特征预测级联翻倍的准确率达 0.795,内容特征很弱。 第一章显示,在中段质量提高 0.1 个标准差使破圈概率乘以 1.89,而发布量翻倍只乘以 2。第二章用分支过程说明转发强度提高一成可使大级联的概率乘以 4.1。第三章说明硬门槛在数据上是斜坡,批次从 100 人增至 3,000 人时放行区间宽度从 0.124 收窄到 0.023;示意基准视频的破圈概率为 0.03585,瓶颈在第四道门。第四章说明 7 秒视频的完播率是 60 秒视频的 2.89 倍,完全来自长度,区分互动率 0.03 与 0.04 需要 5,301 次曝光。 第五章在制作时间约束下求出期望破圈数的内部最优点(示意下每周 1.189 次与 1.334 次),精做方案一个季度零破圈的概率为 0.069,多发方案为 0.687。第六章给出每周十二条、三个因子的零投流普拉克特—伯曼矩阵,主效应方差为单条的 0.333 倍,测出完播率 0.04 的差别需要 11 周。第七章考辨《孙子·形篇》的「先为不可胜」(质量 1.0 的视频破圈概率仅 0.0645)与沃森漏掉的不动点 0.522588。第八章给出周度流程、流量池晋级概率计算器与预注册证伪方案:连乘模型预言第三级上档到达率 0.3449,对照不变比率下的 0.1451,每档 72 条、18 周即可检验。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Promotion as a Product: Four Threshold Gates and the Nonlinear Return to Quality — Chapter 1 shows that in the middle of the quality range, 0.1 standard deviation of quality multiplies the breakout probability by 1.89, while doubling the posting rate multiplies it by 2. (https://5dlibrary.com/answers/en/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c01/) 第1章 连乘的晋级:四道阈值门与质量的非线性回报 — 第一章显示,在中段质量提高 0.1 个标准差使破圈概率乘以 1.89,而发布量翻倍只乘以 2。 (https://5dlibrary.com/answers/zh/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c01/) - Ch.2 Branching Processes: Galton–Watson Cascades, Criticality and Overdispersion — Chapter 2 uses branching processes to show that raising transmission strength by a tenth can multiply the chance of a large cascade by 4.1. (https://5dlibrary.com/answers/en/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c02/) 第2章 分支过程:高尔顿—沃森级联、临界点与过度离散 — 第二章用分支过程说明转发强度提高一成可使大级联的概率乘以 4.1。 (https://5dlibrary.com/answers/zh/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c02/) - Ch.3 Why the Step Becomes a Ramp: Finite Batches, Logistic Promotion and the Bottleneck Gate — Chapter 3 shows that a hard threshold is a ramp in the data, whose width narrows from 0.124 to 0.023 as the batch grows from 100 to 3,000 viewers; the illustrative base video breaks out with probability 0.03585, with the bottleneck at the fourth gate. (https://5dlibrary.com/answers/en/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c03/) 第3章 台阶为何变成斜坡:有限批次、logistic 晋级与瓶颈门 — 第三章说明硬门槛在数据上是斜坡,批次从 100 人增至 3,000 人时放行区间宽度从 0.124 收窄到 0.023;示意基准视频的破圈概率为 0.03585,瓶颈在第四道门。 (https://5dlibrary.com/answers/zh/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c03/) - Ch.4 Completion, Engagement, Replay: the Statistics of Three Rulers — Chapter 4 shows that a 7-second video completes 2.89 times as often as a 60-second one purely because of length, and that telling an interaction rate of 0.03 from 0.04 takes 5,301 views. (https://5dlibrary.com/answers/en/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c04/) 第4章 完播、互动、复播:三把尺子的统计性质 — 第四章说明 7 秒视频的完播率是 60 秒视频的 2.89 倍,完全来自长度,区分互动率 0.03 与 0.04 需要 5,301 次曝光。 (https://5dlibrary.com/answers/zh/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c04/) - Ch.5 More Posts or Better Posts: the Linear Ledger and the Multiplicative Ledger — Chapter 5 finds an interior optimum for expected breakouts under a budget of production time (1.189 and 1.334 a week in the illustrations); the polished plan goes a quarter without a breakout with probability 0.069, the high-volume plan with 0.687. (https://5dlibrary.com/answers/en/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c05/) 第5章 多发与精做:线性账与乘法账 — 第五章在制作时间约束下求出期望破圈数的内部最优点(示意下每周 1.189 次与 1.334 次),精做方案一个季度零破圈的概率为 0.069,多发方案为 0.687。 (https://5dlibrary.com/answers/zh/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c05/) - Ch.6 The Zero-Ad-Spend Test Matrix: a Twelve-Run Plackett–Burman Design and Its Sample Size — Chapter 6 presents a zero-ad-spend Plackett-Burman matrix of twelve videos a week and three factors, whose main-effect variance is 0.333 of a single video's; detecting a completion difference of 0.04 takes 11 weeks. (https://5dlibrary.com/answers/en/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c06/) 第6章 零投流测试矩阵:十二条普拉克特—伯曼设计与样本量 — 第六章给出每周十二条、三个因子的零投流普拉克特—伯曼矩阵,主效应方差为单条的 0.333 倍,测出完播率 0.04 的差别需要 11 周。 (https://5dlibrary.com/answers/zh/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c06/) - Ch.7 Classical Sources East and West: Sunzi's "Disposition" and Galton–Watson — Chapter 7 examines Sunzi's "first make yourself unbeatable" (a video of quality 1.0 breaks out with probability only 0.0645) and the fixed point 0.522588 that Watson missed. (https://5dlibrary.com/answers/en/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c07/) 第7章 东西方原典考辨:《孙子·形篇》与高尔顿—沃森 — 第七章考辨《孙子·形篇》的「先为不可胜」(质量 1.0 的视频破圈概率仅 0.0645)与沃森漏掉的不动点 0.522588。 (https://5dlibrary.com/answers/zh/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c07/) - Ch.8 Closing the Loop: the Weekly Sheet, the Promotion Calculator and a Pre-registered Refutation Plan — Chapter 8 gives the weekly procedure, the Traffic-Pool Promotion Probability Calculator and a pre-registered refutation plan: the product model predicts a top-third rate of 0.3449 at the third level against 0.1451 under an unchanged ratio, testable with 72 videos per third over 18 weeks. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c08/) 第8章 实证闭环:周度测试表、晋级概率计算器与预注册证伪方案 — 第八章给出周度流程、流量池晋级概率计算器与预注册证伪方案:连乘模型预言第三级上档到达率 0.3449,对照不变比率下的 0.1451,每档 72 条、18 周即可检验。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/short-video-traffic-pool-cold-start-step-thresholds-test-matrix/c08/) ## 5DL-B-274 · Compliant Rewriting for Platform-Sensitive Terms: Absolute-Term Automata under the Advertising Law, Meaning-Preserving Rewrites and Conversion Elasticity 中文书名:平台敏感词合规改写、广告法极限词自动机与语义保真转化弹性母典 DOI: pending Cite: Fan, Y. (2026). Compliant Rewriting for Platform-Sensitive Terms: Absolute-Term Automata under the Advertising Law, Meaning-Preserving Rewrites and Conversion Elasticity (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its lexicons, copy, facts and parameters are illustrative; all calculations are generated by its programs; it is not individual legal advice and supplies no way of slipping past platform review. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: Many online shops believe that writing "best", "number one", "top-grade" or "lowest price on the whole web" into a title sells more, and that deleting those words costs conversion. This book refutes the two assumptions behind that belief: that the conversion gain from absolute terms is certain and sizeable, and that plain wording always loses. With illustrative parameters, one absolute term must lift conversion by 0.0333 in relative terms to offset a monthly incident probability of 0.010 times the incident cost; published research suggests that puffery's effect cancels out across buyers, and in a mixture model the lift exceeds break-even in only 0.042 of draws. On the law, the book states only what it is confident of: the Advertising Law prohibits terms such as "national-level", "highest-grade" and "best", and the 2023 enforcement guidelines bring near-synonyms under the same rule while noting that truthful sales figures under stated limits are not covered. It cites no article numbers and no penalty amounts, and a human check before listing governs. Its lexicons, copy, facts and parameters are illustrative; all calculations are generated by its programs; it is not individual legal advice and supplies no way of slipping past platform review. Chapter 1 draws up the risk ledger. Chapter 2 builds an Aho-Corasick multi-pattern automaton: the illustrative lexicon compiles into 35 states, and as the lexicon grows from 10 to 800 terms the term-by-term search rises to 813.46 comparisons per symbol while the automaton never exceeds 1.998 steps. Chapter 3 measures false kills and misses: after audit additions the miss rate in a new week falls from 0.310 to 0.076 with a false-kill rate of 0.074, and bringing the 95 % upper bound on the miss rate below 0.01 needs 299 violating lines even with zero misses. Chapter 4 treats rewriting as constrained optimisation, with sourced facts, a whole-line re-scan and a title-length limit, solved as a multiple-choice knapsack. Chapter 5 shows that surface similarity rewards the violation; a score built on proposition slots has a rank correlation of minus 0.949 with edit-distance similarity. Chapter 6 computes conversion elasticity: in an illustrative logit model a factual rewrite is at worst 0.0286 below the original; the standard error of 0.1071 of a single A/B test dwarfs the spread of true effects, observed and true effects correlate at only 0.226, and rewrites should be tested pooled by class. Chapter 7 examines "if names are not correct, speech does not flow" in the Analects, the Xunzi on correcting names, and Orwell's "Politics and the English Language". Chapter 8 gives the weekly procedure linking a compliant-rewrite library to the persuasive-copy checker, the distilled Compliant Rewriting Automaton, and a pre-registered non-inferiority test: at a relative margin of 0.10 and power of 0.80 each arm needs 39,981 sessions. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:很多网店相信,标题里写上「最佳」「第一」「顶级」「全网最低」就能多卖,删掉就会掉转化。本书证伪这一信念背后的两个假定:极限词的转化增益确定而可观;换成平实说法一定吃亏。在示意参数下,一个极限词要抵消每月 0.010 的出事概率乘以出事成本,需要 0.0333 的相对转化增益;已发表的研究提示吹嘘的作用在不同顾客那里正负相抵,混合模型中增益超过盈亏平衡值的情形只占 0.042。法律层面本书只写确信的层级:《广告法》禁止使用「国家级」「最高级」「最佳」等用语,2023 年的执法指南把近义用语纳入同一口径,并说明限定条件下如实陈述销量等事实信息不在此列;不引用条款序号,不写处罚数额,以上架前人工复核为准。书中的词表、文案、事实与参数均为示意;全部计算由书中程序生成;本书是非个人化法律建议,也不提供任何绕开平台审核的手法。 第一章算风险账。第二章搭建 Aho-Corasick 多模式匹配自动机:示意词表编译成 35 个状态,词表从 10 个词增加到 800 个词时,逐词查找每个符号的比较次数涨到 813.46,自动机始终不超过 1.998 步。第三章量化误杀与漏检:审计补词后,新一周的漏检率从 0.310 降到 0.076,误杀率 0.074;要让漏检率的 95 % 上界低于 0.01,零漏检也要审计 299 行违规文案。第四章把改写写成约束优化:事实要有出处,整行复扫无新命中,标题放得下,用多选背包求解。第五章说明表面相似度奖励违规,按命题槽位打分后与编辑距离相似度的等级相关为负 0.949。 第六章计算转化弹性:示意逻辑模型中事实改写最差只比原文低 0.0286;单次 A/B 测试的标准误 0.1071 远大于真实效应的离散,观测与真实效应的相关只有 0.226,应按改写类别合并检验。第七章考辨《论语·子路》「名不正,则言不顺」、《荀子·正名》与奥威尔的《政治与英语》。第八章给出合规改写库接攻心文案检查器的每周流程、蒸馏小工具「合规改写自动机」与预注册非劣效检验方案:相对界值 0.10、效能 0.80 时每组需要 39,981 次访问。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Risk Ledger of an Absolute Term: How Much Conversion Must One "Best" Buy? — The account of one absolute term reduces to a formula: the break-even uplift equals the incident probability times the incident cost, divided by the monthly gross margin. With the illustrative parameters of program p11, the middle cell requires a relative lift of 0.0333, and an assumed lift of 0.010 is enough in only 4 of the 15 cells. Published research suggests that the gain from puffery cancels out across buyers; in the mixture model of program p12, the lift exceeds the break-even value in only 0.042 of the draws. On the legal side, the Advertising Law's prohibition of terms such as "national-level", "highest-grade" and "best", together with the guidelines' treatment of factual statements under stated limits, points the way to compliant rewriting. What remains is an engineering question. A shop has hundreds of product pages and thousands of lines of copy: how can every absolute term be found in a single pass? The next chapter builds the automaton that does it. (https://5dlibrary.com/answers/en/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c01/) 第1章 极限词的风险账:一个「最佳」要换来多少转化才划算 — 一个极限词的账可以写成一个公式:盈亏平衡增益等于出事概率乘以出事成本、再除以月度毛利。程序 p11 的示意参数下,中间一格要求 0.0333 的相对增益,而假定的 0.010 只在 15 格中的 4 格里够用。已发表的研究提示,吹嘘的增益在不同顾客那里正负相抵,程序 p12 的混合模型里,增益超过盈亏平衡值的情形只占 0.042。法律层面,《广告法》对「国家级」「最高级」「最佳」等用语的禁止,以及执法指南对限定条件下事实陈述的说明,给合规改写指出了方向。剩下的问题是工程问题:一家店有几百个商品页、几千行文案,怎样一遍把所有极限词都找出来?下一章搭建这台自动机。 (https://5dlibrary.com/answers/zh/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c01/) - Ch.2 One Pass for Every Term: the Aho-Corasick Multi-Pattern Automaton — Chapter 2 builds an Aho-Corasick multi-pattern automaton: the illustrative lexicon compiles into 35 states, and as the lexicon grows from 10 to 800 terms the term-by-term search rises to 813.46 comparisons per symbol while the automaton never exceeds 1.998 steps. (https://5dlibrary.com/answers/en/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c02/) 第2章 一遍扫完:Aho-Corasick 多模式匹配自动机 — 第二章搭建 Aho-Corasick 多模式匹配自动机:示意词表编译成 35 个状态,词表从 10 个词增加到 800 个词时,逐词查找每个符号的比较次数涨到 813.46,自动机始终不超过 1.998 步。 (https://5dlibrary.com/answers/zh/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c02/) - Ch.3 False Kills and Misses: Exempt Contexts, Audit Additions and Audit Sample Sizes — Chapter 3 measures false kills and misses: after audit additions the miss rate in a new week falls from 0.310 to 0.076 with a false-kill rate of 0.074, and bringing the 95 % upper bound on the miss rate below 0.01 needs 299 violating lines even with zero misses. (https://5dlibrary.com/answers/en/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c03/) 第3章 误杀与漏检:豁免语境、审计补词与审计样本量 — 第三章量化误杀与漏检:审计补词后,新一周的漏检率从 0.310 降到 0.076,误杀率 0.074;要让漏检率的 95 % 上界低于 0.01,零漏检也要审计 299 行违规文案。 (https://5dlibrary.com/answers/zh/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c03/) - Ch.4 Choosing Rewrites under Constraints: Evidence, Re-scanning and Title Length — Rewriting is not deleting. It is choosing candidates under three constraints: facts must have a source, a whole-line re-scan must find no new hit, and the title must fit. In program p41 a title loses "first-class grade under the national standard", which has no basis on file, and "best in the trade", which fails the re-scan, and ends with a rewrite that has no hits, growing from 16 characters to 21. Program p42 writes the length constraint as a multiple-choice knapsack, and the optimum jumps with the limit: at 14 it keeps the material and drops the price, at 12 the reverse; tied solutions should go to a person. Throughout, the fidelity score has been taken as a given input. The next chapter asks where the score comes from, and whether the usual measures of text similarity can serve as the score directly. (https://5dlibrary.com/answers/en/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c04/) 第4章 候选改写的约束优化:事实凭据、复扫与标题长度 — 改写不是删词,而是在三条约束下选候选:事实要有出处,整行复扫不能有新命中,标题要放得下。程序 p41 在一行标题上依次排除了没有凭据的「国标一等品」与复扫命中的「业内最佳」,得到零命中的改写,长度从 16 个字变成 21 个字。程序 p42 把长度约束写成多选背包,最优选择随上限跳动:上限 14 时留材质、删价格,上限 12 时反过来;打平的解应交给人工。这里的保真分数一直被当作已知输入。下一章要问:分数从哪里来,常用的文本相似度能不能直接拿来当分数? (https://5dlibrary.com/answers/zh/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c04/) - Ch.5 Keeping the Meaning: Why Surface Similarity Rewards the Violation — Chapter 5 shows that surface similarity rewards the violation; a score built on proposition slots has a rank correlation of minus 0.949 with edit-distance similarity. (https://5dlibrary.com/answers/en/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c05/) 第5章 语义保真:表面相似度为何奖励违规 — 第五章说明表面相似度奖励违规,按命题槽位打分后与编辑距离相似度的等级相关为负 0.949。 (https://5dlibrary.com/answers/zh/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c05/) - Ch.6 Conversion Elasticity: How Much Does a Rewrite Move Conversion, and Can a Test Tell? — Chapter 6 computes conversion elasticity: in an illustrative logit model a factual rewrite is at worst 0.0286 below the original; the standard error of 0.1071 of a single A/B test dwarfs the spread of true effects, observed and true effects correlate at only 0.226, and rewrites should be tested pooled by class. (https://5dlibrary.com/answers/en/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c06/) 第6章 转化弹性:改写让转化变了多少,能不能测出来 — 第六章计算转化弹性:示意逻辑模型中事实改写最差只比原文低 0.0286;单次 A/B 测试的标准误 0.1071 远大于真实效应的离散,观测与真实效应的相关只有 0.226,应按改写类别合并检验。 (https://5dlibrary.com/answers/zh/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c06/) - Ch.7 Eastern and Western Sources: "If Names Are Not Correct" and Orwell's Rules for Writing — Confucius's "if names are not correct, speech does not flow" rightly makes the name the precondition of speech. Program p71 turns it into a name-reality gap: three of eight illustrative lines have one, and an absolute name cannot be closed even by third-party evidence, only rewritten. Where Confucius goes too far is in presenting a probabilistic chain of six links as necessary, without any sense of degree. Xunzi's "settled by agreement and custom" explains why the lexicon must be kept by meaning, and his "direct, easy and not contrary" reminds us that rewrites must still read well. Orwell calls empty words meaningless, and in program p72 the rewrites replace 7 puffery words with 5 facts; but his "cut what can be cut" and "short rather than long" do not hold for compliant rewriting, because facts take more room than empty words. The classics give a direction. The next chapter turns it into a procedure that can be run every week and a test plan that can be refuted. (https://5dlibrary.com/answers/en/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c07/) 第7章 东西方原典考辨:名不正则言不顺与奥威尔的写作规则 — 孔子的「名不正,则言不顺」说对了名是言的前提;程序 p71 把它写成名实差,八行示意文案里有三行存在差,绝对化的名即使有第三方凭据也补不平,只能改写。孔子说过头的,是把一条六环的概率链说成必然,并且没有给出程度。荀子的「约定俗成」解释了为什么词表要按意思维护,「径易而不拂」提醒改写还要读得顺。奥威尔把空话称作没有意义的词,程序 p72 中改写把 7 个吹嘘词换成了 5 个事实;但他的「能删就删」「能短就短」在合规改写中并不成立,因为事实比空话占地方。原典给了方向,下一章把方向落成一套每周都能执行的流程与一份可以被推翻的检验方案。 (https://5dlibrary.com/answers/zh/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c07/) - Ch.8 Closing the Loop: a Compliant-Rewrite Library for the Persuasive-Copy Checker, and a Non-Inferiority Test — Chapter 8 gives the weekly procedure linking a compliant-rewrite library to the persuasive-copy checker, the distilled Compliant Rewriting Automaton, and a pre-registered non-inferiority test: at a relative margin of 0.10 and power of 0.80 each arm needs 39,981 sessions. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c08/) 第8章 实证闭环:合规改写库接攻心文案检查器与非劣效检验 — 第八章给出合规改写库接攻心文案检查器的每周流程、蒸馏小工具「合规改写自动机」与预注册非劣效检验方案:相对界值 0.10、效能 0.80 时每组需要 39,981 次访问。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/platform-sensitive-term-compliant-rewriting-absolute-term-automata/c08/) ## 5DL-B-275 · Purchasing-Power Parity across Provinces and Pricing for Lower-Tier Cities: Income Distributions, Engel-Coefficient Calibration and Multi-Tier Package Anchors 中文书名:跨省购买力平价与下沉城市定价、收入分布恩格尔系数校准与多层级城市套餐锚点母典 DOI: pending Cite: Fan, Y. (2026). Purchasing-Power Parity across Provinces and Pricing for Lower-Tier Cities: Income Distributions, Engel-Coefficient Calibration and Multi-Tier Package Anchors (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its city tiers, incomes, price levels, costs and demand parameters are illustrative, and it quotes no specific official statistic; all calculations are generated by its programs; it is not individual business advice. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: When a chain takes a package from first-tier cities into lower-tier ones, the most common pricing method multiplies the first-tier price by the ratio of local to first-tier per-capita income. This book refutes the belief behind that rule: that the price a city can bear is set by its average income. Its thesis is that the price a city can bear depends on quantiles of the income distribution and on the Engel coefficient, and that discounting by the ratio of mean incomes is exactly right only when distributions have the same shape, essentials take the same share and costs scale in proportion. Its city tiers, incomes, price levels, costs and demand parameters are illustrative, and it quotes no specific official statistic; all calculations are generated by its programs; it is not individual business advice. Every price proposed by its small tool must be released by a person and first run through a pricing calculator, prices are posted openly by city, and anchors are only real current prices. Chapter 1 quantifies the thesis on four illustrative tiers: the fourth tier's mean income is 0.402 of the first tier's, but its discretionary income at the 95th percentile is 0.532; its optimal price of 508 yuan is 0.592 of the first tier's, and the proportional discount to 345 yuan gives up 0.077 of profit, erring on the low side. Chapter 2 deals with lognormal incomes and shows that a Pareto tail makes a two-parameter fit overstate the high-income share, 0.0463 against a true 0.0351. Chapter 3 fits an Engel curve to Engel's own table, reads price levels from shifts of the curve (0.768 with equal tastes, jumping to 0.960 with a small taste gap), and shows that Engel's law makes discretionary income more unequal than income, raising the fourth tier's Gini coefficient from 0.389 to 0.468. Chapter 4 shows that a city has no single price level: an online package's cost floor is 150 yuan in every tier, while an in-store service's falls from 667 to 367 yuan. Chapter 5 gives demand curves by tier, the Lerner condition and leakage between tiers; a single national price earns only 0.970 of tiered profit. Chapter 6 prices a three-rung menu by self-selection: fourth-tier prices relative to the first tier are 0.592 for basic, 0.666 for standard and 0.719 for premium, and five rules audit the anchors, catching invented original prices and decoys. Chapter 7 examines Li Kui's grain-purchase policy (food is 0.600 of his farmer's consumption, and "too dear hurts the people, too cheap hurts the farmer" becomes a price corridor that closes in the fourth tier for a package with a visiting expert, floor 2767 yuan against ceiling 2580) and Engel's Belgian family budgets. Chapter 8 gives the quarterly pricing procedure, the City-Tier Package Anchor Calculator and a pre-registered test: the primary test needs 13 pairs of cities. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:连锁品牌把一线城市的套餐卖进下沉城市,最常见的定价办法是一线价格乘以当地人均收入与一线人均收入之比。本书证伪这条规则背后的信念:一座城市能承受多高的价格,由它的平均收入决定。本书的命题是:可承受价取决于收入分布的分位与恩格尔系数,而按平均收入之比打折只在分布形状相同、必需开支份额相同、成本同比缩放时才恰好正确。书中的城市档位、收入、价格水平、成本与需求参数均为示意,不引用任何具体的官方统计数值;全部计算由书中程序生成;本书是非个人化经营建议,书中小工具给出的每一个价格都须由经营者人工放行并先过定价试算器,价格按城市公开标示,锚点只用真实现行价格。 第一章在四档示意城市上量化命题:第四档城市的平均收入是一线的 0.402,余钱在第 95 百分位却是 0.532;它的最优价 508 元是一线的 0.592,按比例打折到 345 元会少赚 0.077,错在定低。第二章讲对数正态收入,并说明顶端的帕累托尾部会使两参数拟合把高收入人口比例从 0.0351 高估到 0.0463。第三章用恩格尔原表拟合恩格尔曲线,用曲线平移反推价格水平(口味相同时估计 0.768,口味差一点就跳到 0.960),并显示恩格尔定律使余钱比收入更不平均,第四档的基尼系数从 0.389 升到 0.468。第四章说明一座城市没有唯一的价格水平:线上套餐的成本底线在各档都是 150 元,到店服务从 667 元降到 367 元。 第五章给出各档需求曲线、勒纳条件与跨层渗漏,全国统一价的利润只有分档定价的 0.970。第六章按自选原理给三档菜单定价:第四档相对一线的价格比,基础档 0.592、标准档 0.666、尊享档 0.719,并用五条规则审计锚点,拦下虚构原价与诱饵。第七章考辨李悝平籴(农户口粮占消费的 0.600,「甚贵伤民,甚贱伤农」写成价格走廊,外请专家的套餐在第四档下限 2767 元高过上限 2580 元)与恩格尔的比利时家庭预算。第八章给出季度定价流程、城市分层套餐锚点计算器与预注册检验方案:主检验需要 13 对城市。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Proportional-Discount Fallacy: Why the Ratio of Mean Incomes Misprices Lower-Tier Cities — Chapter 1 quantifies the thesis on four illustrative tiers: the fourth tier's mean income is 0.402 of the first tier's, but its discretionary income at the 95th percentile is 0.532; its optimal price of 508 yuan is 0.592 of the first tier's, and the proportional discount to 345 yuan gives up 0.077 of profit, erring on the low side. (https://5dlibrary.com/answers/en/cross-province-ppp-lower-tier-city-package-anchors/c01/) 第1章 按比例打折的谬误:均值之比为何定错下沉城市的价格 — 第一章在四档示意城市上量化命题:第四档城市的平均收入是一线的 0.402,余钱在第 95 百分位却是 0.532;它的最优价 508 元是一线的 0.592,按比例打折到 345 元会少赚 0.077,错在定低。 (https://5dlibrary.com/answers/zh/cross-province-ppp-lower-tier-city-package-anchors/c01/) - Ch.2 Lognormal Incomes: Recovering a Whole Distribution from a Mean and a Median — Chapter 2 deals with lognormal incomes and shows that a Pareto tail makes a two-parameter fit overstate the high-income share, 0.0463 against a true 0.0351. (https://5dlibrary.com/answers/en/cross-province-ppp-lower-tier-city-package-anchors/c02/) 第2章 对数正态收入:从均值与中位数还原整条分布 — 第二章讲对数正态收入,并说明顶端的帕累托尾部会使两参数拟合把高收入人口比例从 0.0351 高估到 0.0463。 (https://5dlibrary.com/answers/zh/cross-province-ppp-lower-tier-city-package-anchors/c02/) - Ch.3 Engel Curves: Food Shares, Discretionary Income and Reading Price Levels Backwards — Chapter 3 fits an Engel curve to Engel's own table, reads price levels from shifts of the curve (0.768 with equal tastes, jumping to 0.960 with a small taste gap), and shows that Engel's law makes discretionary income more unequal than income, raising the fourth tier's Gini coefficient from 0.389 to 0.468. (https://5dlibrary.com/answers/en/cross-province-ppp-lower-tier-city-package-anchors/c03/) 第3章 恩格尔曲线:食品份额、可支配余钱与价格水平的反推 — 第三章用恩格尔原表拟合恩格尔曲线,用曲线平移反推价格水平(口味相同时估计 0.768,口味差一点就跳到 0.960),并显示恩格尔定律使余钱比收入更不平均,第四档的基尼系数从 0.389 升到 0.468。 (https://5dlibrary.com/answers/zh/cross-province-ppp-lower-tier-city-package-anchors/c03/) - Ch.4 Purchasing-Power Parity across Provinces: Baskets, the Balassa and Samuelson Effect, and Cost Floors — Chapter 4 shows that a city has no single price level: an online package's cost floor is 150 yuan in every tier, while an in-store service's falls from 667 to 367 yuan. (https://5dlibrary.com/answers/en/cross-province-ppp-lower-tier-city-package-anchors/c04/) 第4章 省际购买力平价:一篮子价格、巴拉萨与萨缪尔森效应和成本底线 — 第四章说明一座城市没有唯一的价格水平:线上套餐的成本底线在各档都是 150 元,到店服务从 667 元降到 367 元。 (https://5dlibrary.com/answers/zh/cross-province-ppp-lower-tier-city-package-anchors/c04/) - Ch.5 Demand Curves by Tier: Revenue-Maximising Prices, the Lerner Condition and Leakage between Tiers — Chapter 5 gives demand curves by tier, the Lerner condition and leakage between tiers; a single national price earns only 0.970 of tiered profit. (https://5dlibrary.com/answers/en/cross-province-ppp-lower-tier-city-package-anchors/c05/) 第5章 分层需求曲线:收益最大价、勒纳条件与跨层渗漏 — 第五章给出各档需求曲线、勒纳条件与跨层渗漏,全国统一价的利润只有分档定价的 0.970。 (https://5dlibrary.com/answers/zh/cross-province-ppp-lower-tier-city-package-anchors/c05/) - Ch.6 Multi-Tier Package Anchors: Self-Selection Menus, Price Endings and a Real-Anchor Audit — Chapter 6 prices a three-rung menu by self-selection: fourth-tier prices relative to the first tier are 0.592 for basic, 0.666 for standard and 0.719 for premium, and five rules audit the anchors, catching invented original prices and decoys. (https://5dlibrary.com/answers/en/cross-province-ppp-lower-tier-city-package-anchors/c06/) 第6章 多层级套餐锚点:自选菜单、价格尾数与真实锚点审计 — 第六章按自选原理给三档菜单定价:第四档相对一线的价格比,基础档 0.592、标准档 0.666、尊享档 0.719,并用五条规则审计锚点,拦下虚构原价与诱饵。 (https://5dlibrary.com/answers/zh/cross-province-ppp-lower-tier-city-package-anchors/c06/) - Ch.7 Eastern and Western Sources Examined: Li Kui's Grain Purchases and Engel's Belgian Family Budgets — Chapter 7 examines Li Kui's grain-purchase policy (food is 0.600 of his farmer's consumption, and "too dear hurts the people, too cheap hurts the farmer" becomes a price corridor that closes in the fourth tier for a package with a visiting expert, floor 2767 yuan against ceiling 2580) and Engel's Belgian family budgets. (https://5dlibrary.com/answers/en/cross-province-ppp-lower-tier-city-package-anchors/c07/) 第7章 东西方原典考辨:李悝平籴与恩格尔的比利时家庭预算 — 第七章考辨李悝平籴(农户口粮占消费的 0.600,「甚贵伤民,甚贱伤农」写成价格走廊,外请专家的套餐在第四档下限 2767 元高过上限 2580 元)与恩格尔的比利时家庭预算。 (https://5dlibrary.com/answers/zh/cross-province-ppp-lower-tier-city-package-anchors/c07/) - Ch.8 Closing the Loop: a City-Tier Package Price Table, Human Release and a Pre-Registered Test — Chapter 8 gives the quarterly pricing procedure, the City-Tier Package Anchor Calculator and a pre-registered test: the primary test needs 13 pairs of cities. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cross-province-ppp-lower-tier-city-package-anchors/c08/) 第8章 实证闭环:城市分层套餐价格表、人工放行与预注册检验 — 第八章给出季度定价流程、城市分层套餐锚点计算器与预注册检验方案:主检验需要 13 对城市。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cross-province-ppp-lower-tier-city-package-anchors/c08/) ## 5DL-B-276 · Head Concentration versus Mid-Tier Portfolios in Creator Matrices: Tail Variance, Sharpe Ratios and the Allocation of a Budget across a Hundred Niche Creators 中文书名:达人矩阵头部集中度与中腰部组合套利、尾部方差夏普比率与百名垂直达人投资组合母典 DOI: pending Cite: Fan, Y. (2026). Head Concentration versus Mid-Tier Portfolios in Creator Matrices: Tail Variance, Sharpe Ratios and the Allocation of a Budget across a Hundred Niche Creators (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its creators, returns, correlations, costs and budgets are illustrative; all calculations are generated by its programs; every collaboration must be honestly labelled as advertising, and bought traffic or fake reviews are not discussed. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Should a brand's creator budget go to one head creator or be spread over a crowd of mid-tier creators? Two contradictory views circulate. This book writes the question as a proposition that can be computed and overturned by data: on the same budget, a mid-tier portfolio spread across verticals swings far less than a single head creator, but correlation, the fixed cost of each collaboration and heavy-tailed returns apply three discounts to that conclusion. The book borrows the variance, correlation, Sharpe ratio and mean-variance frontier of portfolio theory for their arithmetic only; it is about allocating a marketing budget, not about investing in securities, it is not individual financial or investment advice, and it recommends no financial product. Its creators, returns, correlations, costs and budgets are illustrative; all calculations are generated by its programs; every collaboration must be honestly labelled as advertising, and bought traffic or fake reviews are not discussed. Chapter 1 writes down the variance of one budget: with twenty mid-tier creators correlated at 0.20, portfolio variance is only 0.1067 of the head creator's, but the floor is the average covariance. Chapter 2 derives correlation from shared audiences: twenty creators count as only 4.17 independent ones, and the discount on diversification equals the correlation. Chapter 3 moves to skewed returns: a head-creator plan with a positive mean loses money in 0.523 of quarters, and the twenty-creator portfolio loses more than half its budget in only 0.0037 of quarters against 0.2047 for the head creator. Chapter 4 shows that Sharpe ratios cannot be measured precisely, the head creator's estimate over eight quarters averaging only 0.0783. Chapter 5 shows that optimisation on short histories loses to equal allocation: a Sharpe ratio of 0.3638 for equal allocation against 0.3033 when optimising on twelve quarters. Chapter 6 comes down to a hundred niche creators: at a fixed cost of 1000 yuan per collaboration, a matrix of a hundred creators has a Sharpe ratio of -0.2320, and with half the budget on a head creator, the head creator carries 0.80 of the risk. Chapter 7 examines Feng Xuan's three burrows in the Strategies of the Warring States (two tied to Xue, so all fail with probability 0.1000 rather than 0.0640) and Antonio's ships in The Merchant of Venice (with high correlation, losing all four is 656 times as likely as under independence). Chapter 8 gives the quarterly rebalancing procedure, the Creator Portfolio Variance Calculator and a pre-registered test: at a variance ratio of 0.50 each arm needs 53 campaign waves, or 435 once the head creator's kurtosis is counted. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:品牌的达人预算该集中给顶流还是撒给一批中腰部达人,流行着两种互相矛盾的说法。本书把它写成可以计算、可以被数据推翻的命题:同额预算下,跨垂类的中腰部组合,回报波动远小于单一顶流;但相关、单次合作的固定成本与重尾回报会给这一结论打三个折扣。书中借用组合理论的方差、相关、夏普比率与均值—方差前沿,只借其算术,讨论的是营销预算的分配,不是证券投资;本书不构成任何个人化的理财或投资建议,不推荐任何金融产品。书中的达人、回报、相关、成本与预算均为示意;全部计算由书中程序生成;所有合作都须如实标注广告,不讨论购买流量或虚假评价。 第一章写出同额预算的方差公式:二十名中腰部、相关 0.20 时,组合方差只有顶流的 0.1067,但地板是平均协方差。第二章由受众重叠算出相关,二十人只相当于 4.17 个独立的人,分散折扣恰好等于相关。第三章改用偏态回报:均值为正的顶流方案有 0.523 的季度净亏,二十人组合亏掉一半以上预算的季度只有 0.0037,顶流是 0.2047。第四章说明夏普比率测不准,顶流在八个季度里的估计平均只有 0.0783。第五章说明短历史下的优化敌不过等额分配:等额的夏普比率 0.3638,用十二个季度优化只得 0.3033。 第六章落到百名垂直达人:单次合作固定成本 1000 元时,一百人矩阵的夏普比率是 -0.2320;一半预算给顶流时,顶流承担 0.80 的风险。第七章考辨《战国策》冯谖三窟(两窟系于薛,全部失效概率 0.1000 而非 0.0640)与《威尼斯商人》安东尼奥的商船(相关高时四船全损的概率是独立时的 656 倍)。第八章给出季度再平衡流程、达人组合方差计算器与预注册检验:方差比 0.50 时每组需要 53 个波次,计入顶流峰度后要 435 个。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 One Budget, Two Plans: the Variance of a Head Creator and of a Mid-Tier Portfolio — Chapter 1 writes down the variance of one budget: with twenty mid-tier creators correlated at 0.20, portfolio variance is only 0.1067 of the head creator's, but the floor is the average covariance. (https://5dlibrary.com/answers/en/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c01/) 第1章 同一笔预算:顶流的方差与中腰部组合的方差公式 — 第一章写出同额预算的方差公式:二十名中腰部、相关 0.20 时,组合方差只有顶流的 0.1067,但地板是平均协方差。 (https://5dlibrary.com/answers/zh/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c01/) - Ch.2 Shared Audiences Are Correlation: the Discount on Diversification and the Effective Number of Creators — Chapter 2 derives correlation from shared audiences: twenty creators count as only 4.17 independent ones, and the discount on diversification equals the correlation. (https://5dlibrary.com/answers/en/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c02/) 第2章 受众重叠就是相关:分散收益的折扣与有效人数 — 第二章由受众重叠算出相关,二十人只相当于 4.17 个独立的人,分散折扣恰好等于相关。 (https://5dlibrary.com/answers/zh/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c02/) - Ch.3 Tail Variance: Quantiles of Skewed Returns, Expected Shortfall and Power Laws — Chapter 3 moves to skewed returns: a head-creator plan with a positive mean loses money in 0.523 of quarters, and the twenty-creator portfolio loses more than half its budget in only 0.0037 of quarters against 0.2047 for the head creator. (https://5dlibrary.com/answers/en/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c03/) 第3章 尾部方差:偏态回报的分位数、预期亏空与幂律 — 第三章改用偏态回报:均值为正的顶流方案有 0.523 的季度净亏,二十人组合亏掉一半以上预算的季度只有 0.0037,顶流是 0.2047。 (https://5dlibrary.com/answers/zh/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c03/) - Ch.4 The Sharpe Ratio: Hurdle Returns, Estimation Error and the Trap of Skew — Chapter 4 shows that Sharpe ratios cannot be measured precisely, the head creator's estimate over eight quarters averaging only 0.0783. (https://5dlibrary.com/answers/en/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c04/) 第4章 夏普比率:门槛回报、估计误差与偏态的陷阱 — 第四章说明夏普比率测不准,顶流在八个季度里的估计平均只有 0.0783。 (https://5dlibrary.com/answers/zh/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c04/) - Ch.5 The Mean-Variance Frontier: Three Pools, Estimation Error and Shrinkage — Chapter 5 shows that optimisation on short histories loses to equal allocation: a Sharpe ratio of 0.3638 for equal allocation against 0.3033 when optimising on twelve quarters. (https://5dlibrary.com/answers/en/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c05/) 第5章 均值—方差前沿:三池配比、估计误差与收缩 — 第五章说明短历史下的优化敌不过等额分配:等额的夏普比率 0.3638,用十二个季度优化只得 0.3033。 (https://5dlibrary.com/answers/zh/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c05/) - Ch.6 A Hundred Niche Creators: Block Correlation, the Fixed Cost of Each Collaboration and Concentration of Risk — Chapter 6 comes down to a hundred niche creators: at a fixed cost of 1000 yuan per collaboration, a matrix of a hundred creators has a Sharpe ratio of -0.2320, and with half the budget on a head creator, the head creator carries 0.80 of the risk. (https://5dlibrary.com/answers/en/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c06/) 第6章 百名垂直达人:块状相关、单次合作的固定成本与风险集中度 — 第六章落到百名垂直达人:单次合作固定成本 1000 元时,一百人矩阵的夏普比率是 -0.2320;一半预算给顶流时,顶流承担 0.80 的风险。 (https://5dlibrary.com/answers/zh/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c06/) - Ch.7 East and West in the Sources: Feng Xuan's Three Burrows and Antonio's Four Argosies — Chapter 7 examines Feng Xuan's three burrows in the Strategies of the Warring States (two tied to Xue, so all fail with probability 0.1000 rather than 0.0640) and Antonio's ships in The Merchant of Venice (with high correlation, losing all four is 656 times as likely as under independence). (https://5dlibrary.com/answers/en/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c07/) 第7章 东西方原典考辨:冯谖三窟与安东尼奥的四条商船 — 第七章考辨《战国策》冯谖三窟(两窟系于薛,全部失效概率 0.1000 而非 0.0640)与《威尼斯商人》安东尼奥的商船(相关高时四船全损的概率是独立时的 656 倍)。 (https://5dlibrary.com/answers/zh/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c07/) - Ch.8 The Quarterly Rebalancing Loop and a Plan to Falsify the Book — Chapter 8 gives the quarterly rebalancing procedure, the Creator Portfolio Variance Calculator and a pre-registered test: at a variance ratio of 0.50 each arm needs 53 campaign waves, or 435 once the head creator's kurtosis is counted. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c08/) 第8章 季度再平衡闭环与证伪方案 — 第八章给出季度再平衡流程、达人组合方差计算器与预注册检验:方差比 0.50 时每组需要 53 个波次,计入顶流峰度后要 435 个。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/creator-matrix-mid-tier-portfolio-tail-variance-sharpe/c08/) ## 5DL-B-277 · Kaplan-Meier Retention Curves for Livestream Rooms: Locating Swipe-Away Hazard Spikes and Timing the Insertion of Hooks 中文书名:直播间留存卡普兰迈耶生存曲线、划走风险尖峰定位与福利钩子插入秒数母典 DOI: pending Cite: Fan, Y. (2026). Kaplan-Meier Retention Curves for Livestream Rooms: Locating Swipe-Away Hazard Spikes and Timing the Insertion of Hooks (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its rooms, audiences, hazard parameters, hook effects and run sheets are illustrative; all calculations are generated by its programs; the book is not individual business advice. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: The number most often used to review a livestream is average viewing time. This book refutes two beliefs behind it: that the average summarises how well a show keeps its audience, and that benefits or retention segments push the average up as long as they are spread evenly. Using the Kaplan-Meier product-limit estimator, the Nelson-Aalen cumulative hazard, piecewise-constant hazards and Bellman's dynamic programming, it shows that the risk of leaving a livestream room has spikes, one kind following each viewer's own moment of entry and another following segment switches on the programme clock, and that the average folds both into one number. Its rooms, audiences, hazard parameters, hook effects and run sheets are illustrative; all calculations are generated by its programs; the book is not individual business advice. A hook may only be real content or a benefit with real stock and a real closing time, with no resetting countdown and nothing that leads minors to buy, and platform rules as officially published prevail. Chapter 1 shows that censoring biases the average downwards: counting the end of the stream as leaving gives 408.1 seconds, counting only leavers gives 278.0, and the restricted mean from the product-limit estimator is 337.9. Chapter 2 locates the entry spike with the Nelson-Aalen estimator, the hazard in the first two seconds being 133.3 times that of settled viewers, and shows how casually marked spikes produce false reports. Chapter 3 separates two clocks: the switch hazard ratio of settled viewers on the programme clock averages 8.52, squashed to 4.23 when newcomers are mixed in. Chapter 4 compresses the switch spikes into two numbers with a piecewise-constant hazard, a hazard ratio of 9.55. Chapter 5 models what a hook does: a window covering the spike gains 693.4 audience-minutes, only 40.4 when kept viewers take and go, and those a hook keeps stay only 0.68 as long as those who would have stayed anyway. Chapter 6 schedules three hooks by dynamic programming for a gain of 1970.2, against 257.9 on average for one hook every 20 minutes; if the run sheet slips a minute and the hooks stay put, the gain falls to 210.7; scheduling is only 1.29 times as good with no spikes and 9.53 times with very tall ones. Chapter 7 examines the Sunzi's "keen in the morning, slack at midday, homeward in the evening" and Graunt's mortality table of 1662 (hazard in the first six years 1.46 times the mid-life level). Chapter 8 gives the post-show review, the Swipe-Spike Hook Scheduler and a pre-registered test plan with an expected hazard ratio of 0.718. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:直播间复盘最常用的数是平均停留时长。本书证伪它背后的两个信念:平均停留能概括一场直播留人的好坏;福利或留人环节只要均匀撒开就能把平均数往上推。书中用卡普兰与迈耶的乘积极限估计、纳尔逊-阿伦累积风险、分段常数风险与贝尔曼动态规划说明:离开直播间的风险有尖峰,一类跟着每位观众自己的入场时刻,一类跟着节目时钟上的环节切换;平均数把两类尖峰揉成一个数。书中的直播间、观众数、风险参数、钩子效果与节目单均为示意;全部计算由书中程序生成;本书是非个人化经营建议,钩子只能是真实的内容或写明真实库存与真实截止时刻的福利,不用会重置的倒计时,不诱导未成年人消费,平台规则以官方公示为准。 第一章说明截尾让平均停留偏低:示意直播间里把下播当离开得 408.1 秒,只算已离开者得 278.0 秒,乘积极限估计的限定平均停留是 337.9 秒。第二章用纳尔逊-阿伦估计定位入场尖峰,入场头两秒的风险是坐稳之后的 133.3 倍,并说明随手圈尖峰会大量错报。第三章区分两只钟:坐稳观众在节目时钟上的切换风险比平均 8.52 倍,混入新观众后被压到 4.23 倍。第四章用分段常数风险把切换尖峰压成两个数,风险比 9.55。 第五章建立钩子的作用模型:窗口盖住尖峰时收益 693.4 观众分钟,领完就走时只剩 40.4;钩子留下的人此后的停留只有原本会留下者的 0.68。第六章用动态规划排三个钩子,收益 1970.2,均匀每 20 分钟一个平均只有 257.9;节目单推迟一分钟而钩子不动,收益跌到 210.7;没有尖峰时排程只好 1.29 倍,尖峰极高时好 9.53 倍。第七章考辨《孙子·军争》「朝气锐,昼气惰,暮气归」与格朗特 1662 年的死亡表(开头六年风险是中段的 1.46 倍)。第八章给出每场复盘流程、划走尖峰钩子排程器与预注册检验方案,预期风险比 0.718。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Blind Spot of Average Viewing Time: Censoring and the Product Limit — Chapter 1 shows that censoring biases the average downwards: counting the end of the stream as leaving gives 408.1 seconds, counting only leavers gives 278.0, and the restricted mean from the product-limit estimator is 337.9. (https://5dlibrary.com/answers/en/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c01/) 第1章 平均停留时长的盲区:截尾与乘积极限 — 第一章说明截尾让平均停留偏低:示意直播间里把下播当离开得 408.1 秒,只算已离开者得 278.0 秒,乘积极限估计的限定平均停留是 337.9 秒。 (https://5dlibrary.com/answers/zh/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c01/) - Ch.2 From Retention to Hazard: the Nelson-Aalen Estimator and the Entry Spike — Chapter 2 locates the entry spike with the Nelson-Aalen estimator, the hazard in the first two seconds being 133.3 times that of settled viewers, and shows how casually marked spikes produce false reports. (https://5dlibrary.com/answers/en/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c02/) 第2章 从留存到风险:纳尔逊-阿伦估计与入场尖峰定位 — 第二章用纳尔逊-阿伦估计定位入场尖峰,入场头两秒的风险是坐稳之后的 133.3 倍,并说明随手圈尖峰会大量错报。 (https://5dlibrary.com/answers/zh/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c02/) - Ch.3 Two Clocks: Seconds since Entry, the Programme Clock and Delayed Entry — Chapter 3 separates two clocks: the switch hazard ratio of settled viewers on the programme clock averages 8.52, squashed to 4.23 when newcomers are mixed in. (https://5dlibrary.com/answers/en/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c03/) 第3章 两只钟:入场秒数、节目时钟与延迟进入 — 第三章区分两只钟:坐稳观众在节目时钟上的切换风险比平均 8.52 倍,混入新观众后被压到 4.23 倍。 (https://5dlibrary.com/answers/zh/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c03/) - Ch.4 Piecewise-Constant Hazards: Maximum Likelihood, Likelihood Ratios and Finding the Switches — Chapter 4 compresses the switch spikes into two numbers with a piecewise-constant hazard, a hazard ratio of 9.55. (https://5dlibrary.com/answers/en/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c04/) 第4章 分段常数风险:最大似然、似然比与找切换点 — 第四章用分段常数风险把切换尖峰压成两个数,风险比 9.55。 (https://5dlibrary.com/answers/zh/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c04/) - Ch.5 A Model of What a Hook Does: the Window, the Aftermath and Who Stays — Chapter 5 models what a hook does: a window covering the spike gains 693.4 audience-minutes, only 40.4 when kept viewers take and go, and those a hook keeps stay only 0.68 as long as those who would have stayed anyway. (https://5dlibrary.com/answers/en/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c05/) 第5章 钩子的作用模型:窗口、余波与留下来的是谁 — 第五章建立钩子的作用模型:窗口盖住尖峰时收益 693.4 观众分钟,领完就走时只剩 40.4;钩子留下的人此后的停留只有原本会留下者的 0.68。 (https://5dlibrary.com/answers/zh/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c05/) - Ch.6 Dynamic Programming for Hook Insertion: the Best Seconds for K Hooks — Chapter 6 schedules three hooks by dynamic programming for a gain of 1970.2, against 257.9 on average for one hook every 20 minutes; if the run sheet slips a minute and the hooks stay put, the gain falls to 210.7; scheduling is only 1.29 times as good with no spikes and 9.53 times with very tall ones. (https://5dlibrary.com/answers/en/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c06/) 第6章 钩子插入的动态规划:K 个钩子的最优秒数 — 第六章用动态规划排三个钩子,收益 1970.2,均匀每 20 分钟一个平均只有 257.9;节目单推迟一分钟而钩子不动,收益跌到 210.7;没有尖峰时排程只好 1.29 倍,尖峰极高时好 9.53 倍。 (https://5dlibrary.com/answers/zh/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c06/) - Ch.7 Classical Sources East and West: Sunzi on Spirit and Graunt's Table of Mortality — Chapter 7 examines the Sunzi's "keen in the morning, slack at midday, homeward in the evening" and Graunt's mortality table of 1662 (hazard in the first six years 1.46 times the mid-life level). (https://5dlibrary.com/answers/en/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c07/) 第7章 东西方原典考辨:孙子论「气」与格朗特的死亡表 — 第七章考辨《孙子·军争》「朝气锐,昼气惰,暮气归」与格朗特 1662 年的死亡表(开头六年风险是中段的 1.46 倍)。 (https://5dlibrary.com/answers/zh/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c07/) - Ch.8 Closing the Loop: a Hazard Review after Every Show and a Plan to Falsify — Chapter 8 gives the post-show review, the Swipe-Spike Hook Scheduler and a pre-registered test plan with an expected hazard ratio of 0.718. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c08/) 第8章 实证闭环:每场直播的风险曲线复盘与证伪方案 — 第八章给出每场复盘流程、划走尖峰钩子排程器与预注册检验方案,预期风险比 0.718。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/livestream-kaplan-meier-swipe-hazard-spikes-hook-timing/c08/) ## 5DL-B-278 · Cohesion Topology of High-Ticket Peer Boards and First-Quarter Silent Attrition: Degree Centrality, Reciprocal Edge Weights and Ice-Breaking Pairing Algorithms 中文书名:高客单私董社群防退群内聚拓扑、度中心性互惠边权重与破冰配对算法母典 DOI: pending Cite: Fan, Y. (2026). Cohesion Topology of High-Ticket Peer Boards and First-Quarter Silent Attrition: Degree Centrality, Reciprocal Edge Weights and Ice-Breaking Pairing Algorithms (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 1 takes a census of an illustrative board, where 74 directed pairs with contact form only 15 reciprocal edges, and shows with two simulated worlds that risk ratios of 6.09 and 2.69 can both arise where reciprocal edges do nothing. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: The commonest loss in a high-ticket peer board is not a loud resignation but silence: the fee is paid, two meetings are attended, and the member is never seen again. The project blueprint claimed that first-quarter silent attrition comes mainly from members with no reciprocal edge in the board, and that pairing by shared attributes plus centrality builds reciprocal edges better than random ice-breaking. Having checked the sources, this book revises the claim: neither Granovetter's 1973 paper nor the 2006 paper by Backstrom and colleagues measured attrition in any community, and having no reciprocal edge is a measurable structural signal whose causal role only a randomised pairing experiment can settle. Its boards, members, attributes and probabilities are illustrative; all calculations are generated by its programs; the book is not individual business advice, uses only data members have explicitly agreed to share, builds no personal profiles, keeps exit free and honours refunds as agreed. Chapter 1 takes a census of an illustrative board, where 74 directed pairs with contact form only 15 reciprocal edges, and shows with two simulated worlds that risk ratios of 6.09 and 2.69 can both arise where reciprocal edges do nothing. Chapter 2 computes degree, betweenness and constraint; deleting the hub drops the share of member pairs still able to reach each other from 0.758 to 0.229. Chapter 3 compares four tie weights and tests reciprocity against an out-degree-preserving null, with a one-sided p-value of only 0.0747 on a single board. Chapter 4 solves exact maximum-weight matching by bitmask recursion and tests its edge over random pairing in three possible worlds: 0.766 more reciprocal edges per round when the score is right, none when attributes do not matter, and 0.552 fewer when complements attract. Chapter 5 finds stable pairs with Irving's roommates algorithm; under random rankings a board of sixteen has a stable pairing with probability 0.8485. Chapter 6 does the structural accounts of similar and bridging pairs and shows that "no reciprocal edge" used as a risk label is right only 0.240 of the time. Chapter 7 examines the joined lakes of the Dui hexagram and Aristotle on friendship. Chapter 8 gives the intake pairing sheet, the distilled Ice-Breaker Pairing Planner and a pre-registered test plan needing about 34 intake cohorts per arm with reciprocal edges as the primary outcome. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:高客单私董社群最常见的流失不是公开退出,而是沉默:交了费,来过两次,然后不再出现。立项蓝图的命题是,首季沉默流失主要来自在社群里没有一条互惠边的成员,而按共同属性加中心性配对比随机破冰更能建立互惠边。本书核对原文后修正了它:格兰诺维特 1973 年与巴克斯特罗姆等人 2006 年的论文都没有测量社群流失;零互惠边是一个可测的结构信号,它是不是原因,只能由随机化的配对实验回答。书中的社群、成员、属性与概率均为示意;全部计算由书中程序生成;本书是非个人化经营建议,只使用成员明示同意的数据,不做个人画像,退群自由,退款按约定。 第一章普查一个示意社群:74 个有往来的有向对只形成 15 条互惠边,并用两个模拟世界说明,风险比 6.09 与 2.69 都可能出自互惠边毫无作用的世界。第二章计算度、介数与约束系数,删除枢纽成员会让仍能相互到达的成员对从 0.758 跌到 0.229。第三章比较边的四种权重,并以保出度零模型检验互惠,单个社群的单侧 p 值只有 0.0747。第四章用位掩码递推求精确最大权匹配,在三个可能世界中检验它对随机配对的优势:评分正确时每轮多建 0.766 条互惠边,属性无关时为零,互补相吸时少建 0.552 条。 第五章用伊尔文的室友算法求稳定配对,随机意向下十六人社群有 0.8485 的机会存在稳定配对。第六章算同质与跨界配对的结构账,并说明把零互惠边当作风险标签时命中率只有 0.240。第七章考辨《周易·兑》的丽泽讲习与亚里士多德论友爱。第八章给出入群破冰配对表、蒸馏小工具破冰配对器与预注册检验方案:以互惠边为主要结局,每组约 34 个入群批次。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 No Reciprocal Edge: a Structural Signal of First-Quarter Silent Attrition and the Problem of Identifying It — Chapter 1 takes a census of an illustrative board, where 74 directed pairs with contact form only 15 reciprocal edges, and shows with two simulated worlds that risk ratios of 6.09 and 2.69 can both arise where reciprocal edges do nothing. (https://5dlibrary.com/answers/en/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c01/) 第1章 零互惠边:首季沉默流失的结构信号与识别难题 — 第一章普查一个示意社群:74 个有往来的有向对只形成 15 条互惠边,并用两个模拟世界说明,风险比 6.09 与 2.69 都可能出自互惠边毫无作用的世界。 (https://5dlibrary.com/answers/zh/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c01/) - Ch.2 Degree, Betweenness and Structural Holes: Whose Exit Would Cut the Board Apart — Chapter 2 computes degree, betweenness and constraint; deleting the hub drops the share of member pairs still able to reach each other from 0.758 to 0.229. (https://5dlibrary.com/answers/en/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c02/) 第2章 度、介数与结构洞:谁的离开会把社群切开 — 第二章计算度、介数与约束系数,删除枢纽成员会让仍能相互到达的成员对从 0.758 跌到 0.229。 (https://5dlibrary.com/answers/zh/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c02/) - Ch.3 Weighing Reciprocal Edges: Chat Volume, Two-Way Exchange and a Random Baseline — Chapter 3 compares four tie weights and tests reciprocity against an out-degree-preserving null, with a one-sided p-value of only 0.0747 on a single board. (https://5dlibrary.com/answers/en/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c03/) 第3章 互惠边的权重:聊天量、双向往来与随机基准 — 第三章比较边的四种权重,并以保出度零模型检验互惠,单个社群的单侧 p 值只有 0.0747。 (https://5dlibrary.com/answers/zh/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c03/) - Ch.4 Maximum-Weight Matching for Ice-Breaking Pairs: Exact Recursion, Greed, Rotation and Three Possible Worlds — Chapter 4 solves exact maximum-weight matching by bitmask recursion and tests its edge over random pairing in three possible worlds: 0.766 more reciprocal edges per round when the score is right, none when attributes do not matter, and 0.552 fewer when complements attract. (https://5dlibrary.com/answers/en/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c04/) 第4章 破冰配对的最大权匹配:精确递推、贪心、轮转与三个可能世界 — 第四章用位掩码递推求精确最大权匹配,在三个可能世界中检验它对随机配对的优势:评分正确时每轮多建 0.766 条互惠边,属性无关时为零,互补相吸时少建 0.552 条。 (https://5dlibrary.com/answers/zh/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c04/) - Ch.5 Stable Pairing: Irving's Roommates Algorithm, Instances without a Solution and the Price of Satisfaction — Chapter 5 finds stable pairs with Irving's roommates algorithm; under random rankings a board of sixteen has a stable pairing with probability 0.8485. (https://5dlibrary.com/answers/en/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c05/) 第5章 稳定配对:伊尔文室友算法、无解的情形与满意度的代价 — 第五章用伊尔文的室友算法求稳定配对,随机意向下十六人社群有 0.8485 的机会存在稳定配对。 (https://5dlibrary.com/answers/zh/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c05/) - Ch.6 Similarity and Bridging: the Structural Cost of Pairing and the Ethical Limits of Risk Labels — Chapter 6 does the structural accounts of similar and bridging pairs and shows that "no reciprocal edge" used as a risk label is right only 0.240 of the time. (https://5dlibrary.com/answers/en/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c06/) 第6章 同质与跨界:配对的结构代价与风险标签的伦理边界 — 第六章算同质与跨界配对的结构账,并说明把零互惠边当作风险标签时命中率只有 0.240。 (https://5dlibrary.com/answers/zh/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c06/) - Ch.7 Eastern and Western Sources Examined: the Joined Lakes of the Dui Hexagram and Aristotle on Friendship — The "joined lakes" and the "discussing and practising together" of the Dui hexagram can be written as a model of reciprocal bridges between two sub-groups. Program p71 shows that the first reciprocal bridge brings the average number of rounds to reach twelve members down from the cap to 8.61, while even four one-way bridges leave 0.490 of runs unfinished. Book VIII of Aristotle's Nicomachean Ethics puts reciprocity and mutual knowledge at the heart of friendship and distinguishes the grounds of utility, pleasure and character. Program p72 shows that a relationship grounded only in utility survives a year with probability 0.240, against 0.917 for one with all three grounds, and that over twelve meetings a year, raising the per-meeting chance of mutual recognition from 0.10 to 0.20 lifts the chance of mutual acquaintance from 0.718 to 0.931. Neither source discusses how to arrange the first meeting. (https://5dlibrary.com/answers/en/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c07/) 第7章 东西方原典考辨:《周易·兑》的丽泽讲习与亚里士多德论友爱 — 《周易·兑》的「丽泽」与「讲习」,可以写成两子群之间的互惠桥模型。程序 p71 显示,第一条互惠桥把传到十二人的平均轮数从上限降到 8.61,而单向的桥即使有四条也有 0.490 的模拟无法完成。亚里士多德《尼各马可伦理学》第八卷把相互与知晓放在友爱的核心,并区分有用、快乐与品格三种根基;程序 p72 显示,单一出于有用的关系一年后只剩 0.240,三种根基都具备的关系剩 0.917,而一年十二次见面、每次相互认可的概率从 0.10 提高到 0.20,能把相互了解的机会从 0.718 提高到 0.931。两部原典都没有讨论怎样安排第一次相遇。 (https://5dlibrary.com/answers/zh/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c07/) - Ch.8 Closing the Loop: the Intake Pairing Sheet, the Ice-Breaker Pairing Planner and a Pre-Registered Test — Chapter 8 gives the intake pairing sheet, the distilled Ice-Breaker Pairing Planner and a pre-registered test plan needing about 34 intake cohorts per arm with reciprocal edges as the primary outcome. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c08/) 第8章 实证闭环:入群破冰配对表、破冰配对器与预注册检验 — 第八章给出入群破冰配对表、蒸馏小工具破冰配对器与预注册检验方案:以互惠边为主要结局,每组约 34 个入群批次。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/high-ticket-peer-board-cohesion-reciprocal-edges-pairing/c08/) ## 5DL-B-279 · Incentive-Compatible Compliant Referral Programmes: Commission Slopes in a Principal-Agent Model, a Second-Tier Slope of Zero and Maximal Referrals within Anti-Pyramid Limits 中文书名:合规分销激励相容、二级佣金斜率委托代理模型与反传销层级红线内转介绍最大化母典 DOI: pending Cite: Fan, Y. (2026). Incentive-Compatible Compliant Referral Programmes: Commission Slopes in a Principal-Agent Model, a Second-Tier Slope of Zero and Maximal Referrals within Anti-Pyramid Limits (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not individual legal, tax or financial advice, and regulations are given only in gist, subject to human review of the text in force before publication. Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. Abstract: Operators of distribution networks often believe three things: more tiers and a higher second-tier commission sell more; copying competitors' commission rates is safest; and two tiers of commission sit inside the legal red line. This book refutes all three with principal-agent models. The optimal linear commission slope is set by effort cost, risk aversion and noise, and is 0.5263 in the illustrative base case; a second-tier commission paying uplines on downline results mainly buys recruiting, not real retail; and on the most conservative reading adopted here, paying an upline on downline results itself falls among the conduct listed in the 2005 Regulations on the Prohibition of Pyramid Selling, so the second-tier slope of a compliant plan should be zero, and referrals should instead be pursued with single-tier customer rewards paid only for real purchases after the return window. Its parameters are illustrative; all calculations are generated by its programs; the book is not individual legal, tax or financial advice, and regulations are given only in gist, subject to human review of the text in force before publication. Its premise is the anti-pyramid red line: no entry fee, no pay per head, no pay on downline results, no promised returns, and urgency only from real stock and real deadlines. Chapter 1 builds the linear contract model and shows by the informativeness principle that a pure override has no incentive value. In the multitask model of Chapter 2, a second-tier slope of 0.3 makes each seller recruit 2.03 more people a month while real margin falls by 0.20 thousand yuan; only when the real increment in downline reports exceeds 0.50 is a positive second-tier slope worth paying in the model. Chapter 3 uses a branching process to show that referral chains extend on single-tier rewards, the first generation being only 0.4196 of new buyers within six generations, and that in a finite market the largest cohort brings in only 0.876 people each. Chapter 4 finds a profit-maximising referral bonus of 158.5 yuan with 1.1543 referred buyers per seed customer; paid at order, 0.7376 of the bonus goes to fakers, while paid after the return window nobody fakes. Chapter 5 writes the compliance boundary as constraints: when downline reports are mostly self-purchase and displacement, holding the boundary costs almost nothing, and when they are mostly real the cost is about 0.1104, which customer referral can win back. Chapter 6 gives a minimax-regret slope of 0.4705 with a worst regret of 0.1010 under uncertain risk aversion. Chapter 7 examines "The Two Handles" of the Han Feizi and Mackay's crowds: with the same budget, second-tier pay raises early joiners' share from 0.0650 to 0.1085, and with 6 recruits each the last tier holds 0.8333 of everyone. Chapter 8 gives the compliance review sheet, the Commission Slope Calculator and pre-registered tests needing 284 to 1135 sellers per group. Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. 摘要:做分销的经营者常信三句话:层级越多、二级佣金越高卖得越多;佣金率照抄同行最稳;两级分佣在法规红线以内。本书用委托代理模型证伪这三句话。最优的线性佣金斜率由努力成本、风险厌恶与噪音决定,示意基准下是 0.5263;给上线按下线业绩分成的二级佣金,主要买到的是拉人头而不是真实零售;按本书采用的最保守读法,以下线业绩给上线计酬本身落在 2005 年《禁止传销条例》列举的行为之内,合规方案的二级斜率应取零,转介绍改由只为真实购买、退货期满才发放的单层顾客推荐奖励来争取。书中参数为示意;全部计算由书中程序生成;本书是非个人化法律、税务或理财建议,法规只述大意,以上架前人工复核的现行文本为准。全书前提是反传销红线:不收入门费、不按人头计酬、不按下线业绩计酬、不承诺收益,紧迫感只来自真实库存与真实截止。 第一章建立线性合同模型并用信息量原则说明纯抽成没有激励价值。第二章的多任务模型里,二级斜率 0.3 让每人每月多拉 2.03 个下线,真实毛利却少 0.20 千元;只有下线报表的真实增量超过 0.50 时,二级斜率在模型里才值得为正。第三章用分支过程说明推荐链靠单层奖励即可延伸,第一代只占六代以内新买家的 0.4196;有限市场里入会最多的一批人平均只带来 0.876 人。第四章算出利润最大的推荐奖金 158.5 元,每位起点顾客带来 1.1543 位推荐买家;下单即付时奖金的 0.7376 被作假者领走,退货期满才付则无人作假。 第五章把合规边界写成约束:下线报表多为自购与挤占时守住边界几乎没有代价,大半真实时代价约为 0.1104,可改由顾客推荐补回。第六章在风险厌恶不确定时给出极小极大遗憾斜率 0.4705,最坏遗憾 0.1010。第七章考辨《韩非子·二柄》与麦凯笔下的群体狂热:同样预算下,二级计酬让早入会者所得份额从 0.0650 升到 0.1085;每人拉 6 人时最后一层占全体 0.8333。第八章给出佣金方案合规审查表、佣金斜率计算器与预注册检验,样本量每组 284 至 1135 名卖货人。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Linear Contracts: Cost, Risk and Noise Set the Optimal Commission Slope — A linear contract splits a commission plan into incentive, insurance and measurement. With linear contracts, exponential utility and normal noise, the optimal slope is $1/(1+rc\sigma^2)$; in the base case it is 0.5263, and total surplus is the same fraction of what it would be without risk. Noisier results call for lower slopes: doubling noise variance cuts the slope by 0.3214. Losses from missing the optimum grow with the square of the miss, so paying too much is as costly as paying too little. The informativeness principle says that downline results unrelated to the upline's effort belong in the contract at most with a negative weight, to filter common market swings; a positive pure override only adds a risk premium. When the upline can move time from selling to recruiting, is the second-tier commission still pure cost? The next chapter answers with a multitask model. (https://5dlibrary.com/answers/en/compliant-referral-commission-slope-principal-agent-pyramid-limits/c01/) 第1章 线性合同:最优佣金斜率由成本、风险与噪音决定 — 线性合同把一份佣金方案拆成激励、保险与度量三件事。在线性、指数效用、正态噪音的模型里,最优斜率是 $1/(1+rc\sigma^2)$;基准情形下它是 0.5263,总盈余只有不怕风险时的同等比例。噪音越大,斜率越该低:噪音方差加倍,斜率降 0.3214。偏离最优的损失按平方增长,给高和给低同样昂贵。信息量原则说,与上线努力无关的下线业绩,在合同里最多只该以负权重出现,用来滤掉共同的市场波动;正的纯抽成只增加风险补偿。那么,当上线可以把时间从卖货挪到拉人时,二级佣金还是纯成本吗?下一章用多任务模型回答这个问题。 (https://5dlibrary.com/answers/zh/compliant-referral-commission-slope-principal-agent-pyramid-limits/c01/) - Ch.2 Multitasking: a Second-Tier Slope Pushes Effort from Selling to Recruiting — Once the upline can move hours from selling to recruiting, the second-tier commission stops being a harmless pure override and becomes a hand that pushes effort towards recruiting. Under the illustrative parameters, recruiting starts only when the second-tier slope exceeds one fifth of the first-tier slope; at a second-tier slope of 0.3, each seller recruits 2.03 more people a month and reported margin rises by 3.25 thousand yuan, while real margin falls by 0.20 thousand. The jointly optimal second-tier slope is zero whenever the real increment in the downline's report is at most half, and positive only when that increment is very high, which is precisely what is hardest to verify. How many people can a downline grow to, layer by layer, and how many can the last joiners bring in? The next chapter answers with a branching process and a finite market. (https://5dlibrary.com/answers/en/compliant-referral-commission-slope-principal-agent-pyramid-limits/c02/) 第2章 多任务:二级斜率把力气从卖货推向拉人 — 当上线可以把时间从卖货挪到拉人,二级佣金就不再是无害的纯抽成,而是一只把力气推向拉人的手。在示意参数下,二级斜率要超过一级斜率的五分之一,拉人才开始;二级斜率 0.3 时,每人每月多拉 2.03 人,报表毛利多 3.25 千元,真实毛利却少 0.20 千元。联合最优的二级斜率在下线报表的真实增量不超过一半时为零,只在真实增量很高时才为正,而这个增量恰恰最难核实。下线一层一层往下长,到底能长出多少人?最后一批加入的人又能拉到几个?下一章用分支过程与有限市场回答。 (https://5dlibrary.com/answers/zh/compliant-referral-commission-slope-principal-agent-pyramid-limits/c02/) - Ch.3 Referral Trees: Expected Size in a Branching Process and Saturation in a Finite Market — Chapter 3 uses a branching process to show that referral chains extend on single-tier rewards, the first generation being only 0.4196 of new buyers within six generations, and that in a finite market the largest cohort brings in only 0.876 people each. (https://5dlibrary.com/answers/en/compliant-referral-commission-slope-principal-agent-pyramid-limits/c03/) 第3章 推荐树:分支过程的期望人数与有限市场的饱和 — 第三章用分支过程说明推荐链靠单层奖励即可延伸,第一代只占六代以内新买家的 0.4196;有限市场里入会最多的一批人平均只带来 0.876 人。 (https://5dlibrary.com/answers/zh/compliant-referral-commission-slope-principal-agent-pyramid-limits/c03/) - Ch.4 Maximising Referrals: a Bonus Paid Once, Only for a Real Purchase — Chapter 4 finds a profit-maximising referral bonus of 158.5 yuan with 1.1543 referred buyers per seed customer; paid at order, 0.7376 of the bonus goes to fakers, while paid after the return window nobody fakes. (https://5dlibrary.com/answers/en/compliant-referral-commission-slope-principal-agent-pyramid-limits/c04/) 第4章 转介绍最大化:只为真实购买付一次的推荐奖金 — 第四章算出利润最大的推荐奖金 158.5 元,每位起点顾客带来 1.1543 位推荐买家;下单即付时奖金的 0.7376 被作假者领走,退货期满才付则无人作假。 (https://5dlibrary.com/answers/zh/compliant-referral-commission-slope-principal-agent-pyramid-limits/c04/) - Ch.5 The Compliance Boundary: Constrained Optimisation and a Screen for Commission Plans — Chapter 5 writes the compliance boundary as constraints: when downline reports are mostly self-purchase and displacement, holding the boundary costs almost nothing, and when they are mostly real the cost is about 0.1104, which customer referral can win back. (https://5dlibrary.com/answers/en/compliant-referral-commission-slope-principal-agent-pyramid-limits/c05/) 第5章 合规边界:约束优化与佣金方案筛查表 — 第五章把合规边界写成约束:下线报表多为自购与挤占时守住边界几乎没有代价,大半真实时代价约为 0.1104,可改由顾客推荐补回。 (https://5dlibrary.com/answers/zh/compliant-referral-commission-slope-principal-agent-pyramid-limits/c05/) - Ch.6 Estimating the Parameters and a Robust Slope: When Unsure, Err Low — Chapter 6 gives a minimax-regret slope of 0.4705 with a worst regret of 0.1010 under uncertain risk aversion. (https://5dlibrary.com/answers/en/compliant-referral-commission-slope-principal-agent-pyramid-limits/c06/) 第6章 估参数与稳健斜率:估不准时宁低勿高 — 第六章在风险厌恶不确定时给出极小极大遗憾斜率 0.4705,最坏遗憾 0.1010。 (https://5dlibrary.com/answers/zh/compliant-referral-commission-slope-principal-agent-pyramid-limits/c06/) - Ch.7 Classical Sources: the Two Handles of the Han Feizi and Mackay's Crowds — Chapter 7 examines "The Two Handles" of the Han Feizi and Mackay's crowds: with the same budget, second-tier pay raises early joiners' share from 0.0650 to 0.1085, and with 6 recruits each the last tier holds 0.8333 of everyone. (https://5dlibrary.com/answers/en/compliant-referral-commission-slope-principal-agent-pyramid-limits/c07/) 第7章 原典考辨:《韩非子·二柄》的赏罚之柄与麦凯笔下的群体狂热 — 第七章考辨《韩非子·二柄》与麦凯笔下的群体狂热:同样预算下,二级计酬让早入会者所得份额从 0.0650 升到 0.1085;每人拉 6 人时最后一层占全体 0.8333。 (https://5dlibrary.com/answers/zh/compliant-referral-commission-slope-principal-agent-pyramid-limits/c07/) - Ch.8 Closing the Loop: a Compliance Review Sheet, the Commission Slope Calculator and Pre-Registered Tests — Chapter 8 gives the compliance review sheet, the Commission Slope Calculator and pre-registered tests needing 284 to 1135 sellers per group. Marked boxes separate published evidence, runnable Python calculations (17 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/compliant-referral-commission-slope-principal-agent-pyramid-limits/c08/) 第8章 实证闭环:佣金方案合规审查表、佣金斜率计算器与预注册检验 — 第八章给出佣金方案合规审查表、佣金斜率计算器与预注册检验,样本量每组 284 至 1135 名卖货人。书中以标注框区分已发表的证据、可运行的 Python 计算(17 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/compliant-referral-commission-slope-principal-agent-pyramid-limits/c08/) ## 5DL-B-280 · Cross-Border Settlement Frictions and Compliant Treasury Scheduling: Minimum-Loss Network Flows over Multi-Currency Channel Fees and Conversion Slippage 中文书名:跨国多主体结算摩擦与合规资金池调度、多币种通道费率换汇滑点最小损耗网络流母典 DOI: pending Cite: Fan, Y. (2026). Cross-Border Settlement Frictions and Compliant Treasury Scheduling: Minimum-Loss Network Flows over Multi-Currency Channel Fees and Conversion Slippage (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All fees, rates, volatilities, amounts and group structures are illustrative; all calculations are generated by the book's programs; the book is not individual legal, tax or financial advice. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: Cross-border businesses usually convert foreign receipts in one of two ways: each receipt as it lands, or everything at month end. Both rest on one belief: conversion friction is mainly the spread, the spread is charged by amount, and batching hardly matters. This book refutes that belief. On lawful licensed channels with a fixed or minimum fee there is a computable optimal batch between fixed cost, funding and exchange-rate risk, and a group with several entities and currencies must also decide which channel money takes and where it is offset first. All fees, rates, volatilities, amounts and group structures are illustrative; all calculations are generated by the book's programs; the book is not individual legal, tax or financial advice. It discusses only lowering friction on licensed channels with truthful declaration, describes no method of evading exchange controls, structuring to avoid declaration, underground banking or money laundering, and outlines foreign-exchange rules only in general terms, subject to human review before publication. Chapter 1 compares six conversion rules on one month's ledger: with illustrative parameters, converting each order costs 10.93 times the friction of the square-root batch; it also corrects the blueprint, since the square-root law is borrowed from the cash-inventory model of Baumol and Tobin and holds only when there is a fixed fee. Chapter 2 prices exchange-rate risk as variance: the elasticity of the best batch with respect to the fixed fee slides from 0.500 towards one third, and at a risk price of 0.001 the batch shrinks from 42720 to 22833 dollars. Chapter 3 draws a group's currency accounts and licensed channels as a network and finds by minimum-cost flow a route losing 25.69 basis points against 28.15 for direct return, with account shadow prices and bargaining targets for channels. Chapter 4 shows that direct multilateral netting costs only 0.363 of paying invoice by invoice. Chapter 5 lays out a monthly conversion calendar by dynamic programming at 174.2 dollars against 1003.1 for converting only on payment days; with a minimum charge the best batch sits at the kink of 30000 dollars, not the square-root 73993. Chapter 6 tests parameter error, fat-tailed rates and a loss budget, using a tail multiple of 4.3 rather than the normal 3.09. Chapter 7 examines the Quanfu office of the Rites of Zhou (a pool of eight entities takes 0.357 of separate dealings to market) and Fibonacci's chained exchange (a dollar round trip loses 14.1 basis points). Chapter 8 gives the monthly batch table, the Conversion Batch Calculator and a pre-registered test (34 pairs of entity-currency months). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:跨境经营者结汇,常见两种做法:收一笔换一笔,或月底一起换。两者共有一个信念:结汇摩擦主要是点差,点差按金额收,怎么分批都差不多。本书证伪这个信念:在有固定或最低费用的合法持牌通道上,固定成本、资金占用与汇率风险之间存在可计算的最优批量;多主体、多币种的集团还要决定钱走哪条通道、先在哪里抵消。书中的费率、汇率、波动率、金额与主体结构均为示意;全部计算由书中程序生成;本书是非个人化法律、税务或理财建议,只讨论在合法持牌通道上、如实申报前提下降低摩擦,不写任何规避外汇管制、拆分规避申报、地下钱庄或洗钱的手法,外汇管理规定只述大意,以上架前人工复核为准。 第一章用一张月账比较六种结汇规则:示意参数下逐单结汇的摩擦是平方根批量的 10.93 倍;并修正蓝图:平方根律借自鲍莫尔与托宾的现金存货模型,只在有固定费用时成立。第二章把汇率风险按方差计价,最优批量对固定费的弹性从 0.500 滑向三分之一,风险价格 0.001 时批量从 42720 美元缩到 22833 美元。第三章把集团的币种账户与持牌通道画成网络,用最小费用流求出损耗 25.69 个基点的路线,直回为 28.15 个基点,并读出账户影子价格与通道议价目标。第四章说明直接多边净额的成本只是逐张付款的 0.363。 第五章用动态规划排出月度结汇日历,合计 174.2 美元,只在付款日结汇为 1003.1 美元;有最低收费时最优批量钉在拐点 30000 美元而不是平方根的 73993 美元。第六章检验参数误差、汇率厚尾与损失预算:厚尾倍数取 4.3 而不是正态的 3.09。第七章考辨《周礼·泉府》的敛与予(八个主体的资金池到市场的金额为各自交易的 0.357)与斐波那契的连比兑换(美元来回一趟损耗 14.1 个基点)。第八章给出月度结汇批量表、结汇批量计算器与预注册检验方案(34 对主体与币种月份)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Each Receipt or in Batches: One Month's Ledger of Conversion Friction — Chapter 1 compares six conversion rules on one month's ledger: with illustrative parameters, converting each order costs 10.93 times the friction of the square-root batch; it also corrects the blueprint, since the square-root law is borrowed from the cash-inventory model of Baumol and Tobin and holds only when there is a fixed fee. (https://5dlibrary.com/answers/en/cross-border-settlement-friction-treasury-min-cost-flow/c01/) 第1章 逐笔还是批量:一张结汇摩擦的月账 — 第一章用一张月账比较六种结汇规则:示意参数下逐单结汇的摩擦是平方根批量的 10.93 倍;并修正蓝图:平方根律借自鲍莫尔与托宾的现金存货模型,只在有固定费用时成立。 (https://5dlibrary.com/answers/zh/cross-border-settlement-friction-treasury-min-cost-flow/c01/) - Ch.2 The Price of Waiting: How Exchange-Rate Risk Pushes the Square Root towards a Cube Root — Chapter 2 prices exchange-rate risk as variance: the elasticity of the best batch with respect to the fixed fee slides from 0.500 towards one third, and at a risk price of 0.001 the batch shrinks from 42720 to 22833 dollars. (https://5dlibrary.com/answers/en/cross-border-settlement-friction-treasury-min-cost-flow/c02/) 第2章 等待的代价:汇率风险把平方根推向立方根 — 第二章把汇率风险按方差计价,最优批量对固定费的弹性从 0.500 滑向三分之一,风险价格 0.001 时批量从 42720 美元缩到 22833 美元。 (https://5dlibrary.com/answers/zh/cross-border-settlement-friction-treasury-min-cost-flow/c02/) - Ch.3 A Network of Currency Channels: Minimum-Loss Flows and the Shadow Price of Money — Chapter 3 draws a group's currency accounts and licensed channels as a network and finds by minimum-cost flow a route losing 25.69 basis points against 28.15 for direct return, with account shadow prices and bargaining targets for channels. (https://5dlibrary.com/answers/en/cross-border-settlement-friction-treasury-min-cost-flow/c03/) 第3章 多币种通道网络:最小损耗流与账户影子价格 — 第三章把集团的币种账户与持牌通道画成网络,用最小费用流求出损耗 25.69 个基点的路线,直回为 28.15 个基点,并读出账户影子价格与通道议价目标。 (https://5dlibrary.com/answers/zh/cross-border-settlement-friction-treasury-min-cost-flow/c03/) - Ch.4 Offset Before You Transfer: Multilateral Netting and the Fewest Payments — Chapter 4 shows that direct multilateral netting costs only 0.363 of paying invoice by invoice. (https://5dlibrary.com/answers/en/cross-border-settlement-friction-treasury-min-cost-flow/c04/) 第4章 先抵后转:多边净额与最少付款笔数 — 第四章说明直接多边净额的成本只是逐张付款的 0.363。 (https://5dlibrary.com/answers/zh/cross-border-settlement-friction-treasury-min-cost-flow/c04/) - Ch.5 A Month's Conversion Calendar: Dynamic Programming, Minimum Charges and Volume Tiers — Chapter 5 lays out a monthly conversion calendar by dynamic programming at 174.2 dollars against 1003.1 for converting only on payment days; with a minimum charge the best batch sits at the kink of 30000 dollars, not the square-root 73993. (https://5dlibrary.com/answers/en/cross-border-settlement-friction-treasury-min-cost-flow/c05/) 第5章 一个月的结汇日历:动态规划、最低收费与阶梯价 — 第五章用动态规划排出月度结汇日历,合计 174.2 美元,只在付款日结汇为 1003.1 美元;有最低收费时最优批量钉在拐点 30000 美元而不是平方根的 73993 美元。 (https://5dlibrary.com/answers/zh/cross-border-settlement-friction-treasury-min-cost-flow/c05/) - Ch.6 Robustness: Wrong Parameters, Fat-Tailed Rates and a Written Loss Budget — Chapter 6 tests parameter error, fat-tailed rates and a loss budget, using a tail multiple of 4.3 rather than the normal 3.09. (https://5dlibrary.com/answers/en/cross-border-settlement-friction-treasury-min-cost-flow/c06/) 第6章 稳健性:参数猜错、汇率厚尾与成文的损失预算 — 第六章检验参数误差、汇率厚尾与损失预算:厚尾倍数取 4.3 而不是正态的 3.09。 (https://5dlibrary.com/answers/zh/cross-border-settlement-friction-treasury-min-cost-flow/c06/) - Ch.7 Eastern and Western Sources Examined: the Quanfu Office Gathers and Releases, Fibonacci Chains the Rates — Chapter 7 examines the Quanfu office of the Rites of Zhou (a pool of eight entities takes 0.357 of separate dealings to market) and Fibonacci's chained exchange (a dollar round trip loses 14.1 basis points). (https://5dlibrary.com/answers/en/cross-border-settlement-friction-treasury-min-cost-flow/c07/) 第7章 东西方原典考辨:泉府的敛与予,斐波那契的连比兑换 — 第七章考辨《周礼·泉府》的敛与予(八个主体的资金池到市场的金额为各自交易的 0.357)与斐波那契的连比兑换(美元来回一趟损耗 14.1 个基点)。 (https://5dlibrary.com/answers/zh/cross-border-settlement-friction-treasury-min-cost-flow/c07/) - Ch.8 Closing the Loop: the Monthly Batch Table, the Conversion Batch Calculator and a Pre-Registered Test — Chapter 8 gives the monthly batch table, the Conversion Batch Calculator and a pre-registered test (34 pairs of entity-currency months). Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cross-border-settlement-friction-treasury-min-cost-flow/c08/) 第8章 实证闭环:月度结汇批量表、结汇批量计算器与预注册检验 — 第八章给出月度结汇批量表、结汇批量计算器与预注册检验方案(34 对主体与币种月份)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cross-border-settlement-friction-treasury-min-cost-flow/c08/) ## 5DL-B-281 · Multi-Feature Likelihood-Ratio Screening of Pre-Owned Luxury Goods: Fusing Stitch Pitch, Hardware Spectra and Leather Grain, and the Bayesian Boundary between Lab Referral and Instant Purchase 中文书名:奢品真伪多特征似然比鉴定、走线节距五金光谱皮质纹理融合与送检秒收贝叶斯边界母典 DOI: pending Cite: Fan, Y. (2026). Multi-Feature Likelihood-Ratio Screening of Pre-Owned Luxury Goods: Fusing Stitch Pitch, Hardware Spectra and Leather Grain, and the Bayesian Boundary between Lab Referral and Instant Purchase (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Brands are named only generically, the parameters are illustrative, all calculations are generated by the book's programs, and the book is not individual business advice. Chapter 3 gives the effective number of independent features (four features are worth only 2.007 under the illustrative matrix) and a one-number slope of 0.571. Abstract: At the intake desk of a pre-owned luxury store a belief prevails: the more one looks at stitching, hardware, spectra and leather grain, the surer one can be. This book refutes the default premise behind that belief, that these feature readings are independent pieces of evidence. Within one class of items they often share common factors such as wear, photographic lighting, operator habits and batches of parts, so when naive Bayes multiplies their likelihood ratios it records the same evidence several times. Every probability in the book is an appearance-probability reference only, not an official authentication certificate and not a conclusion about authenticity; disputed or high-value items must go to the brand or a professional authentication body. Brands are named only generically, the parameters are illustrative, all calculations are generated by the book's programs, and the book is not individual business advice. Chapter 1 keeps the account of a single feature in likelihood ratios and decibans and adds four features naively. Chapter 2 is the core: with illustrative parameters, items the naive posterior scores above 0.99, at an average of 0.996, are non-conforming in a share of only 0.756; the inflation grows with within-class correlation and reaches 2.28 at a correlation of 0.6; yet the ranking does not suffer, and the damage is concentrated at boundaries far from the prior. Chapter 3 gives the effective number of independent features (four features are worth only 2.007 under the illustrative matrix) and a one-number slope of 0.571. Chapter 4 derives three zones from a loss matrix: an instant buy below a posterior of 0.0539, a referral from 0.0539 to 0.8362, and a decline above; by Good's theorem the gross value of free information is never negative, a real test can change the decision only between 0.0017 and 0.9419, and after fees and seller walk-away the net value of a referral is positive only in the middle band. Chapter 5 shows that cheap extra measurements are worth most at the boundaries (an independent one is worth 444.4 yuan on the instant-buy boundary but only 18.2 yuan in the middle of the band) and that sequential measurement should stop early only outside a wider range, which gives the highest net profit, 2519.4 yuan. Chapter 6 deals with channel priors, verification bias and random audits. Chapter 7 examines the suspected neighbour of the Liezi and Peirce's log odds of 1878. Chapter 8 gives a three-zone intake procedure, the Refer-or-Buy Intake Screener and a pre-registered test plan: testing within-class correlation needs 194 authenticated conforming items, testing log loss 373, and testing the instant-buy zone 386 random audits. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:二手奢品店的收货台流行一个信念:走线、五金、光谱、皮面纹理看得越多,就越有把握。本书证伪这个信念背后的默认前提,即这些特征读数是彼此独立的证据。它们在同一类货品内部往往共享磨损、拍照光线、操作员习惯与零件批次等共同因子;朴素贝叶斯把它们的似然比相乘,就把同一份证据记了好几次。书中全部概率只是外观概率参考,不是官方鉴定证书,不构成真伪结论,涉及争议或高价值货品须送品牌或专业鉴定机构;品牌一律以通称指代;书中参数为示意,全部计算由书中程序生成;本书是非个人化经营建议。 第一章用似然比与分贝班为单一特征记账,并把四个特征朴素相加。第二章是全书核心:示意参数下,朴素后验判到 0.99 以上、平均 0.996 的货,实际不符只有 0.756;放大倍数随类内相关升高,在相关系数 0.6 时达到 2.28;但排序并不吃亏,伤害集中在远离先验的边界上。第三章给出有效独立证据数(示意矩阵下四个特征只相当于 2.007 份)与一个数的斜率收缩 0.571。第四章由损失矩阵导出三区:后验低于 0.0539 秒收,0.0539 到 0.8362 送检,更高则婉拒;按古德的定理,免费信息的毛价值永不为负,真实检验只在 0.0017 到 0.9419 之间能改变决定,扣除费用与卖家流失后,送检净价值只在中间带为正。 第五章说明便宜的加测在边界上最值钱(一项独立加测在秒收边界上值 444.4 元,在中间带正中只值 18.2 元),并说明序贯测量应在更宽的范围外才提前停下,净收益最高为 2519.4 元。第六章处理渠道先验、核实偏倚与随机抽检。第七章考辨《列子·说符》的疑邻盗斧与皮尔士 1878 年的对数几率。第八章给出收货三区流程、送检秒收判定器与预注册检验方案:检验类内相关需 194 件已鉴定的一致货,检验对数损失需 373 件,检验秒收区不符比例需随机抽检 386 件。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Weight of Evidence: Single-Feature Likelihood Ratios, Decibans and the Sum of Four Features — A feature yields a likelihood ratio, its logarithm yields an additive weight of evidence, and the deciban is a convenient unit. With the illustrative parameters, the neutral point for stitch pitch lies at 0.60 spreads, the expected weight per item is 3.127 decibans with a standard deviation of 5.212, and a single feature seldom settles a case. Added naively, four features give an average of 11.94 decibans for a non-conforming item. The addition presumes that the four readings are independent within each hypothesis. The next chapter tests that premise: wear, the lighting of a photograph and the same operator can all push several features at once, and then naive addition records one piece of evidence several times over. (https://5dlibrary.com/answers/en/luxury-authentication-likelihood-ratio-lab-referral-boundary/c01/) 第1章 证据权重:单一特征的似然比、分贝班与四种特征的相加 — 一个特征给出一个似然比,取对数后得到可加的证据权重,单位可以用分贝班。示意参数下,走线针距的中性点是 0.60 个离散单位,单件期望权重 3.127 分贝班,标准差 5.212,单一特征很少能单独定案。四个特征朴素相加时,与规格不符的货平均给出 11.94 分贝班。相加的前提是四个读数在每个假设之内彼此独立。下一章检验这个前提:货品的磨损、拍照光线、同一位操作员,都会让几个特征一起偏,这时朴素相加就会把同一份证据记上好几次。 (https://5dlibrary.com/answers/zh/luxury-authentication-likelihood-ratio-lab-referral-boundary/c01/) - Ch.2 Double Counting: How Correlated Features Make Naive Bayes Overconfident — Chapter 2 is the core: with illustrative parameters, items the naive posterior scores above 0.99, at an average of 0.996, are non-conforming in a share of only 0.756; the inflation grows with within-class correlation and reaches 2.28 at a correlation of 0.6; yet the ranking does not suffer, and the damage is concentrated at boundaries far from the prior. (https://5dlibrary.com/answers/en/luxury-authentication-likelihood-ratio-lab-referral-boundary/c02/) 第2章 重复计数:相关特征让朴素贝叶斯过度自信 — 第二章是全书核心:示意参数下,朴素后验判到 0.99 以上、平均 0.996 的货,实际不符只有 0.756;放大倍数随类内相关升高,在相关系数 0.6 时达到 2.28;但排序并不吃亏,伤害集中在远离先验的边界上。 (https://5dlibrary.com/answers/zh/luxury-authentication-likelihood-ratio-lab-referral-boundary/c02/) - Ch.3 Correcting for Correlation: the Effective Number of Independent Features, Exact Weights and a One-Number Slope — Chapter 3 gives the effective number of independent features (four features are worth only 2.007 under the illustrative matrix) and a one-number slope of 0.571. (https://5dlibrary.com/answers/en/luxury-authentication-likelihood-ratio-lab-referral-boundary/c03/) 第3章 相关校正:有效独立证据数、精确权重与一个数的斜率收缩 — 第三章给出有效独立证据数(示意矩阵下四个特征只相当于 2.007 份)与一个数的斜率收缩 0.571。 (https://5dlibrary.com/answers/zh/luxury-authentication-likelihood-ratio-lab-referral-boundary/c03/) - Ch.4 A Loss Matrix and Three Zones: Instant Buy, Referral and Decline — Chapter 4 derives three zones from a loss matrix: an instant buy below a posterior of 0.0539, a referral from 0.0539 to 0.8362, and a decline above; by Good's theorem the gross value of free information is never negative, a real test can change the decision only between 0.0017 and 0.9419, and after fees and seller walk-away the net value of a referral is positive only in the middle band. (https://5dlibrary.com/answers/en/luxury-authentication-likelihood-ratio-lab-referral-boundary/c04/) 第4章 损失矩阵与三区边界:秒收、送检与婉拒 — 第四章由损失矩阵导出三区:后验低于 0.0539 秒收,0.0539 到 0.8362 送检,更高则婉拒;按古德的定理,免费信息的毛价值永不为负,真实检验只在 0.0017 到 0.9419 之间能改变决定,扣除费用与卖家流失后,送检净价值只在中间带为正。 (https://5dlibrary.com/answers/zh/luxury-authentication-likelihood-ratio-lab-referral-boundary/c04/) - Ch.5 Measuring More or Stopping: Cheap Evidence Is Worth Most at the Boundaries — Chapter 5 shows that cheap extra measurements are worth most at the boundaries (an independent one is worth 444.4 yuan on the instant-buy boundary but only 18.2 yuan in the middle of the band) and that sequential measurement should stop early only outside a wider range, which gives the highest net profit, 2519.4 yuan. (https://5dlibrary.com/answers/en/luxury-authentication-likelihood-ratio-lab-referral-boundary/c05/) 第5章 加测与停测:便宜的证据在边界上最值钱 — 第五章说明便宜的加测在边界上最值钱(一项独立加测在秒收边界上值 444.4 元,在中间带正中只值 18.2 元),并说明序贯测量应在更宽的范围外才提前停下,净收益最高为 2519.4 元。 (https://5dlibrary.com/answers/zh/luxury-authentication-likelihood-ratio-lab-referral-boundary/c05/) - Ch.6 Channel Priors, Audits and Calibration: What Gets Referred Decides What a Store Can Learn — The prior decides which zone the same evidence falls into. With the illustrative parameters, evidence of 4.0 decibans toward N leads to an instant buy in the repeat-consignor channel and to a referral in the private-seller and bulk-lot channels; in bulk lots only 0.400 of conforming items can be bought on the spot, and the instant-buy zone has the highest non-conforming share. Priors may be divided only by business facts such as the channel, never by sellers' personal characteristics. A store learns outcomes only for referred items, which hides miscalibration at the two ends; the referral zone alone suffices to expose the overconfidence of the naive posterior, while the instant-buy and decline zones need random audits, and one item in twenty is far too few. Murphy's decomposition shows that the naive posterior loses almost all its points on reliability. The next chapter leaves the intake desk for two classical sources, to see how earlier thinkers spoke about evidence counted more than once. (https://5dlibrary.com/answers/en/luxury-authentication-likelihood-ratio-lab-referral-boundary/c06/) 第6章 渠道先验、抽检与校准:谁被送检,决定店里能学到什么 — 先验决定同一份证据落在哪一区。示意参数下,往 N 方向 4.0 分贝班的证据,在回头寄卖客渠道秒收,在个人卖家与整批货渠道送检;整批货只有 0.400 的一致货能秒收,秒收区里的不符比例却最高。先验只能按渠道等经营事实来分,不能按卖家的个人特征。店里只知道送检货的结果,这会让两端的失准看不见;送检区本身已足以暴露朴素后验的过度自信,秒收区与婉拒区则需要随机抽检,而二十分之一的比例远远不够。墨菲分解显示,朴素后验的失分几乎全在可靠度上。下一章暂离收货台,到两部原典里看看,古人怎样谈论被重复计数的证据。 (https://5dlibrary.com/answers/zh/luxury-authentication-likelihood-ratio-lab-referral-boundary/c06/) - Ch.7 East and West in the Sources: Suspecting the Neighbour of the Axe, and Peirce's Thermometer of Belief — Chapter 7 examines the suspected neighbour of the Liezi and Peirce's log odds of 1878. (https://5dlibrary.com/answers/en/luxury-authentication-likelihood-ratio-lab-referral-boundary/c07/) 第7章 东西方原典考辨:疑邻盗斧与皮尔士的信念温度计 — 第七章考辨《列子·说符》的疑邻盗斧与皮尔士 1878 年的对数几率。 (https://5dlibrary.com/answers/zh/luxury-authentication-likelihood-ratio-lab-referral-boundary/c07/) - Ch.8 Closing the Loop: a Three-Zone Intake Procedure, the Refer-or-Buy Screener and a Pre-Registered Test — Chapter 8 gives a three-zone intake procedure, the Refer-or-Buy Intake Screener and a pre-registered test plan: testing within-class correlation needs 194 authenticated conforming items, testing log loss 373, and testing the instant-buy zone 386 random audits. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/luxury-authentication-likelihood-ratio-lab-referral-boundary/c08/) 第8章 实证闭环:收货三区流程、送检秒收判定器与预注册检验 — 第八章给出收货三区流程、送检秒收判定器与预注册检验方案:检验类内相关需 194 件已鉴定的一致货,检验对数损失需 373 件,检验秒收区不符比例需随机抽检 386 件。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/luxury-authentication-likelihood-ratio-lab-referral-boundary/c08/) ## 5DL-B-282 · Dynamic Revenue Management and Slot Packing for Space Rentals: Nested Protection Levels and the Elimination of Fragmented Idle Time 中文书名:空间租赁动态收益管理与档期装箱、舱位控制嵌套保护水平与碎片空置消除母典 DOI: pending Cite: Fan, Y. (2026). Dynamic Revenue Management and Slot Packing for Space Rentals: Nested Protection Levels and the Elimination of Fragmented Idle Time (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 1 computes exact expectations for an illustrative month of 20 studio-days: protecting 10 days lifts monthly revenue from 47323.9 yuan to 57432.1 yuan, 21.36 % more, though simulation shows that protection earns less in 0.1900 of months. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: The commonest booking rule in photo studio and venue rental is first come, first served, resting on the intuition that an empty studio is money lost and a cheap short booking beats an idle day. This book refutes that intuition: in seasons when demand exceeds capacity and cheap short bookings arrive first, short bookings accepted first come, first served crowd out dear full-day orders that come later; by Littlewood's 1972 rule, a cheap booking should be accepted only if its revenue is at least the dear fare times the probability that dear demand reaches or exceeds the remaining days. Its studio-days, prices, demands and arrival probabilities are illustrative; all calculations are generated by its programs; the book is not individual business advice, every price is posted in full, scarcity notices come only from the real calendar, and the price table printed by its small tool must be reviewed by the operator before publication. Chapter 1 computes exact expectations for an illustrative month of 20 studio-days: protecting 10 days lifts monthly revenue from 47323.9 yuan to 57432.1 yuan, 21.36 % more, though simulation shows that protection earns less in 0.1900 of months. Chapter 2 compares EMSR-a, EMSR-b, the exact optimum and fixed quotas for four fares: EMSR-b reaches 0.99778 of the exact optimum. Chapter 3 handles interleaved arrivals with bid prices that move over the booking run: bid prices earn 6009.0 yuan a month more than first come, first served, while EMSR-b fixed on total monthly demand earns only 1236.2 yuan more. Chapter 4 packs short bookings with flexible dates into the fewest days by first-fit decreasing, adding 3401.1 yuan a week of full-day revenue. Chapter 5 handles fixed-time bookings by interval partitioning and weighted interval scheduling: first come, first served earns 999.95 yuan a day less than the hindsight optimum. Chapter 6 compares dynamic pricing with a price ladder announced in advance that never falls (0.9622 of the dynamic optimum), and gives a checker for scarcity notices stated only from the real calendar. Chapter 7 examines "follow the season, measure the land" from the Qimin Yaoshu (protecting the same 11 days, the rule by day type expects 57679.4 yuan against 49152.1 for a single rule) and Ricardo's differential rent. Chapter 8 gives the weekly procedure for the studio price table, the Slot Protection Level Calculator and a pre-registered switchback test: 196 weeks for one studio, 25 weeks with eight studios in parallel. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:摄影棚与场地租赁最常见的接单规则是先到先得,背后的直觉是「棚空着就是亏,低价短订也比空着强」。本书证伪这条直觉:在需求超过容量、低价短订先到的季节,先到先得接下的短订会挤掉后来的高价整天订单;按 Littlewood 1972 年的规则,只有当低价收入不低于高价收入乘以高价需求达到或超过剩余档期的概率时才接受低价订单。书中的棚日、价格、需求与到达概率均为示意;全部计算由书中程序生成;本书是非个人化经营建议,所有价格明码标价,稀缺提示只来自真实档期,小工具给出的价格表须经营者人工复核后发布。 第一章在一个 20 个棚日的示意月份里,用精确期望算出保护 10 个棚日可使月收入从 47323.9 元升到 57432.1 元,多 21.36 %,但模拟显示保护在 0.1900 的月份里少收钱。第二章比较四档价格的 EMSR-a、EMSR-b、精确最优与固定配额:EMSR-b 达到精确最优的 0.99778。第三章用随预订期变化的投标价格处理交错到达:投标价格每月比先到先得多收 6009.0 元,而按全月总需求定死的 EMSR-b 只多收 1236.2 元。第四章把弹性日期的短订用首次适应递减装进最少的天数,每周多收 3401.1 元的整天订单收入。 第五章用区间划分与加权区间调度处理定时订单:先到先得每天比事后最优少收 999.95 元。第六章比较动态定价与预先公示、只升不降的价格阶梯(达到动态最优的 0.9622),并给出只按真实档期陈述的稀缺提示检查器。第七章考辨《齐民要术》的「顺天时,量地利」(同样保护 11 天,分日类型的规则期望收入 57679.4 元,一刀切为 49152.1 元)与李嘉图的级差地租。第八章给出摄影棚档期价格表的周度流程、档期保护水平计算器与预注册交替周检验方案:单间棚需要 196 周,八间棚并行需要 25 周。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Cost of First Come, First Served: Littlewood's Two-Class Rule and the Full-Day Orders It Saves — Chapter 1 computes exact expectations for an illustrative month of 20 studio-days: protecting 10 days lifts monthly revenue from 47323.9 yuan to 57432.1 yuan, 21.36 % more, though simulation shows that protection earns less in 0.1900 of months. (https://5dlibrary.com/answers/en/space-rental-revenue-management-protection-levels-slot-packing/c01/) 第1章 先到先得的代价:Littlewood 两档规则与被挤掉的整天订单 — 第一章在一个 20 个棚日的示意月份里,用精确期望算出保护 10 个棚日可使月收入从 47323.9 元升到 57432.1 元,多 21.36 %,但模拟显示保护在 0.1900 的月份里少收钱。 (https://5dlibrary.com/answers/zh/space-rental-revenue-management-protection-levels-slot-packing/c01/) - Ch.2 Nested Protection Levels: EMSR-a, EMSR-b and the Exact Optimum — Chapter 2 compares EMSR-a, EMSR-b, the exact optimum and fixed quotas for four fares: EMSR-b reaches 0.99778 of the exact optimum. (https://5dlibrary.com/answers/en/space-rental-revenue-management-protection-levels-slot-packing/c02/) 第2章 嵌套保护水平:EMSR-a、EMSR-b 与精确最优 — 第二章比较四档价格的 EMSR-a、EMSR-b、精确最优与固定配额:EMSR-b 达到精确最优的 0.99778。 (https://5dlibrary.com/answers/zh/space-rental-revenue-management-protection-levels-slot-packing/c02/) - Ch.3 Bid Prices: How the Value of a Studio-Day Changes over the Booking Run — Chapter 3 handles interleaved arrivals with bid prices that move over the booking run: bid prices earn 6009.0 yuan a month more than first come, first served, while EMSR-b fixed on total monthly demand earns only 1236.2 yuan more. (https://5dlibrary.com/answers/en/space-rental-revenue-management-protection-levels-slot-packing/c03/) 第3章 投标价格:一个棚日的价值随预订期变化 — 第三章用随预订期变化的投标价格处理交错到达:投标价格每月比先到先得多收 6009.0 元,而按全月总需求定死的 EMSR-b 只多收 1236.2 元。 (https://5dlibrary.com/answers/zh/space-rental-revenue-management-protection-levels-slot-packing/c03/) - Ch.4 Slot Packing: First-Fit Decreasing and Fragmented Idle Time — Chapter 4 packs short bookings with flexible dates into the fewest days by first-fit decreasing, adding 3401.1 yuan a week of full-day revenue. (https://5dlibrary.com/answers/en/space-rental-revenue-management-protection-levels-slot-packing/c04/) 第4章 档期装箱:首次适应递减与碎片空置 — 第四章把弹性日期的短订用首次适应递减装进最少的天数,每周多收 3401.1 元的整天订单收入。 (https://5dlibrary.com/answers/zh/space-rental-revenue-management-protection-levels-slot-packing/c04/) - Ch.5 Bookings at Fixed Times: Interval Partitioning and Weighted Interval Scheduling — Chapter 5 handles fixed-time bookings by interval partitioning and weighted interval scheduling: first come, first served earns 999.95 yuan a day less than the hindsight optimum. (https://5dlibrary.com/answers/en/space-rental-revenue-management-protection-levels-slot-packing/c05/) 第5章 定时订单:区间划分与加权区间调度 — 第五章用区间划分与加权区间调度处理定时订单:先到先得每天比事后最优少收 999.95 元。 (https://5dlibrary.com/answers/zh/space-rental-revenue-management-protection-levels-slot-packing/c05/) - Ch.6 Price Ladders Posted in Full: Dynamic Pricing, a Promise Not to Cut, and Real Scarcity — Chapter 6 compares dynamic pricing with a price ladder announced in advance that never falls (0.9622 of the dynamic optimum), and gives a checker for scarcity notices stated only from the real calendar. (https://5dlibrary.com/answers/en/space-rental-revenue-management-protection-levels-slot-packing/c06/) 第6章 明码标价的价格阶梯:动态定价、不降价承诺与真实稀缺 — 第六章比较动态定价与预先公示、只升不降的价格阶梯(达到动态最优的 0.9622),并给出只按真实档期陈述的稀缺提示检查器。 (https://5dlibrary.com/answers/zh/space-rental-revenue-management-protection-levels-slot-packing/c06/) - Ch.7 Classics of East and West: Season and Soil in the Qimin Yaoshu, and Ricardo's Differential Rent — Chapter 7 examines "follow the season, measure the land" from the Qimin Yaoshu (protecting the same 11 days, the rule by day type expects 57679.4 yuan against 49152.1 for a single rule) and Ricardo's differential rent. (https://5dlibrary.com/answers/en/space-rental-revenue-management-protection-levels-slot-packing/c07/) 第7章 东西方原典考辨:《齐民要术》的天时地利与李嘉图的级差地租 — 第七章考辨《齐民要术》的「顺天时,量地利」(同样保护 11 天,分日类型的规则期望收入 57679.4 元,一刀切为 49152.1 元)与李嘉图的级差地租。 (https://5dlibrary.com/answers/zh/space-rental-revenue-management-protection-levels-slot-packing/c07/) - Ch.8 Closing the Loop: The Studio Price Table and a Pre-Registered Test — Chapter 8 gives the weekly procedure for the studio price table, the Slot Protection Level Calculator and a pre-registered switchback test: 196 weeks for one studio, 25 weeks with eight studios in parallel. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/space-rental-revenue-management-protection-levels-slot-packing/c08/) 第8章 实证闭环:摄影棚档期价格表与预注册检验 — 第八章给出摄影棚档期价格表的周度流程、档期保护水平计算器与预注册交替周检验方案:单间棚需要 196 周,八间棚并行需要 25 周。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/space-rental-revenue-management-protection-levels-slot-packing/c08/) ## 5DL-B-283 · Down-Weighting Non-Independent Evidence across Twelve Symbolic Systems: Spurious Resonance from a Shared Birth Input and Orthogonal Latent-Variable Synthesis 中文书名:十二命理体系独立证据降权、同源生辰伪共振剥离与跨东西方符号系统正交合参潜变量母典 DOI: pending Cite: Fan, Y. (2026). Down-Weighting Non-Independent Evidence across Twelve Symbolic Systems: Spurious Resonance from a Shared Birth Input and Orthogonal Latent-Variable Synthesis (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative, and all calculations are generated by its programs. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: When one moment of birth is handed to several symbolic systems at once and most of them "point the same way", people often take it as very strong evidence. This book refutes the premise behind that belief: the systems are not independent witnesses, because they all read one input. It writes twelve systems as deterministic maps of one moment of birth and uses information theory and probability to compute how many bits they read together, how much they repeat one another, how far agreement is spurious resonance, and how correlated evidence should be combined. The book analyses only the structure, information content and probabilities of symbolic systems; it makes no reading of any individual and claims predictive power for no system. Its parameters are illustrative, and all calculations are generated by its programs. Chapter 1 finds that the nominal information of the twelve systems adds to 44.676 bits while their joint entropy is only 18.005 bits, the same as the input itself; in 56 of the 66 pairs of the mutual-information matrix less than 0.05 bit is shared, redundancy gathers in a few blocks of shared cycles, and for day cycles mutual information is set by the greatest common divisor of the periods. Chapter 2 shows that a shared input inflates agreement only when the reader aligns the verdicts: the chance of ten or more agreeing rises to 0.1196, 3.1 times the independent baseline of 0.0385; it also rereads Carlson's double-blind test and Ertel's reanalysis with the book's method. Chapter 3 estimates the effective number of independent pieces of evidence, from 7.55 by participation ratio to 11.41 by Nyholt's formula. Chapter 4 derives the Bayesian combination of correlated evidence: at a correlation of 0.5 the log-likelihood ratio of twelve pieces falls from 1.5000 to 0.2308; with a shared latent factor the naive posterior climbs to 0.9796 while the exact one stops at 0.6000. Chapter 5 finds that a one-factor model fails and instead groups the twelve systems by shared cycles into 5 blocks for orthogonal synthesis. Chapter 6 uses the data-processing inequality to show that in an illustrative model the twelve systems know exactly the 0.5769 millibits that the day of birth itself knows about an outcome, and not a fraction more; fitting after the fact is almost inevitable, since when each statement has only a 0.05 chance of fitting, twelve systems give a 0.9539 chance that at least one fits. Chapter 7 examines Wang Chong's chapter on day selection in the Lunheng and the second book of Cicero's De Divinatione. Chapter 8 gives a read-only down-weighting proposal for synthesis engines, the distilled Shared-Input Evidence Down-Weighter and a pre-registered test plan; in the demonstration a naive probability of 0.917 is down-weighted to 0.750. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:把同一个出生时刻同时交给多套符号体系,若其中多数「同时指向」同一结论,人们常把它当作极强的证据。本书证伪这一信念背后的前提:这些体系并非独立证人,它们读的是同一个输入。书中把十二套体系写成同一出生时刻的确定性映射,用信息论与概率论计算它们合起来读到多少比特、彼此重复多少、同向在多大程度上是伪共振,以及相关证据应当怎样合并。本书只分析符号体系的结构、信息量与概率,不做任何个人占断,不宣称任何体系有预测效力;书中参数为示意,全部计算由书中程序生成。 第一章算出十二套体系的名义信息量相加为 44.676 比特,联合熵只有 18.005 比特,与输入本身相同;互信息矩阵的 66 对里有 56 对共享不到 0.05 比特,冗余集中在共享周期的几块,日周期体系的互信息由周期的最大公约数决定。第二章显示,同源输入只有在解读者把判决对齐时才放大同向:十个以上同向的概率升到 0.1196,是独立基准 0.0385 的 3.1 倍;并以本书方法重读卡尔森的双盲检验与埃特尔的重新分析。第三章估计有效独立证据数,从参与比的 7.55 到奈霍尔特的 11.41 不等。第四章推导相关证据的贝叶斯合并:相关系数 0.5 时,十二路证据的对数似然比从 1.5000 降为 0.2308;共享潜在因子时,朴素后验升到 0.9796,精确后验停在 0.6000。 第五章发现单因子模型不成立,转而按共享周期把十二套体系分成 5 块做正交合参。第六章用数据处理不等式说明,示意模型中十二套体系关于结果的信息恰为出生日本身的 0.5769 毫比特,一点不多;事后对上几乎必然,每句陈述只有 0.05 的机会对上时,十二套体系至少一句对上的概率是 0.9539。第七章考辨王充《论衡·讥日》与西塞罗《论占卜》第二卷。第八章给出合参系统的只读降权提议、蒸馏小工具「同源证据降权器」与预注册检验方案:示范中朴素概率 0.917 经降权为 0.750。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 One Moment of Birth: the Information of Twelve Systems and Their Mutual-Information Matrix — Chapter 1 finds that the nominal information of the twelve systems adds to 44.676 bits while their joint entropy is only 18.005 bits, the same as the input itself; in 56 of the 66 pairs of the mutual-information matrix less than 0.05 bit is shared, redundancy gathers in a few blocks of shared cycles, and for day cycles mutual information is set by the greatest common divisor of the periods. (https://5dlibrary.com/answers/en/twelve-symbolic-systems-shared-input-evidence-down-weighting/c01/) 第1章 同一个出生时刻:十二体系的信息量与互信息矩阵 — 第一章算出十二套体系的名义信息量相加为 44.676 比特,联合熵只有 18.005 比特,与输入本身相同;互信息矩阵的 66 对里有 56 对共享不到 0.05 比特,冗余集中在共享周期的几块,日周期体系的互信息由周期的最大公约数决定。 (https://5dlibrary.com/answers/zh/twelve-symbolic-systems-shared-input-evidence-down-weighting/c01/) - Ch.2 The Illusion of "Pointing the Same Way": Spurious Resonance and the Double-Blind Test — Chapter 2 shows that a shared input inflates agreement only when the reader aligns the verdicts: the chance of ten or more agreeing rises to 0.1196, 3.1 times the independent baseline of 0.0385; it also rereads Carlson's double-blind test and Ertel's reanalysis with the book's method. (https://5dlibrary.com/answers/en/twelve-symbolic-systems-shared-input-evidence-down-weighting/c02/) 第2章 「同时指向」的错觉:伪共振与双盲检验 — 第二章显示,同源输入只有在解读者把判决对齐时才放大同向:十个以上同向的概率升到 0.1196,是独立基准 0.0385 的 3.1 倍;并以本书方法重读卡尔森的双盲检验与埃特尔的重新分析。 (https://5dlibrary.com/answers/zh/twelve-symbolic-systems-shared-input-evidence-down-weighting/c02/) - Ch.3 The Effective Number of Independent Pieces of Evidence: Design Effect, Eigenvalues and Participation Ratio — Chapter 3 estimates the effective number of independent pieces of evidence, from 7.55 by participation ratio to 11.41 by Nyholt's formula. (https://5dlibrary.com/answers/en/twelve-symbolic-systems-shared-input-evidence-down-weighting/c03/) 第3章 有效独立证据数:设计效应、特征值与参与比 — 第三章估计有效独立证据数,从参与比的 7.55 到奈霍尔特的 11.41 不等。 (https://5dlibrary.com/answers/zh/twelve-symbolic-systems-shared-input-evidence-down-weighting/c03/) - Ch.4 Bayesian Combination of Correlated Evidence: Down-Weighting the Log-Odds — Chapter 4 derives the Bayesian combination of correlated evidence: at a correlation of 0.5 the log-likelihood ratio of twelve pieces falls from 1.5000 to 0.2308; with a shared latent factor the naive posterior climbs to 0.9796 while the exact one stops at 0.6000. (https://5dlibrary.com/answers/en/twelve-symbolic-systems-shared-input-evidence-down-weighting/c04/) 第4章 相关证据的贝叶斯合并:对数几率降权 — 第四章推导相关证据的贝叶斯合并:相关系数 0.5 时,十二路证据的对数似然比从 1.5000 降为 0.2308;共享潜在因子时,朴素后验升到 0.9796,精确后验停在 0.6000。 (https://5dlibrary.com/answers/zh/twelve-symbolic-systems-shared-input-evidence-down-weighting/c04/) - Ch.5 One Latent Factor, and Orthogonal Synthesis — Chapter 5 finds that a one-factor model fails and instead groups the twelve systems by shared cycles into 5 blocks for orthogonal synthesis. (https://5dlibrary.com/answers/en/twelve-symbolic-systems-shared-input-evidence-down-weighting/c05/) 第5章 单因子潜变量与正交合参 — 第五章发现单因子模型不成立,转而按共享周期把十二套体系分成 5 块做正交合参。 (https://5dlibrary.com/answers/zh/twelve-symbolic-systems-shared-input-evidence-down-weighting/c05/) - Ch.6 The Data-Processing Inequality: No Number of Systems Can Exceed the Moment of Birth — Chapter 6 uses the data-processing inequality to show that in an illustrative model the twelve systems know exactly the 0.5769 millibits that the day of birth itself knows about an outcome, and not a fraction more; fitting after the fact is almost inevitable, since when each statement has only a 0.05 chance of fitting, twelve systems give a 0.9539 chance that at least one fits. (https://5dlibrary.com/answers/en/twelve-symbolic-systems-shared-input-evidence-down-weighting/c06/) 第6章 数据处理不等式:体系再多也超不过出生时刻 — 第六章用数据处理不等式说明,示意模型中十二套体系关于结果的信息恰为出生日本身的 0.5769 毫比特,一点不多;事后对上几乎必然,每句陈述只有 0.05 的机会对上时,十二套体系至少一句对上的概率是 0.9539。 (https://5dlibrary.com/answers/zh/twelve-symbolic-systems-shared-input-evidence-down-weighting/c06/) - Ch.7 Classical Sources East and West: Wang Chong on Day Selection and Cicero On Divination — The many day books that Wang Chong criticised all read one stem-branch table. Program p71 shows that five day books together hold only 2.765 bits against 3.141 bits added separately, and no number of day books can exceed 5.907 bits; the share of days marked by at least one book is 0.517, below the independent 0.583, because their marks overlap. Cicero's arguments from twins, from the same moment and from Cannae correspond to "same input, different outcomes" and "same outcome, different inputs": each complete twelve-system profile is shared by about 29,660 people a year, and 100 people meeting one end belong on average to 12.00 signs. The gap in the twins argument is the interval between births, with only a 0.250 chance of different double-hour labels 30 minutes apart, and it needs statistical tests like the time-twins study to close it. The next chapter turns the book's conclusions into a down-weighting procedure that can be followed and a pre-registered test plan. (https://5dlibrary.com/answers/en/twelve-symbolic-systems-shared-input-evidence-down-weighting/c07/) 第7章 东西方原典考辨:王充《讥日》与西塞罗《论占卜》 — 王充《讥日》批评的五花八门的日书,读的都是同一张干支表。程序 p71 显示,五本日书合起来只有 2.765 比特,各自相加却是 3.141 比特,再多的日书也超不过 5.907 比特;至少被一本书标出的日子占 0.517,低于独立时的 0.583,因为它们的标记彼此重叠。西塞罗《论占卜》第二卷的孪生子、同一时刻与坎尼战役三条论证,分别对应「同输入异结果」与「同结果异输入」:每份完整的十二体系标签每年由约 29,660 人共享,100 个同一结局的人平均分属 12.00 个星座。孪生子论证的漏洞在于出生间隔,相隔 30 分钟时时辰标签不同的概率只有 0.250,需要时间双胞胎研究那样的统计检验补足。下一章把全书的结论落成一套可以照做的降权流程与预注册检验方案。 (https://5dlibrary.com/answers/zh/twelve-symbolic-systems-shared-input-evidence-down-weighting/c07/) - Ch.8 Closing the Loop: a Down-Weighting Proposal for Synthesis, and Pre-Registered Tests — Chapter 8 gives a read-only down-weighting proposal for synthesis engines, the distilled Shared-Input Evidence Down-Weighter and a pre-registered test plan; in the demonstration a naive probability of 0.917 is down-weighted to 0.750. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/twelve-symbolic-systems-shared-input-evidence-down-weighting/c08/) 第8章 实证闭环:合参权重降权提议与预注册检验 — 第八章给出合参系统的只读降权提议、蒸馏小工具「同源证据降权器」与预注册检验方案:示范中朴素概率 0.917 经降权为 0.750。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/twelve-symbolic-systems-shared-input-evidence-down-weighting/c08/) ## 5DL-B-284 · Error Propagation from Remembered Birth Times and Robust Chart Fixing: Flip Probabilities at Double-Hour Boundaries and the Fewest Past-Event Questions 中文书名:出生时刻记忆误差传播与鲁棒定盘、交界时辰盘面翻转概率与最少前事反推题数母典 DOI: pending Cite: Fan, Y. (2026). Error Propagation from Remembered Birth Times and Robust Chart Fixing: Flip Probabilities at Double-Hour Boundaries and the Fewest Past-Event Questions (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative, and all calculations are generated by its programs. Chapter 1 shows that in the illustrative model 0.2782 of remembered times fall exactly on a double-hour boundary, against 0.0083 without rounding. Abstract: Casting a chart begins with a birth time, and most reported times come from a parent's memory, often as a full hour. This book overturns two default beliefs: that the reported time is the time of the chart, and that a few past-event questions that "fit" amount to a successful rectification. It writes the report as a rounding mixture and computes the posterior of the true time, the flip probability at double-hour boundaries and the information needed to separate candidates. It analyses only the structure of chart rules, error propagation and probability; it makes no reading of any individual and claims no predictive validity for any system. Its parameters are illustrative, and all calculations are generated by its programs. Chapter 1 shows that in the illustrative model 0.2782 of remembered times fall exactly on a double-hour boundary, against 0.0083 without rounding. Chapter 2 checks the blueprint claim: within a quarter hour of a boundary, remembered reports flip with an average probability of 0.4401 and hospital records only 0.1135, while a time reported exactly on a boundary flips half the time whatever its source. Chapter 3 extends the grid to ke, clock hours and rising signs; with the layers stacked, 0.7230 of people have at least one layer that differs. Chapters 4 and 5 use entropy, Chernoff information and expected information gain to bound the number of questions: items right eight times in ten need 7, items right six times in ten need 67. Chapter 6 computes after-the-fact matches under the null: with items of no power, asking until something fits confirms within 40 items with probability 0.9958. Chapter 7 examines Shen Kuo on the water clock and Bessel's personal equation, and shows that in western China the single convention of apparent solar time changes the double hour of 0.9932 of births. Chapter 8 gives the tollbooth birth-time calibration questionnaire, the Boundary Flip Probability Calculator and a pre-registered test: 169 people for two candidates at a true hit rate of six in ten, screening about 676 records. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:排盘的第一步是出生时刻,而多数人报出的时刻来自父母回忆,常常是一个整点。本书推翻两条默认信念:报出的时刻就是排盘的时刻;问几道前事题「对上了」就算校准成功。书中把报告时刻写成取整混合模型,计算真实时刻的后验、时辰交界的翻转概率与区分候选所需的信息量。本书只分析排盘规则的结构、误差传播与概率,不做任何个人占断,不宣称任何体系有预测效力;书中参数为示意,全部计算由书中程序生成。 第一章说明,示意模型中凭记忆报告的时刻有 0.2782 恰好落在时辰分界上,不取整时只有 0.0083。第二章核实立项命题:分界前后一刻钟内,凭记忆报告的平均翻转概率是 0.4401,医院记录只有 0.1135,恰好报在分界上的时刻不论来源都是一半。第三章把格子扩展到刻、整点与上升星座,几层叠加后有 0.7230 的人至少有一层盘面不符。第四、五章用熵、切尔诺夫信息与期望信息增益给出题数下限:正确率八成的题要 7 道,六成的要 67 道。 第六章计算零假设下的事后对上:题目毫无区分力时,问到对上为止,40 题之内的确认概率是 0.9958。第七章考辨沈括论漏刻与贝塞尔的个人方程,并说明在中国西部真太阳时这一个约定能改变 0.9932 的出生时辰。第八章给出收费站出生时刻校准问卷、交界翻转概率计算器与预注册检验:两个候选、真实命中率六成时需要 169 人,约筛查 676 份记录。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Heaping in Reported Times: Full Hours, Half Hours and a Rounding Mixture — Chapter 1 shows that in the illustrative model 0.2782 of remembered times fall exactly on a double-hour boundary, against 0.0083 without rounding. (https://5dlibrary.com/answers/en/birth-time-memory-error-boundary-flip-probability-fewest-questions/c01/) 第1章 报告时刻的堆积:整点、半点与取整混合模型 — 第一章说明,示意模型中凭记忆报告的时刻有 0.2782 恰好落在时辰分界上,不取整时只有 0.0083。 (https://5dlibrary.com/answers/zh/birth-time-memory-error-boundary-flip-probability-fewest-questions/c01/) - Ch.2 A Quarter Hour Either Side of the Boundary: Exact Flip Probabilities — Chapter 2 checks the blueprint claim: within a quarter hour of a boundary, remembered reports flip with an average probability of 0.4401 and hospital records only 0.1135, while a time reported exactly on a boundary flips half the time whatever its source. (https://5dlibrary.com/answers/en/birth-time-memory-error-boundary-flip-probability-fewest-questions/c02/) 第2章 交界前后一刻钟:翻转概率的精确计算 — 第二章核实立项命题:分界前后一刻钟内,凭记忆报告的平均翻转概率是 0.4401,医院记录只有 0.1135,恰好报在分界上的时刻不论来源都是一半。 (https://5dlibrary.com/answers/zh/birth-time-memory-error-boundary-flip-probability-fewest-questions/c02/) - Ch.3 Cell Widths and Stacked Layers: Ke, Double Hours and Rising Signs — Chapter 3 extends the grid to ke, clock hours and rising signs; with the layers stacked, 0.7230 of people have at least one layer that differs. (https://5dlibrary.com/answers/en/birth-time-memory-error-boundary-flip-probability-fewest-questions/c03/) 第3章 格子宽度与多层盘面:刻、时辰与上升星座 — 第三章把格子扩展到刻、整点与上升星座,几层叠加后有 0.7230 的人至少有一层盘面不符。 (https://5dlibrary.com/answers/zh/birth-time-memory-error-boundary-flip-probability-fewest-questions/c03/) - Ch.4 Information Bounds: How Many Bits a Report Leaves and a Question Buys — Chapters 4 and 5 use entropy, Chernoff information and expected information gain to bound the number of questions: items right eight times in ten need 7, items right six times in ten need 67. (https://5dlibrary.com/answers/en/birth-time-memory-error-boundary-flip-probability-fewest-questions/c04/) 第4章 信息论下限:一个报告值几比特,一道题值几比特 — 第四、五章用熵、切尔诺夫信息与期望信息增益给出题数下限:正确率八成的题要 7 道,六成的要 67 道。 (https://5dlibrary.com/answers/zh/birth-time-memory-error-boundary-flip-probability-fewest-questions/c04/) - Ch.5 Bayesian Updating and Choosing Items by Information Gain — Expected information gain gives a transparent rule for choosing items: always ask the one that lowers the posterior entropy most on average. In the illustrative bank no item gains more than 0.1564 bits, far below the prior entropy of 1.5219 bits; a person's posterior can swing widely during a session, and where one stops changes the conclusion. Averaged over ten thousand people, the greedy error after four items is 0.2808 against 0.3648 for a random order; after all twelve, every policy reaches the same floor of about one in five. Respondents' misdating and yea-saying weaken every item further: with a misdating probability of 0.3, an item that needed 7 repetitions needs 15. The next chapter takes up the most dangerous case: items with no power at all, and after-the-fact matches that still appear again and again. (https://5dlibrary.com/answers/en/birth-time-memory-error-boundary-flip-probability-fewest-questions/c05/) 第5章 贝叶斯更新与信息增益选题 — 期望信息增益给出了一个透明的选题规则:每次问能让后验熵平均下降最多的那道题。在示意题库里,单题增益最多 0.1564 比特,远小于先验熵 1.5219 比特;一个人的后验可以在问答中大幅摇摆,停在哪一题会改变结论。按一万人平均,贪心法在四道题时的判错率是 0.2808,随机顺序是 0.3648;问满十二道题,所有策略都落在约两成的地板上。答题者记错年份与顺从倾向会进一步削弱每道题:记错概率为 0.3 时,原来 7 道够用的题要 15 道。下一章讨论最危险的情形:题目其实毫无区分力,而「事后对上」仍然频繁出现。 (https://5dlibrary.com/answers/zh/birth-time-memory-error-boundary-flip-probability-fewest-questions/c05/) - Ch.6 After-the-Fact Matches: Loose Matching, Optional Stopping and Double Counting — Chapter 6 computes after-the-fact matches under the null: with items of no power, asking until something fits confirms within 40 items with probability 0.9958. (https://5dlibrary.com/answers/en/birth-time-memory-error-boundary-flip-probability-fewest-questions/c06/) 第6章 事后对上:宽松匹配、可选停止与重复计数 — 第六章计算零假设下的事后对上:题目毫无区分力时,问到对上为止,40 题之内的确认概率是 0.9958。 (https://5dlibrary.com/answers/zh/birth-time-memory-error-boundary-flip-probability-fewest-questions/c06/) - Ch.7 Classics East and West: Shen Kuo on the Water Clock and Bessel's Personal Equation — Chapter 7 examines Shen Kuo on the water clock and Bessel's personal equation, and shows that in western China the single convention of apparent solar time changes the double hour of 0.9932 of births. (https://5dlibrary.com/answers/en/birth-time-memory-error-boundary-flip-probability-fewest-questions/c07/) 第7章 东西方原典考辨:沈括论漏刻与贝塞尔的个人方程 — 第七章考辨沈括论漏刻与贝塞尔的个人方程,并说明在中国西部真太阳时这一个约定能改变 0.9932 的出生时辰。 (https://5dlibrary.com/answers/zh/birth-time-memory-error-boundary-flip-probability-fewest-questions/c07/) - Ch.8 The Tollbooth Birth-Time Calibration Questionnaire: Read-Only Advice and a Pre-Registered Test — Chapter 8 gives the tollbooth birth-time calibration questionnaire, the Boundary Flip Probability Calculator and a pre-registered test: 169 people for two candidates at a true hit rate of six in ten, screening about 676 records. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/birth-time-memory-error-boundary-flip-probability-fewest-questions/c08/) 第8章 收费站出生时刻校准问卷:只读建议与预注册检验 — 第八章给出收费站出生时刻校准问卷、交界翻转概率计算器与预注册检验:两个候选、真实命中率六成时需要 169 人,约筛查 676 份记录。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/birth-time-memory-error-boundary-flip-probability-fewest-questions/c08/) ## 5DL-B-285 · Attachment Games and Ceasefire Protocols in Intimate Partnerships: Repeated Non-Zero-Sum Games, Emotion-Decay Equations and Ending Cold Wars between Founder Couples 中文书名:亲密关系依恋博弈与冲突停火协议、重复非零和博弈情绪衰减方程与创业伴侣冷战破局母典 DOI: pending Cite: Fan, Y. (2026). Attachment Games and Ceasefire Protocols in Intimate Partnerships: Repeated Non-Zero-Sum Games, Emotion-Decay Equations and Ending Cold Wars between Founder Couples (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative, and all calculations are generated by its programs. The parts on founder couples are not individual business advice. Abstract: A piece of common sense circulates among couples: I treat you the way you treat me. This book refutes the habit of carrying that tit-for-tat rule straight into close relationships. Once there is misreading noise, two partners who both answer in kind fall, after a single misunderstanding, into an echo of blow and counter-blow, spend only a quarter of their time in the long run both turning toward each other, and this share does not depend on how much misreading there is. Using Markov chains for the noisy repeated prisoner's dilemma, evolutionary simulation of reactive strategies, emotion-decay differential equations and absorbing chains, the book derives a safe dose of forgiveness and the structure of a ceasefire protocol. Its parameters are illustrative, and all calculations are generated by its programs. It is not counselling, medical or legal advice, claims no effect on any relationship or on anyone's body or mind, and teaches no manipulation, silent treatment or emotional blackmail; anyone facing domestic violence, threats, a risk of self-harm or severe distress should contact local emergency services and professional help at once. The parts on founder couples are not individual business advice. Chapters 1 and 2 show where the echo comes from and compute the safe window of forgiveness: at an implementation noise of 0.05 the forgiveness probability should lie between 0.0437 and 0.2811, since too little traps a couple in the echo and too much invites exploitation; at a noise of 0.10 the window closes, and it is misreading that must come down. Chapter 3 writes attachment anxiety and avoidance as rates of misreading, forgiveness and withdrawal; in the pursue-and-retreat loop, turning withdrawal into a pause with a return time lowers the chance of escalation from 0.373 to 0.057. The decay equations of Chapter 4 put the median cooling time of a couple at 19.0 minutes; a fixed 20-minute pause suffices for only 0.544 of couples, while a pause set from the slower partner's estimate reaches 0.918. Chapter 5 assembles these results into a ceasefire protocol and uses the shadow of the future to show what makes it self-enforcing. Chapter 6 turns to founder couples: spillover raises the share of weeks with conflict at both company and home from 0.0179 to 0.0961, and a venue rule brings it down to 0.0493; allocating decision rights by domain cuts yearly escalations from 7.49 to 0.77, at the price of about 22.61 overruled decisions. Chapter 7 examines "repay grievance with uprightness" from the Analects and Seneca's On Anger: three rules of response average 2.076, 2.286 and 2.344. Chapter 8 presents a ceasefire card signed by both partners, the Cold-War Ceasefire Card Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:伴侣之间流行一条常识:你怎么待我,我就怎么待你。本书证伪把这条「一报还一报」规则直接搬进亲密关系的做法:只要有误读的噪音,两个奉行对等回应的人就会在一次误会之后陷入轮流出刀的回声,长期只有四分之一的时间处在双方靠近的状态,而且这个比例与误读多少无关。书中用带噪音的重复囚徒困境马尔可夫链、反应策略的演化模拟、情绪衰减微分方程与吸收链,推导宽恕的安全剂量与停火协议的结构。书中参数均为示意,全部计算由书中程序生成。本书不是心理咨询、医疗或法律建议,不宣称任何改变关系或身心状况的效果,也不写任何操控、冷暴力或情感勒索技巧;遇到家庭暴力、威胁、自伤风险或严重心理困扰,请立即联系当地紧急求助渠道与专业机构。涉及创业伴侣的部分属非个人化经营建议。 第一、二章说明回声从哪里来,并算出宽恕的安全窗口:执行噪音为 0.05 时,宽恕概率应在 0.0437 与 0.2811 之间,太少困在回声里,太多被占便宜;噪音到 0.10 时窗口关闭,此时该降低的是误读。第三章把依恋的焦虑与回避写成误读率、宽恕率与退缩率,在追逃循环中把退缩换成有归期的暂停,使升级概率从 0.373 降到 0.057。第四章的衰减方程显示,伴侣冷却时间的中位数是 19.0 分钟,固定暂停 20 分钟只对 0.544 的伴侣够用,按冷却较慢一方的估计取暂停可达 0.918。 第五章把这些结果装配成停火协议,并用未来的影子说明它靠什么自我执行。第六章讨论创业伴侣:外溢使公司与家同时冲突的周数占比从 0.0179 升到 0.0961,场合规则降到 0.0493;按领域分配决策权把每年的升级从 7.49 次降到 0.77 次,代价是约 22.61 次被否决。第七章考辨《论语·宪问》的「以直报怨」与塞涅卡《论愤怒》:三种回应规则的平均收益依次是 2.076、2.286 与 2.344。第八章给出双方签署的停火协议卡、冷战停火协议计算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Echo: an Exact Markov Chain for Tit for Tat under Noise — Chapters 1 and 2 show where the echo comes from and compute the safe window of forgiveness: at an implementation noise of 0.05 the forgiveness probability should lie between 0.0437 and 0.2811, since too little traps a couple in the echo and too much invites exploitation; at a noise of 0.10 the window closes, and it is misreading that must come down. (https://5dlibrary.com/answers/en/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c01/) 第1章 回声报复:噪音下一报还一报的精确马尔可夫链 — 第一、二章说明回声从哪里来,并算出宽恕的安全窗口:执行噪音为 0.05 时,宽恕概率应在 0.0437 与 0.2811 之间,太少困在回声里,太多被占便宜;噪音到 0.10 时窗口关闭,此时该降低的是误读。 (https://5dlibrary.com/answers/zh/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c01/) - Ch.2 A Safe Dose of Forgiveness: Generous Tit for Tat and the Invasion Window — Forgiveness has a safe window. At a noise level of 0.05 the forgiveness probability should lie between 0.0437 and 0.2811: too little traps a couple in the echo and cannot even hold off an alternator, too much is exploited by alternators and unconditional defectors. The more noise, the narrower the window, and at eps of 0.10 it closes; at that point it is the misreading that has to come down, not the forgiveness that has to be tuned. The evolutionary simulation reproduces the four acts in which tit for tat opens the way and generous rules take over, ending with (0.988, 0.433) as the most common strategy. The payoff table shifts the safe dose: when the last word is sweet, the ceiling is only 1/6. The next chapter asks whether the anxiety and avoidance of attachment research can be written as these parameters. (https://5dlibrary.com/answers/en/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c02/) 第2章 宽恕的安全剂量:宽恕型以牙还牙与入侵窗口 — 宽恕有一个安全窗口。在 eps 为 0.05 的噪音下,宽恕概率应在 0.0437 与 0.2811 之间:太少会被困在回声里,连交替者都守不住;太多会被交替者和无条件背叛者占便宜。噪音越大,窗口越窄,到 eps 为 0.10 时窗口关闭,此时该降低的是误读,而不是调整宽恕。演化模拟复现了「一报还一报开路、宽恕型接班」的四幕,最终占比最大的策略是 (0.988, 0.433)。收益表改变安全剂量:「狠话甜」时上限只有 1/6。下一章要问的是:依恋研究里的焦虑与回避,能不能写成这些参数? (https://5dlibrary.com/answers/zh/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c02/) - Ch.3 Attachment as Parameters: Misreading, Forgiveness and the Demand-Withdraw Loop — Chapter 3 writes attachment anxiety and avoidance as rates of misreading, forgiveness and withdrawal; in the pursue-and-retreat loop, turning withdrawal into a pause with a return time lowers the chance of escalation from 0.373 to 0.057. The decay equations of Chapter 4 put the median cooling time of a couple at 19.0 minutes; a fixed 20-minute pause suffices for only 0.544 of couples, while a pause set from the slower partner's estimate reaches 0.918. (https://5dlibrary.com/answers/en/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c03/) 第3章 依恋作为参数:误读率、宽恕率与追逃循环 — 第三章把依恋的焦虑与回避写成误读率、宽恕率与退缩率,在追逃循环中把退缩换成有归期的暂停,使升级概率从 0.373 降到 0.057。 (https://5dlibrary.com/answers/zh/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c03/) - Ch.4 Emotion-Decay Equations: Coupling, Cooling Times and Two Attractors — Emotion decays exponentially, and the cooling time is $\tau \ln(E_0/\theta)$. Two people together feed each other, and once the coupling passes the illustrative critical value of 0.0645 per minute they amplify each other; the first job of a pause is to cut the coupling. Cooling times vary widely: for couples the median is 19.0 minutes and the 95th percentile 39.6 minutes, a fixed 20 minutes suffices for only 0.544 of couples, while taking 1.25 times the slower partner's estimate leaves 0.918 of couples ready on return. The nonlinear difference equations have a positive and a negative attractor, and an opening remark below -2.540 tips the conversation toward the negative one. The next chapter assembles these results into a ceasefire protocol: how long to pause, how to name the return, and what makes the protocol self-enforcing. (https://5dlibrary.com/answers/en/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c04/) 第4章 情绪衰减方程:耦合、冷却时间与两个吸引子 — 情绪按指数衰减,冷却时间是 $\tau \ln(E_0/\theta)$。两人在一起时会相互喂养,耦合系数越过 0.0645 每分钟这一示意临界值后,两人会互相放大;暂停的首要作用是切断耦合。冷却时间在人群中很宽:伴侣的中位数是 19.0 分钟,第 95 百分位是 39.6 分钟,固定 20 分钟只对 0.544 的伴侣够用,而按冷却较慢一方的估计取 1.25 倍,可以让 0.918 的伴侣在恢复时都已准备好。非线性差分方程有正负两个吸引子,开头一句低于 -2.540 就会滑向负的一侧。下一章把这些结果组装成一份停火协议:暂停多长,归期怎么说,协议靠什么才能自我执行? (https://5dlibrary.com/answers/zh/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c04/) - Ch.5 The Structure of a Ceasefire: Pause Length, a Named Return and the Shadow of the Future — A pause is a trade: more total time for less time in flares. In the illustrative model a disagreement without a pause spends 25.0 minutes in a flare, and a 20-minute pause cuts this to 6.3 minutes while the total rises from 45.0 to 53.4 minutes. A return time changes how long the waiting partner can tolerate, so the best pause can lengthen with the cooling time, and the cost falls from 83.8 without a return time to 78.7 with one (tau of 15 minutes). Whether the protocol enforces itself depends on the shadow of the future: under Axelrod's table both partners must expect at least 3.00 further rounds. The next chapter turns to a special kind of couple, partners who are both spouses and co-founders, whose conflicts spill over between two boards. Where should the boundary between the boards be drawn? (https://5dlibrary.com/answers/en/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c05/) 第5章 停火协议的结构:暂停长度、归期与未来的影子 — 暂停是一笔交换:用更长的总时间换更少的升温时间。在示意模型中,不暂停的分歧有 25.0 分钟在升温,暂停 20 分钟时降到 6.3 分钟,但总时长从 45.0 分钟升到 53.4 分钟。归期改变等待者的容忍时长,使最优暂停能随冷却时间变长,代价从无归期时的 83.8 降到 78.7(tau 为 15 分钟)。协议能否自我执行取决于未来的影子:阿克塞尔罗德表下需要预期至少 3.00 个回合。下一章转向一种特殊的伴侣:既是夫妻、又是合伙人的创业伴侣,他们的冲突会在两张棋盘之间外溢,该怎样划界? (https://5dlibrary.com/answers/zh/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c05/) - Ch.6 Founder Couples: Two Boards at Once — Chapter 6 turns to founder couples: spillover raises the share of weeks with conflict at both company and home from 0.0179 to 0.0961, and a venue rule brings it down to 0.0493; allocating decision rights by domain cuts yearly escalations from 7.49 to 0.77, at the price of about 22.61 overruled decisions. (https://5dlibrary.com/answers/en/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c06/) 第6章 创业伴侣:公司与家两张棋盘 — 第六章讨论创业伴侣:外溢使公司与家同时冲突的周数占比从 0.0179 升到 0.0961,场合规则降到 0.0493;按领域分配决策权把每年的升级从 7.49 次降到 0.77 次,代价是约 22.61 次被否决。 (https://5dlibrary.com/answers/zh/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c06/) - Ch.7 Classical Sources: Repaying Grievance with Uprightness, and Seneca on Anger — Chapter 7 examines "repay grievance with uprightness" from the Analects and Seneca's On Anger: three rules of response average 2.076, 2.286 and 2.344. (https://5dlibrary.com/answers/en/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c07/) 第7章 东西方原典考辨:以直报怨与塞涅卡论愤怒 — 第七章考辨《论语·宪问》的「以直报怨」与塞涅卡《论愤怒》:三种回应规则的平均收益依次是 2.076、2.286 与 2.344。 (https://5dlibrary.com/answers/zh/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c07/) - Ch.8 Closing the Loop: a Ceasefire Card Signed by Both Partners, and a Plan to Refute It — Chapter 8 presents a ceasefire card signed by both partners, the Cold-War Ceasefire Card Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c08/) 第8章 实证闭环:双方签署的停火协议卡与证伪方案 — 第八章给出双方签署的停火协议卡、冷战停火协议计算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/attachment-games-ceasefire-protocol-emotion-decay-founder-couples/c08/) ## 5DL-B-286 · Areas under Continuous Glucose Curves, the Two-Compartment Minimal Model of Insulin Action and the Dynamics of the Switch to Ketone Metabolism 中文书名:连续血糖曲线下面积、双室胰岛素最小模型与生酮代谢切换动力学母典 DOI: pending Cite: Fan, Y. (2026). Areas under Continuous Glucose Curves, the Two-Compartment Minimal Model of Insulin Action and the Dynamics of the Switch to Ketone Metabolism (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: It is not medical advice, claims no effect of any kind and offers no dietary, fasting or medication plan; any self-experiment must first be discussed with a licensed physician. Its parameters are illustrative; all calculations are generated by its programs. Abstract: Continuous glucose monitoring brought two popular beliefs with it: how far a food raises glucose is a property of the food, found by looking it up, and some time after eating stops the body "switches" at one moment to burning fat and ketones. This book examines the opposite propositions and breaks every claim into parts that can be computed and refuted by data. It is not medical advice, claims no effect of any kind and offers no dietary, fasting or medication plan; any self-experiment must first be discussed with a licensed physician. Its parameters are illustrative; all calculations are generated by its programs. Chapter 1 shows that the area under a curve is a set of conventions: from one set of readings the total area is 13410.0 and the exactly cut incremental area 2710.0, and each mg/dL of baseline moves the area by about eighty. Chapter 2 starts from the record of 800 people and 46,898 meals by Zeevi and colleagues and splits individual variation into person, person by food and repeat noise: in the illustrative model only 0.610 of a single comparison is real; for two foods with a small average difference only 0.611 of people order them as the average does, against 0.921 for a large difference. Chapter 3 writes out the Bergman minimal model and shows that one curve cannot pin down the decay rate of remote action. Chapter 4 finds that with glucose alone, insulin sensitivity and secretion appear only as a product, both with an elasticity of 0.531. Chapter 5 uses an illustrative dynamic model to show that the metabolic switch is gradual: ketones take 18.2 hours to rise from one tenth to nine tenths, clearance turnover keeps the width above 8.8 hours even when the molecular step is a toggle, and seven definitions of "switched" differ by 12.6 hours. Chapter 6 converts mean absolute relative difference into error bands: a relative standard deviation of 0.1429, or plus or minus 42.0 for one reading at 150 mg/dL. Chapter 7 examines the Suwen and Bernard's criticism of averages: the average curve and the average of individuals share an incremental area of 6461.7, yet individuals spend 37.6 minutes above 140 mg/dL on average while the average curve spends none. Chapter 8 gives a personal log procedure, the Postprandial Curve Area Calculator and a pre-registered plan: a naive test on an A-B-A log raises false alarms at 0.206, randomised blocks of three days at 0.043, and three weeks can detect a log difference of 0.257. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:连续血糖监测流行之后,两种说法随之流行:食物的升糖高低是食物自己的属性,查表即知;停止进食之后,身体会在某一刻「切换」到燃烧脂肪与酮体。本书检验相反的命题,并把每个说法拆成可以计算、可以被数据推翻的部分。本书不是医疗建议,不宣称任何功效,不提供任何饮食、禁食或用药安排;任何自我试验须先咨询执业医师。书中参数为示意;全部计算由书中程序生成。 第一章说明曲线下面积是一套约定:同一串读数,总面积 13410.0,精确切开的增量面积 2710.0,基线每差 1 mg/dL 面积就变八十上下。第二章以泽维等人对 800 人、46,898 餐的记录为出发点,把个体差异分解为人、人与食物交互和重复噪声:示意模型中单次比较只有 0.610 是真实差异;平均差很小的两种食物,只有 0.611 的人个人排序与平均一致,平均差大时为 0.921。第三章写出 Bergman 最小模型并说明一条曲线确定不了远端作用的衰减速率。第四章发现只测血糖时,胰岛素敏感性与分泌只以乘积出现,二者弹性同为 0.531。 第五章用示意动力学模型说明代谢切换是渐变:酮体从一成升到九成用了 18.2 小时,即使分子层面像开关,清除周转也把宽度限制在 8.8 小时以上,七种「切换」定义相差 12.6 小时。第六章把平均绝对相对差换算成误差带:相对标准差 0.1429,在 150 mg/dL 处单读数误差带为正负 42.0。第七章考辨《素问·奇病论》与贝尔纳对平均数的批评:平均曲线与个人平均的增量面积同为 6461.7,个人高于 140 mg/dL 的时间平均 37.6 分钟,平均曲线却为零。第八章给出个人日志的流程、餐后曲线面积计算器与预注册方案:A-B-A 日志的朴素检验误报比例为 0.206,三天一组的随机化区组为 0.043,三周日志可检出的最小对数差为 0.257。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Trapezoids and Increments: Numerical Methods and Errors for the Area under a Continuous Glucose Curve — Chapter 1 shows that the area under a curve is a set of conventions: from one set of readings the total area is 13410.0 and the exactly cut incremental area 2710.0, and each mg/dL of baseline moves the area by about eighty. (https://5dlibrary.com/answers/en/continuous-glucose-auc-minimal-model-ketone-switch/c01/) 第1章 梯形与增量:连续血糖曲线下面积的数值方法与误差 — 第一章说明曲线下面积是一套约定:同一串读数,总面积 13410.0,精确切开的增量面积 2710.0,基线每差 1 mg/dL 面积就变八十上下。 (https://5dlibrary.com/answers/zh/continuous-glucose-auc-minimal-model-ketone-switch/c01/) - Ch.2 Same Meal, Different Curves: Decomposing Individual Variation and the Ceiling on Prediction — Chapter 2 starts from the record of 800 people and 46,898 meals by Zeevi and colleagues and splits individual variation into person, person by food and repeat noise: in the illustrative model only 0.610 of a single comparison is real; for two foods with a small average difference only 0.611 of people order them as the average does, against 0.921 for a large difference. (https://5dlibrary.com/answers/en/continuous-glucose-auc-minimal-model-ketone-switch/c02/) 第2章 同一餐,不同的曲线:个体差异的方差分解与可预测上限 — 第二章以泽维等人对 800 人、46,898 餐的记录为出发点,把个体差异分解为人、人与食物交互和重复噪声:示意模型中单次比较只有 0.610 是真实差异;平均差很小的两种食物,只有 0.611 的人个人排序与平均一致,平均差大时为 0.921。 (https://5dlibrary.com/answers/zh/continuous-glucose-auc-minimal-model-ketone-switch/c02/) - Ch.3 The Bergman Minimal Model: the Mathematical Structure of a Remote Compartment of Insulin Action — The minimal model links glucose and plasma insulin with two equations. Glucose effectiveness $S_G$ plus insulin action $X$ in a remote compartment make up the fractional disappearance rate of glucose, and $S_I = p_3 / p_2$ is the slope of that rate with respect to insulin. The remote compartment makes action lag insulin: in the illustrative injection, plasma insulin peaks at minute 0, action peaks at minute 38, and glucose falls to 76.3 at minute 120, below basal. The profile computation shows that one noisy curve hardly determines $p_2$; for $p_2$ between 0.015 and 0.060 the residuals all sit near the noise floor of 3.98. $S_I$ is comparatively stable, while $S_G$ and $p_2$ trade off. The next chapter replaces the injection by a meal and asks a question closer to everyday readings: of a person's several parameters, which one moves the post-meal curve most? (https://5dlibrary.com/answers/en/continuous-glucose-auc-minimal-model-ketone-switch/c03/) 第3章 Bergman 最小模型:远端胰岛素作用室的数学结构 — 最小模型用两个方程把血糖与血浆胰岛素连起来:葡萄糖效能 $S_G$ 加上远端室的胰岛素作用 $X$,构成葡萄糖的分数消失率;$S_I = p_3 / p_2$ 是这个消失率对胰岛素的斜率。远端室使作用滞后于胰岛素:示意注射中,血浆胰岛素在第 0 分钟最高,作用在第 38 分钟才最高,血糖在第 120 分钟降到 76.3,低于基础值。剖面计算表明,一条带噪声的曲线几乎确定不了 $p_2$,$p_2$ 在 0.015 到 0.060 之间时残差都在噪声底线 3.98 附近;$S_I$ 相对稳定,$S_G$ 与 $p_2$ 此消彼长。下一章把模型从静脉注射换成进餐,问一个更贴近日常读数的问题:在一个人的几个参数里,哪一个最能推动他的餐后曲线。 (https://5dlibrary.com/answers/zh/continuous-glucose-auc-minimal-model-ketone-switch/c03/) - Ch.4 Sensitivity to Personal Parameters: Which Parameter Moves the Post-Meal Curve — Chapter 4 finds that with glucose alone, insulin sensitivity and secretion appear only as a product, both with an elasticity of 0.531. (https://5dlibrary.com/answers/en/continuous-glucose-auc-minimal-model-ketone-switch/c04/) 第4章 个体参数敏感度:哪一个参数推动餐后曲线 — 第四章发现只测血糖时,胰岛素敏感性与分泌只以乘积出现,二者弹性同为 0.531。 (https://5dlibrary.com/answers/zh/continuous-glucose-auc-minimal-model-ketone-switch/c04/) - Ch.5 The Switch to Ketone Metabolism: a Gradient, Not a Toggle — Chapter 5 uses an illustrative dynamic model to show that the metabolic switch is gradual: ketones take 18.2 hours to rise from one tenth to nine tenths, clearance turnover keeps the width above 8.8 hours even when the molecular step is a toggle, and seven definitions of "switched" differ by 12.6 hours. (https://5dlibrary.com/answers/en/continuous-glucose-auc-minimal-model-ketone-switch/c05/) 第5章 生酮代谢切换:渐变而非开关 — 第五章用示意动力学模型说明代谢切换是渐变:酮体从一成升到九成用了 18.2 小时,即使分子层面像开关,清除周转也把宽度限制在 8.8 小时以上,七种「切换」定义相差 12.6 小时。 (https://5dlibrary.com/answers/zh/continuous-glucose-auc-minimal-model-ketone-switch/c05/) - Ch.6 The Measurement Chain: Interstitial Lag, MARD and Sensor Bias — Chapter 6 converts mean absolute relative difference into error bands: a relative standard deviation of 0.1429, or plus or minus 42.0 for one reading at 150 mg/dL. (https://5dlibrary.com/answers/en/continuous-glucose-auc-minimal-model-ketone-switch/c06/) 第6章 测量链:组织间液滞后、MARD 与传感器偏差 — 第六章把平均绝对相对差换算成误差带:相对标准差 0.1429,在 150 mg/dL 处单读数误差带为正负 42.0。 (https://5dlibrary.com/answers/zh/continuous-glucose-auc-minimal-model-ketone-switch/c06/) - Ch.7 Classical Sources East and West: the Suwen on Rich Food and Bernard on Averages — Chapter 7 examines the Suwen and Bernard's criticism of averages: the average curve and the average of individuals share an incremental area of 6461.7, yet individuals spend 37.6 minutes above 140 mg/dL on average while the average curve spends none. (https://5dlibrary.com/answers/en/continuous-glucose-auc-minimal-model-ketone-switch/c07/) 第7章 东西方原典考辨:《素问·奇病论》与贝尔纳的平均数批评 — 第七章考辨《素问·奇病论》与贝尔纳对平均数的批评:平均曲线与个人平均的增量面积同为 6461.7,个人高于 140 mg/dL 的时间平均 37.6 分钟,平均曲线却为零。 (https://5dlibrary.com/answers/zh/continuous-glucose-auc-minimal-model-ketone-switch/c07/) - Ch.8 Closing the Loop: A-B-A Logs, Randomised Blocks and Pre-Registration — Chapter 8 gives a personal log procedure, the Postprandial Curve Area Calculator and a pre-registered plan: a naive test on an A-B-A log raises false alarms at 0.206, randomised blocks of three days at 0.043, and three weeks can detect a log difference of 0.257. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/continuous-glucose-auc-minimal-model-ketone-switch/c08/) 第8章 实证闭环:A-B-A 日志、随机化区组与预注册 — 第八章给出个人日志的流程、餐后曲线面积计算器与预注册方案:A-B-A 日志的朴素检验误报比例为 0.206,三天一组的随机化区组为 0.043,三周日志可检出的最小对数差为 0.257。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/continuous-glucose-auc-minimal-model-ketone-switch/c08/) ## 5DL-B-287 · The Biphasic Dose Window of Photobiomodulation: Arrhenius Thermal Limits and Exposure Times under Pulsed Modulation 中文书名:光生物调节双相剂量窗、阿伦尼乌斯热边界与脉冲调制照射时长母典 DOI: pending Cite: Fan, Y. (2026). The Biphasic Dose Window of Photobiomodulation: Arrhenius Thermal Limits and Exposure Times under Pulsed Modulation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: It writes only physical quantities (irradiance, fluence, temperature rise, duty cycle) and dose ranges reported in published studies; all devices and parameters are illustrative and all calculations are generated by its programs. It is not medical advice and claims no effect of any kind; anyone planning a self-experiment must first consult a licensed physician. Abstract: The most common claim about red and near-infrared light exposure is that longer is better and brighter is better. This book refutes the two beliefs behind it: that dose and reading are monotonically related, and that equal joules per square centimetre make equal exposures. It writes only physical quantities (irradiance, fluence, temperature rise, duty cycle) and dose ranges reported in published studies; all devices and parameters are illustrative and all calculations are generated by its programs. It is not medical advice and claims no effect of any kind; anyone planning a self-experiment must first consult a licensed physician. Chapter 1 writes a product-form biphasic dose function with a closed-form peak. On illustrative readings built to follow the qualitative order reported in a published review, the monotone model's residual sum of squares is 4.3 times the biphasic model's, the peak lies at 0.537 J/cm², and at 30 times the peak time the modelled excess over control falls to 15.6 %. Chapter 2 converts among power, spot, irradiance and fluence and shows that reciprocity often fails: at the same 3 J/cm², different irradiances give different results, so dose needs two dimensions, irradiance and time. Chapter 3 computes tissue attenuation with the diffusion approximation; with an illustrative window ratio of 49.5, a surface kept below the upper edge serves depths only to 18.26 mm at 810 nm. Chapter 4 presents the Arrhenius damage integral: with much-quoted skin parameters as illustration, the threshold takes 24.42 hours at 44 °C and only 60.58 seconds at 54 °C, and at 41 °C it scores 3.50 times the thermal dose of CEM43. Chapter 5 joins irradiance to temperature rise with a lumped model whose time constant is 252 seconds; for an 1800-second exposure the integral boundary is 74.8 mW/cm². Chapter 6 shows that fluence counts only average power, and that two exposures of 144 J/cm² can carry thermal damage integrals 71.72 times apart when one is given in slow pulses. Chapter 7 examines the Suwen's figure of the sun for yang qi (paraphrased after the received text) and Schulz's yeast experiments of 1888; in an illustrative function the zero-equivalent point lies at 6.3 times the peak dose. Chapter 8 gives a device parameter sheet, the distilled Light Dose Converter and a pre-registered test plan for in vitro systems. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:谈到红光与近红外光照射,最常听到的说法是照得越久越好、灯越亮越好。本书证伪这句话背后的两个信念:剂量与读数单调相关;焦耳每平方厘米相同,照射就相同。书中只写物理量(辐照度、能量密度、温升、占空比)与已发表研究报告的剂量范围,所有设备与参数均为示意,全部计算由书中程序生成。本书不是医疗建议,不宣称任何功效;任何自我试验须先咨询执业医师。 第一章写出乘积型双相剂量函数,峰位有闭式解;在按已发表综述定性次序构造的示意读数上,单调模型的残差平方和是双相模型的 4.3 倍,峰值在 0.537 J/cm²,照到峰值时间的 30 倍,模型中高出对照的部分只剩 15.6 %。第二章处理功率、光斑、辐照度与能量密度的换算,并说明互易律常常失效:同为 3 J/cm²,不同辐照度给出的结果不同,剂量须用辐照度与时间两维描述。第三章用扩散近似计算组织衰减,示意剂量窗的上下沿之比为 49.5,表面不越过上沿时 810 nm 最深只能照顾到 18.26 毫米。第四章讲阿伦尼乌斯热损伤积分:以常被转引的皮肤参数作示意,44 °C 达到阈值要 24.42 小时,54 °C 只要 60.58 秒;在 41 °C,它比 CEM43 多记 3.50 倍热剂量。 第五章用时间常数 252 秒的集总模型把辐照度与温升接起来,照射 1800 秒时热损伤积分边界是 74.8 mW/cm²。第六章说明能量密度只认平均功率,同为 144 J/cm²,慢脉冲的热损伤积分可以是连续光的 71.72 倍。第七章考辨《素问·生气通天论》以日喻阳气(据通行本大意)与 Schulz 1888 年的酵母实验,示意函数的零当量点是峰值剂量的 6.3 倍。第八章给出设备参数核算表、蒸馏小工具「光剂量换算器」与离体系统的预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Biphasic Dose Window: the Mathematical Form of the Arndt–Schulz Curve — Chapter 1 writes a product-form biphasic dose function with a closed-form peak. On illustrative readings built to follow the qualitative order reported in a published review, the monotone model's residual sum of squares is 4.3 times the biphasic model's, the peak lies at 0.537 J/cm², and at 30 times the peak time the modelled excess over control falls to 15.6 %. (https://5dlibrary.com/answers/en/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c01/) 第1章 双相剂量窗:Arndt–Schulz 曲线的数学形式 — 第一章写出乘积型双相剂量函数,峰位有闭式解;在按已发表综述定性次序构造的示意读数上,单调模型的残差平方和是双相模型的 4.3 倍,峰值在 0.537 J/cm²,照到峰值时间的 30 倍,模型中高出对照的部分只剩 15.6 %。 (https://5dlibrary.com/answers/zh/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c01/) - Ch.2 Irradiance Times Time: Converting Fluence, and the Failure of Reciprocity — Chapter 2 converts among power, spot, irradiance and fluence and shows that reciprocity often fails: at the same 3 J/cm², different irradiances give different results, so dose needs two dimensions, irradiance and time. (https://5dlibrary.com/answers/en/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c02/) 第2章 辐照度乘时间:能量密度的换算与互易律的失效 — 第二章处理功率、光斑、辐照度与能量密度的换算,并说明互易律常常失效:同为 3 J/cm²,不同辐照度给出的结果不同,剂量须用辐照度与时间两维描述。 (https://5dlibrary.com/answers/zh/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c02/) - Ch.3 Optical Attenuation in Tissue: the Beer–Lambert Law and the Diffusion Approximation — Chapter 3 computes tissue attenuation with the diffusion approximation; with an illustrative window ratio of 49.5, a surface kept below the upper edge serves depths only to 18.26 mm at 810 nm. (https://5dlibrary.com/answers/en/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c03/) 第3章 组织光学衰减:Beer–Lambert 定律与扩散近似 — 第三章用扩散近似计算组织衰减,示意剂量窗的上下沿之比为 49.5,表面不越过上沿时 810 nm 最深只能照顾到 18.26 毫米。 (https://5dlibrary.com/answers/zh/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c03/) - Ch.4 The Arrhenius Damage Integral: the Mathematics of a Temperature Boundary — Chapter 4 presents the Arrhenius damage integral: with much-quoted skin parameters as illustration, the threshold takes 24.42 hours at 44 °C and only 60.58 seconds at 54 °C, and at 41 °C it scores 3.50 times the thermal dose of CEM43. (https://5dlibrary.com/answers/en/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c04/) 第4章 阿伦尼乌斯热损伤积分:温度边界的数学 — 第四章讲阿伦尼乌斯热损伤积分:以常被转引的皮肤参数作示意,44 °C 达到阈值要 24.42 小时,54 °C 只要 60.58 秒;在 41 °C,它比 CEM43 多记 3.50 倍热剂量。 (https://5dlibrary.com/answers/zh/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c04/) - Ch.5 From Irradiance to Temperature Rise: the Thermal Boundary in the Irradiance and Time Plane — Chapter 5 joins irradiance to temperature rise with a lumped model whose time constant is 252 seconds; for an 1800-second exposure the integral boundary is 74.8 mW/cm². (https://5dlibrary.com/answers/en/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c05/) 第5章 从辐照度到温升:辐照度与时间平面上的热边界 — 第五章用时间常数 252 秒的集总模型把辐照度与温升接起来,照射 1800 秒时热损伤积分边界是 74.8 mW/cm²。 (https://5dlibrary.com/answers/zh/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c05/) - Ch.6 Pulsed Modulation: Duty Cycle, Average Power and Temperature Swings — Chapter 6 shows that fluence counts only average power, and that two exposures of 144 J/cm² can carry thermal damage integrals 71.72 times apart when one is given in slow pulses. (https://5dlibrary.com/answers/en/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c06/) 第6章 脉冲调制:占空比、平均功率与温度起伏 — 第六章说明能量密度只认平均功率,同为 144 J/cm²,慢脉冲的热损伤积分可以是连续光的 71.72 倍。 (https://5dlibrary.com/answers/zh/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c06/) - Ch.7 East and West in the Sources: "Yang Qi Is Like the Sky and the Sun" and Schulz's Yeast — Chapter 7 examines the Suwen's figure of the sun for yang qi (paraphrased after the received text) and Schulz's yeast experiments of 1888; in an illustrative function the zero-equivalent point lies at 6.3 times the peak dose. (https://5dlibrary.com/answers/en/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c07/) 第7章 东西方原典考辨:「阳气者若天与日」与 Schulz 的酵母 — 第七章考辨《素问·生气通天论》以日喻阳气(据通行本大意)与 Schulz 1888 年的酵母实验,示意函数的零当量点是峰值剂量的 6.3 倍。 (https://5dlibrary.com/answers/zh/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c07/) - Ch.8 Closing the Loop: a Device Parameter Sheet, the Light Dose Converter and a Pre-registered Test — Chapter 8 gives a device parameter sheet, the distilled Light Dose Converter and a pre-registered test plan for in vitro systems. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c08/) 第8章 实证闭环:设备参数核算表、光剂量换算器与预注册检验 — 第八章给出设备参数核算表、蒸馏小工具「光剂量换算器」与离体系统的预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/photobiomodulation-biphasic-dose-arrhenius-thermal-pulse/c08/) ## 5DL-B-288 · Room-Acoustic Standing-Wave Modes, Low-Frequency Dead Spots, Listening Dose and the Physics of Binaural Beats 中文书名:室内声学驻波简正模态、低频声场死角与双耳节律聆听疲劳母典 DOI: pending Cite: Fan, Y. (2026). Room-Acoustic Standing-Wave Modes, Low-Frequency Dead Spots, Listening Dose and the Physics of Binaural Beats (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its rooms, absorption coefficients, reverberation times, positions and listening habits are illustrative parameters; all calculations are generated by its programs. It is not medical advice, claims no effect, judges no one's hearing, and asks readers to consult a licensed physician before any self-experiment. Abstract: Two pieces of common sense guide loudspeaker placement in studies and small photo studios: a subwoofer in the middle of the room gives the most even bass, and the closer a room is to a cube the better. This book tests both with Rayleigh's modal solution and programs using only the Python standard library. Below one hundred and twenty hertz the illustrative study has only 14 normal modes, the lowest at 40.8 Hz; the middle of every dimension is a node of the odd-order modes in that dimension, and a source at the geometric centre reaches a mean squared coupling of only 0.143 of the corner value. After checking the sources the book corrects its blueprint: Bolt's paper of 1946 concerns proportions and modal spacing, while "the centre sits on nodes" follows from the mode shapes themselves; the model also shows that central placement leaves the bass curve shared by the seating area least flat while making the differences between seats smaller than corner placement does. Its rooms, absorption coefficients, reverberation times, positions and listening habits are illustrative parameters; all calculations are generated by its programs. It is not medical advice, claims no effect, judges no one's hearing, and asks readers to consult a licensed physician before any self-experiment. Chapter 1 lists every low mode and explains the nodes. Chapter 2 compares five proportions at one volume, finding a spacing spread of 7.26 Hz for the cube and 3.23 Hz for the study, and implements the Bonello criterion in its usual later statement. Chapter 3 estimates reverberation time by Sabine and Eyring and derives a modal bandwidth of 4.31 Hz and a Schroeder frequency of about 230 Hz. Chapter 4 computes the bass response at the main seat by modal summation: the standard deviation is 14.4 dB with the subwoofer at the centre of the floor and 10.5 dB in a corner, and one unit at each wall middle lowers the seat-to-seat spread to 0.7 dB at a cost of 8.8 dB in output. Chapter 5 keeps the account of long listening by the NIOSH limit of 85 dB with a 3 dB exchange rate and the WHO-ITU weekly reference: the illustrative listening day has a daily dose of 1.000, of which the loudest hour alone takes 0.500, and when room modes lift the bass by 10 dB the A-weighted reading moves by only 0.02 dB. Chapter 6 shows that a binaural beat exists only in the comparison of the two ears, that over loudspeakers the monaural beat depth at the ears varies with the seat from 0.114 to 0.819, and reports faithfully the inconsistent conclusions of an EEG review (5 of 14 studies supportive, 8 contradictory) and a meta-analysis (pooled effect size 0.45). Chapter 7 examines "sounds respond to one another, and so change arises" from the Record of Music and Chladni's sand figures. Chapter 8 gives a seven-step measurement procedure, the Room Mode Calculator tool and a pre-registered plan needing 27 rooms. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:在书房与小型摄影棚里摆音箱,常见的两条常识是低音箱放在房间正中最均匀、房间越接近立方体越好。本书用瑞利的模态解与纯标准库程序检验这两条常识:示意书房一百二十赫兹以下只有 14 个简正模态,最低在 40.8 赫兹;每一维的正中都是该维奇数阶模态的节点,放在房间几何正中的声源,耦合平方的平均值只有墙角的 0.143。核对原文后本书作了蓝图修正:Bolt 1946 年的论文讨论的是比例与模态间隔,「正中落在节点上」来自模态形状本身;模型还显示,正中摆位让座位区共享的低频曲线最不平,却让座位之间的差异小于墙角摆位。书中的房间、吸声系数、混响时间、位置与聆听习惯均为示意参数;全部计算由书中程序生成。本书非医疗建议,不宣称任何功效,不判断任何人的听力状况,自我试验先咨询执业医师。 第一章列出全部低频模态并说明节点。第二章在同一容积下比较五种比例,立方体的模态间隔标准差为 7.26 赫兹,示意书房为 3.23 赫兹,并按后来文献的通行表述实现 Bonello 判据。第三章用 Sabine 与 Eyring 公式估算混响时间,导出 4.31 赫兹的模态带宽与约 230 赫兹的 Schroeder 频率。第四章用模态叠加计算主座位低频响应,地面正中摆位的标准差为 14.4 分贝,墙角为 10.5 分贝;四面墙中点各一只低音箱把座位间标准差降到 0.7 分贝,输出低 8.8 分贝。 第五章按 NIOSH 的 85 分贝、3 分贝交换率与世卫和国际电联的每周参考剂量为长时间聆听记账:示意聆听日的日剂量为 1.000,最响的一小时单独占 0.500;房间模态把低频抬高 10 分贝时,A 计权读数只变 0.02 分贝。第六章说明双耳拍只存在于两耳比较之中,外放时耳朵处的单耳拍深度随座位从 0.114 变到 0.819,并如实转述脑电综述(14 项研究中 5 项支持、8 项矛盾)与元分析(合并效应量 0.45)的不一致结论。第七章考辨《礼记·乐记》的「声相应,故生变」与克拉尼的振动板沙画。第八章给出七步测量流程、小工具「房间模态计算器」与预注册方案(需要 27 间房间)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Normal Modes of a Rectangular Room: Why the Centre Is a Node — The bass field of a rigid rectangular room consists of a finite set of normal modes, with frequencies from Rayleigh's formula and shapes that are products of three cosines. Below one hundred and twenty hertz the illustrative study has only 14 modes, the lowest at 40.8 Hz and the widest gap 13.1 Hz; this is what "the bass of a small room is governed by a few modes" means. The middle of each dimension is a node of every odd-order mode in that dimension, so a source at the geometric centre can drive only all-even modes and reaches a mean squared coupling of only 0.143 of the corner value; twenty centimetres off the midline, the nodes fall from 9 to 6. After checking the sources, "the centre sits on nodes" belongs to the mode shapes, while Bolt (1946) concerns proportions and modal spacing. The next chapter turns to proportions: with the same volume and different ratios of length, width and height, how do modes crowd together or spread out evenly? (https://5dlibrary.com/answers/en/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c01/) 第1章 矩形房间的简正模态:正中为什么是节点 — 刚性长方体房间的低频声场由有限个简正模态组成,频率由瑞利公式给出,形状是三个余弦的乘积。示意书房一百二十赫兹以下只有 14 个模态,最低的在 40.8 赫兹,最大空档 13.1 赫兹,这就是「小房间低频由少数模态主导」的含义。每一维的中点是该维所有奇数阶模态的节点,所以放在房间几何正中的声源只能推动全偶数阶模态,耦合平方的平均值只有墙角的 0.143;离开中线二十厘米,节点就从 9 个降到 6 个。核对原文后,「正中落在节点上」应归于模态形状,Bolt 1946 讨论的是比例与模态间隔。下一章转向比例:同样的容积,不同的长宽高之比,模态会怎样挤在一起或均匀散开。 (https://5dlibrary.com/answers/zh/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c01/) - Ch.2 Proportions and Coincidence: Bolt's Spacing Statistics and the Bonello Criterion — Chapter 2 compares five proportions at one volume, finding a spacing spread of 7.26 Hz for the cube and 3.23 Hz for the study, and implements the Bonello criterion in its usual later statement. (https://5dlibrary.com/answers/en/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c02/) 第2章 尺寸比与简并:Bolt 的间隔统计与 Bonello 判据 — 第二章在同一容积下比较五种比例,立方体的模态间隔标准差为 7.26 赫兹,示意书房为 3.23 赫兹,并按后来文献的通行表述实现 Bonello 判据。 (https://5dlibrary.com/answers/zh/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c02/) - Ch.3 Reverberation Time and the Schroeder Frequency: the Upper Edge of the Modal Region — Chapter 3 estimates reverberation time by Sabine and Eyring and derives a modal bandwidth of 4.31 Hz and a Schroeder frequency of about 230 Hz. (https://5dlibrary.com/answers/en/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c03/) 第3章 混响时间与 Schroeder 频率:模态区的上界 — 第三章用 Sabine 与 Eyring 公式估算混响时间,导出 4.31 赫兹的模态带宽与约 230 赫兹的 Schroeder 频率。 (https://5dlibrary.com/answers/zh/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c03/) - Ch.4 Low-Frequency Dead Spots: Modal Sums, Seat Maps and Several Subwoofers — Chapter 4 computes the bass response at the main seat by modal summation: the standard deviation is 14.4 dB with the subwoofer at the centre of the floor and 10.5 dB in a corner, and one unit at each wall middle lowers the seat-to-seat spread to 0.7 dB at a cost of 8.8 dB in output. (https://5dlibrary.com/answers/en/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c04/) 第4章 低频声场死角:模态叠加、座位地图与多只低音箱 — 第四章用模态叠加计算主座位低频响应,地面正中摆位的标准差为 14.4 分贝,墙角为 10.5 分贝;四面墙中点各一只低音箱把座位间标准差降到 0.7 分贝,输出低 8.8 分贝。 (https://5dlibrary.com/answers/zh/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c04/) - Ch.5 Sound Exposure Dose: 85 dBA, the 3 dB Exchange Rate and the Blind Spot of A-Weighting — Chapter 5 keeps the account of long listening by the NIOSH limit of 85 dB with a 3 dB exchange rate and the WHO-ITU weekly reference: the illustrative listening day has a daily dose of 1.000, of which the loudest hour alone takes 0.500, and when room modes lift the bass by 10 dB the A-weighted reading moves by only 0.02 dB. (https://5dlibrary.com/answers/en/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c05/) 第5章 声暴露剂量:85 dBA、3 dB 交换率与 A 计权的盲区 — 第五章按 NIOSH 的 85 分贝、3 分贝交换率与世卫和国际电联的每周参考剂量为长时间聆听记账:示意聆听日的日剂量为 1.000,最响的一小时单独占 0.500;房间模态把低频抬高 10 分贝时,A 计权读数只变 0.02 分贝。 (https://5dlibrary.com/answers/zh/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c05/) - Ch.6 Binaural Rhythms: the Physics of Headphones versus Loudspeakers, and the Published Evidence — Chapter 6 shows that a binaural beat exists only in the comparison of the two ears, that over loudspeakers the monaural beat depth at the ears varies with the seat from 0.114 to 0.819, and reports faithfully the inconsistent conclusions of an EEG review (5 of 14 studies supportive, 8 contradictory) and a meta-analysis (pooled effect size 0.45). (https://5dlibrary.com/answers/en/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c06/) 第6章 双耳节律:耳机与外放的物理差别和已发表证据 — 第六章说明双耳拍只存在于两耳比较之中,外放时耳朵处的单耳拍深度随座位从 0.114 变到 0.819,并如实转述脑电综述(14 项研究中 5 项支持、8 项矛盾)与元分析(合并效应量 0.45)的不一致结论。 (https://5dlibrary.com/answers/zh/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c06/) - Ch.7 Classical Sources East and West: "Sounds Respond" in the Record of Music and Chladni's Sand Figures — "Sounds respond to one another, and so change arises" can be rewritten as two testable propositions. A resonator responds noticeably only near one bandwidth of its frequency: the study's 40.8 Hz mode loses 2.7 dB when driven 2.0 Hz off and 13.5 dB at 10.0 Hz off. Its response takes time to build up, 0.164 s in the study and 0.360 s in the studio, set by the reverberation time. Chladni's sand figures of 1787 turned the nodal lines of vibrating plates into visible pictures, and the diagonal lines of square plates come from degeneracy: a square membrane has only 34 distinct frequencies for 64 modes, a 1:1.2 rectangle has 64, and on the diagonal the difference combination has an amplitude of 0.000. The old sources saw the phenomena rightly; modern theory supplied the quantities and the reasons. The last chapter gathers the whole book into a measurement procedure that can be carried out in a photo studio or a study, and writes down a pre-registered test: how many rooms are needed to decide whether central floor placement really is worse? (https://5dlibrary.com/answers/en/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c07/) 第7章 东西方原典考辨:《乐记》的「声相应」与克拉尼的沙画 — 《乐记》「声相应,故生变」可以改写成两个可检验的命题:共振者只对一个带宽附近的声音明显响应,书房 40.8 赫兹模态偏离 2.0 赫兹降 2.7 分贝、偏离 10.0 赫兹降 13.5 分贝;响应需要时间建立,书房 0.164 秒、摄影棚 0.360 秒,由混响时间决定。克拉尼 1787 年的沙画把振动板的节线变成可见图形,方板对角线节线来自简并:正方形膜 64 个模态只有 34 个不同频率,长方形 1:1.2 则有 64 个,对角线上相减组合的振幅为 0.000。原典看对了现象,现代理论补上了数量与原因。最后一章把全书收拢成一套可以在摄影棚与书房里执行的测量流程,并写下预注册的检验方案:要多少间房间,才能判断地面正中摆位是否真的更差? (https://5dlibrary.com/answers/zh/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c07/) - Ch.8 Closing the Loop: Measuring Bass Modes and Placing Loudspeakers in Studios and Studies, with a Pre-Registered Test — Chapter 8 gives a seven-step measurement procedure, the Room Mode Calculator tool and a pre-registered plan needing 27 rooms. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c08/) 第8章 实证闭环:摄影棚与书房的低频模态测量、音箱摆位与预注册检验 — 第八章给出七步测量流程、小工具「房间模态计算器」与预注册方案(需要 27 间房间)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/room-acoustic-standing-wave-modes-low-frequency-dead-spots/c08/) ## 5DL-B-289 · Fickian Release Kinetics of Herbal Polysaccharides: Extraction Curves for Dendrobium and Isobologram Analysis of Combinations 中文书名:草本活性多糖菲克扩散溶出动力学、道地石斛提取曲线与复方等效线图协同母典 DOI: pending Cite: Fan, Y. (2026). Fickian Release Kinetics of Herbal Polysaccharides: Extraction Curves for Dendrobium and Isobologram Analysis of Combinations (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: The book is not medical advice and claims no effect of any kind; it discusses only physical and chemical quantities such as yields, concentrations, released fractions and diffusion coefficients. Its parameters are illustrative and all calculations are generated by its programs; anyone planning a self-experiment should first consult a licensed physician. Abstract: People who boil plant material often believe that the longer it boils, the more comes out. Using the series solutions of Fick's diffusion equation for spheres and plane sheets, the Arrhenius relation and a convolution with chain scission in the liquid, this book refutes that linear picture and restates it in three sentences: the marginal return to time falls, particle size acts squared, and temperature cuts both ways. The book is not medical advice and claims no effect of any kind; it discusses only physical and chemical quantities such as yields, concentrations, released fractions and diffusion coefficients. Its parameters are illustrative and all calculations are generated by its programs; anyone planning a self-experiment should first consult a licensed physician. Chapter 1 shows that with illustrative parameters, going from 30 to 60 minutes multiplies the yield by 1.411 and going from 60 to 120 minutes by only 1.169; at a baseline of 90 degrees, radius 1.5 mm and 30 minutes, doubling the time raises the released fraction by 0.154, ten degrees more by only 0.056, and halving the radius by 0.317. Chapter 2 develops the square law of particle size, slices and sieve mixtures. Chapter 3 shows that chain scission in the liquid has a higher activation energy than diffusion, so the hotter the liquid, the earlier and lower the peak of the measured amount. Chapter 4 compares Noyes–Whitney, Weibull and the diffusion series, showing that nine time points can rule out first-order kinetics but often cannot tell the diffusion solution from Weibull. Chapter 5 draws the yield surface, whose best illustrative cell is 80 degrees, 0.5 mm and 60 minutes at 0.1456 g per gram, and treats daodi, the genuine growing region, only as a variance-components question about differences in content between origins, with no claim about quality. Chapter 6 gives the mathematics of isobolograms for combinations: synergy must be judged against the isobole of Loewe additivity, not by adding readings; in the illustrative in-vitro design, 2 replicate wells per dose make 0.200 of additive experiments look synergistic; and an in-vitro additivity test is not an effect in a person. Chapter 7 examines the decoction rules collected in the preface volume of the Bencao Gangmu and Fick's paper of 1855. Chapter 8 gives the extraction record sheet, the Release Curve Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:煎煮草本材料的人常相信「煮得越久,出来得越多」。本书用 Fick 扩散方程在球形颗粒与平板上的级数解、Arrhenius 温度关系与液相断链损失的卷积,证伪这条线性想象,并把它修正为三句话:时间的边际回报递减,粒径以平方起作用,温度是一把双刃。本书非医疗建议,不宣称任何功效,只讨论得率、浓度、溶出分数与扩散系数这些物理化学量;书中参数为示意,全部计算由书中程序生成;自我试验前应先咨询执业医师。 第一章显示,示意参数下 30 分钟延长到 60 分钟得率乘以 1.411,60 分钟延长到 120 分钟只乘以 1.169;在 90 度、半径 1.5 毫米、30 分钟的基线上,时间加倍使溶出分数增加 0.154,升温十度只增加 0.056,半径减半增加 0.317。第二章展开粒径的平方律、切片与筛分混合物;第三章说明液相断链的活化能高于扩散,温度越高,测得量的峰值越早也越低;第四章比较 Noyes–Whitney、Weibull 与扩散级数解,说明九个时间点能排除一级动力学,却常分不清扩散解与 Weibull;第五章画出得率曲面,示意网格中最高的一格是 80 度、0.5 毫米、60 分钟,每克 0.1456 克,并把「道地」只当作产地与含量差异的方差分量问题,不作品质宣称。 第六章给出复方等效线图的数学:协同须相对于 Loewe 加和性的等效线判定,而不是把读数相加;在示意的体外设计里,每剂量 2 个重复孔会让 0.200 的加和实验被误判为协同;体外加和性检验不等于人体效果。第七章考辨《本草纲目·序例》收录的煮汤法与 Fick 1855 年的论文。第八章给出提取工艺理化记录表、溶出曲线计算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 First-Order Kinetics and the Sphere Series: Why Boiling Time Is Not Linear — Chapter 1 shows that with illustrative parameters, going from 30 to 60 minutes multiplies the yield by 1.411 and going from 60 to 120 minutes by only 1.169; at a baseline of 90 degrees, radius 1.5 mm and 30 minutes, doubling the time raises the released fraction by 0.154, ten degrees more by only 0.056, and halving the radius by 0.317. (https://5dlibrary.com/answers/en/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c01/) 第1章 一级动力学与球形级数:煎煮时间为什么不线性 — 第一章显示,示意参数下 30 分钟延长到 60 分钟得率乘以 1.411,60 分钟延长到 120 分钟只乘以 1.169;在 90 度、半径 1.5 毫米、30 分钟的基线上,时间加倍使溶出分数增加 0.154,升温十度只增加 0.056,半径减半增加 0.317。 (https://5dlibrary.com/answers/zh/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c01/) - Ch.2 The Square Law of Particle Size: Radius, Shape and Sieve Mixtures — The time diffusion needs is proportional to the square of a characteristic length. For a spherical particle the dimensionless times to release a half and nine tenths are 0.03055 and 0.18299, and the time for any radius follows from $\tau a^2/D$: radius 1.00 mm needs 10.2 and 61.0 minutes, radius 4.00 mm needs 162.9 and 975.9 minutes. A large particle starts by the square-root law, so four times the time gives only 1.944 times the release. A sieve mixture is not a single first-order process: its apparent rate constant falls from 0.0738 to 0.0085, and at minute 120 the polysaccharide still in the particles sits in the two coarsest fractions, 0.558 and 0.389 of it respectively. A slice is far slower than a sphere with the same characteristic length, and equal specific surface guarantees only a similar start. After particle size comes temperature. If temperature, acting through the diffusion coefficient, is worth only about a third more time, what role does it play in a boil? Chapter 3 answers, and shows why hotter sometimes yields less. (https://5dlibrary.com/answers/en/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c02/) 第2章 粒径的平方律:半径、形状与筛分混合物 — 扩散所需的时间与特征长度的平方成正比。对球形颗粒,释放一半与九成的无量纲时间分别是 0.03055 与 0.18299,任何半径下的时间都可以按 $\tau a^2/D$ 换算:半径 1.00 毫米需 10.2 与 61.0 分钟,半径 4.00 毫米需 162.9 与 975.9 分钟。大颗粒的开头遵循平方根律,时间乘以四,释放只乘以 1.944。筛分混合物不是一个一级过程:表观速率常数从 0.0738 降到 0.0085,第 120 分钟剩余的多糖有 0.558 与 0.389 分别留在最粗的两档里。切片比同样特征长度的球慢得多,比表面积相同也只保证开头相似。粒径之后是温度:既然温度通过扩散系数只值三成的时间,它在煎煮里到底扮演什么角色?第三章给出答案,并说明为什么更热有时得到更少。 (https://5dlibrary.com/answers/zh/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c02/) - Ch.3 Arrhenius Temperature: Diffusion Racing Chain Scission in the Liquid — Temperature acts through two races. The diffusion coefficient rises with temperature by the Arrhenius relation: with the illustrative activation energy it is multiplied by 1.305 from 90 to 100 degrees, more sensitive than the 1.147 inferred from the viscosity of water, and the diffusion progress of 60 minutes at 100 degrees takes 103.8 minutes at 80 degrees. Scission in the liquid has a higher activation energy, so the hotter the liquid, the earlier and lower the peak of the measured amount: under the first-order model, the peak at 80 degrees comes at 226.5 minutes with 0.137, and at 120 degrees at 43.1 minutes with 0.103. After a full 240 minutes, 0.137 remains at 80 degrees and only 0.021 at 120. We have now used three curves: first-order kinetics, the sphere series and a convolution with loss. Given nine measured time points in a laboratory, which one should be fitted, and can they be told apart at all? Chapter 4 takes up the question. (https://5dlibrary.com/answers/en/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c03/) 第3章 Arrhenius 温度:扩散系数与液相断链的赛跑 — 温度通过两条赛道起作用。扩散系数按 Arrhenius 关系随温度上升:以示意活化能计,从 90 度到 100 度乘以 1.305,比按水的黏度推算的 1.147 更敏感;100 度煮 60 分钟的扩散进度,在 80 度要煮 103.8 分钟。液相断链的活化能更高,于是温度越高,测得量的峰值来得越早、越低:一级模型下,80 度的峰值是 226.5 分钟、0.137,120 度是 43.1 分钟、0.103。煮满 240 分钟时,80 度还剩 0.137,120 度只剩 0.021。到这里,我们已经用了三种曲线:一级动力学、球形级数,以及带损失的卷积。实验室里拿到九个时间点的测量值,究竟该用哪一种去拟合,又能不能分辨它们?第四章讨论这个问题。 (https://5dlibrary.com/answers/zh/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c03/) - Ch.4 A Family of Release Models: Noyes–Whitney, Weibull and What Nine Time Points Can Tell Apart — The Noyes–Whitney equation describes a solid dissolving from its surface, which is not the same as polysaccharide diffusing out of a particle; it fits the redissolving of extract powders. A shrinking surface slows the tail, a dose above the saturation capacity stops at 0.667, and near saturation longer stirring is useless. The Weibull shape exponent is scale-free and carries information about geometry, but it also drifts with the particle distribution and the sampling schedule: the same sphere moves from 0.641 to 0.577 when the schedule changes. Nine time points suffice to rule out first-order kinetics, yet in about a quarter of simulations they cannot tell the diffusion solution from Weibull. The similarity factor is lenient: a radius one fifth larger still scores 57.3. With the model chosen, the next question is to draw all three levers at once, to see the whole landscape of process parameters, and to write "genuine growing region" as a statistical question that can be tested. That is the content of Chapter 5. (https://5dlibrary.com/answers/en/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c04/) 第4章 溶出模型族:Noyes–Whitney、Weibull 与九个时间点能分辨什么 — Noyes–Whitney 方程描述的是固体从表面溶掉,与多糖从颗粒内部扩散出来不是一回事;它适用于提取物粉末的复溶:表面积缩小拖慢尾巴,剂量超过饱和容量时停在 0.667,近饱和时延长搅拌无用。Weibull 形状指数与尺度无关,携带几何信息,但也随颗粒分布与取样表漂移:同一个球,换一张取样表就从 0.641 变成 0.577。九个时间点足以排除一级动力学,却在大约四分之一的模拟里分不清扩散解与 Weibull。相似因子宽松,半径变大两成仍得 57.3。模型选好之后,下一个问题是把三根杠杆同时画出来,看工艺参数的整张地形图,并把「道地」写成一个可以检验的统计问题。这是第五章的内容。 (https://5dlibrary.com/answers/zh/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c04/) - Ch.5 The Yield Surface: Temperature, Size, Time, Liquid Ratio and the Variance Components of "Daodi" — Chapter 5 draws the yield surface, whose best illustrative cell is 80 degrees, 0.5 mm and 60 minutes at 0.1456 g per gram, and treats daodi, the genuine growing region, only as a variance-components question about differences in content between origins, with no claim about quality. (https://5dlibrary.com/answers/en/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c05/) 第5章 提取得率曲面:温度、粒径、时间、液固比与「道地」的方差分量 — 第五章画出得率曲面,示意网格中最高的一格是 80 度、0.5 毫米、60 分钟,每克 0.1456 克,并把「道地」只当作产地与含量差异的方差分量问题,不作品质宣称。 (https://5dlibrary.com/answers/zh/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c05/) - Ch.6 Isobolograms for Combinations: Loewe Additivity, Sham Combinations and the Mathematical Definition of Synergy — Chapter 6 gives the mathematics of isobolograms for combinations: synergy must be judged against the isobole of Loewe additivity, not by adding readings; in the illustrative in-vitro design, 2 replicate wells per dose make 0.200 of additive experiments look synergistic; and an in-vitro additivity test is not an effect in a person. (https://5dlibrary.com/answers/en/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c06/) 第6章 复方等效线图:Loewe 加和性、假组合与协同的数学定义 — 第六章给出复方等效线图的数学:协同须相对于 Loewe 加和性的等效线判定,而不是把读数相加;在示意的体外设计里,每剂量 2 个重复孔会让 0.200 的加和实验被误判为协同;体外加和性检验不等于人体效果。 (https://5dlibrary.com/answers/zh/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c06/) - Ch.7 East and West in the Sources: Decoction Rules in the Preface Volume of the Bencao Gangmu and Fick's Cylinder and Funnel — Chapter 7 examines the decoction rules collected in the preface volume of the Bencao Gangmu and Fick's paper of 1855. (https://5dlibrary.com/answers/en/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c07/) 第7章 东西方原典考辨:《本草纲目·序例》的煮汤法与 Fick 的圆柱和漏斗 — 第七章考辨《本草纲目·序例》收录的煮汤法与 Fick 1855 年的论文。 (https://5dlibrary.com/answers/zh/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c07/) - Ch.8 Closing the Loop: the Extraction Record Sheet, the Release Curve Calculator and a Pre-Registered Test — Chapter 8 gives the extraction record sheet, the Release Curve Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c08/) 第8章 实证闭环:提取工艺理化记录表、溶出曲线计算器与预注册检验 — 第八章给出提取工艺理化记录表、溶出曲线计算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/herbal-polysaccharide-fickian-release-dendrobium-isobologram/c08/) ## 5DL-B-290 · Tail-Risk Option Moats and Quantile Hedging: Jump-Diffusion Models and Fixed-Premium Protection of Core Cash Flows under Kelly Bounds 中文书名:尾部黑天鹅期权护城河与分位数对冲、跳扩散模型与凯利边界下的固定保费现金流保护母典 DOI: pending Cite: Fan, Y. (2026). Tail-Risk Option Moats and Quantile Hedging: Jump-Diffusion Models and Fixed-Premium Protection of Core Cash Flows under Kelly Bounds (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Operators often judge tail protection as an investment: some say it is a moat that pays over the long run, others that it costs money every year and is pure waste. This book overturns the premise both share. Describing a revenue index with Merton's jump-diffusion model and letting premium quotes charge an illustrative premium for jump risk, fixed-premium protection struck at 0.80 costs 1.669 times its average payout and returns -0.401 on average to the buyer; the more exclusively protection covers extremes, the dearer it is relative to its average payout. The drag is invisible in one ledger: a thirty-year record still has a chance of 0.166 of ending without a loss. Its parameters are illustrative; all calculations are generated by its programs; the book is not individual legal, tax or financial advice, recommends no particular financial product, insurance product, institution or trading strategy, promises no return and encourages no leverage. Chapters 2 and 3 show how jumps create a volatility skew and heavy tails: the probability of ending below 0.4 is 173.4 times that of a lognormal with the same variance; expected shortfall at 97.5 % of a year's operating capital is 0.3373, the value-at-risk of stores cannot be added up, and short records systematically understate the tail. Chapter 4 shows that quantile hedging aimed at the success probability abandons the deep tail, so that with the same budget the probability of capital below 0.6 rises from 0.00415 to 0.00480. Chapter 5 uses the Kelly growth rate to bound the premium: when protection is worth buying it raises growth by at most 2.87 basis points a year, its value lies mainly in the path, and the share of twenty-year histories with a fall of more than forty percent drops from 0.2579 to 0.1264; with too much exposure the bound becomes a floor on the cover ratio, to be read alongside Samuelson's criticism. Chapter 6 compares strikes, payout caps, cash reserves and the posterior of the jump intensity; capped protection near the money is the only plan that raises growth. Chapter 7 examines "think of danger in peace; with preparation there is no peril" from the Zuo Zhuan and Aristotle's account of Thales and the olive presses. Chapter 8 gives an annual tail-protection budget sheet, the Tail-Protection Premium Calculator and a pre-registered test plan: backtesting the tail model would need a single record of 1,171 years, so it can be tested only on pooled samples. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:经营者常把尾部保护当作一种投资来评价:有人说它像护城河,长期划算;有人说它年年交钱,纯属浪费。本书推翻两者共有的前提。用默顿跳扩散模型描述营收指数,并让保费报价对跳跃风险收取示意的溢价,行权 0.80 的固定保费保护,保费是平均赔付的 1.669 倍,买方平均收益为 -0.401;越只保极端情形,保费相对平均赔付越贵。这个拖累在一个人的账本里看不见:三十年的记录有 0.166 的机会最后没亏。书中参数为示意;全部计算由书中程序生成;本书是非个人化法律、税务或理财建议,不推荐任何具体金融产品、保险产品、机构或交易策略,不承诺收益,不鼓励杠杆。 第二、三章说明跳跃怎样制造波动率偏斜与厚尾:跌破 0.4 的概率是同方差对数正态的 173.4 倍;经营资本一年的 97.5 % 期望短缺为 0.3373,各分店的在险价值不能相加,短记录系统性低估尾部。第四章说明以成功概率为目标的分位数对冲会放弃深尾,同一预算下资本低于 0.6 的概率反而从 0.00415 升到 0.00480。第五章用凯利增长率给保费定出边界:值得买时增长率最多每年提高 2.87 个基点,保护的价值主要在路径,二十年里回落超过四成的比例从 0.2579 降到 0.1264;敞口过大时边界变成覆盖比例的下限,并须与萨缪尔森的批评并陈。 第六章比较行权水平、赔付上限、现金储备与跳跃强度后验,封顶的接近平值保护是唯一提高增长率的方案。第七章考辨《左传》的「居安思危,有备无患」与亚里士多德笔下泰勒斯的榨油机。第八章给出年度尾部保护预算表、小工具「尾部保护保费计算器」与预注册检验方案:尾部模型的回测需要 1,171 年的单一记录,所以只能在汇总样本上检验。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Premium Drag: Why Buying Tail Protection Year after Year Loses on Average — Writing tail protection as a fixed-premium contract with payoff $(K - S)^+$ and describing the index with Merton's jump-diffusion model, program p11 shows that the ratio of premium to average payout rises from 1.241 at the money to 2.752 at a strike of 0.50, while the buyer's average return falls from -0.194 to -0.637; raising the pricing jump rate from 1.0 to 4.0 times the real one takes the buyer's average return on cover struck at 0.70 from -0.088 to -0.714. Program p12 shows that although the drag is real, an operator with a 30-year record still has a chance of 0.166 of ending without a loss and a chance of 0.025 of never seeing a claim, and only 0.369 of such operators could prove the drag statistically. Bondarenko's overpriced puts puzzle and the warning of Broadie and his co-authors about sampling error both hold inside this model. The next chapter opens Merton's model itself, to see how jumps change prices, the smile and the tail, and why the same mean and variance can hide entirely different tails. (https://5dlibrary.com/answers/en/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c01/) 第1章 保费拖累:长期买尾部保护为什么平均是亏的 — 把尾部保护写成支付 $(K - S)^+$ 的固定保费合约,并用默顿跳扩散模型描述指数,程序 p11 显示:保费与平均赔付之比从平值的 1.241 升到行权 0.50 的 2.752,买方平均收益率从 -0.194 降到 -0.637;把定价世界的跳跃强度从真实值的 1.0 倍提高到 4.0 倍,行权 0.70 的保护的买方平均收益率从 -0.088 降到 -0.714。程序 p12 显示,拖累虽然存在,但一个经营者 30 年的记录里,有 0.166 的机会最后没亏,有 0.025 的机会一次赔付都没见过,能在统计上证明拖累的只有 0.369。邦达连科的「看跌期权定价过高之谜」与布罗迪等人对抽样误差的提醒,在这个模型里同时成立。下一章打开默顿模型本身,看跳跃怎样改变价格、微笑与尾部,以及为什么同样的均值与方差可以藏着完全不同的尾巴。 (https://5dlibrary.com/answers/zh/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c01/) - Ch.2 Jump-Diffusion: Prices, Smiles and Hidden Tails in Merton's Model — Chapters 2 and 3 show how jumps create a volatility skew and heavy tails: the probability of ending below 0.4 is 173.4 times that of a lognormal with the same variance; expected shortfall at 97.5 % of a year's operating capital is 0.3373, the value-at-risk of stores cannot be added up, and short records systematically understate the tail. (https://5dlibrary.com/answers/en/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c02/) 第2章 跳扩散:默顿模型里的价格、微笑与看不见的尾巴 — 第二、三章说明跳跃怎样制造波动率偏斜与厚尾:跌破 0.4 的概率是同方差对数正态的 173.4 倍;经营资本一年的 97.5 % 期望短缺为 0.3373,各分店的在险价值不能相加,短记录系统性低估尾部。 (https://5dlibrary.com/answers/zh/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c02/) - Ch.3 Quantiles and Expected Shortfall: Measuring a Year's Tail for Operating Capital — Once the index is replaced by the operator's capital $M = 1 + A + L(S - 1)$, program p31 shows that under Merton's model expected shortfall at 97.5 % is 0.3373, much deeper than the 0.2588 of a lognormal with the same mean and variance, and that at 99.5 % it reaches 0.4493. In a normal world, value-at-risk at 99 % and expected shortfall at 97.5 % are almost equal at 2.3263 and 2.3378 standard deviations. Program p32 shows two stores each with a 95 % value-at-risk of -2.00 whose combined value-at-risk is 98.00, violating subadditivity, while their combined expected shortfall of 101.26 stays below the sum of 159.20. Program p33 shows that expected shortfall estimated from 20-year records averages only 0.2266 and understates by more than a quarter with a chance of 0.601. The next chapter answers the question that follows: if the tail is hard to estimate and protection is priced with a risk premium, which part of the tail should a limited budget protect first? (https://5dlibrary.com/answers/en/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c03/) 第3章 分位数与期望短缺:怎样量度经营资本的一年尾部 — 把指数换成经营者的资本 $M = 1 + A + L(S - 1)$ 之后,程序 p31 显示:在默顿模型下,97.5 % 的期望短缺是 0.3373,比同均值同方差的对数正态的 0.2588 深得多;在 99.5 % 的水平上,期望短缺是 0.4493。在正态世界里,99 % 的在险价值 2.3263 与 97.5 % 的期望短缺 2.3378 个标准差几乎相等。程序 p32 显示两家各自 95 % 在险价值为 -2.00 的分店,合起来的在险价值是 98.00,违反次可加性;期望短缺合起来是 101.26,小于分开之和 159.20。程序 p33 显示 20 年记录估出的期望短缺平均只有 0.2266,有 0.601 的机会低估超过四分之一。下一章回答一个随之而来的问题:既然尾部难以估准、保护又按风险溢价收费,有限的预算应该先保住尾巴的哪一段? (https://5dlibrary.com/answers/zh/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c03/) - Ch.4 Quantile Hedging: Which Part of the Tail to Protect First on a Limited Budget — Chapter 4 shows that quantile hedging aimed at the success probability abandons the deep tail, so that with the same budget the probability of capital below 0.6 rises from 0.00415 to 0.00480. (https://5dlibrary.com/answers/en/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c04/) 第4章 分位数对冲:预算有限时,先保住尾巴的哪一段 — 第四章说明以成功概率为目标的分位数对冲会放弃深尾,同一预算下资本低于 0.6 的概率反而从 0.00415 升到 0.00480。 (https://5dlibrary.com/answers/zh/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c04/) - Ch.5 The Kelly Bound: Upper and Lower Limits that Long-Run Growth Sets on the Premium — Chapter 5 uses the Kelly growth rate to bound the premium: when protection is worth buying it raises growth by at most 2.87 basis points a year, its value lies mainly in the path, and the share of twenty-year histories with a fall of more than forty percent drops from 0.2579 to 0.1264; with too much exposure the bound becomes a floor on the cover ratio, to be read alongside Samuelson's criticism. (https://5dlibrary.com/answers/en/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c05/) 第5章 凯利边界:长期增长率给保费定的上限与下限 — 第五章用凯利增长率给保费定出边界:值得买时增长率最多每年提高 2.87 个基点,保护的价值主要在路径,二十年里回落超过四成的比例从 0.2579 降到 0.1264;敞口过大时边界变成覆盖比例的下限,并须与萨缪尔森的批评并陈。 (https://5dlibrary.com/answers/zh/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c05/) - Ch.6 How to Spend a Fixed Premium: Strikes, Payout Caps, Cash Reserves and Parameter Uncertainty — Program p61 shows that, spending the same 0.006 of capital, protection struck at 0.70 lowers expected shortfall to 0.2396 and the probability of capital below 0.7 to 0.00058, but lowers growth to 0.13936; capped protection from 0.90 to 0.60 lowers expected shortfall only to 0.2957, yet raises growth to 0.14033, the only plan above no protection. Program p62 shows that to push expected shortfall down to 0.25 a cash reserve gives up 0.0370 a year and protection struck at 0.90 only 0.0033, the break-even point being a business return 0.0136 above the cash rate, with basis risk and liquidity left out. Program p63 shows that operators who saw 0 and 2 jumps in 20 years reach opposite conclusions about the same quote, and that even with a 100-year record the interval of bounds from 0.01517 to 0.02524 still contains the quote. The next chapter turns to the ancients, to see what today's models can test in the Zuo Zhuan's "with preparation there is no peril" and in Aristotle's account of Thales and the olive presses. (https://5dlibrary.com/answers/en/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c06/) 第6章 固定保费怎样花:行权水平、赔付上限、现金储备与参数不确定 — 程序 p61 显示,同样花掉资本的 0.006,行权 0.70 的保护把期望短缺降到 0.2396、把资本跌破 0.7 的概率降到 0.00058,但增长率降到 0.13936;0.90 到 0.60 的封顶保护期望短缺只降到 0.2957,增长率却升到 0.14033,是唯一高于不买保护的方案。程序 p62 显示,把期望短缺压到 0.25,现金储备每年放弃 0.0370,行权 0.90 的保护只放弃 0.0033,两者的盈亏平衡点是经营回报比现金利率高 0.0136,但基差风险与流动性没有计入。程序 p63 显示,20 年里见过 0 次与 2 次跳跃的经营者,对同一报价会得出相反的结论,即使 100 年的记录,边界区间 0.01517 到 0.02524 仍包住报价。下一章回到古人,看《左传》的「有备无患」与亚里士多德笔下泰勒斯的榨油机,能被今天的模型检验出什么。 (https://5dlibrary.com/answers/zh/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c06/) - Ch.7 Eastern and Western Sources: Thinking of Danger in Times of Peace, and Thales's Olive Presses — Wei Jiang's "in peace, think of danger; thinking brings preparation; with preparation there is no peril", recorded in the Zuo Zhuan under the eleventh year of Duke Xiang, gets the order and timing right: preparation costs least when things are calm. Program p71 shows that its literal claim fails: half the capital in cash lowers the probability of falling below 0.75 within a year from 0.034842 to 0.001576 but never to zero, at a cost of 0.0782 of expected return a year, while a tenth in cash plus protection at the Kelly ratio lowers it to 0.014879 at a cost of 0.0182 a year. In the model of program p72, the story of Thales recorded by Aristotle yields zero expected profit and a probability of 0.634 of losing the deposit when there is no edge, and an expected profit of 1.5036 when all three factors are present; Aristotle's attribution of the profit to monopoly is more accurate than the popular reading of "the first option". The next chapter brings the whole book down to an operator's annual routine: a tail-protection budget sheet, a small tool and a test plan that could overturn the book. (https://5dlibrary.com/answers/en/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c07/) 第7章 东西方原典考辨:居安思危与泰勒斯的榨油机 — 《左传·襄公十一年》魏绛的「居安思危,思则有备,有备无患」说对了次序与时机:准备在安的时候做代价最小。程序 p71 显示它的字面断言不成立:五成现金把一年内跌破 0.75 的概率从 0.034842 降到 0.001576,却永远不到零,代价是每年 0.0782 的期望收益;一成现金加凯利比例的保护以每年 0.0182 的代价把它降到 0.014879。亚里士多德记述的泰勒斯故事,在程序 p72 的模型里,没有任何优势时期望利润为零、损失定金的概率是 0.634,三个因素兼有时期望利润为 1.5036;亚里士多德把利润归于垄断,比流行的「第一笔期权」读法更准确。下一章把全书落到经营者的年度流程上:一张尾部保护预算表、一个小工具与一份能够推翻本书的检验方案。 (https://5dlibrary.com/answers/zh/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c07/) - Ch.8 Closing the Loop: The Annual Tail-Protection Budget Sheet and a Pre-Registered Test — Chapter 8 gives an annual tail-protection budget sheet, the Tail-Protection Premium Calculator and a pre-registered test plan: backtesting the tail model would need a single record of 1,171 years, so it can be tested only on pooled samples. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c08/) 第8章 实证闭环:年度尾部保护预算表与预注册检验 — 第八章给出年度尾部保护预算表、小工具「尾部保护保费计算器」与预注册检验方案:尾部模型的回测需要 1,171 年的单一记录,所以只能在汇总样本上检验。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/tail-risk-option-moat-jump-diffusion-kelly-fixed-premium/c08/) ## 5DL-B-291 · Dual-Source Redundancy and Supply-Disruption Real Options in Cross-Border Chains: Optimal Safety Stock and Source Splits under Shipping Delays and Customs Variability 中文书名:跨境供应链双源冗余与断供实物期权、物流延误清关波动下的最优安全库存配比母典 DOI: pending Cite: Fan, Y. (2026). Dual-Source Redundancy and Supply-Disruption Real Options in Cross-Border Chains: Optimal Safety Stock and Source Splits under Shipping Delays and Customs Variability (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its prices, lead times, outage frequencies, margins and capacities are illustrative; all calculations are generated by its programs; the book is not individual legal, tax or financial advice. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Quotation sheets for cross-border sourcing usually compare landed prices only, and safety stock is set by the one-line formula that includes demand variability alone. This book refutes the common sense that follows: that the single cheapest source is the most economical and that dual sourcing merely buys peace of mind. It brings lead-time variance back into the safety-stock formula, enters capital in transit, recurrent shortages and outages in a full annual ledger, and prices near capacity, exchange-rate switching and keeping a supplier warm as real options. Its prices, lead times, outage frequencies, margins and capacities are illustrative; all calculations are generated by its programs; the book is not individual legal, tax or financial advice. Every landed price is assumed to include the duties and taxes due on a truthful declaration; no undervaluation, splitting of consignments to evade duties or grey clearance is discussed, and customs and tariff rules are described only in outline, subject to human review before publication. Chapter 1 shows that with a far lead-time standard deviation of 10 days, stock set by the demand-only formula delivers a nominal 95 % service level of only 0.701 in fact. In Chapter 2's annual ledger the dual split saves 8323 a year against the best single far-source plan, while a naive ledger of landed prices reaches the opposite conclusion; a parameter map shows that with short outages the single far source is still the lowest-cost option. Chapter 3 shows that the ceiling on the best far share (0.55 at the base values) is set by the near source's volume flexibility, and a daily simulation finds that if the inventory system keeps overdue far shipments in transit, annual shortages rise from 11.4 to 259.5 units. Chapter 4 uses the newsvendor model to show that the value of two sources comes from the ability to order the near source after seeing the far outcome. Chapter 5 prices reserved capacity: at a disruption probability of 0.10 the critical reservation fee is 11.2, and the exchange-rate switching right is worth only 1.317 a unit at an annual volatility of 0.10 over 1 year. Chapter 6 shows that customs inspection gives lead times two peaks, so that a normal reorder point targeting 0.95 delivers only 0.9092, and that two sources sharing a port lose the dual saving. Chapter 7 examines "feeding on the enemy" in the Sunzi and Harris's 1913 square-root lot formula, which explains only 0.134 of a single far source's non-purchase cost. Chapter 8 gives the dual-source split sheet, the Dual-Source Safety-Stock Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:跨境货源的报价比较表通常只看到岸价,安全库存只用含需求波动的那一行公式。本书证伪由此而来的常识:单一最低价货源最省钱,双源只是花钱买心安。书中把交期方差请回安全库存公式,把在途资金、反复缺货与断供记进全年总账,并把近源产能、汇率切换与供应商保温当作实物期权定价。书中的价格、交期、断供频率、毛利与产能均为示意;全部计算由书中程序生成;本书是非个人化法律、税务或理财建议;书中每一个到岸价都假定已包含如实申报后应缴的税费,不讨论任何低报、拆单规避或灰色清关手法,海关与关税规则只述大意,以上架前人工复核为准。 第一章说明,远源交期标准差 10 天时,按只含需求项的公式备货,名义 95 % 的服务水平实际只有 0.701。第二章的全年总账中,双源一年比最好的单一远源方案省 8323,而只看到岸价的天真总账得出相反结论;参数地图显示断供短时单一远源仍最省。第三章说明最佳远源份额(基准 0.55)的上限由近源量弹性决定,并用逐日仿真发现:库存系统若不剔除逾期的远源货,每年缺货从 11.4 件涨到 259.5 件。第四章用报童模型说明双源的价值来自先看远源再下近单的反应能力。 第五章给预留产能定价:断供概率 0.10 时临界预留费是 11.2,汇率切换权在年化波动 0.10、期限 1 年时每件只值 1.317。第六章说明清关查验让交期变成双峰,目标 0.95 的正态再订货点实际只有 0.9092,两货源共用港口会抹掉双源的节省。第七章考辨《孙子·作战》的「因粮于敌」与哈里斯 1913 年的批量平方根公式,后者只解释单一远源非采购成本的 0.134。第八章给出采购双源配比表、双源安全库存计算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Lead-Time Variance: the Term Deleted from the Safety-Stock Formula — Chapter 1 shows that with a far lead-time standard deviation of 10 days, stock set by the demand-only formula delivers a nominal 95 % service level of only 0.701 in fact. In Chapter 2's annual ledger the dual split saves 8323 a year against the best single far-source plan, while a naive ledger of landed prices reaches the opposite conclusion; a parameter map shows that with short outages the single far source is still the lowest-cost option. (https://5dlibrary.com/answers/en/cross-border-dual-sourcing-disruption-real-option-safety-stock/c01/) 第1章 交期方差:安全库存公式里被删掉的那一项 — 第一章说明,远源交期标准差 10 天时,按只含需求项的公式备货,名义 95 % 的服务水平实际只有 0.701。 (https://5dlibrary.com/answers/zh/cross-border-dual-sourcing-disruption-real-option-safety-stock/c01/) - Ch.2 The Full Annual Ledger of the Lowest Price: One Far Source, One Near Source, or Two — The quotation sheet records only landed price; the ledger has seven lines. Under this chapter's illustrative parameters, the landed-price advantage of a single far source is eaten away line by line by stock in transit, safety stock, the outage buffer and outage losses. The dual split costs 1560877 a year, 8323 less than the best single far-source plan, and the single near source is dearest; a naive ledger that counts only landed price and a demand-only safety stock reaches the opposite conclusion. The parameter map shows that the size of the effect is set by outage length and the near premium: with short outages the single far source is the lowest-cost option, and once outages average 40 days or more the dual split begins to win. With the number of days lost each year held fixed, the rarer and longer the outages, the more they should be met by a second source rather than by stock. Lead-time variance is a second-order term in cost and a first-order term in service. (https://5dlibrary.com/answers/en/cross-border-dual-sourcing-disruption-real-option-safety-stock/c02/) 第2章 最低价的全年总账:单一远源、单一近源与双源 — 报价单只记到岸价,总账要记七行。在本章的示意参数下,单一远源的到岸价优势被在途库存、安全库存、断供缓冲与断供损失一行行吃掉:双源一年 1560877,比最好的单一远源方案少 8323,单一近源最贵;只算到岸价与只含需求项安全库存的天真总账,会得出相反的结论。参数地图说明,这个结论的分寸由断供长度与近源溢价决定:断供短时单一远源最省,断供平均 40 天以上时双源开始占优;固定每年的停工天数,断供越罕见越长,越应该用货源而不是库存去挡。交期方差在成本上是二阶项,在服务水平上是一阶项。 (https://5dlibrary.com/answers/zh/cross-border-dual-sourcing-disruption-real-option-safety-stock/c02/) - Ch.3 The Split: the Ceiling on the Far Share Is Set by Near-Source Flexibility — Chapter 3 shows that the ceiling on the best far share (0.55 at the base values) is set by the near source's volume flexibility, and a daily simulation finds that if the inventory system keeps overdue far shipments in transit, annual shortages rise from 11.4 to 259.5 units. (https://5dlibrary.com/answers/en/cross-border-dual-sourcing-disruption-real-option-safety-stock/c03/) 第3章 配比:远源份额的上限由近源弹性决定 — 第三章说明最佳远源份额(基准 0.55)的上限由近源量弹性决定,并用逐日仿真发现:库存系统若不剔除逾期的远源货,每年缺货从 11.4 件涨到 259.5 件。 (https://5dlibrary.com/answers/zh/cross-border-dual-sourcing-disruption-real-option-safety-stock/c03/) - Ch.4 The Two-Source Newsvendor: the Critical Fractile of a One-Order Season and the Value of Reacting — With one order per season, the order quantity is set by the critical fractile: the shortage loss divided by the sum of the shortage and leftover losses. Under the illustrative values the fractile is 0.7500, the best order 4004.7 units, and ordering the mean costs about a quarter more; the more asymmetric the cost of shortage, the worse ordering the mean becomes. With two sources, static dual, which splits the order in advance, only compromises between the two single sources. The real value comes from the ability to react: order the far source first, and decide on the near source once it is known whether the far goods will arrive. At a disruption probability of 0.10 the expected profit of reactive dual is 398325, more than thirty thousand above far only. Reacting needs real near capacity when called, and that capacity has to be reserved and paid for. (https://5dlibrary.com/answers/en/cross-border-dual-sourcing-disruption-real-option-safety-stock/c04/) 第4章 报童双源:一季一单的临界分位与反应能力 — 一季一单时,订货量由临界分位决定:缺货损失除以缺货与剩货损失之和。示意参数下临界分位 0.7500,最优订货量 4004.7 件,按均值订货的损失多出约四分之一;缺货代价越不对称,按均值订货越吃亏。两个货源时,事前把订单分成两份的静态双源只是在两个单源之间折中;真正的价值来自反应能力:先订远源,等知道远源能否到货再决定近源。断供概率 0.10 时,反应式双源的期望利润 398325,比只用远源高出三万多。反应能力需要近源在被叫时真有产能,这份产能要花钱预留。 (https://5dlibrary.com/answers/zh/cross-border-dual-sourcing-disruption-real-option-safety-stock/c04/) - Ch.5 Supply-Disruption Real Options: Reserved Capacity, Exchange-Rate Switching and Keeping a Supplier Warm — Chapter 5 prices reserved capacity: at a disruption probability of 0.10 the critical reservation fee is 11.2, and the exchange-rate switching right is worth only 1.317 a unit at an annual volatility of 0.10 over 1 year. (https://5dlibrary.com/answers/en/cross-border-dual-sourcing-disruption-real-option-safety-stock/c05/) 第5章 断供实物期权:预留产能、汇率切换与供应商保温 — 第五章给预留产能定价:断供概率 0.10 时临界预留费是 11.2,汇率切换权在年化波动 0.10、期限 1 年时每件只值 1.317。 (https://5dlibrary.com/answers/zh/cross-border-dual-sourcing-disruption-real-option-safety-stock/c05/) - Ch.6 Clearance Variability and Robustness: Two-Peaked Lead Times, a Handful of Shipments, and a Common Failure Point — Chapter 6 shows that customs inspection gives lead times two peaks, so that a normal reorder point targeting 0.95 delivers only 0.9092, and that two sources sharing a port lose the dual saving. (https://5dlibrary.com/answers/en/cross-border-dual-sourcing-disruption-real-option-safety-stock/c06/) 第6章 清关波动与稳健性:双峰交期、寥寥几票与共同断点 — 第六章说明清关查验让交期变成双峰,目标 0.95 的正态再订货点实际只有 0.9092,两货源共用港口会抹掉双源的节省。 (https://5dlibrary.com/answers/zh/cross-border-dual-sourcing-disruption-real-option-safety-stock/c06/) - Ch.7 The Classics Examined: "Feeding on the Enemy" and Harris's Square Root — Chapter 7 examines "feeding on the enemy" in the Sunzi and Harris's 1913 square-root lot formula, which explains only 0.134 of a single far source's non-purchase cost. (https://5dlibrary.com/answers/en/cross-border-dual-sourcing-disruption-real-option-safety-stock/c07/) 第7章 原典考辨:「因粮于敌」与哈里斯的平方根 — 第七章考辨《孙子·作战》的「因粮于敌」与哈里斯 1913 年的批量平方根公式,后者只解释单一远源非采购成本的 0.134。 (https://5dlibrary.com/answers/zh/cross-border-dual-sourcing-disruption-real-option-safety-stock/c07/) - Ch.8 Closing the Loop: the Dual-Source Split Sheet, the Small Tool and a Pre-Registered Test — Chapter 8 gives the dual-source split sheet, the Dual-Source Safety-Stock Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cross-border-dual-sourcing-disruption-real-option-safety-stock/c08/) 第8章 实证闭环:采购双源配比表、小工具与预注册检验 — 第八章给出采购双源配比表、双源安全库存计算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cross-border-dual-sourcing-disruption-real-option-safety-stock/c08/) ## 5DL-B-292 · Digital Legacy and Perpetual Cross-Generation Contracts for a One-Person Empire: Time-Lock State Machines and the Frictionless Succession of Libraries, Intellectual Property and Automated Systems 中文书名:单兵帝国数字遗产与跨代永续契约、时间锁状态机与书库知识产权自动化系统零摩擦传承母典 DOI: pending Cite: Fan, Y. (2026). Digital Legacy and Perpetual Cross-Generation Contracts for a One-Person Empire: Time-Lock State Machines and the Frictionless Succession of Libraries, Intellectual Property and Automated Systems (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 1 finds in an illustrative model that a will alone loses 0.781 of value, a technical handover 0.150 and a handover with yearly drills 0.030; a chain of five links needs 0.979 per link to be reached nine times in ten. Chapter 5 shows that years 51 to 70 of protection hold only 0.00334 of an illustrative library's present value, while 30 dark days cost 18979 yuan and the first deadline needing a person falls on day 76 after death. Abstract: A press, website or software shop run by one person keeps almost all its assets online, and the usual arrangement for after death is a will that names the heirs. This book refutes the belief behind that comfort: a will answers who owns an asset, not how anyone gets in; an asset must pass along an access chain of recovery mailbox, second factor, device, password vault and platform procedure, and the probabilities along the chain multiply. Its parameters are illustrative; all calculations are generated by its programs; it is not individual legal, tax or financial advice, gives legal rules only in gist subject to human review before publication, leaves actual matters to practising lawyers under local law, and holds that no automation replaces a valid will and notarisation. Chapter 1 finds in an illustrative model that a will alone loses 0.781 of value, a technical handover 0.150 and a handover with yearly drills 0.030; a chain of five links needs 0.979 per link to be reached nine times in ten. Chapter 2 treats a dead man's switch as a sequential test: absences come in stretches, a 30-day window still fires falsely with yearly probability 0.062969, and the tail of absence lengths decides false firing; the best window of a bare switch is 172 days, and a trustee check shortens it to 29. Chapter 3 writes the handover as a state machine and finds a five-step hole in a common design among 87380 event sequences; a three-of-five check never completes with probability 0.02661. Chapter 4 shows a three-of-five threshold missing with probability 0.026612 and confirming wrongly with 0.00115813, holding in year 30 with probability 0.1550 without drills and 0.9992 with yearly drills, and explains threshold splitting on a toy integer. Chapter 5 shows that years 51 to 70 of protection hold only 0.00334 of an illustrative library's present value, while 30 dark days cost 18979 yuan and the first deadline needing a person falls on day 76 after death. Chapter 6 treats decay across generations: without drills 0.1681 remains by the fifth generation, with drills 0.7738; with one successor per generation on average a line survives five generations with probability 0.268. Chapter 7 examines Mencius on a legacy cut off after five generations (0.03125 left if each generation keeps half) and Justinian on seven witnesses, revocation by a later will and the Falcidian quarter (seven witnesses gathered only after death fail with probability 0.6794). Chapter 8 gives the inventory, the yearly drill, the Time-Lock Check-in Window Calculator and a pre-registered test: fixing an illustrative scorecard raises reach from 0.821 to 0.976, test one needs 32 owners and test two 110 owner-years. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:一人经营的出版社、网站与软件小店,资产几乎全在线上。常见的身后安排是立一份遗嘱写明归属。本书证伪这份安心背后的信念:遗嘱回答「归谁」,不回答「怎么进去」;资产要穿过恢复邮箱、第二因素、设备、密码库与平台流程组成的访问链,链上的概率相乘。书中参数为示意;全部计算由书中程序生成;本书是非个人化法律、税务或理财建议,法律规则只写大意、以上架前人工复核为准,具体事务须由执业律师按当地法律办理,任何自动化都不能替代合法有效的遗嘱与公证。 第一章在示意模型中量出只立遗嘱按价值计失联 0.781,技术交接降到 0.150,加年度演练降到 0.030;五环的链要九成可达,每环须 0.979。第二章把死人开关看成序贯检验:缺席成段发生,窗口 30 天时年误触发仍有 0.062969,决定误触发的是缺席时长的尾部;裸开关的最优窗口是 172 天,加上受托人核验缩短到 29 天。第三章把交接写成状态机,在 87380 条事件序列上找到常见设计的五步漏洞;五取三的核验永不达成的概率是 0.02661。第四章显示五取三门限的漏确认 0.026612、误确认 0.00115813,不演练时门限到第 30 年只剩 0.1550 的概率守得住,每年演练则为 0.9992,并用玩具整数说明门限分片的概念。 第五章显示保护期第 51 至 70 年只占示意书库现值的 0.00334,停摆 30 天却损失 18979 元,最早的人工截止在去世后第 76 天。第六章讨论代际衰减:不演练时第五代只剩 0.1681,演练后 0.7738;平均每代一位接班人时谱系五代后仍在的概率只有 0.268。第七章考辨孟子「君子之泽,五世而斩」(每代减半时五代后剩 0.03125)与查士丁尼论七名证人、后立撤销前立与法尔其第亚法(死后才凑七名证人凑不齐的概率 0.6794)。第八章给出数字遗产清单、年度演练、时间锁签到周期计算器与预注册检验:示意记分卡补齐后可达概率由 0.821 升到 0.976,检验一需 32 位主事者,检验二需 110 个主事者年。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Lost Contact: the Access Chain beyond the Will — Chapter 1 finds in an illustrative model that a will alone loses 0.781 of value, a technical handover 0.150 and a handover with yearly drills 0.030; a chain of five links needs 0.979 per link to be reached nine times in ten. (https://5dlibrary.com/answers/en/digital-legacy-time-lock-succession-one-person-empire/c01/) 第1章 失联:遗嘱之外的访问链 — 第一章在示意模型中量出只立遗嘱按价值计失联 0.781,技术交接降到 0.150,加年度演练降到 0.030;五环的链要九成可达,每环须 0.979。 (https://5dlibrary.com/answers/zh/digital-legacy-time-lock-succession-one-person-empire/c01/) - Ch.2 The Check-in Window: Trading False Firing against Late Release — Chapter 2 treats a dead man's switch as a sequential test: absences come in stretches, a 30-day window still fires falsely with yearly probability 0.062969, and the tail of absence lengths decides false firing; the best window of a bare switch is 172 days, and a trustee check shortens it to 29. (https://5dlibrary.com/answers/en/digital-legacy-time-lock-succession-one-person-empire/c02/) 第2章 签到窗口:误触发与迟放行的最优权衡 — 第二章把死人开关看成序贯检验:缺席成段发生,窗口 30 天时年误触发仍有 0.062969,决定误触发的是缺席时长的尾部;裸开关的最优窗口是 172 天,加上受托人核验缩短到 29 天。 (https://5dlibrary.com/answers/zh/digital-legacy-time-lock-succession-one-person-empire/c02/) - Ch.3 The Time-Lock State Machine: Reachability, Safety and Bounded Model Checking — Chapter 3 writes the handover as a state machine and finds a five-step hole in a common design among 87380 event sequences; a three-of-five check never completes with probability 0.02661. (https://5dlibrary.com/answers/en/digital-legacy-time-lock-succession-one-person-empire/c03/) 第3章 时间锁状态机:可达性、安全性与有界模型检验 — 第三章把交接写成状态机,在 87380 条事件序列上找到常见设计的五步漏洞;五取三的核验永不达成的概率是 0.02661。 (https://5dlibrary.com/answers/zh/digital-legacy-time-lock-succession-one-person-empire/c03/) - Ch.4 The Trustee Quorum: Thresholds, Attrition and the Idea of Splitting — Chapter 4 shows a three-of-five threshold missing with probability 0.026612 and confirming wrongly with 0.00115813, holding in year 30 with probability 0.1550 without drills and 0.9992 with yearly drills, and explains threshold splitting on a toy integer. (https://5dlibrary.com/answers/en/digital-legacy-time-lock-succession-one-person-empire/c04/) 第4章 受托人法定人数:多签门限、人员流失与分片的概念 — 第四章显示五取三门限的漏确认 0.026612、误确认 0.00115813,不演练时门限到第 30 年只剩 0.1550 的概率守得住,每年演练则为 0.9992,并用玩具整数说明门限分片的概念。 (https://5dlibrary.com/answers/zh/digital-legacy-time-lock-succession-one-person-empire/c04/) - Ch.5 Automating the Handover of Intellectual Property: Term, Dark Gaps and the Renewal Calendar — Chapter 5 shows that years 51 to 70 of protection hold only 0.00334 of an illustrative library's present value, while 30 dark days cost 18979 yuan and the first deadline needing a person falls on day 76 after death. (https://5dlibrary.com/answers/en/digital-legacy-time-lock-succession-one-person-empire/c05/) 第5章 知识产权的自动化交接:保护期、停摆缺口与续期日历 — 第五章显示保护期第 51 至 70 年只占示意书库现值的 0.00334,停摆 30 天却损失 18979 元,最早的人工截止在去世后第 76 天。 (https://5dlibrary.com/answers/zh/digital-legacy-time-lock-succession-one-person-empire/c05/) - Ch.6 Perpetuity across Generations: Decay of Transfer, Dilution of Shares and the Line of Stewards — Chapter 6 treats decay across generations: without drills 0.1681 remains by the fifth generation, with drills 0.7738; with one successor per generation on average a line survives five generations with probability 0.268. (https://5dlibrary.com/answers/en/digital-legacy-time-lock-succession-one-person-empire/c06/) 第6章 跨代永续:传承衰减、份额稀释与守护者谱系 — 第六章讨论代际衰减:不演练时第五代只剩 0.1681,演练后 0.7738;平均每代一位接班人时谱系五代后仍在的概率只有 0.268。 (https://5dlibrary.com/answers/zh/digital-legacy-time-lock-succession-one-person-empire/c06/) - Ch.7 Sources from East and West: Mencius on Five Generations and Justinian on Wills — Chapter 7 examines Mencius on a legacy cut off after five generations (0.03125 left if each generation keeps half) and Justinian on seven witnesses, revocation by a later will and the Falcidian quarter (seven witnesses gathered only after death fail with probability 0.6794). (https://5dlibrary.com/answers/en/digital-legacy-time-lock-succession-one-person-empire/c07/) 第7章 东西方原典考辨:孟子「五世而斩」与查士丁尼论遗嘱 — 第七章考辨孟子「君子之泽,五世而斩」(每代减半时五代后剩 0.03125)与查士丁尼论七名证人、后立撤销前立与法尔其第亚法(死后才凑七名证人凑不齐的概率 0.6794)。 (https://5dlibrary.com/answers/zh/digital-legacy-time-lock-succession-one-person-empire/c07/) - Ch.8 Closing the Loop: the Digital-Legacy Inventory, the Yearly Drill and a Pre-registered Test — Chapter 8 gives the inventory, the yearly drill, the Time-Lock Check-in Window Calculator and a pre-registered test: fixing an illustrative scorecard raises reach from 0.821 to 0.976, test one needs 32 owners and test two 110 owner-years. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/digital-legacy-time-lock-succession-one-person-empire/c08/) 第8章 实证闭环:数字遗产清单、年度演练与预注册检验 — 第八章给出数字遗产清单、年度演练、时间锁签到周期计算器与预注册检验:示意记分卡补齐后可达概率由 0.821 升到 0.976,检验一需 32 位主事者,检验二需 110 个主事者年。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/digital-legacy-time-lock-succession-one-person-empire/c08/) ## 5DL-B-293 · Normalised Compression Distance and Kolmogorov Complexity: Detecting Rewritten Copies and Clustering Families of Shared-Origin Sales Scripts 中文书名:归一化压缩距离与柯尔莫哥洛夫复杂度、洗稿伪原创识别与同源话术家族聚类母典 DOI: pending Cite: Fan, Y. (2026). Normalised Compression Distance and Kolmogorov Complexity: Detecting Rewritten Copies and Clustering Families of Shared-Origin Sales Scripts (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its corpora, thresholds and parameters are illustrative, every sample text is generated by program or written by the author, and all calculations are generated by its programs. It is not individual business advice and not legal advice, and every flag raised by its small tool needs human review. Abstract: Operators who publish content often hear that replacing three words in ten defeats a duplicate checker, that compression distance is steadier, and that only deep paraphrase fools every tool, so it must be recognised by its numbers and the order of its argument. This book splits that claim into parts that programs can test and tests each on a pseudo-word corpus in which the substitution rate can be set exactly. Its corpora, thresholds and parameters are illustrative, every sample text is generated by program or written by the author, and all calculations are generated by its programs. Its purpose is to protect original work and find copies of it; it does not discuss how to rewrite other people's content to escape duplicate checks. It is not individual business advice and not legal advice, and every flag raised by its small tool needs human review. Chapter 1 finds that the Jaccard index of word 3-grams falls most steeply at the start of substitution and has lost 0.641 of its range by thirty per cent, with a fixed alarm bar of 0.5 crossed at a rate of 0.211; calibrated on unrelated texts, literal fragments still detect 0.957 of rewrites with every content word replaced, while compression distance detects 0.000 and declines more slowly only in the share of signal kept (0.551 against 0.385). Chapter 2 shows that compression only bounds Kolmogorov complexity from above, and that once a text exceeds zlib's window the compression distance of a complete copy jumps to 0.9330. Chapter 3 measures short-text bias: the power to separate is still 0.9894, yet a fixed threshold gives false positives at a rate of 0.900 at twenty words and catches only 0.030 at 640 words. Chapter 4 shows that locality-sensitive hashing designed for near copies misses rewrites, admitting only 0.5088 of thirty per cent rewrites at a knee of 0.4204, so the knee must come down to 0.1250. Chapter 5 takes on deep paraphrase: the area under the curve of compression distance is only 0.0949, the ordered number fingerprint detects 0.923, and the argument-sequence fingerprint detects only 0.050 because of genre conventions. Chapter 6 clusters scripts into families of shared origin by average linkage, with an adjusted Rand index of 0.9962 at a calibrated cut. Chapter 7 examines the colophon of the Cheng house at Meishan in the Song edition of the Dongdu shilüe and the Statute of Anne of 1710; under a half-life model, fourteen years cover nine tenths of sales only if the half-life is at most 4.21 years. Chapter 8 gives 5D Library's weekly monitoring, the de-duplication of direct-answer pages, the Rewrite and Shared-Origin Detector and a pre-registered test plan: library-wide calibration of the number fingerprint cuts false alarms from 96 to 9 at the price of catching only 12 deep paraphrases, and test one needs 155 rewrites. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:做内容的经营者常听说:换掉三成的词,查重就认不出来;压缩距离更稳,只有深度意译才骗得过所有工具,要靠数字与论证顺序去认。本书把这个说法拆成可以用程序检验的几部分,在换词率可以精确控制的伪词语料上逐一检验。书中的语料、阈值与参数均为示意,所有样本文本由程序生成或作者自拟;全部计算由书中程序生成。本书的用途是保护原创与发现抄袭,不讨论如何改写他人内容以逃避查重;本书是非个人化经营建议,不是法律意见,书中小工具的每一条标记都须人工复核。 第一章发现,词三元组杰卡德指数在换词之初下降最陡,到三成时已丢掉 0.641 的可跌幅度,固定报警线 0.5 在换词率 0.211 处被越过;按无关文本校准后,实词全换时字面片段仍有 0.957 的检出率,压缩距离为 0.000,它只在「保留比例」上下降更缓(0.551 对 0.385)。第二章说明压缩只给柯尔莫哥洛夫复杂度的上界,并显示文本一旦超过 zlib 的窗口,完整拷贝的压缩距离会跳到 0.9330。第三章量化短文本偏差:区分能力仍有 0.9894,固定阈值却在二十词上误报 0.900、在 640 词上只检出 0.030。第四章说明为近似拷贝设计的局部敏感哈希会漏掉改写,膝点 0.4204 时三成改写只有 0.5088 进入候选,须把膝点压到 0.1250。 第五章处理深度意译:压缩距离的曲线下面积只有 0.0949,按顺序共享的数字指纹检出 0.923,论证序列指纹因体裁惯例只检出 0.050。第六章用平均联结把话术聚成同源家族,在校准切点处调整兰德指数为 0.9962。第七章考辨宋刻《东都事略》的眉山程舍人宅牌记与 1710 年的安妮法令,按半衰期模型,十四年覆盖九成销量的条件是半衰期不超过 4.21 年。第八章给出 5D Library 的周度防洗稿监测、直答页去重、洗稿同源检测器与预注册检验方案:全库校准的数字指纹把误报从 96 篇压到 9 篇,代价是深度意译只抓到 12 篇;检验一需要 155 篇改写稿。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Sweeping the Substitution Rate: Where Literal Matching Fails — Chapter 1 finds that the Jaccard index of word 3-grams falls most steeply at the start of substitution and has lost 0.641 of its range by thirty per cent, with a fixed alarm bar of 0.5 crossed at a rate of 0.211; calibrated on unrelated texts, literal fragments still detect 0.957 of rewrites with every content word replaced, while compression distance detects 0.000 and declines more slowly only in the share of signal kept (0.551 against 0.385). (https://5dlibrary.com/answers/en/normalised-compression-distance-rewrite-detection-script-families/c01/) 第1章 换词率扫描:字面查重在哪里失效 — 第一章发现,词三元组杰卡德指数在换词之初下降最陡,到三成时已丢掉 0.641 的可跌幅度,固定报警线 0.5 在换词率 0.211 处被越过;按无关文本校准后,实词全换时字面片段仍有 0.957 的检出率,压缩距离为 0.000,它只在「保留比例」上下降更缓(0.551 对 0.385)。 (https://5dlibrary.com/answers/zh/normalised-compression-distance-rewrite-detection-script-families/c01/) - Ch.2 Kolmogorov Complexity and Real Compressors: Upper Bounds, Normality and the Window — Chapter 2 shows that compression only bounds Kolmogorov complexity from above, and that once a text exceeds zlib's window the compression distance of a complete copy jumps to 0.9330. (https://5dlibrary.com/answers/en/normalised-compression-distance-rewrite-detection-script-families/c02/) 第2章 柯尔莫哥洛夫复杂度与压缩器:上界、正规性与窗口 — 第二章说明压缩只给柯尔莫哥洛夫复杂度的上界,并显示文本一旦超过 zlib 的窗口,完整拷贝的压缩距离会跳到 0.9330。 (https://5dlibrary.com/answers/zh/normalised-compression-distance-rewrite-detection-script-families/c02/) - Ch.3 Short-Text Bias and Threshold Sensitivity: One Line Cannot Rule Every Length — Chapter 3 measures short-text bias: the power to separate is still 0.9894, yet a fixed threshold gives false positives at a rate of 0.900 at twenty words and catches only 0.030 at 640 words. (https://5dlibrary.com/answers/en/normalised-compression-distance-rewrite-detection-script-families/c03/) 第3章 短文本偏差与阈值敏感度:一条线管不了所有长度 — 第三章量化短文本偏差:区分能力仍有 0.9894,固定阈值却在二十词上误报 0.900、在 640 词上只检出 0.030。 (https://5dlibrary.com/answers/zh/normalised-compression-distance-rewrite-detection-script-families/c03/) - Ch.4 N-grams, MinHash and Locality-Sensitive Hashing: a First Screen over Millions of Pages — Chapter 4 shows that locality-sensitive hashing designed for near copies misses rewrites, admitting only 0.5088 of thirty per cent rewrites at a knee of 0.4204, so the knee must come down to 0.1250. (https://5dlibrary.com/answers/en/normalised-compression-distance-rewrite-detection-script-families/c04/) 第4章 n 元组、MinHash 与局部敏感哈希:百万篇里的第一轮初筛 — 第四章说明为近似拷贝设计的局部敏感哈希会漏掉改写,膝点 0.4204 时三成改写只有 0.5088 进入候选,须把膝点压到 0.1250。 (https://5dlibrary.com/answers/zh/normalised-compression-distance-rewrite-detection-script-families/c04/) - Ch.5 Deep Paraphrase: Number Fingerprints and Argument-Sequence Fingerprints — Chapter 5 takes on deep paraphrase: the area under the curve of compression distance is only 0.0949, the ordered number fingerprint detects 0.923, and the argument-sequence fingerprint detects only 0.050 because of genre conventions. (https://5dlibrary.com/answers/en/normalised-compression-distance-rewrite-detection-script-families/c05/) 第5章 深度意译:数字指纹与论证序列指纹 — 第五章处理深度意译:压缩距离的曲线下面积只有 0.0949,按顺序共享的数字指纹检出 0.923,论证序列指纹因体裁惯例只检出 0.050。 (https://5dlibrary.com/answers/zh/normalised-compression-distance-rewrite-detection-script-families/c05/) - Ch.6 Hierarchical Clustering of Shared-Origin Script Families — Chapter 6 clusters scripts into families of shared origin by average linkage, with an adjusted Rand index of 0.9962 at a calibrated cut. (https://5dlibrary.com/answers/en/normalised-compression-distance-rewrite-detection-script-families/c06/) 第6章 同源话术家族的层次聚类 — 第六章用平均联结把话术聚成同源家族,在校准切点处调整兰德指数为 0.9962。 (https://5dlibrary.com/answers/zh/normalised-compression-distance-rewrite-detection-script-families/c06/) - Ch.7 The Classics Examined: the Colophon of the Cheng House at Meishan and the Statute of Anne — Chapter 7 examines the colophon of the Cheng house at Meishan in the Song edition of the Dongdu shilüe and the Statute of Anne of 1710; under a half-life model, fourteen years cover nine tenths of sales only if the half-life is at most 4.21 years. (https://5dlibrary.com/answers/en/normalised-compression-distance-rewrite-detection-script-families/c07/) 第7章 原典考辨:眉山程舍人宅牌记与安妮法令 — 第七章考辨宋刻《东都事略》的眉山程舍人宅牌记与 1710 年的安妮法令,按半衰期模型,十四年覆盖九成销量的条件是半衰期不超过 4.21 年。 (https://5dlibrary.com/answers/zh/normalised-compression-distance-rewrite-detection-script-families/c07/) - Ch.8 Closing the Loop: Weekly Monitoring for Rewritten Copies, De-duplicating Direct-Answer Pages and a Pre-registered Test — Chapter 8 gives 5D Library's weekly monitoring, the de-duplication of direct-answer pages, the Rewrite and Shared-Origin Detector and a pre-registered test plan: library-wide calibration of the number fingerprint cuts false alarms from 96 to 9 at the price of catching only 12 deep paraphrases, and test one needs 155 rewrites. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/normalised-compression-distance-rewrite-detection-script-families/c08/) 第8章 实证闭环:防洗稿周度监测、直答页去重与预注册检验 — 第八章给出 5D Library 的周度防洗稿监测、直答页去重、洗稿同源检测器与预注册检验方案:全库校准的数字指纹把误报从 96 篇压到 9 篇,代价是深度意译只抓到 12 篇;检验一需要 155 篇改写稿。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/normalised-compression-distance-rewrite-detection-script-families/c08/) ## 5DL-B-294 · Bloom Filters and HyperLogLog Cardinality Estimation: Counting Unique Visitors and Deduplicating Orders across a Hundred Sites in Kilobytes 中文书名:布隆过滤器与超对数对数基数估计、千字节内存统计百站独立访客与去重订单母典 DOI: pending Cite: Fan, Y. (2026). Bloom Filters and HyperLogLog Cardinality Estimation: Counting Unique Visitors and Deduplicating Orders across a Hundred Sites in Kilobytes (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: In an illustrative network, per-site unique visitors add up to 2.224 times the true number of people. Marked boxes separate published evidence, Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Operators who run a hundred sites often add up the unique visitors on each site's dashboard and call the total the network's audience, keep every order number in an ever-longer table to compare against, and trust the textbook false-positive formula of the Bloom filter as exact. This book refutes all three beliefs. Exact deduplication does need linear memory, and that is proven; but once randomisation and approximation are both accepted, counting any number of distinct visitors takes 3072 bytes at a standard error of 0.0163. In an illustrative network, per-site unique visitors add up to 2.224 times the true number of people. The classical false-positive formula of the Bloom filter is a lower bound, not an exact value: in the smallest counterexample the true rate is 5/8 and the formula gives 9/16. Its sites, visitors, orders and traffic parameters are illustrative; all calculations are generated by its programs; the book is not individual business advice; visitor counting uses only anonymised, key-hashed identifiers, collects no identifying information, and must follow applicable privacy law (summarised only in general terms) and each site's privacy notice. Chapter 1 draws up the memory bill of exact deduplication and measures the overstatement of adding sites up. Chapter 2 opens HyperLogLog, recomputes the limiting constant 1.03896, and shows that the constant is 1.106 at m = 16 and that the abstract's example of 1.5 kilobytes corresponds to a standard error of 0.0230. Chapter 3 handles both ends: below 2.5 times the register count linear counting takes over, with a positive bias of 0.0144 at the switch point, and a 32-bit hash loses 0.1079 of a billion visitors. Chapter 4 measures Bloom filters and explains the optimal probe count and the space factor of 1.443. Chapter 5 rules that the filter routes and never rejects: lookups fall by 0.987, while rejecting on the filter would wrongly refuse 4916 real orders in thirty days; the window rotates in two generations, deletion needs a counting filter, and clearing bits in an ordinary filter makes 0.3944 of live orders read as unseen. Chapter 6 merges the sketches of a hundred sites, shows that at a shared share of 0.10 the intersection estimate has a relative error of 0.421, and explains that registers are not anonymous: with few visitors, adding one more person changes the sketch with probability 0.9935. Chapter 7 examines the registration rule of the Minor Minister in the Zhou li and Laplace's ratio of births, recomputes Laplace's required count of 771469 people, and notes that the same precision needs only 24576 bytes today. Chapter 8 gives a registers-only weekly procedure for tollbooth and direct-answer pages, the Cardinality Estimator and a pre-registered audit plan: in the demonstration week the union of two pages is estimated at 35420 against an audit of 35818, the spread test needs 70 page-days and the order gate 3463 keys known to be new. Marked boxes separate published evidence, Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:经营一百个站点的主事者,常把各站后台的独立访客加在一起当作全网人数,把每张订单号存进越来越长的表里逐条比对,并相信教科书里布隆过滤器的误判率公式是精确的。本书证伪这三个信念。精确去重确实需要线性内存,这有严格证明;但随机化与近似同时接受之后,统计任意多个不同访客只需 3072 字节,标准误 0.0163。示意网络中,各站独立访客相加是真实人数的 2.224 倍。布隆过滤器的经典误判率公式是下界而非精确值:最小反例的真实值是 5/8,公式给出 9/16。书中的站点、访客、订单与流量参数均为示意;全部计算由书中程序生成;本书是非个人化经营建议;访客统计只用匿名化、加密钥哈希的标识,不收集个人身份信息,须遵守所在地个人信息保护法规与站点隐私声明,法规只述大意,以上架前人工复核为准。 第一章算出精确去重的内存账并量出加总的高估。第二章拆开 HyperLogLog,复算极限常数 1.03896,并说明 m = 16 时常数为 1.106、摘要里 1.5 千字节的例子对应标准误 0.0230。第三章处理两端:基数低于 2.5 倍寄存器个数时改用线性计数,切换点有 0.0144 的正偏差;32 位哈希在十亿访客时丢掉 0.1079。第四章实测布隆过滤器,说明最优探查数与 1.443 倍的空间因子。第五章规定过滤器只分流、不拒单:查库量减少 0.987,而凭过滤器拒单三十天会拒错 4916 张真实订单;窗口用两代轮换,删除用计数过滤器,普通过滤器清零会让 0.3944 的有效订单读成没见过。 第六章合并一百个站点的草图,指出共有比例 0.10 时交集估计的相对误差达 0.421,并说明寄存器不是匿名的:访客稀少时再加一人使草图改变的概率是 0.9935。第七章考辨《周礼》小司徒的比法与拉普拉斯的出生比,复算拉普拉斯所需的 771469 人清点规模,并指出同样精度今天只需 24576 字节。第八章给出收费站与直答页只存寄存器的周报流程、基数估计器与预注册审计方案:示范周两页并集估计 35420 人、审计 35818 人,离散检验需要 70 个页日,订单闸门需要 3463 个已知新键。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Memory Bill of Exact Deduplication: Why Unique Visitors Do Not Add Up — The memory of exact deduplication is almost proportional to the count: at a hundred million visitors the information floor is 485.841 MB, 158151 times a sketch of 3072 bytes. This is not a shortfall of engineering but a proven result: a deterministic algorithm needs linear memory even to approximate within a tenth, and a randomised algorithm needs linear memory to be exact; only when randomisation and approximation are both accepted can memory stop depending on the count. Adding up per-site unique visitors overstated the illustrative network by a factor of 2.224; merging a hundred sketches by register maximum gave an error of 0.0221 in 307200 bytes. The next chapter opens HyperLogLog itself: why the position of the first 1-bit can count people, why a harmonic mean beats a geometric one, and where the constant in 1.04/√m comes from and how accurate it really is. (https://5dlibrary.com/answers/en/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c01/) 第1章 精确去重的内存账:独立访客为什么不能相加 — 精确去重的内存与人数几乎成正比:一亿访客时信息论下限是 485.841 MB,是一个 3072 字节草图的 158151 倍。这不是工程不够好,而是有严格证明的结论:确定性算法连一成以内的近似都需要线性内存,随机算法要精确也需要线性内存;只有随机化与近似同时接受,内存才能与人数无关。各站独立访客直接相加,在示意网络中高估了 2.224 倍;一百个草图按寄存器取最大值合并,误差是 0.0221,存储是 307200 字节。下一章拆开 HyperLogLog 本身:首个 1 位的位置为什么能数人,调和平均为什么比几何平均好,1.04/√m 这个常数从哪里来、到底有多准? (https://5dlibrary.com/answers/zh/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c01/) - Ch.2 HyperLogLog: the First 1-Bit, the Harmonic Mean and 1.04/√m — Chapter 2 opens HyperLogLog, recomputes the limiting constant 1.03896, and shows that the constant is 1.106 at m = 16 and that the abstract's example of 1.5 kilobytes corresponds to a standard error of 0.0230. (https://5dlibrary.com/answers/en/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c02/) 第2章 超对数对数:首个 1 位、调和平均与 1.04/√m — 第二章拆开 HyperLogLog,复算极限常数 1.03896,并说明 m = 16 时常数为 1.106、摘要里 1.5 千字节的例子对应标准误 0.0230。 (https://5dlibrary.com/answers/zh/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c02/) - Ch.3 Corrections at Both Ends: Linear Counting, Switch-Point Bias and Hash Width — Chapter 3 handles both ends: below 2.5 times the register count linear counting takes over, with a positive bias of 0.0144 at the switch point, and a 32-bit hash loses 0.1079 of a billion visitors. (https://5dlibrary.com/answers/en/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c03/) 第3章 两端的修正:线性计数、切换点偏差与哈希位数 — 第三章处理两端:基数低于 2.5 倍寄存器个数时改用线性计数,切换点有 0.0144 的正偏差;32 位哈希在十亿访客时丢掉 0.1079。 (https://5dlibrary.com/answers/zh/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c03/) - Ch.4 The Bloom Filter: the False-Positive Formula, the Optimal Probe Count and 1.44 Times the Space — Chapter 4 measures Bloom filters and explains the optimal probe count and the space factor of 1.443. (https://5dlibrary.com/answers/en/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c04/) 第4章 布隆过滤器:误判率公式、最优探查数与 1.44 倍空间 — 第四章实测布隆过滤器,说明最优探查数与 1.443 倍的空间因子。 (https://5dlibrary.com/answers/zh/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c04/) - Ch.5 Order Deduplication: Filters Route but Never Reject, Windows Rotate by Generation, Counters Delete — Chapter 5 rules that the filter routes and never rejects: lookups fall by 0.987, while rejecting on the filter would wrongly refuse 4916 real orders in thirty days; the window rotates in two generations, deletion needs a counting filter, and clearing bits in an ordinary filter makes 0.3944 of live orders read as unseen. (https://5dlibrary.com/answers/en/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c05/) 第5章 订单去重:过滤器只分流、不拒单,窗口分代与计数删除 — 第五章规定过滤器只分流、不拒单:查库量减少 0.987,而凭过滤器拒单三十天会拒错 4916 张真实订单;窗口用两代轮换,删除用计数过滤器,普通过滤器清零会让 0.3944 的有效订单读成没见过。 (https://5dlibrary.com/answers/zh/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c05/) - Ch.6 Merging a Hundred Sites, the Intersection Trap and the Privacy of Registers — Chapter 6 merges the sketches of a hundred sites, shows that at a shared share of 0.10 the intersection estimate has a relative error of 0.421, and explains that registers are not anonymous: with few visitors, adding one more person changes the sketch with probability 0.9935. (https://5dlibrary.com/answers/en/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c06/) 第6章 百站合并、交集陷阱与寄存器的隐私 — 第六章合并一百个站点的草图,指出共有比例 0.10 时交集估计的相对误差达 0.421,并说明寄存器不是匿名的:访客稀少时再加一人使草图改变的概率是 0.9935。 (https://5dlibrary.com/answers/zh/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c06/) - Ch.7 Classical Sources East and West: the Registration Rule of the Minor Minister and Laplace's Ratio of Births — Chapter 7 examines the registration rule of the Minor Minister in the Zhou li and Laplace's ratio of births, recomputes Laplace's required count of 771469 people, and notes that the same precision needs only 24576 bytes today. (https://5dlibrary.com/answers/en/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c07/) 第7章 东西方原典考辨:小司徒的比法与拉普拉斯的出生比 — 第七章考辨《周礼》小司徒的比法与拉普拉斯的出生比,复算拉普拉斯所需的 771469 人清点规模,并指出同样精度今天只需 24576 字节。 (https://5dlibrary.com/answers/zh/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c07/) - Ch.8 Closing the Loop: Registers-Only Weekly Reports for Tollbooth and Direct-Answer Pages, and a Pre-Registered Audit — Chapter 8 gives a registers-only weekly procedure for tollbooth and direct-answer pages, the Cardinality Estimator and a pre-registered audit plan: in the demonstration week the union of two pages is estimated at 35420 against an audit of 35818, the spread test needs 70 page-days and the order gate 3463 keys known to be new. Marked boxes separate published evidence, Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c08/) 第8章 实证闭环:收费站与直答页只存寄存器的周报与预注册审计 — 第八章给出收费站与直答页只存寄存器的周报流程、基数估计器与预注册审计方案:示范周两页并集估计 35420 人、审计 35818 人,离散检验需要 70 个页日,订单闸门需要 3463 个已知新键。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/bloom-filter-hyperloglog-unique-visitors-hundred-sites/c08/) ## 5DL-B-295 · Merkle Hash Trees and Trusted Timestamps: Tamper-Evidence for Monograph Versions and Minimal Proofs of Priority 中文书名:默克尔树哈希链与可信时间戳、专著版本防篡改与优先权举证最小证明母典 DOI: pending Cite: Fan, Y. (2026). Merkle Hash Trees and Trusted Timestamps: Tamper-Evidence for Monograph Versions and Minimal Proofs of Priority (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its manuscripts, timestamps, salts and parameters are illustrative, and all calculations are generated by its programs. Chapter 1 makes single-byte edits to a nine-file illustrative manuscript and guesses the edited file from the overall hash, with a hit rate of 0.1133, no better than the 0.1111 of blind guessing. Abstract: Authors often compute one overall hash of a whole manuscript and assume that this both prevents tampering and proves priority. This book refutes two beliefs behind that habit: that one overall hash is enough to keep a manuscript in order, and that a hash plus a timestamp amounts to proof of copyright. Using Merkle trees, hash chains, anchoring and the birthday bound, it gives calculations that can be checked byte by byte. Its manuscripts, timestamps, salts and parameters are illustrative, and all calculations are generated by its programs. A hash and a timestamp show only that some content existed no later than a certain moment and has not been altered; they are not a copyright registration or proof of ownership, their weight as evidence is for the courts and local law, and the book is not individual legal advice. Chapter 1 makes single-byte edits to a nine-file illustrative manuscript and guesses the edited file from the overall hash, with a hit rate of 0.1133, no better than the 0.1111 of blind guessing. Chapter 2 builds trees by the rules of the Internet standards documents, proves a single chapter with a logarithmic number of hashes, and shows two forgeries against home-made trees that lack domain separation or copy the last node. Chapter 3 finds the least number of comparisons needed to locate a difference and reports a result against the main thesis: once edits scatter over more than 0.1670 of the leaves, tree descent does worse than a flat list. Chapter 4 moves from the linking scheme of Haber and Stornetta to anchoring intervals: weekly anchoring gives a mean proof lag of 3.10 days, and with monthly anchors some version can be quietly rewritten on 0.7534 of the days of a year. Chapter 5 computes the birthday bound; an experiment with hashes cut to 24 bits observes a collision share of 0.4000 against a prediction of 0.3934, and a Yuval-style experiment shows that the claimant is the lowest-cost attacker, producing a collision with only 16384 hash calls. Chapter 6 assembles a minimal bundle of evidence: 365 bytes for nine files and six versions, and only 749 bytes for over a thousand files and five hundred versions, with salting and the limits of the law. Chapter 7 examines the matched fish tallies of the New Book of Tang and Hooke's anagram. Chapter 8 gives the procedure for writing the hash tree of each volume's manuscript and programs into its receipt, the Manuscript Merkle Prover and a pre-registered test plan: the revision-width test needs 116 events, and 31 or more wide revisions refute the empirical thesis. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:写专著的人常给整部书稿算一个总哈希,以为这样既防篡改又证明了先后。本书证伪这个做法背后的两个信念:一个总哈希足以管住整部书稿;哈希加时间戳就等于版权证明。书中用默克尔树、哈希链、锚定与生日界给出可以逐字节核对的计算。书中的书稿、时间戳、盐值与参数均为示意,全部计算由书中程序生成;哈希与时间戳只证明某段内容不晚于某时已经存在且未被改动,不等于著作权登记或权属证明,举证效力以法院与当地法规为准;本书不构成个人化法律建议。 第一章在九个文件的示意书稿上做单字节改动,凭总哈希猜改了哪个文件,命中率 0.1133,与瞎猜的 0.1111 无异。第二章按互联网标准文档搭树,单章证明只要对数个哈希,并演示自制树缺少域分隔或复制末节点时的两种伪造。第三章求差异定位的最少比较次数,并发现与主轴相反的结果:改动分散地超过 0.1670 的叶子时,树的下降不如平铺清单。第四章从哈伯与斯托内塔的链接法讲到锚定间隔:每周锚定的平均证明滞后 3.10 天,每月锚定时一年中有 0.7534 的日子存在可被悄悄改写的版本。 第五章计算生日界,截断到 24 比特的实验观察到碰撞比例 0.4000,预测 0.3934,并用尤瓦尔式实验说明主张者本人是成本最低的攻击者,只花 16384 次哈希就造出碰撞。第六章装配最小证明包:九个文件、六个版本时 365 字节,一千多个文件、五百个版本也只有 749 字节,并说明加盐与法规边界。第七章考辨《新唐书·车服志》的鱼符合契与胡克的字谜。第八章给出把每卷书稿与程序哈希树写入回执的流程、书稿默克尔证明器与预注册检验方案:修订幅度检验需要 116 个事件,31 个以上宽修订即推翻经验命题。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Blind Spot of One Fingerprint: an Overall Hash Only Says Whether Something Changed — Chapter 1 makes single-byte edits to a nine-file illustrative manuscript and guesses the edited file from the overall hash, with a hit rate of 0.1133, no better than the 0.1111 of blind guessing. (https://5dlibrary.com/answers/en/merkle-tree-timestamp-monograph-tamper-evidence-priority/c01/) 第1章 一个指纹的盲区:总哈希只能回答「变没变」 — 第一章在九个文件的示意书稿上做单字节改动,凭总哈希猜改了哪个文件,命中率 0.1133,与瞎猜的 0.1111 无异。 (https://5dlibrary.com/answers/zh/merkle-tree-timestamp-monograph-tamper-evidence-priority/c01/) - Ch.2 The Merkle Tree: Construction, Domain Separation and Inclusion Proofs of Logarithmic Length — Under the rules of RFC 6962, leaf hashes carry a zero prefix, interior hashes a one prefix, and n leaves split at the largest power of two below n. The nine-file tree takes 17 hash calls, and editing one chapter changes only 5 nodes. The audit path of Chapter 5 is 4 sibling hashes, 128 bytes; edited content and a misstated position both fail verification, but a misstated tree size may pass, so the leaf count must be anchored with the root. For every tree size listed, the longest path equals the base-two logarithm rounded up, 12 for 4096 leaves. Without domain separation a 64-byte fake "file" can impersonate a node; with last-node copying, 31 of 32 odd sizes let two different manuscripts share one root. With the tree built, the next questions are these: when two drafts each have a tree, what is the least number of hashes that must be compared to find the edit, and is the tree still worth it when many things change? (https://5dlibrary.com/answers/en/merkle-tree-timestamp-monograph-tamper-evidence-priority/c02/) 第2章 默克尔树:构造、域分隔与对数长度的包含证明 — 按 RFC 6962 的规则,叶子哈希加零字节前缀,内部节点哈希加一字节前缀,n 个叶子在小于 n 的最大 2 的幂处分割。九个文件的树调用 17 次哈希,改一章只动 5 个节点。第五章的审计路径是 4 个兄弟哈希、128 字节;改过的内容、错报的位置都通不过验证,但错报的树大小可能通过,所以叶子数必须与根一起锚定。所有列出的树的大小,最长路径都等于以 2 为底的对数向上取整,4096 个叶子是 12。不加域分隔,64 字节的伪「文件」能冒充一个节点;复制末节点,32 个奇数大小中有 31 个会让两部不同的书稿共享一个根。树搭好了,接下来的问题是:两版书稿各有一棵树时,最少要比较多少个哈希才能找到改动?改动很多时,树还划算吗? (https://5dlibrary.com/answers/zh/merkle-tree-timestamp-monograph-tamper-evidence-priority/c02/) - Ch.3 The Least Number of Hashes to Locate a Difference: the Information Bound, Tree Descent and the Break-Even Point — Chapter 3 finds the least number of comparisons needed to locate a difference and reports a result against the main thesis: once edits scatter over more than 0.1670 of the leaves, tree descent does worse than a flat list. (https://5dlibrary.com/answers/en/merkle-tree-timestamp-monograph-tamper-evidence-priority/c03/) 第3章 差异定位的最少哈希数:信息下界、树形下降与分界点 — 第三章求差异定位的最少比较次数,并发现与主轴相反的结果:改动分散地超过 0.1670 的叶子时,树的下降不如平铺清单。 (https://5dlibrary.com/answers/zh/merkle-tree-timestamp-monograph-tamper-evidence-priority/c03/) - Ch.4 Hash Chains and Linked Timestamps: from the Linking Scheme of Haber and Stornetta to Anchoring Intervals — Chapter 4 moves from the linking scheme of Haber and Stornetta to anchoring intervals: weekly anchoring gives a mean proof lag of 3.10 days, and with monthly anchors some version can be quietly rewritten on 0.7534 of the days of a year. (https://5dlibrary.com/answers/en/merkle-tree-timestamp-monograph-tamper-evidence-priority/c04/) 第4章 哈希链与链式时间戳:从哈伯与斯托内塔的链接法到锚定间隔 — 第四章从哈伯与斯托内塔的链接法讲到锚定间隔:每周锚定的平均证明滞后 3.10 天,每月锚定时一年中有 0.7534 的日子存在可被悄悄改写的版本。 (https://5dlibrary.com/answers/zh/merkle-tree-timestamp-monograph-tamper-evidence-priority/c04/) - Ch.5 The Birthday Bound: Collision Probability, Truncated Fingerprints and the Claimant as Adversary — Chapter 5 computes the birthday bound; an experiment with hashes cut to 24 bits observes a collision share of 0.4000 against a prediction of 0.3934, and a Yuval-style experiment shows that the claimant is the lowest-cost attacker, producing a collision with only 16384 hash calls. (https://5dlibrary.com/answers/en/merkle-tree-timestamp-monograph-tamper-evidence-priority/c05/) 第5章 生日界:碰撞概率、截断长度与「主张者本人」这个对手 — 第五章计算生日界,截断到 24 比特的实验观察到碰撞比例 0.4000,预测 0.3934,并用尤瓦尔式实验说明主张者本人是成本最低的攻击者,只花 16384 次哈希就造出碰撞。 (https://5dlibrary.com/answers/zh/merkle-tree-timestamp-monograph-tamper-evidence-priority/c05/) - Ch.6 A Minimal Bundle of Evidence for Priority: Append-Only Consistency, a Two-Level Design and Salted Leaves — Chapter 6 assembles a minimal bundle of evidence: 365 bytes for nine files and six versions, and only 749 bytes for over a thousand files and five hundred versions, with salting and the limits of the law. (https://5dlibrary.com/answers/en/merkle-tree-timestamp-monograph-tamper-evidence-priority/c06/) 第6章 优先权举证的最小证明包:追加一致性、两层结构与盐值遮蔽 — 第六章装配最小证明包:九个文件、六个版本时 365 字节,一千多个文件、五百个版本也只有 749 字节,并说明加盐与法规边界。 (https://5dlibrary.com/answers/zh/merkle-tree-timestamp-monograph-tamper-evidence-priority/c06/) - Ch.7 Classical Sources East and West: Matched Fish Tallies and Hooke's Anagram — The personal fish tallies of the Treatise on Carriages and Dress, two left halves to one right, and the bronze tallies, three or five left halves to one right, used "beginning with the first and going round again", are an ancient practice of split verification and ordered rotation whose stated purpose was to guard against false summonses. Program p71 shows that with 3 halves in rotation a copy matches again at deliveries 5, 8, 11 and 14, while a Lamport one-way chain rejects every replay. Hooke's anagram of 1676 is a commitment: its sorted letters are exactly ceiiinosssttuv, with 1210809600 orderings, about 30.17 bits; but a small list of 46 words already spells 25 readings, so its binding rests on human judgement. The ancients saw rightly that the verifier must hold a credential first and that one should commit before revealing; they were limited by physical credentials that can be copied, rotation that permits replay and anagrams with more than one reading. The final chapter returns to the publishing house's own procedure: how the hash tree of each volume's manuscript and programs is written into its receipt, how the small tool is used, and what data would refute this book. (https://5dlibrary.com/answers/en/merkle-tree-timestamp-monograph-tamper-evidence-priority/c07/) 第7章 东西方原典考辨:鱼符合契与胡克字谜 — 《新唐书·车服志》的随身鱼符「左二右一」,铜鱼符「左三右一」「左五右一」「用始第一,周而复始」,是分持验证与按序轮换的古代做法,目的是「以防召命之诈」。程序 p71 显示,轮换 3 枚时一份复制品在第 5、8、11、14 次传令都能对上,兰波特单向链则拒绝任何重放。胡克 1676 年的字谜是一个承诺:字母排序后正是 ceiiinosssttuv,有 1210809600 种排列,约 30.17 比特;但一张 46 个词的小词表就能拼出 25 种读法,绑定性靠的是人为判断。古人看对了「核验方先持有凭证」与「先承诺后揭示」,受限于物理凭证可仿制、轮换可重放、字谜读法不唯一。最后一章回到出版局自己的流程:每一卷书稿与程序的哈希树怎样写进回执,小工具怎样用,以及什么样的数据会推翻本书。 (https://5dlibrary.com/answers/zh/merkle-tree-timestamp-monograph-tamper-evidence-priority/c07/) - Ch.8 Closing the Loop: Writing the Hash Tree of Manuscripts and Programs into Receipts, the Small Tool and a Pre-Registered Test — Chapter 8 gives the procedure for writing the hash tree of each volume's manuscript and programs into its receipt, the Manuscript Merkle Prover and a pre-registered test plan: the revision-width test needs 116 events, and 31 or more wide revisions refute the empirical thesis. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/merkle-tree-timestamp-monograph-tamper-evidence-priority/c08/) 第8章 实证闭环:书稿与程序哈希树写入回执、小工具与预注册检验 — 第八章给出把每卷书稿与程序哈希树写入回执的流程、书稿默克尔证明器与预注册检验方案:修订幅度检验需要 116 个事件,31 个以上宽修订即推翻经验命题。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/merkle-tree-timestamp-monograph-tamper-evidence-priority/c08/) ## 5DL-B-296 · Differential Privacy, the Laplace Mechanism and Privacy-Budget Accounting: Publishing Business Statistics without Exposing Any Individual 中文书名:差分隐私拉普拉斯机制与隐私预算会计、公开经营统计而不泄露个人的发布母典 DOI: pending Cite: Fan, Y. (2026). Differential Privacy, the Laplace Mechanism and Privacy-Budget Accounting: Publishing Business Statistics without Exposing Any Individual (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 3 deals with sensitivity: revenue clipped per person has a root mean square error of 6545 yuan in the illustrative data, 4.09 % of revenue, and a customer profile spends the budget only once by parallel composition. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: When owners publish figures they usually follow three unwritten rules: deleting names and phone numbers makes data anonymous, publishing only aggregates exposes no one, and a little jitter sized by feel settles any remaining doubt. This book refutes all three and offers, through differential privacy, an alternative that can be recomputed, audited and overturned. Sweeney estimated that about 87 % of Americans are unique on ZIP code, sex and date of birth, and Golle's re-estimate is about 63 %; the book's cell model shows that in a postcode of ten thousand about 0.843 of residents are unique on date of birth and sex. Two means that differ by one person subtract to that person; 48 subset counts with small errors rebuild most records, while a Laplace release whose total budget is nailed at 1 pushes the adversary back to the coin-toss level of 0.511. Its parameters are illustrative; all calculations are generated by its programs; the book is not individual legal or business advice, differential privacy is a mathematical guarantee and no substitute for legal compliance, laws are stated only in outline, and a human legal review before publication prevails. Chapter 2 defines the Laplace mechanism and epsilon: a count at epsilon 1 has a 95 % half-width of about 2.996 people, and no adversary can beat guessing by more than 0.4621. Chapter 3 deals with sensitivity: revenue clipped per person has a root mean square error of 6545 yuan in the illustrative data, 4.09 % of revenue, and a customer profile spends the budget only once by parallel composition. Chapter 4 computes basic, advanced and optimal composition side by side; for a daily release over a year the optimal bound is only 0.283 of the basic one. Chapter 5 covers randomized response and the local model, whose error at the same epsilon is 24.6 times that of the central model. Chapter 6 reads the choice of epsilon through a posterior ceiling and a table of relative errors, and sets out the dispute over the total budget of 19.61 for the redistricting data of the 2020 United States census: at epsilon 1 a count of 3 has a relative half-width of 0.999. Chapter 7 examines "words that touch the hidden matter" in the Han Feizi and Warner's randomized response of 1965 (a spinner probability of 0.8 implies an epsilon of 1.386). Chapter 8 gives a seven-step noisy release procedure for the public statistics of a pay-per-use service, the distilled tool, the Privacy Budget Accountant, an audit method (100 trials catch an implementation with halved noise scale with probability 0.869) and a pre-registered plan for refutation. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:经营者公开数字时常有三条默认的规矩:删掉姓名电话就是匿名,只发汇总数就不会泄露个人,不放心就凭感觉加一点扰动。本书证伪这三条规矩,并用差分隐私给出一个可以被重算、被审计、被推翻的替代方案。斯威尼估算约 87 % 的美国人仅凭邮编、性别与出生日期即可唯一识别,戈尔的重新估算约为 63 %;书中的分格模型显示,一万人的邮编区里按出生日期与性别唯一的居民约占 0.843。两份只差一人的均值相减就是那个人;48 个带小误差的子集计数可以重构绝大部分明细,而把总预算钉在 1 的拉普拉斯发布让对手退回到 0.511 的抛硬币水平。书中参数为示意;全部计算由书中程序生成;本书是非个人化的法律与经营建议,差分隐私是数学保证,不替代法律合规,法规只述大意,以上架前人工复核为准。 第二章给出拉普拉斯机制与 ε 的定义:计数查询取 ε 为 1 时 95 % 误差半宽约 2.996 人,任何对手比瞎猜最多好 0.4621。第三章处理敏感度:营收按人截断,示意数据中均方根误差 6545 元,占营收 4.09 %;客群分布按并行组合只花一次预算。第四章并排计算基本组合、高级组合与最优组合,每天发布一年时最优组合只有基本组合的 0.283。第五章讨论随机化回答与本地模型,同一 ε 下本地模型的误差是中心模型的 24.6 倍。 第六章用后验上限与相对误差两张表读 ε 的取值,并陈美国 2020 年人口普查选区重划数据总预算 19.61 引发的争论:计数为 3 时 ε 为 1 的相对半宽是 0.999。第七章考辨《韩非子·说难》「语及所匿之事」与沃纳 1965 年的随机化回答(转盘概率 0.8 对应 ε 为 1.386)。第八章给出收费站公开统计的七步加噪发布规程、蒸馏小工具「隐私预算会计器」、审计方法(100 次试验即可以 0.869 的把握查出噪声尺度减半的实现)与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 "Anonymous" Is Not Anonymous: Quasi-Identifiers, Differencing and Reconstruction — Deleting names does not make data anonymous: in a postcode of ten thousand, about 0.843 of residents are unique on date of birth and sex, and only past forty thousand does that fall to one half; Sweeney and Golle report national shares near 87 % and 63 %, by different methods, with the same verdict of "most". Releasing only aggregates is not safe either: two means that differ by one member subtract to the newcomer's 1680.00 yuan. Small errors do not stop many questions: with twelve members and 48 subset counts carrying errors of one unit either way, the adversary still gets 0.919 of the bits right. Only a Laplace release whose total budget is nailed at 1 pushes him back to the coin-toss level of 0.511. The next chapter answers a direct question: what exactly is this "budget", and what does an epsilon of 1 actually mean? (https://5dlibrary.com/answers/en/differential-privacy-laplace-budget-business-statistics/c01/) 第1章 「匿名」不匿名:准标识符、差分攻击与明细重构 — 去掉名字不等于匿名:在一万人的邮编区里,按出生日期与性别唯一的居民约占 0.843,要到四万多人才降到一半;斯威尼与戈尔分别报告 87 % 与 63 % 左右的全国比例,方法不同,结论都是「大多数」。只发汇总也不等于安全:两个只差一人的均值相减,得到的正是新会员的 1680.00 元。小误差挡不住许多次提问:12 位会员、48 个带正负 1 误差的子集计数,对手仍能猜对 0.919。把总预算钉在 1 的拉普拉斯发布,才让对手退回到抛硬币的 0.511。下一章回答一个直接的问题:这个「预算」究竟是什么,ε 等于 1 到底意味着什么? (https://5dlibrary.com/answers/zh/differential-privacy-laplace-budget-business-statistics/c01/) - Ch.2 The Laplace Mechanism: Epsilon Is a Ceiling on the Likelihood Ratio — Chapter 2 defines the Laplace mechanism and epsilon: a count at epsilon 1 has a 95 % half-width of about 2.996 people, and no adversary can beat guessing by more than 0.4621. (https://5dlibrary.com/answers/en/differential-privacy-laplace-budget-business-statistics/c02/) 第2章 拉普拉斯机制:ε 是似然比的天花板 — 第二章给出拉普拉斯机制与 ε 的定义:计数查询取 ε 为 1 时 95 % 误差半宽约 2.996 人,任何对手比瞎猜最多好 0.4621。 (https://5dlibrary.com/answers/zh/differential-privacy-laplace-budget-business-statistics/c02/) - Ch.3 Sensitivity: Clip Revenue First, and Spend the Budget Once on a Customer Profile — Chapter 3 deals with sensitivity: revenue clipped per person has a root mean square error of 6545 yuan in the illustrative data, 4.09 % of revenue, and a customer profile spends the budget only once by parallel composition. (https://5dlibrary.com/answers/en/differential-privacy-laplace-budget-business-statistics/c03/) 第3章 敏感度:营收先截断,客群分布只花一次预算 — 第三章处理敏感度:营收按人截断,示意数据中均方根误差 6545 元,占营收 4.09 %;客群分布按并行组合只花一次预算。 (https://5dlibrary.com/answers/zh/differential-privacy-laplace-budget-business-statistics/c03/) - Ch.4 Composition and Budget Accounting: Basic, Advanced and Optimal Books — Chapter 4 computes basic, advanced and optimal composition side by side; for a daily release over a year the optimal bound is only 0.283 of the basic one. (https://5dlibrary.com/answers/en/differential-privacy-laplace-budget-business-statistics/c04/) 第4章 组合定理与预算会计:基本、高级与最优三种账法 — 第四章并排计算基本组合、高级组合与最优组合,每天发布一年时最优组合只有基本组合的 0.283。 (https://5dlibrary.com/answers/zh/differential-privacy-laplace-budget-business-statistics/c04/) - Ch.5 Randomized Response and the Local Model: Letting Customers Add the Noise — Chapter 5 covers randomized response and the local model, whose error at the same epsilon is 24.6 times that of the central model. (https://5dlibrary.com/answers/en/differential-privacy-laplace-budget-business-statistics/c05/) 第5章 随机化回答与本地模型:把噪声交给顾客自己加 — 第五章讨论随机化回答与本地模型,同一 ε 下本地模型的误差是中心模型的 24.6 倍。 (https://5dlibrary.com/answers/zh/differential-privacy-laplace-budget-business-statistics/c05/) - Ch.6 How Large Should Epsilon Be: the Dispute over the 2020 United States Census and the Price of Small Numbers — Chapter 6 reads the choice of epsilon through a posterior ceiling and a table of relative errors, and sets out the dispute over the total budget of 19.61 for the redistricting data of the 2020 United States census: at epsilon 1 a count of 3 has a relative half-width of 0.999. (https://5dlibrary.com/answers/en/differential-privacy-laplace-budget-business-statistics/c06/) 第6章 ε 取多少:2020 年美国人口普查之争与小数字的代价 — 第六章用后验上限与相对误差两张表读 ε 的取值,并陈美国 2020 年人口普查选区重划数据总预算 19.61 引发的争论:计数为 3 时 ε 为 1 的相对半宽是 0.999。 (https://5dlibrary.com/answers/zh/differential-privacy-laplace-budget-business-statistics/c06/) - Ch.7 Classical Sources East and West: Han Feizi on the Difficulty of Persuasion and Warner's Randomized Response — Chapter 7 examines "words that touch the hidden matter" in the Han Feizi and Warner's randomized response of 1965 (a spinner probability of 0.8 implies an epsilon of 1.386). (https://5dlibrary.com/answers/en/differential-privacy-laplace-budget-business-statistics/c07/) 第7章 东西方原典考辨:韩非《说难》与沃纳的随机化回答 — 第七章考辨《韩非子·说难》「语及所匿之事」与沃纳 1965 年的随机化回答(转盘概率 0.8 对应 ε 为 1.386)。 (https://5dlibrary.com/answers/zh/differential-privacy-laplace-budget-business-statistics/c07/) - Ch.8 A Noisy Release Procedure for the Public Statistics of a Pay-per-Use Service, and a Plan for Refutation — Chapter 8 gives a seven-step noisy release procedure for the public statistics of a pay-per-use service, the distilled tool, the Privacy Budget Accountant, an audit method (100 trials catch an implementation with halved noise scale with probability 0.869) and a pre-registered plan for refutation. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/differential-privacy-laplace-budget-business-statistics/c08/) 第8章 收费站公开统计加噪发布规程与证伪方案 — 第八章给出收费站公开统计的七步加噪发布规程、蒸馏小工具「隐私预算会计器」、审计方法(100 次试验即可以 0.869 的把握查出噪声尺度减半的实现)与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/differential-privacy-laplace-budget-business-statistics/c08/) ## 5DL-B-297 · The Algebra of Check Digits and Verhoeff's Dihedral-Group Code: Detection Rates for Single-Digit Slips and Adjacent Transpositions in Order, Remittance and Book Numbers 中文书名:校验码代数与二面体群韦尔霍夫算法、订单号汇款码书号单字误抄与相邻换位检出率母典 DOI: pending Cite: Fan, Y. (2026). The Algebra of Check Digits and Verhoeff's Dihedral-Group Code: Detection Rates for Single-Digit Slips and Adjacent Transpositions in Order, Remittance and Book Numbers (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Entry volumes, error rates and costs are illustrative; all calculations are generated by the book's programs; the book is not individual business advice. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: When operators add check digits to order numbers, remittance references and internal book numbers, they often borrow Luhn from payment cards or the 1 and 3 weights of product barcodes, believing that any scheme will do. This book refutes that belief by putting the spectrum of human copying errors and the exhaustive detection rates of each scheme on one table. Human errors are mostly single-digit errors and adjacent transpositions, together 0.8926 of Verhoeff's 1969 counts; no scheme built on addition modulo 10 can catch every error of both classes; and the code Verhoeff built from D5, the symmetry group of the pentagon, was the first to do so. Entry volumes, error rates and costs are illustrative; all calculations are generated by the book's programs; the book is not individual business advice. A check digit guards only against miscopying, is not anti-counterfeiting or encryption, and must never serve as a security credential; every number in the book is an obviously artificial test value. Chapter 1 lays out the error spectrum and the exhaustive ledger of seven schemes: weighted by the spectrum, Luhn lets through 117.4 named errors per ten thousand and Verhoeff 18.3. Chapter 2 opens up Luhn's doubling: it misses 09 and 90, but jump transpositions make up 0.696 of its misses. Chapter 3 shows the power of modulus 11 and the cost of its X (0.0909 of numbers need one), and that 13-digit book numbers miss the 0.1111 of swaps differing by 5. Chapter 4 builds D5 from a pentagon and proves by enumeration that no anti-symmetric permutation exists modulo 10, while D5 has 34040. Chapter 5 adds the nameless errors: moving from Luhn to Verhoeff lowers whole-spectrum misses from 207.2 to 107.1, roughly half, and two MOD 97-10 characters reach 6.8. Chapter 6 turns detection rates into a shop's monthly accounts, gives formats for order numbers, remittance references and internal book numbers, and shows that every number invented by a forger who knows the rule is valid. Chapter 7 examines the Rites of Zhou, with great covenants on ritual bronzes and lesser ones on red tablets, and Gauss's congruences. Chapter 8 gives an order desk procedure, the Order Check-Digit Generator and a pre-registered test plan: the one-sided test needs 50 audited discrepancies. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:经营者给订单号、汇款识别码和内部书号加校验位,常照搬信用卡用的 Luhn 或商品条码用的 1、3 权重,并相信哪一种都差不多。本书把人工抄错号码的错误谱与各方案的穷举检出率放在同一张表上,证伪这一信念:人工错误以单字错误与相邻换位为主,韦尔霍夫 1969 年的计数中二者合计 0.8926;模 10 的加法方案不可能同时拦下这两类的全部;韦尔霍夫借五边形对称群 D5 构造的校验码第一次做到了。书中录入量、出错率与成本均为示意参数;全部计算由书中程序生成;本书是非个人化经营建议。校验位只防误抄,不是防伪或加密,不得当作安全凭证;书中号码全是明显的测试值。 第一章摆出错误谱与七种方案的穷举总账:按错误谱加权,有名字的错误每万条 Luhn 漏 117.4 条,韦尔霍夫漏 18.3 条。第二章拆开 Luhn 的加倍:它漏掉 09 与 90,但漏检中跳位换位占 0.696。第三章说明模 11 的威力与 X 的代价(0.0909 的号码需要 X),以及 13 位书号漏掉相差 5 的 0.1111 换位。第四章从五边形造出 D5,穷举证明模 10 里没有反对称置换,而 D5 有 34040 个。第五章补上无名错误:从 Luhn 换到韦尔霍夫,全谱漏检从 207.2 降到 107.1,约减半,两位 MOD 97-10 为 6.8。 第六章把检出率折算成一家店的月度账,给出订单号、汇款识别码与内部书号的格式,并演示知道规则的伪造者编造的号码全部合法。第七章考辨《周礼》「大约剂书于宗彝,小约剂书于丹图」与高斯的同余。第八章给出订单台流程、订单校验位生成器与预注册检验方案:单侧检验需 50 条经审计的差错。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Spectrum of Slips: Two Classes in Twelve Thousand Errors — Chapter 1 lays out the error spectrum and the exhaustive ledger of seven schemes: weighted by the spectrum, Luhn lets through 117.4 named errors per ten thousand and Verhoeff 18.3. (https://5dlibrary.com/answers/en/check-digit-algebra-verhoeff-dihedral-transposition-detection/c01/) 第1章 抄错的谱:一万二千条错误里的两大类 — 第一章摆出错误谱与七种方案的穷举总账:按错误谱加权,有名字的错误每万条 Luhn 漏 117.4 条,韦尔霍夫漏 18.3 条。 (https://5dlibrary.com/answers/zh/check-digit-algebra-verhoeff-dihedral-transposition-detection/c01/) - Ch.2 Luhn's Doubling: Why Mod 10 Misses 09 and 90 — Chapter 2 opens up Luhn's doubling: it misses 09 and 90, but jump transpositions make up 0.696 of its misses. (https://5dlibrary.com/answers/en/check-digit-algebra-verhoeff-dihedral-transposition-detection/c02/) 第2章 Luhn 的加倍:模 10 为什么漏掉 09 与 90 — 第二章拆开 Luhn 的加倍:它漏掉 09 与 90,但漏检中跳位换位占 0.696。 (https://5dlibrary.com/answers/zh/check-digit-algebra-verhoeff-dihedral-transposition-detection/c02/) - Ch.3 Prime Moduli and Weights: The X of ISBN-10 and the Blind Spot of EAN-13 — Chapter 3 shows the power of modulus 11 and the cost of its X (0.0909 of numbers need one), and that 13-digit book numbers miss the 0.1111 of swaps differing by 5. (https://5dlibrary.com/answers/en/check-digit-algebra-verhoeff-dihedral-transposition-detection/c03/) 第3章 素数模与权重:ISBN-10 的 X 与 EAN-13 的盲点 — 第三章说明模 11 的威力与 X 的代价(0.0909 的号码需要 X),以及 13 位书号漏掉相差 5 的 0.1111 换位。 (https://5dlibrary.com/answers/zh/check-digit-algebra-verhoeff-dihedral-transposition-detection/c03/) - Ch.4 The Symmetries of a Pentagon: The Dihedral Group D5 and Anti-Symmetric Permutations — Chapter 4 builds D5 from a pentagon and proves by enumeration that no anti-symmetric permutation exists modulo 10, while D5 has 34040. (https://5dlibrary.com/answers/en/check-digit-algebra-verhoeff-dihedral-transposition-detection/c04/) 第4章 五边形的对称:二面体群 D5 与反对称置换 — 第四章从五边形造出 D5,穷举证明模 10 里没有反对称置换,而 D5 有 34040 个。 (https://5dlibrary.com/answers/zh/check-digit-algebra-verhoeff-dihedral-transposition-detection/c04/) - Ch.5 Verhoeff's Algorithm Dissected: Positions, Twins, Jumps and the Errors with No Name — Chapter 5 adds the nameless errors: moving from Luhn to Verhoeff lowers whole-spectrum misses from 207.2 to 107.1, roughly half, and two MOD 97-10 characters reach 6.8. (https://5dlibrary.com/answers/en/check-digit-algebra-verhoeff-dihedral-transposition-detection/c05/) 第5章 韦尔霍夫算法拆解:位置、孪生、跳位与没有名字的错误 — 第五章补上无名错误:从 Luhn 换到韦尔霍夫,全谱漏检从 207.2 降到 107.1,约减半,两位 MOD 97-10 为 6.8。 (https://5dlibrary.com/answers/zh/check-digit-algebra-verhoeff-dihedral-transposition-detection/c05/) - Ch.6 Order Numbers, Remittance References and Internal Book Numbers: A Month's Accounts — With illustrative parameters of 20,000 manual entries a month, an error rate of 0.00625, 2 yuan per stopped slip and 150 yuan per missed one, a shop with no check loses 18750 yuan a month; EAN-style weights cost 794, Luhn 633, Verhoeff 448, Damm 446 and two-character MOD 97-10 263. As the share of transpositions rises from 0.05 to 0.30, EAN's misses rise from 236.4 to 514.2 while Verhoeff stays at 107.1. The simulated clerk agrees with enumeration; all 1000 of a forger's invented numbers are valid, and blind guesses pass at 0.0995, so a check digit guards only against miscopying and is no security credential. Order numbers, remittance references and internal book numbers use one Verhoeff digit, two MOD 97-10 characters and one Damm digit respectively. The next chapter returns to the classics: the Rites of Zhou put great contracts on ritual bronzes and small ones on red tablets, Gauss wrote down congruence, and each got something right. (https://5dlibrary.com/answers/en/check-digit-algebra-verhoeff-dihedral-transposition-detection/c06/) 第6章 订单号、汇款识别码与内部书号:一个月的账 — 按示意参数,每月 20,000 次人工录入、出错率 0.00625、拦下一笔 2 元、漏过一笔 150 元:没有校验位每月损失 18750 元,EAN 式 794 元,Luhn 633 元,韦尔霍夫 448 元,达姆 446 元,两位 MOD 97-10 为 263 元。换位占比从 0.05 升到 0.30,EAN 的漏检从 236.4 升到 514.2,韦尔霍夫保持 107.1。模拟录入员与穷举相符;伪造者 1000 个编造的号码全部合法,盲猜也有 0.0995 通过,校验位只防误抄,不是安全凭证。订单号、汇款识别码与内部书号分别用韦尔霍夫一位、MOD 97-10 两位与达姆一位。下一章回到原典:《周礼》把大约剂书于宗彝、小约剂书于丹图,高斯写下同余,它们各自说对了什么。 (https://5dlibrary.com/answers/zh/check-digit-algebra-verhoeff-dihedral-transposition-detection/c06/) - Ch.7 Classics East and West: Ritual Bronzes, Red Tablets and Gauss's Congruences — The Overseer of Covenants in the Autumn Offices of the Rites of Zhou wrote great covenants on ritual bronzes and lesser ones on red tablets, a principle of matching protection to the stakes. Program p71's illustrative model shows that, with a slip chance of 0.00499 for a ten-character number, one Verhoeff digit is the lowest-cost choice for small orders (21.18 yuan per thousand), two MOD 97-10 characters are the lowest-cost choice for large orders and remittances (42.44 and 72.55 yuan), and entering numbers twice never pays; but a bronze guards against tampering and loss while a check digit guards against slips, and neither replaces the other. Gauss's congruence is the foundation of the book's arithmetic: over all four-digit amounts, casting out nines detects swaps at 0.0000 and the alternating sum modulo 11 at 1.0000, and every swap changes an amount by a multiple of 9, as 4710 written as 4170 differs by 540. The next chapter turns the whole book into an order desk procedure, a small tool and a pre-registered test plan. (https://5dlibrary.com/answers/en/check-digit-algebra-verhoeff-dihedral-transposition-detection/c07/) 第7章 东西方原典考辨:宗彝丹图与高斯同余 — 《周礼·秋官》的司约把大约剂书于宗彝、小约剂书于丹图,是「按利害分配保护」的原则。程序 p71 的示意模型表明:十位号码出错率 0.00499 时,小额订单最省的是一位韦尔霍夫码(每千个 21.18 元),大额订单与汇款最省的是两位 MOD 97-10(42.44 元与 72.55 元),整号录两遍从来不划算;但宗彝防的是篡改与遗失,校验位防的是手误,二者不可互替。高斯的同余是全书算术的底座:在全部四位数金额上,弃九法对换位的检出率为 0.0000,模 11 交错和为 1.0000,任何换位都使金额变化 9 的倍数,例如 4710 写成 4170 相差 540。下一章把全书落成一套订单台流程、一个小工具和一份预注册检验方案。 (https://5dlibrary.com/answers/zh/check-digit-algebra-verhoeff-dihedral-transposition-detection/c07/) - Ch.8 Closing the Loop: An Order Desk Procedure, the Order Check-Digit Generator and a Pre-Registered Test — Chapter 8 gives an order desk procedure, the Order Check-Digit Generator and a pre-registered test plan: the one-sided test needs 50 audited discrepancies. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/check-digit-algebra-verhoeff-dihedral-transposition-detection/c08/) 第8章 实证闭环:订单台流程、订单校验位生成器与预注册检验 — 第八章给出订单台流程、订单校验位生成器与预注册检验方案:单侧检验需 50 条经审计的差错。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/check-digit-algebra-verhoeff-dihedral-transposition-detection/c08/) ## 5DL-B-298 · Two-Opt Travelling-Salesman Search and Vehicle Routing with Time Windows: Shortest Shopping Loops in Tokyo, Osaka and Seoul under Counter Opening Hours 中文书名:旅行商二元交换与时间窗车辆路径、东京大阪首尔扫货路线最短闭环与专柜开门时间窗母典 DOI: pending Cite: Fan, Y. (2026). Two-Opt Travelling-Salesman Search and Vehicle Routing with Time Windows: Shortest Shopping Loops in Tokyo, Osaka and Seoul under Counter Opening Hours (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 7 examines Shen Kuo's account of Ding Wei's "one move, three jobs" (an illustrative account saves 0.826) and Euler's seven bridges. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: Buyers who shop in Tokyo, Osaka and Seoul for stores and private clients most often plan their routes nearest first. This book refutes two beliefs behind that habit: that a nearest-first route cannot be much longer than the shortest loop, and that a day's count of items depends mainly on how short the walk is. Using nearest neighbour, pairwise exchange (2-opt), segment moves (Or-opt) and Held-Karp dynamic programming for the travelling-salesman problem, it measures the gap in distance; using vehicle routing with time windows, it measures how opening times, limited-release slots and queues affect the number of items. Its coordinates, counters, opening slots, queue times and purchase caps are illustrative and correspond to no real store; all calculations are generated by its programs; the book is not individual business advice, goods carried across borders must be declared truthfully under each customs authority's rules and purchase limits respected, and the book discusses no way of getting around them. Chapter 1 checks and corrects the project blueprint: the literature's "about a quarter above" for nearest neighbour is an excess over the Held-Karp lower bound on large instances; at ten stops nearest neighbour is only 0.092 longer than the exact optimum on average, and only at sixty stops does it reach 0.2501 above the bound. Chapters 2 and 3 open up pairwise exchange and segment moves, and show why the symmetric gain formula gets nearly half its exchanges wrong on one-way streets. Chapter 4 gives the state counts of exact Held-Karp solutions and separates "above the bound" from "above the optimum". Chapter 5 is the turning point: on an illustrative Tokyo day the shortest loop of 20.99 km collects only 12 items, while the plan with the most items walks 0.32 km more and collects all 15; over random days, switching from nearest neighbour to the shortest loop gains only 0.03 items a day, while sequencing by time windows gains 2.93. Chapter 6 deals with two buyers sharing the work and with slack for random queues: planning on average times completes the list on 0.740 of days, and a fifth of slack raises that to 0.915. Chapter 7 examines Shen Kuo's account of Ding Wei's "one move, three jobs" (an illustrative account saves 0.826) and Euler's seven bridges. Chapter 8 gives the buyer-day procedure, the Shopping Route Planner and a pre-registered test plan: the distance test needs about 37 buyer days and the clock test about 55 days per group. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:替门店与私域客户在东京、大阪、首尔扫货的买手,最常见的排线办法是由近及远。本书证伪这个做法背后的两个信念:由近及远的路线不会比最短闭环差多少;一天的件数主要取决于路走得够不够短。书中用旅行商问题的最近邻、二元交换(2-opt)、段移位(Or-opt)与 Held-Karp 动态规划量出距离上的差距,再用带时间窗的车辆路径模型量出专柜开门、限量时段与排队对件数的影响。书中的坐标、专柜、营业时段、排队时长与限购件数均为示意,不对应任何真实店铺;全部计算由书中程序生成;本书是非个人化经营建议,跨境所购物品须按各地海关规定如实申报并遵守限购规定,书中不讨论任何规避做法。 第一章核实并修正立项蓝图:文献里最近邻「约高出四分之一」是大规模实例上相对于 Held-Karp 下界的超出;在十站规模上,最近邻只比精确最优平均长 0.092,到六十站才比下界高 0.2501。第二、三章摊开二元交换与段移位,并说明单行道下对称增益公式为什么会把近一半的交换做错。第四章给出 Held-Karp 精确解的状态数与「高出下界」和「高出最优」的区别。 第五章是全书的转折:示意东京日上,20.99 公里的最短闭环只买到 12 件,件数最多的方案多走 0.32 公里就买齐 15 件;随机日上,把最近邻换成最短闭环每天只多买 0.03 件,按时间窗排序多买 2.93 件。第六章处理两位买手分工与随机排队的余量:按平均时长排,0.740 的日子全部买到,留两成余量升到 0.915。第七章考辨沈括记丁谓「一举而三役济」(示意账本省下 0.826)与欧拉的七桥。第八章给出买手日流程、扫货路线规划器与预注册检验方案:距离检验约需 37 个买手日,时钟检验每组约需 55 天。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Price of Nearest First: How Much Longer than the Shortest Loop — Chapter 1 checks and corrects the project blueprint: the literature's "about a quarter above" for nearest neighbour is an excess over the Held-Karp lower bound on large instances; at ten stops nearest neighbour is only 0.092 longer than the exact optimum on average, and only at sixty stops does it reach 0.2501 above the bound. (https://5dlibrary.com/answers/en/two-opt-tsp-time-windows-buyer-shopping-routes/c01/) 第1章 由近及远的代价:最近邻路线比最短闭环长多少 — 第一章核实并修正立项蓝图:文献里最近邻「约高出四分之一」是大规模实例上相对于 Held-Karp 下界的超出;在十站规模上,最近邻只比精确最优平均长 0.092,到六十站才比下界高 0.2501。 (https://5dlibrary.com/answers/zh/two-opt-tsp-time-windows-buyer-shopping-routes/c01/) - Ch.2 Croes's Pairwise Exchange: Reversing a Stretch, Crossings and Local Optima — Pairwise exchange removes two legs, adds two, walks the stretch between them in reverse, and accepts the change whenever the total falls. On the illustrative Tokyo day, five exchanges cut 25.51 km down to 20.99 km, with gains of 1.52, 0.06, 0.99, 0.10 and 1.85 km in turn. A route with crossings in the plane can always be shortened, but a route without crossings can still be bad; nearest neighbour is an example with zero crossings and a fifth too long. On 200 random days of ten counters, pairwise exchange from the nearest-neighbour start exceeds the optimum by only 0.0093 on average, and the best of 20 random starts lands on the optimum on a share of 0.990 of days. If the walk from $a$ to $b$ is not as long as the walk from $b$ to $a$, or a counter has to be moved as a whole, is pairwise exchange still enough? (https://5dlibrary.com/answers/en/two-opt-tsp-time-windows-buyer-shopping-routes/c02/) 第2章 Croes 的二元交换:反转一段路、交叉与局部最优 — 二元交换拆掉两段腿、换上两段、把中间一段倒过来走,只要总长下降就接受。东京示意日上,五步交换把 25.51 公里缩到 20.99 公里,增益依次是 1.52、0.06、0.99、0.10 和 1.85 公里。平面上交叉的路线一定可以改短,但没有交叉的路线可以很差,最近邻就是一个零交叉却长两成的例子。在 200 张随机的十站日上,最近邻起点的二元交换平均只比最优长 0.0093,20 个随机起点取最好者有 0.990 的日子落在最优上。如果从 $a$ 到 $b$ 与从 $b$ 到 $a$ 不一样长,或者需要把一家专柜整体挪个位置,二元交换还够用吗? (https://5dlibrary.com/answers/zh/two-opt-tsp-time-windows-buyer-shopping-routes/c02/) - Ch.3 Segment Moves and One-Way Streets: the Or-opt Neighbourhood and Asymmetric Travel — Segment moves lift one to three consecutive counters as a block and put them elsewhere, supplying the kind of improvement pairwise exchange cannot make. On the illustrative Osaka day, pairwise exchange stops at 23.41 km, 0.070 above the optimum; four segment moves reach 21.93 km, only 0.002 above the exact optimum of 21.89 km. On 300 days of ten counters, alternating the two moves gives a mean excess of only 0.0009 and a worst day of 0.042. When travel is asymmetric, the symmetric gain formula gets nearly half its exchanges wrong: naive pairwise exchange made 3859 exchanges of which 1893 slowed the route, and forward segment moves avoid that trap. At ten or twelve stops, then, can we skip these approximations altogether and compute the exact shortest loop directly? (https://5dlibrary.com/answers/en/two-opt-tsp-time-windows-buyer-shopping-routes/c03/) 第3章 段移位与单行道:Or-opt 邻域与非对称行程 — 段移位把一到三家连续的专柜整段挪到别处,补上了二元交换做不到的那一类改进。示意大阪日上,二元交换停在 23.41 公里、超出 0.070,段移位四步走到 21.93 公里,只比精确最优的 21.89 公里长 0.002。300 张十站日上,两者交替使用平均只超出 0.0009,最坏一天 0.042。行程不对称时,对称增益公式会把近一半的交换做错,朴素二元交换在 3859 次交换中有 1893 次让路线更慢;只做正向段移位就能避开这个陷阱。那么,在十站、十二站的规模上,能不能干脆不用这些近似办法,直接求出精确最短闭环? (https://5dlibrary.com/answers/zh/two-opt-tsp-time-windows-buyer-shopping-routes/c03/) - Ch.4 Held-Karp Dynamic Programming: Exact Solutions, State Counts and "Above the Bound" — Held-Karp dynamic programming remembers only "where I have been and where I stand", reducing the factorial of brute force to the order of $n^2 2n$. For the nine-stop loop of the illustrative Seoul day, the dynamic programme uses 9216 extensions and brute force 181440 loops, and both arrive at 23.2337 km. At twelve stops it has only 24576 states, at twenty 10485760, and at twenty-five it needs 3355.4 megabytes; for a day of up to fifteen stops the exact solution is at hand. The literature's "about a quarter above" measures the large-scale excess over the Held-Karp bound, and the program reproduces 0.2501 at sixty stops; at twelve stops bound and optimum nearly coincide and nearest neighbour is only 0.1027 above the optimum. With the exact shortest loop in hand, the next question follows: can that shortest loop be walked within the hours when the counters are open? (https://5dlibrary.com/answers/en/two-opt-tsp-time-windows-buyer-shopping-routes/c04/) 第4章 Held-Karp 动态规划:精确解、状态数与「高出下界」 — Held-Karp 动态规划只记「去过哪些、停在哪里」,把穷举的阶乘降到 $n^2 2n$ 量级。示意首尔日的九站闭环,动态规划用 9216 次扩展、穷举用 181440 条闭环,得到同一个 23.2337 公里。十二站时它只有 24576 个状态,二十站时 10485760 个,二十五站时内存要 3355.4 兆字节;一天十五站以内,精确解随手可得。文献里的「高出约四分之一」量的是大规模下相对于 Held-Karp 下界的超出,六十站时程序复现了 0.2501;十二站时下界与最优几乎重合,最近邻只比最优高 0.1027。有了精确最短,下一个问题就来了:最短的那一圈,能不能在专柜开门的时刻里走完? (https://5dlibrary.com/answers/zh/two-opt-tsp-time-windows-buyer-shopping-routes/c04/) - Ch.5 The Counters' Clocks: Time-Window Feasibility and the Day with the Most Items — Chapter 5 is the turning point: on an illustrative Tokyo day the shortest loop of 20.99 km collects only 12 items, while the plan with the most items walks 0.32 km more and collects all 15; over random days, switching from nearest neighbour to the shortest loop gains only 0.03 items a day, while sequencing by time windows gains 2.93. (https://5dlibrary.com/answers/en/two-opt-tsp-time-windows-buyer-shopping-routes/c05/) 第5章 专柜的时钟:时间窗可行性与件数最多的一天 — 第五章是全书的转折:示意东京日上,20.99 公里的最短闭环只买到 12 件,件数最多的方案多走 0.32 公里就买齐 15 件;随机日上,把最近邻换成最短闭环每天只多买 0.03 件,按时间窗排序多买 2.93 件。 (https://5dlibrary.com/answers/zh/two-opt-tsp-time-windows-buyer-shopping-routes/c05/) - Ch.6 Two Buyers and Random Queues: Vehicle Routing with Time Windows and Planning Slack — Chapter 6 deals with two buyers sharing the work and with slack for random queues: planning on average times completes the list on 0.740 of days, and a fifth of slack raises that to 0.915. (https://5dlibrary.com/answers/en/two-opt-tsp-time-windows-buyer-shopping-routes/c06/) 第6章 两位买手与随机排队:带时间窗的车辆路径与计划余量 — 第六章处理两位买手分工与随机排队的余量:按平均时长排,0.740 的日子全部买到,留两成余量升到 0.915。 (https://5dlibrary.com/answers/zh/two-opt-tsp-time-windows-buyer-shopping-routes/c06/) - Ch.7 Classical Sources East and West: Ding Wei's "One Move, Three Jobs" and Euler's Seven Bridges — Chapter 7 examines Shen Kuo's account of Ding Wei's "one move, three jobs" (an illustrative account saves 0.826) and Euler's seven bridges. (https://5dlibrary.com/answers/en/two-opt-tsp-time-windows-buyer-shopping-routes/c07/) 第7章 东西方原典考辨:丁谓「一举而三役济」与欧拉的七桥 — 第七章考辨沈括记丁谓「一举而三役济」(示意账本省下 0.826)与欧拉的七桥。 (https://5dlibrary.com/answers/zh/two-opt-tsp-time-windows-buyer-shopping-routes/c07/) - Ch.8 Closing the Loop: the Shopping Route Planner and a Pre-Registered Test Plan — Chapter 8 gives the buyer-day procedure, the Shopping Route Planner and a pre-registered test plan: the distance test needs about 37 buyer days and the clock test about 55 days per group. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/two-opt-tsp-time-windows-buyer-shopping-routes/c08/) 第8章 实证闭环:扫货路线规划器与预注册检验方案 — 第八章给出买手日流程、扫货路线规划器与预注册检验方案:距离检验约需 37 个买手日,时钟检验每组约需 55 天。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/two-opt-tsp-time-windows-buyer-shopping-routes/c08/) ## 5DL-B-299 · Gale-Shapley Stable Matching: Pairings without Blocking Pairs for Buyers and Clients, Creators and Brands, Photographers and Slots 中文书名:稳定匹配盖尔沙普利算法、买手客户达人品牌摄影师档期的无私奔配对母典 DOI: pending Cite: Fan, Y. (2026). Gale-Shapley Stable Matching: Pairings without Blocking Pairs for Buyers and Clients, Creators and Brands, Photographers and Slots (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 8 gives the weekly procedure, the Stable Pairing Matcher and a pre-registered test: if cancellations fall from 0.15 to 0.08, each group needs 325 pairings, with a simulated power of 0.803. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: When a boutique assigns personal buyers to booked clients, a studio schedules photographers' slots for brands, or a platform pairs creators with brands, the usual procedures are first come, first served or the highest score picks first. This book refutes the belief behind them: that a fair procedure carried out conscientiously yields a sheet that holds. Both procedures ask only one side what it wants and leave blocking pairs, two people who each prefer the other to their assigned partner; the deferred-acceptance algorithm of Gale and Shapley (1962) always yields a sheet without blocking pairs and is optimal for the proposing side. Its people, slots and preferences are illustrative; all calculations are generated by its programs; it offers non-individualised business guidance, uses only preferences the parties themselves state, involves no profiling without consent, and treats every pairing as a suggestion either side may decline. Chapter 1 finds in an illustrative sweep that the more tastes agree, the worse first come does, with blocking pairs rising from 4.14 to 14.12 on average, while score first is stable only when all buyers see clients identically; on its count of eight British regions alone, Roth's comparison gives a one-sided Fisher p-value of 0.2143, so it stands only together with the theorem and the laboratory replication. Chapter 2 proves that deferred acceptance converges to a stable sheet and counts the proposals. Chapter 3 lists all stable matchings and shows that proposer-optimal means receiver-worst, while one extra person on one side shrinks the value of speaking first from 6.02 places to 0.62. Chapter 4 shows that proposers lose nothing by honesty, that 0.233 of receivers can gain by misreporting, and that shortening a list is a gamble that may end with nothing. Chapter 5 extends matching to many-to-one settings such as photographers' slots, tests the rural hospitals theorem and prices stability: in 0.115 of weeks, filling up regardless of rankings books about 1.04 more shoots. Chapter 6 handles incomplete lists and ties and shows that guessed preferences cannot give a stable sheet: with a noise standard deviation of only 0.05, 0.790 of markets keep blocking pairs. Chapter 7 examines "those who elope are not stopped" in the marriage officer of the Rites of Zhou and the original paper of Gale and Shapley. Chapter 8 gives the weekly procedure, the Stable Pairing Matcher and a pre-registered test: if cancellations fall from 0.15 to 0.08, each group needs 325 pairings, with a simulated power of 0.803. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:买手店给预约客户排专属买手、工作室给品牌排摄影师档期、合作平台给达人和品牌撮合主推名额,最常见的排法是先到先得或评分最高先挑。本书证伪这两种排法背后的信念:只要排法公平、执行认真,排出来的表就守得住。两种排法都只问一边想要什么,表上会留下「私奔对」,即一对彼此更想要对方却没被配在一起的人;盖尔与沙普利 1962 年的延迟接受算法一定给出没有私奔对的表,并对提议方最优。书中的人物、档期与偏好均为示意;全部计算由书中程序生成;本书是非个人化经营建议,只使用当事人明示提供的偏好,不做未经同意的画像,配对结果只是建议,双方都可以拒绝。 第一章在示意扫描中发现口味越一致,先到先得越糟,平均私奔对从 4.14 个升到 14.12 个;评分优先只在所有买手看法完全一致时才稳定;罗斯对英国八个地区的比较单看计数,单侧费希尔检验 p 值为 0.2143,须与定理和实验复现一起才站得住。第二章证明延迟接受必然收敛并稳定,并量出提议次数。第三章穷举全部稳定配对,说明提议方最优即接受方最差,而一边只多一个人时,先开口的价值从 6.02 个名次缩到 0.62 个。第四章说明提议方说真话不吃亏,接受方中有 0.233 能靠谎报得益,但缩短名单是一场可能落空的赌注。 第五章把配对推广到摄影师档期这样的多对一,检验偏远医院定理,并算出稳定的代价:0.115 的周里只求塞满可多排约 1.04 场。第六章处理不完整名单与平局,并说明猜出来的偏好配不出稳定的表:噪声标准差仅 0.05 时已有 0.790 的市场留下私奔对。第七章考辨《周礼·地官·媒氏》的「奔者不禁」与盖尔和沙普利的原文。第八章给出周度配对流程、稳定配对器与预注册检验方案:取消比例从 0.15 降到 0.08 时每组需 325 对,模拟检验力 0.803。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 First Come and Highest Score First: Where Blocking Pairs Come From — Chapter 1 finds in an illustrative sweep that the more tastes agree, the worse first come does, with blocking pairs rising from 4.14 to 14.12 on average, while score first is stable only when all buyers see clients identically; on its count of eight British regions alone, Roth's comparison gives a one-sided Fisher p-value of 0.2143, so it stands only together with the theorem and the laboratory replication. (https://5dlibrary.com/answers/en/gale-shapley-stable-matching-buyers-creators-photographers/c01/) 第1章 先到先得与评分优先:私奔对从哪里来 — 第一章在示意扫描中发现口味越一致,先到先得越糟,平均私奔对从 4.14 个升到 14.12 个;评分优先只在所有买手看法完全一致时才稳定;罗斯对英国八个地区的比较单看计数,单侧费希尔检验 p 值为 0.2143,须与定理和实验复现一起才站得住。 (https://5dlibrary.com/answers/zh/gale-shapley-stable-matching-buyers-creators-photographers/c01/) - Ch.2 Deferred Acceptance: Why an Algorithm That Only Holds, Never Commits, Must Converge — Deferred acceptance turns "say yes first" into "hold first". Proposers move down their own lists, and receivers keep only the best offer so far, until nobody is returned any more. Two monotone movements, proposers only down and receivers only up, are enough to prove that the algorithm must stop and must stop without a blocking pair. The illustrative run of p21 reaches a stable sheet after 10 proposals, in which K4 ends with his last choice yet finds no brand willing to elope with him. Program p22 shows that in random markets the average number of proposals stays below Wilson's bound and, at 200 pairs, is only 0.029 of n squared; the worst case of n squared minus n plus 1 is really attained, but only with very special preferences. The order of handling does not affect the result, but the direction of proposing does, and that is the subject of the next chapter. (https://5dlibrary.com/answers/en/gale-shapley-stable-matching-buyers-creators-photographers/c02/) 第2章 延迟接受:只暂留、不拍板的算法为何必然收敛 — 延迟接受把「先答应」改成「先暂留」:提议者沿着自己的名单往下走,接受方只把目前最好的一份留在手里,直到没有人再被退回。两条单调性,提议者只往下、接受方只往上换,足以证明算法一定会停、停下时一定没有私奔对。示意运行 p21 用 10 次提议得到一张稳定的表,其中 K4 落在自己的末位,却找不到任何愿意与他私奔的品牌。p22 显示,随机市场里平均提议次数低于威尔逊的上界,200 对时只占 n 的平方的 0.029;最坏情形 n 的平方减 n 加 1 确实可以达到,但需要极其特殊的偏好。处理顺序不影响结果,提议方向则影响很大,这是下一章的主题。 (https://5dlibrary.com/answers/zh/gale-shapley-stable-matching-buyers-creators-photographers/c02/) - Ch.3 Proposer Optimality: the Lattice of Stable Matchings and the Value of Speaking First — Chapter 3 lists all stable matchings and shows that proposer-optimal means receiver-worst, while one extra person on one side shrinks the value of speaking first from 6.02 places to 0.62. (https://5dlibrary.com/answers/en/gale-shapley-stable-matching-buyers-creators-photographers/c03/) 第3章 提议方最优:稳定配对的格与「谁先开口」的价值 — 第三章穷举全部稳定配对,说明提议方最优即接受方最差,而一边只多一个人时,先开口的价值从 6.02 个名次缩到 0.62 个。 (https://5dlibrary.com/answers/zh/gale-shapley-stable-matching-buyers-creators-photographers/c03/) - Ch.4 Telling the Truth and the Short List: the Proposer's Honesty and the Receiver's Gamble — Chapter 4 shows that proposers lose nothing by honesty, that 0.233 of receivers can gain by misreporting, and that shortening a list is a gamble that may end with nothing. (https://5dlibrary.com/answers/en/gale-shapley-stable-matching-buyers-creators-photographers/c04/) 第4章 说真话与短名单:提议方的诚实与接受方的赌注 — 第四章说明提议方说真话不吃亏,接受方中有 0.233 能靠谎报得益,但缩短名单是一场可能落空的赌注。 (https://5dlibrary.com/answers/zh/gale-shapley-stable-matching-buyers-creators-photographers/c04/) - Ch.5 Many-to-One: Photographers' Slots, Hospitals and Residents — Chapter 5 extends matching to many-to-one settings such as photographers' slots, tests the rural hospitals theorem and prices stability: in 0.115 of weeks, filling up regardless of rankings books about 1.04 more shoots. (https://5dlibrary.com/answers/en/gale-shapley-stable-matching-buyers-creators-photographers/c05/) 第5章 多对一:摄影师档期、医院与住院医 — 第五章把配对推广到摄影师档期这样的多对一,检验偏远医院定理,并算出稳定的代价:0.115 的周里只求塞满可多排约 1.04 场。 (https://5dlibrary.com/answers/zh/gale-shapley-stable-matching-buyers-creators-photographers/c05/) - Ch.6 Short Lists, Ties and Guessed Preferences — Chapter 6 handles incomplete lists and ties and shows that guessed preferences cannot give a stable sheet: with a noise standard deviation of only 0.05, 0.790 of markets keep blocking pairs. (https://5dlibrary.com/answers/en/gale-shapley-stable-matching-buyers-creators-photographers/c06/) 第6章 短名单、平局与猜出来的偏好 — 第六章处理不完整名单与平局,并说明猜出来的偏好配不出稳定的表:噪声标准差仅 0.05 时已有 0.790 的市场留下私奔对。 (https://5dlibrary.com/answers/zh/gale-shapley-stable-matching-buyers-creators-photographers/c06/) - Ch.7 Classical Sources East and West: the Marriage Officer of the Rites of Zhou and Gale and Shapley 1962 — The marriage officer of the Rites of Zhou describes a centralised matching institution: registration, an annual meeting, care for the unmatched, and, during the meeting, "those who elope are not stopped". It got right the importance of centralisation and a single point in time, and it looked after the people of Chapters 5 and 6 who are unmatched under every stable algorithm. It got wrong the use of coercion and punishment, and it organised meetings without prescribing how pairs are formed. Program p71 writes "those who elope are not stopped" as a dynamic: when blocking pairs act one after another, all 1,000 markets reach stability, but the random order breaks 57.37 betrothals on average and 1348 in the longest case. Two specific claims of Gale and Shapley's 1962 paper hold up under p72: the copies method and the quota algorithm agree in 500 markets, and the paper's roommates example has no stable pairing. The next chapter brings the whole book down to a weekly pairing sheet and a pre-registered test plan. (https://5dlibrary.com/answers/en/gale-shapley-stable-matching-buyers-creators-photographers/c07/) 第7章 东西方原典考辨:《周礼》的媒氏与盖尔与沙普利 1962 — 《周礼·地官·媒氏》描述了一个集中的配对机构:登记、统一会期、照看落空者,并在会期内「奔者不禁」。它说对了集中与统一时间点的重要,也照看了第五、六章所说的那批无论用哪种稳定算法都会落空的人;它说错的是强制与处罚,并且只组织相会、不规定怎样配。p71 把「奔者不禁」写成动力过程:放任私奔对一个接一个落实,1,000 个市场全部到达稳定,但随机顺序平均要拆散 57.37 次婚约,最长 1348 次。盖尔与沙普利 1962 年的两条具体说法经 p72 核对成立:分身法与名额算法在 500 个市场里结果一致;室友问题的原文例子没有稳定配对。下一章把全书落到一张周度配对单和一份预注册检验方案上。 (https://5dlibrary.com/answers/zh/gale-shapley-stable-matching-buyers-creators-photographers/c07/) - Ch.8 Closing the Loop: a Weekly Slot Sheet, the Stable Pairing Matcher and a Pre-Registered Test — Chapter 8 gives the weekly procedure, the Stable Pairing Matcher and a pre-registered test: if cancellations fall from 0.15 to 0.08, each group needs 325 pairings, with a simulated power of 0.803. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/gale-shapley-stable-matching-buyers-creators-photographers/c08/) 第8章 实证闭环:周度档期配对单、稳定配对器与预注册检验 — 第八章给出周度配对流程、稳定配对器与预注册检验方案:取消比例从 0.15 降到 0.08 时每组需 325 对,模拟检验力 0.803。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/gale-shapley-stable-matching-buyers-creators-photographers/c08/) ## 5DL-B-300 · Hidden Markov Models and Viterbi Decoding: Inferring a Customer's Hidden Decision Stage from Click Sequences and Timing the Offer 中文书名:隐马尔可夫维特比解码、浏览点击序列反推客户隐藏决策阶段与出手时机母典 DOI: pending Cite: Fan, Y. (2026). Hidden Markov Models and Viterbi Decoding: Inferring a Customer's Hidden Decision Stage from Click Sequences and Timing the Offer (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: Online shops often judge how far a visitor has come from her last click and decide on that basis whether to add a quote. This book refutes that common sense. A customer's decision stage is invisible and each click is only a blurred projection of it; put the whole click sequence into a hidden Markov model and infer the current stage with the forward filter, and the judgement is more accurate, by an amount that can be computed. After checking Rabiner's tutorial of 1989, the book corrects its project blueprint: the tutorial gives algorithms and two optimality criteria but measures no accuracy of judging customer stages, and the stage "now" should be judged by the forward filter, not by the last cell of the Viterbi path. Its stages, page types and probabilities are illustrative; all calculations are generated by its programs; the book is not individual business advice and uses only anonymised on-site data and data customers agreed to provide, with no tracking across sites and no inference of sensitive traits. Decoded stages are never used to press customers, to price differently or to contact customers unprompted, and urgency comes only from real stock and real deadlines. With illustrative parameters, Chapter 1 finds the forward filter 0.0410 more accurate than the last click on four stages, raising ready-or-not accuracy from 0.6990 to 0.7630; the gain varies with stage stickiness and page informativeness from 0.0307 to 0.1187, and the last Viterbi cell is 0.1007 worse when pages are blurred. Chapter 2 unfolds Viterbi on a trellis, shows direct multiplication underflowing near click 368, and finds that the best path holds only 0.0357 of the total probability. Chapter 3 covers forward-backward and the two optimality criteria; judging two clicks later adds 0.0413 in accuracy. Chapter 4 estimates parameters from unlabelled clickstreams with Baum–Welch, and ten random starts decode with accuracy from 0.3877 to 0.5181. Chapter 5 treats label switching, the number of stages and weak identifiability: naming stages by average position within a visit gives an accuracy of 0.4477, close to 0.4623 with true labels, and with blurred pages clicks keep only 0.022 of the information. Chapter 6 gives the timing rule: never speak first, always answer price questions, one price for all, urgency only from reality, and add a quote to other answers only by a calibrated posterior and a cost threshold; counting the inquiry as evidence costs 318.3 per thousand questions against 345.0 from clicks alone, and wrong parameters double premature quotes from 0.0684 to 0.1449. Chapter 7 examines Markov's letter chain in Onegin of 1913 (stationary vowel share 0.4319) and the Chuai chapter of the Guiguzi (one click carries only 0.2680 bits). Chapter 8 gives the weekly review, the Customer Stage Decoder and pre-registered tests: test one needs 372 visits with a self-reported stage, test two 5301 chats per arm. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:网店客服常凭访客最后一次点击判断她走到了哪一步,并据此决定要不要附上报价。本书证伪这条常识:客户的决策阶段看不见,每一次点击只是它的模糊投影;把整条点击序列放进隐马尔可夫模型,用前向滤波推断此刻的阶段,判断更准,而准多少可以算出来。核对拉宾纳 1989 年的教程后,本书修正了立项蓝图:教程给出算法与两种最优准则,没有测量客户阶段的准确率;判断「此刻」应当用前向滤波,而不是维特比路径的最后一格。书中阶段、页面类型与全部概率均为示意;全部计算由书中程序生成;本书是非个人化经营建议,只使用站内匿名化与客户同意的数据,不跨站追踪,不推断敏感属性,阶段解码不用于施压、差别定价或主动联系客户,紧迫感只来自真实库存与真实截止。 第一章在示意参数下量出:前向滤波的四分类准确率比最后一次点击高 0.0410,就绪判别从 0.6990 提高到 0.7630;增益随阶段黏性与页面区分力在 0.0307 到 0.1187 之间变化,维特比最后一格在页面模糊时反而低 0.1007。第二章在格子图上展开维特比,说明直接相乘约在第 368 次点击下溢,最优路径只占全部概率的 0.0357。第三章讲前向后向与两种最优准则,晚两次点击判断可多得 0.0413 的准确率。第四章用 Baum–Welch 从无标签点击流估计参数,十个随机起点的解码准确率从 0.3877 到 0.5181。 第五章讨论标签切换、阶段个数与弱可辨识:按访问中平均位置命名阶段的准确率 0.4477,接近真标签的 0.4623;页面模糊时点击只保留 0.022 的信息。第六章给出出手时机:不先开口、问价必答、同价、紧迫感只来自真实,回答其他问题时按校准后验与代价门槛决定是否附报价,把发问本身算作证据后每千次发问代价 318.3,只用点击是 345.0;用错参数时过早报价从 0.0684 翻到 0.1449。第七章考辨马尔可夫 1913 年的《奥涅金》字母链(平稳元音比例 0.4319)与《鬼谷子·揣篇》(一次点击只携带 0.2680 比特)。第八章给出周度复盘流程、客户阶段解码器与预注册检验:检验一需要 372 次有自报阶段的访问,检验二每组 5301 次对话。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Last-Click Illusion: Whole Sequences against a Single Page — Judging a customer's stage from the last click alone gives, with the illustrative parameters, a four-way error rate of 0.5180 and a ready-or-not accuracy of 0.6990. The forward filter over the whole sequence raises four-way accuracy by 0.0410 and ready-or-not accuracy to 0.7630. The size of the gain depends on stage stickiness and on how much pages reveal, ranging from 0.0307 to 0.1187 in the sweep of p12. The last cell of the Viterbi path is not always better, and with blurred pages it is 0.1007 worse. Rabiner's tutorial separates the question of the most probable state now from that of the most probable whole path, and the next two chapters take them in turn. Chapter 2 shows how Viterbi finds the most probable path on a trellis and why it must work in logarithms. Chapter 3 turns to the forward-backward procedure and to why waiting two more clicks before answering a visitor is often worth it. (https://5dlibrary.com/answers/en/hidden-markov-viterbi-customer-decision-stage-offer-timing/c01/) 第1章 最后一次点击的错觉:整条序列与单点判别的准确率 — 只看最后一次点击判断客户阶段,在示意参数下四分类错误率是 0.5180,就绪判别准确率 0.6990。用整条序列的前向滤波,四分类准确率提高 0.0410,就绪判别提高到 0.7630;提高的幅度取决于阶段黏性与页面区分力,在 p12 的扫描里从 0.0307 到 0.1187。维特比路径的最后一格并不总是更好,页面模糊时反而低 0.1007。拉宾纳的教程区分了「当前最可能的状态」与「最可能的整条路径」两个问题,本书接下来两章分别展开:第二章讲维特比怎样在格子上找出最可能的路径,以及为什么必须在对数域里算;第三章讲前向后向算法,以及为什么回复访客之前「再等两次点击」往往值得。 (https://5dlibrary.com/answers/zh/hidden-markov-viterbi-customer-decision-stage-offer-timing/c01/) - Ch.2 The Viterbi Trellis: Dynamic Programming in Logarithms, Backtracking and Underflow — Chapter 2 unfolds Viterbi on a trellis, shows direct multiplication underflowing near click 368, and finds that the best path holds only 0.0357 of the total probability. (https://5dlibrary.com/answers/en/hidden-markov-viterbi-customer-decision-stage-offer-timing/c02/) 第2章 维特比格子:对数域动态规划、回溯与下溢 — 第二章在格子图上展开维特比,说明直接相乘约在第 368 次点击下溢,最优路径只占全部概率的 0.0357。 (https://5dlibrary.com/answers/zh/hidden-markov-viterbi-customer-decision-stage-offer-timing/c02/) - Ch.3 Forward and Backward: Posterior Marginals, the Best Path and "Wait Two More Clicks" — Chapter 3 covers forward-backward and the two optimality criteria; judging two clicks later adds 0.0413 in accuracy. (https://5dlibrary.com/answers/en/hidden-markov-viterbi-customer-decision-stage-offer-timing/c03/) 第3章 前向后向:后验边际、最优路径与「再等两次点击」 — 第三章讲前向后向与两种最优准则,晚两次点击判断可多得 0.0413 的准确率。 (https://5dlibrary.com/answers/zh/hidden-markov-viterbi-customer-decision-stage-offer-timing/c03/) - Ch.4 Baum–Welch: Estimating Parameters from Unlabelled Clickstreams — Chapter 4 estimates parameters from unlabelled clickstreams with Baum–Welch, and ten random starts decode with accuracy from 0.3877 to 0.5181. (https://5dlibrary.com/answers/en/hidden-markov-viterbi-customer-decision-stage-offer-timing/c04/) 第4章 Baum–Welch:从没有标签的点击流里估计参数 — 第四章用 Baum–Welch 从无标签点击流估计参数,十个随机起点的解码准确率从 0.3877 到 0.5181。 (https://5dlibrary.com/answers/zh/hidden-markov-viterbi-customer-decision-stage-offer-timing/c04/) - Ch.5 Identifiability and Label Switching: What Clickstreams Cannot Say — Chapter 5 treats label switching, the number of stages and weak identifiability: naming stages by average position within a visit gives an accuracy of 0.4477, close to 0.4623 with true labels, and with blurred pages clicks keep only 0.022 of the information. (https://5dlibrary.com/answers/en/hidden-markov-viterbi-customer-decision-stage-offer-timing/c05/) 第5章 可辨识性与标签切换:点击流说不清的东西 — 第五章讨论标签切换、阶段个数与弱可辨识:按访问中平均位置命名阶段的准确率 0.4477,接近真标签的 0.4623;页面模糊时点击只保留 0.022 的信息。 (https://5dlibrary.com/answers/zh/hidden-markov-viterbi-customer-decision-stage-offer-timing/c05/) - Ch.6 Timing the Offer: the Inquiry as Evidence, a Cost Threshold and Calibration — Chapter 6 gives the timing rule: never speak first, always answer price questions, one price for all, urgency only from reality, and add a quote to other answers only by a calibrated posterior and a cost threshold; counting the inquiry as evidence costs 318.3 per thousand questions against 345.0 from clicks alone, and wrong parameters double premature quotes from 0.0684 to 0.1449. (https://5dlibrary.com/answers/en/hidden-markov-viterbi-customer-decision-stage-offer-timing/c06/) 第6章 出手时机:发问即证据、成本门槛与校准 — 第六章给出出手时机:不先开口、问价必答、同价、紧迫感只来自真实,回答其他问题时按校准后验与代价门槛决定是否附报价,把发问本身算作证据后每千次发问代价 318.3,只用点击是 345.0;用错参数时过早报价从 0.0684 翻到 0.1449。 (https://5dlibrary.com/answers/zh/hidden-markov-viterbi-customer-decision-stage-offer-timing/c06/) - Ch.7 Classical Sources East and West: Markov's Letter Chain in Onegin and the Chuai Chapter of the Guiguzi — Chapter 7 examines Markov's letter chain in Onegin of 1913 (stationary vowel share 0.4319) and the Chuai chapter of the Guiguzi (one click carries only 0.2680 bits). (https://5dlibrary.com/answers/en/hidden-markov-viterbi-customer-decision-stage-offer-timing/c07/) 第7章 东西方原典考辨:马尔可夫的《奥涅金》字母链与《鬼谷子·揣篇》 — 第七章考辨马尔可夫 1913 年的《奥涅金》字母链(平稳元音比例 0.4319)与《鬼谷子·揣篇》(一次点击只携带 0.2680 比特)。 (https://5dlibrary.com/answers/zh/hidden-markov-viterbi-customer-decision-stage-offer-timing/c07/) - Ch.8 Closing the Loop: a Weekly Stage Review, the Customer Stage Decoder and Pre-Registered Tests — Chapter 8 gives the weekly review, the Customer Stage Decoder and pre-registered tests: test one needs 372 visits with a self-reported stage, test two 5301 chats per arm. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/hidden-markov-viterbi-customer-decision-stage-offer-timing/c08/) 第8章 实证闭环:周度阶段复盘、客户阶段解码器与预注册检验 — 第八章给出周度复盘流程、客户阶段解码器与预注册检验:检验一需要 372 次有自报阶段的访问,检验二每组 5301 次对话。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/hidden-markov-viterbi-customer-decision-stage-offer-timing/c08/) ## 5DL-B-301 · The Maximum-Entropy Principle and Demand Reconstruction from Missing Data: the Least-Biased Order-Value Distribution Given Only a Mean and Bounds 中文书名:最大熵原理与缺数据需求重建、仅知均价与上下限时最不偏的客单分布推断母典 DOI: pending Cite: Fan, Y. (2026). The Maximum-Entropy Principle and Demand Reconstruction from Missing Data: the Least-Biased Order-Value Distribution Given Only a Mean and Bounds (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: The normal shortcut adds a symmetric shape and a standard deviation; under illustrative parameters, with a mean of 80 and a floor of 30, it puts 0.133 of its probability on impossible values, rising to 0.412 when the mean is 40. Accepting only the bounds and the mean, the maximum-entropy principle proposed by Jaynes in 1957 gives a truncated exponential distribution, with a 90th-percentile order of 146.0 for the illustrative dessert shop. Abstract: Owners who open a new shop or launch a new product often know only three numbers: a floor, a ceiling and an average for order value or demand. The usual moves are "treat it as normal" or "if we know nothing, treat it as uniform", both regarded as the least prejudiced choice. This book overturns that common sense. The normal shortcut adds a symmetric shape and a standard deviation; under illustrative parameters, with a mean of 80 and a floor of 30, it puts 0.133 of its probability on impossible values, rising to 0.412 when the mean is 40. The uniform shortcut contradicts the known mean. Accepting only the bounds and the mean, the maximum-entropy principle proposed by Jaynes in 1957 gives a truncated exponential distribution, with a 90th-percentile order of 146.0 for the illustrative dessert shop. After checking the originals the book corrects its blueprint: the axiomatic case for uniqueness comes from Shore and Johnson in 1980 and holds only relative to the chosen constraints, variable and entropy functional; "least biased" is a minimax theorem under log loss but need not hold for the error of a single quantile. Its parameters are illustrative; all calculations are generated by its programs; it is general business analysis, not personalised business advice. Chapter 1 derives the Lagrange multiplier and runs a sensitivity scan. Chapter 2 moves to discrete price tiers, shows by exact enumeration that maximum-entropy shares occur most often, and shows the answer moving with the constraints. Chapter 3 compares the three approaches under three rulers: across four illustrative truths the log loss of maximum entropy is constant at 7.080 bits and smallest in the worst case, yet the worst error of the 90th percentile is smaller for the normal shortcut, 18.8 against 23.5. Chapter 4 adds a second moment, turning the answer into the truncated normal family. Chapter 5 finds that maximum entropy is more accurate within about thirty orders and should give way to sample quantiles after about fifty, and that adding the sample standard deviation makes matters worse under the illustrative truth. Chapter 6 puts the maximum-entropy quantile into the newsvendor model for new products; for thin-margin lines the normal shortcut's worst regret share reaches 1.018. Chapter 7 reads "knowing that one does not know" in Laozi 71 (Wang Bi and silk texts) and Laplace's principle of insufficient reason, measuring by divergence the pretended knowledge of distributions that contradict the known mean; for the uniform shortcut the discrepancy is 2.385 bits. Chapter 8 gives the cold-start loop for new products, the distilled Maximum-Entropy Demand Reconstructor and a pre-registered plan of refutation, in which detecting an exceedance share of 0.20 in the calibration test needs 86 new products. Boxes separate published evidence, runnable Python (20 programs), classical sources and hypotheses. 摘要:开新店、上新品的经营者常常只知道三个数:客单或需求的下限、上限与一个平均数。常见做法是「就当正态」或「不知道就当均匀」,两者都被视为最不带偏见的选择。本书推翻这一常识:正态捷径私加了对称形状与标准差,示意参数下均值 80、下限 30 时它把 0.133 的概率放在不可能的区间,均值 40 时升到 0.412;均匀捷径隐含均值在区间中点,直接违背已知均值。只承认上下限与均值时,杰恩斯 1957 年提出的最大熵原则给出截断指数分布,示意甜品店的第 90 百分位客单为 146.0 元。核对原文后本书做了蓝图修正:「唯一」的公理化论证来自肖尔与约翰逊 1980 年,且只相对于所选约束、变量与熵泛函成立;「最不偏」在对数损失下是极小极大定理,在单个分位数误差下未必成立。书中参数均为示意,全部计算由书中程序生成;本书是非个人化经营建议。 第一章推导拉格朗日乘子并做敏感度扫描。第二章转到离散价格档位,用精确整数穷举说明最大熵份额在频率上最常出现,并演示约束改变时答案随之改变。第三章以三把尺子比较三种做法:在四个示意真相下,最大熵的对数损失恒为 7.080 比特,最坏情形最小;第 90 百分位的最坏误差却是正态捷径更小,18.8 对 23.5 元。第四章加入第二矩,答案变为截断正态族。第五章发现约三十笔订单以内最大熵更准、约五十笔以后应换成样本分位数,且示意真相下加入样本标准差反而更差。 第六章把最大熵分位数放进新品报童模型,薄利品类中正态捷径的最坏后悔比例可达 1.018。第七章考辨《老子》第七十一章「知不知」的王弼本与帛书本异文,以及拉普拉斯《概率的哲学随笔》的不充分理由原则,用散度量出违背已知均值的「冒充的知识」,均匀捷径的差额为 2.385 比特。第八章给出新品冷启动闭环、蒸馏小工具「最大熵需求重建器」与预注册证伪方案,校准检验分辨超出比例 0.20 需 86 个新品。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Lagrange Multiplier of Maximum Entropy: the Truncated Exponential When Only a Mean and Bounds Are Known — The closer the mean sits to the lower bound, the worse the normal shortcut becomes. With a mean of 40 it puts 0.412 of its probability below 30, and its 10th percentile is a negative order value of $-17.7$, impossible in any shop. With a mean of 50 that share is 0.328, and with a mean of 80 it is still 0.133. Meanwhile the gap between the two answers for the 90th percentile is $-44.6$ at a mean of 40, where the normal overstates the upper tail, and 55.6 at a mean of 130, where it understates it; somewhere in between the gap passes through zero. At a mean of 80 the gap is only 8.4, which shows that at certain points the normal shortcut's upper-tail figure happens to be passable. Chapter 3 deals with this directly: the figure is passable because two errors cancel, not because the shortcut has better grounds. (https://5dlibrary.com/answers/en/maximum-entropy-demand-reconstruction-mean-and-bounds/c01/) 第1章 最大熵的拉格朗日乘子:只知均价与上下限时的截断指数分布 — 均值越靠近下限,正态捷径越离谱。均值为 40 时,正态分布把 0.412 的概率放在 30 元以下,第 10 百分位是负数 $-17.7$ 元,这在任何店里都不可能;均值为 50 时这一份额是 0.328,均值为 80 时仍有 0.133。与此同时,两种方法对第 90 百分位的判断差距在均值 40 时为 $-44.6$ 元(正态高估),在均值 130 时为 55.6 元(正态低估),中间某处差距穿过零。均值 80 处差距只有 8.4 元,这说明在某些点上正态捷径的上尾数字碰巧不坏,第三章会专门处理这一点:它不坏,是因为两种错误相互抵消,而不是因为它更有根据。 (https://5dlibrary.com/answers/zh/maximum-entropy-demand-reconstruction-mean-and-bounds/c01/) - Ch.2 Discrete Maximum Entropy over Price Tiers: Wallis's Count, the Concentration Theorem and the Limits of Uniqueness — Raising the average from 79 to 85 lifts the share at 149 and above from 0.1770 to 0.2100; lowering it to 73 cuts the share to 0.1438. A difference of six in the average shifts the large-order share by three to four percentage points, so the maximum-entropy answer is quite sensitive to the accuracy of the mean. Chapter 5 deals with the bias that arises when the mean itself comes from a small sample. Adding a tier at 399 moves the share at 149 and above from 0.1770 to 0.1695 even though the average is unchanged, and the lowest tier goes from 0.2672 to 0.2752; dropping the 299 tier moves the large-order share to 0.1931. The support is part of the constraints: which prices appear on the menu is itself information. (https://5dlibrary.com/answers/en/maximum-entropy-demand-reconstruction-mean-and-bounds/c02/) 第2章 离散最大熵与价格档位:沃利斯计数、集中定理与「唯一」的边界 — 均价从 79 改成 85,149 元及以上的份额从 0.1770 涨到 0.2100;改成 73,降到 0.1438。均价只差六元,大单份额就差出三到四个百分点,这说明最大熵答案对均值的准确度相当敏感,第五章会专门讨论均值本身来自小样本时的偏差。增加一个 399 元的档位,即使均价不变,149 元及以上的份额也从 0.1770 变成 0.1695,最低档从 0.2672 变成 0.2752;删去 299 元档,大单份额变成 0.1931。支撑是约束的一部分:菜单上有哪些价位,本身就是一条信息。 (https://5dlibrary.com/answers/zh/maximum-entropy-demand-reconstruction-mean-and-bounds/c02/) - Ch.3 The Cost of the Normal and Uniform Shortcuts: Quantiles, Log Loss and the Minimax — Chapter 3 compares the three approaches under three rulers: across four illustrative truths the log loss of maximum entropy is constant at 7.080 bits and smallest in the worst case, yet the worst error of the 90th percentile is smaller for the normal shortcut, 18.8 against 23.5. (https://5dlibrary.com/answers/en/maximum-entropy-demand-reconstruction-mean-and-bounds/c03/) 第3章 正态与均匀两条捷径的代价:分位数、对数损失与极小极大 — 第三章以三把尺子比较三种做法:在四个示意真相下,最大熵的对数损失恒为 7.080 比特,最坏情形最小;第 90 百分位的最坏误差却是正态捷径更小,18.8 对 23.5 元。 (https://5dlibrary.com/answers/zh/maximum-entropy-demand-reconstruction-mean-and-bounds/c03/) - Ch.4 Adding the Second Moment: the Truncated Normal Family, Convex Exponents and What One Standard Deviation Buys — At standard deviations of 60 and 75 the maximum-entropy answer becomes convex. The 90th percentile is pushed to 166.1 and 236.8 while the median drops to 57.0 and 47.9: polarisation empties the middle. The uncut normal breaks down completely here. Its 10th percentiles are 3.1 and $-16.1$, its out-of-range shares 0.202 and 0.252, and its 90th percentiles 156.9 and 176.1, both far below the maximum-entropy answer. At a standard deviation of 75 entropy falls by 0.497 bits, the largest drop in the table, because a standard deviation far above the implied value is a very surprising piece of information. (https://5dlibrary.com/answers/en/maximum-entropy-demand-reconstruction-mean-and-bounds/c04/) 第4章 加入第二矩:截断正态族、凸指数与一条标准差能买到多少信息 — 标准差 60 与 75 时,最大熵答案变成凸形,第 90 百分位分别推到 166.1 与 236.8 元,中位数却降到 57.0 与 47.9 元:两极分化让中间变空。不截断的正态在这里完全失效,它的第 10 百分位分别是 3.1 元和 $-16.1$ 元,越界份额 0.202 与 0.252,上尾第 90 百分位 156.9 与 176.1 元,都远低于最大熵答案。熵在标准差 75 时下降了 0.497 比特,是表中最多的,因为一个远大于隐含值的标准差是一条很「意外」的信息。 (https://5dlibrary.com/answers/zh/maximum-entropy-demand-reconstruction-mean-and-bounds/c04/) - Ch.5 After the Sample Arrives: Plug-in Bias, the Second-Moment Trap and the Handover Point — Look first at the root mean square error. With only 5 orders, maximum entropy errs by 45.2, the sample quantile by 52.4 and the normal by 50.0. With 10 orders the figures are 34.1, 44.4 and 38.0; with 20 they are 24.8, 29.8 and 28.5. Up to twenty orders maximum entropy is always the most accurate. The reason is direct. The sample quantile uses only the top one or two orders and is very noisy; maximum entropy uses only the sample mean, which averages over every order and is far less noisy, and it adds the free information of the bounds. (https://5dlibrary.com/answers/en/maximum-entropy-demand-reconstruction-mean-and-bounds/c05/) 第5章 样本到来之后:插值偏差、第二矩的陷阱与换手点 — 先看均方根误差。样本只有 5 笔时,最大熵的误差是 45.2 元,样本分位数 52.4 元,正态 50.0 元;10 笔时分别是 34.1、44.4 与 38.0 元;20 笔时是 24.8、29.8 与 28.5 元。在二十笔以内,最大熵始终最准。原因很直接:样本分位数只用到排在最上面的一两笔订单,噪声极大;最大熵只用样本均值,均值是所有订单的平均,噪声小得多,再加上上下限这条免费的信息。 (https://5dlibrary.com/answers/zh/maximum-entropy-demand-reconstruction-mean-and-bounds/c05/) - Ch.6 Stocking a New Product with the Newsvendor Model: Ordering at a Maximum-Entropy Quantile and Worst-Case Regret — Chapter 6 puts the maximum-entropy quantile into the newsvendor model for new products; for thin-margin lines the normal shortcut's worst regret share reaches 1.018. (https://5dlibrary.com/answers/en/maximum-entropy-demand-reconstruction-mean-and-bounds/c06/) 第6章 新品备货的报童模型:最大熵分位数订货与最坏后悔 — 第六章把最大熵分位数放进新品报童模型,薄利品类中正态捷径的最坏后悔比例可达 1.018。 (https://5dlibrary.com/answers/zh/maximum-entropy-demand-reconstruction-mean-and-bounds/c06/) - Ch.7 East and West in the Classics: "Knowing That One Does Not Know" in Laozi 71 and Laplace's Philosophical Essay on Probabilities — Chapter 7 reads "knowing that one does not know" in Laozi 71 (Wang Bi and silk texts) and Laplace's principle of insufficient reason, measuring by divergence the pretended knowledge of distributions that contradict the known mean; for the uniform shortcut the discrepancy is 2.385 bits. (https://5dlibrary.com/answers/en/maximum-entropy-demand-reconstruction-mean-and-bounds/c07/) 第7章 东西方原典考辨:《老子·第七十一章》「知不知」与拉普拉斯《概率的哲学随笔》 — 第七章考辨《老子》第七十一章「知不知」的王弼本与帛书本异文,以及拉普拉斯《概率的哲学随笔》的不充分理由原则,用散度量出违背已知均值的「冒充的知识」,均匀捷径的差额为 2.385 比特。 (https://5dlibrary.com/answers/zh/maximum-entropy-demand-reconstruction-mean-and-bounds/c07/) - Ch.8 Closing the Loop: Cold-Start Demand for a New Product, the Maximum-Entropy Demand Reconstructor and a Pre-Registered Plan of Refutation — Chapter 8 gives the cold-start loop for new products, the distilled Maximum-Entropy Demand Reconstructor and a pre-registered plan of refutation, in which detecting an exceedance share of 0.20 in the calibration test needs 86 new products. Boxes separate published evidence, runnable Python (20 programs), classical sources and hypotheses. (https://5dlibrary.com/answers/en/maximum-entropy-demand-reconstruction-mean-and-bounds/c08/) 第8章 实证闭环:新品冷启动需求分布、最大熵需求重建器与预注册证伪方案 — 第八章给出新品冷启动闭环、蒸馏小工具「最大熵需求重建器」与预注册证伪方案,校准检验分辨超出比例 0.20 需 86 个新品。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/maximum-entropy-demand-reconstruction-mean-and-bounds/c08/) ## 5DL-B-302 · Simpson's Paradox and the Back-Door Criterion: Reversals When Branch Data Are Pooled and De-Confounding with Causal Graphs 中文书名:辛普森悖论与后门调整准则、多分店数据合并结论反转与因果图去混杂母典 DOI: pending Cite: Fan, Y. (2026). Simpson's Paradox and the Back-Door Criterion: Reversals When Branch Data Are Pooled and De-Confounding with Causal Graphs (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Branches, plans and conversions are illustrative; all calculations are generated by the book's programs. This is general analysis, not individualised business advice. Abstract: Every branch shows plan A converting better, yet the pooled table shows plan B ahead: should the owner believe the strata or the total? The two popular answers, that the larger pooled sample is more reliable and that finer is truer, both assume the answer is hidden in the data. The central claim of this book is that the causal graph, not the data, decides which table to believe. Before writing, the book checked the original sources and corrects the blueprint: Simpson's note of 1951 was not the first, since Yule discussed in 1903 how mixing universes creates association; Bickel and colleagues concluded in 1975 that after stratifying by department there is no bias against women and the aggregate gap comes from where people applied; and the graph cannot be learned from the table, so it must be stated and tested with randomised trials. Branches, plans and conversions are illustrative; all calculations are generated by the book's programs. This is general analysis, not individualised business advice. Chapter 1 recomputes the reversal for Berkeley's six largest departments: pooled admission rates are 0.445 for men and 0.304 for women, while standardised to the applicant mix they are 0.387 and 0.430; in an illustrative three-branch report plan A leads by 0.050 in every branch yet trails by 0.2493 when pooled, and the exact condition for a reversal is proved. Chapter 2 finds the smallest reversal, which needs nine customers, and a seeded simulation gives a reversal frequency of 0.01648 for random tables, in line with the published figure of about one in sixty. Chapter 3 implements the back-door formula and inverse-probability weighting. Chapter 4 pairs one table with confounder, mediator and collider graphs, whose correct answers are +0.0500 and -0.2493. Chapter 5 enumerates back-door sets and shows that adjusting for a pre-treatment collider creates a bias equal to 66.0 % of the effect. Chapter 6 covers intervals, power and sparse strata. Chapter 7 examines the Lüshi Chunqiu line that many words seem wrong but are right, and many seem right but are wrong, together with Yule's paper of 1903. Chapter 8 sets out the closed loop of forced stratification and adjustment for multi-branch reports, the distilled Stratified Reversal Detector and a pre-registered randomised holdout: when the branch is the only confounder the adjusted estimate is falsified in 0.033 of replications, and with a hidden confounder in 0.637. The book contains 23 runnable Python programs. 摘要:各分店都显示甲方案成交更好,合并成总表后却是乙方案更好,这时该信分层还是合并?流行的两种回答,一说总表样本大最可信,一说分得越细越真实,都以为答案藏在数据里。本书的核心命题是:该信哪张表由因果图决定,而不是由数据决定。落笔前本书核对原文并作蓝图修正:Simpson 1951 年的短文并非最早,Yule 1903 年已讨论混合宇宙制造关联;Bickel 等 1975 年的结论是按院系分层后不存在对女性的偏差,汇总差距来自申请去向;因果图本身无法从表中学出,必须写明并用随机化试行检验。书中门店、方案与成交数均为示意;全部计算由书中程序生成。本书属于非个人化经营建议。 第一章在伯克利六个最大院系上复算反转:合并后男性录取率 0.445、女性 0.304,按申请者构成标准化后男性 0.387、女性 0.430;三店示意报表中甲在每店都高 0.050,合并后却落后 0.2493,并给出反转可能的充要条件。第二章找出只需九个顾客的最小反例,并用固定种子模拟得到随机表反转频率 0.01648,与约六十分之一的文献结果一致。第三章实现后门调整公式与逆概率加权。第四章用同一张表配混杂、中介、对撞三张因果图,正确答案分别是 +0.0500 与 -0.2493。第五章枚举后门集合,并说明对一个处理前的对撞变量作调整会带来相当于效应 66.0 % 的偏差。第六章讨论区间、检验力与稀疏格子。 第七章考辨《吕氏春秋·察传》「辞多类非而是,多类是而非」与 Yule 1903。第八章给出多店报表强制分层与调整的闭环、蒸馏小工具「分层反转检测器」与预注册的随机留出检验:只有分店一种混杂时,调整估计被留出数据推翻的比例为 0.033,存在隐藏混杂时升到 0.637。书中含 23 段可运行的 Python 程序。 Chapter answers: - Ch.1 Every Branch Wins, the Total Loses: the Reversal Condition for 2 x 2 x k Tables — Chapter 1 recomputes the reversal for Berkeley's six largest departments: pooled admission rates are 0.445 for men and 0.304 for women, while standardised to the applicant mix they are 0.387 and 0.430; in an illustrative three-branch report plan A leads by 0.050 in every branch yet trails by 0.2493 when pooled, and the exact condition for a reversal is proved. (https://5dlibrary.com/answers/en/simpson-paradox-back-door-adjustment-branch-data/c01/) 第1章 分店都赢、合并却输:二乘二乘 k 表的反转条件 — 第一章在伯克利六个最大院系上复算反转:合并后男性录取率 0.445、女性 0.304,按申请者构成标准化后男性 0.387、女性 0.430;三店示意报表中甲在每店都高 0.050,合并后却落后 0.2493,并给出反转可能的充要条件。 (https://5dlibrary.com/answers/zh/simpson-paradox-back-door-adjustment-branch-data/c01/) - Ch.2 How Common Are Reversals? The Smallest Example, Random Tables and Sampling Noise — Chapter 2 finds the smallest reversal, which needs nine customers, and a seeded simulation gives a reversal frequency of 0.01648 for random tables, in line with the published figure of about one in sixty. (https://5dlibrary.com/answers/en/simpson-paradox-back-door-adjustment-branch-data/c02/) 第2章 反转有多常见:最小反例、随机表频率与抽样噪声 — 第二章找出只需九个顾客的最小反例,并用固定种子模拟得到随机表反转频率 0.01648,与约六十分之一的文献结果一致。 (https://5dlibrary.com/answers/zh/simpson-paradox-back-door-adjustment-branch-data/c02/) - Ch.3 The Back-Door Adjustment Formula: Standardisation, Inverse Probability Weighting and a Structural Model with a Known Answer — Of all customers, the flagship holds 41/112, the community branch 5/16 and the outlet 9/28. With this mix, the conversion rate if everyone received plan A is 223/448, or 0.4978; if everyone received plan B it is 1003/2240, or 0.4478; the average effect is exactly 1/20, that is, 0.0500. Weighted by plan A's mix, A is at 0.3455 and B at 0.2955; weighted by plan B's mix, A is at 0.6447 and B at 0.5947. Under all three weightings the effect is plus 0.0500, because in this illustrative example the within-branch gap is the same in every branch. The last line prints the naive pooled gap again, minus 0.2493, for comparison. (https://5dlibrary.com/answers/en/simpson-paradox-back-door-adjustment-branch-data/c03/) 第3章 后门调整公式:标准化、逆概率加权与已知真值的结构模型 — 全体顾客中,旗舰店占 41/112,社区店占 5/16,奥特莱斯占 9/28。按这个构成,全员用甲的成交率是 223/448,即 0.4978;全员用乙是 1003/2240,即 0.4478;平均效应恰好是 1/20,也就是 0.0500。按甲组的构成加权,甲是 0.3455、乙是 0.2955;按乙组的构成加权,甲是 0.6447、乙是 0.5947。三种权重下的效应都是正的 0.0500,因为在这个示意例子里各店的店内差完全相同。最后一行再次打印朴素的合并差,负的 0.2493,作为对照。 (https://5dlibrary.com/answers/zh/simpson-paradox-back-door-adjustment-branch-data/c03/) - Ch.4 One Table, Three Causal Graphs: Confounder, Mediator and Collider — Chapter 4 pairs one table with confounder, mediator and collider graphs, whose correct answers are +0.0500 and -0.2493. (https://5dlibrary.com/answers/en/simpson-paradox-back-door-adjustment-branch-data/c04/) 第4章 同一张表,三张因果图:混杂、中介与对撞 — 第四章用同一张表配混杂、中介、对撞三张因果图,正确答案分别是 +0.0500 与 -0.2493。 (https://5dlibrary.com/answers/zh/simpson-paradox-back-door-adjustment-branch-data/c04/) - Ch.5 Enumerating Back-Door Sets: d-Separation, M-Bias and Counting Causal Graphs — Chapter 5 enumerates back-door sets and shows that adjusting for a pre-treatment collider creates a bias equal to 66.0 % of the effect. (https://5dlibrary.com/answers/en/simpson-paradox-back-door-adjustment-branch-data/c05/) 第5章 后门集合的枚举:d 分离、M 型偏差与因果图的计数 — 第五章枚举后门集合,并说明对一个处理前的对撞变量作调整会带来相当于效应 66.0 % 的偏差。 (https://5dlibrary.com/answers/zh/simpson-paradox-back-door-adjustment-branch-data/c05/) - Ch.6 Sampling Error in Stratified Estimates: Intervals, Power and Sparse Strata — Detecting an effect of 0.05 with 80 % power needs 393.5 customers per plan per branch, rounded up to 394, or 2364 customers across three branches and two plans. The table shows how power grows with sample size: 0.292 at 100 customers per plan per branch, 0.515 at 200, 0.806 at 400, 0.933 at 600, 0.979 at 800 and 0.998 at 1200. The simulated power at 400 is 0.814, within simulation error of the closed-form 0.806. (https://5dlibrary.com/answers/en/simpson-paradox-back-door-adjustment-branch-data/c06/) 第6章 分层估计的抽样误差:区间、检验力与稀疏格子 — 要以 80 % 的检验力看出 0.05 的效应,每店每方案需要 393.5 人,向上取整为 394 人,三店两方案合计 2364 人。表格给出了检验力随样本量的变化:每店每方案 100 人时只有 0.292,200 人时 0.515,400 人时 0.806,600 人时 0.933,800 人时 0.979,1200 人时 0.998。模拟出的 400 人时的检验力是 0.814,与闭式的 0.806 相差在模拟误差之内。 (https://5dlibrary.com/answers/zh/simpson-paradox-back-door-adjustment-branch-data/c06/) - Ch.7 Classics East and West: the Lüshi Chunqiu on Examining Hearsay, and Yule 1903 — Chapter 7 examines the Lüshi Chunqiu line that many words seem wrong but are right, and many seem right but are wrong, together with Yule's paper of 1903. (https://5dlibrary.com/answers/en/simpson-paradox-back-door-adjustment-branch-data/c07/) 第7章 东西方原典考辨:《吕氏春秋·察传》与 Yule 1903 — 第七章考辨《吕氏春秋·察传》「辞多类非而是,多类是而非」与 Yule 1903。 (https://5dlibrary.com/answers/zh/simpson-paradox-back-door-adjustment-branch-data/c07/) - Ch.8 Forced Stratification and Adjustment of Multi-Branch Reports: the Closed Loop, the Tool and a Pre-Registered Test — Chapter 8 sets out the closed loop of forced stratification and adjustment for multi-branch reports, the distilled Stratified Reversal Detector and a pre-registered randomised holdout: when the branch is the only confounder the adjusted estimate is falsified in 0.033 of replications, and with a hidden confounder in 0.637. The book contains 23 runnable Python programs. (https://5dlibrary.com/answers/en/simpson-paradox-back-door-adjustment-branch-data/c08/) 第8章 多店报表强制分层与调整:实证闭环、小工具与预注册方案 — 第八章给出多店报表强制分层与调整的闭环、蒸馏小工具「分层反转检测器」与预注册的随机留出检验:只有分店一种混杂时,调整估计被留出数据推翻的比例为 0.033,存在隐藏混杂时升到 0.637。书中含 23 段可运行的 Python 程序。 (https://5dlibrary.com/answers/zh/simpson-paradox-back-door-adjustment-branch-data/c08/) ## 5DL-B-303 · Regression-Discontinuity Designs at Membership-Tier Thresholds: Local Causal Identification on Either Side of a Spend-Based Upgrade Line 中文书名:断点回归设计与会员等级门槛、消费满额升级线两侧的局部因果识别母典 DOI: pending Cite: Fan, Y. (2026). Regression-Discontinuity Designs at Membership-Tier Thresholds: Local Causal Identification on Either Side of a Spend-Based Upgrade Line (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations are generated by the book's programs. This is general analysis, not individualised business advice, and the book neither recommends nor supplies misleading threshold copy designed to provoke top-ups. Abstract: Comparing customers who reached the VIP spending line with those who did not is the commonest way membership reports measure "the effect of the benefits", and the one most prone to overstatement: customers who can reach the line already visit more and spend more freely. This book argues that only a comparison of customers who fall just on either side of the line identifies the causal effect of the benefits, which is the regression-discontinuity design introduced by Thistlethwaite and Campbell in the Journal of Educational Psychology in 1960 and set out as a practical guide by Imbens and Lemieux in the Journal of Econometrics in 2008. Before writing, the book checked the originals and corrects the blueprint: the 1960 paper studies certificates of merit in a scholarship programme, not membership schemes, and the 2008 guide is a methods paper; neither reports any effect size for membership benefits, so every overstatement ratio in the book comes from illustrative data with fixed seeds. The claim must also carry its limits: only the local effect at the threshold is identified, the design fails when customers can top up precisely, estimates are sensitive to the bandwidth, and global high-order polynomials should not be used. Parameters are illustrative; all calculations are generated by the book's programs. This is general analysis, not individualised business advice, and the book neither recommends nor supplies misleading threshold copy designed to provoke top-ups. Chapter 1 finds a naive gap of 19.005 on illustrative membership data against a true effect at the line of 2.5769, an overstatement of 7.37 times, while the local linear estimate with a bandwidth of 8 is 2.764. Chapter 2 covers triangular-kernel weighted least squares, boundary bias and the local nature of the effect. Chapter 3 covers the bias-variance trade in bandwidth choice and boundary cross-validation. Chapter 4 builds a McCrary-style density test by hand, whose z reaches 24.14 on data with top-ups, and shows the swapping kind of sorting it cannot see, where the predetermined covariate test gives z of 13.14. Chapter 5 shows the bias reaching 1.709 as top-ups grow and examines imprecise control and the donut. Chapter 6 discusses high-order polynomials, placebo thresholds and fuzzy designs. Chapter 7 examines the threshold incentives of the rank-for-count rules in the Book of Lord Shang and the original paper of Thistlethwaite and Campbell. Chapter 8 gives a complete procedure for evaluating a membership-tier threshold, the distilled Threshold Discontinuity Effect Estimator and a pre-registered test. The book contains 23 runnable Python programs. 摘要:把达到 VIP 消费线的客人与没达到的客人相比,是会员报表里最常见的「权益效果」算法,也是最容易高估的算法:能达到线的客人本来就更常来、更舍得花钱。本书主张,只有比较恰好落在消费线两侧的客人,才能识别权益的因果效应,这正是 Thistlethwaite 与 Campbell 1960 年在《教育心理学杂志》上提出、Imbens 与 Lemieux 2008 年在《计量经济学杂志》上整理成实践指南的断点回归设计。落笔前本书核对原文并作蓝图修正:1960 年的论文研究的是奖学金荣誉证书而非会员体系,2008 年的指南是方法论文,二者都没有给出任何会员权益的效应量,书中的高估倍数全部来自固定种子的示意数据;命题也必须带边界陈述:只识别门槛处的局部效应,客人能精确凑单时设计失效,估计对带宽敏感,不宜用全局高阶多项式。书中参数为示意;全部计算由书中程序生成。本书属于非个人化经营建议,不建议也不提供任何为制造凑单而设计的误导性门槛文案。 第一章在示意会员数据上算出朴素差距 19.005,而线上真值为 2.5769,前者是后者的 7.37 倍;带宽为 8 的局部线性估计为 2.764。第二章讲三角核加权最小二乘、边界偏差与效应的局部性。第三章讲带宽选择的偏差方差交换与边界交叉验证。第四章手写 McCrary 式密度检验,凑单数据上 z 值达到 24.14,并演示它看不见的对调式排序,此时前定协变量检验的 z 值为 13.14。第五章显示凑单比例升高时偏差可达 1.709,并检验不精确控制与甜甜圈剔除。第六章讨论高阶多项式、安慰剂断点与模糊断点。 第七章考辨《商君书·境内》以计数定爵的门槛激励结构与 Thistlethwaite 与 Campbell 的原论文。第八章给出会员等级门槛效果评估的完整流程、蒸馏小工具「门槛断点效应估计器」与预注册检验方案。书中含 23 段可运行的 Python 程序。 Chapter answers: - Ch.1 Either Side of the VIP Line: Selection Bias, the Naive Gap and a Local Comparison at the Threshold — Chapter 1 finds a naive gap of 19.005 on illustrative membership data against a true effect at the line of 2.5769, an overstatement of 7.37 times, while the local linear estimate with a bandwidth of 8 is 2.764. (https://5dlibrary.com/answers/en/regression-discontinuity-membership-tier-thresholds/c01/) 第1章 VIP 线两侧:选择偏差、朴素差距与断点处的局部比较 — 第一章在示意会员数据上算出朴素差距 19.005,而线上真值为 2.5769,前者是后者的 7.37 倍;带宽为 8 的局部线性估计为 2.764。 (https://5dlibrary.com/answers/zh/regression-discontinuity-membership-tier-thresholds/c01/) - Ch.2 Local Linear Regression: Triangular-Kernel Weighted Least Squares, Boundary Bias and an Effect That Holds Only at the Threshold — At the line of 60 the discontinuity estimate is 2.764 and the true value at the line is 2.5769. Yet the average benefit over all VIPs is 3.063, and over all customers (if everyone had the benefits) it is 2.001; all three differ. Moving the line to 70 raises the average stickiness of the customers at the line, the true value becomes 3.1538, and the discontinuity estimate on that sample is 2.904. Meanwhile the share of customers who reach the line falls from 0.2488 to 0.0823. (https://5dlibrary.com/answers/en/regression-discontinuity-membership-tier-thresholds/c02/) 第2章 局部线性回归:三角核加权最小二乘、边界偏差与只在门槛处成立的效应 — 在 60 这条线上,断点估计是 2.764,线上真值是 2.5769。可是全部 VIP 的平均权益是 3.063,全部客人(假如人人都有权益)的平均权益是 2.001,三者都不相同。如果把线挪到 70,线上客人的平均黏性更高,真值变成 3.1538,那份样本上的断点估计是 2.904。与此同时,能达到线的客人比例从 0.2488 降到 0.0823。 (https://5dlibrary.com/answers/zh/regression-discontinuity-membership-tier-thresholds/c02/) - Ch.3 Choosing the Bandwidth: Trading Bias for Variance, a Monte Carlo Optimum and Boundary Cross-Validation — At a bandwidth of 2 only 1801 customers are used; the estimate is 2.723 with standard error 0.479, and the interval from 1.784 to 3.662 is wide. From 4 to 16 the estimate moves between 2.676 and 2.954, every interval covers the truth, and the standard error falls from 0.345 to 0.180. At 24 the estimate drops to 2.217 and the upper end of the interval, 2.519, no longer reaches the truth; at 40 nearly all 19581 customers are used, the estimate is just 0.873 and the interval from 0.604 to 1.142 is far from the truth. (https://5dlibrary.com/answers/en/regression-discontinuity-membership-tier-thresholds/c03/) 第3章 带宽选择:偏差与方差的交换、蒙特卡洛最优与边界交叉验证 — 带宽为 2 时只用了 1801 人,估计 2.723,标准误 0.479,区间从 1.784 到 3.662,很宽。带宽从 4 到 16,估计在 2.676 到 2.954 之间摆动,区间全部包住真值,标准误从 0.345 降到 0.180。带宽放到 24,估计掉到 2.217,区间上限 2.519 已经够不到真值;放到 40,几乎用上了全部 19581 人,估计只剩 0.873,区间从 0.604 到 1.142,离真值很远。 (https://5dlibrary.com/answers/zh/regression-discontinuity-membership-tier-thresholds/c03/) - Ch.4 The McCrary Density Test: Local Lines on Either Side of a Histogram, Power and Blind Spots — Chapter 4 builds a McCrary-style density test by hand, whose z reaches 24.14 on data with top-ups, and shows the swapping kind of sorting it cannot see, where the predetermined covariate test gives z of 13.14. (https://5dlibrary.com/answers/en/regression-discontinuity-membership-tier-thresholds/c04/) 第4章 McCrary 密度检验:直方图两侧的局部线性、检验力与盲区 — 第四章手写 McCrary 式密度检验,凑单数据上 z 值达到 24.14,并演示它看不见的对调式排序,此时前定协变量检验的 z 值为 13.14。 (https://5dlibrary.com/answers/zh/regression-discontinuity-membership-tier-thresholds/c04/) - Ch.5 Topping Up to Upgrade: How Precisely Manipulable Spending Breaks the Design — Chapter 5 shows the bias reaching 1.709 as top-ups grow and examines imprecise control and the donut. (https://5dlibrary.com/answers/en/regression-discontinuity-membership-tier-thresholds/c05/) 第5章 凑单升级:可被精确操纵的消费额如何让设计失效 — 第五章显示凑单比例升高时偏差可达 1.709,并检验不精确控制与甜甜圈剔除。 (https://5dlibrary.com/answers/zh/regression-discontinuity-membership-tier-thresholds/c05/) - Ch.6 Three Robustness Checks: High-Order Polynomials, Placebo Thresholds and the Fuzzy Design — If the benefits really act at the line of 60, then at other points where no rule changes, say 48 or 72, next year's spending should not jump. Program p62 places fake thresholds at 36, 42, 48 and 54 below the line and at 66, 72 and 78 above it, each using only customers on the same side (non-VIPs for fake thresholds below the line, VIPs for those above) so that the real jump cannot leak in, with a bandwidth of 6. It then repeats the seven tests on 20 fresh samples and counts the false alarms. (https://5dlibrary.com/answers/en/regression-discontinuity-membership-tier-thresholds/c06/) 第6章 稳健性三件事:高阶多项式、安慰剂断点与模糊断点 — 如果权益真的在 60 这条线上起作用,那么在其他没有任何规则变化的位置,例如 48 或 72,明年消费不应当出现跳跃。程序 p62 在线下方的 36、42、48、54 与线上方的 66、72、78 设置假的断点,各自只用同一侧的客人(线下的假断点只用非 VIP,线上的只用 VIP),以免真实的跳跃渗进来;带宽取 6。然后在另外 20 份新样本上重复这七个检验,统计误报的频率。 (https://5dlibrary.com/answers/zh/regression-discontinuity-membership-tier-thresholds/c06/) - Ch.7 Sources East and West: Rank for Count in the Book of Lord Shang and the Scholarship Discontinuity of 1960 — Under rule A there are 666 units at a count of 24 and 675 at 25, a ratio of 1.014; the distribution is smooth. Under rule B the number at 25 jumps to 2656 while only 65 remain at 24, a ratio of 40.862; every count from 20 to 24 has been hollowed out, and 1981 units in all have been pushed onto the threshold. (https://5dlibrary.com/answers/en/regression-discontinuity-membership-tier-thresholds/c07/) 第7章 东西方原典考辨:《商君书·境内》的计数定爵与 1960 年的奖学金断点 — 规则甲之下,计数在 24 与 25 上的单位分别是 666 个与 675 个,两者之比 1.014,分布平滑。规则乙之下,25 上的单位跳到 2656 个,而 24 上只剩 65 个,比值 40.862;从 20 到 24 的各档都被掏空了一块,共有 1981 个单位被推到门槛上。 (https://5dlibrary.com/answers/zh/regression-discontinuity-membership-tier-thresholds/c07/) - Ch.8 Closing the Loop: Evaluating a Membership-Tier Threshold, the Distilled Tool and a Pre-Registered Plan — Chapter 8 gives a complete procedure for evaluating a membership-tier threshold, the distilled Threshold Discontinuity Effect Estimator and a pre-registered test. The book contains 23 runnable Python programs. (https://5dlibrary.com/answers/en/regression-discontinuity-membership-tier-thresholds/c08/) 第8章 实证闭环:会员等级门槛效果评估、蒸馏小工具与预注册方案 — 第八章给出会员等级门槛效果评估的完整流程、蒸馏小工具「门槛断点效应估计器」与预注册检验方案。书中含 23 段可运行的 Python 程序。 (https://5dlibrary.com/answers/zh/regression-discontinuity-membership-tier-thresholds/c08/) ## 5DL-B-304 · Synthetic Control and the Causal Effect of a Single-Site Redesign: Rebuilding the Counterfactual from Weighted Donor Sites and Placebo Permutation Tests 中文书名:合成控制法与单站改版因果、捐赠站加权重建反事实与安慰剂置换检验母典 DOI: pending Cite: Fan, Y. (2026). Synthetic Control and the Causal Effect of a Single-Site Redesign: Rebuilding the Counterfactual from Weighted Donor Sites and Placebo Permutation Tests (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: On an illustrative panel of thirty-four sites with a true lift of twelve orders a week, before-after reports 37.77, mostly because of the season; weighting thirty-three untouched donor sites with simplex-constrained weights into a synthetic site gives 8.83. Over two hundred panels the root mean squared error is 35.61 for before-after and 5.42 for synthetic control. Abstract: People who run several sites commonly judge a single-site redesign by taking a few weeks before and after the change and subtracting. This book tests that habit with reproducible calculation. On an illustrative panel of thirty-four sites with a true lift of twelve orders a week, before-after reports 37.77, mostly because of the season; weighting thirty-three untouched donor sites with simplex-constrained weights into a synthetic site gives 8.83. Over two hundred panels the root mean squared error is 35.61 for before-after and 5.42 for synthetic control. The method is the synthetic control of Abadie, Diamond and Hainmueller in the Journal of the American Statistical Association in 2010. After checking the original, the book corrects the blueprint: "testable" means a placebo permutation test in which each donor in turn plays the treated site, with a smallest p-value set by the number of donors, and synthetic control holds only when the pre-period fit is good, the donor pool is free of spillover and interpolation bias is under control (Abadie's 2021 review in the Journal of Economic Literature). Parameters are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Chapter 1 takes apart season, market and noise in a before-after comparison. Chapter 2 finds simplex-constrained least-squares weights by projected gradient and by exact enumeration, and shows that with only thirty-six pre weeks the error of free regression rises to 18.22 while simplex weights stay at 5.34. Chapter 3 treats the pre-period root mean squared prediction error, the convex hull and hold-out checks. Chapter 4 gives the placebo permutation test: in the illustrative panel the redesigned site's post/pre ratio ranks second, p is 0.0588, short of five per cent. Chapter 5 treats spillover, the number of donors and interpolation bias. Chapter 6 computes power and in-time placebos and shows that a before-after t-test reports significance often even when there is no effect. Chapter 7 examines "take by kind, give by kind" from the Lesser Pick chapter of the Mozi and Mill's method of difference in A System of Logic. Chapter 8 gives a thirty-four-site evaluation sheet, the distilled Synthetic Control Evaluator and a pre-registered protocol: for a lift of five per cent of volume the single-site test has power of about 0.780 at thirteen weeks; several sites can be pooled by the sum of their ranks, and five sites that each rank third give a combined p of 0.000066. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. 摘要:经营多个站点的人普遍用「改版前后各取几周、相减」来衡量单站改版的效果。本书用可复现的计算检验这条常识。在一个三十四站的示意面板上,真实提升设为每周十二单,前后对比报出 37.77,主要误差来自季节;用单纯形约束权重把三十三个未改版的捐赠站加权成合成对照站,估计为 8.83。两百个面板上,前后对比的均方根误差为 35.61,合成对照为 5.42。方法来自 Abadie、Diamond 与 Hainmueller 2010 年发表于《美国统计协会杂志》的合成控制法,本书核对原文后作蓝图修正:「可检验」指逐个捐赠站当处理站的安慰剂置换检验,其最小 p 值由捐赠站数决定;合成对照只在前期拟合良好、捐赠池无溢出、插值偏差可控时成立(Abadie 2021 年《经济文献杂志》综述)。书中参数为示意;全部计算由书中程序生成;本书属于非个人化经营建议。 第一章拆解前后对比的季节、大盘与噪声。第二章用投影梯度与精确枚举求单纯形约束最小二乘权重,并说明前期只有三十六周时自由回归的估计误差升到 18.22,单纯形权重仍为 5.34。第三章讨论前期均方根预测误差、凸包与留出检验。第四章给出安慰剂置换检验:示意面板中改版站的后前比排第二,p 为 0.0588,未达百分之五。 第五章讨论溢出、捐赠站数与插值偏差。第六章计算检验力与时间安慰剂,并说明前后对比的 t 检验在没有效应时也频繁报出显著。第七章考辨《墨子·小取》「以类取,以类予」与密尔《逻辑体系》的差异法。第八章给出三十四站单站改版评估表、蒸馏小工具「合成对照评估器」与预注册方案:提升为体量百分之五时,单站检验在十三周的检验力约为 0.780;多个站按名次之和合并检验,五个站各排第三时合并 p 为 0.000066。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Three Confounders in a Before-After Comparison: How Season, Market and Noise Pose as a Redesign Effect — Before-after has a mean estimate of 45.64, a bias of 33.64 and a root mean squared error of 35.61. In 0.980 of the panels its error exceeds 6 orders, and in 0.970 of them it reports more than twice the true lift. At this point of the calendar, then, before-after inflates the redesign almost every time. Difference-in-differences lowers the bias to 10.26 with a standard deviation of only 3.75 and a root mean squared error of 10.92; it is stable, but stably wrong, and 0.320 of the panels report more than double. Synthetic control has a mean estimate of 15.90, a bias of 3.90, a standard deviation of 3.77 and a root mean squared error of 5.42; only 0.015 of the panels exceed twice the truth. Measured by root mean squared error, before-after is 6.56 times worse than synthetic control and difference-in-differences 2.01 times worse. (https://5dlibrary.com/answers/en/synthetic-control-single-site-redesign-placebo-tests/c01/) 第1章 前后对比的三重混杂:季节、大盘与噪声怎样冒充改版效应 — 前后对比的平均估计是 45.64,偏差 33.64,均方根误差 35.61;两百个面板里有 0.980 的比例误差超过 6 单,有 0.970 的比例报出了真实提升两倍以上的数。这说明在这样的改版时点,前后对比几乎每一次都会把改版吹得很大。双重差分的偏差降到 10.26,标准差只有 3.75,均方根误差 10.92;它稳定,但稳定地偏,0.320 的面板报出两倍以上的提升。合成对照的平均估计是 15.90,偏差 3.90,标准差 3.77,均方根误差 5.42;超过真实值两倍的只有 0.015。按均方根误差算,前后对比是合成对照的 6.56 倍,双重差分是 2.01 倍。 (https://5dlibrary.com/answers/zh/synthetic-control-single-site-redesign-placebo-tests/c01/) - Ch.2 Simplex-Constrained Least Squares: How Donor Weights Are Found and Why They May Not Be Negative — Chapter 2 finds simplex-constrained least-squares weights by projected gradient and by exact enumeration, and shows that with only thirty-six pre weeks the error of free regression rises to 18.22 while simplex weights stay at 5.34. (https://5dlibrary.com/answers/en/synthetic-control-single-site-redesign-placebo-tests/c02/) 第2章 单纯形约束最小二乘:捐赠站权重怎样求,为什么不许为负 — 第二章用投影梯度与精确枚举求单纯形约束最小二乘权重,并说明前期只有三十六周时自由回归的估计误差升到 18.22,单纯形权重仍为 5.34。 (https://5dlibrary.com/answers/zh/synthetic-control-single-site-redesign-placebo-tests/c02/) - Ch.3 Pre-Period Fit Is the Entry Ticket: Root Mean Squared Prediction Error, the Convex Hull and Hold-Out Checks — On panel zero, synthetic control has a fitting error of 6.39 and a hold-out error of 4.00. Free regression fits to 3.11 but holds out at 15.51. The averages over two hundred panels are clearer: synthetic control fits at 6.57 and holds out at 9.29, a ratio of 1.41; free regression fits at 2.96 and holds out at 19.72, a ratio of 6.66. The plain donor mean fits at 18.09 and holds out at 19.55; the shifted mean fits at 13.63 and holds out at 15.23; the best single donor fits at 13.82 and holds out at 16.39. The last three have ratios close to one because they learn almost nothing during fitting and so cannot overfit, but their absolute errors are all much larger than those of synthetic control. (https://5dlibrary.com/answers/en/synthetic-control-single-site-redesign-placebo-tests/c03/) 第3章 前期拟合是入场券:均方根预测误差、凸包与留出检验 — 在第零号面板上,合成对照的拟合期误差是 6.39,留出期误差 4.00。自由回归的拟合期误差只有 3.11,留出期却是 15.51。两百个面板的平均更清楚:合成对照拟合 6.57、留出 9.29,留出与拟合之比 1.41;自由回归拟合 2.96、留出 19.72,比值 6.66。捐赠站简单平均的拟合误差 18.09,留出 19.55;平移后的平均拟合 13.63,留出 15.23;最好的单个站拟合 13.82,留出 16.39。后三者的比值都接近一,因为它们几乎没有在拟合期「学」到什么,自然也不存在过拟合,但它们的绝对误差都比合成对照大得多。 (https://5dlibrary.com/answers/zh/synthetic-control-single-site-redesign-placebo-tests/c03/) - Ch.4 The Placebo Permutation Test: the Post/Pre Ratio, Ranks and the Floor of the p-Value — Chapter 4 gives the placebo permutation test: in the illustrative panel the redesigned site's post/pre ratio ranks second, p is 0.0588, short of five per cent. (https://5dlibrary.com/answers/en/synthetic-control-single-site-redesign-placebo-tests/c04/) 第4章 安慰剂置换检验:后前比、名次与 p 值的地板 — 第四章给出安慰剂置换检验:示意面板中改版站的后前比排第二,p 为 0.0588,未达百分之五。 (https://5dlibrary.com/answers/zh/synthetic-control-single-site-redesign-placebo-tests/c04/) - Ch.5 The Donor Pool: No Spillover, Enough Donors and Interpolation Bias — With three donors the mean pre-period error is 19.95, 0.745 of the panels fail the fit, and the lift estimate has a root mean squared error of 31.81, hardly better than before-after; the floor of the p-value is 0.2500, so significance is impossible however large the effect. With five donors the failure share is still 0.495 and the floor 0.1667. With ten it drops to 0.125, the lift error is 9.47 and the floor 0.0909, still short of five per cent. With twenty the pre-period error is 7.52, failures 0.030, the lift error 6.78 and the floor 0.0476, crossing the five per cent line for the first time. With thirty-three the pre-period error is 6.84, failures 0.000, the lift error 5.42 and the floor 0.0294. (https://5dlibrary.com/answers/en/synthetic-control-single-site-redesign-placebo-tests/c05/) 第5章 捐赠池:无溢出、捐赠站数与插值偏差 — 只有三个捐赠站时,前期误差平均 19.95,0.745 的面板拟合不及格,提升估计的均方根误差高达 31.81,比前后对比好不了多少;p 值的地板是 0.2500,无论效应多大都不可能显著。五个站时拟合不及格的比例仍有 0.495,地板 0.1667。十个站时拟合不及格降到 0.125,提升误差 9.47,地板 0.0909,仍然够不着百分之五。二十个站时,前期误差 7.52,不及格 0.030,提升误差 6.78,地板 0.0476,第一次跨过百分之五的门槛。三十三个站时,前期误差 6.84,不及格 0.000,提升误差 5.42,地板 0.0294。 (https://5dlibrary.com/answers/zh/synthetic-control-single-site-redesign-placebo-tests/c05/) - Ch.6 Power and the Length of the Evaluation: How Many Weeks to See Five Per Cent — Start with the first row, zero lift. Requiring first place, the false alarm rate is 0.053 at four weeks, 0.040 at eight and 0.020 at twelve, near or below the nominal level, in line with Chapter 4. With a lift of 6 orders a week, two and a half per cent of size, power is only 0.240 to 0.267 and barely grows with more weeks. With 12 orders it is 0.633 at four weeks, 0.733 at eight and 0.773 at twelve; allowing the top two, it reaches 0.867 at twelve weeks. With 18 orders, power is already 0.933 at four weeks. With 24 orders it is 1.000 at twelve weeks. (https://5dlibrary.com/answers/en/synthetic-control-single-site-redesign-placebo-tests/c06/) 第6章 检验力与评估周数:多少周才看得出百分之五 — 先看第一行,提升为零。要求排第一时,四周的误报率是 0.053,八周 0.040,十二周 0.020,都在名义水平附近或以下,与第四章的结论一致。提升为每周 6 单,也就是体量的百分之二点五,检验力只有 0.240 到 0.267,加周数也几乎不涨。提升 12 单时,四周 0.633,八周 0.733,十二周 0.773;放宽到前二名,十二周为 0.867。提升 18 单时,四周就有 0.933。提升 24 单时,十二周为 1.000。 (https://5dlibrary.com/answers/zh/synthetic-control-single-site-redesign-placebo-tests/c06/) - Ch.7 Eastern and Western Sources: "Take by Kind, Give by Kind" in the Mozi and Mill's Method of Difference — Literal twins number only 0.325 per panel on average, and only 0.260 of the panels have at least one. The kind of instance Mill asks for simply does not exist in about three shops in four with thirty-four sites. Before-after, as two moments of the same object, has a bias of +33.64 and a root mean squared error of 35.61. The single twin has a pre-period error of 9.43 and a bias of only +1.86, but a root mean squared error of 8.63. Its bias is small because the shift removes the difference in level, but it brings all of its own noise with it, and the random swings of a single site pass straight into the estimate. The synthetic site has a pre-period error of 6.84, a bias of +3.90 and a root mean squared error of 5.42, the smallest of the three. By averaging several sites it dilutes the noise of any one of them, at the cost of a little bias from interpolation. (https://5dlibrary.com/answers/en/synthetic-control-single-site-redesign-placebo-tests/c07/) 第7章 东西方原典考辨:《墨子·小取》「以类取,以类予」与密尔的差异法 — 字面意义上的孪生站,平均每个面板只有 0.325 个,有至少一个的面板只占 0.260。密尔要求的那种事例,在三十四个站的店里大约四次里有三次根本不存在。前后对比作为同一对象的两个时刻,偏差 +33.64,均方根误差 35.61。单个孪生站的前期误差 9.43,偏差只有 +1.86,均方根误差却有 8.63:它的偏差小,是因为平移去掉了水平差,可它带着自己全部的噪声,单个站的随机起伏原样进入了估计。合成站的前期误差 6.84,偏差 +3.90,均方根误差 5.42,三者最小。合成站用多个站的加权平均摊薄了单个站的噪声,代价是一点插值造成的偏差。 (https://5dlibrary.com/answers/zh/synthetic-control-single-site-redesign-placebo-tests/c07/) - Ch.8 Closing the Loop: Evaluating Single-Site Redesigns across Thirty-Four Sites, with a Pre-Registered Protocol — Chapter 8 gives a thirty-four-site evaluation sheet, the distilled Synthetic Control Evaluator and a pre-registered protocol: for a lift of five per cent of volume the single-site test has power of about 0.780 at thirteen weeks; several sites can be pooled by the sum of their ranks, and five sites that each rank third give a combined p of 0.000066. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/synthetic-control-single-site-redesign-placebo-tests/c08/) 第8章 实证闭环:三十四站单站改版评估与预注册方案 — 第八章给出三十四站单站改版评估表、蒸馏小工具「合成对照评估器」与预注册方案:提升为体量百分之五时,单站检验在十三周的检验力约为 0.780;多个站按名次之和合并检验,五个站各排第三时合并 p 为 0.000066。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/synthetic-control-single-site-redesign-placebo-tests/c08/) ## 5DL-B-305 · CUSUM Control Charts and Average Run Length: the Fastest Alarm within a False-Alarm Budget for Latency Drift and Silent Conversion Drops 中文书名:累积和控制图与平均运行长度、延迟漂移与转化暗降的最快无误报告警母典 DOI: pending Cite: Fan, Y. (2026). CUSUM Control Charts and Average Run Length: the Fastest Alarm within a False-Alarm Budget for Latency Drift and Silent Conversion Drops (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice, and Chapter 8 is an offline model that reads no server. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. Abstract: People who run online services commonly believe that one well-chosen fixed red line each for P99 latency and conversion is enough. This book overturns that belief with average run lengths: with false alarms kept equally rare, a fixed threshold reacts very slowly to small persistent drifts, while a cumulative sum (CUSUM) chart cuts the detection delay to a fraction. The project blueprint attributed the claim to papers by Page of 1954 and Moustakides of 1986; after checking, the book corrects the blueprint: Page proposed the CUSUM, the worst-case delay criterion and asymptotic optimality come from Lorden in 1971, and Moustakides proved exact optimality; the specific ratios are consequences computed by the book's programs, not figures from the papers; and "without false positives" in the working title means fastest within an agreed false-alarm budget, not free of false alarms. Parameters are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice, and Chapter 8 is an offline model that reads no server. Chapter 1 shows that, with the same budget of one false alarm in 370 days, a drift of half a standard deviation is found by a fixed threshold after 88.9 days on average and by a CUSUM after 27.7 days, a ratio of 3.21; for a large jump of three standard deviations the ratio reverses to 0.76. Chapter 2 shows that the one-line CUSUM recursion is the log-likelihood ratio maximised over all change points. Chapter 3 solves the Markov chain of Brook and Evans by hand-written Gaussian elimination and shows that a CUSUM with one false alarm a year on average still raises at least one within a year with probability 0.628. Chapter 4 designs limits with Siegmund's approximation and shows that detection delay grows only with the logarithm of the false-alarm interval. Chapter 5 gives design tables for the reference value and the limit and prices a wrong guess of the shift and a wrong baseline. Chapter 6 marks the boundaries of optimality: autocorrelation, residual charts and large jumps. Chapter 7 examines the Han Feizi's dike that collapses through ant holes and Shewhart's control chart of 1931. Chapter 8 sets out a daily CUSUM routine for a toll gate's P99 latency and conversion: when conversion falls by a tenth, the CUSUM alarms after 13.7 days on average and a fixed count rule after 52.8; it also presents the distilled CUSUM Alarm Designer and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. 摘要:经营在线服务的人普遍相信,给 P99 延迟和转化率各画一条合理的固定红线就够用了。本书用平均运行长度推翻这条常识:在误报一样稀少的前提下,固定阈值对小幅持续漂移反应极慢,而累积和控制图把检测延迟缩短到几分之一。立项蓝图把这一命题挂在佩奇 1954 年与穆斯塔基德斯 1986 年的论文名下;核对之后,本书作蓝图修正:累积和由佩奇提出,「最坏情况延迟」准则与渐近最优出自洛登 1971 年,穆斯塔基德斯给出精确最优;具体倍数是由程序算出的推论,并非原文数字;书名中的「无误报」只指在约定误报预算下最快,不指不误报。书中参数为示意;全部计算由书中程序生成;本书属于非个人化经营建议,第八章只做离线模型,不访问任何服务器。 第一章证明,在同为 370 天误报一次的预算下,半个标准差的漂移被固定阈值平均 88.9 天发现,被累积和 27.7 天发现,倍数 3.21;三个标准差的大跳变则相反,倍数 0.76。第二章说明累积和的一行递推就是对所有变点取最大值的对数似然比。第三章用手写高斯消元求解布鲁克与埃文斯马尔可夫链,并指出平均一年误报一次的累积和在一年内至少误报一次的概率是 0.628。第四章用西格蒙德近似设计告警限,并说明检测延迟只随误报间隔的对数增长。 第五章给出参考值与告警限的设计表,量化猜错漂移与估错基线的代价。第六章划定最优性的边界:自相关、残差图与大幅跳变。第七章考辨《韩非子·喻老》「千丈之堤,以蝼蚁之穴溃」与休哈特 1931 年的控制图。第八章给出收费站 P99 延迟与转化率的每日累积和流程,转化率下降一成时累积和平均 13.7 天告警,固定计数规则需 52.8 天;并给出蒸馏小工具「累积和告警器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Sluggish Red Line: at Equal False-Alarm Rates the CUSUM Is Several Times Faster — Chapter 1 shows that, with the same budget of one false alarm in 370 days, a drift of half a standard deviation is found by a fixed threshold after 88.9 days on average and by a CUSUM after 27.7 days, a ratio of 3.21; for a large jump of three standard deviations the ratio reverses to 0.76. (https://5dlibrary.com/answers/en/cusum-control-chart-latency-drift-conversion-alarm/c01/) 第1章 红线的迟钝:同等误报率下累积和快出数倍 — 第一章证明,在同为 370 天误报一次的预算下,半个标准差的漂移被固定阈值平均 88.9 天发现,被累积和 27.7 天发现,倍数 3.21;三个标准差的大跳变则相反,倍数 0.76。 (https://5dlibrary.com/answers/zh/cusum-control-chart-latency-drift-conversion-alarm/c01/) - Ch.2 One Line of Recursion: the CUSUM Is a Running Log-Likelihood Ratio — To keep the same false-alarm budget with two sides, the limit must rise from 4.096 to 4.775, since now two sums can each raise a false alarm. The price is paid in detection speed. For a shift of half a standard deviation the average run length grows from 27.7 days to 35.3 days, 7.6 days more; for a quarter of a standard deviation it grows from 81.9 days to 121.6 days, 39.7 days more. The larger the shift, the smaller the price: at one standard deviation it is only 1.4 days. (https://5dlibrary.com/answers/en/cusum-control-chart-latency-drift-conversion-alarm/c02/) 第2章 一行递推:累积和就是在线的对数似然比 — 双侧要守住同样的误报预算,告警限必须从 4.096 提高到 4.775,因为现在有两条累积和都可能误报。代价落在检测速度上:漂移半个标准差时,平均运行长度从 27.7 天增加到 35.3 天,多等 7.6 天;漂移四分之一个标准差时,从 81.9 天增加到 121.6 天,多等 39.7 天;漂移越大,代价越小,到一个标准差只多 1.4 天。 (https://5dlibrary.com/answers/zh/cusum-control-chart-latency-drift-conversion-alarm/c02/) - Ch.3 The Markov Chain of Brook and Evans: the Average Run Length as a System of Linear Equations — Chapter 3 solves the Markov chain of Brook and Evans by hand-written Gaussian elimination and shows that a CUSUM with one false alarm a year on average still raises at least one within a year with probability 0.628. (https://5dlibrary.com/answers/en/cusum-control-chart-latency-drift-conversion-alarm/c03/) 第3章 布鲁克与埃文斯的马尔可夫链:把平均运行长度写成线性方程组 — 第三章用手写高斯消元求解布鲁克与埃文斯马尔可夫链,并指出平均一年误报一次的累积和在一年内至少误报一次的概率是 0.628。 (https://5dlibrary.com/answers/zh/cusum-control-chart-latency-drift-conversion-alarm/c03/) - Ch.4 Siegmund's Approximation and Lorden's Bound: Detection Delay Grows Only with the Logarithm of the False-Alarm Interval — The information per reading is 0.125. With each tenfold increase in the false-alarm interval, the CUSUM's delay rises by 16.2, then 18.1, then 18.4 days, approaching the slope of the bound, the natural logarithm of ten divided by the information, which is 18.4 days. The fixed threshold grows in an entirely different way: each tenfold increase in the false-alarm interval multiplies its delay by 7.1, 7.5 and 7.7. With a false-alarm interval of a hundred thousand days, the CUSUM raises its alarm after 67.5 days on average and the fixed threshold after 12003.2 days. (https://5dlibrary.com/answers/en/cusum-control-chart-latency-drift-conversion-alarm/c04/) 第4章 西格蒙德近似与洛登下界:检测延迟只随误报间隔的对数增长 — 每个观测的信息量是 0.125。误报间隔每扩大十倍,累积和的延迟依次增加 16.2、18.1、18.4 天,逐渐逼近下界的斜率,即十的自然对数除以信息量,等于 18.4 天。固定阈值则完全是另一种增长方式:误报间隔每扩大十倍,它的延迟乘以 7.1、7.5、7.7 倍。误报间隔为十万天时,累积和平均 67.5 天报警,固定阈值要 12003.2 天。 (https://5dlibrary.com/answers/zh/cusum-control-chart-latency-drift-conversion-alarm/c04/) - Ch.5 Design Tables for the Reference Value and the Limit: the Cost of a Wrong Shift and a Wrong Baseline — Now the mean. With a baseline of 14 days, the standard error of the mean is 0.267 standard deviations. If the mean happens to be estimated one standard error too low, the false-alarm interval shrinks to 74.9 days, about one every two and a half months; if one standard error too high, it becomes 2646.6 days, and small drifts take a very long time to be found. With a baseline of 30 days the two directions give 118.2 and 1377.3 days; with 90 days, 186.7 and 776.5 days; and even a full year of 365 days leaves a swing from 261.4 to 531.1 days. (https://5dlibrary.com/answers/en/cusum-control-chart-latency-drift-conversion-alarm/c05/) 第5章 参考值与告警限的设计表:猜错漂移与估错基线的代价 — 再看均值。用 14 天做基线,均值的标准误是 0.267 个标准差;若均值恰好估低了一个标准误,误报间隔只剩 74.9 天,大约两个半月一次;若估高了一个标准误,误报间隔变成 2646.6 天,小幅漂移要很久才能被发现。用 30 天做基线,两个方向分别是 118.2 天和 1377.3 天;用 90 天,分别是 186.7 天和 776.5 天;即使用满一年 365 天,也还有 261.4 天到 531.1 天的摆动。 (https://5dlibrary.com/answers/zh/cusum-control-chart-latency-drift-conversion-alarm/c05/) - Ch.6 The Boundaries of Optimality: Autocorrelation, Residual Charts and Large Jumps — In the combined rule the CUSUM limit rises from 4.096 to 4.148, and the false-alarm interval is still 370.0 days. The cost falls on small drifts and is small: for half a standard deviation the combined rule needs 28.0 days and the pure CUSUM 27.7. The gain falls on large jumps: for four standard deviations the combined rule needs 1.3 days, the pure CUSUM 1.7 and the fixed threshold 1.1; for five standard deviations the three need 1.1, 1.3 and 1.0 days. Nowhere in the table is the combined rule the slowest of the three, and almost nowhere is it noticeably slower than the fastest. (https://5dlibrary.com/answers/en/cusum-control-chart-latency-drift-conversion-alarm/c06/) 第6章 最优性的边界:自相关、残差图与大幅跳变 — 组合规则的累积和告警限从 4.096 上调到 4.148,误报间隔仍为 370.0 天。代价落在小漂移上,而且很小:半个标准差漂移时,组合规则 28.0 天,纯累积和 27.7 天。收益落在大跳变上:漂移四个标准差时,组合规则 1.3 天,纯累积和 1.7 天,固定阈值 1.1 天;漂移五个标准差时,三者分别是 1.1、1.3、1.0 天。组合规则在全表中从未是最慢的那一个,也几乎从未明显慢于最快的那一个。 (https://5dlibrary.com/answers/zh/cusum-control-chart-latency-drift-conversion-alarm/c06/) - Ch.7 Eastern and Western Sources: the Dike of a Thousand Zhang and Shewhart's Three Sigma — Chapter 7 examines the Han Feizi's dike that collapses through ant holes and Shewhart's control chart of 1931. (https://5dlibrary.com/answers/en/cusum-control-chart-latency-drift-conversion-alarm/c07/) 第7章 东西方原典考辨:千丈之堤与休哈特的三倍标准差 — 第七章考辨《韩非子·喻老》「千丈之堤,以蝼蚁之穴溃」与休哈特 1931 年的控制图。 (https://5dlibrary.com/answers/zh/cusum-control-chart-latency-drift-conversion-alarm/c07/) - Ch.8 A Daily CUSUM at the Toll Gate: P99 Latency, Silent Conversion Drops and a Falsification Protocol — Chapter 8 sets out a daily CUSUM routine for a toll gate's P99 latency and conversion: when conversion falls by a tenth, the CUSUM alarms after 13.7 days on average and a fixed count rule after 52.8; it also presents the distilled CUSUM Alarm Designer and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cusum-control-chart-latency-drift-conversion-alarm/c08/) 第8章 收费站的每日累积和:P99 延迟、转化暗降与证伪方案 — 第八章给出收费站 P99 延迟与转化率的每日累积和流程,转化率下降一成时累积和平均 13.7 天告警,固定计数规则需 52.8 天;并给出蒸馏小工具「累积和告警器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cusum-control-chart-latency-drift-conversion-alarm/c08/) ## 5DL-B-306 · Conformal Prediction and Distribution-Free Intervals: Finite-Sample Coverage Guarantees for Order Counts and Order Values 中文书名:共形预测与无分布区间、订单量与客单价有限样本覆盖保证母典 DOI: pending Cite: Fan, Y. (2026). Conformal Prediction and Distribution-Free Intervals: Finite-Sample Coverage Guarantees for Order Counts and Order Values (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. Abstract: Shop owners widely believe that the mean plus or minus one point nine six standard deviations is a 95 per cent prediction interval. This book tests that belief with exact calculation. On thirty heavy-tailed past values the average two-sided coverage of the normal interval is only slightly below nominal, but the right-tail miss rate rises from a nominal 0.025 to 0.0585, and a one-sided 95 per cent stocking bound misses 0.0731 of the time; with an infinite sample the right-tail miss rate is still 0.0382, so a wrong shape does not fade as data accumulate. Split conformal prediction guarantees finite-sample marginal coverage of at least one minus alpha when the data are exchangeable (Lei and colleagues, Journal of the American Statistical Association, 2018; Vovk and colleagues, 2005), and on the same data its mean coverage sits on the theoretical 0.9524 across seven shapes. The book corrects the blueprint: the gap of normal intervals is a simulation result of the book's programs and lies in the right tail and in one-sided bounds, and the Eastern source sentence comes from the Biography of Guo Shoujing in the History of Yuan. Parameters are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Chapter 1 compares the real coverage of both intervals. Chapter 2 proves that the rank must be the calibration size plus one, times one minus alpha, rounded up; without the plus one, twenty calibration values cover only 0.9048. Chapter 3 prices the guarantee and the shape of the score. Chapter 4 shows that coverage given a calibration set follows a Beta law; with twenty calibration values the chance that true coverage is below 0.90 is 0.1216. Chapter 5 separates marginal from conditional coverage: high-traffic days can fall to 0.8798, and a score scaled by traffic brings all three bands back near 0.95. Chapter 6 shows how weekend effects, growth drift and promotion shocks break exchangeability: pooled calibration covers Saturdays only 0.8252 of the time, a frozen calibration set misses 0.3602 of the time under growth, and weekday scaling and recent windows repair it. Chapter 7 examines Guo Shoujing's principle that the calendar rests on measurement and Neyman's confidence intervals of 1937, and shows that a confidence interval for the mean, used as a prediction interval, covers next week only 0.0912 of the time after three hundred weeks of history. Chapter 8 gives a weekly order-count interval sheet, the distilled Conformal Interval Calculator and a pre-registered test; stocking to a one-sided 95 per cent bound, the normal bound runs out 0.0657 of the time and the conformal bound 0.0382. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. 摘要:店主普遍相信「平均数加减一点九六个标准差就是九成五的预测区间」。本书用精确计算检验这条常识。在三十个重尾历史值上,正态区间的双侧平均覆盖率只略低于名义值,但右尾漏网率从名义的 0.025 升到 0.0585,单侧九成五备货上界的漏网率为 0.0731;样本无穷大时右尾漏网率仍是 0.0382,形状错误不会随数据增多而消失。分裂共形预测在数据可交换时保证至少一减 α 的有限样本边际覆盖(Lei 等 2018 年《美国统计协会杂志》;Vovk 等 2005 年专著),同一批数据上七种形状的平均覆盖率都贴着理论值 0.9524。本书作蓝图修正:正态区间的缺口是本书程序的模拟结果,集中在右尾与单侧上界;东方原典句出自《元史·郭守敬传》。书中参数为示意;全部计算由书中程序生成;本书属于非个人化经营建议。 第一章比较两种区间的真实覆盖率。第二章证明名次必须是校准集大小加一后乘以一减 α 再向上取整,少加一时二十个校准值的覆盖率只有 0.9048。第三章计算保证的价钱与得分形状。第四章说明给定校准集时覆盖率服从贝塔分布,二十个校准值时真实覆盖率低于 0.90 的概率为 0.1216。第五章区分边际与条件覆盖,高流量日的覆盖率可跌到 0.8798,按流量缩放的得分把三档都拉回 0.95 附近。 第六章演示周末效应、增长漂移与促销冲击怎样破坏可交换性:合并校准时周六的覆盖率只有 0.8252,冻结的校准集在增长中漏网率升到 0.3602,按星期缩放与近期窗口可以修补。第七章考辨郭守敬「历之本在于测验」与奈曼 1937 年的置信区间,并说明把均值的置信区间当作预测区间时,三百周历史下对下一周的覆盖率只有 0.0912。第八章给出每周订单量预测区间表、蒸馏小工具「共形区间计算器」与预注册检验方案;按九成五单侧上界备货时,正态上界的缺货率为 0.0657,共形上界为 0.0382。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Coverage Gap of Normal Intervals: Heavy-Tailed Orders Push Every Miss to the Right — The Pareto rows tell the same story. With tail index 2.1, where the variance barely exists, two-sided coverage is 0.9429, the right-tail miss rate is 0.0571 and the one-sided bound misses 0.0690 of the time. Note the last column as well: the share of histories, out of the eight thousand, whose true coverage is below 0.90. It is 0.1378 in the log-normal row with shape one and only 0.0770 in the normal row. The sample standard deviation of heavy-tailed data is highly unstable. A history that happens to contain no large order underestimates it, and the interval becomes absurdly narrow. The owner cannot see this, because she holds only one history. (https://5dlibrary.com/answers/en/conformal-prediction-distribution-free-order-intervals/c01/) 第1章 正态区间的覆盖缺口:重尾订单把失误全部推向右尾 — 帕累托分布的几行讲的是同一件事。尾指数为 2.1 时,方差勉强存在,双侧覆盖率 0.9429,右尾漏网率 0.0571,单侧漏网 0.0690。值得注意的是最后一列:在八千组历史中,真实覆盖率不到 0.90 的历史所占比例,对数正态一的一行为 0.1378,正态一行只有 0.0770。重尾数据的样本标准差极不稳定,碰上一段恰好没有大单的历史,标准差被低估,区间就窄得离谱;店主看不到这一点,因为她手里只有一段历史。 (https://5dlibrary.com/answers/zh/conformal-prediction-distribution-free-order-intervals/c01/) - Ch.2 The Correct Rank: Why "Plus One" Matters and Where the Finite-Sample Guarantee Comes From — Chapter 2 proves that the rank must be the calibration size plus one, times one minus alpha, rounded up; without the plus one, twenty calibration values cover only 0.9048. (https://5dlibrary.com/answers/en/conformal-prediction-distribution-free-order-intervals/c02/) 第2章 正确的名次:为什么要「加一」,以及有限样本覆盖保证从哪里来 — 第二章证明名次必须是校准集大小加一后乘以一减 α 再向上取整,少加一时二十个校准值的覆盖率只有 0.9048。 (https://5dlibrary.com/answers/zh/conformal-prediction-distribution-free-order-intervals/c02/) - Ch.3 The Price of the Guarantee: Interval Width, the Shape of the Score and the Extra Paid When Normality Holds — In the log-normal row with parameter one half, the oracle width is 2.289, the median width of the normal interval 2.367 and that of the conformal interval 2.960. In the log-normal row with parameter one, the oracle is 6.958, the normal median 7.008 and the conformal median 10.301. The last line divides the median widths for Pareto 2.1 by the oracle: 0.99 for the normal interval and 1.49 for the conformal one. (https://5dlibrary.com/answers/en/conformal-prediction-distribution-free-order-intervals/c03/) 第3章 保证的价钱:区间宽度、得分形状与正态成立时多付的那一截 — 对数正态零点五一行,神谕宽度是 2.289,正态区间的中位宽度是 2.367,共形区间是 2.960。对数正态一的一行,神谕 6.958,正态中位 7.008,共形中位 10.301。最后一行把帕累托 2.1 的中位宽度除以神谕:正态 0.99,共形 1.49。 (https://5dlibrary.com/answers/zh/conformal-prediction-distribution-free-order-intervals/c03/) - Ch.4 How Large a Calibration Set: Coverage Given the Calibration Set Follows a Beta Law — Chapter 4 shows that coverage given a calibration set follows a Beta law; with twenty calibration values the chance that true coverage is below 0.90 is 0.1216. (https://5dlibrary.com/answers/en/conformal-prediction-distribution-free-order-intervals/c04/) 第4章 校准集要多大:给定校准集时覆盖率服从贝塔分布 — 第四章说明给定校准集时覆盖率服从贝塔分布,二十个校准值时真实覆盖率低于 0.90 的概率为 0.1216。 (https://5dlibrary.com/answers/zh/conformal-prediction-distribution-free-order-intervals/c04/) - Ch.5 Marginal Is Not Conditional: Traffic Bands, Customer Segments and Intervals Too Wide to Use — Chapter 5 separates marginal from conditional coverage: high-traffic days can fall to 0.8798, and a score scaled by traffic brings all three bands back near 0.95. (https://5dlibrary.com/answers/en/conformal-prediction-distribution-free-order-intervals/c05/) 第5章 边际不是条件:流量分档、客群分层与宽到没用的区间 — 第五章区分边际与条件覆盖,高流量日的覆盖率可跌到 0.8798,按流量缩放的得分把三档都拉回 0.95 附近。 (https://5dlibrary.com/answers/zh/conformal-prediction-distribution-free-order-intervals/c05/) - Ch.6 When Exchangeability Breaks: Weekend Effects, Growth Drift, Promotion Shocks and Three Repairs — Chapter 6 shows how weekend effects, growth drift and promotion shocks break exchangeability: pooled calibration covers Saturdays only 0.8252 of the time, a frozen calibration set misses 0.3602 of the time under growth, and weekday scaling and recent windows repair it. (https://5dlibrary.com/answers/en/conformal-prediction-distribution-free-order-intervals/c06/) 第6章 可交换性破裂:周末效应、增长漂移、促销冲击与三种修补 — 第六章演示周末效应、增长漂移与促销冲击怎样破坏可交换性:合并校准时周六的覆盖率只有 0.8252,冻结的校准集在增长中漏网率升到 0.3602,按星期缩放与近期窗口可以修补。 (https://5dlibrary.com/answers/zh/conformal-prediction-distribution-free-order-intervals/c06/) - Ch.7 Classical Sources East and West: Guo Shoujing's "The Calendar Rests on Measurement" and Neyman's Confidence Intervals — Chapter 7 examines Guo Shoujing's principle that the calendar rests on measurement and Neyman's confidence intervals of 1937, and shows that a confidence interval for the mean, used as a prediction interval, covers next week only 0.0912 of the time after three hundred weeks of history. (https://5dlibrary.com/answers/en/conformal-prediction-distribution-free-order-intervals/c07/) 第7章 东西方原典考辨:郭守敬「历之本在于测验」与奈曼的置信区间 — 第七章考辨郭守敬「历之本在于测验」与奈曼 1937 年的置信区间,并说明把均值的置信区间当作预测区间时,三百周历史下对下一周的覆盖率只有 0.0912。 (https://5dlibrary.com/answers/zh/conformal-prediction-distribution-free-order-intervals/c07/) - Ch.8 Closing the Loop: the Weekly Order-Count Interval Sheet, the Conformal Interval Calculator and a Pre-Registered Test — Chapter 8 gives a weekly order-count interval sheet, the distilled Conformal Interval Calculator and a pre-registered test; stocking to a one-sided 95 per cent bound, the normal bound runs out 0.0657 of the time and the conformal bound 0.0382. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/conformal-prediction-distribution-free-order-intervals/c08/) 第8章 实证闭环:每周订单量预测区间表、共形区间计算器与预注册检验 — 第八章给出每周订单量预测区间表、蒸馏小工具「共形区间计算器」与预注册检验方案;按九成五单侧上界备货时,正态上界的缺货率为 0.0657,共形上界为 0.0382。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/conformal-prediction-distribution-free-order-intervals/c08/) ## 5DL-B-307 · Multiple Comparisons and False-Discovery-Rate Control: the Benjamini-Hochberg Step-Up for Testing Dozens of Copy Rules at Once 中文书名:多重比较与伪发现率控制、同时检验数十条文案规则的本杰明尼霍赫伯格阶梯母典 DOI: pending Cite: Fan, Y. (2026). Multiple Comparisons and False-Discovery-Rate Control: the Benjamini-Hochberg Step-Up for Testing Dozens of Copy Rules at Once (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Conversion rates, visitor counts and numbers of rules are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. Abstract: Owners often test dozens of copy rules at once, judge each at p below 0.05, and trust that the champions so chosen are broadly reliable. This book overturns that belief with exact probabilities: when forty rules that all do nothing are tested together, two false champions appear in an average week, and at least one appears with probability 0.8715; in simulation the best false champion of each week claims a relative lift of 34.2 % on average while its true lift is zero. The project blueprint attributed the claim to Benjamini and Hochberg's paper of 1995; after checking, the book corrects the blueprint: "two false champions" is binomial arithmetic, while that paper's contribution was the false discovery rate as a target and a proof that the step-up procedure controls it for independent tests; BH is more permissive than Bonferroni because its target is weaker, not because it is cleverer about the same target; the limits under correlated tests come from Benjamini and Yekutieli's paper of 2001. Conversion rates, visitor counts and numbers of rules are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Chapter 1 gives the exact distribution of false champions and the winner's curse. Chapter 2 compares the rejection sets and adjusted p-values of Bonferroni, Holm, BH and BY on one week of data and explains why the ladder steps up. Chapter 3 separates the false discovery rate from the family-wise error rate and shows that the former is an average. Chapter 4 sweeps the share of real effects, the effect size and the number of rules, draws power curves and tries an adaptive variant that estimates the share of useless rules. Chapter 5 treats positive correlation from a shared control, two-sided tests and negative correlation, and constructs exactly a counterexample in which BH's false discovery rate reaches 0.0750, showing why the BY factor is needed. Chapter 6 converts correction into visitors per arm: 10,960 without correction, 26,476 under Bonferroni, and about 18,555 under BH when a fifth of rules are real. Chapter 7 examines "three men speaking make a tiger" from the Strategies of the Warring States and Bonferroni's inequalities of 1936: three independent reports push the posterior to 0.4219, while reports copied from one source leave it at 0.0089. Chapter 8 adds multiple-rule correction to the weekly review radar, where e-BH lowers the false discovery rate from 0.0896 to 0.0023, and presents the distilled FDR Step-Up Ladder and a pre-registered falsification plan. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. 摘要:经营者常同时检验几十条文案规则,每条以 p 小于 0.05 判胜负,并相信这样选出的冠军大体可信。本书用精确概率推翻这条常识:四十条全部无效的规则同时检验,平均每周冒出两条假冠军,至少一条的概率为 0.8715;模拟中每周最佳假冠军宣称的相对提升平均为 34.2 %,而真实提升为零。立项蓝图把这一命题挂在本杰明尼与霍赫伯格 1995 年的论文名下;核对之后作蓝图修正:「两条假冠军」是二项分布的算术,那篇论文的贡献是提出伪发现率这一目标并证明阶梯算法在独立检验下控制它;BH 比邦费罗尼宽松,是因为目标较弱,而非同一目标下更聪明;相关检验的边界由本杰明尼与耶库铁利 2001 年的论文给出。书中转化率、访客数与规则条数均为示意参数;全部计算由书中程序生成;本书属于非个人化经营建议。 第一章给出假冠军条数的精确分布与赢家诅咒。第二章在同一周数据上比较邦费罗尼、霍尔姆、BH 与 BY 的拒绝集与调整后 p 值,并说明阶梯为何向上走。第三章区分伪发现率与家族错误率,说明前者是平均数。第四章扫描真效应比例、效应大小与规则条数,画出功效曲线,并试验估计无效比例的自适应做法。第五章处理共享对照组的正相关、双侧与负相关,并精确构造一个让 BH 的伪发现率达到 0.0750 的反例,说明 BY 因子的必要。第六章把校正换算成每组访客数:不校正 10,960 名,邦费罗尼 26,476 名,BH 在真规则占两成时约 18,555 名。 第七章考辨《战国策》「三人言而成虎」与邦费罗尼 1936 年的不等式:三份独立报告把后验推到 0.4219,同源转述则停在 0.0089。第八章给周度复盘雷达加上多规则校正,e-BH 把伪发现率从 0.0896 降到 0.0023,并给出蒸馏小工具「伪发现率阶梯器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Two False Champions among Forty Useless Rules — The shape matters more. The chance of no false champion at all is only 0.1285, so out of a hundred such weeks only about thirteen give a clean leaderboard, and the chance of at least one is 0.8715. Exactly one and exactly two occur with probabilities 0.2706 and 0.2777, and two is the most likely count. Three or more happens with probability 0.3233, and even five or more still has 0.0480, close to one week in twenty. (https://5dlibrary.com/answers/en/false-discovery-rate-benjamini-hochberg-copy-rules/c01/) 第1章 四十条无用规则里的两个假冠军 — 更值得看的是分布的形状。一条假冠军都没有的概率只有 0.1285,换句话说,一百个这样的星期里,大约只有十三个星期的冠军榜是干净的;至少一条的概率是 0.8715。恰好一条与恰好两条的概率分别是 0.2706 与 0.2777,最可能出现的条数就是两条。三条及以上的概率是 0.3233,五条及以上也还有 0.0480,接近二十个星期里有一个。 (https://5dlibrary.com/answers/zh/false-discovery-rate-benjamini-hochberg-copy-rules/c01/) - Ch.2 Four Rulers: Bonferroni, Holm, the BH Step-Up and the BY Correction — Rows five and six show where the extra two come from. R18 has p-value 0.001665, above the Bonferroni bar of 0.00125, so Bonferroni turns it down; but it sits fifth, and BH sets the fifth rung at 0.00625, which it clears easily. R33 has p-value 0.002019 against a sixth rung of 0.00750, and clears too. R03, seventh, is a real rule with p-value 0.019500, above the seventh rung of 0.00875, and none of the four rulers rejects it. Real rules can also land outside the bar because of that week's random fluctuation, and no method escapes this. (https://5dlibrary.com/answers/en/false-discovery-rate-benjamini-hochberg-copy-rules/c02/) 第2章 四把尺子:邦费罗尼、霍尔姆、BH 阶梯与 BY 修正 — 看第五、六行就能明白多出来的两条从哪里来。R18 的 p 值是 0.001665,高于邦费罗尼门槛 0.00125,所以邦费罗尼不要它;但它是排第五的规则,BH 给第五级的门槛是 0.00625,它轻松过线。R33 的 p 值是 0.002019,第六级门槛是 0.00750,同样过线。排第七的 R03 是一条真规则,p 值 0.019500,高于第七级门槛 0.00875,四把尺子都没有拒绝它。真规则也会因为当周的随机波动而落在门槛之外,这是任何方法都躲不开的。 (https://5dlibrary.com/answers/zh/false-discovery-rate-benjamini-hochberg-copy-rules/c02/) - Ch.3 False Discovery Rate and Family-Wise Error Rate: Two Different Targets — Without correction the share of weeks with a false champion is 0.8722, matching the exact value of 0.8715, with 1.9939 false champions per week on average. Bonferroni and Holm both come in at 0.0469; when everything is useless the two are identical, because Holm's first rung is the Bonferroni bar and a failed first rung stops everything. BH comes in at 0.0481, with 0.0529 false discoveries on average, only slightly above Bonferroni. BY is at just 0.0115, clearly more cautious. (https://5dlibrary.com/answers/en/false-discovery-rate-benjamini-hochberg-copy-rules/c03/) 第3章 伪发现率与家族错误率:两个不同的目标 — 不校正时,出现假冠军的比例是 0.8722,与精确值 0.8715 相符,平均每周 1.9939 条。邦费罗尼与霍尔姆的比例都是 0.0469,在这种全部无效的情形下两者完全相同,因为霍尔姆的第一级门槛就是邦费罗尼门槛,第一级不过线就全部停止。BH 的比例是 0.0481,平均假发现 0.0529 条,只比邦费罗尼略高一点。BY 只有 0.0115,明显更保守。 (https://5dlibrary.com/answers/zh/false-discovery-rate-benjamini-hochberg-copy-rules/c03/) - Ch.4 Power: Sweeping the Share of Real Effects, the Effect Size and the Number of Rules — The Bonferroni column barely moves, staying between 0.298 and 0.309, because its bar depends only on the number of rules and not on how many are real. Holm rises slightly, from 0.309 to 0.325: the more real rules, the more its bar loosens after the first rungs are rejected. BH's power rises fastest with the share of real rules: 0.347 at 0.05, 0.483 at 0.20, 0.647 at 0.50, 0.706 at 0.70. With more real rules the last passing rung moves further up the ladder and every rung's bar is wider. Meanwhile BH's false discovery rate stays below 0.05 throughout, falling from 0.0460 to 0.0145, in line with the previous chapter's "share of useless rules times q". (https://5dlibrary.com/answers/en/false-discovery-rate-benjamini-hochberg-copy-rules/c04/) 第4章 功效:真效应比例、效应大小与规则条数的扫描 — 邦费罗尼一列几乎不动,始终在 0.298 到 0.309 之间,因为它的门槛只取决于规则条数,与有多少真规则无关。霍尔姆略有上升,从 0.309 升到 0.325:真规则越多,前几级被拒绝后门槛放宽得越多。BH 的功效随真规则比例上升最快:占 0.05 时是 0.347,占 0.20 时是 0.483,占 0.50 时是 0.647,占 0.70 时是 0.706。真规则多,阶梯上过线的位置就靠后,每一级的门槛也就更宽。与此同时,BH 的伪发现率始终在 0.05 以下,从 0.0460 降到 0.0145,与上一章说的「无效比例乘以 q」一致。 (https://5dlibrary.com/answers/zh/false-discovery-rate-benjamini-hochberg-copy-rules/c04/) - Ch.5 Correlated Tests: Shared Controls, Positive Regression Dependency and the BY Correction — Chapter 5 treats positive correlation from a shared control, two-sided tests and negative correlation, and constructs exactly a counterexample in which BH's false discovery rate reaches 0.0750, showing why the BY factor is needed. (https://5dlibrary.com/answers/en/false-discovery-rate-benjamini-hochberg-copy-rules/c05/) 第5章 相关检验:共享对照、正回归相依与 BY 修正 — 第五章处理共享对照组的正相关、双侧与负相关,并精确构造一个让 BH 的伪发现率达到 0.0750 的反例,说明 BY 因子的必要。 (https://5dlibrary.com/answers/zh/false-discovery-rate-benjamini-hochberg-copy-rules/c05/) - Ch.6 Sample Size: Visitors per Rule, and How to Split Traffic around a Shared Control — Chapter 6 converts correction into visitors per arm: 10,960 without correction, 26,476 under Bonferroni, and about 18,555 under BH when a fifth of rules are real. (https://5dlibrary.com/answers/en/false-discovery-rate-benjamini-hochberg-copy-rules/c06/) 第6章 样本量:每条规则要多少访客,共享对照怎么分流 — 第六章把校正换算成每组访客数:不校正 10,960 名,邦费罗尼 26,476 名,BH 在真规则占两成时约 18,555 名。 (https://5dlibrary.com/answers/zh/false-discovery-rate-benjamini-hochberg-copy-rules/c06/) - Ch.7 Eastern and Western Sources: Three Men Make a Tiger, and Bonferroni's Inequalities — Chapter 7 examines "three men speaking make a tiger" from the Strategies of the Warring States and Bonferroni's inequalities of 1936: three independent reports push the posterior to 0.4219, while reports copied from one source leave it at 0.0089. (https://5dlibrary.com/answers/en/false-discovery-rate-benjamini-hochberg-copy-rules/c07/) 第7章 东西方原典考辨:三人成虎与邦费罗尼的不等式 — 第七章考辨《战国策》「三人言而成虎」与邦费罗尼 1936 年的不等式:三份独立报告把后验推到 0.4219,同源转述则停在 0.0089。 (https://5dlibrary.com/answers/zh/false-discovery-rate-benjamini-hochberg-copy-rules/c07/) - Ch.8 Closing the Loop: Multiple-Rule Correction for the Weekly Review Radar, and a Falsification Plan — Chapter 8 adds multiple-rule correction to the weekly review radar, where e-BH lowers the false discovery rate from 0.0896 to 0.0023, and presents the distilled FDR Step-Up Ladder and a pre-registered falsification plan. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/false-discovery-rate-benjamini-hochberg-copy-rules/c08/) 第8章 实证闭环:周度复盘雷达的多规则校正与证伪方案 — 第八章给周度复盘雷达加上多规则校正,e-BH 把伪发现率从 0.0896 降到 0.0023,并给出蒸馏小工具「伪发现率阶梯器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/false-discovery-rate-benjamini-hochberg-copy-rules/c08/) ## 5DL-B-308 · The Bootstrap and Permutation Tests: Resampling Inference for Small Shop Experiments without Normality Assumptions 中文书名:自助法与置换检验、小样本门店实验不靠正态假设的重抽样推断母典 DOI: pending Cite: Fan, Y. (2026). The Bootstrap and Permutation Tests: Resampling Inference for Small Shop Experiments without Normality Assumptions (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Shops run experiments on few and skewed numbers, yet owners habitually click the t test and read the p-value. This book uses programs to overturn the default trust that the t test stays reliable on small samples, while computing its robust side just as carefully. After checking the project blueprint, the book corrects it: the distortion of the t test concentrates in the one-sample case, one-sided tests and unequal groups, and with equal groups, the same shape and two-sided tests the test is conservative; Fisher (1935) and Efron (1979) gave the randomization test and the bootstrap respectively, and that a permutation test controls its type I error under exchangeability is a theorem of testing theory, meaning not above the nominal level rather than exactly equal to it. Parameters are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Chapter 1 finds actual two-sided false alarm rates of 0.1013 for a one-sample t test on ten exponential observations and 0.1840 on five lognormal ones, falling to conservative levels such as 0.0289 for equal groups of the same shape. Chapter 2 moves from the lady's chance of 1/70 to an exact permutation test on twelve trading days, with a two-sided p of 0.0043 where the pooled t test gives 0.1094. Chapter 3 explains the add-one rule and the number of shuffles for Monte Carlo permutation: when the exact p is 0.05901, ninety-nine shuffles call it significant with probability 0.299. Chapter 4 gives percentile and simplified BCa bootstrap intervals; with five exponential observations a nominal ninety-five per cent percentile interval covers 0.798 of the time. Chapter 5 shows that permutation tests also lose control when exchangeability fails: unequal variances with unequal groups give a false alarm rate of 0.250, and a before-and-after comparison gives 0.209 even with no trend. Chapter 6 treats no peeking, no early stopping and pairing: looking every day raises the false alarm rate to 0.1714, and an exact sign-flip test on weekly pairs gives p of 0.0234 against 0.8008 unpaired. Chapter 7 examines the Lüshi Chunqiu line on tasting one slice of meat and Fisher's lady tasting tea. Chapter 8 gives the analysis template for small shop experiments, the distilled Permutation Tester and a pre-registered refutation plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:门店做实验,数据少而偏,店主却习惯点一下 t 检验看 p 值。本书用程序推翻「t 检验在小样本上照样可靠」的默认信任,同时把它稳健的一面算清楚。核对立项蓝图后作蓝图修正:t 检验的失真集中在单样本、单侧与两组不等量的情形,等量、同形、双侧时偏保守;费希尔 1935 年与埃夫龙 1979 年分别给出随机化检验与自助法,「置换检验在交换性下控制第一类错误」是由检验理论证明的定理,并且是不超过名义水平而非恰好等于。书中参数为示意;全部计算由书中程序生成;本书属于非个人化经营建议。 第一章测得,指数总体十个观测的单样本 t 检验两侧实际误报率为 0.1013,对数正态五个观测为 0.1840;两组同形等量时则降到 0.0289 一类的保守水平。第二章从女士品茶的 1/70 讲到十二个营业日的精确置换检验,双侧 p 值为 0.0043,而合并方差 t 检验只有 0.1094。第三章说明蒙特卡洛置换的加一校正与抽样次数,精确 p 为 0.05901 时,只抽九十九次有 0.299 的机会误判为显著。第四章给出百分位与简化 BCa 自助区间,五个指数观测时名义百分之九十五的百分位区间实际覆盖 0.798。 第五章说明交换性被破坏时置换检验同样失控:方差不等且样本量不等时误报率为 0.250,先旧后新的对照在没有趋势时也达 0.209。第六章讲不偷看、不早停与配对:每天看一次使误报率升到 0.1714,按周配对的精确符号翻转检验 p 为 0.0234,不配对则为 0.8008。第七章考辨《吕氏春秋·察今》的「尝一脟肉」与费希尔的女士品茶。第八章给出门店小实验分析模板、蒸馏小工具「置换检验器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Nominal Five Per Cent, Actual False Alarms: the Type I Error of the t Test on Small Skewed Samples — Chapter 1 finds actual two-sided false alarm rates of 0.1013 for a one-sample t test on ten exponential observations and 0.1840 on five lognormal ones, falling to conservative levels such as 0.0289 for equal groups of the same shape. (https://5dlibrary.com/answers/en/bootstrap-permutation-tests-small-shop-experiments/c01/) 第1章 名义五分与实际误报:偏态小样本上 t 检验的第一类错误率 — 第一章测得,指数总体十个观测的单样本 t 检验两侧实际误报率为 0.1013,对数正态五个观测为 0.1840;两组同形等量时则降到 0.0289 一类的保守水平。 (https://5dlibrary.com/answers/zh/bootstrap-permutation-tests-small-shop-experiments/c01/) - Ch.2 From Eight Teacups to Twelve Trading Days: Why the Exact Permutation Test Does Not Over-Report — Chapter 2 moves from the lady's chance of 1/70 to an exact permutation test on twelve trading days, with a two-sided p of 0.0043 where the pooled t test gives 0.1094. (https://5dlibrary.com/answers/en/bootstrap-permutation-tests-small-shop-experiments/c02/) 第2章 从八只茶杯到十二个营业日:精确置换检验为何不会多报 — 第二章从女士品茶的 1/70 讲到十二个营业日的精确置换检验,双侧 p 值为 0.0043,而合并方差 t 检验只有 0.1094。 (https://5dlibrary.com/answers/zh/bootstrap-permutation-tests-small-shop-experiments/c02/) - Ch.3 Splits Too Many to List: Seeded Monte Carlo Permutation and the Add-One Rule — Chapter 3 explains the add-one rule and the number of shuffles for Monte Carlo permutation: when the exact p is 0.05901, ninety-nine shuffles call it significant with probability 0.299. (https://5dlibrary.com/answers/en/bootstrap-permutation-tests-small-shop-experiments/c03/) 第3章 列不完的分组:固定种子蒙特卡洛置换与加一校正 — 第三章说明蒙特卡洛置换的加一校正与抽样次数,精确 p 为 0.05901 时,只抽九十九次有 0.299 的机会误判为显著。 (https://5dlibrary.com/answers/zh/bootstrap-permutation-tests-small-shop-experiments/c03/) - Ch.4 Bootstrap Intervals: Percentile, Bias-Corrected and Accelerated, and Undercoverage in Small Samples — Chapter 4 gives percentile and simplified BCa bootstrap intervals; with five exponential observations a nominal ninety-five per cent percentile interval covers 0.798 of the time. (https://5dlibrary.com/answers/en/bootstrap-permutation-tests-small-shop-experiments/c04/) 第4章 自助法的区间:百分位、偏差校正加速与小样本的覆盖不足 — 第四章给出百分位与简化 BCa 自助区间,五个指数观测时名义百分之九十五的百分位区间实际覆盖 0.798。 (https://5dlibrary.com/answers/zh/bootstrap-permutation-tests-small-shop-experiments/c04/) - Ch.5 When Exchangeability Fails: Unequal Variances, Before-and-After Comparisons and Correlated Days Break the Permutation Test Too — Chapter 5 shows that permutation tests also lose control when exchangeability fails: unequal variances with unequal groups give a false alarm rate of 0.250, and a before-and-after comparison gives 0.209 even with no trend. (https://5dlibrary.com/answers/en/bootstrap-permutation-tests-small-shop-experiments/c05/) 第5章 交换性失效:方差不等、先后对照与日间相关让置换检验同样失控 — 第五章说明交换性被破坏时置换检验同样失控:方差不等且样本量不等时误报率为 0.250,先旧后新的对照在没有趋势时也达 0.209。 (https://5dlibrary.com/answers/zh/bootstrap-permutation-tests-small-shop-experiments/c05/) - Ch.6 Designing a Shop Experiment: No Peeking, No Early Stopping, Pair First, Size by Simulation — Chapter 6 treats no peeking, no early stopping and pairing: looking every day raises the false alarm rate to 0.1714, and an exact sign-flip test on weekly pairs gives p of 0.0234 against 0.8008 unpaired. (https://5dlibrary.com/answers/en/bootstrap-permutation-tests-small-shop-experiments/c06/) 第6章 门店实验的设计:不偷看、不早停、先配对、按模拟定天数 — 第六章讲不偷看、不早停与配对:每天看一次使误报率升到 0.1714,按周配对的精确符号翻转检验 p 为 0.0234,不配对则为 0.8008。 (https://5dlibrary.com/answers/zh/bootstrap-permutation-tests-small-shop-experiments/c06/) - Ch.7 Eastern and Western Sources: Tasting One Slice of Meat in the Lüshi Chunqiu and Fisher's Lady Tasting Tea — With d equal to zero, the simulated rejection rates of the four designs are 0.0521, 0.0151, 0.0408 and 0.0358, matching the exact values 0.0500, 0.0143, 0.0400 and 0.0352 in the first row, which shows the simulation is sound. With d equal to one, the eight-cup design passes her only 0.1412 of the time and the twelve-cup design 0.3595; with d equal to two, 0.4934 and 0.8313; with d equal to three, 0.8349 and 0.9878. (https://5dlibrary.com/answers/en/bootstrap-permutation-tests-small-shop-experiments/c07/) 第7章 东西方原典考辨:《吕氏春秋·察今》的「尝一脟肉」与费希尔的女士品茶 — d 为零时,四种设计的模拟拒绝率分别是 0.0521、0.0151、0.0408、0.0358,与第一行的精确值 0.0500、0.0143、0.0400、0.0352 相符,说明模拟正确。d 为一时,八杯设计只有 0.1412 的机会通过,十二杯设计有 0.3595;d 为二时,八杯设计 0.4934,十二杯设计 0.8313;d 为三时,八杯设计 0.8349,十二杯设计 0.9878。 (https://5dlibrary.com/answers/zh/bootstrap-permutation-tests-small-shop-experiments/c07/) - Ch.8 An Analysis Template for Small Shop Experiments, the Permutation Tester and a Plan to Refute the Book — Chapter 8 gives the analysis template for small shop experiments, the distilled Permutation Tester and a pre-registered refutation plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/bootstrap-permutation-tests-small-shop-experiments/c08/) 第8章 门店小实验分析模板、置换检验器与证伪方案 — 第八章给出门店小实验分析模板、蒸馏小工具「置换检验器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/bootstrap-permutation-tests-small-shop-experiments/c08/) ## 5DL-B-309 · Condorcet's Jury Theorem with Correlated Errors: When Multi-Model Voting Gets More Accurate and When Shared Origins Make Everyone Wrong Together 中文书名:孔多塞陪审团定理与相关误差、多模型投票何时越投越准何时同源共错母典 DOI: pending Cite: Fan, Y. (2026). Condorcet's Jury Theorem with Correlated Errors: When Multi-Model Voting Gets More Accurate and When Shared Origins Make Everyone Wrong Together (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is general business analysis, not business advice. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. Abstract: Many owners let several artificial-intelligence models vote, or ask one model several times and take the majority, believing that a few more votes will make a verdict safe. The practice rests on Condorcet's jury theorem of 1785: if each vote is right with probability above one half and the votes are independent, the more votes there are, the more likely the majority is right. This book overturns the popular version of that theorem in multi-model voting: when models share origins and their errors are correlated, majority accuracy has a ceiling below 1. At single-vote accuracy 0.70, five independent votes give a majority accuracy of 0.8369; in the exchangeable Beta-binomial model the ceiling is 0.8996 at a correlation of 0.1 and only 0.7731 at 0.3. After checking the sources the book corrects its blueprint: Ladha's paper of 1992 gives sufficient conditions under which the theorem survives weak correlation, the ceiling below 1 comes from the models of Berg (1993) and Boland (1989), and under weak correlation more votes can still bring more accuracy. Its parameters are illustrative; all calculations are generated by its programs; it is general business analysis, not business advice. Chapter 1 works out Condorcet's independent case and the votes it requires. Chapter 2 computes the ceiling with the Beta-binomial and opinion-leader models. Chapter 3 measures correlation from a right-wrong table and shows how plurality voting loses its edge when wrong answers collide. Chapter 4 compares decorrelation strategies: resampling one model has a ceiling of only 0.7366, five different models a ceiling of 0.8257, and four models plus a standard-library check that decides wherever it applies reach 0.8887. Chapter 5 treats log-odds weighting and the double counting of models from one family. Chapter 6 works out how many models and how many labelled questions are needed. Chapter 7 examines "when three read, follow the two" in the Great Plan chapter of the Book of Documents and Condorcet's Essai, analysing the voting structure only and making no reading of any kind: under independence a margin of 5 makes the majority right with probability 0.9857, but at a correlation of 0.1 with a hundred and one votes only 0.5819; when two of three agree at a correlation of 0.3, the majority is right with probability 0.6077. Chapter 8 gives the empirical loop for decorrelated multi-model verdicts, the distilled Correlated Voting Accuracy Calculator and a pre-registered plan of refutation. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:很多经营者让多个人工智能模型投票,或让同一个模型多答几次再取多数,以为多投几票就稳了。这个做法依据的是孔多塞 1785 年的陪审团定理:每一票判对的概率超过一半且各票独立时,票数越多,多数越可能正确。本书推翻它在多模型投票中的常识版本:模型同源、错误相关时,多数准确率有一个低于 1 的上限。单票准确率 0.70 时,五张独立票的多数准确率是 0.8369;在可交换的贝塔二项模型里,相关系数 0.1 时上限是 0.8996,相关系数 0.3 时只有 0.7731。核对原文之后,本书做了蓝图修正:拉达 1992 年的论文给出的是相关足够弱时定理仍然成立的充分条件,上限低于 1 的结论来自伯格 1993 年与博兰 1989 年的模型;弱相关之下多数仍可越投越准。书中参数为示意;全部计算由书中程序生成;本书是非个人化的经营分析,不构成经营建议。 第一章算孔多塞的独立情形与所需票数。第二章用贝塔二项模型与意见领袖模型算出上限。第三章从对错表中量出相关,并说明错答扎堆时相对多数会失去优势。第四章比较去相关策略:同一模型重采样的上限只有 0.7366,五家不同模型的上限是 0.8257,让标准库验算在能覆盖的题上直接拍板,四个模型加验算可达 0.8887。第五章讨论对数几率加权与同族模型的重复计票。第六章算所需模型数与样本量。 第七章考辨《尚书·洪范》「三人占,则从二人之言」与孔多塞的《论文》,只分析投票结构,不做任何占断:独立时票差为 5 的多数判对概率是 0.9857,相关系数 0.1、一百零一票时只有 0.5819;三人中二人一致时,相关系数 0.3 下多数判对的概率是 0.6077。第八章给出多模型裁决去相关的实证闭环、蒸馏小工具「相关投票准确率计算器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Condorcet's Binomial Majority: Why Independent Votes Grow More Accurate — The five columns of panel A correspond to single-vote accuracies of 0.45, 0.55, 0.60, 0.70 and 0.80. Take the column for 0.70. One vote gives 0.7000, three give 0.7840, five give 0.8369, fifteen give 0.9500 and fifty-one give 0.9986. With enough voters, the majority almost never errs. That is the bright side of Condorcet's theorem: independent judgements that are only modestly better than a coin toss can be pooled into something close to certainty. (https://5dlibrary.com/answers/en/condorcet-jury-theorem-correlated-multi-model-voting/c01/) 第1章 孔多塞的二项多数:独立投票为什么越投越准 — A 段的五列分别对应单票准确率 0.45、0.55、0.60、0.70 与 0.80。看 0.70 这一列:一票时是 0.7000,三票时升到 0.7840,五票时 0.8369,十五票时 0.9500,五十一票时 0.9986。只要投票者够多,多数意见几乎不会出错。这就是孔多塞定理的正面:独立的、略好于抛硬币的判断,积少成多,可以变成近乎确定的判断。 (https://5dlibrary.com/answers/zh/condorcet-jury-theorem-correlated-multi-model-voting/c01/) - Ch.2 Correlated Errors and the Ceiling: the Beta-Binomial Model, Opinion Leaders and Shared Origins — Whenever errors share a source, majority accuracy has a ceiling below 1. In the Beta-binomial model the ceiling is the share of questions on the right side of one half; in the opinion-leader model it depends on whether the copying probability crosses $1-1/(2p)$. Under weak correlation, more votes still bring more accuracy, at a discount; under moderately strong correlation the ceiling lies far below 1, and more votes only mark time or even make matters worse. The next chapter turns to a practical question: what is the correlation in the owner's own pool of models, and how can it be measured from data? (https://5dlibrary.com/answers/en/condorcet-jury-theorem-correlated-multi-model-voting/c02/) 第2章 相关误差与上限:贝塔二项模型、意见领袖与同源共错 — 只要错误有共同来源,多数准确率就有一个低于 1 的上限。这个上限在贝塔二项模型里等于难度低于一半的题所占的比例,在意见领袖模型里取决于跟随概率是否越过 $1-1/(2p)$。弱相关时,越投越准仍然成立,但收益被打了折扣;相关稍强时,上限远低于 1,再多的票也只是在原地踏步,甚至越投越差。下一章讨论一个实际问题:店主手里的模型池,相关系数到底是多少,怎样从数据里量出来。 (https://5dlibrary.com/answers/zh/condorcet-jury-theorem-correlated-multi-model-voting/c02/) - Ch.3 Measuring Correlation: Pairwise Correlation, the Q Statistic, Effective Votes and Colliding Wrong Answers — Correlation can be measured, and it should be. The phi coefficient and the moment estimate correspond directly to $\rho$ in the model; the Q statistic is more sensitive but runs on a different scale; the both-wrong share is the most intuitive. Once $\rho$ is known, the cap $1/\rho$ on effective votes tells the owner how much independent information remains to be bought, and multiple-choice reviews need one more measurement, the collision rate. The next chapter turns to action: since correlation comes from common causes, which methods can break it, and how much does each break? (https://5dlibrary.com/answers/en/condorcet-jury-theorem-correlated-multi-model-voting/c03/) 第3章 怎样量出相关:两两相关、Q 统计量、等效票数与错答撞车 — 相关可以量,而且应该量。phi 系数与矩估计直接对应模型里的 $\rho$,Q 统计量更敏感但刻度不同,同时判错的比例最直观。量出 $\rho$ 之后,等效票数的上界 $1/\rho$ 告诉店主还有多少独立信息可买;多选题还要再量一笔撞车率。下一章转向行动:既然相关来自共同原因,有哪些办法能把它拆掉,各能拆掉多少。 (https://5dlibrary.com/answers/zh/condorcet-jury-theorem-correlated-multi-model-voting/c03/) - Ch.4 Comparing Ways to Decorrelate: Resampling, Different Prompts, Different Models and Standard-Library Checks — Chapter 4 compares decorrelation strategies: resampling one model has a ceiling of only 0.7366, five different models a ceiling of 0.8257, and four models plus a standard-library check that decides wherever it applies reach 0.8887. (https://5dlibrary.com/answers/en/condorcet-jury-theorem-correlated-multi-model-voting/c04/) 第4章 去相关策略比较:重采样、异提示、异模型与标准库验算 — 第四章比较去相关策略:同一模型重采样的上限只有 0.7366,五家不同模型的上限是 0.8257,让标准库验算在能覆盖的题上直接拍板,四个模型加验算可达 0.8887。 (https://5dlibrary.com/answers/zh/condorcet-jury-theorem-correlated-multi-model-voting/c04/) - Ch.5 Weighted Majorities: Log-Odds Weights and the Double Counting of Correlated Clusters — Under independence, log-odds weighting is the optimal rule, and when one vote's weight exceeds the sum of all others it should decide alone, which gives Chapter 4's deciding check its theoretical footing. Under correlation, log-odds weights count the same evidence twice, and family votes should be merged into one before weighting; the boundary between merging and counting lies roughly between within-family correlations of 0.4 and 0.5. The next chapter returns to the budget: below the ceiling, how many models does a target accuracy require, and how many questions with known answers are needed to measure accuracy and correlation precisely? (https://5dlibrary.com/answers/en/condorcet-jury-theorem-correlated-multi-model-voting/c05/) 第5章 加权多数:对数几率权重与相关簇的重复计票 — 独立时,按对数几率加权是最优规则;当某一票的权重超过其余所有票之和,它就应当一票定夺,这为第四章的验算拍板提供了理论依据。相关时,对数几率权重会把同一份证据算两遍,此时应先把同族的票合成一票再加权,合与不合的分界大致在族内相关 0.4 到 0.5 之间。下一章回到预算问题:在上限之下,要多少个模型才能达到目标准确率,又要多少道带标准答案的题,才能把准确率与相关系数量准。 (https://5dlibrary.com/answers/zh/condorcet-jury-theorem-correlated-multi-model-voting/c05/) - Ch.6 How Many Models and How Many Labelled Questions: the Vote Ledger and the Labelling Ledger below the Ceiling — Under correlation the vote ledger fills with "never": the higher the target, the less correlation it tolerates, and beyond the threshold no number of extra calls helps. The labelling ledger reminds the owner that all these judgements rest on estimates whose precision improves only with the square root of the number of questions; three hundred questions are the starting point and a thousand the check near a threshold. The next chapter leaves the shop floor for a while and returns to two classical sources: the majority rule "when three read, follow the two" in the Great Plan chapter of the Book of Documents, and Condorcet's Essai of 1785. (https://5dlibrary.com/answers/en/condorcet-jury-theorem-correlated-multi-model-voting/c06/) 第6章 所需模型数与样本量:上限之下的票数账与标注账 — 相关之下,票数账里出现了大片的 never:目标越高,能容忍的相关越低,越过门槛之后多调用多少次都无济于事。标注账则提醒店主,所有这些判断都建立在估计值上,估计值的精度按题量的平方根改善;三百道题是起点,一千道题是靠近门槛时的复核量。下一章暂时离开经营现场,回到两部原典:《尚书·洪范》里「三人占,则从二人之言」的多数决,以及孔多塞 1785 年的《论文》。 (https://5dlibrary.com/answers/zh/condorcet-jury-theorem-correlated-multi-model-voting/c06/) - Ch.7 Classical Sources East and West: "When Three Read, Follow the Two" in the Great Plan and Condorcet's Essai of 1785 — Chapter 7 examines "when three read, follow the two" in the Great Plan chapter of the Book of Documents and Condorcet's Essai, analysing the voting structure only and making no reading of any kind: under independence a margin of 5 makes the majority right with probability 0.9857, but at a correlation of 0.1 with a hundred and one votes only 0.5819; when two of three agree at a correlation of 0.3, the majority is right with probability 0.6077. (https://5dlibrary.com/answers/en/condorcet-jury-theorem-correlated-multi-model-voting/c07/) 第7章 东西方原典考辨:《尚书·洪范》「三人占,则从二人之言」与孔多塞 1785 年的《论文》 — 第七章考辨《尚书·洪范》「三人占,则从二人之言」与孔多塞的《论文》,只分析投票结构,不做任何占断:独立时票差为 5 的多数判对概率是 0.9857,相关系数 0.1、一百零一票时只有 0.5819;三人中二人一致时,相关系数 0.3 下多数判对的概率是 0.6077。 (https://5dlibrary.com/answers/zh/condorcet-jury-theorem-correlated-multi-model-voting/c07/) - Ch.8 Closing the Loop on Decorrelated Multi-Model Verdicts: a Procedure, the Correlated Voting Accuracy Calculator and a Pre-Registered Plan of Refutation — Chapter 8 gives the empirical loop for decorrelated multi-model verdicts, the distilled Correlated Voting Accuracy Calculator and a pre-registered plan of refutation. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/condorcet-jury-theorem-correlated-multi-model-voting/c08/) 第8章 多模型裁决去相关的实证闭环:流程、相关投票准确率计算器与预注册证伪方案 — 第八章给出多模型裁决去相关的实证闭环、蒸馏小工具「相关投票准确率计算器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/condorcet-jury-theorem-correlated-multi-model-voting/c08/) ## 5DL-B-310 · Arrow's Impossibility Theorem and the Aggregation of Multiple Reviewers: Why No Perfect Arbiter Exists When Kit Refusal Codes Conflict 中文书名:阿罗不可能定理与多评审规则聚合、五件套拒绝码冲突时为何不存在完美裁决器母典 DOI: pending Cite: Fan, Y. (2026). Arrow's Impossibility Theorem and the Aggregation of Multiple Reviewers: Why No Perfect Arbiter Exists When Kit Refusal Codes Conflict (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Reviewer names and parameters are illustrative, all calculations are generated by the book's programs, and the book is general business analysis, not business advice. Chapter 6 builds a scenario of conflicting kit refusal codes with illustrative rule names, computes deadlock risks and recommends a hybrid rule that vetoes first and then compares grades in written priority order. Abstract: When several checks disagree, the usual hope is that a clever enough merging rule will always produce a fair, consistent and unimpeachable verdict. This book overturns that belief. Arrow proved in 1950 and 1951 that with three or more options no rule for merging rankings satisfies unrestricted domain, weak Pareto, independence of irrelevant alternatives and non-dictatorship at once. After checking the original sources the book corrects its blueprint: the common four-axiom statement follows the second edition of 1963, the first edition used different conditions, and Blau showed in 1957 that its proof was flawed; the cycle probability of one in eighteen holds only under the impartial culture, which is exactly where cycles are most likely, and real ballots rarely show cycles; Borda's memoir appears in the volume for 1781, printed in 1784; and a lexicographic priority keeps the other three axioms by giving up non-dictatorship. Reviewer names and parameters are illustrative, all calculations are generated by the book's programs, and the book is general business analysis, not business advice. Chapter 1 enumerates all 262,144 rules that satisfy Pareto and independence for three reviewers and three options and finds that only dictators guarantee transitivity. Chapter 2 computes the exact probability of a Condorcet cycle: 0.0556 for three reviewers, rising towards 0.0877 as the panel grows, and rising fast as options are added. Chapter 3 recomputes Borda's twenty-one electors and finds a spoiler in a share of 0.2604 of profiles with three reviewers and four options. Chapter 4 shows how single-peaked preferences and Sen's value restriction remove cycles, and how a little off-axis noise brings them back. Chapter 5 audits the lexicographic priority and graded rules and illustrates the Gibbard and Satterthwaite theorem with a census of manipulation: with four options Borda is manipulable in a share of 0.5122 of profiles, the priority rule in none. Chapter 6 builds a scenario of conflicting kit refusal codes with illustrative rule names, computes deadlock risks and recommends a hybrid rule that vetoes first and then compares grades in written priority order. Chapter 7 examines the sealed names, recopied papers and "follow either the first or the review" of the Treatise on Selection in the Song History, and Borda's equal-step scoring. Chapter 8 gives a seven-step verdict procedure, the distilled Reviewer Aggregation Paradox Detector and a pre-registered plan of refutation. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:多道检查意见不一时,经营者最常见的想法是:只要合成规则设计得足够聪明,总能得到一个公正、一致、无可挑剔的裁决。本书推翻这个常识。阿罗 1950 年与 1951 年证明,选项不少于三个时,不存在同时满足无限制域、弱帕累托、无关选项独立与非独裁的排序合成规则。核对原始文献后,本书做了蓝图修正:通行的四公理是 1963 年第二版之后的写法,初版条件不同,布劳 1957 年指出初版证明有缺陷;十八分之一的循环概率只在不偏文化下成立,而不偏文化恰是循环最易出现的情形,真实选票中循环少见;博尔达备忘录载于 1781 年卷、1784 年印行;字典序优先级保住其余三条,放弃的正是非独裁。书中检查名与参数均为示意,全部计算由书中程序生成;本书是非个人化的经营分析,不构成经营建议。 第一章在三评审三选项上穷举所有满足帕累托与无关选项独立的 262,144 种规则,能保证传递的只有独裁者。第二章精确算出孔多塞循环的概率:三评审为 0.0556,评审增多时趋向 0.0877,选项增多时迅速上升。第三章重算博尔达的二十一名选举人,并发现三评审四选项时 0.2604 的剖面存在搅局者。第四章说明单峰偏好与森的值域限制让循环消失,而少量离轴噪音又把它带回。 第五章审计字典序优先级与分档规则,并用操纵普查说明吉巴德与萨特思韦特定理:四选项时博尔达在 0.5122 的剖面可被操纵,优先级规则为零。第六章用示意规则名搭出五件套拒绝码冲突场景,计算死锁概率,并推荐先一票否决、再按书面优先级比较分档的混合规则。第七章考辨《宋史·选举志》的糊名、誊录与「或从初,或从覆」,以及博尔达的等差计分。第八章给出七步裁决流程、蒸馏小工具「评审聚合悖论检测器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Four Axioms and an Empty Table: Arrow's Theorem Enumerated for Three Reviewers and Three Options — Chapter 1 enumerates all 262,144 rules that satisfy Pareto and independence for three reviewers and three options and finds that only dictators guarantee transitivity. (https://5dlibrary.com/answers/en/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c01/) 第1章 四条公理与一张空表:在三评审三选项上穷举阿罗定理 — 第一章在三评审三选项上穷举所有满足帕累托与无关选项独立的 262,144 种规则,能保证传递的只有独裁者。 (https://5dlibrary.com/answers/zh/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c01/) - Ch.2 The Exact Probability of a Condorcet Cycle: 216 Profiles, Panel Size and the Number of Options — Chapter 2 computes the exact probability of a Condorcet cycle: 0.0556 for three reviewers, rising towards 0.0877 as the panel grows, and rising fast as options are added. (https://5dlibrary.com/answers/en/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c02/) 第2章 孔多塞循环的精确概率:216 种剖面、评审人数与选项个数 — 第二章精确算出孔多塞循环的概率:三评审为 0.0556,评审增多时趋向 0.0877,选项增多时迅速上升。 (https://5dlibrary.com/answers/zh/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c02/) - Ch.3 The Borda Count and Irrelevant Options: Twenty-One Electors, Decoys and Spoilers — Chapter 3 recomputes Borda's twenty-one electors and finds a spoiler in a share of 0.2604 of profiles with three reviewers and four options. (https://5dlibrary.com/answers/en/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c03/) 第3章 博尔达计分与无关选项:二十一名选举人、诱饵方案与搅局者 — 第三章重算博尔达的二十一名选举人,并发现三评审四选项时 0.2604 的剖面存在搅局者。 (https://5dlibrary.com/answers/zh/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c03/) - Ch.4 Single-Peaked Preferences and Value Restriction: When Cycles Disappear — With $w$ at zero the cycle probability is zero, as Black's theorem requires. Raising $w$ only to 0.1 gives 0.0117 for a panel of three and 0.0116 for a panel of five; at 0.5 the figures are 0.0434 and 0.0509; at one they return to the impartial-culture values of 0.0556 and 0.0694. The last two lines give an intuitive ratio: one tenth of off-axis noise brings back 0.2102 of the full no-consensus cycle risk for a panel of three, and 0.1677 for a panel of five. (https://5dlibrary.com/answers/en/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c04/) 第4章 单峰偏好与值域限制:循环在什么条件下消失 — $w$ 为零时循环概率为零,与布莱克定理一致。$w$ 只要升到 0.1,三人组的循环概率就是 0.0117,五人组是 0.0116;$w$ 到 0.5 时,三人组是 0.0434,五人组是 0.0509;$w$ 为一时回到不偏文化的 0.0556 与 0.0694。最后两行给出一个直观的比例:十分之一的离轴噪音,就把三人组完全无共识时循环风险的 0.2102 带了回来,五人组是 0.1677。 (https://5dlibrary.com/answers/zh/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c04/) - Ch.5 The Lexicographic Priority Arbiter: Which Axiom It Gives Up and What It Buys — Chapter 5 audits the lexicographic priority and graded rules and illustrates the Gibbard and Satterthwaite theorem with a census of manipulation: with four options Borda is manipulable in a share of 0.5122 of profiles, the priority rule in none. (https://5dlibrary.com/answers/en/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c05/) 第5章 字典序优先级裁决器:它放弃了哪条公理,换来了什么 — 第五章审计字典序优先级与分档规则,并用操纵普查说明吉巴德与萨特思韦特定理:四选项时博尔达在 0.5122 的剖面可被操纵,优先级规则为零。 (https://5dlibrary.com/answers/zh/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c05/) - Ch.6 Conflicting Kit Refusal Codes: an Illustrative Scenario, Deadlock Risk and the Priority Order — The table shows two trends running in opposite directions. More checks make deadlock easier: at a refusal rate of 0.10 with two actions, three checks lock up with probability 0.0734, five checks with 0.1677. More actions make deadlock harder: with the same three checks and rate 0.10, the probability is 0.0734 for two actions and only 0.0054 for four. Yet Chapter 2 has just shown that more actions make majority voting more likely to cycle. Adding candidate actions reduces deadlock while increasing cycles, and the two cannot both be pushed to zero. (https://5dlibrary.com/answers/en/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c06/) 第6章 五件套拒绝码冲突:示意场景、死锁概率与优先级顺序 — 表中有两个方向相反的趋势。检查越多,死锁越容易:拒绝率 0.10、两个动作时,三道检查的死锁概率是 0.0734,五道检查升到 0.1677。动作越多,死锁越难:同样是三道检查、拒绝率 0.10,两个动作时 0.0734,四个动作时只有 0.0054。可是第二章刚刚算过,动作越多,多数决越容易转圈。增加备选动作在减少死锁的同时增加了循环,两者无法同时压到零。 (https://5dlibrary.com/answers/zh/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c06/) - Ch.7 Eastern and Western Sources Examined: First Grader, Review Grader and Arbiter in the Song History, and Borda's Equal Steps — Part A uses Borda's own twenty-one electors. With $s$ at 1, which amounts to each elector casting only a vote against the least wanted candidate, C wins with 20 points; with $s$ at 2, the Borda count, C wins with 26; with $s$ at 3 and 6 C still leads, but by less and less, and at 6 the three candidates have 48, 49 and 50 points; at exactly 7 all three have 56 points and tie; at 8 A overtakes with 64. Once the weight of first place is large enough, scoring by rank degenerates into plurality and again elects A, the candidate who loses every pairwise contest. (https://5dlibrary.com/answers/en/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c07/) 第7章 东西方原典考辨:《宋史·选举志》的初考、覆考与详定,博尔达的等差计分 — A 段用博尔达本人的二十一名选举人。$s$ 为 1 时,相当于每人只投「最不想要谁」的反对票,C 以 20 分胜出;$s$ 为 2 就是博尔达计分,C 以 26 分胜出;$s$ 为 3 与 6 时 C 仍然领先,但优势逐步缩小,$s$ 为 6 时三人分别是 48、49、50 分;$s$ 恰好为 7 时三人都是 56 分,完全打平;$s$ 为 8 时 A 以 64 分反超。也就是说,只要首位的分量足够大,按名次计分就会退化成相对多数,重新选出那个两两对决全输的 A。 (https://5dlibrary.com/answers/zh/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c07/) - Ch.8 Closing the Loop: Turning Multi-Reviewer Verdicts in the Kits into an Explicit Priority, with a Pre-Registered Plan of Refutation — Chapter 8 gives a seven-step verdict procedure, the distilled Reviewer Aggregation Paradox Detector and a pre-registered plan of refutation. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c08/) 第8章 实证闭环:把五件套的多评审裁决改为显式优先级,并预注册证伪方案 — 第八章给出七步裁决流程、蒸馏小工具「评审聚合悖论检测器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/arrow-impossibility-multi-reviewer-aggregation-kit-conflicts/c08/) ## 5DL-B-311 · The Minimum Description Length Principle and Ockham's Razor for Rule Bases: Two-Part Codes that Decide Whether a New Rule Earns Its Place 中文书名:最小描述长度原则与规则库奥卡姆剃刀、两部编码判定新规则值不值得入库母典 DOI: pending Cite: Fan, Y. (2026). The Minimum Description Length Principle and Ockham's Razor for Rule Bases: Two-Part Codes that Decide Whether a New Rule Earns Its Place (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All decision records are synthetic, all parameters illustrative, and all calculations generated by the book's programs; it is general business analysis, not personalised business advice. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. Abstract: Whoever runs a rule base is tempted to believe that more rules mean better coverage. This book overturns that piece of common sense: a new rule deserves a place only if the bits it saves in encoding the exceptions exceed the bits needed to describe the rule itself. The grounds are Rissanen's principle of shortest data description, published in Automatica in 1978, and Grünwald's monograph of 2007. After checking the sources, the book corrects its blueprint: Rissanen's paper concerns statistical model selection, and its use for rule bases is later work; "saves more than it costs" is the criterion of crude two-part coding and depends on the choice of code; and minimum description length does not guarantee the least prediction error. All decision records are synthetic, all parameters illustrative, and all calculations generated by the book's programs; it is general business analysis, not personalised business advice. Chapter 1 covers universal integer codes, the Kraft inequality and four ways to write an exception list, showing that every code is a prior. Chapter 2 sets up the two-part ledger: a narrow rule that is right more than eight times in ten but fixes only 7 exceptions net cannot earn its own length; on a synthetic ledger the gate admits only 3 of the proposer's 12 rules, the third with a margin of just 2.9 bits. Chapter 3 draws the overfitting curve: training exceptions keep falling while hold-out error climbs back from 0.1973 to 0.2333; over twenty repetitions, the base that stops at the shortest total length has a hold-out error of 0.1916 against 0.2223 with no gate. Chapter 4 reconciles the principle with the Bayesian information criterion and shows that the latter puts no price on picking one rule out of a thousand candidates, and so admits too many. Chapter 5 measures sensitivity to the code: in twenty repetitions six reasonable codes never once choose the same rule count, every verdict with a margin below 4 bits flips with the code, and a minimum margin of 16 bits makes every admission verdict code-proof. Chapter 6 brings the gate to the cross-volume axiom pruner: a gate that only catches repeats admits 46 rules with a hold-out error of 0.4155, while the description-length gate admits only the 2 true rules. Chapter 7 examines the webbed toes and extra finger of the Zhuangzi and Ockham's "in vain through more" in the Summa Logicae, noting that "entities must not be multiplied beyond necessity" is not Ockham's own wording. Chapter 8 gives the admission check-up, the distilled Rule Admission Description-Length Referee and a pre-registered plan of refutation; testing whether the gate beats no gate needs 2,733 hold-out records per arm. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:规则库的主事者很自然地相信规则越多越周全。本书推翻这个常识:一条新规则只有在它节省的例外编码长度超过它自身的描述长度时,才值得入库。依据是里沙宁 1978 年在《自动学》上提出的最短数据描述原则与格伦瓦尔德 2007 年的专著。核对原文之后,本书做了蓝图修正:里沙宁的论文讨论的是统计模型选择,用于规则库是后人的工作;「节省超过自身」是粗糙两部编码的判据,依赖于编码方案的选择;最小描述长度也不保证预测误差最小。书中的决策记录全部合成、参数均为示意,全部计算由书中程序生成;本书是非个人化的经营分析,不构成经营建议。 第一章讲通用整数码、克拉夫特不等式与例外名单的四种写法,说明每一种码都是一种先验。第二章建立两部编码账本:一条准确率超过八成、却只净修正 7 条例外的窄规则挣不回自己的篇幅;在合成台账上,门槛从提议者的 12 条规则中只放进 3 条,第三条的余量仅 2.9 个比特。第三章画出过拟合曲线:训练例外一路下降,留出误差却从 0.1973 回升到 0.2333;二十次重复中,按最短总长度停下的规则库留出误差 0.1916,不设门槛时 0.2223。第四章与贝叶斯信息准则对账,说明后者不为「从上千候选中挑出这一条」计价,因而多收规则。 第五章做编码方案敏感度:六种合理的码在二十次重复中没有一次选出相同的规则数,余量不足 4 个比特的裁决全部随编码翻转,16 个比特的余量下限使全部入库裁决不再翻转。第六章把门槛搬到跨卷公理剪枝器上:只查重复的门收进 46 条规则,留出误差 0.4155,描述长度门只收 2 条真规则。第七章考辨《庄子·骈拇》的「骈拇枝指」与奥卡姆《逻辑大全》的「徒然以多为之」,并指出「如无必要勿增实体」并非奥卡姆原话。第八章给出入库体检流程、蒸馏小工具「规则入库描述长度裁判」与预注册证伪方案,检验门槛是否优于不设门槛,每组需 2,733 条留出记录。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Even Integers Cost Something: Universal Codes, the Kraft Inequality and the Bits of an Exception List — Program p13 compares the four at two scales, 600 records and 6,000. With 600 records and 30 exceptions, flagging costs 600 bits, listing positions 309, and the enumerative code only 177.1, just 5.25 bits above Shannon's limit of 171.8. Per exception, the enumerative code spends 5.90 bits and position listing 10.30. With 6,000 records and 600 exceptions, listing positions takes 7,819 bits, the enumerative code 2,827.1 and the limit is 2,814.0. (https://5dlibrary.com/answers/en/minimum-description-length-rule-base-ockham-two-part-code/c01/) 第1章 整数也要付费:通用整数码、克拉夫特不等式与例外名单的比特数 — 程序 p13 把四种写法放在 600 条记录与 6,000 条记录两个规模上比较。在 600 条记录、30 条例外时,逐条打标要 600 个比特,列编号要 309 个,枚举码只要 177.1 个,离香农下限 171.8 只差 5.25 个比特。平均到每一条例外,枚举码花 5.90 个比特,列编号花 10.30 个。到了 6,000 条记录、600 条例外,列编号要 7,819 个比特,枚举码只要 2,827.1 个,下限是 2,814.0。 (https://5dlibrary.com/answers/zh/minimum-description-length-rule-base-ockham-two-part-code/c01/) - Ch.2 The Two-Part Ledger: How Many Exceptions a Rule Must Fix to Earn Its Own Length — Chapter 2 sets up the two-part ledger: a narrow rule that is right more than eight times in ten but fixes only 7 exceptions net cannot earn its own length; on a synthetic ledger the gate admits only 3 of the proposer's 12 rules, the third with a margin of just 2.9 bits. (https://5dlibrary.com/answers/en/minimum-description-length-rule-base-ockham-two-part-code/c02/) 第2章 两部编码账本:一条规则要修正多少条例外才算挣回自己的篇幅 — 第二章建立两部编码账本:一条准确率超过八成、却只净修正 7 条例外的窄规则挣不回自己的篇幅;在合成台账上,门槛从提议者的 12 条规则中只放进 3 条,第三条的余量仅 2.9 个比特。 (https://5dlibrary.com/answers/zh/minimum-description-length-rule-base-ockham-two-part-code/c02/) - Ch.3 The Overfitting Curve: Rule Count, Total Description Length and Hold-out Error — Chapter 3 draws the overfitting curve: training exceptions keep falling while hold-out error climbs back from 0.1973 to 0.2333; over twenty repetitions, the base that stops at the shortest total length has a hold-out error of 0.1916 against 0.2223 with no gate. (https://5dlibrary.com/answers/en/minimum-description-length-rule-base-ockham-two-part-code/c03/) 第3章 过拟合曲线:规则数、总描述长度与留出误差的三条线 — 第三章画出过拟合曲线:训练例外一路下降,留出误差却从 0.1973 回升到 0.2333;二十次重复中,按最短总长度停下的规则库留出误差 0.1916,不设门槛时 0.2223。 (https://5dlibrary.com/answers/zh/minimum-description-length-rule-base-ockham-two-part-code/c03/) - Ch.4 Reconciling with the Bayesian Information Criterion: Parametric Complexity, Normalised Maximum Likelihood and Where Rule Bases Part Ways — On 150 records the Bayesian information criterion admits 9.5 rules on average, with a hold-out error of 0.2512; minimum description length admits only 1.4, with an error of 0.2218; the optimum is 2.2 rules with an error of 0.1951. On 600 records the three are 6.9, 2.6 and 2.4 rules, with errors of 0.1929, 0.1849 and 0.1830. At every size the Bayesian information criterion admits several rules more than the optimum, and its error is higher every time. The reason is the one just given: it pays for each rule as though it were one parameter, and nothing for having picked this particular rule out of 1347 candidates. The richer the structure of the rules, the larger this unpaid structural cost. (https://5dlibrary.com/answers/en/minimum-description-length-rule-base-ockham-two-part-code/c04/) 第4章 与贝叶斯信息准则对账:参数复杂度、归一化最大似然与规则库上的分歧 — 在 150 条记录上,贝叶斯信息准则平均收了 9.5 条规则,留出误差是 0.2512;最小描述长度只收了 1.4 条,误差 0.2218;最优是 2.2 条,误差 0.1951。在 600 条记录上,三者分别是 6.9 条、2.6 条与 2.4 条,误差是 0.1929、0.1849 与 0.1830。贝叶斯信息准则在每一个规模上都比最优多收好几条规则,而且误差也都更高。原因正是上面说的:它只为每条规则付了「一个参数」的钱,却没有为「从 1347 条候选里挑中了这一条」付钱。规则的结构越丰富,这个漏掉的结构代价就越大。 (https://5dlibrary.com/answers/zh/minimum-description-length-rule-base-ockham-two-part-code/c04/) - Ch.5 Sensitivity to the Code: Six Reasonable Codes, Six Verdicts, and How Many Bits of Slack to Reserve — Chapter 5 measures sensitivity to the code: in twenty repetitions six reasonable codes never once choose the same rule count, every verdict with a margin below 4 bits flips with the code, and a minimum margin of 16 bits makes every admission verdict code-proof. (https://5dlibrary.com/answers/en/minimum-description-length-rule-base-ockham-two-part-code/c05/) 第5章 编码方案敏感度:六种合理的码,六种裁决,余量要留多少比特 — 第五章做编码方案敏感度:六种合理的码在二十次重复中没有一次选出相同的规则数,余量不足 4 个比特的裁决全部随编码翻转,16 个比特的余量下限使全部入库裁决不再翻转。 (https://5dlibrary.com/answers/zh/minimum-description-length-rule-base-ockham-two-part-code/c05/) - Ch.6 The Pruner's Admission Threshold: a Gate that Only Catches Repeats and a Gate that Counts Bits — Chapter 6 brings the gate to the cross-volume axiom pruner: a gate that only catches repeats admits 46 rules with a hold-out error of 0.4155, while the description-length gate admits only the 2 true rules. (https://5dlibrary.com/answers/en/minimum-description-length-rule-base-ockham-two-part-code/c06/) 第6章 剪枝器的入库门槛:只查重复的门与描述长度的门 — 第六章把门槛搬到跨卷公理剪枝器上:只查重复的门收进 46 条规则,留出误差 0.4155,描述长度门只收 2 条真规则。 (https://5dlibrary.com/answers/zh/minimum-description-length-rule-base-ockham-two-part-code/c06/) - Ch.7 Sources East and West: the Extra Finger of the Zhuangzi and Ockham's "In Vain through More" — "Lengthening the duck's legs" replaces a short rule by a longer version, which narrows a true rule. The rule half grows by only 2.0 bits, but the records narrowed away become exceptions again, the exception half grows by 45.6 bits, the total by 47.6 bits, and the hold-out error rises by 0.0307. "Cutting the crane's legs" shortens a long rule by removing a condition, which widens a true rule. The rule half actually saves 5.3 bits, but the widened rule wrongly acts on many records it should leave alone, the exception half grows by 152.5 bits, the total by 147.2 bits, and the hold-out error rises by 0.1408. "Amputation" deletes a true rule outright: the rules save 10.5 bits, the exceptions grow by 74.6 bits, and the hold-out error rises by 0.0507. (https://5dlibrary.com/answers/en/minimum-description-length-rule-base-ockham-two-part-code/c07/) 第7章 东西方原典考辨:《庄子·骈拇》的枝指与奥卡姆的「徒然以多为之」 — 「续凫胫」是把一条短规则换成更长的版本,即把真规则收窄。规则那一半只多了 2.0 个比特,可是被收窄掉的那些记录重新变成了例外,例外那一半多出 45.6 个比特,总长度多出 47.6 个比特,留出误差上升 0.0307。「断鹤胫」是把一条长规则截短,即去掉一个条件,把真规则放宽。规则那一半反而省了 5.3 个比特,可是放宽以后误判了大量本不该动作的记录,例外那一半多出 152.5 个比特,总长度多出 147.2 个比特,留出误差上升 0.1408。「截肢」是把一条真规则整条删掉,规则省了 10.5 个比特,例外多出 74.6 个比特,留出误差上升 0.0507。 (https://5dlibrary.com/answers/zh/minimum-description-length-rule-base-ockham-two-part-code/c07/) - Ch.8 Closing the Loop: the Admission Check-up, the Rule Admission Description-Length Referee and a Pre-registered Plan of Refutation — Chapter 8 gives the admission check-up, the distilled Rule Admission Description-Length Referee and a pre-registered plan of refutation; testing whether the gate beats no gate needs 2,733 hold-out records per arm. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/minimum-description-length-rule-base-ockham-two-part-code/c08/) 第8章 实证闭环:入库体检、规则入库描述长度裁判与预注册的证伪方案 — 第八章给出入库体检流程、蒸馏小工具「规则入库描述长度裁判」与预注册证伪方案,检验门槛是否优于不设门槛,每组需 2,733 条留出记录。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/minimum-description-length-rule-base-ockham-two-part-code/c08/) ## 5DL-B-312 · Active Learning and Uncertainty Sampling: the Fewest Human Labels a Customer-Intent Classifier Needs to Reach Its Target 中文书名:主动学习与不确定性采样、最少人工标注让客服意图分类器达标母典 DOI: pending Cite: Fan, Y. (2026). Active Learning and Uncertainty Sampling: the Fewest Human Labels a Customer-Intent Classifier Needs to Reach Its Target (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Intent priors, word probabilities and costs are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. Abstract: Online shop owners commonly believe that training a customer-intent classifier takes a few thousand randomly drawn past messages to label, and that more is always better. The counterintuitive claim tested here is that the accuracy random labeling reaches only after more than a thousand labels can often be reached with a fraction of them if the model picks the messages it is least sure of and hands those to a person; but messages picked this way carry a sampling bias, and using them to estimate intent shares, to grade the model or to decide when to stop goes wrong. After checking the sources the book corrects its blueprint: the SIGIR paper of Lewis and Gale from 1994 measured F1 on rare news categories, where 999 uncertainty-selected examples matched a hundred thousand or more random ones in most categories and five hundredfold was the best case for a few; the report of Settles from 2009 is a survey, not an experiment; and the fundamental nature of the bias is shown by Dasgupta and Hsu in 2008. Intent priors, word probabilities and costs are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Chapter 1 recomputes the comparison on synthetic customer data with four intents: to reach the target accuracy random sampling needs a median of 1,600 labels and margin sampling 640, a ratio of 0.400; the model trained on every label scores 0.893, and the ratio changes as the target nears the ceiling. Chapter 2 writes a multinomial logistic regression from scratch, with a gradient check and the Bayes ceiling. Chapter 3 compares least confidence, smallest margin and entropy. Chapter 4 measures the bias: the margin-picked labeled set lies at a total variation distance of 0.208 from the traffic, and a model graded on the messages it picked scores only 0.463 against a true accuracy of 0.876; in a one-dimensional example uncertainty sampling stays at the wrong threshold with probability 0.513 even after three hundred labels, against 0.068 for random sampling. Chapter 5 draws with known inclusion probabilities and weights by their inverse, after Horvitz and Thompson, bringing the grade of a model whose true accuracy is 0.809 back from the plain mean of 0.657 to the truth, and shows that weighted training and prior-only adjustment can do worse in finite samples. Chapter 6 covers labeling budgets, stopping rules and audit sample sizes. Chapter 7 examines the Xueji line on tackling the easy parts before the knots and Bacon's crucial instances in the New Organon. Chapter 8 gives a labeling queue for customer intents, the distilled Labeling Priority Sorter and a pre-registered test; the classifier only routes and suggests a hand-over and takes no adverse decision about a customer automatically. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. 摘要:网店店主普遍相信,训练一个客服意图分类器要随机抽几千条历史消息来标注,抽得越多越准。本书检验的反常识命题是:随机抽样要上千条才达到的准确率,让模型挑出自己最拿不准的消息交给人标注,常常只需其中一部分;但这样挑出的样本带有抽样偏差,用它估计意图占比、给模型打分或决定何时停止都会出错。核对原文之后,本书作蓝图修正:刘易斯与盖尔 1994 年的 SIGIR 论文衡量的是稀有新闻类别上的 F1,不确定性采样的 999 条在多数类别上抵得上十万条以上的随机样本,五百倍只是个别类别的最好情形;塞特尔斯 2009 年的报告是综述而非实验;抽样偏差的根本性由达斯古普塔与许 2008 年的论文说明。书中意图先验、词概率与成本均为示意参数;全部计算由书中程序生成;本书属于非个人化经营建议。 第一章在合成的四意图客服数据上重算:达到目标准确率,随机抽样的中位数是 1,600 条,最小间隔采样是 640 条,后者只用前者的 0.400;全部标注训练的模型准确率为 0.893,这一倍数随目标逼近天花板而变化。第二章从零写出多类逻辑回归、梯度检验与贝叶斯天花板。第三章比较最小置信、最小间隔与熵。第四章测量偏差:间隔挑选的已标注集合与客流的总变差距离为 0.208,模型在自己挑出的消息上自评准确率只有 0.463,而真实准确率为 0.876;在一维例子里,不确定性采样即使标到三百条仍有 0.513 的概率困在错误阈值上,随机抽样只有 0.068。 第五章以已知入选概率抽样并按霍维茨与汤普森的倒数加权,把一个真实准确率 0.809 的模型的打分从直接平均的 0.657 校正回真值,并说明加权训练与只校正先验在有限样本下可能更差。第六章讨论标注预算、停止规则与审计样本量。第七章考辨《礼记·学记》「先其易者,后其节目」与培根《新工具》的关键实例。第八章给出客服意图标注队列、蒸馏小工具「标注优先级排序器」与预注册检验方案;分类器只做分派与转人工建议,不对客户自动做出不利决定。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Thousands versus Hundreds: the Counterintuitive Ledger of Uncertainty Sampling — Chapter 1 recomputes the comparison on synthetic customer data with four intents: to reach the target accuracy random sampling needs a median of 1,600 labels and margin sampling 640, a ratio of 0.400; the model trained on every label scores 0.893, and the ratio changes as the target nears the ceiling. (https://5dlibrary.com/answers/en/active-learning-uncertainty-sampling-intent-classifier-labels/c01/) 第1章 几千条与几百条:不确定性采样的反常识账本 — 第一章在合成的四意图客服数据上重算:达到目标准确率,随机抽样的中位数是 1,600 条,最小间隔采样是 640 条,后者只用前者的 0.400;全部标注训练的模型准确率为 0.893,这一倍数随目标逼近天花板而变化。 (https://5dlibrary.com/answers/zh/active-learning-uncertainty-sampling-intent-classifier-labels/c01/) - Ch.2 A Hand-Written Multinomial Logistic Regression: Stochastic Gradient Descent on Integer Bag-of-Words Features — A hand-written logistic regression together with synthetic data whose ceiling can be computed exactly gives a clean test bench: the gradient has passed a numerical check, the training path is repeatable, the ceiling is 0.891, and the model trained on every label already touches it. We can now return to the central question of active learning. Of the four probabilities the model gives, which kind of doubt is most worth sending to a person? Will least confidence, smallest margin and entropy pick the same messages? (https://5dlibrary.com/answers/en/active-learning-uncertainty-sampling-intent-classifier-labels/c02/) 第2章 手写多类逻辑回归:词袋整数特征上的随机梯度下降 — 一个手写的逻辑回归,加上可以精确算出上限的合成数据,给了我们一个干净的实验台:梯度经过数值检验,训练轨迹可重复,天花板为 0.891,用全部标注的模型已经贴着它。现在可以回到主动学习的核心问题:模型给出的四个概率里,哪一种「拿不准」最值得送去标注?最小置信、最小间隔和熵,挑出来的会是同一批消息吗? (https://5dlibrary.com/answers/zh/active-learning-uncertainty-sampling-intent-classifier-labels/c02/) - Ch.3 Three Rulers of Doubt: Least Confidence, Smallest Margin and Entropy — On the book's data the first hazard is mild: of the 200 messages one margin batch picks, 197 are distinct, and the whole pool of 6,000 messages contains 133 exact duplicates. Yet the make-up of the batch is quite unlike that of a random one. The random batch holds 127 tracking messages, 40 refunds and 10 hand-overs; the margin batch holds only 83 tracking messages, 69 refunds and 26 hand-overs, and its messages are shorter, 5.17 words on average against 5.85 for the random batch. Margin sampling concentrates the taggers' time on the border between refunds and tracking and on the rare intent, which is why it saves labels and also why it brings bias. (https://5dlibrary.com/answers/en/active-learning-uncertainty-sampling-intent-classifier-labels/c03/) 第3章 三把尺子量「拿不准」:最小置信、最小间隔与熵 — 在本书的数据上,第一个隐患不严重:最小间隔一次挑出的 200 条里有 197 条互不相同,整个池子 6,000 条中完全重复的消息有 133 条。但这一批的构成与随机批截然不同:随机批里查物流 127 条、退款 40 条、转人工 10 条;最小间隔批里查物流只有 83 条,退款 69 条,转人工 26 条,消息平均也更短,为 5.17 个词,随机批是 5.85 个。最小间隔把标注员的时间集中在退款与查物流的交界和稀有意图上,这正是它省标注的原因,也正是它带来偏差的原因。 (https://5dlibrary.com/answers/zh/active-learning-uncertainty-sampling-intent-classifier-labels/c03/) - Ch.4 Sampling Bias: Picked Messages No Longer Stand for the Traffic — Chapter 4 measures the bias: the margin-picked labeled set lies at a total variation distance of 0.208 from the traffic, and a model graded on the messages it picked scores only 0.463 against a true accuracy of 0.876; in a one-dimensional example uncertainty sampling stays at the wrong threshold with probability 0.513 even after three hundred labels, against 0.068 for random sampling. (https://5dlibrary.com/answers/en/active-learning-uncertainty-sampling-intent-classifier-labels/c04/) 第4章 抽样偏差:被挑出来的消息不再代表客流 — 第四章测量偏差:间隔挑选的已标注集合与客流的总变差距离为 0.208,模型在自己挑出的消息上自评准确率只有 0.463,而真实准确率为 0.876;在一维例子里,不确定性采样即使标到三百条仍有 0.513 的概率困在错误阈值上,随机抽样只有 0.068。 (https://5dlibrary.com/answers/zh/active-learning-uncertainty-sampling-intent-classifier-labels/c04/) - Ch.5 Importance Weighting and Prior Adjustment: Computing the Bias Away — Chapter 5 draws with known inclusion probabilities and weights by their inverse, after Horvitz and Thompson, bringing the grade of a model whose true accuracy is 0.809 back from the plain mean of 0.657 to the truth, and shows that weighted training and prior-only adjustment can do worse in finite samples. (https://5dlibrary.com/answers/en/active-learning-uncertainty-sampling-intent-classifier-labels/c05/) 第5章 重要性加权与先验校正:把偏差算回来 — 第五章以已知入选概率抽样并按霍维茨与汤普森的倒数加权,把一个真实准确率 0.809 的模型的打分从直接平均的 0.657 校正回真值,并说明加权训练与只校正先验在有限样本下可能更差。 (https://5dlibrary.com/answers/zh/active-learning-uncertainty-sampling-intent-classifier-labels/c05/) - Ch.6 Labeling Budgets, Stopping Rules and Audit Sample Sizes — If the aim is to show that accuracy is at least 0.850 when the true value is 0.880, a one-sided test at significance 0.05 with power 0.80 needs 824 messages. In terms of the Wilson lower bound, with an observed 0.880 the bound reaches 0.850 at 550 messages, 0.860 at 1,160 and 0.870 at 4,350. The last line of the program corresponds to S2 above: an audit of 400 with an observed 0.880 has a lower bound of only 0.844, below 0.850, so the rule does not fire. (https://5dlibrary.com/answers/en/active-learning-uncertainty-sampling-intent-classifier-labels/c06/) 第6章 标注预算、停止规则与审计样本量 — 如果目的是证明「准确率不低于 0.850」,而真实值是 0.880,单侧检验在显著性 0.05、功效 0.80 下需要 824 条。如果用 Wilson 下界的写法:观测到 0.880 时,下界不低于 0.850 需要 550 条,不低于 0.860 需要 1,160 条,不低于 0.870 需要 4,350 条。程序最后一行对应上一节的 S2:400 条审计在观测到 0.880 时下界只有 0.844,低于 0.850,所以规则不触发。 (https://5dlibrary.com/answers/zh/active-learning-uncertainty-sampling-intent-classifier-labels/c06/) - Ch.7 Eastern and Western Classics: the Xueji's Easy-First Order and Bacon's Crucial Instances — Program p72 trains two candidate models, A and B, each on 160 random labels that do not overlap, and asks how many labels are needed to decide which is more accurate. The two models disagree on only 0.201 of the messages; where they agree, no label can separate them, whatever it says. Among the 0.184 of messages on which exactly one of them is right, A is right on a share of 0.432. Deciding that this share differs from one half with a two-sided test at the 0.05 level and power 0.80 needs 417 informative instances; labeling only the disagreements needs about 456 labels, and labeling at random about 2,266. The disagreement filter saves 0.799 of the labels. Bacon's fingerpost here acquires a precise quantity: it divides the number of observations needed by roughly the share of disagreement. (https://5dlibrary.com/answers/en/active-learning-uncertainty-sampling-intent-classifier-labels/c07/) 第7章 东西方原典考辨:《学记》的先易后难与培根的关键实例 — 程序 p72 让两个候选模型 A 与 B 各用 160 条互不重叠的随机标注训练,然后问:要判定谁更准,需要标多少条?两个模型只在 0.201 的消息上意见不一;在两者意见一致的消息上,无论标注结果如何,都不能区分它们。在恰有一个判对的 0.184 的消息里,A 判对的比例为 0.432。用双侧检验在 0.05 水平、功效 0.80 下判定这个比例不等于一半,需要 417 条有效实例;只标分歧消息,约需 456 条;随机标注则约需 2,266 条。分歧过滤省下了 0.799 的标注。培根的路标在这里有一个精确的数量含义:它把需要的观察次数缩小到分歧比例分之一左右。 (https://5dlibrary.com/answers/zh/active-learning-uncertainty-sampling-intent-classifier-labels/c07/) - Ch.8 Closing the Loop: a Labeling Queue for Customer Intents and a Plan to Refute the Claim — Chapter 8 gives a labeling queue for customer intents, the distilled Labeling Priority Sorter and a pre-registered test; the classifier only routes and suggests a hand-over and takes no adverse decision about a customer automatically. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/active-learning-uncertainty-sampling-intent-classifier-labels/c08/) 第8章 实证闭环:客服意图标注队列与证伪方案 — 第八章给出客服意图标注队列、蒸馏小工具「标注优先级排序器」与预注册检验方案;分类器只做分派与转人工建议,不对客户自动做出不利决定。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/active-learning-uncertainty-sampling-intent-classifier-labels/c08/) ## 5DL-B-313 · Moving Festival Calendars in Multicultural Malaysia: Overlaps of Hari Raya, Chinese New Year and Deepavali and the Alignment of Promotional Schedules 中文书名:马来西亚多元节庆移动日历叠加、开斋节农历新年屠妖节浮动日期与促销排期对齐母典 DOI: pending Cite: Fan, Y. (2026). Moving Festival Calendars in Multicultural Malaysia: Overlaps of Hari Raya, Chinese New Year and Deepavali and the Alignment of Promotional Schedules (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Sales series and lead times are illustrative; all calculations are generated by the book's programs. Every festival date follows the official announcement; the book treats the religious and cultural meaning of each festival with respect, passes no judgment on it, and is general analysis, not individualised business advice. Abstract: The three largest moments of Malaysian retail, Aidilfitri, Chinese New Year and Deepavali, are fixed by the Islamic, the Chinese and the Hindu calendars and move through the Gregorian year. Comparing with the same date or the same month last year reads the movement of a festival as the quality of a business. This book uses the arithmetic Islamic calendar, the mean synodic month with periodic terms, the sun's longitude and the sidereal zodiac to turn the dates of the three festivals into testable approximations, checked year by year against verified official dates. Before writing, the book checked its sources and corrects the blueprint: the Islamic year is on average 10.8758 days shorter than the Gregorian year, and a circuit of the seasons takes 32.58 Gregorian years rather than thirty-three; Calendrical Calculations is cited in its fourth edition of 2018, which supplies calendar arithmetic, while the evidence on distorted comparisons comes from the time-series literature on moving holidays. Sales series and lead times are illustrative; all calculations are generated by the book's programs. Every festival date follows the official announcement; the book treats the religious and cultural meaning of each festival with respect, passes no judgment on it, and is general analysis, not individualised business advice. Chapter 1 computes the drift of Aidilfitri with the arithmetic calendar and checks it against the announcements of the Keeper of the Rulers' Seal from 2020 to 2026. Chapter 2 measures the distortion: in an illustrative shop, Aidilfitri moving from April into March makes the same-date comparison plus 0.289 for March and minus 0.218 for April, while true growth is only four per cent; with real festival dates the mean absolute error over twelve comparisons is 0.180. Chapter 3 computes Chinese New Year as the second new moon after the winter solstice and finds the years where the rule fails. Chapter 4 approximates Malaysian Deepavali as the day before the first new moon after the sun enters sidereal Libra. Chapter 5 shows that all three festivals are tied to the new moon, so meeting years come in clusters and a uniform-phase model badly underestimates how often festivals meet. Chapter 6 builds a festival-aligned monthly comparison that cuts the mean absolute error over the year from 0.054 to 0.017. Chapter 7 examines the rule of the Jingchu Suishi Ji that counts Cold Food from the winter solstice and the nineteen-year cycle of Bede's The Reckoning of Time. Chapter 8 gives promotion calendars for Malaysia, Singapore and Taiwan, the distilled Moving Festival Aligner and a pre-registered test, and shows that the claim fails when the festival lift is below about 0.145 of the illustrative model's. The book contains 24 runnable Python programs. 摘要:马来西亚零售业的三个最大消费节点,开斋节、农历新年与屠妖节,分别由伊斯兰历、中国农历与印度历决定,每年在公历上移动。按「去年同日」或「去年同月」做同比,会把节日的移动误读为经营的好坏。本书用伊斯兰算术历、平均朔望月加周期修正项、太阳黄经与恒星黄道,把三个节日的日期算成可检验的近似,并逐年与核实过的官方日期比对。落笔前本书核对原文并作蓝图修正:伊斯兰历平均每年比公历短 10.8758 天,绕行四季一周需 32.58 个公历年而非三十三年;《Calendrical Calculations》以 2018 年第四版为准,它提供历法算术,而同比失真的证据来自移动节日的时间序列文献。书中销售序列与提前天数为示意;全部计算由书中程序生成。所有节日日期以官方公布为准;本书对各节日的宗教与文化内涵保持尊重,不作评判,属于非个人化经营建议。 第一章用算术历计算开斋节的漂移,并与 2020 年至 2026 年掌玺大臣的宣布比对。第二章量化同比失真:在一家示意门店中,开斋节从四月移入三月,使三月同日同比为加 0.289、四月为减 0.218,而真实增长只有百分之四;用真实节日日期,十二对比较的平均绝对误差为 0.180。第三章以冬至后第二个朔日规则计算农历新年,并找出规则失效的年份。第四章以太阳入恒星天秤宫后第一个朔日的前一天近似马来西亚的屠妖节。第五章说明三个节日都依附于朔,相遇年份成组出现,均匀相位模型严重低估相遇频率。第六章给出按节庆对齐的月度同比,全年平均绝对误差由 0.054 降到 0.017。 第七章考辨《荆楚岁时记》以冬至数日定寒食的规则与比德《论时间的计算》的十九年周期。第八章给出马新台三地的促销排期表、蒸馏小工具「移动节庆对齐器」与预注册检验方案,并指出节日提升幅度低于示意模型的约 0.145 倍时命题不成立。书中含 24 段可运行的 Python 程序。 Chapter answers: - Ch.1 Eleven Days of Drift: the Arithmetic Islamic Calendar, the Thirty-Year Leap Cycle and the Official Announcement of Aidilfitri — Chapter 1 computes the drift of Aidilfitri with the arithmetic calendar and checks it against the announcements of the Keeper of the Rulers' Seal from 2020 to 2026. (https://5dlibrary.com/answers/en/malaysia-moving-festival-calendars-promotion-alignment/c01/) 第1章 十一天的漂移:伊斯兰算术历、三十年闰周与开斋节的官方宣布 — 第一章用算术历计算开斋节的漂移,并与 2020 年至 2026 年掌玺大臣的宣布比对。 (https://5dlibrary.com/answers/zh/malaysia-moving-festival-calendars-promotion-alignment/c01/) - Ch.2 The Same-Date Trap: How Moving Festivals Distort Year-on-Year Comparisons — Chapter 2 measures the distortion: in an illustrative shop, Aidilfitri moving from April into March makes the same-date comparison plus 0.289 for March and minus 0.218 for April, while true growth is only four per cent; with real festival dates the mean absolute error over twelve comparisons is 0.180. (https://5dlibrary.com/answers/en/malaysia-moving-festival-calendars-promotion-alignment/c02/) 第2章 去年同日的陷阱:节庆移动下的同比失真 — 第二章量化同比失真:在一家示意门店中,开斋节从四月移入三月,使三月同日同比为加 0.289、四月为减 0.218,而真实增长只有百分之四;用真实节日日期,十二对比较的平均绝对误差为 0.180。 (https://5dlibrary.com/answers/zh/malaysia-moving-festival-calendars-promotion-alignment/c02/) - Ch.3 The Mean Synodic Month and Chinese New Year: the Solstice Rule, Periodic Terms and Exceptional Years — The next two lines show the basic tension the Chinese calendar must resolve. Twelve mean months come to 354.367 days, 10.875 days short of the tropical year, so an intercalary month must be inserted every two or three years if the calendar is not to drift away like the Islamic one. The seventh line is the classic arithmetic of intercalation: 235 months make 6939.688 days and nineteen tropical years 6939.602 days, a difference of only 0.087 days, which is why seven intercalary months in nineteen years works so well. It is the same coincidence on which Bede's Easter cycle in Chapter 7 relies. (https://5dlibrary.com/answers/en/malaysia-moving-festival-calendars-promotion-alignment/c03/) 第3章 平均朔望月与农历新年:冬至规则、修正项与例外年份 — 最后三行回答本节开头的问题。在 2020 年至 2030 年的 136 次朔中,真实朔相对平均朔最早提前 13.5 小时,最晚推后 14.2 小时,平均绝对偏离 6.3 小时;偏离达到半天以上的有 12 次。对一个以「哪一天」为单位的节日来说,这意味着只用平均朔日时,大约每十次就有一次会落在相邻的日子上。 (https://5dlibrary.com/answers/zh/malaysia-moving-festival-calendars-promotion-alignment/c03/) - Ch.4 The Sidereal Window of Deepavali: Entry into Libra, the Eve of the New Moon and the Gap between Time Zones — Of the 151 Deepavali new moons between 1950 and 2100, the date at UTC+8 differs from the date at UTC+5:30 in 12 years, a share of 0.079, and from the date in universal time in a share of 0.318. If the hour of the new moon were spread evenly over the day, the theoretical shares would be 0.104 and 0.333, close to the counts. The years of disagreement after 2000 are listed as well. (https://5dlibrary.com/answers/en/malaysia-moving-festival-calendars-promotion-alignment/c04/) 第4章 屠妖节的恒星窗口:入天秤宫、朔日前夕与时区之差 — 屠妖节在 1950 年至 2100 年之间,按规则最早是 1963 年的 10 月 16 日,最晚是 2020 年的 11 月 14 日。2020 年这个最晚日期,正是官方核实过的日期之一。1950 年至 2000 年的平均年中日序为 304.14,2050 年至 2100 年为 305.04,一个世纪后移约一天,与入宫时刻的滑动一致。 (https://5dlibrary.com/answers/zh/malaysia-moving-festival-calendars-promotion-alignment/c04/) - Ch.5 How Often Festivals Meet: Why a Uniform-Phase Model Fails — Aidilfitri falls within seven days of Chinese New Year in 14 years, and they come in clusters: 1964 to 1966, 1996 to 1998, 2029 to 2031, 2062 to 2063, 2094 to 2096. The clusters are thirty-two or thirty-three years apart, the circuit of Aidilfitri computed in Chapter 1. The program's 1996 to 1998 matches the Kongsi Raya years of the public record, and the next cluster, 2029 to 2031, is close at hand. (https://5dlibrary.com/answers/en/malaysia-moving-festival-calendars-promotion-alignment/c05/) 第5章 节庆相遇的概率:均匀相位模型为何失效 — 开斋节与农历新年相隔七天以内的年份共 14 个,成组出现:1964 年至 1966 年,1996 年至 1998 年,2029 年至 2031 年,2062 年至 2063 年,2094 年至 2096 年。组与组之间相隔三十二三年,正是第一章算出的开斋节绕行周期。程序算出的 1996 年至 1998 年与公共记录中的 Kongsi Raya 年份吻合;下一组是 2029 年至 2031 年,就在眼前。 (https://5dlibrary.com/answers/zh/malaysia-moving-festival-calendars-promotion-alignment/c05/) - Ch.6 Festival-Aligned Comparisons: an Event-Time Map, the Split of Ramadan and the Cost of Date Errors — Chapter 6 builds a festival-aligned monthly comparison that cuts the mean absolute error over the year from 0.054 to 0.017. (https://5dlibrary.com/answers/en/malaysia-moving-festival-calendars-promotion-alignment/c06/) 第6章 按节庆对齐的同比:事件时间映射、斋月切分与日期误差 — 第六章给出按节庆对齐的月度同比,全年平均绝对误差由 0.054 降到 0.017。 (https://5dlibrary.com/answers/zh/malaysia-moving-festival-calendars-promotion-alignment/c06/) - Ch.7 Eastern and Western Classics Examined: the Jingchu Suishi Ji and Bede's The Reckoning of Time — Counting from the Council of Nicaea in 325 to Bede's writing in 725, the equinox fixed on 21 March in the Julian calendar had drifted 3.12 days from the true equinox, and the moon of the cycle had fallen 1.30 days behind. By the Gregorian reform of 1582 the drift of the equinox had grown to 9.82 days and that of the moon to 4.08 days. The reform of 1582 removed ten days at once, matching the 9.82 days here. (https://5dlibrary.com/answers/en/malaysia-moving-festival-calendars-promotion-alignment/c07/) 第7章 东西方原典考辨:《荆楚岁时记》与比德《论时间的计算》 — 从尼西亚会议的 325 年算起,到比德写书的 725 年,按儒略历固定在 3 月 21 日的春分已经比真实春分偏离了 3.12 天,周期推算的月亮落后了 1.30 天。到 1582 年格里高利历改革时,春分偏离累积到 9.82 天,月亮偏离 4.08 天。1582 年的改革一次删去十天,与这里的 9.82 天正好对应。 (https://5dlibrary.com/answers/zh/malaysia-moving-festival-calendars-promotion-alignment/c07/) - Ch.8 Closing the Loop: a Promotion Calendar for Malaysia, Singapore and Taiwan, the Moving Festival Aligner and a Pre-Registered Test — Chapter 8 gives promotion calendars for Malaysia, Singapore and Taiwan, the distilled Moving Festival Aligner and a pre-registered test, and shows that the claim fails when the festival lift is below about 0.145 of the illustrative model's. The book contains 24 runnable Python programs. (https://5dlibrary.com/answers/en/malaysia-moving-festival-calendars-promotion-alignment/c08/) 第8章 实证闭环:马新台促销排期、移动节庆对齐器与预注册检验 — 第八章给出马新台三地的促销排期表、蒸馏小工具「移动节庆对齐器」与预注册检验方案,并指出节日提升幅度低于示意模型的约 0.145 倍时命题不成立。书中含 24 段可运行的 Python 程序。 (https://5dlibrary.com/answers/zh/malaysia-moving-festival-calendars-promotion-alignment/c08/) ## 5DL-B-314 · Cash-on-Delivery Refusal Games and the Trust Deficit in Southeast Asian E-Commerce: Deposit Thresholds and Prepayment-Discount Incentive Design 中文书名:货到付款拒收博弈与东南亚电商信任赤字、押金阈值与预付折扣激励设计母典 DOI: pending Cite: Fan, Y. (2026). Cash-on-Delivery Refusal Games and the Trust Deficit in Southeast Asian E-Commerce: Deposit Thresholds and Prepayment-Discount Incentive Design (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: No peer-reviewed estimate of refusal rates in Southeast Asia was found, so all parameters are illustrative, and all calculations are generated by the book's programs. This is general analysis, not individualised business or legal advice. Abstract: Cash on delivery lets a stranger order with confidence and change his mind at the door for free. This book takes one order apart: with illustrative parameters an accepted order earns a margin of 4.5342, a refused one costs 4.3522 in two shipping legs, damage and tied-up capital, and the break-even refusal rate is 0.4202; under the cautious assumption that any payment by the buyer is voluntary, collection breaks even only when the product of reach and payment rates reaches 0.2068, which is rarely achieved. The central claim is that refusal losses cannot be recovered by chasing buyers afterwards, and that a small deposit or a prepayment discount offered at checkout lets buyers separate themselves by type. Before writing, the book checked the original papers and corrects the blueprint: Rothschild and Stiglitz (1976) study competitive insurance markets, the transfer to cash on delivery is this book's analogy, and they prove that a separating equilibrium may fail to exist; the statement that collection cannot recover the loss is this book's own model result, not theirs; the closer precedent for screening with a deposit is Bester's (1985) screening with collateral. No peer-reviewed estimate of refusal rates in Southeast Asia was found, so all parameters are illustrative, and all calculations are generated by the book's programs. Chapter 2 solves the separating menu exactly in fractions: committed buyers post a deposit of 4.1108 while casual buyers stay with plain cash on delivery, and when casual buyers are no more than 0.1900 of the market a single pooling contract breaks the menu. The trust deficit is the price of screening: a deposit that stays within the refusal cost and still benefits committed buyers exists only inside a window of trust costs. Chapter 3 turns a sweep of deposit thresholds into a rule that uses only the buyer's own order record. Chapter 4 finds an optimal prepayment discount of 0.5556, about 0.0278 of the price. Chapter 5 separates refusals by fault, keeps a deposit only for refusals the buyer personally caused, and measures how often a flag wrongly catches committed buyers. Chapter 6 simulates four policies with fixed seeds; the deposit rule with appeal and review earns 3535.44 more than cash on delivery alone. Chapter 7 examines the line from the market chapter of the Rites of Zhou on settling trust with written contracts to stop lawsuits, and Book XX of Montesquieu's The Spirit of the Laws. Chapter 8 gives the payment-method rule table, the distilled COD Deposit Threshold Calculator and a pre-registered test. The rules use no protected trait, never treat areas differently, always offer appeal and human review, and state when deposits are refunded. This is general analysis, not individualised business or legal advice. The book contains 19 runnable Python programs. 摘要:货到付款让陌生买家敢于下单,也让他可以在门口免费反悔。流行的看法要么把拒收交给事后追讨,要么把它当作固定成本平摊给所有人。本书把一单的账拆开:示意参数下,签收一单的毛利为 4.5342,拒收一单要付出双程运费、货损与资金占用共 4.3522,盈亏平衡拒收率为 0.4202;在买家自愿付款的保守假设下,追讨要打平需要联系率与付款率之积达到 0.2068,现实中很难做到。核心命题是:拒收损失不能靠事后催收弥补,应在下单时用小额押金或预付折扣让买家按类型自我分离。落笔前本书核对原文并作蓝图修正:Rothschild 与 Stiglitz 1976 年研究的是竞争性保险市场,移植到货到付款是本书的类比,而且他们证明分离均衡可能不存在;「催收补不回」是本书模型的结论,不出自原文;押金筛选更直接的先例是 Bester 1985 年的抵押品筛选。东南亚拒收率没有找到经同行评议的估计,书中参数全部为示意,全部计算由书中程序生成。 第二章用分数精确求出分离菜单:守信买家的押金为 4.1108,随手下单的买家留在普通货到付款;当随手下单买家的占比不超过 0.1900 时,一份混同合同就能打破菜单。信任赤字是筛选的代价:押金不超过拒收成本、又让守信买家得益,只在一个信任成本窗口里成立。第三章把押金阈值随客单价与拒收率扫描,压缩为只用买家本人订单记录的押金规则。第四章求出最优预付折扣 0.5556,约为售价的 0.0278。第五章区分拒收责任,说明只对买家本人造成的拒收扣留押金,并计算标记规则对守信买家的误伤。第六章以固定种子模拟四种政策,押金规则加申诉复核比只做货到付款多赚 3535.44。 第七章考辨《周礼·地官·司市》「以质剂结信而止讼」与孟德斯鸠《论法的精神》第二十篇。第八章给出订单付款方式规则表、蒸馏小工具「货到付款押金阈值计算器」与预注册检验方案。规则不得使用任何受保护群体特征,不按地区差别对待,必须提供申诉与人工复核,押金写清退还条件。本书属于非个人化经营或法律建议。书中含 19 段可运行的 Python 程序。 Chapter answers: - Ch.1 The Ledger of One Cash-on-Delivery Order: Two Shipping Legs, Damage, Tied-Up Capital and the Arithmetic of Collection — Expected profit falls fast with the refusal rate. At a rate of 0.05 the expected profit per order is 3.2899; at 0.20 it is only 1.9570; at 0.30 it is 1.0683; at 0.40 it is just 0.1797. The break-even refusal rate is 0.4202. The last line puts it another way: each refusal wipes out 0.960 of an accepted margin. The column headed loss share of margin is even more direct: at a refusal rate of 0.20 the refusal losses already eat 0.240 of the margin earned on accepted orders, and at 0.30 they eat 0.411. (https://5dlibrary.com/answers/en/cash-on-delivery-refusal-deposit-prepay-discount-screening/c01/) 第1章 一单货到付款的账本:双程运费、货损、资金占用与催收的算术 — 期望利润随拒收率下降得很快。拒收率 0.05 时,每单期望利润 3.2899;到 0.20 时只剩 1.9570;到 0.30 时是 1.0683;到 0.40 时只有 0.1797。盈亏平衡的拒收率是 0.4202。最后一行换一种说法:每发生一次拒收,就抵掉 0.960 单签收的毛利。表中「损失占毛利的比例」一列更直观:拒收率 0.20 时,拒收损失已经吃掉签收毛利的 0.240;拒收率 0.30 时吃掉 0.411。 (https://5dlibrary.com/answers/zh/cash-on-delivery-refusal-deposit-prepay-discount-screening/c01/) - Ch.2 Screening Equilibria: Carrying Rothschild and Stiglitz from the Insurance Market to the Courier's Doorstep — Chapter 2 solves the separating menu exactly in fractions: committed buyers post a deposit of 4.1108 while casual buyers stay with plain cash on delivery, and when casual buyers are no more than 0.1900 of the market a single pooling contract breaks the menu. The trust deficit is the price of screening: a deposit that stays within the refusal cost and still benefits committed buyers exists only inside a window of trust costs. (https://5dlibrary.com/answers/en/cash-on-delivery-refusal-deposit-prepay-discount-screening/c02/) 第2章 筛选均衡:把 Rothschild 与 Stiglitz 的保险市场搬到快递员的门口 — 第二章用分数精确求出分离菜单:守信买家的押金为 4.1108,随手下单的买家留在普通货到付款;当随手下单买家的占比不超过 0.1900 时,一份混同合同就能打破菜单。信任赤字是筛选的代价:押金不超过拒收成本、又让守信买家得益,只在一个信任成本窗口里成立。 (https://5dlibrary.com/answers/zh/cash-on-delivery-refusal-deposit-prepay-discount-screening/c02/) - Ch.3 Deposit Thresholds: Sweeping Basket Values and Refusal Rates, and a Rule That Reads Only the Buyer's Own Record — The first line lists six illustrative records and the second the matching posterior means. A buyer with two orders and one refusal has a posterior mean of 0.167 but fewer than five orders, so he still gets plain cash on delivery. Buyers with five orders and no refusal, or five orders and one refusal, have posterior means of 0.067 and 0.133, both below 0.15. A buyer with five orders and two refusals has a posterior mean of 0.200 and starts to be asked for a deposit: 3.0 units on a basket of 20 and 4.0 units on a basket of 45. A buyer with eight orders and four refusals, at 0.278, is asked for 3.5 to 4.5 units. A buyer with ten orders and nine refusals, at 0.500, has his deposit capped by the refusal cost on baskets up to 30 and is also offered prepayment. The basket of 8 is prepay-only for everyone. (https://5dlibrary.com/answers/en/cash-on-delivery-refusal-deposit-prepay-discount-screening/c03/) 第3章 押金阈值:随客单价与拒收率的扫描,以及一条只看买家本人记录的规则 — 第一行是六种示意记录,第二行是对应的后验均值。只下过两单、拒收过一次的买家,后验均值是 0.167,但订单不满五次,仍然得到普通货到付款。五单零拒收与五单一拒收的买家,后验均值分别为 0.067 与 0.133,都在 0.15 以下。五单两拒收的买家后验均值为 0.200,开始被要求押金:客单价 20 时押金 3.0 个单位,客单价 45 时 4.0 个单位。八单四拒收的买家后验均值 0.278,押金 3.5 到 4.5 个单位。十单九拒收的买家后验均值 0.500,在客单价不超过 30 的篮子上,押金被拒收成本封顶,同时提供预付。售价 8 的篮子对所有人只开放预付。 (https://5dlibrary.com/answers/zh/cash-on-delivery-refusal-deposit-prepay-discount-screening/c03/) - Ch.4 The Optimal Prepayment Discount: Fees Saved, Fewer Second Thoughts and Money Given Away — Chapter 4 finds an optimal prepayment discount of 0.5556, about 0.0278 of the price. (https://5dlibrary.com/answers/en/cash-on-delivery-refusal-deposit-prepay-discount-screening/c04/) 第4章 最优预付折扣:省下的代收费、更少的反悔与让出去的钱 — 第四章求出最优预付折扣 0.5556,约为售价的 0.0278。 (https://5dlibrary.com/answers/zh/cash-on-delivery-refusal-deposit-prepay-discount-screening/c04/) - Ch.5 Whose Failure? Deterrence by Deposit, Fair Forfeiture and the Error Rate of a Flag Built on the Buyer's Own Record — Look at deterrence first. Without a deposit, casual buyers refuse in person at 0.3250 and, with the other two kinds, fail at 0.3500 in total. With the deposit of 4.111 computed in Chapter 2, personal refusals fall to 0.1638 and total failures to 0.1888; the second-to-last line shows that the personal refusals of casual buyers fall by 0.496, close to half. With a deposit of 5.000 they fall to 0.1412. The marginal effect of deterrence declines: raising the deposit beyond 4.111 buys little further reduction, while the cost borne by committed buyers keeps rising. (https://5dlibrary.com/answers/en/cash-on-delivery-refusal-deposit-prepay-discount-screening/c05/) 第5章 谁的失败:押金的威慑、公平没收与只凭本人记录的误伤率 — 先看威慑。不交押金时,随手买家本人拒收的比率是 0.3250,加上另外两类,总失败率为 0.3500。押金取第二章算出的 4.111 时,本人拒收降到 0.1638,总失败率降到 0.1888;倒数第二行说明,随手买家本人拒收减少了 0.496,接近一半。押金升到 5.000 时,本人拒收降到 0.1412。威慑的边际效果递减,押金从 4.111 再加大,换来的减少有限,而守信买家承担的成本却在上升。 (https://5dlibrary.com/answers/zh/cash-on-delivery-refusal-deposit-prepay-discount-screening/c05/) - Ch.6 A Fixed-Seed Order Simulation: Four Payment Policies Applied to the Same Buyers — Chapter 6 simulates four policies with fixed seeds; the deposit rule with appeal and review earns 3535.44 more than cash on delivery alone. (https://5dlibrary.com/answers/en/cash-on-delivery-refusal-deposit-prepay-discount-screening/c06/) 第6章 固定种子的订单模拟:四种付款政策在同一批买家身上的账 — 第六章以固定种子模拟四种政策,押金规则加申诉复核比只做货到付款多赚 3535.44。 (https://5dlibrary.com/answers/zh/cash-on-delivery-refusal-deposit-prepay-discount-screening/c06/) - Ch.7 Classics East and West: Written Contracts in the Market Chapter of the Rites of Zhou and Montesquieu on Commerce — The deposit lowers personal refusals from 0.20 to 0.1131, with a loss of 0.6183 per order. Under coercion, if nothing is ever recovered, the loss is 1.9573 per order; with a recovery probability of 0.40 it is 1.2039; even with recovery as likely as 0.80, the loss is 0.7058, still 0.0875 more than with the deposit. The last line says that coercion after the fact catches up with the deposit only if recovery succeeds with probability 0.8920. Chapter 1 explained that in cash on delivery a refusal usually creates no enforceable debt, so real recovery rates are far below that figure. Montesquieu's approach was to make up for missing trust with coercion; this book's conclusion is to replace coercion with choice. (https://5dlibrary.com/answers/en/cash-on-delivery-refusal-deposit-prepay-discount-screening/c07/) 第7章 东西方原典考辨:《周礼·司市》的质剂与孟德斯鸠论商业 — 事前押金把本人拒收从 0.20 降到 0.1131,每单损失 0.6183。事后强制下,若一分钱也收不回,每单损失 1.9573;收回概率 0.40 时,损失 1.2039;即使收回概率高达 0.80,损失仍是 0.7058,比押金多 0.0875。最后一行说,事后强制要追上事前押金,收回概率必须达到 0.8920。第一章已经说明,在货到付款里,拒收通常并不构成可以强制执行的债务,现实中的收回概率远低于这个数。孟德斯鸠的思路是用强制补足信任,本书的结论是用选择代替强制。 (https://5dlibrary.com/answers/zh/cash-on-delivery-refusal-deposit-prepay-discount-screening/c07/) - Ch.8 Closing the Loop: the Payment-Method Rule Table, the Deposit Threshold Calculator and a Pre-Registered Test — Chapter 8 gives the payment-method rule table, the distilled COD Deposit Threshold Calculator and a pre-registered test. The rules use no protected trait, never treat areas differently, always offer appeal and human review, and state when deposits are refunded. This is general analysis, not individualised business or legal advice. The book contains 19 runnable Python programs. (https://5dlibrary.com/answers/en/cash-on-delivery-refusal-deposit-prepay-discount-screening/c08/) 第8章 实证闭环:订单付款方式规则表、押金阈值计算器与预注册检验 — 第八章给出订单付款方式规则表、蒸馏小工具「货到付款押金阈值计算器」与预注册检验方案。规则不得使用任何受保护群体特征,不按地区差别对待,必须提供申诉与人工复核,押金写清退还条件。本书属于非个人化经营或法律建议。书中含 19 段可运行的 Python 程序。 (https://5dlibrary.com/answers/zh/cash-on-delivery-refusal-deposit-prepay-discount-screening/c08/) ## 5DL-B-315 · Loyalty-Point Liabilities and Breakage Actuarial Science: Issuance, Redemption Rates and the Insolvency Boundary of Deferred Revenue 中文书名:会员积分负债与沉睡率精算、积分发行兑付率与递延收入的破产边界母典 DOI: pending Cite: Fan, Y. (2026). Loyalty-Point Liabilities and Breakage Actuarial Science: Issuance, Redemption Rates and the Insolvency Boundary of Deferred Revenue (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs; this is not individualised accounting, tax or legal advice, and a practising accountant's opinion prevails. Chapter 1 reworks the allocation and the cumulative catch-up with the numbers of the standard's illustrative example. Abstract: Loyalty points are often treated as a rounding error in the marketing budget; this book treats them as a liability that the issuer prints at will and must one day honour. Before writing, the book checked the standard and corrects the blueprint: IFRS 15 treats points that give the customer a material right as a separate performance obligation, allocates the transaction price on relative stand-alone selling prices, and requires the estimate for the points to reflect the likelihood of exercise (application guidance B39 to B43); the provisions on unexercised rights, or breakage, are in B44 to B47, where expected breakage to which the entity is entitled is recognised in proportion to the pattern of rights exercised. The claim that ignoring breakage misstates the liability needs a direction: ignoring breakage altogether overstates the liability by 0.473 in the illustrative data, while counting unredeemed points in immature data understates the deferral per point by 0.272, the more dangerous error. Parameters are illustrative and all calculations are generated by the book's programs; this is not individualised accounting, tax or legal advice, and a practising accountant's opinion prevails. Chapter 1 reworks the allocation and the cumulative catch-up with the numbers of the standard's illustrative example. Chapter 2 handles censoring with the Kaplan-Meier curve and shows that the same data, projected to expiry, give breakage of 0.083, 0.179 and 0.250 under geometric, Weibull and mixture models, against a true 0.286. Chapter 3 gives Wilson intervals on matured cohorts, the gap between breakage by points (0.169) and by members (0.422), and a reversal constraint met by booking a posterior percentile. Chapter 4 separates the selling-price liability from the present value of fulfilment cost and shows that a clear reminder before expiry lowers breakage from 0.2862 to 0.2717. Chapter 5 simulates points inflation and the redemption run that follows a devaluation notice. Chapter 6 draws the insolvency boundary: keeping the two-year chance of overdraft below one in a hundred needs a reserve ratio of 0.492, and a single expiry reminder takes 0.121 of the profit of a scheme that depends on breakage. Chapter 7 examines flying cash in the Xin Tang Shu and Smith's account of receipts at the Bank of Amsterdam, turning lapsed receipts into a breakage model that predicts gold receipts lapse 2.0 times as often as silver. Chapter 8 gives a monthly loyalty-liability report, the distilled Loyalty Liability Actuary and a pre-registered test, and states plainly that scheme rules must not use hidden expiry terms to create breakage and that members should be told before points lapse. The book contains 21 runnable Python programs. 摘要:会员积分常被当作营销费用里的零头,本书把它当作一种由发行者单方面印制、日后必须兑付的负债来精算。落笔前本书核对准则并作蓝图修正:IFRS 15 把给予客户重大权利的积分视为单独的履约义务,按单独售价的相对比例分摊交易价格,积分单独售价的估计要反映被行使的可能性(应用指南 B39 至 B43);关于客户未行使权利即沉睡的规定在 B44 至 B47,预期有权获得的沉睡金额按客户行使权利的模式按比例确认为收入。「忽略沉睡会错估负债」须分方向:完全忽略沉睡在示意数据中把负债多记 0.473,而用未成熟数据直接计数则把每分递延额少记 0.272,后者更危险。书中参数为示意;全部计算由书中程序生成;本书是非个人化会计、税务或法律建议,以执业会计师意见为准。 第一章用准则释例的数字演示分摊与累计追补。第二章用 Kaplan–Meier 曲线处理删失,并显示同一份数据外推到到期日,几何、Weibull 与混合模型给出 0.083、0.179 与 0.250 的沉睡率,真实值 0.286。第三章给出成熟批次的 Wilson 区间、按点数(0.169)与按人数(0.422)的沉睡率之别,以及按后验分位数入账的收入转回约束。第四章区分售价负债与履约成本现值,并显示到期前清楚提醒可使沉睡率从 0.2862 降到 0.2717。第五章模拟积分通胀与贬值公告引发的兑付挤兑。第六章画出破产边界:两年内透支概率不超过百分之一需要 0.492 的准备金比例,而一封到期提醒就会使依赖沉睡的计划少赚 0.121 的利润。 第七章考辨《新唐书·食货志》的飞钱与斯密笔下阿姆斯特丹银行的存单,把收据失效写成沉睡模型,预测黄金收据失效频率是白银的 2.0 倍。第八章给出会员积分负债月报、蒸馏小工具「积分负债精算器」与预注册检验方案,并明确主张积分规则不得以隐藏过期条款制造沉睡,应提前告知到期。书中含 21 段可运行的 Python 程序。 Chapter answers: - Ch.1 Issuing Points Creates a Liability: Material Rights, Stand-Alone Selling Prices and the Cumulative Catch-Up — Program p11 reworks the pattern of the illustrative example. The stand-alone selling price of the points is the 9,500 points expected to be redeemed times the value per point; the product's stand-alone selling price is 100,000; the sum is the denominator, so the points receive 8,676 and the product 91,324. In the first period 4,500 points are redeemed, so revenue on points is 4,500 over 9,500 times 8,676, or 4,110, and the closing contract liability is 4,566. In the second period cumulative redemptions reach 8,500 and the estimate of total expected redemptions is revised to 9,700. History is not rewritten: the cumulative share gives cumulative revenue of 7,603, the 4,110 already recognised is subtracted, and second-period revenue is 3,493, leaving a liability of 1,073. (https://5dlibrary.com/answers/en/loyalty-point-liability-breakage-deferred-revenue/c01/) 第1章 积分发行即负债:重大权利、单独售价分摊与累计追补 — 程序 p11 重算准则释例的格局。积分的单独售价等于预期兑换的 9,500 分乘以每分价值,商品的单独售价是 100,000,二者相加为分母,积分分到 8,676,商品分到 91,324。第一期兑换 4,500 分,积分收入等于 4,500 除以 9,500 再乘以 8,676,即 4,110,期末合同负债 4,566。第二期累计兑换 8,500 分,同时把预期兑换总数修订为 9,700 分;这时不重算历史,而是用累计比例重算累计收入 7,603,再减去第一期已确认的 4,110,得到第二期收入 3,493,期末负债 1,073。 (https://5dlibrary.com/answers/zh/loyalty-point-liability-breakage-deferred-revenue/c01/) - Ch.2 Redemption Survival Curves: Censoring, the Product-Limit Estimator and Extrapolation to Expiry — Chapter 2 handles censoring with the Kaplan-Meier curve and shows that the same data, projected to expiry, give breakage of 0.083, 0.179 and 0.250 under geometric, Weibull and mixture models, against a true 0.286. (https://5dlibrary.com/answers/en/loyalty-point-liability-breakage-deferred-revenue/c02/) 第2章 兑付生存曲线:删失、乘积限估计与到期日外推 — 第二章用 Kaplan–Meier 曲线处理删失,并显示同一份数据外推到到期日,几何、Weibull 与混合模型给出 0.083、0.179 与 0.250 的沉睡率,真实值 0.286。 (https://5dlibrary.com/answers/zh/loyalty-point-liability-breakage-deferred-revenue/c02/) - Ch.3 Estimating Expected Breakage: Matured Cohorts, Point Weighting and the Reversal Constraint — Chapter 3 gives Wilson intervals on matured cohorts, the gap between breakage by points (0.169) and by members (0.422), and a reversal constraint met by booking a posterior percentile. (https://5dlibrary.com/answers/en/loyalty-point-liability-breakage-deferred-revenue/c03/) 第3章 预期沉睡率的估计与区间:成熟批次、按点加权与收入转回约束 — 第三章给出成熟批次的 Wilson 区间、按点数(0.169)与按人数(0.422)的沉睡率之别,以及按后验分位数入账的收入转回约束。 (https://5dlibrary.com/answers/zh/loyalty-point-liability-breakage-deferred-revenue/c03/) - Ch.4 Present Value and Expiry Terms: Discounted Fulfilment Cost, the Expiry Horizon and Advance Notice — Chapter 4 separates the selling-price liability from the present value of fulfilment cost and shows that a clear reminder before expiry lowers breakage from 0.2862 to 0.2717. (https://5dlibrary.com/answers/en/loyalty-point-liability-breakage-deferred-revenue/c04/) 第4章 负债现值与到期条款:履约成本折现、到期期限与提前告知 — 第四章区分售价负债与履约成本现值,并显示到期前清楚提醒可使沉睡率从 0.2862 降到 0.2717。 (https://5dlibrary.com/answers/zh/loyalty-point-liability-breakage-deferred-revenue/c04/) - Ch.5 Points Inflation and Redemption Runs: Issuance Speed, Expected Devaluation and the Two-Month Window — With a flat earn rate, outstanding points expected to redeem number 12,633,312, the liability is 0.126 of monthly sales and the month's redemption cost is 0.0038 of sales; the long-run steady state is 15,000,000 points. If the earn rate grows by 0.01 a month, after three years there are 15,912,370 points outstanding, the liability is 0.159 of monthly sales and redemption cost 0.0047; bringing that cost back to the flat-rate level needs a cut of 0.200 in point value. At 0.03 a month the liability is 0.257 of monthly sales and the cut needed is 0.496. At 0.05 a month there are 42,255,277 points outstanding, the liability is 0.423 of monthly sales, redemption cost 0.0120, and point value must be cut by 0.688. (https://5dlibrary.com/answers/en/loyalty-point-liability-breakage-deferred-revenue/c05/) 第5章 积分通胀与兑付挤兑:发放速度、贬值预期与两个月的窗口 — 送分比例不增长时,在外的预期兑换积分是 12,633,312 分,负债相当于月销售额的 0.126,当月兑换成本是销售额的 0.0038;长期稳态是 15,000,000 分。送分比例每月增长 0.01,三年后在外积分 15,912,370 分,负债占月销售额 0.159,兑换成本占 0.0047;若要把兑换成本压回不增长时的水平,积分价值要下调 0.200。每月增长 0.03,负债占 0.257,需要下调 0.496;每月增长 0.05,在外积分 42,255,277 分,负债占 0.423,兑换成本占 0.0120,积分价值要砍掉 0.688。 (https://5dlibrary.com/answers/zh/loyalty-point-liability-breakage-deferred-revenue/c05/) - Ch.6 The Insolvency Boundary: Reserve-to-Liability Ratios, Dependence on Breakage and a Two-Dimensional Stress Map — Chapter 6 draws the insolvency boundary: keeping the two-year chance of overdraft below one in a hundred needs a reserve ratio of 0.492, and a single expiry reminder takes 0.121 of the profit of a scheme that depends on breakage. (https://5dlibrary.com/answers/en/loyalty-point-liability-breakage-deferred-revenue/c06/) 第6章 破产边界:准备金 / 负债比、沉睡依赖与二维压力图 — 第六章画出破产边界:两年内透支概率不超过百分之一需要 0.492 的准备金比例,而一封到期提醒就会使依赖沉睡的计划少赚 0.121 的利润。 (https://5dlibrary.com/answers/zh/loyalty-point-liability-breakage-deferred-revenue/c06/) - Ch.7 Sources East and West: Flying Cash in Chang'an and Receipts in Amsterdam — Chapter 7 examines flying cash in the Xin Tang Shu and Smith's account of receipts at the Bank of Amsterdam, turning lapsed receipts into a breakage model that predicts gold receipts lapse 2.0 times as often as silver. (https://5dlibrary.com/answers/en/loyalty-point-liability-breakage-deferred-revenue/c07/) 第7章 东西方原典考辨:长安的飞钱与阿姆斯特丹的存单 — 第七章考辨《新唐书·食货志》的飞钱与斯密笔下阿姆斯特丹银行的存单,把收据失效写成沉睡模型,预测黄金收据失效频率是白银的 2.0 倍。 (https://5dlibrary.com/answers/zh/loyalty-point-liability-breakage-deferred-revenue/c07/) - Ch.8 Closing the Loop: a Monthly Loyalty-Liability Report, the Loyalty Liability Actuary and a Pre-Registered Test — Chapter 8 gives a monthly loyalty-liability report, the distilled Loyalty Liability Actuary and a pre-registered test, and states plainly that scheme rules must not use hidden expiry terms to create breakage and that members should be told before points lapse. The book contains 21 runnable Python programs. (https://5dlibrary.com/answers/en/loyalty-point-liability-breakage-deferred-revenue/c08/) 第8章 实证闭环:会员积分负债月报、积分负债精算器与预注册检验 — 第八章给出会员积分负债月报、蒸馏小工具「积分负债精算器」与预注册检验方案,并明确主张积分规则不得以隐藏过期条款制造沉睡,应提前告知到期。书中含 21 段可运行的 Python 程序。 (https://5dlibrary.com/answers/zh/loyalty-point-liability-breakage-deferred-revenue/c08/) ## 5DL-B-316 · Assurance Contracts and Pre-Sale Group Thresholds: Refund Promises and Free-Riding Equilibria in Public Goods 中文书名:保证合约预售成团阈值、退款承诺与公共品搭便车均衡母典 DOI: pending Cite: Fan, Y. (2026). Assurance Contracts and Pre-Sale Group Thresholds: Refund Promises and Free-Riding Equilibria in Public Goods (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations are generated by the book's programs. The book is general analysis, not individualised business or legal advice. Abstract: When a limited pre-sale fails to gather enough orders, free-riding is usually blamed, and "full refund if the threshold is not reached" is often offered as the ready answer. This book breaks that claim into propositions that can be computed. Before writing, the original papers were checked and the blueprint corrected: the 1989 paper by Bagnoli and Lipman is a theoretical result proving that under a refund rule the undominated perfect equilibrium outcomes coincide with the core, and it measures no probability of success; Palfrey and Rosenthal had shown in 1984 that, looking only at Nash equilibria, the equilibria under a refund rule include those under a no-refund rule; the experimental evidence points both ways; and Hume's meadow comes from Book 3 of the Treatise, published in 1740. Parameters are illustrative; all calculations are generated by the book's programs. The book is general analysis, not individualised business or legal advice. Chapter 1 enumerates a five-person threshold game exactly: the share of equilibria that lead to production is 0.909 under the keep rule and 0.625 under the refund rule; in symmetric mixed equilibrium the refund removes the low equilibrium with a chance of production of 0.2161, and the remaining equilibrium has a chance of 0.8256. Chapter 2 discretises the continuous contribution game, compares it with the core, and reproduces with an arrival model the crowdfunding pattern of failures that miss by a lot and successes that pass by a little. Chapter 3 computes the chance of success before a deadline: only 0.5133 when expected orders equal the threshold. Chapter 4 computes the expected cost of the refund bonus in a dominant assurance contract, whose share of pledged value never exceeds the ratio of bonus to price. Chapter 5 shows that expected profit is highest when the threshold equals break-even. Chapter 6 gives rules for displaying true progress, deadlines, places left and refunds, with an audit program. Chapter 7 examines the Li Yun line that goods are hated when thrown away upon the ground but need not be stored for oneself, and Hume's meadow: if everyone must take part and each slips with chance one in a hundred, joint action by a thousand people succeeds with chance only 0.000043. Chapter 8 gives the operating procedure for a limited pre-sale page, the distilled Threshold Chance Calculator and a pre-registered test whose whole-campaign design needs 392 pre-sales per arm. The book contains 20 runnable Python programs. 摘要:限量预售凑不齐人,常被归咎于搭便车,而「达不到门槛全额退款」常被当作现成的解药。本书把这句话拆成可以计算的命题。落笔前本书核对原文并作蓝图修正:Bagnoli 与 Lipman 1989 年的论文是一项理论结果,证明退款规则下的非劣完美均衡结果恰好等于核,并未测量达标概率;Palfrey 与 Rosenthal 1984 年已证明,只看纳什均衡,退款规则的均衡集合包含不退规则的均衡集合;实验证据有正有反;休谟的草地一例出自 1740 年出版的《人性论》第三卷。书中参数为示意;全部计算由书中程序生成。本书属于非个人化经营或法律建议。 第一章精确枚举五人门槛博弈:能开工的均衡所占比例在不退规则下为 0.909,在退款规则下为 0.625;在对称混合均衡中,退款清除了开工概率为 0.2161 的低均衡,剩余均衡的开工概率为 0.8256。第二章把连续出资博弈离散化并与核对照,并以到达模型重现众筹数据中「失败差得远、成功超得少」的形状。第三章计算截止前的达标概率:预期订单恰等于门槛时只有 0.5133。第四章计算优势保证合约中退款奖励的期望成本,其占认捐价款的比例不超过奖励与单价之比。第五章说明期望利润在门槛等于保本线时最大。第六章给出真实进度、截止、余量与退款的展示规则及审计程序。 第七章考辨《礼记·礼运》「货恶其弃于地也,不必藏于己」与休谟的草地排水一例:在人人须参加且每人有百分之一疏忽时,一千人行动成功的概率只有 0.000043。第八章给出限量预售成团页的操作流程、蒸馏小工具「成团门槛概率计算器」与预注册检验方案,整场设计每组需要 392 场预售。书中含 20 段可运行的 Python 程序。 Chapter answers: - Ch.1 The Equilibrium List of a Threshold Game: What a Refund Promise Actually Changes — Chapter 1 enumerates a five-person threshold game exactly: the share of equilibria that lead to production is 0.909 under the keep rule and 0.625 under the refund rule; in symmetric mixed equilibrium the refund removes the low equilibrium with a chance of production of 0.2161, and the remaining equilibrium has a chance of 0.8256. (https://5dlibrary.com/answers/en/assurance-contract-presale-group-threshold-refund/c01/) 第1章 门槛博弈的均衡清单:退款承诺到底改变了什么 — 第一章精确枚举五人门槛博弈:能开工的均衡所占比例在不退规则下为 0.909,在退款规则下为 0.625;在对称混合均衡中,退款清除了开工概率为 0.2161 的低均衡,剩余均衡的开工概率为 0.8256。 (https://5dlibrary.com/answers/zh/assurance-contract-presale-group-threshold-refund/c01/) - Ch.2 From the Core to Crowdfunding: Refined Theory, Laboratory Evidence and Platform Data — The model's output is close in shape to the paper's figures. The share reaching the goal is 0.469 in the model, against 0.489 computed from the paper's counts. Failed projects raise on average 0.112 of the goal, against 0.103; the share of failures raising three tenths or more is 0.115, against 0.100; the share raising half or more is 0.063, against 0.030; the share of funded projects ending less than three per cent over is 0.241, against 0.250; and the share ending less than a tenth over is 0.503, consistent with the paper's "about half". (https://5dlibrary.com/answers/en/assurance-contract-presale-group-threshold-refund/c02/) 第2章 从核到众筹:理论精炼、实验室证据与平台数据 — 模型的输出与论文的数字在形状上相当接近。模型中达到目标的比例为 0.469,论文中按计数算为 0.489;失败项目平均筹到目标的 0.112,论文为 0.103;失败项目中筹到三成以上的占 0.115,论文为 0.100;筹到一半以上的占 0.063,论文为 0.030;成功项目中超出不到百分之三的占 0.241,论文为 0.250;超出不到一成的占 0.503,与论文「约一半」相符。 (https://5dlibrary.com/answers/zh/assurance-contract-presale-group-threshold-refund/c02/) - Ch.3 Before the Deadline: Success Probabilities under Poisson Arrivals, Binomial Audiences and Uncertain Rates — Chapter 3 computes the chance of success before a deadline: only 0.5133 when expected orders equal the threshold. (https://5dlibrary.com/answers/en/assurance-contract-presale-group-threshold-refund/c03/) 第3章 截止之前:泊松到达、二项受众与速率不确定下的达标概率 — 第三章计算截止前的达标概率:预期订单恰等于门槛时只有 0.5133。 (https://5dlibrary.com/answers/zh/assurance-contract-presale-group-threshold-refund/c03/) - Ch.4 The Refund Bonus: Mechanism, Equilibria and Expected Cost of the Dominant Assurance Contract — The table shows three regularities. First, the expected bonus is largest when the chance of accepting puts the average number of accepters just below the threshold: at q of 0.15, an average of 45 people, the chance of failure is 0.9884 and the expected bonus 442.85; at q of 0.18 it is 411.45. Second, once the average passes the threshold the expected bonus falls quickly to nothing: 102.01 at q of 0.22, only 9.97 at 0.25, and 0.02 at 0.30. Third, the expected bonus as a share of the expected value of pledges never exceeds the ratio of bonus to price, 10 to 200, which is 0.0500, and at q of 0.20, where the average exactly equals the threshold, it is only 0.0215. (https://5dlibrary.com/answers/en/assurance-contract-presale-group-threshold-refund/c04/) 第4章 退款奖励:优势保证合约的机制、均衡与期望成本 — 表中有三个规律。其一,期望补偿在接受概率使平均人数略低于门槛时最大:q 为 0.15、平均 45 人时,失败概率为 0.9884,期望补偿为 442.85;q 为 0.18 时为 411.45。其二,当平均人数超过门槛之后,期望补偿迅速归零:q 为 0.22 时只有 102.01,q 为 0.25 时只有 9.97,q 为 0.30 时只有 0.02。其三,期望补偿占认捐价款期望的比例,在任何情况下都不超过补偿额与单价之比,也就是 10 比 200 的 0.0500,并且在 q 为 0.20、平均恰好等于门槛时只有 0.0215。 (https://5dlibrary.com/answers/zh/assurance-contract-presale-group-threshold-refund/c04/) - Ch.5 Where to Put the Threshold: Break-Even, Sensitivity and the Weight of Pilot Data — The table gives a clear conclusion: expected profit is highest when the threshold equals the break-even count exactly, at 1420.6. Lowering the threshold to 100 raises the chance of success from 0.6194 to 0.7777 but lowers expected profit to 1339.1, and a confirmed batch then loses money with chance 0.2035; lowering it to 60 gives a chance of success as high as 0.9723 but an expected profit of only 980.4, and a confirmed batch then loses money with chance 0.3629. Raising the threshold to 140 or more also lowers expected profit, because more campaigns fall short, and many of those would have made money. (https://5dlibrary.com/answers/en/assurance-contract-presale-group-threshold-refund/c05/) 第5章 门槛设在哪里:保本线、敏感度与试销数据的分量 — 表中有一个清楚的结论:期望利润在门槛恰好等于保本线时最大,为 1420.6。门槛降到 100,成团概率从 0.6194 升到 0.7777,期望利润却降到 1339.1,并且成团之后仍然亏损的概率为 0.2035;门槛降到 60,成团概率高达 0.9723,期望利润只剩 980.4,成团之后亏损的概率达 0.3629。门槛升到 140 以上,期望利润也下降,因为凑不齐的场次增多,而这些场次里本来有不少是能赚钱的。 (https://5dlibrary.com/answers/zh/assurance-contract-presale-group-threshold-refund/c05/) - Ch.6 Rules for Displaying True Progress: Auditing Counts, Deadlines, Remaining Stock and Refunds — Two features of the table deserve attention. First, the straight line overestimates early and underestimates late. On day 3 the planned count is 42 and the straight line gives 196.0, far above the real expectation; on day 12 the planned count is 84 and the straight line gives only 98.0, apparently short, yet with 31.1 orders still expected the chance of success is 0.9989. Second, once the true count falls behind plan, the chance of success drops quickly: at day 10 with only eight tenths of the planned count, 60 orders, the chance is 0.5464; at day 12 with eight tenths, 67 orders, it is only 0.3888. (https://5dlibrary.com/answers/en/assurance-contract-presale-group-threshold-refund/c06/) 第6章 真实进度展示规则:计数、截止、余量与退款的逐条审计 — 这张表有两点值得注意。第一,线性外推在开头高估、在后段低估。第 3 天按计划进度有 42 单,线性外推得 196.0 单,远高于真实预期;第 12 天按计划进度有 84 单,线性外推只得 98.0 单,似乎凑不够,可是按剩余的预期 31.1 单计算,达标概率为 0.9989。第二,真实计数一旦落后计划,达标概率下降得很快:第 10 天若只有计划的八成,即 60 单,达标概率为 0.5464;第 12 天若只有计划的八成,即 67 单,只剩 0.3888。 (https://5dlibrary.com/answers/zh/assurance-contract-presale-group-threshold-refund/c06/) - Ch.7 Classical Sources East and West: "Goods Hate to Lie on the Ground" and Hume's Meadow — Chapter 7 examines the Li Yun line that goods are hated when thrown away upon the ground but need not be stored for oneself, and Hume's meadow: if everyone must take part and each slips with chance one in a hundred, joint action by a thousand people succeeds with chance only 0.000043. (https://5dlibrary.com/answers/en/assurance-contract-presale-group-threshold-refund/c07/) 第7章 东西方原典考辨:「货恶其弃于地」与休谟的草地 — 第七章考辨《礼记·礼运》「货恶其弃于地也,不必藏于己」与休谟的草地排水一例:在人人须参加且每人有百分之一疏忽时,一千人行动成功的概率只有 0.000043。 (https://5dlibrary.com/answers/zh/assurance-contract-presale-group-threshold-refund/c07/) - Ch.8 Closing the Loop: Operating a Limited Pre-Sale Page, the Calculator and a Pre-Registered Test — Chapter 8 gives the operating procedure for a limited pre-sale page, the distilled Threshold Chance Calculator and a pre-registered test whose whole-campaign design needs 392 pre-sales per arm. The book contains 20 runnable Python programs. (https://5dlibrary.com/answers/en/assurance-contract-presale-group-threshold-refund/c08/) 第8章 实证闭环:限量预售成团页的操作流程、计算器与预注册检验 — 第八章给出限量预售成团页的操作流程、蒸馏小工具「成团门槛概率计算器」与预注册检验方案,整场设计每组需要 392 场预售。书中含 20 段可运行的 Python 程序。 (https://5dlibrary.com/answers/zh/assurance-contract-presale-group-threshold-refund/c08/) ## 5DL-B-317 · Phonetic, Graphic and Semantic Distance between Trademarks: Pre-Screening Brand-Name Collisions from Fanqie Spellings to Sound Codes 中文书名:商标音近形近义近三维距离、从反切到音码的品牌撞名预检母典 DOI: pending Cite: Fan, Y. (2026). Phonetic, Graphic and Semantic Distance between Trademarks: Pre-Screening Brand-Name Collisions from Fanqie Spellings to Sound Codes (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: In an illustrative model, with the sound channel at eight tenths, the AUC of the sound code is 0.851 against 0.817 for letter edit distance. Parameters are illustrative; all calculations are generated by the book's programs; the book is general analysis, not individualised legal advice, and judgements of trademark similarity rest with official searches in each jurisdiction and professional agents. Abstract: People who name a new brand often run a single check: if the names do not look alike, they are safe. This book overturns that conclusion. Names unlike in shape can collide in sound: same-sound respellings in English, Chinese characters that differ only in tone, and homophones across scripts between Chinese and Latin names all slip past a shape check. The project blueprint claimed that sound-code edit distance predicts confusion better than shape similarity; after checking studies of drug-name confusion and consumer memory, the book corrects it: similarity of sound and of shape are independent risk factors, phonetic measures lead only where hearing dominates, only sound can compare names across scripts, and a combination beats any single dimension. In an illustrative model, with the sound channel at eight tenths, the AUC of the sound code is 0.851 against 0.817 for letter edit distance. Parameters are illustrative; all calculations are generated by the book's programs; the book is general analysis, not individualised legal advice, and judgements of trademark similarity rest with official searches in each jurisdiction and professional agents. Chapter 1 uses three rulers, edit distance, a sound skeleton and a visual distance, to show that the channel decides which ruler is better. Chapter 2 implements Soundex from the U.S. National Archives guide and checks it example by example; on a synthetic lexicon 0.436 of same-code pairs differ by more than half in sound, and 0.093 of sound-keeping respellings change the code, all through the initial letter. Chapter 3 computes a Chinese syllable distance from initial, final and tone and handles homophones across languages. Chapter 4 measures graphic likeness of characters with a component table and shows that Unicode normalisation must precede any distance. Chapter 5 adds meaning and shows that a weighted mean dilutes a collision in one dimension; in seven illustrative pairs, 5 are caught only by the maximum. Chapter 6 chooses thresholds with ROC curves and cost ratios. Chapter 7 examines Lu Fayan's preface to the Qieyun and Russell's 1918 patent that preceded Soundex, with a program that reads ancient fanqie spellings in modern sounds. Chapter 8 gives a seven-step pre-screening procedure for new brand names, the distilled Brand Name Collision Pre-Screener and a pre-registered test whose primary comparison needs about 1820 name pairs, stating which results would overturn the central claim. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:给新品牌起名的人常常只做一道检查:写出来不像,就算安全。本书推翻这个结论:字形不像的名字可能读音撞车,英文里的同音换写、汉语里只差声调的字、汉字名与拉丁字母名之间的跨文字谐音,都逃得过字形检查。立项蓝图称音码编辑距离比字形相似度更能预测混淆,本书核对药名混淆与消费者记忆研究后作蓝图修正:音近与形近都是独立的危险因素,读音指标只在听觉通道主导时领先,跨文字时只能靠读音,组合优于任一单维。在一个示意模型中,声音通道占八成时音码的 AUC 为 0.851,高于字母编辑距离的 0.817。书中参数为示意;全部计算由书中程序生成;本书是非个人化法律建议,商标近似判断以各地官方检索与专业代理意见为准。 第一章用编辑距离、音骨架与视觉距离三把尺子说明通道决定哪把尺子更好。第二章按美国国家档案馆指南实现 Soundex 并逐例核对,在合成词表上同码名字对中有 0.436 在读音上相差过半,保留读音的改写有 0.093 改变了编码,且全部来自首字母。第三章按声母、韵母、声调计算汉语音节距离并处理跨语言谐音。第四章用部件表量汉字形近,并说明量距离前必须先做 Unicode 归一。第五章加入义近,说明加权平均会稀释单维撞车,示意七对中有 5 对只有取最大值才能标出。第六章用 ROC 曲线与成本比选阈值。 第七章考辨陆法言《切韵·序》与 Russell 1918 年 Soundex 前身专利,并用程序以今音拼读古代反切。第八章给出新品牌命名的七步预检流程、蒸馏小工具「品牌撞名预检器」与预注册检验方案,主要检验约需 1820 对名字,并写明什么结果会推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Far in Spelling, Near in Sound: Edit Distance, Sound Skeletons and the Channels of Confusion — Look first at the top row, where the sound channel has a share of zero and every confusion comes from text. The visual ruler reaches an AUC of 0.906, plain letter edit distance 0.880, and sound code only 0.683. Now the bottom row, where every confusion comes from sound: the sound code rises to 0.893, letter distance falls to 0.789 and the visual ruler to 0.767. The lines cross in between. At a spoken share of 0.5, sound code stands at 0.789 against 0.825 for letters and 0.826 for the visual ruler, so the shape side still leads slightly; at a share of 0.8, sound code at 0.851 overtakes letter distance at 0.817. (https://5dlibrary.com/answers/en/trademark-phonetic-graphic-semantic-distance-brand-collision/c01/) 第1章 字形远、读音近:编辑距离、音骨架与混淆的通道 — 先看最上面一行:声音通道占比为零,混淆全部来自文字,这时视觉尺子的 AUC 是 0.906,普通字母编辑距离是 0.880,音码只有 0.683。再看最下面一行:混淆全部来自声音,音码的 AUC 升到 0.893,字母编辑距离降到 0.789,视觉尺子降到 0.767。两条线在中间交叉:声音占比 0.5 时,音码为 0.789,字母距离为 0.825,视觉为 0.826,字形一侧仍略占上风;声音占比 0.8 时,音码以 0.851 超过字母距离的 0.817。 (https://5dlibrary.com/answers/zh/trademark-phonetic-graphic-semantic-distance-brand-collision/c01/) - Ch.2 Soundex: a Century-Old Sound Code, Too Wide and Too Narrow — Chapter 2 implements Soundex from the U.S. National Archives guide and checks it example by example; on a synthetic lexicon 0.436 of same-code pairs differ by more than half in sound, and 0.093 of sound-keeping respellings change the code, all through the initial letter. (https://5dlibrary.com/answers/en/trademark-phonetic-graphic-semantic-distance-brand-collision/c02/) 第2章 Soundex:一百年前的音码,太宽又太窄 — 第二章按美国国家档案馆指南实现 Soundex 并逐例核对,在合成词表上同码名字对中有 0.436 在读音上相差过半,保留读音的改写有 0.093 改变了编码,且全部来自首字母。 (https://5dlibrary.com/answers/zh/trademark-phonetic-graphic-semantic-distance-brand-collision/c02/) - Ch.3 Initial, Final and Tone: a Distance for Chinese Syllables and Homophones across Languages — qing2-yue4 and qing4-yue4, each differing by one tone, are 0.075 from the base; without tone they are 0.000, identical. qin1-yue4, which differs by a nasal ending, is 0.100. With an initial that is not in the merger table, jing1-yue4 and xing1-yue4 both stand at 0.225. Adding a syllable gives qing1-yue4-xing1 a distance of 0.333, because the insertion costs 1 and is divided by three. qing1-shuo1, with its second syllable replaced, is at 0.500, and ming2-le4, with both replaced, at 0.725. Taking 0.25 as the line, 5 of the eight names lie close to the base. (https://5dlibrary.com/answers/en/trademark-phonetic-graphic-semantic-distance-brand-collision/c03/) 第3章 声母、韵母、声调:汉语音节距离与跨语言谐音 — 只差一个声调的 qing2-yue4 与 qing4-yue4,名字距离都是 0.075;不看声调时是 0.000,完全相同。前后鼻音之差的 qin1-yue4 是 0.100。换了一个不在混读表里的声母,jing1-yue4 与 xing1-yue4 都是 0.225。在基准名后面加一个音节,qing1-yue4-xing1 的距离是 0.333,因为插入一个音节的代价为 1,再除以三。第二个音节完全换掉的 qing1-shuo1 是 0.500,两个音节都换掉的 ming2-le4 是 0.725。按 0.25 为界,八个名字中有 5 个离基准名很近。 (https://5dlibrary.com/answers/zh/trademark-phonetic-graphic-semantic-distance-brand-collision/c03/) - Ch.4 Component Distance: a Table of Look-Alike Characters and Unicode Normalisation — The three pairs that are near in sound only deserve a namer's attention. N1 and N3, and N2 and N3, are 1.000 apart in shape, with not a single shared component, yet only 0.075 apart in sound. N4 and N8 are 0.667 apart in shape and 0.300 in sound. A shape-only check would wave all three through. (https://5dlibrary.com/answers/en/trademark-phonetic-graphic-semantic-distance-brand-collision/c04/) 第4章 部件距离:形近字表与 Unicode 归一 — 「只在读音上近」的三对最值得命名者注意。清悦与庆月、晴悦与庆月,字形距离都是 1.000,没有一个共同部件,读音距离却只有 0.075。请说与清脱字形距离 0.667,读音距离 0.300。这三对如果只做字形检查,全部会被放行。 (https://5dlibrary.com/answers/zh/trademark-phonetic-graphic-semantic-distance-brand-collision/c04/) - Ch.5 Meaning and a Three-Way Score: How a Weighted Mean Dilutes a One-Dimension Collision — Chapter 5 adds meaning and shows that a weighted mean dilutes a collision in one dimension; in seven illustrative pairs, 5 are caught only by the maximum. (https://5dlibrary.com/answers/en/trademark-phonetic-graphic-semantic-distance-brand-collision/c05/) 第5章 义近与三维打分:加权平均会稀释单维撞车 — 第五章加入义近,说明加权平均会稀释单维撞车,示意七对中有 5 对只有取最大值才能标出。 (https://5dlibrary.com/answers/zh/trademark-phonetic-graphic-semantic-distance-brand-collision/c05/) - Ch.6 ROC Curves and Thresholds: Making the Pre-Screen a Calibrated Classifier — Now consider the two ways of combining. Their AUCs are almost equal; the bootstrap interval runs from -0.016 to 0.005 and contains zero. As a ranking, the two cannot be told apart. The difference appears at fixed thresholds. With a threshold of 0.6 for both, the weighted mean has a recall of only 0.295 against 0.797 for the maximum; the price is the false-alarm rate, 0.021 against 0.268. With a threshold of 0.8 for both, the weighted mean catches only 0.152 of the confusions and the maximum 0.531. (https://5dlibrary.com/answers/en/trademark-phonetic-graphic-semantic-distance-brand-collision/c06/) 第6章 ROC 与阈值:把预检做成可校准的判别 — 再看两种合成方式。它们的 AUC 几乎一样,自助区间从 -0.016 到 0.005,包含零。换句话说,就整体排序而言,二者没有可分辨的差别。差别出现在固定阈值上。阈值都取 0.6 时,加权平均的查全率只有 0.295,取最大值是 0.797;代价是误报率,前者 0.021,后者 0.268。阈值都取 0.8 时,加权平均只抓住 0.152,取最大值抓住 0.531。 (https://5dlibrary.com/answers/zh/trademark-phonetic-graphic-semantic-distance-brand-collision/c06/) - Ch.7 East and West in the Sources: the Preface to the Qieyun and Russell's Patent of 1918 — Chapter 7 examines Lu Fayan's preface to the Qieyun and Russell's 1918 patent that preceded Soundex, with a program that reads ancient fanqie spellings in modern sounds. (https://5dlibrary.com/answers/en/trademark-phonetic-graphic-semantic-distance-brand-collision/c07/) 第7章 东西方原典考辨:《切韵·序》与 Russell 1918 年专利 — 第七章考辨陆法言《切韵·序》与 Russell 1918 年 Soundex 前身专利,并用程序以今音拼读古代反切。 (https://5dlibrary.com/answers/zh/trademark-phonetic-graphic-semantic-distance-brand-collision/c07/) - Ch.8 A Pre-Screening Loop for New Brand Names, and a Plan to Refute the Book — Chapter 8 gives a seven-step pre-screening procedure for new brand names, the distilled Brand Name Collision Pre-Screener and a pre-registered test whose primary comparison needs about 1820 name pairs, stating which results would overturn the central claim. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/trademark-phonetic-graphic-semantic-distance-brand-collision/c08/) 第8章 新品牌命名预检闭环与证伪方案 — 第八章给出新品牌命名的七步预检流程、蒸馏小工具「品牌撞名预检器」与预注册检验方案,主要检验约需 1820 对名字,并写明什么结果会推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/trademark-phonetic-graphic-semantic-distance-brand-collision/c08/) ## 5DL-B-318 · The Conservation Algebra of Four-Pillar Settlement and Double-Entry Bookkeeping: Invariants for Error-Free Reconciliation of Buying-Agent Ledgers 中文书名:四柱结算与复式记账的守恒代数、代购流水零差错对账不变量母典 DOI: pending Cite: Fan, Y. (2026). The Conservation Algebra of Four-Pillar Settlement and Double-Entry Bookkeeping: Invariants for Error-Free Reconciliation of Buying-Agent Ledgers (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: With runnable programs this book shows that a single-entry list contains no test of itself: on an illustrative day, leave out a balance payment of 1,500.00 yuan and the spreadsheet stays perfectly consistent; a gap appears only against the bank balance. Parameters are illustrative; all calculations are generated by the book's programs; the book is general analysis, not individualised accounting, tax or legal advice. Abstract: Owners of buying-agent businesses often believe that if every item is written down and the column is added each night, the books cannot be wrong. With runnable programs this book shows that a single-entry list contains no test of itself: on an illustrative day, leave out a balance payment of 1,500.00 yuan and the spreadsheet stays perfectly consistent; a gap appears only against the bank balance. On checking, the blueprint's proposition is corrected in two places: when a whole entry is left out, debits and credits fall together and the double-entry trial balance is just as blind, so omissions are found only by an independent count; and the locating power of double entry stops at a pair of accounts, errors between the same pair forming the null space of the incidence matrix. Parameters are illustrative; all calculations are generated by the book's programs; the book is general analysis, not individualised accounting, tax or legal advice. Chapters 1 and 2 enumerate nine types of error on the illustrative day: the trial balance finds 12.9 %, four pillars on counted cash find 49.1 %, and the full double-entry close, comparing each account with outside evidence, finds 87.6 %. Chapter 3 uses the incidence matrix to compute the blind spots of balance checks and the search window set by the closing interval. Chapter 4 shows why floating point must not keep books and makes conservation exact with an exchange-difference account and largest-remainder sharing. Chapter 5 designs nine accounts, seven daily invariants and an order subledger for a buying agent. Chapter 6 computes weekly hand-sampling sizes from the hypergeometric law. Chapter 7 examines the Song four-pillar settlement and the double-entry treatise Pacioli printed in 1494, testing with programs what the four-pillar chain and the cross-checks among three books can each find. Chapter 8 gives the owner's five-step daily close, the distilled Four-Pillar Double-Entry Reconciler and a pre-registered planted-error trial, stating which results would overturn the central claim. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:做代购的店主常以为,每笔都记、晚上加一加,账就不会错。本书用可运行的程序说明,单式流水本身不含任何检验:在一个示意日上漏记一笔 1,500.00 元的尾款,表格依旧自洽,只有拿银行余额来比才出现差额。立项蓝图的命题经核对作两处修正:整笔漏记时借贷同减,复式的试算平衡同样看不见,发现漏记靠的是独立盘点;复式的定位能力止于账户对,同一对账户之间的错误构成关联矩阵的零空间。书中参数为示意;全部计算由书中程序生成;本书属于非个人化的会计、税务或法律建议。 第一、二章在示意日上穷举九类错账,试算平衡只发现 12.9 %,只盘现金的四柱发现 49.1 %,逐账户对外部证据的完整复式对账发现 87.6 %。第三章用关联矩阵算出余额检验的盲区与结账频率的搜寻窗口。第四章说明浮点数为何不能记账,并用汇差科目与最大余数法让守恒精确成立。第五章为代购店设计九个科目、七条每日不变量与订单子账。第六章用超几何分布计算每周人工抽查的样本量。 第七章考辨宋代四柱结算法与 Pacioli 1494 年刊印的复式记账论述,用程序检验四柱链条与三本账之间的核对各能发现什么。第八章给出店主每日收工五步、蒸馏小工具「四柱复式对账器」与预注册的植入错账试验,写明什么结果会推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 What a Single-Entry List Cannot See: Conservation Tests Only Against a Count — Chapters 1 and 2 enumerate nine types of error on the illustrative day: the trial balance finds 12.9 %, four pillars on counted cash find 49.1 %, and the full double-entry close, comparing each account with outside evidence, finds 87.6 %. (https://5dlibrary.com/answers/en/four-pillar-settlement-double-entry-reconciliation-invariants/c01/) 第1章 单式流水看不见漏记:守恒只有对着盘点才是检验 — 第一、二章在示意日上穷举九类错账,试算平衡只发现 12.9 %,只盘现金的四柱发现 49.1 %,逐账户对外部证据的完整复式对账发现 87.6 %。 (https://5dlibrary.com/answers/zh/four-pillar-settlement-double-entry-reconciliation-invariants/c01/) - Ch.2 The Detection Matrix of Nine Error Types: What the Trial Balance Catches and What It Lets Through — The column headed cases gives the number of forms each type can take on this day. Each of the twelve entries can be left out once, duplicated once or reversed once. Misposting moves each leg of each entry to each account not already in that entry, giving 172 forms; compensating errors take one leg from each of two different entries, add ten yuan to one and subtract ten from the other, giving 286 forms. Altogether there are 581 forms. (https://5dlibrary.com/answers/en/four-pillar-settlement-double-entry-reconciliation-invariants/c02/) 第2章 九类错账的检出矩阵:试算平衡抓得住什么、放得过什么 — 表中「cases」一列是每类错误在这一天能生成的全部形态数。十二笔分录可以各漏一次、各重复一次、各颠倒一次;串户要把每一条借贷腿改到每一个不在该分录里的账户上,共 172 种;互相抵消要从两笔不同分录里各挑一条腿,一条加十元、一条减十元,共 286 种。九类合计 581 种。 (https://5dlibrary.com/answers/zh/four-pillar-settlement-double-entry-reconciliation-invariants/c02/) - Ch.3 The Algebra of Location: Incidence Matrix, Null Space and the Closing Window — The program prints a basis of the null space; each line is an error that leaves every balance unchanged. In the first line T01 is understated by 45000 fen while T02 is overstated by 45000 fen. Both entries debit cash and credit customer deposits; one too small and one too large, they leave both accounts where they were. In the second line T01 and T09 are each overstated by 20000 fen; T01 receives a deposit and T09 refunds one, so overstating both by the same amount cancels in cash, in and out, and in deposits, credit and debit. The remaining three lines shift amounts between the two purchases, the two freight bills and two of the deposits. (https://5dlibrary.com/answers/en/four-pillar-settlement-double-entry-reconciliation-invariants/c03/) 第3章 定位的代数:关联矩阵、零空间与结账窗口 — 程序列出了零空间的一组基,每一行都是一种「余额完全不变的错误」。第一行:T01 少记 45000 分,同时 T02 多记 45000 分。这两笔都是「现金借、预收订金贷」,一笔少一笔多,现金与预收订金两个账户的余额都不变。第二行:T01 与 T09 各多记 20000 分;T01 是收订金,T09 是退订金,同时多记同样的数,现金一进一出抵消,预收订金也一贷一借抵消。后面三行分别是两笔采购之间、两笔运费之间、两笔订金之间的互相挪移。 (https://5dlibrary.com/answers/zh/four-pillar-settlement-double-entry-reconciliation-invariants/c03/) - Ch.4 Exact Arithmetic: Integer Fen, an Exchange-Difference Account and Largest-Remainder Sharing — The fourth and fifth lines show a case a buying agent will meet in earnest. An order is priced at 750.40 yuan; the customer pays a deposit of 300.10 yuan and then a balance of 450.30 yuan. Testing in floating point whether deposit plus balance equals price gives False, because the float total is 750.4000000000001. In integer fen, 30010 plus 45030 equals 75040 and the test passes. (https://5dlibrary.com/answers/en/four-pillar-settlement-double-entry-reconciliation-invariants/c04/) 第4章 精确算术:整数分、汇差科目与最大余数分摊 — 第四、五行是一个代购店会真实遇到的情形:一张订单售价 750.40 元,客户先付订金 300.10 元,再付尾款 450.30 元。用浮点数检查「订金加尾款等于售价」,结果是 False,因为浮点合计是 750.4000000000001。改用整数分,30010 加 45030 等于 75040,检查通过。 (https://5dlibrary.com/answers/zh/four-pillar-settlement-double-entry-reconciliation-invariants/c04/) - Ch.5 Accounts for a Buying Agent: Four Pillars for Every Account and Seven Daily Invariants — The table lays out the illustrative day of Chapter 1 account by account. Every row is a set of four pillars: old balance, debits (new receipts), credits (issues) and present balance. For cash the old balance is 15,000.00 yuan, debits 2,850.00 yuan, credits 3,116.00 yuan and the present balance 14,734.00 yuan, exactly the cash pillars of Chapter 1. Stock starts at 4,200.00 yuan, takes in purchases of 2,748.60 yuan, sends out 1,700.00 yuan on delivery and ends at 5,248.60 yuan. (https://5dlibrary.com/answers/en/four-pillar-settlement-double-entry-reconciliation-invariants/c05/) 第5章 代购流水的科目设计:每个科目一套四柱,七条每日不变量 — 这张表把第一章的示意日按科目展开。每一行都是一套四柱:旧管、借方(新收)、贷方(开除)、实在。以现金为例,旧管 15,000.00 元,借方 2,850.00 元,贷方 3,116.00 元,实在 14,734.00 元,与第一章现金四柱的结果完全相同。存货旧管 4,200.00 元,采购借入 2,748.60 元,交付转出 1,700.00 元,实在 5,248.60 元。 (https://5dlibrary.com/answers/zh/four-pillar-settlement-double-entry-reconciliation-invariants/c05/) - Ch.6 How Many Orders to Re-check Each Week: Discovery Sampling and Value Strata — The program takes a week of 400 orders. If one order in ten is bad, 40 orders, a sample of 22 gives ninety per cent assurance of meeting at least one, and 28 give ninety-five per cent. If five in a hundred are bad, 20 orders, ninety-five per cent needs 55. With two in a hundred bad, 8 orders, it needs 124; with one in a hundred, 4 orders, it needs 211, more than half the week. (https://5dlibrary.com/answers/en/four-pillar-settlement-double-entry-reconciliation-invariants/c06/) 第6章 每周抽查多少单:发现抽样与价值分层 — 程序取一周 400 张订单。如果错单占一成,即 40 张,要有九成把握碰到至少一张,只需抽 22 张;要有九成五把握,抽 28 张。如果错单占百分之五,即 20 张,九成五把握需要 55 张。错单占百分之二,即 8 张,需要 124 张;错单占百分之一,即 4 张,需要 211 张,已经超过总量的一半。 (https://5dlibrary.com/answers/zh/four-pillar-settlement-double-entry-reconciliation-invariants/c06/) - Ch.7 Eastern and Western Sources: the Song Four-Pillar Settlement and Pacioli's Three Books — Chapter 7 examines the Song four-pillar settlement and the double-entry treatise Pacioli printed in 1494, testing with programs what the four-pillar chain and the cross-checks among three books can each find. (https://5dlibrary.com/answers/en/four-pillar-settlement-double-entry-reconciliation-invariants/c07/) 第7章 东西方原典考辨:宋代四柱结算与 Pacioli 的三本账 — 第七章考辨宋代四柱结算法与 Pacioli 1494 年刊印的复式记账论述,用程序检验四柱链条与三本账之间的核对各能发现什么。 (https://5dlibrary.com/answers/zh/four-pillar-settlement-double-entry-reconciliation-invariants/c07/) - Ch.8 Closing the Loop: the Daily Close, the Four-Pillar Double-Entry Reconciler and a Pre-registered Test — Chapter 8 gives the owner's five-step daily close, the distilled Four-Pillar Double-Entry Reconciler and a pre-registered planted-error trial, stating which results would overturn the central claim. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/four-pillar-settlement-double-entry-reconciliation-invariants/c08/) 第8章 实证闭环:每日收工对账、四柱复式对账器与预注册检验 — 第八章给出店主每日收工五步、蒸馏小工具「四柱复式对账器」与预注册的植入错账试验,写明什么结果会推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/four-pillar-settlement-double-entry-reconciliation-invariants/c08/) ## 5DL-B-319 · Bunching at Tax Thresholds and the Distribution of Turnover: Behavioural Elasticities Revealed by Piling-Up below the Line 中文书名:起征点聚束估计与营业额分布、门槛下方堆积揭示的行为弹性母典 DOI: pending Cite: Fan, Y. (2026). Bunching at Tax Thresholds and the Distribution of Turnover: Behavioural Elasticities Revealed by Piling-Up below the Line (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and the threshold is always written T; all calculations are generated by the book's programs. The book does statistical analysis and compliance checking only, teaches no way of staying under a threshold, and is general analysis, not individualised tax, legal or financial advice. Abstract: Abnormal piling-up of turnover below a registration or exemption threshold is often called proof of evasion, and just as often taken to show that owners ignore the levy. This book treats it as a measurable statistic: a polynomial fitted outside a window gives the counterfactual distribution, the excess mass is read off, and it is converted into an elasticity of declared turnover with respect to the net-of-tax share. Before writing, the book checked the original papers and corrects the blueprint: Saez (2010) studies income tax kinks and finds clear bunching only among the self-employed at the first kink of the earned income tax credit; Kleven and Waseem (2013) find that bunching at notches is large but frictions are large too and structural elasticities small; the evidence on turnover thresholds comes from studies of United Kingdom VAT registration and Spanish large-taxpayer monitoring; Blomquist and co-authors (2021) prove that bunching does not identify the elasticity without further assumptions. Parameters are illustrative and the threshold is always written T; all calculations are generated by the book's programs. The book does statistical analysis and compliance checking only, teaches no way of staying under a threshold, and is general analysis, not individualised tax, legal or financial advice. Chapter 1 estimates an excess mass of 942.0 firms and an observed elasticity of 0.1378 on an illustrative distribution from a fixed seed, against a structural elasticity of 0.25, the gap coming from frictions and the approximate counterfactual. Chapter 2 turns to notches: with the same elasticity the marginal buncher lies 20.731 units above the threshold, the bin-by-bin estimate is 19.064 and the estimated friction share is 0.274. Chapter 3 examines the polynomial degree, the integration constraint and the bin width. Chapter 4 gives residual bootstrap intervals and sensitivity to the excluded window. Chapter 5 discusses round-number heaping, the limits of identification and the use of third-party traces to separate real from reporting responses. Chapter 6 treats a registration threshold as a notch that varies by trade and simulates how it holds back growth. Chapter 7 examines the Mencius line that all the ancient levies were one tenth in substance and Smith's four maxims of taxation, with a program showing that at equal revenue the excess burden of a notch is 2.40 times that of a kink. Chapter 8 gives a compliance self-check of multi-shop turnover, the distilled Threshold Bunching Estimator and a pre-registered test: when abnormal piling-up is found, an accountant checks the real declarations, and no declaration is adjusted. The book contains 22 runnable Python programs. 摘要:营业额在登记或免税门槛下方异常堆积,常被说成逃税的铁证,也常被说成经营者对税负毫不在乎。本书把这种堆积当作一个可以测量的统计量:用窗口外的多项式拟合反事实分布,读出超额质量,再换算成申报营业额对税后留存比例的弹性。落笔前本书核对原文并作蓝图修正:Saez 2010 年研究的是所得税拐点,清楚的聚束只见于劳动所得抵免第一拐点的自雇者;Kleven 与 Waseem 2013 年发现跳跃下聚束大而摩擦也大、结构弹性很小;营业额门槛的证据来自英国增值税登记与西班牙大型纳税人监管两项研究;Blomquist 等 2021 年证明聚束不加假设识别不出弹性。书中参数为示意,门槛一律记作 T;全部计算由书中程序生成。本书只做统计分析与合规自查,不教任何规避门槛的手法,属于非个人化税务、法律或理财建议。 第一章在一份固定种子的示意分布上估出超额质量 942.0 家、观察弹性 0.1378,结构弹性为 0.25,差距来自摩擦与反事实近似。第二章转向跳跃:同一弹性下边际聚束者离门槛 20.731 个单位,逐格积分的估计为 19.064,摩擦份额估计 0.274。第三章拆解多项式阶数、积分约束与分格宽度。第四章给出残差自助法区间与排除窗宽敏感度。第五章讨论整数堆积、识别极限,以及用第三方痕迹区分真实反应与申报反应。第六章把登记门槛看成因行业而异的跳跃,并模拟门槛对成长的拖累。 第七章考辨《孟子·滕文公上》「其实皆什一也」与斯密赋税四原则,并用程序算出在同等税收下跳跃的超额负担是拐点的 2.40 倍。第八章给出多店营业额合规自查流程、蒸馏小工具「门槛聚束估计器」与预注册检验方案:一旦发现异常堆积,交由会计师核对真实申报,而不是调整申报。书中含 22 段可运行的 Python 程序。 Chapter answers: - Ch.1 The Spike at a Kink: Excess Mass, Polynomial Counterfactuals and the Elasticity — Chapter 1 estimates an excess mass of 942.0 firms and an observed elasticity of 0.1378 on an illustrative distribution from a fixed seed, against a structural elasticity of 0.25, the gap coming from frictions and the approximate counterfactual. (https://5dlibrary.com/answers/en/tax-threshold-bunching-turnover-elasticity/c01/) 第1章 拐点上的尖峰:超额质量、多项式反事实与弹性换算 — 第一章在一份固定种子的示意分布上估出超额质量 942.0 家、观察弹性 0.1378,结构弹性为 0.25,差距来自摩擦与反事实近似。 (https://5dlibrary.com/answers/zh/tax-threshold-bunching-turnover-elasticity/c01/) - Ch.2 Notches and Dominated Ranges: Optimization Frictions and the Marginal Buncher — Chapter 2 turns to notches: with the same elasticity the marginal buncher lies 20.731 units above the threshold, the bin-by-bin estimate is 19.064 and the estimated friction share is 0.274. (https://5dlibrary.com/answers/en/tax-threshold-bunching-turnover-elasticity/c02/) 第2章 跳跃与受支配区间:优化摩擦与边际聚束者 — 第二章转向跳跃:同一弹性下边际聚束者离门槛 20.731 个单位,逐格积分的估计为 19.064,摩擦份额估计 0.274。 (https://5dlibrary.com/answers/zh/tax-threshold-bunching-turnover-elasticity/c02/) - Ch.3 Three Components of the Counterfactual: Polynomial Degree, Integration Constraint and Bin Width — The results are very stable. Measured in bins, b varies inversely with the width, from 6.228 to 0.785, but once converted into a distance in units, the four widths give dz between 3.114 and 3.138 and elasticities between 0.1374 and 0.1385. The excess mass itself barely moves, between 940.0 and 945.3. (https://5dlibrary.com/answers/en/tax-threshold-bunching-turnover-elasticity/c03/) 第3章 反事实的三个零件:多项式阶数、积分约束与分格宽度 — 结果非常稳定。以格数计的 b 随格宽成反比变化,从 6.228 到 0.785,但换算成以单位计的距离之后,四种格宽给出的 dz 在 3.114 与 3.138 之间,弹性在 0.1374 与 0.1385 之间。超额质量本身也几乎不动,在 940.0 与 945.3 之间。 (https://5dlibrary.com/answers/zh/tax-threshold-bunching-turnover-elasticity/c03/) - Ch.4 Is the Spike Noise? Bootstrap Intervals, the Excluded Window and Power — The point estimate is still a b of 3.123 bins and an elasticity of 0.1378. The standard deviation of the two hundred bootstrap re-estimates is 0.205 and the t statistic is 15.21, so the spike is far from noise. The 95 % interval for b runs from 2.534 to 3.279, and for the elasticity from 0.1121 to 0.1446. (https://5dlibrary.com/answers/en/tax-threshold-bunching-turnover-elasticity/c04/) 第4章 尖峰是不是噪声:自助法区间、排除窗宽与检验力 — 点估计仍是 b 为 3.123 格、弹性 0.1378。两百次自助法重估的标准差是 0.205,t 值为 15.21,尖峰远非噪声。b 的 95 % 区间从 2.534 到 3.279,弹性的区间从 0.1121 到 0.1446。 (https://5dlibrary.com/answers/zh/tax-threshold-bunching-turnover-elasticity/c04/) - Ch.5 Three Confounds: Round Numbers, the Limits of Identification and Reporting Responses — The baseline ratio is 0.6007. In world R, the bin from 98 to 100 holds 884 firms and a ratio of 0.5991, no different from the baseline; in world M, the same bin holds 843 firms and the ratio rises to 0.6315, 13.0 standard errors above the baseline. A real response lowers sales and card receipts together and leaves the ratio unchanged; a reporting response lowers only the denominator and pushes the ratio up. (https://5dlibrary.com/answers/en/tax-threshold-bunching-turnover-elasticity/c05/) 第5章 三种混淆:整数堆积、识别极限与申报反应 — 基准比值是 0.6007。R 世界里,98 到 100 那一格挤进了 884 家,比值 0.5991,与基准无异;M 世界里,同一格有 843 家,比值升到 0.6315,比基准高出 13.0 个标准误。真实反应让销售与刷卡一起下降,比值不变;申报反应只压低了分母,比值就抬了起来。 (https://5dlibrary.com/answers/zh/tax-threshold-bunching-turnover-elasticity/c05/) - Ch.6 Registration Thresholds: Notches That Vary by Trade and Growth Held Back — In the first year, 0.103 of owners have crossed T in the world without a threshold and only 0.040 in the world with one. The gap peaks in year six at 0.147 and then narrows slowly, as the potential turnover of more and more owners grows past T plus D. By year twelve the shares above the line are 0.673 and 0.556, still 0.117 apart. Of the ten thousand owners, 5129 stay below the threshold for at least one year, for an average of 2.93 years. The last line gives a deterministic benchmark: with steady growth of 4 % a year, crossing the bunching range takes 4.71 years. (https://5dlibrary.com/answers/en/tax-threshold-bunching-turnover-elasticity/c06/) 第6章 登记门槛:因行业而异的跳跃与被拖住的成长 — 第一年,无门槛世界里有 0.103 的经营者越过 T,有门槛世界里只有 0.040。差距在第六年达到最大,0.147,此后缓慢收窄,因为越来越多经营者的潜在营业额长过了 T 加 D 的范围。到第十二年,两个世界越线的比例分别是 0.673 与 0.556,差距仍有 0.117。一万家经营者中,有 5129 家至少有一年停在门槛下方,平均停了 2.93 年。最后一行给出一个确定性的参照:若每年稳定增长 4 %,穿过这段聚束区间需要 4.71 年。 (https://5dlibrary.com/answers/zh/tax-threshold-bunching-turnover-elasticity/c06/) - Ch.7 Classical Sources East and West: "One Tenth in Substance" and Smith's Four Maxims — Chapter 7 examines the Mencius line that all the ancient levies were one tenth in substance and Smith's four maxims of taxation, with a program showing that at equal revenue the excess burden of a notch is 2.40 times that of a kink. (https://5dlibrary.com/answers/en/tax-threshold-bunching-turnover-elasticity/c07/) 第7章 东西方原典考辨:「其实皆什一也」与斯密的赋税四原则 — 第七章考辨《孟子·滕文公上》「其实皆什一也」与斯密赋税四原则,并用程序算出在同等税收下跳跃的超额负担是拐点的 2.40 倍。 (https://5dlibrary.com/answers/zh/tax-threshold-bunching-turnover-elasticity/c07/) - Ch.8 Closing the Loop: a Multi-Shop Compliance Self-Check, the Threshold Bunching Estimator and a Pre-Registered Test — Chapter 8 gives a compliance self-check of multi-shop turnover, the distilled Threshold Bunching Estimator and a pre-registered test: when abnormal piling-up is found, an accountant checks the real declarations, and no declaration is adjusted. The book contains 22 runnable Python programs. (https://5dlibrary.com/answers/en/tax-threshold-bunching-turnover-elasticity/c08/) 第8章 实证闭环:多店营业额合规自查、门槛聚束估计器与预注册检验 — 第八章给出多店营业额合规自查流程、蒸馏小工具「门槛聚束估计器」与预注册检验方案:一旦发现异常堆积,交由会计师核对真实申报,而不是调整申报。书中含 22 段可运行的 Python 程序。 (https://5dlibrary.com/answers/zh/tax-threshold-bunching-turnover-elasticity/c08/) ## 5DL-B-320 · The Nash Bargaining Solution and Co-Branding Revenue Splits: Threat Points from Outside Options and Allocation under Risk Attitudes 中文书名:纳什议价解与联名分润、外部选项威胁点与风险态度的分配母典 DOI: pending Cite: Fan, Y. (2026). The Nash Bargaining Solution and Co-Branding Revenue Splits: Threat Points from Outside Options and Allocation under Risk Attitudes (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Under one set of illustrative parameters, with outside options of 10 for the creator and 30 for the brand, fifty-fifty keeps only 0.8889 of the maximum Nash product. Parameters are illustrative; all calculations are generated by the book's programs; the book is general analysis, not individualised business or legal advice. Abstract: When brands and creators negotiate a co-branding split, the most common answer is fifty-fifty. Using the axiomatic solution of Nash's 1950 paper "The Bargaining Problem", this book overturns the belief that fifty-fifty is the fair point: the Nash solution first returns to each side what it would get if talks failed, then divides the surplus of the collaboration, and it is shaped by the concavity of utility, that is, by attitudes to risk. Under one set of illustrative parameters, with outside options of 10 for the creator and 30 for the brand, fifty-fifty keeps only 0.8889 of the maximum Nash product. The book corrects its blueprint: Nash's 1950 paper takes the disagreement point as given, and threat strategies appear in his 1953 paper; whether an outside option is the threat point depends on how talks can end, and work by Binmore and co-authors in 1986 and 1989 shows that outside options are the disagreement point when talks risk breaking down, while under the freedom to walk away only options worth more than bargaining itself count; risk aversion is a disadvantage when riskless outcomes are divided, but Roth and Rothblum proved in 1982 that this need not hold for risky agreements. Parameters are illustrative; all calculations are generated by the book's programs; the book is general analysis, not individualised business or legal advice. Chapters 1 and 2 compute the symmetric Nash solution and compare three readings of the outside option. Chapter 3 uses power utility to show that the more risk-averse side gives up share, and computes that on a two-point risky pool a guarantee raises the creator's certainty equivalent by 2.8433 and the brand's expectation by 0.8384. Chapter 4 uses Rubinstein's 1982 alternating offers to show that patience is bargaining power, converging as offers come faster to the asymmetric Nash solution with weight 0.2000. Chapter 5 sets the 1975 solution of Kalai and Smorodinsky beside Nash's. Chapter 6 weighs the evidence that equal splits are common in ultimatum experiments and explains when fifty-fifty is a convenient focal point. Chapter 7 examines Bai Gui's rule "take what others discard, give out what others take" in the Records of the Grand Historian and the contract curve of Edgeworth's Mathematical Psychics of 1881. Chapter 8 gives a negotiation table for co-branding and creator splits, the distilled Co-Branding Split Bargainer and a pre-registered test needing 122 deals per group, stating which results would overturn the central claim; its negotiating advice asks only for the disclosure of true outside options and teaches no overstatement or deceptive threats. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:品牌与达人谈联名分润,最常见的答案是五五开。本书用纳什 1950 年《议价问题》的公理化解推翻「五五开就是公平点」:纳什解先让双方各自拿回谈不成时的所得,再分合作剩余,并受效用凹度即风险态度影响。在一组示意参数下,达人外部选项 10、品牌 30 时,五五开只保住最大纳什积的 0.8889。本书作蓝图修正:纳什 1950 年把分歧点当作给定,威胁策略见于 1953 年论文;外部选项是否就是威胁点取决于谈判破裂的机制,Binmore 等人 1986 年与 1989 年的研究表明,有破裂风险时外部选项是分歧点,可随时离场时只有超过谈判所得的外部选项才起作用;风险厌恶在无风险分配中不利,但 Roth 与 Rothblum 1982 年证明在有风险的协议中不一定。书中参数为示意;全部计算由书中程序生成;本书是非个人化经营或法律建议。 第一、二章求对称纳什解并比较外部选项的三种读法。第三章用幂效用说明更厌恶风险的一方让出份额,并算出在两点风险的分润池上,保底合同让达人的确定性等值多 2.8433、品牌期望多 0.8384。第四章用 Rubinstein 1982 年的交替出价说明耐心就是议价力,出价间隔趋零时收敛到权重为 0.2000 的非对称纳什解。第五章以 Kalai 与 Smorodinsky 1975 年的解作对照。第六章并陈最后通牒实验中对半分常见的证据,并说明五五开何时是省事的聚点。 第七章考辨《史记·货殖列传》白圭「人弃我取,人取我与」与 Edgeworth 1881 年《数理心理学》的契约曲线。第八章给出联名与达人分润谈判表、蒸馏小工具「联名分润议价器」与每组 122 单的预注册检验方案,写明什么结果会推翻核心命题;谈判建议只要求披露真实外部选项,不教虚报与欺骗性威胁。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Fifty-Fifty Is Not the Fair Point: Nash's Four Axioms and the Nash Product — The third and fourth rows diverge. When the creator would get ten and the brand thirty without a deal, the Nash split gives the creator 40.00, not fifty; fifty-fifty then yields a Nash product of only 800.00 against a maximum of 900.00, a ratio of 0.8889. Swap the two disagreement payoffs and the Nash split gives the creator sixty. The same pool and the same contributions produce a difference of twenty units purely because one side has a better fallback. (https://5dlibrary.com/answers/en/nash-bargaining-cobranding-revenue-split-threat-points/c01/) 第1章 五五开不是公平点:纳什的四条公理与纳什积 — 第三、四行开始分叉。达人谈不成拿十、品牌谈不成拿三十时,纳什解给达人 40.00,而不是五十;此时五五开的纳什积只有 800.00,最大值是 900.00,比值 0.8889。把两边的分歧所得对调,纳什解就变成给达人六十。同一个分润池,同样的「贡献」,仅仅因为谈不成时谁更有退路,分配就差了二十个单位。 (https://5dlibrary.com/answers/zh/nash-bargaining-cobranding-revenue-split-threat-points/c01/) - Ch.2 Where the Threat Point Comes From: Outside Options, Payoffs during Disagreement and the Risk of Breakdown — With a breakdown probability of one half, the first proposer has a large advantage: proposing first, the creator takes 70.0000, but only 50.0000 when the brand proposes first. At a probability of one tenth the two figures close in to 61.5789 and 58.4211; at one in a thousand, to 60.0150 and 59.9850. The midpoint is exactly 60.0000 in every row, the Nash split with outside options as the disagreement point. (https://5dlibrary.com/answers/en/nash-bargaining-cobranding-revenue-split-threat-points/c02/) 第2章 威胁点从哪里来:外部选项、僵持收益与破裂风险 — 破裂概率为一半时,先出价的一方优势很大:达人先出价拿 70.0000,品牌先出价时达人只拿 50.0000。破裂概率降到十分之一,两个数收拢到 61.5789 与 58.4211;降到千分之一,收拢到 60.0150 与 59.9850。两者的中点在每一行都正好是 60.0000,也就是以外部选项为分歧点的纳什解。 (https://5dlibrary.com/answers/zh/nash-bargaining-cobranding-revenue-split-threat-points/c02/) - Ch.3 Attitudes to Risk: How the Curvature of Utility Hands Money to the Other Side — Chapter 3 uses power utility to show that the more risk-averse side gives up share, and computes that on a two-point risky pool a guarantee raises the creator's certainty equivalent by 2.8433 and the brand's expectation by 0.8384. (https://5dlibrary.com/answers/en/nash-bargaining-cobranding-revenue-split-threat-points/c03/) 第3章 风险态度:效用的凹度怎样把钱让给对方 — 第三章用幂效用说明更厌恶风险的一方让出份额,并算出在两点风险的分润池上,保底合同让达人的确定性等值多 2.8433、品牌期望多 0.8384。 (https://5dlibrary.com/answers/zh/nash-bargaining-cobranding-revenue-split-threat-points/c03/) - Ch.4 The Asymmetric Nash Solution and Alternating Offers: Patience Is Bargaining Power — Chapter 4 uses Rubinstein's 1982 alternating offers to show that patience is bargaining power, converging as offers come faster to the asymmetric Nash solution with weight 0.2000. (https://5dlibrary.com/answers/en/nash-bargaining-cobranding-revenue-split-threat-points/c04/) 第4章 非对称纳什与交替出价:耐心就是议价力 — 第四章用 Rubinstein 1982 年的交替出价说明耐心就是议价力,出价间隔趋零时收敛到权重为 0.2000 的非对称纳什解。 (https://5dlibrary.com/answers/zh/nash-bargaining-cobranding-revenue-split-threat-points/c04/) - Ch.5 The Kalai-Smorodinsky Comparison: Monotonicity against Independence of Irrelevant Alternatives — Chapter 5 sets the 1975 solution of Kalai and Smorodinsky beside Nash's. (https://5dlibrary.com/answers/en/nash-bargaining-cobranding-revenue-split-threat-points/c05/) 第5章 Kalai–Smorodinsky 对照:单调性与无关选项独立 — 第五章以 Kalai 与 Smorodinsky 1975 年的解作对照。 (https://5dlibrary.com/answers/zh/nash-bargaining-cobranding-revenue-split-threat-points/c05/) - Ch.6 Why Fifty-Fifty Is Everywhere: Ultimatums, Focal Points and the Cost of Haggling — At an offer of zero, only responders whose threshold is zero accept, and the proposer expects just 10.00. As the offer rises, the probability of acceptance climbs faster than the proposer's own part falls: at an offer of 30 the acceptance probability is 0.70 and the expectation 49.00; at 40 it is 0.90 and the expectation 54.00, the best in this table. An offer of fifty is certainly accepted, but the expectation falls to 50.00. (https://5dlibrary.com/answers/en/nash-bargaining-cobranding-revenue-split-threat-points/c06/) 第6章 五五开为什么到处都是:最后通牒、聚点与讨价还价的成本 — 出价为零时,只有门槛为零的回应者会接受,提议者期望只拿 10.00。出价逐步提高,接受概率上升得比留给自己的部分下降得快;出价 30 时接受概率为 0.70,期望 49.00;出价 40 时接受概率为 0.90,期望 54.00,是这张表里的最优。出价五十时一定被接受,但期望降到 50.00。 (https://5dlibrary.com/answers/zh/nash-bargaining-cobranding-revenue-split-threat-points/c06/) - Ch.7 Classical Sources East and West: Bai Gui's "Take What Others Discard" and Edgeworth's Contract Curve — Chapter 7 examines Bai Gui's rule "take what others discard, give out what others take" in the Records of the Grand Historian and the contract curve of Edgeworth's Mathematical Psychics of 1881. (https://5dlibrary.com/answers/en/nash-bargaining-cobranding-revenue-split-threat-points/c07/) 第7章 东西方原典考辨:白圭「人弃我取」与 Edgeworth 的契约曲线 — 第七章考辨《史记·货殖列传》白圭「人弃我取,人取我与」与 Edgeworth 1881 年《数理心理学》的契约曲线。 (https://5dlibrary.com/answers/zh/nash-bargaining-cobranding-revenue-split-threat-points/c07/) - Ch.8 The Co-Branding Negotiation Table: an Empirical Loop, the Bargainer and a Pre-registered Test — Chapter 8 gives a negotiation table for co-branding and creator splits, the distilled Co-Branding Split Bargainer and a pre-registered test needing 122 deals per group, stating which results would overturn the central claim; its negotiating advice asks only for the disclosure of true outside options and teaches no overstatement or deceptive threats. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/nash-bargaining-cobranding-revenue-split-threat-points/c08/) 第8章 联名分润谈判表:实证闭环、议价器与预注册检验 — 第八章给出联名与达人分润谈判表、蒸馏小工具「联名分润议价器」与每组 122 单的预注册检验方案,写明什么结果会推翻核心命题;谈判建议只要求披露真实外部选项,不教虚报与欺骗性威胁。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/nash-bargaining-cobranding-revenue-split-threat-points/c08/) ## 5DL-B-321 · Shanxi Merchants' Human-Capital Shares and the Algebra of Early-Stage Dilution: Valuation Caps, Discounts and Founder Ownership 中文书名:晋商身股与早期股权稀释代数、估值上限与折扣取孰优的创始人持股母典 DOI: pending Cite: Fan, Y. (2026). Shanxi Merchants' Human-Capital Shares and the Algebra of Early-Stage Dilution: Valuation Caps, Discounts and Founder Ownership (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: On one illustrative path, compounding the headlines leaves the founders 52.02 %, while compounding the effective dilution leaves only 39.56 %. Parameters are illustrative; all calculations are generated by the book's programs; the book is general analysis, not individualised legal, tax or financial advice. Abstract: Start-up partners often multiply the headline fractions of their financing rounds and believe that this tells them what they still own. Using exact fractions, this book overturns that conclusion: each round issues more new shares than the new money buys, and the option pool enlargement and the conversion of safes dilute existing holders just as surely. On one illustrative path, compounding the headlines leaves the founders 52.02 %, while compounding the effective dilution leaves only 39.56 %. The project blueprint gave no source, and the book corrects it: no peer-reviewed study directly measuring founders' estimation error was found; what can be verified is Y Combinator's 2018 user guide for the post-money safe, which states that the original safe made dilution hard to calculate, and the 2020 study by Gornall and Strebulaev showing that headline valuations overvalue common shares. The book therefore computes the mechanical gap exactly and leaves the estimation error as a pre-registered hypothesis. Parameters are illustrative; all calculations are generated by the book's programs; the book is general analysis, not individualised legal, tax or financial advice. Chapters 1 to 3 show that dilution can only multiply, that placing the pool before the money costs existing holders the pool target times the new money's fraction more than placing it after, and that the switch point between a safe's cap and discount lies at the cap over one minus the discount, 12.50 million in the example. Chapter 4 places the pool, two safes and new money in one exact Series A cap table: the founders' mental estimate is 58.07 %, the exact figure 52.73 %. Chapter 5 uses the liquidation waterfall to show that under an illustrative distribution of sale prices a common share is expected to be worth only 0.7087 of the headline price. Chapter 6 discusses vesting, the cliff and the return of human-capital shares. Chapter 7 examines the human shares of the Shanxi remittance houses recorded in Shanxi Piaohao Shiliao and Smith's treatment of joint-stock companies in The Wealth of Nations, with a program showing how growing human shares dilute capital shares. Chapter 8 gives a partners' cap table, the distilled Dilution Algebra Calculator and a pre-registered test, stating which results would overturn the central claim. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:创业合伙人常把每一轮融资的标题比例相乘,以为这样就知道自己还剩多少。本书用精确分数推翻这个结论:每一轮实际发行的新股不止新钱买走的那一份,期权池扩充与 SAFE 转股同样摊薄现有股东。在一条示意路径上,按标题连乘创始人剩 52.02 %,按实际稀释只剩 39.56 %。立项蓝图没有给出处,本书作蓝图修正:没有找到直接测量创始人估计偏差的同行评议研究,能核实的是 Y Combinator 2018 年融资后 SAFE 使用指南对旧式 SAFE 难以计算稀释的说明,以及 Gornall 与 Strebulaev 2020 年关于标题估值高估普通股的研究;因此本书把机制差距精确算出,把估计偏差留作预注册假说。书中参数为示意;全部计算由书中程序生成;本书是非个人化法律、税务或理财建议。 第一至三章说明稀释只能相乘,期权池放在钱前比放在钱后让现有股东多付期权池目标乘新钱比例,SAFE 的上限与折扣取孰优的切换点在上限除以一减折扣,示例中为 12.50 百万。第四章把期权池、两张 SAFE 与新钱放进一张精确的 A 轮股权表,创始人心算 58.07 %,精确值 52.73 %。第五章用清算瀑布说明,在一组示意出售分布下普通股每股期望价值只有标题价格的 0.7087。第六章讨论人力股的兑现、断崖与回流。 第七章考辨《山西票号史料》所录晋商票号的身股制度与斯密《国富论》论股份公司,并用程序展示身股增长如何摊薄银股。第八章给出合伙人股权表、蒸馏小工具「稀释代数计算器」与预注册检验方案,写明什么结果会推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Dilution Multiplies: the Two Terms a Headline Leaves Out — Chapters 1 to 3 show that dilution can only multiply, that placing the pool before the money costs existing holders the pool target times the new money's fraction more than placing it after, and that the switch point between a safe's cap and discount lies at the cap over one minus the discount, 12.50 million in the example. (https://5dlibrary.com/answers/en/shanxi-human-capital-shares-early-dilution-cap-discount/c01/) 第1章 稀释是乘法:标题百分比漏掉的两项 — 第一至三章说明稀释只能相乘,期权池放在钱前比放在钱后让现有股东多付期权池目标乘新钱比例,SAFE 的上限与折扣取孰优的切换点在上限除以一减折扣,示例中为 12.50 百万。 (https://5dlibrary.com/answers/zh/shanxi-human-capital-shares-early-dilution-cap-discount/c01/) - Ch.2 The Option Pool Before or After the Money: a Gap of p Times f — With the pool before the money the price per share is 0.7778, the investor receives 2,571,429 shares, the new pool has 1,285,714 shares, the existing holders own 70.00 %, the investor 20.00 % and the pool a tenth. With the pool after the money the price is 0.8889, the investor receives 2,250,000 shares, the pool 1,250,000 shares, the existing holders own 72.00 % and the investor 18.00 %. (https://5dlibrary.com/answers/en/shanxi-human-capital-shares-early-dilution-cap-discount/c02/) 第2章 期权池放在钱前还是钱后:一个 p 乘 f 的缺口 — 放在钱前时,每股价格为 0.7778,新投资人得到 2,571,429 股,新设期权池 1,285,714 股,现有股东合计持股 70.00 %,投资人 20.00 %,期权池一成。放在钱后时,每股价格为 0.8889,投资人得到 2,250,000 股,期权池 1,250,000 股,现有股东持股 72.00 %,投资人 18.00 %。 (https://5dlibrary.com/answers/zh/shanxi-human-capital-shares-early-dilution-cap-discount/c02/) - Ch.3 Cap or Discount, Whichever Is Better: the Switch Sits at the Cap over One Minus the Discount — At a round price of 1.50 the cap price of one is lower than the discount price, the cap applies, and the safe receives one million shares; the note converting after twelve months receives 1,060,000 shares and after thirty-six months 1,180,000 shares, 18.00 % more. At a round price of 1.10 the discount price is 0.8800, below the cap price of 1.0000, so the discount applies: the safe receives 1,136,364 shares and the note after thirty-six months 1,340,909. (https://5dlibrary.com/answers/en/shanxi-human-capital-shares-early-dilution-cap-discount/c03/) 第3章 估值上限与折扣取孰优:切换点在上限除以一减折扣 — 定价轮每股价格为 1.50 时,上限价格一元比折扣价格更低,上限起作用,SAFE 得到一百万股;票据在十二个月后转股得到 1,060,000 股,三十六个月后得到 1,180,000 股,多出 18.00 %。定价轮每股价格为 1.10 时,折扣价格为 0.8800,低于上限价格 1.0000,折扣起作用,SAFE 得到 1,136,364 股,票据在三十六个月后得到 1,340,909 股。 (https://5dlibrary.com/answers/zh/shanxi-human-capital-shares-early-dilution-cap-discount/c03/) - Ch.4 One Exact Cap Table: Safes, the Pool and New Money Arrive Together — Chapter 4 places the pool, two safes and new money in one exact Series A cap table: the founders' mental estimate is 58.07 %, the exact figure 52.73 %. (https://5dlibrary.com/answers/en/shanxi-human-capital-shares-early-dilution-cap-discount/c04/) 第4章 一张精确的股权表:SAFE、期权池与新钱同时入场 — 第四章把期权池、两张 SAFE 与新钱放进一张精确的 A 轮股权表,创始人心算 58.07 %,精确值 52.73 %。 (https://5dlibrary.com/answers/zh/shanxi-human-capital-shares-early-dilution-cap-discount/c04/) - Ch.5 The Liquidation Waterfall: What a Common Share Is Worth under One Headline Valuation — Chapter 5 uses the liquidation waterfall to show that under an illustrative distribution of sale prices a common share is expected to be worth only 0.7087 of the headline price. (https://5dlibrary.com/answers/en/shanxi-human-capital-shares-early-dilution-cap-discount/c05/) 第5章 优先清算的瀑布:同一个标题估值下普通股值多少 — 第五章用清算瀑布说明,在一组示意出售分布下普通股每股期望价值只有标题价格的 0.7087。 (https://5dlibrary.com/answers/zh/shanxi-human-capital-shares-early-dilution-cap-discount/c05/) - Ch.6 Vesting, the Cliff and the Return of Human-Capital Shares — The lower part of the output covers partner C's departure. Before it, A, B and C hold 40.00 %, 30.00 % and 20.00 %. If C leaves in month eleven, all two million shares are cancelled, and A rises to 50.00 % and B to 37.50 %. If C leaves in month eighteen and keeps the vested 37.50 %, seven hundred and fifty thousand shares, A holds 45.71 %, B 34.29 % and C 8.57 %. If C leaves in month thirty, A holds 43.24 %, B 32.43 % and C 13.51 %. (https://5dlibrary.com/answers/en/shanxi-human-capital-shares-early-dilution-cap-discount/c06/) 第6章 人力股的兑现、断崖与回流 — 表的下半部分是合伙人 C 离开的情形。离开之前,A、B、C 分别持有 40.00 %、30.00 %、20.00 %。若 C 在第十一个月离开,其两百万股全部注销,A 升到 50.00 %,B 升到 37.50 %。若 C 在第十八个月离开,保留 37.50 % 的兑现部分,即七十五万股,A 为 45.71 %,B 为 34.29 %,C 为 8.57 %。若 C 在第三十个月离开,A 为 43.24 %,B 为 32.43 %,C 为 13.51 %。 (https://5dlibrary.com/answers/zh/shanxi-human-capital-shares-early-dilution-cap-discount/c06/) - Ch.7 Classical Sources East and West: the Human Shares of the Shanxi Merchants and Smith on Joint-Stock Companies — An owner-manager running his own money achieves the maximum, 100.000 %. One of three equal partners, receiving a third, achieves only 55.556 % of the maximum. A remittance-house manager holding one of forty-four shares in the fifth period has a fraction of 2.273 % and achieves only 4.494 %. A director of a joint-stock company holding one share in a thousand has 0.100 % and achieves just 0.200 %. The last two lines compute that for this manager to reach half the maximum surplus he would need thirteen shares, a fraction of 29.545 %. (https://5dlibrary.com/answers/en/shanxi-human-capital-shares-early-dilution-cap-discount/c07/) 第7章 东西方原典考辨:晋商身股与斯密论股份公司 — 一位自己出钱自己经营的东主,总剩余是最优的 100.000 %。三位平等合伙人之一,分得三分之一,总剩余只有最优的 55.556 %。一位在第五个账期持有四十四股中一股的票号经营者,比例为 2.273 %,总剩余只有 4.494 %。一位持有千分之一股份的股份公司董事,比例 0.100 %,总剩余只剩 0.200 %。程序最后两行算出,要让这位票号经营者达到最优剩余的一半,他需要持有十三股,即 29.545 %。 (https://5dlibrary.com/answers/zh/shanxi-human-capital-shares-early-dilution-cap-discount/c07/) - Ch.8 The Partners' Cap Table: an Empirical Loop, a Tool and a Pre-registered Test — Chapter 8 gives a partners' cap table, the distilled Dilution Algebra Calculator and a pre-registered test, stating which results would overturn the central claim. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/shanxi-human-capital-shares-early-dilution-cap-discount/c08/) 第8章 合伙人股权表:实证闭环、小工具与预注册检验 — 第八章给出合伙人股权表、蒸馏小工具「稀释代数计算器」与预注册检验方案,写明什么结果会推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/shanxi-human-capital-shares-early-dilution-cap-discount/c08/) ## 5DL-B-322 · The Ringelmann Effect and Brooks's Law: the Optimal Size of Teams and Agent Squads under Coordination Overhead 中文书名:林格曼效应与布鲁克斯定律、团队与智能体军团规模的协调开销最优边界母典 DOI: pending Cite: Fan, Y. (2026). The Ringelmann Effect and Brooks's Law: the Optimal Size of Teams and Agent Squads under Coordination Overhead (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Chapter 4 takes this bureau's four volumes in parallel per wave as an illustrative scene: capacity peaks at 6.78 windows, four windows already reach 89.6 % of the peak, and the coordinator's review queue holds the ceiling at four. Abstract: "Many hands make light work" is the intuition team owners use most, and now that agents can be doubled at the press of a button it can for the first time be pushed to its extreme and tested. This book uses standard-library programs to overturn its range: when headcount doubles, output per member falls; channels grow as n(n-1)/2, 1,225 of them for fifty people; serial steps do not shrink as people are added. The book corrects its blueprint: "a group of eight gives about half of solo force per member" comes from a summary table with no stated task in Ringelmann's 1913 report (0.49 at eight), and the 1986 article by Kravitz and Martin located and described the original rather than reanalysing it; the rope-pulling dynamometer experiment the report does describe gave groups of seven and fourteen 0.762 and 0.720 of solo force per man, a loss that saturates with group size. Parameters are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Chapter 1 sets the summary table beside the rope data, the pseudo-group experiments of Ingham and colleagues and Steiner's process loss. Chapter 2 computes the threshold at which adding people to a late project stops making it later. Chapter 3 joins Amdahl's serial fraction with the coherence term of the universal scalability law and finds the capacity peak. Chapter 4 takes this bureau's four volumes in parallel per wave as an illustrative scene: capacity peaks at 6.78 windows, four windows already reach 89.6 % of the peak, and the coordinator's review queue holds the ceiling at four. Chapter 5 treats merge conflicts as a birthday problem: one hot file raises the conflict probability of four windows to 0.6875, while private directories bring it to zero. Chapter 6 compares the channels of mesh, star and cells and builds a work table by longest job first. Chapter 7 examines the Shen Fen chapter of the Lüshi Chunqiu, "working in common is slow, because there is room to hide one's strength", and Ringelmann's report of 1913. Chapter 8 gives the empirical loop for agent squad ceilings and work tables, the distilled Optimal Team Size Calculator and a pre-registered test: a ratio per member below 0.781 rejects the hypothesis of no loss. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:「人多力量大」是团队主事者最常用的直觉,智能体可以按一下按钮就翻倍之后,这条直觉第一次能被推到极端去检验。本书用纯标准库程序推翻它的适用范围:人数翻倍,人均出力下降;通道按 n(n-1)/2 增长,五十人有 1,225 条;串行环节不会因人多而变短。本书作蓝图修正:「八人组人均约为单人一半」出自林格曼 1913 年报告中一张未注明任务的汇总表(八人 0.49),Kravitz 与 Martin 1986 年的文章是找到并介绍原文,不是重新分析;原文写明的拉绳测力实验,七人组与十四人组的人均拉力为单人的 0.762 与 0.720,损失随人数饱和。书中参数为示意;全部计算由书中程序生成;本书属于非个人化经营建议。 第一章并陈汇总表、拉绳数据、Ingham 等人的伪群体实验与斯坦纳的过程损失。第二章计算布鲁克斯定律下向迟到项目加人的门槛。第三章把阿姆达尔串行比例与通用可扩展律的一致性项合并,求出容量峰值。第四章以本出版局每波四卷并行为示意场景:容量峰值在 6.78 个窗口,四个窗口已拿到峰值的 89.6 %,协调者的审稿队列把上限压在四。第五章把合并冲突写成生日问题:一个热点文件让四个窗口的冲突概率达到 0.6875,私有目录则为零。第六章比较网状、星形与分组的通道数,并用最长任务优先排出分工表。 第七章考辨《吕氏春秋·审分》「公作则迟,有所匿其力也」与林格曼 1913 年报告。第八章给出智能体并行数上限与分工表的实证闭环、蒸馏小工具「团队规模最优计算器」与预注册检验方案:人均比值低于 0.781 即拒绝「无损失」假设。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Ringelmann's Two Sets of Numbers: Where "Half at Eight" Comes From and What the Rope Experiment Says — In groups of seven each man pulled 0.762 of his solo force; in groups of fourteen, 0.720. The two differ by four percentage points while the group size doubled. Fitted to these two points, the loss ceiling L is 0.291. The saturating model has a root mean square error of only 0.011 on the two data points; the linear model has 0.493. The saturating model predicts 0.746 for a group of eight, far from the 0.49 of the summary table, and 0.855 and 0.806 for groups of two and three, clearly different from the direction in which the straight line through 0.93 and 0.85 is heading. In other words the two sets of numbers tell different stories. The summary table says losses accumulate linearly; the rope experiment says they saturate quickly. (https://5dlibrary.com/answers/en/ringelmann-brooks-law-team-agent-squad-optimal-size/c01/) 第1章 林格曼的两组数字:八人「一半」从哪里来,拉绳实验又说了什么 — 七人组的人均拉力是单人的 0.762,十四人组是 0.720,两者只差四个百分点,而人数翻了一倍。用这两点拟合,损失上限 L 为 0.291。饱和模型在两个数据点上的均方根误差只有 0.011,线性模型高达 0.493。饱和模型对八人组的预测是 0.746,与汇总表的 0.49 相差甚远;对两人与三人组的预测是 0.855 与 0.806,也明显高于汇总表的 0.93 与 0.85 所隐含的那条直线在远处的走向。换句话说,两组数字讲的不是同一个故事:汇总表说损失线性累积,拉绳实验说损失很快饱和。 (https://5dlibrary.com/answers/zh/ringelmann-brooks-law-team-agent-squad-optimal-size/c01/) - Ch.2 Brooks's Law: n(n-1)/2 Channels and the Newcomers on a Late Project — Two people have one channel, three have three, four have six, eight have twenty-eight, twelve have sixty-six, and fifty have 1,225. When the team doubles, channels rise sixfold from two people to four, by 4.67 times from four to eight, and approach a factor of four as the team grows, because the channel count is essentially the square of headcount. Each member maintains n-1 channels, so the time each member spends on channels is proportional to team size: with four members each spends 0.037 of the week on channels, with eight 0.087, with twenty 0.237, with fifty 0.613. With this illustrative parameter, at eighty-one members channels consume everyone's entire week. (https://5dlibrary.com/answers/en/ringelmann-brooks-law-team-agent-squad-optimal-size/c02/) 第2章 布鲁克斯定律:n(n-1)/2 条通道与迟到项目里的新人 — 两人只有一条通道,三人三条,四人六条,八人二十八条,十二人六十六条,五十人则是 1,225 条。人数翻倍时,通道数从两人翻到四人是六倍,从四人翻到八人是 4.67 倍,人数越多越接近四倍,因为通道数本质上是人数的平方。每个成员要维持的通道是 n-1 条,所以每人花在通道上的时间与人数成正比:四人时每人每周 0.037 的时间花在通道上,八人时 0.087,二十人时 0.237,五十人时 0.613。按这个示意参数,团队到八十一人时,每个人的全部时间都会被通道吃掉。 (https://5dlibrary.com/answers/zh/ringelmann-brooks-law-team-agent-squad-optimal-size/c02/) - Ch.3 Serial Fractions and Pairwise Coherence: Amdahl's Ceiling and the Optimal Size under the Universal Scalability Law — The first line has no coherence cost, only contention of 0.05; capacity rises monotonically towards 20.0 and has no peak. From the second line on, the coherence term b is positive and capacity peaks somewhere. With contention 0.05 and coherence 0.002 the peak lies at 21.79 members, the best integer is 22 and peak capacity is 7.40; doubling to forty-four members lowers capacity to 6.35. With coherence raised to 0.01 the peak moves down to 9.75 members and peak capacity to 4.26. With contention 0.10 and coherence 0.03 the peak lies at 5.48, so six members, with peak capacity 2.50 and efficiency per member 0.417; doubling to twelve lowers capacity to 1.98. With contention 0.20 and coherence 0.05 the peak is at exactly 4.00, and four members together are worth only 1.82 solo members. (https://5dlibrary.com/answers/en/ringelmann-brooks-law-team-agent-squad-optimal-size/c03/) 第3章 串行比例与成对协调:阿姆达尔上限与通用可扩展律的最优人数 — 第一行是没有一致性成本的情形,只有争用项 0.05,容量随人数单调上升,趋于 20.0,没有峰值。从第二行起,一致性项 b 大于零,容量都在某处达到峰值。争用 0.05、一致性 0.002 时,峰值在 21.79 人,整数最优为 22 人,峰值容量 7.40;人数加倍到四十四人,容量反而降到 6.35。一致性升到 0.01,峰值位置降到 9.75 人,峰值容量 4.26。争用 0.10、一致性 0.03 时,峰值在 5.48 人,取六人,峰值容量 2.50,人均效率 0.417;加倍到十二人,容量降到 1.98。争用 0.20、一致性 0.05 时,峰值恰好在 4.00 人,四个人加起来只抵 1.82 个单人。 (https://5dlibrary.com/answers/zh/ringelmann-brooks-law-team-agent-squad-optimal-size/c03/) - Ch.4 The Parallel Ceiling of an Agent Squad: How Many Volumes per Wave Pay Best — Chapter 4 takes this bureau's four volumes in parallel per wave as an illustrative scene: capacity peaks at 6.78 windows, four windows already reach 89.6 % of the peak, and the coordinator's review queue holds the ceiling at four. (https://5dlibrary.com/answers/en/ringelmann-brooks-law-team-agent-squad-optimal-size/c04/) 第4章 智能体军团的并行上限:出版局每波几卷最划算 — 第四章以本出版局每波四卷并行为示意场景:容量峰值在 6.78 个窗口,四个窗口已拿到峰值的 89.6 %,协调者的审稿队列把上限压在四。 (https://5dlibrary.com/answers/zh/ringelmann-brooks-law-team-agent-squad-optimal-size/c04/) - Ch.5 Merge Conflicts as a Birthday Problem: Shared Files, Hot Files and Private Directories — Chapter 5 treats merge conflicts as a birthday problem: one hot file raises the conflict probability of four windows to 0.6875, while private directories bring it to zero. (https://5dlibrary.com/answers/en/ringelmann-brooks-law-team-agent-squad-optimal-size/c05/) 第5章 合并冲突的生日问题:共享文件、热点文件与私有目录 — 第五章把合并冲突写成生日问题:一个热点文件让四个窗口的冲突概率达到 0.6875,私有目录则为零。 (https://5dlibrary.com/answers/zh/ringelmann-brooks-law-team-agent-squad-optimal-size/c05/) - Ch.6 Work Tables and the Pruning of Channels: Mesh, Star, Cells and Longest Job First — With two windows the work finishes in thirty-eight hours, a speed-up of 1.97; with three, in twenty-five hours, a speed-up of 3.00, an exact division; with four, in nineteen hours, a speed-up of 3.95 and an efficiency per window of 0.99. Up to this point the division costs almost nothing. With five windows longest job first gives seventeen hours against a theoretical bound of fifteen, a speed-up of only 4.41 and an efficiency of 0.88, because twelve tasks of unequal length cannot be shared evenly among five. Six windows return to thirteen hours and an efficiency of 0.96; seven give twelve hours and 0.89. With eight windows the largest window load falls to eleven hours, but the coordinator's twelve hours of review become the bottleneck, so completion still takes twelve hours, the speed-up stays at 6.25 and efficiency drops to 0.78. (https://5dlibrary.com/answers/en/ringelmann-brooks-law-team-agent-squad-optimal-size/c06/) 第6章 分工表与通道修剪:网状、星形、小组与最长任务优先 — 两个窗口时完工时间是三十八小时,加速比 1.97;三个窗口二十五小时,加速比 3.00,正好整除;四个窗口十九小时,加速比 3.95,每个窗口效率 0.99。到这里,分工几乎没有损失。五个窗口时最长任务优先给出十七小时,而理论下界是十五小时,加速比只有 4.41,效率降到 0.88,因为十二个长短不一的任务无法被五个人均匀分开。六个窗口又回到十三小时,效率 0.96;七个窗口十二小时,效率 0.89;八个窗口时,窗口的最大负荷降到十一小时,但协调者的十二小时审阅成了瓶颈,完工时间仍是十二小时,加速比停在 6.25,效率跌到 0.78。 (https://5dlibrary.com/answers/zh/ringelmann-brooks-law-team-agent-squad-optimal-size/c06/) - Ch.7 Classical Sources East and West: "Working in Common Is Slow" in the Lüshi Chunqiu and Ringelmann in 1913 — Chapter 7 examines the Shen Fen chapter of the Lüshi Chunqiu, "working in common is slow, because there is room to hide one's strength", and Ringelmann's report of 1913. (https://5dlibrary.com/answers/en/ringelmann-brooks-law-team-agent-squad-optimal-size/c07/) 第7章 东西方原典考辨:《吕氏春秋·审分》「公作则迟」与林格曼 1913 — 第七章考辨《吕氏春秋·审分》「公作则迟,有所匿其力也」与林格曼 1913 年报告。 (https://5dlibrary.com/answers/zh/ringelmann-brooks-law-team-agent-squad-optimal-size/c07/) - Ch.8 Closing the Loop: Agent Squad Ceilings, Work Tables and a Pre-Registered Test — Chapter 8 gives the empirical loop for agent squad ceilings and work tables, the distilled Optimal Team Size Calculator and a pre-registered test: a ratio per member below 0.781 rejects the hypothesis of no loss. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/ringelmann-brooks-law-team-agent-squad-optimal-size/c08/) 第8章 实证闭环:智能体并行数上限、分工表与预注册检验 — 第八章给出智能体并行数上限与分工表的实证闭环、蒸馏小工具「团队规模最优计算器」与预注册检验方案:人均比值低于 0.781 即拒绝「无损失」假设。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/ringelmann-brooks-law-team-agent-squad-optimal-size/c08/) ## 5DL-B-323 · Chinese Readability from Character Frequency and Sentence Length: Reading Difficulty of Landing Pages and Monograph Paragraphs 中文书名:中文可读性公式与字频句长、落地页与专著段落的阅读难度母典 DOI: pending Cite: Fan, Y. (2026). Chinese Readability from Character Frequency and Sentence Length: Reading Difficulty of Landing Pages and Monograph Paragraphs (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations are generated by the book's programs; the book offers general rather than individualised business advice. Chapters 2 and 3 treat common-character lists, an illustrative frequency-tier table and the rules for cutting sentences. Abstract: Writing rules say that shorter sentences read more easily. This book overturns the rule in its absolute form: the difficulty of reading Chinese comes at least from sentence length and from how common the characters are, and length alone misjudges in two directions. Among four small texts written by the author, short four-character sentences packed with rare characters have the shortest sentences yet are the hardest under the book's illustrative difficulty formula, scoring 19.15, against 0.45 for equally short sentences in common characters. Published research supports the direction: the Chinese readability models of Sung and colleagues rank the number of difficult words and word frequency as leading predictors, and Soh found that the readability of Singapore's Chinese-language lessons depends more on character and word features than on sentence features. After checking the sources the book corrects its blueprint: readability has far more than two features, the book's two-feature formula is a teaching skeleton, and early works whose year or volume could not be verified are not cited. Parameters are illustrative; all calculations are generated by the book's programs; the book offers general rather than individualised business advice. Chapter 1 derives the probability that length alone misorders a pair of texts: 0.109 when length and rare share correlate at 0.8, rising to 0.390 at a correlation of minus 0.6. Chapters 2 and 3 treat common-character lists, an illustrative frequency-tier table and the rules for cutting sentences. Chapter 4 combines the two terms into a simplified linear formula, calibrates it on a synthetic corpus and shows how collinearity makes a length formula look sufficient. Chapter 5 compares two levers on a landing page, cutting sentences and swapping characters: the original scores 9.22 and doing both brings it to 1.16. Chapter 6 proposes a glossary exemption and a rare-character budget for monographs; the sample paragraph falls from 4.67 to 3.53. Chapter 7 examines the four principles of choosing characters in Liu Xie's Wenxin Diaolong and Flesch's 1948 readability formula. Chapter 8 plugs readability hints into a persuasive-copy checker and a book engine, presents the distilled Chinese Readability Scorer and a pre-registered within-reader test needing 197 readers, and states which results would overturn the core claim. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:写作守则说句子越短越好读。本书推翻这条守则的绝对形式:中文的阅读难度至少来自句长与字的常见程度两处,只看句长会在两个方向上误判。作者自拟的四篇小样本中,四字短句堆僻字的文本平均句长最短,按本书的示意难度公式却最难,得分 19.15;同样短句、全用常用字的文本只有 0.45。已发表研究支持这一方向:宋曜廷等人的中文可读性模型以难词数与词频为首要预测变项,苏启祯发现新加坡华文课文的可读性受字词特质影响大于语句特质。核对原文后本书作蓝图修正:可读性特征远不止两项,本书的二元公式只是教学骨架;未能核实年份与卷期的早期文献不予引用。书中参数为示意;全部计算由书中程序生成;本书属于非个人化经营建议。 第一章推导单看句长排错文本对的概率:句长与少见字比例的相关系数为 0.8 时错误率 0.109,为负 0.6 时升到 0.390。第二、三章讨论常用字表、示意字频等级表与切句规则。第四章把两项合成简化线性公式并在合成语料上标定,揭示共线性如何让句长公式看似够用。第五章在落地页上比较砍句子与换字两条杠杆,原稿难度 9.22,两者兼施降到 1.16。第六章提出专著的术语豁免与少见字预算,样例段落从 4.67 降到 3.53。 第七章考辨《文心雕龙·练字》的四条选字原则与 Flesch 一九四八年的可读性公式。第八章把可读性提示接入攻心文案检查器与写书母机,给出蒸馏小工具「中文可读性评分器」,以及需要 197 名读者的组内预注册检验方案,并写明什么结果会推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Two Misjudgements of "Shorter Reads Easier": the Plane of Sentence Length and Common-Character Share — Chapter 1 derives the probability that length alone misorders a pair of texts: 0.109 when length and rare share correlate at 0.8, rising to 0.390 at a correlation of minus 0.6. (https://5dlibrary.com/answers/en/chinese-readability-character-frequency-sentence-length/c01/) 第1章 「句子越短越好读」的两种误判:句长与常用字比例的二维平面 — 第一章推导单看句长排错文本对的概率:句长与少见字比例的相关系数为 0.8 时错误率 0.109,为负 0.6 时升到 0.390。 (https://5dlibrary.com/answers/zh/chinese-readability-character-frequency-sentence-length/c01/) - Ch.2 The Long Tail of Character Frequency: the General Standard Table, a Zipf Model and an Illustrative Frequency-Tier Table — The program first checks that the three levels add up: 3500 plus 3000 plus 1605 equals 8105. The exponent that meets the target coverage of 0.99 is s = 1.3840. In this model the first 1000 characters cover 0.9676, the first 2500 cover 0.9850 and level one covers 0.9900, so 1.00 character in every hundred falls outside level one; levels one and two together, 6500 characters, cover 0.9977. (https://5dlibrary.com/answers/en/chinese-readability-character-frequency-sentence-length/c02/) 第2章 字频的长尾:通用规范汉字表、齐普夫模型与一份示意字频等级表 — 程序先核对三级字数相加:3500 加 3000 加 1605,等于 8105。满足目标覆盖率 0.99 的指数是 s = 1.3840。在这个模型里,前 1000 个字覆盖 0.9676,前 2500 个字覆盖 0.9850,一级字表覆盖 0.9900,即每百字有 1.00 个字落在一级字表之外;一、二级合计 6500 字覆盖 0.9977。 (https://5dlibrary.com/answers/zh/chinese-readability-character-frequency-sentence-length/c02/) - Ch.3 Where to Cut a Sentence: Segmentation Rules, Mean and Median — The three rules yield 5, 6 and 16 units. Under rule A the mean sentence length is 26.0 characters, the median 20.0 and the longest sentence 69; rule B splits that sixty-nine-character sentence at its semicolon, the mean falls to 21.7, the median actually rises to 22.5, and the longest unit becomes 42; rule C cuts into clauses with a mean of only 8.1 characters, a median of 9.0, the longest clause 16, and no unit above thirty characters. (https://5dlibrary.com/answers/en/chinese-readability-character-frequency-sentence-length/c03/) 第3章 句长从哪里切:切分规则、均值与中位数 — 三种规则得到的单位数分别是 5、6 与 16 个。规则甲的平均句长是 26.0 字,中位数 20.0 字,最长一句 69 字;规则乙把那句六十九字的长句在分号处一分为二,平均句长降到 21.7 字,中位数反而升到 22.5 字,最长一句变为 42 字;规则丙按小句切,平均长度只有 8.1 字,中位数 9.0 字,最长的小句 16 字,没有一个单位超过三十字。 (https://5dlibrary.com/answers/zh/chinese-readability-character-frequency-sentence-length/c03/) - Ch.4 A Simplified Linear Readability Formula: Term-by-Term Decomposition, Calibration and the Collinearity Trap — Across the 120 texts, the correlation between mean sentence length and rare share is 0.464: both rise with grade, so they are correlated with each other. Regressing on length alone gives grade equals minus 0.620 plus 0.2030 times mean sentence length, with a coefficient of determination of 0.653 and a root mean square error of 1.007 grades. Adding the rare term gives grade equals minus 0.694 plus 0.1757 times mean sentence length plus 0.6457 times rare characters per hundred; the coefficient of determination rises to 0.696 and the error falls to 0.942 grades. (https://5dlibrary.com/answers/en/chinese-readability-character-frequency-sentence-length/c04/) 第4章 简化线性可读性公式:逐项分解、标定与共线性陷阱 — 逐行看。短句常用字版两项都很小,D 为 0.45,属易读。卖点段平均句长 28.50 字,没有少见字,句长项 2.85 就是全部,属平实。长句常用字版句长项 3.70,同样平实。落地页标题的平均句长只有 8.00 字,句长项 0.80,但每百字 12.50 个少见字贡献了 3.75,合计 4.55,进入偏难档。专著段落两项都不小,句长项 3.53,少见字项 1.13,合计 4.67,也是偏难。售后条款没有一个少见字,但一句 72.00 字的长句让句长项达到 7.20,独自进入密集档。长句僻字版合计 11.93,短句僻字版合计 19.15,其中少见字项 18.75 占了绝大部分。 (https://5dlibrary.com/answers/zh/chinese-readability-character-frequency-sentence-length/c04/) - Ch.5 Reading Difficulty of a Landing Page: Headline, Selling Points and Terms, and Two Levers for Rewriting — Chapter 5 compares two levers on a landing page, cutting sentences and swapping characters: the original scores 9.22 and doing both brings it to 1.16. (https://5dlibrary.com/answers/en/chinese-readability-character-frequency-sentence-length/c05/) 第5章 落地页的阅读难度:标题、卖点与条款三段,以及两条改写杠杆 — 第五章在落地页上比较砍句子与换字两条杠杆,原稿难度 9.22,两者兼施降到 1.16。 (https://5dlibrary.com/answers/zh/chinese-readability-character-frequency-sentence-length/c05/) - Ch.6 Monograph Paragraphs: Glossary Exemption and a Rare-Character Budget — Chapter 6 proposes a glossary exemption and a rare-character budget for monographs; the sample paragraph falls from 4.67 to 3.53. (https://5dlibrary.com/answers/en/chinese-readability-character-frequency-sentence-length/c06/) 第6章 专著段落:术语豁免与少见字预算 — 第六章提出专著的术语豁免与少见字预算,样例段落从 4.67 降到 3.53。 (https://5dlibrary.com/answers/zh/chinese-readability-character-frequency-sentence-length/c06/) - Ch.7 Eastern and Western Classics Examined: the "Choosing Characters" Chapter of Wenxin Diaolong and Flesch 1948 — The passage of short sentences and long words has a mean sentence length of 2.00 words, 5.5000 syllables per word, and a Reading Ease of minus 260.49, far outside the usual range of zero to a hundred. The passage of one long sentence and short words has a mean sentence length of 34.00 words, 1.0294 syllables per word and a score of 85.24, at the easy end. Flesch's formula does not commit the misjudgement of Chapter 1: its word-length term rates short sentences of long words as extremely hard. (https://5dlibrary.com/answers/en/chinese-readability-character-frequency-sentence-length/c07/) 第7章 东西方原典考辨:《文心雕龙·练字》与 Flesch 1948 — 第一组对应「避诡异」:短句僻字版每百字 62.50 个少见字,长句常用字版为零,与第一章一致。第二组对应「省联边」:「江河湖海波浪滔滔」八个字全是三点水,连续长度为 8,远超刘勰的三接之限,被标记;同一句后半的「江上有风」连续长度为 1,通过。第三组对应「权重出」:一句十三字的话,第 5、6、9、10 个位置上的字都在前六字之内出现过,十三字中四字重复,是「问题」二字反复出现造成的。第四组对应「调单复」:「一人上山」平均 2.25 画,「大水小日」平均 3.50 画,都是瘦字;「警露翻疆」平均 19.25 画,「黯霸鑫矗」平均 22.50 画,都是肥字。 (https://5dlibrary.com/answers/zh/chinese-readability-character-frequency-sentence-length/c07/) - Ch.8 Closing the Loop: Readability Hints for the Copy Checker and the Book Engine, the Chinese Readability Scorer and a Pre-registered Reader Test — Chapter 8 plugs readability hints into a persuasive-copy checker and a book engine, presents the distilled Chinese Readability Scorer and a pre-registered within-reader test needing 197 readers, and states which results would overturn the core claim. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/chinese-readability-character-frequency-sentence-length/c08/) 第8章 实证闭环:攻心文案检查器与写书母机的可读性提示、中文可读性评分器与预注册读者检验 — 第八章把可读性提示接入攻心文案检查器与写书母机,给出蒸馏小工具「中文可读性评分器」,以及需要 197 名读者的组内预注册检验方案,并写明什么结果会推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/chinese-readability-character-frequency-sentence-length/c08/) ## 5DL-B-324 · Simulating Colour-Vision Deficiency and Colour Discriminability: Buttons and Charts as Seen by Red-Green Colour-Deficient Users 中文书名:色觉缺陷模拟与配色可辨性、红绿色弱用户眼中的按钮与图表母典 DOI: pending Cite: Fan, Y. (2026). Simulating Colour-Vision Deficiency and Colour Discriminability: Buttons and Charts as Seen by Red-Green Colour-Deficient Users (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: Many teams believe that once button labels reach the WCAG contrast ratio of four and a half to one, accessibility is done. This book overturns that reasoning with colour-vision simulation and colour-difference calculation. According to Birch's review of large surveys, about 8 per cent of European Caucasian men have inherited red-green colour-vision deficiency, and between four and six and a half per cent of Chinese and Japanese men; contrast looks only at luminance, while the red-green distinction lies in hue. A common red and green button pair passes with white labels and differs by 64.28 in normal view, but only by 9.06 after Machado's severity-one deutan simulation; among nine thousand red-green pairs with passing white labels, the share below a difference of five after simulation is 0.200. The book corrects the blueprint: the population is given in the source's wording, European Caucasian, and "indistinguishable" becomes "a substantial share close to indistinguishable, the rest much compressed". Parameters are illustrative; all calculations are generated by the book's programs; the simulation stands for no particular person, the book makes no judgement about any individual's colour vision, and it is general business analysis, not individualised business advice. Chapters 1 to 4 present population shares and a counterexample, sRGB linearisation and WCAG contrast, Machado's three simulation matrices and confusion pairs, and an implementation of CIELAB and CIEDE2000 checked digit by digit against published test data. Chapter 5 searches safe colours for right and wrong buttons: one dark and one light, one with a white and one with a black label, 3.01 apart in luminance, still differ by 21.85 under the deutan simulation. Chapter 6 computes discriminability matrices of chart palettes: the first five colours of a common ten-colour default have a smallest difference of only 1.25, the palette of Okabe and Ito 11.61. Chapter 7 examines the paired colours of the painting passage in the Kaogong ji and Dalton's 1798 account of his own colour vision. Chapter 8 writes the rules as four colour-vision checks for a front-end style library, gives the distilled Colour-Vision-Safe Palette Checker and a pre-registered test, and states which results would overturn the core claim. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:许多团队相信:按钮上的文字对比度达到 WCAG 规定的四点五比一,可访问性就算合格。本书用色觉模拟与色差计算推翻这条推理。据 Birch 对各国大样本调查的综述,欧洲白人男性约百分之八有先天红绿色觉缺陷,中国与日本男性在百分之四到百分之六点五之间;对比度只看亮度,而红绿之分在色相上。常见的一对红绿按钮白字都合格,正常视觉下色差为 64.28,用 Machado 严重度为一的第二型矩阵模拟后只剩 9.06;在九千对白字合格的红绿组合中,模拟后色差低于五的比例为 0.200。本书作蓝图修正:人群口径改用原文的「欧洲白人」,并把「不可区分」改为「相当一部分组合接近不可区分、其余差别被大幅压缩」。书中参数为示意;全部计算由书中程序生成;模拟不对应任何具体的人,本书不对任何个人的色觉作判断,属于非个人化经营建议。 第一至四章依次给出人群比例与反例、sRGB 线性化与 WCAG 对比度、Machado 三个模拟矩阵与混淆对、CIELAB 与 CIEDE2000 的实现及与公开检验数据的逐位核对。第五章为对错按钮搜索安全配色:一深一浅、一白字一黑字、彼此亮度比 3.01 的组合在第二型模拟下仍有 21.85 的色差。第六章计算图表色板的可辨性矩阵:常见十色默认色板前五色的最小色差只有 1.25,Okabe 与 Ito 的色板为 11.61。 第七章考辨《考工记·画缋之事》的五色相次与道尔顿 1798 年对自身色觉的记述。第八章把规则写成前端样式库的四条色觉模拟检查,给出蒸馏小工具「色觉安全配色检查器」与预注册检验方案,并写明什么结果会推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Eight Per Cent and Four and a Half to One: Why Red and Green Buttons That Pass Contrast Still Collide — Chapters 1 to 4 present population shares and a counterexample, sRGB linearisation and WCAG contrast, Machado's three simulation matrices and confusion pairs, and an implementation of CIELAB and CIEDE2000 checked digit by digit against published test data. (https://5dlibrary.com/answers/en/colour-vision-deficiency-simulation-button-chart-discriminability/c01/) 第1章 八个百分点与四点五比一:对比度合格的红绿按钮为何仍会撞色 — 第一至四章依次给出人群比例与反例、sRGB 线性化与 WCAG 对比度、Machado 三个模拟矩阵与混淆对、CIELAB 与 CIEDE2000 的实现及与公开检验数据的逐位核对。 (https://5dlibrary.com/answers/zh/colour-vision-deficiency-simulation-button-chart-discriminability/c01/) - Ch.2 Linear Light and Relative Luminance: What WCAG Contrast Computes and What It Leaves Out — The lower part of the table deserves more attention. Code 128 looks like "middle grey" and normalises to 0.5020, yet it carries only 0.2159 of the linear light of white, about a fifth. Half the linear light needs code 188, whose linear value is 0.5029. Code 119 has linear value 0.1845, a number that will keep returning: it lies very close to the luminance at which white and black labels both barely pass. Code 64 has linear value 0.0513, and code 11 only 0.0033. (https://5dlibrary.com/answers/en/colour-vision-deficiency-simulation-button-chart-discriminability/c02/) 第2章 线性光与相对亮度:WCAG 对比度算了什么、没算什么 — 表的下半部分更值得注意。编码 128 看起来是「中灰」,归一化后是 0.5020,可它携带的线性光只有 0.2159,约为白色的五分之一。要得到一半的线性光,需要编码 188,其线性值为 0.5029。编码 119 的线性值为 0.1845,这个数后面会反复出现:它非常接近白字与黑字都勉强合格的那个亮度。编码 64 的线性值是 0.0513,编码 11 只有 0.0033。 (https://5dlibrary.com/answers/zh/colour-vision-deficiency-simulation-button-chart-discriminability/c02/) - Ch.3 The Machado Matrices: Simulating Three Kinds of Dichromacy in Linear sRGB — The 216 colours form 23220 pairs. The protan simulation yields 46 confusion pairs, a share of 0.0020; the most extreme is the dark green #006633 against the bright red #FF0033, which differ by 70.61 in normal view and by only 1.51 after simulation. The deutan simulation yields 120 pairs, a share of 0.0052; the most extreme is the turquoise #00CC99 against the pink #FF6699, a normal difference of 77.30 reduced to 2.02. The tritan simulation yields no confusion pairs under these thresholds. (https://5dlibrary.com/answers/en/colour-vision-deficiency-simulation-button-chart-discriminability/c03/) 第3章 Machado 矩阵:在线性 sRGB 里模拟三类二色视 — 216 种颜色共有 23220 对。第一型模拟下有 46 对混淆,占 0.0020;最极端的一对是深绿 #006633 与亮红 #FF0033,正常视觉下色差 70.61,模拟后只剩 1.51。第二型模拟下有 120 对,占 0.0052;最极端的是青绿 #00CC99 与粉红 #FF6699,正常色差 77.30,模拟后 2.02。第三型模拟下,在这个阈值组合下没有混淆对。 (https://5dlibrary.com/answers/zh/colour-vision-deficiency-simulation-button-chart-discriminability/c03/) - Ch.4 CIELAB and CIEDE2000: Turning "a Visible Difference" into a Number — The two formulas diverge most at middling thresholds. At a threshold of five, CIE76 finds only 0.068 of the pairs colliding while CIEDE2000 finds 0.200, about three times as many; at ten, CIE76 gives 0.264 and CIEDE2000 0.555. At three the two give 0.025 and 0.073. At the very small threshold of one both stay below one per cent, and at the large threshold of twenty they give 0.672 and 0.919. The last line says that 209 pairs are ten or more apart under CIE76 but less than five apart under CIEDE2000: a check based on CIE76 would call these pairs "clearly separable", while under the formula closer to human judgement they are "hard to separate". (https://5dlibrary.com/answers/en/colour-vision-deficiency-simulation-button-chart-discriminability/c04/) 第4章 CIELAB 与 CIEDE2000:把「看得出差别」写成一个数 — 两个公式的分歧在中等阈值处最大。以五为阈值,CIE76 认为只有 0.068 的组合撞色,CIEDE2000 认为有 0.200,相差约三倍;以十为阈值,CIE76 为 0.264,CIEDE2000 为 0.555。以三为阈值,两者分别为 0.025 与 0.073。在很小的阈值一处,两者都不到百分之一;在很大的阈值二十处,两者分别为 0.672 与 0.919。最后一行说明,有 209 对组合在 CIE76 下相差十以上,按 CIEDE2000 却不到五:用 CIE76 做检查,这些组合会被判为「明显可分」,而它们在更接近人眼的公式下属于「很难分开」。 (https://5dlibrary.com/answers/zh/colour-vision-deficiency-simulation-button-chart-discriminability/c04/) - Ch.5 Safe Colours for Right and Wrong Buttons: a Hue Search and a Luminance-Only Scheme — Chapter 5 searches safe colours for right and wrong buttons: one dark and one light, one with a white and one with a black label, 3.01 apart in luminance, still differ by 21.85 under the deutan simulation. (https://5dlibrary.com/answers/en/colour-vision-deficiency-simulation-button-chart-discriminability/c05/) 第5章 对错按钮的安全配色:色相搜索与只靠亮度差的方案 — 第五章为对错按钮搜索安全配色:一深一浅、一白字一黑字、彼此亮度比 3.01 的组合在第二型模拟下仍有 21.85 的色差。 (https://5dlibrary.com/answers/zh/colour-vision-deficiency-simulation-button-chart-discriminability/c05/) - Ch.6 Chart Palettes: Discriminability Matrices and How Many Series Colour Can Carry — Chapter 6 computes discriminability matrices of chart palettes: the first five colours of a common ten-colour default have a smallest difference of only 1.25, the palette of Okabe and Ito 11.61. (https://5dlibrary.com/answers/en/colour-vision-deficiency-simulation-button-chart-discriminability/c06/) 第6章 图表系列色板:可辨性矩阵与颜色能承载几条序列 — 第六章计算图表色板的可辨性矩阵:常见十色默认色板前五色的最小色差只有 1.25,Okabe 与 Ito 的色板为 11.61。 (https://5dlibrary.com/answers/zh/colour-vision-deficiency-simulation-button-chart-discriminability/c06/) - Ch.7 Classical Sources East and West: the Paired Colours of the Kaogong ji and Dalton's 1798 Account — Chapter 7 examines the paired colours of the painting passage in the Kaogong ji and Dalton's 1798 account of his own colour vision. (https://5dlibrary.com/answers/en/colour-vision-deficiency-simulation-button-chart-discriminability/c07/) 第7章 东西方原典考辨:《考工记·画缋之事》的五色相次与道尔顿 1798 — 第七章考辨《考工记·画缋之事》的五色相次与道尔顿 1798 年对自身色觉的记述。 (https://5dlibrary.com/answers/zh/colour-vision-deficiency-simulation-button-chart-discriminability/c07/) - Ch.8 Closing the Loop: a Colour-Vision Check in the Front-End Style Library, the Colour-Vision-Safe Palette Checker and a Pre-Registered Test — Chapter 8 writes the rules as four colour-vision checks for a front-end style library, gives the distilled Colour-Vision-Safe Palette Checker and a pre-registered test, and states which results would overturn the core claim. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/colour-vision-deficiency-simulation-button-chart-discriminability/c08/) 第8章 实证闭环:前端样式库的色觉模拟检查、色觉安全配色检查器与预注册检验 — 第八章把规则写成前端样式库的四条色觉模拟检查,给出蒸馏小工具「色觉安全配色检查器」与预注册检验方案,并写明什么结果会推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/colour-vision-deficiency-simulation-button-chart-discriminability/c08/) ## 5DL-B-325 · Inverse-Square Lighting and Area-Source Falloff in the Studio: Mixed Colour Temperatures and the Algebra of Lighting Ratios 中文书名:摄影棚布光平方反比与面光源衰减、色温混合与光比代数母典 DOI: pending Cite: Fan, Y. (2026). Inverse-Square Lighting and Area-Source Falloff in the Studio: Mixed Colour Temperatures and the Algebra of Lighting Ratios (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: Studios repeat a rule: move the light twice as far and a quarter remains, so open two stops. This book uses illuminance integrals to overturn the rule's range: it holds only for point sources. On the axis of a uniform Lambertian disc illuminance is pi L times R² over R² plus d², and close in it falls far more slowly than the inverse square; at a distance of one diameter, doubling the distance costs only 1.766 stops. The project blueprint gave no source, and the book corrects it: the evidence is the five times rule of a light measurement handbook and the configuration-factor catalogue of radiative heat transfer, and the book shows that the rule read as five diameters has an error of 0.0100, misread as five radii 0.0400; the accuracy of McCamy's cubic for correlated colour temperature is checked by the book itself against CIE standard illuminants, not quoted from secondary sources; only the canon text of the Mohist passage is quoted, its explanation summarised. Parameters are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Chapters 1 to 3 rebuild the disc and rectangle formulas from rings and grids, introduce a local falloff exponent and show that a strip light falls off like a line at intermediate distances. Chapter 4 is ratio algebra: placing the fill by the inverse square turns a three-to-one target into 2.80:1. Chapter 5 tests McCamy's formula and mired steps. Chapter 6 mixes two colour temperatures on one face: with half the light from daylight the correlated colour temperature is 4142 K, not the kelvin mean of 4400 K, and the mixture lies below the Planckian locus; at three to one a white balance of 3809 K equalises the errors of lit and shadow sides. Chapter 7 examines the Mohist account of the inverted pinhole image and Lambert's Photometria. Chapter 8 gives lighting cards for a StudioTaiwan set, the distilled Lighting Ratio Calculator and a pre-registered test, stating which measurements would overturn the core claim. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:摄影棚里流传一句口诀:灯离主体远一倍,光剩四分之一,要补两档。本书用照度积分推翻这条口诀的适用范围:它只对点光源成立。均匀朗伯圆盘轴上照度为 πL 乘 R² 除以 R² 与 d² 之和,近距离衰减远慢于平方反比;距离等于一个直径时,「距离加倍」只损失 1.766 档。立项蓝图没有给出处,本书作蓝图修正:以光测量手册的五倍法则与辐射换热角系数目录为证据,并指出五倍法则按五倍直径读误差为 0.0100,误读为五倍半径则为 0.0400;McCamy 相关色温三次式的精度由本书用标准照明体色度自行检验,不转引二手数字;《墨经》只直引经文,经说述其大意。书中参数为示意;全部计算由书中程序生成;本书属于非个人化经营建议。 第一至三章用圆环与网格积分重建圆盘与矩形照度公式,引入局部衰减指数,并说明细长条灯在中等距离上衰减得像一条线。第四章是光比代数:按平方反比摆放辅光,三比一的目标实际只得到 2.80:1。第五章检验 McCamy 公式与 mired 档位。第六章计算两种色温在一张脸上的混合:日光份额一半时相关色温为 4142 K,不是开尔文平均的 4400 K,混合点落在普朗克轨迹下方;三比一时把白平衡设在 3809 K 可使亮暗两面偏差相等。 第七章考辨《墨经》「景到,在午有端」与朗伯《光度学》。第八章给出 StudioTaiwan 摄影棚布光卡、蒸馏小工具「布光光比计算器」与预注册检验方案,并写明什么测量结果会推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Inverse Square Belongs to Point Sources: the Real Falloff of a 120 cm Softbox at Every Distance — Distances are in units of the diameter, so the table holds for a softbox of any size. When the distance is only 0.250 of the diameter, doubling it reduces illuminance by a factor of just 1.600, a loss of 0.678 stops, more than a stop short of two. At half a diameter the loss is 1.322 stops. At one diameter it is 1.766 stops, 0.234 stops less than the rule. Only at two diameters does the loss reach 1.935 stops; at five diameters it is 1.989 stops, just 0.011 stops short of two. (https://5dlibrary.com/answers/en/studio-lighting-inverse-square-area-source-colour-temperature/c01/) 第1章 平方反比只属于点光源:一米二柔光箱在各个距离上的真实衰减 — 表里的距离以直径为单位,因此对任何尺寸的柔光箱都成立。当距离只有直径的 0.250 倍时,距离加倍只让照度降为原来的 1.600 分之一,损失 0.678 档,离两档差了一档多。距离等于半个直径时,损失 1.322 档。距离等于一个直径时,损失 1.766 档,比口诀少 0.234 档。要到两倍直径,损失才达到 1.935 档;五倍直径时为 1.989 档,与两档只差 0.011 档。 (https://5dlibrary.com/answers/zh/studio-lighting-inverse-square-area-source-colour-temperature/c01/) - Ch.2 A Disc Built from Rings: the Illuminance Integral of an Area Source, Evenness on the Subject Plane and the Local Falloff Exponent — For a disc 1.2 m across, at 0.2 m n is only 0.2000, so illuminance hardly changes with distance. At 0.6 m, half a diameter, n equals 1.0000 and the falloff resembles a line source rather than a point. At 1.2 m, one diameter, n is 1.6000; not until 2.4 m does it reach 1.8824, and at 3.6 m it is 1.9459. The foot of the table turns the question round: n reaches 1.5 at 1.039 m, about 0.866 diameters, and 1.9 at 2.615 m, about 2.179 diameters. (https://5dlibrary.com/answers/en/studio-lighting-inverse-square-area-source-colour-temperature/c02/) 第2章 从圆环积出圆盘:面光源的照度积分、主体平面的均匀度与局部衰减指数 — 一米二的圆盘,在 0.2 m 处 n 只有 0.2000,几乎不随距离衰减;在 0.6 m 处,即半个直径,n 等于 1.0000,衰减像一条线光源而不是点光源;在 1.2 m 处,即一个直径,n 为 1.6000;要到 2.4 m,n 才到 1.8824;3.6 m 时为 1.9459。表的末尾把问题倒过来:n 达到 1.5 需要 1.039 m,约 0.866 倍直径;n 达到 1.9 需要 2.615 m,约 2.179 倍直径。 (https://5dlibrary.com/answers/zh/studio-lighting-inverse-square-area-source-colour-temperature/c02/) - Ch.3 Rectangular Boxes and Strip Lights: Configuration Factors of a Rectangle, and Why a Thin Source Falls Off Like a Line — Now the last column. At half a metre the large box gives only 0.411 of the point-model prediction, at one metre 0.731, at two metres 0.915. The small box, about half the size, is closer to a point at the same distances: 0.679 at half a metre and 0.971 at two metres. The strip gives 0.395 at half a metre, below even the large box; 0.675 at one metre, again below the large box; and 0.883 at two metres, likewise below the large box's 0.915. The strip has only half the area of the large box, so why is it less point-like at every distance? Because of its long side. A length of one metre eighty shrinks to a point only from very far away; until then the strip remains an "unfolded" source along its length. (https://5dlibrary.com/answers/en/studio-lighting-inverse-square-area-source-colour-temperature/c03/) 第3章 方箱与条灯:矩形面光源的角系数,以及细长光源为何衰减得像一条线 — 再看最后一列。大方箱在半米处的照度只有点光源模型预测值的 0.411,一米处为 0.731,两米处为 0.915。小方箱尺寸只有大方箱的一半左右,同样的距离上更接近点光源:半米处为 0.679,两米处为 0.971。条灯在半米处为 0.395,比大方箱更低;一米处为 0.675,也低于大方箱;两米处为 0.883,同样低于大方箱的 0.915。条灯的面积其实比大方箱小一半,为什么在每个距离上都比大方箱更不像点光源?原因在于它的长边。一米八的长度,要到很远的地方才能缩成一个点;在那之前,条灯沿长边方向始终是一个「展开」的光源。 (https://5dlibrary.com/answers/zh/studio-lighting-inverse-square-area-source-colour-temperature/c03/) - Ch.4 The Algebra of Lighting Ratios: Key, Fill, Stops and Background, Repositioned by the Area-Source Law — Chapter 4 is ratio algebra: placing the fill by the inverse square turns a three-to-one target into 2.80:1. (https://5dlibrary.com/answers/en/studio-lighting-inverse-square-area-source-colour-temperature/c04/) 第4章 光比代数:主光、辅光、档数与背景,按面光源规律重新摆位 — 第四章是光比代数:按平方反比摆放辅光,三比一的目标实际只得到 2.80:1。 (https://5dlibrary.com/answers/zh/studio-lighting-inverse-square-area-source-colour-temperature/c04/) - Ch.5 Colour Temperature Is Not a Thermometer Reading: the Planckian Locus, a Test of McCamy's Cubic and Mired Steps — Now the main table. The second and third columns are blackbody chromaticities computed with the analytic fit. The fit is not identical to the standard tables, so at 2856 K x comes out as 0.4462 against the standard 0.44757, and McCamy's recovered temperature carries the fitting error: at 2856 K it returns 2888.5 K, 32.5 K high; at 3200 K, 3235.4 K; at 5600 K, 5612.2 K, 12.2 K high; at 6504 K, 6510.0 K. The last column converts these errors to mired: only -3.42 mired at 3200 K and -0.39 mired at 5600 K. The next section shows that, measured in mired, such errors are negligible for photography. (https://5dlibrary.com/answers/en/studio-lighting-inverse-square-area-source-colour-temperature/c05/) 第5章 色温不是温度计读数:普朗克轨迹、McCamy 三次式的检验与 mired 档位 — 再看主表。第二列与第三列是用解析拟合函数算出的黑体色度。拟合函数并不完全等于标准表格,所以在 2856 K 处得到 x 为 0.4462,而标准值是 0.44757;由此,McCamy 公式还原出的温度带上了拟合误差:2856 K 处还原为 2888.5 K,偏高 32.5 K;3200 K 处还原为 3235.4 K;5600 K 处为 5612.2 K,偏高 12.2 K;6504 K 处为 6510.0 K。最后一列把这些误差换算成 mired:3200 K 处只差 -3.42 mired,5600 K 处只差 -0.39 mired。下一节会说明,以 mired 计,这样的误差在摄影上微不足道。 (https://5dlibrary.com/answers/zh/studio-lighting-inverse-square-area-source-colour-temperature/c05/) - Ch.6 Two Colour Temperatures on One Face: the Chromaticity of Mixed Light, a Quantified White-Balance Error and the Price of a Correction Gel — Chapter 6 mixes two colour temperatures on one face: with half the light from daylight the correlated colour temperature is 4142 K, not the kelvin mean of 4400 K, and the mixture lies below the Planckian locus; at three to one a white balance of 3809 K equalises the errors of lit and shadow sides. (https://5dlibrary.com/answers/en/studio-lighting-inverse-square-area-source-colour-temperature/c06/) 第6章 一张脸上的两种色温:混合光的色度、白平衡偏差的量化与校色纸的代价 — 第六章计算两种色温在一张脸上的混合:日光份额一半时相关色温为 4142 K,不是开尔文平均的 4400 K,混合点落在普朗克轨迹下方;三比一时把白平衡设在 3809 K 可使亮暗两面偏差相等。 (https://5dlibrary.com/answers/zh/studio-lighting-inverse-square-area-source-colour-temperature/c06/) - Ch.7 Eastern and Western Classics Examined: the Mohist "Inverted Image Through a Point" and Lambert's Photometria — With the figure two metres away and a box 0.25 m deep, the image is -212.5 mm tall, the minus sign marking inversion, "jing dao" exactly; with a box one metre deep it is -850.0 mm. The best hole lies between 0.546 mm and 1.036 mm, with blur from about 0.614 mm to 1.419 mm and equivalent f-numbers between four hundred and a thousand, far smaller apertures than photographic lenses use. The last three lines show that with the same hole, doubling the depth of the box darkens the image by 2.00 stops, and quadrupling it by 4.00 stops. This is the inverse square at work inside the box: double the distance from hole to screen and the illuminance on the screen falls to a quarter. (https://5dlibrary.com/answers/en/studio-lighting-inverse-square-area-source-colour-temperature/c07/) 第7章 东西方原典考辨:《墨经》的「景到,在午有端」与朗伯的《光度学》 — 第一行,半球积分给出 3.14159298,圆周率为 3.14159265,两者在第七位小数上一致,余下的差别来自中点法的截断误差。这个 π 正是圆盘公式在贴近光源时的上限 πL 的来源。第二部分,受照面倾斜三十度,照度降为 0.8660,损失 0.208 档;四十五度降为 0.7071,损失 0.500 档;六十度降为 0.5000,损失 1.000 档;七十五度降为 0.2588,损失 1.950 档。这对人像布光很有用:脸的侧面相对主光倾斜得越厉害,照度损失越大,这部分明暗过渡是余弦定律的产物,与距离无关。第三部分,距离为 0.6 m 时半张角为 45.00 度,两种写法都给出 1.57080;距离为 2.4 m 时半张角为 14.04 度,两种写法都给出 0.18480,完全一致。 (https://5dlibrary.com/answers/zh/studio-lighting-inverse-square-area-source-colour-temperature/c07/) - Ch.8 Closing the Loop: StudioTaiwan Lighting Cards, the Lighting Ratio Calculator and a Pre-Registered Test — Chapter 8 gives lighting cards for a StudioTaiwan set, the distilled Lighting Ratio Calculator and a pre-registered test, stating which measurements would overturn the core claim. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/studio-lighting-inverse-square-area-source-colour-temperature/c08/) 第8章 实证闭环:StudioTaiwan 摄影棚布光卡、布光光比计算器与预注册检验 — 第八章给出 StudioTaiwan 摄影棚布光卡、蒸馏小工具「布光光比计算器」与预注册检验方案,并写明什么测量结果会推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/studio-lighting-inverse-square-area-source-colour-temperature/c08/) ## 5DL-B-326 · The Multiplicative Law of Fan-Out Tail Latency and Hedged Requests: How Percentile Latency Amplifies When One Page Makes Many Calls 中文书名:扇出尾延迟乘法律与对冲请求、单页多调用时百分位延迟的放大母典 DOI: pending Cite: Fan, Y. (2026). The Multiplicative Law of Fan-Out Tail Latency and Hedged Requests: How Percentile Latency Amplifies When One Page Makes Many Calls (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Service times, slow shares and costs are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: Website owners commonly believe that if every backend service keeps its slow requests to one per cent, the page experience cannot be too bad. This book overturns that belief with probability models. A page waits for all its calls, so its latency is their maximum, and one slow request in a hundred at each service becomes a slow-page probability of 0.634 with a hundred parallel calls. The number comes from "The Tail at Scale" by Dean and Barroso in Communications of the ACM, 2013; after checking the original, the book corrects the blueprint: it is a consequence computed under independence for a hypothetical scenario, not a measurement, and the article's own measured table does not fit the independent model. The idea of the law of the minimum, the Western classic, already appears in a paper by Sprengel in 1828, and the barrel image is a later popularisation. Service times, slow shares and costs are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Chapter 1 recomputes the multiplicative law: one slow request in ten thousand across two thousand calls still gives slow pages with probability 0.181, and a page 99th percentile allows each call a slow share of only 0.000100. Chapter 2 writes the page as a maximum and tests independence against the measured table: the independent model puts the ratio of "ninety-five per cent of leaves returned" to "all returned" at 0.087, while the measurement gives 0.500. Chapter 3 presents both directions of correlation: bad epochs lower the slow-page probability to 0.4033, fixed bad machines visited without replacement raise it to 0.6740, and synchronised background work presses it down to 0.0104. Chapter 4 covers hedged requests: sending a second copy at the single call's 95th percentile adds 0.0500 extra requests and cuts the page miss probability to 0.0100; when slowness lies in the request itself, hedging fails. Chapter 5 uses queueing theory to show how utilisation lengthens the tail and how fixed-delay hedging can push a loaded system into instability. Chapter 6 splits the page budget across calls, and splitting it unequally by cost saves 0.0597 of the capacity cost. Chapter 7 examines "a journey of a hundred li is half done at ninety" and the law of the minimum. Chapter 8 gives an offline fan-out audit for the pages of a paid web service, the distilled Fan-Out Tail Latency Calculator and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:网站经营者普遍相信,只要每个后端服务的慢请求控制在百分之一,页面体验就差不到哪里去。本书用概率模型推翻这条常识:页面要等全部调用返回,延迟是各调用的最大值,单服务百分之一的慢请求在一百个并行调用下变成 0.634 的慢页面。这个数出自迪恩与巴罗索 2013 年在《美国计算机协会通讯》发表的《规模化下的尾部》;核对原文后,本书作蓝图修正:它是假想情形下按独立性假设算出的推论,不是实测,原文的实测表反而与独立模型不符。西方原典最小因子律的思想早见于施普伦格尔 1828 年的论文,木桶比喻是后人的普及说法。书中服务时间、慢请求比例与成本均为示意参数;全部计算由书中程序生成;本书属于非个人化经营建议。 第一章复算乘法律:万分之一的慢请求在两千个调用下仍给出 0.181 的慢页面,页面第九十九百分位要求每个调用只允许 0.000100 的慢份额。第二章把页面写成最大值,并用实测表检验独立性:独立模型预测「百分之九十五叶节点返回」与「全部返回」的时刻之比为 0.087,实测为 0.500。第三章并陈相关性的两个方向:坏时段让慢页面降到 0.4033,固定坏机器配合不放回访问让它升到 0.6740,同步后台任务把它压到 0.0104。 第四章讲对冲请求:在单调用第九十五百分位补发一份,额外请求为 0.0500,页面超时概率降到 0.0100;若慢来自请求本身,对冲失效。第五章用排队论说明利用率把尾巴拉长,并指出固定延迟的对冲在高负载下会把系统推向不稳定。第六章把页面预算拆给各调用,按成本不等分可省下 0.0597 的容量成本。第七章考辨「行百里者半于九十」与最小因子律。第八章给出收费站页面扇出离线审计、蒸馏小工具「扇出尾延迟计算器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Multiplicative Law: How One Slow Call in a Hundred Becomes Sixty-Three Slow Pages in a Hundred — Chapter 1 recomputes the multiplicative law: one slow request in ten thousand across two thousand calls still gives slow pages with probability 0.181, and a page 99th percentile allows each call a slow share of only 0.000100. (https://5dlibrary.com/answers/en/fan-out-tail-latency-hedged-requests-percentile-amplification/c01/) 第1章 乘法律:百分之一的慢调用如何变成六成三的慢页面 — 第一章复算乘法律:万分之一的慢请求在两千个调用下仍给出 0.181 的慢页面,页面第九十九百分位要求每个调用只允许 0.000100 的慢份额。 (https://5dlibrary.com/answers/zh/fan-out-tail-latency-hedged-requests-percentile-amplification/c01/) - Ch.2 The Page Is a Maximum: Order Statistics and the Last Few Leaves — Chapter 2 writes the page as a maximum and tests independence against the measured table: the independent model puts the ratio of "ninety-five per cent of leaves returned" to "all returned" at 0.087, while the measurement gives 0.500. (https://5dlibrary.com/answers/en/fan-out-tail-latency-hedged-requests-percentile-amplification/c02/) 第2章 页面是最大值:次序统计量与最后几片叶子 — 第二章把页面写成最大值,并用实测表检验独立性:独立模型预测「百分之九十五叶节点返回」与「全部返回」的时刻之比为 0.087,实测为 0.500。 (https://5dlibrary.com/answers/zh/fan-out-tail-latency-hedged-requests-percentile-amplification/c02/) - Ch.3 When Independence Fails: Correlated Slowness, Bad Machines and Synchronised Background Work — Chapter 3 presents both directions of correlation: bad epochs lower the slow-page probability to 0.4033, fixed bad machines visited without replacement raise it to 0.6740, and synchronised background work presses it down to 0.0104. (https://5dlibrary.com/answers/en/fan-out-tail-latency-hedged-requests-percentile-amplification/c03/) 第3章 独立性不成立时:相关慢、坏机器与同步后台任务 — 第三章并陈相关性的两个方向:坏时段让慢页面降到 0.4033,固定坏机器配合不放回访问让它升到 0.6740,同步后台任务把它压到 0.0104。 (https://5dlibrary.com/answers/zh/fan-out-tail-latency-hedged-requests-percentile-amplification/c03/) - Ch.4 Hedged Requests: Sending a Second Copy at the 95th Percentile — Chapter 4 covers hedged requests: sending a second copy at the single call's 95th percentile adds 0.0500 extra requests and cuts the page miss probability to 0.0100; when slowness lies in the request itself, hedging fails. (https://5dlibrary.com/answers/en/fan-out-tail-latency-hedged-requests-percentile-amplification/c04/) 第4章 对冲请求:在第九十五百分位补发一份 — 第四章讲对冲请求:在单调用第九十五百分位补发一份,额外请求为 0.0500,页面超时概率降到 0.0100;若慢来自请求本身,对冲失效。 (https://5dlibrary.com/answers/zh/fan-out-tail-latency-hedged-requests-percentile-amplification/c04/) - Ch.5 Queueing Tails: Utilisation Stretches the Tail, and Hedges Are Load Too — Consider utilisation first. At a utilisation of 0.30, a call's mean response time is 7.1 milliseconds and its 99th percentile 32.9 milliseconds; at 0.70 they rise to 16.7 and 76.8 milliseconds; at 0.90 to 50.0 and 230.3 milliseconds; at 0.95 the mean is 100.0 milliseconds and the 99th percentile 460.5 milliseconds. Going from 0.5 to 0.9 multiplies every percentile by 5.0, because every percentile is inversely proportional to one minus rho. The last few points of utilisation are the most expensive: going from 0.90 to 0.95 adds only five points of load but doubles the tail. (https://5dlibrary.com/answers/en/fan-out-tail-latency-hedged-requests-percentile-amplification/c05/) 第5章 排队尾分布:利用率把尾巴拉长,对冲也是负载 — 先看利用率的作用。利用率 0.30 时,单个调用的平均响应时间是 7.1 毫秒,第九十九百分位是 32.9 毫秒;利用率 0.70 时,分别升到 16.7 与 76.8 毫秒;利用率 0.90 时,升到 50.0 与 230.3 毫秒;利用率 0.95 时,平均是 100.0 毫秒,第九十九百分位是 460.5 毫秒。从 0.5 到 0.9,每一个百分位都乘以 5.0,因为所有百分位都与「一减 rho」成反比。利用率最后那几个百分点最贵:从 0.90 到 0.95 只多用了五个百分点,尾部却翻了一倍。 (https://5dlibrary.com/answers/zh/fan-out-tail-latency-hedged-requests-percentile-amplification/c05/) - Ch.6 Budget Allocation: Splitting a Page Percentile Across Every Call — Chapter 6 splits the page budget across calls, and splitting it unequally by cost saves 0.0597 of the capacity cost. (https://5dlibrary.com/answers/en/fan-out-tail-latency-hedged-requests-percentile-amplification/c06/) 第6章 预算分配:把页面百分位拆到每一个调用 — 第六章把页面预算拆给各调用,按成本不等分可省下 0.0597 的容量成本。 (https://5dlibrary.com/answers/zh/fan-out-tail-latency-hedged-requests-percentile-amplification/c06/) - Ch.7 The Classics East and West: Half the Journey at Ninety, and the Law of the Minimum — The result is an almost literal confirmation. In the light-tailed model with a hundred leaves, the ninetieth leaf returns on average at 8.78 milliseconds and the last at 17.56 milliseconds, a ratio of exactly 0.500: at ninety, only half the journey is done. The last ten leaves take 0.500 of the page's mean waiting time. When the first leaf returns only 0.116 of the waiting is done; at the fiftieth, only 0.231; at the ninety-fifth, 0.610; at the ninety-ninth, 0.829. Waiting time is not spread evenly across leaves; each further leaf is dearer than the one before. (https://5dlibrary.com/answers/en/fan-out-tail-latency-hedged-requests-percentile-amplification/c07/) 第7章 东西方原典考辨:行百里者半于九十与最小因子律 — 结果几乎是逐字的印证。在一百个叶节点的轻尾模型里,第九十个叶节点返回的期望时刻是 8.78 毫秒,全部返回的期望时刻是 17.56 毫秒,两者之比恰好是 0.500:走到九十,只走了一半。最后十个叶节点,占去了页面平均等待时间的 0.500。第一个叶节点返回时,只走完了 0.116;第五十个返回时,也只走完了 0.231;到第九十五个,是 0.610;到第九十九个,是 0.829。等待的时间不是均匀地分布在各个叶节点之间,而是越往后越贵。 (https://5dlibrary.com/answers/zh/fan-out-tail-latency-hedged-requests-percentile-amplification/c07/) - Ch.8 A Fan-Out Audit of a Paid Web Service, and a Plan for Refutation — Chapter 8 gives an offline fan-out audit for the pages of a paid web service, the distilled Fan-Out Tail Latency Calculator and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/fan-out-tail-latency-hedged-requests-percentile-amplification/c08/) 第8章 收费站页面扇出审计与证伪方案 — 第八章给出收费站页面扇出离线审计、蒸馏小工具「扇出尾延迟计算器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/fan-out-tail-latency-hedged-requests-percentile-amplification/c08/) ## 5DL-B-327 · Cache Hit Ratios and the Che Approximation: Optimal Capacity for Caching Computed Charts 中文书名:缓存命中率与车氏近似、排盘结果缓存容量的最优边界母典 DOI: pending Cite: Fan, Y. (2026). Cache Hit Ratios and the Che Approximation: Optimal Capacity for Caching Computed Charts (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Under Zipf requests the hit ratio of a least-recently-used (LRU) cache grows slowly with capacity: with illustrative parameters, a hundred-and-twenty-eight-fold increase in capacity raises the hit ratio only 6.80 times. Chapter 4 derives the optimal capacity from slope and cost, 4187 slots or 0.2094 of the catalogue under illustrative parameters, with a flat cost curve around the optimum. Abstract: Owners of pay-per-computation services commonly believe that doubling a cache doubles its hit ratio. This book overturns that instinct. Under Zipf requests the hit ratio of a least-recently-used (LRU) cache grows slowly with capacity: with illustrative parameters, a hundred-and-twenty-eight-fold increase in capacity raises the hit ratio only 6.80 times. The Che approximation, using nothing but request probabilities and capacity, computes the LRU hit ratio in advance, with a largest gap of 0.0013 from request-by-request simulation. After checking the sources the book corrects the project blueprint: the 2002 paper by Che, Tung and Wang concerns hierarchical caching, with the characteristic time as its modelling device; slow growth of hit ratio with capacity is more directly due to Breslau and colleagues in 1999; and the accuracy of the approximation is explained by Fagin in 1977 and by Fricker and colleagues in 2012. Parameters are illustrative; all calculations are generated by the book's programs; the advice is read-only and touches no server; "chart" is only the name of a computed result, and no personal reading of any kind is made; the book is general business analysis, not individualised business advice. Chapter 1 shows the doubling illusion and confronts the Che approximation with simulation. Chapter 2 solves for the characteristic time by bisection and tests accuracy with batch-means intervals and the exact solution of a small catalogue. Chapter 3 takes Belady's optimal replacement as a lower bound on misses and the static top C as an upper bound on hits, and finds that LRU needs 4.48 times the slots for not knowing popularity. Chapter 4 derives the optimal capacity from slope and cost, 4187 slots or 0.2094 of the catalogue under illustrative parameters, with a flat cost curve around the optimum. Chapter 5 shows that under popularity drift the stationary approximation can only overestimate, that fast turnover can cut the hit ratio from 0.4483 to 0.3488, that frozen lists age quickly, and that refill after a flush runs on the characteristic time. Chapter 6 shows that multi-node caches must route by key, that blind replication knocks the hit ratio back to the level of a single node, and that tiering adds no hits. Chapter 7 examines "a state without a reserve of nine years is said to be short" from the Royal Regulations of the Book of Rites alongside Belady's optimal replacement of 1966, uses a granary model to show that spoilage decides whether a nine-year reserve is reachable, and confirms optimal replacement by exhaustive search. Chapter 8 gives a read-only planning workflow for a toll booth's chart-result cache, the distilled Cache Hit Ratio Estimator and a pre-registered test, and flags for drift any capacity whose characteristic time reaches 9.58 days. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:按次计算的服务经营者普遍相信,缓存容量翻倍,命中率也会跟着翻倍。本书用标准库程序推翻这条直觉:在齐普夫请求下,最近最少使用(LRU)缓存的命中率随容量增长得很慢,示意参数下容量放大一百二十八倍,命中率只放大 6.80 倍;而车氏近似只凭请求概率与容量,就能事先算出 LRU 命中率,与逐条模拟的最大差距为 0.0013。核对出处后,本书作蓝图修正:车皓、董烨与王志军 2002 年的论文主题是分层缓存,特征时间是其建模技术;「命中率随容量缓慢增长」的直接出处是布雷斯劳等人 1999 年的研究;近似为何准确,引费金 1977 年与弗里克等人 2012 年的论证。书中参数为示意;全部计算由书中程序生成;本书给出只读的容量建议,不访问任何服务器、不写任何配置;「排盘」仅作一种计算结果的名称,不做任何命理推断;属于非个人化经营建议。 第一章展示翻倍的错觉,并让车氏近似与模拟对质。第二章用二分法求特征时间,以批均值区间与小目录的精确解检验精度。第三章以贝拉迪最优替换为下界、静态前 C 名为上界,算出 LRU 为不知道流行度要多付 4.48 倍槽位。第四章由斜率与成本求出最优容量,示意参数下为 4187 个槽位、占目录 0.2094,且最优点附近成本平坦。第五章说明流行度漂移时稳态近似只会高估,快速换人可使命中率从 0.4483 跌到 0.3488,冻结名单迅速老化,清空后的回填按特征时间计时。 第六章说明多节点必须按键路由,盲目复制会把命中率打回单节点水平,分层不增加命中。第七章考辨《礼记·王制》「国无九年之蓄曰不足」与贝拉迪 1966 年的最优替换,用储粮模型显示损耗率决定九年之蓄能否达到,并以穷举证实最优替换。第八章给出收费站排盘结果缓存的只读规划流程、蒸馏小工具「缓存命中率估算器」与预注册检验方案,特征时间达 9.58 天的容量须标出漂移提示。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Doubling Illusion: Hit-Ratio Curves under Zipf Requests — At the four capacities the simulated hit ratios are 0.1380, 0.2624, 0.4496 and 0.7374; the Che approximation gives 0.1373, 0.2613, 0.4483 and 0.7367. The gaps are -0.0007, -0.0010, -0.0013 and -0.0007, and the largest absolute gap is only 0.0013. The approximation has no parameter fitted to the simulation; it uses only the request probabilities and the capacity. For a capacity-planning problem this error is far below the error in estimating the request probabilities themselves; Chapter 8 shows that a slightly misestimated Zipf exponent alone costs an order of magnitude more. (https://5dlibrary.com/answers/en/cache-hit-ratio-che-approximation-chart-cache-capacity/c01/) 第1章 翻倍的错觉:齐普夫请求下的命中率曲线 — 四个容量下,模拟命中率分别是 0.1380、0.2624、0.4496 与 0.7374,车氏近似给出的是 0.1373、0.2613、0.4483 与 0.7367,差距分别是 -0.0007、-0.0010、-0.0013 与 -0.0007,最大的绝对差距只有 0.0013。车氏近似没有任何从模拟里拟合出来的参数,它只用了请求概率和容量两样输入。对一个容量规划问题来说,这样的误差已经远小于请求概率本身的估计误差:第八章会看到,单是齐普夫指数估得偏一点,带来的误差就比这里大一个数量级。 (https://5dlibrary.com/answers/zh/cache-hit-ratio-che-approximation-chart-cache-capacity/c01/) - Ch.2 The Characteristic Time: Solving the Che Approximation by Bisection, and How Accurate It Is — At all three capacities the Che value falls inside the interval. At two hundred and fifty slots the simulated mean is 0.1930, the interval runs from 0.1921 to 0.1940, and the approximation gives 0.1931. At one thousand the mean is 0.3448, the interval runs from 0.3428 to 0.3468, and the approximation gives 0.3450. At four thousand the mean is 0.5769, the interval runs from 0.5749 to 0.5789, and the approximation gives 0.5767. The standard deviation between batches lies between 0.0012 and 0.0024. Even under perfectly stationary requests, then, a hit ratio measured over a few tens of thousands of requests moves by a few thousandths; before comparing it with a forecast, that wobble must be acknowledged. (https://5dlibrary.com/answers/en/cache-hit-ratio-che-approximation-chart-cache-capacity/c02/) 第2章 特征时间:车氏近似的二分求解与精度 — 三个容量下,车氏近似都落在区间之内。容量二百五十时,模拟均值是 0.1930,区间是从 0.1921 到 0.1940,车氏近似给出 0.1931;容量一千时,均值 0.3448,区间从 0.3428 到 0.3468,近似值 0.3450;容量四千时,均值 0.5769,区间从 0.5749 到 0.5789,近似值 0.5767。批与批之间的标准差在 0.0012 到 0.0024 之间,这说明即使在完全稳定的请求下,一天几万次请求测出来的命中率本身就会上下晃动千分之几;拿它与预测对比时,要先承认这份晃动。 (https://5dlibrary.com/answers/zh/cache-hit-ratio-che-approximation-chart-cache-capacity/c02/) - Ch.3 Belady's Bound: Optimal Replacement, the Static Bound and the Capacity Price of LRU — Chapter 3 takes Belady's optimal replacement as a lower bound on misses and the static top C as an upper bound on hits, and finds that LRU needs 4.48 times the slots for not knowing popularity. (https://5dlibrary.com/answers/en/cache-hit-ratio-che-approximation-chart-cache-capacity/c03/) 第3章 贝拉迪下界:最优替换、静态上界与 LRU 的容量代价 — 第三章以贝拉迪最优替换为下界、静态前 C 名为上界,算出 LRU 为不知道流行度要多付 4.48 倍槽位。 (https://5dlibrary.com/answers/zh/cache-hit-ratio-che-approximation-chart-cache-capacity/c03/) - Ch.4 Marginal Returns: the Slope of the Capacity Curve and the Cost-Optimal Boundary — Chapter 4 derives the optimal capacity from slope and cost, 4187 slots or 0.2094 of the catalogue under illustrative parameters, with a flat cost curve around the optimum. (https://5dlibrary.com/answers/en/cache-hit-ratio-che-approximation-chart-cache-capacity/c04/) 第4章 边际收益:容量与命中率曲线的斜率和成本最优边界 — 第四章由斜率与成本求出最优容量,示意参数下为 4187 个槽位、占目录 0.2094,且最优点附近成本平坦。 (https://5dlibrary.com/answers/zh/cache-hit-ratio-che-approximation-chart-cache-capacity/c04/) - Ch.5 Non-Stationary Requests: Popularity Drift, Frozen Lists and Cold Starts — Chapter 5 shows that under popularity drift the stationary approximation can only overestimate, that fast turnover can cut the hit ratio from 0.4483 to 0.3488, that frozen lists age quickly, and that refill after a flush runs on the characteristic time. (https://5dlibrary.com/answers/en/cache-hit-ratio-che-approximation-chart-cache-capacity/c05/) 第5章 非平稳请求:流行度漂移、冻结名单与冷启动 — 第五章说明流行度漂移时稳态近似只会高估,快速换人可使命中率从 0.4483 跌到 0.3488,冻结名单迅速老化,清空后的回填按特征时间计时。 (https://5dlibrary.com/answers/zh/cache-hit-ratio-che-approximation-chart-cache-capacity/c05/) - Ch.6 Partitions and Tiers: the Same Slots, Arranged Differently — Then the Che column. The front-level hit ratios, 0.1931, 0.2613, 0.3450 and 0.4027, match a single cache of the same size exactly; this level is fine. The trouble lies at the back. The approximation for both levels together gives 0.4319, 0.4190, 0.4177 and 0.4309, systematically above the simulation, by about six points at the even split. The reason is that the front level's miss stream is not a stream of independent references. An object that has just missed the front is now in the front, so for a while its requests hit there and never reach the back. Requests reaching the back are thus spread out in time, with the requests for any one object pulled apart. Treating them as independent references, the Che approximation overstates the back level's chance to hit. (https://5dlibrary.com/answers/en/cache-hit-ratio-che-approximation-chart-cache-capacity/c06/) 第6章 分片与分层:同样的槽位,不同的摆法 — 再看车氏近似一栏。前一级的命中率 0.1931、0.2613、0.3450、0.4027,与单一缓存在同样容量下的值完全一致,这一级没有问题。问题出在后一级:两级合计的近似值是 0.4319、0.4190、0.4177 与 0.4309,比模拟值系统地偏高,对半分时偏高约六个百分点。原因在于,第一级的缺失流并不是独立引用:一个对象刚刚在第一级缺失,它就被放进了第一级,接下来一段时间里它的请求都会在第一级命中,不会再到达第二级;也就是说,到达第二级的请求在时间上是「打散」过的,同一对象的请求被拉开了间隔。车氏近似把它当作独立引用,就高估了第二级的命中机会。 (https://5dlibrary.com/answers/zh/cache-hit-ratio-che-approximation-chart-cache-capacity/c06/) - Ch.7 Eastern and Western Classics: the Nine-Year Reserve of the Royal Regulations and Belady's Optimal Replacement — Chapter 7 examines "a state without a reserve of nine years is said to be short" from the Royal Regulations of the Book of Rites alongside Belady's optimal replacement of 1966, uses a granary model to show that spoilage decides whether a nine-year reserve is reachable, and confirms optimal replacement by exhaustive search. (https://5dlibrary.com/answers/en/cache-hit-ratio-che-approximation-chart-cache-capacity/c07/) 第7章 东西方原典考辨:《礼记·王制》的九年之蓄与贝拉迪的最优替换 — 第七章考辨《礼记·王制》「国无九年之蓄曰不足」与贝拉迪 1966 年的最优替换,用储粮模型显示损耗率决定九年之蓄能否达到,并以穷举证实最优替换。 (https://5dlibrary.com/answers/zh/cache-hit-ratio-che-approximation-chart-cache-capacity/c07/) - Ch.8 Closing the Loop: Read-Only Capacity Planning for a Toll Booth's Chart-Result Cache — Chapter 8 gives a read-only planning workflow for a toll booth's chart-result cache, the distilled Cache Hit Ratio Estimator and a pre-registered test, and flags for drift any capacity whose characteristic time reaches 9.58 days. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cache-hit-ratio-che-approximation-chart-cache-capacity/c08/) 第8章 实证闭环:收费站排盘结果缓存的只读容量规划 — 第八章给出收费站排盘结果缓存的只读规划流程、蒸馏小工具「缓存命中率估算器」与预注册检验方案,特征时间达 9.58 天的容量须标出漂移提示。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cache-hit-ratio-che-approximation-chart-cache-capacity/c08/) ## 5DL-B-328 · Low-Rank Matrix Completion and Alternating Least Squares: Inferring Unstated Preferences from Sparse Purchase Records 中文书名:低秩矩阵补全与交替最小二乘、稀疏购买记录推断未表态偏好母典 DOI: pending Cite: Fan, Y. (2026). Low-Rank Matrix Completion and Alternating Least Squares: Inferring Unstated Preferences from Sparse Purchase Records (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. Abstract: Small-shop owners and buyers commonly believe that purchase records are too sparse to infer the preferences customers have never stated, and that recommendation needs the vast data of a large platform. This book overturns that belief with low-rank matrix completion: if tastes are governed by a few hidden dimensions, the empty cells of a table that is more than nine-tenths empty are mathematically determined, and the observations needed depend on the degrees of freedom times a logarithmic factor, not on the number of cells. The project blueprint attributed "about O(n r log n) observations suffice" to Candès and Recht's 2009 paper in Foundations of Computational Mathematics; after checking, the book corrects the blueprint: that paper proved a sufficient condition of order n to the 6/5 times r log n, while n r log n is the necessary floor proved by Candès and Tao in 2010 and reached under stronger assumptions at the cost of extra logarithmic factors. Customer and product counts, taste dimensions and purchase probabilities are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice; its workflow uses only a shop's own data. Chapter 1 sweeps the observed share: a 40 by 40 matrix of rank two reaches a median error of 0.0021 when observations are 2.05 times the degrees of freedom, and exact binomial tails show the ratio of the observations needed to n r ln n falling slowly from 0.677 to 0.436, so the logarithmic factor cannot be removed. Chapter 2 writes alternating least squares by hand; its objective never rises, and ridge and rank must be chosen together. Chapter 3 checks the Eckart-Young error by power iteration and deflation; a spectral start brings the error of alternating least squares to 0.00108, while a random start drives it to 12.23857. Chapter 4 shows that recovery fails without incoherence: a small niche has a median error of 0.9134. Chapter 5 asks what happens when only purchases are seen: under strong selection the unbought cells are predicted 0.4245 too high, and propensity weights cut this to 0.1671; a model fitted to bought cells alone hits only 0.286, while weighted alternating least squares that counts the unbought cells reaches 0.743. Chapter 6 treats cold start: a new customer with eight records reaches an error of 0.2434 by folding in with shrinkage toward the category prior. Chapter 7 examines "things of a kind gather together, creatures divide into groups" from the Great Commentary of the Book of Changes and Spearman's general factor of 1904, showing that pure groups and mixed membership each favour a different model. Chapter 8 gives the buyer's recommendation sheet, the distilled Preference Completion Recommender and a pre-registered test needing 3156 customers per arm. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. 摘要:小店店主与买手普遍相信,购买记录太稀疏,推断不了顾客没有表态的偏好,要做推荐必须先有大平台那样的海量数据。本书用低秩矩阵补全推翻这条常识:顾客口味若由少数隐藏维度决定,一张九成以上是空格的表,其空格在数学上可以被确定,所需观测数取决于自由度再乘一个对数因子,而不是格子总数。立项蓝图把「约 O(n r log n) 个观测即可恢复」挂在坎德斯与雷希特 2009 年《计算数学基础》的论文名下;核对之后,本书作蓝图修正:该文证明的充分条件是 n 的 6/5 次方乘 r log n 的量级,n r log n 是坎德斯与陶哲轩 2010 年证明的必要下限,并在更强前提下以多出对数因子的代价达到。书中顾客数、商品数、口味维度与购买概率均为示意参数;全部计算由书中程序生成;本书属于非个人化经营建议,推荐流程只用店铺自有数据,不外传个人数据,不做跨平台追踪。 第一章扫描观测比例:40 乘 40、秩二的矩阵在观测数为自由度 2.05 倍时误差中位数降到 0.0021,并用精确二项尾概率说明所需观测与 n r ln n 之比从 0.677 缓降到 0.436,对数因子去不掉。第二章手写交替最小二乘,目标函数从不上升,岭参数与秩须一起选。第三章用幂迭代与紧缩核对埃卡特与杨的误差,谱初值使交替最小二乘误差降到 0.00108,随机初值却升到 12.23857。第四章说明非相干性不成立时恢复失败:冷门小众品项的误差中位数为 0.9134。 第五章说明只看到买了的会怎样:选择偏强时未买格子的预测偏高 0.4245,倾向加权降到 0.1671;只拟合已买格子的模型命中率仅 0.286,把没买的格子也计入的加权交替最小二乘达到 0.743。第六章讲冷启动,新顾客有八条记录时向品类先验收缩的折入法误差为 0.2434。第七章考辨《周易·系辞上》「方以类聚,物以群分」与斯皮尔曼 1904 年的一般因素,指出纯分群与混合成员下两种模型各有胜负。第八章给出买手推荐表、蒸馏小工具「偏好补全推荐器」与预注册检验方案,每组需 3156 位顾客。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Sparse Yet Completable: How Many Entries a Low-Rank Matrix Needs — Chapter 1 sweeps the observed share: a 40 by 40 matrix of rank two reaches a median error of 0.0021 when observations are 2.05 times the degrees of freedom, and exact binomial tails show the ratio of the observations needed to n r ln n falling slowly from 0.677 to 0.436, so the logarithmic factor cannot be removed. (https://5dlibrary.com/answers/en/low-rank-matrix-completion-als-sparse-purchase-preferences/c01/) 第1章 稀疏也能补全:低秩矩阵需要多少个观测条目 — 第一章扫描观测比例:40 乘 40、秩二的矩阵在观测数为自由度 2.05 倍时误差中位数降到 0.0021,并用精确二项尾概率说明所需观测与 n r ln n 之比从 0.677 缓降到 0.436,对数因子去不掉。 (https://5dlibrary.com/answers/zh/low-rank-matrix-completion-als-sparse-purchase-preferences/c01/) - Ch.2 Alternating Least Squares: Breaking a Hard Problem into a Chain of Small Systems — There are 343 observed cells. With a ridge parameter of 0.001 the training error is 0.3886 but the test error is as high as 4.2879: with almost no penalty, customers and products with few records are fitted into the noise, and predictions drift far off on the unobserved cells. This row's training error is even higher than the 0.3361 obtained with 0.100, a sign that without a penalty alternating least squares does not converge cleanly either. The test error is lowest, 0.7649, at 1.000; beyond that the penalty overwhelms the data, training and test errors rise together, and at 30.000 they stand at 1.7089 and 1.7581, a model that has learned almost nothing. (https://5dlibrary.com/answers/en/low-rank-matrix-completion-als-sparse-purchase-preferences/c02/) 第2章 交替最小二乘:把一个难题拆成一串小方程组 — 观测格子共 343 个。岭参数取 0.001 时,训练误差是 0.3886,检验误差却高达 4.2879:几乎不加惩罚,记录少的顾客与商品被拟合到噪声里去,预测在未观测格子上大幅外飘。这一行的训练误差反而比岭参数 0.100 时的 0.3361 还高,说明几乎不加惩罚时交替最小二乘也收敛得不干净。岭参数取 1.000 时检验误差最低,是 0.7649;再往上加,惩罚压过了数据,训练与检验误差一起上升,到 30.000 时两者分别是 1.7089 与 1.7581,模型已经接近什么也没学到。 (https://5dlibrary.com/answers/zh/low-rank-matrix-completion-als-sparse-purchase-preferences/c02/) - Ch.3 Power Iteration and the Leading Singular Vector: A Good Start for Alternating Least Squares — Chapter 3 checks the Eckart-Young error by power iteration and deflation; a spectral start brings the error of alternating least squares to 0.00108, while a random start drives it to 12.23857. (https://5dlibrary.com/answers/en/low-rank-matrix-completion-als-sparse-purchase-preferences/c03/) 第3章 幂迭代与主奇异向量:给交替最小二乘找一个好起点 — 第三章用幂迭代与紧缩核对埃卡特与杨的误差,谱初值使交替最小二乘误差降到 0.00108,随机初值却升到 12.23857。 (https://5dlibrary.com/answers/zh/low-rank-matrix-completion-als-sparse-purchase-preferences/c03/) - Ch.4 Incoherence: How Best-Sellers, Small Niches and Long Tails Defeat Completion — Chapter 4 shows that recovery fails without incoherence: a small niche has a median error of 0.9134. (https://5dlibrary.com/answers/en/low-rank-matrix-completion-als-sparse-purchase-preferences/c04/) 第4章 非相干性:爆款、冷门小众与长尾让补全失败 — 第四章说明非相干性不成立时恢复失败:冷门小众品项的误差中位数为 0.9134。 (https://5dlibrary.com/answers/zh/low-rank-matrix-completion-als-sparse-purchase-preferences/c04/) - Ch.5 Missing Not at Random: How Far Seeing Only Purchases Leads Inference Astray — Chapter 5 asks what happens when only purchases are seen: under strong selection the unbought cells are predicted 0.4245 too high, and propensity weights cut this to 0.1671; a model fitted to bought cells alone hits only 0.286, while weighted alternating least squares that counts the unbought cells reaches 0.743. (https://5dlibrary.com/answers/en/low-rank-matrix-completion-als-sparse-purchase-preferences/c05/) 第5章 缺失非随机:只看到买了的,会把推断带偏多少 — 第五章说明只看到买了的会怎样:选择偏强时未买格子的预测偏高 0.4245,倾向加权降到 0.1671;只拟合已买格子的模型命中率仅 0.286,把没买的格子也计入的加权交替最小二乘达到 0.743。 (https://5dlibrary.com/answers/zh/low-rank-matrix-completion-als-sparse-purchase-preferences/c05/) - Ch.6 Cold Start: Catching New Customers and New Products with Means and Category Priors — Chapter 6 treats cold start: a new customer with eight records reaches an error of 0.2434 by folding in with shrinkage toward the category prior. (https://5dlibrary.com/answers/en/low-rank-matrix-completion-als-sparse-purchase-preferences/c06/) 第6章 冷启动:用均值与品类先验接住新顾客与新商品 — 第六章讲冷启动,新顾客有八条记录时向品类先验收缩的折入法误差为 0.2434。 (https://5dlibrary.com/answers/zh/low-rank-matrix-completion-als-sparse-purchase-preferences/c06/) - Ch.7 Eastern and Western Sources: "Things of a Kind Gather Together" and Spearman's General Factor — Chapter 7 examines "things of a kind gather together, creatures divide into groups" from the Great Commentary of the Book of Changes and Spearman's general factor of 1904, showing that pure groups and mixed membership each favour a different model. (https://5dlibrary.com/answers/en/low-rank-matrix-completion-als-sparse-purchase-preferences/c07/) 第7章 东西方原典考辨:「方以类聚,物以群分」与斯皮尔曼的一般因素 — 第七章考辨《周易·系辞上》「方以类聚,物以群分」与斯皮尔曼 1904 年的一般因素,指出纯分群与混合成员下两种模型各有胜负。 (https://5dlibrary.com/answers/zh/low-rank-matrix-completion-als-sparse-purchase-preferences/c07/) - Ch.8 The Buyer's Recommendation Sheet and a Plan to Refute the Book: A Loop on the Shop's Own Data — Chapter 8 gives the buyer's recommendation sheet, the distilled Preference Completion Recommender and a pre-registered test needing 3156 customers per arm. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/low-rank-matrix-completion-als-sparse-purchase-preferences/c08/) 第8章 买手推荐表与证伪方案:只用店内自有数据的闭环 — 第八章给出买手推荐表、蒸馏小工具「偏好补全推荐器」与预注册检验方案,每组需 3156 位顾客。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/low-rank-matrix-completion-als-sparse-purchase-preferences/c08/) ## 5DL-B-329 · Acceptance Sampling Plans and Operating-Characteristic Curves: How Many Items a Buyer Must Inspect on Arrival 中文书名:验收抽样方案与操作特性曲线、买手到货验货抽检件数母典 DOI: pending Cite: Fan, Y. (2026). Acceptance Sampling Plans and Operating-Characteristic Curves: How Many Items a Buyer Must Inspect on Arrival (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Lot sizes, defect shares and costs are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: Buyers and resellers commonly believe that inspecting a tenth of every delivery is a fair and safe rule: more for large lots, less for small. This book overturns that belief with exact probabilities. Whether inspection stops bad lots depends on the number of items sampled and the number of defectives allowed, and hardly at all on the sample's share of the lot. The project blueprint attributed this claim to Dodge and Romig's 1929 paper in the Bell System Technical Journal; after checking, the book corrects the blueprint: that paper introduced the lot tolerance, the consumer's risk and the principle of least inspection, while "protection follows the count, not the share" is a consequence of the hypergeometric distribution, and the average outgoing quality limit and the full tables appear in the authors' long article of 1941. The relevant clause of the Qin Statutes on Crafts from Shuihudi could not be checked word for word and is only summarised. Lot sizes, defect shares and costs are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Chapter 1 shows that drawing a tenth from a lot of fifty still accepts a lot with a tenth defective with probability 0.577, and that the lot tolerance falls from 0.3600 on small lots to 0.0044 on large ones; a fixed sample of fifty gives acceptance probabilities that barely change with lot size, with a large-lot limit of 0.736; over one season a fixed plan inspecting 154 items gives more even protection than the tenth rule's 477. Chapter 2 gives the exact hypergeometric operating-characteristic curve. Chapter 3 searches for the smallest plan under two risks and shows that zero acceptance wrongly rejects normal lots at one in a hundred with probability 0.3638. Chapter 4 computes average outgoing quality: one plan for lots of a thousand has a limit of 0.01625, and the plan with the least inspection at the same limit inspects 88.0 items per lot on average. Chapter 5 shows that double sampling with the same protection inspects only 103.2 items on average, 28.8 fewer than the single plan. Chapter 6 writes inspection as expected cost: the best plan costs 327.82 per lot, 71.32 less than the tenth rule, and lot size enters through cost, not proportion. Chapter 7 examines the Qin requirement that like vessels match and the protection constraint of Dodge and Romig; a gauge offset can multiply the defect share 8.4 times. Chapter 8 gives the buyer's receiving sheet, the distilled Inbound Inspection Sampling Plan Calculator and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:买手与代购店主普遍相信「到货抽一成验货」是公平而稳妥的规则:批大多看,批小少看。本书用精确概率推翻这条常识。抽检能否拦住坏批,取决于样本件数与允许的不良件数,几乎不取决于样本占整批的比例。立项蓝图把这一命题挂在道奇与罗米格 1929 年《贝尔系统技术杂志》的论文名下;核对之后,本书作蓝图修正:该文提出的是批容许不良率、消费者风险与最少检验量原则,「保护力看件数不看比例」是由超几何分布推出的结论;平均出厂质量上限及成套抽样表见于两人 1941 年的长文。东方原典睡虎地秦简《工律》的相关条文因未能逐字核对,只述其大意。书中批量、不良率与成本均为示意参数;全部计算由书中程序生成;本书属于非个人化经营建议。 第一章证明,五十件的批抽一成时,一成不良的批仍有 0.577 的概率被接收,批容许不良率从小批的 0.3600 降到大批的 0.0044;固定抽五十件时,接收概率几乎不随批量变化,大批极限为 0.736;同一季到货,固定方案只用 154 件就比抽一成的 477 件给出更均匀的保护。第二章给出超几何操作特性曲线的精确算法。第三章在两种风险约束下搜索最小方案,并指出零接收数方案让百分之一不良的正常批被误退 0.3638。第四章计算平均出厂质量,一个千件批方案的上限为 0.01625,同样上限下检验最少的方案每批平均检验 88.0 件。 第五章说明双次抽样在同样保护下平均只需检验 103.2 件,比单次方案少 28.8 件。第六章把验货写成期望成本,最优方案每批 327.82,比抽一成省 71.32,并说明批量通过成本而非比例进入方案。第七章考辨秦律「同物必等」与道奇、罗米格的保护约束,量具偏差可使不良率放大 8.4 倍。第八章给出买手到货验货表、蒸馏小工具「验货抽样方案计算器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Illusion of the Tenth: Protection Follows the Count, Not the Share — Chapter 1 shows that drawing a tenth from a lot of fifty still accepts a lot with a tenth defective with probability 0.577, and that the lot tolerance falls from 0.3600 on small lots to 0.0044 on large ones; a fixed sample of fifty gives acceptance probabilities that barely change with lot size, with a large-lot limit of 0.736; over one season a fixed plan inspecting 154 items gives more even protection than the tenth rule's 477. (https://5dlibrary.com/answers/en/acceptance-sampling-oc-curve-buyer-inbound-inspection/c01/) 第1章 抽一成的错觉:保护力看样本件数而不看比例 — 第一章证明,五十件的批抽一成时,一成不良的批仍有 0.577 的概率被接收,批容许不良率从小批的 0.3600 降到大批的 0.0044;固定抽五十件时,接收概率几乎不随批量变化,大批极限为 0.736;同一季到货,固定方案只用 154 件就比抽一成的 477 件给出更均匀的保护。 (https://5dlibrary.com/answers/zh/acceptance-sampling-oc-curve-buyer-inbound-inspection/c01/) - Ch.2 The Hypergeometric Operating-Characteristic Curve: Acceptance Probabilities Computed Item by Item — Read the table from both ends toward the middle. With no defectives the acceptance probability is of course 1.0000. At a defect share of 0.010, five defectives in five hundred, it is 0.9194, short of 0.95; to reach 0.95 the share must fall to 0.006, three defectives. At 0.020 the acceptance probability drops to 0.7365, so a supplier running steadily at two in a hundred will see roughly one lot in four rejected. At 0.050 it is 0.2636, still a better than one-in-four chance of passing; at 0.100 it is 0.0278, and only then are such lots essentially stopped. (https://5dlibrary.com/answers/en/acceptance-sampling-oc-curve-buyer-inbound-inspection/c02/) 第2章 超几何操作特性曲线:一件一件精确算出接收概率 — 读这张表要从两端往中间看。不良率为零时,接收概率当然是 1.0000。不良率为 0.010,即五百件里有五件不良,接收概率是 0.9194,并没有达到 0.95;要让接收概率至少 0.95,不良率要降到 0.006,也就是三件不良。不良率升到 0.020,接收概率降到 0.7365,这意味着一个长期稳定在百分之二不良的供应商,每四批左右就会被判退一批。不良率到 0.050,接收概率是 0.2636,仍有四分之一以上的机会放行;到 0.100,接收概率 0.0278,才算基本拦住。 (https://5dlibrary.com/answers/zh/acceptance-sampling-oc-curve-buyer-inbound-inspection/c02/) - Ch.3 Two Risks at Once: Searching for the Smallest Sample and Acceptance Number — Chapter 3 searches for the smallest plan under two risks and shows that zero acceptance wrongly rejects normal lots at one in a hundred with probability 0.3638. (https://5dlibrary.com/answers/en/acceptance-sampling-oc-curve-buyer-inbound-inspection/c03/) 第3章 两种风险的双重约束:搜索最小的样本量与接收数 — 第三章在两种风险约束下搜索最小方案,并指出零接收数方案让百分之一不良的正常批被误退 0.3638。 (https://5dlibrary.com/answers/zh/acceptance-sampling-oc-curve-buyer-inbound-inspection/c03/) - Ch.4 Average Outgoing Quality and Its Limit: What Remains after Rejected Lots Are Screened — Chapter 4 computes average outgoing quality: one plan for lots of a thousand has a limit of 0.01625, and the plan with the least inspection at the same limit inspects 88.0 items per lot on average. (https://5dlibrary.com/answers/en/acceptance-sampling-oc-curve-buyer-inbound-inspection/c04/) 第4章 平均出厂质量与其上限:被拒批全检之后还剩多少不良 — 第四章计算平均出厂质量,一个千件批方案的上限为 0.01625,同样上限下检验最少的方案每批平均检验 88.0 件。 (https://5dlibrary.com/answers/zh/acceptance-sampling-oc-curve-buyer-inbound-inspection/c04/) - Ch.5 Double Sampling and the Average Sample Number: Look at a Few, Then More If Needed — Chapter 5 shows that double sampling with the same protection inspects only 103.2 items on average, 28.8 fewer than the single plan. (https://5dlibrary.com/answers/en/acceptance-sampling-oc-curve-buyer-inbound-inspection/c05/) 第5章 双次抽样与平均样本量:先看少数,必要时再多看 — 第五章说明双次抽样在同样保护下平均只需检验 103.2 件,比单次方案少 28.8 件。 (https://5dlibrary.com/answers/zh/acceptance-sampling-oc-curve-buyer-inbound-inspection/c05/) - Ch.6 The Expected Cost of Inbound Inspection: How Many Items Are Worth Checking — Chapter 6 writes inspection as expected cost: the best plan costs 327.82 per lot, 71.32 less than the tenth rule, and lot size enters through cost, not proportion. (https://5dlibrary.com/answers/en/acceptance-sampling-oc-curve-buyer-inbound-inspection/c06/) 第6章 到货验货的期望成本:验多少件才划算 — 第六章把验货写成期望成本,最优方案每批 327.82,比抽一成省 71.32,并说明批量通过成本而非比例进入方案。 (https://5dlibrary.com/answers/zh/acceptance-sampling-oc-curve-buyer-inbound-inspection/c06/) - Ch.7 Eastern and Western Sources: the Qin Statute That Like Vessels Must Match, and the Protection Constraint of Dodge and Romig — Chapter 7 examines the Qin requirement that like vessels match and the protection constraint of Dodge and Romig; a gauge offset can multiply the defect share 8.4 times. (https://5dlibrary.com/answers/en/acceptance-sampling-oc-curve-buyer-inbound-inspection/c07/) 第7章 东西方原典考辨:秦《工律》的「同物必等」与道奇、罗米格的保护约束 — 第七章考辨秦律「同物必等」与道奇、罗米格的保护约束,量具偏差可使不良率放大 8.4 倍。 (https://5dlibrary.com/answers/zh/acceptance-sampling-oc-curve-buyer-inbound-inspection/c07/) - Ch.8 The Buyer's Receiving Sheet and a Plan to Refute the Book — Chapter 8 gives the buyer's receiving sheet, the distilled Inbound Inspection Sampling Plan Calculator and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/acceptance-sampling-oc-curve-buyer-inbound-inspection/c08/) 第8章 买手到货验货表与证伪方案 — 第八章给出买手到货验货表、蒸馏小工具「验货抽样方案计算器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/acceptance-sampling-oc-curve-buyer-inbound-inspection/c08/) ## 5DL-B-330 · First-Order Ageing Kinetics of Pu-erh Tea under Storage Temperature and Humidity: Equivalent Age and the Detection of Misstated Vintages 中文书名:普洱陈化一级动力学与仓储温湿度、等效陈化年龄与年份误报检出母典 DOI: pending Cite: Fan, Y. (2026). First-Order Ageing Kinetics of Pu-erh Tea under Storage Temperature and Humidity: Equivalent Age and the Detection of Misstated Vintages (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book makes no claim of any wellbeing or other bodily effect of tea and is not personalised investment advice. Chapter 1 gives first-order loss with an illustrative half-life of 8.66 years and shows that ten years of data cannot tell curve shapes apart. Abstract: The pu-erh trade holds that "the older, the more valuable". This book tests only the chemical part of the saying and finds that it is not monotonic: the loss of ester catechins in storage is approximately first order, its rate varies with temperature through the Arrhenius relation, humidity acts in addition, a hot and humid store can produce in a few years the chemical state of a dozen or more dry years, and the prized "aged character" rises and then falls. After checking the sources, the book corrects the blueprint's statement that the rate is set by temperature and humidity through the Arrhenius relation: Arrhenius governs temperature only, humidity acts through moisture and microbes, microbe-driven change departs from Arrhenius, and the first-order model is an approximation set beside counter-examples. Its parameters are illustrative; all calculations are generated by its programs; the book makes no claim of any wellbeing or other bodily effect of tea and is not personalised investment advice. Chapter 1 gives first-order loss with an illustrative half-life of 8.66 years and shows that ten years of data cannot tell curve shapes apart. Chapter 2 uses consecutive reactions to show that when cakes aged one to thirty years in one store are compared in pairs, 0.653 of the pairs have the older cake less aged. Chapter 3, at the middle case of an activation energy of 60 and a humidity slope of 0.030, finds that a wet store at 30 °C and 80 % relative humidity runs 4.10 times as fast, matching fifteen dry years in 3.65 years. Chapter 4 shows that seasonal swings speed a store up: ten years in a wet coastal store equal 32.0 reference years, and a cake kept four years wet is read on a dry-store assumption as 18.31 years old. Chapter 5 uses two markers with different activation energies to separate temperature from time in part, and explains why humidity stays tied to time. Chapter 6 gives a likelihood-ratio test of claimed vintages: without a retained sample of the batch the power against wet tea passed off as dry is only 0.488, and in a market where misstatement runs at 0.05 a flag has a positive predictive value of just 0.339, while with a retained sample the power rises to 1.000. Chapter 7 examines the yu box of Lu Yu's Tea Classic and Arrhenius's paper of 1889; humidity control in the manner of the yu saves 4.98 equivalent years over ten years. Chapter 8 gives the routine for a storage ledger, the distilled Equivalent Ageing Age Calculator and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:普洱交易里流行「越陈越值钱」。本书只检验其中的化学部分,结论是它并不单调:贮藏中酯型儿茶素的减少近似一级动力学,速率随温度按阿伦尼乌斯关系变化,湿度另有作用,高温高湿的仓可以在几年内做出干仓十几年的化学状态,而被看重的「陈韵」先升后降。核对原文后,本书把立项蓝图「速率由温湿度按阿伦尼乌斯关系决定」修正为:阿伦尼乌斯只管温度,湿度通过含水量与微生物起作用,微生物驱动的部分偏离阿伦尼乌斯,一级模型只是近似并与反例并陈。书中参数为示意;全部计算由书中程序生成;本书不谈饮茶的任何保健或其他身体作用,也不是个人化投资建议。 第一章给出一级衰减,示意半衰期 8.66 年,并说明十年数据分不清曲线形状。第二章用连串反应说明,同一仓中一至三十年的两饼茶两两比年份,有 0.653 的配对「更老的反而更不陈」。第三章在活化能 60、湿度系数 0.030 的中间情形下算出 30 °C、相对湿度 80 % 的湿仓倍数为 4.10,追平干仓十五年只需 3.65 年。第四章说明季节波动让仓变快,沿海湿仓十年相当于参照仓 32.0 年,湿仓四年的茶会被按干仓推断为 18.31 年。 第五章用两个活化能不同的标志物把温度与时间部分分开,并说明湿度仍与时间绑在一起。第六章给出年份声称的似然比检验:没有同批留样时,对「湿仓冒充干仓」的检出功效只有 0.488,在误报占 0.05 的市场里一次报警的阳性预测值仅 0.339,有留样后功效升至 1.000。第七章考辨陆羽《茶经》的「育」与阿伦尼乌斯 1889 年的论文,按「育」控湿十年可少 4.98 个等效年。第八章给出仓储账流程、蒸馏小工具「等效陈化年龄计算器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 First-Order Kinetics: the Loss of Ester Catechins and the Gain of Gallic Acid — Chapter 1 gives first-order loss with an illustrative half-life of 8.66 years and shows that ten years of data cannot tell curve shapes apart. (https://5dlibrary.com/answers/en/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c01/) 第1章 一级动力学:酯型儿茶素的衰减与没食子酸的累积 — 第一章给出一级衰减,示意半衰期 8.66 年,并说明十年数据分不清曲线形状。 (https://5dlibrary.com/answers/zh/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c01/) - Ch.2 Consecutive Reactions and the Peak of Aged Character: Why "Older Is Better" Is Not Monotonic in Chemistry — Chapter 2 uses consecutive reactions to show that when cakes aged one to thirty years in one store are compared in pairs, 0.653 of the pairs have the older cake less aged. (https://5dlibrary.com/answers/en/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c02/) 第2章 连串反应与陈香峰值:「越陈越值钱」为什么在化学上不单调 — 第二章用连串反应说明,同一仓中一至三十年的两饼茶两两比年份,有 0.653 的配对「更老的反而更不陈」。 (https://5dlibrary.com/answers/zh/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c02/) - Ch.3 Arrhenius and the Wet Store: How Many Dry Years a Few Hot, Humid Years Are Worth — Chapter 3, at the middle case of an activation energy of 60 and a humidity slope of 0.030, finds that a wet store at 30 °C and 80 % relative humidity runs 4.10 times as fast, matching fifteen dry years in 3.65 years. (https://5dlibrary.com/answers/en/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c03/) 第3章 阿伦尼乌斯与湿热仓:几年高温高湿抵得上多少年干仓 — 第三章在活化能 60、湿度系数 0.030 的中间情形下算出 30 °C、相对湿度 80 % 的湿仓倍数为 4.10,追平干仓十五年只需 3.65 年。 (https://5dlibrary.com/answers/zh/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c03/) - Ch.4 Store Logs and Equivalent Age: Seasonal Swings, Humidity Corrections and Storage Histories — Chapter 4 shows that seasonal swings speed a store up: ten years in a wet coastal store equal 32.0 reference years, and a cake kept four years wet is read on a dry-store assumption as 18.31 years old. (https://5dlibrary.com/answers/en/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c04/) 第4章 温湿度记录与等效年龄:季节波动、湿度修正与仓储史 — 第四章说明季节波动让仓变快,沿海湿仓十年相当于参照仓 32.0 年,湿仓四年的茶会被按干仓推断为 18.31 年。 (https://5dlibrary.com/answers/zh/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c04/) - Ch.5 Two-Marker Inversion: Separating Time from Heat — With a gap of only 10, the inverted temperature ranges from 17.6 °C to 34.8 °C, almost the whole span from a cold store to a wet one, and the age from 3.78 to 17.03 years, a width of 13.25 years: a useless result in trade. With a gap of 20, the temperature narrows to 21.7 to 30.4 °C and the age to 5.33 to 12.09, a width of 6.76 years. With 40, the temperature lies between 23.8 and 28.2 °C and the age between 6.26 and 10.12, a width of 3.86 years. With gaps of 60 and 80 the age intervals are 2.98 and 2.47 years wide, and the temperature interval shrinks to within two or three degrees. (https://5dlibrary.com/answers/en/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c05/) 第5章 双标志物反演:把时间与热量分开 — 活化能之差只有 10 时,反演温度的区间从 17.6 °C 到 34.8 °C,几乎覆盖了从冷仓到湿仓的全部范围;年份区间从 3.78 到 17.03 年,宽度 13.25 年,这样的结果在交易中毫无用处。差到 20 时,温度区间收窄到 21.7 至 30.4 °C,年份区间 5.33 至 12.09,宽度 6.76 年。差到 40 时,温度 23.8 至 28.2 °C,年份 6.26 至 10.12,宽度 3.86 年。差到 60 与 80 时,年份区间宽度分别是 2.98 与 2.47 年,温度区间收窄到两三度之内。 (https://5dlibrary.com/answers/zh/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c05/) - Ch.6 Detecting Misstated Vintages: Likelihood Ratios, Intervals and Base Rates — Chapter 6 gives a likelihood-ratio test of claimed vintages: without a retained sample of the batch the power against wet tea passed off as dry is only 0.488, and in a market where misstatement runs at 0.05 a flag has a positive predictive value of just 0.339, while with a retained sample the power rises to 1.000. (https://5dlibrary.com/answers/en/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c06/) 第6章 年份误报检出:似然比、区间与基础比率 — 第六章给出年份声称的似然比检验:没有同批留样时,对「湿仓冒充干仓」的检出功效只有 0.488,在误报占 0.05 的市场里一次报警的阳性预测值仅 0.339,有留样后功效升至 1.000。 (https://5dlibrary.com/answers/zh/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c06/) - Ch.7 Eastern and Western Sources Examined: Lu Yu's Tea Classic and Arrhenius 1889 — Chapter 7 examines the yu box of Lu Yu's Tea Classic and Arrhenius's paper of 1889; humidity control in the manner of the yu saves 4.98 equivalent years over ten years. (https://5dlibrary.com/answers/en/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c07/) 第7章 东西方原典考辨:陆羽《茶经》与阿伦尼乌斯 1889 — 第七章考辨陆羽《茶经》的「育」与阿伦尼乌斯 1889 年的论文,按「育」控湿十年可少 4.98 个等效年。 (https://5dlibrary.com/answers/zh/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c07/) - Ch.8 Closing the Loop: Storage Logs for Collected Tea and a Pre-Registered Test — Chapter 8 gives the routine for a storage ledger, the distilled Equivalent Ageing Age Calculator and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c08/) 第8章 实证闭环:收藏茶仓储温湿度记录与预注册检验 — 第八章给出仓储账流程、蒸馏小工具「等效陈化年龄计算器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/puerh-ageing-first-order-kinetics-equivalent-age-vintage/c08/) ## 5DL-B-331 · Error Bars in Thermoluminescence and Radiometric Dating: Multiple Solutions on Calibration Plateaus and When to Send Antiques for Testing 中文书名:热释光与放射性断代误差棒、校准曲线平台多解与古玩送检边界母典 DOI: pending Cite: Fan, Y. (2026). Error Bars in Thermoluminescence and Radiometric Dating: Multiple Solutions on Calibration Plateaus and When to Send Antiques for Testing (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Objects, prices and parameters are illustrative, the calibration curve is a hand-set illustrative line rather than IntCal data, and all calculations are generated by the book's programs. Chapter 6 writes testing as a problem of expected loss; under illustrative parameters a buyer should not buy untested below a prior of 0.655, and a test is worth most at middling priors. Abstract: The antiques trade widely believes that a laboratory test returns an exact year. This book writes that belief as runnable arithmetic and tests it: a date is a probability distribution with error bars; in the plateau periods of the radiocarbon calibration curve one measurement maps onto several separate historical ranges; and porcelain bodies, lacking datable organic carbon, can be dated only by thermoluminescence and related methods whose zero is the last heating. After checking the originals the book corrects its project blueprint: the 1949 Science note of Libby, Anderson and Arnold tested the global uniformity of natural radiocarbon, while the check against samples of known age is a second note by Arnold and Libby the same year; and "porcelain can be dated only by thermoluminescence" is narrowed to "the firing of a porcelain body can be dated only by luminescence, which cannot separate adjacent reigns and does not see assembly or reheating". Objects, prices and parameters are illustrative, the calibration curve is a hand-set illustrative line rather than IntCal data, and all calculations are generated by the book's programs. The book reaches no conclusion about any particular object, gives no authentication opinion, and is not personalised investment, legal, tax or financial advice. Chapter 1 treats exponential decay and the half-life conventions, with a Cambridge-to-Libby ratio of 1.0291, and shows how counting error passes into age. Chapter 2 demonstrates a plateau on the illustrative curve: with the same error of plus or minus 25 years, a plateau measurement calibrates to two ranges with probabilities 0.627 and 0.328, and four times the precision shrinks the plateau width only to 0.79 of what it was. Chapter 3 covers highest-density ranges, consistency tests before pooling and independent prior windows. Chapter 4 sets the thermoluminescence age equal to accumulated dose over annual dose, with a relative error of about 0.089 for excavated pieces rising to 0.123 without a burial context. Chapter 5 matches materials to clocks and shows that thermoluminescence cleanly separates new copies from old pieces but can hardly separate two periods a century apart: at a relative error of 0.25, the ninety-five per cent range excludes the adjacent period with probability only 0.109. Chapter 6 writes testing as a problem of expected loss; under illustrative parameters a buyer should not buy untested below a prior of 0.655, and a test is worth most at middling priors. Chapter 7 examines Zhao Xigu's Dongtian Qinglu and the 1949 notes of Libby and colleagues, modelling crackle on qin lacquer as a clock with an uncertain rate, under which "crackle means over five hundred years" is right with probability 0.835. Chapter 8 gives a decision table for testing antiques, the distilled Dating Error-Interval Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:古玩行普遍相信,把东西送进实验室就能得到一个精确年份。本书把这一信念写成可以运行的算术并加以检验:断代给出的是带误差棒的概率分布;在放射性碳校准曲线的平台年代,一个测值会对应几段彼此分离的历史区间;瓷胎缺少可测年的有机碳,烧成年代只能靠热释光一类以最后一次受热为零点的方法。核对原文后,本书对立项蓝图作了修正:Libby、Anderson 与 Arnold 1949 年的《科学》短文检验的是天然放射性碳的全球均一性,用已知年龄样品作核对的是同年 Arnold 与 Libby 的另一篇;「瓷器只能靠热释光」收窄为「瓷胎烧成年代只能靠释光方法,但它分不开相邻两朝,也看不见拼接与重烧」。书中器物、价格与参数均为示意,校准曲线是手设示意折线而非 IntCal 数据;全部计算由书中程序生成。本书不对任何具体藏品下结论,不构成鉴定意见,也不是个人化的投资、法律、税务或理财建议。 第一章算指数衰变与半衰期约定,剑桥值与 Libby 值之比为 1.0291,并说明计数误差如何传到年龄上。第二章在示意曲线上演示平台:同样正负 25 年的测量误差,平台上的测值校准为两段,概率 0.627 与 0.328;精度提高四倍,平台区间宽度只缩到原来的 0.79。第三章讲高密度区间、合并前的一致性检验与独立先验窗口。第四章写热释光年龄等于累积剂量除以年剂量,出土器物的相对误差约 0.089,无出土背景时升到 0.123。 第五章讨论材料与时钟的对应,说明热释光能干净地分开新仿与老器,却很难分开相隔一百年的两个时期,相对误差为 0.25 时九成五区间能排除相邻时期的概率只有 0.109。第六章把送检写成期望损失问题,示意参数下先验把握低于 0.655 时不检测就不应购买,检测在把握居中时最值钱。第七章考辨赵希鹄《洞天清录》与 Libby 等人的 1949 年短文,把古琴断纹写成速率不确定的时钟,「有断纹即逾五百年」的正确率为 0.835。第八章给出古玩送检决策表、蒸馏小工具「断代误差区间计算器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Exponential Decay and Error Propagation: from Counts to the Error Bar on an Age — Chapter 1 treats exponential decay and the half-life conventions, with a Cambridge-to-Libby ratio of 1.0291, and shows how counting error passes into age. (https://5dlibrary.com/answers/en/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c01/) 第1章 指数衰变与误差传播:从计数到年龄的误差棒 — 第一章算指数衰变与半衰期约定,剑桥值与 Libby 值之比为 1.0291,并说明计数误差如何传到年龄上。 (https://5dlibrary.com/answers/zh/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c01/) - Ch.2 Calibration Plateaus: Why One Measurement Maps onto Several Historical Ranges — Chapter 2 demonstrates a plateau on the illustrative curve: with the same error of plus or minus 25 years, a plateau measurement calibrates to two ranges with probabilities 0.627 and 0.328, and four times the precision shrinks the plateau width only to 0.79 of what it was. (https://5dlibrary.com/answers/en/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c02/) 第2章 校准曲线平台:一个测值为何对应多个历史区间 — 第二章在示意曲线上演示平台:同样正负 25 年的测量误差,平台上的测值校准为两段,概率 0.627 与 0.328;精度提高四倍,平台区间宽度只缩到原来的 0.79。 (https://5dlibrary.com/answers/zh/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c02/) - Ch.3 A Date Is a Distribution: Highest-Density Ranges, Consistency Tests and Prior Windows — A distribution with several peaks cannot be summarised by one number. The 68.3 % and 95.4 % levels can give different numbers of segments, and on a plateau the mean, median and mode can lie a century apart. Before several measurements are pooled they must pass a consistency test, or one outlier slips into the result wearing a smaller error. What genuinely narrows a range is an independent prior unrelated to the claim under test, and a stylistic judgement cannot play that role. Here the story of radiocarbon pauses. The next chapter turns to the clock that porcelain and pottery actually rely on, which measures not carbon but the radiation dose that mineral grains have accumulated since they were last heated. (https://5dlibrary.com/answers/en/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c03/) 第3章 年代是一条分布:高密度区间、合并检验与先验窗口 — 一条多峰分布不能用一个数字概括。68.3 % 与 95.4 % 两个水平可以给出不同的段数;均值、中位数与众数在平台上可以相差一百年。合并多个测值前必须先做一致性检验,否则一个离群值会以更小的误差混进结果。真正能收窄区间的,是与待检验说法无关的独立先验,而风格判断不能充当这种先验。到这里,放射性碳的故事暂告一段落。下一章转向瓷器与陶器真正依靠的那一类时钟:它测的不是碳,而是矿物颗粒自上一次受热以来累积的辐射剂量。 (https://5dlibrary.com/answers/zh/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c03/) - Ch.4 The Thermoluminescence Age: Accumulated Dose over Annual Dose, and Its Error Budget — Chapter 4 sets the thermoluminescence age equal to accumulated dose over annual dose, with a relative error of about 0.089 for excavated pieces rising to 0.123 without a burial context. (https://5dlibrary.com/answers/en/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c04/) 第4章 热释光年龄:累积剂量除以年剂量及其误差预算 — 第四章写热释光年龄等于累积剂量除以年剂量,出土器物的相对误差约 0.089,无出土背景时升到 0.123。 (https://5dlibrary.com/answers/zh/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c04/) - Ch.5 Why Porcelain Carries No Carbon Clock: Materials, Sampling and the Limits of Method — Chapter 5 matches materials to clocks and shows that thermoluminescence cleanly separates new copies from old pieces but can hardly separate two periods a century apart: at a relative error of 0.25, the ninety-five per cent range excludes the adjacent period with probability only 0.109. (https://5dlibrary.com/answers/en/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c05/) 第5章 瓷器为何没有碳钟:材料、取样与方法边界 — 第五章讨论材料与时钟的对应,说明热释光能干净地分开新仿与老器,却很难分开相隔一百年的两个时期,相对误差为 0.25 时九成五区间能排除相邻时期的概率只有 0.109。 (https://5dlibrary.com/answers/zh/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c05/) - Ch.6 The Expected Loss of Testing: When to Test and What a Test Can Answer — Chapter 6 writes testing as a problem of expected loss; under illustrative parameters a buyer should not buy untested below a prior of 0.655, and a test is worth most at middling priors. (https://5dlibrary.com/answers/en/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c06/) 第6章 送检的期望损失:什么时候测、测了能回答什么 — 第六章把送检写成期望损失问题,示意参数下先验把握低于 0.655 时不检测就不应购买,检测在把握居中时最值钱。 (https://5dlibrary.com/answers/zh/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c06/) - Ch.7 Classical Sources East and West: Zhao Xigu's Dongtian Qinglu and Libby, Anderson and Arnold 1949 — Chapter 7 examines Zhao Xigu's Dongtian Qinglu and the 1949 notes of Libby and colleagues, modelling crackle on qin lacquer as a clock with an uncertain rate, under which "crackle means over five hundred years" is right with probability 0.835. (https://5dlibrary.com/answers/en/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c07/) 第7章 东西方原典考辨:赵希鹄《洞天清录》与 Libby, Anderson & Arnold 1949 — 第七章考辨赵希鹄《洞天清录》与 Libby 等人的 1949 年短文,把古琴断纹写成速率不确定的时钟,「有断纹即逾五百年」的正确率为 0.835。 (https://5dlibrary.com/answers/zh/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c07/) - Ch.8 Closing the Loop: a Decision Table for Testing Antiques, the Distilled Tool and a Pre-Registered Plan — Chapter 8 gives a decision table for testing antiques, the distilled Dating Error-Interval Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c08/) 第8章 实证闭环:古玩送检决策表、蒸馏工具与预注册方案 — 第八章给出古玩送检决策表、蒸馏小工具「断代误差区间计算器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/thermoluminescence-radiometric-dating-error-bars-calibration-plateaus/c08/) ## 5DL-B-332 · Topological Ordering of the Theorem-Dependency Graph of Three Hundred Volumes and a Convergence Criterion for the Self-Evolving Engine: Halting Conditions between Proposal Noise and Pruning Gain 中文书名:三百卷定理依赖图拓扑排序与母机收敛判据、立项噪音与剪枝增益的闭环停机条件母典 DOI: pending Cite: Fan, Y. (2026). Topological Ordering of the Theorem-Dependency Graph of Three Hundred Volumes and a Convergence Criterion for the Self-Evolving Engine: Halting Conditions between Proposal Noise and Pruning Gain (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative and contain no production information; all calculations are generated by its programs; it is general business analysis, not business advice. Chapter 4 sorts with Kahn's algorithm and finds and repairs cycles with strongly connected components; in a synthetic citation graph of three hundred volumes, 18,828 pairs are comparable, a share of 0.420. Abstract: A five-ring self-evolving engine that keeps proposing projects, writing books and distilling rules is expected to grow wiser as its rules multiply. This book overturns that piece of common sense: the engine does not converge on its own. When proposals outpace pruning, the potential conflict pairs among rules grow with the square of the rule count, reaching 4,498,500 pairs among three thousand rules; under a fixed audit budget, an engine that proposes and never retires sees its unchecked pairs multiply by 4.23 when time doubles. The engine converges only if pruning gain covers proposal noise. After checking Ashby's An Introduction to Cybernetics of 1956, the book corrects its blueprint: the law of requisite variety is a lower bound on the variety of outcomes and says nothing of rates or convergence, so its translation into a quarterly information ledger is the author's hypothesis; what grows as a square is the number of potential pairs, and whether realised conflicts do must be measured; and pruning gain exceeding proposal noise is only a necessary condition. Its parameters are illustrative and contain no production information; all calculations are generated by its programs; it is general business analysis, not business advice. Chapter 1 works out the square law and the backlog under a fixed budget. Chapter 2 enumerates Ashby's regulation game exhaustively and writes the convergence condition as noise inflow not exceeding pruning capacity; in a sensitivity scan, simulation and theory agree in 56 of 56 cells. Chapter 3 writes difference equations with audit delay: the stable limit of pruning gain is 1.0000 with a one-quarter delay and 0.6180 with two; with too little pruning capacity, conflicts grow 3.78-fold in twenty quarters, falling to 1.78-fold once obsolete rules retire. Chapter 4 sorts with Kahn's algorithm and finds and repairs cycles with strongly connected components; in a synthetic citation graph of three hundred volumes, 18,828 pairs are comparable, a share of 0.420. Chapter 5 shows that audit scope matters more than audit staff: checking only neighbours in the dependency graph costs about 7.2 checks per new rule against 2949.5 for the whole base, and greedy pruning by conflict count hits wrong rules at a rate of 0.818 against 0.330 for random pruning. Chapter 6 derives the halting size, 2,000 rules under illustrative parameters, and designs a strict curvature alarm that cuts the naive rule's false-alarm rate of 0.785 to 0.030. Chapter 7 examines "generation upon generation is called change" from the Great Commentary on the Book of Changes and Wiener's Cybernetics of 1948. Chapter 8 gives the quarterly convergence check-up, the distilled Engine Convergence Criterion Calculator and a pre-registered plan of refutation, in which testing the effect of pruning needs 7,785 audited pairs per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:一台不断立项、成书、蒸馏规则的五环自进化母机,人们自然期待它越长越聪明,规则越多越周全。本书推翻这个常识:母机不会自动收敛。立项速率超过剪枝速率时,规则之间的潜在冲突对数按规则数的平方增长,三千条规则之间有 4,498,500 对;在固定审计预算下只立项不退役,未查对数在时间翻倍时变成 4.23 倍。只有剪枝增益盖过立项噪音,母机才会收敛。核对艾什比 1956 年《控制论导论》原书之后,本书做了蓝图修正:必要多样性定律本身是关于结果多样性的下界,不讲速率与收敛,把它译成季度信息账本是作者的假说;平方增长的是潜在对数,实际冲突是否平方增长须实测;「剪枝增益超过立项噪音」只是必要条件。书中母机参数均为示意,不涉及任何服务器、域名、密钥或生产信息;全部计算由书中程序生成;本书是非个人化的经营分析,不构成经营建议。 第一章算平方律与固定预算下的积压。第二章逐一枚举艾什比的调节博弈,并把收敛条件写成噪音流入不超过剪枝容量,敏感度扫描中模拟与理论 56 格全部吻合。第三章写带审计延迟的差分方程:延迟一季时剪枝增益的稳定上限是 1.0000,两季时是 0.6180;剪枝容量不足时冲突在二十个季度里增长到 3.78 倍,加上过时退役后降到 1.78 倍。第四章用卡恩算法排序、用强连通分量查环修环,三百卷合成引用图中可比对为 18,828 对,占 0.420。 第五章说明审计范围比审计人数更重要:只查依赖图近邻时每条新规则约查 7.2 对,全库核对要 2949.5 对;按冲突数贪心剪枝的命中率是 0.818,随机剪枝只有 0.330。第六章推导停机规模,示意参数下为 2,000 条,并设计严格曲率报警,把朴素规则 0.785 的误报压到 0.030。第七章考辨《周易·系辞上》「生生之谓易」与维纳 1948 年的《控制论》。第八章给出季度母机收敛体检、蒸馏小工具「母机收敛判据计算器」与预注册证伪方案,检验剪枝效果每组需审计 7,785 对规则。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Square Law of Conflict Pairs: n Choose 2, a Fixed Audit Budget and the Backlog — With $q$ at 0.001, one clash per thousand pairs, the base sounds clean. Yet three hundred rules carry 44.85 expected conflicts and twenty-four hundred rules carry 2878.8. At a hundred rules a base with $q$ of 0.0001 still has a 0.6096 chance of being free of conflict; once $q$ rises to 0.001 that chance falls to 0.0071. (https://5dlibrary.com/answers/en/three-hundred-volume-dependency-graph-engine-convergence-criterion/c01/) 第1章 冲突对数的平方律:n 选 2、固定审计预算与积压 — $q$ 取 0.001,也就是一千对里只有一对冲突,听上去已经很干净。可是三百条规则时预期冲突是 44.85 个,两千四百条规则时是 2878.8 个。一百条规则时,$q$ 为 0.0001 的规则库仍有 0.6096 的机会完全无冲突;$q$ 一旦升到 0.001,这个概率就跌到 0.0071。 (https://5dlibrary.com/answers/zh/three-hundred-volume-dependency-graph-engine-convergence-criterion/c01/) - Ch.2 Requisite Variety: When the Information Budget of Pruning Covers the Noise of Proposals — Chapter 2 enumerates Ashby's regulation game exhaustively and writes the convergence condition as noise inflow not exceeding pruning capacity; in a sensitivity scan, simulation and theory agree in 56 of 56 cells. (https://5dlibrary.com/answers/en/three-hundred-volume-dependency-graph-engine-convergence-criterion/c02/) 第2章 必要多样性:剪枝的信息预算何时盖过立项噪音 — 第二章逐一枚举艾什比的调节博弈,并把收敛条件写成噪音流入不超过剪枝容量,敏感度扫描中模拟与理论 56 格全部吻合。 (https://5dlibrary.com/answers/zh/three-hundred-volume-dependency-graph-engine-convergence-criterion/c02/) - Ch.3 Difference Equations and Stability: Pruning Gain, Audit Delay and Oscillation — Chapter 3 writes difference equations with audit delay: the stable limit of pruning gain is 1.0000 with a one-quarter delay and 0.6180 with two; with too little pruning capacity, conflicts grow 3.78-fold in twenty quarters, falling to 1.78-fold once obsolete rules retire. (https://5dlibrary.com/answers/en/three-hundred-volume-dependency-graph-engine-convergence-criterion/c03/) 第3章 差分方程与稳定性:剪枝增益、审计延迟与震荡 — 第三章写带审计延迟的差分方程:延迟一季时剪枝增益的稳定上限是 1.0000,两季时是 0.6180;剪枝容量不足时冲突在二十个季度里增长到 3.78 倍,加上过时退役后降到 1.78 倍。 (https://5dlibrary.com/answers/zh/three-hundred-volume-dependency-graph-engine-convergence-criterion/c03/) - Ch.4 The Theorem-Dependency Graph: Kahn's Topological Sort, Cycle Detection and a Synthetic Graph of Three Hundred Volumes — Chapter 4 sorts with Kahn's algorithm and finds and repairs cycles with strongly connected components; in a synthetic citation graph of three hundred volumes, 18,828 pairs are comparable, a share of 0.420. (https://5dlibrary.com/answers/en/three-hundred-volume-dependency-graph-engine-convergence-criterion/c04/) 第4章 定理依赖图:卡恩拓扑排序、环检测与三百卷合成图 — 第四章用卡恩算法排序、用强连通分量查环修环,三百卷合成引用图中可比对为 18,828 对,占 0.420。 (https://5dlibrary.com/answers/zh/three-hundred-volume-dependency-graph-engine-convergence-criterion/c04/) - Ch.5 Local Audits and Pruning Gain: How Many Pairs Each New Rule Costs and How Many Conflicts Each Cut Removes — Chapter 5 shows that audit scope matters more than audit staff: checking only neighbours in the dependency graph costs about 7.2 checks per new rule against 2949.5 for the whole base, and greedy pruning by conflict count hits wrong rules at a rate of 0.818 against 0.330 for random pruning. (https://5dlibrary.com/answers/en/three-hundred-volume-dependency-graph-engine-convergence-criterion/c05/) 第5章 局部审计与剪枝增益:每条新规则查多少对,每剪一条消多少冲突 — 第五章说明审计范围比审计人数更重要:只查依赖图近邻时每条新规则约查 7.2 对,全库核对要 2949.5 对;按冲突数贪心剪枝的命中率是 0.818,随机剪枝只有 0.330。 (https://5dlibrary.com/answers/zh/three-hundred-volume-dependency-graph-engine-convergence-criterion/c05/) - Ch.6 Halting Conditions: the Halting Size and a Curvature Alarm on Conflicts — Chapter 6 derives the halting size, 2,000 rules under illustrative parameters, and designs a strict curvature alarm that cuts the naive rule's false-alarm rate of 0.785 to 0.030. (https://5dlibrary.com/answers/en/three-hundred-volume-dependency-graph-engine-convergence-criterion/c06/) 第6章 停机条件:停机规模与冲突曲率警报 — 第六章推导停机规模,示意参数下为 2,000 条,并设计严格曲率报警,把朴素规则 0.785 的误报压到 0.030。 (https://5dlibrary.com/answers/zh/three-hundred-volume-dependency-graph-engine-convergence-criterion/c06/) - Ch.7 Eastern and Western Sources: "Generation upon Generation" in the Great Commentary and Wiener's Cybernetics — Chapter 7 examines "generation upon generation is called change" from the Great Commentary on the Book of Changes and Wiener's Cybernetics of 1948. (https://5dlibrary.com/answers/en/three-hundred-volume-dependency-graph-engine-convergence-criterion/c07/) 第7章 东西方原典考辨:《周易·系辞上》「生生之谓易」与维纳《控制论》 — 第七章考辨《周易·系辞上》「生生之谓易」与维纳 1948 年的《控制论》。 (https://5dlibrary.com/answers/zh/three-hundred-volume-dependency-graph-engine-convergence-criterion/c07/) - Ch.8 The Quarterly Convergence Check-up: Procedure, the Engine Convergence Criterion Calculator and a Pre-registered Plan of Refutation — Chapter 8 gives the quarterly convergence check-up, the distilled Engine Convergence Criterion Calculator and a pre-registered plan of refutation, in which testing the effect of pruning needs 7,785 audited pairs per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/three-hundred-volume-dependency-graph-engine-convergence-criterion/c08/) 第8章 季度母机收敛体检:流程、母机收敛判据计算器与预注册证伪方案 — 第八章给出季度母机收敛体检、蒸馏小工具「母机收敛判据计算器」与预注册证伪方案,检验剪枝效果每组需审计 7,785 对规则。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/three-hundred-volume-dependency-graph-engine-convergence-criterion/c08/) ## 5DL-B-333 · Dark-Ads Whitelist Licensing Economics, Three-Second Hook Survival Analysis and Single-Tier Scout Attribution Games 中文书名:暗投白名单授权经济学、三秒钩子生存分析与单层星探归因博弈母典 DOI: pending Cite: Fan, Y. (2026). Dark-Ads Whitelist Licensing Economics, Three-Second Hook Survival Analysis and Single-Tier Scout Attribution Games (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative, all calculations are generated by its programs, and it is general analysis, not business or legal advice. Chapter 6 prices the bypass option on a binomial tree, where a damages clause at the reasonable pre-estimate of 20.82 deters every bypass in the illustrative model. Abstract: In creator whitelisting, known in the trade as dark posts, a brand runs paid ads from a creator's account; the posts do not appear on the public profile but must carry the platform's paid-partnership disclosure. The popular belief is that last-touch dashboard ROAS tells whether a licence fee was worth paying and how much the scout deserves. This book overturns that belief: pricing licences and paying scouts by dashboard ROAS systematically rewards noise. At the magnitudes reported by Lewis and Rao, a test with a million people per group still leaves an ROI confidence interval 138.6 percentage points wide; a single month's dashboard reading has a reliability of only 0.0471; and in simulation the dashboard credits 24.6 times the true incremental orders. After checking the sources the book corrects its blueprint: judging profit needs tens of millions of person-weeks, not millions of impressions; observational methods often fail, but the source does not claim a constant direction of error; the UK Supreme Court now applies a legitimate-interest proportionality test; and there is no three-second law in the literature. Its parameters are illustrative, all calculations are generated by its programs, and it is general analysis, not business or legal advice. Chapter 1 measures noise and selection bias. Chapter 2 prices licences along a wear-out curve: a quarter is worth only 1.70 times the first month, so a quarterly quote at 2.5 times overpays by a factor of 1.47, while a wear-out ladder raises the brand's expected surplus to 30,619. Chapter 3 analyses the three-second hook with Kaplan-Meier curves and a Weibull fit, with a shape parameter of 0.582, and finds kept orders highest at a hook intensity of 0.26. Chapter 4 runs the cold start with Thompson sampling, which after three thousand clicks picks the best creative in a share of only 0.765. Chapter 5 settles single-tier scout commission by Shapley value, with a scout share of 0.1458; the advertiser nets 0.2000 under last touch and 0.3561 under Shapley credit, and a null touch earns nothing. Chapter 6 prices the bypass option on a binomial tree, where a damages clause at the reasonable pre-estimate of 20.82 deters every bypass in the illustrative model. Chapter 7 examines Han Feizi's parable of buying the box and returning the pearl and Hopkins' Scientific Advertising. Chapter 8 gives a monthly review, the distilled Whitelist Licence Pricing and Hook Decay Calculator and a pre-registered plan of refutation, in which testing the central claim needs 13 contracts with holdout experiments. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:达人白名单授权,行内叫暗投:品牌借达人账号投放广告,帖子不在主页公开发布,但须按平台规则标明付费合作。流行说法认为,看后台末次触点 ROAS,就知道授权费值不值、星探该分多少。本书推翻这个常识:按后台 ROAS 给授权定价、给星探分成,会系统性奖励噪音。按刘易斯与拉奥所报量级,每组百万人时,回报率置信区间仍宽 138.6 个百分点;单月后台读数信度只有 0.0471;模拟中后台记功订单是真实增量 24.6 倍。核对原文之后,本书做了蓝图修正:判定盈亏常需千万人周而非百万级曝光;观察法常测不准,但原文未说偏差方向恒定;英国最高法院已改用正当利益比例测试;学界并无三秒定律。书中参数均为示意,全部计算由书中程序生成;本书是非个人化经营分析,不构成经营或法律建议。 第一章算噪音与选择偏差。第二章按磨损曲线给授权定价:一季只值首月 1.70 倍,按月价 2.5 倍报季度价高估 1.47 倍;磨损阶梯合约让品牌期望剩余升到 30,619。第三章用卡普兰与迈耶曲线、威布尔拟合分析三秒钩子,形状参数 0.582,留住订单在钩子强度 0.26 时最高。第四章用汤普森抽样做冷启动,三千点击后挑对最优素材比例只有 0.765。 第五章以沙普利值做单层星探分成,星探份额 0.1458,末次触点下广告主净得 0.2000,沙普利值下升到 0.3561,零贡献触点分成为零。第六章用二叉树为跳单期权定价,合理预估 20.82 已拦住示意模型中全部跳单。第七章考辨《韩非子》买椟还珠与霍普金斯《科学广告》。第八章给出月度复盘流程、蒸馏小工具「白名单授权定价与钩子衰减计算器」与预注册证伪方案,检验核心命题需 13 张配有留出实验之合约。书中以标注框区分已发表证据、可运行计算(20 段程序)、原典与作者假说。 Chapter answers: - Ch.1 Dashboard Attribution Cannot Price a Licence — Start with panel A. With 1,000,000 people per group, the standard error of ROI is 0.354 and the confidence interval is 138.6 percentage points wide. Put differently, this experiment cannot tell a true ROI of zero from one that doubles the money. With 100,000 people per group the interval widens to 438.3 points; even with 10,000,000 per group it is still 43.8 points wide. (https://5dlibrary.com/answers/en/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c01/) 第1章 后台归因,给不了授权价 — 先看 A 段。每组 1,000,000 人时,回报率标准误为 0.354,置信区间宽 138.6 个百分点。换句话说,真实回报率是零,还是翻倍,这场实验分不清。每组 100,000 人时,区间宽到 438.3;每组 10,000,000 人时,仍有 43.8。 (https://5dlibrary.com/answers/zh/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c01/) - Ch.2 Pricing a Licence along the Wear-Out Curve — Chapter 2 prices licences along a wear-out curve: a quarter is worth only 1.70 times the first month, so a quarterly quote at 2.5 times overpays by a factor of 1.47, while a wear-out ladder raises the brand's expected surplus to 30,619. (https://5dlibrary.com/answers/en/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c02/) 第2章 磨损曲线,给授权定价 — 第二章按磨损曲线给授权定价:一季只值首月 1.70 倍,按月价 2.5 倍报季度价高估 1.47 倍;磨损阶梯合约让品牌期望剩余升到 30,619。 (https://5dlibrary.com/answers/zh/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c02/) - Ch.3 The Three-Second Hook as a Survival Problem — Chapter 3 analyses the three-second hook with Kaplan-Meier curves and a Weibull fit, with a shape parameter of 0.582, and finds kept orders highest at a hook intensity of 0.26. (https://5dlibrary.com/answers/en/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c03/) 第3章 三秒钩子,生存分析 — 第三章用卡普兰与迈耶曲线、威布尔拟合分析三秒钩子,形状参数 0.582,留住订单在钩子强度 0.26 时最高。 (https://5dlibrary.com/answers/zh/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c03/) - Ch.4 The Pixel Cold Start and Thompson Sampling — Chapter 4 runs the cold start with Thompson sampling, which after three thousand clicks picks the best creative in a share of only 0.765. (https://5dlibrary.com/answers/en/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c04/) 第4章 像素冷启动,汤普森抽样 — 第四章用汤普森抽样做冷启动,三千点击后挑对最优素材比例只有 0.765。 (https://5dlibrary.com/answers/zh/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c04/) - Ch.5 Single-Tier Scout Commission by Shapley Value — Chapter 5 settles single-tier scout commission by Shapley value, with a scout share of 0.1458; the advertiser nets 0.2000 under last touch and 0.3561 under Shapley credit, and a null touch earns nothing. (https://5dlibrary.com/answers/en/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c05/) 第5章 单层星探,沙普利分成 — 第五章以沙普利值做单层星探分成,星探份额 0.1458,末次触点下广告主净得 0.2000,沙普利值下升到 0.3561,零贡献触点分成为零。 (https://5dlibrary.com/answers/zh/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c05/) - Ch.6 Liquidated Damages against Bypass Deals, and Statistical Power — Chapter 6 prices the bypass option on a binomial tree, where a damages clause at the reasonable pre-estimate of 20.82 deters every bypass in the illustrative model. (https://5dlibrary.com/answers/en/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c06/) 第6章 防跳单违约金,与检测功效 — 第六章用二叉树为跳单期权定价,合理预估 20.82 已拦住示意模型中全部跳单。 (https://5dlibrary.com/answers/zh/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c06/) - Ch.7 Buying the Box and Returning the Pearl, and Scientific Advertising — In the illustrative data the rank correlation is 0.629 with a permutation $p$-value of 0.0169. Twelve creatives are enough to see the direction, but only because the effect in the illustrative data was set large. If the true correlation is only 0.4, the test needs 38 creatives; if it is 0.6, it needs 16. A brand may not license that many creatives in a quarter. (https://5dlibrary.com/answers/en/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c07/) 第7章 买椟还珠,与科学广告 — 示意数据里,等级相关为 0.629,置换 $p$ 值 0.0169。十二条素材,就能看出方向。可这是因为,示意数据里,效应设得很大。真实效应若只有 0.4,要 38 条素材;若为 0.6,要 16 条。品牌一个季度,未必有这么多授权素材。 (https://5dlibrary.com/answers/zh/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c07/) - Ch.8 The Monthly Review and a Pre-Registered Refutation — Chapter 8 gives a monthly review, the distilled Whitelist Licence Pricing and Hook Decay Calculator and a pre-registered plan of refutation, in which testing the central claim needs 13 contracts with holdout experiments. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c08/) 第8章 月度复盘,与预注册证伪 — 第八章给出月度复盘流程、蒸馏小工具「白名单授权定价与钩子衰减计算器」与预注册证伪方案,检验核心命题需 13 张配有留出实验之合约。书中以标注框区分已发表证据、可运行计算(20 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/dark-ads-whitelist-licensing-hook-survival-single-tier-scout-attribution/c08/) ## 5DL-B-334 · Genealogical DAG Merge Algorithms, Generation-Poem Congruence Coding and the Institutional Economics of the Fan Clan Charitable Estate 中文书名:宗族世系有向无环图合龙算法、字辈同余编码与范氏义庄跨代治理母典 DOI: pending Cite: Fan, Y. (2026). Genealogical DAG Merge Algorithms, Generation-Poem Congruence Coding and the Institutional Economics of the Fan Clan Charitable Estate (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All data are synthetic or illustrative, all calculations come from its programs, and it is general analysis, not legal, financial or business advice; it compares no lineages or ethnic groups. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Among genealogists a popular belief holds that two island branch books can be joined wherever they record a man of the same surname, given name, generation character and native place, and that a clan with plenty of sons will naturally endure. This book overturns both beliefs. Where names follow a generation poem only one character is free, so under illustrative parameters eleven men of one generation already give an even chance of a repeated name, and joining books of 60 and 40 men by full name alone gets only a share of 0.208 of its joins right. After checking the sources the book corrects its blueprint: Watson's 1875 conclusion that every surname must die out is wrong, and the considerable but sub-unit probability of extinction is a later correction; the thousand mu of the Fan estate come from Qian Gongfu's record, while Twitchett also gives more than three thousand mou, so no round figure is adopted; and its duration is taken only as the span 1050 to 1760 studied by Twitchett, not eight hundred years. All data are synthetic or illustrative, all calculations come from its programs, and it is general analysis, not legal, financial or business advice; it compares no lineages or ethnic groups. Chapter 1 computes same-generation name collisions. Chapter 2 uses Fellegi-Sunter likelihood ratios with two thresholds to cut false merges from 36.33 to 0.612 at the cost of reviewing 15.4 pairs, and estimates the parameters by expectation-maximisation. Chapter 3 checks descent as a directed acyclic graph with Kahn's sort and Tarjan's components; among single misattachments, same-generation namesakes invisible to the structure make up 0.136. Chapter 4 treats the generation poem as a congruence code that catches every single dropped generation when no character repeats, falling to 0.9500 with repeats; restarting an eight-character poem makes cross-cycle namesakes 0.517 of all. Chapter 5 models chain migration to the Southern Seas as a branching process in which villagers crowd into a few ports, cutting the weight of port agreement from a naive 5.17 bits to 2.00. Chapter 6 computes extinction and a ration ledger for the estate: with 1.2604 sons per man a single line still dies out with probability 0.8193, and with population growing at 0.00442 a year the headroom lasts 47.6 years; it also compares Ostrom's eight principles. Chapter 7 examines Fan Zhongyan's rules and the extinction problem of Galton and Watson. Chapter 8 gives a quarterly merge check-up, the distilled Island Genealogy Merge Matcher and a pre-registered plan of refutation needing 385 reviewed pairs and 435 blindly checked pairs per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:修谱圈里,流行说法认为:同姓、同名、同字辈、同籍贯,两本孤岛支谱就能合龙;人丁兴旺,宗族自然长久。本书推翻这两句话。按字辈起名,同代人只剩一字可变,示意参数下一代 11 人就有一半机会撞名;两本 60 人对 40 人之支谱按全名硬接,接对者只占 0.208。核对原文之后,本书做了蓝图修正:沃森一八七五年原文断言各姓终将消亡,是错的,断绝概率可观而小于一,是后人修正;范氏义庄千亩之说出自钱公辅《义田记》,杜希德另有三千亩以上之说,本书不取整数;延续年代只用杜希德所考一〇五〇年至一七六〇年,不说八百年。书中数据全为合成或示意,不含任何真实在世者资料;全部计算由书中程序生成;本书是非个人化分析,不构成法律、理财或经营建议,也不做任何血统高下或族群比较。 第一章算同名同辈碰撞。第二章用费莱吉与森特之似然比与两条阈值,把误合从 36.33 对压到 0.612 对,代价是复核 15.4 对,并用期望最大化算法估参数。第三章把世系写成有向无环图,用卡恩拓扑排序与塔扬强连通分量查错;单处误挂父亲,同代同名、结构看不见者占 0.136。第四章把字辈诗当同余码,不重字之诗单边漏代全部检出,有重字时降到 0.9500;八字诗用完从头再用,跨轮同名占 0.517。 第五章把下南洋写成链式迁徙分支过程,同村之人挤进少数埠头,埠头一致之分量从朴素 5.17 比特降到 2.00 比特。第六章算断嗣与义庄口粮账:平均男嗣 1.2604 时单房仍有 0.8193 之机会断绝,人口年增 0.00442 时余量只撑 47.6 年,并对照奥斯特罗姆八条原则。第七章考辨范仲淹义庄规矩与高尔顿、沃森之断嗣问题。第八章给出季度合龙体检、蒸馏小工具「孤岛家谱合龙匹配器」与预注册证伪方案,核心检验需核 385 对、每组盲核 435 对。书中以标注框区分已发表证据、可运行计算(20 段程序)、原典与作者假说。 Chapter answers: - Ch.1 Same Name, Same Generation: Collisions Are Far Commoner Than They Look — Start with panel A. With a Zipf exponent of 0.8, the sum of the squared character probabilities is 0.01520. Its reciprocal, the effective pool of characters, is only 65.8. Four hundred candidates look generous, but fewer than seventy actually do the work. The ten most common characters carry a share of 0.294 between them. With an exponent of 0.6 the effective pool is still only 159.6; with an exponent of 1.0 it shrinks to 26.3. (https://5dlibrary.com/answers/en/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c01/) 第1章 同名同辈,撞名远比想象多 — 先看 A 段。齐夫指数取 0.8,各字概率平方和为 0.01520。平方和倒数,叫有效字池,只有 65.8 个字。四百个候选字,看着宽,实际起作用,不到七十个。最常用十个字,就占了 0.294。指数取 0.6,有效字池也只有 159.6;取 1.0,缩到 26.3。 (https://5dlibrary.com/answers/zh/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c01/) - Ch.2 Keeping a Ledger of Likelihood Ratios: Two Thresholds and a Review Zone — Chapter 2 uses Fellegi-Sunter likelihood ratios with two thresholds to cut false merges from 36.33 to 0.612 at the cost of reviewing 15.4 pairs, and estimates the parameters by expectation-maximisation. (https://5dlibrary.com/answers/en/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c02/) 第2章 似然比记账:两条阈值与复核区 — 第二章用费莱吉与森特之似然比与两条阈值,把误合从 36.33 对压到 0.612 对,代价是复核 15.4 对,并用期望最大化算法估参数。 (https://5dlibrary.com/answers/zh/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c02/) - Ch.3 A Lineage Is a Directed Acyclic Graph: Topological Sorting and Cycle Detection — Chapter 3 checks descent as a directed acyclic graph with Kahn's sort and Tarjan's components; among single misattachments, same-generation namesakes invisible to the structure make up 0.136. (https://5dlibrary.com/answers/en/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c03/) 第3章 世系是有向无环图:拓扑排序与查环 — 第三章把世系写成有向无环图,用卡恩拓扑排序与塔扬强连通分量查错;单处误挂父亲,同代同名、结构看不见者占 0.136。 (https://5dlibrary.com/answers/zh/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c03/) - Ch.4 The Generation Poem Is a Congruence Code: Dropped Generations, Extra Generations and Collisions Across Cycles — Chapter 4 treats the generation poem as a congruence code that catches every single dropped generation when no character repeats, falling to 0.9500 with repeats; restarting an eight-character poem makes cross-cycle namesakes 0.517 of all. (https://5dlibrary.com/answers/en/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c04/) 第4章 字辈诗是同余码:漏代、多代与轮回撞名 — 第四章把字辈诗当同余码,不重字之诗单边漏代全部检出,有重字时降到 0.9500;八字诗用完从头再用,跨轮同名占 0.517。 (https://5dlibrary.com/answers/zh/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c04/) - Ch.5 Going to the Southern Seas Was Chain Migration: Branching Processes and the Crowding of Ports — Chapter 5 models chain migration to the Southern Seas as a branching process in which villagers crowd into a few ports, cutting the weight of port agreement from a naive 5.17 bits to 2.00. (https://5dlibrary.com/answers/en/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c05/) 第5章 下南洋是链式迁徙:分支过程与埠头集中 — 第五章把下南洋写成链式迁徙分支过程,同村之人挤进少数埠头,埠头一致之分量从朴素 5.17 比特降到 2.00 比特。 (https://5dlibrary.com/answers/zh/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c05/) - Ch.6 The Perpetual Charitable Estate: Extinction Probabilities, a Ration Ledger and the Risk of Governance — Chapter 6 computes extinction and a ration ledger for the estate: with 1.2604 sons per man a single line still dies out with probability 0.8193, and with population growing at 0.00442 a year the headroom lasts 47.6 years; it also compares Ostrom's eight principles. (https://5dlibrary.com/answers/en/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c06/) 第6章 义庄永续:断嗣概率、口粮账与治理风险 — 第六章算断嗣与义庄口粮账:平均男嗣 1.2604 时单房仍有 0.8193 之机会断绝,人口年增 0.00442 时余量只撑 47.6 年,并对照奥斯特罗姆八条原则。 (https://5dlibrary.com/answers/zh/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c06/) - Ch.7 The Classics Examined: Fan Zhongyan's Rules for the Charitable Estate and the Extinction Problem of Galton and Watson — Panel C: with a mean of exactly one son, extinction is certain but slow. At the tenth generation a line still has a chance of 0.1582 of surviving, against 0.2000 from the asymptotic formula; at the fiftieth, 0.0376 against 0.0400; at the hundredth, 0.0194 against 0.0200; at the two hundredth, 0.0098 against 0.0100. In the critical case extinction is certain but takes very many generations. (https://5dlibrary.com/answers/en/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c07/) 第7章 原典考辨:范仲淹之义庄规矩与高尔顿、沃森之断嗣问题 — C 段:平均恰好一子时,断绝必然,但很慢。传到第十代,仍有 0.1582 之机会未断,渐近式给出 0.2000;第五十代,0.0376 对 0.0400;第一百代,0.0194 对 0.0200;第二百代,0.0098 对 0.0100。临界情形,断绝是必然,可要很多代。 (https://5dlibrary.com/answers/zh/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c07/) - Ch.8 The Quarterly Merge Check-up: Procedure, Distilled Tool and Pre-registered Refutation — Chapter 8 gives a quarterly merge check-up, the distilled Island Genealogy Merge Matcher and a pre-registered plan of refutation needing 385 reviewed pairs and 435 blindly checked pairs per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c08/) 第8章 季度合龙体检:流程、小工具与预注册证伪 — 第八章给出季度合龙体检、蒸馏小工具「孤岛家谱合龙匹配器」与预注册证伪方案,核心检验需核 385 对、每组盲核 435 对。书中以标注框区分已发表证据、可运行计算(20 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/genealogical-dag-merge-generation-poem-congruence-fan-clan-estate/c08/) ## 5DL-B-335 · Strata-Title Commons Games, Sinking-Fund Ruin Actuarial Science and High-Intent Property Lead Routing 中文书名:分层地契公地博弈、储备金破产精算与高意向置业线索派单母典 DOI: pending Cite: Fan, Y. (2026). Strata-Title Commons Games, Sinking-Fund Ruin Actuarial Science and High-Intent Property Lead Routing (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All data are synthetic or illustrative, all calculations come from its programs, and it is general analysis, not legal, tax or investment advice; statutes await human review. Boxes separate evidence, runnable calculations (20 programs), classical sources and hypotheses. Abstract: In Malaysian strata buildings a popular belief holds that a sinking fund funded at the statutory ten per cent will cover major works, and that late payment is a matter of a few owners' character. This book overturns both beliefs. Major works come in lumps, at random times and with occasional heavy-tailed accidents; for an illustrative building, even with charges raised in line with inflation, funding at ten per cent leaves a 0.5715 chance of running dry within thirty years, and a contribution exactly equal to average spending still leaves 0.3479. Late payment is a public goods game in which sanctions have a threshold and arrears are contagious. After checking the sources the book corrects its blueprint: the ten per cent floor is in the Strata Management Act 2013 itself; the self-reinforcement of arrears past a critical point is a consequence of threshold models and is treated as a hypothesis; and the published ratios on lead response compare different companies and serve only as upper bounds. All data are synthetic or illustrative, all calculations come from its programs, and it is general analysis, not legal, tax or investment advice; statutes await human review. Chapter 1 uses public goods experiments and threshold models to show that a sanction below the private gain is no sanction, that above it the paid share rises to 0.9320, and that once the mean threshold falls below 0.2275 arrears jump past one half. Chapter 2 computes finite-horizon ruin by compound Poisson simulation and the Lundberg bound, and shows where heavy tails defeat the bound. Chapter 3 prices a delayed strata title with two clocks, about 0.0568 of value for a three-year wait, and shows that dropping still-waiting projects shrinks the mean wait from 4.33 to 2.42 years. Chapter 4 shows that the meta-analytic band premium of 0.042 is not the premium at the door, and that omitting highways inflates the station effect, the estimated gap rising from 0.0533 to 0.0827. Chapter 5 staffs a lead desk by the Erlang C formula, a fourth agent lifting the warmth left from 0.8802 to 0.9728, and a single queue served by the longest-idle agent yields 384.5 expected sales against 353.4 for strict rotation. Chapter 6 treats diverted leads as an inspection game in which raising detection beats raising penalties, cutting the call-backs needed per thousand leads from 1,000 to 370, using consent-based, tamper-evident logs only. Chapter 7 examines Mencius, Longzi and Aristotle. Chapter 8 gives the quarterly sinking-fund check-up, the monthly routing review, the distilled Reserve Ruin and Lead Dispatch Calculator and a pre-registered plan of refutation; the illustrative building must raise its rate from ten per cent to 0.2060, and the speed test needs 1,035 leads per arm. Boxes separate evidence, runnable calculations (20 programs), classical sources and hypotheses. 摘要:马来西亚分层公寓里,流行说法认为:储备金按法定一成计提,大修就有钱;迟缴只是少数人品行问题。本书推翻这两句话。大修支出块状、时点随机、偶有重尾意外;示意大楼即使管理费随通胀上调,按一成计提,三十年内见底机会仍有 0.5715,计提恰好等于平均支出,也有 0.3479。迟缴是公共物品博弈,惩罚有门槛,迟缴会传染。核对原文之后,本书做了蓝图修正:一成下限写在 2013 年「分层管理法令」本身;迟缴越过临界点自我强化,是门槛模型之推论,列为假说;线索响应之倍数,比较之是不同公司,只能当上限。书中楼盘、线索与参数全为合成或示意;全部计算由书中程序生成;本书是非个人化分析,不构成法律、税务或投资建议,法条与比例以上架前人工复核为准;线索只处理经同意者。 第一章用公共物品实验与门槛模型,说明罚则低于私人收益等于没有罚则,罚则过线后缴费比例升到 0.9320,平均门槛低过 0.2275,迟缴率就跳到一半以上。第二章用复合泊松模拟与伦德伯格上界,算出有限期破产概率,并说明重尾意外下上界失效。第三章用两只时钟写出产权证迟发之风险中性折价,等三年约为价值之 0.0568;删去仍在等待之项目,平均等待会从 4.33 年被说成 2.42 年。第四章说明轨交荟萃分析之带内平均 0.042 不能当门口溢价,漏掉公路时车站效应虚高,估计差从 0.0533 升到 0.0827。 第五章用厄朗 C 式排班,三人加到四人,线索剩余热度从 0.8802 升到 0.9728;同一队列、空闲最久者接单,期望成交 384.5 宗,严格轮转只有 353.4 宗。第六章写防飞单之检查博弈,提高查获率比加重罚则省钱,每千条线索所需回访从 1,000 条降到 370 条,并只用同意制留痕与防篡改日志。第七章考辨孟子井田、龙子论贡与助,以及亚里士多德论公有。第八章给出季度储备金体检、月度派单复盘、蒸馏小工具「储备金破产概率与派单计算器」与预注册证伪方案,示意大楼计提率须由一成升到 0.2060,回电速度检验每组需 1,035 条线索。书中以标注框区分已发表证据、可运行计算(20 段程序)、原典与作者假说。 Chapter answers: - Ch.1 The Game of Late Payment: Conditional Cooperation, the Sanction Threshold and the Critical Arrears Rate — Chapter 1 uses public goods experiments and threshold models to show that a sanction below the private gain is no sanction, that above it the paid share rises to 0.9320, and that once the mean threshold falls below 0.2275 arrears jump past one half. (https://5dlibrary.com/answers/en/strata-title-commons-sinking-fund-ruin-property-lead-routing/c01/) 第1章 迟缴之博弈:条件合作、惩罚门槛与临界迟缴率 — 第一章用公共物品实验与门槛模型,说明罚则低于私人收益等于没有罚则,罚则过线后缴费比例升到 0.9320,平均门槛低过 0.2275,迟缴率就跳到一半以上。 (https://5dlibrary.com/answers/zh/strata-title-commons-sinking-fund-ruin-property-lead-routing/c01/) - Ch.2 The Actuarial Ruin of a Sinking Fund: Lumpy Works, Heavy-Tailed Costs and Finite-Horizon Ruin — The second block keeps the charges frozen. At ten per cent, the chance that the fund runs dry within ten years is 0.1771; within twenty years, 0.6053; within thirty years, as high as 0.9454. Even at the rate that exactly matches average spending, 0.1659, the chance of running dry within thirty years is 0.3479. Raising the rate to 0.15 gives a thirty-year ruin probability of 0.5172; at 0.20 it falls to 0.1169; at 0.25 it is only 0.0250; and at 0.30 it is 0.0107. (https://5dlibrary.com/answers/en/strata-title-commons-sinking-fund-ruin-property-lead-routing/c02/) 第2章 储备金破产精算:块状大修、重尾成本与有限期破产概率 — 第二部分,管理费原地不动。按一成计提:十年内见底,机会 0.1771;二十年内,0.6053;三十年内,高达 0.9454。即使计提率,恰好等于平均支出,也就是 0.1659,三十年内见底,机会仍有 0.3479。计提率提到 0.15,三十年见底机会 0.5172;提到 0.20,降到 0.1169;提到 0.25,只有 0.0250;提到 0.30,是 0.0107。 (https://5dlibrary.com/answers/zh/strata-title-commons-sinking-fund-ruin-property-lead-routing/c02/) - Ch.3 The Discount for a Delayed Strata Title: Risk-Neutral Pricing and Censored Data — Chapter 3 prices a delayed strata title with two clocks, about 0.0568 of value for a three-year wait, and shows that dropping still-waiting projects shrinks the mean wait from 4.33 to 2.42 years. (https://5dlibrary.com/answers/en/strata-title-commons-sinking-fund-ruin-property-lead-routing/c03/) 第3章 产权证迟发之折价:风险中性定价与删失数据 — 第三章用两只时钟写出产权证迟发之风险中性折价,等三年约为价值之 0.0568;删去仍在等待之项目,平均等待会从 4.33 年被说成 2.42 年。 (https://5dlibrary.com/answers/zh/strata-title-commons-sinking-fund-ruin-property-lead-routing/c03/) - Ch.4 The Walking-Distance Gradient around Rail Stations: Reading a Meta-Analysis, Fitting and Confounding — Chapter 4 shows that the meta-analytic band premium of 0.042 is not the premium at the door, and that omitting highways inflates the station effect, the estimated gap rising from 0.0533 to 0.0827. (https://5dlibrary.com/answers/en/strata-title-commons-sinking-fund-ruin-property-lead-routing/c04/) 第4章 轨交步行衰减梯度:荟萃分析之读法、拟合与混淆 — 第四章说明轨交荟萃分析之带内平均 0.042 不能当门口溢价,漏掉公路时车站效应虚高,估计差从 0.0533 升到 0.0827。 (https://5dlibrary.com/answers/zh/strata-title-commons-sinking-fund-ruin-property-lead-routing/c04/) - Ch.5 Routing High-Intent Leads: Response Decay, Erlang Queues and Fair Rotation — Chapter 5 staffs a lead desk by the Erlang C formula, a fourth agent lifting the warmth left from 0.8802 to 0.9728, and a single queue served by the longest-idle agent yields 384.5 expected sales against 353.4 for strict rotation. (https://5dlibrary.com/answers/en/strata-title-commons-sinking-fund-ruin-property-lead-routing/c05/) 第5章 高意向线索派单:响应衰减、厄朗排队与公平轮转 — 第五章用厄朗 C 式排班,三人加到四人,线索剩余热度从 0.8802 升到 0.9728;同一队列、空闲最久者接单,期望成交 384.5 宗,严格轮转只有 353.4 宗。 (https://5dlibrary.com/answers/zh/strata-title-commons-sinking-fund-ruin-property-lead-routing/c05/) - Ch.6 Central Control against Diverted Leads: the Inspection Game, Sampled Call-Backs and Tamper-Evident Logs — Chapter 6 treats diverted leads as an inspection game in which raising detection beats raising penalties, cutting the call-backs needed per thousand leads from 1,000 to 370, using consent-based, tamper-evident logs only. (https://5dlibrary.com/answers/en/strata-title-commons-sinking-fund-ruin-property-lead-routing/c06/) 第6章 中控防飞单:检查博弈、抽样回访与防篡改留痕 — 第六章写防飞单之检查博弈,提高查获率比加重罚则省钱,每千条线索所需回访从 1,000 条降到 370 条,并只用同意制留痕与防篡改日志。 (https://5dlibrary.com/answers/zh/strata-title-commons-sinking-fund-ruin-property-lead-routing/c06/) - Ch.7 Eastern and Western Classics Examined: Mencius on the Well-Field, Aristotle on Common Property — Panel B reads "support one another in sickness" as a mutual-aid pool. Suppose each household suffers a loss of 30 units in a year with probability 0.1, independently of the others. Alone, a household's burden has a standard deviation of 9.0000, and the chance that its burden exceeds 10 is 0.1000. With two households pooling, the standard deviation falls to 6.3640, yet the chance of a burden above 10 rises to 0.1900: when one household is hit, each of the two carries half, which is still over the line. With four households the chance is 0.0523. With eight, the standard deviation is 3.1820 and the chance of exceeding 10 is only 0.0381. (https://5dlibrary.com/answers/en/strata-title-commons-sinking-fund-ruin-property-lead-routing/c07/) 第7章 东西方原典考辨:孟子井田、亚里士多德论公有 — B 段,读「疾病相扶持」。设每户每年,以 0.1 之机会,遭遇 30 单位损失,各户独立。独自承担,负担标准差 9.0000,负担超过 10 之机会,0.1000。两户互助,标准差 6.3640,可负担超过 10 之机会,反升到 0.1900:一户出事,两家各担一半,就过线。四户,0.0523。八户,标准差 3.1820,超过 10 之机会,只剩 0.0381。 (https://5dlibrary.com/answers/zh/strata-title-commons-sinking-fund-ruin-property-lead-routing/c07/) - Ch.8 Closing the Empirical Loop: the Quarterly Sinking-Fund Check-Up, the Monthly Routing Review and Pre-Registered Refutation — Chapter 8 gives the quarterly sinking-fund check-up, the monthly routing review, the distilled Reserve Ruin and Lead Dispatch Calculator and a pre-registered plan of refutation; the illustrative building must raise its rate from ten per cent to 0.2060, and the speed test needs 1,035 leads per arm. Boxes separate evidence, runnable calculations (20 programs), classical sources and hypotheses. (https://5dlibrary.com/answers/en/strata-title-commons-sinking-fund-ruin-property-lead-routing/c08/) 第8章 实证闭环:季度储备金体检、月度派单复盘与预注册证伪 — 第八章给出季度储备金体检、月度派单复盘、蒸馏小工具「储备金破产概率与派单计算器」与预注册证伪方案,示意大楼计提率须由一成升到 0.2060,回电速度检验每组需 1,035 条线索。书中以标注框区分已发表证据、可运行计算(20 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/strata-title-commons-sinking-fund-ruin-property-lead-routing/c08/) ## 5DL-B-336 · Zero-Trust Cybersecurity for a One-Commander Silicon Empire: Origin Cloaking, Secret Hygiene and Anti-Scraper Defense 中文书名:单兵硅基帝国零信任网安架构、源站隧道隐身与反爬虫防勒索母典 DOI: pending Cite: Fan, Y. (2026). Zero-Trust Cybersecurity for a One-Commander Silicon Empire: Origin Cloaking, Secret Hygiene and Anti-Scraper Defense (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: The book is defensive only and describes no attack, intrusion, bypass or evasion technique; all hosts, domains and tokens are synthetic, parameters are illustrative, and all calculations are generated by its programs. It is general business analysis, not a security audit or legal advice, and any production change requires human review and sign-off. Abstract: For a one-person business running a few dozen containers, the popular advice is to hide the origin behind a tunnel and open no public port. This book falsifies that advice: for small operators the largest attack surface is usually not a port but a credential. Using the structure of the attack surface metric, a tunnel removes 0.286 of the surface, but the credential dimension does not move, and its share of what remains rises to between 0.549 and 0.630. A large measurement of public repositories found secret leakage pervasive and persistent: a pushed string became searchable after a median of twenty seconds, and after sixteen days of monitoring 0.81 of detected secrets were still in place. After checking the originals the book corrects its blueprint: the twenty seconds measure searchability, not exploitation; the sample covers private keys and eleven platforms with distinctive key formats; the old incident handling guide was withdrawn and its third revision is cited instead; the honeypot monograph appeared in 2002; and the line from Sunzi follows the edition with eleven commentators, with the variant noted. Chapter 1 quantifies the attack surface. Chapter 2 computes blast radius on a reachability graph: with narrowed tokens, dropped privileges and read-only roots, an intruder reaches on average 2.50 containers from any entry. Chapter 3 shows that entropy cannot separate hexadecimal keys from commit hashes, while a format with a check value raises 1.5 alarms a month, and compresses revocation lists into a Bloom filter. Chapter 4 derives the exposure window and a square-root rule for rotation, and caps loss by lifetime times rate. Chapter 5 is defensive only: token buckets, trap pages, canary records and sequential tests. Chapter 6 shows that copies sharing a credential count as one, that a backup drilled once a year is broken at the moment of an incident with probability 0.7331, and that scripting the first three stop-bleed steps cuts 1,200 seconds to ten. Chapter 7 examines Sunzi on the nine variations and Kerckhoffs on military cryptography. Chapter 8 gives a monthly check-up, the distilled One-Commander Zero-Trust Check-up Calculator and a pre-registered test plan. The book is defensive only and describes no attack, intrusion, bypass or evasion technique; all hosts, domains and tokens are synthetic, parameters are illustrative, and all calculations are generated by its programs. It is general business analysis, not a security audit or legal advice, and any production change requires human review and sign-off. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:单人经营,几十个容器,流行说法是:源站藏进隧道,公网不开端口,就安全了。本书证伪这句话:对小型经营者,最大攻击面往往不是端口,而是凭据。按攻击面三元组之结构计算,隧道削去攻击面 0.286,可凭据一维纹丝不动,它在剩余攻击面里之份额,升到 0.549 至 0.630。公开仓库之大规模测量显示,密钥泄露普遍而持续,推送以后中位数二十秒即可检索,监测十六天,0.81 仍在原处。核对原文以后,本书作了蓝图修正:二十秒是可检索时间,不是攻击者利用时间;样本只覆盖私钥与十一类格式鲜明之平台密钥;旧版事件处置指南已撤回,改引第三版;蜜罐专著出版于二〇〇二年;九变篇原句依十一家注本,并注明异文。 第一章量化攻击面。第二章用可达图算爆炸半径:令牌收窄、去特权、只读根文件系统齐备以后,任意入口平均可达 2.50 个容器。第三章说明熵判别分不开十六进制密钥与提交哈希,格式加校验码之扫描,月报警只有 1.5 条;吊销名单用布隆过滤器压缩。第四章推导暴露窗口与轮换周期之平方根律,以寿命乘速率封顶损失。第五章只写防守:令牌桶、陷阱页、金丝雀记录与序贯检验。第六章说明共用凭据之副本只算一份,一年才演练一回之备份,事件发生时已坏之机会为 0.7331;止血前三步写成脚本,从 1,200 秒压到十秒。第七章考辨孙子九变篇与柯克霍夫斯之军事密码学。第八章给出月度体检、蒸馏小工具「单兵零信任体检计算器」与预注册检验方案。 全书只写防守,不写任何攻击、入侵、绕过或规避检测手法;书中主机、域名、令牌全为合成,参数为示意,全部计算由书中程序生成;本书为非个人化经营分析,不构成安全审计或法律意见,任何生产变更须经人工复核放行。书中以标注框区分已发表之证据、可运行之 Python 计算(20 段程序)、原典与作者之假说。 Chapter answers: - Ch.1 The Attack Surface: Beyond the Ports, Credentials Are the Front Door — Now let tokens leak. With a yearly leak probability of 0.05, the open layout totals 33.600 and the tunnel layout 24.350, and the token share rises from 0.415 to 0.572. With a leak probability of 0.20, the totals are 37.400 and 28.150, and the token shares 0.474 and 0.630. The higher the leak probability, the smaller the fraction the tunnel can remove: from 0.286, to 0.275, to 0.247. (https://5dlibrary.com/answers/en/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c01/) 第1章 攻击面:端口之外,凭据才是大门 — 再看令牌会泄露。每年泄露机会取 0.05,开放布局合计 33.600,隧道布局 24.350,令牌份额从 0.415 升到 0.572。泄露机会取 0.20,两者合计 37.400 与 28.150,令牌份额 0.474 与 0.630。泄露机会越高,隧道削减比例越小:从 0.286,降到 0.275,再降到 0.247。 (https://5dlibrary.com/answers/zh/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c01/) - Ch.2 Blast Radius: the Cage of Least Privilege — Chapter 2 computes blast radius on a reachability graph: with narrowed tokens, dropped privileges and read-only roots, an intruder reaches on average 2.50 containers from any entry. (https://5dlibrary.com/answers/en/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c02/) 第2章 爆炸半径:最小权限之囚笼 — 第二章用可达图算爆炸半径:令牌收窄、去特权、只读根文件系统齐备以后,任意入口平均可达 2.50 个容器。 (https://5dlibrary.com/answers/zh/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c02/) - Ch.3 Credential Hygiene: Format Checks and Revocation Lists — Chapter 3 shows that entropy cannot separate hexadecimal keys from commit hashes, while a format with a check value raises 1.5 alarms a month, and compresses revocation lists into a Bloom filter. (https://5dlibrary.com/answers/en/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c03/) 第3章 凭据卫生:格式校验与吊销名单 — 第三章说明熵判别分不开十六进制密钥与提交哈希,格式加校验码之扫描,月报警只有 1.5 条;吊销名单用布隆过滤器压缩。 (https://5dlibrary.com/answers/zh/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c03/) - Ch.4 Token Lifetime: the Exposure Window and the Ceiling on Loss — The second half of p42 uses a burst of 20 and a rate of 60 per hour. With a lifetime of one hour, the worst case is 80 requests and the average 50. With six hours, 380 and 195. With one day, 1,460 and 666. With one week, 10,100 and 2,668. With one year, 525,620 and 4,304. (https://5dlibrary.com/answers/en/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c04/) 第4章 令牌寿命:暴露窗口,损失上界 — p42 后半段,取突发 20,每小时 60 次。寿命一小时,最坏 80 次,平均 50 次。寿命六小时,最坏 380 次,平均 195 次。寿命一天,最坏 1,460 次,平均 666 次。寿命一周,最坏 10,100 次,平均 2,668 次。寿命一年,最坏 525,620 次,平均 4,304 次。 (https://5dlibrary.com/answers/zh/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c04/) - Ch.5 Anti-Scraper Defence: Token Buckets and Trap Pages — With ten canaries, a leaked copy of 1,000 records contains at least one with probability 0.0956; a copy of 10,000, 0.6513; a copy of 50,000, 0.9990. With fifty canaries, a copy of 1,000 records gives 0.3951 and a copy of 10,000 gives 0.9949. With two hundred canaries, a copy of 1,000 records gives 0.8663. For a copy of 1,000 records to contain at least one canary with probability 0.95, 298 canaries are needed, and such a copy holds on average 2.980 of them. (https://5dlibrary.com/answers/en/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c05/) 第5章 反爬虫:令牌桶与陷阱页 — 金丝雀十条:外流副本 1,000 条时,检出机会 0.0956;10,000 条,0.6513;50,000 条,0.9990。金丝雀五十条:1,000 条副本,0.3951;10,000 条,0.9949。金丝雀两百条:1,000 条副本,0.8663。要让 1,000 条之副本,以 0.95 之机会,至少含一条金丝雀,需埋 298 条,副本里平均有 2.980 条。 (https://5dlibrary.com/answers/zh/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c05/) - Ch.6 Ransomware: Common-Cause Failure and the Ten-Second Stop — Chapter 6 shows that copies sharing a credential count as one, that a backup drilled once a year is broken at the moment of an incident with probability 0.7331, and that scripting the first three stop-bleed steps cuts 1,200 seconds to ten. (https://5dlibrary.com/answers/en/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c06/) 第6章 防勒索:共因失效与十秒止血 — 第六章说明共用凭据之副本只算一份,一年才演练一回之备份,事件发生时已坏之机会为 0.7331;止血前三步写成脚本,从 1,200 秒压到十秒。 (https://5dlibrary.com/answers/zh/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c06/) - Ch.7 The Classics Examined: Sunzi and Kerckhoffs — Three example rows follow. With an attack probability of 0.02, hoping costs 2.0 in expectation and preparing 5.2, so not preparing is cheaper; yet the chance of at least one attack within ten years is already 0.1829. With an attack probability of 0.10, hoping costs 10.0 and preparing 6.0, and the chance of at least one attack within ten years is 0.6513. With an attack probability of 0.30, the two cost 30.0 and 8.0, and the ten-year chance is 0.9718. (https://5dlibrary.com/answers/en/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c07/) 第7章 原典考辨:孙子与柯克霍夫斯 — 再看三行例子。遭攻机会 0.02,指望不来,期望成本 2.0,准备 5.2,不准备反而便宜;可十年之内,至少遭攻一回之机会,已有 0.1829。遭攻机会 0.10,指望不来,期望成本 10.0,准备 6.0;十年之内,至少遭攻一回,0.6513。遭攻机会 0.30,两者 30.0 对 8.0;十年之内,0.9718。 (https://5dlibrary.com/answers/zh/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c07/) - Ch.8 Closing the Loop: a Monthly Check-up and Pre-Registered Tests — Test one needs 67 incidents under the normal approximation, with the null hypothesis rejected if at least 41 have a credential as root cause. But the exact binomial power at that size is only 0.7841, short of 0.8. Searching upwards, the first sample size whose exact power reaches the target is 69 incidents, with a threshold of 42 and a power of 0.8021. The protocol registers 69 incidents. (https://5dlibrary.com/answers/en/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c08/) 第8章 实证闭环:月度体检与预注册检验 — 检验一:正态近似,需要 67 起事件;拒绝原假设之门槛,是至少 41 起根因为凭据;可按精确二项分布算,检验力只有 0.7841,不足 0.8。往上逐个试,精确检验力,首次达标之样本量,是 69 起事件,门槛 42 起,检验力 0.8021。方案登记 69 起。 (https://5dlibrary.com/answers/zh/zero-trust-one-commander-origin-cloaking-secret-hygiene-anti-scraper/c08/) ## 5DL-B-337 · Heavyweight Cotton Areal-Density Mechanics, Heat-Transfer Coating Wash-Cycle Durability and Micro-Batch Custom Apparel Economics 中文书名:重磅棉织物克重力学、热转印涂层耐洗疲劳与小批量柔性定制经济学母典 DOI: pending Cite: Fan, Y. (2026). Heavyweight Cotton Areal-Density Mechanics, Heat-Transfer Coating Wash-Cycle Durability and Micro-Batch Custom Apparel Economics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable calculations (20 programs), sources and the author's hypotheses, each stating what would refute it. Abstract: This volume tests three popular beliefs for small brands and print studios that make heavyweight cotton T-shirts: the heavier the fabric the stiffer it is, the thicker the heat-transfer film the better it washes, and the bigger the run the bigger the profit. Each belief turns a conditional relation into an unconditional law. The fabric, care, cost and demand parameters in the book are illustrative, and all calculations come from its programs; the book is non-personalised business guidance, its durability conclusions are conditional and testable forecasts rather than promises, and washing and dimensional standards are cited only by number and scope. Chapters 1 and 2 use Peirce's cantilever and a numerical elastica to show that a quarter more weight multiplies bending rigidity by between 1.250 and 1.953, so the cube law is only a ceiling, and that stand follows rigidity while drape follows rigidity relative to weight, so the drape coefficient changes far less than rigidity. Chapter 3 splits shrinkage into relaxation and consolidation (cotton does not felt); under tumble drying the first wash covers only 0.643 of the shrinkage limit, and pattern allowances must divide rather than add. Chapter 4 describes wash failure of prints with Miner's cumulative damage and Weibull lives; under mixed care, averaging lives overstates the result 2.086 times. Chapter 5 finds an interior optimum in film thickness: doubling the film leaves only 0.257 of the optimal B10 life, and near the optimum the care condition matters more than thickness. Chapter 6 uses the newsvendor model to show that run length turns on the critical fractile, not on the unit-cost curve: the break-even coefficient of variation moves with mean demand, about 0.593 at 120 pieces and 0.392 at 40 under the illustrative parameters, and the plan that fills a 250-piece price break has the lowest unit cost and the lowest profit. Chapter 7 reads Song Yingxing's chapter on clothing in the Tiangong Kaiwu and Book I, Chapter III of Adam Smith's Wealth of Nations, and writes the distillation spec for the four kits. Chapter 8 gives the batch record, the distilled Micro-Batch Break-Even and Wash-Durability Calculator, interval-censored life estimation and three pre-registered tests. Marked boxes separate published evidence, runnable calculations (20 programs), sources and the author's hypotheses, each stating what would refute it. 摘要:本卷为做重磅棉 T 恤定制的小品牌与印花工作室检验三句流行说法:克重越高越挺,热转印膜越厚越耐洗,批量越大越赚钱。三句话都把有条件的关系说成了无条件的定律。书中的布料、洗法、成本与需求参数均为示意,全部计算由书中程序生成;本书是非个人化经营建议,耐久结论一律写成条件化、可检验的预测,不作任何绝对化承诺,洗涤与尺寸标准只引编号与适用范围。 第一、二章用皮尔斯悬臂法与弹性线数值解说明:克重加四分之一时,弯曲刚度的倍数落在 1.250 到 1.953 之间,三次方律只是上界;「挺」看刚度,「垂」看刚度与自重之比,悬垂系数的变化远小于刚度。第三章把缩水拆成松弛与压实两段,棉没有毡缩;滚筒烘干时第一次洗涤只走完缩水极限的 0.643,纸样补偿要用除法。第四章用迈纳累积损伤与韦布尔寿命描述印花水洗失效,混合洗法下平均寿命的算法高估 2.086 倍。第五章算出膜厚的内点最优,膜厚翻倍时 B10 寿命只剩最优值的 0.257,且在最优附近洗法比膜厚更要紧。 第六章用报童模型说明批量决策取决于临界分位而非单价曲线:盈亏临界变异系数随均值需求变化,示意参数下均值 120 件约为 0.593、40 件约为 0.392,为凑价格档做满 250 件的方案单价最低、利润也最低。第七章考辨宋应星《天工开物·乃服》与亚当·斯密《国富论》第一篇第三章,并写出给四件套的蒸馏规格。第八章给出批次记录、蒸馏小工具「小批量定制盈亏与耐洗试算器」、区间删失的寿命估计与三项预注册检验。书中以标注框区分已发表的证据、可运行的计算(20 段程序)、原典与作者的假说,每条假说都写明什么结果会推翻它。 Chapter answers: - Ch.1 Areal Density and Bending Rigidity: Peirce's Cantilever and the Limits of the Cube Law — Chapters 1 and 2 use Peirce's cantilever and a numerical elastica to show that a quarter more weight multiplies bending rigidity by between 1.250 and 1.953, so the cube law is only a ceiling, and that stand follows rigidity while drape follows rigidity relative to weight, so the drape coefficient changes far less than rigidity. (https://5dlibrary.com/answers/en/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c01/) 第1章 克重与弯曲刚度:皮尔斯悬臂与三次方律的边界 — 第一、二章用皮尔斯悬臂法与弹性线数值解说明:克重加四分之一时,弯曲刚度的倍数落在 1.250 到 1.953 之间,三次方律只是上界;「挺」看刚度,「垂」看刚度与自重之比,悬垂系数的变化远小于刚度。 (https://5dlibrary.com/answers/zh/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c01/) - Ch.2 Stand and Drape: Why the Drape Coefficient Lags Behind Rigidity — Stand follows rigidity, drape follows rigidity relative to weight, and they are two axes. With a quarter more weight, the cube law adds 95 % to rigidity and only about a fifth to the drape coefficient; under the linear law the drape coefficient does not change at all. Hand is multidimensional and cannot be captured by a weight figure. Fabrics twenty grams apart need dozens of instrument strips to be told apart. (https://5dlibrary.com/answers/en/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c02/) 第2章 挺与垂:悬垂系数为什么跟不上刚度 — 挺看刚度,垂看刚度与自重之比,两者是两根轴。克重加 25 % 时,三次方律下刚度加 95 %,悬垂系数只加 21 %;一次方律下悬垂系数完全不变。手感是多维的,一个克重数字代表不了它。二十克之差,仪器要几十条布条才分得开。 (https://5dlibrary.com/answers/zh/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c02/) - Ch.3 Shrinkage in Two Stages: Relaxation, Consolidation and the Design of Preshrinking — Chapter 3 splits shrinkage into relaxation and consolidation (cotton does not felt); under tumble drying the first wash covers only 0.643 of the shrinkage limit, and pattern allowances must divide rather than add. (https://5dlibrary.com/answers/en/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c03/) 第3章 缩水的两段:松弛、压实与预缩率设计 — 第三章把缩水拆成松弛与压实两段,棉没有毡缩;滚筒烘干时第一次洗涤只走完缩水极限的 0.643,纸样补偿要用除法。 (https://5dlibrary.com/answers/zh/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c03/) - Ch.4 A Little Damage per Wash: Cumulative Damage and Life Distributions of Heat-Transfer Film — Chapter 4 describes wash failure of prints with Miner's cumulative damage and Weibull lives; under mixed care, averaging lives overstates the result 2.086 times. (https://5dlibrary.com/answers/en/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c04/) 第4章 洗一次伤一点:热转印膜的累积损伤与寿命分布 — 第四章用迈纳累积损伤与韦布尔寿命描述印花水洗失效,混合洗法下平均寿命的算法高估 2.086 倍。 (https://5dlibrary.com/answers/zh/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c04/) - Ch.5 Does a Thicker Film Wash Better? Two Failure Modes and an Interior Optimum in Film Thickness — Chapter 5 finds an interior optimum in film thickness: doubling the film leaves only 0.257 of the optimal B10 life, and near the optimum the care condition matters more than thickness. (https://5dlibrary.com/answers/en/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c05/) 第5章 膜越厚越耐洗吗:两种失效与膜厚的内点最优 — 第五章算出膜厚的内点最优,膜厚翻倍时 B10 寿命只剩最优值的 0.257,且在最优附近洗法比膜厚更要紧。 (https://5dlibrary.com/answers/zh/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c05/) - Ch.6 Does a Bigger Run Make More Money? The Newsvendor Fractile and the Break-Even of Zero Inventory — Chapter 6 uses the newsvendor model to show that run length turns on the critical fractile, not on the unit-cost curve: the break-even coefficient of variation moves with mean demand, about 0.593 at 120 pieces and 0.392 at 40 under the illustrative parameters, and the plan that fills a 250-piece price break has the lowest unit cost and the lowest profit. (https://5dlibrary.com/answers/en/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c06/) 第6章 批量越大越赚钱吗:报童临界分位与零库存的盈亏临界 — 第六章用报童模型说明批量决策取决于临界分位而非单价曲线:盈亏临界变异系数随均值需求变化,示意参数下均值 120 件约为 0.593、40 件约为 0.392,为凑价格档做满 250 件的方案单价最低、利润也最低。 (https://5dlibrary.com/answers/zh/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c06/) - Ch.7 Reading the Sources: Song Yingxing's "Clothing" and Book I, Chapter III of The Wealth of Nations — On this illustrative curve, a twist multiplier of 2.0 gives a relative strength of only 0.629; at 3.0 it rises to 0.929, the optimum lies near 3.85, and at 4.0 the strength is 0.998. Beyond that it falls: 0.923 at 5.0, 0.804 at 6.0 and only 0.688 at 7.0. Song's sentence holds completely on the left half of the curve and fails on the right half. Twist the yarn too hard and the cloth is not stronger but weaker, and harsher in the hand as well. (https://5dlibrary.com/answers/en/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c07/) 第7章 原典考辨:《天工开物·乃服》与《国富论》第一篇第三章 — 按这条示意曲线,捻系数 2.0 时相对强力只有 0.629,3.0 时升到 0.929,最优点在 3.85 附近,4.0 时是 0.998;再往上就下降,5.0 时 0.923,6.0 时 0.804,7.0 时只剩 0.688。宋应星那句话在曲线左半边完全成立,在右半边就不成立了:捻得过紧,布反而不结实,而且手感发硬。 (https://5dlibrary.com/answers/zh/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c07/) - Ch.8 The Empirical Loop: Batch Records, Pre-Registered Tests and the Distillation Spec — Chapter 8 gives the batch record, the distilled Micro-Batch Break-Even and Wash-Durability Calculator, interval-censored life estimation and three pre-registered tests. Marked boxes separate published evidence, runnable calculations (20 programs), sources and the author's hypotheses, each stating what would refute it. (https://5dlibrary.com/answers/en/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c08/) 第8章 实证闭环:批次记录、预注册检验与蒸馏规格 — 第八章给出批次记录、蒸馏小工具「小批量定制盈亏与耐洗试算器」、区间删失的寿命估计与三项预注册检验。书中以标注框区分已发表的证据、可运行的计算(20 段程序)、原典与作者的假说,每条假说都写明什么结果会推翻它。 (https://5dlibrary.com/answers/zh/heavyweight-cotton-gsm-bending-heat-transfer-wash-fatigue-micro-batch-economics/c08/) ## 5DL-B-338 · Cross-Border Dental and Wellness Service Pricing, Multi-Stage Quality Signalling and Client Escrow Mechanics 中文书名:跨境高端健康与齿科服务定价、多阶段品质信号甄别与客户信任托管母典 DOI: pending Cite: Fan, Y. (2026). Cross-Border Dental and Wellness Service Pricing, Multi-Stage Quality Signalling and Client Escrow Mechanics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its prices, probabilities and costs are illustrative and describe no real provider, country or price; all calculations are generated by its programs. It is general information on medicine, finance and law, not advice, and it makes no claim of benefit or success rate; plans must be decided in person with a licensed dentist or physician, and client money must be held by a licensed third party. Abstract: The trade in cross-border dentistry and premium wellness care rests on three popular sayings: a plan priced at half the local figure abroad must be the better deal; an expensive clinic is a good clinic, because price is a signal; and paying in full for a discount is simplest. This book refutes each saying or bounds it, and deals only with pricing, information and contract structure. Its prices, probabilities and costs are illustrative and describe no real provider, country or price; all calculations are generated by its programs. It is general information on medicine, finance and law, not advice, and it makes no claim of benefit or success rate; plans must be decided in person with a licensed dentist or physician, and client money must be held by a licensed third party. Chapter 1 completes the quotation as a five-line account: in an illustrative two-stay plan the sticker gap of 14,000 keeps only 0.159 of its size once stays, return trips, expected rework and the present value of upkeep are counted, and the plan abroad is cheaper only while its rework probability is below a threshold of 0.2280 that is highly sensitive to travel costs. In Chapter 2, twenty years of upkeep under Weibull wear are worth 0.378 of the sticker, and with short-lived components the advantage abroad turns over around year 10. In Chapter 3's lemons market of clinics, invisible quality cuts the surplus to 0.250, a well-placed pass-or-fail certificate restores it to 1.000, price alone pools in a one-off sale, and the repeat-business signal window for distant clients is 0.0094 of its local size. Chapter 4 compares five signals; consecutive case records have a likelihood ratio of 5.500, while a provider that publishes only clean cases leaves readers with an average confidence of 0.960 in a false conclusion, against 0.0232 for the complete series. Chapter 5 treats prepayment as credit risk: at a stop probability of 0.05 per stage and a recovery rate of 0.2, expected loss is 1,476 when paying in full, 764 when paying stage by stage and zero with a licensed escrow agent; after a fair cost of capital and the escrow fee, escrow leaves the client better off only above a stop probability of 0.0364, which corrects the blueprint's claim of a strict improvement. Chapter 6 prices a ten-year upkeep plan at a level net premium of 157.9 a year; a pool charged at expected cost runs dry within ten years about half the time, and a loading of 0.20 brings a pool of two hundred down to 0.054. Chapter 7 examines Sun Simiao's warning against flaunting one's name and laws 215 to 223 of the Code of Hammurabi, whose heavy penalties push a healer's break-even success probability to 0.909 and above. Chapter 8 sets out the comparison-and-escrow ledger, the Cross-Border Service Total-Cost and Escrow Release Calculator and a pre-registered test plan: the total-cost test needs 71 paired records and the signal test 181 plans per group. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:跨境齿科与高端健康服务的生意靠三句流行说法撑着:境外标价只有一半,越便宜越划算;贵的机构品质好,价格就是信号;一次付清拿折扣最省心。本书逐条证伪或立界,只讨论定价、信息与契约结构。书中价格、概率与成本均为示意参数,不对应任何真实机构、国家或价格,也不点名评价任何诊所、医院或中介;全部计算由书中程序生成。本书是非个人化的医疗、财务与法律信息,不是建议,不写任何功效或成功率承诺;方案须由本人与持牌牙医或医师当面决定,资金托管须由持牌第三方办理。 第一章把报价补成五项总账:示意两段式方案标价差 14,000,计入停留、回程、返修期望与维保现值后只剩标价差的 0.159;境外方案只在返修概率低于阈值 0.2280 时更便宜,而这个阈值对差旅费用极其敏感。第二章用韦布尔磨损与更新方程计算维保,二十年维保现值相当于标价的 0.378,部件寿命短时境外方案的优势在第十年前后翻转。第三章把柠檬市场搬到诊所之间:看不见品质时市场剩余只剩 0.250,一张及格线合适的认证就恢复到 1.000;一次性交易里只看价格必然混同,靠回头客的信号窗口在远途跨境市场只剩境内常规客户的 0.0094。第四章比较五种信号的似然比,连续病例记录高达 5.500,而只公开无返修病例的机构能让读者以 0.960 的平均信心误判,看到完整序列时只有 0.0232。 第五章把预付写成信用风险:每段中止概率 0.05、回收率 0.2 时,一次付清的期望损失是 1,476,按段预付 764,持牌托管为零;扣除公平资金成本与托管费后,托管只在每段中止概率高于 0.0364 时对客户净划算,这是对立项蓝图「严格改善」一说的修正。第六章为十年维保计划定价,均衡净保费每年 157.9,按期望收费的资金池约一半会在十年内用光,附加 0.20 时两百人的池子降到 0.054。第七章考辨孙思邈《大医精诚》反对「炫耀声名」与《汉谟拉比法典》第 215 至 223 条按结果计酬,后者重罚使医者的盈亏平衡成功率高达 0.909 以上。第八章给出比价与托管台账、蒸馏小工具「跨境服务总成本与托管放款试算器」与预注册检验方案:总成本检验需 71 份配对记录,信号检验每组需 181 份方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Illusion of the Sticker Gap: Expected Total Cost and the Rework Threshold — Chapter 1 completes the quotation as a five-line account: in an illustrative two-stay plan the sticker gap of 14,000 keeps only 0.159 of its size once stays, return trips, expected rework and the present value of upkeep are counted, and the plan abroad is cheaper only while its rework probability is below a threshold of 0.2280 that is highly sensitive to travel costs. In Chapter 2, twenty years of upkeep under Weibull wear are worth 0.378 of the sticker, and with short-lived components the advantage abroad turns over around year 10. In Chapter 3's lemons market of clinics, invisible quality cuts the surplus to 0.250, a well-placed pass-or-fail certificate restores it to 1.000, price alone pools in a one-off sale, and the repeat-business signal window for distant clients is 0.0094 of its local size. (https://5dlibrary.com/answers/en/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c01/) 第1章 标价差的幻觉:期望总成本与返修概率阈值 — 第一章把报价补成五项总账:示意两段式方案标价差 14,000,计入停留、回程、返修期望与维保现值后只剩标价差的 0.159;境外方案只在返修概率低于阈值 0.2280 时更便宜,而这个阈值对差旅费用极其敏感。 (https://5dlibrary.com/answers/zh/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c01/) - Ch.2 The Upkeep Annuity: Wear, Discounting and the Year the Gap Turns Over — Upkeep consists of a routine cost and random replacements, and replacements are described by Weibull wear and the discrete renewal equation. Over a twenty-year holding period, with a scale of 15 years and $k$ equal to 2, the expected number of replacements is 1.112 and the present value of upkeep at 0.03 is 10,594, which is 0.378 of the home sticker; the naive estimate from the yearly fee is only 8,000. The plan abroad starts 4,400 ahead. With upkeep taken over at home it turns over in year 20 when the component scale is 15 years, in year 10 when it is 10 years, and in year 6 when it is 7 years; flying back to the original provider delays or removes the turn. Component life and the place of upkeep decide the fate of the plan abroad more than its sticker does. (https://5dlibrary.com/answers/en/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c02/) 第2章 维保年金:部件磨损、折现与价差翻转之年 — 维保由例行费用与随机更换两部分组成;更换用韦布尔磨损与离散更新方程描述。二十年持有期里,尺度 15 年、$k$ 取 2 时期望更换 1.112 次,按利率 0.03 折现的维保现值是 10,594,相当于境内标价的 0.378;朴素的年费估计只有 8,000。境外方案在第零年领先 4,400,境内接手维保、部件尺度 15 年时第二十年翻转;尺度 10 年时第十年翻转;尺度 7 年时第六年翻转;飞回原机构维保则能把翻转推迟或消除。部件寿命与维保地点,比标价更能决定境外方案的成败。 (https://5dlibrary.com/answers/zh/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c02/) - Ch.3 Lemons and Price Signals: When a High Price Only Pools — The lemons market works among clinics as it does among used cars. In an illustrative market of five grades, with quality invisible only the two lowest grades trade and the surplus is 0.250 of its full-information level; a certificate with its pass mark at grade 3 restores the share to 1.000, while pass marks at grade 2 or grade 4 recover only 0.700 and 0.800. In a one-off sale with price as the only visible variable, the poor provider always copies the high price and the market pools. A review holdback can separate the two, but at a high price of 12,000 with a failure probability of 0.35 it must reach 0.952, and with larger gaps it cannot do the job at all. The window for signals that rest on returning clients is, for distant cross-border clients, only 0.0094 of its size for routine clients at home. (https://5dlibrary.com/answers/en/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c03/) 第3章 柠檬市场与价格信号:高价何时只是混同 — 柠檬市场在诊所之间同样成立:五个等级的示意市场里,看不见品质时只有最低两级成交,剩余只剩完全信息时的 0.250;一张及格线定在三级的认证就把份额找回到 1.000,线划在二级或四级则只有 0.700 与 0.800。一次性交易里,只看价格时差机构总会模仿高价,市场混同;复查留存款能把它们分开,但高价 12,000、差机构失手概率 0.35 时需要留存 0.952,价差再大就挡不住。靠回头客的信号窗口,在远途跨境市场里只剩境内常规客户的 0.0094。 (https://5dlibrary.com/answers/zh/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c03/) - Ch.4 Credence Goods and Multi-Stage Signals: Credentials, Consecutive Cases and the Bayesian Posterior — Chapter 4 compares five signals; consecutive case records have a likelihood ratio of 5.500, while a provider that publishes only clean cases leaves readers with an average confidence of 0.960 in a false conclusion, against 0.0232 for the complete series. (https://5dlibrary.com/answers/en/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c04/) 第4章 信任品与多阶段信号:资质、连续病例与贝叶斯后验 — 第四章比较五种信号的似然比,连续病例记录高达 5.500,而只公开无返修病例的机构能让读者以 0.960 的平均信心误判,看到完整序列时只有 0.0232。 (https://5dlibrary.com/answers/zh/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c04/) - Ch.5 Staged Escrow: Stop Probability, Recovery Rate and the Review Holdback — Chapter 5 treats prepayment as credit risk: at a stop probability of 0.05 per stage and a recovery rate of 0.2, expected loss is 1,476 when paying in full, 764 when paying stage by stage and zero with a licensed escrow agent; after a fair cost of capital and the escrow fee, escrow leaves the client better off only above a stop probability of 0.0364, which corrects the blueprint's claim of a strict improvement. (https://5dlibrary.com/answers/en/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c05/) 第5章 分阶段托管:中止概率、回收率与复查留存款 — 第五章把预付写成信用风险:每段中止概率 0.05、回收率 0.2 时,一次付清的期望损失是 1,476,按段预付 764,持牌托管为零;扣除公平资金成本与托管费后,托管只在每段中止概率高于 0.0364 时对客户净划算,这是对立项蓝图「严格改善」一说的修正。 (https://5dlibrary.com/answers/zh/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c05/) - Ch.6 The Upkeep Reserve: Actuarial Values, the Reserve Curve and the Probability of Ruin — Chapter 6 prices a ten-year upkeep plan at a level net premium of 157.9 a year; a pool charged at expected cost runs dry within ten years about half the time, and a loading of 0.20 brings a pool of two hundred down to 0.054. (https://5dlibrary.com/answers/en/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c06/) 第6章 维保准备金:精算现值、准备金曲线与破产概率 — 第六章为十年维保计划定价,均衡净保费每年 157.9,按期望收费的资金池约一半会在十年内用光,附加 0.20 时两百人的池子降到 0.054。 (https://5dlibrary.com/answers/zh/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c06/) - Ch.7 Classical Sources East and West: Sun Simiao on Flaunting a Name, and Hammurabi on Payment by Result — Chapter 7 examines Sun Simiao's warning against flaunting one's name and laws 215 to 223 of the Code of Hammurabi, whose heavy penalties push a healer's break-even success probability to 0.909 and above. (https://5dlibrary.com/answers/en/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c07/) 第7章 东西方原典考辨:孙思邈论「炫耀声名」与汉谟拉比按结果计酬 — 第七章考辨孙思邈《大医精诚》反对「炫耀声名」与《汉谟拉比法典》第 215 至 223 条按结果计酬,后者重罚使医者的盈亏平衡成功率高达 0.909 以上。 (https://5dlibrary.com/answers/zh/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c07/) - Ch.8 Closing the Loop: The Comparison-and-Escrow Ledger, the Calculator and a Pre-Registered Test — Chapter 8 sets out the comparison-and-escrow ledger, the Cross-Border Service Total-Cost and Escrow Release Calculator and a pre-registered test plan: the total-cost test needs 71 paired records and the signal test 181 plans per group. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c08/) 第8章 实证闭环:比价与托管台账、试算器与预注册检验 — 第八章给出比价与托管台账、蒸馏小工具「跨境服务总成本与托管放款试算器」与预注册检验方案:总成本检验需 71 份配对记录,信号检验每组需 181 份方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cross-border-dental-wellness-pricing-quality-signalling-client-escrow/c08/) ## 5DL-B-339 · Fickian Transdermal Diffusion of Botanical Massage Oils, Thermal Comfort Kinetics and Wellness Studio Retention 中文书名:草本芳香按摩油菲克透皮扩散、微循环热舒适度与理疗空间留存母典 DOI: pending Cite: Fan, Y. (2026). Fickian Transdermal Diffusion of Botanical Massage Oils, Thermal Comfort Kinetics and Wellness Studio Retention (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All concentrations, amounts, vapour pressures, temperatures and rating parameters are illustrative, and every calculation is generated by the book's programs. The book is not medical advice: it claims no effect for any oil, carrier, temperature or technique, gives no recommendation on dose or concentration, and leaves dilution, patch testing and use in pregnancy or with children to qualified professionals. Abstract: The aromatherapy massage trade repeats three claims: a stronger essential oil goes deeper; warmer is always more comfortable; and whether a client returns depends on her average satisfaction across the session. This book uses Fickian diffusion, heat transfer and peak-end evaluation to turn each claim into a proposition that data can refute, and tests it. Ginger and frankincense appear only as two volatile oils whose molecular weights, partition coefficients and vapour pressures are discussed. All concentrations, amounts, vapour pressures, temperatures and rating parameters are illustrative, and every calculation is generated by the book's programs. The book is not medical advice: it claims no effect for any oil, carrier, temperature or technique, gives no recommendation on dose or concentration, and leaves dilution, patch testing and use in pregnancy or with children to qualified professionals. Nor is it individual business advice, and it names no real brand or studio. Chapters 1 to 3 correct the project blueprint. Terpenes mix with medium-chain (C8/C10) triglyceride in any proportion, so there is no saturation ceiling; the neat oil has about 18.12 times the activity of a 2 % blend. An illustrative stratum corneum has a lag time of about 62.5 minutes, and doubling the concentration does not push the profile deeper. In a sixty-minute open session 99.994 % of the alpha-pinene evaporates, while zingiberene loses 49.385 % to the air and leaves 0.2445 % in the stratum corneum; raising the blend from 1 % to 10 % moves the share reaching the skin only from 0.2447 % to 0.2583 %. Chapter 4 shows that warmth comes from hands, towels and stones rather than hot oil: a mass of oil at 45 degrees Celsius brings the skin only to 36.75 degrees, a stone at 50 degrees reaches 43.77 degrees, and a starting temperature no higher than 48.7 degrees keeps a stone at 43 degrees. Chapter 5 searches temperature plans with an adaptation model: the balanced optimum is a neutral base, a late warm passage and an easing down, with a mean of 6.84 and a peak-end score of 8.55, while the plan that chases peak and end alone has a mean of only 4.88 and is rejected. Chapter 6 weighs the peak-end rule against a meta-analysis of 2022: adding ten minutes of tea after a rushed ending moves the mean only from 6.92 to 6.93 but raises the peak-end score from 6.50 to 8.00, and the ratio between the repurchase elasticities of ending and middle runs between 15.25 and 1.00 with the memory weight, which a studio must measure. Chapter 7 examines "massage and medicated wine" in the Basic Questions and Fick's paper of 1855 and closes with clauses for the four kits. Chapter 8 gives the service record, the Massage-Oil Flux and Peak-End Retention Calculator and a pre-registered test: detecting a gap of 0.35 against 0.45 in repurchase needs 376 first-time clients per group. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:芳香按摩行业有三条流行说法:精油浓度越高渗得越深;按摩时越热越舒服;客人回不回来取决于整场服务的平均满意度。本书用菲克扩散、传热学与峰终评价把三条说法写成可以被数据推翻的命题,逐条检验。生姜与乳香只作为两种挥发油讨论其分子量、分配系数与蒸气压。书中浓度、用量、蒸气压、温度与评分参数均为示意,全部计算由书中程序生成;本书不是医疗建议,不宣称任何精油、载体、温度或手法的功效,不给任何剂量或浓度建议,稀释、斜贴测试、孕期与儿童是否适用交由合格专业人员判断;本书也不是个人化经营建议,不点名任何真实品牌或门店。 第一至三章修正立项蓝图:萜烯与中链甘油三酯(C8 / C10 辛癸酸甘油三酯)任意比例互溶,不存在饱和溶解度的天花板,纯油活度约为 2 % 配方的 18.12 倍;示意角质层的滞后时间约 62.5 分钟,浓度加倍不会让剖面更深。60 分钟开放按摩里,α-蒎烯 99.994 % 挥发,姜烯 49.385 % 挥发、0.2445 % 留在角质层;浓度从 1 % 加到 10 %,进入皮肤的比例只从 0.2447 % 变到 0.2583 %。第四章说明温热来自手、毛巾与石块而非热油:45 摄氏度的整块油只把皮肤拉到 36.75 摄氏度,50 摄氏度的石块达到 43.77 摄氏度,石块起始温度不高于 48.7 摄氏度才能守住 43 摄氏度。 第五章用适应模型搜索温度安排:平衡最优是中性底色、靠后温热段再回落,平均 6.84、峰终 8.55;只追峰终的方案平均仅 4.88,本书拒绝采用。第六章核对峰终定律与 2022 年荟萃分析,在仓促结尾后加 10 分钟茶歇,平均只从 6.92 变到 6.93,峰终却从 6.50 升到 8.00;结尾与中段的复购弹性之比随记忆权重在 15.25 与 1.00 之间变化,须在门店实测。第七章考辨《素问·血气形志》的「按摩醪药」与菲克 1855 年原文,章末给出四件套蒸馏条款。第八章给出服务记录、按摩油配方通量与峰终留存试算器与预注册检验:检出 0.35 对 0.45 的复购差距,每组需 376 位首次客人。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者假说。 Chapter answers: - Ch.1 Fick's Laws and the Membrane Model: Steady Flux, Lag Time and the Myth of "Going Deeper" — Chapters 1 to 3 correct the project blueprint. Terpenes mix with medium-chain (C8/C10) triglyceride in any proportion, so there is no saturation ceiling; the neat oil has about 18.12 times the activity of a 2 % blend. An illustrative stratum corneum has a lag time of about 62.5 minutes, and doubling the concentration does not push the profile deeper. In a sixty-minute open session 99.994 % of the alpha-pinene evaporates, while zingiberene loses 49.385 % to the air and leaves 0.2445 % in the stratum corneum; raising the blend from 1 % to 10 % moves the share reaching the skin only from 0.2447 % to 0.2583 %. (https://5dlibrary.com/answers/en/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c01/) 第1章 菲克定律与膜模型:稳态通量、滞后时间与「渗得更深」的误会 — 第一至三章修正立项蓝图:萜烯与中链甘油三酯(C8 / C10 辛癸酸甘油三酯)任意比例互溶,不存在饱和溶解度的天花板,纯油活度约为 2 % 配方的 18.12 倍;示意角质层的滞后时间约 62.5 分钟,浓度加倍不会让剖面更深。60 分钟开放按摩里,α-蒎烯 99.994 % 挥发,姜烯 49.385 % 挥发、0.2445 % 留在角质层;浓度从 1 % 加到 10 %,进入皮肤的比例只从 0.2447 % 变到 0.2583 %。 (https://5dlibrary.com/answers/zh/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c01/) - Ch.2 Partition Coefficients and Carriers: the Activity of Ginger and Frankincense Constituents in Medium-Chain (C8/C10) Triglyceride — The partition coefficient governs a molecule's tendency to leave its vehicle. Under this book's illustrative split of the Potts-Guy regression, the partition coefficient between stratum corneum and MCT is $10^{-0.29 \log K_{ow}}$: 0.0530 for alpha-pinene and 0.0182 for zingiberene, so that the more lipophilic the constituent, the less willing it is to leave the film. Illustrative lag times are about 60.4 minutes for the monoterpenes and about 145.3 minutes for the sesquiterpenes. Volume fraction is not activity: at the same 2 %, the essential oil has a mole fraction of 0.0552 in MCT and 0.0960 in a long-chain triglyceride, and the neat oil has 18.12 times the activity of the 2 % MCT blend, a ceiling rather than a saturation solubility. Even in a sealed film, no more than 0.2688 % of the applied amount crosses the illustrative stratum corneum in sixty minutes. (https://5dlibrary.com/answers/en/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c02/) 第2章 分配系数与载体:生姜、乳香成分在中链甘油三酯(C8 / C10)里的活度 — 分配系数决定分子离开载体的倾向。按本书对 Potts 与 Guy 经验式的示意拆法,角质层与 MCT 之间的分配系数是 $10^{-0.29 \log K_{ow}}$:α-蒎烯 0.0530,姜烯 0.0182,越亲脂越不愿离开油膜。单萜的示意滞后约 60.4 分钟,倍半萜约 145.3 分钟。体积分数不是活度:同为 2 %,精油在 MCT 里的摩尔分数是 0.0552,在长链甘油三酯里是 0.0960;纯油的活度是 2 % MCT 配方的 18.12 倍,这是天花板,而不是饱和溶解度。即使油膜封闭不挥发,60 分钟里穿过示意角质层的量也不超过用量的 0.2688 %。 (https://5dlibrary.com/answers/zh/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c02/) - Ch.3 Evaporation from an Open Surface: the Mass Balance of a Massage — Evaporation follows Raoult's law and gas-side mass transfer. On the illustrative film the monoterpenes have evaporation half-lives under a minute and the sesquiterpenes about 48.72 to 81.20 minutes; the evaporation rate constant is hundreds to more than twenty thousand times the uptake rate constant. In a sixty-minute open session 99.994 % of the alpha-pinene goes to the air, while zingiberene loses 49.385 % to evaporation, keeps 50.369 % in the film and leaves 0.2445 % in the stratum corneum, with only 0.00187 % through it. Raising the concentration from 1 % to 10 % raises the absolute amount in the skin roughly in proportion, while the share moves only from 0.2447 % to 0.2583 %. Covering with a towel changes where the molecules go far more than strengthening the blend. (https://5dlibrary.com/answers/en/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c03/) 第3章 开放表面的挥发:一场按摩的质量平衡 — 挥发服从拉乌尔定律与气侧传质:在示意油膜上,单萜的挥发半衰期不到一分钟,倍半萜约 48.72 到 81.20 分钟;挥发速率常数是透过速率常数的数百到两万多倍。60 分钟的开放按摩里,α-蒎烯 99.994 % 进入空气;姜烯 49.385 % 挥发、50.369 % 留在油膜、0.2445 % 留在角质层,穿过的只有 0.00187 %。浓度从 1 % 加到 10 %,进入皮肤的绝对量大致成比例增加,比例却只从 0.2447 % 变到 0.2583 %;盖毛巾改变去向的力度,比加浓度大得多。 (https://5dlibrary.com/answers/zh/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c03/) - Ch.4 First-Order Response of Skin Temperature: Contact Temperature, the Heat Capacity of an Oil Film and a Time-Temperature Red Line — Chapter 4 shows that warmth comes from hands, towels and stones rather than hot oil: a mass of oil at 45 degrees Celsius brings the skin only to 36.75 degrees, a stone at 50 degrees reaches 43.77 degrees, and a starting temperature no higher than 48.7 degrees keeps a stone at 43 degrees. (https://5dlibrary.com/answers/en/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c04/) 第4章 皮肤温度的一阶响应:接触温度、油膜热容与时间温度红线 — 第四章说明温热来自手、毛巾与石块而非热油:45 摄氏度的整块油只把皮肤拉到 36.75 摄氏度,50 摄氏度的石块达到 43.77 摄氏度,石块起始温度不高于 48.7 摄氏度才能守住 43 摄氏度。 (https://5dlibrary.com/answers/zh/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c04/) - Ch.5 The Comfort Band and Adaptation: an Optimal Temperature Trajectory, Not "the Hotter the Better" — Chapter 5 searches temperature plans with an adaptation model: the balanced optimum is a neutral base, a late warm passage and an easing down, with a mean of 6.84 and a peak-end score of 8.55, while the plan that chases peak and end alone has a mean of only 4.88 and is rejected. (https://5dlibrary.com/answers/en/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c05/) 第5章 舒适区间与适应:最优温度轨迹,而不是越热越好 — 第五章用适应模型搜索温度安排:平衡最优是中性底色、靠后温热段再回落,平均 6.84、峰终 8.55;只追峰终的方案平均仅 4.88,本书拒绝采用。 (https://5dlibrary.com/answers/zh/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c05/) - Ch.6 The Peak-End Rule and Repurchase: the Elasticity of the Ending and Retention Curves — Chapter 6 weighs the peak-end rule against a meta-analysis of 2022: adding ten minutes of tea after a rushed ending moves the mean only from 6.92 to 6.93 but raises the peak-end score from 6.50 to 8.00, and the ratio between the repurchase elasticities of ending and middle runs between 15.25 and 1.00 with the memory weight, which a studio must measure. (https://5dlibrary.com/answers/en/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c06/) 第6章 峰终定律与复购:结尾段的弹性与留存曲线 — 第六章核对峰终定律与 2022 年荟萃分析,在仓促结尾后加 10 分钟茶歇,平均只从 6.92 变到 6.93,峰终却从 6.50 升到 8.00;结尾与中段的复购弹性之比随记忆权重在 15.25 与 1.00 之间变化,须在门店实测。 (https://5dlibrary.com/answers/zh/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c06/) - Ch.7 Classical Sources: "Massage and Medicated Wine" in the Basic Questions, and Fick in 1855 — Chapter 7 examines "massage and medicated wine" in the Basic Questions and Fick's paper of 1855 and closes with clauses for the four kits. (https://5dlibrary.com/answers/en/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c07/) 第7章 原典考辨:《素问·血气形志》的「按摩醪药」与菲克 1855 — 第七章考辨《素问·血气形志》的「按摩醪药」与菲克 1855 年原文,章末给出四件套蒸馏条款。 (https://5dlibrary.com/answers/zh/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c07/) - Ch.8 The Empirical Loop and the Distillation Spec: Service Records, the Calculator and a Refutation after One Quarter — Chapter 8 gives the service record, the Massage-Oil Flux and Peak-End Retention Calculator and a pre-registered test: detecting a gap of 0.35 against 0.45 in repurchase needs 376 first-time clients per group. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c08/) 第8章 实证闭环与蒸馏规格:服务记录、试算器与一个季度后的证伪 — 第八章给出服务记录、按摩油配方通量与峰终留存试算器与预注册检验:检出 0.35 对 0.45 的复购差距,每组需 376 位首次客人。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/botanical-massage-oil-fickian-diffusion-thermal-comfort-studio-retention/c08/) ## 5DL-B-340 · Distributed Multi-City Team Coordination, Asynchronous Deliverable Verification and Regional Talent Allocation across KL, Taipei and HCMC 中文书名:跨时区多城市分布式团队协同、异步交付核验与区域人才配置效率母典 DOI: pending Cite: Fan, Y. (2026). Distributed Multi-City Team Coordination, Asynchronous Deliverable Verification and Regional Talent Allocation across KL, Taipei and HCMC (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its schedules, lunch breaks, cost coefficients, exchange-rate volatilities and price levels are illustrative, and all calculations are generated by its programs. It is general management guidance, not individual, legal or tax advice; labour, tax and work-pass matters in the three cities must be confirmed by licensed local advisers. Abstract: Teams working at once from Kuala Lumpur, Taipei and Ho Chi Minh City inherit three popular beliefs: Kuala Lumpur and Taipei share UTC+8 and Ho Chi Minh City is only an hour behind, so coordination costs nothing; the more acceptance checks the better; and the steeper the incentive slope the better. This book refutes each belief or draws its boundary. Its schedules, lunch breaks, cost coefficients, exchange-rate volatilities and price levels are illustrative, and all calculations are generated by its programs. It is general management guidance, not individual, legal or tax advice; labour, tax and work-pass matters in the three cities must be confirmed by licensed local advisers. It labels no group by city or nationality and involves no multi-level profit sharing. Chapters 1 to 3 show that a one-hour offset removes at most 1.00 hour of a 9-hour office day, while Kuala Lumpur and Taipei, on one clock, lose 2.00 hours to misaligned lunches; that 0.791 of the slowdown of cross-city work comes from the number of hands, and a pooled review board cuts waits by a factor of 3.71; that task granularity has an interior optimum, 4 pieces in one office but 2 across three cities; and that handoff notes are best batched every 1.83 hours, while handoffs sent from Ho Chi Minh City to Kuala Lumpur in its last hour wait 14.99 hours on average. Chapters 4 and 5 show that for a stable process the kp rule admits only full or no inspection, and that an interior optimum comes only from the value of information when defect rates are unstable: sampling 15 items before deciding saves 38.3 hours per lot; the sample worth paying for is 42 items, and under drift one samples every 2.32 working days. Chapter 6 shows that when handoff notes go uncounted the optimal slope falls from 0.667 to 0.364, that exchange-rate noise lowers it only to 0.663, but that price levels make one slope worth between 1.67 and 2.50 baskets of goods. Chapter 7 examines the Guanzi on dividing the people well, Babbage and Taylor; Chapter 8 gives a weekly ledger, the Three-City Async Handoff and Sampling Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:在吉隆坡、台北与胡志明市同时开工的团队,常听到三条流行说法:吉隆坡与台北同为 UTC+8、胡志明市只差一小时,所以没有协同成本;验收抽查越勤越好;激励斜率越陡越好。本书逐条证伪或立界。书中的作息、午休、成本系数、汇率波动率与价格水平均为示意参数,全部计算由书中程序生成;本书是非个人化管理建议,不构成法律或税务建议,三城劳动、税务与准证事项须由当地持牌顾问确认;书中不以城市或国籍给任何人群贴标签,也不涉及任何多层分润。 第一至三章说明,一小时偏移最多削掉 9 个小时办公日里的 1.00 个小时,同时区的吉隆坡与台北因午休错开反而少了 2.00 个小时;跨城工作变慢,0.791 来自经手人数,合池评审能把等待缩短为 3.71 分之一;切分粒度存在内点最优,同一份工作在办公室切 4 块、在三城切 2 块;交接说明的最优批量间隔是 1.83 个小时,胡志明市最后一小时发往吉隆坡的交接平均要等 14.99 个小时。 第四、五章说明,过程稳定时 kp 规则只认全检或零检,内点最优只来自缺陷率不稳定时的信息价值:先抽 15 件再决定,每批省下 38.3 个工时;值得付钱的样本是 42 件,漂移时每 2.32 个工作日抽一次。第六章说明,交接说明不被计数时最优斜率从 0.667 降到 0.364,汇率噪声只让它降到 0.663,价格水平差异却让同一斜率的实际强度在 1.67 到 2.50 篮商品之间。第七章考辨《管子·乘马》「善分民」与巴贝奇、泰勒,第八章给出周度台账、三城异步交接与抽检频率试算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 One Time Zone Is Not Zero Coordination Cost: Shared Windows, Hands and Review Queues — Chapters 1 to 3 show that a one-hour offset removes at most 1.00 hour of a 9-hour office day, while Kuala Lumpur and Taipei, on one clock, lose 2.00 hours to misaligned lunches; that 0.791 of the slowdown of cross-city work comes from the number of hands, and a pooled review board cuts waits by a factor of 3.71; that task granularity has an interior optimum, 4 pieces in one office but 2 across three cities; and that handoff notes are best batched every 1.83 hours, while handoffs sent from Ho Chi Minh City to Kuala Lumpur in its last hour wait 14.99 hours on average. (https://5dlibrary.com/answers/en/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c01/) 第1章 同一时区不等于零协同成本:重叠窗口、交接人数与评审队列 — 第一至三章说明,一小时偏移最多削掉 9 个小时办公日里的 1.00 个小时,同时区的吉隆坡与台北因午休错开反而少了 2.00 个小时;跨城工作变慢,0.791 来自经手人数,合池评审能把等待缩短为 3.71 分之一;切分粒度存在内点最优,同一份工作在办公室切 4 块、在三城切 2 块;交接说明的最优批量间隔是 1.83 个小时,胡志明市最后一小时发往吉隆坡的交接平均要等 14.99 个小时。 (https://5dlibrary.com/answers/zh/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c01/) - Ch.2 An Interior Optimum for Task Granularity: Handoff Loss, Brooks's Channels and Cross-City Assignment — Task granularity has an interior optimum, and the optimal number of pieces moves as $\sqrt{((1-f)-\delta)\,T/H}$. The same work is best cut into 4 pieces in one office, 2 in a three-city team without a handoff protocol, and 3 with one. Brooks's channel count reminds us that with people split evenly over three cities, two thirds of the channels in an all-to-all structure cross a city line, while a three-hub structure cuts the number of channels to about three tenths. Cross-city assignment must write the handoff charge into its objective, or the exact optimum on hourly cost alone will create crossing after crossing. (https://5dlibrary.com/answers/en/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c02/) 第2章 切分粒度的内点最优:交接损耗、Brooks 通道与跨城指派 — 切分粒度有内点最优,最优块数按 $\sqrt{((1-f)-\delta)\,T/H}$ 变化。同一份工作,在办公室里最优是 4 块,在没有交接协议的三城团队里是 2 块,加上协议回到 3 块。Brooks 的通道数提醒我们,人数平分在三座城市时,全连通结构里有三分之二的通道跨城;三中枢结构把通道压到约三成。跨城指派要把交接费写进目标,否则精确的单价最优会制造频繁跨城。 (https://5dlibrary.com/answers/zh/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c02/) - Ch.3 A Protocol for Asynchronous Handoff: Batch Interval, Inbox Rhythm and the Edge of the Day — The optimal interval between batches of handoff notes is 1.83 hours, and handing everything over at the end of the day costs 2.57 times as much. Once inbox rhythm, lunch and the edge of the day are layered on top, a one-hour offset can be cancelled out or blown up into a whole night: handoffs that Ho Chi Minh City sends to Kuala Lumpur in its last hour wait 14.99 hours on average. Writing complete notes takes less effort than reading the inbox more often. The protocol decides H, and it decides delta too. (https://5dlibrary.com/answers/en/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c03/) 第3章 异步交接协议:批量间隔、收件节拍与日界 — 交接说明的最优批量间隔是 1.83 个小时,下班前集中交接的成本是它的 2.57 倍。收件节拍、午休与日界叠加以后,一小时时差可能抵消,也可能放大成一整夜:胡志明市最后一小时发往吉隆坡的交接,平均要等 14.99 个小时。把说明写完整,比加密收件节拍更省力。协议决定了 H,也决定了 δ。 (https://5dlibrary.com/answers/zh/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c03/) - Ch.4 Acceptance Sampling: Operating Characteristics, the kp Rule and the Boundaries of All or None — Chapters 4 and 5 show that for a stable process the kp rule admits only full or no inspection, and that an interior optimum comes only from the value of information when defect rates are unstable: sampling 15 items before deciding saves 38.3 hours per lot; the sample worth paying for is 42 items, and under drift one samples every 2.32 working days. (https://5dlibrary.com/answers/en/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c04/) 第4章 抽样验收:接受概率曲线、kp 规则与全检零检的边界 — 第四、五章说明,过程稳定时 kp 规则只认全检或零检,内点最优只来自缺陷率不稳定时的信息价值:先抽 15 件再决定,每批省下 38.3 个工时;值得付钱的样本是 42 件,漂移时每 2.32 个工作日抽一次。 (https://5dlibrary.com/answers/zh/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c04/) - Ch.5 When the Defect Rate Is Unknown or Drifts: Bayesian Updating, Sample Size and the Inspection Interval — When the defect rate is unknown, what a sample buys is information, and the sample size worth paying for has an interior optimum: 42 items with the illustrative parameters, on a curve that is flat between 20 and 60 items. Accumulated data create inertia, letting an old good record suppress a new warning, so old data must retire or a single-week alarm line must be set. When the defect rate can drift, the inspection interval has an interior optimum too, 2.32 days with the illustrative parameters, and it moves with the square root of the drift rate. "The more often the better" fails in time just as it fails in space. (https://5dlibrary.com/answers/en/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c05/) 第5章 缺陷率未知或会漂移:贝叶斯更新、样本量与检验间隔 — 缺陷率未知时,抽样买到的是信息,值得付钱的样本量有内点最优:示意参数下是 42 件,曲线在 20 到 60 件之间很平。累积数据会产生惯性,让旧的好记录压住新的警讯,所以要让旧数据退出,或者设单周报警线。缺陷率会漂移时,检验间隔也有内点最优,示意参数下是 2.32 天,并按漂移速率的平方根变化。「检验越频繁越好」在时间维度上同样不成立。 (https://5dlibrary.com/answers/zh/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c05/) - Ch.6 The Incentive Slope: Noise, Risk Aversion, Multitask Crowding Out and Several Currencies — Chapter 6 shows that when handoff notes go uncounted the optimal slope falls from 0.667 to 0.364, that exchange-rate noise lowers it only to 0.663, but that price levels make one slope worth between 1.67 and 2.50 baskets of goods. (https://5dlibrary.com/answers/en/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c06/) 第6章 激励斜率:噪声、风险厌恶、多任务挤出与多币种 — 第六章说明,交接说明不被计数时最优斜率从 0.667 降到 0.364,汇率噪声只让它降到 0.663,价格水平差异却让同一斜率的实际强度在 1.67 到 2.50 篮商品之间。 (https://5dlibrary.com/answers/zh/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c06/) - Ch.7 Classical Sources East and West: Dividing the People Well in the Guanzi, and the Division of Labour in Babbage and Taylor — Chapter 7 examines the Guanzi on dividing the people well, Babbage and Taylor; (https://5dlibrary.com/answers/en/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c07/) 第7章 东西方原典考辨:《管子·乘马》「善分民」与巴贝奇、泰勒的分工论 — 第七章考辨《管子·乘马》「善分民」与巴贝奇、泰勒,第八章给出周度台账、三城异步交接与抽检频率试算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c07/) - Ch.8 Closing the Empirical Loop: a Weekly Three-City Ledger, a Calculator and Pre-Registered Tests — The weekly ledger records the flow of work; the calculator turns the ledger into five recommendations, on the number of pieces, the batch interval, the sample size, the inspection interval and the incentive slope; and the pre-registered plan states how large the samples must be: for H1, a few tens of deliverables per group for a moderate effect; for H2, a full quarter of lots; for H3, eight to eleven handoffs a week. Each of the three core propositions carries an explicit condition of refutation, and the data that could refute them can only come from readers' own ledgers. (https://5dlibrary.com/answers/en/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c08/) 第8章 实证闭环:三城周度台账、试算器与预注册检验 — 周度台账记下工作的流动,试算器把台账变成切分块数、批量间隔、抽检样本、检验间隔与激励斜率五条建议,预注册方案说明了检验需要多少样本:H1 在中等效应下每组几十件,H2 需要一季度的批次,H3 每周八到十一次交接。三条核心命题都写明了推翻条件,推翻它们的数据,只能来自读者自己的台账。 (https://5dlibrary.com/answers/zh/kl-taipei-hcmc-distributed-teams-async-verification-talent-allocation/c08/) ## 5DL-B-341 · Real-Options Valuation of Geo-Exact Domain Portfolios, Optimal Renewal Pruning and the Moats of Organic Search 中文书名:精确地域域名实物期权估值、续费剪枝最优停时与自然搜索垄断护城河母典 DOI: pending Cite: Fan, Y. (2026). Real-Options Valuation of Geo-Exact Domain Portfolios, Optimal Renewal Pruning and the Moats of Organic Search (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 2 examines the price evidence: in the illustrative model a repeat-sales estimate is biased upwards by 0.1284, and applying the index's growth of 0.0723 a year leads to systematic under-pruning. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Owners of a pool of forty-odd domain names commonly believe two things: good names appreciate and renewal fees hardly matter, and a name that spells out a city and a trade is favoured by search engines. This book refutes both popular beliefs. An unbuilt name is valued mainly as a development option, and the renewal fee is the carrying cost of that option; the carrying cost splits the single trigger for building into two boundaries, letting the name lapse below the lower one and building only above the upper one. Under illustrative parameters the trigger multiple is 2.893, and with a fee of 15 the lower boundary is 599 and the upper 16,849; of 42 illustrative names, 23 should lapse, the option accounts for 0.861 of best-use value, and sensible renewals need only a small part of the full bill of 1,340. After checking the sources, the book corrects its blueprint: the exact-match-domain update of 2012 targeted low-quality sites, so "no ranking bonus" is a robustness assumption rather than a fact; Lindenthal's repeat-sales index describes only names that traded; and traffic that does not depend on rank is about 0.309 of the total under illustrative parameters, a dependable floor rather than the bulk of the flow. All parameters are illustrative and all calculations are generated by the book's programs; the book is general business and investment analysis, not personalised investment, legal or tax advice, and statements on trademarks, surnames and platform rules must be checked by a person before publication. Chapter 1 derives the value of waiting and the two boundaries under a renewal fee. Chapter 2 examines the price evidence: in the illustrative model a repeat-sales estimate is biased upwards by 0.1284, and applying the index's growth of 0.0723 a year leads to systematic under-pruning. Chapter 3 solves the choice among renewing, lapsing and building year by year on a lattice. Chapter 4 writes pruning as a knapsack and runs a ten-year back-test: prune and wait has a mean present value of 11,121 against 1,344 for keeping everything and building by net present value, though waiting also widens the spread. Chapter 5 separates rank, clicks and direct navigation with the cascade model; detecting a fifth more willingness to click needs 3,837 impressions per group. Chapter 6 covers entity consistency in generative search, with structured data that visitors can see and no cloaking, hidden text, link schemes or site networks. Chapter 7 examines Bai Gui's "take what others discard" in the Records of the Grand Historian and Henry George's Progress and Poverty on land speculation. Chapter 8 gives the annual renewal audit, the distilled Domain Pool Renewal Pruner and a pre-registered plan of refutation: the test of the lapse rule needs 52 released names and is refuted if 6 cross the line. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:手握四十多枚域名的经营者普遍相信两件事:好名字会升值,续费不值一提;名字里带上城市与行业,搜索引擎就会偏爱它。本书推翻这两条流行说法。一枚没建站的名字,价值主要是一份开发期权,续费就是期权的持有成本;持有成本把一条建站触发线拆成两条边界,跌破下边界就弃续,越过上边界才立项。示意参数下,触发倍数为 2.893,续费 15 时下边界为 599、上边界为 16,849;42 枚示意名字里有 23 枚应当弃续,期权部分占最优价值的 0.861,合理续费只需全额 1,340 中的一小部分。核对原文之后,本书做了蓝图修正:2012 年的精确匹配域名更新针对的是低质量网站,名字不带排名加成只是稳健性假定;林登塔尔的重复交易指数只反映成交过的名字;不依赖排名的访问在示意参数下约占 0.309,是可靠的地板而不是流量的大头。书中参数均为示意,全部计算由书中程序生成;本书是非个人化的经营与投资分析,不构成投资、法律或税务建议,涉及商标、姓氏与平台规则处以上架前人工复核为准。 第一章推导等待的价值与续费下的两条边界。第二章检验价格证据:在示意模型里,重复交易估计偏高 0.1284,套用指数的 0.0723 年涨幅会系统性少剪。第三章用二叉树逐年求解续费、弃续与立项三择。第四章把剪枝写成背包,并做十年回测:剪枝加等待的平均现值为 11,121,全部续费加净现值法则只有 1,344,但等待也加大了离散。第五章用级联模型拆解排名、点击与直接输入,检出点击意愿提升两成每组需要 3,837 次展示。第六章讲生成式搜索中的实体一致性,只写看得见的结构化数据,不写伪装、隐藏文本、刷链与站群。第七章考辨《史记·货殖列传》白圭「人弃我取」与亨利·乔治《进步与贫困》论土地投机。第八章给出年度续费审计、蒸馏小工具「域名池续费剪枝器」与预注册证伪方案:放手规则的检验需要 52 枚放手名字,6 枚越线即推翻。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Carrying Cost and the Development Option: the Value of Waiting, Renewal Fees and Two Boundaries — In the first row of Part A the fee is nearly zero, the lower boundary is close to zero and the upper boundary is the 17,360 of p11: the model reduces to the case without fees, as it should. With a fee of 15, the present value of renewing forever is 188, the lower boundary rises to 599, the upper boundary falls to 16,849, and the trigger multiple falls to 2.808. With a fee of 120, the present value of fees is 1,500, the lower boundary rises to 2,168, the upper boundary falls to 13,906 and the multiple falls to 2.318. (https://5dlibrary.com/answers/en/geo-exact-domain-real-options-renewal-pruning-search-moats/c01/) 第1章 持有成本与开发期权:等待的价值、续费与两条边界 — A 段第一行续费几乎为零,下边界接近零,上边界就是 p11 的 17,360,模型退回无续费的情形。续费为 15 时,十年、百年续下去的现值是 188,下边界升到 599,上边界降到 16,849,触发倍数降到 2.808。续费为 120 时,现值是 1,500,下边界升到 2,168,上边界降到 13,906,倍数降到 2.318。 (https://5dlibrary.com/answers/zh/geo-exact-domain-real-options-renewal-pruning-search-moats/c01/) - Ch.2 The Price Evidence: Repeat-Sales Indices, Survivorship Bias and the Cost of Holding — Chapter 2 examines the price evidence: in the illustrative model a repeat-sales estimate is biased upwards by 0.1284, and applying the index's growth of 0.0723 a year leads to systematic under-pruning. (https://5dlibrary.com/answers/en/geo-exact-domain-real-options-renewal-pruning-search-moats/c02/) 第2章 域名价格的证据:重复交易指数、幸存者偏差与持有成本 — 第二章检验价格证据:在示意模型里,重复交易估计偏高 0.1284,套用指数的 0.0723 年涨幅会系统性少剪。 (https://5dlibrary.com/answers/zh/geo-exact-domain-real-options-renewal-pruning-search-moats/c02/) - Ch.3 A Finite-Horizon Lattice: Renewing, Lapsing and Building Year by Year — Part A places the perpetual boundaries of Chapter 1 in the dimension of time. With one year left, the lower boundary is as high as 3,052 and the upper boundary as low as 10,692. With two years left they are 2,157 and 11,844; with five, 1,303 and 13,807; with ten, 918 and 14,802; and with forty, 665 and 15,397, close to the perpetual values of 599 and 16,849 from Chapter 1. (https://5dlibrary.com/answers/en/geo-exact-domain-real-options-renewal-pruning-search-moats/c03/) 第3章 有限期二叉树:逐年续费、弃续与立项的动态规划 — A 段把第一章的两条永续边界放进了时间维度。只剩一年时,下边界高达 3,052,上边界低到 10,692;剩两年时是 2,157 与 11,844;剩五年时是 1,303 与 13,807;剩十年时是 918 与 14,802;剩四十年时是 665 与 15,397,已接近第一章的永续值 599 与 16,849。 (https://5dlibrary.com/answers/zh/geo-exact-domain-real-options-renewal-pruning-search-moats/c03/) - Ch.4 Pruning under a Budget: the Knapsack, the Greedy Ratio and a Ten-Year Back-Test of Policies — Chapter 4 writes pruning as a knapsack and runs a ten-year back-test: prune and wait has a mean present value of 11,121 against 1,344 for keeping everything and building by net present value, though waiting also widens the spread. (https://5dlibrary.com/answers/en/geo-exact-domain-real-options-renewal-pruning-search-moats/c04/) 第4章 预算约束下的剪枝:背包、贪心比率与十年政策回测 — 第四章把剪枝写成背包,并做十年回测:剪枝加等待的平均现值为 11,121,全部续费加净现值法则只有 1,344,但等待也加大了离散。 (https://5dlibrary.com/answers/zh/geo-exact-domain-real-options-renewal-pruning-search-moats/c04/) - Ch.5 Rank, Clicks and Direct Navigation: Where the Moat of a Geo-Exact Name Comes From — Chapter 5 separates rank, clicks and direct navigation with the cascade model; detecting a fifth more willingness to click needs 3,837 impressions per group. (https://5dlibrary.com/answers/en/geo-exact-domain-real-options-renewal-pruning-search-moats/c05/) 第5章 排名、点击与直接输入:地域精确名的护城河从哪里来 — 第五章用级联模型拆解排名、点击与直接输入,检出点击意愿提升两成每组需要 3,837 次展示。 (https://5dlibrary.com/answers/zh/geo-exact-domain-real-options-renewal-pruning-search-moats/c05/) - Ch.6 Generative Search and Entity Consistency: Only Structured Data that Visitors Can See — Part B adds the owner's own page, always accurate and marked up with visible structured data. With five other sources, the chance of a correct answer rises from 0.5094 to 0.7206 when $q$ is 0.40, from 0.7463 to 0.8869 when $q$ is 0.55, and from 0.9104 to 0.9703 when $q$ is 0.70. The owner's page is the only source entirely under the owner's control, and it does most good when consistency elsewhere is poor. (https://5dlibrary.com/answers/en/geo-exact-domain-real-options-renewal-pruning-search-moats/c06/) 第6章 生成式搜索与实体一致性:只写看得见的结构化数据 — B 段加上经营者自己的页面,内容始终准确,并用看得见的结构化数据标注。五个来源下,$q$ 为 0.40 时说对的机会从 0.5094 升到 0.7206,$q$ 为 0.55 时从 0.7463 升到 0.8869,$q$ 为 0.70 时从 0.9104 升到 0.9703。自己的页面是唯一完全由自己掌控的来源,它在一致率低的时候作用最大。 (https://5dlibrary.com/answers/zh/geo-exact-domain-real-options-renewal-pruning-search-moats/c06/) - Ch.7 Classics East and West: Bai Gui's "Take What Others Discard" and Henry George on Land Speculation — With a fee of 5 the lower boundary is only 294: 18 of the forty-two names are let go, 24 are held idle and none is built. With a fee of 15, 22 are let go and 20 held idle. With 60, 33 are let go and 9 held idle. With 240, 38 are let go, 3 held idle and 1 built; with 480, 40 are let go, 1 held idle and 1 built, and the upper boundary has fallen to 8,982. The higher the cost of holding, the fewer names sit idle and the more are either released or built, which is exactly the direction George predicted: a cost of holding drives resources either into use or back onto the market. (https://5dlibrary.com/answers/en/geo-exact-domain-real-options-renewal-pruning-search-moats/c07/) 第7章 东西方原典考辨:白圭「人弃我取」与亨利·乔治论土地投机 — 续费 5 时,下边界只有 294,四十二枚名字里 18 枚放手、24 枚闲置持有、没有一枚建站。续费 15 时,22 枚放手、20 枚闲置。续费 60 时,33 枚放手、9 枚闲置。续费 240 时,38 枚放手、3 枚闲置、1 枚建站;续费 480 时,40 枚放手、1 枚闲置、1 枚建站,上边界降到 8,982。持有成本越高,闲置的名字越少,放手与建站越多,这正是乔治预言的方向:持有的代价推动资源要么投入使用,要么回到市场。 (https://5dlibrary.com/answers/zh/geo-exact-domain-real-options-renewal-pruning-search-moats/c07/) - Ch.8 The Annual Renewal Audit and a Plan for Refutation — Chapter 8 gives the annual renewal audit, the distilled Domain Pool Renewal Pruner and a pre-registered plan of refutation: the test of the lapse rule needs 52 released names and is refuted if 6 cross the line. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/geo-exact-domain-real-options-renewal-pruning-search-moats/c08/) 第8章 年度续费审计闭环与证伪方案 — 第八章给出年度续费审计、蒸馏小工具「域名池续费剪枝器」与预注册证伪方案:放手规则的检验需要 52 枚放手名字,6 枚越线即推翻。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/geo-exact-domain-real-options-renewal-pruning-search-moats/c08/) ## 5DL-B-342 · Social-Graph Anti-Spam Detection Bounds, Compliant Affiliate Bridge-Page Funnels and the Kinetics of Recovery after Reach Throttling 中文书名:公域社交图谱反垃圾检测边界、合规联盟桥接页转化与限流自愈动力学母典 DOI: pending Cite: Fan, Y. (2026). Social-Graph Anti-Spam Detection Bounds, Compliant Affiliate Bridge-Page Funnels and the Kinetics of Recovery after Reach Throttling (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is general business analysis, not personalised business, legal or financial advice, and rules must be checked by a human reviewer before publication. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Two sentences circulate among operators of public social feeds: automated posting goes unnoticed as long as the volume is low, and more links mean more commission. This book refutes both, and it is written throughout from the side of the detector and of the compliant operator: its detection mathematics serves only to judge whether one's own posting would be recognised as spam or manipulation, so that it can be changed to a compliant cadence. It discusses no method of evading platform enforcement, requires affiliate links to be disclosed honestly, and treats recovery after a throttle only as stopping the violation, natural recovery and official appeal. After checking the sources it corrects its blueprint: human activity is bursty, and the machine's fingerprint is the absence of burstiness. Its parameters are illustrative; all calculations are generated by its programs; it is general business analysis, not personalised business, legal or financial advice, and rules must be checked by a human reviewer before publication. Chapter 1 shows that burstiness does not depend on posting frequency: −0.969 for a fixed timer, 0.300 for a human-like rhythm. Chapter 2 computes sample sizes with the likelihood ratio and Wald's sequential test; even a Poisson timer needs only 15 gaps. Chapter 3 uses shingles and minimum hashing: template posts resemble each other at about 0.522, and two of them are caught by banded detection with a chance of 0.915309. Chapter 4 shows that link share and clicks follow a curve that rises and then collapses, and an account that is all links is identified from 3 sampled posts. Chapter 5 prices the compliant bridge page: 308.4 orders a year for an honestly disclosed page against 132.0 for a springboard, and only 59 disclosed items out of 59 support a coverage of at least 0.95. Chapter 6 writes first-order recovery: an account that stops regains nine tenths of its reach in 30 days, and in a year of accounts bridging overtakes blasting between weeks 26 and 37. Chapter 7 examines "rumours stop with those who know" from the Xunzi and Aristotle on character: with three checks the critical share of careful readers is 0.429, and once concealment is discovered trust falls from 0.732 to 0.505 or below. Chapter 8 gives the monthly check-up, the distilled Public-Feed Compliance Check-up Sheet and a pre-registered plan of refutation needing 112 topics per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:公域流量圈子里流传两句话:自动发帖只要量少,平台就发现不了;外链越多,佣金越多。本书证伪这两句流行说法,并且始终站在检测方与合规经营者一边:书中的检测数学只用来判断自己的发帖行为会不会被识别为垃圾或操纵,从而改成合规的节奏;本书不讨论任何规避平台风控的手法,联盟链接必须如实披露,限流之后只写停止违规、自然恢复与官方申诉。核对原文之后,本书修正了立项蓝图:人的活动才是突发的,机器的指纹是突发的缺失。书中参数为示意;全部计算由书中程序生成;本书是非个人化的经营分析,不构成经营、法律或理财建议,法规与平台规则以上架前人工复核为准。 第一章说明突发性系数与发帖频率无关:固定定时器为 −0.969,类人节奏为 0.300。第二章用似然比与瓦尔德序贯检验计算样本量,泊松定时器也只需 15 个间隔。第三章用瓦片与最小哈希说明模板帖子两两相似度约为 0.522,只发 2 条就有 0.915309 的机会被分带检测抓到。第四章说明外链占比与点击是一条先升后崩的曲线,全是外链的账号 3 条样本即可识别。 第五章算合规桥接页:如实披露的页面全年 308.4 单,跳板页面 132.0 单;59 条全有披露才能说覆盖率不低于 0.95。第六章写一阶恢复:停止违规的账号 30 天回到九成触达,一年期账中桥接策略在第 26 到 37 周反超。第七章考辨《荀子·大略》「流言止于知者」与亚里士多德的品格论:三项查证时知者的临界比例为 0.429,隐瞒关系一经发现,信任从 0.732 降到 0.505 以下。第八章给出月度公域合规体检、蒸馏小工具「公域合规体检表」与预注册证伪方案,每组需 112 个话题。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Timing Fingerprint: Burstiness, Memory and the Illusion of Safety in Small Numbers — Chapter 1 shows that burstiness does not depend on posting frequency: −0.969 for a fixed timer, 0.300 for a human-like rhythm. (https://5dlibrary.com/answers/en/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c01/) 第1章 间隔指纹:突发性系数、记忆系数与「量少就安全」的错觉 — 第一章说明突发性系数与发帖频率无关:固定定时器为 −0.969,类人节奏为 0.300。 (https://5dlibrary.com/answers/zh/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c01/) - Ch.2 A Few Posts Are Enough: Likelihood Ratios, Sequential Tests and Change-Point Alarms — Chapter 2 computes sample sizes with the likelihood ratio and Wald's sequential test; even a Poisson timer needs only 15 gaps. (https://5dlibrary.com/answers/en/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c02/) 第2章 看几条就够:似然比、序贯检验与变点报警 — 第二章用似然比与瓦尔德序贯检验计算样本量,泊松定时器也只需 15 个间隔。 (https://5dlibrary.com/answers/zh/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c02/) - Ch.3 The Fingerprint of Words: Shingles, the Jaccard Coefficient and Minimum Hashing — Chapter 3 uses shingles and minimum hashing: template posts resemble each other at about 0.522, and two of them are caught by banded detection with a chance of 0.915309. (https://5dlibrary.com/answers/en/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c03/) 第3章 文字的指纹:瓦片、杰卡德系数与最小哈希 — 第三章用瓦片与最小哈希说明模板帖子两两相似度约为 0.522,只发 2 条就有 0.915309 的机会被分带检测抓到。 (https://5dlibrary.com/answers/zh/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c03/) - Ch.4 The Price of Links: Demotion Thresholds, Information Posts and the Binomial Test — Chapter 4 shows that link share and clicks follow a curve that rises and then collapses, and an account that is all links is identified from 3 sampled posts. (https://5dlibrary.com/answers/en/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c04/) 第4章 外链的代价:降权阈值、信息帖与二项检验 — 第四章说明外链占比与点击是一条先升后崩的曲线,全是外链的账号 3 条样本即可识别。 (https://5dlibrary.com/answers/zh/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c04/) - Ch.5 The Compliant Bridge Page: Funnels, Honest Disclosure and the Attribution Window — Chapter 5 prices the compliant bridge page: 308.4 orders a year for an honestly disclosed page against 132.0 for a springboard, and only 59 disclosed items out of 59 support a coverage of at least 0.95. (https://5dlibrary.com/answers/en/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c05/) 第5章 合规桥接页:漏斗、如实披露与佣金归因窗口 — 第五章算合规桥接页:如实披露的页面全年 308.4 单,跳板页面 132.0 单;59 条全有披露才能说覆盖率不低于 0.95。 (https://5dlibrary.com/answers/zh/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c05/) - Ch.6 After the Throttle: First-Order Recovery, Half-Lives and a Year of Commission Accounts — Chapter 6 writes first-order recovery: an account that stops regains nine tenths of its reach in 30 days, and in a year of accounts bridging overtakes blasting between weeks 26 and 37. (https://5dlibrary.com/answers/en/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c06/) 第6章 限流之后:一阶恢复、半衰期与一年期的佣金账 — 第六章写一阶恢复:停止违规的账号 30 天回到九成触达,一年期账中桥接策略在第 26 到 37 周反超。 (https://5dlibrary.com/answers/zh/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c06/) - Ch.7 Classical Sources Examined: "Rumours Stop at Those Who Know" in the Xunzi and Aristotle on Ethos in the Rhetoric — Chapter 7 examines "rumours stop with those who know" from the Xunzi and Aristotle on character: with three checks the critical share of careful readers is 0.429, and once concealment is discovered trust falls from 0.732 to 0.505 or below. (https://5dlibrary.com/answers/en/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c07/) 第7章 东西方原典考辨:《荀子·大略》「流言止于知者」与亚里士多德《修辞学》的品格论 — 第七章考辨《荀子·大略》「流言止于知者」与亚里士多德的品格论:三项查证时知者的临界比例为 0.429,隐瞒关系一经发现,信任从 0.732 降到 0.505 以下。 (https://5dlibrary.com/answers/zh/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c07/) - Ch.8 Closing the Loop: a Monthly Public-Feed Compliance Check-up, a Distilled Tool and a Pre-Registered Plan of Refutation — Chapter 8 gives the monthly check-up, the distilled Public-Feed Compliance Check-up Sheet and a pre-registered plan of refutation needing 112 topics per arm. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c08/) 第8章 实证闭环:月度公域合规体检、蒸馏小工具与预注册证伪方案 — 第八章给出月度公域合规体检、蒸馏小工具「公域合规体检表」与预注册证伪方案,每组需 112 个话题。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/social-graph-antispam-bounds-compliant-bridge-pages-shadowban-recovery/c08/) ## 5DL-B-343 · Flagship Talent Scarcity Pricing, Brand-Dilution Spillover Penalties and Exclusive Representation Game Theory 中文书名:超模与头部艺人稀缺溢价保护、代言降格污染惩罚与独家经纪博弈母典 DOI: pending Cite: Fan, Y. (2026). Flagship Talent Scarcity Pricing, Brand-Dilution Spillover Penalties and Exclusive Representation Game Theory (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All parameters are illustrative and all calculations are generated by the book's programs. The book evaluates no living person and is not personalised legal, financial or investment advice; contract terms are described only as structure and game logic, subject to review before publication. Abstract: Talent agencies widely believe that a performer's fee rises with her exposure and that every commercial offer is worth taking. This book writes that belief as runnable arithmetic and tests it. Under Rosen's economics of superstars, flagship fees are a convex function of rank in talent; because they are convex, the cash from one down-market deal is often smaller than the depreciation it imposes on future luxury fees; and without investment protection, an agency with exclusive rights invests less than is jointly efficient. After checking the originals the book corrects its project blueprint: the convexity is in talent, not exposure, and is contested by Adler and Hamlen; published endorsement event studies measure firms' share prices, not performers' fees, so down-market dilution is presented as a model and a pre-registered hypothesis; the classic statement of hold-up is due to Klein, Crawford and Alchian, while Grossman and Hart analyse residual control; and covert image placement is replaced throughout by disclosed split image licensing. All parameters are illustrative and all calculations are generated by the book's programs. The book evaluates no living person and is not personalised legal, financial or investment advice; contract terms are described only as structure and game logic, subject to review before publication. Chapter 1 shows that a ten percent edge in talent is worth a fee 1.464 times higher on a convex schedule with k equal to 4, and uses Krueger's concert-revenue concentration to test the direction of convexity. Chapter 2 writes the fee ladder as a Markov chain and computes the power of event studies. Chapter 3 builds a reputation-stock model: under illustrative parameters one down-market deal carries a dilution charge of 977.4, which is 3.672 times the linear estimate of 266.2, against only 167.5 for a mid-market performer, and luxury partners can bear an external share of 0.722. Chapter 4 sets the charge at the eightieth percentile under uncertainty, 1,236.4, and adds a slot cost of 553.9 for category exclusivity. Chapter 5 is the hold-up game of exclusive representation: an indexed buy-out lifts total surplus to 0.9924 of the efficient level, a lock leaves only 0.7171, and splitting image control by channel reaches 0.9829. Chapter 6 shows that, net of dilution, a menu of channel prices, bundle prices and a published eligibility floor earns 260 more than pricing for gross revenue. Chapter 7 examines the Zhanguoce story of buying a horse's bones and Veblen's conspicuous consumption. Chapter 8 gives the quarterly fee and deal checkup, the distilled Talent Deal Downgrade Cost Calculator and a pre-registered test plan that needs 112 performers per group. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:经纪行普遍相信,艺人越红,报价就按曝光往上加,商单来者不拒。本书把这一说法写成可以运行的算术并加以检验:按 Rosen 的超级明星经济学,头部报价是能力排名的凸函数;正因为凸,一单低端商单带来的现金常常小于它对未来高奢报价的折旧;独家经纪若没有投入保护,经纪方的投入会低于合作最优。核对原文后,本书对立项蓝图作了修正:凸性是对能力而非对曝光,且有 Adler 与 Hamlen 的反方证据;已发表的代言事件研究测的是企业股价,不是艺人报价,降格污染因此写成模型与预注册假说;套牢的经典表述来自 Klein、Crawford 与 Alchian,Grossman 与 Hart 讨论的是剩余控制权;「肖像暗投」一律改写为如实披露的肖像拆分授权。书中参数均为示意,全部计算由书中程序生成;本书不点名评价任何在世人士,不是个人化的法律、财务或投资建议,合同条款只写结构与博弈逻辑,以上架前人工复核为准。 第一章说明能力领先一成,在 k 为 4 的凸报价下报价是 1.464 倍,并用 Krueger 的演唱会收入集中度检验凸性方向。第二章把报价阶梯写成马尔可夫链,并计算事件研究的检验功效。第三章建立声誉存量模型:示意参数下一单低端商单的稀释费为 977.4,是线性估计 266.2 的 3.672 倍,腰部艺人同条件只有 167.5;高端合作品牌承担的外部份额可达 0.722。第四章说明参数不确定时按第八十百分位收稀释费,即 1,236.4,品类独家另有档位成本 553.9。 第五章是独家经纪的套牢博弈:指数化买断价使总剩余达到最优的 0.9924,锁死规则只有 0.7171;肖像控制权按渠道拆分时为 0.9829。第六章说明扣除稀释代价后,分渠道价、套装价与公开资格线并存的菜单比按毛收入选价多赚 260。第七章考辨《战国策》「千金买马骨」与 Veblen 的炫耀性消费。第八章给出季度报价与接单体检流程、蒸馏小工具「艺人接单降格成本计算器」与预注册检验方案,降格检验每组需要 112 位艺人。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Convex Superstar Pricing: a Small Edge in Talent, a Large Edge in Fees — Chapter 1 shows that a ten percent edge in talent is worth a fee 1.464 times higher on a convex schedule with k equal to 4, and uses Krueger's concert-revenue concentration to test the direction of convexity. (https://5dlibrary.com/answers/en/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c01/) 第1章 超级明星凸报价:能力差一点,收入差一截 — 第一章说明能力领先一成,在 k 为 4 的凸报价下报价是 1.464 倍,并用 Krueger 的演唱会收入集中度检验凸性方向。 (https://5dlibrary.com/answers/zh/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c01/) - Ch.2 The Fee Ladder: a Markov Chain of Tiers and the Eyesight of Event Studies — The fee ladder is a Markov chain. Raise the chance of slipping by a few points and lower the chance of climbing by a few points, and the expected time to the top becomes three to five times longer while the present value of twenty quarters of income falls. For a performer on the way up, the cash from down-market deals would have to reach between 0.686 and 0.781 of tier income to compensate for the lost chance to climb. Event studies confirm that endorsements create value and that scandals destroy it, but their eyesight is limited, and the experience of a single performer is nearly blind to the cost of downgrades. Since the cost cannot be seen, it must be modelled. How, exactly, does a down-market deal depreciate a performer's luxury fees, and how much should be charged for that depreciation on the day the contract is signed? (https://5dlibrary.com/answers/en/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c02/) 第2章 报价阶梯:跃迁马尔可夫链与事件研究的视力 — 报价阶梯是一条马尔可夫链。下滑概率升几个点、爬升概率降几个点,到顶时间就拉长三到五倍,二十个季度的收入现值反而更少;低端商单的现金要达到档位收入的 0.686 到 0.781 倍,才能补偿爬坡艺人失去的机会。事件研究证实代言有价值、丑闻有代价,但它的视力有限,单个艺人的经验更看不清降格成本。既然看不清,就要建模:一单低端商单究竟怎样折旧艺人的高奢报价,这笔折旧又该在签约那天收多少钱? (https://5dlibrary.com/answers/zh/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c02/) - Ch.3 Down-Market Dilution: a Reputation Stock, a Signing-Day Charge and the Loss Borne by Others — Chapter 3 builds a reputation-stock model: under illustrative parameters one down-market deal carries a dilution charge of 977.4, which is 3.672 times the linear estimate of 266.2, against only 167.5 for a mid-market performer, and luxury partners can bear an external share of 0.722. (https://5dlibrary.com/answers/en/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c03/) 第3章 降格污染:声誉存量、稀释费与外部损失 — 第三章建立声誉存量模型:示意参数下一单低端商单的稀释费为 977.4,是线性估计 266.2 的 3.672 倍,腰部艺人同条件只有 167.5;高端合作品牌承担的外部份额可达 0.722。 (https://5dlibrary.com/answers/zh/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c03/) - Ch.4 Deciding on a Deal: Thresholds under Uncertainty and the Slot that Exclusivity Locks — Chapter 4 sets the charge at the eightieth percentile under uncertainty, 1,236.4, and adds a slot cost of 553.9 for category exclusivity. (https://5dlibrary.com/answers/en/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c04/) 第4章 接单决策:参数不确定时的门槛与独家锁定的档位 — 第四章说明参数不确定时按第八十百分位收稀释费,即 1,236.4,品类独家另有档位成本 553.9。 (https://5dlibrary.com/answers/zh/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c04/) - Ch.5 The Hold-Up Game in Exclusive Representation: Investment Protection, Buy-Out Prices and Control — Chapter 5 is the hold-up game of exclusive representation: an indexed buy-out lifts total surplus to 0.9924 of the efficient level, a lock leaves only 0.7171, and splitting image control by channel reaches 0.9829. (https://5dlibrary.com/answers/en/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c05/) 第5章 独家经纪的套牢博弈:投入保护、买断价格与控制权 — 第五章是独家经纪的套牢博弈:指数化买断价使总剩余达到最优的 0.9924,锁死规则只有 0.7171;肖像控制权按渠道拆分时为 0.9829。 (https://5dlibrary.com/answers/zh/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c05/) - Ch.6 Split Image Licensing: Pricing by Channel, Territory and Term, and When to Bundle — Chapter 6 shows that, net of dilution, a menu of channel prices, bundle prices and a published eligibility floor earns 260 more than pricing for gross revenue. (https://5dlibrary.com/answers/en/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c06/) 第6章 肖像拆分授权:按渠道、地区、期限定价,以及何时打包 — 第六章说明扣除稀释代价后,分渠道价、套装价与公开资格线并存的菜单比按毛收入选价多赚 260。 (https://5dlibrary.com/answers/zh/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c06/) - Ch.7 Classical Sources: Buying a Horse's Bones for Five Hundred, and the Conspicuous Consumption of the Leisure Class — With an exclusivity index of 1.0, the effective slope of demand with respect to price is 0.40, the brand's profit-maximising price is 125.0 and its revenue 6,250.0. When the index falls to 0.9, the slope rises to 0.46, the profit-maximising price falls to 108.7 and revenue to 5,434.8, only 0.870 of its former level. At an index of 0.8 revenue is 0.769 of the original; at 0.5 it is only 0.571. (https://5dlibrary.com/answers/en/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c07/) 第7章 原典考辨:千金买马骨与有闲阶级的炫耀性消费 — 独家性指数为 1.0 时,需求对价格的有效斜率是 0.40,品牌的最优价格 125.0,收入 6,250.0。指数降到 0.9,斜率升到 0.46,最优价格 108.7,收入 5,434.8,只剩原来的 0.870;指数 0.8 时收入只剩 0.769;指数 0.5 时只剩 0.571。 (https://5dlibrary.com/answers/zh/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c07/) - Ch.8 Closing the Loop: a Quarterly Fee and Deal Checkup, the Distilled Tool and a Pre-Registered Test — Chapter 8 gives the quarterly fee and deal checkup, the distilled Talent Deal Downgrade Cost Calculator and a pre-registered test plan that needs 112 performers per group. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c08/) 第8章 实证闭环:季度报价与接单体检、小工具与预注册检验 — 第八章给出季度报价与接单体检流程、蒸馏小工具「艺人接单降格成本计算器」与预注册检验方案,降格检验每组需要 112 位艺人。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/flagship-talent-scarcity-pricing-brand-dilution-exclusive-representation/c08/) ## 5DL-B-344 · Concentrated-Liquidity Loss-Versus-Rebalancing, On-Chain Sandwich Bounds and Decentralised Vault Risk Control 中文书名:集中流动性做市商重平衡损耗(LVR)、链上三明治抢跑边界与去中心化金库风控母典 DOI: pending Cite: Fan, Y. (2026). Concentrated-Liquidity Loss-Versus-Rebalancing, On-Chain Sandwich Bounds and Decentralised Vault Risk Control (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. This is general investment risk analysis, not investment advice; it recommends no token, protocol or trade, and crypto-assets are high risk and can be lost in full. Abstract: The most popular sentence in on-chain market making says that a liquidity provider earns fees while bearing an "impermanent loss" that vanishes once the price comes back. This book refutes it. A constant-product pool is continuously short gamma: arbitrageurs take a small slice at every new price, the slice depends only on the squared size of each move, and it does not come back when the price returns home. Under geometric Brownian motion the loss-versus-rebalancing (LVR) of a constant-product pool is σ²/8 per unit of value per unit of time, about 3.125 basis points a day at a daily volatility of five per cent. After checking the original papers the book corrects its blueprint: the σ²/8 dependence first appeared in the analysis of Angeris and co-authors, and the 2022 paper generalised it into a pathwise decomposition; it is a frictionless cap, and with a 30 basis point fee and twelve-second blocks arbitrageurs actually take about 0.120 of it; the amplification of concentrated liquidity holds only inside the range; and the claim that fees often fail to cover LVR holds only where volatility is high and uninformed turnover thin. Parameters are illustrative and all calculations are generated by the book's programs. This is general investment risk analysis, not investment advice; it recommends no token, protocol or trade, and crypto-assets are high risk and can be lost in full. Chapter 1 uses the value function V = 2L√p to show that market making is selling gamma. Chapter 2 derives σ²/8 and measures the discount from fees and block times. Chapter 3 gives the amplification factor of concentrated ranges, 21.49 inside a range of plus or minus ten per cent, and the odds of leaving the range. Chapter 4 computes the break-even volatility, the noise in a measured coverage ratio and the crowding effect of concentration. Chapter 5 takes only the victim's side: a sandwich can take no more than the slippage tolerance times the quote, and a first-order model gives the tolerance below which no adversary profits, about 0.00427 for a trade of one per cent of a pool one million deep at a 0.30 % fee. Chapter 6 treats contract tail risk as Poisson jumps and derives a Kelly cap with jumps, which falls to 0.3406 when a total loss arrives at a rate of 0.03 a year, together with a rebalancing band for vaults. Chapter 7 reads the Guanzi on the levelling of prices and Bagehot's Lombard Street of 1873. Chapter 8 gives a monthly on-chain position check-up, the distilled Market-Making Loss and Slippage Check-up and a pre-registered refutation plan whose main test needs 56 position-months. The book is defensive only and describes no building or running of front-running or sandwich bots. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:链上做市圈最流行的说法是:流动性提供者一边赚手续费,一边承担「无常损失」,价格回到原点损失就消失。本书证伪这个说法。恒定乘积池在持续卖出伽马,套利者每次都按新价格拿走一小块差额,这块差额只与价格走过的平方步长有关,价格回家也拿不回来。几何布朗运动下,恒定乘积池相对重平衡组合的损耗(LVR)每单位价值每单位时间为 σ²/8,日波动五个百分点时约为每天 3.125 个基点。核对原文之后,本书做了蓝图修正:σ²/8 的依赖关系最早见于安杰里斯等人的分析,2022 年的论文把它推广为逐条路径成立的分解;它是无摩擦上限,费率 30 个基点、十二秒出块时套利者实际只拿走约 0.120;集中流动性的放大只在区间之内成立;「手续费常常不足以覆盖 LVR」只在高波动、低不知情成交的参数区成立。书中参数为示意,全部计算由书中程序生成。本书是非个人化投资风险分析,不构成投资建议,不推荐任何代币、协议或交易;加密资产高风险,可能全部损失。 第一章由价值函数 V = 2L√p 说明做市等于卖出伽马。第二章推导 σ²/8 并量化手续费与出块间隔的折扣。第三章给出集中区间的放大因子,正负一成的区间在区间内放大到 21.49 倍,并计算出界概率。第四章算盈亏平衡波动率、覆盖率的测量噪声与集中的拥挤效应。第五章只站在受害者一侧:三明治夹击的损失不超过滑点容忍度乘以报价,一阶模型给出对手无利可图的容忍度,例如池深一百万、费率 0.30 %、交易占池深百分之一时约为 0.00427。 第六章把合约尾部风险写成泊松跳跃,推出带跳跃的 Kelly 仓位上限,例如事故年率 0.03 且可能全损时 Kelly 分数降到 0.3406,并给出金库的再平衡带宽。第七章考辨《管子·国蓄》的轻重平准与白芝浩 1873 年的《伦巴第街》。第八章给出月度链上仓位体检、蒸馏小工具「做市损耗与滑点体检器」与预注册证伪方案,主检验需要 56 个头寸月。本书只做防御与风险度量,不写任何抢跑或三明治机器人的搭建与运营。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Concavity of Constant Product: Market Making Is Selling Gamma — Chapter 1 uses the value function V = 2L√p to show that market making is selling gamma. (https://5dlibrary.com/answers/en/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c01/) 第1章 恒定乘积的凹性:做市就是持续卖出伽马 — 第一章由价值函数 V = 2L√p 说明做市等于卖出伽马。 (https://5dlibrary.com/answers/zh/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c01/) - Ch.2 The σ²/8 Rate of Loss-Versus-Rebalancing, Discounted by Fees and Block Times — Chapter 2 derives σ²/8 and measures the discount from fees and block times. (https://5dlibrary.com/answers/en/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c02/) 第2章 重平衡损耗的速率 σ²/8:手续费与出块间隔怎样给它打折 — 第二章推导 σ²/8 并量化手续费与出块间隔的折扣。 (https://5dlibrary.com/answers/zh/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c02/) - Ch.3 Concentrated Liquidity: the Amplification Factor, the Range Value Function and the Odds of Leaving the Range — Chapter 3 gives the amplification factor of concentrated ranges, 21.49 inside a range of plus or minus ten per cent, and the odds of leaving the range. (https://5dlibrary.com/answers/en/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c03/) 第3章 集中流动性:放大因子、区间价值函数与出界概率 — 第三章给出集中区间的放大因子,正负一成的区间在区间内放大到 21.49 倍,并计算出界概率。 (https://5dlibrary.com/answers/zh/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c03/) - Ch.4 Fees Against LVR: Break-Even Volatility, Measurement Noise and the Crowded Range — Break-even needs two numbers, the fee yield $f\,u\,365$ and $\sigma^2/8$. The threshold volatility is $\sqrt{8\,f\,u\,365}$. At thirty basis points and daily turnover of 0.05 the frictionless threshold is 0.662, and an asset with annual volatility 0.8 has a coverage ratio of only 0.684. Blocks and fees raise the threshold, so the slogan that fees often fail to cover LVR must be replaced by a more careful statement: in pools with high volatility and low uninformed turnover, fees are likely to fall short. Measurement is noisy as well; with fat-tailed returns, one month of data has a 0.416 chance of reading a losing position as a winning one. Concentration pays for the individual and costs the industry. (https://5dlibrary.com/answers/en/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c04/) 第4章 手续费与 LVR 的盈亏平衡:覆盖率、测量噪声与拥挤的集中 — 盈亏平衡只要两个数:手续费年化 $f\,u\,365$ 与 $\sigma^2/8$。门槛波动率是 $\sqrt{8\,f\,u\,365}$;30 个基点、每天换手 0.05 倍时,无摩擦门槛是 0.662,年化波动 0.8 的资产覆盖率只有 0.684。出块与手续费会抬高门槛,所以「手续费常常盖不住 LVR」要换成更诚实的说法:高波动、低不知情换手的池子,手续费很可能盖不住。测量本身也很吵,厚尾收益下一个月的数据有 0.416 的机会把亏损仓位读成盈利。集中在个体层面划算、在行业层面亏损。 (https://5dlibrary.com/answers/zh/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c04/) - Ch.5 Bounding the Sandwich: Your Slippage Tolerance Is the Largest Loss You Offer — Chapter 5 takes only the victim's side: a sandwich can take no more than the slippage tolerance times the quote, and a first-order model gives the tolerance below which no adversary profits, about 0.00427 for a trade of one per cent of a pool one million deep at a 0.30 % fee. (https://5dlibrary.com/answers/en/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c05/) 第5章 三明治夹击的上界:滑点容忍度就是写给对手的最大损失 — 第五章只站在受害者一侧:三明治夹击的损失不超过滑点容忍度乘以报价,一阶模型给出对手无利可图的容忍度,例如池深一百万、费率 0.30 %、交易占池深百分之一时约为 0.00427。 (https://5dlibrary.com/answers/zh/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c05/) - Ch.6 Pricing Smart-Contract Tail Risk with Poisson Jumps, Kelly Position Caps and Vault Rebalancing Bands — Chapter 6 treats contract tail risk as Poisson jumps and derives a Kelly cap with jumps, which falls to 0.3406 when a total loss arrives at a rate of 0.03 a year, together with a rebalancing band for vaults. (https://5dlibrary.com/answers/en/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c06/) 第6章 合约尾部风险的泊松定价、Kelly 仓位上限与金库再平衡带宽 — 第六章把合约尾部风险写成泊松跳跃,推出带跳跃的 Kelly 仓位上限,例如事故年率 0.03 且可能全损时 Kelly 分数降到 0.3406,并给出金库的再平衡带宽。 (https://5dlibrary.com/answers/zh/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c06/) - Ch.7 Classical Sources: the Guanzi on Light-and-Heavy Price Levelling and Bagehot's Lombard Street — Chapter 7 reads the Guanzi on the levelling of prices and Bagehot's Lombard Street of 1873. (https://5dlibrary.com/answers/en/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c07/) 第7章 原典考辨:《管子》轻重平准与白芝浩的《伦巴第街》 — 第七章考辨《管子·国蓄》的轻重平准与白芝浩 1873 年的《伦巴第街》。 (https://5dlibrary.com/answers/zh/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c07/) - Ch.8 Closing the Loop: a Monthly On-Chain Position Check-up, the Distilled Tool and a Pre-registered Refutation — Chapter 8 gives a monthly on-chain position check-up, the distilled Market-Making Loss and Slippage Check-up and a pre-registered refutation plan whose main test needs 56 position-months. The book is defensive only and describes no building or running of front-running or sandwich bots. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c08/) 第8章 实证闭环:月度链上仓位体检、蒸馏小工具与预注册证伪 — 第八章给出月度链上仓位体检、蒸馏小工具「做市损耗与滑点体检器」与预注册证伪方案,主检验需要 56 个头寸月。本书只做防御与风险度量,不写任何抢跑或三明治机器人的搭建与运营。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/concentrated-liquidity-lvr-mev-sandwich-bounds-vault-risk-control/c08/) ## 5DL-B-345 · Paid-Ad Scaling for Diagnostic Paywalls: Meta and TikTok Stopping Rules, Server-Side CAPI Calibration and Hundred-Order Daily Unit Economics 中文书名:测评与命理收费站买量精算、服务端转化回传校准与日入百单扩量母典 DOI: pending Cite: Fan, Y. (2026). Paid-Ad Scaling for Diagnostic Paywalls: Meta and TikTok Stopping Rules, Server-Side CAPI Calibration and Hundred-Order Daily Unit Economics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Under illustrative parameters one icebreaker buyer on tarotmalaysia is worth 2.68 times the icebreaker price, and a first-order break-even stop rule captures only 0.222 of the attainable profit across forty synthetic creatives. Parameters are illustrative and all calculations are generated by the book's programs. Abstract: Assessment, astrology and tarot paywalls give a free result, sell a deeper report and live on a cheap icebreaker and a dearer upsell; most must buy traffic. Four rules of thumb circulate among media buyers: cut anything with first-order ROAS below 1; cut after a round sum with zero sales, or wait for the learning phase to finish; the more server-side events the better; never add more than a fifth to a budget and nothing will break. This book refutes or bounds each of them. Under illustrative parameters one icebreaker buyer on tarotmalaysia is worth 2.68 times the icebreaker price, and a first-order break-even stop rule captures only 0.222 of the attainable profit across forty synthetic creatives. What can be said after zero sales depends on how many multiples of the target acquisition cost have been spent, and the 95 % red line is 2.996 multiples; a small budget can call creatives clearly bad but never clearly good. Checking the sources corrected the blueprint: Gordon and co-authors (2019) say only that observational methods often fail to reproduce experiments; the media mix model of Jin and co-authors (2017) is a Google technical report; platform figures follow the official help pages. Parameters are illustrative and all calculations are generated by the book's programs. This is general business analysis, not personalised advice; it makes no personal prediction and promises no income. Chapter 1 derives the expected funnel value and the break-even acquisition cost. Chapter 2 builds zero-conversion red lines from the rule of three and gamma priors. Chapter 3 applies Wald's sequential test: calling bad is easy, calling good hard. Chapter 4 compares an even split with Thompson sampling, which turns an illustrative two-week profit of -706 into 1,766, and computes how many new creatives a week a budget can screen. Chapter 5 covers consent, deduplication by event identifier and net-value reporting for server-side events: reporting only the icebreaker price yields an illustrative profit of -1,432 and reporting net value 552, and a randomised holdout measures an illustrative calibration factor of 0.615. Chapter 6 uses Hill saturation and geometric carry-over to show that average ROAS hides marginal ROAS: raising 2,000 ringgit a day by a fifth earns only 0.769 on the increment while average ROAS slips from 1.110 to 1.053; a marginal-return control law beats the twenty-per-cent rule, and "a hundred orders a day" is only an engineering target. Chapter 7 reads the Sunzi on waging war and Hopkins's Scientific Advertising of 1923 and writes the distillation spec for the four kits. Chapter 8 gives a three-site ad ledger, the distilled Ad Stop-Loss and Scaling Ladder Calculator and a pre-registered refutation plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:测评类与命理塔罗类收费站免费出结果、付费解锁深层报告,靠低客单破冰与高客单升单赚钱,几乎都要买量。买量圈子里流传四句口诀:首单 ROAS 不到 1 就砍;零成交花满某个整数就砍,或必须跑满学习期再判;服务端回传发得越多越准;每次加预算不超过两成就不会崩。本书逐条证伪或立界。按示意参数,tarotmalaysia 一个破冰买家的期望净值是破冰价的 2.68 倍,按首单回本线止损,在四十条合成素材上只拿到可得利润的 0.222。零成交时能说的话由花了几倍目标获客成本决定,95 % 把握的红线是 2.996 倍;小预算只能判明显坏,不能判明显好。核对原文以后,本书做了蓝图修正:戈登等人 2019 年的研究只说观察法往往无法复现实验结果,没有给出统一方向;金等人 2017 年的媒介组合模型是谷歌技术报告;平台机制数字只按官方帮助文档陈述。书中参数为示意,全部计算由书中程序生成。本书是非个人化经营建议,不做任何个人占断,不承诺任何收入。 第一章给出漏斗期望净值与盈亏平衡获客成本。第二章用三分之规则、伽马先验与两段漏斗写零成交红线。第三章用信噪比与瓦尔德序贯检验说明判坏容易判好难。第四章比较均分与汤普森抽样,两周示意利润由 -706 变为 1,766,并计算一周能筛几条新素材。第五章写服务端回传的合规前提、事件编号去重与净值回传:只回传破冰标价时示意利润为 -1,432,回传净值时为 552;随机留出组测得示意校准系数 0.615。 第六章用 Hill 饱和曲线与几何滞留说明平均 ROAS 掩盖边际 ROAS:从每天 2,000 令吉加两成,新增部分的 ROAS 只有 0.769,平均 ROAS 只从 1.110 降到 1.053;按边际回报控制预算的律优于两成口诀,「日入百单」只作工程目标反推口袋数与上新速度。第七章考辨《孙子兵法·作战篇》与霍普金斯 1923 年的《科学的广告》,并写出给四件套的蒸馏规格。第八章给出三站买量台账、蒸馏小工具「买量止损与扩量阶梯试算器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Funnel Unit Economics: Why First-Order Break-Even Is Not a Stop Line — Of the forty creatives, 23 truly make money, and keeping exactly those earns 17,241 ringgit. Rule A kills 15 good creatives and keeps 2 bad ones, and earns only 3,822, or 0.222 of the attainable profit. Rule B does much better: it kills 7 good creatives but keeps 13 bad ones and earns 12,295, a share of 0.713. Rule C makes no mistakes and takes the full 1.000. (https://5dlibrary.com/answers/en/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c01/) 第1章 漏斗单位经济:首单回本线为什么不是止损线 — 四十条素材里有 23 条真正赚钱,全部留下时的利润是 17,241 令吉。规则 A 误杀了 15 条好素材,又放过 2 条坏素材,只拿到 3,822,占可得利润的 0.222。规则 B 好得多,误杀 7 条,可它放过了 13 条坏素材,拿到 12,295,占 0.713。规则 C 一条不错,拿满 1.000。 (https://5dlibrary.com/answers/zh/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c01/) - Ch.2 Stopping at Zero: the Rule of Three, Gamma-Poisson Posteriors and Spend Red Lines — Part B gives the 90 % red lines. At 0.45 per click the line needs 111 clicks, a spend of 49.95 ringgit or 1.88 multiples of target. At 0.90 per click it needs 43 clicks, 38.70 ringgit, 1.46 multiples. At 1.80 per click it needs only 9 clicks, 16.20 ringgit, 0.61 multiples. With the same zero unlocks, a creative with expensive clicks should be stopped earlier, because it demands a higher unlock rate. (https://5dlibrary.com/answers/en/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c02/) 第2章 零成交止损:三分之规则、伽马—泊松后验与止损红线 — B 段给出 90 % 把握的红线。点击单价 0.45 时要 111 次点击,花 49.95 令吉,是 1.88 倍目标;点击单价 0.90 时要 43 次点击,38.70 令吉,1.46 倍;点击单价 1.80 时只要 9 次点击,16.20 令吉,0.61 倍。同样零解锁,贵点击的素材应该停得更早,因为它对解锁率的要求更苛刻。 (https://5dlibrary.com/answers/zh/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c02/) - Ch.3 Signal, Noise and Sequential Tests: Small Budgets Can Only Call "Clearly Bad" — For $\rho$ of 3.0 only 6 sales are needed, which at a target acquisition cost of 26.51 is 159 ringgit, less than two days on a budget of one hundred a day. For $\rho$ of 2.0 the requirement is 13 sales and 345 ringgit. For $\rho$ of 1.5 it is 38 sales and 1,007 ringgit, about ten days. For $\rho$ of 1.2 it is 186 sales and 4,931 ringgit, about 49.3 days. For $\rho$ of 1.1 it is 681 sales and 18,053 ringgit, half a year at one hundred a day. (https://5dlibrary.com/answers/en/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c03/) 第3章 信噪比与序贯检验:小预算只能判明显坏 — $\rho$ 为 3.0 时只要 6 单,按目标获客成本 26.51 折合 159 令吉,每天一百令吉的预算不到两天就够。$\rho$ 为 2.0 时要 13 单、345 令吉。$\rho$ 为 1.5 时要 38 单、1,007 令吉,大约十天。$\rho$ 为 1.2 时要 186 单、4,931 令吉,大约 49.3 天。$\rho$ 为 1.1 时要 681 单、18,053 令吉,每天一百令吉要跑半年。 (https://5dlibrary.com/answers/zh/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c03/) - Ch.4 Many Creatives on a Small Budget: Thompson Sampling, Even Splits and Screening Speed — Chapter 4 compares an even split with Thompson sampling, which turns an illustrative two-week profit of -706 into 1,766, and computes how many new creatives a week a budget can screen. (https://5dlibrary.com/answers/en/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c04/) 第4章 小预算多素材:汤普森抽样、均分预算与筛选速度 — 第四章比较均分与汤普森抽样,两周示意利润由 -706 变为 1,766,并计算一周能筛几条新素材。 (https://5dlibrary.com/answers/zh/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c04/) - Ch.5 Calibrating Server-Side Events: Deduplication, Net-Value Signals and the Report Inflation Factor — Chapter 5 covers consent, deduplication by event identifier and net-value reporting for server-side events: reporting only the icebreaker price yields an illustrative profit of -1,432 and reporting net value 552, and a randomised holdout measures an illustrative calibration factor of 0.615. (https://5dlibrary.com/answers/en/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c05/) 第5章 服务端回传校准:去重、净值回传与报表膨胀系数 — 第五章写服务端回传的合规前提、事件编号去重与净值回传:只回传破冰标价时示意利润为 -1,432,回传净值时为 552;随机留出组测得示意校准系数 0.615。 (https://5dlibrary.com/answers/zh/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c05/) - Ch.6 Control Laws for Scaling: Saturation, Carry-Over and Marginal Returns — Chapter 6 uses Hill saturation and geometric carry-over to show that average ROAS hides marginal ROAS: raising 2,000 ringgit a day by a fifth earns only 0.769 on the increment while average ROAS slips from 1.110 to 1.053; a marginal-return control law beats the twenty-per-cent rule, and "a hundred orders a day" is only an engineering target. (https://5dlibrary.com/answers/en/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c06/) 第6章 扩量控制律:饱和曲线、滞留效应与边际回报 — 第六章用 Hill 饱和曲线与几何滞留说明平均 ROAS 掩盖边际 ROAS:从每天 2,000 令吉加两成,新增部分的 ROAS 只有 0.769,平均 ROAS 只从 1.110 降到 1.053;按边际回报控制预算的律优于两成口诀,「日入百单」只作工程目标反推口袋数与上新速度。 (https://5dlibrary.com/answers/zh/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c06/) - Ch.7 Classical Sources: the Sunzi on Waging War and Hopkins's Scientific Advertising — Chapter 7 reads the Sunzi on waging war and Hopkins's Scientific Advertising of 1923 and writes the distillation spec for the four kits. (https://5dlibrary.com/answers/en/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c07/) 第7章 原典考辨:《孙子兵法·作战篇》与霍普金斯的《科学的广告》 — 第七章考辨《孙子兵法·作战篇》与霍普金斯 1923 年的《科学的广告》,并写出给四件套的蒸馏规格。 (https://5dlibrary.com/answers/zh/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c07/) - Ch.8 Closing the Loop: a Three-Site Ad Ledger, the Distilled Tool and a Pre-Registered Refutation — Chapter 8 gives a three-site ad ledger, the distilled Ad Stop-Loss and Scaling Ladder Calculator and a pre-registered refutation plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c08/) 第8章 实证闭环:三站买量台账、蒸馏小工具与预注册证伪 — 第八章给出三站买量台账、蒸馏小工具「买量止损与扩量阶梯试算器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/paid-ad-scaling-assessment-paywalls-zero-conversion-stopping-capi-calibration/c08/) ## 5DL-B-346 · Programmatic SEO Long-Tail Topology, Rapid Indexation and High-Intent Organic Traffic Harvesting across Multi-Domain Fleets 中文书名:多域名程序化SEO长尾拓扑、新站沙盒极速收录与高意图自然流量收割母典 DOI: pending Cite: Fan, Y. (2026). Programmatic SEO Long-Tail Topology, Rapid Indexation and High-Intent Organic Traffic Harvesting across Multi-Domain Fleets (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations come from the book's programs. This is general business analysis, not personalised advice: property content is not legal, tax or investment advice, the tarot material offers no personal readings, and all data are synthetic. Abstract: Starting a site with no advertising budget comes with four popular sentences: multiply templates by keywords and the more pages you publish the more long-tail traffic you get; more internal links are always better and value spreads evenly; new sites sit in a sandbox for three to six months; more traffic is always better. This book rewrites each as a model that can be computed and refutes it or bounds it. After checking the original papers and the platforms' public documentation, the book corrects its blueprint: a 64-bit SimHash with a Hamming threshold of 3 folds only pages that are almost identical word for word, so that 0.9750 of page pairs differing in a single place name fall within the threshold, and the real boundary is information gain and the spam policies; Google is not on the public IndexNow participant list; Google does not use the changefreq and priority fields of sitemaps; and the Indexing API covers only job-posting and livestream structured data. Parameters are illustrative and all calculations come from the book's programs. This is general business analysis, not personalised advice: property content is not legal, tax or investment advice, the tarot material offers no personal readings, and all data are synthetic. Chapter 1 uses the Zipf distribution and intent weighting to show that long-tail transactional terms are valuable because they are specific and little contested. Chapter 2 applies three cuts that reduce 2,880 keyword cells to 393 pages. Chapter 3, the spine of the book, derives from shingle resemblance, SimHash and a site-mean model a floor on unique facts per page, from 1.72 to 8.31 facts in the book's setting; adding 1,000 pages with a single fact each to a site of 400 pages lowers modelled traffic by 0.5091. Chapter 4 compares four internal-link topologies; the funnel lifts the inquiry page to 7.12 times the uniform share and visitors reach it in 2.84 clicks on average. Chapter 5 splits the "sandbox" into discovery delay and host queueing: an orphan page has a median first crawl of 89.33 days, which falls to 7.30 days once it is in the sitemap, and enriching pages before releasing them in value order gives a 30-day indexed value share of 0.7038 against 0.0654 for launching everything at once. Chapter 6 treats the independent value and division of queries in a fleet of domains, raising covered queries from 1,548 to 1,782, and shows how to maintain hreflang clusters. Chapter 7 reads Zheng Qiao's treatise on classes in the Tongzhi and Vannevar Bush's associative trails of 1945 and gives the distillation spec for the four kits. Chapter 8 supplies the batch ledger, the distilled Long-Tail Page Information-Gain and Indexation Scheduler and a pre-registered refutation plan whose crossover design needs 31 sections. The book describes white-hat practice only. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:零广告费冷启动流传着四句流行说法:模板乘关键词,页面越多长尾流量越多;内链越多越好,权重平均分;新站有沙盒,要等三到六个月;流量越大越好。本书把每一句写成可以计算的模型,逐句证伪或立界。核对原文与平台公开文档之后,本书做了蓝图修正:64 位 SimHash 加汉明距离 3 只折叠几乎逐字相同的页面,只替换一处区名的页面对有 0.9750 落在阈值内,真正的边界是信息增益与垃圾政策;IndexNow 的公开参与名单里没有 Google;Google 不使用站点地图的 changefreq 与 priority;Indexing API 只适用于招聘与直播结构化数据。书中参数为示意,全部计算由书中程序生成。本书是非个人化经营建议;房产内容不构成法律、税务或投资建议,塔罗部分不做个人占断,宗祠资料须经本人同意,书中数据全部合成。 第一章用齐普夫分布与意图加权说明长尾交易词的价值来自具体与少竞争。第二章用三道剪刀把 2,880 个关键词格子剪到 393 页。第三章是全书主轴:用片段相似、SimHash 与站点均值模型推出每页独有事实数的下界,本书设定下从 1.72 条到 8.31 条;一座 400 页的站点一次加 1,000 页、每页只有 1 条事实,模型流量下降 0.5091。第四章比较四种内链拓扑,漏斗把询盘页推到均匀分布的 7.12 倍,访客平均 2.84 次点击到达。 第五章把「沙盒」拆成发现延迟与主机队列:孤儿页首抓中位数 89.33 天,进站点地图后降到 7.30 天;先补写事实再按价值分批上线,30 天收录价值份额 0.7038,一次全上只有 0.0654。第六章讨论多域名舰队的独立价值与查询分工,分工后覆盖查询从 1,548 条增到 1,782 条,并给出 hreflang 簇的维护方法。第七章考辨郑樵《通志·校雠略》的类例与布什 1945 年的联想路径,并给出四件套蒸馏规格。第八章给出批次台账、蒸馏小工具「长尾页信息增益与收录排程器」与预注册证伪方案,交叉设计需要 31 个栏目。本书只写白帽做法。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Zipf Long Tail: Head, Tail and Intent-Weighted Cash — Start with the row for $s = 1.0$. The top 100 queries take 0.4291 of all searches, and the queries beyond position 1,000 together take 0.3809. The head is heavy and the tail is long, and the two are of the same order. The last column adds a sharper fact: from position 82,712 onwards a query expects fewer than one search a month, so 82,711 queries reach at least one a month. The lower half of the table makes the decline concrete. The top query receives 82,711.99 searches a month, the query at position 1,000 only 82.71, the one at 10,000 receives 8.27, and the one at 50,000 receives 1.65. (https://5dlibrary.com/answers/en/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c01/) 第1章 齐普夫长尾:头部、尾部与按意图加权的现金 — 先看 $s = 1.0$ 这一行。前 100 条查询拿走 0.4291 的搜索,排在第 1,000 位以后的查询合计拿走 0.3809。头部确实重,尾部也确实长,两者量级相当。再看最后一列:排第 82,712 位起,每条查询每月期望不到一次搜索;能达到每月一次的查询有 82,711 条。下半张表更直观:第 1 位每月 82,711.99 次,第 1,000 位只剩 82.71 次,第 10,000 位 8.27 次,第 50,000 位 1.65 次。 (https://5dlibrary.com/answers/zh/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c01/) - Ch.2 The Keyword Matrix: Cartesian Products, Feasible Cells and the First Hundred Pages — Chapter 2 applies three cuts that reduce 2,880 keyword cells to 393 pages. (https://5dlibrary.com/answers/en/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c02/) 第2章 关键词矩阵:笛卡尔积、可行组合与先建哪一百页 — 第二章用三道剪刀把 2,880 个关键词格子剪到 393 页。 (https://5dlibrary.com/answers/zh/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c02/) - Ch.3 A Floor on Information Gain: Shingles, SimHash and Unique Facts per Page — Chapter 3, the spine of the book, derives from shingle resemblance, SimHash and a site-mean model a floor on unique facts per page, from 1.72 to 8.31 facts in the book's setting; adding 1,000 pages with a single fact each to a site of 400 pages lowers modelled traffic by 0.5091. (https://5dlibrary.com/answers/en/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c03/) 第3章 信息增益下界:片段相似、SimHash 与每页独有事实数 — 第三章是全书主轴:用片段相似、SimHash 与站点均值模型推出每页独有事实数的下界,本书设定下从 1.72 条到 8.31 条;一座 400 页的站点一次加 1,000 页、每页只有 1 条事实,模型流量下降 0.5091。 (https://5dlibrary.com/answers/zh/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c03/) - Ch.4 Internal-Link Topology: the PageRank Funnel and the Stationary Share of the Inquiry Page — Chapter 4 compares four internal-link topologies; the funnel lifts the inquiry page to 7.12 times the uniform share and visitors reach it in 2.84 clicks on average. (https://5dlibrary.com/answers/en/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c04/) 第4章 内链拓扑:PageRank 漏斗与询盘页的稳态概率 — 第四章比较四种内链拓扑,漏斗把询盘页推到均匀分布的 7.12 倍,访客平均 2.84 次点击到达。 (https://5dlibrary.com/answers/zh/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c04/) - Ch.5 The Indexation Schedule: Poisson Discovery, Host Queues and Batch Launches in the First Thirty Days — Chapter 5 splits the "sandbox" into discovery delay and host queueing: an orphan page has a median first crawl of 89.33 days, which falls to 7.30 days once it is in the sitemap, and enriching pages before releasing them in value order gives a 30-day indexed value share of 0.7038 against 0.0654 for launching everything at once. (https://5dlibrary.com/answers/en/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c05/) 第5章 收录排程:泊松发现、主机队列与前三十天的分批上线 — 第五章把「沙盒」拆成发现延迟与主机队列:孤儿页首抓中位数 89.33 天,进站点地图后降到 7.30 天;先补写事实再按价值分批上线,30 天收录价值份额 0.7038,一次全上只有 0.0654。 (https://5dlibrary.com/answers/zh/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c05/) - Ch.6 A Fleet of Domains: Independent Value, a Division of Queries and hreflang Clusters — Chapter 6 treats the independent value and division of queries in a fleet of domains, raising covered queries from 1,548 to 1,782, and shows how to maintain hreflang clusters. (https://5dlibrary.com/answers/en/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c06/) 第6章 多域名舰队:独立价值、查询分工与 hreflang 簇 — 第六章讨论多域名舰队的独立价值与查询分工,分工后覆盖查询从 1,548 条增到 1,782 条,并给出 hreflang 簇的维护方法。 (https://5dlibrary.com/answers/zh/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c06/) - Ch.7 Reading the Sources: Zheng Qiao's Classes in the Tongzhi and Vannevar Bush's Associative Trails — Chapter 7 reads Zheng Qiao's treatise on classes in the Tongzhi and Vannevar Bush's associative trails of 1945 and gives the distillation spec for the four kits. (https://5dlibrary.com/answers/en/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c07/) 第7章 原典考辨:郑樵《通志·校雠略》的类例与布什的联想路径 — 第七章考辨郑樵《通志·校雠略》的类例与布什 1945 年的联想路径,并给出四件套蒸馏规格。 (https://5dlibrary.com/answers/zh/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c07/) - Ch.8 Closing the Loop: the Batch Ledger, the Information-Gain and Indexation Scheduler, and Pre-Registered Refutation — Chapter 8 supplies the batch ledger, the distilled Long-Tail Page Information-Gain and Indexation Scheduler and a pre-registered refutation plan whose crossover design needs 31 sections. The book describes white-hat practice only. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c08/) 第8章 实证闭环:批次台账、信息增益与收录排程器、预注册证伪 — 第八章给出批次台账、蒸馏小工具「长尾页信息增益与收录排程器」与预注册证伪方案,交叉设计需要 31 个栏目。本书只写白帽做法。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/programmatic-seo-long-tail-topology-information-gain-rapid-indexation/c08/) ## 5DL-B-347 · Active Ad-Library Prospecting, B2B Outbound Deliverability Physics and High-Ticket Whitelist Deal Closing 中文书名:公开广告库金主反查、B2B冷触达进箱物理与高客单暗投签单母典 DOI: pending Cite: Fan, Y. (2026). Active Ad-Library Prospecting, B2B Outbound Deliverability Physics and High-Ticket Whitelist Deal Closing (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: A live ad in a public ad library is right-censored data and must be read with the Kaplan-Meier product-limit estimator; in a synthetic library the rule "live and at least 30 days old" raises the share of steady advertisers from 0.313 to 0.902. Parameters are illustrative and all calculations are generated by the book's programs. Abstract: Business-services teams that go looking for paying brands tend to believe four popular claims: the bigger the list the better, the more you send the more replies you get, a long letter is more professional, and a high-ticket sale needs several rounds of talk before money. This book refutes each in its general form with standard-library programs and marks the boundary within which a narrower version holds. A live ad in a public ad library is right-censored data and must be read with the Kaplan-Meier product-limit estimator; in a synthetic library the rule "live and at least 30 days old" raises the share of steady advertisers from 0.313 to 0.902. After checking the sources the book corrects its blueprint: advertisers cannot measure their returns precisely, so days in flight signal budget commitment rather than, necessarily, profit; empty reasons work only for small requests and their replication is disputed; a deposit is better only when serious brands accept it; and "zero mail in spam" is restated as the boundary of inbox placement. Parameters are illustrative and all calculations are generated by the book's programs. This is a general analysis of business and legal questions, not individualised advice; laws are stated at the level of principle, to be checked by a qualified person before publication. Chapters 1 and 2 score brands with survival analysis and proportional hazards, and show the complaint rate rising to 0.298 % when the whole directory is contacted. Chapters 3 and 4 write sender reputation as a Beta posterior on complaints: once volume exceeds the matched audience, day-30 inbox placement can fall from 0.974 to 0.054; warm-up is a rate controller that reads the posterior, and one-click unsubscribe is a physical device for lowering complaints. Chapter 5 shows that in a multiplicative funnel the three-sentence opener yields about 2.742 times as many pilots as the long letter, and bounds it with the copying-machine experiment and its replication dispute. Chapter 6 analyses a symmetric pilot deposit with signalling and screening theory: with illustrative parameters the separating line is about 1,022.73, and serious brands must accept at a rate above 0.585 for total revenue not to fall. Chapter 7 reads Mao Sui's "awl in a bag" in the Records of the Grand Historian and Aristotle's three modes of persuasion. Chapter 8 gives a weekly ledger, the distilled Ad-Library Prospect Score and Outbound Inbox Budget, and four pre-registered tests. The book describes only compliant practice: ad libraries are consulted under platform terms, cold outreach goes only to published business contacts with truthful identity and a working unsubscribe, authorised creator ads always carry the paid-partnership label, and no filter-evasion technique is described. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:主动寻找付费品牌方的企业服务团队,常信四句流行说法:名单越大越好,发得越多回复越多,长信才专业,高客单要多聊几轮再谈钱。本书用纯标准库程序逐条证伪它们的普遍形式,并给每一句划出成立的边界。公开广告库里仍在投的广告是右删失数据,须用卡普兰与迈耶的乘积极限估计去读;「仍在投且已投满 30 天」这条规则在合成广告库里把持续投放者的比例从 0.313 抬到 0.902。核对原文之后,本书做了蓝图修正:广告主难以测准投放回报,在投时长首先是预算承诺的信号,不一定是盈利信号;空洞理由只对小请求有效且复制有争议;定金只在认真品牌方愿意接受时才更优;「零进垃圾箱」一律改写为进箱率边界。书中参数为示意,全部计算由书中程序生成。本书是非个人化的经营与法律议题分析,法规只写到原则层级,以上架前人工复核为准。 第一、二章用生存分析与比例风险模型给品牌打分,并说明撒网到整份名录时投诉率升到 0.298 %。第三、四章把发信信誉写成投诉率的贝塔后验:发信量越过匹配人群,第 30 天进箱率可从 0.974 跌到 0.054;预热是读后验的速率控制器,一键退订是压低投诉的物理装置。第五章说明乘积漏斗下三句式破冰信的产出约为长信的 2.742 倍,并以复印机实验及其复制争议为它立界。 第六章用信号与筛选理论分析对称试点定金,示意参数下的分离线约为 1,022.73,认真品牌方接受比例须高于 0.585 才不减少总收入。第七章考辨《史记》毛遂「锥之处囊中」与亚里士多德《修辞学》的说服三途。第八章给出周度台账、蒸馏小工具「金主反查评分与冷触达进箱预算器」与四项预注册检验。本书只写合规做法:公开广告库按平台条款查看,冷触达只发公开的企业联系人并如实标明身份、提供有效退订,授权投放一律标明付费合作,不写任何规避过滤的手法。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Days in Flight: Survival Curves in the Ad Library — Chapters 1 and 2 score brands with survival analysis and proportional hazards, and show the complaint rate rising to 0.298 % when the whole directory is contacted. (https://5dlibrary.com/answers/en/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c01/) 第1章 在投时长:广告库里的生存曲线 — 第一、二章用生存分析与比例风险模型给品牌打分,并说明撒网到整份名录时投诉率升到 0.298 %。 (https://5dlibrary.com/answers/zh/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c01/) - Ch.2 Scoring Paying Advertisers: Creative Refresh, Proportional Hazards and List Size — The directory in this run contains 143 buyers, a base rate of 0.071. Taking the top 50 by score yields 34 hits, a precision of 0.680 and a lift of 9.5 over the base rate; we expect 5.60 replies per hundred letters and a complaint rate of 0.098 %. At the top 100 the precision is 0.590 and the complaint rate rises to 0.112 %. At the top 200 the precision is 0.410 and the complaint rate 0.139 %. (https://5dlibrary.com/answers/en/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c02/) 第2章 金主评分:素材更新、比例风险与名单规模 — 这份名录实际抽出 143 家买家,基准比例 0.071。按分数取前 50 家,命中 34 家,精度 0.680,是基准的 9.5 倍;每百封信预计回复 5.60 封,投诉率 0.098 %。取前 100 家,精度 0.590,投诉率升到 0.112 %。取前 200 家,精度 0.410,投诉率 0.139 %。 (https://5dlibrary.com/answers/zh/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c02/) - Ch.3 The Physics of the Inbox: Authentication, Complaint Rates and a Reputation Posterior — Chapters 3 and 4 write sender reputation as a Beta posterior on complaints: once volume exceeds the matched audience, day-30 inbox placement can fall from 0.974 to 0.054; warm-up is a rate controller that reads the posterior, and one-click unsubscribe is a physical device for lowering complaints. (https://5dlibrary.com/answers/en/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c03/) 第3章 进箱物理:认证、投诉率与信誉后验 — 第三、四章把发信信誉写成投诉率的贝塔后验:发信量越过匹配人群,第 30 天进箱率可从 0.974 跌到 0.054;预热是读后验的速率控制器,一键退订是压低投诉的物理装置。 (https://5dlibrary.com/answers/zh/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c03/) - Ch.4 Warm-Up as Feedback Rate Control — Start with the clean list. Over forty-five days the ladder sends 11,324 messages and lands 4,948 in the inbox, draws 15.4 complaints, and ends the last week with inbox placement of only 0.354. The feedback policy sends 5,423 messages, fewer than half as many, yet lands 4,964 in the inbox, slightly more than the ladder; it draws only 3.2 complaints and ends the last week with placement of 0.911. Both reach 150 a day between day 9 and day 11. The difference is that the ladder refuses to stop. It climbs on to 300 a day and sends all the extra mail to poorly matched recipients. (https://5dlibrary.com/answers/en/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c04/) 第4章 预热:带反馈的速率控制 — 先看干净名单。阶梯策略四十五天发出 11,324 封,进箱 4,948 封,投诉 15.4 次,最后一周进箱率只有 0.354。反馈策略只发 5,423 封,不到前者一半,进箱却有 4,964 封,略多于阶梯策略;投诉只有 3.2 次,最后一周进箱率 0.911。两者都在第 9 到第 11 天达到每天 150 封,差别在于阶梯策略不肯停,一路加到 300 封,把多出来的信全送给了匹配差的人。 (https://5dlibrary.com/answers/zh/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c04/) - Ch.5 The Three-Sentence Opener and the Multiplicative Funnel — Chapter 5 shows that in a multiplicative funnel the three-sentence opener yields about 2.742 times as many pilots as the long letter, and bounds it with the copying-machine experiment and its replication dispute. (https://5dlibrary.com/answers/en/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c05/) 第5章 三句式破冰信与乘积漏斗 — 第五章说明乘积漏斗下三句式破冰信的产出约为长信的 2.742 倍,并以复印机实验及其复制争议为它立界。 (https://5dlibrary.com/answers/zh/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c05/) - Ch.6 The Pilot Deposit: Separating Two Kinds of Brand — Chapter 6 analyses a symmetric pilot deposit with signalling and screening theory: with illustrative parameters the separating line is about 1,022.73, and serious brands must accept at a rate above 0.585 for total revenue not to fall. (https://5dlibrary.com/answers/en/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c06/) 第6章 试点定金:分离两类品牌方 — 第六章用信号与筛选理论分析对称试点定金,示意参数下的分离线约为 1,022.73,认真品牌方接受比例须高于 0.585 才不减少总收入。 (https://5dlibrary.com/answers/zh/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c06/) - Ch.7 The Awl in the Bag and the Three Modes of Persuasion: Mao Sui and Aristotle — The differences among the four letters are plain. The three-sentence letter, full marks on ethos and logos and respectful in tone, draws 33.90 replies and 1.23 complaints per thousand. The same facts and identity with a pressuring tone draw only 18.89 replies and 4.07 complaints, 3.31 times as many complaints. The anonymous "hook" letter, with almost no ethos, draws 2.87 replies and 13.39 complaints. The long company profile, middling on ethos and without argument, draws 4.07 replies and 2.24 complaints. (https://5dlibrary.com/answers/en/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c07/) 第7章 锥处囊中与说服三途:毛遂与亚里士多德 — 四封信的差别一目了然。三句式信品格满分、论证满分、语气尊重,每千封回复 33.90 封,投诉 1.23 次。同样的事实与身份,换成施压的语气,回复降到 18.89 封,投诉升到 4.07 次,是原来的 3.31 倍。匿名的「钩子信」几乎没有品格,回复 2.87 封,投诉 13.39 次。长篇公司介绍品格一般、没有论证,回复 4.07 封,投诉 2.24 次。 (https://5dlibrary.com/answers/zh/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c07/) - Ch.8 The Weekly Ledger and the Refutation Loop — Chapter 8 gives a weekly ledger, the distilled Ad-Library Prospect Score and Outbound Inbox Budget, and four pre-registered tests. The book describes only compliant practice: ad libraries are consulted under platform terms, cold outreach goes only to published business contacts with truthful identity and a working unsubscribe, authorised creator ads always carry the paid-partnership label, and no filter-evasion technique is described. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c08/) 第8章 周度台账与证伪闭环 — 第八章给出周度台账、蒸馏小工具「金主反查评分与冷触达进箱预算器」与四项预注册检验。本书只写合规做法:公开广告库按平台条款查看,冷触达只发公开的企业联系人并如实标明身份、提供有效退订,授权投放一律标明付费合作,不写任何规避过滤的手法。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/ad-library-prospecting-b2b-outbound-deliverability-pilot-deposit-closing/c08/) ## 5DL-B-348 · Five-Minute High-Intent Lead Conversion, Offline Deal-Leakage Auditing and Referral Commission Lock-In 中文书名:高意向线索黄金五分钟转化、线下派单防飞单审计与转介绍佣金锁死母典 DOI: pending Cite: Fan, Y. (2026). Five-Minute High-Intent Lead Conversion, Offline Deal-Leakage Auditing and Referral Commission Lock-In (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. This is general analysis of law, tax, investment and business operations, not personalised advice. Abstract: Turning web forms into money comes with four popular sentences: a call within five minutes is a hundred times more likely to reach the lead; a WhatsApp opener should go out as early and as often as possible, ideally from a bot; a lead passed to an offline partner agent is a job done; and one heavy fine for one leaked deal ends leakage. This book refutes or bounds each of them. The "hundred times" figure comes from an industry study of 2007 that was never peer reviewed and compares odds, not probabilities; the 2011 Harvard Business Review article reports smaller multiples, and the magazine is edited rather than refereed, with one author tied to the software company behind the study, which is the book's correction to its blueprint. A two-group curve fitted to the two multiples of 2011 implies a five-minute versus thirty-minute ratio of only about 2.33. Confounding drives the inflation: in a fixed-seed simulation the pooled comparison reports 3.206 while the true multiple for the same two delay bands is 1.441. Parameters are illustrative and all calculations are generated by the book's programs. This is general analysis of law, tax, investment and business operations, not personalised advice. Chapters 1 and 2 write contact as a hazard function of delay and show how randomised delays and a shift boundary recover the causal decay. Chapter 3 uses the M/M/c queue to size the desk for a five-minute promise and shows that triage, a person first for high-intent leads and a disclosed automatic message followed by a person for the rest, beats both people for everyone and a bot for everyone. Chapter 4 times the first message within consent, the twenty-four-hour window and the template rules. Chapter 5 treats leakage as an agency game in which the value of a repeated relationship is the main deterrent, and reads the evidence on agents' premiums when they sell their own homes. Chapter 6 gives the operating characteristic curve of a sampling audit, the incentive conditions for buyer confirmation and the expected referral share under registration, a protection period and escrow; with illustrative parameters, three callbacks among forty files reported as not closed detect a single leak with probability only 0.0675, so it is the value of the relationship that deters. Chapter 7 examines the two handles of reward and punishment in the Han Feizi and Adam Smith on directors who manage other people's money. Chapter 8 delivers a lead and hand-off ledger for the MPE Malaysian property matrix and the distribution board, the distilled Five-Minute Lead Triage and Deal-Leakage Audit Sampler and a pre-registered refutation plan whose main test needs 1,004 leads per arm. Only registered Malaysian estate agencies collect and share commission; board rewards are single-tier, by price band; WhatsApp contact follows consent and automatic messages say so; callbacks are agreed in advance. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:把表单变成现金,行内有四句流行说法:五分钟内回电联系率高一百倍;WhatsApp 破冰越早越多越好,最好全交给机器人;线索派给线下合作经纪就算办完;抓到一次飞单重罚一次就能杜绝。本书逐条证伪或立界。「一百倍」出自 2007 年一份未经同行评审的行业研究,比的是几率而不是概率;2011 年《哈佛商业评论》短文报告的倍数较小,而且该刊是编辑审稿的管理杂志,作者之一与研究所用软件公司有利益关联,这是本书的蓝图修正。用两组人群的混合曲线拟合 2011 年的两个倍数,五分钟对三十分钟只差约 2.33 倍。倍数被高估主要来自混杂:固定种子模拟里混合比较得出 3.206 倍,同样两个延迟区间的真实倍数只有 1.441。书中参数为示意,全部计算由书中程序生成。本书是非个人化的法律、税务、投资与经营分析,不构成任何建议。 第一、二章把联系概率写成随延迟衰减的危险函数,演示随机分派与断点设计怎样估出因果衰减。第三章用 M/M/c 排队算五分钟达标所需人手,并证明高意向先由真人首触、其余先发表明自动的消息再由真人跟进的分级方案优于全真人与全机器人。第四章在用户同意、二十四小时客服窗口与模板消息规则之内求最优首触时点。第五章写飞单的委托代理博弈与重复合作的关系价值,并读经纪出售自家房屋的溢价证据。第六章给出验收抽样的 OC 曲线、买家双向确认的激励相容条件,以及引荐登记、保护期与持牌机构托管下的期望佣金;在示意参数下,三次回访对四十份未成交档案的查获概率只有 0.0675,靠的是关系价值才足以威慑。 第七章考辨《韩非子·二柄》的刑德两柄与亚当·斯密论股份公司董事管理他人之钱。第八章给出 MPE 马来西亚房产矩阵与分销盘可直接使用的线索与派单台账、蒸馏小工具「线索黄金五分钟分级与防飞单抽检器」与预注册证伪方案,主检验每组需要 1,004 条线索。马来西亚房地产经纪佣金只能由依法注册的经纪机构与人员收取与分配;分销盘推荐奖励只给直接推荐人、按客单价档位计、绝无多层计佣;WhatsApp 只在用户同意后联系,自动消息表明自动;回访与确认事先约定并取得同意。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Where "a Hundred Times" Comes From: Hazard Functions of Lead Decay — "Call within five minutes and you are a hundred times more likely to make contact" comes from an industry study that was not peer reviewed and that compared odds rather than probabilities. The Harvard Business Review article reports multiples on a different yardstick and carries the same commercial interest. An odds ratio of one hundred usually corresponds to a probability ratio of a few dozen, and at high baselines to about ten. Once the probability of contact is written as a hazard function, the published multiples cannot all come from one smooth curve: calibrating on twenty-one produces astronomical multiples at later windows, while calibrating on the two 2011 multiples leaves five minutes ahead of thirty by only 2.33 times. (https://5dlibrary.com/answers/en/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c01/) 第1章 一百倍从哪里来:线索衰减的危险函数 — 「五分钟内回电,联系率高一百倍」来自一份没有经过同行评审的行业研究,比的是几率而不是概率;《哈佛商业评论》那篇短文报告的是另一组口径的倍数,同样有利益关联。几率比一百,换成概率比往往只有几十,在高基准下只有十倍左右。把联系概率写成危险函数后,已发表的几组倍数无法同时出自一条平滑曲线:用二十一倍定标,后面的窗口出现天文数字;用 2011 年的两个倍数定标,五分钟对三十分钟只差 2.33 倍。 (https://5dlibrary.com/answers/zh/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c01/) - Ch.2 Fast Teams Are Better Teams: Confounding, Randomised Delays and a Shift Boundary — Fast teams are better teams, so comparing fast teams with slow teams inflates the effect of delay. In the known world of program p21 the pooled comparison gives 3.206, the true multiple over the same bins is 1.441, and stratifying within teams brings the estimate back to 1.417. Randomising three thousand leads gives a multiple of 1.235, with an interval that covers the truth and excludes the pooled figure. The handover discontinuity is nearly unbiased over repeated samples, though any single estimate is noisy. (https://5dlibrary.com/answers/en/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c02/) 第2章 快的团队本来就好:混杂、随机分派与断点 — 快团队本来就好,所以快团队与慢团队的比较会放大延迟的效应。在程序 p21 的已知世界里,混合比较得出 3.206 倍,同样区间的真实倍数是 1.441,团队内分层把估计拉回 1.417。随机分派的三千条线索给出 1.235 倍,区间罩住了真实值并排除了混合倍数。交班断点在重复抽样下几乎无偏,单次估计却波动明显。 (https://5dlibrary.com/answers/zh/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c02/) - Ch.3 How Many People Does Five Minutes Take? Queues, Triage and the Desk — The five-minute promise is a queueing problem. With 18 forms an hour and first touches of 8 minutes, meeting the target for nine leads in ten takes 5 people; a campaign peak takes 6, and nineteen in twenty takes 7. The price of the promise is slack: in a quiet hour each person is busy only 0.2667 of the time. Triage keeps the promise with fewer people. Three people who triage already outperform four who serve in arrival order, and arrival order needs six people to catch up with four who triage. (https://5dlibrary.com/answers/en/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c03/) 第3章 五分钟要几个人:排队、分级与值班人手 — 五分钟承诺是一道排队题。每小时 18 条表单、首触平均 8 分钟时,九成达标需要 5 个人;投放高峰需要 6 个人,九成五要 7 个人。达标的代价是闲置,安静时段的利用率只有 0.2667。分级能用更少的人守住承诺:三个人做分级,价值已超过四个人做先来先服务;先来先服务要六个人,才追得上四个人的分级。 (https://5dlibrary.com/answers/zh/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c03/) - Ch.4 The First Message after Consent: Timing, the 24-Hour Window and Disclosed Automation — A WhatsApp opener is bounded three ways: by the user's consent, by the platform's rules and by an honest statement that a message is automatic. Within those bounds the reply rate is set by two opposing forces, decay and preparation, and the best moment for a human first message is around minute 7.18, with a reply rate of 0.4625. A self-identified automatic acknowledgement at night lifts the daily first-reply rate from 0.3182 to 0.4376. On costs and benefits, the plan in which only tier A receives a human first touch has the highest net value, 20,513.8, ahead of the all-bot plan at 16,560.0 and the all-human plan at 14,508.8. (https://5dlibrary.com/answers/en/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c04/) 第4章 同意之后的第一条消息:破冰时点、二十四小时窗口与表明自动 — WhatsApp 破冰受三道边界约束:用户同意、平台规则、诚实表明自动。在边界之内,回复率由衰减与准备两股力共同决定,真人首条消息的最佳时点约在第 7.18 分钟,回复率 0.4625;夜间先发表明身份的自动确认,把全天首条回复率从 0.3182 提到 0.4376。成本收益上,只让 A 级由真人首触的分级方案净值最高,为 20,513.8,高于全自动的 16,560.0 与全真人的 14,508.8。 (https://5dlibrary.com/answers/zh/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c04/) - Ch.5 The Arithmetic of Deal Leakage: Agency, Information and the Value of the Relationship — Leakage is an arithmetic problem. The referral share the partner keeps is the gain; the detection probability times the sum of repayment, agreed damages and future relationship value is the cost. Under trust alone the threshold is 1.0000 and leaking always pays. Under chance discovery alone the expected gain stays positive even when damages and relationship value are both written into the contract. What really lowers the threshold is a steady, continuing relationship, and that deterrent melts as the relationship approaches its end. An information advantage pulls the agent's interest away from the principal's, and Holmström tells us to put every signal that carries information about the agent's action into the contract. (https://5dlibrary.com/answers/en/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c05/) 第5章 飞单的算术:委托代理、信息优势与关系价值 — 飞单是一道算术题。合作方省下的引荐份额是收益,查获概率乘以退款、违约赔偿与未来合作价值之和是成本。只靠信任时,门槛是 1.0000,飞单永远划算;只靠偶然发现时,哪怕赔偿与合作价值都写进合约,期望收益仍然为正。真正压低门槛的是稳定而持续的合作关系,但它在合作终点附近会融化。信息优势让代理人的利益天然偏离,霍姆斯特伦告诉我们应当把一切带有行动信息的信号写进合约。 (https://5dlibrary.com/answers/zh/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c05/) - Ch.6 Sampling Audits, Buyer Confirmation and Locking In the Referral Share — Chapter 6 gives the operating characteristic curve of a sampling audit, the incentive conditions for buyer confirmation and the expected referral share under registration, a protection period and escrow; with illustrative parameters, three callbacks among forty files reported as not closed detect a single leak with probability only 0.0675, so it is the value of the relationship that deters. (https://5dlibrary.com/answers/en/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c06/) 第6章 抽检、买家确认与佣金锁死 — 第六章给出验收抽样的 OC 曲线、买家双向确认的激励相容条件,以及引荐登记、保护期与持牌机构托管下的期望佣金;在示意参数下,三次回访对四十份未成交档案的查获概率只有 0.0675,靠的是关系价值才足以威慑。 (https://5dlibrary.com/answers/zh/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c06/) - Ch.7 Classical Sources: the Two Handles of the Han Feizi and Adam Smith on Other People's Money — An owner with a share of 1.000 puts in 39.00 units of effort, spends nothing on comforts and creates a value of 108.56. A partner agency handling a referred lead must pass 0.25 of the commission to the referring agency and keeps 0.750; its effort falls to 29.00, which is 74.36 % of the owner's, although the value lost is small and the result is still 0.9830 of the best. At a share of 0.100 effort collapses to 3.00, comforts rise to 9.00 and only 0.4003 of the best value survives. (https://5dlibrary.com/answers/en/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c07/) 第7章 原典考辨:《韩非子·二柄》的刑德与斯密论「他人之钱」 — 份额为 1.000 的东家投入 39.00 单位努力,不花一分冤枉钱,价值 108.56。合作机构拿到引荐线索,要把 0.25 的佣金分给引荐方,自己留 0.750,努力降到 29.00,是东家的 74.36 %;价值只少了一点,是最优值的 0.9830。份额降到 0.100,努力只剩 3.00,舒适开销升到 9.00,价值只剩最优值的 0.4003。 (https://5dlibrary.com/answers/zh/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c07/) - Ch.8 The Empirical Loop: the Lead and Hand-off Ledger, the Audit Sampler and Pre-registered Refutation — Chapter 8 delivers a lead and hand-off ledger for the MPE Malaysian property matrix and the distribution board, the distilled Five-Minute Lead Triage and Deal-Leakage Audit Sampler and a pre-registered refutation plan whose main test needs 1,004 leads per arm. Only registered Malaysian estate agencies collect and share commission; board rewards are single-tier, by price band; WhatsApp contact follows consent and automatic messages say so; callbacks are agreed in advance. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c08/) 第8章 实证闭环:线索与派单台账、抽检器与预注册证伪 — 第八章给出 MPE 马来西亚房产矩阵与分销盘可直接使用的线索与派单台账、蒸馏小工具「线索黄金五分钟分级与防飞单抽检器」与预注册证伪方案,主检验每组需要 1,004 条线索。马来西亚房地产经纪佣金只能由依法注册的经纪机构与人员收取与分配;分销盘推荐奖励只给直接推荐人、按客单价档位计、绝无多层计佣;WhatsApp 只在用户同意后联系,自动消息表明自动;回访与确认事先约定并取得同意。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/five-minute-lead-decay-offline-deal-leakage-audit-referral-lock-in/c08/) ## 5DL-B-349 · Paywall Abandonment Recovery Timing, Personalised Curiosity Recall and High-Ticket Upsell Ratchets 中文书名:诊断收费站弃单挽回时序、个性化悬念召回与高客单升单棘轮母典 DOI: pending Cite: Fan, Y. (2026). Paywall Abandonment Recovery Timing, Personalised Curiosity Recall and High-Ticket Upsell Ratchets (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: The site is treated as a paid report product, with no claim of predictive accuracy and no fear; all visitor and order data are synthetic or illustrative. Marked boxes separate published evidence, runnable Python calculations (25 programs), classical sources and the author's hypotheses. Abstract: People who run paid content websites hear two sentences more than any others: "more reminders recover more" and "the sooner the recall the better". This book refutes both. Some leavers return by themselves within the first hour, and a message at the first minute lands on them and repeats what is still on screen. In the book's illustrative model a first recall wave at one minute wins back 0.00609 of consented leavers, against 0.03732 at fifteen minutes; three waves packed into one hour are worth RM 1.1790 net per leaver, against RM 1.9419 for waves at fifteen minutes, twenty-four hours and seventy-two hours; and a fifth wave is already at -0.0121. The 95 per cent leak is wide, but three recall waves win back only 0.0237 of it; the big lever is the upgrade after the first order. RM 5,000 a day is a target for working out requirements, not a promise: selling only the entry report needs 5,290 visitors a day, crediting the first order toward the buy-out volume for one buyer in ten needs 2,784, and adding three recall waves needs 1,920. After checking the sources the book corrects its blueprint: Ebbinghaus's forgetting data show only that decay is fast at first and slow later, and fix no sending time; fifteen minutes, twenty-four hours and seventy-two hours reach only 0.807 to 0.965 of the best schedule, and a third wave at forty-eight hours (RM 1.9934) beats seventy-two; the inverted-U dose comes from Kang and co-authors' experiments over confidence, not directly from Loewenstein's theory, and the illustrative net value peaks at a dose of 0.36; Google's published 0.10 % and 0.30 % are one mailbox provider's lines, the book adds its own brake at 0.08 %, and a clean record needs 2,995 messages to show a rate below 0.10 %; and in a static model an upgrade ladder adds almost nothing. The chapters cover the cash equation; intent decay and self-return; a curiosity dose built only from the visitor's own free result; stopping rules, complaint brakes and the consent law of Malaysia, Singapore and the United States; crediting a first RM 19.90 toward RM 199 within a window that never resets; sequential tests, bandits and the false discovery rate; the Han Feizi and Aristotle's Rhetoric; and a Recall Scheduler and Upsell Trigger with a pre-registered refutation plan. The site is treated as a paid report product, with no claim of predictive accuracy and no fear; all visitor and order data are synthetic or illustrative. Marked boxes separate published evidence, runnable Python calculations (25 programs), classical sources and the author's hypotheses. 摘要:经营付费内容网站的人最常听到两句话:「提醒越多,挽回越多」与「召回越早越好」。本书证伪这两句话。离开收费页的访客里,有一小部分本来就会自己回来,他们集中在头一个小时;一分钟时送达的消息落在这些人身上,又只复述屏幕上还开着的内容。按本书的示意模型,第一波召回放在一分钟时只挽回同意离开者的 0.00609,放在十五分钟时挽回 0.03732;把三波压在一小时之内,每位离开者净值 RM 1.1790,按十五分钟、二十四小时、七十二小时发送时为 RM 1.9419;第五波的净值已是 -0.0121。九成五的漏口很宽,三波召回只挽回其中 0.0237;真正的大杠杆是首单之后的升单。每天 RM 5,000 是倒推需求的目标,不是承诺:只卖入门报告需要每天 5,290 位访客,一成买家把首单抵扣进买断卷后需要 2,784 位,再加三波召回需要 1,920 位。 核对原文之后,本书做了几处蓝图修正:艾宾浩斯的遗忘数据只说明衰减先快后慢,不能推出发送时刻;十五分钟、二十四小时、七十二小时只达到最优排法的 0.807 到 0.965,第三波放在四十八小时(RM 1.9934)胜过七十二小时;倒 U 形剂量来自康等人在把握程度上的实验,不是罗文斯坦理论的直接推论,示意模型的净值峰值在剂量 0.36;谷歌公布的 0.10 % 与 0.30 % 是一家邮箱服务商的线,本书另设 0.08 % 的内部刹车,零投诉要累计 2,995 封才说明低于 0.10 %;静态模型里的升单阶梯几乎不增加收入。 第一章写现金方程与每日波动,第二章写意向衰减、自返生存曲线与三波召回的边际挽回,第三章写只用访客自己免费结果的悬念剂量与真话过滤器,第四章写停发规则、投诉刹车与马来西亚、新加坡、美国三地的同意法规,第五章写首单 RM 19.90 抵扣进 RM 199 的期望值、退款风险与永不重置的窗口,第六章写序贯检验、多臂老虎机与错误发现率,第七章考辨《韩非子·说难》与亚里士多德《修辞学》并给出四件套蒸馏规格,第八章给出可接入现网的「召回排程与升单触发器」与预注册证伪方案。书中网站被当作付费报告产品讨论,不宣称任何预测准确性,不用恐惧招揽;所有访客与订单数据均为合成或示意。书中以标注框区分已发表的证据、可运行的 Python 计算(25 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The 95 Per Cent Leak: The Cash Arithmetic of an RM 5,000 Daily Target — Part B is about the noise in measurement itself. If a conversion of 0.05 is measured on n visitors, the width of the Wilson 95 % interval shrinks with n: with 200 visitors the interval runs from 0.0274 to 0.0896, a width of 0.0622, which says almost nothing; with 1,000 visitors from 0.0381 to 0.0653, width 0.0272; with 5,000 from 0.0443 to 0.0564, width 0.0121; with 20,000 from 0.0471 to 0.0531, width 0.0060. If a week of data holds only one or two thousand visitors, a difference of one or two tenths of a point in conversion may very well be noise, and no change can be declared effective on that basis. (https://5dlibrary.com/answers/en/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c01/) 第1章 九成五的漏口:每天 RM 5,000 目标的现金算术 — B 段讨论测量本身的噪声。若在 n 位访客上测得转化率 0.05,威尔逊 95 % 区间的宽度随 n 缩小:200 位访客时区间从 0.0274 到 0.0896,宽 0.0622,几乎没有信息;1,000 位时从 0.0381 到 0.0653,宽 0.0272;5,000 位时从 0.0443 到 0.0564,宽 0.0121;20,000 位时从 0.0471 到 0.0531,宽 0.0060。一周的数据若只有一两千位访客,转化率差一两个千分点完全可能只是噪声,不能据此宣布某个改动有效。 (https://5dlibrary.com/answers/zh/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c01/) - Ch.2 The Decay of Intent: Three Recall Waves at Fifteen Minutes, Twenty-Four Hours and Seventy-Two Hours — Part B moves only the first wave. At one minute the first wave wins back only 0.00609, worth +0.0677 net, and the three waves together 0.7938. At five minutes it wins back 0.02284 and the total is 1.4113; at fifteen minutes 0.03732 and 1.9419; at twenty-five minutes 0.03869, worth +1.2823 net, with a total of 1.9941, the highest in the table; at one hour it falls back to 0.03301 with a total of 1.7972; at six hours it is only 0.01412 with a total of 1.1494. The curve peaks between fifteen and twenty-five minutes and falls on both sides: too early, and novelty is lacking, self-returners are many and unsubscriptions high; too late, and the intent of impulse visitors has dispersed. (https://5dlibrary.com/answers/en/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c02/) 第2章 意向衰减:十五分钟、二十四小时与七十二小时三波召回 — B 段只移动第一波。放在一分钟时,第一波只挽回 0.00609,净值 +0.0677,三波合计 0.7938;五分钟时挽回 0.02284,合计 1.4113;十五分钟时 0.03732,合计 1.9419;二十五分钟时 0.03869,净值 +1.2823,合计 1.9941,是表中最高;一小时时回落到 0.03301,合计 1.7972;六小时时只有 0.01412,合计 1.1494。曲线在十五到二十五分钟之间达到顶点,两侧都下降:太早,新意不足、自返者多、退订高;太晚,冲动型访客的意向已经散了。 (https://5dlibrary.com/answers/zh/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c02/) - Ch.3 The Dose of the Information Gap: Curiosity Recalls Written Only from the Visitor's Own Free Result — With $\gamma$ at 1.5, the pull is greatest at $g^{*}=1/(1+\gamma)=0.400$. In the table, a dose of 0 gives only the base click rate of 0.0200 and a net of +0.0080; a dose of 0.2 raises clicks to 0.0893; a dose of 0.4 reaches the top at 0.1100, with complaints at 0.00059 and a net of +0.1083; at a dose of 0.6 clicks fall back to 0.0935 while complaints rise to 0.00105; at a dose of 0.8 clicks are down to 0.0546, complaints are at 0.00194 and the net turns negative at -0.0119; at a dose of 1.0 clicks are back at 0.0200, complaints at 0.00340 and the net at -0.1120. (https://5dlibrary.com/answers/en/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c03/) 第3章 信息缺口的剂量:只用访客自己的免费结果写悬念召回 — $\gamma$ 取 1.5 时,拉力在 $g^{*}=1/(1+\gamma)=0.400$ 处最大。表中,剂量 0 时点击率只有基础值 0.0200,净值 +0.0080;剂量 0.2 时点击率升到 0.0893;剂量 0.4 时达到顶点 0.1100,投诉 0.00059,净值 +0.1083;剂量 0.6 时点击率回落到 0.0935,投诉升到 0.00105;剂量 0.8 时点击率只剩 0.0546,投诉 0.00194,净值转负为 -0.0119;剂量 1.0 时点击率又回到 0.0200,投诉 0.00340,净值 -0.1120。 (https://5dlibrary.com/answers/zh/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c03/) - Ch.4 The Red Lines of Unsubscription and Complaint: A Stopping Rule, a Complaint Brake and the Consent Law of Three Jurisdictions — With a true rate of 0.0006, 1,000 messages a week touch the line in 0.451 of weeks; at 7,000 the chance of touching the brake is 0.247 and of touching the 0.10 % line 0.132; at 35,000 the chances are 0.082 and 0.003. With a true rate of 0.0012 the list itself is already over the 0.10 % line: at 7,000 messages a week, 0.843 of weeks touch the brake and 0.733 touch the 0.10 % line; at 35,000 the figures are 0.991 and 0.879. Here the brake is not giving a false alarm; it is reporting the truth. (https://5dlibrary.com/answers/en/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c04/) 第4章 退订与投诉的红线:停发规则、投诉刹车与三地同意法规 — 真实投诉率为 0.0006 时,每周 1,000 封有 0.451 的星期碰线,7,000 封时碰刹车线的机会是 0.247,碰 0.10 % 线的机会 0.132;35,000 封时分别是 0.082 与 0.003。真实投诉率为 0.0012 时,名单本身已经越过了 0.10 % 的线:每周 7,000 封时,0.843 的星期会碰刹车线,0.733 的星期碰 0.10 % 的线;35,000 封时分别是 0.991 与 0.879。这时刹车不是误报,而是如实报告。 (https://5dlibrary.com/answers/zh/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c04/) - Ch.5 The Upgrade Ratchet: Expected Value and Refund Risk of Crediting a First RM 19.90 Toward RM 199 — With λ at 0.5 the picture changes completely. With the middle rung at RM 49, the ladder's revenue rises to RM 36.793, +13.501 above the single step, and 0.090 of first-order buyers read the middle rung and then upgrade to the top. With the middle rung at RM 59 the ladder adds +11.355, and 0.079 climb the whole way. With the middle rung at RM 79 the ladder adds +11.796. With λ at 1.0, the RM 49 rung adds +27.399 and 0.182 climb the whole way. (https://5dlibrary.com/answers/en/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c05/) 第5章 升单棘轮:首单 RM 19.90 抵扣进 RM 199 的期望值与退款风险 — λ 为 0.5 时,情况完全不同。中间档定在 RM 49,阶梯收入升到 RM 36.793,比单级多 +13.501,首单买家中有 0.090 读完中间档后升到顶级。中间档定在 RM 59,阶梯多 +11.355,走完全程者占 0.079。中间档定在 RM 79,阶梯多 +11.796。λ 为 1.0 时,RM 49 一档多 +27.399,走完全程者占 0.182。 (https://5dlibrary.com/answers/zh/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c05/) - Ch.6 Testing Message Variants: Sequential Tests, Multi-Armed Bandits and the False Discovery Rate — The upper boundary is a log-likelihood ratio of 2.7726 and the lower one -1.5581. A one-sided fixed-sample test with the same error rates needs 1,985 messages. When the true recovery rate is 0.030, the sequential test stops after only 982 messages on average, with a median of 697 and a ninetieth percentile of 2,078, and wrongly declares an effect in 0.045 of runs, below 0.05. When the true rate is 0.040, it needs 1,262 messages on average, with a median of 1,008, and correctly declares an effect in 0.795 of runs, close to the 0.80 that was set. (https://5dlibrary.com/answers/en/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c06/) 第6章 检验消息变体:序贯检验、多臂老虎机与错误发现率 — 上界是对数似然比 2.7726,下界是 -1.5581。达到同样错误率的单侧固定样本检验,需要 1,985 封消息。序贯检验在真实挽回率为 0.030 时,平均只用 982 封就停下,中位数 697,第九十百分位 2,078,错误地判为有效的比例 0.045,低于 0.05。真实挽回率为 0.040 时,平均用 1,262 封,中位数 1,008,正确判为有效的比例 0.795,接近设定的 0.80。 (https://5dlibrary.com/answers/zh/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c06/) - Ch.7 A Study of Classics East and West: The Han Feizi on the Difficulties of Persuasion and Aristotle's Rhetoric — The parameters are c of 40 seconds and $T_0$ of 20 seconds. With a gap of 0.1, understanding takes only 0.4 seconds and the probability of understanding is 0.980, but the value is only 0.0980, because there was little to learn; with a gap of 0.3 the value is 0.2506; with a gap of 0.5 it reaches its peak of 0.3033, with 10.0 seconds needed and a probability of understanding of 0.607; with a gap of 0.7 understanding takes 19.6 seconds, its probability falls to 0.375 and the value is 0.2627; with a gap of 1.0 understanding takes 40.0 seconds and only 0.135 of readers hold on long enough to get there. (https://5dlibrary.com/answers/en/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c07/) 第7章 东西方原典考辨:《韩非子·说难》与亚里士多德《修辞学》 — 参数取 c 为 40 秒、$T_0$ 为 20 秒。缺口 0.1 时,理解只需 0.4 秒,理解概率 0.980,但价值只有 0.0980,因为没学到什么;缺口 0.3 时价值 0.2506;缺口 0.5 时达到顶点 0.3033,理解需要 10.0 秒,理解概率 0.607;缺口 0.7 时理解需要 19.6 秒,理解概率降到 0.375,价值 0.2627;缺口 1.0 时理解需要 40.0 秒,只有 0.135 的读者能撑到明白。 (https://5dlibrary.com/answers/zh/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c07/) - Ch.8 The Recall Scheduler and Upsell Trigger: Integration Spec, Example Runs and a Pre-Registered Refutation Plan — The two quantiles are 1.9600 and 0.8416. H1 tests whether the second and third waves add anything: the group that receives only the first wave is expected to win back 0.03732 and the group that receives all three 0.05917, and each group needs 1,509 consented leavers. H2 tests the timing of the first wave: 0.00609 for the one-minute group against 0.03732 for the fifteen-minute group, a large gap that needs only 341 per group. H3 tests the credit on the first order: an acceptance of 0.1069 without the credit against 0.1300 with it, which needs 3,071 first-order buyers per group. H4 is the five-arm dose experiment of Chapter 3, with 8,000 messages an arm and 40,000 in all. H5 tests the complaint rate: a clean record of 2,995 messages is needed to show a rate below 0.10 %. (https://5dlibrary.com/answers/en/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c08/) 第8章 召回排程与升单触发器:接入规格、示意运行与预注册证伪方案 — 两个分位数是 1.9600 与 0.8416。H1 检验第二、第三波是否带来增量:只发第一波的组,期望挽回 0.03732;发三波的组,期望挽回 0.05917;每组需要 1,509 位同意离开者。H2 检验第一波的时刻:一分钟组 0.00609,十五分钟组 0.03732,差距很大,每组只需 341 人。H3 检验首单抵扣:不抵扣的接受率 0.1069,抵扣的 0.1300,每组需要 3,071 位首单买家。H4 是第三章的五臂剂量实验,每臂 8,000 封,合计 40,000 封。H5 检验投诉率:零投诉要累计 2,995 封,才能说明投诉率低于 0.10 %。 (https://5dlibrary.com/answers/zh/paywall-abandonment-recall-timing-curiosity-gap-dosing-upsell-ratchet/c08/) ## 5DL-B-350 · Faceless Case-Screen Short-Video Production, Three-Second Hook Tournaments and Direct-to-Paywall Traffic Funnels 中文书名:零露脸案例录屏短视频量产、前三秒痛点钩子赛马与私域收费站导流母典 DOI: pending Cite: Fan, Y. (2026). Faceless Case-Screen Short-Video Production, Three-Second Hook Tournaments and Direct-to-Paywall Traffic Funnels (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: In a synthetic year of each strategy, steady case videos matched to their audience bring 16.51 times as many payers a day as betting on viral hits. It is a target, not a promise: at an illustrative baseline it needs 280,112 plays a day. Abstract: The sentence short-video marketers hear most often is: go viral once and the money follows. This book refutes it. Between one play and one payment lie five gates, completion, profile visit, bio-link tap, bridge survival and payment, and revenue is the product of plays and five ratios. In a synthetic year of each strategy, steady case videos matched to their audience bring 16.51 times as many payers a day as betting on viral hits. The book's subject is a live chart-diagnosis tollbooth whose traffic comes entirely from faceless case-screen videos, with a target of RM 5,000 of cash a day and no advertising spend. It is a target, not a promise: at an illustrative baseline it needs 280,112 plays a day. Checking the sources corrected the blueprint: "the first three seconds" is no platform's rule and distribution staircases are illustrative; the much-quoted research on first seconds studied video advertisements; page-load evidence is peer-reviewed only. Chapter 1 breaks the target into five ratios and shows how noisy a single day is. Chapter 2 gives the contrast hook and five-beat script; contrasts concern only rules of calculation, cases come only from public records, synthetic charts or consent, and no predictive accuracy is claimed. Chapter 3 reads retention curves with the product-limit estimator: departures in the first three seconds are 0.587 of all departures, and handling censoring corrects a naive completion of 0.299 to 0.318. Chapter 4 runs hooks as a batched bandit tournament; Thompson sampling cuts the plays lost from 6,883 to 952; with twelve candidates the naive rule produces at least one false winner with probability 0.252, false-discovery-rate control brings the share of false winners among winners to 0.066, peeking pushes the false-positive rate to 0.270 and a sequential test brings it back to 0.010; the minimum sample is 2,311 plays per arm. Chapter 5 takes the bridge apart: only 0.405 of those who tap reach the paywall through a two-hop aggregator, against 0.857 on a no-hop report page; the first screen is budgeted at two seconds and about 366 KB; cloaking, hidden redirects and missing consent or unsubscribe are blocked outright. Chapter 6 schedules six genuine accounts: three operators make 13.50 videos a day, and over sixty posts a single format averages a completion of only 0.245 against 0.375 for four rotating formats. Chapter 7 reads Li Yu, Liu Xie, Aristotle and Horace on openings and gives the distillation spec for the four kits. Chapter 8 gives the distilled Hook Tournament Scorer, Bridge Leakage Audit and Daily Schedule and a pre-registered refutation plan, one test of which needs 85 videos. All parameters are illustrative and all calculations come from the book's 22 standard-library programs; the book promises no income and teaches no fake engagement, cloaking or manipulation. 摘要:做短视频导流的人最常听到的一句话是:先爆一条,钱自己会来。本书证伪这句话。从一次播放到一笔付款要经过完播、点进主页、点开简介链接、桥页存活与付款五道门,收入是播放与五个比例的乘积;爆款带来的宽泛观众看得多、付得少。两种打法各跑一年的合成实验里,对口的稳定案例录屏日均付款人数是赌爆款的 16.51 倍。本书的对象是一个已上线的命盘诊断收费站,流量全部来自不露脸的案例录屏,目标是零广告费每天现金 RM 5,000。这是目标,不是承诺:在示意基线下,它需要每天 280,112 次播放。核对出处后本书做了几处蓝图修正:「前三秒」不是任何平台公布的规则,平台分发阶梯只作示意;常被引用的前几秒研究对象是网络视频广告;网页加载只引用同行评审研究,约两秒的可容忍等待只作保守预算。 第一章把目标拆成五个比例并说明单日数据有多吵。第二章给出反差钩子与五拍脚本,反差只落在计算规则上,案例只来自公众人物公开出生记录、合成命盘或书面同意,不宣称预测准确率。第三章用乘积极限估计读留存曲线,前三秒的离开占全部离开的 0.587,截尾把朴素完播率 0.299 修正为 0.318。第四章把钩子当作分批老虎机来赛马,汤普森抽样把损失的播放从 6,883 次降到 952 次;十二个候选同台时朴素规则每场至少一个假赢家的概率为 0.252,错误发现率控制把赢家中的假赢家比例压到 0.066;偷看把假阳性率推到 0.270,序贯检验压回 0.010;每臂最低样本 2,311 次。 第五章逐步拆解桥页漏损:两跳聚合页只有 0.405 的点开者看到付费区,零跳报告页为 0.857;首屏预算两秒、约 366 KB;伪装、隐藏跳转、缺少同意与退订一律拦回。第六章排定六个真实账号的日程:三位操作员每天产能 13.50 条,单一格式六十条的平均完播率只有 0.245,四种格式轮换为 0.375。第七章考辨李渔《闲情偶寄》论家门与冲场、刘勰《文心雕龙》论首尾、亚里士多德《诗学》与贺拉斯《诗艺》论开端,并写出四件套蒸馏规格。第八章给出蒸馏小工具「钩子赛马计分、桥页漏损审计与日排程器」与预注册证伪方案,其中一项检验需要 85 条视频。全部参数为示意,计算由书中 22 段只用标准库的程序生成;本书不承诺任何收入,不教任何刷量、伪装或操纵手法。 Chapter answers: - Ch.1 From Plays to RM 5,000: the Multiplication of the Funnel — This chapter has broken RM 5,000 into a single multiplication: about two hundred and eighty thousand plays a day pass through five gates and leave one hundred payers. Each gate has an elasticity of one, so the choice of which gate to work on depends only on its cost and its headroom, and improving all five by a fifth more than halves the plays needed. With long-tailed plays the daily count of payers is very noisy, so a change must be judged on the pass rate of its own gate. (https://5dlibrary.com/answers/en/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c01/) 第1章 从播放到 RM 5,000:转化漏斗的乘法 — 本章把 RM 5,000 拆成一条乘法:每天约二十八万次播放,经过五道门,剩下一百位付款人。每道门的弹性都是一,所以挑哪一道来改,只看成本与剩余空间;五道门各改善两成,播放需求就能减半以上。单日付款人数在长尾播放下噪声很大,判断改动要看每道门自己的通过率。 (https://5dlibrary.com/answers/zh/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c01/) - Ch.2 The Script Formula for Case-Screen Videos: the Contrast Hook and the Gate before the Screen — The gate has three rules. The first concerns the source. It accepts only the public birth records of public figures, synthetic charts, and charts whose owners have given written consent to public teaching. C03 is a private chart that a customer paid to have computed, and it is rejected outright, whether or not it is blurred or renamed. The second rule concerns precision. Any conclusion that depends on the hour, such as the rising axis or the hour pillar, requires a birth time exact to the minute. C04 has an unknown birth time and C08 falls on the boundary between two hours, so neither may discuss an hour axis; C09 is synthetic but has the required precision and may. The third rule concerns wording. C06 says "accurate", a claim of predictive accuracy. C07 uses a "bad chart" as its hook, a fear tactic. C10 states an inference as fact. All three are rejected. (https://5dlibrary.com/answers/en/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c02/) 第2章 案例录屏的脚本公式:反差钩子与上屏闸门 — 闸门有三条规则。第一条管来源:只接受公众人物的公开出生记录、合成命盘、以及本人书面同意公开讲解的命盘;C03 是客户付费生成的私人命盘,直接拒绝,无论打不打码、改不改名。第二条管精度:凡是依赖时辰的结论,比如上升轴线或时柱,出生时间必须精确到分钟;C04 出生时间未知、C08 落在时辰交界,都不能讲时辰轴线,C09 虽是合成命盘,精度达标,可以讲。第三条管说法:C06 写了「准确」,属于预测准确率宣称;C07 用「凶盘」做钩子,属于恐惧话术;C10 把推断说成事实,三者一律拒绝。 (https://5dlibrary.com/answers/zh/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c02/) - Ch.3 A Retention Curve Is a Survival Curve: the Hazard of the First Three Seconds and the Completion Threshold — Chapter 3 reads retention curves with the product-limit estimator: departures in the first three seconds are 0.587 of all departures, and handling censoring corrects a naive completion of 0.299 to 0.318. (https://5dlibrary.com/answers/en/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c03/) 第3章 留存曲线是生存曲线:前三秒风险与完播门槛 — 第三章用乘积极限估计读留存曲线,前三秒的离开占全部离开的 0.587,截尾把朴素完播率 0.299 修正为 0.318。 (https://5dlibrary.com/answers/zh/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c03/) - Ch.4 Hook Tournaments: Batched Bandits, Multiple Comparisons and Honest Stopping — Chapter 4 runs hooks as a batched bandit tournament; Thompson sampling cuts the plays lost from 6,883 to 952; with twelve candidates the naive rule produces at least one false winner with probability 0.252, false-discovery-rate control brings the share of false winners among winners to 0.066, peeking pushes the false-positive rate to 0.270 and a sequential test brings it back to 0.010; the minimum sample is 2,311 plays per arm. (https://5dlibrary.com/answers/en/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c04/) 第4章 钩子赛马:分批老虎机、多重比较与诚实停止 — 第四章把钩子当作分批老虎机来赛马,汤普森抽样把损失的播放从 6,883 次降到 952 次;十二个候选同台时朴素规则每场至少一个假赢家的概率为 0.252,错误发现率控制把赢家中的假赢家比例压到 0.066;偷看把假阳性率推到 0.270,序贯检验压回 0.010;每臂最低样本 2,311 次。 (https://5dlibrary.com/answers/zh/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c04/) - Ch.5 From Profile to Paywall: Step-by-Step Leakage on the Bridge Page and the Zero-Loss Jump — Chapter 5 takes the bridge apart: only 0.405 of those who tap reach the paywall through a two-hop aggregator, against 0.857 on a no-hop report page; the first screen is budgeted at two seconds and about 366 KB; cloaking, hidden redirects and missing consent or unsubscribe are blocked outright. (https://5dlibrary.com/answers/en/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c05/) 第5章 从主页到付费区:桥页的逐步漏损与零损跳转 — 第五章逐步拆解桥页漏损:两跳聚合页只有 0.405 的点开者看到付费区,零跳报告页为 0.857;首屏预算两秒、约 366 KB;伪装、隐藏跳转、缺少同意与退订一律拦回。 (https://5dlibrary.com/answers/zh/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c05/) - Ch.6 A Small, Honest Account Matrix: Capacity, Format Fatigue and a Standard Schedule — Chapter 6 schedules six genuine accounts: three operators make 13.50 videos a day, and over sixty posts a single format averages a completion of only 0.245 against 0.375 for four rotating formats. (https://5dlibrary.com/answers/en/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c06/) 第6章 小型诚实账号矩阵:产能、格式疲劳与标准日程 — 第六章排定六个真实账号的日程:三位操作员每天产能 13.50 条,单一格式六十条的平均完播率只有 0.245,四种格式轮换为 0.375。 (https://5dlibrary.com/answers/zh/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c06/) - Ch.7 Classical Sources East and West: Li Yu on the Opening and the Entrance, Liu Xie on Beginnings and Ends, Aristotle and Horace on Where to Begin — That the lead should not come on too late has an exact price in Part A. If the chart appears on screen at the first second, 0.835 of plays are still there, and 8,353 of every ten thousand plays see it. At the third second, 0.583 remain, or 5,827. At the fifth, 0.538; at the eighth, 0.477, or 4,771; at the twelfth, 0.427; at the twentieth only 0.379, or 3,791. In a faceless recording the chart is the named role. Li Yu warns that the audience will "take the guest for the host" if someone else comes on first; in a screen recording, if an animated intro or an unrelated picture appears first, that is what viewers remember, not the chart. (https://5dlibrary.com/answers/en/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c07/) 第7章 东西方原典考辨:李渔论家门与冲场,刘勰论首尾,亚里士多德与贺拉斯论开端 — 主角不宜出得太迟,这一条在 A 段有精确的价格。盘在第一秒上屏,0.835 的播放还在,每一万次播放有 8,353 次看到它;第三秒上屏,剩 0.583,即 5,827 次;第五秒 0.538;第八秒 0.477,即 4,771 次;第十二秒 0.427;第二十秒只剩 0.379,即 3,791 次。不露脸的录屏里,盘就是有名脚色。李渔说观众会把先上场的人「反認為主」,录屏里先上屏的若是一段片头动画或一张无关图片,观众记住的就是它,而不是那张盘。 (https://5dlibrary.com/answers/zh/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c07/) - Ch.8 Closing the Loop: a Daily Conversion Check-up, the Distilled Tool and a Pre-registered Refutation Plan — Chapter 8 gives the distilled Hook Tournament Scorer, Bridge Leakage Audit and Daily Schedule and a pre-registered refutation plan, one test of which needs 85 videos. All parameters are illustrative and all calculations come from the book's 22 standard-library programs; the book promises no income and teaches no fake engagement, cloaking or manipulation. (https://5dlibrary.com/answers/en/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c08/) 第8章 实证闭环:每日转化体检、蒸馏小工具与预注册证伪 — 第八章给出蒸馏小工具「钩子赛马计分、桥页漏损审计与日排程器」与预注册证伪方案,其中一项检验需要 85 条视频。全部参数为示意,计算由书中 22 段只用标准库的程序生成;本书不承诺任何收入,不教任何刷量、伪装或操纵手法。 (https://5dlibrary.com/answers/zh/faceless-case-screen-shorts-three-second-hook-tournaments-paywall-funnel/c08/) ## 5DL-B-351 · Southeast Asian Social-Commerce Whitelist Amplification, Cross-Channel Attribution and Dual-Track Commission Growth 中文书名:东南亚社交电商暗投带货联动、站内外归因叠加与双轨抽佣增长母典 DOI: pending Cite: Fan, Y. (2026). Southeast Asian Social-Commerce Whitelist Amplification, Cross-Channel Attribution and Dual-Track Commission Growth (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 1 takes apart an illustrative order priced at RM 59.90: the goods margin is 0.6837, but after platform fees, returns and refused cash-on-delivery parcels the contribution rate is 0.3467. The book is not financial, legal or tax advice and promises no income or return. Abstract: Brands and agencies selling through TikTok Shop and Shopee in Malaysia and Southeast Asia often end their weekly review the same way: the dashboards for in-app ads, creator affiliates and off-site ads all clear the line, so every budget goes up. This book refutes the popular claim behind that habit: add up the ROAS each platform reports and you have the true return. The same order is claimed by several dashboards, and the claimed orders include many purchases that would have happened anyway. In a synthetic week where every truth is known, the orders claimed by three dashboards add up to 1.112 times the orders placed, while the ads truly caused only 0.312 of them; the summed dashboard ROAS is 4.81 and the true incremental ROAS only 1.35, below the break-even line of 2.884. Checking the sources corrected the blueprint: spillover evidence points both ways, the Shapley value divides credit but does not set budgets, user-level holdouts give way to geographic experiments across Malaysia's sixteen regions, fee rates are illustrative inputs, and the 2024 electronic-trade regulations replace those of 2012. All calculations come from the book's programs. Chapter 1 takes apart an illustrative order priced at RM 59.90: the goods margin is 0.6837, but after platform fees, returns and refused cash-on-delivery parcels the contribution rate is 0.3467. Chapter 2 shows that break-even ROAS is the reciprocal of the contribution rate, so that the same order puts the line anywhere between 1.463 and 3.882, and finds where a flat licensing fee beats commission on whitelisted orders. Chapter 3 models whitelisting spillover, mega-sale bias and marginal return. Chapter 4 measures double counting. Chapter 5 covers holdout sizes, geographic experiments and the Shapley split. Chapter 6 designs the dual-track contract between agency and brand: a fixed monthly fee plus a commission in marginal tiers on incremental net GMV measured by experiment, with clawback of returns and a holdback of about 0.13; the optimal commission share of a linear contract falls with measurement noise, 0.357 on dashboards and 0.762 with a geographic experiment. In this book the dual-track commission is a single-tier business-to-business service fee and involves no multi-level commission of any kind. RM 5,000 a day of agency fee income is treated as a planning target, never a promise. Chapter 7 reads Bai Gui and "greedy merchants make three, fair merchants five" in the Shiji, the medieval Mediterranean commenda, and Adam Smith on joint-stock companies and piece rates. Chapter 8 specifies a Break-Even ROAS, Attribution Dedup and Dual-Track Contract Calculator for the live sites and a pre-registered refutation plan. Marked boxes separate published evidence, runnable Python calculations (24 programs), classical sources and the author's hypotheses. The book is not financial, legal or tax advice and promises no income or return. 摘要:在马来西亚与东南亚做 TikTok Shop、Shopee 带货的品牌与代理商,每周复盘最常见的结论是:站内广告、达人联盟与站外广告三个后台的 ROAS 都过线,预算全部加码。本书证伪这背后的流行说法:把各平台后台报告的 ROAS 加起来,就是真实回报。同一张订单会被几个后台同时认领,认领的订单里又有大量本来就会发生的购买。在一个所有真相都已知的合成周里,三个后台认领的订单合计是实际订单的 1.112 倍,真正由广告带来的增量只占实际订单的 0.312;后台 ROAS 之和是 4.81,真实增量 ROAS 只有 1.35,低于盈亏平衡线 2.884。核对原文之后,本书做了蓝图修正:付费向自然流量的外溢在随机实验里有正有负,本书把它写成可正可负的净效应;沙普利值适合分功劳,不适合定预算;用户级留出组对多数东南亚品牌太大,重点改为全马十六个地区的地区实验;平台费率无法逐项核实,全部作为示意输入;马来西亚电子交易条例已由 2024 年新条例取代。书中参数均为示意,全部计算由书中程序生成。 第一章拆开一张标价 RM 59.90 的示意订单,货品毛利 0.6837,扣掉平台费用、退货与货到付款拒收以后,贡献率只剩 0.3467。第二章由此推出盈亏平衡 ROAS 等于贡献率的倒数,同一张订单按四种口径落在 1.463 到 3.882 之间,并比较暗投同时计佣与改付固定授权费的打平点。第三章讨论暗投:截流与外溢的日度模型、大促造成的前后对比偏差与饱和曲线上的边际回报。第四章量化重复计数并检验本书要推翻的那句话。第五章讲留出组样本量、地区实验的功效与精确随机化检验,以及沙普利分配。 第六章设计代理商与品牌之间的双轨合同:固定月费加按实验测得的增量净成交额边际分档的抽佣,并设退货追回与约 0.13 的留存比例;线性合同的最优抽佣比例随测量噪声下降,只看后台时是 0.357,地区实验时是 0.762。双轨抽佣在本书中专指单层企业间服务费,不涉及任何多层计佣。代理业务每天 RM 5,000 的服务费收入只作为推算条件的目标,不是承诺。第七章考辨《史记·货殖列传》的白圭与「贪贾三之,廉贾五之」、地中海中世纪的康曼达契约与亚当·斯密论合股公司及计件报酬。第八章给出可接入网站的「盈亏平衡 ROAS、归因去重与双轨合同试算器」与预注册证伪方案。全书以标注框区分已发表的证据、可运行的 Python 计算(24 段程序)、原典与作者的假说;本书不构成财务、法律或税务意见,不承诺任何收入或回报。 Chapter answers: - Ch.1 The Account of One Order: Platform Fees, Returns and Cash on Delivery — Chapter 1 takes apart an illustrative order priced at RM 59.90: the goods margin is 0.6837, but after platform fees, returns and refused cash-on-delivery parcels the contribution rate is 0.3467. (https://5dlibrary.com/answers/en/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c01/) 第1章 一张订单的账:平台费用、退货与货到付款 — 第一章拆开一张标价 RM 59.90 的示意订单,货品毛利 0.6837,扣掉平台费用、退货与货到付款拒收以后,贡献率只剩 0.3467。 (https://5dlibrary.com/answers/zh/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c01/) - Ch.2 The Break-Even Point of Advertising: Contribution Rate, Sensitivity and Four Routes — Chapter 2 shows that break-even ROAS is the reciprocal of the contribution rate, so that the same order puts the line anywhere between 1.463 and 3.882, and finds where a flat licensing fee beats commission on whitelisted orders. (https://5dlibrary.com/answers/en/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c02/) 第2章 广告盈亏平衡点:贡献率、敏感度与四种来路 — 第二章由此推出盈亏平衡 ROAS 等于贡献率的倒数,同一张订单按四种口径落在 1.463 到 3.882 之间,并比较暗投同时计佣与改付固定授权费的打平点。 (https://5dlibrary.com/answers/zh/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c02/) - Ch.3 Whitelist Amplification: Paid-to-Organic Spillover, Mega-Sale Confounding and Marginal Return — The net spillover $\rho$ of whitelisting is set by two forces, interception and spillover, and can be positive or negative. In the illustrative model, interception of three tenths with no spillover leaves a true return of only 2.240, below the line; moderate spillover brings it back to 3.508; small interception with strong spillover lifts it to 5.416. Measuring spillover during a mega-sale by before and after reads +0.037 in a world with no spillover and +0.340 in a world whose true $\rho$ is +0.102; difference-in-differences reads -0.201 and +0.104, close to the truth. On a saturating curve the marginal ROAS may have fallen through the line long before the average ROAS does; the best spend is RM 1,273 a day, while adding budget by average ROAS goes on all the way to RM 2,806. (https://5dlibrary.com/answers/en/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c03/) 第3章 暗投放大:付费向自然流量的外溢、大促混淆与边际回报 — 暗投的净外溢 $\rho$ 由截流与外溢两股力量决定,可正可负。示意模型里,截流三成且无外溢时真实回报只有 2.240,低于盈亏线;中等外溢时回到 3.508;截流小而外溢强时高达 5.416。大促期间用前后对比测外溢,在没有外溢的世界里读出 +0.037,在真实 $\rho$ 为 +0.102 的世界里读出 +0.340;双重差分分别读出 -0.201 与 +0.104,接近真值。饱和曲线上,平均 ROAS 过线时边际 ROAS 可能早已跌破;最佳投放额是每天 RM 1,273,而按平均 ROAS 加预算会一路加到 RM 2,806。 (https://5dlibrary.com/answers/zh/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c03/) - Ch.4 Attribution Stacking: Double Counting On and Off the Platform, Deduplication Rules and the Sum of ROAS — In the synthetic week, the orders claimed by three dashboards add up to 1.112 times the orders actually placed, while the true increment is only 0.312 of the orders placed. Deduplication brings the total down to 8,556 and removes the double counting, yet it is still more than twice the true increment of 3,549, and last touch credits the off-site ads with 3.16 times their true increment. In money, the summed dashboard ROAS is 4.81, the blended ROAS 4.33 and the true incremental ROAS only 1.35, below the break-even line of 2.884. The dashboards say the week made RM 100,176; the experiment says it lost RM 79,835. The book's central proposition has been tested in this world: the sum of the dashboards is not the true return. (https://5dlibrary.com/answers/en/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c04/) 第4章 归因叠加:站内外重复计数、去重规则与 ROAS 之和 — 合成周里,三个后台认领的订单合计是实际订单的 1.112 倍,真实增量只占实际订单的 0.312。去重把总数降到 8,556 张,消除了重复计数,却仍高于真实增量 3,549 张一倍多,末次触点给站外广告的功劳是其真实增量的 3.16 倍。换成钱,后台 ROAS 之和是 4.81,混合 ROAS 4.33,真实增量 ROAS 只有 1.35,低于盈亏线 2.884;后台说这一周赚 RM 100,176,实验说亏 RM 79,835。本书的核心命题在这个世界里被检验:后台之和不是真实回报。 (https://5dlibrary.com/answers/zh/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c04/) - Ch.5 Incrementality Experiments: Holdouts, Geographic Tests and the Shapley Split — The number of people a holdout needs grows with the inverse square of ad spend per user. At a coefficient of variation of 10 and RM 0.02 per user, merely telling whether incremental ROAS reaches the break-even line needs 1,061,621 people in each group. A ghost-ads design removes the factor of 11.1 that exposure dilution brings, but needs platform support. A geographic experiment across Malaysia's sixteen regions has a power of only 0.734 at a budget share of 0.03, rising to 0.999 when the share doubles to 0.06; one experiment estimates 2.390 with an exact randomisation p-value of 0.0005. The Shapley values of a factorial experiment divide an increment of RM 288,000 into 124,500, 102,000 and 61,500, exactly the whole; but A's return with the other channels running is 3.600, above its Shapley 3.112, so dividing money and spending it need two different rods. (https://5dlibrary.com/answers/en/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c05/) 第5章 增量实验:留出组、地区实验与沙普利分配 — 留出组所需人数随人均广告费的平方反比增长:变异系数 10、人均 RM 0.02 时,仅分辨增量 ROAS 是否达到盈亏线,每组就要 1,061,621 人。幽灵广告式的设计能省掉曝光稀释带来的 11.1 倍,但需要平台支持。全马十六个地区的地区实验在预算比例 0.03 时功效只有 0.734,比例加倍到 0.06 时升到 0.999;一次实验估计出 2.390,精确随机化检验的 p 值是 0.0005。析因实验的沙普利值把 RM 288,000 的增量分成 124,500、102,000 与 61,500,合计不多不少;但 A 在其余渠道都开着时的回报是 3.600,高于沙普利值的 3.112,分钱与花钱要用两把尺子。 (https://5dlibrary.com/answers/zh/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c05/) - Ch.6 The Dual-Track Contract: Fixed Licensing Fee, Tiered Commission and the Arithmetic of Principal and Agent — Chapter 6 designs the dual-track contract between agency and brand: a fixed monthly fee plus a commission in marginal tiers on incremental net GMV measured by experiment, with clawback of returns and a holdback of about 0.13; the optimal commission share of a linear contract falls with measurement noise, 0.357 on dashboards and 0.762 with a geographic experiment. In this book the dual-track commission is a single-tier business-to-business service fee and involves no multi-level commission of any kind. RM 5,000 a day of agency fee income is treated as a planning target, never a promise. (https://5dlibrary.com/answers/en/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c06/) 第6章 双轨合同:固定授权费、分档抽佣与委托代理的算术 — 第六章设计代理商与品牌之间的双轨合同:固定月费加按实验测得的增量净成交额边际分档的抽佣,并设退货追回与约 0.13 的留存比例;线性合同的最优抽佣比例随测量噪声下降,只看后台时是 0.357,地区实验时是 0.762。双轨抽佣在本书中专指单层企业间服务费,不涉及任何多层计佣。代理业务每天 RM 5,000 的服务费收入只作为推算条件的目标,不是承诺。 (https://5dlibrary.com/answers/zh/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c06/) - Ch.7 The Classics Examined: Bai Gui, the Fair Merchant, the Commenda and Adam Smith — Bai Gui's "take what others discard" reads as counter-cyclical buying: in the illustrative quarter, the same budget of RM 39,000 loses RM 17,362 when put entirely into mega-sale weeks and gains RM 4,446 when put entirely into quiet weeks. "The greedy merchant makes three, the fair merchant makes five" can be read as turnover: a mark-up of 0.25 earns a yearly return of 0.500 and a mark-up of 0.60 only 0.296, yet the latter has the lower break-even ROAS, so a dashboard praises the wrong merchant. The commenda shared losses by capital and profits by contribution; the dual-track contract adds a fixed fee on top and lowers the agency's standard deviation to 6.90. Smith saw that those who manage other people's money are prone to negligence, and that piece rates make people overdraw; the first calls for transparent costs, the second for clawback of returns. (https://5dlibrary.com/answers/en/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c07/) 第7章 原典考辨:白圭、廉贾、康曼达与斯密 — 白圭「人弃我取」读作逆周期投放:示意季度里,同样 RM 39,000 的预算,全押大促周亏 RM 17,362,全放冷清周赚 RM 4,446。「贪贾三之,廉贾五之」可以读作周转:加价率 0.25 的年回报 0.500,加价率 0.60 只有 0.296,可后者的盈亏平衡 ROAS 更低,后台会夸错人。康曼达把亏损按出资分担、利润按贡献分配;双轨合同在此之上加一笔固定费用,让代理商的收益标准差降到 6.90。斯密指出管理别人钱的人难免疏忽,计件报酬会让人透支;前者要求费用透明,后者要求退货追回。 (https://5dlibrary.com/answers/zh/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c07/) - Ch.8 The Conversion Tool: Break-Even, Attribution Dedup and Dual-Track Contract Calculator, with a Pre-Registered Test — Chapter 8 specifies a Break-Even ROAS, Attribution Dedup and Dual-Track Contract Calculator for the live sites and a pre-registered refutation plan. Marked boxes separate published evidence, runnable Python calculations (24 programs), classical sources and the author's hypotheses. The book is not financial, legal or tax advice and promises no income or return. (https://5dlibrary.com/answers/en/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c08/) 第8章 转化工具:盈亏平衡、归因去重与双轨合同试算器及预注册检验 — 第八章给出可接入网站的「盈亏平衡 ROAS、归因去重与双轨合同试算器」与预注册证伪方案。全书以标注框区分已发表的证据、可运行的 Python 计算(24 段程序)、原典与作者的假说;本书不构成财务、法律或税务意见,不承诺任何收入或回报。 (https://5dlibrary.com/answers/zh/sea-social-commerce-whitelist-amplification-cross-channel-attribution-dual-commission/c08/) ## 5DL-B-352 · Productised B2B Service Packaging, Low-Friction Pilot Deposit Anchoring and Forty-Eight-Hour Deal Velocity 中文书名:企业服务产品化封装、低门槛试点定金锚定与四十八小时闪电成交母典 DOI: pending Cite: Fan, Y. (2026). Productised B2B Service Packaging, Low-Friction Pilot Deposit Anchoring and Forty-Eight-Hour Deal Velocity (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All parameters are illustrative and every calculation comes from 22 standard-library programs; deadlines must be real and never reset, marketing messages need prior consent and a one-click unsubscribe, and no multi-level commission is involved. Abstract: The most popular saying among small teams that sell marketing services is that only a custom quotation lands a big deal. This book refutes it. Put the chance of winning, the selling hours and the gross margin into one account, and with illustrative parameters a custom contract is 4.68 times the weighted ticket of a fixed package, yet earns only USD 13.78 of gross profit per selling hour against USD 87.94 for the package, a gap of 6.38 times; a custom deal would have to grow to about USD 26,800 to catch up. The book packages complex marketing services into three fixed prices, USD 350, USD 950 and USD 2,400, written in scope units that promise deliverables and never enquiries, sales or returns. RM 5,000 a day is a planning target used to work out the sales and meetings required, not a promise of income. After checking its sources the book corrects its blueprint in five places: the "zero-risk pilot" becomes a pilot with bounded risk; "too many options frighten buyers" is not a law, so the case for three tiers rests on the compromise effect, true anchors and deferral, and the book reports that in a rational model the middle tier adds only USD 4.75 per potential buyer; forty-eight hours turns from a sales rule into a hypothesis to be tested, with each day's delay multiplying the chance of a sale by 0.7558 under illustrative decay; card payment earns no more per meeting than an invoice issued at the start, USD 269.96 against USD 275.95, and its value is early money; and prepayments are a liability, with about RM 93,942 unearned on an average day. Chapter 1 computes the cash cost of pitches and payment terms; Chapter 2 sets out scope units, delivery hours and tier margins; Chapter 3 measures anchoring, compromise and decoy effects, and under loss aversion adding the third tier raises the middle tier's share from 0.518 to 0.576; Chapter 4 uses a screening model to set the pilot deposit window at USD 343.34 to USD 417.16, where USD 350 keeps 0.942 of serious buyers and admits only 0.095 of browsers; Chapter 5 treats deal velocity with Little's law and a decay model; Chapter 6 gives a seven-item checklist for a one-meeting close, an honest script, written consent and payment risk; Chapter 7 studies Shang Yang's log in the Shiji and earnest money in Justinian's Institutes; Chapter 8 presents the "Three-Tier Quote Builder, Pilot Deposit Calculator and 48-Hour Pipeline Monitor" and a pre-registered testing plan. All parameters are illustrative and every calculation comes from 22 standard-library programs; deadlines must be real and never reset, marketing messages need prior consent and a one-click unsubscribe, and no multi-level commission is involved. 摘要:卖营销服务的小团队最流行的说法是:定制报价才能拿下大单。本书证伪这个说法。把赢面、销售小时与毛利放进同一张账,示意参数下定制单金额是固定套餐加权客单价的 4.68 倍,每销售小时毛利却只有 USD 13.78,套餐是 USD 87.94,相差 6.38 倍;定制单要涨到约 USD 26,800 才能追平。本书把复杂营销服务封装成三档固定价格 USD 350、USD 950、USD 2,400,以范围单位写明交付物,只承诺交付物,不承诺询盘、销量或回报。每日 RM 5,000 只是用来倒推成交与会面数的规划目标,不是收入承诺。 核对文献之后,本书对立项蓝图做了五处修正:所谓「零风险试点」改为风险有上限的试点;「选项过多会吓跑买方」不是定律,三档的理由改为折中效应、真实锚点与推迟倾向,并如实报告理性模型中第二档每位潜在买方只多带来 USD 4.75;四十八小时从成交纪律改为待检验假说,示意衰减下每晚一天成交概率乘以 0.7558;刷卡不比开工日发票更赚钱,每场会面期望净额 USD 269.96 对 USD 275.95,它的价值在于钱早到;预付款是负债,平均约 RM 93,942 尚未赚到。 第一章算比稿与账期的现金代价;第二章写范围单位、交付小时与三档毛利;第三章量化锚定、折中与诱饵效应,损失厌恶下加入第三档让第二档份额从 0.518 升到 0.576;第四章以筛选模型定出试点定金的窗口 USD 343.34 到 USD 417.16,USD 350 留住 0.942 的认真买方、只放进 0.095 的观望者;第五章以利特尔定律与衰减模型处理成交速度;第六章给出一次会面成交的七项清单、诚实话术、书面同意与付款风险;第七章考辨《史记·商君列传》徙木立信与查士丁尼《法学阶梯》论定金;第八章给出可接入现网的「三档报价、试点定金与四十八小时管线体检器」与预注册检验方案。书中参数均为示意,全部计算由 22 段标准库程序生成;截止日必须真实且不重置,营销信息须事先同意并可一键退订,不涉及任何多层返佣。 Chapter answers: - Ch.1 The Cost of Pitch Contests and Payment Terms: Cash Arithmetic for RM 5,000 a Day — "Custom quotes win bigger deals" is only half right. Custom deals are indeed bigger, 4.68 times the package in the illustrative parameters; but they earn only USD 13.78 of gross profit per selling hour against USD 87.94 for packages, a gap of 6.38 times, and a custom deal would have to grow to about USD 26,800 to catch up. Pitches also slow buyers down, because options that are hard to rank make people put decisions off. RM 5,000 a day means 1.260 sales and 4.20 meetings a day. Sixty-day terms lock up about RM 385,000 of receivables, of which only 0.099 has arrived by day sixty. The advantage of a card deposit is not lower fees but early money and no chasing. (https://5dlibrary.com/answers/en/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c01/) 第1章 比稿与账期的代价:日进 RM 5,000 的现金算术 — 「定制报价才能拿大单」只说对了一半:定制单的金额确实大,示意参数下是套餐的 4.68 倍;可每销售小时毛利只有 USD 13.78,套餐是 USD 87.94,相差 6.38 倍,定制单要涨到约 USD 26,800 才能追平。比稿还拖慢买方,因为难分高下的选项会让人推迟决定。日进 RM 5,000 意味着每天成交 1.260 单、开 4.20 场会;60 天账期会压住约 RM 385,000 的应收账款,到第六十天只收回 0.099。卡付定金的好处不是手续费更低,而是钱早到、不必催款。 (https://5dlibrary.com/answers/zh/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c01/) - Ch.2 Productised Packaging and the Three-Tier Price Architecture: Scope Units, Delivery Hours and Margin per Tier — Productisation means writing scope as units, lead time as days and price on the page, and promising deliverables, never results. With the illustrative parameters the three tiers have gross margins of 0.558, 0.572 and 0.577, and the unit price falls by 0.095 and then 0.053 from tier to tier; the floors that keep a 0.50 margin are USD 307.02, USD 804.89 and USD 2,004.89. The tier mix hardly changes the delivery hours needed but decides the number of sales a day: 3.23 for the first tier alone and 1.26 for the middle-weighted mix. Scope creep drags the margin of custom projects from a planned 0.413 to an average of 0.269, with 0.467 of them below 0.30; once scope is written in units with change orders, that probability falls to zero. (https://5dlibrary.com/answers/en/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c02/) 第2章 产品化封装与三档价格架构:范围单位、交付小时与每档毛利 — 产品化是把范围写成单位、把工期写成天数、把价格写在页面上,并且只承诺交付物,不承诺结果。示意参数下,三档的毛利率分别是 0.558、0.572、0.577,单位价格逐档下降 0.095 与 0.053;毛利率 0.50 的价格底线分别是 USD 307.02、USD 804.89、USD 2,004.89。档位结构几乎不改变交付工时,却决定每天要成交的笔数:只卖第一档每天 3.23 单,以中间档为主 1.26 单。范围蔓延让定制项目的毛利率从计划的 0.413 跌到平均 0.269,有 0.467 的项目低于 0.30;写成单位与变更单以后,这个概率降为零。 (https://5dlibrary.com/answers/zh/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c02/) - Ch.3 Anchoring, Compromise and Decoy Effects Measured: When the Middle Tier Is Chosen — Chapter 3 measures anchoring, compromise and decoy effects, and under loss aversion adding the third tier raises the middle tier's share from 0.518 to 0.576; (https://5dlibrary.com/answers/en/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c03/) 第3章 锚定、折中与诱饵效应的量化:中间档在什么条件下被选中 — 第三章量化锚定、折中与诱饵效应,损失厌恶下加入第三档让第二档份额从 0.518 升到 0.576; (https://5dlibrary.com/answers/zh/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c03/) - Ch.4 The Pilot Boundary and the Deposit as a Screen: Why USD 350 Keeps the Browsers Out — Chapter 4 uses a screening model to set the pilot deposit window at USD 343.34 to USD 417.16, where USD 350 keeps 0.942 of serious buyers and admits only 0.095 of browsers; (https://5dlibrary.com/answers/en/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c04/) 第4章 试点边界与定金筛选:USD 350 为什么恰好挡住观望者 — 第四章以筛选模型定出试点定金的窗口 USD 343.34 到 USD 417.16,USD 350 留住 0.942 的认真买方、只放进 0.095 的观望者; (https://5dlibrary.com/answers/zh/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c04/) - Ch.5 Forty-Eight-Hour Deal Velocity: The Pipeline Is a Queue and the Chance of a Sale Decays — Little's law shows that the length of the pipeline is set by the arrival rate and the length of stay. With a stay of two days the package route carries 8.40 open opportunities; at nine days it carries 37.81, and follow-up rises from 1.26 to 5.67 hours a day without a single extra sale. In the illustrative model each day's delay multiplies the chance of a sale by 0.7558, and the synthetic closing rates inside and outside forty-eight hours are 0.317 and 0.092; but the speed of decay depends on τ, and with τ at 30 days the gain shrinks to 0.088. An observational ledger cannot separate speed from urgency, so the test must randomise the timing of quotations. Deadlines must come from real slots and must neither raise the price nor reset when they pass. Taking payment at the start turns the cash cycle into minus 18 days and releases RM 203,300 of working capital compared with custom work on 60-day terms. (https://5dlibrary.com/answers/en/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c05/) 第5章 四十八小时成交速度:管线是一条队列,成交概率会衰减 — 利特尔定律说明管线长度由进入速度与停留时间决定:套餐路线停留两天,管线挂着 8.40 个机会,拖到九天就是 37.81 个,跟进时间从 1.26 小时涨到 5.67 小时,成交一单没多。示意模型中每晚一天,成交概率乘以 0.7558,四十八小时内外的合成成交率是 0.317 对 0.092;可衰减速度取决于 τ,τ 为 30 天时增益只剩 0.088。观察台账分不开速度与急迫,检验必须随机安排报价时间。截止日必须来自真实档期,过期不涨价、不重置。开工即收款让现金周期变为负 18 天,比定制 60 天账期释放 RM 203,300 的营运资金。 (https://5dlibrary.com/answers/zh/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c05/) - Ch.6 The One-Meeting Close and the Mathematics of Payment Risk: An Honest Script, Card Mechanics and a Checklist — "Closing means pushing" is replaced by a sentence that can be tested: a one-meeting close depends on seven conditions being in place before the meeting ends. At 0.95 for each item the meeting closes with probability 0.698; at 0.80 for each, only 0.210; raising the weakest item from 0.60 to 0.95 multiplies the probability by 1.583. The script only makes sure that the information is delivered: scope, price, refunds and a real slot, with no invented places, no resetting countdown and no guarantee of results. Consent to the order and consent to marketing are kept apart, and marketing messages require consent and a one-click unsubscribe. Card payment brings an expected net of USD 269.96 per meeting, no more than an invoice; its value lies in money arriving in two days. For a small merchant the chargeback ratio is noisy, and the probability of crossing the illustrative threshold in at least one month a year is 0.837. On average RM 93,942 of prepayments is unearned at any time; it is a liability, not income. (https://5dlibrary.com/answers/en/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c06/) 第6章 一次会面成交与付款风险:诚实话术、刷卡机制与清单 — 「逼单才是成交」被换成一句可以检验的话:一次会面成交,靠七个条件在会面结束前齐备。每项 0.95 时当场成交 0.698,每项 0.80 时只有 0.210;最弱一项从 0.60 补到 0.95,成交概率乘以 1.583。话术只负责把信息讲完:范围、价格、退款、真实档期,不虚构名额,不重置倒计时,不保证效果。订单同意与营销同意分开,营销信息须有同意与一键退订。卡付款每场会面的期望净额 USD 269.96,不比账期发票高,它的价值在于两天到账;小商户的拒付比例噪声大,一年里至少一个月触及示意阈值的概率是 0.837。预付款平均有 RM 93,942 尚未赚到,是负债,不是收入。 (https://5dlibrary.com/answers/zh/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c06/) - Ch.7 A Study of Classics East and West: Shang Yang's Log and the Institutes of Justinian on Earnest Money — Shang Yang paid fifty pieces of gold for one carried log; in the model this raises the people's trust from 0.100 to 0.400 and lowers the smallest reward that brings a response to later decrees from 20.00 pieces of gold to 5.00. But he deceived Prince Ang, a broken promise of the same weight pulled trust back to 0.300, and in the end the state of Wei refused to take him in. Book 3, Title 23 of the Institutes makes earnest money the evidence that a sale has been concluded, makes a buyer who backs out lose it and a seller who backs out restore double, and requires the price to be certain. In the model, with earnest at 0.10 of the price, the two-sided rule cuts the seller's rate of backing out from 0.261 under the one-sided rule to 0.057. One classic teaches paying a cost first to build trust, the other teaches both sides staking equally; together they form the skeleton of this book's pilot design. (https://5dlibrary.com/answers/en/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c07/) 第7章 东西方原典考辨:商鞅徙木立信与《法学阶梯》论定金 — 商鞅以五十金兑现一次搬木,模型里把百姓的信任从 0.100 抬到 0.400,此后响应政令的最低赏格从 20.00 金降到 5.00 金;可他欺骗公子卬,同等分量的失信让信任回落到 0.300,最终被魏国拒之门外。《法学阶梯》第三卷第二十三题规定定金是买卖成立的证据,买方反悔丧失定金,卖方反悔返还双倍,并要求价格确定;模型中定金取价格的 0.10 时,双向规则把卖方反悔从单向规则下的 0.261 降到 0.057。两部原典一个教人先付代价立信,一个教人双方同等押注,合起来就是本书试点设计的骨架。 (https://5dlibrary.com/answers/zh/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c07/) - Ch.8 The Conversion Tool and the Empirical Loop: Three-Tier Quote Builder, Pilot Deposit Calculator and 48-Hour Pipeline Monitor — Chapter 8 presents the "Three-Tier Quote Builder, Pilot Deposit Calculator and 48-Hour Pipeline Monitor" and a pre-registered testing plan. All parameters are illustrative and every calculation comes from 22 standard-library programs; deadlines must be real and never reset, marketing messages need prior consent and a one-click unsubscribe, and no multi-level commission is involved. (https://5dlibrary.com/answers/en/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c08/) 第8章 转化工具与实证闭环:三档报价、试点定金与四十八小时管线体检器 — 第八章给出可接入现网的「三档报价、试点定金与四十八小时管线体检器」与预注册检验方案。书中参数均为示意,全部计算由 22 段标准库程序生成;截止日必须真实且不重置,营销信息须事先同意并可一键退订,不涉及任何多层返佣。 (https://5dlibrary.com/answers/zh/productised-b2b-service-tiers-pilot-deposit-anchoring-48-hour-deal-velocity/c08/) ## 5DL-B-353 · Repeat-Buyer Reactivation in Private-Channel Commerce, Threshold Bundle Margin Design and Flash-Clearance Dynamics 中文书名:私域代购熟客复购唤醒、满额凑单毛利配比与限时清仓脉冲闪购母典 DOI: pending Cite: Fan, Y. (2026). Repeat-Buyer Reactivation in Private-Channel Commerce, Threshold Bundle Margin Design and Flash-Clearance Dynamics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and correspond to no real customer, order or supplier of HOMIE-X, every price in the book is illustrative, and all calculations are generated by the book's programs. This is general business analysis, not personalised business, tax or legal advice. Abstract: Three sayings circulate among resellers and cross-border personal shoppers: the longer a regular has been away, the harder you should push coupons; set the free-shipping threshold a little above the average order, because the more customers top up, the more you earn; and the shorter and more urgent a flash sale, the faster the stock clears. This book tests each one. It handles right censoring with the Kaplan–Meier product-limit estimator, estimates the probability of being alive with buy-till-you-die models, and writes the reactivation coupon as an incremental ledger: the probability of being alive times the extra conversion and the margin, minus the leak to regulars who would have returned anyway and the cost of the touch. Under illustrative parameters a coupon of RM 6 has positive net value only for gap ratios between 0.85 and 11.37, so both recent buyers and long-dormant ones are worth less than nothing, and redemption accounting without a control group overstates the increment 12.2 times. The first saying is therefore largely refuted. A free-shipping threshold is a price. Piecewise expected margin shows that revenue peaks at a threshold of RM 83 and margin at RM 170, with margins of 20.621 and 23.109 per visitor, in line with the 2006 finding of Lewis and colleagues that free shipping raises sales without necessarily raising profit. A ladder of "spend 150 save 10, spend 250 save 25, spend 400 save 50" cuts margin by 2.493 per visitor, because most of the savings leak to customers who were already above the step; a ladder is worth having only when the top-up zone holds genuine tail stock. Tail-stock clearance is treated with finite-horizon dynamic programming and Poisson sell-out probabilities: a two-hour window sells 9.21 units where a half-hour window sells only 4.79, and when flash sales come too often regulars learn to wait, so the number per month must be capped. Chapter 7 examines Ji Ran's reversal of dear and cheap in the Shiji and Dupuit's tiered bridge tolls of 1844. Chapter 8 gives the weekly ledger of regulars, the pricing and closing tool specification and three pre-registered tests. Parameters are illustrative and correspond to no real customer, order or supplier of HOMIE-X, every price in the book is illustrative, and all calculations are generated by the book's programs. This is general business analysis, not personalised business, tax or legal advice. Reactivation messages go only to people who agreed to receive them and stop at an opt-out; flash-sale counts and deadlines are real and verifiable; referral rewards are single-level and paid only to the direct referrer. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:代购与跨境买手圈流传三句经验:越久没买的老客越要猛发优惠券;免邮门槛定在客单价稍上方,凑单越多越赚;限时闪购越短越急,库存清得越快。本书逐条检验。用卡普兰与迈耶的乘积极限估计处理右删失,用买到死模型估计仍活跃概率,再把唤醒券写成增量账:仍活跃概率乘以增量转化与毛利,减去送给必回客的漏损与触达成本。示意参数下,面额 RM 6 的券只在间隔比 0.85 到 11.37 之间净值为正,刚买客与久睡客都是负值;不设对照组的核销记账会把增量高估 12.2 倍。第一句流行说法因此基本被证伪。 免邮门槛是一道价格。分段期望毛利显示,营收最高的门槛在 RM 83,毛利最高的门槛在 RM 170,每位访客毛利分别为 20.621 与 23.109,与刘易斯等人 2006 年「免邮带动销售却未必带来利润」的发现方向一致。一张「满 150 减 10、满 250 减 25、满 400 减 50」的阶梯让每位访客毛利少 2.493,让利大半漏给本来就在门槛上方的人;阶梯只在凑单区有真正尾货时值得设。尾单清仓用有限期动态规划与泊松售罄概率:两小时窗口卖出 9.21 件,半小时只卖出 4.79 件;闪购太频繁,熟客学会等待,每月次数必须设上限。第七章考辨《史记·货殖列传》计然的贵贱反转与杜皮特 1844 年的分级桥费。第八章给出熟客周度台账、定价与成交工具规格与三项预注册检验。 书中参数为示意,不对应 HOMIE-X 任何真实客户、订单或供货商,书中价格一律为示意;全部计算由书中程序生成。本书是非个人化经营分析,不构成个人化经营、税务或法律建议。唤醒只发给已同意接收者,退订即停;闪购件数与截止真实可核对;推荐奖励只给直接推荐人一层。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Survival Curves of Purchase Gaps: Censoring, Kaplan–Meier and the Misreading of Dormancy — The ledger holds 2,899 gaps. Of these, 2,299 ended in an observed order and 600 are censored, exactly one per buyer. The survival curve says that 0.8178 of gaps last longer than 15 days, 0.5736 longer than 30 days, and only 0.2563 longer than 60 days. By 120 and 180 days the curve has almost stopped moving, resting near 0.1142 and 0.1072. That flat stretch matters a great deal, and the next section returns to it. (https://5dlibrary.com/answers/en/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c01/) 第1章 购买间隔生存曲线:删失、Kaplan-Meier 与沉睡误读 — 台账里共有 2,899 段间隔,2,299 段看到了终点,600 段被删失,恰好每人一段。生存曲线告诉我们:间隔超过 15 天的占 0.8178,超过 30 天的占 0.5736,超过 60 天的只剩 0.2563。到了 120 天与 180 天,曲线几乎不动,停在 0.1142 和 0.1072 附近。这一截平台很重要,下一节再细说。 (https://5dlibrary.com/answers/zh/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c01/) - Ch.2 Buy-Till-You-Die: the Probability of Being Alive and Next Quarter's Expected Orders — The most striking part of the table is its lower right corner. A customer with 12 repeat purchases still has a 0.583 probability of being alive after 8 weeks of silence, but this falls to 0.126 after 14 weeks and to 0.009 after 20 weeks. A customer with a single repeat purchase still has 0.587 after 20 weeks of silence and 0.314 after 34 weeks. The second table compresses the point into one column: the silence needed before the probability of being alive drops below one half is 26 weeks for one repeat purchase, 21 weeks for two, 16 weeks for four, only 12 weeks for eight and only 9 weeks for twelve. (https://5dlibrary.com/answers/en/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c02/) 第2章 买到死模型:仍活跃概率与下一季期望单数 — 表里最醒目的是右下角。复购 12 次的顾客,沉默 8 周时仍活跃概率还有 0.583,沉默 14 周就跌到 0.126,沉默 20 周只剩 0.009。复购只有 1 次的顾客,沉默 20 周仍有 0.587,沉默 34 周还有 0.314。第二个表把这层意思压成一行:仍活跃概率跌破一半所需的沉默,复购 1 次要 26 周,复购 2 次要 21 周,复购 4 次要 16 周,复购 8 次只要 12 周,复购 12 次只要 9 周。 (https://5dlibrary.com/answers/zh/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c02/) - Ch.3 The Incremental Ledger of a Reactivation Coupon: Survival, Leakage to Sure Returners and the Coupon Band — A reactivation coupon has four destinations, and only one of them is incremental. Regulars who have just ordered are the most likely to come back anyway, so a coupon sent to them is mostly a gift. Long-dormant regulars are unlikely to be alive, so a coupon sent to them is mostly a leaflet into an empty room. Incremental value peaks in the middle band: under illustrative parameters a coupon of RM 6 pays only for gap ratios between 0.85 and 11.37, and the band is narrower for regulars with clearer rhythms. Redemption accounting overstates the increment more than tenfold, so one group must always receive nothing. (https://5dlibrary.com/answers/en/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c03/) 第3章 唤醒券增量账:存活概率、必回客漏损与发券区间 — 唤醒券有四个去处,只有一个是增量。刚下过单的熟客必回概率最高,券多半是白送;久睡熟客仍活跃概率太低,券多半是空投。增量价值在中段最高,按示意参数,面额 RM 6 的券在间隔比 0.85 到 11.37 之间才划算,节奏越清楚的熟客区间越窄。核销记账会把增量高估十倍以上,必须留一组不发券。 (https://5dlibrary.com/answers/zh/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c03/) - Ch.4 Free-Shipping Thresholds: Piecewise Expected Margin and the Revenue Trap — A free-shipping threshold is a price: customers on either side of it face different total prices. The piecewise expected margin shows that the revenue-best threshold sits close to the median basket while the margin-best threshold lies well above it; under illustrative parameters the first is RM 83 and the second RM 170. The higher the margin on added items and the lower the shipping cost, the lower the best threshold. Under illustrative parameters, free shipping on everything always loses to never shipping free. An exact check with fractions on the empirical distribution prevents rounding from reversing the ranking and removes the need for a distributional assumption. (https://5dlibrary.com/answers/en/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c04/) 第4章 免邮门槛:分段期望毛利与营收陷阱 — 免邮门槛是一道价格,门槛两侧客人面对不同总价。分段期望毛利显示,营收最优门槛贴近篮子中位数,毛利最优门槛远高于它;按示意参数,前者是 RM 83,后者是 RM 170。凑单品毛利率越高、运费越低,最优门槛越低。全部免邮在示意参数下总是输给不免邮。用 Fraction 在经验分布上精确核对,可以避免舍入颠倒排序,也免去分布假设。 (https://5dlibrary.com/answers/zh/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c04/) - Ch.5 Margin Design of Spend-and-Save Ladders: Where the Savings Leak — Most of what a spend-and-save ladder gives away leaks to customers who were already above the step. Under illustrative parameters, the respectable ladder "spend 150 save 10, spend 250 save 25, spend 400 save 50" raises revenue slightly and lowers margin clearly; every ringgit of margin earned back from added items costs 4.27 ringgit of savings on the first step and 21.90 on the third. A ladder designed for revenue almost always loses margin. A ladder designed for margin has small savings and high steps, and is worth having only when the top-up zone holds real tail stock. (https://5dlibrary.com/answers/en/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c05/) 第5章 满减阶梯毛利配比:让利漏给了谁 — 满减阶梯的让利,大半漏给了本来就在门槛上面的人。示意参数下,一张「满 150 减 10、满 250 减 25、满 400 减 50」的体面阶梯,让营收略增、毛利明显下降;每换回一块钱凑单毛利,第一档要让出 4.27 块,第三档要让出 21.90 块。按营收设计的阶梯几乎一定亏毛利;按毛利设计的阶梯让利很小、档位很高,而且只在凑单区有真正尾货时才值得设。 (https://5dlibrary.com/answers/zh/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c05/) - Ch.6 Clearing Tail Stock: Finite-Horizon Pricing, Two-Hour Pulses and Regulars Who Wait — The price path for clearing tail stock depends on the units left and the time left: the fewer the units and the more the time, the higher the price. When stock is plentiful relative to the flow of buyers, a fixed price is already close to optimal, and even at the last moment the lot should not be dumped at a floor price. Under illustrative parameters a two-hour pulse sells 9.21 units and a half-hour pulse only 4.79: when the window is too short, too few people see it. When flash sales come too often, regulars learn to wait and full-price repurchases are eroded; if they learn quickly, one sale a month is the most a shop should run. (https://5dlibrary.com/answers/en/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c06/) 第6章 尾单清仓:有限期定价、两小时脉冲与等待熟客 — 尾单清仓的价格路径取决于剩余件数与剩余时间:件数越少、时间越多,价格越高;库存相对客流充足时,固定价格已经接近最优,最后时刻也不该甩到地板价。两小时脉冲在示意参数下卖出 9.21 件,半小时只卖出 4.79 件,窗口太短,看见的人太少。闪购太频繁,熟客学会等待,正价复购被侵蚀;学得较快时,每月只宜一场。 (https://5dlibrary.com/answers/zh/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c06/) - Ch.7 Ji Ran and Dupuit: the Reversal of Dear and Cheap and Tiered Tolls in the Classical Sources — Chapter 7 examines Ji Ran's reversal of dear and cheap in the Shiji and Dupuit's tiered bridge tolls of 1844. (https://5dlibrary.com/answers/en/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c07/) 第7章 计然与杜皮特:贵贱反转与分级收费原典考辨 — 第七章考辨《史记·货殖列传》计然的贵贱反转与杜皮特 1844 年的分级桥费。 (https://5dlibrary.com/answers/zh/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c07/) - Ch.8 The Empirical Loop: the Regulars' Ledger and the Pricing and Closing Tool Specification — The regulars' ledger has four tables: exported on Monday, acted on during the week, filled in at the end of the week, tested every quarter. The distilled tool combines the formulas of p31 in Chapter 3, p41 in Chapter 4 and p62 in Chapter 6. With its illustrative defaults it says: send a coupon of RM 6 only to regulars with gap ratios between 0.85 and 11.37, set the free-shipping threshold at RM 170, price a 10-unit tail lot at RM 85 for a two-hour flash sale, and run at most 1 flash sale a month. The three pre-registered tests need 3,100 regulars, 9,588 visitors and 3,551 regulars per group respectively, and their results are recorded whether positive or negative. (https://5dlibrary.com/answers/en/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c08/) 第8章 实证闭环:熟客台账与定价与成交工具规格 — 熟客台账分四张表,周一导出、周内执行、周末回填、每季检验。蒸馏小工具把第三章 p31、第四章 p41 与第六章 p62 的公式合在一起,输入示意默认值时给出:面额 RM 6 的券只发间隔比 0.85 到 11.37 的熟客,免邮门槛 RM 170,10 件尾单两小时闪购定价 RM 85,每月闪购最多 1 场。三项预注册检验分别需要每组 3,100 位熟客、9,588 位访客与 3,551 位熟客,结果不论正负都要记录。 (https://5dlibrary.com/answers/zh/repeat-buyer-reactivation-survival-threshold-bundle-margin-flash-clearance/c08/) ## 5DL-B-354 · Text-First Social Feed Traffic Interception, Pinned-Comment Bridge Funnels and High-Commission Affiliate Monetisation 中文书名:公域图文社交信息流截流、评论区置顶桥接与电商联盟高佣变现母典 DOI: pending Cite: Fan, Y. (2026). Text-First Social Feed Traffic Interception, Pinned-Comment Bridge Funnels and High-Commission Affiliate Monetisation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. This is general business analysis, not legal, tax or financial advice, and it promises no income. Abstract: Four popular sentences rule text-first feed monetisation: controversy pays; links belong in the comments because platforms penalise outbound links; pick the highest commission rate; and a matrix of accounts that boost one another need not fear bans. This book refutes or bounds each. The published research supports only an inverted U between controversy and discussion and does not measure purchases; once a downstream conversion penalty is multiplied in, the illustrative discussion peak lies at a controversy score of 6.00 and the revenue peak at 3.50, and posting where the noise is loudest loses 0.268. Checking the sources corrects the blueprint: high-arousal emotion, not controversy, drives sharing; moral-emotional diffusion is in doubt after reanalysis; the five-factor EPC is this book's accounting identity; and no major platform publishes a link penalty, while a demotion of engagement bait is documented. Parameters are illustrative and all calculations are generated by the book's programs. This is general business analysis, not legal, tax or financial advice, and it promises no income. Chapter 1 computes the distance between the peaks of discussion and revenue. Chapter 2 splits a post into hook, evidence, stance and action, gives the exact condition for the position of the link and the revenue band after honest disclosure. Chapter 3 computes the funnel product of body links, pinned comments and bridge pages: the break-even penalty is 0.150 for a plain pinned comment and 0.490 for a bridge page. Chapter 4 shows that earnings per click are a product of five factors; among the illustrative items the rank correlation between commission rate and EPC is -0.262, the item chosen first by rate earns 0.776 of the best, and the chapter gives a selection frontier and an exact choice under a time budget. Chapter 5 models audience fatigue, shows that pulsed schedules beat even ones and draws up weekly plans for three real brand accounts that each state who runs them. Chapter 6 monitors account health with a sequential probability ratio test and computes the expected value of a boosting network: once the probability of identification exceeds 0.00019, a five-account network is worth less than one compliant account. Chapter 7 reads Xunzi on being good at borrowing things and Aristotle's three modes of persuasion. Chapter 8 gives the weekly ledger, the daily, weekly and monthly rhythm with thresholds, the pricing and closing tool specification, a check of single-level referral rewards and pre-registered tests. Written for compliant operators only, the book requires prominent affiliate disclosure and describes no boosting, disguise or evasion. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:做图文信息流变现的人常听到四条流行说法:争议越大越赚钱;平台给外链降权,所以链接全放评论区没有损失;选佣金率最高的商品;多开几个号组成矩阵互推就不怕封号。本书逐条证伪或立界。已发表的研究只支持争议与讨论之间的倒 U 形,并没有测量购买;把下游转化惩罚乘上去,示意参数下讨论顶点在争议度 6.00,收入顶点在 3.50,按热闹发帖要损失 0.268。核对原文之后,本书做了蓝图修正:推动分享的是高唤醒情绪而不是争议本身,「道德情绪词促进扩散」在重新分析后存疑,联盟推荐费文献讨论的是按线索还是按成交付费,EPC 的五乘数分解是本书的记账恒等式;主要平台没有公开外链降权幅度,能核对到的是对互动诱饵的降权。书中参数为示意,全部计算由书中程序生成。本书是非个人化经营分析,不构成法律、税务或理财建议,不承诺任何收入。 第一章算出讨论顶点与收入顶点的距离。第二章把帖子拆成钩子、证据、立场、行动四段,给出链接位置的精确条件与如实披露之后的收入区间。第三章算正文外链、置顶评论与桥接页的漏斗乘积:普通置顶评论的盈亏平衡降权为 0.150,桥接页为 0.490。第四章说明每次点击收益是五个乘数的乘积,示意商品里佣金率与 EPC 的排序相关为 -0.262,按佣金率挑出的第一名只拿到最优的 0.776,并给出选品前沿与时间预算下的精确解。 第五章写受众疲劳,说明脉冲排程优于平均排程,为三个各自写明归属的真实品牌账号排周计划。第六章用序贯概率比检验监测账号健康,并算出互刷网络的期望值:识别概率高于 0.00019 时,五个号的网络期望值就不如一个合规账号。第七章考辨《荀子·劝学》的「善假于物」与亚里士多德《修辞学》的三种说服方式。第八章给出周度台账、日周月节奏与阈值、定价与成交工具规格、单层推荐奖励的核对与预注册检验。本书只从合规经营者角度写作:联盟关系须显著披露,不写任何互刷、伪装或规避平台规则的手法。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Inverted U of Controversy: the Peak of Discussion Is Not the Peak of Sales — "The more controversial, the more profitable" treats discussion, reach and revenue as one curve. The published research supports only a part of that chain: controversy and discussion form an inverted U, what drives sharing is high-arousal emotion rather than controversy itself, and the diffusion effect of moral-emotional words is in doubt after reanalysis. Multiply in a downstream conversion penalty and the revenue peak must fall to the left of the discussion peak, in the illustrative case from 6.00 to 3.50, with a loss of 0.268 for posting where the noise is loudest. The peak has to be estimated from a ledger; with 30 posts the interval is too wide to use, and only at about 240 does it become usable. (https://5dlibrary.com/answers/en/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c01/) 第1章 争议度倒 U:讨论顶点不等于成交顶点 — 「争议越大越赚钱」把讨论、曝光与收入当成同一条曲线。已发表的研究只支持其中一段:争议与讨论之间是倒 U,带动分享的是高唤醒情绪而不是争议本身,道德情绪词的扩散效应在重新分析后存疑。把下游转化惩罚乘上去,收入顶点必然落在讨论顶点的左边,示意参数下从 6.00 移到 3.50,按热闹发帖要损失 0.268。顶点要靠台账去估,30 条帖子的区间宽到没有用处,240 条才勉强可用。 (https://5dlibrary.com/answers/zh/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c01/) - Ch.2 The Four-Segment Post: Hook, Evidence, Stance, Action and Honest Disclosure — A text post is a four-segment survival curve. Moving the link up puts it in front of more readers, but readers who skipped the evidence click less and buy less. The early link pays only when the product of click and conversion rates after the move is at least the retention it skips, which is 0.72 in the illustrative case. Disclosure is a requirement, not a variable. Its effect on revenue can go either way, and in the illustrative bands the worst corner keeps between 0.760 and 0.900 of the base without changing the order of priority among post types. (https://5dlibrary.com/answers/en/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c02/) 第2章 帖子四段论:钩子、证据、立场、行动与如实披露 — 一条图文帖是一条四段生存曲线。链接提前,看到链接的人多了,没读证据的人却点得少、买得少;提前划算的条件是提前后的点击率乘成交率不低于证据段与立场段留存乘积,示意参数下为 0.72 倍。披露是硬要求,不是变量;它对收入的影响有正有负,示意区间下最坏只保留 0.760 到 0.900,而且不改变帖型的优先次序。 (https://5dlibrary.com/answers/zh/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c02/) - Ch.3 Where the Link Goes: the Funnel Product of Body Links, Pinned Comments and Bridge Pages — Chapter 3 computes the funnel product of body links, pinned comments and bridge pages: the break-even penalty is 0.150 for a plain pinned comment and 0.490 for a bridge page. (https://5dlibrary.com/answers/en/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c03/) 第3章 链接放哪里:正文、置顶评论与桥接页的漏斗乘积 — 第三章算正文外链、置顶评论与桥接页的漏斗乘积:普通置顶评论的盈亏平衡降权为 0.150,桥接页为 0.490。 (https://5dlibrary.com/answers/zh/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c03/) - Ch.4 Earnings per Click: Four Factors beyond the Commission Rate and the Selection Frontier — Chapter 4 shows that earnings per click are a product of five factors; among the illustrative items the rank correlation between commission rate and EPC is -0.262, the item chosen first by rate earns 0.776 of the best, and the chapter gives a selection frontier and an exact choice under a time budget. (https://5dlibrary.com/answers/en/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c04/) 第4章 每次点击收益:佣金率之外的四个乘数与选品前沿 — 第四章说明每次点击收益是五个乘数的乘积,示意商品里佣金率与 EPC 的排序相关为 -0.262,按佣金率挑出的第一名只拿到最优的 0.776,并给出选品前沿与时间预算下的精确解。 (https://5dlibrary.com/answers/zh/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c04/) - Ch.5 Posting Cadence: Audience Fatigue, Pulsed Schedules and a Weekly Plan for Three Real Brand Accounts — Audience fatigue makes the marginal value of extra posts shrink fast. In the illustrative case the third post of the day adds only 114.9 to reach and the fourth subtracts. For the same average volume, a pulsed rhythm beats an even one: random Poisson posting already gains between 1.020 and 1.044 times, deliberately arranged pulses gain more, and the best weekly plan of the main brand account beats the flat plan by more than a tenth. The only kind of multi-account scheduling this book describes is several real brand accounts, each stating who runs it, each with its own audience and content, each keeping its own rhythm within the platform's published limits. (https://5dlibrary.com/answers/en/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c05/) 第5章 发帖节奏:受众疲劳、脉冲排程与三个真实品牌账号的周计划 — 受众疲劳让多发的边际收益急剧缩水:示意参数下,每天第三条只多出 114.9 的触达,第四条为负。同样的平均发帖量,脉冲节奏胜过平均节奏,泊松节奏已高出 1.020 到 1.044 倍,刻意安排的脉冲更多,主品牌账号的最优周计划比平铺高出一成多。所谓「防封号矩阵排程」,本书只写一种:几个真实品牌账号各写明归属、各有独立受众与内容、在平台公开限额之内各排各的节奏。 (https://5dlibrary.com/answers/zh/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c05/) - Ch.6 Account Health: the Sequential Probability Ratio Test and the Expected Value of a Boosting Network — Chapter 6 monitors account health with a sequential probability ratio test and computes the expected value of a boosting network: once the probability of identification exceeds 0.00019, a five-account network is worth less than one compliant account. (https://5dlibrary.com/answers/en/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c06/) 第6章 账号健康:序贯概率比检验与互刷网络的期望值 — 第六章用序贯概率比检验监测账号健康,并算出互刷网络的期望值:识别概率高于 0.00019 时,五个号的网络期望值就不如一个合规账号。 (https://5dlibrary.com/answers/zh/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c06/) - Ch.7 Classical Sources: Xunzi on Borrowing Things and Aristotle's Three Modes of Persuasion — Xunzi's "beckoning from a height", "shouting with the wind" and "being good at borrowing things" describe amplification. Position, timing and instruments are multipliers that leave the content unchanged; under the illustrative activity curve the best hour beats the worst by 7.35 times, but this is an upper bound that ignores competition within the hour. What is borrowed must be legitimate, and impersonation is not borrowing. Aristotle held that ethos should come from the speech rather than from prior reputation; the synthetic data show that 40 posts overstate the effect of follower count almost twofold and that 160 are needed to tell the two apart. Read together, the two books say that reach comes from legitimate borrowing and trust from checkable evidence. (https://5dlibrary.com/answers/en/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c07/) 第7章 原典考辨:《荀子·劝学》的善假于物与亚里士多德的三种说服 — 《荀子·劝学》的「登高而招」「顺风而呼」「善假于物」说的是放大:位置、时段与工具都是乘数,不改变内容本身;示意活跃曲线下最好钟点是最差钟点的 7.35 倍,但这是忽略同时段竞争的上限。借来的必须是正当的外物,冒名不是假物。亚里士多德说品格应来自言辞而非先前声望;合成数据显示,40 条帖子会把粉丝数的作用高估近一倍,160 条才分得清。两部旧书合起来,就是触达靠正当借势、信任靠可核对的证据。 (https://5dlibrary.com/answers/zh/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c07/) - Ch.8 Closing the Loop: the Weekly Post Ledger and the Pricing and Closing Tool Specification — Chapter 8 gives the weekly ledger, the daily, weekly and monthly rhythm with thresholds, the pricing and closing tool specification, a check of single-level referral rewards and pre-registered tests. Written for compliant operators only, the book requires prominent affiliate disclosure and describes no boosting, disguise or evasion. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c08/) 第8章 实证闭环:图文帖周度台账与定价与成交工具规格 — 第八章给出周度台账、日周月节奏与阈值、定价与成交工具规格、单层推荐奖励的核对与预注册检验。本书只从合规经营者角度写作:联盟关系须显著披露,不写任何互刷、伪装或规避平台规则的手法。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/text-first-feed-interception-pinned-comment-bridge-affiliate-epc/c08/) ## 5DL-B-355 · Three-Question Property Lead Qualification, Same-Day Multi-Viewing Closing Kinetics and Exclusive Landlord Mandates 中文书名:房产高意向客资质三问秒筛、多盘同天连看逼单与房东独家委托母典 DOI: pending Cite: Fan, Y. (2026). Three-Question Property Lead Qualification, Same-Day Multi-Viewing Closing Kinetics and Exclusive Landlord Mandates (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: The third needs boundaries: an exclusive mandate serves the owner only up to a commission rate of 0.0287 and a best term of 45 days in the illustrative model. Parameters are illustrative and every number comes from the book's programs. Abstract: Klang Valley property-lead work runs on three popular sayings: the more a form asks, the better it finds real buyers; on a same-day viewing the last unit closes best, so stage a weak unit first as a decoy; and an owner in a hurry should be pushed down on price, while an exclusive mandate is bad for the owner. This book tests each. The first is refuted: extra questions lower completion while adding ever less accuracy, and under illustrative parameters three questions, on cash, the debt service ratio and the timetable, bring the expected net value per lead to its peak of 129.21, though the optimum moves between two and five with the keep-rate per question. The second is half right: anchoring is real, and Northcraft and Neale's figures imply an anchor weight of 0.4850, but decoy effects usually vanish in experienced choices, and a decoy beats a genuine comparable only if its effect exceeds 0.1200. The third needs boundaries: an exclusive mandate serves the owner only up to a commission rate of 0.0287 and a best term of 45 days in the illustrative model. Blueprint corrections are recorded in the Preface. Parameters are illustrative and every number comes from the book's programs. This is general legal, tax, financial, lending and investment analysis, not advice; statutes, central-bank figures and stamp duty bands are to be checked by hand before publication. The chapters cover the stopping point under a cost threshold of 0.3333 and Wald's sequential test; the debt service ratio, its stress test and the affordable price; the itemised cash threshold, 72,250 or 0.1445 of an illustrative price; viewing routes, anchoring and how many units to see; bounds on decoys and choice overload, and a review period of 58.5 hours for a compliant close; the owner's net price, Rubinstein bargaining and rental vacancy; the Rites of Zhou and Cicero on records and disclosure; and a ledger, trigger rules, the pricing and closing specification of the distilled Three-Question Lead Triage and Exclusive Mandate Net-Price Comparator, and pre-registered tests. Closing always means honest comparison, real deadlines and time to check documents. The book contains 22 runnable Python programs. 摘要:巴生谷做房产线索的人常信三句流行说法:表单问得越多越能筛出真买家;同天连看三盘,最后一盘最好成交,先放一套差房当诱饵;急售业主最好压价,签独家对业主不利。本书逐条检验。第一句被证伪:每多问一题,完成率下降,准确率的边际增益递减;示意参数下,现金、偿债比与时间表三问使每条线索的期望净价值达到峰值 129.21,但最优题数随每题留存比例在二到五题之间移动,三问是区间结论而不是定律。第二句只对了一半:锚定在房价判断中真实存在,诺斯克拉夫特与尼尔的数字隐含锚定权重 0.4850;诱饵效应在真实可感的选择里通常不出现,示意账本中诱饵要胜过真比较房,效应须超过 0.1200。第三句需要立界:独家委托解决经纪努力的道德风险,却不自动对业主有利;示意参数下对业主有利的佣金率上限是 0.0287,最优期限 45 天。核对原文后,本书在序中写明蓝图修正。书中参数均为示意,全部计算由书中程序生成。本书是非个人化的法律、税务、财务、贷款与投资分析,不构成建议;法条、国家银行口径与印花税阶梯以上架前人工复核为准。 第一章用似然比、误判代价阈值 0.3333 与沃尔德序贯检验求停问点。第二章用年金公式算偿债比、做压力测试并反推可负担房价。第三章逐项算现金门槛:示意房价五十万、成数九成时需要现金 72,250,占房价 0.1445;并说明三问答案相关时朴素贝叶斯的校准偏差。第四章穷举带时间窗的看房顺序,量化锚定与一天看几套。第五章给诱饵效应与选择过载立界,并算出合规成交推进的考虑期,示意模型下 58.5 小时最优。第六章比较独家与开放委托的业主期望净价,计算急售业主的鲁宾斯坦议价分配与租房空置。第七章考辨《周礼·地官·质人》的书面凭证与西塞罗《论义务》第三卷论卖方披露。第八章给出台账字段、日周月节奏与触发动作、蒸馏小工具「房产线索三问分级与独家委托净价比较器」的定价与成交工具规格,以及预注册检验与推翻条件。书中「逼单」一律指合规成交推进:如实比较、真实截止、给客户考虑与核对文件的时间。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Information Value of Three Questions: Question Count, Completion and the Stopping Point — The best number of questions is set by two curves. Accuracy rises with the number of questions, but by ever smaller steps; completion falls geometrically. Under the illustrative parameters the value per lead peaks at three questions, at 129.21, and every question after that loses value. Change the keep-rate, though, and the optimum moves anywhere between two and five. The cost of errors sets the threshold, and the threshold sets the tiers. Sequential testing shows that when the first two answers agree the conversation can stop at once, and that when they conflict, supplementary questions cannot move the verdict, so the case should go to a phone conversation instead. (https://5dlibrary.com/answers/en/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c01/) 第1章 三问的信息价值:题数、完成率与停问点 — 题数的最优点由两条曲线决定:准确率随题数递减地上升,完成率随题数几何地下降。示意参数下,每条线索的价值在三题处达到峰值 129.21,此后每题都在亏;但留存比例一变,最优题数就在二到五之间移动。误判代价决定阈值,阈值决定分层。序贯检验告诉我们,前两题一致时可以立刻停问,答案打架时补充题也推不动结论,应当改由电话面谈。 (https://5dlibrary.com/answers/zh/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c01/) - Ch.2 The Arithmetic of the Debt Service Ratio: Annuity Instalments, Stress Tests and the Affordable Price — The instalment follows from the annuity formula, and the debt service ratio is all monthly repayments divided by monthly net income. The illustrative household has a ratio of 0.4654 at a rate of 0.040 over 35 years; mistaking gross income for the denominator understates it at 0.3723. The stress test shows that a higher income is not the same as safety: H4 crosses 0.70 after a rise of only 0.0023. When the calculation is inverted to find an affordable price, both the loan side and the cash side must be checked, and for many households the binding constraint is cash. (https://5dlibrary.com/answers/en/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c02/) 第2章 偿债比的算术:月供年金、压力测试与可负担房价反推 — 月供由年金公式决定,偿债比是全部月供除以月净收入。示意家庭在利率 0.040、年期 35 年时偿债比 0.4654,把毛收入误作分母会低估到 0.3723。压力测试显示,收入高不等于安全,H4 只需利率上浮 0.0023 就越过 0.70。反推房价时要同时看贷款与现金两侧,许多家庭真正受约束的是现金。 (https://5dlibrary.com/answers/zh/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c02/) - Ch.3 The Cash Threshold: Stamp Duty Bands, Transaction Costs and a Combined Three-Question Score — At an illustrative price of 500,000 and a margin of 0.90, the cash threshold behind the first question is 72,250, or 0.1445 of the price, which is 22,250 more than the ten per cent down payment. The stamp duty schedule is progressive and moves smoothly across band edges without a cliff, so any suggestion to understate the contract price or record part of it elsewhere has no basis in the arithmetic and is not compliant. When the three answers are correlated, naive Bayes overstates the posterior for agreeing answers, but the ranking power of the "two, two, one" score is unaffected; probabilities should be calibrated from the backfilled ledger. (https://5dlibrary.com/answers/en/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c03/) 第3章 现金门槛:印花税阶梯、交易费用与三问合成评分 — 示意房价 500,000、成数 0.90 时,第一问背后的现金门槛是 72,250,占房价 0.1445,比一成首付多出 22,250。印花税阶梯是累进的,跨档时平滑过渡,没有悬崖,任何压低合约价或价外安排的说法都没有算术依据,也不合规。三道题答案相关时,朴素贝叶斯会高估一致答案的后验,但「二二一」打分的排序能力不受影响;概率应当用台账回填校准。 (https://5dlibrary.com/answers/zh/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c03/) - Ch.4 The Same-Day Viewing: Routes under Time Windows, the Weight of the Anchor and How Many Units to See — Of the 120 orders in which five units can be visited, only 8 are feasible, and the best finishes fifty minutes before the worst feasible order; a staged order costs a quarter of an hour. Northcraft and Neale's figures imply an anchor weight of 0.4850, and experts are no exception. On a same-day viewing the previous unit's asking price forms a second anchor, and a comparison sheet handed out in advance cuts its effect to about a third. How many units to see is set by the declining marginal value of comparison and by the cost of time; under the illustrative parameters three is best, but when the time cost changes, the optimum moves anywhere from two to eight. (https://5dlibrary.com/answers/en/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c04/) 第4章 同天连看:时间窗路线、锚定权重与一天看几套 — 五套房的 120 种顺序里只有 8 种能走通,最优顺序比最差的可行顺序早五十分钟结束;刻意的布局顺序多花一刻钟。诺斯克拉夫特与尼尔的数字隐含锚定权重 0.4850,专家也不例外。同天连看时,前一套的叫价构成第二个锚,事前比较表能把它的影响压到原来的三分之一左右。看几套由比较价值的边际递减与时间成本共同决定,示意参数下三套最优,但时间成本一变,最优就在两套到八套之间移动。 (https://5dlibrary.com/answers/zh/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c04/) - Ch.5 Bounding the Decoy Effect: an Effect that May Be Positive or Negative, Choice Overload and Compliant Progress to a Close — The decoy effect exists on numerical attribute tables in the laboratory, usually fails to appear in choices that people actually experience and sometimes reverses; with zero occurrences in 19 natural-stimulus studies, the upper bound on how often it appears is 0.1459. In the illustrative account of a viewing day the decoy must exceed +0.1200 to beat a genuine comparable, and it costs buyers 6.0 deals per hundred viewing days. The mean effect of choice overload is close to zero, the prediction interval for a new setting runs from -0.6670 to 0.7070, and its direction cannot be known in advance. Compliant progress to a close gives the client time to consider; under the illustrative model 58.5 hours is best, completing 5.40 more deals than pressing for an immediate signature. (https://5dlibrary.com/answers/en/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c05/) 第5章 诱饵效应立界:可正可负的效应量、选择过载与合规成交推进 — 诱饵效应在实验室的数字属性表上存在,在真实可感的选择里通常不出现,有时还会反向;19 项自然刺激研究零次出现,允许的出现比例上界是 0.1459。在示意的带看账本里,诱饵要胜过真比较房,效应必须超过 +0.1200,而诱饵让买家每一百个带看日少成交 6.0 宗。选择过载的平均效应接近零,新情境的预测区间从 -0.6670 到 0.7070,方向无法事先断定。合规成交推进给客户考虑期,示意模型下 58.5 小时最优,比当场逼签多完成 5.40 宗。 (https://5dlibrary.com/answers/zh/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c05/) - Ch.6 Exclusive Mandates: Agent Effort, the Owner's Net Price and Bargaining for an Owner in a Hurry — An exclusive mandate solves the moral hazard of agent effort but does not automatically serve the owner. In the illustrative model, competition under an open listing produces a higher probability of a sale, at the cost of duplicated work. The agent receives only 0.0200 of any increase in price and is less willing than the owner to wait for a better one. Comparing the owner's expected net prices gives two ceilings: under the illustrative parameters the commission rate should not exceed 0.0287, and the best term is 45 days. The share of an owner in a hurry falls quickly as patience falls, and the agent's duty is to win patience back for the owner. Rental vacancy follows a geometric distribution, and before cutting a rent the value of the shorter vacancy should be computed. (https://5dlibrary.com/answers/en/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c06/) 第6章 独家委托:经纪努力、业主净价与急售业主的议价 — 独家委托解决经纪努力的道德风险,却不自动对业主有利:在示意模型里,开放委托的竞争反而带来更高的成交概率,代价是重复劳动。经纪从房价增量中只分到 0.0200,比业主更不愿意为好价多等。业主期望净价的比较给出两个上限:示意参数下佣金率不超过 0.0287,期限以 45 天为最优。急售业主的议价份额随耐心下降而快速缩水,经纪的职责是替业主争回耐心。租房空置期服从几何分布,降租之前先算空置缩短的价值。 (https://5dlibrary.com/answers/zh/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c06/) - Ch.7 Classical Sources East and West: the Market-Certificate Office of the Rites of Zhou and Cicero on the Seller's Disclosure — The threshold is 22,222. For a deal worth 5,000 the expected total cost of the short form is 25.0 against 63.8 for the long form, so the short form is cheaper. At 20,000 the short form costs 70.0 and the long form 75.0, still a win for the short form. At 50,000 the short form costs 160.0 and the long form 97.5, and the long form wins. At a house price such as 600,000 the expected cost of the short form is 1,810.0 and that of the long form only 510.0. "Large deals with a zhi, small deals with a ji" is not red tape; it is a rule that stands up on expected cost. (https://5dlibrary.com/answers/en/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c07/) 第7章 东西方原典考辨:《周礼》质人的书面凭证与西塞罗论卖方披露 — 阈值是 22,222。交易额 5,000 时,短式的期望总成本 25.0,长式 63.8,短式划算;20,000 时,短式 70.0,长式 75.0,仍是短式划算;50,000 时,短式 160.0,长式 97.5,长式胜出;到了 600,000 这样的房价,短式的期望成本 1,810.0,长式只有 510.0。「大市以质,小市以剂」不是繁文缛节,而是一条在期望成本上站得住的规则。 (https://5dlibrary.com/answers/zh/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c07/) - Ch.8 The Empirical Loop and the Pricing and Closing Tool Specification: Ledger, Rhythm, Comparator and Pre-Registered Tests — There are two ledgers. The lead ledger records the three answers, the cash threshold, the stress ratio, the tier, the viewings and the signing; the mandate ledger records the type, the commission rate, the term, the commitments, the days on the market and the sale price. The rhythm runs daily, weekly, monthly and quarterly, and every threshold has a trigger action. The comparator gathers the formulas of Chapters 1, 2, 3 and 6 into a single file, and its results for the demonstration leads and mandate match those of p81 exactly. Among the pre-registered tests, the test of question count needs 209 leads per group and can finish in three weeks; the test of the review period needs 1,143 offers per group and must accumulate across quarters; and the test of small differences in exclusive mandates needs data pooled across several partner agencies. (https://5dlibrary.com/answers/en/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c08/) 第8章 实证闭环与定价与成交工具规格:台账、节奏、比较器与预注册检验 — 台账两张:线索台账记三问、现金门槛、压力偿债比、分层、带看与签约;委托台账记类型、佣金率、期限、承诺、在市天数与成交价。节奏按日、周、月、季度展开,每个阈值都对应一个触发动作。比较器把第一、二、三、六章的公式收进一个文件,示范线索与委托的结果与 p81 逐字相同。预注册检验里,题数检验每组 209 条、三周可完成;考虑期检验每组 1,143 次,需要跨季度累积;独家委托的小差距检验需要多家合作经纪合并数据。 (https://5dlibrary.com/answers/zh/three-question-dsr-lead-triage-same-day-viewing-exclusive-mandates/c08/) ## 5DL-B-356 · Niche Keyword Positioning for Digital Monograph Catalogues, High-Ticket Thematic Bundle Direct Sales and Automated Knowledge Cashflow 中文书名:百卷数字专著蓝海词卡位、高客单主题合集直售与全自动知识收银母典 DOI: pending Cite: Fan, Y. (2026). Niche Keyword Positioning for Digital Monograph Catalogues, High-Ticket Thematic Bundle Direct Sales and Automated Knowledge Cashflow (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. This is general business analysis, not personalised business, legal or tax advice, and it promises no income; the catalogue is described publicly only as the 100 published titles, with "more than 330 manuscripts" as an on-site statement. Abstract: Three popular sentences promise to turn a large bilingual library into daily cash: target the keywords with the largest search volume; sell cheap single titles and the upgrade to a high-priced collection will follow; automate checkout and delivery and the business runs with no human work and no risk. This book refutes or bounds each of them. Expected sales from a keyword are roughly demand times the chance of a front position times click-through and conversion, and for a head term the second factor is almost zero: in the illustrative data a phrase searched twenty thousand times a week brings about 0.0254 copies a week, while a niche phrase brings 3.1445. The long-tail share is acutely sensitive to the power-law exponent, and the share beyond rank one hundred thousand can fall from 0.411 to 0.046. Corrections to the blueprint: the rank-to-sales exponent is illustrative, not a constant; KDP keyword rules follow the live help page (up to seven keywords, no unassociated author names, no brands one does not own); AI-generated content must be disclosed; no work by Dan Ariely is cited. Parameters are illustrative and all calculations are generated by the book's programs. This is general business analysis, not personalised business, legal or tax advice, and it promises no income; the catalogue is described publicly only as the 100 published titles, with "more than 330 manuscripts" as an on-site statement. Chapters 1 to 4 cover the power laws of rank and search demand, a blue-ocean score built on a Plackett-Luce ranking model, an exhaustive choice of the seven slots after a KDP rule filter, and Jaccard clustering of the catalogue with a measure of overlap with past purchases. Chapter 5 shows that bundling loses its edge as valuations become correlated: for eight volumes the bundle-to-separate revenue ratio falls from 1.593 to 1.029 as the correlation rises from zero to 0.9, and at a flat price only 0.187 of readers who own four volumes upgrade, against 0.868 of those who own none. Chapter 6 shows that tiers at 99 and 299 sort readers only when the content gap is at least 200, that a left-digit ending is worth a few per cent, and that the multi-step funnel from a KDP title to the site's collections yields an average direct net of only 95.3 per thousand readers, against royalties of 7,000. Chapter 7 reads the Hanshu on Liu Xiang's collation and Naudé's advice of 1627 on founding a library. Chapter 8 gives a weekly cash ledger, the pricing and closing tool specification of the distilled Catalogue Blue-Ocean Keyword and Theme-Bundle Upgrade Planner, and pre-registered tests: the five steps of automated delivery all succeed unattended with probability 0.9654, and checking only the email and the link costs 1.485 per order. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:把大型双语书库变成每日现金流,流行说法有三句:选搜索量最大的关键词;单本低价引流再推高价合集,升单一定会来;收银与交付全自动就是零人工、零风险。本书逐条证伪或立界。对小作者,关键词带来的期望销量约等于搜索需求、进入前几位的概率、点击率与转化率之积;大词的进前位概率近乎为零,示意数据里周搜索两万次的大词只带来约 0.0254 本周销量,冷僻短语却有 3.1445 本。长尾份额对幂律指数极其敏感,排名十万以后的份额可以从 0.411 跌到 0.046。核对原文后,本书做了蓝图修正:排名换销量的指数是示意参数而非定值;KDP 关键词规则以官方帮助页为准,每本书至多七个关键词,不得用与本书无关的作者名或自己不拥有的品牌;AI 生成内容须如实披露;本书不引用艾瑞里的任何研究。书中参数为示意,全部计算由书中程序生成。本书是非个人化经营分析,不构成经营、法律或税务建议,不承诺任何收入;对外书目口径只写已发布的 100 部与站内「330 余部书稿」。 第一章写排名与销量、搜索词需求的两条幂律,并说明估计指数要用最大似然。第二章用普拉基特·卢斯排名模型与级联点击模型给出蓝海词评分。第三章把 KDP 规则写成过滤器,在七个槽里穷举最优组合。第四章用杰卡德相似度把书库聚成主题,测量合集与读者已购书目的重叠。第五章说明捆绑优势随估值相关性上升而消失,八卷书相关系数从零升到 0.9 时,捆绑对分别定价的收入比从 1.593 降到 1.029;已买四卷的读者按统一价只有 0.187 会升单,一卷未买者为 0.868。 第六章说明 99 与 299 两档只有在内容价值差不小于 200 时才能把读者分开,左位数效应只值几个百分点,从 KDP 单本走到自营合集的多步漏斗,每千名读者的直接净收入平均只有 95.3,同样读者的版税是 7,000。第七章考辨《汉书·艺文志》叙刘向校书与诺代 1627 年的《建立图书馆的建议》。第八章给出书库现金流周度台账、蒸馏小工具「书库蓝海词评分与主题合集升单配比器」的定价与成交工具规格与预注册检验:交付五环无人值守时全部成功的概率是 0.9654,只复核邮件与链接两环时每笔期望成本 1.485。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Power Laws of Search Demand: Rank, Sales and the Share of the Long Tail — The relationship between rank and sales is a power law, and every step on the rank ladder is expensive. The long-tail share is acutely sensitive to the exponent: when it moves from 0.871 to 1.2, the share beyond rank one hundred thousand falls from 0.411 to 0.046. Search demand is also a power law; the top twenty phrases take more than half of all queries, and the exponent should be estimated by maximum likelihood rather than by a straight line on a log-log chart. Multiply the two power laws and the conclusion follows: a small author should not chase the biggest phrases, but the phrases that people do search for and whose front page is not crowded. (https://5dlibrary.com/answers/en/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c01/) 第1章 搜索需求的幂律:排名、销量与长尾份额 — 排名与销量之间是幂律,排名每一步都很贵;长尾份额对指数极其敏感,指数从 0.871 升到 1.2,十万名以后的份额就从 0.411 降到 0.046。搜索词需求也是幂律,前二十个短语拿走一半以上查询,估计指数时要用最大似然,不要用对数坐标上的直线。两条幂律相乘,结论是:小作者不该追大词,应该追「有人搜、但前排不拥挤」的词。 (https://5dlibrary.com/answers/zh/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c01/) - Ch.2 Front-Page Probability and the Blue-Ocean Keyword Score — The numbers speak plainly. On the head term, the new title's probability of a top-three position rounds to 0.0000 with no reviews, is 0.0001 with five reviews and only 0.0002 with twenty. Put more intuitively, a new title with five reviews appears in the top three for about one search in 13,657. The middle term is somewhat better, at 0.0117 with five reviews. The niche term is another world: 0.1559 with no reviews at all, 0.3753 with five and 0.6015 with twenty. With the same five reviews, the chance on the niche term is 5,126 times the chance on the head term. (https://5dlibrary.com/answers/en/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c02/) 第2章 进前位概率与蓝海词评分 — 数字说得很直白。大词上,新书评价为零时进前三的概率四舍五入是 0.0000,五条评价时是 0.0001,二十条评价也只有 0.0002。换算成更直观的说法:五条评价的新书,大约每 13,657 次搜索才有一次出现在前三位。中等词好一些,五条评价时是 0.0117。冷僻词完全是另一个世界:零评价时已有 0.1559,五条评价时是 0.3753,二十条评价时是 0.6015。同样五条评价,冷僻词上的机会是大词的 5,126 倍。 (https://5dlibrary.com/answers/zh/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c02/) - Ch.3 Seven Keyword Slots: a Compliance Filter and an Exact Knapsack — The sweep over the decay factor points the same way. With a factor of 0.3, meaning that same-intention phrases duplicate each other heavily, the exact optimum beats the naive set by 1.067; with 0.9, meaning that same-intention phrases are nearly independent, the advantage shrinks to 1.027. The heavier the overlap, the more the top-seven habit loses. (https://5dlibrary.com/answers/en/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c03/) 第3章 七个关键词槽:合规过滤与背包选词 — 衰减系数的扫描给出同一个方向。系数取 0.3,表示同意图短语重复得很厉害,精确最优比天真做法高出 1.067 倍;系数取 0.9,同意图短语几乎各管各的,优势缩到 1.027 倍。重叠越重,「按分数取前七」越吃亏。 (https://5dlibrary.com/answers/zh/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c03/) - Ch.4 Clustering the Catalogue into Themes and Measuring Bundle Overlap — Jaccard similarity of keyword sets with average linkage clusters a catalogue into themes automatically; in the illustrative catalogue the similarity inside clusters is 0.269 and across clusters only 0.007. Target readers of a themed collection still lack, on average, more than sixty per cent of its value, which makes it a better offer than a cross-theme collection; yet nearly a quarter of them already own more than half of it, and a full-price offer does almost nothing for them. (https://5dlibrary.com/answers/en/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c04/) 第4章 书库主题聚类与合集重叠 — 用关键词集合的杰卡德相似度与平均连接法,可以把书库自动聚成主题,示意书库里类内相似度 0.269,类间只有 0.007。主题合集的目标读者平均还有六成以上的价值没拿到,比跨主题合集更值得推;可是将近四分之一的目标读者已买过合集的一半以上,原价推销对他们几乎无效。 (https://5dlibrary.com/answers/zh/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c04/) - Ch.5 Bundle Valuation: Correlation, Marginal Cost and Past Purchases — Chapter 5 shows that bundling loses its edge as valuations become correlated: for eight volumes the bundle-to-separate revenue ratio falls from 1.593 to 1.029 as the correlation rises from zero to 0.9, and at a flat price only 0.187 of readers who own four volumes upgrade, against 0.868 of those who own none. (https://5dlibrary.com/answers/en/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c05/) 第5章 捆绑估值:相关性、边际成本与已购重叠 — 第五章说明捆绑优势随估值相关性上升而消失,八卷书相关系数从零升到 0.9 时,捆绑对分别定价的收入比从 1.593 降到 1.029;已买四卷的读者按统一价只有 0.187 会升单,一卷未买者为 0.868。 (https://5dlibrary.com/answers/zh/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c05/) - Ch.6 Two Price Tiers: Screening at 99 and 299, the Left Digit, and a Multi-Step Funnel — Chapter 6 shows that tiers at 99 and 299 sort readers only when the content gap is at least 200, that a left-digit ending is worth a few per cent, and that the multi-step funnel from a KDP title to the site's collections yields an average direct net of only 95.3 per thousand readers, against royalties of 7,000. (https://5dlibrary.com/answers/en/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c06/) 第6章 两档价格:99 与 299 的甄别、左位数与多步漏斗 — 第六章说明 99 与 299 两档只有在内容价值差不小于 200 时才能把读者分开,左位数效应只值几个百分点,从 KDP 单本走到自营合集的多步漏斗,每千名读者的直接净收入平均只有 95.3,同样读者的版税是 7,000。 (https://5dlibrary.com/answers/zh/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c06/) - Ch.7 Classical Sources: the Treatise on Literature of the Hanshu and Naudé's Advice on Establishing a Library — Chapter 7 reads the Hanshu on Liu Xiang's collation and Naudé's advice of 1627 on founding a library. (https://5dlibrary.com/answers/en/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c07/) 第7章 原典考辨:《汉书·艺文志》与诺代《建立图书馆的建议》 — 第七章考辨《汉书·艺文志》叙刘向校书与诺代 1627 年的《建立图书馆的建议》。 (https://5dlibrary.com/answers/zh/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c07/) - Ch.8 The Empirical Loop: a Weekly Cash Ledger for the Catalogue and the Pricing and Closing Tool Specification — Chapter 8 gives a weekly cash ledger, the pricing and closing tool specification of the distilled Catalogue Blue-Ocean Keyword and Theme-Bundle Upgrade Planner, and pre-registered tests: the five steps of automated delivery all succeed unattended with probability 0.9654, and checking only the email and the link costs 1.485 per order. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c08/) 第8章 实证闭环:书库现金流周度台账与定价与成交工具规格 — 第八章给出书库现金流周度台账、蒸馏小工具「书库蓝海词评分与主题合集升单配比器」的定价与成交工具规格与预注册检验:交付五环无人值守时全部成功的概率是 0.9654,只复核邮件与链接两环时每笔期望成本 1.485。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/niche-keyword-kdp-positioning-thematic-bundle-direct-knowledge-cashflow/c08/) ## 5DL-B-357 · The Thousand-Ringgit Single-Site Minimum Loop, Headless Core Multi-Skin Cloning and Five-Site Revenue Multiplication 中文书名:单站日入千币最小闭环、中央母核多皮裂变与五站矩阵复制母典 DOI: pending Cite: Fan, Y. (2026). The Thousand-Ringgit Single-Site Minimum Loop, Headless Core Multi-Skin Cloning and Five-Site Revenue Multiplication (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Order values, conversion rates, fees, incident rates and market parameters are illustrative, and all calculations are generated by the book's programs. Revenue figures are illustrative arithmetic, not a promise of income, and not investment, legal or tax advice. Abstract: People who run websites as a business often hear a popular belief: get one site to RM 1,000 a day, copy it five times, and you have RM 5,000 a day. This book tries to refute the belief piece by piece and then draws the boundary within which it holds. Its core thesis is that five cloned sites do not give five times the income: overlap of keywords and audiences, the smaller own demand of new sites, correlated failure on a shared core and the noise in one site's daily revenue each cut away part of the multiplier. Order values, conversion rates, fees, incident rates and market parameters are illustrative, and all calculations are generated by the book's programs. Revenue figures are illustrative arithmetic, not a promise of income, and not investment, legal or tax advice. The book discusses only single-tier, direct referral relationships and involves no multi-level profit sharing or commission. Chapters 1 to 4 show that under the illustrative parameters RM 1,000 of revenue leaves RM 733.37 of contribution, and RM 1,000 of contribution needs RM 1,364 of revenue; that daily revenue has a coefficient of variation of 0.604, so a site truly averaging RM 850 has a 0.218 chance of showing a thousand on a 7-day mean; and that a clone gate combining a negative-binomial sequential test with at least 28 days and a mean of at least RM 950 cuts the false-pick rate to 0.090. With one site a central core has scope economies of -0.404, breaks even at 1.90 sites, and should wait whenever the first site succeeds with probability below 0.409; copied forks miss 23.05 ports of fixes a year. Of five candidate skins, the payback rule opens three, with combined revenue of RM 1,770.00 a day, none close to the original site. Chapters 5 to 8 show that a partition with a Jaccard coefficient of only 0.048 still has a volume-weighted overlap of 0.209; that with new sites at 0.8 of the first site's size the five-site multiplier is only 3.902, and RM 5,000 requires a first site of RM 1,284.09; that a shared core leaves the average loss to downtime unchanged but raises the worst one-per-cent monthly loss from RM 672.47 to RM 1,259.38; and that canary releases cut the monthly loss from bad versions from RM 51.37 to RM 14.66. Chapter 7 examines Sunzi on division, signals and the regular and the unexpected, and Smith on the extent of the market; Chapter 8 gives a 72-hour clone pipeline (a 0.9578 chance of finishing within 72 hours), a weekly ledger, a threshold table, the Five-Site Matrix Cloning Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:做网站生意的人常听到一条流行说法:先把一个站做到日入 RM 1,000,再照样复制五个,就是日入 RM 5,000。本书逐条证伪这句话,再为它划出成立的边界。核心命题是:五个克隆站不会给出五倍收入,关键词与人群的重叠、新站偏小的独有需求、共用母核的相关故障与单站日营收的噪声,各自削掉一截倍数。书中的客单价、转化率、手续费、事故率与各市场参数均为示意,全部计算由书中程序生成;营收数字是示意算术,不是收入承诺,也不构成投资、法律或税务建议;书中只讨论单层、直接的推荐关系,不涉及任何多层分润或返佣。 第一至四章说明,示意参数下 RM 1,000 营收只留下 RM 733.37 的贡献,要贡献 RM 1,000 营收得到 RM 1,364;日营收的变异系数是 0.604,看 7 天均值,真实平均 RM 850 的站有 0.218 的机会显示为千币;负二项序贯检验加至少 28 天、日均不低于 RM 950 的复制门禁,误挑率降到 0.090。单站时母核方案的范围经济是 −0.404,到 1.90 个站才打平,首站成功的概率低于 0.409 时应先建独立站;拷贝副本一年漏掉 23.05 次修复移植。五张候选皮肤按回本规则只开三张,合计每天营收 RM 1,770.00,没有一张接近原站。 第五至八章说明,只数个数的 Jaccard 系数只有 0.048 的切分方案,按搜索量加权的重叠仍有 0.209;新站规模为首站的 0.8 时,五站倍数只有 3.902,要凑齐 RM 5,000,首站得做到 RM 1,284.09。共用母核不改变平均停机损失,最坏百分之一月份的损失却从 RM 672.47 升到 RM 1,259.38;金丝雀发布把坏版本的月损失从 RM 51.37 压到 RM 14.66。第七章考辨《孙子·兵势》的「分数」「形名」「奇正」与斯密「分工受市场范围限制」,第八章给出 72 小时复制流水线(72 小时内完工概率 0.9578)、周度台账、阈值表、五站矩阵复制试算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Ask What a Thousand a Day Leaves: The Minimum Loop, Unit Contribution and Break-Even — Chapters 1 to 4 show that under the illustrative parameters RM 1,000 of revenue leaves RM 733.37 of contribution, and RM 1,000 of contribution needs RM 1,364 of revenue; that daily revenue has a coefficient of variation of 0.604, so a site truly averaging RM 850 has a 0.218 chance of showing a thousand on a 7-day mean; and that a clone gate combining a negative-binomial sequential test with at least 28 days and a mean of at least RM 950 cuts the false-pick rate to 0.090. With one site a central core has scope economies of -0.404, breaks even at 1.90 sites, and should wait whenever the first site succeeds with probability below 0.409; copied forks miss 23.05 ports of fixes a year. Of five candidate skins, the payback rule opens three, with combined revenue of RM 1,770.00 a day, none close to the original site. (https://5dlibrary.com/answers/en/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c01/) 第1章 日入千币先问剩多少:最小闭环、单位贡献与盈亏平衡 — 第一至四章说明,示意参数下 RM 1,000 营收只留下 RM 733.37 的贡献,要贡献 RM 1,000 营收得到 RM 1,364;日营收的变异系数是 0.604,看 7 天均值,真实平均 RM 850 的站有 0.218 的机会显示为千币;负二项序贯检验加至少 28 天、日均不低于 RM 950 的复制门禁,误挑率降到 0.090。单站时母核方案的范围经济是 −0.404,到 1.90 个站才打平,首站成功的概率低于 0.409 时应先建独立站;拷贝副本一年漏掉 23.05 次修复移植。五张候选皮肤按回本规则只开三张,合计每天营收 RM 1,770.00,没有一张接近原站。 (https://5dlibrary.com/answers/zh/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c01/) - Ch.2 A Good Week Is Not a Stable Thousand: Daily Noise, Sequential Tests and the Clone Gate — Under the illustrative parameters, one site's daily revenue has a coefficient of variation of 0.604, so a single day says almost nothing, and a 7-day mean gives a site that truly earns RM 850 a 0.218 chance of showing a thousand. A sequential test written for the negative binomial settles most cases in a little over 20 days with error rates as designed; wrongly assuming Poisson orders pushes the false-pick rate to 0.208. Of the four gates, the sequential test plus 28 days plus a floor of RM 950, with a false-pick rate of 0.090 and a recognition rate of 0.942, is the one this book recommends. (https://5dlibrary.com/answers/en/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c02/) 第2章 七天好成绩不等于站稳千币:日营收噪声、序贯检验与复制门禁 — 单站日营收的变异系数在示意参数下是 0.604,单日几乎不说明任何事;7 天均值会让 RM 850 的站有 0.218 的机会显示为千币。按负二项分布写的序贯检验平均 20 多天就能定案,错误率符合设计;错用泊松假设会把误判率推到 0.208。四种门禁里,序贯检验加 28 天再加 RM 950 底线,误挑率 0.090、认出率 0.942,是本书的建议。 (https://5dlibrary.com/answers/zh/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c02/) - Ch.3 A Central Core Is Not Always Worth Building First: Headless Architecture, Information Hiding and Economies of Scope — A core is not better the earlier it is built. Under the illustrative parameters, the core plan has scope economies of -0.404 with one site, breaks even at 1.90 sites and rises to 0.883 at five; when the first site succeeds with probability below 0.409, it is cheaper to build a standalone site first and extract the core after it succeeds. Copied forks miss 23.05 ports a year, and four copies are almost never all fully patched. Logic about money, identity, consent and risk belongs in the core; content about language and markets stays in the skins. (https://5dlibrary.com/answers/en/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c03/) 第3章 中央母核不是越早越好:无头架构、信息隐藏与范围经济 — 母核不是越早越好。示意参数下,单站时母核方案的范围经济是 −0.404,到 1.90 个站才打平,五个站时升到 0.883;首站成功的概率低于 0.409 时,先建独立站、成功后再抽母核更省钱。拷贝副本一年漏掉 23.05 次移植,四个副本几乎不可能全部补齐。与钱、身份、同意和风险有关的逻辑进母核,与语言和市场有关的内容留在皮肤。 (https://5dlibrary.com/answers/zh/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c03/) - Ch.4 A New Skin Is Not a Copy of the Revenue: Per-Skin Accounts by Language, Country and Niche — A new skin is not a copy of the revenue. Under the illustrative parameters, the five candidate skins bring in between RM 299.52 and RM 600.60 a day, none of them close to the original site; under the rule of "payback within 120 days and at least RM 250 a day at full traffic", three open, with combined revenue of RM 1,770.00 and a net of RM 1,292.38. The best payment mix differs by market: market A needs three methods, market B all five. Translation is not localisation, and a skin that is only machine-translated may also breach search-engine policy. Our advice is to keep accounts skin by skin and to open skin by skin. (https://5dlibrary.com/answers/en/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c04/) 第4章 换一张皮不等于复制一份收入:语种、国别与小众人群的逐皮账 — 换皮不等于复制收入。示意参数下,五张候选皮肤的日营收在 RM 299.52 到 RM 600.60 之间,没有一张接近原站;按「120 天回本、满载每天至少 RM 250」的规则开出三张,合计营收 RM 1,770.00,净额 RM 1,292.38。付款方式的最优组合因市场而异,市场 A 只需三种,市场 B 五种全接。翻译不是本地化,只做机器翻译的皮肤还可能违反搜索引擎政策。我们的建议是逐皮记账,逐皮开张。 (https://5dlibrary.com/answers/zh/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c04/) - Ch.5 Five Sites That Do Not Raid Each Other: Orthogonal Partition of Keywords and Audiences, and the Multiplier — Chapters 5 to 8 show that a partition with a Jaccard coefficient of only 0.048 still has a volume-weighted overlap of 0.209; that with new sites at 0.8 of the first site's size the five-site multiplier is only 3.902, and RM 5,000 requires a first site of RM 1,284.09; that a shared core leaves the average loss to downtime unchanged but raises the worst one-per-cent monthly loss from RM 672.47 to RM 1,259.38; and that canary releases cut the monthly loss from bad versions from RM 51.37 to RM 14.66. (https://5dlibrary.com/answers/en/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c05/) 第5章 五站不互抢:关键词与人群的正交切分和倍数公式 — 第五至八章说明,只数个数的 Jaccard 系数只有 0.048 的切分方案,按搜索量加权的重叠仍有 0.209;新站规模为首站的 0.8 时,五站倍数只有 3.902,要凑齐 RM 5,000,首站得做到 RM 1,284.09。共用母核不改变平均停机损失,最坏百分之一月份的损失却从 RM 672.47 升到 RM 1,259.38;金丝雀发布把坏版本的月损失从 RM 51.37 压到 RM 14.66。 (https://5dlibrary.com/answers/zh/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c05/) - Ch.6 The Price of a Shared Core: Correlated Failure, Blast Radius and the Chargeback Pool — A shared core does not change the average: under the illustrative parameters both architectures lose about RM 291 a month, and each site's availability is 0.99806 in both. What it changes is the tail: the one-per-cent level rises from RM 672.47 to RM 1,259.38, and in 0.696 of months the five sites stop together. Canary releases cut the loss from bad versions from RM 51.37 to RM 14.66 a month at almost no cost; bulkheads, timeouts and circuit breakers keep one failure in one place. The chargeback pool should be divided by risk: settled sites together, the new skin on its own. (https://5dlibrary.com/answers/en/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c06/) 第6章 共用母核的代价:相关故障、爆炸半径与拒付池 — 共用母核不改变平均:示意参数下两种架构的月均损失都在 RM 291 左右,单站可用率都是 0.99806。它改变的是尾部:百分之一分位从 RM 672.47 升到 RM 1,259.38,0.696 的月份五站会一起停。金丝雀发布把坏版本损失从每月 RM 51.37 压到 RM 14.66,几乎不花钱;舱壁、超时与断路器把一处故障关在一处。拒付池要按风险分格,成熟站合、新皮肤分。 (https://5dlibrary.com/answers/zh/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c06/) - Ch.7 East-West Classical Sources: Sunzi on Division and Variation, and Smith on the Extent of the Market — Chapter 7 examines Sunzi on division, signals and the regular and the unexpected, and Smith on the extent of the market; (https://5dlibrary.com/answers/en/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c07/) 第7章 东西方原典考辨:《孙子·兵势》「分数」「奇正」与斯密「分工受市场范围限制」 — 第七章考辨《孙子·兵势》的「分数」「形名」「奇正」与斯密「分工受市场范围限制」,第八章给出 72 小时复制流水线(72 小时内完工概率 0.9578)、周度台账、阈值表、五站矩阵复制试算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c07/) - Ch.8 System Architecture and Operations Spec: The 72-Hour Clone Pipeline, the Weekly Ledger and the Five-Site Matrix Cloning Calculator — The system has three layers, core, skins and ledger, and the eight ledger tables hold summed data only. The critical path of the 72-hour clone pipeline has an expected 53.00 hours; Monte Carlo gives a 0.9578 chance of finishing within 72 hours and only 0.2099 within 48. Under the default conditions the five sites can expect RM 3,893.70 a day, and reaching RM 5,000 requires a first site of RM 1,284.09. The pre-registered tests show that the multiplier needs about 153 days of observation and that cannibalisation can be tested only at the level of search terms. Our advice is: pass the gate first, partition next, clone after that, and write every step into the ledger. (https://5dlibrary.com/answers/en/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c08/) 第8章 系统架构与运营规格:72 小时复制流水线、周度台账与五站矩阵复制试算器 — 系统分母核、皮肤与台账三层,台账八张表只存汇总数据。72 小时复制流水线的关键路径期望 53.00 个小时,蒙特卡洛给出 72 小时内完工的概率 0.9578,48 小时只有 0.2099。默认条件下五站期望 RM 3,893.70,要凑齐 RM 5,000 首站得做到 RM 1,284.09。预注册检验表明,倍数要观察约 153 天,互抢只能在搜索词层面检验。我们的建议是:先过门禁,再切分,再复制,每一步都记在台账上。 (https://5dlibrary.com/answers/zh/thousand-ringgit-single-site-minimum-loop-headless-core-multi-skin-five-site-cloning/c08/) ## 5DL-B-358 · Bootstrapping Industry Accreditation Standards, Spence Signalling Tiers and Inbound Creator Certification Magnetics 中文书名:行业认证标准自举、斯宾塞信号分级与达人主动申领动力学母典 DOI: pending Cite: Fan, Y. (2026). Bootstrapping Industry Accreditation Standards, Spence Signalling Tiers and Inbound Creator Certification Magnetics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its numbers of creators and brands, scoring errors, costs, premiums and audit parameters are illustrative, and all calculations are generated by its programs. It is general business guidance, not individual, legal, tax or investment advice. Abstract: When a private platform sets out to run a certificate that creators apply for of their own accord and that brands actually use as a filter, it inherits three popular beliefs: the lower the pass rate the better; the more holders the greater the success; the scarcer the places the more valuable the certificate. This book refutes each belief or draws its boundary. Its core proposition is that a certificate is valuable only while it separates different types of creator and stays credible. Its numbers of creators and brands, scoring errors, costs, premiums and audit parameters are illustrative, and all calculations are generated by its programs. It is general business guidance, not individual, legal, tax or investment advice. The certificate it designs has public, measurable criteria, a fee that never depends on passing, no suggestion of any government or statutory qualification, revocation by audit and no false scarcity; referral rewards are single-tier and paid only to the direct referrer, and no multi-level profit sharing is involved. Chapters 1 to 3 show that when two types of creator face different costs of meeting a standard there is a whole interval of separating standards, and that a certificate passing anyone who pays never separates anyone; holders admitted by a measurable standard are 1.254 standard deviations above the pool, places by lottery carry no premium, and when half of those who fail buy the certificate only 0.212 remains. With noisy scores, brands will use the certificate at pass rates between 0.18 and 0.23, with the best at 0.22 and a precision of 0.873; once the cost of applying reaches 0.241 two displayed levels beat exact scores, and from 0.315 even pass or fail beats them; grade inflation leads only in the second quarter, and the total premium over nine quarters falls from 305,243 to 134,913. Chapters 4 to 6 show that cross-sectional comparison inflates the holder premium to a ratio of 2.612 when the truth is 1.105; that the opportunity cost of not being certified peaks at 0.446 near a holder share of 0.15 and vanishes at 0.50; and that a certificate without audits settles at a precision of only 0.514, while keeping 0.85 requires re-auditing 0.282 of holders each quarter. Chapter 7 examines the line of the Book of Rites on inscribing the maker's name to examine sincerity and English hallmarking, showing that a named maker has no reason to cut corners once detection reaches 0.067; Chapter 8 gives the application flow, data tables, an append-only public registry, a weekly ledger, the Certification Pass-Rate and Tier Cut Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:私营平台想办一张达人主动申请、品牌方真正拿来筛人的认证时,常听到三条流行说法:录取率越低越好;持证者越多越成功;名额越稀缺越值钱。本书逐条证伪或立界,核心命题是:证书只有在能把不同类型的达人分开、并且一直可信时才有价值。书中的达人人数、品牌数、打分误差、成本、溢价与审计参数均为示意值,全部计算由书中程序生成;本书是非个人化经营建议,不构成法律、税务或投资建议;书中的认证标准公开可测,费用与是否通过无关,不暗示任何政府或法定资格,可以经审计撤销,不制造虚假稀缺,推荐奖励只限单层、只给直接推荐人,不涉及任何多层分润。 第一至三章说明,两类达人达标成本不同时存在一整段分离区间,交钱即过的证书永远分不开人;按可测标准录取的持证者高出 1.254 个标准差,抽签名额毫无溢价,一半落选者买证后只剩 0.212。打分有误差时,品牌方愿意使用的录取率在 0.18 到 0.23 之间,最优 0.22,精度 0.873;申请成本达到 0.241 后两级展示胜过精确分数,达到 0.315 后只分通过与否也胜过精确分数;分数膨胀只在第二季度领先,九个季度的溢价总量从 305,243 降到 134,913。 第四至六章说明,横截面比较把持证溢价夸大为 2.612 倍,真实只有 1.105 倍;不持证的机会成本只在持证者占比 0.15 附近达到 0.446,占比到 0.50 即归零;不审计的证书稳态精度只有 0.514,维持 0.85 每季度须重审 0.282 的持证者。第七章考辨《礼记·月令》的「物勒工名,以考其诚」与英格兰金银器检验印记,刻名者在检查率 0.067 时即无偷工动机;第八章给出申请流程、数据表、只追加的公开登记册、周度台账、认证录取率与分级切点试算器及预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Signalling Equilibrium: When a Certificate Separates Creators and When It Pools Them — Chapters 1 to 3 show that when two types of creator face different costs of meeting a standard there is a whole interval of separating standards, and that a certificate passing anyone who pays never separates anyone; holders admitted by a measurable standard are 1.254 standard deviations above the pool, places by lottery carry no premium, and when half of those who fail buy the certificate only 0.212 remains. With noisy scores, brands will use the certificate at pass rates between 0.18 and 0.23, with the best at 0.22 and a precision of 0.873; once the cost of applying reaches 0.241 two displayed levels beat exact scores, and from 0.315 even pass or fail beats them; grade inflation leads only in the second quarter, and the total premium over nine quarters falls from 305,243 to 134,913. (https://5dlibrary.com/answers/en/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c01/) 第1章 信号均衡:证书什么时候能把达人分开,什么时候会混同 — 第一至三章说明,两类达人达标成本不同时存在一整段分离区间,交钱即过的证书永远分不开人;按可测标准录取的持证者高出 1.254 个标准差,抽签名额毫无溢价,一半落选者买证后只剩 0.212。打分有误差时,品牌方愿意使用的录取率在 0.18 到 0.23 之间,最优 0.22,精度 0.873;申请成本达到 0.241 后两级展示胜过精确分数,达到 0.315 后只分通过与否也胜过精确分数;分数膨胀只在第二季度领先,九个季度的溢价总量从 305,243 降到 134,913。 (https://5dlibrary.com/answers/zh/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c01/) - Ch.2 The Early Pass Rate: Where to Set the Bar When Scores Are Noisy — When scores carry error, a lower pass rate gives higher precision, and on this point the popular belief is right; but recall, the number of people available and the information carried all fall. Program p21 shows that information is greatest when the pass rate is close to the base rate of strong creators, peaking at 0.35 for an error standard deviation of 0.5. Program p22 shows that once the decisions of brands are included, the best pass rate is 0.22, and the range within nine tenths of the best runs only from 0.18 to 0.23. Program p23 shows that after 10 early re-audits with no failure, the upper bound on the false-pass rate is still 0.238, so the platform cannot yet claim any precision. (https://5dlibrary.com/answers/en/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c02/) 第2章 早期录取率:打分有误差时,门槛定在哪里 — 打分有误差时,录取率越低,精度越高,这一点流行说法说对了;可是召回、可用人数与信息量都在下降。p21 说明,信息量在录取率接近强者基础比例时最大,误差标准差 0.5 时峰值在 0.35;p22 说明,加入品牌方的使用决策以后,最优录取率是 0.22,质量提升在九成以内的区间只是从 0.18 到 0.23;p23 说明,早期重审 10 人零失败,误收率的上界仍有 0.238,平台不能据此宣称精度。 (https://5dlibrary.com/answers/zh/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c02/) - Ch.3 Grades: How Many Levels, Where to Cut, and Why Grade Inflation Is Sweet Before It Turns Bitter — The number of levels is set by the precision of scoring. Program p31 shows that with an error standard deviation of 0.3 three levels are best, with 0.5 two levels, and with 1.0 a single level. Program p32 shows that once the cost of applying reaches 0.241 two levels beat exact scores, and once it reaches 0.315 even pass or fail beats them. Program p33 shows that grade inflation is ahead only in the second quarter, and that the total premium over nine quarters falls from 305,243 to 134,913. (https://5dlibrary.com/answers/en/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c03/) 第3章 分级:分几级、切在哪里,以及分数膨胀为什么先甜后苦 — 级数由打分精度决定。p31 说明,误差标准差 0.3 时最好分三级,0.5 时两级,1.0 时只能有一个级别;p32 说明,申请成本达到 0.241 以后两级方案胜过精确分数,达到 0.315 以后连只分通过与否都胜过精确分数;p33 说明,分数膨胀只在第二季度略占上风,九个季度的溢价总量从 305,243 降到 134,913。 (https://5dlibrary.com/answers/zh/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c03/) - Ch.4 The Holder Premium: How Many Inquiries a Badge Brings, and How to Measure It Honestly — Chapters 4 to 6 show that cross-sectional comparison inflates the holder premium to a ratio of 2.612 when the truth is 1.105; that the opportunity cost of not being certified peaks at 0.446 near a holder share of 0.15 and vanishes at 0.50; and that a certificate without audits settles at a precision of only 0.514, while keeping 0.85 requires re-auditing 0.282 of holders each quarter. (https://5dlibrary.com/answers/en/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c04/) 第4章 持证溢价:徽章到底带来多少询价,怎样量才诚实 — 第四至六章说明,横截面比较把持证溢价夸大为 2.612 倍,真实只有 1.105 倍;不持证的机会成本只在持证者占比 0.15 附近达到 0.446,占比到 0.50 即归零;不审计的证书稳态精度只有 0.514,维持 0.85 每季度须重审 0.282 的持证者。 (https://5dlibrary.com/answers/zh/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c04/) - Ch.5 Filtering by Brands and the Opportunity Cost of Not Certifying: When It Is Real and When It Is Exaggerated — The opportunity cost of not being certified is a conditional quantity. Program p51 shows that when holders make up only 0.02 of creators, non-holders lose only 0.067 of their inquiries; the loss is greatest at 0.15, at 0.446; at 0.50 brands no longer filter and the loss is zero; and the range in which the loss exceeds a fifth runs only from 0.10 to 0.30. Program p52 shows that both the cost of applying and naive brands stop unravelling, so that the gain from certification for a creator of quality 0.60 can fall from 0.560 to 0.240 from one market to another. A platform may publish these quantities, but must not use them to press anyone into applying. (https://5dlibrary.com/answers/en/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c05/) 第5章 品牌方的筛选与不持证的机会成本:什么时候真实,什么时候被夸大 — 不持证的机会成本是一个依赖条件的量。p51 说明,持证者占比只有 0.02 时,未持证者只损失 0.067 的询价;在 0.15 时损失最大,0.446;到 0.50,品牌方不再筛人,损失归零;损失超过两成的区间只是 0.10 到 0.30。p52 说明,申请成本与天真的品牌方都会让信息拆解停下来,质量 0.60 的达人在不同市场里的认证收益可以从 0.560 降到 0.240。平台可以公布这些量,但不能拿它们催人申请。 (https://5dlibrary.com/answers/zh/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c05/) - Ch.6 The Badge Display Cascade: Thresholds, Informational Cascades, Audits and Revocation — Adoption depends on the distribution of thresholds across the whole population. Program p61 reproduces Granovetter's example, in which replacing one person in a hundred changes the outcome from everyone acting to one person acting; with a threshold spread of 0.05 the start must be 0.131, with 0.10 only 0.038. Program p62 shows that an informational cascade among brands locks in the wrong direction with probability 0.155 when private signals are correct with probability 0.7, and that an insufficiently accurate audit report pulls the probability of a correct cascade down from 0.845 to 0.750. Program p63 shows that a certificate with no audits has a steady-state precision of only 0.514, and that keeping it at 0.85 requires auditing at least 0.282 of holders each quarter. (https://5dlibrary.com/answers/en/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c06/) 第6章 徽章展示的连锁反应:阈值、信息瀑布与审计撤销 — 普及取决于整个人群的阈值分布。p61 复现了 Granovetter 的例子:一百个人里换掉一个,结局从全员行动变成一人行动;阈值离散程度为 0.05 时起点要 0.131,0.10 时只要 0.038。p62 说明,品牌方之间的信息瀑布在信号正确率 0.7 时有 0.155 的概率锁在错误方向,而一份不够准的审计报告反而会把正确瀑布的概率从 0.845 拉到 0.750。p63 说明,完全不审计的证书稳态精度只有 0.514,要保持 0.85,每季度至少审计 0.282 的持证者。 (https://5dlibrary.com/answers/zh/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c06/) - Ch.7 Classical Sources East and West: "Inscribe the Maker's Name to Examine Sincerity" in the Book of Rites, and English Hallmarking — Chapter 7 examines the line of the Book of Rites on inscribing the maker's name to examine sincerity and English hallmarking, showing that a named maker has no reason to cut corners once detection reaches 0.067; (https://5dlibrary.com/answers/en/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c07/) 第7章 东西方原典考辨:《礼记·月令》「物勒工名,以考其诚」与英格兰金银器检验印记 — 第七章考辨《礼记·月令》的「物勒工名,以考其诚」与英格兰金银器检验印记,刻名者在检查率 0.067 时即无偷工动机; (https://5dlibrary.com/answers/zh/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c07/) - Ch.8 System Architecture and Operating Specification: Application Flow, Public Registry, Weekly Ledger and Pre-Registered Tests — Chapter 8 gives the application flow, data tables, an append-only public registry, a weekly ledger, the Certification Pass-Rate and Tier Cut Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c08/) 第8章 系统架构与运行规格:申请流程、公开登记册、周度台账与预注册检验 — 第八章给出申请流程、数据表、只追加的公开登记册、周度台账、认证录取率与分级切点试算器及预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/industry-accreditation-bootstrapping-spence-signalling-tiers-inbound-creator-certification/c08/) ## 5DL-B-359 · Algorithmic Heterogeneity across RedNote, Threads, X, Instagram and Facebook: Cross-Platform Adaptation and Creator Aggregation 中文书名:五大社交图谱算法异质性、跨平台同源异构分发与达人聚拢母典 DOI: pending Cite: Fan, Y. (2026). Algorithmic Heterogeneity across RedNote, Threads, X, Instagram and Facebook: Cross-Platform Adaptation and Creator Aggregation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Every platform claim is labelled as a dated official disclosure, peer-reviewed research or the book's illustrative model; its weights and funnel parameters are illustrative, not any platform's settings, and all calculations are generated by its programs. It is general business guidance, not legal, tax or investment advice. Abstract: People who run RedNote, Threads, X, Instagram and Facebook at once hear three beliefs: good content wins everywhere, so a post pasted onto five platforms earns five times the exposure; the algorithm counts likes; and growth needs shortcuts such as bought reach and mass messages to recruit creators. This book refutes each belief or draws its boundary. Every platform claim is labelled as a dated official disclosure, peer-reviewed research or the book's illustrative model; its weights and funnel parameters are illustrative, not any platform's settings, and all calculations are generated by its programs. It is general business guidance, not legal, tax or investment advice. It involves no fake accounts, bots, engagement pods, bought engagement, data scraping or unsolicited mass messages; sponsored content is disclosed; creator pay is single-tier and direct, with no multi-level profit sharing. Chapter 1 shows that the five platforms draw candidates from different graphs, that no illustrative post type wins everywhere, and that choosing the type per platform scores 1.496 times one type throughout; in X's 2023 weights a reply weighs 27 times a like and an author-engaged reply 150 times, but weights multiply each viewer's predicted probability, not counts. Chapter 2: a RedNote note aimed at search earns 2.26 times the half-year views of a burst note, long-tail phrases bring 2.80 times the clicks of head phrases, and an informative cover earns 1.50 times the saves of a loud one. Chapter 3: a provocative post turns negative once its probability of negative feedback exceeds 0.0020, and the best and worst posting hours differ by a factor of 20.0. Chapter 4: an Instagram week built on Stories with a few Reels is worth 20.01 against 9.90 for Reels alone, and a month of inbound messages yields 1.18 to 6.77 purchases. Chapter 5: seeds on people who span circles reach 15.36 people against 7.89 for seeds in one circle. Chapter 6 gives the rewriting formula, a fixed core with a surface rewritten per platform: the optimal split of a 10-hour week is worth 71.83, 3.32 times pasting unchanged and only 1.051 times an equal split; when platform parameters are badly misjudged, spread rewriting loses 0.126 on average against 0.352 for a home platform with copies. Chapter 7 examines the Sunzi's "water has no constant shape", Liu Xie's "the form has constants, the variations have no fixed number" and Aristotle's Rhetoric, and shows that chasing noise in a stable setting earns only 0.930 of the oracle. Chapter 8 specifies a content-atom store, platform adapters, a core gate, a creator intake funnel, a weekly ledger, a pre-registered test and the One Source, Five Adaptations and Creator Intake Funnel Calculator, whose illustrative twelve-week run gathers 244.8 consented applications and 23.07 active creators. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:同时经营小红书、Threads、X、Instagram 与 Facebook 的人,常听到三条流行说法:好内容到处都火,一稿原样贴到五个平台就是五倍曝光;算法就是数点赞;要做大就得买量、互推、群发私讯拉达人。本书逐条证伪或立界。书中把每条平台说法标为三类:带日期的官方披露、同行评审研究、本书的示意模型;书中的权重、概率与漏斗参数均为示意,不是任何平台的真实设定,全部计算由书中程序生成。本书是一般性经营参考,不构成法律、税务或投资建议;不涉及假账号、机器人、互赞群、买互动、抓取个人资料或群发私讯;合作内容一律标明;达人酬劳只有单层直付,不涉及任何多层分润。 第一章说明五个平台从五张不同的图挑候选,五种示意帖型没有一种通吃,按平台挑帖型的总分是只用一种的 1.496 倍;X 在 2023 年公开的权重里回复是点赞的 27 倍、作者回应的回复是 150 倍,但权重乘的是每位读者的预测概率,不是累计次数。第二章说明对准搜索需求的小红书笔记半年累计浏览是爆款笔记的 2.26 倍,长尾词组合的点击是大词的 2.80 倍,信息型封面的收藏是抢眼封面的 1.50 倍。第三章说明负面反馈概率超过 0.0020 时挑衅帖子在公开权重下变成负分,最好与最差发帖时刻相差 20.0 倍。第四章说明 Instagram 以快拍为主、辅以少量 Reels 的一周价值 20.01,只做 Reels 是 9.90,私讯漏斗一个月的购买区间是 1.18 到 6.77。第五章说明种子放在跨圈者身上触达 15.36 人,同圈种子只有 7.89 人。 第六章给出一鱼五吃的改写公式:核心不变、表层按平台重写;一周 10 个小时的最优分配价值 71.83,是原样转贴的 3.32 倍,只比平均分配高 1.051 倍;平台参数大幅估错时,分散改写平均损失 0.126,主平台精做加转贴损失 0.352。第七章考辨《孙子兵法·虚实》「兵无常势,水无常形」、《文心雕龙·通变》「体有常,数无方」与亚里士多德《修辞学》,并说明环境不变时追着噪声换形态只拿到先知的 0.930。第八章给出内容原子库、平台适配器、核心检查门、达人入库漏斗、周度台账、预注册检验方案与「一鱼五吃改写与达人入库漏斗试算器」,示意运行十二周积累 244.8 份带同意记录的申请与 23.07 位活跃达人。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Five Graphs, Five Objectives: Interest, Search, Follow, Conversation and Friend Graphs — Chapter 1 shows that the five platforms draw candidates from different graphs, that no illustrative post type wins everywhere, and that choosing the type per platform scores 1.496 times one type throughout; in X's 2023 weights a reply weighs 27 times a like and an author-engaged reply 150 times, but weights multiply each viewer's predicted probability, not counts. (https://5dlibrary.com/answers/en/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c01/) 第1章 五张图,五个目标:兴趣图、搜索图、关注图、对话图与熟人图 — 第一章说明五个平台从五张不同的图挑候选,五种示意帖型没有一种通吃,按平台挑帖型的总分是只用一种的 1.496 倍;X 在 2023 年公开的权重里回复是点赞的 27 倍、作者回应的回复是 150 倍,但权重乘的是每位读者的预测概率,不是累计次数。 (https://5dlibrary.com/answers/zh/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c01/) - Ch.2 RedNote: Search Intent, Saves and Long-Tail Retrieval, and the Arithmetic of Titles, Keywords and Covers — Chapter 2: a RedNote note aimed at search earns 2.26 times the half-year views of a burst note, long-tail phrases bring 2.80 times the clicks of head phrases, and an informative cover earns 1.50 times the saves of a loud one. (https://5dlibrary.com/answers/en/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c02/) 第2章 小红书:搜索意图、收藏与长尾检索,以及标题、关键词与封面的算术 — 第二章说明对准搜索需求的小红书笔记半年累计浏览是爆款笔记的 2.26 倍,长尾词组合的点击是大词的 2.80 倍,信息型封面的收藏是抢眼封面的 1.50 倍。 (https://5dlibrary.com/answers/zh/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c02/) - Ch.3 Threads and X: Reply Chains, Reply Weighting, Early Velocity and Time Decay — Chapter 3: a provocative post turns negative once its probability of negative feedback exceeds 0.0020, and the best and worst posting hours differ by a factor of 20.0. (https://5dlibrary.com/answers/en/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c03/) 第3章 Threads 与 X:回复链、回复权重、早期速度与时间衰减 — 第三章说明负面反馈概率超过 0.0020 时挑衅帖子在公开权重下变成负分,最好与最差发帖时刻相差 20.0 倍。 (https://5dlibrary.com/answers/zh/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c03/) - Ch.4 Instagram: Visual Consistency, Reels, Carousels and Stories, and the Direct-Message Conversion Funnel — Chapter 4: an Instagram week built on Stories with a few Reels is worth 20.01 against 9.90 for Reels alone, and a month of inbound messages yields 1.18 to 6.77 purchases. (https://5dlibrary.com/answers/en/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c04/) 第4章 Instagram:视觉一致性、Reels、组图与快拍,以及私讯转化漏斗 — 第四章说明 Instagram 以快拍为主、辅以少量 Reels 的一周价值 20.01,只做 Reels 是 9.90,私讯漏斗一个月的购买区间是 1.18 到 6.77。 (https://5dlibrary.com/answers/zh/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c04/) - Ch.5 Facebook: Groups, Local Trust and Weak Ties, and Reach through Friend Networks — Chapter 5: seeds on people who span circles reach 15.36 people against 7.89 for seeds in one circle. (https://5dlibrary.com/answers/en/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c05/) 第5章 Facebook:小组、本地信任与弱连接,以及经由朋友网络的触达 — 第五章说明种子放在跨圈者身上触达 15.36 人,同圈种子只有 7.89 人。 (https://5dlibrary.com/answers/zh/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c05/) - Ch.6 One Source, Five Adaptations: the Rewriting Formula, Allocation under a Time Budget and Robustness to Unknown Weights — Chapter 6 gives the rewriting formula, a fixed core with a surface rewritten per platform: the optimal split of a 10-hour week is worth 71.83, 3.32 times pasting unchanged and only 1.051 times an equal split; when platform parameters are badly misjudged, spread rewriting loses 0.126 on average against 0.352 for a home platform with copies. (https://5dlibrary.com/answers/en/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c06/) 第6章 一鱼五吃:改写公式、时间预算下的分配与对未知权重的稳健性 — 第六章给出一鱼五吃的改写公式:核心不变、表层按平台重写;一周 10 个小时的最优分配价值 71.83,是原样转贴的 3.32 倍,只比平均分配高 1.051 倍;平台参数大幅估错时,分散改写平均损失 0.126,主平台精做加转贴损失 0.352。 (https://5dlibrary.com/answers/zh/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c06/) - Ch.7 Eastern and Western Sources Examined: the Sunzi on Water Without Constant Shape, Liu Xie on Continuity and Change, and Aristotle's Rhetoric — Chapter 7 examines the Sunzi's "water has no constant shape", Liu Xie's "the form has constants, the variations have no fixed number" and Aristotle's Rhetoric, and shows that chasing noise in a stable setting earns only 0.930 of the oracle. (https://5dlibrary.com/answers/en/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c07/) 第7章 东西方原典考辨:《孙子兵法·虚实》「水无常形」、《文心雕龙·通变》与亚里士多德《修辞学》 — 第七章考辨《孙子兵法·虚实》「兵无常势,水无常形」、《文心雕龙·通变》「体有常,数无方」与亚里士多德《修辞学》,并说明环境不变时追着噪声换形态只拿到先知的 0.930。 (https://5dlibrary.com/answers/zh/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c07/) - Ch.8 System Architecture and Operations Specification: Content-Atom Store, Platform Adapters, Creator Intake Funnel, Weekly Ledger and Pre-Registered Test — Chapter 8 specifies a content-atom store, platform adapters, a core gate, a creator intake funnel, a weekly ledger, a pre-registered test and the One Source, Five Adaptations and Creator Intake Funnel Calculator, whose illustrative twelve-week run gathers 244.8 consented applications and 23.07 active creators. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c08/) 第8章 系统架构与运营规格:内容原子库、平台适配器、达人入库漏斗、周度台账与预注册检验 — 第八章给出内容原子库、平台适配器、核心检查门、达人入库漏斗、周度台账、预注册检验方案与「一鱼五吃改写与达人入库漏斗试算器」,示意运行十二周积累 244.8 份带同意记录的申请与 23.07 位活跃达人。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/five-social-graph-algorithm-heterogeneity-cross-platform-adaptation-creator-aggregation/c08/) ## 5DL-B-360 · Mission-Driven Co-Creation Networks, Paid-Seat Identity Commitment and Voluntary Contribution Flywheels 中文书名:使命感召型共创网络、付费席位身份沉没与志愿贡献飞轮母典 DOI: pending Cite: Fan, Y. (2026). Mission-Driven Co-Creation Networks, Paid-Seat Identity Commitment and Voluntary Contribution Flywheels (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All parameters are illustrative, all calculations come from its programs, and revenue figures are illustrative arithmetic, not forecasts or guarantees. It is general guidance, not legal, tax or investment advice. Abstract: Community builders inherit three popular beliefs: founding seats sell on their perks; the deeper members invest, the more loyal they become, so they should keep raising their stake; and paying volunteers either makes them keener or destroys their enthusiasm. This book refutes each belief or draws its boundary. All parameters are illustrative, all calculations come from its programs, and revenue figures are illustrative arithmetic, not forecasts or guarantees. It is general guidance, not legal, tax or investment advice. A paid seat here is a membership and recognition product, never an investment: it promises no return and states its refund and exit rules. Referral rewards go to one direct referrer only; there is no multi-level profit sharing or downline. The word gongde, merit, is used only as naming and cultural background, with no promise of any spiritual return. Chapters 1 and 2 show that in an illustrative community only 13 members would pay RM 1,000 for a seat on its perks alone, while public recognition and a transparent use of funds raise the number to 1,786, with 0.807 of buyers' willingness to pay coming on average from recognition and mission; a cap is a promise to holders, not a revenue tool. The IKEA effect holds only on completion: the valuation of a first task in one block peaks at 4.24 hours, and with the member's own time counted the task should take only 2.66 hours. Renewal letters that recite sunk investment, combined with automatic renewal, make 0.364 of renewals come from members who regret them; a cooling-off period, a usage statement and opt-in renewal push the share down to 0.040. Chapters 3 to 6 show that four plaque tiers raise total gifts by a factor of 1.163; that an honour planned for 5 % ends up held by 0.209 and raises contribution hours the most; that a small public payment lowers participation from 0.615 to 0.433, and that 0.230 of the gain a reward brings is moved from a nearby project; that gentle invitation has more holders from month 35 and more cumulative revenue from month 47, while a merged points ledger fills 0.960 of the honour roll with payers; and that votes by money let payers pass almost every proposal most members oppose, while a cap of a tenth on payers' weight brings capture down to 0.070, the refund reserve is set at the 99th percentile, 0.182 of the fund, and single-tier referral rewards come to 0.084 of kept sales. Chapter 7 examines the Record of Rites, the Platform Sutra, Mauss's essay on the gift and Franklin's subscription library, whose dues came to 12.5 times its founding fund, and shows that 7 days between a gift and an invitation cut obligation-driven purchases to 0.170; Chapter 8 gives a system architecture, a weekly ledger, the Dual-Track Funnel and Seat Ratchet Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. 摘要:做社群、开放项目与知识平台的人,常听到三条流行说法:创始席位靠权益卖出,权益越多越好卖;成员投入越深越忠诚,所以要让他们不断加码;给志愿者付钱只会让他们更积极,或者反过来,一给钱就毁掉热情。本书逐条证伪或立界。书中的成员数、价格、续费率、退款率与时数均为示意参数,全部计算由书中程序生成,收入数字只是示意算术,不是收益预测或保证;本书是非个人化经营建议,不构成法律、税务或投资建议。付费席位在书中只是会员与署名产品,不是投资,不承诺回报,写明冷静期退款与退出规则;推荐奖励只给一位直接推荐人,不涉及任何多层分润、下线或团队提成;「功德」只作命名与文化背景,不附带任何精神或超自然回报。 第一、二章说明,单凭权益愿为 RM 1,000 席位付钱的示意成员只有 13 位,加上公开署名与透明用途升到 1,786 位,买家愿付价格平均 0.807 来自署名与使命;限量是对持有人的承诺,不是增收工具。宜家效应只在完成时成立,整块入门任务的估值在 4.24 个小时见顶,按成员自身时间成本只该做 2.66 个小时;续费邮件背诵沉没投入加上自动续费,让 0.364 的续费出自后悔者,冷静期、使用报告与主动续费把这一比例压到 0.040。 第三至六章说明,四档名牌让捐款合计升到 1.163 倍,荣誉计划发给 5 % 时最终持有 0.209、贡献时数增幅最大;公开场合的小额报酬让参与率从 0.615 跌到 0.433,奖励带来的增加里有 0.230 是从邻近项目挪来的;温和邀请从第 35 个月起持有人更多、第 47 个月起累计收入更高,合并积分账让付费者占满荣誉榜的 0.960;按出钱分票几乎让付费者通过所有多数反对的提案,付费者权重封顶十分之一时俘获率降到 0.070,退款准备金按第 99 百分位留出基金的 0.182,单层推荐奖励占保留销售额的 0.084。第七章考辨《礼记·礼运》、《六祖坛经》、莫斯《论礼物》与富兰克林的会员图书馆,全部会费是创始基金的 12.5 倍,赠品与邀请相隔 7 天让义务驱动的购买降到 0.170;第八章给出九部件系统架构、周度台账、共创双轨漏斗与席位棘轮试算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A Seat Is Not Sold on Perks: How Much of Willingness to Pay Is Identity, Recognition and Mission — Chapters 1 and 2 show that in an illustrative community only 13 members would pay RM 1,000 for a seat on its perks alone, while public recognition and a transparent use of funds raise the number to 1,786, with 0.807 of buyers' willingness to pay coming on average from recognition and mission; a cap is a promise to holders, not a revenue tool. The IKEA effect holds only on completion: the valuation of a first task in one block peaks at 4.24 hours, and with the member's own time counted the task should take only 2.66 hours. Renewal letters that recite sunk investment, combined with automatic renewal, make 0.364 of renewals come from members who regret them; a cooling-off period, a usage statement and opt-in renewal push the share down to 0.040. (https://5dlibrary.com/answers/en/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c01/) 第1章 席位卖的不是权益:身份、署名与使命各占多少愿付价格 — 第一、二章说明,单凭权益愿为 RM 1,000 席位付钱的示意成员只有 13 位,加上公开署名与透明用途升到 1,786 位,买家愿付价格平均 0.807 来自署名与使命;限量是对持有人的承诺,不是增收工具。宜家效应只在完成时成立,整块入门任务的估值在 4.24 个小时见顶,按成员自身时间成本只该做 2.66 个小时;续费邮件背诵沉没投入加上自动续费,让 0.364 的续费出自后悔者,冷静期、使用报告与主动续费把这一比例压到 0.040。 (https://5dlibrary.com/answers/zh/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c01/) - Ch.2 The IKEA Effect and the Ratchet of Sunk Investment: Effort Raises Value, and Also Breeds Regret — The IKEA effect is real, but what carries value is finishing, not working. In the illustrative model of p21, the valuation of a first task in one block peaks at 4.24 hours, and with the member's own time deducted the task should take only 2.66 hours; cut into milestones, net value rises from 106.8 to 111.4 points. The ratchet of sunk cost is real too: in p22, almost all of the extra money the ratchet design collects over four years comes from 5,610 regretful renewals, and the three defences of a cooling-off period, a transparent use of funds and opt-in renewal push the share of regretful renewals from 0.364 down to 0.040. (https://5dlibrary.com/answers/en/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c02/) 第2章 宜家效应与沉没投入的棘轮:投入抬高估值,也会制造后悔 — 宜家效应是真的,但值钱的是做成,而不是做。p21 的示意模型里,整块入门任务的估值在 4.24 个小时达到高点,扣掉成员自己的时间只该做 2.66 个小时;切成里程碑以后,净值从 106.8 分升到 111.4 分。沉没成本的棘轮也是真的,p22 里棘轮设计四年多收的钱,几乎全部来自 5,610 次后悔续费;冷静期、透明用途与主动续费三道防线,把后悔续费的比例从 0.364 压到 0.040。 (https://5dlibrary.com/answers/zh/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c02/) - Ch.3 Honours That Cost No Money: Honour Ladders, Public Plaques and the Saturation of Honours — Chapters 3 to 6 show that four plaque tiers raise total gifts by a factor of 1.163; that an honour planned for 5 % ends up held by 0.209 and raises contribution hours the most; that a small public payment lowers participation from 0.615 to 0.433, and that 0.230 of the gain a reward brings is moved from a nearby project; that gentle invitation has more holders from month 35 and more cumulative revenue from month 47, while a merged points ledger fills 0.960 of the honour roll with payers; and that votes by money let payers pass almost every proposal most members oppose, while a cap of a tenth on payers' weight brings capture down to 0.070, the refund reserve is set at the 99th percentile, 0.182 of the fund, and single-tier referral rewards come to 0.084 of kept sales. (https://5dlibrary.com/answers/en/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c03/) 第3章 不花钱的荣誉:荣誉阶梯、公开名牌与荣誉的饱和 — 第三至六章说明,四档名牌让捐款合计升到 1.163 倍,荣誉计划发给 5 % 时最终持有 0.209、贡献时数增幅最大;公开场合的小额报酬让参与率从 0.615 跌到 0.433,奖励带来的增加里有 0.230 是从邻近项目挪来的;温和邀请从第 35 个月起持有人更多、第 47 个月起累计收入更高,合并积分账让付费者占满荣誉榜的 0.960;按出钱分票几乎让付费者通过所有多数反对的提案,付费者权重封顶十分之一时俘获率降到 0.070,退款准备金按第 99 百分位留出基金的 0.182,单层推荐奖励占保留销售额的 0.084。 (https://5dlibrary.com/answers/zh/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c03/) - Ch.4 Does Money Crowd Out Enthusiasm? The Boundaries of Crowding Out — Paying is neither always better nor always worse. In the model of p41, participation in a public setting falls from 0.615 to 0.433 under a small payment, and the payment must reach 1.16 before participation returns to where it began; in a private setting, it rises almost monotonically. Program p42 shows that the option to pass the payment on brings participation in the image-conscious group back from 0.441 to 0.552, and that of the gross gain of 392 people brought by a reward, 0.230 were moved from a nearby project. Whether to pay depends on the list of boundary conditions: visibility, amount, how the payment is tied, the nature of the work and the option to pass it on. (https://5dlibrary.com/answers/en/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c04/) 第4章 给钱会不会挤掉热情:挤出效应的边界 — 给钱既不总是更好,也不总是更坏。p41 的模型里,公开场合的参与率在小额报酬下从 0.615 跌到 0.433,报酬要到 1.16 才回到原点;私下场合则基本单调上升。p42 显示,转赠选项把看重形象的人群的参与率从 0.441 拉回到 0.552;奖励带来的 392 人毛增加里,有 0.230 是从邻近项目挪过来的。是否付钱,要看边界清单:可见性、金额、挂钩方式、工作性质与转赠选项。 (https://5dlibrary.com/answers/zh/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c04/) - Ch.5 The Dual-Track Funnel: Free Contributors, Paid Seats and Fairness Between the Tracks — The free track is not a lobby for the paid track. Lakhani and von Hippel show that free contribution rests on a low marginal cost and on contributors' own benefits. Program p51 shows that a hard push sells more seats in the first year, is overtaken by gentle invitation from month 35, and after five years has 446 fewer holders and only a little more than half as many active contributors; program p52 shows that 1 invitation a year is best, and that raising it to 6 loses 0.323 of contribution hours; program p53 shows that a merged points ledger fills 0.960 of the honour roll with payers. Never merging the two ledgers is the floor of fairness between the tracks. (https://5dlibrary.com/answers/en/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c05/) 第5章 双轨漏斗:免费贡献者、付费席位与两条轨道之间的公平 — 免费轨道不是付费轨道的前厅。Lakhani 与 von Hippel 说明,免费贡献靠的是低边际成本与贡献者自身的收益。p51 显示,硬推在头一年多卖席位,第 35 个月起让温和邀请反超,五年后少了 446 位持有人,活跃贡献者只剩温和邀请时的一半多一点;p52 显示,一年邀请 1 次最合适,提高到 6 次会损失 0.323 的贡献时数;p53 显示,合并积分账让付费者占满荣誉榜的 0.960。两本账永不合并,是双轨公平的底线。 (https://5dlibrary.com/answers/zh/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c05/) - Ch.6 The Community Governance Contract: Design Principles, Protection Against Capture by Payers, Refund Reserves and Single-Tier Referral — Ostrom's eight design principles can be written directly into a community contract: clear boundaries, rules that fit conditions, members taking part in changing the rules, independent monitoring, graduated sanctions, low-cost resolution of disputes, outside recognition and nested governance. Program p61 shows that allocating votes by money lets payers pass 0.999 of the proposals most members oppose, that a cap at a third still leaves 0.715, and that only a cap at a tenth or one member one vote suppresses capture. Program p62 shows that the refund reserve should be set at the 99th percentile, 0.182 of the fund. Program p63 shows that a single-tier referral ledger has only one level, that all rewards come to 0.084 of kept sales, and that a reward depends only on what was sold. (https://5dlibrary.com/answers/en/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c06/) 第6章 社群治理契约:设计原则、防止付费者俘获、退款准备金与单层推荐 — 奥斯特罗姆的八条设计原则可以直接写成社群契约:边界清楚,规则相符,成员参与修改,独立监督,分级制裁,低成本纠纷解决,外部承认,嵌套治理。p61 显示,按出钱分票会让 0.999 的多数反对提案由付费者表决通过,封顶三分之一仍有 0.715,只有封顶十分之一或一人一票才把俘获压下去。p62 显示,退款准备金要按第 99 百分位留,即基金的 0.182。p63 显示,单层推荐台账里只有一层,全部奖励占保留销售额的 0.084,奖励只取决于卖出了什么。 (https://5dlibrary.com/answers/zh/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c06/) - Ch.7 East-West Classical Sources: The Record of Rites on the Great Unity, the Platform Sutra, Mauss's Essay on the Gift and Franklin's Subscription Library — Chapter 7 examines the Record of Rites, the Platform Sutra, Mauss's essay on the gift and Franklin's subscription library, whose dues came to 12.5 times its founding fund, and shows that 7 days between a gift and an invitation cut obligation-driven purchases to 0.170; (https://5dlibrary.com/answers/en/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c07/) 第7章 东西方原典考辨:《礼记·礼运》、《六祖坛经》、莫斯《论礼物》与富兰克林的会员图书馆 — 第七章考辨《礼记·礼运》、《六祖坛经》、莫斯《论礼物》与富兰克林的会员图书馆,全部会费是创始基金的 12.5 倍,赠品与邀请相隔 7 天让义务驱动的购买降到 0.170; (https://5dlibrary.com/answers/zh/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c07/) - Ch.8 System Architecture and the Operating Loop: Registries, Ledgers, the Honour Engine, the Seat Lifecycle and Pre-Registered Tests — Chapter 8 gives a system architecture, a weekly ledger, the Dual-Track Funnel and Seat Ratchet Calculator and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (18 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c08/) 第8章 系统架构与运营闭环:登记、台账、荣誉引擎、席位生命周期与预注册检验 — 第八章给出九部件系统架构、周度台账、共创双轨漏斗与席位棘轮试算器与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(18 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/mission-driven-co-creation-paid-seat-identity-commitment-voluntary-contribution-flywheels/c08/) ## 5DL-B-361 · Cross-Lingual Command Channel Distortion, High-Low Context Alignment and Zero-Ambiguity Multinational Operations 中文书名:跨语种指令信道失真、高低语境语义对齐与跨国团队零误解作战母典 DOI: pending Cite: Fan, Y. (2026). Cross-Lingual Command Channel Distortion, High-Low Context Alignment and Zero-Ambiguity Multinational Operations (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. This is general operating analysis, not legal or human-resources advice, and it labels no individual by nationality, ethnicity, religion or language. Abstract: Three pieces of folk wisdom govern dispatch in multinational teams: hire a good translator and say things clearly and nothing will go wrong; people from high-context cultures hear what is left unsaid, so sort dispatches by nationality; once the other side has replied "OK", "好的" or "Baik", the instruction is confirmed. This book refutes each of them or draws its boundary. An instruction passes a language gate and a cultural filter, which form a cascade of channels; the data processing inequality guarantees that a cascade can only lose information, so no translator can win back what the cultural filter removes. After checking the original papers the book corrects its blueprint in several places listed in the preface; for instance, readback errors are rare per transmission, yet readback, hearback and missing readbacks are major categories among reported errors. Parameters are illustrative and all calculations are generated by the book's programs. This is general operating analysis, not legal or human-resources advice, and it labels no individual by nationality, ethnicity, religion or language. Chapter 1 writes a dispatch as a cascade of channels and quantifies the pull of habitual practice and the sender's overconfidence. Chapter 2 writes a dispatch as the receiver's set of possible actions, 3,888 of them when seven illustrative fields are left open, and shows how a four-field dispatch (Goal, Boundary, Acceptance, Counter-example) cuts its conditional entropy. Chapter 3 builds a readback protocol: residual misunderstanding equals the misread rate times the combined chance of concealment and a missed hearback; a bare "OK" carries 0.0016 bits while an own-words readback carries 0.2240 bits, and under illustrative parameters the misreadings that slip through fall below a quarter of those under one-word confirmation, by a factor of 4.17. Chapter 4 builds a multilingual term-lock table, finds Vietnamese words that collide without tone marks and false friends across varieties, and gives the table a version fingerprint. Chapter 5 uses the overlapping coefficient of two normal distributions to show that labelling people by nationality is wrong about 0.3493 of the time; in an illustrative experiment dispatching by nationality is misread 0.1371 of the time and dispatching by four-field structure 0.0500. Chapter 6 computes the break-even between rework and writing and sets a triage threshold for each of the AIX, HOMIE-X and MPE business lines. Chapter 7 reads the chain of rectified names in the Analects, Sunzi's two rounds of drill in the Records of the Grand Historian, Babel in Genesis and Clausewitz on friction. Chapter 8 delivers standard operating templates and a checker specification, the distilled Four-Field Dispatch and Readback Checker, and a pre-registered refutation plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:跨国团队派单有三句流行说法:找个好翻译、把话说清楚就不会做偏;高语境的人听得懂弦外之音,按国籍分类派单即可;对方回了「OK」「好的」「Baik」就算确认。本书逐条证伪或立界。指令要过语种转写与文化过滤两道关,二者是串联信道,数据处理不等式保证串联只会更差,好翻译补不回文化那一道损耗;发令者又系统高估自己被听懂的程度。核对原文之后,本书做了蓝图修正:说话者高估的实验是实验室朗读任务,只借方向不借幅度;接收方同样高估自己读懂;国家之间的方差占比取自工作价值观的元分析,用作语境维度的类比;读回错误按单次通话并不频繁,但在已报告的差错里,读回、听回与缺失读回是大类。书中参数为示意,全部计算由书中程序生成。本书是非个人化经营建议,不是法律或人力资源建议,不按国籍、民族、宗教或语言给任何个人贴标签。 第一章把派单写成串联信道,量化默认做法的拉偏与发令者的自信偏差。第二章把派单写成执行方的动作集合,七个示意字段全空时有 3,888 种做法,四格派单(目标、边界、验收、反例)逐格削减条件熵。第三章建立回述确认协议:残余误解等于读错率乘以隐藏与听回漏检的合成概率,一句「OK」只携带 0.0016 比特,自述复述携带 0.2240 比特,示意参数下溜走的误解降到单字确认的四分之一以下,倍数 4.17。第四章做多语种术语锁定表,找出去掉声调后撞车的越南文词与跨地区同形异义词,并给锁定表加版本指纹。 第五章用两个正态分布的重叠系数说明,按国籍给人贴标签的错误率约为 0.3493,示意实验里按国籍派单读错率 0.1371,按四格结构派单 0.0500,并演示生态谬误。第六章算返工与书写的盈亏平衡,给 AIX、HOMIE-X、MPE 三条业务线各定分诊阈值。第七章考辨《论语·子路》的正名链、《史记》孙子教战的两轮问责、《创世记》巴别塔与克劳塞维茨论摩擦。第八章交付标准作业模板与检查器规格、蒸馏小工具「四格派单与回述确认检查器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Channels in Series: Why a Good Translator Cannot Win Back the Cultural Loss — On average the senders expect 0.8659 of their dispatches to be executed as intended; the realised share is 0.6674, a gap of 0.1985. The bias appears in every band of the sender's estimate. The batch in the highest band, where the mean estimate is 0.9755, is in fact carried out correctly only 0.8476 of the time. Senders rated 6,729 dispatches as clear, and 1,350 of those were still misread, a share of 0.2006. (https://5dlibrary.com/answers/en/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c01/) 第1章 串联信道:好翻译为什么补不回文化那一道损耗 — 发令者平均以为 0.8659 的派单会被正确执行,实际只有 0.6674,差了 0.1985。按估计值分段看,偏差在每一段都存在。估计值落在最高一档、平均 0.9755 的那批派单,实际做对的只有 0.8476。发令者自评「很清楚」的派单共 6,729 张,其中 1,350 张照样做偏,比例 0.2006。 (https://5dlibrary.com/answers/zh/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c01/) - Ch.2 The Four-Field Dispatch: Cutting the Receiver's Action Set with Conditional Entropy — Chapter 2 writes a dispatch as the receiver's set of possible actions, 3,888 of them when seven illustrative fields are left open, and shows how a four-field dispatch (Goal, Boundary, Acceptance, Counter-example) cuts its conditional entropy. (https://5dlibrary.com/answers/en/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c02/) 第2章 四格派单:用条件熵削掉执行方的动作集合 — 第二章把派单写成执行方的动作集合,七个示意字段全空时有 3,888 种做法,四格派单(目标、边界、验收、反例)逐格削减条件熵。 (https://5dlibrary.com/answers/zh/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c02/) - Ch.3 Readback Confirmation: the Readback Error Rate Times the Hearback Miss Rate — Chapter 3 builds a readback protocol: residual misunderstanding equals the misread rate times the combined chance of concealment and a missed hearback; a bare "OK" carries 0.0016 bits while an own-words readback carries 0.2240 bits, and under illustrative parameters the misreadings that slip through fall below a quarter of those under one-word confirmation, by a factor of 4.17. (https://5dlibrary.com/answers/en/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c03/) 第3章 回述确认:读回错误率乘听回漏检率 — 第三章建立回述确认协议:残余误解等于读错率乘以隐藏与听回漏检的合成概率,一句「OK」只携带 0.0016 比特,自述复述携带 0.2240 比特,示意参数下溜走的误解降到单字确认的四分之一以下,倍数 4.17。 (https://5dlibrary.com/answers/zh/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c03/) - Ch.4 The Term-Lock Table: Collisions without Tone Marks, False Friends across Varieties and a Version Fingerprint — The old tag is e9c30e1b. The new version changes only the entry for the deposit, and its tag is completely different; recomputing the old table gives the same tag again. Among four hundred illustrative open dispatches, 128 still carry the old tag, a share of 0.320. Of these, 52 also use the edited term, a share of 0.130. Re-confirming only those 52 instead of all 128 old-tag dispatches saves 0.594 of the work. (https://5dlibrary.com/answers/en/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c04/) 第4章 术语锁定表:去掉声调的撞车、跨地区的同形异义与版本指纹 — 旧版标签是 e9c30e1b,新版只改了「定金」一条,标签就完全不同;同一张表重算一次,标签不变。示意的四百张未结派单里,128 张还挂着旧标签,比例 0.320。其中同时用到改过那条术语的有 52 张,比例 0.130。只对这 52 张重新确认,而不是把旧标签的 128 张全部重来,可以省下 0.594 的工作量。 (https://5dlibrary.com/answers/zh/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c04/) - Ch.5 High and Low Context: the Overlap of Country Means and the Ecological Fallacy — Chapter 5 uses the overlapping coefficient of two normal distributions to show that labelling people by nationality is wrong about 0.3493 of the time; in an illustrative experiment dispatching by nationality is misread 0.1371 of the time and dispatching by four-field structure 0.0500. (https://5dlibrary.com/answers/en/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c05/) 第5章 高低语境:国别均值的重叠系数与生态谬误 — 第五章用两个正态分布的重叠系数说明,按国籍给人贴标签的错误率约为 0.3493,示意实验里按国籍派单读错率 0.1371,按四格结构派单 0.0500,并演示生态谬误。 (https://5dlibrary.com/answers/zh/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c05/) - Ch.6 Breaking Even between Rework and Writing: a Triage Threshold for Each Business Line — Cost must be counted for both sides together, and a message must be complete while each field stays short. With four fields and a readback the residual misread rate is 0.0120 at an extra 8.5 minutes per dispatch; at an ordinary misread rate of 0.10, the protocol pays whenever rework exceeds 1.6098 hours. The readback alone has the lowest break-even and should be the default for cross-language dispatches, while four-field rewrites are triaged by dispatch score. Under illustrative parameters the best threshold is 7 for AIX, readbacks only for HOMIE-X and 6 for MPE, with weekly net savings of 34.65, 14.07 and 14.59 hours respectively. The threshold curve is flat near its optimum, so precision in setting it matters little. (https://5dlibrary.com/answers/en/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c06/) 第6章 返工与书写的盈亏平衡:给每条业务线定分诊阈值 — 成本要按双方合计来算,消息要全但每格要短。四格加回述的残余读错率是 0.0120,每单多花 8.5 分钟;平常读错率 0.10 时,返工超过 1.6098 小时即划算。只回述的盈亏平衡点最低,应当成为跨语种派单的默认动作;四格重写按派单得分分诊。示意参数下,AIX 的最佳阈值是 7,HOMIE-X 是全部只回述,MPE 是 6,三条线每周分别净省 34.65、14.07、14.59 小时。阈值曲线在最优点附近很平,不必算得很精。 (https://5dlibrary.com/answers/zh/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c06/) - Ch.7 Classical Sources East and West: Rectified Names, a Drill, Babel and Friction — The chain of names, the two rounds of blame, friction and Babel each capture one side of cross-language dispatch, and each turns a probabilistic relation into a necessity. The programs translate them into testable numbers: little information survives five links in series; locking terms raises the probability that a six-term dispatch is read correctly throughout from 0.7828 to 0.9704; a job of twelve hand-offs succeeds only 0.6938 of the time; and after a matching readback the probability that the dispatch was still at fault falls to 0.2707. The ancients supplied the direction, and today's ledger supplies the scale. (https://5dlibrary.com/answers/en/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c07/) 第7章 东西方原典考辨:正名、教战、巴别与摩擦 — 正名链、两轮问责、摩擦与巴别塔,各自说中了跨语种派单的一个侧面,也各自把概率关系写成了必然。程序把它们翻成可以检验的数字:五环串联后信息只剩一小部分,术语锁定让六个术语的派单全部读对的概率从 0.7828 升到 0.9704,十二次交接的成功率只有 0.6938,回述一致之后派单仍有问题的概率降到 0.2707。古人的洞见给了方向,今天的台账给出刻度。 (https://5dlibrary.com/answers/zh/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c07/) - Ch.8 Standard Operating Templates and Checker Specification: the Weekly Ledger, the Four-Field Checker and a Pre-Registered Refutation — Chapter 8 delivers standard operating templates and a checker specification, the distilled Four-Field Dispatch and Readback Checker, and a pre-registered refutation plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c08/) 第8章 标准作业模板与检查器规格:周度台账、四格派单检查器与预注册证伪 — 第八章交付标准作业模板与检查器规格、蒸馏小工具「四格派单与回述确认检查器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/cross-lingual-command-distortion-high-low-context-zero-ambiguity-operations/c08/) ## 5DL-B-362 · Eliminating Attention Residue in Multi-Arena Command, Ultradian Energy Pacing and Rapid Context-Switching Protocols 中文书名:多战线指挥官注意力残留清零、超日节律能量调配与跨场景秒级切换母典 DOI: pending Cite: Fan, Y. (2026). Eliminating Attention Residue in Multi-Arena Command, Ultradian Energy Pacing and Rapid Context-Switching Protocols (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. This is not medical, sleep, nutritional or mental-health advice, and it claims no health effect. Abstract: One person running several lines of business hears three popular sentences about mental pacing: switching costs only seconds, so switch freely; people work in ninety-minute ultradian cycles and must stop when a cycle ends; willpower is like blood sugar, so sugar keeps the mind at its peak. After checking the sources, this book refutes or bounds each one. The direct cost measured in task-switching experiments is indeed a few hundred milliseconds, but with sixteen switches a day it is only 0.270 % of the total cost of switching; the bulk is attention residue left by unfinished work. Ninety minutes comes from the sleep cycle, and the waking evidence is weak; the deliberate-practice literature gives a daily ceiling of about four hours, not a pendulum. The ego-depletion effect behind the glucose account had a pooled effect size of only 0.04 in a preregistered replication across twenty-three laboratories. The book makes no claim about any food, supplement or caffeine. Parameters are illustrative and all calculations are generated by the book's programs. This is not medical, sleep, nutritional or mental-health advice, and it claims no health effect. Chapter 1 separates the two ledgers of switching cost and simulates a five-arena day with a fixed seed. Chapter 2 derives a residue decay equation from an activation model of goal memory, with time constants of about 22.33, 8.68 and 6.45 minutes for open, noted and closed exits, and a design for estimating one's own τ from a log. Chapter 3 solves for the optimal sprint length, set by the alertness decay constant and the switching cost and close to the square root of twice their product; across more than two thousand parameter combinations a fixed ninety minutes is near the optimum in a share of only 0.450. Chapter 4 takes the three-minute reset apart: on the residue ledger a single written next-step line nets 1.77 minutes per switch, the full reset only 0.67. Chapter 5 draws an alertness curve from the two-process model, marks a personal high-alertness window and makes "sleep length is never compressed" a precondition of scheduling; greedily filling the window with hard matters loses at worst 2.56 % against dynamic programming. Chapter 6 builds a load budget for the decision day: irreversible decisions count double, a switch cap follows from a residue budget, and batching the five arenas cuts residue from 77.1 minutes to 11.3. Chapter 7 examines the Xunzi, the Zhuangzi, Seneca's second letter and William James on attention, and turns them into a distillation spec for four kits. Chapter 8 gives standard operating templates, the checker specification of the distilled Switch-Residue Budget and Sprint Scheduler, and a preregistered test whose main hypothesis needs 250 switches, about 6.2 weeks. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:一人统帅多条战线时,脑力管理有三句流行说法:切换只损失几秒,多切无妨;人天生按九十分钟的超日节律工作,到点必须休息;意志力像血糖,多吃糖就能撑满全天。本书逐条核对原文后证伪或立界。任务切换实验测到的直接代价确是几百毫秒,但一天切十六回时,直接代价只占切换总成本的 0.270 %,大头是上一件事没收尾留下的注意力残留。九十分钟源于睡眠周期,清醒期的证据并不稳固;刻意练习研究给出的是每天约四小时的总量上限,不是钟摆。葡萄糖假说依附的自我损耗效应,在二十三个实验室的预注册复现里合并效应量只有 0.04。本书不写任何饮食、补充剂或咖啡因的功效。书中参数为示意,全部计算由书中程序生成。本书不是医疗、睡眠、营养或心理健康建议,不宣称任何功效。 第一章拆开切换代价的两本账,并用固定种子模拟五条战线的一日。第二章由目标记忆的激活模型推出残留衰减方程:敞口离场、写下一步与已收口三种方式的时间常数分别约为 22.33、8.68 与 6.45 分钟,并给出从台账估计个人 τ 的设计。第三章求最优冲刺长度,它由警觉衰减常数与切换成本决定,近似为二者乘积两倍的平方根;在两千多种参数组合里,固定九十分钟只在 0.450 的格子里接近最优。第四章拆解三分钟重置:在残留账上,只写一行下一步每回净赚 1.77 分钟,完整重置只净赚 0.67 分钟。 第五章用双过程模型画警觉曲线,标出个人高警觉窗,并把「睡眠时长不得压缩」写成排程前提;把难事贪心地塞进高窗,与动态规划相比最坏只差 2.56 %。第六章做决策日的负荷预算:不可逆决策负荷加倍、切换上限由残留预算推出,五条战线分批推进把残留从 77.1 分钟降到 11.3 分钟。第七章考辨《荀子·解蔽》《庄子·达生》、塞涅卡第二封信与詹姆斯论注意,并落成四件套蒸馏规格。第八章给出标准作业模板、蒸馏小工具「切换残留预算与冲刺排程器」的检查器规格与预注册检验,主检验需要 250 回切换、约 6.2 周。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者假说。 Chapter answers: - Ch.1 Two Ledgers: the Millisecond Switch and the Minute-Scale Residue — Switching cost is kept in two ledgers. The laboratory measures the change of hands, a few hundred milliseconds, real and small. In a working day the large item is attention residue, counted in minutes and highly sensitive to the way one leaves. Program p11 shows that with sixteen switches a day residue outweighs the direct cost by about 369 to one; program p12 shows that the note and finishing first both help, and that the cost of finishing first can be passed on to the people who are kept waiting. (https://5dlibrary.com/answers/en/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c01/) 第1章 两本账:毫秒级的切换动作与分钟级的注意力残留 — 切换代价有两本账。实验室测到的是换手动作,几百毫秒,真实而微小。工作日里真正的大头是注意力残留,按分钟计,随离场方式大幅变化。p11 显示,一天切十六回时,残留与直接代价之比约为 369 倍;p12 显示,写下一步与先做完各有用处,先做完的代价会转嫁给等待的人。 (https://5dlibrary.com/answers/zh/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c01/) - Ch.2 The Residue Decay Equation: Goal Memory, Closing Moves and the Time Constant τ — Chapter 2 derives a residue decay equation from an activation model of goal memory, with time constants of about 22.33, 8.68 and 6.45 minutes for open, noted and closed exits, and a design for estimating one's own τ from a log. (https://5dlibrary.com/answers/en/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c02/) 第2章 残留衰减方程:目标记忆、收尾动作与时间常数 τ — 第二章由目标记忆的激活模型推出残留衰减方程:敞口离场、写下一步与已收口三种方式的时间常数分别约为 22.33、8.68 与 6.45 分钟,并给出从台账估计个人 τ 的设计。 (https://5dlibrary.com/answers/zh/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c02/) - Ch.3 Sprint Length: Alertness Decay, Switching Cost and Where Ninety Minutes Came From — Chapter 3 solves for the optimal sprint length, set by the alertness decay constant and the switching cost and close to the square root of twice their product; across more than two thousand parameter combinations a fixed ninety minutes is near the optimum in a share of only 0.450. (https://5dlibrary.com/answers/en/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c03/) 第3章 冲刺长度:警觉度衰减、切换成本与九十分钟的来历 — 第三章求最优冲刺长度,它由警觉衰减常数与切换成本决定,近似为二者乘积两倍的平方根;在两千多种参数组合里,固定九十分钟只在 0.450 的格子里接近最优。 (https://5dlibrary.com/answers/zh/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c03/) - Ch.4 The Three-Minute Reset: Taking Apart Standing Up, the Sound-and-Light Anchor and the Written Close — Chapter 4 takes the three-minute reset apart: on the residue ledger a single written next-step line nets 1.77 minutes per switch, the full reset only 0.67. (https://5dlibrary.com/answers/en/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c04/) 第4章 三分钟重置协议:起身换位、声光锚点与书面收尾的成分拆解 — 第四章拆解三分钟重置:在残留账上,只写一行下一步每回净赚 1.77 分钟,完整重置只净赚 0.67 分钟。 (https://5dlibrary.com/answers/zh/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c04/) - Ch.5 The Two-Process Alertness Curve: Sleep Pressure, Circadian Rhythm and the High-Alertness Window — Chapter 5 draws an alertness curve from the two-process model, marks a personal high-alertness window and makes "sleep length is never compressed" a precondition of scheduling; greedily filling the window with hard matters loses at worst 2.56 % against dynamic programming. (https://5dlibrary.com/answers/en/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c05/) 第5章 双过程警觉曲线:睡眠压力、昼夜节律与高警觉窗 — 第五章用双过程模型画警觉曲线,标出个人高警觉窗,并把「睡眠时长不得压缩」写成排程前提;把难事贪心地塞进高窗,与动态规划相比最坏只差 2.56 %。 (https://5dlibrary.com/answers/zh/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c05/) - Ch.6 A Load Budget for the Decision Day: Reversibility Weights, a Switch Cap and Batching the Arenas — Chapter 6 builds a load budget for the decision day: irreversible decisions count double, a switch cap follows from a residue budget, and batching the five arenas cuts residue from 77.1 minutes to 11.3. (https://5dlibrary.com/answers/en/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c06/) 第6章 决策日负荷预算:可逆性加权、切换上限与战线分批 — 第六章做决策日的负荷预算:不可逆决策负荷加倍、切换上限由残留预算推出,五条战线分批推进把残留从 77.1 分钟降到 11.3 分钟。 (https://5dlibrary.com/answers/zh/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c06/) - Ch.7 Classical Sources: the Xunzi on Dispelling Obsession, the Zhuangzi on Mastering Life, Seneca's Second Letter and James on Attention — With p of 0.990 the mind drifts after 6.67 minutes on average, and the chance of no drift in fifty minutes is only 0.001. With p of 0.995 the average is 13.33 minutes and the chance 0.024. With p of 0.998, 33.33 minutes and 0.223. With p of 0.999, 66.67 minutes and 0.472. For a fifty-minute sprint to have an even chance of no drift, p has to reach 0.99908, and in those fifty minutes about 750 acts of pulling back take place. (https://5dlibrary.com/answers/en/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c07/) 第7章 原典考辨:《荀子·解蔽》《庄子·达生》、塞涅卡第二封信与詹姆斯论注意 — p 为 0.990 时,平均 6.67 分钟就会飘走,五十分钟内不飘走的概率只有 0.001。p 为 0.995 时,平均 13.33 分钟,概率 0.024。p 为 0.998 时,平均 33.33 分钟,概率 0.223。p 为 0.999 时,平均 66.67 分钟,概率 0.472。要让五十分钟冲刺有一半机会不飘走,p 要达到 0.99908;在这五十分钟里,主动拉回大约发生 750 回。 (https://5dlibrary.com/answers/zh/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c07/) - Ch.8 Closing the Loop: Standard Operating Templates and Checker Specification — Chapter 8 gives standard operating templates, the checker specification of the distilled Switch-Residue Budget and Sprint Scheduler, and a preregistered test whose main hypothesis needs 250 switches, about 6.2 weeks. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c08/) 第8章 实证闭环:标准作业模板与检查器规格 — 第八章给出标准作业模板、蒸馏小工具「切换残留预算与冲刺排程器」的检查器规格与预注册检验,主检验需要 250 回切换、约 6.2 周。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/attention-residue-ultradian-pacing-rapid-context-switching-protocols/c08/) ## 5DL-B-363 · Demographic Cognitive Segmentation, Elder-to-Gen-Z Interface Divergence and Audience-Adaptive Conversion UX 中文书名:客群认知画像分层、适老与新世代界面分化及千人千面转化交互母典 DOI: pending Cite: Fan, Y. (2026). Demographic Cognitive Segmentation, Elder-to-Gen-Z Interface Divergence and Audience-Adaptive Conversion UX (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative, all calculations come from the book's programs, and no real site is described. This is not legal advice and not personalised business advice. Abstract: Operators repeat three sayings: generational differences are huge, so cut the interface by year of birth; older visitors only need bigger type; Generation Z has no patience, so the flashier the better, and rational buyers want data, so the more the better. This book refutes or bounds each one. Meta-analyses find small mean differences between birth cohorts; in a simulated population deliberately tilted towards age, a four-band age label explains 0.1757 of the latent need while the visitors' own interaction signals explain 0.7746, and routing everyone aged sixty or more to a large-print template sends 7,453 people there, 4,144 of whom do not need it. After checking the originals the book corrects its blueprint: the generational meta-analysis studied work attitudes, not web behaviour; the CREATE study examined predictors of technology adoption and did not measure losses at folded menus, so that claim becomes a hypothesis to be tested; the accessibility guidelines set 24 pixels as the level AA minimum for targets and 44 pixels at level AAA, which this book adopts as a stricter house rule; and the mean effect of choice overload is close to zero. Parameters are illustrative, all calculations come from the book's programs, and no real site is described. This is not legal advice and not personalised business advice. Chapter 1 refutes routing by year of birth with a variance decomposition. Chapter 2 does the accounting of Hick and Fitts: grouping adds information rather than removing it, the cost of folding lies in discovery, and for visitors who prefer the calm template a zero-fold linear path completes +0.3076 more journeys. Chapter 3 computes sRGB relative luminance and contrast exactly; the fashion template's decorative grey reaches only 3.55 and must be lightened, and the old and new linearisation thresholds change no verdict for any 8-bit colour. Chapter 4 writes checkout as a product of continuation rates; scan-to-fill breaks even in the calm template at a success rate as low as 0.2956. Chapter 5 shows that choice overload is conditional: an illustrative pooled mean sits near zero while the prediction interval for a new setting runs from -0.5195 to +0.6687, and filters with a buyer-chosen sort make comparing listings 5.69 times as fast. Chapter 6 builds a naive Bayes segmentation engine and prices its mistakes: inference alone loses 44.0 completions per thousand visits, a default with a visible one-tap switch 20.2, inference with the switch 14.2; a SHA-256 check enforces identical prices and terms across templates. Chapter 7 reads the Analects on teaching by temperament, the Book of Rites on courtesy to elders and Aristotle's Rhetoric on young and old audiences, and gives the distillation spec for four kits. Chapter 8 gives standard operating templates, the distilled Audience Template Friction and Accessibility Checker and four pre-registered tests, the largest needing 9,775 visits per arm. Marked boxes separate published evidence, Python calculations (20 programs), classical sources and hypotheses. 摘要:多站经营圈流行三句话:代际差异巨大,按出生年份切界面就行;长辈只要字大就够了;Z 世代没耐心所以越炫越好,理性买家要数据所以越多越好。本书逐条证伪或立界。世代元分析显示出生世代之间的均值差很小;在一个刻意偏向年龄的模拟人群里,四段年龄标签只解释潜在需要的 0.1757,访客自己留下的交互信号解释 0.7746;按六十岁一刀切分流会把 7,453 人送进长辈版,其中 4,144 人并不需要。核对原文之后,本书做了蓝图修正:世代元分析研究的是工作态度而非网页行为;CREATE 研究是技术采纳的预测因素研究,没有直接测量折叠菜单的流失,相关论断改为待检验的作者假说;网页可及性指南 AA 级点按目标下限是 24 像素,44 像素属 AAA 级,本书把 44 像素定为更严的自家规则;选择过载的平均效应接近零。书中参数为示意,全部计算由书中程序生成,不对应任何真实站点;非法律建议,非个人化经营建议。 第一章用方差分解拆掉「按出生年份切界面」。第二章算希克与菲茨的账:分组并不减少信息,折叠的代价在于发现,对偏好清晰版的访客,零折叠线性路径多走完 +0.3076 的旅程。第三章逐位计算 sRGB 相对亮度与对比度,暗色时尚版的装饰灰只有 3.55,须调亮;新旧线性化阈值在全部 8 位颜色上不改变任何判定。第四章把结账写成留存率乘积,扫码免填在清晰版上的盈亏平衡成功率低至 0.2956。 第五章说明选择过载是有条件的效应:示意合并均值接近零,新情境的预测区间却从 -0.5195 到 +0.6687;筛选加买家自选排序让比较楼盘快 5.69 倍。第六章搭朴素贝叶斯分层引擎并算误分层代价:只靠推断每千次访问损失 44.0 单,默认版加可见一键切换 20.2 单,推断加切换 14.2 单;并以 SHA-256 哈希检查落实各模板价格与条款逐字一致。第七章考辨《论语》因材施教、《礼记·曲礼》敬老之礼与亚里士多德《修辞学》论青年与老年听众,并给出四件套蒸馏规格。第八章给出标准作业模板、蒸馏小工具「客群模板摩擦与可及性检查器」与四项预注册检验,最大一项每组需 9,775 次访问。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Year of Birth Is Not an Interface Variable: the Variance of Cohort Labels and Measurable Capabilities — A birth cohort is a weak label. Meta-analyses show that the mean differences in work attitudes between generations are small, and a label that corresponds to a gap of 0.5 standard deviations still explains less than six per cent of the variance. Age-related changes in capability are real, but they are continuous and vary widely between individuals, and what visitors do on the page is closer to the quantity we want to measure than the year in which they were born. Routing by age sends many visitors to the wrong place; a one-tap choice made by the visitor needs no guessing at all. (https://5dlibrary.com/answers/en/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c01/) 第1章 出生年份不是界面变量:世代标签的方差与可测能力 — 出生世代是一个很弱的标签。元分析显示世代之间在工作态度上的均值差很小,一个标签哪怕对应 0.5 个标准差的差异,也只能解释不到百分之六的方差。年龄相关的能力变化真实存在,但它是连续的,个体差异很大,访客自己的动作比出生年份更接近我们要测的量。按年龄分流会把大量访客送错地方,一键自选则不必猜。 (https://5dlibrary.com/answers/zh/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c01/) - Ch.2 Hick and Fitts: an Account of Operating Time for Folded Menus and Linear Paths — Chapter 2 does the accounting of Hick and Fitts: grouping adds information rather than removing it, the cost of folding lies in discovery, and for visitors who prefer the calm template a zero-fold linear path completes +0.3076 more journeys. (https://5dlibrary.com/answers/en/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c02/) 第2章 希克与菲茨:折叠菜单与线性路径的操作时间账 — 第二章算希克与菲茨的账:分组并不减少信息,折叠的代价在于发现,对偏好清晰版的访客,零折叠线性路径多走完 +0.3076 的旅程。 (https://5dlibrary.com/answers/zh/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c02/) - Ch.3 The Exact Arithmetic of Contrast: sRGB Linearisation, Dark Themes and High-Contrast Themes — Chapter 3 computes sRGB relative luminance and contrast exactly; the fashion template's decorative grey reaches only 3.55 and must be lightened, and the old and new linearisation thresholds change no verdict for any 8-bit colour. (https://5dlibrary.com/answers/en/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c03/) 第3章 对比度的精确算术:sRGB 线性化、暗色主题与高对比主题 — 第三章逐位计算 sRGB 相对亮度与对比度,暗色时尚版的装饰灰只有 3.55,须调亮;新旧线性化阈值在全部 8 位颜色上不改变任何判定。 (https://5dlibrary.com/answers/zh/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c03/) - Ch.4 The Checkout Friction Funnel: Field Counts, Scan-to-Fill and the Product of Abandonment Rates — Chapter 4 writes checkout as a product of continuation rates; scan-to-fill breaks even in the calm template at a success rate as low as 0.2956. (https://5dlibrary.com/answers/en/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c04/) 第4章 结账摩擦漏斗:字段数、扫码免填与放弃率乘积 — 第四章把结账写成留存率乘积,扫码免填在清晰版上的盈亏平衡成功率低至 0.2956。 (https://5dlibrary.com/answers/zh/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c04/) - Ch.5 Assortment Size and the Probability of Buying: the Uncertain Inverted U — Chapter 5 shows that choice overload is conditional: an illustrative pooled mean sits near zero while the prediction interval for a new setting runs from -0.5195 to +0.6687, and filters with a buyer-chosen sort make comparing listings 5.69 times as fast. (https://5dlibrary.com/answers/en/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c05/) 第5章 选项数与购买概率:倒 U 型的不确定区间 — 第五章说明选择过载是有条件的效应:示意合并均值接近零,新情境的预测区间却从 -0.5195 到 +0.6687;筛选加买家自选排序让比较楼盘快 5.69 倍。 (https://5dlibrary.com/answers/zh/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c05/) - Ch.6 The Segmentation Engine: Observable Signals, Naive Bayes and a Cost Matrix for Misrouting — Chapter 6 builds a naive Bayes segmentation engine and prices its mistakes: inference alone loses 44.0 completions per thousand visits, a default with a visible one-tap switch 20.2, inference with the switch 14.2; a SHA-256 check enforces identical prices and terms across templates. (https://5dlibrary.com/answers/en/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c06/) 第6章 分层引擎:可观测信号、朴素贝叶斯与误分层代价矩阵 — 第六章搭朴素贝叶斯分层引擎并算误分层代价:只靠推断每千次访问损失 44.0 单,默认版加可见一键切换 20.2 单,推断加切换 14.2 单;并以 SHA-256 哈希检查落实各模板价格与条款逐字一致。 (https://5dlibrary.com/answers/zh/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c06/) - Ch.7 Classical Sources: Teaching by Temperament in the Analects, Courtesy to Elders in the Book of Rites, and Aristotle on Young and Old Audiences — At $d$ = 0.2, in the range of the small effects of the generational research cited in Chapter 1, the two distributions overlap by 0.9203, and the typology misorders a random pair of one young and one old person 0.4438 of the time, barely better than a coin. At $d$ = 0.5, the overlap is 0.8026 and the misordering rate 0.3618. Even at $d$ = 1.2, a large effect seldom seen in that literature, the overlap is 0.5485 and the typology misorders a random pair 0.1981 of the time, about one pair in five. (https://5dlibrary.com/answers/en/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c07/) 第7章 原典考辨:《论语》因材施教、《礼记》敬老之礼与亚里士多德论青年与老年听众 — $d$ 为 0.2 时,也就是第一章所引世代研究的小效应范围,两个分布重叠 0.9203,类型学把随机一老一少排错的概率是 0.4438,只比掷硬币好一点。$d$ 为 0.5 时,重叠 0.8026,排错率 0.3618。即便 $d$ 达到 1.2,这在那批文献里很少见到,重叠仍有 0.5485,类型学排错随机一对的概率是 0.1981,大约五对错一对。 (https://5dlibrary.com/answers/zh/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c07/) - Ch.8 Standard Operating Templates and Checker Specification: the Weekly Audience-Template Ledger, Pre-Registered Tests and Refutation Conditions — Chapter 8 gives standard operating templates, the distilled Audience Template Friction and Accessibility Checker and four pre-registered tests, the largest needing 9,775 visits per arm. Marked boxes separate published evidence, Python calculations (20 programs), classical sources and hypotheses. (https://5dlibrary.com/answers/en/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c08/) 第8章 标准作业模板与检查器规格:客群界面周度台账、预注册检验与推翻条件 — 第八章给出标准作业模板、蒸馏小工具「客群模板摩擦与可及性检查器」与四项预注册检验,最大一项每组需 9,775 次访问。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/demographic-cognitive-segmentation-elder-genz-interface-adaptive-conversion-ux/c08/) ## 5DL-B-364 · Quantum-Inspired Combinatorial Search, Abductive Intuition and Computationally Verified Breakthrough Monograph Engineering 中文书名:量子启发式组合寻优、拉马努金式直觉猜想与突破性著作推演母典 DOI: pending Cite: Fan, Y. (2026). Quantum-Inspired Combinatorial Search, Abductive Intuition and Computationally Verified Breakthrough Monograph Engineering (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. Abstract: Three popular sentences dominate talk about original research: quantum computers try every possibility at once, so "quantum inspiration" means faster exhaustive search; the intuition of genius cannot be learned, so Ramanujan-style conjectures come from inspiration alone; and the more distant an analogy, the better the idea. This book refutes each sentence or draws its boundary. After checking the original papers it corrects its own blueprint: the quantum speedup for unstructured search is only a square root and is proved unbeatable; the advantage of eight orders of magnitude reported for a hardware annealer was a constant factor against single-core simulated annealing on crafted instances, which other classical heuristics solve about as fast; and dequantisation shows that part of the exponential advantage claimed for quantum recommendation came from its input assumptions. In this book "quantum-inspired" means only three classical ideas: annealing and restarts, sampling weighted by value, and spectra and mappings that expose shared structure. Every program is a classical simulation and no quantum speedup is claimed. Parameters are illustrative and all calculations are generated by the book's programs. The book offers no investment advice, no advice on academic review and no advice on quantum hardware. Chapter 1 simulates amplitude amplification with a complex vector and computes the square-root wall. Chapter 2 compares greedy descent, random restarts and annealing on a frustrated spin glass: greedy descent stalls at a hit rate of 0.060, random restarts rise to 0.985 with the budget, and several short anneals do as well. Chapter 3 sets out the burden of proof for a speedup: time to solution, constant factors versus scaling laws, speedups manufactured by selecting instances, and dequantisation. Chapter 4 turns amplification by value into importance sampling and allocation by upper confidence bounds. Chapter 5 separates structural from verbal similarity with spectral fingerprints and exhaustive mappings and rebuilds a simplified typicality z-score for combinations. Chapter 6 builds a pipeline for Ramanujan-style conjectures: Euler-type candidates from a template are refuted at a rate of 0.9970, and the Pólya conjecture survives every small check although its smallest counterexample is 906,150,257. Chapter 7 examines Cook Ding in the Zhuangzi, Shen Kuo's rule for stacked piles, Peirce on abduction and one sentence of Hardy, and checks Shen Kuo's correction on 27,000 piles. Chapter 8 supplies a weekly ledger of original propositions, an optimisation log, the checker specification of the distilled Remote-Analogy Conjecture Refutation Pipeline and pre-registered tests. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:谈原创研究,最流行的说法有三句:量子计算能瞬间穷举,所以「量子启发」就是更快的穷举;天才的直觉不可学,拉马努金式的猜想全凭灵感;类比的两端隔得越远,点子越好。本书逐条证伪或立界。核对原文之后,本书做了蓝图修正:无结构搜索的量子加速只有平方根量级,且已被证明不可超越;硬件退火器报告的八个数量级优势,是在专门构造的实例上相对单核模拟退火的常数倍,其他经典启发式能以相近速度解出多数实例;去量子化结果说明,部分「量子推荐」的指数优势来自输入假设。书中「量子启发」只指三种经典思想:退火与重启、按价值放大采样、用谱与映射揭示结构同构。所有程序都是经典模拟,不宣称任何量子加速。书中参数为示意,全部计算由书中程序生成。本书不构成投资、学术评审或量子硬件方面的建议。 第一章用复数向量模拟振幅放大,算出平方根之墙。第二章在受挫自旋玻璃上比较贪心、随机重启与退火:贪心命中率停在 0.060,随机重启随预算升到 0.985,切碎预算的多段短退火同样可达。第三章讲加速的举证责任:求解时间、常数倍与标度律、挑选实例造出的虚假加速与去量子化。第四章把按价值放大落成重要性采样与置信上界分配。第五章用谱指纹与穷举映射区分结构相似与字面相似,并复现简化版组合典型性 z 分数。 第六章搭建拉马努金式猜想流水线:模板生成的欧拉式候选证伪率 0.9970;Pólya 猜想在小范围内毫发无损,最小反例却在 906,150,257。第七章考辨《庄子》庖丁解牛、沈括隙积术、皮尔士的溯因与哈代的一句话,沈括的修正公式在 27,000 个堆上逐一核对。第八章给出原创命题周度台账、寻优登记表、蒸馏小工具「远距类比猜想证伪流水线」的检查器规格与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Square-Root Wall: a Classical Simulation of Amplitude Amplification and the Lower Bound for Unstructured Search — The classical cost of unstructured search is proportional to N and the quantum cost is proportional to √N. These are proved upper and lower bounds, not a temporary engineering gap. Program p11 simulated the amplitudes round by round on an ordinary computer, confirmed the closed form sin²((2k + 1)θ), and showed that rotating too far makes the probability fall back. Program p12 showed that the square-root wall turns 2 to the n into 2 to the n/2, leaving the exponent in place; p13 showed that an unknown number of solutions roughly doubles the cost, and that mistuned rounds can yield nothing at all. (https://5dlibrary.com/answers/en/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c01/) 第1章 平方根之墙:振幅放大的经典模拟与无结构搜索的下界 — 无结构搜索的经典代价与 N 成正比,量子代价与 √N 成正比,这是已被证明的上下界,不是工程上暂时的差距。程序 p11 在普通计算机上逐轮模拟振幅,确认了 sin²((2k + 1)θ) 的闭式解,也显示了转过头会让概率回落。程序 p12 说明平方根之墙把 2 的 n 次方变成 2 的 n/2 次方,指数还在;p13 说明不知道解的个数时,代价约翻一倍,轮数调错可能一无所获。 (https://5dlibrary.com/answers/zh/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c01/) - Ch.2 Annealing, Restarts and the Metropolis Rule: Classical Recipes for Escaping Local Optima — Chapter 2 compares greedy descent, random restarts and annealing on a frustrated spin glass: greedy descent stalls at a hit rate of 0.060, random restarts rise to 0.985 with the budget, and several short anneals do as well. (https://5dlibrary.com/answers/en/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c02/) 第2章 退火、重启与 Metropolis 准则:逃离局部最优的经典配方 — 第二章在受挫自旋玻璃上比较贪心、随机重启与退火:贪心命中率停在 0.060,随机重启随预算升到 0.985,切碎预算的多段短退火同样可达。 (https://5dlibrary.com/answers/zh/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c02/) - Ch.3 The Burden of Proof for a Speedup: Time to Solution, Scaling Laws and Dequantisation — The time to solution turns a hit rate into hours: a method that succeeds with probability 0.01 must be repeated 458.2 times for 99 % confidence. A constant-factor advantage does not change with size, while a slightly worse scaling law eventually swallows a lead of 10 to the 8. Among a thousand instances without any real speedup, picking the 241 on which the device did well turns a median of 1.036 into 2.893. Dequantisation shows that the input assumption of squared-value sampling carries much of the supposed speedup, and that classical algorithms can use it too. (https://5dlibrary.com/answers/en/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c03/) 第3章 加速的举证责任:求解时间、标度律与去量子化 — 求解时间把命中率换成钟点:成功率 0.01 的方法要重复 458.2 次才有 99 % 把握。常数倍优势不随规模变化;标度律稍差一点,规模一大,10 的 8 次方倍领先也会被吞掉。一千个没有真实加速的实例里,挑出表现好的 241 个,中位数就从 1.036 变成 2.893。去量子化说明,按平方抽样这一输入假设本身就携带了大部分「加速」,经典算法同样能用。 (https://5dlibrary.com/answers/zh/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c03/) - Ch.4 Amplifying Samples by Value: Importance Sampling and Upper Confidence Bounds — Classical statistics has two mature ways of amplifying samples by value. Importance sampling switches to a distribution aimed at the target and corrects with weights: in estimating a normal tail probability, a shift of 4.5 cuts the relative standard deviation of a single draw from 542.5 to 2.3, while too small or too large a shift spoils the gain. The multi-armed bandit learns while sampling: with ten directions and two thousand checks, epsilon-greedy has a regret of 77.9, UCB1 188.0 and even allocation 300.0. Both depend on knowledge of value; prior knowledge and the speed of learning are the real sources of amplification. (https://5dlibrary.com/answers/en/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c04/) 第4章 按价值放大采样:重要性采样与置信上界 — 按价值放大采样,经典统计有两种成熟做法。重要性采样换一个对准目标的分布、用权重纠偏:估计正态尾部概率时,平移量取 4.5,单次相对标准差从 542.5 降到 2.3;平移不足或过头,效果都打折。多臂老虎机边抽边学:十个方向两千次检验,ε 贪心的遗憾 77.9,UCB1 为 188.0,平均分配为 300.0。两者都依赖对价值的知识,先验与学习速度才是放大的真正来源。 (https://5dlibrary.com/answers/zh/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c04/) - Ch.5 Shared Structure: Spectral Fingerprints, Exhaustive Mappings and the Typicality of Combinations — Structural similarity and verbal similarity are different things. Heat conduction and the circuit have identical singular values and a spectral distance of 0.0000 while sharing no words; the kitchen scene shares words at 0.308 yet lies at a spectral distance of 1.6179. Exhaustive mapping confirms that all 6 relations align from heat conduction to the circuit, while only 2 align with the kitchen and 90 mappings tie. Equal spectra do not imply isomorphism, as the star graph and the four-cycle with an isolated vertex show. Uzzi's typicality z-score distinguishes the conventional from the novel, but its threshold depends on the make-up of the corpus. (https://5dlibrary.com/answers/en/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c05/) 第5章 结构同构:谱指纹、穷举映射与组合典型性 — 结构相似与字面相似是两回事。热传导与电路的奇异值完全相同,谱距离 0.0000,词表却毫不重合;厨房场景用词重合 0.308,谱距离却有 1.6179。穷举映射确认热传导到电路 6 条关系全部对齐,到厨房只对上 2 条,且有 90 个映射并列。谱相同不等于同构,星形图与四边形加孤立点就是反例。Uzzi 的典型性 z 分数能区分常规与新颖,但它的门槛依赖语料构成。 (https://5dlibrary.com/answers/zh/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c05/) - Ch.6 A Pipeline for Ramanujan-Style Conjectures: Abundant Generation, Exhaustive Refutation and Exact Bounds — Chapter 6 builds a pipeline for Ramanujan-style conjectures: Euler-type candidates from a template are refuted at a rate of 0.9970, and the Pólya conjecture survives every small check although its smallest counterexample is 906,150,257. (https://5dlibrary.com/answers/en/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c06/) 第6章 拉马努金式猜想流水线:高产生成、穷举证伪与精确定界 — 第六章搭建拉马努金式猜想流水线:模板生成的欧拉式候选证伪率 0.9970;Pólya 猜想在小范围内毫发无损,最小反例却在 906,150,257。 (https://5dlibrary.com/answers/zh/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c06/) - Ch.7 Following the Natural Joints and Reasoning to Hypotheses: Cook Ding, Shen Kuo, Peirce and Hardy — Chapter 7 examines Cook Ding in the Zhuangzi, Shen Kuo's rule for stacked piles, Peirce on abduction and one sentence of Hardy, and checks Shen Kuo's correction on 27,000 piles. (https://5dlibrary.com/answers/en/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c07/) 第7章 依乎天理与溯因:庖丁、沈括、皮尔士与哈代 — 第七章考辨《庄子》庖丁解牛、沈括隙积术、皮尔士的溯因与哈代的一句话,沈括的修正公式在 27,000 个堆上逐一核对。 (https://5dlibrary.com/answers/zh/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c07/) - Ch.8 Closing the Empirical Loop: Standard Operating Templates and Checker Specification — Chapter 8 supplies a weekly ledger of original propositions, an optimisation log, the checker specification of the distilled Remote-Analogy Conjecture Refutation Pipeline and pre-registered tests. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c08/) 第8章 实证闭环:标准作业模板与检查器规格 — 第八章给出原创命题周度台账、寻优登记表、蒸馏小工具「远距类比猜想证伪流水线」的检查器规格与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/quantum-inspired-search-abductive-conjecture-verified-monograph-engineering/c08/) ## 5DL-B-365 · Negative Working-Capital Flywheels, Upfront Customer Float and Zero-Capital Bootstrap Scaling 中文书名:负营运资金飞轮、预收定金周转动力学与零本金商业放大母典 DOI: pending Cite: Fan, Y. (2026). Negative Working-Capital Flywheels, Upfront Customer Float and Zero-Capital Bootstrap Scaling (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Revenues, margins, payment terms, refund rates, acquisition costs and prices are illustrative, and all calculations are generated by the book's programs. Chapters 4 to 6 use fixed-seed Monte Carlo to show that single shocks rarely break a presale business while stacked shocks do; under illustrative parameters the cautious reserve-ratio red line is 0.474. Abstract: Owners with little capital hear three popular beliefs: without capital you cannot expand; once customers pay first, working capital turns negative and becomes free money, so the bigger the business the safer it is; and if you roll the first customers' prepayments into advertising, any payback ratio above one lets you scale without limit. This book refutes each belief or draws its boundary. "Zero capital" in the title means no borrowing and no outside equity; it does not mean absence of risk. Revenues, margins, payment terms, refund rates, acquisition costs and prices are illustrative, and all calculations are generated by the book's programs. This is general business guidance, not legal, accounting, tax or investment advice; local rules on prepayments, deposits and stored value must be reviewed by a qualified person before publication, and the book involves no profit sharing or commissions of any kind. Chapters 1 to 3 show that the cash structure is governed by the revenue-weighted cash cycle K rather than the textbook CCC: one group-buying business has a CCC of 10.5 days but a K of 6.18 days, and when the supplier must be prepaid a thin-margin group buy needs a deposit share of 0.7872 to push K below zero. With K positive, self-financed growth has a ceiling; with K negative, growth releases float. But float is a liability that flows out at the same rate when revenue falls: ten times the size means ten times the drain and the same month of rupture, and prepaid buyers mirror bank depositors, so a run is an equilibrium whenever reserves fall short of the full float. Chapters 4 to 6 use fixed-seed Monte Carlo to show that single shocks rarely break a presale business while stacked shocks do; under illustrative parameters the cautious reserve-ratio red line is 0.474. The share of each deposit that may fund acquisition before delivery is capped at 0.475, with payback counted only in realised margin, and the chapters give trust anchors, the deposit-ratio trade-off and a weekly test for new-pays-old. Chapter 7 examines Jiran and Bai Gui in the Shiji, Smith and Bagehot; in an illustrative simulation, same-week refunds lead to 216.7 requests and stalling to 686.9. Chapter 8 gives the Negative Working-Capital Flywheel and Customer-Float Safety Calculator, a weekly ledger, traffic-light actions and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:本金有限的经营者常听到三条流行说法:没有本金就不能扩张;让客户先付钱,营运资金变成负数就等于免费资金,规模越大越安全;用首波客户的预付款滚动买广告,只要回本率大于一就能无限放大。本书逐条证伪或立界。书名中的「零本金」只指不借债、不融资,不等于没有风险;书中的营收、毛利、账期、退款率、获客成本与价格均为示意参数,全部计算由书中程序生成;本书是非个人化经营建议,不构成法律、会计、税务或投资建议,各地预付款、定金与储值规则以上架前人工复核为准,本书不涉及任何分润或返佣。 第一至三章说明,决定现金结构的是按营收计的现金加权周期 K,而不是教科书的 CCC:同一笔团购生意 CCC 为 10.5 天、K 为 6.18 天;供应商要求预付时,低毛利的预购成团要收 0.7872 的定金才能把 K 压负。K 为正时自筹增长有上限,K 为负时增长释放浮存;可是浮存是负债,营收下滑时按同一系数流出,规模扩大十倍,抽水也扩大十倍,断裂月份不变;预付买家与存款人同构,只要准备金不足全额,挤兑就是一个均衡。 第四至六章用固定种子蒙特卡洛说明,单一冲击很少压垮预售生意,叠加冲击才是准备金要防的东西,示意参数下审慎的准备金率红线是 0.474;每单定金交付前可动用于获客的比例上界是 0.475,回本只按已实现毛利计;并给出信任锚、定金比例权衡与逐周检出以新补旧的规则。第七章考辨《史记·货殖列传》计然、白圭与斯密、白芝浩,示意模拟中当周退款使申请为 216.7 笔,拖延则升到 686.9 笔。第八章给出「负营运资金飞轮与预收资金安全边界试算器」、周度台账、亮灯动作与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 When the Cash Conversion Cycle Turns Negative: Four Prepaid Business Forms Term by Term — Chapters 1 to 3 show that the cash structure is governed by the revenue-weighted cash cycle K rather than the textbook CCC: one group-buying business has a CCC of 10.5 days but a K of 6.18 days, and when the supplier must be prepaid a thin-margin group buy needs a deposit share of 0.7872 to push K below zero. With K positive, self-financed growth has a ceiling; with K negative, growth releases float. But float is a liability that flows out at the same rate when revenue falls: ten times the size means ten times the drain and the same month of rupture, and prepaid buyers mirror bank depositors, so a run is an equilibrium whenever reserves fall short of the full float. (https://5dlibrary.com/answers/en/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c01/) 第1章 现金转换周期何时为负:四种预收业务的分项代入 — 第一至三章说明,决定现金结构的是按营收计的现金加权周期 K,而不是教科书的 CCC:同一笔团购生意 CCC 为 10.5 天、K 为 6.18 天;供应商要求预付时,低毛利的预购成团要收 0.7872 的定金才能把 K 压负。K 为正时自筹增长有上限,K 为负时增长释放浮存;可是浮存是负债,营收下滑时按同一系数流出,规模扩大十倍,抽水也扩大十倍,断裂月份不变;预付买家与存款人同构,只要准备金不足全额,挤兑就是一个均衡。 (https://5dlibrary.com/answers/zh/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c01/) - Ch.2 Float-Driven Growth: Self-Financeable Growth and Monthly Difference Equations — With a cycle of -20 days and the same growth of 0.10, cash reaches 2,766.90 by month 24. At 0.20 it reaches 16,529.65. The faster the business grows, the more comfortable its bank balance becomes, and in both growing cases the lowest point falls in the very first month, before growth has had time to release any float. There is no dip to survive. (https://5dlibrary.com/answers/en/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c02/) 第2章 浮存驱动增长:自筹资金增长率与逐月差分方程 — k 是 1.50 个月、经营现金率 0.05 时,每月最多增长 0.033,一年约 0.482;这已经是不少代购生意的处境。k 缩短到 0.25 个月、经营现金率 0.15 时,月度上限是 0.600,一年的数字大得失去了意义,280.475 倍。这说明一件事:周期接近零时,营运资金不再是增长的约束,真正的约束早已换成了供给、交付与需求。 (https://5dlibrary.com/answers/zh/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c02/) - Ch.3 Flywheel Reversal: The Cash Drain of Falling Revenue and Its Isomorphism with Bank Runs — Float is a liability, not a profit. When revenue grows it flows in at |k|; when revenue falls it flows out through the same coefficient. Program p31 shows that ten times the size means ten times the drain while the month of rupture does not move at all; the only thing that delays the break is the reserve kept back during growth. Program p32 shows that whenever the reserve ratio is below one, a refund run is an equilibrium that can appear at any moment, and that full escrow removes that equilibrium altogether. Between the chronic drain and the acute one sits the same variable, ρ, and the same lesson: what protects a prepaid business is not how much customer money passes through it but how much of that money it has left untouched. (https://5dlibrary.com/answers/en/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c03/) 第3章 飞轮反转:营收下滑时的抽水量与挤兑同构 — 浮存是负债,不是利润。营收增长时它按 |k| 流入,营收下滑时它按同一个系数流出;p31 说明,规模扩大十倍,抽水量也扩大十倍,断裂月份却纹丝不动,能推迟断裂的只有增长期留下的准备金。p32 说明,只要准备金比例小于一,退款挤兑就是一个随时可能出现的均衡,而全额托管能把这个均衡消掉。 (https://5dlibrary.com/answers/zh/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c03/) - Ch.4 Rupture Probability under Three Shocks and the Ring-Fenced Reserve Ratio — Chapters 4 to 6 use fixed-seed Monte Carlo to show that single shocks rarely break a presale business while stacked shocks do; under illustrative parameters the cautious reserve-ratio red line is 0.474. The share of each deposit that may fund acquisition before delivery is capped at 0.475, with payback counted only in realised margin, and the chapters give trust anchors, the deposit-ratio trade-off and a weekly test for new-pays-old. (https://5dlibrary.com/answers/en/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c04/) 第4章 三种冲击下的资金链断裂概率与圈存准备金率 — 第四至六章用固定种子蒙特卡洛说明,单一冲击很少压垮预售生意,叠加冲击才是准备金要防的东西,示意参数下审慎的准备金率红线是 0.474;每单定金交付前可动用于获客的比例上界是 0.475,回本只按已实现毛利计;并给出信任锚、定金比例权衡与逐周检出以新补旧的规则。 (https://5dlibrary.com/answers/zh/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c04/) - Ch.5 Paying for Acquisition with Customer Float: Payback, Delivery Lag and the Usable-Float Ceiling — The part of a deposit that can be used before delivery is this order's future expected margin, a share of 1 − r_min. Acquisition spending can push that ceiling upward only if it pays back in realised margin before delivery; payback counted in new deposits treats a fresh liability as income. The program p52 shows that rolling advertising beyond the ceiling leads to a break within weeks of growth stalling, while paying old with new begins more than ten weeks before the break, concealed by the continuing inflow of new deposits. (https://5dlibrary.com/answers/en/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c05/) 第5章 用预收款付获客成本:回本周期、交付时滞与可动用浮存上界 — 一单定金里能在交付前动用的部分,就是这一单将来的期望毛利,比例是 1 − r_min。获客支出只有按已实现毛利在交付之前回本,才能把上界往上推;按新收定金计的回本,是把新的负债当作收入。p52 说明,超出上界的滚动投放,会在增长停下之后数周内断裂,而以新补旧早在断裂之前十几周就已经开始,只是被持续流入的新定金遮住了。 (https://5dlibrary.com/answers/zh/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c05/) - Ch.6 Trust Anchors, Deposit Ratios and Detecting New-Pays-Old — A buyer pays first because of trust anchors: a delivery record, third-party escrow, a published delivery deadline and published refund terms. The program p62 shows that without escrow a new seller needs a dozen or more clean deliveries before a cautious buyer will prepay, while with escrow two or three suffice. The program p61 shows that under illustrative curves the best deposit is only about a fifth of order value, far below what a negative cycle would need, so supplier terms must do the compressing. The program p63 gives a weekly rule that flags new-pays-old weeks before cash turns negative. (https://5dlibrary.com/answers/en/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c06/) 第6章 信任锚、定金比例与以新补旧检出 — 买家先付钱,靠的是信任锚:履约记录、第三方托管、公开交付期限与公开退款条款。p62 说明,没有托管时,新卖家要累积十几单干净记录,谨慎的买家才肯先付;有托管时,两三单就够。p61 说明,在示意曲线下,期望净现金最大的定金比例只有两成左右,远低于把周期压负所需的比例,所以压周期要靠供应商账期,不能全靠定金。p63 给出逐周检出以新补旧的规则,它在现金转负之前数周就亮起红灯。 (https://5dlibrary.com/answers/zh/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c06/) - Ch.7 Eastern and Western Sources: Jiran, Bai Gui, Smith and Bagehot — Chapter 7 examines Jiran and Bai Gui in the Shiji, Smith and Bagehot; in an illustrative simulation, same-week refunds lead to 216.7 requests and stalling to 686.9. (https://5dlibrary.com/answers/en/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c07/) 第7章 东西方原典考辨:计然、白圭、斯密与白芝浩 — 第七章考辨《史记·货殖列传》计然、白圭与斯密、白芝浩,示意模拟中当周退款使申请为 216.7 笔,拖延则升到 686.9 笔。 (https://5dlibrary.com/answers/zh/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c07/) - Ch.8 Actuarial Tools and Execution Checklist: Weekly Ledger, Pre-Registration and Refutation — Chapter 8 gives the Negative Working-Capital Flywheel and Customer-Float Safety Calculator, a weekly ledger, traffic-light actions and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c08/) 第8章 精算工具与执行清单:周度台账、预注册检验与推翻条件 — 第八章给出「负营运资金飞轮与预收资金安全边界试算器」、周度台账、亮灯动作与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/negative-working-capital-flywheel-upfront-customer-float-zero-capital-bootstrap/c08/) ## 5DL-B-366 · Click-to-Messenger Conversion Funnels, Chat Attribution and Voice-Note Closing in Southeast Asian Commerce 中文书名:东南亚即时通讯成交漏斗、点击直达对话归因与语音权威收口母典 DOI: pending Cite: Fan, Y. (2026). Click-to-Messenger Conversion Funnels, Chat Attribution and Voice-Note Closing in Southeast Asian Commerce (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its conversion rates, latencies, reply rates, sample sizes and amounts are illustrative, and all calculations are generated by its programs. It is general business guidance, not legal or marketing-compliance advice, and promises no income. Abstract: Selling high-ticket goods and services through WhatsApp in Malaysia, Zalo in Vietnam and LINE in Taiwan comes with four popular beliefs: a direct-to-chat ad merely saves a landing page; the last click tells which ad a conversation came from; the more follow-ups the better; and the longer a real-person voice note, the more sincere it sounds. This book refutes each belief or draws its boundary. Its conversion rates, latencies, reply rates, sample sizes and amounts are illustrative, and all calculations are generated by its programs. It is general business guidance, not legal or marketing-compliance advice, and promises no income. It describes only consented contact; voice notes are recorded only by the real person; payment uses only licensed providers' public features; urgency comes only from real stock and real deadlines; referral rewards go only to the direct referrer, one level. Chapters 1 and 2 derive the condition for halving friction loss: under illustrative parameters the first-line rate after the jump must reach 0.650; a direct ad multiplies conversations by 2.500 while the high-intent share falls from 0.571 to 0.327, and the two paths tie at a handling cost of 4.94 units per conversation. Fitting two published ratios with a decay curve towards a floor gives a time constant of about 29.2 minutes; when volume exceeds staffing, answering hot chats first raises one agent's daily closes from 4.10 to 12.44. Chapters 3 and 4 show that last click can credit the final touch with 2.36 times its Shapley share, that severed sources over-credit ads to 1.20 times and prefilled reference codes cut this to 1.06, and that detecting a one-tenth increment in a cold audience needs 822,401 people per group, while a conversation-level test still needs 8,158 conversations per group. Chapters 5 and 6 show that reply rates decay and block risks rise with each follow-up, giving illustrative stop points of three follow-ups for hot chats, one for warm and none for cold; that 60 seconds is not an optimal voice-note length but a robust cap, keeping 0.999 of the best gain in the illustrative model with a median regret of 0.0005 across 135 parameter settings; and that a payment link valid until the real deadline yields payment before it from 0.996 of intending buyers. Chapter 7 examines the Guiguzi, the Analects, Aristotle and Cicero: reminders about 4 hours apart give the highest reply rate, 0.953, and spoken sources agree with logged codes at a kappa of only 0.511. Chapter 8 provides a weekly ledger, the Click-to-Chat Funnel and Three-Label Follow-Up Cadence Checker and pre-registered tests. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:在马来西亚用 WhatsApp、在越南用 Zalo、在台湾用 LINE 谈高客单生意,常听到四条流行说法:广告直跳私信只是省掉一个落地页;看最后一次点击就知道对话来自哪条广告;追单越勤越好;真人语音条越长越真诚。本书逐条证伪或立界。书中的转化率、时延、回复率、样本量与金额均为示意参数,全部计算由书中程序生成;本书是非个人化经营建议,不构成法律或营销合规意见,不承诺任何收入;只写有同意记录的触达,语音条只由真人本人录制,收款只用持牌支付机构的公开功能,紧迫感只来自真实库存与真实截止,推荐奖励只给直接推荐人一层。 第一、二章推导摩擦损耗减半的条件:示意参数下直跳后首句发出率须达 0.650;直跳让对话数涨到 2.500 倍,高意向占比从 0.571 跌到 0.327,每段接待成本 4.94 单位时两条路径打平。把两个已发表的比值拟合成带下限的衰减曲线,时间常数约 29.2 分钟;对话量超过人手时,热的先答让一名接待员每天的成交从 4.10 笔升到 12.44 笔。第三、四章说明,末次点击记给最后触点的功劳可达 Shapley 份额的 2.36 倍,来源被切断时广告被多记到 1.20 倍,预填参考代码降到 1.06 倍;冷受众测出一成增量,每组要 822,401 人,对话层面每组也要 8,158 段。 第五、六章说明,回复率逐次衰减、屏蔽风险逐次上升,示意最优停追点为热三次、温一次、冷零次;语音条 60 秒不是最优值而是稳健上限,示意模型下保留最佳增益的 0.999,扫描 135 组参数的遗憾中位数为 0.0005;收款链接开到真实截止,截止前付款比例为 0.996。第七章考辨《鬼谷子·反应》《论语·公冶长》、亚里士多德与西塞罗,提醒间隔约 4 小时时回复率最高,为 0.953,口头来源与代码的一致系数只有 0.511。第八章给出周度台账、点击直达对话漏斗与三标签追单节奏检查器与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Funnel Products and Halving Friction Loss: the Break-Even of Direct-to-Chat Ads — Chapters 1 and 2 derive the condition for halving friction loss: under illustrative parameters the first-line rate after the jump must reach 0.650; a direct ad multiplies conversations by 2.500 while the high-intent share falls from 0.571 to 0.327, and the two paths tie at a handling cost of 4.94 units per conversation. Fitting two published ratios with a decay curve towards a floor gives a time constant of about 29.2 minutes; when volume exceeds staffing, answering hot chats first raises one agent's daily closes from 4.10 to 12.44. (https://5dlibrary.com/answers/en/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c01/) 第1章 漏斗乘积与摩擦损耗减半:直跳私信的盈亏平衡 — 第一、二章推导摩擦损耗减半的条件:示意参数下直跳后首句发出率须达 0.650;直跳让对话数涨到 2.500 倍,高意向占比从 0.571 跌到 0.327,每段接待成本 4.94 单位时两条路径打平。把两个已发表的比值拟合成带下限的衰减曲线,时间常数约 29.2 分钟;对话量超过人手时,热的先答让一名接待员每天的成交从 4.10 笔升到 12.44 笔。 (https://5dlibrary.com/answers/zh/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c01/) - Ch.2 The Decay Curve of First-Response Latency: Triage Before Staff — The published study supplies two ratios: contacting a lead within an hour makes qualification nearly 7 times as likely as contacting it an hour later, and more than 60 times as likely as contacting it after a day. Fitted as exponential decay towards a floor, the curve has a time constant of about 29.2 minutes, falls fastest in the first half hour and runs along the floor after two hours; depending on the assumed bin points, the time constant moves between 29.2 and 51.1 minutes. The queue simulation shows that when volume exceeds staffing, answering hot chats first acts before extra staff do: one agent who changes the ordering raises daily closes from 4.10 to 12.44. (https://5dlibrary.com/answers/en/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c02/) 第2章 首响时延的衰减曲线:先分层,再加人 — 已发表的研究给了两个比值:一小时内联系,资格率接近一小时后的 7 倍,是一天后的 60 多倍。拟合成带下限的指数衰减后,时间常数约 29.2 分钟,前半小时掉得最快,过了两小时贴着下限走;时间常数随格点假设在 29.2 到 51.1 分钟之间浮动。排队模拟说明,在对话量超过人手时,热的先答比加人更先起作用:一名接待员改排法,每天成交从 4.10 笔升到 12.44 笔。 (https://5dlibrary.com/answers/zh/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c02/) - Ch.3 Chat Attribution: Last Click, the Shapley Value and Severed Sources — Chapters 3 and 4 show that last click can credit the final touch with 2.36 times its Shapley share, that severed sources over-credit ads to 1.20 times and prefilled reference codes cut this to 1.06, and that detecting a one-tenth increment in a cold audience needs 822,401 people per group, while a conversation-level test still needs 8,158 conversations per group. (https://5dlibrary.com/answers/en/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c03/) 第3章 对话归因:末次点击、Shapley 值与被切断的来源 — 第三、四章说明,末次点击记给最后触点的功劳可达 Shapley 份额的 2.36 倍,来源被切断时广告被多记到 1.20 倍,预填参考代码降到 1.06 倍;冷受众测出一成增量,每组要 822,401 人,对话层面每组也要 8,158 段。 (https://5dlibrary.com/answers/zh/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c03/) - Ch.4 Increments and Sample Sizes: Why Most Increments Cannot Be Measured — Increments can only be measured by holding out a control group, and the samples required are often startling. Detecting a relative increment of 0.10 in a cold audience requires 822,401 people per group; comparing close rates of 0.050 and 0.060 at the conversation level requires 8,158 conversations per group, and with 2,000 per group the power is only 0.287. Revenue is noisier still: with a coefficient of variation of 10, detecting an increment of 0.05 requires 627,911 people per group. What a small or mid-sized seller can do is measure upstream metrics, look only for large effects, alternate by week, borrow the platform's scale and register in advance. (https://5dlibrary.com/answers/en/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c04/) 第4章 增量与样本量:为什么多数增量测不出来 — 增量只能靠留出对照来测,而所需样本量往往大得惊人。冷受众要测出 0.10 的相对增量,每组要 822,401 人;在对话层面比较 0.050 与 0.060 的成交率,每组也要 8,158 段对话,每组 2,000 段时检验力只有 0.287。营业额的噪声更大:变异系数为 10 时,测 0.05 的增量每组要 627,911 人。中小经营者能做的,是测上游指标、只测大效应、按周轮换、借平台规模,并且事先登记。 (https://5dlibrary.com/answers/zh/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c04/) - Ch.5 Three Labels and the Optimal Stop Point: the Mathematics of Consent-Based Follow-Up — Chapters 5 and 6 show that reply rates decay and block risks rise with each follow-up, giving illustrative stop points of three follow-ups for hot chats, one for warm and none for cold; that 60 seconds is not an optimal voice-note length but a robust cap, keeping 0.999 of the best gain in the illustrative model with a median regret of 0.0005 across 135 parameter settings; and that a payment link valid until the real deadline yields payment before it from 0.996 of intending buyers. (https://5dlibrary.com/answers/en/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c05/) 第5章 三标签分层与最优停追点:同意制追单的数学 — 第五、六章说明,回复率逐次衰减、屏蔽风险逐次上升,示意最优停追点为热三次、温一次、冷零次;语音条 60 秒不是最优值而是稳健上限,示意模型下保留最佳增益的 0.999,扫描 135 组参数的遗憾中位数为 0.0005;收款链接开到真实截止,截止前付款比例为 0.996。 (https://5dlibrary.com/answers/zh/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c05/) - Ch.6 The Length Boundary of Real-Person Voice Notes and One-Tap Payment in the Chat — A voice conveys thought and humanity better than text, but the longer a voice note runs, the fewer people hear it to the end. In the illustrative model the best length is 65 seconds and the net gain at 60 seconds is 0.999 of the best; across 135 parameter combinations the best length is 60 seconds or less in only 0.459 of cases, yet the median regret of a 60-second cap is just 0.0005. Sixty seconds is therefore not an optimum but a robust upper limit. Each step removed from a payment link gains the reciprocal of that step's pass rate; link validity should equal the real deadline, and a link valid for the full 72.00 hours yields payment before the deadline from 0.996 of intending buyers. (https://5dlibrary.com/answers/en/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c06/) 第6章 真人语音条的时长边界与对话内一键收款 — 声音比文字更能传达思考与人性,但语音条越长,听完的人越少。示意模型下,最佳时长 65 秒,60 秒的净增益是最佳值的 0.999;扫描 135 组参数,最佳时长不超过 60 秒的只占 0.459,可是 60 秒上限的遗憾中位数只有 0.0005。所以 60 秒不是最优值,而是一条稳健的上限。收款链接每省一步,增益等于那一步通过率的倒数;链接时效应当等于真实截止,开到 72.00 小时的链接,截止前付款比例是 0.996。 (https://5dlibrary.com/answers/zh/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c06/) - Ch.7 Classical Sources East and West: Silence in the Guiguzi, Watching Deeds in the Analects, and the Voice in Classical Rhetoric — Chapter 7 examines the Guiguzi, the Analects, Aristotle and Cicero: reminders about 4 hours apart give the highest reply rate, 0.953, and spoken sources agree with logged codes at a kappa of only 0.511. (https://5dlibrary.com/answers/en/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c07/) 第7章 东西方原典考辨:鬼谷子论反默、论语论观行与古典修辞论声音 — 第七章考辨《鬼谷子·反应》《论语·公冶长》、亚里士多德与西塞罗,提醒间隔约 4 小时时回复率最高,为 0.953,口头来源与代码的一致系数只有 0.511。 (https://5dlibrary.com/answers/zh/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c07/) - Ch.8 Actuarial Tools and Execution Checklist: the Weekly Chat Ledger, the Cadence Checker and Pre-Registered Tests — Chapter 8 provides a weekly ledger, the Click-to-Chat Funnel and Three-Label Follow-Up Cadence Checker and pre-registered tests. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c08/) 第8章 精算工具与执行清单:对话成交周度台账、节奏检查器与预注册检验 — 第八章给出周度台账、点击直达对话漏斗与三标签追单节奏检查器与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/click-to-messenger-funnels-chat-attribution-voice-note-closing-southeast-asia/c08/) ## 5DL-B-367 · Stacking Monthly Recurring Revenue, Involuntary-Churn Dunning Recovery and Upfront Annual Cash Lock-In 中文书名:月度经常性收入叠砖、扣款失败自动挽回与年付现金流锁定母典 DOI: pending Cite: Fan, Y. (2026). Stacking Monthly Recurring Revenue, Involuntary-Churn Dunning Recovery and Upfront Annual Cash Lock-In (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its subscriber counts, prices, churn, recovery and discount rates are illustrative, all calculations are generated by its programs, and none corresponds to a real customer or transaction. It is not legal, accounting, tax or investment advice and promises no level of income; "steady" means only that revenue can be anticipated, with premises and failure conditions stated. Abstract: Three popular beliefs promise to end the anxiety of revenue that "resets to zero on the first of the month": a subscription is just a one-off price split into twelve, so grow the subscriber count and retention will follow; all churn is customers who no longer want the product, and a failed charge only needs retrying; and an annual discount is a giveaway. This book refutes each belief or draws its boundary. Its subscriber counts, prices, churn, recovery and discount rates are illustrative, all calculations are generated by its programs, and none corresponds to a real customer or transaction. It is not legal, accounting, tax or investment advice and promises no level of income; "steady" means only that revenue can be anticipated, with premises and failure conditions stated. Chapters 1 to 3 show that steady-state revenue equals monthly sign-ups times revenue per subscriber divided by monthly churn, so sign-ups set the slope and churn sets the ceiling; that with heterogeneous subscribers aggregate retention rises with age and the shifted-beta-geometric model projects it from a few months of cohort data; that the first two renewals are a sieve where each intervention returns 1.40 times as much as later, while the three-year gain lies in the tail, 18,598 against 4,760 with illustrative parameters; and that a subscription layer stands only when must-have uses clearly exceed the value threshold. Chapter 4 classifies passive churn into insufficient funds, expired cards, risk declines and hard declines and schedules retries within public card-network caps, lifting recovery of all failures from 0.232 to 0.627; recovering seven tenths lifts the ceiling to 1.762 times its level. Hard declines are never retried, and card details stay with licensed payment institutions. Chapter 5 solves the annual tipping point, a monthly churn of 0.0242 when money costs 0.01 a month and two months are free; annual takers lean to low churn and behave as if their churn were 0.0220, so a naive gain of 257.5 becomes -35.8, and prepayments are recognised monthly with a refund reserve. Chapter 6 shows that retention targets should be chosen by lift, with clear pre-renewal reminders and cancellation as easy as subscription. Chapter 7 examines Mencius, Guanzi, Franklin and Aesop; Chapter 8 gives the monthly subscription cash ledger, the Subscription Stacking, Dunning Recovery and Annual Tipping Calculator, an execution checklist and a pre-registered test plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:想消灭「每月一号收入归零」的焦虑,常听到三条流行说法:订阅就是把买断价拆成十二期,先做大订阅数,留存自然会来;流失都是客户不想要了,扣款失败多试几次就行;年付打折就是让利。本书逐条证伪或立界。书中的订阅数、客单、流失率、挽回率、折现率与价格均为示意参数,全部计算由书中程序生成,不对应任何真实客户或成交数据;本书是非法律、会计、税务或投资建议,不承诺任何收入水平,「稳定」只指收入可预期并写明前提与失效条件。 第一至三章说明,稳态收入等于每月新增乘以客单再除以月流失率,新增定斜率,流失率定天花板;订阅者异质时总体留存随月龄上升,移位贝塔几何模型用几个月的队列数据就能外推;前两次续费是筛子,单次干预回报是后段的 1.40 倍,三年总增益却在长尾,示意参数下是 18,598 对 4,760;订阅层只在刚需使用次数明显高于值回票价的门槛时成立。第四章把被动流失按余额不足、卡片过期、风控拒绝与硬拒绝分类,在卡组织公开上限之内排重试,全部失败的挽回率从 0.232 升到 0.627,挽回七成可把天花板抬到 1.762 倍;硬拒绝不重试,卡片信息只由持牌支付机构保管。 第五章解出年付拐点,资金成本每月 0.01、送两个月时为月流失率 0.0242;选年付的人天然偏向低流失,接受者的行为相当于 0.0220,朴素增益 257.5 修正后为 -35.8,预收款按月确认并预留退款。第六章说明挽留对象按增益挑选,续费前清楚提醒、取消与订阅同等便捷。第七章考辨孟子、管子、富兰克林与伊索,第八章给出订阅现金流月度台账、订阅叠砖留存与扣款挽回年付拐点试算器、执行清单与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Bathtub Steady State: New Sign-ups Set the Slope, Churn Sets the Ceiling — Chapters 1 to 3 show that steady-state revenue equals monthly sign-ups times revenue per subscriber divided by monthly churn, so sign-ups set the slope and churn sets the ceiling; that with heterogeneous subscribers aggregate retention rises with age and the shifted-beta-geometric model projects it from a few months of cohort data; that the first two renewals are a sieve where each intervention returns 1.40 times as much as later, while the three-year gain lies in the tail, 18,598 against 4,760 with illustrative parameters; and that a subscription layer stands only when must-have uses clearly exceed the value threshold. (https://5dlibrary.com/answers/en/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c01/) 第1章 浴缸稳态:新增只定斜率,流失率才定天花板 — 第一至三章说明,稳态收入等于每月新增乘以客单再除以月流失率,新增定斜率,流失率定天花板;订阅者异质时总体留存随月龄上升,移位贝塔几何模型用几个月的队列数据就能外推;前两次续费是筛子,单次干预回报是后段的 1.40 倍,三年总增益却在长尾,示意参数下是 18,598 对 4,760;订阅层只在刚需使用次数明显高于值回票价的门槛时成立。 (https://5dlibrary.com/answers/zh/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c01/) - Ch.2 Stacking Cohorts: the Shifted-Beta-Geometric Model and the First Two Renewals — Retention in the first period is 0.847, which means first-month churn of 0.153. Every later period is steadier than the one before: 0.870 in the second, 0.917 in the sixth, 0.947 in the twelfth, 0.969 in the twenty-fourth and 0.978 in the thirty-sixth. The surviving share falls from 0.847 to 0.330 at month twelve and 0.204 at month twenty-four. The running product and the closed form agree to three decimals at every horizon. (https://5dlibrary.com/answers/en/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c02/) 第2章 队列叠砖:移位贝塔几何模型与前两次续费 — 第一期留存率是 0.847,也就是首月流失 0.153。此后每一期都比上一期稳:第二期 0.870,第六期 0.917,第十二期 0.947,第二十四期 0.969,第三十六期 0.978。存活比例从 0.847 降到第十二个月的 0.330,第二十四个月的 0.204。逐期连乘与闭式解在三位小数上完全一致。 (https://5dlibrary.com/answers/zh/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c02/) - Ch.3 Carving Out a Subscription Layer: the Ratio of Must-Have Use to the Billing Period — A subscription layer stands up when must-have uses in each billing period clearly exceed the value threshold. In an uneven population the ratio must reach about 2 for a first-renewal rate of seven in ten; in an even population about 1.31 suffices. Even at high usage, part of an uneven population does not use the product at all in a period, so first-month churn has a floor set by composition. With illustrative parameters, a monthly layer beats an outright sale only at about six must-have uses a month, and much of the extra money that quarterly billing collects comes from customers who barely use the product, which makes it the most fragile revenue in the base. (https://5dlibrary.com/answers/en/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c03/) 第3章 切出订阅层:刚需频率与计费周期之比 — 订阅层成立的条件,是每个计费周期内的刚需使用次数明显高于值回票价的门槛。在参差人群里,比值要到 2 左右,首次续费率才到七成;人群整齐时,1.31 倍左右即可。即使使用频率很高,参差人群里仍有一部分人整个周期不用,首月流失有一个由构成决定的底。示意参数下,按月订阅层要到每月 6 次刚需使用才胜过买断;按季计费多收的钱,有相当一部分来自不怎么用的客户,这部分收入最脆弱。 (https://5dlibrary.com/answers/zh/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c03/) - Ch.4 Failed Charges: Decline Reasons and Retry Schedules for Involuntary Churn — Chapter 4 classifies passive churn into insufficient funds, expired cards, risk declines and hard declines and schedules retries within public card-network caps, lifting recovery of all failures from 0.232 to 0.627; recovering seven tenths lifts the ceiling to 1.762 times its level. Hard declines are never retried, and card details stay with licensed payment institutions. (https://5dlibrary.com/answers/en/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c04/) 第4章 扣款失败:被动流失的原因分类与重试排程 — 第四章把被动流失按余额不足、卡片过期、风控拒绝与硬拒绝分类,在卡组织公开上限之内排重试,全部失败的挽回率从 0.232 升到 0.627,挽回七成可把天花板抬到 1.762 倍;硬拒绝不重试,卡片信息只由持牌支付机构保管。 (https://5dlibrary.com/answers/zh/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c04/) - Ch.5 Annual Lock-In: the NPV Tipping Point of Two Free Months and the Self-Selection Correction — Chapter 5 solves the annual tipping point, a monthly churn of 0.0242 when money costs 0.01 a month and two months are free; annual takers lean to low churn and behave as if their churn were 0.0220, so a naive gain of 257.5 becomes -35.8, and prepayments are recognised monthly with a refund reserve. (https://5dlibrary.com/answers/en/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c05/) 第5章 年付锁定:送两个月的净现值拐点与自选择修正 — 第五章解出年付拐点,资金成本每月 0.01、送两个月时为月流失率 0.0242;选年付的人天然偏向低流失,接受者的行为相当于 0.0220,朴素增益 257.5 修正后为 -35.8,预收款按月确认并预留退款。 (https://5dlibrary.com/answers/zh/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c05/) - Ch.6 Whom to Retain and Transparent Renewal: Risk times Recoverability, and Revenue That Does Not Rest on Inattention — Retention targets should be chosen by lift, not by risk. With illustrative parameters, selecting by risk performs almost exactly like selecting at random, while selecting by estimated lift raises net value from 19,328 to 29,686. The highest-risk subscribers have on average only a third of the lift of middling-risk ones. Perhaps a third of subscription revenue comes from subscribers who have not noticed they want to leave; this ice melts at card replacement, at reminders and under regulation, and planning should use the churn rate measured with reminders. The cancel path must be as easy as the subscribe path, and contrast, tap height and step count can all be checked by a program. (https://5dlibrary.com/answers/en/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c06/) 第6章 挽留对象与透明续费:风险乘可挽回性,以及不靠疏忽的收入 — 挽留对象按增益挑,不按风险挑:示意参数下,按风险挑与随机挑几乎一样,按估计增益挑能把净值从 19,328 提高到 29,686。最高风险的人平均增益只有中等风险者的三分之一。订阅收入里可能有三成多来自没察觉的订阅者,这层冰会在换卡、提醒与监管面前化掉;规划要用发提醒后的流失率。取消路径必须与订阅路径同等便捷,对比度、点按高度与步数都可以用程序检验。 (https://5dlibrary.com/answers/zh/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c06/) - Ch.7 Classical Sources East and West: Steady Livelihood and Steady Heart, Granaries, Breeding Money and the Ant's Store — Chapter 7 examines Mencius, Guanzi, Franklin and Aesop; (https://5dlibrary.com/answers/en/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c07/) 第7章 东西方原典考辨:恒产与恒心、仓廪、生息之钱与蚂蚁的储备 — 第七章考辨孟子、管子、富兰克林与伊索,第八章给出订阅现金流月度台账、订阅叠砖留存与扣款挽回年付拐点试算器、执行清单与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c07/) - Ch.8 Actuarial Tools and Execution Checklist: the Monthly Subscription Cash Ledger, the Calculator and Pre-Registered Tests — The ledger has one row per month and one sheet per subscription layer; it separates voluntary from passive churn, classifies failures into four reasons, and keeps annual plans and deferred revenue on lines of their own. The calculator turns the ledger into a steady-state ceiling, recovery by reason, next month's retry days, the annual tipping point and alerts; on the demonstration ledger the ceiling is 251,440 units and two alerts light up. The execution checklist fixes actions by day, week, month and quarter. The pre-registered tests give sample sizes: 356 failures per arm for the retry schedule, 691 subscribers per arm for the first renewal, and a cohort of 680 newcomers for the test of rising retention. (https://5dlibrary.com/answers/en/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c08/) 第8章 精算工具与执行清单:订阅现金流月度台账、试算器与预注册检验 — 台账每月一行,按订阅层分表,主动与被动流失分开、失败按四类原因登记、年付与递延收入单列。试算器按台账算出稳态收入、各原因挽回率、下月重试日、年付拐点与警示,示范台账的稳态收入是 251,440 个单位,亮了两盏灯。执行清单把动作落到日、周、月、季。预注册检验给出样本量:重试排程每组 356 笔,首次续费每组 691 位,留存上升检验一个队列 680 位。 (https://5dlibrary.com/answers/zh/mrr-stacking-involuntary-churn-dunning-recovery-annual-prepay-lock-in/c08/) ## 5DL-B-368 · Demo-First B2B Closing, Forty-Eight-Hour Turnkey System Delivery and Scope-Creep Containment 中文书名:半成品演示先行逼单、四十八小时交钥匙系统交付与防需求蔓延母典 DOI: pending Cite: Fan, Y. (2026). Demo-First B2B Closing, Forty-Eight-Hour Turnkey System Delivery and Scope-Creep Containment (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its win rates, hours, margins, upfront shares, fees and prices are illustrative, counted in units that match no currency, client or real deal, and all calculations are generated by its programs. It is not legal advice and not individual business or financial advice; contract terms need review by a locally admitted lawyer. Abstract: Turning rapid site building and automation into a large upfront payment plus a long-term hosting fee runs into four popular beliefs: a thick proposal makes a vendor look professional; forty-eight-hour delivery is a matter of working through the night; free small changes buy goodwill; and a hosting fee is either a pleasant extra or something kept by holding the client's domain and data. This book refutes each belief or draws its boundary. Its win rates, hours, margins, upfront shares, fees and prices are illustrative, counted in units that match no currency, client or real deal, and all calculations are generated by its programs. It is not legal advice and not individual business or financial advice; contract terms need review by a locally admitted lawyer. It contains no pressure copy, no resetting countdown and no promise of income, and referral rewards are single-level only. Chapters 1 and 2 derive the break-even line of demo-first selling: a 3-hour preview against a margin of 4,400 units needs a win-rate jump of only 0.0409. But the endowment effect is disputed and the personalisation evidence concerns e-mail opens; detecting a jump from 0.20 to 0.30 at 12 pitches a month takes 49.0 months, so the book reads posterior probabilities of clearing the line instead. The optimal preview completeness is about 0.289, and a complete preview does worse than none. Chapters 3 and 4 turn forty-eight hours into a probability: a seven-module order planned at 12.50 hours is on time with probability only 0.846; the illustrative catalogue caps orders at 6 modules; an end-of-chain buffer commits to 14.73 hours with probability 0.997, against 0.721 for padding every task; and one off-catalogue integration drags the probability down to 0.632. Chapter 5 shows that free revisions push nearly half of all orders below a margin of 4,000 units, while charging a single unit per change cuts the request rate from 6.0 to 2.49, and that a change-pricing lock stretches the range of a fixed price from 4.34 changes to 66.59. Chapter 6 shows that three tenths upfront gives the highest expected net value and that a hosting client with all visible maintenance items has a lifetime margin of 2,589 units; retention rests on visible value alone, with domain, accounts and data owned by the client. Chapter 7 examines the Mozi, the Kaogong ji, the Gospel of Luke and Vitruvius; Chapter 8 delivers the actuarial tool, the execution checklist and pre-registered tests. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:把极速建站与自动化能力变成高额首付款加长期托管费,路上有四条流行说法:先写厚方案书才显得专业;四十八小时交付靠熬夜硬拼;免费小改能换好感;托管月费只是锦上添花,或者要靠扣住域名与数据才留得住。本书逐条证伪或立界。书中的赢率、工时、毛利、首付比例、月费与价格均为示意参数,以「单位」计,不对应任何币种、真实客户或真实成交;全部计算由书中程序生成。本书不是法律意见,也不是个人化经营或理财建议,合同条款须由当地执业律师审阅;书中不写威胁催促、不写会重置的倒计时、不写收益承诺,推荐奖励只写单层。 第一、二章推导演示先行的盈亏线:3 小时预览配 4,400 单位毛利,赢率只需抬高 0.0409;但禀赋效应有争议,姓名个性化证据来自邮件打开率,小工作室要验出 0.20 到 0.30 的跃升,按每月 12 次报价要攒 49.0 个月,所以改看后验越线概率。预览的最优完成度约为 0.289,做满反而不如不做。第三、四章把四十八小时改写成准时概率:单点计划 12.50 小时的七模块订单,准时概率只有 0.846;示意目录的接单上限是 6 个模块;链尾缓冲承诺 14.73 小时、准时概率 0.997,逐项加垫只有 0.721;一个目录外接入就能把准时概率拖到 0.632。 第五章说明,免费改稿让近一半订单毛利跌破 4,000 单位,只要每次收 1 单位,请求率就从 6.0 次降到 2.49 次;变更计价锁把固定价的适用范围从 4.34 次变更延伸到 66.59 次。第六章说明,首付三成的期望净值最高,托管客户给满可见运维项目时终身毛利达 2,589 单位,留存只靠可见价值,域名、账号与数据归属客户。第七章考辨《墨子·公输》《考工记》《路加福音》与维特鲁威,第八章给出精算工具、执行清单与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Break-Even Line of Demo-First Selling: Win-Rate Jumps, the Endowment Dispute and Name Personalisation — Chapters 1 and 2 derive the break-even line of demo-first selling: a 3-hour preview against a margin of 4,400 units needs a win-rate jump of only 0.0409. But the endowment effect is disputed and the personalisation evidence concerns e-mail opens; detecting a jump from 0.20 to 0.30 at 12 pitches a month takes 49.0 months, so the book reads posterior probabilities of clearing the line instead. The optimal preview completeness is about 0.289, and a complete preview does worse than none. (https://5dlibrary.com/answers/en/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c01/) 第1章 演示先行的盈亏线:赢率跃升、禀赋之争与姓名个性化 — 第一、二章推导演示先行的盈亏线:3 小时预览配 4,400 单位毛利,赢率只需抬高 0.0409;但禀赋效应有争议,姓名个性化证据来自邮件打开率,小工作室要验出 0.20 到 0.30 的跃升,按每月 12 次报价要攒 49.0 个月,所以改看后验越线概率。预览的最优完成度约为 0.289,做满反而不如不做。 (https://5dlibrary.com/answers/zh/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c01/) - Ch.2 How Complete Should a Preview Be? The Saturating Win-Rate Curve, the Optimal Completeness and Preview Compliance — The win rate of a preview rises along a saturating curve, its hours rise along a convex one, and expected net value peaks in between. Under the illustrative parameters the optimal completeness is about 0.289, and a complete preview loses more than no preview at all. The higher the margin, the higher the optimum; the faster the win rate saturates, the earlier the preview should stop. The IKEA effect requires the client's own labour and a finished result, so it cannot be borrowed for half-built previews and must instead be tested as a hypothesis about client participation. Before any client sees a preview, it must pass five checks: source of materials, trademarks, indexing, contrast and tap height. (https://5dlibrary.com/answers/en/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c02/) 第2章 预览做到几成:完成度饱和曲线、最优完成度与预览合规 — 预览的赢率沿饱和曲线上升,工时沿凸曲线上升,期望净值在中间达到最大。示意参数下最优完成度约为 0.289,做满反而比不做还亏。毛利越高,最优完成度越高;赢率越快饱和,越应早停。宜家效应要求客户自己出力且劳动完成,不能直接借来论证半成品预览,只能写成「客户参与」的假说去检验。预览给客户看之前,要过素材来源、商标、索引、对比度与点按高度五道检查。 (https://5dlibrary.com/answers/zh/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c02/) - Ch.3 Forty-Eight Hours Is a Distribution: Sums of Module Hours, the Probability of Being on Time and the Planning Fallacy — Chapters 3 and 4 turn forty-eight hours into a probability: a seven-module order planned at 12.50 hours is on time with probability only 0.846; the illustrative catalogue caps orders at 6 modules; an end-of-chain buffer commits to 14.73 hours with probability 0.997, against 0.721 for padding every task; and one off-catalogue integration drags the probability down to 0.632. (https://5dlibrary.com/answers/en/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c03/) 第3章 四十八小时是一个分布:模块工时之和、准时概率与规划谬误 — 第三、四章把四十八小时改写成准时概率:单点计划 12.50 小时的七模块订单,准时概率只有 0.846;示意目录的接单上限是 6 个模块;链尾缓冲承诺 14.73 小时、准时概率 0.997,逐项加垫只有 0.721;一个目录外接入就能把准时概率拖到 0.632。 (https://5dlibrary.com/answers/zh/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c03/) - Ch.4 Put the Buffer at the End of the Chain: Critical Chain, Fat-Tailed Modules and the Limits of Intake — Padding every task bloats the plan and, because early finishes never hand over, digests only bad news: the padded plan of the illustrative six-module order is 15.91 hours, with a probability of being on time of only 0.721. An end-of-chain buffer, with tasks at their medians and immediate handover, commits to 14.73 hours with a probability of 0.997. The root-sum-square rule underestimates the buffer when two lanes run in parallel and serves only as a lower bound. Off-catalogue modules have fat tails: a single custom integration planned at 3 hours can drag the probability of being on time down to 0.632, so every such module leaves the forty-eight-hour order. (https://5dlibrary.com/answers/en/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c04/) 第4章 缓冲放在链尾:关键链、肥尾模块与接单边界 — 逐项加垫让计划虚胖,又因为早完成不交棒而只吃坏消息:示意六模块订单的加垫计划 15.91 小时,准时概率只有 0.721。链尾缓冲按中位排、早完成即交棒,承诺 14.73 小时,准时概率 0.997。平方和开方法会低估两条线并行时的缓冲,只能当下限。目录外模块有肥尾,加一个计划 3 小时的定制接入就能把准时概率拖到 0.632,所以一律移出四十八小时订单。 (https://5dlibrary.com/answers/zh/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c04/) - Ch.5 Free Revisions and Scope Creep: Poisson Request Rates, the Change-Pricing Lock and the Choice of Contract — Chapter 5 shows that free revisions push nearly half of all orders below a margin of 4,000 units, while charging a single unit per change cuts the request rate from 6.0 to 2.49, and that a change-pricing lock stretches the range of a fixed price from 4.34 changes to 66.59. (https://5dlibrary.com/answers/en/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c05/) 第5章 零价改稿与需求蔓延:泊松请求率、变更计价锁与合同选型 — 第五章说明,免费改稿让近一半订单毛利跌破 4,000 单位,只要每次收 1 单位,请求率就从 6.0 次降到 2.49 次;变更计价锁把固定价的适用范围从 4.34 次变更延伸到 66.59 次。 (https://5dlibrary.com/answers/zh/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c05/) - Ch.6 Upfront Payments, Milestones and the Hosting Retainer: Expected Cash and Retention Without Hostages — Chapter 6 shows that three tenths upfront gives the highest expected net value and that a hosting client with all visible maintenance items has a lifetime margin of 2,589 units; retention rests on visible value alone, with domain, accounts and data owned by the client. (https://5dlibrary.com/answers/en/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c06/) 第6章 首付款、里程碑与托管月费:期望现金与不扣押的留存 — 第六章说明,首付三成的期望净值最高,托管客户给满可见运维项目时终身毛利达 2,589 单位,留存只靠可见价值,域名、账号与数据归属客户。 (https://5dlibrary.com/answers/zh/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c06/) - Ch.7 Classical Sources East and West: Mozi on Gongshu Ban, the Kaogong ji, the Gospel of Luke and Vitruvius — Chapter 7 examines the Mozi, the Kaogong ji, the Gospel of Luke and Vitruvius; (https://5dlibrary.com/answers/en/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c07/) 第7章 东西方原典考辨:《墨子·公输》、《考工记》、《路加福音》与维特鲁威 — 第七章考辨《墨子·公输》《考工记》《路加福音》与维特鲁威,第八章给出精算工具、执行清单与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c07/) - Ch.8 The Empirical Loop: Actuarial Tools and Execution Checklist for Demo and Delivery — The actuarial tool combines the formulas of the first six chapters into one checker, and run without arguments it reproduces the demonstration above. The execution checklist sets out the ledger fields, the daily, weekly and monthly rhythm, and what to do, and what to stop, when each threshold is crossed. The pre-registered tests fix sample sizes and refutation conditions in advance: 231 pitches per arm for the preview jump, 59 orders for the forty-eight-hour promise and 6 orders per arm for the zero-price effect. (https://5dlibrary.com/answers/en/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c08/) 第8章 实证闭环:演示与交付的精算工具与执行清单 — 精算工具把前六章的公式合成一个检查器,无参数运行即复现示范结果。执行清单规定台账字段、每日每周每月的节奏,以及每个阈值亮灯以后做什么、停什么。预注册检验事先写明了样本量与推翻条件:预览跃升每组 231 次报价,四十八小时承诺 59 份订单,零价效应每组 6 份订单。 (https://5dlibrary.com/answers/zh/demo-first-b2b-closing-48-hour-turnkey-delivery-scope-creep-containment/c08/) ## 5DL-B-369 · Local-Pack Search Ranking Factors, Review-Velocity Signalling and High-Urgency Geo-Intent Lead Capture 中文书名:本地搜索地图排名因子、评论速度信号权重与同城高急迫线索截流母典 DOI: pending Cite: Fan, Y. (2026). Local-Pack Search Ranking Factors, Review-Velocity Signalling and High-Urgency Geo-Intent Lead Capture (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. This is not legal or marketing-compliance advice; platform policies and laws are governed by their current official texts and change over time. Abstract: Same-city businesses repeat three popular claims: more good reviews mean a higher place on the map, a small gift for five stars is the fastest way to get reviews, and once the listing ranks the phone only has to be picked up. This book refutes or bounds each one. Google files local ranking under relevance, distance and prominence, and reviews are only part of prominence; a practitioner survey ranks the primary category and distance ahead of reviews, though it records expert opinion rather than a causal experiment. The book writes the three factors as an illustrative multiplicative score whose weights are illustrative parameters, not Google's weights: a store two kilometres away with a mis-registered category would need 8,511 reviews to match the effect of correcting its primary category. Review incentives, invitations screened for satisfied customers, keyword-stuffed names and virtual addresses all breach platform policy and appear only as counter-examples; in expectation, inviting only satisfied customers lifts the displayed mean by 0.443 stars. Parameters are illustrative and all calculations are generated by the book's programs. This is not legal or marketing-compliance advice; platform policies and laws are governed by their current official texts and change over time. Chapters 1 and 2 cover the product structure of the three factors, Huff's gravity model and distance decay, and how to read a five-by-five rank grid and its scan noise. Chapter 3 covers prominence: the logarithmic margin of review counts, the shock of one bad review under a Bayesian average, and half-life weighting of recent reviews. Chapter 4 covers compliant review velocity: an invitation funnel that treats every customer alike, exact Poisson intervals, and "top three on the map" written as a simulated quantile; at 12 reviews a month the probability of reaching the goal within a year is 0.843. Chapter 5 covers capture: at only four calls an hour a single answering seat picks up just 0.6812 of calls within twenty seconds, and fixing the answer rate and the call-back deadline multiplies bookings by 1.437, about as much as one more place in the ranking. Chapter 6 gives decay constants and a call-back order for three tiers of urgency, today, this week and this month, with urgency drawn only from the customer's own deadline and genuine availability. Chapter 7 examines Mencius on timing, place and harmony, the Analects on virtue that finds neighbours, Proverbs 22:1 and Tobler's first law of geography; only 0.1250 of random draws satisfy Mencius's strict ordering, and the chapter closes with a distillation spec for the four kits. Chapter 8 delivers a weekly ledger, the distilled Same-City Map Exposure and Enquiry Capture Calculator, red, amber and green risk-control tables and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:同城生意最流行的三句话是:好评越多地图排名越高,送小礼物换五星最快,排上去以后电话接起来就行。本书逐条证伪或立界。谷歌把本地排名归为相关性、距离与知名度,评论只是知名度的一部分;从业者问卷把主类目与距离排在评论之前,但那只是专家意见,不是因果实验。本书把三因子写成乘积式示意打分,全部权重都是示意参数,不是谷歌真实权重:一家距离两公里、类目登记错误的示意门店,要靠评论追平改对主类目的效果,需要 8,511 条评论。激励好评、只邀请满意客人、名称堆词与虚拟地址都违反平台政策,只作反例;按期望值计算,只邀请满意客人会把显示均值抬高 0.443 星。书中参数为示意,全部计算由书中程序生成。本书不构成法律或营销合规意见,平台政策与法规以官方现行条文为准并会变动。 第一、二章讲三因子的乘积结构、赫夫引力与距离衰减,以及 5 × 5 方格名次图的读法与扫描噪声。第三章讲知名度:评论数的对数边际、贝叶斯平均下一条差评的冲击,以及近期评论的半衰期加权。第四章讲合规评论速度:全部客人一视同仁的邀请漏斗、精确泊松区间,以及把「挤进前三名」写成模拟分位数:每月 12 条评论时,一年内达标的概率是 0.843。第五章讲承接:每小时只有 4 通电话时,单人值守的二十秒接起率也只有 0.6812;把接起率与回拨时限改好,预约量变成 1.437 倍,与名次升一位相当。 第六章讲今天、本周、本月三层急迫意图的衰减常数与回拨排序,急迫感只来自客人自述的时限与真实档期。第七章考辨《孟子》「天时不如地利,地利不如人和」、《论语》「德不孤,必有邻」、《箴言》第 22 章第 1 节与托布勒的地理学第一定律,抽样检验后只有 0.1250 的抽样满足孟子的严格排序,章末给出四件套蒸馏规格。第八章交出同城地图周度台账、蒸馏小工具「同城地图曝光与询盘承接试算器」、红黄绿风控表与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Three Factors Multiplied: the Product Structure of Relevance, Distance and Prominence — Chapters 1 and 2 cover the product structure of the three factors, Huff's gravity model and distance decay, and how to read a five-by-five rank grid and its scan noise. (https://5dlibrary.com/answers/en/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c01/) 第1章 三因子相乘:相关性、距离与知名度的乘积结构 — 第一、二章讲三因子的乘积结构、赫夫引力与距离衰减,以及 5 × 5 方格名次图的读法与扫描噪声。 (https://5dlibrary.com/answers/zh/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c01/) - Ch.2 Distance Decay: Gravity Curves, the Edge of the Trade Area and a Five-by-Five Rank Grid — The shape of distance decay is not published, and the exponential, power and Gaussian forms each have a case. Whichever is used, ring area puts a large share of exposure in the outer rings: under the exponential curve only 0.1489 of exposure falls within one kilometre, and half is passed only at three kilometres. On a rank grid, the category lever lifts the whole map, while the review lever works only at the edge. Scans are noisy, and telling apart a change of three points needs 8 scans in each arm. (https://5dlibrary.com/answers/en/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c02/) 第2章 距离衰减:引力曲线、商圈边缘与 5 × 5 方格名次图 — 距离衰减的形状没有公开,指数、幂函数与高斯三种写法各有道理。不管哪条,圆环面积都让曝光大量落在外圈:在指数曲线下,1 公里内只有 0.1489 的曝光,3 公里内才过半。方格名次图上,类目杠杆整体抬升,评论杠杆只在边缘点起作用。扫描本身有噪声,分辨三个点的变化,每组需要 8 次扫描。 (https://5dlibrary.com/answers/zh/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c02/) - Ch.3 Prominence: the Logarithmic Margin of Review Counts, Bayesian Averages and Recent Velocity — Prominence has three parts. The count term is logarithmic: the marginal gain from the next review falls from 0.03008 to 0.00010, and an established store cannot pile up enough reviews to offset half a kilometre. The rating term is sensitive to bad reviews: with only 10 reviews, a single one-star review needs 63 further reviews to absorb. The velocity term is half-life weighted: the index bought by a burst is a remnant after six months, and only a steady flow of customers gives a steady index. (https://5dlibrary.com/answers/en/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c03/) 第3章 知名度:评论数的对数边际、贝叶斯平均与近期速度 — 知名度由三块组成。条数项是对数式的,下一条评论的边际增益从 0.03008 一路跌到 0.00010,老店堆评论抵不了半公里的距离。评分项对差评敏感,只有 10 条评论时,1 条一星要 63 条新评论才能消化。速度项是半衰期加权,突击换来的指数半年后只剩零头,稳定的来客才有稳定的指数。 (https://5dlibrary.com/answers/zh/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c03/) - Ch.4 Compliant Review Velocity: the Invitation Funnel, Poisson Intervals and "Top Three" as a Simulated Quantile — Chapter 4 covers compliant review velocity: an invitation funnel that treats every customer alike, exact Poisson intervals, and "top three on the map" written as a simulated quantile; at 12 reviews a month the probability of reaching the goal within a year is 0.843. (https://5dlibrary.com/answers/en/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c04/) 第4章 合规评论速度:邀请漏斗、泊松区间与「前三名」的模拟分位数 — 第四章讲合规评论速度:全部客人一视同仁的邀请漏斗、精确泊松区间,以及把「挤进前三名」写成模拟分位数:每月 12 条评论时,一年内达标的概率是 0.843。 (https://5dlibrary.com/answers/zh/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c04/) - Ch.5 Capture: Erlang Queues, the Twenty-Second Answer Rate and Answering Seats — Chapter 5 covers capture: at only four calls an hour a single answering seat picks up just 0.6812 of calls within twenty seconds, and fixing the answer rate and the call-back deadline multiplies bookings by 1.437, about as much as one more place in the ranking. (https://5dlibrary.com/answers/en/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c05/) 第5章 承接:爱尔朗排队、二十秒接起率与值守席位 — 第五章讲承接:每小时只有 4 通电话时,单人值守的二十秒接起率也只有 0.6812;把接起率与回拨时限改好,预约量变成 1.437 倍,与名次升一位相当。 (https://5dlibrary.com/answers/zh/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c05/) - Ch.6 Tiers of Urgency: Today, This Week, This Month and the Order of Call-Backs — Same-city enquiries fall into three tiers by the customer's stated deadline: today, this week and this month. The today tier loses 0.02222 of its value per minute, nearly a hundred times the rate of the month tier. Call-backs should go to the most urgent first and, within a tier, to the freshest, but no call may exceed its tier's deadline. The weekly simulation shows that with the owner alone, switching to most urgent first lifts the today tier's retention from 0.4885 to 0.5905, while adding one duty seat lifts it to 0.9636. The rule is free; the seat is the main course. (https://5dlibrary.com/answers/en/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c06/) 第6章 急迫度分层:今天、本周、本月三层意图与回拨排序 — 同城询盘按客人自述的时限分三层:今天、本周、本月。今天层每分钟流失 0.02222,是本月层的近百倍。回拨顺序应当急者优先,同一层里先救还新鲜的,但任何一通都不能超过本层时限。一周模拟显示,单人值守时改用急者优先,今天层保留从 0.4885 升到 0.5905;加 1 个值守席位,则升到 0.9636。改规则免费,加席位是主菜。 (https://5dlibrary.com/answers/zh/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c06/) - Ch.7 Classical Sources: Timing, Place and Harmony, Virtue with Neighbours, and a Good Name over Great Riches — Chapter 7 examines Mencius on timing, place and harmony, the Analects on virtue that finds neighbours, Proverbs 22:1 and Tobler's first law of geography; only 0.1250 of random draws satisfy Mencius's strict ordering, and the chapter closes with a distillation spec for the four kits. (https://5dlibrary.com/answers/en/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c07/) 第7章 原典考辨:天时地利人和、德不孤必有邻与「美名胜过大财」 — 第七章考辨《孟子》「天时不如地利,地利不如人和」、《论语》「德不孤,必有邻」、《箴言》第 22 章第 1 节与托布勒的地理学第一定律,抽样检验后只有 0.1250 的抽样满足孟子的严格排序,章末给出四件套蒸馏规格。 (https://5dlibrary.com/answers/zh/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c07/) - Ch.8 Closing the Loop: the Weekly Same-City Map Ledger, Calculators and Risk-Control Tables — Chapter 8 delivers a weekly ledger, the distilled Same-City Map Exposure and Enquiry Capture Calculator, red, amber and green risk-control tables and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c08/) 第8章 实证闭环:同城地图周度台账、试算工具与风控表 — 第八章交出同城地图周度台账、蒸馏小工具「同城地图曝光与询盘承接试算器」、红黄绿风控表与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/local-pack-ranking-factors-review-velocity-geo-intent-lead-capture/c08/) ## 5DL-B-370 · Long-Form Video Search Half-Lives, Deep-Trust Pre-Selling and Multi-Year Passive Acquisition Assets 中文书名:长视频搜索半衰期红利、深度信任预售与多年期被动获客资产母典 DOI: pending Cite: Fan, Y. (2026). Long-Form Video Search Half-Lives, Deep-Trust Pre-Selling and Multi-Year Passive Acquisition Assets (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. This is general business and marketing analysis, not marketing or investment advice; platform rules change. Abstract: Three sayings circulate among people who win customers with content: short videos get the traffic and nobody watches long ones; long videos have poor completion, so a video that cannot hold people for 30 seconds is worthless; and nobody clicks the links in a description box. This book falsifies or bounds each of them, and it rewrites the slogans "rots within 48 hours" and "brings in clients for three years" as testable half-life parameters rather than facts or promises. Under illustrative curves a short clip collects 0.750 of its three-year views in the first 48 hours, while a long video with a search floor collects only 0.059 and still draws 5.00 views a day on day 1,095; when the half-life is ten times longer, a short item needs a peak 9.96 times higher to match the present value. Checking the originals corrects the blueprint: the 2007 data of Cha and colleagues show both fast decay per video and daily requests dominated by videos over a month old; the link between referrer types and popularity trends is in the 2014 extended paper of Figueiredo and colleagues; The Long Tail is a popular book; and the sleeper effect has not replicated reliably. Parameters are illustrative and all calculations are generated by the book's programs. This is general business and marketing analysis, not marketing or investment advice; platform rules change. Chapter 1 compares flow with stock. Chapter 2 separates the push burst from the search floor and shows that judging from the first month puts the three-year total at 0.512 of the truth. Chapter 3 treats long-tail question terms with the Zipf distribution and topic scoring; ten videos on obscure questions beat two on big terms with probability 0.722. Chapter 4 writes retention as a Weibull survival curve; an honest-preview opening ends with 1.125 times as many high-intent viewers as a broad-promise opening. Chapter 5 describes deep trust as Beta updating; with high-quality evidence a twelve-minute video leaves about 13.2 viewers per thousand starters ready to buy. Chapter 6 writes the description box as a chain of multiplications and computes coupled elasticities, sample sizes and the uncertainty of three months of data, 0.967 of which comes from the sales stage. Chapter 7 reads the Xunzi, the Analects, Ecclesiastes and Aristotle's Rhetoric; a long-video channel that stops still receives 0.377 of its three-year views after the stop, and the chapter ends with the distillation spec for the four kits. Chapter 8 gives the monthly ledger of long-video assets, the distilled Long-Video Asset Half-Life and Description-Box Conversion Calculator, red-yellow-green risk tables and pre-registered tests. The book teaches no buying of traffic, inflated engagement, misleading titles or padded metadata, and describes the study hall only by its reading list and course format. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:内容获客圈流传三句话:短视频流量大、长视频没人看;长视频完播率低、前 30 秒留不住人就没有价值;描述栏放链接没人点。本书逐条证伪或立界,并把「48 小时速朽」与「三年持续进客」两句口号改写成可检验的半衰期参数,而不当作事实或承诺。示意曲线下,短视频头 48 小时拿到三年观看的 0.750,带搜索地板的长视频只拿到 0.059,第 1,095 天仍有每天 5.00 次观看;半衰期翻十倍时,短内容峰值要高出 9.96 倍才打平现值。核对原文后,本书做了蓝图修正:卡等人 2007 年的数据同时显示单支视频衰减快与每日请求多来自上线超过一个月的老视频;按引流来源对应热度趋势的分析在菲格雷多等人 2014 年的扩展版;《长尾》是通俗读物;睡眠者效应的重复并不稳定。书中参数为示意,全部计算由书中程序生成。本书是非个人化的经营与营销分析,不构成营销或投资建议;平台规则与监管指南以使用时官方页面为准。 第一章比较流量与存量。第二章拆开推荐脉冲与搜索地板,说明只看头一个月会把三年总量估成真值的 0.512。第三章用 Zipf 分布与选题打分处理长尾问题词,十支冷门问题视频胜过两支大词视频的概率是 0.722。第四章把留存写成 Weibull 生存曲线,如实预告式开场在结尾留下的高意图观众是大承诺式的 1.125 倍。第五章用 Beta 信念更新描述深度信任,高质量证据下 12 分钟视频每千位起播观众约 13.2 位准备购买。 第六章把描述栏写成串联乘积,算出耦合弹性、样本量与三个月数据的不确定性,其中 0.967 来自成交环节。第七章考辨《荀子·劝学》《论语·为政》《传道书》与亚里士多德《修辞学》,长视频频道停更后仍有 0.377 的三年观看到来,并给出四件套蒸馏规格。第八章给出长视频资产月度台账、蒸馏小工具「长视频资产半衰期与描述栏转化试算器」、红黄绿风控表与预注册检验。书中不教买量、刷量、误导标题与堆砌元数据,涉及学堂只写书目与课程形态。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Half-Lives: Rewriting "Rots in 48 Hours" and "Brings Clients for Three Years" as Parameters — Part A gives monthly views when the floor starts at 40 a day. In the first month the clips bring 49,766 views and the long videos 18,571, so the clips bring more than twice as many. By month 12 the clips bring 51,937 and the long videos 49,308, almost level. In month 24 the long videos bring 60,149 and in month 36 they bring 65,568, while the clips stay at 51,937 throughout. (https://5dlibrary.com/answers/en/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c01/) 第1章 半衰期:把「48 小时速朽」与「三年持续进客」写成参数 — A 段是地板为每天 40 次时的月观看。第一个月,短视频 49,766,长视频 18,571,前者是后者的两倍多。到第 12 个月,短视频是 51,937,长视频 49,308,几乎追平。第 24 个月,长视频 60,149,第 36 个月 65,568,短视频始终停在 51,937。 (https://5dlibrary.com/answers/zh/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c01/) - Ch.2 Two Sources: the Push Burst, the Search Floor and Early Half-Lives That Come Out Too Short — Chapter 2 separates the push burst from the search floor and shows that judging from the first month puts the three-year total at 0.512 of the truth. (https://5dlibrary.com/answers/en/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c02/) 第2章 两种来源:推荐脉冲、搜索地板与偏短的早期半衰期 — 第二章拆开推荐脉冲与搜索地板,说明只看头一个月会把三年总量估成真值的 0.512。 (https://5dlibrary.com/answers/zh/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c02/) - Ch.3 Long-Tail Question Terms: Zipf Demand, Topic Scoring and the Portfolio Risk of "Small per Video, Large in Total" — Chapter 3 treats long-tail question terms with the Zipf distribution and topic scoring; ten videos on obscure questions beat two on big terms with probability 0.722. (https://5dlibrary.com/answers/en/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c03/) 第3章 长尾问题词:Zipf 需求、选题打分与「单支小、合计大」的组合风险 — 第三章用 Zipf 分布与选题打分处理长尾问题词,十支冷门问题视频胜过两支大词视频的概率是 0.722。 (https://5dlibrary.com/answers/zh/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c03/) - Ch.4 Retention Is a Survival Curve: Weibull Negative Ageing, the Filtering Opening and Stayers as the Denominator — Chapter 4 writes retention as a Weibull survival curve; an honest-preview opening ends with 1.125 times as many high-intent viewers as a broad-promise opening. (https://5dlibrary.com/answers/en/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c04/) 第4章 留存是生存曲线:Weibull 负老化、开场筛选与留存者分母 — 第四章把留存写成 Weibull 生存曲线,如实预告式开场在结尾留下的高意图观众是大承诺式的 1.125 倍。 (https://5dlibrary.com/answers/zh/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c04/) - Ch.5 Deep Trust: Beta Updating, Checkable Evidence Segments and the Denominator of a Pre-Sale — Chapter 5 describes deep trust as Beta updating; with high-quality evidence a twelve-minute video leaves about 13.2 viewers per thousand starters ready to buy. (https://5dlibrary.com/answers/en/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c05/) 第5章 深度信任:Beta 信念更新、可核验证据片段与预售的分母 — 第五章用 Beta 信念更新描述深度信任,高质量证据下 12 分钟视频每千位起播观众约 13.2 位准备购买。 (https://5dlibrary.com/answers/zh/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c05/) - Ch.6 The Description-Box Tollbooth: a Chain of Multiplications, Coupled Elasticities, Sample Sizes and Disclosure — Chapter 6 writes the description box as a chain of multiplications and computes coupled elasticities, sample sizes and the uncertainty of three months of data, 0.967 of which comes from the sales stage. (https://5dlibrary.com/answers/en/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c06/) 第6章 描述栏收费站:串联乘积、耦合弹性、样本量与披露 — 第六章把描述栏写成串联乘积,算出耦合弹性、样本量与三个月数据的不确定性,其中 0.967 来自成交环节。 (https://5dlibrary.com/answers/zh/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c06/) - Ch.7 Classical Sources: the Xunzi, the Analects, Ecclesiastes and Aristotle's Rhetoric — Chapter 7 reads the Xunzi, the Analects, Ecclesiastes and Aristotle's Rhetoric; a long-video channel that stops still receives 0.377 of its three-year views after the stop, and the chapter ends with the distillation spec for the four kits. (https://5dlibrary.com/answers/en/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c07/) 第7章 原典考辨:《荀子·劝学》《论语·为政》《传道书》与亚里士多德《修辞学》 — 第七章考辨《荀子·劝学》《论语·为政》《传道书》与亚里士多德《修辞学》,长视频频道停更后仍有 0.377 的三年观看到来,并给出四件套蒸馏规格。 (https://5dlibrary.com/answers/zh/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c07/) - Ch.8 Closing the Loop: the Long-Video Asset Ledger, Calculators and Risk-Control Tables — Chapter 8 gives the monthly ledger of long-video assets, the distilled Long-Video Asset Half-Life and Description-Box Conversion Calculator, red-yellow-green risk tables and pre-registered tests. The book teaches no buying of traffic, inflated engagement, misleading titles or padded metadata, and describes the study hall only by its reading list and course format. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c08/) 第8章 实证闭环:长视频资产台账、试算工具与风控表 — 第八章给出长视频资产月度台账、蒸馏小工具「长视频资产半衰期与描述栏转化试算器」、红黄绿风控表与预注册检验。书中不教买量、刷量、误导标题与堆砌元数据,涉及学堂只写书目与课程形态。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/long-form-video-search-half-life-deep-trust-preselling-passive-acquisition/c08/) ## 5DL-B-371 · Joint-Venture Audience Monetisation, Tech-for-Traffic Barter Structures and Co-Launch Revenue Waterfalls 中文书名:异业私域联营拆账、技术换流量零本金借船与联合发售脉冲母典 DOI: pending Cite: Fan, Y. (2026). Joint-Venture Audience Monetisation, Tech-for-Traffic Barter Structures and Co-Launch Revenue Waterfalls (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative, all calculations come from the book's programs, and nothing describes a real partner. This is not legal, tax or investment advice; contract clauses are templates, and data-protection provisions change. Abstract: Joint ventures between audience owners and operators run on three popular beliefs: an even split is fairest; a channel's groups together will reach tens of thousands of qualified buyers in three days; and more rounds of messages sell more. This book refutes each belief or draws its boundary. The Nash bargaining solution first returns each side's disagreement point and then halves the gain, so an even split is the special case of symmetric fallbacks; with a channel that earns 2,000 alone and an operator that earns 6,000, the channel's share is 0.433. Three-party ventures are divided with exact Shapley values and set against the overseas commission bands: single level only, the direct referrer only, and a rate fixed by the product's price band. The models decide whether to sign, which product band to sell and the interval for a non-cash consideration; they never change the rate. Reach is a chained product of audience size, delivery, opening and clicking: a single community needs 31,566 members for twenty thousand to open at least once at mid-range rates, and three channels together reach twenty thousand with probability 0.0005. On frequency, eight pushes bring less than twice the orders of three but 7.453 times the lost members, and a one-off operator's preferred schedule costs 0.1468 of twelve-month joint value. Blueprint correction: Ansari and Mela (2003) studied content customisation, so the frequency evidence rests on Kumar and co-authors and on Zhang, Kumar and Cosguner; whether too few or too many messages costs more varies by list. Chapters 1 and 2 cover disagreement points, three bargaining solutions, exact Shapley values, the bands and the core. Chapter 3 prices tech for traffic: zero principal is not zero cost, and the consideration must be one fixed amount agreed once. Chapter 4 rewrites the reach slogan as bounds and lets a channel audit its size by sampling without handing over its list. Chapter 5 derives a 72-hour push cap from a fatigue model. Chapter 6 schedules a 72-hour co-launch with real stock and a real deadline, a sell-out probability of 0.3195 and a refund reserve of 1,656 at the 99% quantile. Chapter 7 reads Xunzi, Sunzi, Ecclesiastes and Aristotle. Chapter 8 gives the revenue waterfall, the Joint-Venture Share and Frequency-Protection Contract Checker, red, yellow and green risk-control tables and a pre-registered test. Parameters are illustrative, all calculations come from the book's programs, and nothing describes a real partner. This is not legal, tax or investment advice; contract clauses are templates, and data-protection provisions change. All launch copy is a compliant sample with no medical claims, no pressure, no promised returns and no resetting countdown. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:异业联营有三条流行说法:对半分最公平;渠道方的群加起来三天就能触达数万精准买家;多发几轮就多卖几单。本书逐条证伪或立界。纳什议价解先还双方的破裂点,再平分合作红利,五五分只是破裂点对称时的特例;示意参数下,渠道方单干 2,000、操盘方单干 6,000 时,渠道方份额是 0.433。三方联营按夏普利值精确计算,并与海外轨分润档位比对:只许单层、只给直接推荐人,费率按商品客单价档位定值,模型只用来判断签不签、选哪一档商品、技术换流量的非现金对价落在哪个区间,绝不改费率。触达是私域规模、送达、打开与点击的串联乘积,单个社群要有 31,566 人,才能在区间中点下让两万人至少打开一次;三类渠道合起来,达到两万的概率也只有 0.0005。频次方面,示意名单推 8 次比推 3 次成交多不到一倍,流失成员却是 7.453 倍;一次性操盘方的最优推送会让十二个月联合价值损失 0.1468。核对原文之后,本书做了蓝图修正:安萨里与梅拉 2003 年的论文研究的是内容个性化,频次证据改由库马尔等人与张、库马尔、科斯古纳的研究承担,后者发现最优频次居中,而少发与多发哪边更亏随名单而变。 第一章讲破裂点、纳什解、卡莱与斯莫罗丁斯基解和鲁宾斯坦的耐心。第二章算三方夏普利值、逐商品过档位并检验核心。第三章给技术换流量作价:零本金不等于零成本,对价必须一次议定、固定一笔。第四章把「三天触达数万」改写为上下界,并让渠道方在不交出名单的前提下抽样核对规模。第五章建立疲劳模型,求 72 小时推送上限,写出退订即停、名单不外流与冷却期。第六章排一场 72 小时联合快闪:真实库存与真实截止下售罄概率 0.3195,退款准备金按 99% 分位计提为 1,656。第七章考辨《荀子·劝学》借舟、《孙子·九地》同舟、《传道书》二人胜一与亚里士多德的比例分配,并给出四件套蒸馏规格。第八章给出收入瀑布、联营分润与频次保护契约检查器、红黄绿风控表与预注册检验。 书中参数为示意,全部计算由书中程序生成,不对应任何真实合作方、达人或客户。本书非法律、税务或投资建议,契约条款只作模板参考,个人资料保护法条文会变动。联合发售文案一律为合规样本:不写医疗功效、威胁催促与收益担保,不用会重置的倒计时。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Disagreement Points Set the Split: Fifty-Fifty Is the Symmetric Special Case — At a cap of 1.00 there is no constraint and both solutions give the channel 14,000, a share of 0.350. When the cap falls to 0.50 and then 0.40, the Nash solution stays at 14,000, because the cap does not touch it. The Kalai and Smorodinsky solution, by contrast, falls to 12,286 and 10,842, shares of 0.307 and 0.271, and the operator's part rises to 19,714 and 21,158. At a cap of 0.30 the Nash solution binds for the first time and drops to 12,000, while the Kalai and Smorodinsky solution falls to 9,059. At a cap of 0.15 the two give 6,000 and 5,429. (https://5dlibrary.com/answers/en/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c01/) 第1章 破裂点决定拆账:五五分只是对称特例 — 上限为 1.00 时没有任何约束,两种解都给渠道方 14,000,份额 0.350。上限降到 0.50 与 0.40,纳什解仍是 14,000,因为上限没有卡住它;卡莱与斯莫罗丁斯基解却降到 12,286 与 10,842,份额 0.307 与 0.271,操盘方相应拿到 19,714 与 21,158。上限降到 0.30,纳什解第一次被卡住,只剩 12,000;卡莱与斯莫罗丁斯基解降到 9,059。上限 0.15 时两者分别是 6,000 与 5,429。 (https://5dlibrary.com/answers/zh/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c01/) - Ch.2 Exact Three-Party Contributions and the Commission Bands: the Model Chooses the Product, Never the Rate — A three-party joint venture can be divided by the Shapley value, which in this chapter's illustrative figures gives the channel 0.3380. The band rate is fixed; the Shapley value is used only to choose the product and to decide whether to sign. The entry band is generous to the channel, the deep band pays a little less than contribution, an add-on cannot carry a co-launch, and an unclassified product must be classified first. The core test warns that when the channel's outside options improve, the venture loses stability first. (https://5dlibrary.com/answers/en/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c02/) 第2章 三方贡献的精确分数与分润档位:模型只挑商品,不改费率 — 三方联营按夏普利值分,渠道方在本章示意数字下占 0.3380。档位费率是定值,夏普利值只用来挑商品与判断签不签:引流档对渠道方慷慨,深度档略低于贡献,加购档撑不起一场联营,未归档商品先归档。核心检查提醒,渠道方外部选项变好时,联营先失去稳定。 (https://5dlibrary.com/answers/zh/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c02/) - Ch.3 Tech for Traffic: Zero Principal Is Not Zero Cost — A tech-for-traffic price has two boundaries: the operator's opportunity cost below and the channel's replacement cost above. With illustrative parameters the package is worth between 2,874 and 4,254, and exchanging it for 2 pushes suits both sides best. Zero principal is not zero cost: the operator's opportunity cost must be recorded, and so must the channel's reputation risk. The package must be a single fixed consideration agreed once, never charged per order and never paid for with a list. (https://5dlibrary.com/answers/en/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c03/) 第3章 技术换流量:零本金不等于零成本 — 技术换流量的作价有两条边界:操盘方的机会成本是下界,渠道方的替代成本是上界。示意参数下技术包值 2,874 到 4,254,换 2 次推送对双方最有利。零本金不等于零成本,操盘方的机会成本要记账,渠道方的信用风险也要记账。技术包必须是一次议定的固定对价,不按单计,不换名单。 (https://5dlibrary.com/answers/zh/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c03/) - Ch.4 Reach Is a Chained Product: Bounds on "Tens of Thousands in Three Days" — In the simulation, a sample of 100 finds 48 active members, a share of 0.480 with a Wilson interval of 0.385 to 0.577, so the effective size lies between 4,616 and 6,922, a wide range. A sample of 385 finds 154 active, a share of 0.400 with an interval of 0.352 to 0.450 and an effective size of 4,227 to 5,396. A sample of 1,068 finds 443 active, a share of 0.415 with an interval of 0.386 to 0.445 and an effective size of 4,627 to 5,335. (https://5dlibrary.com/answers/en/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c04/) 第4章 触达是串联乘积:「三天触达数万」的上下界 — 模拟里抽 100 人,活跃 48 人,比例 0.480,威尔逊区间 0.385 到 0.577,有效规模在 4,616 到 6,922 之间,区间很宽。抽 385 人,活跃 154 人,比例 0.400,区间 0.352 到 0.450,有效规模 4,227 到 5,396。抽 1,068 人,活跃 443 人,比例 0.415,区间 0.386 到 0.445,有效规模 4,627 到 5,335。 (https://5dlibrary.com/answers/zh/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c04/) - Ch.5 Frequency Protection: the Fatigue Curve and a Push Cap for 72 Hours — Chapter 5 derives a 72-hour push cap from a fatigue model. (https://5dlibrary.com/answers/en/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c05/) 第5章 频次保护:疲劳曲线与 72 小时推送上限 — 第五章建立疲劳模型,求 72 小时推送上限,写出退订即停、名单不外流与冷却期。 (https://5dlibrary.com/answers/zh/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c05/) - Ch.6 A 72-Hour Co-Launch: Real Stock, a Real Deadline and a Refund Reserve — Chapter 6 schedules a 72-hour co-launch with real stock and a real deadline, a sell-out probability of 0.3195 and a refund reserve of 1,656 at the 99% quantile. (https://5dlibrary.com/answers/en/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c06/) 第6章 72 小时联合快闪:真实库存、真实截止与退款准备金 — 第六章排一场 72 小时联合快闪:真实库存与真实截止下售罄概率 0.3195,退款准备金按 99% 分位计提为 1,656。 (https://5dlibrary.com/answers/zh/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c06/) - Ch.7 Reading the Classics: Xunzi's Borrowed Boat, Sunzi's Shared Boat, the Two of Ecclesiastes and Aristotle's Proportion — Xunzi was right that a borrowed boat crosses rivers, and silent about returning it. Sunzi was right that people in one boat help each other, and silent about what happens ashore. Ecclesiastes was right that two are better than one, and silent about by how much. Aristotle was right that justice is proportion, and silent about who sets the denominator. Programs p71 and p72 turned two of these sentences into parameters: the years of waiting saved by borrowing a boat, and the number of repeat ventures needed for keeping one's word to pay. (https://5dlibrary.com/answers/en/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c07/) 第7章 原典考辨:荀子借舟、孙子同舟、《传道书》二人与亚里士多德的比例 — 荀子说对了借舟能渡河,没说船要还;孙子说对了同舟会相救,没说下船之后怎样;《传道书》说对了二人胜过一人,没说胜多少;亚里士多德说对了公正即比例,没说比例的分母由谁定。p71 与 p72 把其中两句写成了参数:借船省下的等待年数,以及守约所需的重复合作次数。 (https://5dlibrary.com/answers/zh/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c07/) - Ch.8 Closing the Loop: the Partnership Ledger, Calculators and Risk-Control Tables — The revenue waterfall is checked with fractions and its parts must add up exactly. The contract checker marks each line red, yellow or green and never outputs a new rate. The risk-control tables give the ledger fields, the criteria, the actions and a daily, weekly and monthly rhythm. The pre-registered test needs about 2,193 members per group, scaled up for clustering by channel, and 25 flash sales. The three results that would overturn the book are written down in advance. (https://5dlibrary.com/answers/en/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c08/) 第8章 实证闭环:联营台账、试算工具与风控表 — 收入瀑布用分数核对,分项必须加总相等;契约检查器逐条判红黄绿,永远不输出新费率;风控表给出台账字段、判据、动作与日周月节奏;预注册检验需要每组约 2,193 名成员(按渠道聚类放大)与 25 场快闪。三条推翻条件事先写明。 (https://5dlibrary.com/answers/zh/joint-venture-audience-revenue-split-tech-for-traffic-barter-co-launch/c08/) ## 5DL-B-372 · Bootstrap Cashflow Discipline, Seventy-Two-Hour Kill-Switch Dashboards and Reinvestment Compounding to Five Thousand Daily 中文书名:翻身战现金流铁血纪律、七十二小时止血红绿灯与日进五千复利重投母典 DOI: pending Cite: Fan, Y. (2026). Bootstrap Cashflow Discipline, Seventy-Two-Hour Kill-Switch Dashboards and Reinvestment Compounding to Five Thousand Daily (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: The north star becomes daily net cash banked (cash received minus real outlays minus the refund reserve); in a fixed-seed synthetic ledger, revenue grows by a factor of 4.05 while net cash banked keeps sinking. Parameters are illustrative; all figures come from the book's programs. Abstract: Operators starting from nothing hear three popular sayings: rising revenue means profit; a new test needs either patience or a blunt seventy-two-hour deadline; and reinvesting everything is the fastest way to five thousand a day. This book refutes or bounds each one. The north star becomes daily net cash banked (cash received minus real outlays minus the refund reserve); in a fixed-seed synthetic ledger, revenue grows by a factor of 4.05 while net cash banked keeps sinking. The seventy-two hours become Wald's sequential boundaries plus a hard cap plus a preset stop-loss amount; without a correction for conversion lag, the sequential test gives a red light to 0.967 of good tests, and with an exposure correction the share of green lights returns to 0.890. Reinvestment becomes fill first, split later: six months of essential living outlays go into a personal account kept physically apart from the business account, and only then is profit reinvested at a fractional-Kelly share bounded by a ruin constraint. Blueprint corrections: the original study behind "82% of small businesses fail because of cash flow" could not be found, so the figure is treated as a popular saying, and Staw's escalation experiment, with mixed replications, is only supporting material. Parameters are illustrative; all figures come from the book's programs. This is general business analysis, not personalised investment, financial, tax or legal advice, and the author is not a licensed adviser; "five thousand a day" is only a target value or a simulated quantile, never a promised return. Chapters 1 and 2 decompose where revenue goes and compute the cash locked up by settlement lags and purchase-agent advances. Chapter 3 draws Wald's decision lines and prices hourly peeking. Chapter 4 builds the three-gate lights with a Beta-binomial payback probability and a lag correction, and compares four testing policies. Chapter 5 sizes the cushion: with large swings, three months leaves 0.4181 of paths forced to stop, six months 0.0290. Chapter 6 chooses the reinvestment share under a ruin constraint: half Kelly keeps 0.753 of the growth rate of full Kelly, and with a cushion and half-Kelly reinvestment the simulated median time for daily net cash to first reach the target value of five thousand is 76 months with a closure share of 0.009, against 0.111 at full Kelly. The Kelly share applies only to reinvestment in one's own business, never to stocks, crypto-assets, foreign exchange, betting or borrowing. Chapter 7 examines the Book of Rites, the Great Learning, Joseph's fifth in Genesis and Micawber's sixpence. Chapter 8 gives the ledger fields, the light criteria and actions, the distilled 72-Hour Kill-Switch Light and Reinvestment Share Calculator and a pre-registered plan. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:白手起步的操盘者身边流传三句话:营收在涨、粉丝在涨就是在赚钱;新测试要么坚持给足时间,要么七十二小时不回本立刻关;利润全部再投最快滚到日进五千。本书逐条证伪或立界。北极星指标换成每日落袋净现金,即当日到账减当日真实付出再减退款准备计提;固定种子合成账本里,营收涨了 4.05 倍,落袋净现金却一路走低。七十二小时改写为沃尔德序贯边界加硬上限加预设止损额度;不修正转化滞后时,序贯检验会把好测试的 0.967 判成红灯,按暴露修正后绿灯比例回到 0.890。重投改写为先填垫、后分成:六个月生活必需开支放进与公司账户物理隔离的个人账户,填满之后按带破产约束的分数凯利比例重投。核对原文后本书做了蓝图修正:「82% 的小企业死于现金流」查不到原始研究,降级为流行说法;斯陶的升级承诺实验重复结果不一,只作旁证。书中参数为示意,全部计算由书中程序生成。本书是非个人化的经营分析,不构成投资、理财、税务或法律建议,作者不是持牌顾问;「日进五千」在书中只是目标值或模拟分位数,不是收益承诺。 第一章定义每日落袋净现金并拆解营收去向。第二章计算结算延迟锁住的在途资金与代购垫款的现金转换周期。第三章用沃尔德序贯概率比检验给出两条判决线与平均停时,并量化每小时偷看的误报代价。第四章组装三道闸的止血红绿灯,加入 Beta–二项回本概率与转化滞后修正,并在同一批合成测试上比较四种策略。第五章计算安全垫的填满时间:爬坡期里三个月垫子在波动较大时有 0.4181 的路径被迫停业,六个月降到 0.0290。 第六章在破产约束下求重投比例:半凯利保留全凯利增长率的 0.753;有垫子、按半凯利重投时,每日净现金首次达到五千这一目标值的模拟中位数是 76 个月,被迫停业比例 0.009,全凯利则升到 0.111。凯利比例只用于自有业务重投,不涉及股票、加密货币、外汇、博彩或借贷加杠杆。第七章考辨《礼记·王制》的三年之蓄、《大学》的生众食寡、《创世记》约瑟的五分之一与狄更斯笔下米考伯的六便士。第八章给出每日现金总控台的台账字段表、红黄绿判据与触发动作、蒸馏小工具「七十二小时止血红绿灯与重投比例试算器」与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 A New North Star: Daily Net Cash Banked versus Vanity Metrics — Revenue, impressions and followers describe heat; daily net cash banked describes survival. In the synthetic ledger revenue rose by a factor of 4.05 and followers by 46.41, while cumulative net cash banked over twelve weeks was -64,018 and the correlation between weekly revenue and weekly net cash was -0.348. The expected-value decomposition shows that unit costs, ads and money in transit together absorb more than the whole of revenue. Once ad spend passes the marginal break-even line of 1,257 a day, every extra unit raises the heat and lowers the profit. (https://5dlibrary.com/answers/en/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c01/) 第1章 北极星换人:每日落袋净现金与虚荣指标 — 营收、曝光与粉丝描述热度,每日落袋净现金描述生存。合成账本里,营收涨了 4.05 倍,粉丝涨了 46.41 倍,十二周累计落袋净现金却是 -64,018,周营收与周净现金的相关系数是 -0.348。期望值拆解显示,单位成本、广告与在途资金合起来吞掉了全部营收还不够;广告一旦越过每天 1,257 这条边际盈亏线,每多花一分钱,热度都在涨,利润都在退。 (https://5dlibrary.com/answers/zh/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c01/) - Ch.2 Settlement Lags and the Cash Conversion Cycle: Why Booked Profit Eats Cash First — Settlement delays and advances both make booked profit appear before cash. At steady daily sales of 5,000, a mixed set of channels with a weighted settlement time of 5.9 days locks up 29,500; during growth, the slowest channel mix collects only 0.640 of booked revenue. With a 15-day cycle, a purchase-agent line ties up 5,100 of working capital for each order per day; at 0.01 growth a day the balance never returns to its starting level within 180 days, while the same growth under a 7-day cycle pays back by day 59. Growth itself costs money, and the cycle decides how much. (https://5dlibrary.com/answers/en/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c02/) 第2章 回款延迟与现金转换周期:账面利润为何先吃掉现金 — 结算延迟与垫款都让账面利润先于现金出现。日销 5,000 的稳态下,加权结算 5.9 天的混合渠道锁住 29,500;增长期里,最慢的渠道组合只收回预订营收的 0.640。代购线周期 15 天时,每日每单要压住 5,100 的营运资金;日增长 0.01 时,180 天内余额再也回不到起点,而把周期缩到 7 天,同样的增长在第 59 天就回本。增长本身要花钱,周期决定这笔钱有多少。 (https://5dlibrary.com/answers/zh/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c02/) - Ch.3 Sequential Boundaries in Seventy-Two Hours: Wald's Test and the Price of Peeking — Seventy-two hours are not a moment of decision but a hard cap. The real decision comes from Wald's boundaries. With the parameters of this chapter, clearly good and clearly bad ads are decided in about 16 hours on average, almost half the time a fixed-sample test needs. Adding the seventy-two-hour cap barely affects the power of the test; only tests near break-even run into the cap, and they are closed as planned. Conversely, peeking every hour with a fixed threshold declares an ad with no advantage a winner 0.325 of the time, and running a bad ad for the full fourteen days loses almost five times as much as closing it at seventy-two hours. (https://5dlibrary.com/answers/en/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c03/) 第3章 七十二小时的序贯边界:沃尔德检验与偷看的代价 — 七十二小时不是一个判决时刻,而是一道硬上限。真正的判决由沃尔德边界给出:本章参数下,明显的好广告与坏广告平均 16 小时左右就能判决,比固定样本检验省下近一半;加上七十二小时上限,检验力几乎不受影响,只有落在盈亏平衡附近的测试会撞上上限,按预设关停。反过来,每小时用固定门槛偷看一次,会把毫无优势的广告以 0.325 的比例误判为赢家;坚持跑满十四天,坏广告的期望亏损是七十二小时关停的近五倍。 (https://5dlibrary.com/answers/zh/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c03/) - Ch.4 The Kill-Switch Lights: Payback Probability, Conversion Lag and a Preset Stop-Loss — The kill-switch lights consist of three gates: the sequential boundary decides, the seventy-two-hour cap closes the grey zone, and the preset stop-loss amount catches whatever is left. The payback probability is a reading for people; the decision is still made by the boundary. Conversion lag makes an uncorrected test give red to 0.967 of good ads; corrected by exposure, the share of green lights returns to 0.890. On the same batch of synthetic tests, the lights and the blind deadline tie on the bottom line, while the money at risk during testing under the lights is only about a fifth of the other; chasing losses is the worst policy of the four. (https://5dlibrary.com/answers/en/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c04/) 第4章 止血红绿灯:回本概率、转化滞后与预设止损额度 — 止血红绿灯由三道闸组成:序贯边界判决,七十二小时上限收尾,预设止损额度兜底。回本概率是给人看的读数,判决仍由边界做。转化滞后会让不修正的检验把好广告的 0.967 判成红灯,按暴露修正后绿灯比例回到 0.890。在同一批合成测试上,红绿灯与一刀切的总账几乎打平,测试期间的风险资金却只有后者的五分之一左右;追亏是最糟的策略。 (https://5dlibrary.com/answers/zh/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c04/) - Ch.5 A Six-Month Cushion: Fill First, Split Later, and Keep the Accounts Apart — Chapter 5 sizes the cushion: with large swings, three months leaves 0.4181 of paths forced to stop, six months 0.0290. (https://5dlibrary.com/answers/en/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c05/) 第5章 六个月安全垫:先填垫、后分成与账户物理隔离 — 第五章计算安全垫的填满时间:爬坡期里三个月垫子在波动较大时有 0.4181 的路径被迫停业,六个月降到 0.0290。 (https://5dlibrary.com/answers/zh/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c05/) - Ch.6 Reinvestment under a Ruin Constraint: Fractional Kelly and the Time Distribution to Five Thousand a Day — Chapter 6 chooses the reinvestment share under a ruin constraint: half Kelly keeps 0.753 of the growth rate of full Kelly, and with a cushion and half-Kelly reinvestment the simulated median time for daily net cash to first reach the target value of five thousand is 76 months with a closure share of 0.009, against 0.111 at full Kelly. The Kelly share applies only to reinvestment in one's own business, never to stocks, crypto-assets, foreign exchange, betting or borrowing. (https://5dlibrary.com/answers/en/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c06/) 第6章 带破产约束的重投比例:分数凯利与日进五千的时间分布 — 第六章在破产约束下求重投比例:半凯利保留全凯利增长率的 0.753;有垫子、按半凯利重投时,每日净现金首次达到五千这一目标值的模拟中位数是 76 个月,被迫停业比例 0.009,全凯利则升到 0.111。凯利比例只用于自有业务重投,不涉及股票、加密货币、外汇、博彩或借贷加杠杆。 (https://5dlibrary.com/answers/zh/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c06/) - Ch.7 Classical Sources: Three Years of Stores, Many Producers and Few Consumers, Joseph's Fifth and Micawber's Sixpence — If a year of plenty yields 1.2 times normal needs, seven years store 1.68 years of needs, and the store suffices as long as a lean year yields at least 0.76. If plenty means 1.5 times normal, 2.10 years are stored and lean years need to yield only 0.70. If plenty means 2.0 times normal, 2.80 years are stored and lean years need to yield only 0.60. If the lean years produced nothing at all, covering them would require storing 0.833, 0.667 or 0.500 of each good year's harvest, far more than a fifth. (https://5dlibrary.com/answers/en/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c07/) 第7章 原典考辨:三年之蓄、生众食寡、约瑟的五分之一与米考伯的六便士 — 丰年收成是正常的 1.2 倍时,七年存下 1.68 年的所需,荒年收成只要不低于 0.76 就够;丰年是 1.5 倍时,存下 2.10 年,荒年收成不低于 0.70 就够;丰年是 2.0 倍时,存下 2.80 年,荒年收成不低于 0.60 就够。若荒年颗粒无收,要覆盖它,丰年就得存下 0.833、0.667 或 0.500,远多于五分之一。 (https://5dlibrary.com/answers/zh/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c07/) - Ch.8 Closing the Loop: the Daily Cash Control Desk, Calculators and Risk-Control Tables — Chapter 8 gives the ledger fields, the light criteria and actions, the distilled 72-Hour Kill-Switch Light and Reinvestment Share Calculator and a pre-registered plan. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c08/) 第8章 实证闭环:每日现金总控台、试算工具与风控表 — 第八章给出每日现金总控台的台账字段表、红黄绿判据与触发动作、蒸馏小工具「七十二小时止血红绿灯与重投比例试算器」与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/bootstrap-cashflow-kill-switch-dashboard-reinvestment-compounding-runway/c08/) ## 5DL-B-373 · Gendered Consumption Decision Heterogeneity, the Selectivity Hypothesis and High-Conversion Architecture for Female Markets 中文书名:性别消费决策异质性、综合信息加工假说与女性高客单转化工程母典 DOI: pending Cite: Fan, Y. (2026). Gendered Consumption Decision Heterogeneity, the Selectivity Hypothesis and High-Conversion Architecture for Female Markets (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. Chapter 4 turns regret theory into a purchase probability: under illustrative parameters four regret cushions lift it from 0.451 to 0.808, a bare countdown leaves buyers with regret of 0.556, and the refund break-even of a return window is 0.289. Abstract: Three popular sentences circulate in high-ticket selling: women are emotional and men rational, so sell to women with feeling; women form a single market; and expensive sales run on scarcity and pressure. This book refutes or bounds each. Meyers-Levy's selectivity hypothesis concerns breadth of processing, not feeling against reason: heuristic processors read a few salient cues, comprehensive processors integrate them all, and the difference vanishes when cues agree or are very faint or very prominent. Hyde's gender similarities hypothesis and later metasynthesis show that most psychological gender differences are small and that variation within groups far exceeds the gap between them; at d = 0.30 the best accuracy of guessing processing style from gender is 0.560, and routing by gender matches a layered page only at d = 1.79. After checking the sources, the preface lists six corrections to the blueprint. Parameters are illustrative and all calculations are generated by the book's programs. Chapter 1 writes the selectivity hypothesis as cue-integration and threshold models. Chapter 2 computes the misclassification bound of routing by gender with the overlapping coefficient and addresses transgender and non-binary customers. Chapter 3 infers style from four on-site behaviour counts only, reaching a simulated accuracy of 0.864, with almost nothing gained from a gender prior. Chapter 4 turns regret theory into a purchase probability: under illustrative parameters four regret cushions lift it from 0.451 to 0.808, a bare countdown leaves buyers with regret of 0.556, and the refund break-even of a return window is 0.289. Chapter 5 shows that 7 concrete usage scenes let a comprehensive reader decide where generic slogans would need 88, and that an exposed undisclosed paid unboxing pushes the half-width back from 0.591 to 1.349. Chapter 6 sizes one-to-one seats (5 under the illustrative load), finds a repurchase tie line of 0.245 and compares the yearly net cash of two funnels by Monte Carlo. Chapter 7 reads the Analects, the Doctrine of the Mean, Wollstonecraft and Aristotle's three modes of persuasion. Chapter 8 gives the weekly ledger, field tools and checker specifications, the distilled Processing-Style and Regret-Cushion Calculator, and pre-registered tests (3,213 enquiries per arm for the conversion test) with refutation conditions. The book never recommends pricing by gender, collects no sensitive data, sells to no one at a fragile moment and makes no claim of any effect on the body; it is not psychological, medical, legal or personalised business advice. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. 摘要:高客单生意里流传三句流行说法:女人感性、男人理性,所以对女性要煽情;女性客群是同一群人;高客单靠稀缺与催促。本书逐句证伪或立界。梅耶斯-列维的选择性加工假说讲的不是感性对理性,而是加工广度:启发式加工者看少数显著线索,综合加工者把全部线索都算进去,线索彼此一致或过弱过强时差异消失。海德的性别相似性假说与后续元合成显示,多数心理变量的性别差异很小,组内差异远大于组间差异;d 为 0.30 时,按性别猜加工风格的最好命中率只有 0.560,按性别分流要追平一张分层页面,d 须达到 1.79。核对原文后,本书在序里写明六处蓝图修正。书中参数为示意,全部计算由书中程序生成。 第一章把选择性假说写成线索积分与门槛模型。第二章用重叠系数计算按性别分流的误分上界,并写明跨性别与非二元客户。第三章只用站内四类行为计数做贝叶斯判别,模拟准确率 0.864,加回性别先验几乎不再提升。第四章把后悔理论写成购买概率:示意参数下四项后悔安全垫把购买概率从 0.451 推到 0.808,单挂倒计时的买家后悔升到 0.556;退换窗口的退款打平线是 0.289。第五章显示具体情境叙事 7 条即可让综合加工者想清楚,空泛口号要 88 条;未披露的付费开箱被揭穿后,区间半宽从 0.591 弹回 1.349。 第六章计算私密承接席位(示意负载下 5 个)与复购追平线 0.245,并以蒙特卡洛比较两条漏斗一年的净现金。第七章考辨《论语》「性相近,习相远」、《中庸》「致广大而尽精微」、沃斯通克拉夫特与亚里士多德的三种说服。第八章给出周度台账、实战工具与检查器规格、蒸馏小工具「加工风格判别与后悔安全垫试算器」,以及预注册检验(例如转化检验每组 3,213 次询价)与推翻条件。本书不写按性别差别定价,不采集敏感资料,不在客户脆弱时成交,不作任何身体效果宣称;非心理咨询、医疗、法律或个人化经营建议。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Selectivity Hypothesis: a Difference in Breadth of Processing, Not Feeling against Reason — Block A shows that the gap is largest in the middle. At a prominence of 0.5, a share of 0.500 of group A elaborates, against 0.382 of group B, a gap of 0.118, the largest on the grid. Towards either end the gap closes quickly. At a prominence of 0.0 the shares are just 0.023 and 0.011, a gap of 0.012; at 1.0 they are 0.977 and 0.955, a gap of 0.022. A cue that is too faint is read by nobody; a cue that is loud enough is read by everybody. (https://5dlibrary.com/answers/en/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c01/) 第1章 选择性加工假说:差别在加工广度,不在感性与理性 — A 段显示,差距在中段最大。显著性为 0.5 时,A 组有 0.500 的人细读,B 组有 0.382,差距 0.118,是表上的最大值。往两端走,差距迅速收窄:显著性为 0.0 时,两组分别只有 0.023 与 0.011,差距 0.012;显著性为 1.0 时,两组分别是 0.977 与 0.955,差距 0.022。线索太弱,谁都不看;线索太强,谁都会看。 (https://5dlibrary.com/answers/zh/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c01/) - Ch.2 Gender Similarities: the Overlapping Coefficient and the Misclassification Bound of Routing by Gender — The magnitudes in the literature are clear: on most psychological variables the difference between men and women lies in the close-to-zero or small range, and variation within groups is much larger than the gap between them. At d = 0.30, routing by gender beats a coin toss by six percentage points, and within-group variance is 44.4 times the between-group variance; for rule G to draw level with a layered page, d would have to reach 1.79. The lopsided composition of the extreme tail is a geometric effect of normal tails and cannot justify labelling the majority. (https://5dlibrary.com/answers/en/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c02/) 第2章 性别相似性:重叠系数与按性别分流的误分上界 — 文献给出的量级很清楚:多数心理变量的性别差异落在接近零或小效应区间,组内差异远大于组间差异。d 为 0.30 时,按性别分流只比抛硬币高出六个百分点,组内方差是组间方差的 44.4 倍;规则 G 要追平一张分层页面,d 得大到 1.79。尖端组成的悬殊,是正态尾部的几何效应,不能拿来给多数人贴标签。 (https://5dlibrary.com/answers/zh/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c02/) - Ch.3 Reading Processing Style from Browsing Behaviour: Bayesian Updating Overwhelms the Gender Prior — Chapter 3 infers style from four on-site behaviour counts only, reaching a simulated accuracy of 0.864, with almost nothing gained from a gender prior. (https://5dlibrary.com/answers/en/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c03/) 第3章 按浏览行为判别加工风格:贝叶斯更新盖过性别先验 — 第三章只用站内四类行为计数做贝叶斯判别,模拟准确率 0.864,加回性别先验几乎不再提升。 (https://5dlibrary.com/answers/zh/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c03/) - Ch.4 Regret Theory and Regret Cushions: Returns, Trials, Transparent Instalments and After-Sales Promises — Chapter 4 turns regret theory into a purchase probability: under illustrative parameters four regret cushions lift it from 0.451 to 0.808, a bare countdown leaves buyers with regret of 0.556, and the refund break-even of a return window is 0.289. (https://5dlibrary.com/answers/en/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c04/) 第4章 后悔理论与后悔安全垫:退换、试用、分期透明与售后承诺 — 第四章把后悔理论写成购买概率:示意参数下四项后悔安全垫把购买概率从 0.451 推到 0.808,单挂倒计时的买家后悔升到 0.556;退换窗口的退款打平线是 0.289。 (https://5dlibrary.com/answers/zh/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c04/) - Ch.5 Usage Scenes and Peer Unboxings: How Cues Converge, and the Duty to Disclose — Chapter 5 shows that 7 concrete usage scenes let a comprehensive reader decide where generic slogans would need 88, and that an exposed undisclosed paid unboxing pushes the half-width back from 0.591 to 1.349. (https://5dlibrary.com/answers/en/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c05/) 第5章 情境叙事与同侪开箱:线索收敛与披露义务 — 第五章显示具体情境叙事 7 条即可让综合加工者想清楚,空泛口号要 88 条;未披露的付费开箱被揭穿后,区间半宽从 0.591 弹回 1.349。 (https://5dlibrary.com/answers/zh/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c05/) - Ch.6 One-to-One Conversations, Repurchase, and the Net Cash of Two Funnels — Chapter 6 sizes one-to-one seats (5 under the illustrative load), finds a repurchase tie line of 0.245 and compares the yearly net cash of two funnels by Monte Carlo. (https://5dlibrary.com/answers/en/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c06/) 第6章 私密承接、复购与两条漏斗的净现金 — 第六章计算私密承接席位(示意负载下 5 个)与复购追平线 0.245,并以蒙特卡洛比较两条漏斗一年的净现金。 (https://5dlibrary.com/answers/zh/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c06/) - Ch.7 The Classical Sources: "Near by Nature, Far by Practice", "Exhaust the Fine Detail", Wollstonecraft, and Aristotle's Three Modes of Persuasion — Each of the four old sentences gets part of the truth. "Near by nature, far by practice" puts differences down to practice, and p71 shows further that acquired differences land first on individuals: even at the strongest socialisation, two people from the same group are still 3.6 group gaps apart. "Reach the broad and exhaust the fine" describes comprehensive processing but cannot be treated as a fixed attribute of a kind of person. Wollstonecraft rewrote "women are emotional" as a question of education and put the customer's autonomy first. Aristotle's three modes of persuasion become, in p72, three channels of cues, and copy that relies on feeling alone scores lowest with comprehensive readers. (https://5dlibrary.com/answers/en/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c07/) 第7章 原典考辨:「性相近,习相远」、「尽精微」、沃斯通克拉夫特与亚里士多德的三种说服 — 四句旧话,各自说对了一部分。「性相近,习相远」把差异归于后天,p71 进一步说明后天差异首先落在个人身上:社会化最强时,同组两人仍相隔 3.6 个组间差。「致广大而尽精微」描述了综合加工,却不能当作某类人的固定属性。沃斯通克拉夫特把「女人感性」改写成教育问题,并把客户自主放到首位。亚里士多德的三种说服在 p72 里成了三条线索通道,只押情感的文案在综合加工者那里得分最低。 (https://5dlibrary.com/answers/zh/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c07/) - Ch.8 Closing the Loop: the Weekly High-Ticket Ledger, Field Tools and Checker Specifications — Chapter 8 gives the weekly ledger, field tools and checker specifications, the distilled Processing-Style and Regret-Cushion Calculator, and pre-registered tests (3,213 enquiries per arm for the conversion test) with refutation conditions. The book never recommends pricing by gender, collects no sensitive data, sells to no one at a fragile moment and makes no claim of any effect on the body; it is not psychological, medical, legal or personalised business advice. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c08/) 第8章 实证闭环:高客单周度台账、实战工具与检查器规格 — 第八章给出周度台账、实战工具与检查器规格、蒸馏小工具「加工风格判别与后悔安全垫试算器」,以及预注册检验(例如转化检验每组 3,213 次询价)与推翻条件。本书不写按性别差别定价,不采集敏感资料,不在客户脆弱时成交,不作任何身体效果宣称;非心理咨询、医疗、法律或个人化经营建议。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/gendered-consumption-selectivity-hypothesis-female-high-ticket-conversion/c08/) ## 5DL-B-374 · Co-Evolutionary Human-AI Writing Spirals, Falsification-Gated Ratchets and Symbiotic Capability Compounding 中文书名:人机互哺双螺旋进化、实证门禁稳健棘轮与共生创作复利母典 DOI: pending Cite: Fan, Y. (2026). Co-Evolutionary Human-AI Writing Spirals, Falsification-Gated Ratchets and Symbiotic Capability Compounding (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. Chapter 2 treats the jagged frontier: with illustrative parameters, clicking confirm outside the frontier is right 0.604 of the time against 0.800 for answering alone, while falsifying first and then comparing leads on both sides. Abstract: Three claims about writing with machines are widely repeated: a model that feeds on its own output can improve without limit; with a good model the human only has to click "confirm"; human intuition is unreliable, so everything should go to the data. This book refutes each claim or draws its boundary. Feeding a model its own output multiplies the expected variance by (n-1)/n per generation, and the tails vanish before the variance does; mixing in a real source gives the variance a steady state, and with 100 samples per generation the real share must be at least about 0.0909. After checking the sources the book corrects its blueprint: Gerstgrasser and colleagues show that accumulating rather than replacing real data keeps the error bounded, so collapse is conditional, not fate; the jagged-frontier study appeared first as a Harvard Business School working paper and later in Organization Science. Parameters are illustrative and all calculations are generated by the book's programs. The book is neither software-engineering nor investment advice and makes no claim about the internal workings of any specific commercial model. Chapter 1 computes variance collapse and the early loss of tails. Chapter 2 treats the jagged frontier: with illustrative parameters, clicking confirm outside the frontier is right 0.604 of the time against 0.800 for answering alone, while falsifying first and then comparing leads on both sides. Chapter 3 computes the gain from injected human outliers: a pool of 20 model candidates has an expected best of 0.934, two human hypotheses lift it to 1.186, and the gain holds only when the hypotheses are testable. Chapter 4 builds the falsification gate: a sequential test needs 98 to 118 cases on average where a fixed design needs 211 at the same error rates, and the leak rate is computed by candidate source. Chapter 5 treats the ratchet as a random walk with misclassification, adds rollback audits and states the condition for compounding: with illustrative parameters the feedback strength must exceed 0.0556. Chapter 6 computes multi-agent falsification: correlated same-source agents add votes without adding protection, and the independent prediction of 0.0579 for five votes holds only under independence. Chapter 7 reads the Record of Learning, the Exhortation to Learning in the Xunzi, Bacon's idols of the tribe and Popper's conjectures and refutations, and writes the distillation spec for the four kits. Chapter 8 gives the field tools and checker specifications: a daily ratchet ledger, red-yellow-green criteria, the distilled Ratchet Gate and Collapse Alarm Calculator and a pre-registered refutation plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:人机协作写作流传着三条说法:AI 自产自学就能无限进化;有了 AI,人只要点确认;人类直觉靠不住,全交给数据。本书逐条证伪或立界。模型拿自己的产出一代代回喂,期望方差每代乘以 (n-1)/n,尾部比方差先消失;每代掺进真源,方差就有稳态,每代 100 个样本时真源比例下界约为 0.0909。核对原文后本书做了蓝图修正:Gerstgrasser 等人显示,真实数据累积而不替换时误差有上界,所以坍缩是条件命题而不是宿命;锯齿边界一文先是哈佛商学院工作论文,后刊于 Organization Science。书中参数为示意,全部计算由书中程序生成。本书不是软件工程意见,也不是投资意见,不对任何具体商业模型的内部机制下断言。 第一章算方差坍缩与尾部先走。第二章算锯齿边界:示意参数下,边界外只点确认的正确率是 0.604,低于独立作答的 0.800,先证伪再比对在两侧都领先。第三章算人类异常值的注入增益:20 条候选全是模型产出时最优值期望 0.934,掺进两条人类假说升到 1.186,但增益只在假说可检验时成立。第四章写证伪门禁:序贯检验在同样错误率下平均只要 98 到 118 个案例,固定样本要 211 个,并按候选来源算出漏网率。 第五章把棘轮写成带误判的随机游走,加上回退审计,并给出复利条件:示意参数下反馈强度要大于 0.0556。第六章算多代理证伪:同源代理相关时,多投票不等于多保险,五票独立预测的误判率 0.0579 只在独立时成立。第七章考辨《礼记·学记》《荀子·劝学》、培根的族类假象与波普尔的猜想反驳,并写出四件套蒸馏规格。第八章给出实战工具与检查器规格:每日棘轮台账、红黄绿判据、蒸馏小工具「人机互哺棘轮门禁与坍缩警戒试算器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Variance Collapse under Self-Feeding: Why the Tails Go First — Self-feeding multiplies the expected variance by (n-1)/n per generation, typical paths shrink faster still, and the median half-life is about half the expected one. The tails leave before the variance does: a fifth less variance costs almost four fifths of the one-in-a-thousand tail. Mixing in a real source gives the variance a steady state, and keeping old real data gives the error an upper bound. Collapse is a conditional statement, not a fate. (https://5dlibrary.com/answers/en/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c01/) 第1章 自产回喂的方差坍缩:尾部为什么先消失 — 自产回喂让期望方差每代乘以 (n-1)/n,典型路径缩得更快,中位半衰期约为期望半衰期的一半。尾部比方差先走:方差缩两成,千分之一尾部就丢掉将近八成。每代掺进真源,方差就有稳态;不扔掉旧的真实数据,误差就有上界。坍缩是条件命题,不是宿命。 (https://5dlibrary.com/answers/zh/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c01/) - Ch.2 The Jagged Frontier: Why Clicking Confirm Loses Accuracy, and Why Better Individuals Make a Narrower Crowd — Chapter 2 treats the jagged frontier: with illustrative parameters, clicking confirm outside the frontier is right 0.604 of the time against 0.800 for answering alone, while falsifying first and then comparing leads on both sides. (https://5dlibrary.com/answers/en/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c02/) 第2章 锯齿边界:只点确认为何掉分,个体变好为何集体变窄 — 第二章算锯齿边界:示意参数下,边界外只点确认的正确率是 0.604,低于独立作答的 0.800,先证伪再比对在两侧都领先。 (https://5dlibrary.com/answers/zh/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c02/) - Ch.3 The Gain from Injected Outliers: How Worse-on-Average Candidates Raise the Best — Chapter 3 computes the gain from injected human outliers: a pool of 20 model candidates has an expected best of 0.934, two human hypotheses lift it to 1.186, and the gain holds only when the hypotheses are testable. (https://5dlibrary.com/answers/en/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c03/) 第3章 异常值注入增益:平均更差的候选为何能抬高最优解 — 第三章算人类异常值的注入增益:20 条候选全是模型产出时最优值期望 0.934,掺进两条人类假说升到 1.186,但增益只在假说可检验时成立。 (https://5dlibrary.com/answers/zh/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c03/) - Ch.4 The Falsification Gate: Sequential Boundaries, Sample Sizes and the Leak Rate — Chapter 4 builds the falsification gate: a sequential test needs 98 to 118 cases on average where a fixed design needs 211 at the same error rates, and the leak rate is computed by candidate source. (https://5dlibrary.com/answers/en/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c04/) 第4章 证伪门禁:序贯检验的边界、样本量与漏网率 — 第四章写证伪门禁:序贯检验在同样错误率下平均只要 98 到 118 个案例,固定样本要 211 个,并按候选来源算出漏网率。 (https://5dlibrary.com/answers/zh/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c04/) - Ch.5 A Robust Ratchet: Random Walks with Misclassification, Rollback Audits and the Condition for Compounding — Chapter 5 treats the ratchet as a random walk with misclassification, adds rollback audits and states the condition for compounding: with illustrative parameters the feedback strength must exceed 0.0556. (https://5dlibrary.com/answers/en/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c05/) 第5章 稳健棘轮:带误判的随机游走、回退审计与复利条件 — 第五章把棘轮写成带误判的随机游走,加上回退审计,并给出复利条件:示意参数下反馈强度要大于 0.0556。 (https://5dlibrary.com/answers/zh/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c05/) - Ch.6 Multi-Agent Falsification: Majority Error and the Effective Number of Votes under Correlation — Chapter 6 computes multi-agent falsification: correlated same-source agents add votes without adding protection, and the independent prediction of 0.0579 for five votes holds only under independence. (https://5dlibrary.com/answers/en/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c06/) 第6章 多代理证伪:多数票的误判率与相关误判下的有效票数 — 第六章算多代理证伪:同源代理相关时,多投票不等于多保险,五票独立预测的误判率 0.0579 只在独立时成立。 (https://5dlibrary.com/answers/zh/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c06/) - Ch.7 Classical Sources: Teaching and Learning in the Xueji, Blue from Indigo in the Xunzi, Bacon's Idols and Popper's Conjectures — The closed loop is worse. If each round's consensus becomes the next round's prior, the shared bias accumulates round by round. With no outside anchor, the bias is 0.500 after 10 rounds, 2.500 after 50 and 10.000 after 200, with no upper limit. If a share of 0.05 of each round is anchored on outside evidence, the bias settles at 0.950; with an anchored share of 0.20, at 0.200; with 0.50, at 0.050, which is the single-round bias itself. (https://5dlibrary.com/answers/en/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c07/) 第7章 原典考辨:学记论教学相长、劝学论青出于蓝、培根四假象与波普尔猜想反驳 — 更麻烦的是闭环。若每一轮的共识成为下一轮的先验,共享偏差会逐轮累积。没有外部锚点时,偏差 10 轮后是 0.500,50 轮后 2.500,200 轮后 10.000,没有上限。每轮有 0.05 的比例锚定在外部证据上,偏差最终稳定在 0.950;锚定比例 0.20,稳定在 0.200;锚定比例 0.50,稳定在 0.050,也就是单轮偏差本身。 (https://5dlibrary.com/answers/zh/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c07/) - Ch.8 Field Tools and Checker Specifications: the Daily Ratchet Ledger, Red-Yellow-Green Criteria and a Pre-registered Refutation — Chapter 8 gives the field tools and checker specifications: a daily ratchet ledger, red-yellow-green criteria, the distilled Ratchet Gate and Collapse Alarm Calculator and a pre-registered refutation plan. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c08/) 第8章 实战工具与检查器规格:每日棘轮台账、红黄绿判据与预注册证伪 — 第八章给出实战工具与检查器规格:每日棘轮台账、红黄绿判据、蒸馏小工具「人机互哺棘轮门禁与坍缩警戒试算器」与预注册证伪方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/human-ai-co-evolution-falsification-gated-ratchet-symbiotic-compounding/c08/) ## 5DL-B-375 · Supermodular Weapon-Arsenals, Adjacent-Possible Recombination Graphs and Cross-Asset Cashflow Chaining 中文书名:存量武器库超模组合代数、邻近可能联想图谱与跨业务火力串联母典 DOI: pending Cite: Fan, Y. (2026). Supermodular Weapon-Arsenals, Adjacent-Possible Recombination Graphs and Cross-Asset Cashflow Chaining (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: In the illustrative model, twelve of the fifteen pairs of six assets are complements and three are substitutes, and with all six switched on the monthly value is 23,104 against a stand-alone sum of 13,668. Parameters are illustrative and correspond to no real business, customer, creator or sale; all calculations are generated by the book's programs. Abstract: Operators who hold several assets hear three popular sayings: more assets are always better and each should simply be made strong on its own; new combinations come from inspiration; and the more you chain, the better, so that zero new development cost will double the cash flow. This book refutes or bounds each of them. Its core tool is the cross-difference: the value of two assets together minus the value of each alone; positive means complements, negative means substitutes. In the illustrative model, twelve of the fifteen pairs of six assets are complements and three are substitutes, and with all six switched on the monthly value is 23,104 against a stand-alone sum of 13,668. Parameters are illustrative and correspond to no real business, customer, creator or sale; all calculations are generated by the book's programs. This is general business analysis, not investment, legal or tax advice. Chapter 1 computes the matrix of pairwise gains and its Möbius decomposition. Chapter 2 presents supermodular functions on lattices and Topkis's monotone comparative statics: three reach assets cause 239 violations of supermodularity, substitution makes the best set swap members as conditions change, and complementary connections with high integration costs form a valley in which single pilots lose money while the joint connection gains 963. Chapter 3 draws a graph of the adjacent possible, showing that realisable combinations lie within one step and that unfamiliar combinations widen the spread and lower the median. Chapter 4 selects the best chaining set under caps on hours and contact: on supermodular problems the greedy method's worst case reaches only 0.2188 of the optimum, pair-aware selection lifts it to 0.5358, and any sensitivity to contact makes the best number of chains finite. Chapter 5 treats funnel bottlenecks and quantiles for the doubling slogan: chaining everything gives a median ratio to each-on-its-own of 1.07 and a doubling probability of 0.0804, while chaining only complements gives a median of 1.37 and cuts the chance of doing worse from 0.4563 to 0.1554; commission is single-level, paid only to the direct referrer by product band. Chapter 6 designs a 2×2 factorial test; detecting a cross-difference of 0.010 needs 8,379 visitors per cell. Chapter 7 examines the Sunzi, Aristotle and Adam Smith and closes with a distillation spec for the four kits. Chapter 8 delivers field tools and checker specifications, a pre-registered test and refutation conditions. Chaining never moves customer data between businesses, and the domain matrix is described only by public site names and content links. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:手握多项资产的操盘者常听到三句流行说法:资产越多越好、每项各自做强就行;新组合靠灵感;串联越多越好,零新增研发成本就能让现金流倍增。本书逐条证伪或立界。核心工具是交叉差分:两项资产一起的价值减去各自的价值,为正即互补,为负即替代。在示意模型里,六项资产两两配对的十五对中有十二对互补、三对替代,全部打开时月度价值 23,104,各自相加只有 13,668。书中参数为示意,不对应任何真实业务、客户、达人或成交;全部计算由书中程序生成。本书是非个人化经营分析,不构成投资、法律或税务意见。 第一章算配对增益矩阵与默比乌斯分解。第二章讲格上的超模函数与托普基斯单调比较静态:三项拉新资产造成 239 对超模违反,替代关系会让最优组合随条件「换人」;整合成本较高的互补接入会形成山谷,单项试点各自亏损,联合接入多赚 963。第三章画邻近可能图谱,说明能实现的组合只在一步之内,陌生组合抬高波动、压低中位数。第四章在工时与打扰上限下选最优串联集:贪心法在超模问题上最差只得最优值的 0.2188,配对感知挑法提到 0.5358;只要客户对打扰敏感,最优串联数就是有限的。 第五章写漏斗瓶颈与倍增口号的分位数:全部串联相对各自为战的比值中位数 1.07,达到两倍的概率 0.0804;只串互补的中位数 1.37,变差的概率从 0.4563 降到 0.1554;佣金只付直接推荐人,按商品档位单层计算。第六章设计 2×2 析因检验,检测 0.010 的交叉差分每格需要 8,379 位访客。第七章考辨《孙子·兵势》、亚里士多德与亚当·斯密,章末给出四件套蒸馏规格。第八章交付实战工具与检查器规格、预注册检验与推翻条件。跨业务串联不得搬运客户个人资料,多域名矩阵只写公开站名与内容互链。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Matrix of Pairwise Gains: Cross-Differences Separate Complements, Independents and Substitutes — The column the ledger lacks is the cross-difference. When it is positive the two assets are complements and should be run together; when it is negative they are substitutes and stacking them is waste; when it is near zero each can run on its own. In the illustrative model, twelve of fifteen pairs are complements and three are substitutes, and the combined value of 23,104 exceeds the stand-alone sum of 13,668 by nearly ten thousand. Second-order terms carry most of the surplus, but plans in which three assets are useless without one another still need to be tested as a bundle. (https://5dlibrary.com/answers/en/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c01/) 第1章 配对增益矩阵:交叉差分判别互补、独立与替代 — 账本少的那一栏是交叉差分。它为正,两项资产互补,应当一起经营;它为负,两项资产替代,叠加是浪费;它接近零,各做各的即可。在本书的示意模型里,十五对中有十二对互补、三对替代,组合价值 23,104 比各自相加的 13,668 多出近一万。二阶项占了组合剩余的绝大部分,但三项缺一不可的方案仍需整体检验。 (https://5dlibrary.com/answers/zh/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c01/) - Ch.2 Supermodular Functions on a Lattice: Topkis's Monotone Comparative Statics and the Single-Practice Pilot Trap — Chapter 2 presents supermodular functions on lattices and Topkis's monotone comparative statics: three reach assets cause 239 violations of supermodularity, substitution makes the best set swap members as conditions change, and complementary connections with high integration costs form a valley in which single pilots lose money while the joint connection gains 963. (https://5dlibrary.com/answers/en/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c02/) 第2章 格上的超模函数:托普基斯单调比较静态与单项试点的陷阱 — 第二章讲格上的超模函数与托普基斯单调比较静态:三项拉新资产造成 239 对超模违反,替代关系会让最优组合随条件「换人」;整合成本较高的互补接入会形成山谷,单项试点各自亏损,联合接入多赚 963。 (https://5dlibrary.com/answers/zh/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c02/) - Ch.3 A Graph of the Adjacent Possible: One-Step Combinations and the Trade-off of Unfamiliarity — Section B arranges ten trials into a quarter. With only familiar combinations ($u = 0.1$), the mean total is 11.77, the chance of at least one breakthrough is only 0.0057 and the worst tenth of quarters totals 9.49. With 1 unfamiliar trial ($u = 0.8$) the breakthrough chance rises to 0.0390 and the worst tenth falls to 9.16. With 2, the breakthrough chance is 0.0712 and the worst tenth 8.97. With 3, the figures are 0.1023 and 8.72. With all ten unfamiliar, the breakthrough chance reaches 0.2927 and the worst tenth sinks to 7.22. (https://5dlibrary.com/answers/en/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c03/) 第3章 邻近可能图谱:一步可达的组合与陌生度的取舍 — B 段把十次试验排成一个季度的组合。全部用熟悉组合($u = 0.1$),总价值均值 11.77,至少一次突破的概率只有 0.0057,最差一成的总价值是 9.49。放进 1 次陌生试验($u = 0.8$),突破概率升到 0.0390,最差一成降到 9.16;放进 2 次,突破概率 0.0712,最差一成 8.97;放进 3 次,突破概率 0.1023,最差一成 8.72。十次全部陌生,突破概率 0.2927,最差一成跌到 7.22。 (https://5dlibrary.com/answers/zh/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c03/) - Ch.4 The Best Chaining Set: Choosing Under Caps on Working Hours and Contact Frequency — Chapter 4 selects the best chaining set under caps on hours and contact: on supermodular problems the greedy method's worst case reaches only 0.2188 of the optimum, pair-aware selection lifts it to 0.5358, and any sensitivity to contact makes the best number of chains finite. (https://5dlibrary.com/answers/en/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c04/) 第4章 最优串联集:工时与打扰频次约束下的选择 — 第四章在工时与打扰上限下选最优串联集:贪心法在超模问题上最差只得最优值的 0.2188,配对感知挑法提到 0.5358;只要客户对打扰敏感,最优串联数就是有限的。 (https://5dlibrary.com/answers/zh/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c04/) - Ch.5 Chaining Cash Flow: the Bottleneck of a Cross-Asset Funnel, Quantiles for the Doubling Slogan and Single-Level Commission — Chapter 5 treats funnel bottlenecks and quantiles for the doubling slogan: chaining everything gives a median ratio to each-on-its-own of 1.07 and a doubling probability of 0.0804, while chaining only complements gives a median of 1.37 and cuts the chance of doing worse from 0.4563 to 0.1554; commission is single-level, paid only to the direct referrer by product band. (https://5dlibrary.com/answers/en/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c05/) 第5章 现金流串联:跨资产漏斗的瓶颈、倍增口号的分位数与单层分润 — 第五章写漏斗瓶颈与倍增口号的分位数:全部串联相对各自为战的比值中位数 1.07,达到两倍的概率 0.0804;只串互补的中位数 1.37,变差的概率从 0.4563 降到 0.1554;佣金只付直接推荐人,按商品档位单层计算。 (https://5dlibrary.com/answers/zh/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c05/) - Ch.6 Testing Complementarity: the 2×2 Factorial Test, Standard Errors and Sample Size — Chapter 6 designs a 2×2 factorial test; detecting a cross-difference of 0.010 needs 8,379 visitors per cell. (https://5dlibrary.com/answers/en/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c06/) 第6章 互补性检验:2×2 析因试验、标准误与样本量 — 第六章设计 2×2 析因检验,检测 0.010 的交叉差分每格需要 8,379 位访客。 (https://5dlibrary.com/answers/zh/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c06/) - Ch.7 Classical Sources: Regular and Unusual Forces in the Sunzi, Aristotle on Wholes and Adam Smith on the Extent of the Market — With 5,000 visitors a month the best number of modules is zero: no specialisation can support itself. With 10,000 visitors the best is 1 module, gaining 220 a month; with 20,000, 3 modules gaining 2,151; with 40,000, 5 modules gaining 7,680; with 80,000, 7 modules; with 160,000, 10; and with 320,000, 12 modules gaining 103,490. The break-even audience for the first module is 7,839 visitors. (https://5dlibrary.com/answers/en/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c07/) 第7章 原典考辨:《孙子·兵势》正奇之分、亚里士多德论整体与亚当·斯密论市场范围 — 每月访客 5,000 时,最优模块数是零:任何专门化都养不活。访客 10,000 时最优是 1 个,月增益 220;20,000 时 3 个,增益 2,151;40,000 时 5 个,增益 7,680;80,000 时 7 个;160,000 时 10 个;320,000 时 12 个,增益 103,490。第一个模块的盈亏平衡访客数是 7,839。 (https://5dlibrary.com/answers/zh/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c07/) - Ch.8 Field Tools and Checker Specifications: the Monthly Chaining Ledger, a Pre-Registered Test and Refutation Conditions — Chapter 8 delivers field tools and checker specifications, a pre-registered test and refutation conditions. Chaining never moves customer data between businesses, and the domain matrix is described only by public site names and content links. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c08/) 第8章 实战工具与检查器规格:月度串联台账、预注册检验与推翻条件 — 第八章交付实战工具与检查器规格、预注册检验与推翻条件。跨业务串联不得搬运客户个人资料,多域名矩阵只写公开站名与内容互链。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/supermodular-arsenal-adjacent-possible-recombination-cross-asset-chaining/c08/) ## 5DL-B-376 · Stage-Bottleneck Triage, Weighted Shortest-Processing-Time Scheduling and Level-of-Detail Execution Velocity 中文书名:阶段瓶颈动态定焦、价值工时比优先级排序与认知视锥渲染加速母典 DOI: pending Cite: Fan, Y. (2026). Stage-Bottleneck Triage, Weighted Shortest-Processing-Time Scheduling and Level-of-Detail Execution Velocity (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Smith proved in 1956 that when one person faces a list, the order that minimises the weighted cost of delay sorts tasks by value divided by duration, not by value; on seven illustrative tasks the importance order pays 0.311 more than the exhaustive optimum. Frustum culling and levels of detail from computer graphics show that a fixed frame budget should not be spent on doing everything finely: doing all eight illustrative tasks finely finishes only three and reaches 0.783 of the best plan. Abstract: Three sayings dominate advice on planning a day: do the most important thing first, do every task to the highest standard, and run several lines at once because parallel is faster. This book turns each into a computable claim and tests it. Smith proved in 1956 that when one person faces a list, the order that minimises the weighted cost of delay sorts tasks by value divided by duration, not by value; on seven illustrative tasks the importance order pays 0.311 more than the exhaustive optimum. Frustum culling and levels of detail from computer graphics show that a fixed frame budget should not be spent on doing everything finely: doing all eight illustrative tasks finely finishes only three and reaches 0.783 of the best plan. The output of a serial funnel is set by the stage with the smallest converted capacity: adding a fifth to each of five stages in turn raises weekly output from 15.00 to 18.00 only at the engagement stage. Parameters are illustrative and all calculations are generated by the book's programs. This is not management consulting and not health advice; it concerns one person scheduling their own work. After checking the original sources the book corrects its blueprint. The c-mu rule is optimal for a single server with linear costs, and the generalised c-mu rule is only asymptotically optimal in heavy traffic. Zero marginal value of non-bottleneck improvements holds for output, not for flow time. A single bottleneck applies to capacities only; yields multiply, so when demand binds every conversion step matters equally. Sprengel stated the substance of the law of the minimum in 1828. Greedy level selection by benefit per cost reaches 0.9960 of the optimum on average under concave value curves, but is not optimal. Chapters 1 to 6 cover value per minute, the c-mu rule and the Gittins index, phase bottlenecks and Little's law, frustum culling, multiple-choice knapsack levels, and log-normal tails with a pooled buffer; the focus cap is never stretched, rest is written into the window, and a plan that does not fit is cut or downgraded. Chapter 7 examines Sunzi on emptiness and fullness, the Mencius on declining to act, Seneca's second letter and the law of the minimum, and ends with a distillation spec for the four kits. Chapter 8 gives field tools and checker specifications: the daily scheduling ledger, the distilled Bottleneck Focus and Value-per-Hour Scheduler, red, yellow and green checks, daily, weekly and monthly rhythms, and a pre-registered paired-week test with refutation conditions; in the illustrative simulation the five-step pipeline averages 79.74 points a day against 42.85 for the habit of doing the most valuable task first and filling the cap. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:关于每天怎样排事,流行着三句说法:先做最重要的事,每件事都做到满分,多线并进才快。本书把三句话写成可以计算的命题逐一检验。史密斯 1956 年证明,单人面对一张清单,使加权延迟代价最小的次序是按价值除以耗时排序,而不是按价值排序;七件示意任务上,按重要性排比穷举最优多付 0.311。图形学的视锥剔除与细节层次说明,固定帧预算下不该每件都精做:八件示意任务全部精做只做完三件,总价值只有最优安排的 0.783。串联漏斗的产出由换算后产能最小的环节决定:示意漏斗里五个环节轮流加两成产能,只有承接环节那一次让周产出从 15.00 升到 18.00。书中参数为示意,全部计算由书中程序生成。本书非管理咨询,也非健康意见,只写一个人给自己排程。 核对原文以后,本书做了蓝图修正:cμ 法则的最优性限于单服务台与线性代价,广义 cμ 法则只在重负荷下渐近最优;非瓶颈改进零边际只对产出成立,对流转时间不成立;唯一瓶颈只对产能约束成立,转化率相乘,需求不足时每道转化同等重要;最小因子律的要义施普伦格尔 1828 年已经写出;按收益成本比的贪心定档在凹形价值曲线下平均达到最优的 0.9960,但不是最优。第一至六章依次讲价值工时比、cμ 法则与吉廷斯指数、阶段瓶颈与利特尔定律、视锥剔除、多选背包定档、对数正态尾部与合并缓冲;专注上限固定不许拉长,休息写进窗口,排不下就砍计划或降档。 第七章考辨《孙子·虚实篇》、《孟子·离娄下》、塞涅卡第二封信与最小因子律,章末给出四件套蒸馏规格。第八章给出实战工具与检查器规格:每日作战排程台账字段表、蒸馏小工具「瓶颈定焦与价值工时比排程器」的参数与输出、红黄绿检查器、日周月节奏、预注册成对周检验与推翻条件;示意模拟中五步流程平均每天 79.74 点,排满上限的「最有价值先做」习惯只有 42.85 点。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Why "Most Important First" Is Wrong: Value per Hour and the Adjacent Interchange — With k equal to zero or 0.3 the ratio order barely changes. Its expected cost is about 292.8 points, roughly 0.318 below the importance order. With k equal to 0.6 the planned duration of the shelving task shrinks from 180 minutes to 93 minutes, and the task jumps to second place in the ratio order. The expected cost rises to 316.3 points, 0.080 more than ordering by the true ratios. Even so, the biased ratio order still beats the importance order, by 0.220. (https://5dlibrary.com/answers/en/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c01/) 第1章 先做最重要的事为什么不对:价值工时比与相邻交换 — k 等于零与 0.3 时,按比值排的次序几乎不变,期望代价约 292.8 点,比按重要性排省下 0.318 左右。k 等于 0.6 时,书库上架的计划耗时从 180 分钟被压到 93 分钟,它在比值排序里一下子跳到第二位,期望代价升到 316.3 点,比真实比值排序多付 0.080。即便如此,它仍比按重要性排好,差距 0.220。 (https://5dlibrary.com/answers/zh/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c01/) - Ch.2 Streams of Arriving Work: the c-mu Rule, the Generalised c-mu Rule and the Gittins Index — Read the results at three loads. At a load of 0.70, the fixed rule that always gives questions priority costs 2.560 points per task on average and the generalised rule costs 2.769, so the fixed rule is cheaper. At 0.85 the fixed rule costs 6.022 and the generalised rule 6.420, and the fixed rule still wins narrowly. Only at 0.95 does the generalised rule, at 16.998, overtake the fixed rule at 17.348. First come, first served is poor at all three loads and reaches 127.846 per task at 0.95; giving publishing tasks priority is worse still, at 444.840. (https://5dlibrary.com/answers/en/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c02/) 第2章 源源不断的任务流:cμ 法则、广义 cμ 与吉廷斯指数 — 结果分三档负荷读。负荷 0.70 时,固定让咨询优先的规则每件平均代价 2.560 点,广义 cμ 是 2.769 点,固定规则反而更省;负荷 0.85 时,固定规则 6.022 点,广义规则 6.420 点,仍是固定规则略好;负荷 0.95 时,广义规则 16.998 点,才反超固定规则的 17.348 点。先来先做在三档负荷下都很差,到 0.95 时每件 127.846 点;让出版局任务优先更糟,到 444.840 点。 (https://5dlibrary.com/answers/zh/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c02/) - Ch.3 The Single Bottleneck of a Phase: a Five-Stage Funnel, Little's Law and the Moving Constraint — All four cases show the same weekly output, 14.03 orders, because that is how many come in and every case can cope with them. The differences lie in flow time and work in progress. In the baseline, an order spends on average 1.402 weeks in the chain, there are 19.63 orders in progress, and 12.79 of them are waiting in front of engagement. Speeding acquisition up by half cuts the flow time to 1.302 weeks, while the queue in front of engagement is 12.84 orders. Speeding delivery up by half cuts the flow time to 1.224 weeks, with 12.79 orders still waiting in front of engagement. Only when engagement is sped up by a fifth does the flow time fall to 0.673 weeks and work in progress to 9.42 orders, with just 2.75 waiting in front of engagement. (https://5dlibrary.com/answers/en/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c03/) 第3章 阶段唯一瓶颈:五环节漏斗、利特尔定律与瓶颈转移 — 四种情形的周产出都是 14.03 单,因为进来的就是这么多,四种情形都接得住。差别在流转时间与在制品。基线下,一张订单平均在链上停留 1.402 周,在制品 19.63 张,其中 12.79 张堵在承接门前。获客提速一半,流转时间降到 1.302 周,承接门前反而是 12.84 张;交付提速一半,流转时间降到 1.224 周,承接门前仍是 12.79 张。只有承接提速两成,流转时间才降到 0.673 周,在制品降到 9.42 张,承接门前只剩 2.75 张。 (https://5dlibrary.com/answers/zh/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c03/) - Ch.4 Frustum Culling: Tasks with No Causal Path to Cash — With no misreading at all, culling raises the day's value from 110.7 points to 219.1, an increase of 0.979, nearly double. With a misreading rate of 0.1 the increase is 0.846; at 0.3 it is still 0.548; even at 0.5, when every other task is misread, culling still adds 0.200. The price appears in the last column: the value left sitting in the parked pile rises with the misreading rate, from 17.8 points to 155.7. (https://5dlibrary.com/answers/en/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c04/) 第4章 视锥剔除:与现金流没有因果路径的任务 — 完全不误判时,剔除让每天的价值从 110.7 点升到 219.1 点,多出 0.979,接近翻倍。误判率 0.1 时多出 0.846,误判率 0.3 时仍多出 0.548,误判率高到 0.5,也就是每两件看错一件,剔除仍然多出 0.200。代价写在最后一列:留在挂起区里的价值随误判率上升,从 17.8 点升到 155.7 点。 (https://5dlibrary.com/answers/zh/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c04/) - Ch.5 Levels of Detail: a Multiple-Choice Knapsack inside a Fixed Focus Budget — Doing everything finely is the most revealing case. Only three of the eight tasks get done, for a total value of 101, which is 0.783 of the optimal plan. Doing everything coarsely finishes all eight tasks but reaches a total of only 58, which is 0.450 of the optimum. Doing everything at medium finishes seven tasks for a total of 114, 0.884 of the optimum: not bad, but still short. The optimal plan reaches 129. (https://5dlibrary.com/answers/en/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c05/) 第5章 细节分档:固定专注时长里的多选背包 — 全部精做的结果最能说明问题:八件只做完三件,总价值 101,是最优安排的 0.783。全部粗做八件都做完了,总价值只有 58,是最优的 0.450。全部中做做完七件,总价值 114,是最优的 0.884,已经不坏,但仍差一截。最优安排的总价值是 129。 (https://5dlibrary.com/answers/zh/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c05/) - Ch.6 No Stalls: Log-Normal Tails, Pooled Buffers and a Focus Budget That Includes Rest — The first row is the most favourable case: no systematic bias and little spread, with s at 0.3. Even so, a day without a buffer runs over with probability 0.590. A buffer of 0.15 brings this down to 0.231, a buffer of 0.30 to 0.058, and a buffer of half the plan to 0.006. The second row widens the spread to 0.6. Without a buffer the probability of running over is 0.669; with a buffer of 0.30 it is still 0.326, and with half the plan as buffer it is 0.177. (https://5dlibrary.com/answers/en/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c06/) 第6章 防卡顿:对数正态尾部、合并缓冲与含休息的专注预算 — 第一行是最乐观的情形:估计没有系统偏差,波动也小,离散参数 0.3。即便如此,不留缓冲的日子有 0.590 的概率超时;缓冲 0.15 时降到 0.231,缓冲 0.30 时降到 0.058,缓冲一半时只剩 0.006。第二行把波动放大到 0.6,不留缓冲的超时概率是 0.669,缓冲 0.30 时仍有 0.326,缓冲一半时 0.177。 (https://5dlibrary.com/answers/zh/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c06/) - Ch.7 The Classical Sources: Sunzi on Emptiness and Fullness, Mencius on Declining, Seneca's Second Letter and the Law of the Minimum — In the demand-bound week, demand is 12.00 orders, capacity is spare and orders are 12.00. Raising visitors, the first yield or the second yield lifts orders to 13.20 in each case, an elasticity of 1.00; the elasticities of both capacities are 0.00. The three conversion factors matter equally, and no single link matters more than the others. In the capacity-bound week, demand is already 27.00 orders but engagement capacity is only 15.00. Now only engagement capacity has an elasticity of 1.00, and raising it by a tenth lifts orders to 16.50; the three conversion factors and delivery capacity all have elasticities of 0.00. (https://5dlibrary.com/answers/en/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c07/) 第7章 原典考辨:《孙子·虚实篇》、《孟子·离娄下》、塞涅卡第二封信与最小因子律 — 需求不足的那一周,需求是 12.00 单,产能有富余,订单就是 12.00 单。这时访客、第一道转化率、第二道转化率,无论抬高哪一个,订单都升到 13.20,弹性都是 1.00;两项产能的弹性都是 0.00。三个转化因素同等重要,不存在唯一要紧的一环。产能受限的那一周,需求已有 27.00 单,承接产能只有 15.00 单。这时只有承接产能的弹性是 1.00,抬高一成订单升到 16.50;三个转化因素与交付产能的弹性全是 0.00。 (https://5dlibrary.com/answers/zh/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c07/) - Ch.8 Field Tools and Checker Specifications: the Daily Scheduling Ledger, the Scheduler and a Pre-registered Test — Chapter 8 gives field tools and checker specifications: the daily scheduling ledger, the distilled Bottleneck Focus and Value-per-Hour Scheduler, red, yellow and green checks, daily, weekly and monthly rhythms, and a pre-registered paired-week test with refutation conditions; in the illustrative simulation the five-step pipeline averages 79.74 points a day against 42.85 for the habit of doing the most valuable task first and filling the cap. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c08/) 第8章 实战工具与检查器规格:每日作战排程台账、排程器与预注册检验 — 第八章给出实战工具与检查器规格:每日作战排程台账字段表、蒸馏小工具「瓶颈定焦与价值工时比排程器」的参数与输出、红黄绿检查器、日周月节奏、预注册成对周检验与推翻条件;示意模拟中五步流程平均每天 79.74 点,排满上限的「最有价值先做」习惯只有 42.85 点。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/stage-bottleneck-triage-wspt-cmu-scheduling-level-of-detail-velocity/c08/) ## 5DL-B-377 · Next-Generation Knowledge Distillation, Context-Parametric Compilation and Live-Conversion Feedback Calibration 中文书名:高阶知识蒸馏进化、条件参数化编译与实盘转化反馈校准母典 DOI: pending Cite: Fan, Y. (2026). Next-Generation Knowledge Distillation, Context-Parametric Compilation and Live-Conversion Feedback Calibration (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. This is general business analysis, not investment, legal or medical advice, and it describes no real customer or sale. Abstract: Three popular beliefs govern the distillation of monographs into rules: a good rule holds wherever you put it, more rules make a pack more accurate, and a threshold once set can be left alone. This book refutes each of them or draws its boundary. Simpson in 1951 and Bickel and co-authors in 1975 showed that a direction holding in aggregate can reverse within strata: pooled over the six largest Berkeley departments, men were admitted at 0.445 and women at 0.304, yet women did better in four departments. Rules must therefore be written as maps from conditions to parameters, with a condition vector of price tier, traffic source, device, first or return visit, and time slot. Parameters are illustrative and all calculations are generated by the book's programs. This is general business analysis, not investment, legal or medical advice, and it describes no real customer or sale. Chapter 1 prices the single-sentence rule: on an illustrative decision table, "always B" is wrong for 0.5433 of traffic. Chapter 2 compiles the 72-cell table into a decision list and an ordered decision diagram whose size ranges from 6 to 23 nodes depending on variable order. Chapter 3 scans for conflicts across volumes; just 9 rules put 0.7366 of traffic in conflict, and certainty factors, when five books restate one observation, inflate the posterior to 8.19 times its correct value, so evidence must be deduplicated by observation. Chapter 4 calibrates rules online with Beta posteriors and Thompson sampling, whose mean regret over forty thousand sessions is 50.2; detecting a rise in pay rate from 0.040 to 0.044 needs 39,475 visitors per arm. Chapter 5 compares discounting, sliding windows and a cumulative-sum alarm; under a step change the sliding window's total regret is 33.0. Chapter 6 checks calibration with the Brier partition and binned calibration error, re-scaling cuts the error from 0.0210 to 0.0035, and a whole-pipeline simulation shows online calibration without a minimum-sample guard losing to the static pack in 164 of 200 worlds. Chapter 7 reads Wheelwright Bian in the Zhuangzi, Mencius on law that cannot carry itself out, and Aristotle on equity; the unspoken residual share in the illustrative model is 0.291. Chapter 8 gives the Conditional-Parametric Rule Compiler and Online Calibrator, checker specifications, a risk-control ledger and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:把专著蒸馏成规则,最流行的三句说法是:好规则放到哪里都成立,规则越多越准,阈值定好就不用动。本书逐条证伪或划界。辛普森 1951 年与比克尔等人 1975 年说明,汇总层成立的方向分层后可以反转:伯克利六个最大院系合计,男性录取率 0.445、女性 0.304,分院系看却有四个院系女性更高。因此规则必须写成条件到参数的形式,条件向量包括客单价档、流量来源、设备、首访或回访与时段。书中参数为示意,全部计算由书中程序生成。本书是非个人化经营分析,不构成投资、法律或医疗意见,不对应任何真实客户或成交。 第一章量化单句规则的代价,在示意决策表上「一律用 B」对 0.5433 的流量是错的。第二章把 72 格决策表编译成决策列表与有序决策图,变量次序不同,节点数从 6 个到 23 个不等。第三章做跨卷冲突扫描,9 条规则就让 0.7366 的流量落入冲突区;确定性因子在五卷书重述同一观察时,把后验抬到正确值的 8.19 倍,必须按观察编号去重。第四章用贝塔后验与汤普森抽样在线校准,四万次会话的平均遗憾为 50.2;要检出付款率从 0.040 升到 0.044,每组需要 39,475 人。 第五章比较折扣、滑动窗口与累积和报警,滑动窗口在台阶变化下合计遗憾 33.0。第六章用布莱尔分解与分箱校准误差检验校准度,再标定把误差从 0.0210 降到 0.0035;全流程模拟显示,不设最小样本门槛的在线校准在 200 个世界里有 164 个不如静态规则包。第七章考辨《庄子》轮扁斫轮、《孟子》徒法不能以自行与亚里士多德的公道论,说不出的残差份额在示意模型中为 0.291。第八章给出条件参数化规则编译与在线校准器、检查器规格、风控台账与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Simpson's Reversal: Why a Rule Without Its Context Can Point the Wrong Way — Chapter 1 prices the single-sentence rule: on an illustrative decision table, "always B" is wrong for 0.5433 of traffic. (https://5dlibrary.com/answers/en/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c01/) 第1章 辛普森反转:脱离环境的单句规则为何会指错方向 — 第一章量化单句规则的代价,在示意决策表上「一律用 B」对 0.5433 的流量是错的。 (https://5dlibrary.com/answers/zh/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c01/) - Ch.2 Context-Parametric Compilation: Squeezing a Decision Table into Decision Lists and Ordered Decision Diagrams — Chapter 2 compiles the 72-cell table into a decision list and an ordered decision diagram whose size ranges from 6 to 23 nodes depending on variable order. (https://5dlibrary.com/answers/en/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c02/) 第2章 条件参数化编译:把决策表压成决策列表与有序决策图 — 第二章把 72 格决策表编译成决策列表与有序决策图,变量次序不同,节点数从 6 个到 23 个不等。 (https://5dlibrary.com/answers/zh/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c02/) - Ch.3 Resolving Conflicts Across Volumes: Do More Rules Make a Pack More Accurate? — Chapter 3 scans for conflicts across volumes; just 9 rules put 0.7366 of traffic in conflict, and certainty factors, when five books restate one observation, inflate the posterior to 8.19 times its correct value, so evidence must be deduplicated by observation. (https://5dlibrary.com/answers/en/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c03/) 第3章 跨卷冲突裁决:规则越多越准吗 — 第三章做跨卷冲突扫描,9 条规则就让 0.7366 的流量落入冲突区;确定性因子在五卷书重述同一观察时,把后验抬到正确值的 8.19 倍,必须按观察编号去重。 (https://5dlibrary.com/answers/zh/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c03/) - Ch.4 Online Bayesian Calibration: Giving Every Rule a Beta Distribution That Grows — Chapter 4 calibrates rules online with Beta posteriors and Thompson sampling, whose mean regret over forty thousand sessions is 50.2; detecting a rise in pay rate from 0.040 to 0.044 needs 39,475 visitors per arm. (https://5dlibrary.com/answers/en/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c04/) 第4章 在线贝叶斯校准:给每条规则一个会长大的贝塔分布 — 第四章用贝塔后验与汤普森抽样在线校准,四万次会话的平均遗憾为 50.2;要检出付款率从 0.040 升到 0.044,每组需要 39,475 人。 (https://5dlibrary.com/answers/zh/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c04/) - Ch.5 Threshold Drift: Discounted Beta Updates, Sliding Windows and a Cumulative-Sum Alarm — Chapter 5 compares discounting, sliding windows and a cumulative-sum alarm; under a step change the sliding window's total regret is 33.0. (https://5dlibrary.com/answers/en/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c05/) 第5章 阈值漂移:折扣贝塔、滑动窗口与累积和报警 — 第五章比较折扣、滑动窗口与累积和报警,滑动窗口在台阶变化下合计遗憾 33.0。 (https://5dlibrary.com/answers/zh/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c05/) - Ch.6 Calibration and the Whole Pipeline: the Brier Partition, Binned Calibration Error and a Net-Benefit Simulation — Chapter 6 checks calibration with the Brier partition and binned calibration error, re-scaling cuts the error from 0.0210 to 0.0035, and a whole-pipeline simulation shows online calibration without a minimum-sample guard losing to the static pack in 164 of 200 worlds. (https://5dlibrary.com/answers/en/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c06/) 第6章 校准度与全流程:布莱尔分解、分箱校准误差与净收益模拟 — 第六章用布莱尔分解与分箱校准误差检验校准度,再标定把误差从 0.0210 降到 0.0035;全流程模拟显示,不设最小样本门槛的在线校准在 200 个世界里有 164 个不如静态规则包。 (https://5dlibrary.com/answers/zh/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c06/) - Ch.7 Classical Sources: Wheelwright Bian and Aristotle on Equity — Chapter 7 reads Wheelwright Bian in the Zhuangzi, Mencius on law that cannot carry itself out, and Aristotle on equity; the unspoken residual share in the illustrative model is 0.291. (https://5dlibrary.com/answers/en/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c07/) 第7章 原典考辨:轮扁斫轮与亚里士多德的公道 — 第七章考辨《庄子》轮扁斫轮、《孟子》徒法不能以自行与亚里士多德的公道论,说不出的残差份额在示意模型中为 0.291。 (https://5dlibrary.com/answers/zh/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c07/) - Ch.8 Closing the Loop: Field Tools and Checker Specifications, the Risk Ledger and a Pre-registered Test — Chapter 8 gives the Conditional-Parametric Rule Compiler and Online Calibrator, checker specifications, a risk-control ledger and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c08/) 第8章 实证闭环:实战工具与检查器规格、风控台账与预注册检验 — 第八章给出条件参数化规则编译与在线校准器、检查器规格、风控台账与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/knowledge-distillation-context-parametric-decision-graphs-live-conversion-bayesian-calibration/c08/) ## 5DL-B-378 · Beyond Distillation: Executable Business-Organ Synthesis, Generative Code Crystallisation and Living-System Bootstrapping 中文书名:超越蒸馏:可执行商业器官合成、生成式代码结晶与活体系统自举母典 DOI: pending Cite: Fan, Y. (2026). Beyond Distillation: Executable Business-Organ Synthesis, Generative Code Crystallisation and Living-System Bootstrapping (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. This is general business analysis, not investment, legal or medical advice, and the code it generates is teaching material deployed to no site. Abstract: Three sayings are popular: the harder you distil, the better; generated code is right if it runs; a live system only needs maintenance. This book refutes the first, bounds the second and rewrites the third as a testable claim. Compression keeps only the regularities already in the data. Of twelve rules distilled from a mother book, three hard limits account for 0.0920 of rule firings in the home market and are the first to go when one distils by frequency, yet they account for 0.3780 in the target market; an old price list covers only 0.067 of a new market's legal prices. The book argues for upward synthesis: compiling a monograph's theorems and a target market's constraints into small organs that carry their own specifications and invariants. Blueprint corrections after checking the literature: distilled models generalise within their original constraints; a fingerprint self-check misses deliberate tampering, so an outside anchor is required; the mutation score is a proxy; a 44 pixel target is WCAG level AAA; and bootstrapping means self-checks and version fingerprints, never copying or spreading. Parameters are illustrative and all calculations are generated by the book's programs. This is general business analysis, not investment, legal or medical advice, and the code it generates is teaching material deployed to no site. Chapter 1 shows what compression keeps and what synthesis supplies. Chapter 2 writes the margin floor, price endings, early-bird ladders and a single-layer referral payout as invariants of a pricing engine and attacks them with property tests. Chapter 3 writes the front end as a state machine, enumerates reachability, deadlocks and traps, and checks contrast and tap height on every walk to the paid state. Chapter 4 draws an objection tree that counts a leaf as covered only when it has an answer, evidence and a pass on the copy red lines. Chapter 5 watches conversion and error rates with cumulative-sum probes and writes alarms back into the mother book, which avoids 0.7764 of recurring anomalies in simulation. Chapter 6 tests the organ itself with mutation testing, a crystallinity index and fingerprint self-checks: an outside anchor catches all of 1,000 random changes, while a replaced self-check catches none. Chapter 7 reads the Kaogongji, Aristotle's four causes, the Laozi and Goethe, and distils them into clauses for the four kits. Chapter 8 gives the organ-synthesis ledger, a red, yellow and green checker specification, the distilled Business-Organ Synthesis and Self-Check Tool and a pre-registered test: synthesis without a checker lets at least one defect escape within ninety days with probability 0.9530, and a checker lowers that to 0.2218. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:知识产品圈有三句流行说法:蒸馏得越狠越好;生成的代码能跑就算对;系统上线之后只需维护。本书证伪第一句,给第二句立界,把第三句改写成可检验的命题。压缩只能保留数据里已有的规律:母书十二条规则里,三条硬约束在原市场只占 0.0920 的触发频次,按频次蒸馏时最先被砍掉,到了目标市场却占 0.3780;一张跑了两年的价目表压缩后只覆盖新市场合法价位的 0.067。本书主张向上合成:把专著定理连同目标市场的约束(计价单位、价格带、合规红线、界面门禁)正向编译成自带规格与不变量的小器官。核对原始文献后,本书做了蓝图修正:知识蒸馏可以在原约束内泛化,新约束必须作为额外规格输入;指纹自检只抓意外漂移,抓不住蓄意篡改,须设器官写不到的外部锚点;变异得分是代理指标;44 像素点按目标属于 WCAG 的 AAA 级,比 AA 级更严;「活体系统」只作类比,自举只指自检与版本指纹,不指复制或传播。书中参数为示意,全部计算由书中程序生成。本书是非个人化经营建议,不构成投资、法律或医疗意见;书中生成的代码只作教学样例,不部署到任何站点。 第一章用描述长度、规则覆盖与版本空间说明压缩保留什么、合成补上什么。第二章把毛利下限、尾数、早鸟阶梯与单层转介绍报酬写成计价引擎的不变量,并用性质测试去撞。第三章把前端写成状态机,穷举可达性、死锁与陷阱,对每条通往付款态的走法检查对比度与点按高度。第四章画异议拆解树,只把有回答、有证据、过文案红线的叶子计入覆盖。第五章用累积和探针盯转化率与报错率,并把报警反向扩写进母书,模拟中可避免 0.7764 的同类异常。 第六章用变异测试、结晶度与指纹自检检验器官本身,随机改动 1,000 次全部被外部锚点抓到,而被替换的自检一次也抓不到。第七章考辨《考工记》「天有时,地有气,材有美,工有巧」、亚里士多德四因、《老子》第四十二章与歌德植物变形论,并落成四件套蒸馏规格。第八章给出器官合成台账、红黄绿检查器规格、小工具「商业器官合成与自检器」与预注册检验:不带检查器的合成在九十天里至少逃逸一个缺陷的概率是 0.9530,带检查器降到 0.2218。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Ceiling of Distillation: Compression Keeps, Synthesis Generates — Now read the table. With all twelve rules kept, the description is 522 bytes long and nothing in the target market goes unhandled. Cut to ten rules, the description shrinks to 415 bytes, a ratio of 0.795, and home-market coverage is still 0.9439, which looks almost lossless. Yet in the target market 0.2266 of cases now have no rule to follow, and 0.6045 of the hard-limit cases are missed. Cut to eight rules, home-market coverage still reads 0.8758, while the share of hard-limit cases missed reaches 1.0000: every one of them is lost. Cut to two rules, the description is down to 75 bytes, a ratio of 0.144, and 0.6673 of target-market cases are handled by nobody. (https://5dlibrary.com/answers/en/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c01/) 第1章 蒸馏的天花板:压缩只能保留,合成才能生成 — 再看表格。保留全部十二条规则,描述长度 522 字节,目标市场零遗漏。砍到十条,描述只剩 415 字节,压缩比 0.795,原市场覆盖率仍有 0.9439,看上去几乎无损;可目标市场已有 0.2266 的场景无规则可依,硬约束场景漏掉 0.6045。砍到八条,原市场覆盖率还有 0.8758,硬约束场景却漏掉 1.0000,也就是全部漏掉。砍到两条,描述只剩 75 字节,压缩比 0.144,目标市场有 0.6673 的场景无人处理。 (https://5dlibrary.com/answers/zh/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c01/) - Ch.2 The Pricing Engine: Constraint Solving, Invariants and Property Tests — In M1 the cost converts to 38.00, the lower bound is 70.91 and the four tiers run from 71.90 to 113.90. In M2 the cost converts to 167.20, the lower bound is 299.47, the ending rule is "whole tens minus one" and the tiers run from 309.00 to 489.00. In M3 the cost converts to 7.98, the lower bound is 17.58, the ending is ninety-nine cents and the tiers run from 17.99 to 28.99. The lowest-tier margins in the three markets are 0.3076, 0.3174 and 0.3607, each at or above its own floor; the floor in M3 is 0.35, while M1 and M2 keep 0.30. (https://5dlibrary.com/answers/en/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c02/) 第2章 计价引擎:约束求解、不变量与性质测试 — M1 的成本折算为 38.00,下界 70.91,四档价格从 71.90 到 113.90。M2 的成本折算为 167.20,下界 299.47,尾数规则是整十减一,四档从 309.00 到 489.00。M3 的成本折算为 7.98,下界 17.58,尾数九毛九,四档从 17.99 到 28.99。三个市场的最低档净毛利分别是 0.3076、0.3174、0.3607,全部不低于各自下限,其中 M3 的下限是 0.35。 (https://5dlibrary.com/answers/zh/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c02/) - Ch.3 The Front-End Statechart: Reachability, Deadlock and an Interface Gate on Every Path — With five orthogonal regions, each drawn once in a statechart, there are 17 states and 27 transitions. Flattened into a single machine, there are 360 states and 2,652 transitions. Two coupling constraints across regions (payment may leave idle only when the cart is locked; paid or refunded requires consent first) cut the legal states to 144, a share of 0.4000. (https://5dlibrary.com/answers/en/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c03/) 第3章 前端状态机:可达性、死锁与每条路径的界面门禁 — 五个正交分区,在状态图里各画一次,一共 17 个状态、27 条转移。拼成平铺状态机,状态数是 360,转移数是 2,652。两条跨分区的耦合约束(购物车未锁定时付款不得离开空闲态;已付款或已退款必须先有同意)把合法状态砍到 144 个,占 0.4000。 (https://5dlibrary.com/answers/zh/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c03/) - Ch.4 The Objection Tree: Coverage, Depth and the Order of Repairs — With budgets of 2, 4 and 6 hours, greedy and optimum agree exactly, and coverage rises through 0.7300, 0.7725 and 0.8325. With 8 hours, greedy reaches 0.8725 and the optimum 0.8775, a ratio of 0.9943. The difference lies in the choice. Greedy picks free alternative, reviews and no time; the optimum picks free alternative, reviews and instalments. Greedy takes a small, cheap gap early, and the remaining hours can then no longer fit the larger combination. At 10 and 12 hours the ratios are 0.9973 and 0.9948. At 14 hours every gap is closed and coverage reaches 1.0000. (https://5dlibrary.com/answers/en/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c04/) 第4章 异议拆解树:覆盖率、深度与补洞次序 — 预算 2、4、6 小时时,贪心与最优完全相同,覆盖率依次升到 0.7300、0.7725、0.8325。预算 8 小时,贪心得到 0.8725,最优是 0.8775,比值 0.9943。差别在于选择:贪心选了免费替代品、评价、没时间三项,最优选了免费替代品、评价、分期三项。贪心先拿了一个性价比高的小洞,结果剩下的时间凑不出更大的组合。预算 10 小时与 12 小时时,比值分别是 0.9973 与 0.9948;14 小时全部补完,覆盖率 1.0000。 (https://5dlibrary.com/answers/zh/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c04/) - Ch.5 The Live Probe: CUSUM, False-Alarm Run Lengths and Writing Back into the Mother Book — Chapter 5 watches conversion and error rates with cumulative-sum probes and writes alarms back into the mother book, which avoids 0.7764 of recurring anomalies in simulation. (https://5dlibrary.com/answers/en/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c05/) 第5章 实盘探针:CUSUM、假警间隔与反向扩写母书 — 第五章用累积和探针盯转化率与报错率,并把报警反向扩写进母书,模拟中可避免 0.7764 的同类异常。 (https://5dlibrary.com/answers/zh/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c05/) - Ch.6 Crystallinity and Bootstrapping: Mutation Testing, a Composite Index and the Limits of Self-Fingerprinting — Chapter 6 tests the organ itself with mutation testing, a crystallinity index and fingerprint self-checks: an outside anchor catches all of 1,000 random changes, while a replaced self-check catches none. (https://5dlibrary.com/answers/en/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c06/) 第6章 结晶度与自举:变异测试、合成指标与指纹自检的边界 — 第六章用变异测试、结晶度与指纹自检检验器官本身,随机改动 1,000 次全部被外部锚点抓到,而被替换的自检一次也抓不到。 (https://5dlibrary.com/answers/zh/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c06/) - Ch.7 Classical Sources East and West: the Kaogongji's Four Conditions, the Laozi's Generative Sequence and Aristotle's Four Causes — The lower half of the table gives each condition a probability of holding. With all four at 0.9, the AND rule's yield of good work is only 0.6561, while the scoring rule reports 0.9477. With one condition weak at 0.6 and the others at 0.95, the AND yield is 0.5144 and the scoring rule still reports 0.9386. The scoring rule overstates the yield systematically. The last line shows how much the weakest link weighs: raising the 0.6 to 0.95 lifts the yield from 0.5144 to 0.8145. (https://5dlibrary.com/answers/en/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c07/) 第7章 东西方原典考辨:《考工记》四材、《老子》生成序与亚里士多德四因 — 下半张表给每个条件一个成立概率。四项都是 0.9 时,与规则的良品率只有 0.6561,打分规则却报 0.9477。一项弱到 0.6、其余 0.95 时,与规则的良品率是 0.5144,打分规则仍报 0.9386。打分规则在系统性地高估。最后一行说明短板的分量:把那项 0.6 提到 0.95,良品率从 0.5144 跳到 0.8145。 (https://5dlibrary.com/answers/zh/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c07/) - Ch.8 The Empirical Loop: Field Tools and Checker Specifications — Chapter 8 gives the organ-synthesis ledger, a red, yellow and green checker specification, the distilled Business-Organ Synthesis and Self-Check Tool and a pre-registered test: synthesis without a checker lets at least one defect escape within ninety days with probability 0.9530, and a checker lowers that to 0.2218. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c08/) 第8章 实证闭环:实战工具与检查器规格 — 第八章给出器官合成台账、红黄绿检查器规格、小工具「商业器官合成与自检器」与预注册检验:不带检查器的合成在九十天里至少逃逸一个缺陷的概率是 0.9530,带检查器降到 0.2218。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/executable-business-organ-synthesis-generative-code-crystallisation-living-system-bootstrapping/c08/) ## 5DL-B-379 · Zero-Friction First-Minute Onboarding, Micro-Interaction Habit Loops and High-Stickiness Interface Engineering 中文书名:首分钟零摩擦上手、微交互成瘾回路与产品超高粘性留存工程母典 DOI: pending Cite: Fan, Y. (2026). Zero-Friction First-Minute Onboarding, Micro-Interaction Habit Loops and High-Stickiness Interface Engineering (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Enlarging a tap target from 24 to 44 pixels lowers the illustrative miss rate from 0.0865 to 0.0017. Parameters are illustrative and correspond to no real user or sale; all calculations are generated by the book's programs. Abstract: Product teams repeat three popular sayings: register people first or you will not keep them; the richer the feedback the better; a habit takes twenty-one days and stickiness comes from bombarding people with notifications. This book rewrites each saying as parameters, computes the consequences with standard-library programs, and states what result, after back-filling with one's own data, would refute its conclusions. A first-minute funnel multiplies: under illustrative parameters a registration wall takes 9 steps and only 3,477 visitors in ten thousand see a first result, while play first, save later takes 2 steps and 8,924 see one, about 2.57 times as many; the two flows yield the same number of accounts once the save offer is accepted by 0.508 of those who saw a result. Feedback depends on latency, not decoration. Enlarging a tap target from 24 to 44 pixels lowers the illustrative miss rate from 0.0865 to 0.0017. Parameters are illustrative and correspond to no real user or sale; all calculations are generated by the book's programs. The book is not medical or psychological advice and not personalised business advice. After checking the original sources the book corrects its blueprint: the median time to a habit plateau reported by Lally and colleagues is about 66 days, and the upper figure of 254 days is an extrapolation; a 2025 meta-analysis found no general memory advantage for unfinished tasks, so the book designs dismissible resumption cards rather than betting on memory; and a 44-pixel tap height belongs to the strictest level of the accessibility guidelines, which the house front-end style kit adopts. The "habit loop" of the title is only an analogy from behaviour design. The book does not teach anyone to build addiction, does not target minors and does not exploit compulsive use; every feedback channel can be switched off with one action and notification frequency is capped; dark patterns appear only as counter-examples. In the illustrative model a second daily notification buys only 0.96 active days while driving several hundred more members per ten thousand away. Chapters 1 to 6 cover the first-minute funnel, feedback within a hundred milliseconds, tapping and choosing, suspense and ratchets, the ethical ceiling of variable reinforcement and the DAU/MAU retention equation; moving the stickiness ratio from 0.2 to 0.5 requires the next-day return rate to rise from 0.8196 to 0.9549. Chapter 7 examines the Laozi on beginning hard things with the easy, the Xunzi on small paces, and Aristotle and William James on habit: the easy-to-hard order completes with probability 0.5757 and the hard-to-easy order with 0.4805. Chapter 8 supplies field tools and checker specifications: a ledger, a red-amber-green checklist, the distilled First-Minute Onboarding and Stickiness Equation Calculator and a pre-registered test in which 3,857 visitors per arm cover three hypotheses. Marked boxes separate published evidence, 23 runnable Python programs, classical sources and the author's hypotheses. 摘要:产品圈有三句流行说法:先注册才留得住人;反馈越华丽越好;二十一天养成习惯,粘性靠推送轰炸。本书把它们写成参数,用纯标准库程序计算,并写明用自己的数据回填后什么结果会推翻结论。首分钟漏斗是乘法:示意参数下,注册墙要走 9 步,每一万名访客只有 3,477 人看到第一个结果;先玩后存只要 2 步,有 8,924 人看到,约为前者的 2.57 倍,存档接受率达到 0.508 时两者账号数打平。回馈的关键是时延而不是特效:时延是分布,主战场在尾部。点按目标从 24 像素放大到 44 像素,示意误点率从 0.0865 降到 0.0017。书中参数为示意,不对应任何真实用户或成交;全部计算由书中程序生成。本书不是医疗或心理干预意见,也非个人化经营建议。 核对原文之后,本书做了蓝图修正:拉利等人的习惯平台期中位数约 66 天,上端 254 天为外推值;2025 年元分析显示未完成任务的记忆优势并不普遍,因此本书设计可关闭的续接卡片,而不押注记忆;44 像素点按高度在无障碍指南里属最高级,帝国前端样式库取更严的这一档。书名里的「成瘾回路」只是行为设计的类比:本书不教制造成瘾,不针对未成年人,不利用强迫性使用;所有回馈可以一键关闭,推送频次封顶;暗黑模式只作反例。示意模型里,每天第二条推送只多换来 0.96 个活跃日,却多送走每一万人里数百位成员。 第一至六章依次讨论首分钟漏斗、百毫秒回馈、点按与选择、悬念与棘轮、可变强化的伦理上限与 DAU/MAU 留存方程;把粘性比从 0.2 推到 0.5,需要次日回访率从 0.8196 升到 0.9549。第七章考辨《老子》「天下难事必作于易」、《荀子》「不积跬步」、亚里士多德与威廉·詹姆斯论习惯:从易到难的顺序完成率 0.5757,从难到易 0.4805。第八章给出实战工具与检查器规格:台账、红黄绿检查表、蒸馏小工具「首分钟上手与粘性方程试算器」与预注册检验,每组 3,857 名访客即可覆盖三项假设。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者假说。 Chapter answers: - Ch.1 The First-Minute Funnel: a Registration Wall Puts Friction Before Value — Chapters 1 to 6 cover the first-minute funnel, feedback within a hundred milliseconds, tapping and choosing, suspense and ratchets, the ethical ceiling of variable reinforcement and the DAU/MAU retention equation; moving the stickiness ratio from 0.2 to 0.5 requires the next-day return rate to rise from 0.8196 to 0.9549. (https://5dlibrary.com/answers/en/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c01/) 第1章 首分钟漏斗:注册墙把价值兑现推迟到摩擦之后 — 第一至六章依次讨论首分钟漏斗、百毫秒回馈、点按与选择、悬念与棘轮、可变强化的伦理上限与 DAU/MAU 留存方程;把粘性比从 0.2 推到 0.5,需要次日回访率从 0.8196 升到 0.9549。 (https://5dlibrary.com/answers/zh/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c01/) - Ch.2 Feedback Within a Hundred Milliseconds: Latency Thresholds Beat Lavish Effects — Latency is a distribution, and the battle is fought in its tail. Two versions whose medians differ by only 0.35 seconds can differ by more than a factor of ten in the share of responses above one second. A skeleton screen shortens no latency, yet in the illustrative model it reduces the share of visitors who leave by 0.25; optimistic updates suit only operations that rarely fail and can be fully restored. Serial requests push the P95 over one second quickly, and with these parameters no more than 2 serial requests fit before the first result. Effects are welcome, but they come after latency, and they can always be switched off. (https://5dlibrary.com/answers/en/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c02/) 第2章 百毫秒回馈:时延阈值胜过华丽特效 — 时延是分布,主战场在尾部。中位数只差 0.35 秒的两个版本,超过一秒的比例可以差十倍以上。骨架屏不缩短时延,却在示意模型里让离开比例下降 0.25;乐观更新只适合失败率低、可完整恢复的操作。串行请求让 P95 迅速越过一秒门槛,首个结果之前最多只容得下 2 次串行请求。特效可以有,但必须排在时延之后,而且可以关闭。 (https://5dlibrary.com/answers/zh/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c02/) - Ch.3 Tapping and Choosing: Fitts' Law, Hick's Law and the Tappable Elements of the First Screen — Fitts' law shows that the logarithm flattens the effect of target size on movement time, while the effect on misses is large: under the illustrative parameters, a 24-pixel target is missed with probability 0.0865 and a 44-pixel target with probability 0.0017. Hick's law combined with the height of the screen gives an optimum of about 8 options per screen, 2.908 seconds in all, and the time jumps once the options are squeezed into a scrolling list. Contrast must be computed, and 4.48 is not 4.5. The first screen should be sparse, large and legible. (https://5dlibrary.com/answers/en/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c03/) 第3章 点按与选择:Fitts 定律、Hick 定律与首屏可点元素 — Fitts 定律说明,目标尺寸对移动时间的影响被对数压扁,对误点率的影响却很大:在示意参数下,24 像素目标的误点率是 0.0865,44 像素只有 0.0017。Hick 定律与屏幕高度合起来,给出每屏约 8 个选项的最优,合计 2.908 秒;选项挤到要滚动时,耗时跳升。对比度要用公式算,4.48 不是 4.5。首屏要少、要大、要清楚。 (https://5dlibrary.com/answers/zh/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c03/) - Ch.4 Suspense and Ratchets: Endowed Progress, the Goal Gradient and Personal Assets — Endowed progress works, but only through perception. In the illustrative model card B is completed with probability 0.4117 against 0.3119 for card A, and when effort depends only on the purchases actually made the two cards are indistinguishable. The memory advantage of unfinished tasks is not general according to the meta-analysis, while the tendency to resume is stable, so the product should provide a resumption card rather than bet on memory. Personal assets reduce churn, mostly through value and partly through lock-in; the book gives up the lock-in part and allows export throughout. (https://5dlibrary.com/answers/en/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c04/) 第4章 悬念与棘轮:赠予进度、目标梯度与个人资产 — 赠予进度有效,但效应依赖感知;示意模型里,B 卡完成率 0.4117 对 A 卡 0.3119,而当努力只看自己买了几次时,两卡毫无差别。未完成任务的记忆优势在元分析里并不普遍,续做倾向却稳定,所以要设计续接卡片,不要押注记忆。个人资产降低流失,大部分来自价值,小部分来自锁定;本书放弃锁定,一律允许导出。 (https://5dlibrary.com/answers/zh/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c04/) - Ch.5 The Ethical Ceiling of Variable Reinforcement: Frequency Caps, One-Tap Off Switches and Dark-Pattern Counter-Examples — Variable reinforcement works through prediction error. In the illustrative model the error when a reward arrives is 0.000 under a fixed ratio and 0.791 under a variable ratio, and after rewards stop the first learner gives up after 5 misses while the second needs 12. This force may be used only on real diversity of content. The second daily notification buys only 0.96 active days in the model while driving away 654 more members per ten thousand, so the book caps notifications at one a day. The eight dark-pattern rules serve only as counter-examples, and any hit is red. (https://5dlibrary.com/answers/en/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c05/) 第5章 可变强化的伦理上限:频次封顶、一键关闭与暗黑模式反例 — 可变强化之所以有效,在于预测误差:示意模型里,固定比率下回馈到手时误差为 0.000,可变比率下为 0.791;回馈停止后,前者落空 5 次就放弃,后者要 12 次。这股力量只能用在内容的真实多样性上。推送的第二条在示意模型里只多换来 0.96 个活跃日,却多送走每一万人里 654 位成员,所以本书封顶每天 1 条。暗黑模式八条规则全部只作反例,任何命中即判红。 (https://5dlibrary.com/answers/zh/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c05/) - Ch.6 The DAU/MAU Retention Equation: Visit-Frequency Distributions and the sBG Retention Curve — DAU/MAU is determined by the distribution of visit frequency. With the mean held fixed, the ratio can move from 0.105 to 0.416 simply because light users drop out of the denominator; the lower bound follows from the mean and the upper bound is 1.000. The sBG model shows that retention rates rise with cohort age, and that a cohort's DAU/MAU rises too, from 0.191 towards 0.334, so the more young cohorts a product has, the lower its overall ratio. Moving the ratio from 0.2 to 0.5 requires the next-day return rate to rise from 0.8196 to 0.9549, which is the shape of a daily utility and not a target for every product. (https://5dlibrary.com/answers/en/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c06/) 第6章 DAU/MAU 留存方程:访问频率分布与 sBG 留存曲线 — DAU/MAU 由访问频率分布决定。均值不变时,比值可以从 0.105 变到 0.416,只因轻度用户从分母里消失;下界由均值算出,上界是 1.000。sBG 模型说明,留存率随群组年龄上升,同一群组的 DAU/MAU 也从 0.191 升向 0.334,新群组越多,整体比值越低。把比值从 0.2 推到 0.5,需要次日回访率从 0.8196 升到 0.9549,那是日常工具的形态,不是所有产品的目标。 (https://5dlibrary.com/answers/zh/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c06/) - Ch.7 Source Criticism: the Laozi, the Xunzi, Aristotle and William James on Beginnings and Habit — Chapter 7 examines the Laozi on beginning hard things with the easy, the Xunzi on small paces, and Aristotle and William James on habit: the easy-to-hard order completes with probability 0.5757 and the hard-to-easy order with 0.4805. (https://5dlibrary.com/answers/en/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c07/) 第7章 原典考辨:《老子》《荀子》与亚里士多德、威廉·詹姆斯论起手与习惯 — 第七章考辨《老子》「天下难事必作于易」、《荀子》「不积跬步」、亚里士多德与威廉·詹姆斯论习惯:从易到难的顺序完成率 0.5757,从难到易 0.4805。 (https://5dlibrary.com/answers/zh/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c07/) - Ch.8 Field Tools and Checker Specifications: the First-Minute and Stickiness Ledger, the Calculator and a Pre-registered Test — Chapter 8 supplies field tools and checker specifications: a ledger, a red-amber-green checklist, the distilled First-Minute Onboarding and Stickiness Equation Calculator and a pre-registered test in which 3,857 visitors per arm cover three hypotheses. Marked boxes separate published evidence, 23 runnable Python programs, classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c08/) 第8章 实战工具与检查器规格:首分钟与粘性台账、试算器与预注册检验 — 第八章给出实战工具与检查器规格:台账、红黄绿检查表、蒸馏小工具「首分钟上手与粘性方程试算器」与预注册检验,每组 3,857 名访客即可覆盖三项假设。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/zero-friction-first-minute-onboarding-micro-interaction-habit-loops-stickiness-engineering/c08/) ## 5DL-B-380 · Commercial Storytelling Mechanics, Counter-Arguing Suppression Arcs and Seamless Narrative-to-Checkout Bridges 中文书名:商业故事劝买动力学、心防卸载叙事弧与无痕切入成交母典 DOI: pending Cite: Fan, Y. (2026). Commercial Storytelling Mechanics, Counter-Arguing Suppression Arcs and Seamless Narrative-to-Checkout Bridges (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative, and all calculations are generated by the book's programs. This is general business analysis, not legal or medical advice; every story sample is an illustrative fiction, and no testimonial or real case is fabricated. Abstract: People who sell online repeat three beliefs: the plainer the pitch the better, so the hard sell works best; the longer and more tearful the story the better; and a story should be told first, with the sales pitch in a separate paragraph afterwards. This book refutes or bounds each of them. Research on psychological reactance and on persuasion knowledge shows that a recognised selling intent provokes counter-arguing and resistance; research on narrative transportation shows that stories lower counter-arguing. After checking the original papers, the book corrects its blueprint: the persuasive advantage of narrative is steady and moderate, with pooled correlations of roughly 0.17 to 0.23 in meta-analysis and only 0.063 in health communication; the identifiable-victim effect comes from donation experiments rather than purchases; adherence to narrative-arc structure is unrelated to popularity and can be cited only as a descriptive regularity; and applying the peak-end rule and the curiosity gap to the timing of a call to action is this book's inference, to be settled by pre-registered tests. Parameters are illustrative, and all calculations are generated by the book's programs. This is general business analysis, not legal or medical advice; every story sample is an illustrative fiction, and no testimonial or real case is fabricated. Chapter 1 sets out the counter-arguing and transport model and computes the narrow boundary inside which a hard sell still wins: readers already intending to buy and an argument close to flawless. Chapter 2 derives a dose ceiling for transportation; in the illustrative model the best story segment is about 600 hanzi with light melodrama, while a long, tearful version keeps only 33.9 % of the best value; it also sets the ratio of an identifiable person to statistics. Chapter 3 covers six emotional arcs, the inverted U of reversal density and the share of concrete sentences. Chapter 4 chooses openings by similarity times intensity and rejects every opening built on grief or fear. Chapter 5 gives the bridge formula: place the call to action at the emotional peak, keep the jump in voice small, and answer the question the story left open. Chapter 6 sets five compliance red lines and simulates three landing pages: per thousand visitors, 9.8 payments for the hard sell, 10.5 for a story with a hard cut and 18.4 for a story with a seamless bridge (illustrative). Chapter 7 examines Han Feizi on the difficulty of persuasion, Zou Ji's parable in the Zhanguo ce, and Aristotle's Poetics and Rhetoric, and distils clauses for four house kits. Chapter 8 delivers a story-page ledger, the Narrative Purchase Arc and Checkout Bridge Checker, a checker specification table, a pre-registered test (3,581 visitors per arm for the primary comparison) and the conditions under which the book's claims are refuted. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:做线上生意的人常说三句话:卖点越直白越好,硬推销最有效;故事越长、越煽情越好;故事讲完再另起一段推销。本书逐条证伪或立界。心理抗拒与说服知识的研究说明,被识破的推销意图会激起反驳与抗拒;叙事传输研究说明,故事能压低反驳。核对原文之后,本书做了蓝图修正:叙事的说服优势稳健而温和,元分析给出的合并相关大约在 0.17 到 0.23 之间,健康传播里只有 0.063;可识别个体效应来自捐款实验而非购买实验;叙事弧的结构与作品受欢迎程度无关,只能作为描述性规律;峰终规则与好奇缺口搬到行动召唤的时机上,是本书的推论,须经预注册检验。书中参数为示意,全部计算由书中程序生成。本书是非个人化经营建议,不构成法律或医疗意见;所有故事样本均为示意虚构,不伪造任何证言或真实案例。 第一章写反驳与传输模型,并算出硬推销仍然胜出的窄边界:读者早已准备购买、论证接近满分。第二章写传输的剂量上限,示意模型里最佳故事段约 600 字、轻度煽情,又长又催泪的版本只剩最佳值的 33.9 %;并给出具体人物与统计句的配比。第三章写六种情绪弧、转折密度的倒 U 形与具象句比例,并说明单一模型给出的精确最佳点不可信。第四章用相似度乘强度系数选择开头,悲伤与恐惧开头一律拒用。第五章给出桥接公式:行动召唤放在情绪峰值处,语气跳变小,回答故事留下的问题。 第六章写医疗功效、威胁催促、收益担保、会重置的倒计时与未标注证言五条红线,并用蒙特卡洛模拟三种落地页:每千位访客成交,硬推销 9.8 笔,故事后硬切 10.5 笔,故事无痕桥接 18.4 笔(示意)。第七章考辨《韩非子·说难》、《战国策》邹忌讽齐王与亚里士多德《诗学》《修辞学》,并给出四件套蒸馏规格。第八章给出故事落地页台账、叙事劝买弧与成交桥接检查器、检查器规格表、预注册检验(主检验每组 3,581 位访客)与推翻条件。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Counter-Arguing and Transport: Why the Hard Sell So Often Fails — The trouble with the hard sell is not that it speaks too directly. It is that it places its intent exactly where the reader is best equipped to argue against it. The value of a story is not emotional intensity either; it lies in lowering the counter-arguing rate so that the argument gets a chance to land. Published meta-analyses show the advantage to be steady and moderate, with pooled correlations of roughly 0.17 to 0.23. The hard sell wins only in a narrow region, where readers have already decided to buy and the argument is close to flawless. (https://5dlibrary.com/answers/en/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c01/) 第1章 反驳与传输:硬推销为什么常常失灵 — 硬推销的问题不在于话说得太直,而在于它把意图摆在读者最容易反驳的位置。故事的价值也不在于煽情,而在于降低反驳率,让论证有机会落地。已发表的元分析显示,这种优势稳健而温和,合并相关大约在 0.17 到 0.23 之间。硬推销只在一个窄区间里胜出:读者早已准备购买,而且论证质量接近满分。 (https://5dlibrary.com/answers/zh/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c01/) - Ch.2 The Dose of Transportation: Longer and Sadder Is Not Better — Chapter 2 derives a dose ceiling for transportation; in the illustrative model the best story segment is about 600 hanzi with light melodrama, while a long, tearful version keeps only 33.9 % of the best value; it also sets the ratio of an identifiable person to statistics. (https://5dlibrary.com/answers/en/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c02/) 第2章 传输的剂量:故事不是越长越好、越煽情越好 — 第二章写传输的剂量上限,示意模型里最佳故事段约 600 字、轻度煽情,又长又催泪的版本只剩最佳值的 33.9 %;并给出具体人物与统计句的配比。 (https://5dlibrary.com/answers/zh/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c02/) - Ch.3 The Parameters of a Narrative Arc: Emotional Curves, Reversal Density and Concreteness — Shape is harder to tune than length. Of the six emotional arcs, the three that end at their peak are best suited to a call to action, and a story's mean valence has almost no bearing on its peak-end score. Reversal density follows an inverted U, but a precise optimum from any single model deserves little trust: with the illustrative data, the intervals from both fits missed the truth. The candidate band for the share of concrete sentences is 0.45 to 0.70. (https://5dlibrary.com/answers/en/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c03/) 第3章 叙事弧的参数:情绪曲线、转折密度与具象比例 — 故事的形状比长度更难调。六种情绪弧里,结尾停在峰值的三种最适合接上行动召唤;平均情绪的高低几乎不影响峰终分数。转折密度呈倒 U 形,可单一模型给出的精确最佳点并不可信,示意数据里两种拟合的区间都没包住真值。具象句比例的候选带是 0.45 到 0.70。 (https://5dlibrary.com/answers/zh/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c03/) - Ch.4 The Empathic Opening: How a Story Lets Readers In — With 100 impressions per arm, only 10 of the two hundred tests pick the right opening and 190 reach no decision. At 400 impressions per arm, 36 pick correctly; at 1,600, 109; at 3,200, 156, with 44 still undecided. Notice that the column for wrong picks is 0 throughout. The strict 95 % rule almost never crowns the wrong opening; the price is that it often declines to crown any. (https://5dlibrary.com/answers/en/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c04/) 第4章 困境共情起点:开头怎样让读者走进故事 — 每组 100 次曝光时,两百次检验里只有 10 次选对,190 次没有结论。到 400 次曝光,选对 36 次;1,600 次曝光,选对 109 次;3,200 次曝光,选对 156 次,仍有 44 次没有结论。值得注意的是,「选错」一栏始终为 0:严格的 95 % 规则几乎不会误判,代价是常常给不出结论。 (https://5dlibrary.com/answers/zh/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c04/) - Ch.5 The Bridge Formula: From Climax to Call to Action — The story paragraph averages 13.7 words a sentence, uses no second person, has a past-tense share of 0.244 and contains no sales words. The hard-cut bridge collapses to 3.0 words a sentence with a sales-word density of 0.778, a jump of 15.67, and is marked red. The soft bridge averages 11.5 words, switches to the second person (a share of 0.130) and to the present tense (the past-tense share falls to 0.043), for a jump of 2.56, marked yellow. The seamless bridge keeps third-person, past-tense narration (0.138), averages 14.5 words a sentence and uses no sales words; its jump is 1.06 and it is marked green. (https://5dlibrary.com/answers/en/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c05/) 第5章 桥接公式:从高潮到行动召唤 — 故事段平均句长 13.7 词,没有第二人称,过去时标记占 0.244,推销词为零。硬切桥接句长骤降到 3.0,推销词密度 0.778,跳变 15.67,判红。温和桥接句长 11.5,改用第二人称(占 0.130),时态切到现在时(过去时只剩 0.043),跳变 2.56,判黄。无痕桥接保持第三人称与过去时叙述(0.138),句长 14.5,没有推销词,跳变 1.06,判绿。 (https://5dlibrary.com/answers/zh/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c05/) - Ch.6 Compliance Red Lines and a Full-Funnel Simulation: A Smooth Bridge Must Not Become a Trapdoor — Chapter 6 sets five compliance red lines and simulates three landing pages: per thousand visitors, 9.8 payments for the hard sell, 10.5 for a story with a hard cut and 18.4 for a story with a seamless bridge (illustrative). (https://5dlibrary.com/answers/en/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c06/) 第6章 合规红线与全流程模拟:顺畅的桥不能变成暗道 — 第六章写医疗功效、威胁催促、收益担保、会重置的倒计时与未标注证言五条红线,并用蒙特卡洛模拟三种落地页:每千位访客成交,硬推销 9.8 笔,故事后硬切 10.5 笔,故事无痕桥接 18.4 笔(示意)。 (https://5dlibrary.com/answers/zh/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c06/) - Ch.7 Eastern and Western Sources Examined: The Difficulty of Persuasion, Zou Ji, Reversal and the Three Proofs — When pitch and listener match, the fit lies between 0.98 and 0.99 and the attitude shifts are 0.735 and 0.738. When they are mismatched, the fit drops to 0.38 and 0.43 and the shifts to 0.563 and 0.581. The mixed pitch fits the two listeners at 0.88 and 0.86, with shifts of 0.707 and 0.703: never the best, but always close. When an owner faces a crowd of readers whose minds differ, a mixed opening is often safer than a bet on any single mind. (https://5dlibrary.com/answers/en/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c07/) 第7章 东西方原典考辨:说难、邹忌、突转与三种说服 — 说法与听者对路时,契合度在 0.98 到 0.99 之间,态度位移 0.735 与 0.738。错位时,契合度掉到 0.38 与 0.43,位移降到 0.563 与 0.581。兼顾两头的说法,对两位听者的契合度是 0.88 与 0.86,位移 0.707 与 0.703,不是最好,却最稳。经营者面对一群心思各异的读者时,兼顾式开头往往比押注某一种心思更安全。 (https://5dlibrary.com/answers/zh/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c07/) - Ch.8 Field Tools and Checker Specifications: The Story-Page Ledger, a Pre-Registered Test and the Conditions of Refutation — Chapter 8 delivers a story-page ledger, the Narrative Purchase Arc and Checkout Bridge Checker, a checker specification table, a pre-registered test (3,581 visitors per arm for the primary comparison) and the conditions under which the book's claims are refuted. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c08/) 第8章 实战工具与检查器规格:故事落地页台账、预注册检验与推翻条件 — 第八章给出故事落地页台账、叙事劝买弧与成交桥接检查器、检查器规格表、预注册检验(主检验每组 3,581 位访客)与推翻条件。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/commercial-storytelling-narrative-transport-counterarguing-suppression-checkout-bridges/c08/) ## 5DL-B-381 · Information Theory of Code-Switching, Lexical-Gap Compensation and Monolingual Precision Equivalence in Multilingual Markets 中文书名:多语参杂语码转换信息论、词汇空缺补偿与单语种等效精准表达母典 DOI: pending Cite: Fan, Y. (2026). Information Theory of Code-Switching, Lexical-Gap Compensation and Monolingual Precision Equivalence in Multilingual Markets (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs; nothing corresponds to a real customer, conversation or sale, and Malay, Hokkien, Cantonese, English and Mandarin are treated as equals. This is not linguistic doctrine and not business advice to any individual or company. Abstract: Three popular beliefs surround mixed-language speech and writing: that slipping in a foreign word reveals poor ability or laziness; that the inserted word is a random habit with no information; and that a text that must stay in one language has to lose precision or wear translationese. This book tests each. The first is refuted: switch points respect the equivalence constraint and the matrix language frame, random switching would break word order with probability 0.5714 while natural corpora rarely do, and intra-sentential switching is most typical of speakers fluent in both languages. The second is refuted too: switches favour positions where the native word is hard to predict. After checking the original papers the book corrects its blueprint: published work finds the inserted English word harder to predict than its Chinese synonym, so insertion usually does not save bits; the book therefore books its benefit as code length plus residual ambiguity, and notes that the link with surprisal may also reflect cognitive load. Refutation is followed by a boundary. Insertion pays only when readers share the bilingual lexicon. Writing monolingual expression, insertion and insertion with a gloss as three cost lines in coverage gives, under illustrative parameters, a written-channel threshold of 0.844 for kiasu; when readers do not ask, failures become silent misreadings; and words worth inserting on regional pages can all fall below threshold on one mixed page. Insertion density has an optimum and an upper bound, 5.10 per hundred characters at a coverage of 0.95. The third belief fails as well: recombining meaning at the level of the phrase, with non-equivalence strategies, dynamic equivalence and construction grammar, lets monolingual constructions reach the target feature coverage with translationese below the manuscript line of 18.44. Chapter 7 examines Yan Fu's three difficulties, the Analects on correct names, Jerome's sense for sense and Schleiermacher's two paths. Chapter 8 gives the insertion ledger, checker specifications for four kits, the distilled Code-Switching Net-Gain and Monolingual-Equivalence Calculator and a pre-registered test. Parameters are illustrative and all calculations are generated by the book's programs; nothing corresponds to a real customer, conversation or sale, and Malay, Hokkien, Cantonese, English and Mandarin are treated as equals. This is not linguistic doctrine and not business advice to any individual or company. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. 摘要:关于多语参杂,流行着三条说法:夹杂外语是语言能力差、偷懒;夹进来的词只是随机习惯,没有信息量;必须写纯中文或纯英文时,只能接受精度损失或翻译腔。本书逐条检验。第一条被证伪:转换点守着两套语法的等价约束与基质语框架,随机换语言会以 0.5714 的概率破坏语序,自然语料却很少违约,句内转换最常见于两种语言都流利的说话人。第二条也被证伪:转换点偏向母语词难以预测、信息量高的位置。核对原文之后,本书做了蓝图修正:已发表研究显示,转换点上插入的英文词比同义中文词更难预测,嵌词本身通常并不省比特;本书因此改按码长加剩余歧义的总账记嵌词收益,并指出转换与惊异度的关联也可能来自认知负荷。 证伪之后要立界。嵌词收益只在读者共享双语词库时成立:把单语表达、嵌入与嵌入加注释写成覆盖率的三条成本线,示意参数下 kiasu 在书面渠道的覆盖率阈值为 0.844;读者不发问时,失败会变成沉默的误读;各地区分页面值得嵌的词,合成混合页面后可能全部跌破阈值。嵌词密度有最优值与上界,覆盖率 0.95 时上界为每百字 5.10 个。第三条流行说法同样站不住:用非对等策略、动态对等与构式语法在短语层面重组,单语构式能达到目标特征覆盖率,翻译腔标记可以降到书稿线 18.44 以下。 第七章考辨严复的译事三难、《论语》的正名、哲罗姆的意译与施莱尔马赫的两条路。第八章给出嵌词台账、四件套检查器规格、蒸馏小工具「多语嵌词净收益与单语构式等效试算器」与预注册检验。书中参数为示意,全部计算由书中程序生成,不对应任何真实客户、对话或成交;马来语、闽南语、粤语、英语、华语一律平等对待。本书不是语言学定论,也不构成针对个人或企业的经营建议。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Not Laziness: Switch Points Obey Two Grammars — This figure gives the laziness hypothesis a sharp consequence. If speakers changed language at random positions, each switch would break the equivalence constraint with probability 0.5714. In a corpus of 200 switches the random model expects about 114.3 violations. Poplack reported the opposite: violations of the constraint were rare in her data. The last line of the program turns this into a likelihood ratio. Suppose we observe 2 violations, and give the constrained model a small allowance for slips (eps 0.01). The base-ten logarithm of the likelihood ratio between the constrained model and the random one is 68.5. This is not a close contest. (https://5dlibrary.com/answers/en/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c01/) 第1章 不是偷懒:转换点守着两套语法 — 这组数字给出偷懒假说的一个直接推论。假如说话人在随机位置换语言,每一次转换违反等价约束的概率就是 0.5714。一份收录 200 次转换的语料,随机模型预期约 114.3 次违规。波普拉克报告的情形恰好相反:她的语料里,违反约束的转换极少。程序最后一行把这件事写成似然比:假设观察到 2 次违规,守约模型容许百分之一的偶发失误(eps 0.01),两种模型的似然比取以十为底的对数是 68.5。这不是一个接近的比赛。 (https://5dlibrary.com/answers/zh/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c01/) - Ch.2 The Information at a Switch Point: Surprisal, Word Length and Uniform Density — The circumlocution runs to 16 syllables with a peak of 2.67 bits and a variance of only 0.10; no syllable exceeds capacity. The inserted word shortens the utterance to 12 syllables and carries 35.00 bits in total, slightly less than the 36.00 of the circumlocution. But its peak rises to 7.50, its variance to 4.20, and the bits above capacity add up to 5.00. The squared cost, one way the uniform density view penalises peaks, rises from 82.67 to 152.50. (https://5dlibrary.com/answers/en/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c02/) 第2章 转换点的信息量:惊异度、词长与信息均匀 — 迂回表达总长 16 个音节,峰值 2.67 比特,方差只有 0.10,没有任何音节超出容量。嵌入词把总长压到 12 个音节,总信息量 35.00 比特,略低于迂回表达的 36.00 比特,可峰值升到 7.50,方差升到 4.20,超出容量的部分合计 5.00 比特。平方代价(信息均匀理论惩罚峰值的一种写法)从 82.67 升到 152.50。 (https://5dlibrary.com/answers/zh/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c02/) - Ch.3 The Bit Ledger and the Degree of a Lexical Gap — The corrected bit ledger adds code length and residual ambiguity. Under the illustrative parameters, two of the six gap words are longer in code length when inserted, yet five are cheaper on the full account and only one is more expensive; the bits saved range from +1.33 to +7.22. Gap degree can be measured as the feature mismatch of the nearest word, but the depth of a gap and the benefit of insertion do not move together: it also matters how many similar concepts surround the nearest word. Each language trades simplicity against precision, gaps arise from those trades, and a bilingual speaker can borrow the category boundaries of another language to cut meaning more finely. (https://5dlibrary.com/answers/en/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c03/) 第3章 比特账与词汇空缺度 — 修正后的比特账把码长与剩余歧义加在一起。在示意参数下,六个空缺词里有两个嵌入后码长更长,可总账上五个更省,只有一个更贵;省下的比特从 +1.33 到 +7.22 不等。空缺度可以用最近邻词的特征失配率来量,但空缺深浅与嵌词收益并不同步,还要看最近邻词周围有多少相似概念。语言各自在简单与精确之间取舍,空缺因此而生;双语者借用另一种语言的分类边界,就能把意思切得更细。 (https://5dlibrary.com/answers/zh/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c03/) - Ch.4 The Audience-Coverage Threshold: When an Inserted Word Becomes a Decoding Failure — The benefit of insertion is conditional: enough people have to know the word. When monolingual expression, insertion and insertion with a gloss are written as three cost lines in coverage, the illustrative threshold at which insertion breaks even runs from 0.789 to 0.952 in live chat and from 0.844 to 0.967 in written messages; words with a negative precision account never pay, whatever their coverage. The gloss has a transition window for only a few words. When readers do not ask, failures become invisible misreadings, whose risk rises linearly as coverage falls. Words that deserve insertion on regional pages may all fall below threshold once those pages are merged into one mixed page. (https://5dlibrary.com/answers/en/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c04/) 第4章 听众覆盖率阈值:嵌词何时变成解码失败 — 嵌词收益有条件:认得的人要足够多。把单语、嵌入、嵌入加注释写成覆盖率的三条成本线,在示意参数下,嵌入与单语打平的阈值 p* 在即时聊天里从 0.789 到 0.952,在书面私讯里从 0.844 到 0.967;精度账为负的词,不论覆盖率多高都不划算。注释只在少数词上有一段过渡窗口。读者不发问时,失败会变成看不见的误读,其风险随覆盖率下降线性上升。各地区分页面值得嵌的词,合成一张混合页面后可能全部跌破阈值。 (https://5dlibrary.com/answers/zh/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c04/) - Ch.5 An Upper Bound on Insertion Density, and the Register Signal — Insertion density is not a switch but a continuous quantity. The register benefit has diminishing returns, the readability loss accelerates, and the cost of decoding failure is proportional to density and to the share of readers who do not know the words. Under illustrative parameters, at a coverage of 0.95 the optimum is 2.16 inserted words per hundred characters with an upper bound of 5.10, and at a coverage of 0.50 or below nothing should be inserted. The register signal carried by insertion can be measured as mutual information, and in the synthetic table it removes only a small part of the uncertainty about register. Under a reliability requirement for single messages the allowed density is far below the average optimum, and one-to-one messages should take the smaller of the two. (https://5dlibrary.com/answers/en/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c05/) 第5章 嵌词密度上界与语域信号 — 嵌词密度不是开关,而是连续量。语域收益边际递减,可读性损失加速上升,解码失败成本与密度和不认得的比例成正比;在示意参数下,覆盖率 0.95 时最优密度为每百字 2.16 个、上界 5.10 个,覆盖率不高于 0.50 时完全不该嵌。嵌词携带的语域信号可以用互信息量,在合成表里它只消除了语域不确定性的一小部分。按单条消息的可靠性要求,允许的密度比平均最优低得多,一对一私讯应取两者中较小的一个。 (https://5dlibrary.com/answers/zh/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c05/) - Ch.6 Monolingual Constructional Equivalence: Precision Without Insertion — The third popular belief holds that a writer who must stay in Chinese or in English can only choose between a loss of precision and translationese. The non-equivalence strategies of translation studies, dynamic equivalence and the oblique procedures, together with the argument of construction grammar that constructions carry meaning of their own, give a third route: recombination at the level of the phrase. Once precision is written as feature coverage, all three gap concepts in the illustrative examples reach a coverage of 0.80 with monolingual constructions of four to twelve units. Exhaustive search finds skeletons shorter than greedy or hand-written candidates, and rebuilt versions can bring the translationese markers down to zero. The price is length, and what it buys is a text everyone understands. (https://5dlibrary.com/answers/en/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c06/) 第6章 单语构式等效:不嵌词也能精准 — 第三条流行说法认为,写纯中文或纯英文时只能在精度损失与翻译腔之间二选一。翻译学的非对等策略、动态对等与间接程序,加上构式语法关于构式自身携带意义的论证,给出第三条路:在短语层面重组。把精准写成特征覆盖率之后,示意例子里三个空缺概念都能用 4 到 12 个单位的单语构式达到 0.80 的覆盖率;穷举搜索给出的骨架比贪心与手写候选更短;重组版的翻译腔标记可以降到零。代价是长度,换来的是人人都懂。 (https://5dlibrary.com/answers/zh/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c06/) - Ch.7 Classical Sources: Yan Fu's Three Difficulties, Confucius on Names, Jerome's Sense for Sense and Schleiermacher's Two Paths — Of the seven renderings, four are Pareto-optimal, in the sense that no other rendering beats them on every score: the literal rendering, the borrowed-sense compound, the full-coverage construction and plain insertion. The word-for-word rendering wins on nothing. Its xin is 0.800, its da only 0.587 and its ya only 0.40, and it is beaten on every score by the construction with a moved pragmatic marker. Plain insertion has a da of 0.767, but its xin is only 0.820, because more than one reader in ten does not know the word. Under the Yan Fu rule, among renderings whose da clears 0.6, the one with the highest xin is the full-coverage construction hyper+, with xin of 1.000, da of 0.670 and ya of 0.95. An equally weighted average of the three scores picks the same rendering, with a mean of 0.873. (https://5dlibrary.com/answers/en/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c07/) 第7章 原典考辨:严复三难、孔子正名、哲罗姆意译与施莱尔马赫两条路 — 七种译法里,帕累托意义上不被别人全面压过的有四种:直译、借义复合、全覆盖构式与直接嵌入。逐词硬译(word-for-word)在三项上都不占优:「信」0.800,「达」只有 0.587,「雅」只有 0.40,被语用标记移位的构式全面压过。直接嵌入的「达」是 0.767,「信」却只有 0.820,因为一成多读者不认得这个词。按严复规则,「达」过了 0.6 线的译法里,「信」最高的是全覆盖构式 hyper+,「信」1.000、「达」0.670、「雅」0.95。改用三项等权平均,结果仍是它,平均分 0.873。 (https://5dlibrary.com/answers/zh/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c07/) - Ch.8 Field Tools and Checker Specifications: the Insertion Ledger, Four-Kit Checkers and a Pre-registered Test — Chapter 8 gives the insertion ledger, checker specifications for four kits, the distilled Code-Switching Net-Gain and Monolingual-Equivalence Calculator and a pre-registered test. Parameters are illustrative and all calculations are generated by the book's programs; nothing corresponds to a real customer, conversation or sale, and Malay, Hokkien, Cantonese, English and Mandarin are treated as equals. This is not linguistic doctrine and not business advice to any individual or company. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c08/) 第8章 实战工具与检查器规格:嵌词台账、四件套检查器与预注册检验 — 第八章给出嵌词台账、四件套检查器规格、蒸馏小工具「多语嵌词净收益与单语构式等效试算器」与预注册检验。书中参数为示意,全部计算由书中程序生成,不对应任何真实客户、对话或成交;马来语、闽南语、粤语、英语、华语一律平等对待。本书不是语言学定论,也不构成针对个人或企业的经营建议。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/code-switching-information-theory-lexical-gap-compensation-monolingual-precision/c08/) ## 5DL-B-382 · Classical Chinese Etymological Philology, Paratactic Imagery Topology and East-West Aesthetic Asymmetry in High-Authority Writing 中文书名:古汉语字源本义训诂、意合留白拓扑与中西意境张力转化母典 DOI: pending Cite: Fan, Y. (2026). Classical Chinese Etymological Philology, Paratactic Imagery Topology and East-West Aesthetic Asymmetry in High-Authority Writing (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Anachronistic reading yields an illustrative 1.307 expected misreadings per thousand graphs, and with only three faulty sentences a checker must read 0.632 of a manuscript. Parameters are illustrative and all calculations come from the book's 23 programs. Abstract: Three popular sentences dominate talk about Classical Chinese: characters are pictographs, one picture per word; Classical Chinese has no grammar; the mood of a Chinese text is necessarily lost in English. This book refutes the first two and rewrites the third as a constraint that can be computed. Chinese characters form a morphosyllabic script that records a language, with semantic-phonetic graphs as its main body; in the opening passage of the Analects, graphs whose glossed original sense is detached from their use in the line account for 0.400 of thirty tokens, and the commonest graphs are the least pictorial. Classical grammar lives in word order, function words and parallelism, and the gap in function-word density between prose and verse is one of genre. The published counts of the Penn Discourse TreeBank show that 0.5453 of relations in English newswire are unmarked; the rough figure for Chinese newswire is 0.82, a difference of degree. Corrections to the blueprint: widely repeated six-category counts could not be verified in the original works, so no percentage is cited; the Chinese implicit share comes from modern newswire and cannot be transferred to Classical Chinese; Humboldt is quoted from the manuscript edition and his ranking of languages is rejected; cultural comparison serves only as a weak group-level prior, with variation within groups exceeding variation between them; disputed Shuowen accounts of graphs such as dong, wei, wang and bu are given with all sides. Chapters 3 to 6 draw sentences as graphs, define the gap ratio and the ambiguity entropy of each juxtaposition edge (8.273 bits for the Analects passage), and search all English plans under a traceability of at least 0.90: Wen Tingyun's couplet keeps at most 0.314 of its gap and the Analects passage 0.371, and under perturbation the optimum holds the line with probability only 0.598. Anachronistic reading yields an illustrative 1.307 expected misreadings per thousand graphs, and with only three faulty sentences a checker must read 0.632 of a manuscript. Chapter 7 examines the Great Commentary, Wang Bi, Xu Shen, Aristotle and Humboldt. Chapter 8 gives a two-track ledger, field tools and checker specifications for the distilled Parataxis Gap, Ambiguity Entropy and Bilingual Traceability Calculator, and pre-registered tests needing 1,456 readers. Parameters are illustrative and all calculations come from the book's 23 programs. This is not a ruling in philology and not legal or business advice; inscription samples are public-domain or synthetic and concern no real family. 摘要:谈古汉语与中西表达,流行说法有三句:汉字是象形文字,一字一图;古汉语没有语法;中文意境译成英文必然丢失。本书证伪前两句,把第三句改写成一条可以计算的约束。汉字是记录语言的语素音节文字,形声字是主体;《论语》开篇一章三十个字次里,本义与句中用法脱节的借用字形占 0.400,越常用的字越不像画。古汉语的语法在语序、虚词与对偶里,散文与诗曲的语法词密度差别来自文体。宾州篇章树库的公开计数显示,英文新闻也有 0.5453 的篇章关系没有连接词,汉语新闻语料的粗略数字是 0.82,差异是程度上的。 蓝图修正:流传的六书分类计数未能在原书核实,不引具体百分比;汉语隐性关系比例来自现代新闻语料,不能直接套到古汉语;洪堡书信引自手稿整理本,其语言高下框架不予采纳;文化比较只作群体层面弱先验,组内差异大于组间差;《说文》对「东、为、王、不」等字的说解有争议,书中一律并陈各方出处。 第三至六章把句子画成图,定义留白比例与每条并置边的歧义熵(《论语》开篇一章合计 8.273 比特),在英文可追溯率不低于 0.90 的约束下穷举最优方案:温庭筠一联最多保留 0.314 的留白,《论语》开篇一章保留 0.371;参数扰动下最优方案守住红线的概率只有 0.598。以今律古按示意参数每千字有 1.307 次期望误读,错句只有三句时抽检须读全稿的 0.632。第七章考辨《周易·系辞》、王弼、许慎与亚里士多德、洪堡。第八章给出双轨意境转换台账、蒸馏小工具「意合留白歧义熵与中英双轨可追溯率试算器」的实战工具与检查器规格,以及需要 1,456 位读者的预注册检验。书中参数为示意,全部计算由书中程序生成(23 段程序)。本书不是文字学或训诂学定论,不构成法律或经营建议;碑铭样本均为公有领域或合成文本,不涉及任何真实家族。 Chapter answers: - Ch.1 Graphs Are Not Pictures: the Shuowen's Six Categories, Phonetic Loans and a Morphosyllabic Script — Chinese graphs are not one picture per word. Xu Shen spends 14.27 characters of language on each head graph. The semantic-phonetic design lets a few hundred components spell close to ten thousand morphemes. In the opening passage of the Analects, 0.400 of the tokens are written with graphs whose glossed original sense is detached from their use. The pictorial portrait captures a small pictographic layer and misses both the body of the writing system and the skeleton of its grammar. (https://5dlibrary.com/answers/en/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c01/) 第1章 字不是画:《说文》六书、假借与语素音节文字 — 汉字不是一字一图。许慎用 14.27 个字的语言解说一个字头;形声结构让几百个部件就能拼写近万个语素;《论语》开篇一章里,0.400 的字次出自本义与用法脱节的借用字形。画像论只抓住了少数象形字,却漏掉了文字系统的主体与语法的骨架。 (https://5dlibrary.com/answers/zh/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c01/) - Ch.2 Classical Chinese Has a Grammar: Word Order, Function Words and the Hard Constraint of Parallelism — The texts fall into two sharply distinct groups. In the two prose passages function words are dense: 46.7 per hundred graphs in the Analects, 40.0 in the Mencius, and the predicate-linking connectives alone reach 13.3 per hundred in the Mencius line. In the two verse texts function words all but vanish: none in Wen Tingyun's couplet, and a single locative verb among the twenty-eight graphs of the song, 3.6 per hundred. (https://5dlibrary.com/answers/en/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c02/) 第2章 古汉语有语法:语序、虚词与对偶的硬约束 — 结果分成截然不同的两组。两段散文里,语法词极其密集:《论语》每百字 46.7 个,《孟子》每百字 40.0 个;连接谓语的连词,《孟子》那句话里每百字有 13.3 个。两段诗曲里,语法词几乎消失:温庭筠那一联一个都没有,《天净沙》二十八个字里只有一个「在」,每百字 3.6 个。 (https://5dlibrary.com/answers/zh/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c02/) - Ch.3 The Hypotactic Tree and the Paratactic Network: the Gap Ratio and the Graph of Juxtaposition — Chapters 3 to 6 draw sentences as graphs, define the gap ratio and the ambiguity entropy of each juxtaposition edge (8.273 bits for the Analects passage), and search all English plans under a traceability of at least 0.90: Wen Tingyun's couplet keeps at most 0.314 of its gap and the Analects passage 0.371, and under perturbation the optimum holds the line with probability only 0.598. Anachronistic reading yields an illustrative 1.307 expected misreadings per thousand graphs, and with only three faulty sentences a checker must read 0.632 of a manuscript. (https://5dlibrary.com/answers/en/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c03/) 第3章 形合树与意合网:留白比例与并置边的图结构 — 第三至六章把句子画成图,定义留白比例与每条并置边的歧义熵(《论语》开篇一章合计 8.273 比特),在英文可追溯率不低于 0.90 的约束下穷举最优方案:温庭筠一联最多保留 0.314 的留白,《论语》开篇一章保留 0.371;参数扰动下最优方案守住红线的概率只有 0.598。以今律古按示意参数每千字有 1.307 次期望误读,错句只有三句时抽检须读全稿的 0.632。 (https://5dlibrary.com/answers/zh/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c03/) - Ch.4 Ambiguity Entropy: the Distribution of Readings on Each Juxtaposition Edge and the Burden on Readers — Once the distribution of readings is counted, the ambiguity of a paratactic sentence acquires a value. The opening passage of the Analects carries 8.273 bits and Wen Tingyun's couplet 7.806 bits, 1.182 and 1.115 bits per edge respectively. The probability that two readers agree on the whole sentence is below one in a hundred, and a pair disagrees on more than three edges on average. The most valuable place for an added function word is the edge of highest entropy; by the third added word only 0.571 of the gap remains. (https://5dlibrary.com/answers/en/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c04/) 第4章 歧义熵:每条并置边上的读法分布与读者负担 — 把读法的分布算进来,意合句的歧义就有了数值。《论语》开篇一章合计 8.273 比特,温庭筠一联 7.806 比特,平均每条边分别是 1.182 与 1.115 比特。两位读者全句读成一样的概率不到百分之一,平均在三条多的边上分歧。加一个虚词最值钱的位置,是熵最高的那条边;加到第三个,留白只剩 0.571。 (https://5dlibrary.com/answers/zh/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c04/) - Ch.5 The Two-Track Transfer: the Largest Gap That Survives an English Traceability Constraint — Between spelling out everything and juxtaposing everything lies a frontier that can be computed. Under a traceability of at least 0.90, Wen Tingyun's couplet keeps at most 0.314 of its gap and the opening of the Analects 0.371. The clause effect s sets the position of the frontier; at an s of 0.7 the gap kept reaches 0.514. The optimum sits right on the red line and holds it with a probability of only 0.598 under perturbation, so plans need a margin. (https://5dlibrary.com/answers/en/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c05/) 第5章 双轨转换:英文可追溯率约束下的最大留白 — 全部写明与全部并置之间,存在一条可以计算的前沿。在可追溯率不低于 0.90 的约束下,温庭筠一联最多保留 0.314 的留白,《论语》开篇一章保留 0.371。分句的效果 s 决定前沿的位置,s 取 0.7 时留白可达 0.514。最优方案正好压在红线上,扰动之下只有 0.598 的概率守住红线,规划时要留余量。 (https://5dlibrary.com/answers/zh/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c05/) - Ch.6 Reading the Past with Today's Senses: a List of Misreadable Words, the Cost of Sampling and the Form of Inscriptions — Reading the past with today's senses is an invisible error. Under the illustrative parameters a thousand graphs of classical prose yield 1.307 expected misreadings, and the three riskiest graphs carry 0.489 of the risk. The cost of sampling rises as errors become fewer: with only three faulty sentences, 0.95 confidence requires checking 0.632 of the manuscript. The form of inscriptions can be written as four rules, and the fourth couplet of the synthetic text is marked for revision because its structure does not correspond. (https://5dlibrary.com/answers/en/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c06/) 第6章 以今律古:易误读字清单、抽检成本与碑铭体式 — 以今律古是一种看不见的错误。按示意参数,每千字古文里有 1.307 次期望误读,前三个高风险字承担 0.489 的风险。抽检的成本与错误数目成反比:错句只有 3 句时,要有 0.95 的把握需抽查全稿的 0.632。碑铭体式可以写成四条规则,合成文本的第四联因结构不对应被标为需修改。 (https://5dlibrary.com/answers/zh/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c06/) - Ch.7 Classical Sources: the Great Commentary and Wang Bi on Words, Images and Intent; Humboldt and Aristotle on Grammar and Signs — The entropy of the intent is 2.0000 bits. The image retains 1.1524 bits and the words only 0.5511 bits. The information reaching the words is 0.276 of the intent; the information reaching the image is 0.576. That is the quantitative form of "writing does not exhaust speech, speech does not exhaust intent": along the chain information can only decrease, never increase, a result known in information theory as the data processing inequality (Shannon 1948). (https://5dlibrary.com/answers/en/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c07/) 第7章 原典考辨:《系辞》与王弼论言象意,洪堡与亚里士多德论语法和符号 — 意的熵是 2.0000 比特。象保留了 1.1524 比特,言只保留了 0.5511 比特。抵达言的信息只占意的 0.276,抵达象的占 0.576。这正是「书不尽言,言不尽意」的定量形式:每经过一环,信息只会减少,不会增加,这在信息论里叫数据处理不等式 (Shannon 1948)。 (https://5dlibrary.com/answers/zh/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c07/) - Ch.8 Closing the Loop: a Two-Track Ledger for Carrying Mood across Languages, Field Tools and Checker Specifications — Chapter 8 gives a two-track ledger, field tools and checker specifications for the distilled Parataxis Gap, Ambiguity Entropy and Bilingual Traceability Calculator, and pre-registered tests needing 1,456 readers. Parameters are illustrative and all calculations come from the book's 23 programs. This is not a ruling in philology and not legal or business advice; inscription samples are public-domain or synthetic and concern no real family. (https://5dlibrary.com/answers/en/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c08/) 第8章 实证闭环:双轨意境转换台账、实战工具与检查器规格 — 第八章给出双轨意境转换台账、蒸馏小工具「意合留白歧义熵与中英双轨可追溯率试算器」的实战工具与检查器规格,以及需要 1,456 位读者的预注册检验。书中参数为示意,全部计算由书中程序生成(23 段程序)。本书不是文字学或训诂学定论,不构成法律或经营建议;碑铭样本均为公有领域或合成文本,不涉及任何真实家族。 (https://5dlibrary.com/answers/zh/classical-chinese-etymology-paratactic-imagery-topology-east-west-asymmetry/c08/) ## 5DL-B-383 · Human-AI Joint Asset and IP Protection, Automated Contract-Loophole Auditing and Full-Spectrum Anti-Erosion Architecture 中文书名:人机联防资产与知识产权护城河、契约暗坑机检与全方位防侵蚀母典 DOI: pending Cite: Fan, Y. (2026). Human-AI Joint Asset and IP Protection, Automated Contract-Loophole Auditing and Full-Spectrum Anti-Erosion Architecture (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: But machine screening is a recall-first screen: on the book's synthetic labelled set only 0.380 of reworded risky clauses are caught, and the costlier a miss, the more a lawyer must read in full. Parameters are illustrative and all calculations are generated by the book's programs. Abstract: Three popular sentences guide small owners: a signed contract is safe; once a work is copied there is no way to trace it; a complicated password keeps an account safe. This book refutes or bounds each one. Automated contract screening is well supported: legal experts marked 41 categories of review points across 510 commercial contracts, and potentially unfair terms of service can be flagged by machine. But machine screening is a recall-first screen: on the book's synthetic labelled set only 0.380 of reworded risky clauses are caught, and the costlier a miss, the more a lawyer must read in full. Hidden fingerprints can trace a leak, within physical bounds: Tardos code length grows with the square of the number of colluders, and with the original constant fifty copies against two colluders already need 4,400 marked positions; once more than eight tenths of a text is reworded, a fingerprint of 400 positions traces its source in only 0.28 of leaks. The real risk to accounts lies in recovery channels and concentrated control: a recovery mailbox protected only by a password drags a two-factor account back to a yearly compromise probability of 0.1065, and a single mailbox can carry 0.829 of the risk to a domain. Checking the sources corrected the blueprint: the journal version of Tardos's code dates from 2008 and its constant of 100 is a cautious upper bound; the current digital identity guideline is the fourth revision, finalised in July 2025; Cubic Electronics was decided in 2018 and reported in 2019; and Malaysia and the United States set opposite defaults for copyright in commissioned works. Chapters 1 to 6 give the eight-dimension scan matrix, the expected-loss gap of one-way caps (changing only how a cap is written moves expected net value by 14.47), the quasi-rent bound of hold-up (below a salvage fraction of 0.40 a good project finds no investor), fingerprint length and the decay of rewording, the control-path graph of sovereign separation, and access red lines for outside partners (at technical access of 0.6 the customer red line is 0.241) and staged delivery with escrow (largest loss 17.50). Chapter 7 examines the Sunzi, the Han Feizi, Hobbes and the Vienna Convention; Chapter 8 gives a weekly ledger, the distilled Contract Pitfall Scanner and Asset Sovereignty Compromise Calculator, four-kit checker specifications and pre-registered tests. Parameters are illustrative and all calculations are generated by the book's programs. This is not legal advice and not a security audit; statutes are labelled by jurisdiction and are subject to amendment, and any real contract must be reviewed by a practising lawyer. The book is defensive only, names no real account, key or server, uses fingerprints only on one's own works, and uses only synthetic clauses. Boxes separate evidence, runnable Python (22 programs), classical sources and hypotheses. 摘要:小生意谈资产防护,最流行的三句话是:合同签了就安全;作品发出去被抄了无从追查;密码设得够复杂账户就守住了。本书逐条证伪或立界。合同机检确有扎实证据:法律专家在 510 份商业合同里标注了 41 类审阅要点,在线服务条款里的八类潜在不公平条款可被机器标出;但机检是召回优先的初筛,在本书的合成标注集上,改写过的风险条款只有 0.380 被抓到,漏报代价越高,律师越要通读全文。隐式指纹可以追溯,但有物理上界:塔尔多什码长随合谋人数按平方增长,原始常数下五十份副本、两人合谋就需 4,400 个标记位;改写超过八成时,400 位指纹的溯源成功率只剩 0.28。真正的账户风险在恢复通道与权限集中:恢复通道若是只有口令的邮箱,两因子账户的年失陷概率也会被拉回 0.1065;一个邮箱可以贡献域名风险的 0.829。 核对原文之后本书做了蓝图修正:塔尔多什期刊版为 2008 年,常数 100 是保守上界;数字身份指南现行为 2025 年 7 月定稿的第四版;Cubic Electronics 一案 2018 年宣判、2019 年见于判例汇编;《孙子·九变》原文作「恃吾有以待之」;马来西亚与美国对委托作品版权的默认归属方向相反。第一至六章依次给出八维合同扫描矩阵、单向上限的期望损失差(只改上限写法,期望净值相差 14.47)、敲竹杠的准租金上界(残值比例低于 0.40 时好项目无人投入)、指纹码长与改写衰减、主权隔离的控制路径图、外部合作的接触比例红线(技术接触 0.6 时客户接触红线 0.241)与分段托管(最大损失压到 17.50)。第七章考辨《孙子兵法·九变》《韩非子·二柄》、霍布斯《利维坦》与《维也纳条约法公约》第 26 条,并给出四件套蒸馏规格。第八章给出周度台账、小工具「合同暗坑扫描与资产主权隔离失陷概率试算器」、四件套检查器规格与预注册检验。 书中参数为示意,全部计算由书中程序生成。本书不是法律意见,也不是安全审计;法条注明法域且会修订,任何真实合同须由执业律师复核。本书只写防御,不写任何攻击、入侵、破解或社会工程做法,不写任何真实账户、口令、密钥或服务器;指纹只用于自家作品溯源;合同样例全部合成。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Contract-Pitfall Scan Matrix: Eight Risk Dimensions and a Recall-First Screen — Chapters 1 to 6 give the eight-dimension scan matrix, the expected-loss gap of one-way caps (changing only how a cap is written moves expected net value by 14.47), the quasi-rent bound of hold-up (below a salvage fraction of 0.40 a good project finds no investor), fingerprint length and the decay of rewording, the control-path graph of sovereign separation, and access red lines for outside partners (at technical access of 0.6 the customer red line is 0.241) and staged delivery with escrow (largest loss 17.50). (https://5dlibrary.com/answers/en/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c01/) 第1章 合同暗坑扫描矩阵:八个风险维度与召回优先的初筛 — 第一至六章依次给出八维合同扫描矩阵、单向上限的期望损失差(只改上限写法,期望净值相差 14.47)、敲竹杠的准租金上界(残值比例低于 0.40 时好项目无人投入)、指纹码长与改写衰减、主权隔离的控制路径图、外部合作的接触比例红线(技术接触 0.6 时客户接触红线 0.241)与分段托管(最大损失压到 17.50)。 (https://5dlibrary.com/answers/zh/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c01/) - Ch.2 Lopsided Accounts: One-Way Liability Caps, Stipulated Sums and Automatic Renewal — The cost of a one-way liability cap sits in the tail. Under one fixed set of risks, changing only how the cap is written swings expected net value from -2.63 to 11.84, and the net payment in the worst year in a thousand exceeds the annual fee more than elevenfold. A stipulated sum is a ceiling and not a floor: writing five times the mean loss recovers very little more in expectation and must still survive scrutiny for reasonableness. The risk of automatic renewal comes mainly from management; two independent reminders and a named person cut the expected time bound to a bad partner over six years from 1.395 years to 0.908 years. (https://5dlibrary.com/answers/en/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c02/) 第2章 不对等的账:单向赔偿上限、违约金与自动续约 — 单向赔偿上限的代价藏在尾部:同一组风险下,只改上限写法,期望净值从 -2.63 摆到 11.84,千分之一坏年份的净付超过年费十一倍。违约金是上限而非保底,写到五倍均值也只能多拿回很少,且要经受合理性审查。自动续约的风险主要来自管理,两道独立提醒加专人负责,就能把六年绑定从 1.395 年压到 0.908 年。 (https://5dlibrary.com/answers/zh/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c02/) - Ch.3 Specific Investment and Hold-Up: Ownership of Intellectual Property as the Foundation of the Moat — Specific investment creates quasi-rent, and quasi-rent invites hold-up. Under the illustrative parameters of p31, once the salvage fraction falls below 0.40, a good project finds no investor because hold-up is expected. Ownership is an outside option: p32 shows that when the intellectual property sits with the other side and we hold no licence, investment is only a quarter of the first best and the loss of surplus is 0.250. The default ownership of commissioned works points in opposite directions in Malaysia and the United States, and three of the eight synthetic cases in p33 are red, one of them only because the words "work made for hire" were applied to software, which lies outside the nine kinds. (https://5dlibrary.com/answers/en/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c03/) 第3章 专用投入与敲竹杠:知识产权归属是护城河的地基 — 专用投入产生准租金,准租金招来敲竹杠。在 p31 的示意参数下,残值比例低于 0.40,好项目就会因预期被敲竹杠而无人投入。所有权是外部选项:p32 显示,知识产权在对方手里且无许可时,投入只有第一优解的四分之一,剩余损失 0.250。默认归属在马来西亚与美国方向相反,p33 的八个合成情形里有三个判红,其中一个只因「雇佣作品」字样用在了九类之外的软件上。 (https://5dlibrary.com/answers/zh/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c03/) - Ch.4 Hidden Fingerprints: Floors on Code Length and the Decay of Rewording — "Once a work is out, there is no way to trace it" is wrong: a marked copy carries the trace of its source. "Everything can be traced" is wrong too. Tardos code length grows with the square of the number of colluders; with the original constant, none of the twelve combinations fits inside a book's budget of 700 positions; and once rewording passes 0.8 of the text, a fingerprint of 400 positions traces its source in only 0.28 of leaks. A fingerprint is a tool with physical bounds, and the rules for using it must be written in terms of batches, positions and the strength of rewording. (https://5dlibrary.com/answers/en/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c04/) 第4章 隐式指纹:码长下界与改写衰减 — 「作品发出去就无从追查」是错的:做过标记的副本带着来源的痕迹。「百分之百可追溯」也是错的:塔尔多什码长随合谋人数按平方增长,原始常数下十二种组合全都装不进一本书 700 位的预算;改写比例超过 0.8 时,400 位标记的溯源成功率跌到 0.28。指纹是有物理上界的工具,规则要按批次、按位数、按改写强度来写。 (https://5dlibrary.com/answers/zh/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c04/) - Ch.5 Sovereign Separation: Single Points of Compromise, Correlated Failure and Threshold Control — A complicated password cannot protect the recovery channel or prevent concentrated control. Current guidance asks for long passwords, no imposed composition rules, no change on a schedule, and comparison with a blocklist. The multiplicative advantage of multiple factors is eroded by correlated failure: at beta of 1.00, the chance of compromise returns to the weaker factor's 0.0500. If recovery goes through a mailbox guarded only by a password, total compromise reaches 0.1065. Threshold control relies on independence: (2,3) beats a single holder, but two pieces on one device collapse back to a single point. The control-path graph shows the mailbox carrying 0.829 of the risk to the domain, and hardening three classes while leaving one improves the whole by only 1.6 times. (https://5dlibrary.com/answers/en/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c05/) 第5章 主权隔离:单点失陷、相关失效与门限分权 — 复杂密码守不住恢复通道与权限集中。现行指南要求口令长、不强加组合规则、不定期更换、与黑名单比对。多因子的乘法优势会被相关失效抵消:贝塔为 1.00 时,失守概率退回较弱因子本身的 0.0500。恢复通道若是只有口令的邮箱,总失陷 0.1065。门限分权要靠独立:(2,3) 优于一人独握,两份同处一台设备就退回单点。控制路径图显示,邮箱贡献了域名风险的 0.829;加固三类留下一类,整体只改善到 1.6 倍。 (https://5dlibrary.com/answers/zh/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c05/) - Ch.6 Guarding Against Hollowing-Out by Outside Partners: Access Ratios, Staged Delivery and Minimal Disclosure — How much a departing partner can take depends first on its access ratio. Under the illustrative parameters of p61, once technical access reaches 0.6, customer access above 0.241 turns the cell red, and a contractual sanction with an expected value of only 30.0 cannot hold back a partner with high access. Staged delivery handles "paid but not delivered", source-code escrow handles "delivered but impossible to take over", and twelve stages with escrow cut the largest loss from 100.00 to 17.50. Minimal necessary disclosure shows only 0.55 when there is no protection and can safely show 0.91 when protection is in place. (https://5dlibrary.com/answers/en/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c06/) 第6章 外部合作防掏空:可接触比例、分段交付与最小披露 — 合作方离开时能带走多少,首先取决于接触比例:在 p61 的示意参数下,技术接触到 0.6 时,客户接触超过 0.241 就判红;合同的期望制裁只有 30.0,挡不住高接触比例的合作方。分段交付管「已付未交」,源代码托管管「已交难接」,十二段加托管把最大损失从 100.00 压到 17.50。最小必要披露在没有保护时只展示 0.55,有保护时可放心展示到 0.91。 (https://5dlibrary.com/answers/zh/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c06/) - Ch.7 Classical Sources: the Nine Variations of the Sunzi, the Two Handles of the Han Feizi and Hobbes's Leviathan — Chapter 7 examines the Sunzi, the Han Feizi, Hobbes and the Vienna Convention; (https://5dlibrary.com/answers/en/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c07/) 第7章 原典考辨:《孙子·九变》《韩非子·二柄》与霍布斯《利维坦》 — 第七章考辨《孙子兵法·九变》《韩非子·二柄》、霍布斯《利维坦》与《维也纳条约法公约》第 26 条,并给出四件套蒸馏规格。 (https://5dlibrary.com/answers/zh/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c07/) - Ch.8 The Empirical Loop: a Weekly Anti-Erosion Ledger and Field Tools and Checker Specifications — The ledger turns the book into four tables: contracts, assets, distribution and partners. The tool uses the same formulas as the book's programs to give every row a band. The checker specification is grouped by the four kits, and each provision carries a threshold, a trigger action and a source. The pre-registered tests give their sample sizes: 55 risky clauses, 26 rehearsals, 21 simulated leaks and 49 ended partnerships in each group. (https://5dlibrary.com/answers/en/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c08/) 第8章 实证闭环:资产防侵蚀周度台账与实战工具与检查器规格 — 台账把全书落成四张表:合同、资产、分发、合作。小工具用与书中程序相同的公式,把每一行变成红黄绿。检查器规格按四件套分组,每条都有阈值、触发动作与出处。预注册检验给出样本量:55 条风险条款、26 次演练、21 份模拟外流、每组 49 个合作案例。 (https://5dlibrary.com/answers/zh/human-ai-asset-ip-moat-contract-loophole-audit-anti-erosion/c08/) ## 5DL-B-384 · When the Shortest Path Fails: Brachistochrone Detours, Second-Order Consequence Mapping and Holographic Deep Thinking 中文书名:非最短路径最优解、二阶连锁效应全息推演与周密思考复利母典 DOI: pending Cite: Fan, Y. (2026). When the Shortest Path Fails: Brachistochrone Detours, Second-Order Consequence Mapping and Holographic Deep Thinking (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. This is not investment advice or advice on any particular business decision, and it promises no return. Abstract: Three popular sayings treat the shortest path as the best one: a straight line is the fastest route between two points; choosing the best step every time adds up to the best result; and thinking too much is procrastination, so ship first. This book refutes or bounds each in turn. In a gravitational field the quickest route is the cycloid, not the line: in the square case the cycloid needs only 0.9128 of the line's time. This is the strict result of Bernoulli's challenge of 1696; every business analogy in the book is structural only and states when it breaks. After checking the sources the book corrects its blueprint: the structures on which greed is optimal for every weighting are exactly the matroids, though greed can still happen to be optimal on a non-matroid instance; with linear delays a new shortcut can slow selfish routing to at most 4/3 of the optimum; the claim that fixing costs grow geometrically by phase is contradicted by Menzies and colleagues in 2017, so the interest rate on rework must be measured; and the optimal depth of deliberation is the author's hypothesis, stated with the conditions that would refute it. Parameters are illustrative and all calculations are generated by the book's programs. This is not investment advice or advice on any particular business decision, and it promises no return. Chapter 1 compares the line, the circular arc and the cycloid using conservation of energy and numerical integration, and marks where the advantage vanishes. Chapter 2 turns building the road first into a schedule model: with illustrative parameters the best preparation takes 20.6% of the schedule, saves 39.2% and breaks even at 59.94. Chapter 3 sets greed against dynamic programming on matroids and beyond; in the knapsack trap greed obtains 0.0002 of the optimum. Chapter 4 computes Braess's paradox, in which a free shortcut raises the equilibrium time from 65.00 to 80.00 minutes, and shows it appears only in a band of demand. Chapter 5 writes the second-order rework of a first-order shortcut as a liability with an implied interest rate ranging from -51.5% to +26.8%; in a year-long simulation the careful lane spends 6.4% of its hours on rework against 22.6% for the fast lane. Chapter 6 builds consequence trees and finds the optimal depth of deliberation; leaving out opposing evidence makes a judgement of 0.643 look like 0.900. Chapter 7 examines the Sunzi on manoeuvring, the twenty-second chapter of the Laozi and Bernoulli's challenge. Chapter 8 gives the holographic decision ledger, the distilled Detour Break-even and Second-Order Rework Liability Calculator, checker specifications for the four kits and a pre-registered test whose primary comparison needs 198 decisions per arm. "Holographic" means only multi-dimensional coverage, with no mystical sense; deliberation budgets include rest. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. 摘要:关于「最短即最优」有三句流行说法:两点之间直线最快;每一步都挑眼前最优,加起来就是全局最优;想太多就是拖延,先上线再说。本书逐条证伪或立界。重力场里最快的路线是摆线而非直线,边长相等时摆线用时只有直线的 0.9128;这是伯努利 1696 年征题的严格结论,书中的商业类比只作结构类比,每次都写明失效条件。核对原文之后,本书做了蓝图修正:贪心对任意权重都最优的结构恰是拟阵,非拟阵实例上贪心仍可能碰巧最优;线性延迟下,加一条捷径最多让自私路由慢到最优的 4/3 倍;修复成本随阶段几何增长之说有门齐斯等人 2017 年的反证,返工利率只能实测;「推演深度有最优值」是作者假说,书中给出推翻条件。书中参数为示意,全部计算由书中程序生成。本书不是投资意见,也不是任何具体经营决策意见,不承诺任何收益。 第一章用能量守恒与数值积分比较直线、圆弧与摆线,并标出目标在正下方与摩擦过大时优势消失。第二章把先修路再赶路写成工期模型,示意参数下最优铺垫占总工期 20.6%,省 39.2%,盈亏平衡铺垫为 59.94。第三章对照拟阵与非拟阵上的贪心,背包陷阱里贪心只拿到最优的 0.0002。第四章算布雷斯悖论,加一条免费捷径使均衡用时从 65.00 分钟升到 80.00 分钟,并标出悖论只在一段需求带里出现。 第五章把一阶捷径的二阶返工写成负债现值与隐含利率,利率随暴露速度与成本增长从 -51.5% 变到 +26.8%;一年模拟里慢车道的返工工时占比 6.4%,快车道 22.6%。第六章搭后果树,求推演深度的最优值,漏记反方证据会把 0.643 的判断看成 0.900。第七章考辨《孙子·军争》的「以迂为直」、《老子》第二十二章的「曲则全,枉则直」与伯努利征题。第八章给出全息推演决策台账、蒸馏小工具「迂回铺垫盈亏平衡与二阶返工负债推演器」、四件套检查器规格与预注册检验,主检验每组需要 198 项决策。书中「全息」只指多维全覆盖,不含任何玄学含义;深度推演的时间预算包含休息,不以透支换产出。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Brachistochrone: Why the Straight Line Loses to the Cycloid — In a gravitational field the quickest route is the cycloid, not the straight line. Program p11 finds that in the square case the cycloid takes 0.9128 of the line's time, and the advantage grows with horizontal distance. Program p12 shows that dropping first and travelling second has a best depth, and that dropping too far is slower than the line. Program p13 marks two boundaries: the advantage vanishes when the target is directly below, and the cycloid stalls halfway once friction passes 0.793. Business analogies borrow the structure, never the numbers, and each use has to check whether potential can be converted, whether speed persists, and whether the target is far enough away. (https://5dlibrary.com/answers/en/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c01/) 第1章 最速降线:直线为什么输给摆线 — 重力场里,最快的路线是摆线,而不是直线。程序 p11 在边长相等时算出摆线用时是直线的 0.9128,水平距离越长优势越大。程序 p12 显示「先降再走」有最佳深度,降过头反而比直线慢。程序 p13 标出两条边界:目标在正下方时优势消失,摩擦越过 0.793 时摆线停在半路。商业类比只借结构,不借数字,每次使用都要核对势能能否转换、速度能否保持、目标是否够远。 (https://5dlibrary.com/answers/zh/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c01/) - Ch.2 Roundabout Preparation: the Break-even of Building the Road First — Chapter 2 turns building the road first into a schedule model: with illustrative parameters the best preparation takes 20.6% of the schedule, saves 39.2% and breaks even at 59.94. (https://5dlibrary.com/answers/en/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c02/) 第2章 迂回铺垫:先修路再赶路的盈亏平衡 — 第二章把先修路再赶路写成工期模型,示意参数下最优铺垫占总工期 20.6%,省 39.2%,盈亏平衡铺垫为 59.94。 (https://5dlibrary.com/answers/zh/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c02/) - Ch.3 The Limits of Greed: Inside and Outside the Matroid — Chapter 3 sets greed against dynamic programming on matroids and beyond; in the knapsack trap greed obtains 0.0002 of the optimum. (https://5dlibrary.com/answers/en/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c03/) 第3章 贪心的边界:拟阵之内与拟阵之外 — 第三章对照拟阵与非拟阵上的贪心,背包陷阱里贪心只拿到最优的 0.0002。 (https://5dlibrary.com/answers/zh/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c03/) - Ch.4 One More Road and Everyone Is Slower: Braess's Paradox and Selfish Routing — Chapter 4 computes Braess's paradox, in which a free shortcut raises the equilibrium time from 65.00 to 80.00 minutes, and shows it appears only in a band of demand. (https://5dlibrary.com/answers/en/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c04/) 第4章 加一条路全体变慢:布雷斯悖论与自私路由 — 第四章算布雷斯悖论,加一条免费捷径使均衡用时从 65.00 分钟升到 80.00 分钟,并标出悖论只在一段需求带里出现。 (https://5dlibrary.com/answers/zh/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c04/) - Ch.5 The Second-Order Interest on First-Order Shortcuts: Hidden Rework Liability — Chapter 5 writes the second-order rework of a first-order shortcut as a liability with an implied interest rate ranging from -51.5% to +26.8%; in a year-long simulation the careful lane spends 6.4% of its hours on rework against 22.6% for the fast lane. (https://5dlibrary.com/answers/en/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c05/) 第5章 一阶捷径的二阶利息:返工隐性负债 — 第五章把一阶捷径的二阶返工写成负债现值与隐含利率,利率随暴露速度与成本增长从 -51.5% 变到 +26.8%;一年模拟里慢车道的返工工时占比 6.4%,快车道 22.6%。 (https://5dlibrary.com/answers/zh/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c05/) - Ch.6 The Optimal Depth of Deliberation: Consequence Trees, Evidence For and Against, and Tail Stress — Chapter 6 builds consequence trees and finds the optimal depth of deliberation; leaving out opposing evidence makes a judgement of 0.643 look like 0.900. (https://5dlibrary.com/answers/en/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c06/) 第6章 推演深度的最优值:后果树、正反合证据与尾部压力 — 第六章搭后果树,求推演深度的最优值,漏记反方证据会把 0.643 的判断看成 0.900。 (https://5dlibrary.com/answers/zh/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c06/) - Ch.7 Classical Sources: the Sunzi on Manoeuvring, the Twenty-Second Chapter of the Laozi, and Bernoulli's Challenge — With L/H at 1.0 the line takes 0.6386 seconds and the cycloid 0.5829 seconds, so the traveller on the cycloid can leave 0.0557 seconds later, 8.7% of the line's time. The further away the target, the later one can leave: with L/H at 2.0 the head start is 0.2041 seconds, or 20.2%; at 3.0 it is 0.4095 seconds, or 28.7%. When the target is close there is little room: with L/H at 0.5 the head start is only 0.0141 seconds, or 2.8%. (https://5dlibrary.com/answers/en/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c07/) 第7章 原典考辨:《孙子·军争》《老子》第二十二章与伯努利征题 — L/H 为 1.0 时,直线要 0.6386 秒,摆线要 0.5829 秒,走摆线的人可以晚出发 0.0557 秒,占直线用时的 8.7%。目标越远,可以晚出发得越多:L/H 为 2.0 时可晚 0.2041 秒,占 20.2%;L/H 为 3.0 时可晚 0.4095 秒,占 28.7%。目标较近时,余地很小:L/H 为 0.5 时只能晚 0.0141 秒,占 2.8%。 (https://5dlibrary.com/answers/zh/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c07/) - Ch.8 Closing the Loop: the Holographic Decision Ledger, Field Tools and Checker Specifications — Chapter 8 gives the holographic decision ledger, the distilled Detour Break-even and Second-Order Rework Liability Calculator, checker specifications for the four kits and a pre-registered test whose primary comparison needs 198 decisions per arm. "Holographic" means only multi-dimensional coverage, with no mystical sense; deliberation budgets include rest. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c08/) 第8章 实证闭环:全息推演决策台账、实战工具与检查器规格 — 第八章给出全息推演决策台账、蒸馏小工具「迂回铺垫盈亏平衡与二阶返工负债推演器」、四件套检查器规格与预注册检验,主检验每组需要 198 项决策。书中「全息」只指多维全覆盖,不含任何玄学含义;深度推演的时间预算包含休息,不以透支换产出。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/brachistochrone-detours-second-order-consequence-mapping-deep-thinking/c08/) ## 5DL-B-385 · Reasoning-Trajectory Profiling, Process-Tracing Games and Separating Cognitive Structure from Stated Intentions and Outcome Luck 中文书名:思维脉络溯源识人、过程追踪博弈与剥离意图结果假象母典 DOI: pending Cite: Fan, Y. (2026). Reasoning-Trajectory Profiling, Process-Tracing Games and Separating Cognitive Structure from Stated Intentions and Outcome Luck (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: After-the-fact explanations are often folk theories: in a synthetic experiment, retrospective self-reports correlate only -0.013 with the weights used but 0.826 with an observer's guess; concurrent thinking aloud keeps order and leaps. Parameters are illustrative and all calculations come from the book's programs. Abstract: Two popular beliefs guide the choice of partners, buyers, clients and seat applicants: hear a person explain why he acted and you can judge him; count his deals won and lost and you know his judgement. This book refutes the first on its own ground and bounds the second. After-the-fact explanations are often folk theories: in a synthetic experiment, retrospective self-reports correlate only -0.013 with the weights used but 0.826 with an observer's guess; concurrent thinking aloud keeps order and leaps. Track records mix judgement with luck: when success rates differ by a tenth, 82 decisions are needed before the weaker decider's chance of showing the better record falls below one in ten. Blueprint corrections: people can perform reasoning and onlookers judge lies barely better than a coin, so the book drops the "unforgeable fingerprint of thought" and claims only that forgery cost rises with probing depth, with bounds and a refutation condition. Selection validity is anchored to revised estimates: structured interviews rank first at about 0.42, so any single method is limited and several must be combined by rule. Parameters are illustrative and all calculations come from the book's programs. The book is not a psychometric instrument, not legal advice on hiring, partnership or personal data, and not personalised business advice. Chapter 1 separates after-the-fact explanation from thinking aloud. Chapter 2 writes the four process-tracing tests as likelihood ratios and corrects for correlated clues. Chapter 3 turns thinking aloud into a reasoning graph with leap rate, citation share, backtrack rate, counterexample latency and revision rate. Chapter 4 tests its stability under three pressures, and rater agreement, with intraclass correlation and kappa. Chapter 5 models forgery cost: under illustrative parameters, probing to level six flags a performer with probability 0.784 and a genuine reasoner with 0.241, while the claim fails where scripts are ample and reasoning trees narrow, so a flag can only trigger review. Chapter 6 strips outcome luck with exact binomials, the sequential probability ratio test and multi-method composites. Chapter 7 reads the Analects on seeing, watching and examining, and on listening to words and watching conduct, Bayes's essay of 1763 and Aristotle on choice. Chapter 8 gives field tools and checker specifications: ledger fields, five gates, the distilled Reasoning-Trajectory Process-Tracing and Outcome-Luck Separation Scorer, checks grouped by the four kits, and a pre-registered plan whose binding validity test needs 347 people. All reasoning traces are collected with prior notice and consent; the book describes no covert lie detection, inducement, deception or extraction of trade secrets, uses no protected characteristic, and keeps outcomes in a separate column that never enters the reasoning score. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:遴选合伙人、采购骨干、大客户与年度席位申请者时,最流行的两句说法是:听他自述为什么这样做就能识人;看他做成几单、搞砸几单就知道判断好坏。本书逐条证伪或立界。事后解释常常是一套通俗理论:合成实验里,事后自述与真实权重的相关只有 -0.013,与旁观者凭常识的猜测却高达 0.826;同步出声思考保留了次序与跳跃,要求边做边解释则会改变被测过程。成败记录混合了判断与运气:成功率相差一成时,要 82 次决策才能把差者战绩反超的概率压到一成以下。核对文献后本书做了蓝图修正:人能表演推理,旁人判别真假的正确率只比抛硬币高一点,因此不主张「不可伪造的思维指纹」,只主张伪造成本随追问深度上升,并写出上下界与推翻条件;甄选效度以修订后的估计为锚,结构化访谈居首也只有 0.42 上下,单一方法有限,须按规则多方法合成。书中参数为示意,全部计算由书中程序生成;本书不是心理测评工具,也不构成雇佣、合伙或个人资料方面的法律意见,非个人化经营建议。 第一章用合成实验区分事后解释与出声思考。第二章把过程追踪的风中草、环形检验、确凿证据与双重决定写成似然比,并演示线索相关时朴素相乘的高估。第三章把出声思考转写成推理图,定义跳跃率、引证占比、回溯率、反例时延与修正率。第四章用组内相关与卡帕检验指纹在利益冲突、反例冲击、信息缺失三类压力下的稳定性与评分者一致性。第五章给出伪造成本模型:示意参数下第六层追问时表演者与真推理者的标记概率为 0.784 与 0.241,而在脚本充分、推理树狭窄的角落里命题失效,所以一次标记只能触发复查。 第六章用二项分布、序贯概率比检验与多方法合成剥离结果运气。第七章考辨《论语》「视其所以,观其所由,察其所安」与「听其言而观其行」、贝叶斯 1763 年的遗作和亚里士多德论选择。第八章给出实战工具与检查器规格:台账字段、五道闸门、蒸馏小工具「推理轨迹过程追踪与结果运气剥离评分器」、按四件套分组的检查条款与预注册方案,起约束作用的效度检验需要 347 人。一切推理轨迹采集须事先告知并取得同意,不写隐蔽测谎、诱导、欺骗或套取商业秘密,不以任何受保护特征作判别,结果成败单列、永不进入推理评分。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Limits of Self-Report: After-the-Fact Reasons, Thinking Aloud and Reactivity — Thinking aloud keeps 0.919 of the steps, its Kendall correlation with the real order is 0.977, and 0.968 of reports still show at least one leap. Retrospective recall keeps only 0.648 of the steps, its correlation falls to 0.833, and only 0.235 of reports still show a leap. (https://5dlibrary.com/answers/en/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c01/) 第1章 听其自述的边界:事后解释、出声思考与反应性 — 出声思考保住了 0.919 的步骤,与真实次序的肯德尔相关是 0.977,0.968 的报告仍能看出至少一处跳跃。事后回忆只记起 0.648 的步骤,相关降到 0.833,能看出跳跃的报告只剩 0.235。 (https://5dlibrary.com/answers/zh/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c01/) - Ch.2 Four Process-Tracing Tests Written as Likelihood Ratios — With ρ equal to zero, the true ratio for seeing both clues is 4.000, matching naive multiplication, and the posterior is 0.800. At ρ of 0.25 the true ratio falls to 3.024 and the posterior to 0.752; at 0.5 the values are 2.519 and 0.716; at 0.75 they are 2.209 and 0.688. At ρ equal to one the second clue merely echoes the first, the true ratio returns to the single-clue value of 2.000, and the posterior is 0.667. Naive multiplication reports 0.800 throughout. (https://5dlibrary.com/answers/en/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c02/) 第2章 过程追踪四类检验:把线索写成似然比 — ρ 为零时,两条线索同时出现的真实似然比是 4.000,与朴素相乘一致,后验 0.800。ρ 升到 0.25,真实似然比降到 3.024,后验 0.752;ρ 为 0.5 时是 2.519 与 0.716;ρ 为 0.75 时是 2.209 与 0.688;ρ 为一时,第二条线索纯属复读,真实似然比回到单条的 2.000,后验 0.667。朴素相乘却始终报告 0.800。 (https://5dlibrary.com/answers/zh/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c02/) - Ch.3 The Reasoning Graph: Nodes, Leaps, Backtracks and Revisions — At twelve edges the interval runs from 0.063 to 0.483, with a half-width of 0.210, which says almost nothing. At 48 edges it narrows to 0.111 to 0.333; at 96 edges it is 0.132 to 0.291; at 192 edges the half-width is 0.056; at 384 edges the interval runs from 0.163 to 0.243 and the half-width is 0.040. (https://5dlibrary.com/answers/en/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c03/) 第3章 推理图:节点、跳跃、回溯与反例修正 — 十二条边时,区间从 0.063 到 0.483,半宽 0.210,几乎什么都说明不了。边数翻到 48 条,区间收到 0.111 到 0.333;96 条时是 0.132 到 0.291;192 条时半宽 0.056;384 条时是 0.163 到 0.243,半宽 0.040。 (https://5dlibrary.com/answers/zh/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c03/) - Ch.4 Fingerprints under Pressure: Intraclass Correlation and Rater Agreement — With a single-scene coefficient of 0.30, the mean of three scenes reaches only 0.562 and the mean of six 0.720; reaching 0.70 takes 6 scenes and reaching 0.80 takes 10. With a coefficient of 0.45, three scenes give 0.711 and six give 0.831, so 3 and 5 scenes are needed respectively. With 0.60, two scenes suffice for 0.70 and three for 0.80. (https://5dlibrary.com/answers/en/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c04/) 第4章 压力场景下的推理指纹:组内相关与评分者一致性 — 单场景系数 0.30 时,三场平均只有 0.562,六场 0.720;要到 0.70 需要 6 场,要到 0.80 需要 10 场。系数 0.45 时,三场 0.711,六场 0.831,分别需要 3 场与 5 场。系数 0.60 时,两场就够 0.70,三场够 0.80。 (https://5dlibrary.com/answers/zh/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c04/) - Ch.5 The Cost of Forgery: Probing Depth, Memory Load and Bounds on Detection — Chapter 5 models forgery cost: under illustrative parameters, probing to level six flags a performer with probability 0.784 and a genuine reasoner with 0.241, while the claim fails where scripts are ample and reasoning trees narrow, so a flag can only trigger review. (https://5dlibrary.com/answers/en/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c05/) 第5章 伪造成本:追问深度、记忆负荷与检出上下界 — 第五章给出伪造成本模型:示意参数下第六层追问时表演者与真推理者的标记概率为 0.784 与 0.241,而在脚本充分、推理树狭窄的角落里命题失效,所以一次标记只能触发复查。 (https://5dlibrary.com/answers/zh/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c05/) - Ch.6 Stripping Out Outcome Luck: Track Records, Sequential Tests and Multi-Method Composites — Records of success and failure mix the quality of decisions with luck. When sound and weak decision makers differ by a tenth in their success rates, the weak one still shows the better record about three times in ten after ten decisions, and getting that below one in ten takes more than eighty decisions; a sequential test saves about half and still needs more than a hundred. Classic validity estimates have been revised downward, and structured interviews, at the top, reach only about 0.42. Combining several methods by rule beats overall impression, and the increment from a reasoning-graph score depends on how much it overlaps with other methods; even in the most optimistic case it is modest. So outcomes are recorded separately and never enter the reasoning score, and the reasoning score is only ever one component of a composite. (https://5dlibrary.com/answers/en/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c06/) 第6章 剥离结果运气:成败记录、序贯检验与多方法合成 — 成败记录混合了决策质量与运气。好决策者与差决策者的成功率相差一成时,十次决策下差者战绩反超的概率仍有三成,要压到一成以下需要八十多次;序贯检验能省一半,仍需一百多次。经典甄选效度经修订后下调,结构化访谈居首也只有 0.42 上下;多方法按规则合成优于整体印象,推理图评分的增量取决于它与其他方法的相关,最乐观时也不大。所以,结果成败单独记、永不进推理评分,推理评分只作合成中的一项。 (https://5dlibrary.com/answers/zh/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c06/) - Ch.7 Classical Sources: the Analects on Seeing, Watching and Examining; Bayes and Aristotle on Choice — Chapter 7 reads the Analects on seeing, watching and examining, and on listening to words and watching conduct, Bayes's essay of 1763 and Aristotle on choice. (https://5dlibrary.com/answers/en/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c07/) 第7章 原典考辨:《论语》视观察三层与贝叶斯、亚里士多德论选择 — 第七章考辨《论语》「视其所以,观其所由,察其所安」与「听其言而观其行」、贝叶斯 1763 年的遗作和亚里士多德论选择。 (https://5dlibrary.com/answers/zh/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c07/) - Ch.8 Field Tools and Checker Specifications: the Reasoning-Trajectory Selection Ledger, the Scorer, Pre-Registration and Refutation — Chapter 8 gives field tools and checker specifications: ledger fields, five gates, the distilled Reasoning-Trajectory Process-Tracing and Outcome-Luck Separation Scorer, checks grouped by the four kits, and a pre-registered plan whose binding validity test needs 347 people. All reasoning traces are collected with prior notice and consent; the book describes no covert lie detection, inducement, deception or extraction of trade secrets, uses no protected characteristic, and keeps outcomes in a separate column that never enters the reasoning score. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c08/) 第8章 实战工具与检查器规格:推理轨迹甄选台账、评分器、预注册与证伪 — 第八章给出实战工具与检查器规格:台账字段、五道闸门、蒸馏小工具「推理轨迹过程追踪与结果运气剥离评分器」、按四件套分组的检查条款与预注册方案,起约束作用的效度检验需要 347 人。一切推理轨迹采集须事先告知并取得同意,不写隐蔽测谎、诱导、欺骗或套取商业秘密,不以任何受保护特征作判别,结果成败单列、永不进入推理评分。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/reasoning-trajectory-profiling-process-tracing-intention-outcome-luck/c08/) ## 5DL-B-386 · Mid-Stream Premise Seeding, Enthymematic Self-Derivation and Zero-Reactance Persuasion Along Reasoning Paths 中文书名:推演中途前提植入、省略三段论自发归结与零阻抗顺势说服母典 DOI: pending Cite: Fan, Y. (2026). Mid-Stream Premise Seeding, Enthymematic Self-Derivation and Zero-Reactance Persuasion Along Reasoning Paths (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative or synthetic, and all calculations are generated by the book's programs. The book is not psychological counselling, not a guarantee of marketing results and not individual business advice. Abstract: Two opposite sayings dominate talk about persuasion: the plainer the conclusion, the more persuasive the message; and the expert never states the conclusion, only sets out premises and lets the listener infer it. This book refutes the word "always" in each. After checking the original studies it records corrections to its blueprint: O'Keefe's meta-analysis of 13,754 participants found explicit conclusions more persuasive on average, with a mean correlation of 0.110; unstated conclusions win only when involvement is high and the listener can complete the inference; and the "zero reactance" of the title is not a demonstrable fact but a set of bounds and simulated quantiles on the reactance component. Parameters are illustrative or synthetic, and all calculations are generated by the book's programs. The book is not psychological counselling, not a guarantee of marketing results and not individual business advice. Chapter 1 converts the meta-analysis into readable units and explains why small tests flip direction so often. Chapter 2 draws the listener's inference as a reasoning-path graph, defines the value of a premise as the gain in the probability of the target conclusion, and uses an ethical filter to block premises that would erode an honest exit. Chapter 3 compares premises delivered before, at and after a branch point, and finds lateness costlier than earliness. Chapter 4 maps the region on the axes of motivation and ability where unstated conclusions win; with illustrative parameters the boundary is 0.6087 and the region covers only 0.0891, and across three weighted audiences the unstated conclusion has an average disadvantage of 0.4272, so local advantage and average disadvantage hold together. Chapter 5 computes the net loss from reactance with the intertwined model, where controlling wording lowers acceptance by 0.1772, and rewrites "zero reactance" as the engineering target of a threat component no greater than 0.041. Chapter 6 simulates how self-derived conclusions survive counterargument, with a crossover at 2.058 rounds, and gives a protocol for measuring restatement. Chapter 7 reads the Analects on not opening the way until the learner is eager, the Record of Learning on guiding without pulling, Aristotle on the enthymeme and Plato's Meno, and turns them into clauses for the four kits. Chapter 8 gives the ledger, the distilled Branch-Point Premise Value and Reactance Risk Calculator, checker specifications and pre-registered tests. The book covers only informed, refusable persuasion: premises are true and checkable, with no false scarcity, dark patterns, threats or pressure, and the other party may say no at any time. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:关于说服流传着两条相反的说法:结论讲得越直白越有说服力;高手从不说出结论,只摆前提让对方自己推出来。本书证伪两条说法里的「总是」。核对原文以后,本书写下蓝图修正:O'Keefe 的元分析汇总 13,754 名被试,明示结论平均更有说服力,平均相关 0.110;省略结论只在高卷入、听者有能力自行推出时占优;书名里的「零阻抗」不是可证事实,只能写成逆反分量的上下界与模拟分位数。书中参数为示意或合成数据,全部计算由书中程序生成。本书不是心理咨询,不是营销效果保证,也不是个人化经营建议。 第一章把元分析换成可读单位,并说明小样本测试为何频繁翻转方向。第二章把听者的推导画成推理路径图,以目标结论概率的增量定义前提价值,并用伦理过滤拦下压低正当退出的前提。第三章比较前提在分岔点之前、之时与之后投放,延后的损失大于提前。第四章在动机与能力两轴上求省略结论的占优区,示意参数下边界为 0.6087,占优区只占 0.0891;三类受众加权后留白平均劣势 0.4272,局部占优与平均劣势同时成立。 第五章以交织模型计算逆反净损失,控制型措辞让接受率下降 0.1772,并把「零阻抗」改写为威胁分量不超过 0.041 的工程目标。第六章模拟自行推出的结论在反驳下的保持率,交叉点在 2.058 轮,并给出复述率的测量办法。第七章考辨《论语》「不愤不启」、《礼记·学记》「道而弗牵」、亚里士多德论省略三段论与柏拉图《美诺篇》,章末落成四件套条款。第八章给出台账、蒸馏小工具「推理分岔点前提价值与逆反风险试算器」、检查器规格与预注册检验。全书只写知情、可拒绝的说服:前提真实可核验,不写虚假稀缺、暗黑模式与威胁催促,对方随时可以说不。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Stated or Unstated: the Meta-Analysis of Conclusion Explicitness and a Corrected Blueprint — The meta-analysis gives a clear average answer. Explicit conclusions are on average more persuasive, with a mean correlation of 0.110: a small effect, but a stable direction. Both camps can point to personal experience because the direction of a single small test flips quite often; with 100 readers the probability of a reversal is still 0.138. The advantage of leaving the conclusion unstated appears only in a region of high involvement and high ability. (https://5dlibrary.com/answers/en/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c01/) 第1章 明示还是留白:结论明示度元分析与蓝图修正 — 元分析给出的平均答案很清楚:明示结论平均更有说服力,平均相关 0.110,属于小效应但方向稳定。两派说法都能找到亲身经历,是因为单次小测试的方向本来就会翻转;样本 100 人时翻转概率仍有 0.138。留白的优势只在高卷入、高能力的区域里出现。 (https://5dlibrary.com/answers/zh/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c01/) - Ch.2 The Reasoning-Path Graph: Proposition Nodes, Completion Probabilities and Branch Points — The reasoning-path graph breaks persuasion down into two kinds of estimable number: which route a reader takes, and whether she can complete a step unaided. In this chapter's illustrative graph, the most valuable premise sits at the branch point where incomplete information produces a misunderstanding, and its increment is 2.222 times the baseline. Premises that would lower the honest exit are never stated. Premises that help a reader leave honestly are stated even when they lower the probability of the target. (https://5dlibrary.com/answers/en/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c02/) 第2章 推理路径图:命题节点、补全概率与分岔点 — 推理路径图把「说服」拆成可估的两类数:读者选哪条路,读者能否自己补全这一步。在本章的示意图上,最值钱的前提出现在信息不全导致误会的分岔点,增量是基线的 2.222 倍;能压低正当退出的前提一律不说;能帮助读者正当退出的前提即使降低目标概率,也照样说。 (https://5dlibrary.com/answers/zh/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c02/) - Ch.3 Timing: Premises Before, At and After the Branch Point — The same true premise works best when it arrives exactly as the reader reaches the branch point. One round early keeps 0.750 of the gain; one round late keeps only 0.550. Where the reader has already declared her position, saying it late costs more than saying it early. When the arrival round is unknown, the best fixed round is one round before the mean arrival; sincere check-in questions add a little more, and hand the reader's own words back to the author. (https://5dlibrary.com/answers/en/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c03/) 第3章 投放时机:前提在分岔点之前、之时与之后 — 同一条真实前提,恰好在读者走到分岔点时投放,效果最好;提前一轮保留 0.750 的增量,延后一轮只保留 0.550。在读者已经公开表态的场景里,晚说比早说损失更大。不知道读者何时到达时,最优的固定轮次比平均到达轮次早一轮;真诚的检查式提问还能再多一点,并且把读者自己的话交回作者手里。 (https://5dlibrary.com/answers/zh/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c03/) - Ch.4 Where the Enthymeme Wins: a Boundary on the Axes of Motivation and Ability — Chapter 4 maps the region on the axes of motivation and ability where unstated conclusions win; with illustrative parameters the boundary is 0.6087 and the region covers only 0.0891, and across three weighted audiences the unstated conclusion has an average disadvantage of 0.4272, so local advantage and average disadvantage hold together. (https://5dlibrary.com/answers/en/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c04/) 第4章 省略三段论的占优区:动机与能力两轴上的边界 — 第四章在动机与能力两轴上求省略结论的占优区,示意参数下边界为 0.6087,占优区只占 0.0891;三类受众加权后留白平均劣势 0.4272,局部占优与平均劣势同时成立。 (https://5dlibrary.com/answers/zh/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c04/) - Ch.5 The Reactance Component: Bounds on Threat to Freedom, Anger and Negative Cognitions — Chapter 5 computes the net loss from reactance with the intertwined model, where controlling wording lowers acceptance by 0.1772, and rewrites "zero reactance" as the engineering target of a threat component no greater than 0.041. (https://5dlibrary.com/answers/en/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c05/) 第5章 逆反分量:威胁自由、愤怒与负面认知的上下界 — 第五章以交织模型计算逆反净损失,控制型措辞让接受率下降 0.1772,并把「零阻抗」改写为威胁分量不超过 0.041 的工程目标。 (https://5dlibrary.com/answers/zh/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c05/) - Ch.6 The Ownership Gain: How Self-Derived Conclusions Survive Counterargument — Chapter 6 simulates how self-derived conclusions survive counterargument, with a crossover at 2.058 rounds, and gives a protocol for measuring restatement. (https://5dlibrary.com/answers/en/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c06/) 第6章 自我归属增益:自行推出的结论在反驳下的保持率 — 第六章模拟自行推出的结论在反驳下的保持率,交叉点在 2.058 轮,并给出复述率的测量办法。 (https://5dlibrary.com/answers/zh/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c06/) - Ch.7 Classical Sources: the Analects, the Record of Learning, Aristotle and Plato — The program adds an interesting side note. The first seven convergents of the continued fraction for the square root of eight are 2, 3, 14/5, 17/6, 82/29, 99/35 and 478/169, and their squares fall alternately on either side of 8, closer each time; the square of 99/35, for example, is 8.000816. The boy's second answer, 3, is precisely the first convergent after the starting point. The program also confirms that among fractions with denominators up to 1,000, none has a square exactly equal to 8; the square root of eight is about 2.8284. In other words, the answer to the problem is not a fraction at all. However long the boy tried numbers, he would never find it. The answer has to be a construction: the diagonal. (https://5dlibrary.com/answers/en/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c07/) 第7章 原典考辨:《论语》《学记》与亚里士多德、柏拉图 — 程序还给出了一个有趣的旁证。根号八的连分数渐近分数,前七个依次是 2、3、14/5、17/6、82/29、99/35 与 478/169,平方交替落在 8 的两侧,越来越近,例如 99/35 的平方是 8.000816。童仆第二次答的 3,正是起点之后第一个渐近分数。程序也验证了分母不超过 1,000 的分数里,没有一个平方恰好等于 8;根号八约为 2.8284。换句话说,这道题的答案不是任何分数,童仆再怎么试数也试不出来;答案必须是一个作图,对角线。 (https://5dlibrary.com/answers/zh/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c07/) - Ch.8 Field Tools and Checker Specifications: the Persuasion Ledger, the Calculator and Pre-Registration — Chapter 8 gives the ledger, the distilled Branch-Point Premise Value and Reactance Risk Calculator, checker specifications and pre-registered tests. The book covers only informed, refusable persuasion: premises are true and checkable, with no false scarcity, dark patterns, threats or pressure, and the other party may say no at any time. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c08/) 第8章 实战工具与检查器规格:顺势说服台账、试算器与预注册 — 第八章给出台账、蒸馏小工具「推理分岔点前提价值与逆反风险试算器」、检查器规格与预注册检验。全书只写知情、可拒绝的说服:前提真实可核验,不写虚假稀缺、暗黑模式与威胁催促,对方随时可以说不。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/mid-stream-premise-seeding-enthymeme-self-derivation-zero-reactance/c08/) ## 5DL-B-387 · Dissolving Entrenched Priors, Shattering the Illusion of Explanatory Depth and Anti-Backfire Persuasion Mechanics 中文书名:顽固先验偏见消解、解释深度错觉破除与反逆火效应沟通母典 DOI: pending Cite: Fan, Y. (2026). Dissolving Entrenched Priors, Shattering the Illusion of Explanatory Depth and Anti-Backfire Persuasion Mechanics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. Not counselling, no clinical service, not personalised business advice; examples concern only beliefs about goods, never political, religious or ethnic questions. Abstract: A popular saying holds that the more evidence you show people, the more firmly they hold their view: the backfire effect. This book refutes it first. Corrections to the blueprint: Nyhan and Reifler reported backfire, but Wood and Porter found none across 52 issues and more than ten thousand participants, Swire-Thompson and colleagues pointed to unreliable measurement, and Nyhan himself revised the conclusion; direct correction works on average, and the real obstacles are motivated discounting, which shrinks updates, and rebound after a correction. On the illusion of explanatory depth the book claims only that probing lowers self-rated understanding; three preregistered replications with 1,054 people in total reproduced no moderation of attitudes, so any attitude effect is treated as small; the paper by Voelkel and colleagues appeared in Comprehensive Results in Social Psychology, not in Cognition as the blueprint stated. Parameters are illustrative and all calculations are generated by the book's programs. Not counselling, no clinical service, not personalised business advice; examples concern only beliefs about goods, never political, religious or ethnic questions. Chapter 1 shows that when nobody truly backfires, a scale with reliability 0.50 still makes 0.351 of people appear to backfire, and 0.179 at reliability 0.95; that chance alone yields 2.60 significant backfires among fifty-two issues; and that under mixed evidence two camps drift apart whenever disagreeable items weigh less than 0.4286, with no backfire at all. Chapter 2 models stubbornness with a Beta prior and a discount weight, of which a trajectory identifies only the ratio. Chapter 3 covers decay, spacing and half-life. Chapter 4 uses a seven-point scale and a price chain: a felt markup of 0.4980 against a true margin of 0.0700, the gap lying in links customers never considered. Chapter 5: under high threat, affirming first and then giving four items lowers confidence by 0.0730, more than 0.0439 from five items; a benign thought experiment states its exit, and the best length for mockery is zero. Chapter 6 casts the four moves as a Markov decision process: from the hardest start the expected value is 10.555, the full path survives two thousand perturbations in a share of 0.8270, and keeping the baseline policy costs a mean relative regret of only 0.0003, so the result is called a simulated optimum. Chapter 7 reads the Zhuangzi, the Analects, Bacon's Aphorism 46 and Plato's Apology. Chapter 8 gives the ledger, the distilled Stubborn-Prior and Question-Sequence Calculator, four-kit checker specifications and pre-registered tests: the reliable change index is 22.31 points and the backfire-rate test needs 169 records. Boxes separate published evidence, Python calculations (22 programs), classical sources and hypotheses. 摘要:客服、答疑页与团队复盘里流传一句流行说法:对方认定的事,越摆证据越认定,这叫逆火效应。本书先证伪这句话。核对原文之后,蓝图修正如下:奈恩与赖夫勒报告过逆火,但伍德与波特在 52 个议题、上万被试中没有复现,斯怀尔·汤普森等人指出测量信度问题,奈恩本人也修正了结论;直接纠正平均有效,真正的障碍是动机性折扣让更新变小,以及纠正之后的回弹。解释深度错觉只主张追问能降低自评理解度,三次预注册复现合计 1,054 人都没有复现态度温和化,态度作用只写成小效应;福尔克尔等人的论文发表在《社会心理学综合结果》而非蓝图所写的《认知》。书中参数为示意,全部计算由书中程序生成;本书非心理咨询、不提供任何临床服务、非个人化经营建议,例子只用商品与消费认知,不涉及政治、宗教与族群议题。 第一章说明,没有任何人真逆火时,重测信度 0.50 的量表也会让 0.351 的人看上去在逆火,信度 0.95 时仍有 0.179;五十二个议题里光凭偶然平均就有 2.60 个显著逆火;混合证据下,不合意材料权重低于 0.4286,两派就会拉开,不需要任何逆火。第二章用贝塔先验与折扣系数写顽固,证明单看轨迹只能识别二者之比。第三章给出纠正的衰减、间隔与半衰期估计。第四章用七级量表与价格链说明自以为懂,感受加价 0.4980 与真实毛利 0.0700 的差全在没想到的环节上。 第五章说明高威胁下先肯定再给四件证据,降幅 0.0730,大于连给五件的 0.0439;善意思想实验明说可退出,讥讽的最佳长度是零步。第六章把追问、肯定、证据与思想实验写成马尔可夫决策过程,示意参数下最难起点的期望收益 10.555,参数扰动两千次,完整路线不变的比例为 0.8270,沿用基准策略的平均相对遗憾只有 0.0003,所以只称模拟最优。第七章考辨《庄子·秋水》、《论语·为政》、培根《新工具》格言四十六与柏拉图《申辩篇》。第八章给出先验消解沟通台账、蒸馏小工具「顽固度先验与拆墙提问序列试算器」、按四件套分组的检查器规格与预注册检验:可靠变化指数 22.31 分,逆火率检验需要 169 条记录。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Bounding the Backfire Effect: Corrections Work on Average, Reversals Are Rare — Chapter 1 shows that when nobody truly backfires, a scale with reliability 0.50 still makes 0.351 of people appear to backfire, and 0.179 at reliability 0.95; that chance alone yields 2.60 significant backfires among fifty-two issues; and that under mixed evidence two camps drift apart whenever disagreeable items weigh less than 0.4286, with no backfire at all. (https://5dlibrary.com/answers/en/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c01/) 第1章 逆火之辨:纠正平均有效,逆火罕见 — 第一章说明,没有任何人真逆火时,重测信度 0.50 的量表也会让 0.351 的人看上去在逆火,信度 0.95 时仍有 0.179;五十二个议题里光凭偶然平均就有 2.60 个显著逆火;混合证据下,不合意材料权重低于 0.4286,两派就会拉开,不需要任何逆火。 (https://5dlibrary.com/answers/zh/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c01/) - Ch.2 Motivated Discounting: Beta Priors, Pseudo-Samples and Evidence Weights — Stubbornness comes in two kinds, a strong prior and a heavy discount. On a single person's trajectory they cannot be told apart, because the mean depends only on $w/n_0$. For a person with a pseudo-sample of 80 who keeps a fifth of the weight, the first document moves the mean by only 0.0020; of the twenty-four cells with a positive weight, only 3 cross 0.5 within one conversation. Separating discounting from backfire, when the difference is a point or two, takes hundreds of people, and the number computed from the expectation gives only even odds. (https://5dlibrary.com/answers/en/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c02/) 第2章 动机性折扣:贝塔先验、伪样本量与证据权重 — 固执有两种:先验强与折扣重。二者在单人轨迹上无法区分,因为均值只依赖 $w/n_0$。伪样本量 80、打两成折扣的人,第一件材料只挪动 0.0020;二十四个正折扣格子里,只有 3 个能在一次沟通内把信念推过 0.5。分清打折与逆火需要的样本,在差别只有一两分时动辄数百人,而按期望算出的人数只有一半把握。 (https://5dlibrary.com/answers/zh/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c02/) - Ch.3 Rebound and Softening: the Decay of Corrections, Spacing and Half-Life — Corrections work, but they fade. With a half-life of 3 days, a single correction in four weeks reduces the average by only 0.0484; daily corrections give the largest total reduction, but each buys only 0.0185. When the pull is weaker and the half-life is 10 days, weekly corrections already achieve a reduction of 0.3118. The half-life must be estimated from one's own records, and ten rounds of data give an interval from 1.41 to 2.26 rounds, which is not narrow. A correction must come as a set, withdrawing the conclusion while offering a checkable alternative account, or the gap will pull the old claim back. (https://5dlibrary.com/answers/en/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c03/) 第3章 回弹与软化:纠正的衰减、间隔与半衰期 — 纠正有效,但会褪色。半衰期 3 天时,四周只纠正一回,平均降幅只有 0.0484;天天纠正,总降幅最大,可每次只换来 0.0185。拉力弱、半衰期 10 天时,每周一回就有 0.3118 的降幅。半衰期要从自己的记录里估,十轮数据估出的区间从 1.41 到 2.26 轮,并不窄。纠正还要成套,撤回结论的同时给出可核对的替代解释,否则空缺会把旧说法拉回来。 (https://5dlibrary.com/answers/zh/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c03/) - Ch.4 The Illusion of Explanatory Depth: the Drop on a Seven-Point Scale and the Small Attitude Effect — Chapter 4 uses a seven-point scale and a price chain: a felt markup of 0.4980 against a true margin of 0.0700, the gap lying in links customers never considered. (https://5dlibrary.com/answers/en/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c04/) 第4章 解释深度错觉:七级量表上的跌幅与态度小效应 — 第四章用七级量表与价格链说明自以为懂,感受加价 0.4980 与真实毛利 0.0700 的差全在没想到的环节上。 (https://5dlibrary.com/answers/zh/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c04/) - Ch.5 Self-Affirmation and Benign Thought Experiments: Affirm First, Then Evidence, with an Exit Always Open — Chapter 5: under high threat, affirming first and then giving four items lowers confidence by 0.0730, more than 0.0439 from five items; a benign thought experiment states its exit, and the best length for mockery is zero. (https://5dlibrary.com/answers/en/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c05/) 第5章 自尊缓冲与善意思想实验:先肯定,再给证据,随时可退出 — 第五章说明高威胁下先肯定再给四件证据,降幅 0.0730,大于连给五件的 0.0439;善意思想实验明说可退出,讥讽的最佳长度是零步。 (https://5dlibrary.com/answers/zh/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c05/) - Ch.6 Question Sequences as a Decision Process: a Simulated Optimum under Illustrative Parameters and Its Sensitivity — Chapter 6 casts the four moves as a Markov decision process: from the hardest start the expected value is 10.555, the full path survives two thousand perturbations in a share of 0.8270, and keeping the baseline policy costs a mean relative regret of only 0.0003, so the result is called a simulated optimum. (https://5dlibrary.com/answers/en/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c06/) 第6章 拆墙提问序列:示意参数下的模拟最优与敏感度 — 第六章把追问、肯定、证据与思想实验写成马尔可夫决策过程,示意参数下最难起点的期望收益 10.555,参数扰动两千次,完整路线不变的比例为 0.8270,沿用基准策略的平均相对遗憾只有 0.0003,所以只称模拟最优。 (https://5dlibrary.com/answers/zh/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c06/) - Ch.7 Classical Sources: the Frog in the Well and Knowing What One Knows, Bacon's Aphorism 46 and Socrates' Knowledge of His Ignorance — The respondent answers 23 of the forty questions correctly, a share of 0.575. With no pull-back, average confidence is 0.8115, which is 0.2365 above the actual hit rate, and the Brier score is 0.2908, worse than the 0.2500 one would get by answering "fifty-fifty" to everything. Pulling back by half gives average confidence of 0.6966 and a Brier score of 0.2475, just better than fifty-fifty. Pulling back fully gives average confidence of 0.5816, only 0.0066 above the actual rate, and a Brier score of 0.2316. (https://5dlibrary.com/answers/en/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c07/) 第7章 原典考辨:井蛙与「知之为知之」,培根的格言四十六与苏格拉底的自知无知 — 四十题答对 23 题,比例 0.575。不做任何收回时,平均信心 0.8115,高出实际 0.2365,布里尔分数 0.2908,比对所有题都说「一半一半」的 0.2500 还差。收回一半时,平均信心 0.6966,布里尔分数 0.2475,刚好胜过一半一半。完全收回时,平均信心 0.5816,高出实际只有 0.0066,布里尔分数 0.2316。 (https://5dlibrary.com/answers/zh/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c07/) - Ch.8 Closing the Loop: the Prior-Dissolving Ledger, Field Tools and Checker Specifications — Chapter 8 gives the ledger, the distilled Stubborn-Prior and Question-Sequence Calculator, four-kit checker specifications and pre-registered tests: the reliable change index is 22.31 points and the backfire-rate test needs 169 records. Boxes separate published evidence, Python calculations (22 programs), classical sources and hypotheses. (https://5dlibrary.com/answers/en/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c08/) 第8章 实证闭环:先验消解沟通台账、实战工具与检查器规格 — 第八章给出先验消解沟通台账、蒸馏小工具「顽固度先验与拆墙提问序列试算器」、按四件套分组的检查器规格与预注册检验:可靠变化指数 22.31 分,逆火率检验需要 169 条记录。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/entrenched-priors-explanatory-depth-illusion-anti-backfire-persuasion/c08/) ## 5DL-B-388 · Dilution-Coacervation Fragrance Deposition, Surfactant-Polymer Golden Ratios and Skin-Feel Tribology in Luxury Body Wash 中文书名:高定香氛沐浴体系稀释絮凝留香、表活阳离子黄金配比与丝滑摩擦学母典 DOI: pending Cite: Fan, Y. (2026). Dilution-Coacervation Fragrance Deposition, Surfactant-Polymer Golden Ratios and Skin-Feel Tribology in Luxury Body Wash (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. Chapter 5 covers the salt curve and wormlike micelles, and sets batches at the peak measured with perfume present (illustrative coefficient of variation 0.027). Abstract: Three popular sayings surround scented body wash: lasting scent comes from adding more perfume, the more slippery the more luxurious, and extra perfume changes only the smell. This book refutes or bounds each of them. How much fragrance a rinse-off product leaves on skin depends mainly on how much oily phase the complex coacervate, formed by a cationic polymer and an anionic surfactant at the moment of dilution, carries onto the skin, and on the oil-water partition coefficient and vapour pressure of each fragrance material. Under illustrative parameters, doubling the perfume multiplies perceived intensity at four hours by only 1.464, while adding a cationic polymer at the same perfume dose multiplies it by 1.740; high log P materials are retained about 3.86 times as well as without polymer, while low log P materials hardly benefit. After checking the literature the book corrects its blueprint: the leftward shift and lowering of the salt-curve peak by perfume was measured directly by Parker and Fieber in 2013; four further source corrections are listed in the preface. Parameters are illustrative and all calculations are generated by the book's programs. Chapters 1 to 4 cover the coacervate phase map, Langmuir-type deposition kinetics, three-way partition by log P with micelles competing for fragrance, and Raoult evaporation with Stevens perception. Chapter 5 covers the salt curve and wormlike micelles, and sets batches at the peak measured with perfume present (illustrative coefficient of variation 0.027). Chapter 6 separates squeaky, velvety and falsely slippery films by the Hersey number and turns the golden ratio into a feasible region of 85 out of 1,440 grid formulas, with a charge ratio between about 0.0048 and 0.0087, each edge carrying a refutation condition. Chapter 7 reads the Xiang Pu, the Liji, Theophrastus and Pliny. Chapter 8 gives the ledger, a field calculator, checker tables and pre-registered tests. The book deals only with cleansing, skin feel, scent and sensory experience and makes no medical claim; it is not a manufacturing instruction, a safety assessment or a regulatory opinion. Any formula needs qualified stability, preservative-challenge, patch and toxicological safety assessment; fragrance must follow the IFRA standards and local allergen labelling rules, which are amended and must be checked in their current official texts. The book gives no home-made instructions, names or rates no brand and quotes no prices. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. 摘要:香氛沐浴有三句流行说法:留香靠多加香精,越滑越高级,香精加多了只影响气味。本书逐条证伪或立界。冲洗型产品在皮肤上留下多少香料,主要取决于稀释那一刻阳离子聚合物与阴离子表活形成的复合凝聚相能把多少油相带上皮肤,以及每种香料的油水分配系数与蒸汽压。在示意参数下,香精加倍只让四小时感知强度乘以 1.464,同样香精量下加入阳离子聚合物则乘以 1.740;高 log P 原料的冲洗后留存约为不加聚合物时的 3.86 倍,低 log P 原料几乎不受益。核对文献后本书做了蓝图修正:香精让盐曲线峰位左移、峰高下降已有帕克与菲伯 2013 年的直接测量,不再写成作者假说;另有四处出处更正写在序里。书中参数为示意,全部计算由书中程序生成。 第一章画复合凝聚相图,说明稀释路径为何在对数坐标上是斜线;第二章用朗缪尔型速率方程说明沉积受动力学限制,冲洗习惯能把沉积量拉开约三倍;第三章按 log P 计算香料在水、胶束与凝聚相之间的分配,指出胶束在争夺香料;第四章用拉乌尔定律与史蒂文斯幂律写前中后调的挥发与感知;第五章写盐曲线、蠕虫状胶束与香精左移,把批次定在加香后的峰上,示意批间变异系数为 0.027;第六章用赫西数区分涩、丝绒与假滑,把黄金配比写成 1,440 个网格配方中 85 个组成的可行域,电荷当量比约在 0.0048 到 0.0087 之间,每条边界都附推翻条件。 第七章考辨洪芻《香譜》、《陳氏香譜》、《禮記·內則》、泰奥弗拉斯托斯《论气味》与普林尼《博物志》,并蒸馏出四件套条款。第八章给出配方与肤感台账、周度核对、实战工具「稀释絮凝留香与肤感润滑区试算器」、四件套检查器规格表与预注册检验。本书只写清洁、肤感、香气与感官体验,不宣称任何医疗功效;不是配方生产指导、安全评估或法规意见;任何配方须由合格人员完成稳定性、防腐挑战、斑贴与毒理安全评估,香料须遵守 IFRA 标准与各地过敏原标示规定,法规会修订,以官方现行文本为准;不提供在家自制做法,不点名评价任何品牌,不写任何带币种价格。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Dilution-Induced Coacervation: the Phase Map and the Dilution Path — On logarithmic axes the dilution path is a diagonal line along which the charge ratio stays fixed. Separation happens because free surfactant falls from "enough to resolubilise" to "not enough", and then below the aggregation threshold. The illustrative formula enters the two-phase region at 6.67 and leaves at 600, a crossing of 1.95 decades. When the models of the later chapters are chained, better deposition beats doubled perfume under the illustrative parameters. (https://5dlibrary.com/answers/en/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c01/) 第1章 稀释絮凝:复合凝聚相图与稀释路径 — 稀释路径在对数坐标上是一条斜线,电荷当量比沿途不变,分相来自游离表活从「足以再增溶」降到「不足以再增溶」,再降到「低于聚集浓度」。示意配方在 6.67 倍进入两相区、600 倍离开,穿越 1.95 个数量级。把后续章节的模型串起来,改进沉积的效果在示意参数下超过加倍香精。 (https://5dlibrary.com/answers/zh/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c01/) - Ch.2 Deposition Kinetics: Langmuir Adsorption and the Length of the Rinse — The Langmuir-type rate equation shows that deposition from the illustrative formula is far from equilibrium. The window occupies 0.56 of the rinse, the coverage stops at 0.4182 and the deposited share is 0.0627. Rinsing habit can spread deposition by roughly a factor of three, and a higher polymer dose and charge density narrow the spread. Deposition is a kinetic problem, not an equilibrium one. (https://5dlibrary.com/answers/en/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c02/) 第2章 沉积动力学:朗缪尔吸附与冲洗时长 — 朗缪尔型速率方程显示,示意配方的沉积远未达到平衡:窗口只占冲洗时长的 0.56,覆盖率停在 0.4182,沉积份额 0.0627。冲洗习惯能把沉积量拉开三倍左右,聚合物用量与电荷密度越高,分布越窄。沉积是动力学问题,不是平衡问题。 (https://5dlibrary.com/answers/zh/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c02/) - Ch.3 Fragrance Partition: log P, Micellar Competition and What Remains after Rinsing — Fragrance materials fall into two groups by log P. Those below 2 leave almost entirely with the water, and the polymer barely helps them. Those above 4 can be carried onto skin by the coacervate, and with the polymer they are retained about 3.86 times as well as without it. Micelles and coacervate compete for the same fragrance, and more surfactant means less retention. The benefit of hydrophobicity is capped beyond a log P of about four by the volume ratio of coacervate to micelles. (https://5dlibrary.com/answers/en/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c03/) 第3章 香料分配:log P、胶束竞争与冲洗后留存 — 香料按 log P 分成两群:低于 2 的分子几乎全部随水流走,加聚合物也帮不上忙;高于 4 的分子能被凝聚相带上皮肤,加聚合物后留存约为不加时的 3.86 倍。胶束与凝聚相争夺同一批香料,表活越多,留存反而越少。疏水性的好处在 log P 约为四之后被凝聚相与胶束的体积比封顶。 (https://5dlibrary.com/answers/zh/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c03/) - Ch.4 Top, Heart and Base Notes: Raoult Evaporation and the Power Law of Stevens — Raoult's law and first-order kinetics give each fragrance material a half-life on skin that follows its vapour pressure, from a third of an hour for limonene to hundreds of hours for hexyl cinnamal. The power law of Stevens compresses what the nose reports: at four hours the physical signal has fallen to 0.3547 while perceived intensity has fallen only to 0.5655. Non-ideal mixing bends the rule that evaporation follows vapour pressure without breaking it, and it penalises water-loving materials a second time. (https://5dlibrary.com/answers/en/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c04/) 第4章 前中后调:拉乌尔挥发与史蒂文斯幂律 — 拉乌尔定律与一阶动力学给每种香料一个跟着蒸汽压走的皮肤半衰期,从柠檬烯的三分之一小时到己基肉桂醛的数百小时。史蒂文斯幂律压缩了鼻子的报告:四小时时物理信号降到 0.3547,感知强度只降到 0.5655。非理想混合让「挥发跟着蒸汽压走」的规则弯曲而不折断,并让亲水原料再吃一次亏。 (https://5dlibrary.com/answers/zh/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c04/) - Ch.5 Micellar Viscosity: the Salt Curve, Wormlike Micelles and the Leftward Shift Caused by Perfume — Chapter 5 covers the salt curve and wormlike micelles, and sets batches at the peak measured with perfume present (illustrative coefficient of variation 0.027). (https://5dlibrary.com/answers/en/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c05/) 第5章 胶束黏度:盐曲线、蠕虫状胶束与香精左移 — 第五章写盐曲线、蠕虫状胶束与香精左移,把批次定在加香后的峰上,示意批间变异系数为 0.027; (https://5dlibrary.com/answers/zh/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c05/) - Ch.6 Skin-Feel Tribology: the Stribeck Curve, the Velvet Zone and the Feasible Region behind the Golden Ratio — Chapter 6 separates squeaky, velvety and falsely slippery films by the Hersey number and turns the golden ratio into a feasible region of 85 out of 1,440 grid formulas, with a charge ratio between about 0.0048 and 0.0087, each edge carrying a refutation condition. (https://5dlibrary.com/answers/en/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c06/) 第6章 丝滑摩擦学:斯特里贝克曲线、丝绒区与黄金配比可行域 — 第六章用赫西数区分涩、丝绒与假滑,把黄金配比写成 1,440 个网格配方中 85 个组成的可行域,电荷当量比约在 0.0048 到 0.0087 之间,每条边界都附推翻条件。 (https://5dlibrary.com/answers/zh/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c06/) - Ch.7 Classical Sources: the Xiang Pu, the Neize of the Liji, Theophrastus and Pliny — The book's illustrative perfume, which looks balanced by weight, is not balanced by perception. Its evenness starts at 0.634, falls to 0.244 at two hours, when the heart notes dominate, and recovers to 0.396 by eight hours; its mean over the day is 0.395. The top-heavy variant is more even early, at 0.636 at the start and still 0.697 at half an hour, but it collapses later and averages 0.476. The base-heavy variant is the most even through the day, with 0.732 at the start, 0.458 at two hours and 0.613 at eight hours, and a mean of 0.573. (https://5dlibrary.com/answers/en/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c07/) 第7章 原典考辨:《香譜》《禮記·內則》与泰奥弗拉斯托斯、普林尼 — 本书的示意香精按重量看很均衡,按感知却不均衡。它的均匀度起步 0.634,两小时降到 0.244,此时心调独大,八小时回升到 0.396,全天平均 0.395。偏前调变体在早段更均匀,起步 0.636,半小时仍有 0.697,后段却塌下去,平均 0.476。偏后调变体全天最均匀,起步 0.732,两小时 0.458,八小时 0.613,平均 0.573。 (https://5dlibrary.com/answers/zh/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c07/) - Ch.8 The Empirical Loop: the Formula and Skin-Feel Ledger, Field Tools and Checker Specifications — Chapter 8 gives the ledger, a field calculator, checker tables and pre-registered tests. The book deals only with cleansing, skin feel, scent and sensory experience and makes no medical claim; it is not a manufacturing instruction, a safety assessment or a regulatory opinion. Any formula needs qualified stability, preservative-challenge, patch and toxicological safety assessment; fragrance must follow the IFRA standards and local allergen labelling rules, which are amended and must be checked in their current official texts. The book gives no home-made instructions, names or rates no brand and quotes no prices. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c08/) 第8章 实证闭环:配方与肤感台账、实战工具与检查器规格 — 第八章给出配方与肤感台账、周度核对、实战工具「稀释絮凝留香与肤感润滑区试算器」、四件套检查器规格表与预注册检验。本书只写清洁、肤感、香气与感官体验,不宣称任何医疗功效;不是配方生产指导、安全评估或法规意见;任何配方须由合格人员完成稳定性、防腐挑战、斑贴与毒理安全评估,香料须遵守 IFRA 标准与各地过敏原标示规定,法规会修订,以官方现行文本为准;不提供在家自制做法,不点名评价任何品牌,不写任何带币种价格。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/dilution-coacervation-fragrance-deposition-surfactant-polymer-skin-tribology/c08/) ## 5DL-B-389 · Water-Hardness Chemistry in Personal-Care Formulations, Calcium-Magnesium Chelation and Surfactant Foam-Deposition Enhancement 中文书名:洗护配方水质硬度化学、钙镁离子螯合软化与表活泡沫沉积增效母典 DOI: pending Cite: Fan, Y. (2026). Water-Hardness Chemistry in Personal-Care Formulations, Calcium-Magnesium Chelation and Surfactant Foam-Deposition Enhancement (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Chapter 1 covers units of hardness, six illustrative waters and the activity of calcium. Parameters are illustrative and all calculations are generated by the book's programs. Abstract: Three popular sentences circulate in the shampoo and shower categories: hard water makes a whole bottle stop working; adding more chelant to a formula softens hard water in the shower; and a deionised-water base marks a premium formula. This book refutes them one by one or draws the boundary inside which they hold. Calcium and magnesium first lower the critical micelle concentration and precipitate calcium salts only in an intermediate band of dilution during rinsing, so the loss is graded by surfactant system: soaps lose most, alkyl sulfates next, acyl glutamates sit in the middle, and ether sulfates, amphoteric and nonionic surfactants are hardly affected; in the illustrative W4 water an alkyl sulfate loses a share of 0.632 in the early rinse. A chelant binds calcium and magnesium one to one, so with illustrative inputs one dose softens only 10.5 mL of hard water and does not even cover the lathering water; softening a whole shower would take 1,524 times the chelant in the formula, and a whole bottle covers about 3.16 seconds of shower flow. After checking the sources the book corrects its blueprint: the ceiling on coverage is tighter than the blueprint said; two small in-vitro studies of hard water and hair strength disagree, and no health conclusion is drawn from them; and no direct literature settles how calcium shifts the deposition of cationic conditioning polymers, so the direction is stated as a hypothesis to be tested. Chapter 1 covers units of hardness, six illustrative waters and the activity of calcium. Chapter 2 covers the Corrin-Harkins line, solubility products and the protection offered by micelles. Chapter 3 converts losses into foam height and half-life. Chapter 4 carries the main argument: the stoichiometry of chelation and its ceiling. Chapter 5 computes conditional stability constants, which at pH 5.5 fall to 5.06 for EDTA and 2.90 for GLDA, and shows that during rinsing a chelant does not postpone precipitation by a single step. Chapter 6 covers the deposition of cationic polymers, where screening and bridging give coacervation-window width ratios of 0.677 and 1.323. Chapter 7 reads Lu Yu's Classic of Tea, the Guanzi on water, the Hippocratic treatise on airs, waters and places and Clark's soap test. Chapter 8 delivers a weekly water-fit ledger, the distilled Hard-Water Loss and Chelant Dosing Calculator, checker specifications for the four kits and pre-registered tests; a triangle test needs 42 assessors to detect pd = 0.3, and the smoothness study enrols 64. Parameters are illustrative and all calculations are generated by the book's programs. This is not a formula safety assessment or regulatory advice: every formula needs assessment by qualified people, and regulations must be read in their current official text. The book gives no home recipes, rates no brand and makes no claim about skin or hair beyond lather and feel. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:洗护品类流行三句话:硬水让洗发水、沐浴露整瓶失效;配方里多加螯合剂就能在淋浴时把硬水变软;去离子水做底就是高端配方。本书逐条证伪或立界。钙镁离子先压低临界胶束浓度,只在冲洗稀释的中间段才析出钙盐,折损按表活体系分档:皂基最重,烷基硫酸盐次之,酰基谷氨酸盐居中,醚硫酸盐与两性、非离子表活几乎不受影响;示意 W4 水里,烷基硫酸盐在冲洗初段的折损份额为 0.632。螯合剂按摩尔一比一结合钙镁,示意剂量下一次用量只软化 10.5 毫升硬水,连起泡兑水都覆盖不全;全部软化一次淋浴需要配方含量的 1,524 倍,一整瓶只够花洒流约 3.16 秒。核对原文之后,本书做了蓝图修正:覆盖上界比蓝图更紧;硬水与发丝强度的两项小样本体外研究结论相反,不外推任何健康结论;钙离子对阳离子调理聚合物沉积的方向没有直接文献,写成待检验假说。 第一章讲硬度单位、六种示意水样与钙离子活度。第二章讲科林与哈金斯直线、溶度积与胶束保护。第三章把折损换成泡沫高度与半衰期。第四章是主轴:螯合计量与覆盖上界。第五章讲条件稳定常数,酸碱度 5.5 时乙二胺四乙酸降到 5.06、谷氨酸二乙酸降到 2.90,冲洗稀释时螯合剂一档也推迟不了析出。第六章讲阳离子聚合物沉积,屏蔽与桥连两种机制给出的凝聚窗口宽度比为 0.677 与 1.323。 第七章考辨陆羽《茶经》、《管子·水地》、希波克拉底《气候水土论》与克拉克的皂液测定法。第八章给出水质适配周度台账、蒸馏小工具「硬水折损与螯合配比试算器」、四件套检查器规格与预注册检验,三角检验检出 pd = 0.3 需 42 人次,顺滑度评分登记 64 名评价员。书中参数为示意,全部计算由书中程序生成。本书不是配方安全评估,也不是法规意见:任何配方须由合格人员做稳定性、相容性、防腐挑战、斑贴与毒理安全评估,法规以官方现行文本为准;不写在家自制,不点名评价任何品牌,不对皮肤与发丝作泡沫与触感以外的任何宣称。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Four Units of Hardness and Six Illustrative Waters: the Effective Concentration of Calcium in Shower Water — Part B shows the six illustrative waters. W1 holds only 0.4 mmol/L, or 40.0 mg/L, and is soft. W2 is 100.1 mg/L and moderately hard. W3 is 160.1 mg/L and hard. W4, W5 and W6 come to 250.2, 350.3 and 500.4 mg/L, and all three fall into the "very hard" class. W4 is the water used most often in later chapters; it equals 14.02 German degrees or 17.55 Clark degrees. (https://5dlibrary.com/answers/en/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c01/) 第1章 硬度的四种单位与六种示意水样:钙离子在淋浴水里的有效浓度 — B 段是六种示意水样。W1 只有 0.4 毫摩尔每升,折合 40.0 毫克每升,属于软水;W2 是 100.1,属于中等硬度;W3 是 160.1,属于硬水;W4 到 W6 分别是 250.2、350.3 与 500.4,全都落在「极硬」一档。W4 是后面各章最常用的那一种,折合 14.02 德国度或 17.55 克拉克度。 (https://5dlibrary.com/answers/zh/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c01/) - Ch.2 Critical Micelle Concentration, Solubility Products and the Loss Curves of Five Surfactant Classes — The first table reads as follows. Each cell is the total calcium concentration, in millimoles per litre, at which precipitation starts for the stated surfactant level. The fatty-acid soap stays below 0.001 in all four cells: any hard water makes it precipitate. The alkyl sulfate tolerates 10.050 mmol/L of calcium in a dense lather of 100 mmol/L; diluted to 20 mmol/L it tolerates 2.050, and diluted to 5 mmol/L only 0.550. Diluted further to 1 mmol/L, its tolerance climbs back to 1.000. The acyl glutamate gives 12.967, 2.896, 1.007 and 1.259: the same shape, somewhat higher overall. The ether sulfate is the most tolerant of all; even its lowest cell is 50.399. (https://5dlibrary.com/answers/en/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c02/) 第2章 临界胶束浓度、溶度积与五类表活的折损曲线 — 第一张表读法如下。每一格是该表活浓度下,开始析出时所需的总钙浓度,单位毫摩尔每升。脂肪酸皂四格都低于 0.001,任何硬水都让它析出。烷基硫酸盐在 100 毫摩尔每升的浓稠泡沫里能耐 10.050 的钙,稀释到 20 降为 2.050,再稀释到 5 只剩 0.550;继续稀释到 1,反而回升到 1.000。酰基谷氨酸盐的四格是 12.967、2.896、1.007 与 1.259,形状相同,整体略高。醚硫酸盐最耐钙,最低一格仍有 50.399。 (https://5dlibrary.com/answers/zh/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c02/) - Ch.3 From Lost Shares to Foam Height: Precipitated Particles and Foam Half-Life — The first table shows the thin state at 5 mmol/L, which corresponds to the first seconds of rinsing. In soft water all four classes give an initial height of 128.6 mm, because their surfactant concentration is the same. In hard water the four curves separate quickly. The fatty-acid soap falls to 97.3 already in W1, to 39.0 in W3 and to 15.4 in W4, and reaches zero in W6, a ratio of 0.000. The alkyl sulfate is unaffected in W1, falls to 59.0 in W3, 24.4 in W4 and 9.3 in W6, a ratio of 0.073. The acyl glutamate still gives 116.2 in W3, then 80.7 in W4 and 25.5 in W6, a ratio of 0.199. The ether sulfate gives 128.6 in every water, a ratio of 1.000. (https://5dlibrary.com/answers/en/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c03/) 第3章 从折损份额到泡沫高度:析出颗粒与泡沫半衰期 — 第一张表是 5 毫摩尔每升的稀薄状态,对应冲洗最初几秒。软水里四类表活的初始高度都是 128.6 毫米,因为表活浓度相同。进入硬水以后,四条曲线迅速分开。脂肪酸皂在 W1 已降到 97.3,W3 降到 39.0,W4 为 15.4,W6 归零,比值 0.000。烷基硫酸盐在 W1 不受影响,W3 降到 59.0,W4 为 24.4,W6 为 9.3,比值 0.073。酰基谷氨酸盐在 W3 仍有 116.2,W4 为 80.7,W6 为 25.5,比值 0.199。醚硫酸盐四档都是 128.6,比值 1.000。 (https://5dlibrary.com/answers/zh/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c03/) - Ch.4 The Stoichiometry of Chelation and the Ceiling on Coverage: How Much Water Can a Formula's Chelant Soften? — A chelant binds calcium and magnesium one to one, mole for mole. With illustrative inputs, a single dose softens only 10.5 mL of W4 water and does not cover the 40 mL of lathering water; softening the water of one shower would take 1,524 times the chelant the formula carries; and a whole bottle covers three or four seconds of shower flow. The real purpose of a chelant in a formula is stability inside the bottle, not the softening of shower water. (https://5dlibrary.com/answers/en/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c04/) 第4章 螯合计量与覆盖上界:配方里的螯合剂能软化多少水 — 螯合剂按摩尔一比一结合钙镁。示意剂量下,一次用量只够软化 10.5 毫升 W4 水,连 40 毫升起泡兑水都覆盖不全;全部软化一次淋浴需要配方含量的 1,524 倍;一整瓶只够花洒流三四秒。螯合剂在配方里的真正用途是瓶内稳定,而不是软化淋浴水。 (https://5dlibrary.com/answers/zh/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c04/) - Ch.5 Conditional Stability Constants, the Contest between Calcium and Magnesium, and Dilution by Rinsing: Why a Chelant Neither Grips Hard nor Keeps Up — Chapter 5 computes conditional stability constants, which at pH 5.5 fall to 5.06 for EDTA and 2.90 for GLDA, and shows that during rinsing a chelant does not postpone precipitation by a single step. (https://5dlibrary.com/answers/en/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c05/) 第5章 条件稳定常数、钙镁之争与冲洗稀释:螯合剂为何握不紧也追不上 — 第五章讲条件稳定常数,酸碱度 5.5 时乙二胺四乙酸降到 5.06、谷氨酸二乙酸降到 2.90,冲洗稀释时螯合剂一档也推迟不了析出。 (https://5dlibrary.com/answers/zh/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c05/) - Ch.6 The Deposition of Cationic Polymers: Charge Screening, Competition for Sites and the Coacervation Window — Chapter 6 covers the deposition of cationic polymers, where screening and bridging give coacervation-window width ratios of 0.677 and 1.323. (https://5dlibrary.com/answers/en/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c06/) 第6章 阳离子聚合物沉积:电荷屏蔽、位点竞争与凝聚窗口 — 第六章讲阳离子聚合物沉积,屏蔽与桥连两种机制给出的凝聚窗口宽度比为 0.677 与 1.323。 (https://5dlibrary.com/answers/zh/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c06/) - Ch.7 Classical Sources: Lu Yu on Waters, the Guanzi on Water, Hippocrates on Hard Water and Clark's Soap Test — The probability that a mountain water is softer than a river water is 0.8525, against an exact value of 0.8539, so the simulation and the analytic result agree. The probability that a river water is softer than a well water is 0.8464, and that a mountain water is softer than a well water is 0.9802. The probability that the whole order holds for all three is only 0.7025. The mean Kendall rank correlation per set of three is 0.7861. The last line converts rainwater to 5.0 mg/L as calcium carbonate. (https://5dlibrary.com/answers/en/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c07/) 第7章 原典考辨:陆羽评水、《管子·水地》、希波克拉底论硬水与克拉克皂液测定法 — 山水比江水软的概率是 0.8525,解析值 0.8539,两者吻合;江水比井水软的概率 0.8464;山水比井水软的概率 0.9802。三者次序全部成立的概率只有 0.7025。每组三份水的平均肯德尔秩相关系数是 0.7861。最后一行把雨水折成 5.0 毫克每升碳酸钙。 (https://5dlibrary.com/answers/zh/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c07/) - Ch.8 The Empirical Loop: a Weekly Water-Fit Ledger, Field Tools and Checker Specifications, and Pre-Registered Tests — Chapter 8 delivers a weekly water-fit ledger, the distilled Hard-Water Loss and Chelant Dosing Calculator, checker specifications for the four kits and pre-registered tests; a triangle test needs 42 assessors to detect pd = 0.3, and the smoothness study enrols 64. Parameters are illustrative and all calculations are generated by the book's programs. This is not a formula safety assessment or regulatory advice: every formula needs assessment by qualified people, and regulations must be read in their current official text. The book gives no home recipes, rates no brand and makes no claim about skin or hair beyond lather and feel. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c08/) 第8章 实证闭环:水质适配周度台账、实战工具与检查器规格、预注册检验 — 第八章给出水质适配周度台账、蒸馏小工具「硬水折损与螯合配比试算器」、四件套检查器规格与预注册检验,三角检验检出 pd = 0.3 需 42 人次,顺滑度评分登记 64 名评价员。书中参数为示意,全部计算由书中程序生成。本书不是配方安全评估,也不是法规意见:任何配方须由合格人员做稳定性、相容性、防腐挑战、斑贴与毒理安全评估,法规以官方现行文本为准;不写在家自制,不点名评价任何品牌,不对皮肤与发丝作泡沫与触感以外的任何宣称。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/water-hardness-chelation-surfactant-cmc-foam-deposition-in-shower-softening/c08/) ## 5DL-B-390 · Apparel Graphic Gestalt Capture, Long-Range Contrast Legibility and Streetwear Visual Memetic Retention 中文书名:服饰图腾完形视知觉捕获、远距高对比辨识与街头视觉模因驻留母典 DOI: pending Cite: Fan, Y. (2026). Apparel Graphic Gestalt Capture, Long-Range Contrast Legibility and Streetwear Visual Memetic Retention (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. Not trademark or copyright advice; graphics are the author's geometric forms, no real brand is named or judged, no subliminal claim is made, and street tests record anonymous answers only, never faces or personal data. Abstract: Streetwear lore says that a graphic seen three times is stamped on memory, that more detail catches more eyes, and that brighter colours jump out of a crowd. This book rewrites each saying as a refutable proposition and bounds it with standard-library programs. Blueprint corrections: Krugman's "three" (1972) describes psychological exposures, not a measured threshold; mere-exposure liking is an inverted U for visual stimuli; millisecond recognition results are laboratory findings, not street capture probabilities; contrast sensitivity peaks near 4 cycles per degree; and the web accessibility contrast formula is used on cloth only as an analogy. Parameters are illustrative and all calculations are generated by the book's programs. Not trademark or copyright advice; graphics are the author's geometric forms, no real brand is named or judged, no subliminal claim is made, and street tests record anonymous answers only, never faces or personal data. Chapter 1 works out the geometry of visual angle: at fifteen metres a one-arcminute stroke is 4.36 mm, the parts of a graphic vanish in order of width, and fabric folds narrow strokes by nearly a third in one view out of ten. Chapter 2 converts pattern periods into cycles per degree; 4 mm hatching keeps only 0.092 of its modulation at five metres, while the silhouette survives as average luminance. Chapter 3 computes relative luminance from sRGB: hot pink on signal red has a contrast of only 1.36, a distant graphic becomes its linear-light average, and white ink on black needs an ink share of only 0.123 to reach the 3:1 analogy band whereas black ink on white needs 0.708. Chapter 4 uses small centre-surround grids and an illustrative logistic model to show that vivid colour catches the eye only with luminance contrast, and that the unit picked out at a distance is the colour block of the whole garment. Chapter 5 shows distance closing gaps and merging letter spacing. Chapter 6 rewrites "stamped on memory" as a recognition probability: with the same three exposures, spread exposure retains 2.38 times as much as massed exposure after a month, and among street encounters 0.311 of passers-by truly see a high-contrast graphic three times against only 0.027 for a low-contrast one. Chapter 7 examines Xie He's bone method, Han Feizi on painting dogs and horses, Leonardo on the small parts that vanish first and the opening sentence of Wertheimer's Gestalt paper, and distils the results for four operational kits. Chapter 8 delivers the Apparel Graphic Long-Range Legibility and Exposure Threshold Calculator, checker specifications grouped by kit, a weekly ledger and a pre-registered test for which a single-graphic check needs about 69 anonymous respondents. Boxes separate evidence, runnable Python (22 programs), classical sources and hypotheses. 摘要:街头服饰圈流传三句话:图案让路人看三次就会深印脑海;细节越丰富越高级、越抓眼;颜色越艳越能从人群里跳出来。本书把三句话逐条改写成可被数据推翻的命题,并用纯标准库程序算出上下界。核对原文后做了蓝图修正:「三次」出自克鲁格曼 1972 年的解释性短文,说的是心理上的三次,不是测出的阈值;单纯曝光效应的好感先升后降,视觉刺激呈倒 U 形,峰值随条件移动;快速识别的毫秒级结果来自实验室刺激,不等于街头抓眼概率;对比敏感度在每度约 4 周处最高;网页无障碍指南的亮度对比公式只作为布料上的类比尺。书中参数为示意,全部计算由书中程序生成。本书不是商标或著作权法律意见,示意图案全部为作者自拟几何形,不点名评价任何真实品牌,不写任何潜意识影响宣称,街头测试只记录匿名对错,不采集人脸与个人资料。 第一章算视角几何:十五米外一角分笔画是 4.36 毫米,图案部件按宽度从小到大依次消失,布料褶皱让一成视角把笔画压窄近三成。第二章把图案周期折成每度周数,4 毫米排线五米外只剩 0.092 的调制,剪影却以平均亮度留下。第三章从 sRGB 算相对亮度:玫红印信号红的对比只有 1.36,远看图案化成线性光平均,白墨印黑衣只要 0.123 的墨色占比就达到 3:1 的类比档,黑墨印白衣要 0.708。第四章用中心与周边差分的小网格与示意逻辑斯蒂模型说明,艳色要配上亮度对比才抓眼,远处被挑中的单位是整件衣服的色块。第五章算距离怎样合上缺口、压并字距,紧凑闭合的形状最能保住轮廓。 第六章把「深印脑海」改写成再认概率:同样三次,分散曝光一个月后的保持是集中曝光的 2.38 倍;街头偶遇里,高对比图案有 0.311 的路人真正看见三次,低对比图案只有 0.027。第七章考辨谢赫骨法用笔、韩非子画犬马、达芬奇论远处先失小部件与韦特海默的完形开篇句,并给出四件套蒸馏规格。第八章交出蒸馏小工具「服饰图腾远距辨识与曝光阈值试算器」、按四件套分组的检查器规格、周度台账与预注册检验,单款检验约需 69 名匿名受访者。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Geometry of Visual Angle: How Many Millimetres a Stroke Keeps at Fifteen Metres — Chapter 1 works out the geometry of visual angle: at fifteen metres a one-arcminute stroke is 4.36 mm, the parts of a graphic vanish in order of width, and fabric folds narrow strokes by nearly a third in one view out of ten. (https://5dlibrary.com/answers/en/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c01/) 第1章 视角几何:十五米外,一道笔画还剩几毫米 — 第一章算视角几何:十五米外一角分笔画是 4.36 毫米,图案部件按宽度从小到大依次消失,布料褶皱让一成视角把笔画压窄近三成。 (https://5dlibrary.com/answers/zh/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c01/) - Ch.2 Spatial Frequency: Fine Stripes Leave First, the Silhouette Leaves Last — Chapter 2 converts pattern periods into cycles per degree; 4 mm hatching keeps only 0.092 of its modulation at five metres, while the silhouette survives as average luminance. (https://5dlibrary.com/answers/en/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c02/) 第2章 空间频率:细条纹先走,剪影最后离场 — 第二章把图案周期折成每度周数,4 毫米排线五米外只剩 0.092 的调制,剪影却以平均亮度留下。 (https://5dlibrary.com/answers/zh/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c02/) - Ch.3 Luminance Contrast: Vivid Is Not the Same as Contrasting — Chapter 3 computes relative luminance from sRGB: hot pink on signal red has a contrast of only 1.36, a distant graphic becomes its linear-light average, and white ink on black needs an ink share of only 0.123 to reach the 3:1 analogy band whereas black ink on white needs 0.708. (https://5dlibrary.com/answers/en/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c03/) 第3章 亮度对比:艳色不等于对比 — 第三章从 sRGB 算相对亮度:玫红印信号红的对比只有 1.36,远看图案化成线性光平均,白墨印黑衣只要 0.123 的墨色占比就达到 3:1 的类比档,黑墨印白衣要 0.708。 (https://5dlibrary.com/answers/zh/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c03/) - Ch.4 Salience: Which Item Is Picked Out Within a Glance — D1 is an emblem 26 centimetres across with a silhouette contrast of 9.13 and a closure of 0.9, corresponding to white ink at half coverage on a black shirt in Chapter 3. Its illustrative probability is 0.877 at one metre, 0.590 at five, 0.418 at ten and 0.324 at fifteen. D2 has the same size and closure but the contrast of hot pink on red, 1.36, and keeps only 0.121 at ten metres. D3 shrinks the emblem to 12 centimetres and reaches 0.249 at ten metres. D4 lowers closure to 0.3 and reaches 0.226 at ten metres. D5 is a large graphic of 40 centimetres with a contrast of 4.25 and a closure of 0.6; at ten metres it reaches 0.266, so size alone did not buy much. (https://5dlibrary.com/answers/en/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c04/) 第4章 显著性:一瞥之内,谁先被挑中 — D1 是 26 厘米、剪影对比 9.13、闭合度 0.9 的徽记,对应第三章白墨半版印黑衣。它在一米外的示意概率是 0.877,五米 0.590,十米 0.418,十五米 0.324。D2 尺寸与闭合度相同,对比换成玫红印红的 1.36,十米外只剩 0.121。D3 把尺寸缩到 12 厘米,十米外 0.249。D4 把闭合度降到 0.3,十米外 0.226。D5 是 40 厘米的大图案,对比 4.25、闭合度 0.6,十米外 0.266,大并没有换来更多。 (https://5dlibrary.com/answers/zh/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c04/) - Ch.5 Gestalt Closure: Distance Closes the Gap for the Designer — Distance closes gaps for the designer. On a ring 200 mm across, a notch of 2 degrees closes at 12.00 m, and a notch of 5 degrees reads as open out to 10.00 m. Letter spacing, bar spacing and group spacing merge in order along the same ruler: letter spacing of 2.5 mm holds the three-arcminute threshold only to 2.86 m. Compact, closed shapes best keep their outlines, while a thin ring and a thin cross start to be lost at ten metres, and to keep half the contrast at fifteen metres a part needs to be at least 5.89 mm wide. (https://5dlibrary.com/answers/en/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c05/) 第5章 完形闭合:距离替设计师合上缺口 — 距离会替设计师合上缺口。外径 200 毫米的圆环,2 度的缺口在 12.00 米处合上,5 度的缺口能读成开口到 10.00 米。字距、条距与组距按同一把尺依次合并:2.5 毫米的字距三角分只撑到 2.86 米。紧凑闭合的形状最能保住轮廓,细环与细十字在十米外开始丢失,十五米外要保住一半对比,部件至少 5.89 毫米宽。 (https://5dlibrary.com/answers/zh/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c05/) - Ch.6 Retention in Memory: Three Encounters Are Not Three Exposures — Chapter 6 rewrites "stamped on memory" as a recognition probability: with the same three exposures, spread exposure retains 2.38 times as much as massed exposure after a month, and among street encounters 0.311 of passers-by truly see a high-contrast graphic three times against only 0.027 for a low-contrast one. (https://5dlibrary.com/answers/en/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c06/) 第6章 记忆驻留:三次偶遇不等于三次曝光 — 第六章把「深印脑海」改写成再认概率:同样三次,分散曝光一个月后的保持是集中曝光的 2.38 倍;街头偶遇里,高对比图案有 0.311 的路人真正看见三次,低对比图案只有 0.027。 (https://5dlibrary.com/answers/zh/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c06/) - Ch.7 Classical Sources: The Bone Method, Dogs and Horses, and the Small Parts That Vanish First — A leg 0.10 m wide drops below one arcminute at 343.8 m; the head, 0.25 m wide, at 859.4 m; the neck, 0.30 m, at 1,031.3 m; the body, 0.70 m deep, at 2,406.4 m; and the whole length of 2.40 m at 8,250.6 m. The order of disappearance is leg, head, neck, body, whole length, in full agreement with Leonardo's "legs before head". The vanishing distance of the whole length is 24 times that of a leg. (https://5dlibrary.com/answers/en/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c07/) 第7章 原典考辨:骨法、犬马与远处先失的小部件 — 腿宽 0.10 米,在 343.8 米外跌破一角分;头宽 0.25 米,到 859.4 米;颈宽 0.30 米,到 1,031.3 米;躯干厚 0.70 米,到 2,406.4 米;全长 2.40 米,到 8,250.6 米。消失顺序是腿、头、颈、躯干、全长,与达芬奇说的「先失腿,后失头」完全一致。全长与腿的消失距离之比是 24 倍。 (https://5dlibrary.com/answers/zh/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c07/) - Ch.8 Field Tools and Checker Specifications: A Weekly Long-Range Legibility Ledger and a Pre-Registered Test — Chapter 8 delivers the Apparel Graphic Long-Range Legibility and Exposure Threshold Calculator, checker specifications grouped by kit, a weekly ledger and a pre-registered test for which a single-graphic check needs about 69 anonymous respondents. Boxes separate evidence, runnable Python (22 programs), classical sources and hypotheses. (https://5dlibrary.com/answers/en/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c08/) 第8章 实战工具与检查器规格:周度远距辨识台账与预注册检验 — 第八章交出蒸馏小工具「服饰图腾远距辨识与曝光阈值试算器」、按四件套分组的检查器规格、周度台账与预注册检验,单款检验约需 69 名匿名受访者。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/apparel-graphic-gestalt-capture-long-range-contrast-legibility-memetic-retention/c08/) ## 5DL-B-391 · Multi-Sensory Luxury Unboxing Engineering, Lid-Friction Air-Damping Physics and Haptic Weight Price Anchoring 中文书名:高奢包装五感开箱工程、天地盖阻尼排气物理与触觉重量溢价母典 DOI: pending Cite: Fan, Y. (2026). Multi-Sensory Luxury Unboxing Engineering, Lid-Friction Air-Damping Physics and Haptic Weight Price Anchoring (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Errors in the gap are magnified by the cube, so a conventional narrow-gap design keeps only 0.897 of its boxes inside the illustrative window; the target is a window, not "slower is better". Weight is a cue to quality, but people weigh the difference from expectation, and the score of an illustrative expectation model peaks at a ratio of actual to expected weight of 1.416. Abstract: Three sentences circulate among people who package expensive goods: the slower a telescoping lid sinks, the more luxurious the box; the heavier the box, the more precious it seems; and the more one invests in the senses, the more customers will pay. This book refutes or bounds each. As a lid descends, the air inside can leave only through the slit between skirt and base, and plane Poiseuille flow gives a closed-form drop-time equation: for one illustrative box the lid takes 17.080 seconds at a gap of 0.20 mm and only 1.093 seconds at 0.50 mm, a factor of 15.625. Errors in the gap are magnified by the cube, so a conventional narrow-gap design keeps only 0.897 of its boxes inside the illustrative window; the target is a window, not "slower is better". Weight is a cue to quality, but people weigh the difference from expectation, and the score of an illustrative expectation model peaks at a ratio of actual to expected weight of 1.416. The honest uses of weight are stability and stiffness: seating a bottle 20 mm lower adds 2.79 degrees to the tipping angle, and hidden ballast is always flagged red. Packaging investment shows diminishing returns; in the illustrative case the turning point lies at 12.876, and doubling the investment lowers the margin by 6.476. Chapter 1 derives the drop-time equation and checks it by numerical integration; Chapter 2 treats tolerance, the just-noticeable difference and paired comparisons; Chapter 3 the slide-angle method, roughness and sticking skirts; Chapter 4 weight, the centre of gravity, board stiffness and the size-weight illusion; Chapter 5 tear-strip acoustics and the discrete Fourier transform; Chapter 6 the turning point of investment, a checker for the over-packaging limits of the Chinese standard GB 23350-2021 and EU Regulation 2025/40, and the rule that anchors must be genuine references: a fictitious original price makes regret purchases 0.481 of all purchases. Chapter 7 reads the Han Feizi story of the casket and the pearl, the Kaogongji and Veblen. Chapter 8 gives a weekly five-sense unboxing ledger, a distilled tool, checker specifications grouped by four kits and a pre-registered test plan whose main test needs at least 199 people. All dimensions, weights, costs, preferences and willingness to pay are illustrative parameters that describe no real brand, product or sale; costs are in illustrative units without currency, and all calculations are generated by the book's programs. This is not legal advice on packaging regulation or consumer protection, nor personalised business advice; regulatory texts must be read in their current official versions and will be revised. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:高客单商品的包装圈流行三句话:天地盖越慢落下越高级,盒子越重越显贵,五感投入越多顾客越肯多付钱。本书逐条证伪或立界。盖子下落时,盒内空气只能从盖裙与地盒之间的窄缝排出,狭缝泊肃叶流给出闭式下落时间方程:同一只示意盒子,缝宽 0.20 毫米时落 17.080 秒,0.50 毫米时只要 1.093 秒,相差 15.625 倍;缝宽误差按三次方放大,常规窄缝方案只有 0.897 的盒子落在示意可行区间里,所以要找的是区间而不是「越慢越好」。重量是品质线索,但人掂的是重量与预期之差,示意期望模型的品质分在实际与预期之比 1.416 处见顶;重量的正当用处是稳与硬,瓶子坐低 20 毫米让倾倒角多出 2.79 度,藏配重一律判红。包装投入与支付意愿边际递减,示意情形的拐点在 12.876,投入加倍反使毛利少 6.476。 第一章推导下落时间方程并用数值积分核对;第二章写公差、最小可觉差与成对比较;第三章写斜面法、粗糙度与盖裙卡滞;第四章写重量、重心、板材刚度与大小重量错觉;第五章写撕封声学与离散傅里叶变换;第六章求投入拐点,把国家标准 GB 23350-2021 与欧盟 2025/40 的过度包装红线写成检查器,并说明锚定只用真实对照:虚构原价让后悔购买占购买者 0.481。第七章考辨《韩非子》买椟还珠、《考工记》与凡勃伦。第八章给出开箱五感周度台账、蒸馏小工具、按四件套分组的检查器规格与预注册检验方案,主检验至少需要 199 人。 书中全部尺寸、重量、成本、偏好与支付意愿均为示意参数,不对应任何真实品牌、商品或成交,成本与支付意愿一律用示意单位,不写币种;全部计算由书中程序生成。本书不是包装法规或消费者保护方面的法律意见,也不是个人化经营建议,法规条文以官方现行文本为准并会修订。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Lid Air Damping: Plane Poiseuille Flow in the Slit and the Drop-Time Equation — The drop time of a telescoping lid is governed by plane Poiseuille flow in the slit, with the closed form $T = 12\mu A^2(L_0 S + S2/2)/(w h^3 m g)$. Between gaps of 0.20 mm and 0.50 mm the drop time changes by a factor of 15.625; the log-log slope is -3.000; the numerical solution with inertia differs from the closed form by milliseconds; above 0.50 mm the losses at entry and exit must be counted. A heavier lid falls faster, which contradicts the intuition that a lid should be both heavy and slow. (https://5dlibrary.com/answers/en/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c01/) 第1章 天地盖排气:狭缝泊肃叶流与下落时间方程 — 天地盖的下落时间由狭缝泊肃叶流决定,闭式解为 $T = 12\mu A^2(L_0 S + S2/2)/(w h^3 m g)$。缝宽 0.20 毫米与 0.50 毫米之间,下落时间相差 15.625 倍;对数斜率为 -3.000;带惯性的数值解与闭式解之差只有毫秒量级;缝宽超过 0.50 毫米,进出口损失必须计入。盖子越重落得越快,这一条与「又重又慢才高级」的直觉相冲突。 (https://5dlibrary.com/answers/zh/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c01/) - Ch.2 Slower Is Not Better: Cubic Amplification of Tolerance and the Feasible Window of Drop Time — Errors in the gap pass into the drop time with a power of three. A narrow-gap design with a nominal drop time of 5.061 seconds, under a standard deviation of 0.03 mm, has fifth and ninety-fifth percentiles that differ by a factor of 2.687, with a share of 0.897 inside the window. Widening the gap while shortening the open perimeter narrows the scatter markedly, but the losses at entry and exit rise, and design C is the sensible compromise. At an illustrative Weber fraction of 0.15, two narrow-gap boxes from the same batch can be told apart with probability 0.745. Paired comparisons and the Bradley-Terry model turn "how slow is good" into a testable inverted-U hypothesis. (https://5dlibrary.com/answers/en/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c02/) 第2章 慢不等于好:公差立方放大与下落时间可行区间 — 缝宽误差按三次方传到下落时间上。名义 5.061 秒的窄缝方案,在 0.03 毫米标准差下,第五与第九十五百分位相差 2.687 倍,区间内比例 0.897;放宽缝宽、缩短走气周长后,离散明显收窄,但进出口损失上升,方案 C 是较稳妥的折中。按 0.15 的示意韦伯分数,同批两只窄缝盒子有 0.745 的概率被分出快慢。成对比较与布拉德利-特里模型可以把「慢多少才好」变成可检验的倒 U 形假说。 (https://5dlibrary.com/answers/zh/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c02/) - Ch.3 Paper Friction and Touch: the Slide-Angle Method, Roughness and a Sticking Skirt — Friction arises from real contact at asperities, and its coefficient belongs to a pair of materials. The slide-angle method measures the static coefficient as $\tan\theta$; the steeper the angle, the more errors are magnified, so grippy surfaces such as fabric need more readings. The illustrative data show touch ratings following an inverted U in roughness, which a straight line misreads as "the smoother the better". When a skirt rubs, friction takes part of the weight and the drop lengthens by $mg/(mg - \mu_k N)$; at a side load of 5.450 N the lid sticks, and it grows ever less stable as it approaches that point. (https://5dlibrary.com/answers/en/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c03/) 第3章 纸面摩擦与触感:斜面法、粗糙度与盖裙卡滞 — 摩擦来自微凸体上的真实接触,系数是材料对的性质。斜面法用 $\tan\theta$ 测静摩擦系数,角度越陡误差放大越多,布面这类涩表面要多读几次。示意数据显示触感评分随粗糙度呈倒 U 形,直线拟合会误读成「越光滑越好」。盖裙擦壁时,摩擦分走重量,下落时间按 $mg/(mg - \mu_k N)$ 拉长,侧压 5.450 牛时盖子卡死,越接近卡死点越不稳定。 (https://5dlibrary.com/answers/zh/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c03/) - Ch.4 Weight, Centre of Gravity and Expectation: Honest Heft and the Size-Weight Illusion — Weight is a cue to quality, and both experiments and observational studies support that; but the evidence compares only a few levels of weight, and its effect on willingness to pay is indirect. People weigh the difference between weight and expectation, as both Charpentier's size-weight illusion and Flanagan's lifting experiments show. The quality score of the illustrative expectation model peaks at a ratio of 1.416. The honest uses of weight are stability and stiffness: seating the bottle 20 mm lower raises the tipping angle by 2.79 degrees, and going from 1.5 mm to 2.5 mm board shrinks the dip under a press to 0.216 of its value. Hidden ballast is mechanically indistinguishable from real structure, but it misleads, and it is flagged red without exception. (https://5dlibrary.com/answers/en/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c04/) 第4章 重量、重心与期望:真材实料的重与大小重量错觉 — 重量是品质线索,实验与观察研究都支持,但现有证据只比较了少数几个重量水平,对支付意愿的作用是间接的。人掂的是重量与预期之差,夏庞蒂埃的大小重量错觉与弗拉纳根的提举实验都说明这一点。示意期望模型的品质分在比值 1.416 处见顶。重量的正当用处是稳与硬:瓶子坐低 20 毫米让翻倒角增加 2.79 度,灰板从 1.5 毫米加到 2.5 毫米让按压凹陷缩为 0.216 倍。藏配重在力学上与真实结构无异,但属误导,一律判红。 (https://5dlibrary.com/answers/zh/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c04/) - Ch.5 The Acoustics of Tearing a Seal: Perforation Pitch, the Discrete Fourier Transform and Sound Ratings — The dominant frequency of a tear is $f_0 = v/p$: the pitch is supplied by the designer, the speed by the customer. Summed directly from its definition, the discrete Fourier transform is fast enough on 1,024 samples and locates the 250.0 Hz fundamental and the 546.5 Hz spectral centroid exactly. The illustrative model shows that sound has a real but modest effect on ratings: one octave is worth only 0.533 of the noise standard deviation, and the rater-level $R^2$ is only 0.295. Carrying the crisp experiments over to packaging is an analogy that must be tested with one's own data. (https://5dlibrary.com/answers/en/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c05/) 第5章 撕封声学:穿孔节距、离散傅里叶与声音评分 — 撕封声的主频由 $f_0 = v/p$ 决定:节距是设计者给的,速度是顾客给的。离散傅里叶变换直接按定义求和,在 1,024 个样点上足够快,能准确找到 250.0 赫兹的主频与 546.5 赫兹的谱质心。示意模型显示,声音对评分有真实但不大的作用,一个倍频程只相当于噪声标准差的 0.533,评分者层面 $R^2$ 只有 0.295。薯片实验迁移到包装,是类比,要用自己的数据检验。 (https://5dlibrary.com/answers/zh/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c05/) - Ch.6 The Turning Point of Packaging Investment, the Red Lines Against Over-Packaging and Genuine Reference Prices — Chapter 6 the turning point of investment, a checker for the over-packaging limits of the Chinese standard GB 23350-2021 and EU Regulation 2025/40, and the rule that anchors must be genuine references: a fictitious original price makes regret purchases 0.481 of all purchases. (https://5dlibrary.com/answers/en/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c06/) 第6章 投入拐点、过度包装红线与真实对照锚定 — 第六章求投入拐点,把国家标准 GB 23350-2021 与欧盟 2025/40 的过度包装红线写成检查器,并说明锚定只用真实对照:虚构原价让后悔购买占购买者 0.481。 (https://5dlibrary.com/answers/zh/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c06/) - Ch.7 Classical Sources: the Casket and the Pearl, the Kaogongji and Veblen — A swing of 10 percentage points in humidity moves the gap by 0.030 mm. If the gap widens, the drop time falls to 3.802 seconds; if it narrows, it rises to 6.942 seconds. A swing of 30 percentage points moves the gap by 0.090 mm, and the drop ranges from 2.304 seconds at fastest to 14.755 seconds at slowest, beyond both ends of the illustrative window of Chapter 2. The same box can behave like two different boxes in the dry season and the wet one. That is "heaven has its seasons": material and craft unchanged, but the season changes how the object performs. The program also records the threshold used later in this chapter's hypothesis, a relative change of 0.15, the illustrative Weber fraction of Chapter 2. (https://5dlibrary.com/answers/en/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c07/) 第7章 原典考辨:买椟还珠、《考工记》与凡勃伦 — 湿度变化 10 个百分点,缝宽漂移 0.030 毫米:缝变宽时下落时间降到 3.802 秒,变窄时升到 6.942 秒。湿度变化 30 个百分点,缝宽漂移 0.090 毫米:最快 2.304 秒,最慢 14.755 秒,已经跨出第二章的示意区间两端。同一只盒子,干燥季节与潮湿季节,可以像两只不同的盒子。这就是「天有时」:材料与工艺都没变,时令一变,器物表现就变。 (https://5dlibrary.com/answers/zh/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c07/) - Ch.8 The Empirical Loop: a Weekly Five-Sense Unboxing Ledger and Field Tools and Checker Specifications — Chapter 8 gives a weekly five-sense unboxing ledger, a distilled tool, checker specifications grouped by four kits and a pre-registered test plan whose main test needs at least 199 people. (https://5dlibrary.com/answers/en/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c08/) 第8章 实证闭环:开箱五感周度台账与实战工具与检查器规格 — 第八章给出开箱五感周度台账、蒸馏小工具、按四件套分组的检查器规格与预注册检验方案,主检验至少需要 199 人。 (https://5dlibrary.com/answers/zh/luxury-unboxing-lid-air-damping-haptic-weight-price-anchoring/c08/) ## 5DL-B-392 · Cross-Modal Coupling of Acoustic Rhythms and Olfactory Cues, Proustian Memory Retrieval and Rapid State-Conditioning Anchors 中文书名:声学节律与嗅觉分子跨模态耦合、普鲁斯特记忆唤醒与瞬时状态锚定母典 DOI: pending Cite: Fan, Y. (2026). Cross-Modal Coupling of Acoustic Rhythms and Olfactory Cues, Proustian Memory Retrieval and Rapid State-Conditioning Anchors (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations come from the book's programs. This is not medical, psychological or regulatory advice and claims no effect on body or mind. Abstract: Spatial branding repeats three claims: smell goes straight to the emotional centre and changes a person's state instantly; a sound and a scent presented together a few times plant an instant anchor; and stronger scent with livelier music works better. This book checks each claim and refutes it or sets bounds on it. Odour-evoked memories feel more emotional, but there is no evidence that they are more accurate; the effect is a shift of a distribution, not a switch, and it peaks before the age of ten. The "anchoring" of neuro-linguistic programming lacks empirical support: among the relevant empirical studies counted by Witkowski, only 27.3 % were supportive. The testable mechanism is evaluative conditioning, with a meta-analytic mean effect of d = 0.52 that is larger when people are aware of the pairing. Matching the arousal of fragrance and music matters more than raising it; doubling a concentration buys only about 1.464 times the perception, and olfactory adaptation lowers that within minutes. After checking the sources the book records its blueprint corrections: stronger emotion, unverified accuracy; Chu and Downes appeared in Cognition; Witkowski is read by the counts in the body of the paper; pitch-odour matching is driven by pleasantness and complexity; the current IFRA Standards are the 51st Amendment. Parameters are illustrative and all calculations come from the book's programs. This is not medical, psychological or regulatory advice and claims no effect on body or mind. Chapter 1 covers the olfactory pathway and the Proust effect; Chapter 2 the age peak of odour memory and how to test it when the distribution is bimodal. Chapter 3 uses first-order evaporation, Raoult's law and a box model to time note hand-overs and expose over-diffusion; with illustrative parameters the top note hands over at 182.2 minutes. Chapter 4 places musical borders with zero error by dynamic programming, shows by simulation that the hand-over moves between 135.8 and 246.1 minutes from the 5th to the 95th percentile, and checks volume against the WHO safe-listening standard. Chapter 5 computes cross-modal congruence as a centred cosine. Chapter 6 rewrites the "instant anchor" as Rescorla-Wagner learning and extinction curves. Chapter 7 examines the Record of Music, the Jun Chen chapter of the Book of Documents, Baudelaire's Correspondances and Proust's madeleine passage, and ends with the four-kit distillation specification. Chapter 8 gives the distilled Sound-and-Scent Signature Alignment and Congruence Calculator, checker specifications grouped by the four kits, a ledger and a pre-registered test needing 276 people per group. Scent and music in a space must be announced in advance, with a way to leave and care for people sensitive to smell, and no design aimed at minors. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:空间品牌圈流行三句话:气味直通情绪中枢,闻到就能瞬间改变状态;声音和香味同时出现几次,就能建立瞬时锚点;香越浓、音乐越嗨,效果越好。本书逐条核对并证伪或立界。气味唤起的记忆情绪更浓,却没有证据显示更准确,效应是分布的平移而不是开关,峰值落在十岁以前。神经语言程序学所说的「锚定」缺乏实证:维特科夫斯基统计的相关实证研究中,支持者只占 27.3 %;可检验的机制是评价性条件作用,元分析平均效应 d 为 0.52,参与者意识到配对时更大。香氛与音乐唤醒度相配比强度高更要紧,浓度翻倍只换来约 1.464 倍感受,嗅觉适应还会在几分钟内压低它。核对原文后,本书写明蓝图修正:情绪更强、准确度未经证实;朱与唐斯论文发表在《认知》;维特科夫斯基一文以正文计数为准;音高与气味的对应由愉悦度与复杂度驱动;IFRA 现行为第 51 次修订。书中参数为示意,全部计算由书中程序生成。本书不是医疗、心理咨询或法规意见,不宣称任何身心功效。 第一章讲嗅觉通路与普鲁斯特效应,第二章讲气味记忆的年龄峰与双峰分布下的检验口径。第三章用一阶蒸发、拉乌尔定律与箱体模型算香调交班与超量扩香,示意参数下前调在 182.2 分钟交给中调。第四章用动态规划把乐段边界排到零误差,又用模拟显示交班时刻日间第 5 到第 95 百分位跨越 135.8 到 246.1 分钟,并按世卫安全聆听标准核对音量。第五章以居中余弦计算跨感官一致度。第六章用雷斯科拉-瓦格纳规则把「瞬时锚定」改写为习得与消退曲线。 第七章考辨《礼记·乐记》、《尚书·君陈》、波德莱尔《应和》与普鲁斯特玛德莲段落,章末给出四件套蒸馏规格。第八章给出蒸馏小工具「声香签名对齐与跨感官一致度试算器」、按四件套分组的检查器规格、台账与预注册检验,主检验每组 276 人。空间投放香氛与音乐须事先公示、可选择离开、顾及气味敏感人群,不针对未成年人。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Olfactory Pathway and the Proust Effect: a Dial, Not a Switch — The primary olfactory route has no thalamic relay; there is anatomical support for that. Odour-evoked memories feel more emotional and have been visited less often; there is experimental support for that. The step from there to "you smell it and your state changes instantly" skips two links: the effect is a shift, not a switch, and it lands on emotional intensity, not on accuracy. Program p11 shows that pushing most people past the upper range of the control condition would need effects far larger than experiments usually find; program p12 shows that a clear difference in emotion can coexist with no difference at all in accuracy. (https://5dlibrary.com/answers/en/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c01/) 第1章 气味通路与普鲁斯特效应:旋钮而非开关 — 嗅觉初级通路不经丘脑中继,这一点有解剖依据;气味唤起的记忆情绪更浓、久未触碰,这一点有实验依据。可从这里跳到「闻到就会瞬间改变状态」,中间缺了两环:效应是平移而不是开关,效应落在情绪强度而不是准确度。程序 p11 说明,要让多数人都越过对照组的高位,所需效应量远超常见实验;程序 p12 说明,情绪差清晰时,准确度差可以完全为零。 (https://5dlibrary.com/answers/zh/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c01/) - Ch.2 The Reminiscence Bump: Which Part of Life a Smell Returns You To — Odour cues move the centre of autobiographical memory about a decade earlier: the peak falls before the age of ten, while word cues peak in adolescence and early adulthood. This is a shift in a distribution, not a switch: recent memories are equally common under both cues. Program p22 shows that with a bimodal distribution the median is useless and a share is the right summary; program p23 shows that the required sample grows with the inverse square of the gap between groups. For brands, the path by which scent helps memory is more likely attention than the revival of the past. (https://5dlibrary.com/answers/en/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c02/) 第2章 年龄峰:气味把人带回哪一段人生 — 气味线索把自传记忆的分布重心向前推了约十年,峰值落在十岁以前,词语线索的峰值则在青少年到成年早期。这是分布的平移,不是开关:近期记忆在两种线索下同样多。程序 p22 显示,双峰分布下中位数毫无用处,占比才是合适的指标;程序 p23 显示,组间差距越小,所需样本按平方增长。品牌层面,香氛帮助记忆的路径更可能是注意,而不是唤起往事。 (https://5dlibrary.com/answers/zh/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c02/) - Ch.3 The Evaporation Timeline: When Top, Heart and Base Notes Hand Over — Chapter 3 uses first-order evaporation, Raoult's law and a box model to time note hand-overs and expose over-diffusion; with illustrative parameters the top note hands over at 182.2 minutes. (https://5dlibrary.com/answers/en/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c03/) 第3章 挥发时间轴:前中后调何时交班 — 第三章用一阶蒸发、拉乌尔定律与箱体模型算香调交班与超量扩香,示意参数下前调在 182.2 分钟交给中调。 (https://5dlibrary.com/answers/zh/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c03/) - Ch.4 Beat Alignment: Scheduling Musical Sections Against Scent Hand-overs — Chapter 4 places musical borders with zero error by dynamic programming, shows by simulation that the hand-over moves between 135.8 and 246.1 minutes from the 5th to the 95th percentile, and checks volume against the WHO safe-listening standard. (https://5dlibrary.com/answers/en/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c04/) 第4章 节拍对位:音乐段落与香调交班的排程 — 第四章用动态规划把乐段边界排到零误差,又用模拟显示交班时刻日间第 5 到第 95 百分位跨越 135.8 到 246.1 分钟,并按世卫安全聆听标准核对音量。 (https://5dlibrary.com/answers/zh/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c04/) - Ch.5 Cross-Modal Congruence: Fit Matters More Than Force — Crossmodal correspondences are stable, but what drives them is first of all liking and complexity, so a congruence calculation needs a separate pleasantness dimension. Cosines on positive rating scales must be centred first, or every combination appears to fit. Field experiments and the illustrative model point the same way: matching arousal matters more than raising it; high-with-high and low-with-low hardly differ, and a crossed combination can do worse than a single cue. The power law and olfactory adaptation show that doubling the concentration buys only about 1.464 times the perception, which adaptation then lowers within minutes; "stronger is better" usually grows out of the operator's own adaptation. (https://5dlibrary.com/answers/en/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c05/) 第5章 跨感官一致度:相配比浓烈更要紧 — 跨感官对应是稳定的,但驱动它的首先是好恶与复杂度,计算一致度时须单列愉悦度维度。正数量表上的余弦必须先居中,否则所有组合都显得相配。零售现场实验与示意模型都指向同一件事:唤醒度相配比唤醒度高更要紧,高配高与低配低几乎没有差别,交叉搭配可能比单一线索更差。心理物理幂律与嗅觉适应说明,浓度翻倍只换来约 1.464 倍感受,几分钟后又被适应压低;「越浓越好」往往源于经营者自己的适应。 (https://5dlibrary.com/answers/zh/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c05/) - Ch.6 From "Instant Anchors" to Learning Curves: the Ledger of Evaluative Conditioning — NLP "anchoring" lacks empirical support: of the 33 relevant studies counted by Witkowski, the supportive share is only 0.273. The testable alternative is evaluative conditioning, with a meta-analytic mean effect of d = 0.52 that is larger when participants are aware of the pairing. The Rescorla-Wagner rule turns "a few pairings" into a learning curve: at a learning rate of 0.10 it takes 7 pairings to pass one half, and unpleasant pairings pull the strength back by the same rule. Spacing has an effective region between too dense and too sparse, whose location must be filled in from data. Three red lines hold throughout: nothing covert, nothing aimed at minors, and no describing liking as a state. (https://5dlibrary.com/answers/en/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c06/) 第6章 从「瞬时锚定」到习得曲线:评价性条件作用的账本 — NLP 所说的「锚定」缺乏实证:维特科夫斯基统计的 33 项相关研究中,支持者只占 0.273。可检验的替代机制是评价性条件作用,元分析平均效应 d 为 0.52,并在参与者意识到配对时更大。雷斯科拉-瓦格纳规则把「几次配对」写成习得曲线:学习率 0.10 时要 7 次才过半,不愉快配对又会让强度按同样规则消退。配对间隔有一个太密与太疏之间的有效区间,具体位置须由数据回填。三条红线:不隐蔽、不针对未成年人、不把好感说成状态。 (https://5dlibrary.com/answers/zh/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c06/) - Ch.7 Eastern and Western Sources: Music Born of the Heart, Fragrant Virtue and Correspondences — The Record of Music says that sound arises from the heart; the expressive direction holds, the claim of necessary cause does not. The five tones obtained by removing or adding a third can be recomputed exactly with fractions, and twelve steps overshoot by 23.46 cents. Jun Chen uses fragrance as a metaphor for virtue; its date must be handled as that of a forged Old Text chapter, and its "it is not the millet that is fragrant" agrees with the extinction rule of conditioning. Baudelaire's answering senses are supported by research on crossmodal correspondences, while the metaphysical part cannot be tested. Proust's surfacing memory is delayed and effortful, and program p72 shows that the advantage of smell for memories thirty years old depends on the age of the person remembering. (https://5dlibrary.com/answers/en/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c07/) 第7章 东西方原典考辨:乐由心生、明德惟馨与万物相应 — 《乐记》讲声由心生,表达方向成立,必然因果不成立;三分损益的五音比例可以用分数精确复算,十二步之后多出 23.46 音分。《君陈》以馨喻德,应按伪古文处理年代,其「黍稷非馨」与条件作用的消退规则相通。波德莱尔的感官应答为跨感官对应研究所支持,形而上学部分不可检验。普鲁斯特的回忆浮现有延迟、需努力,程序 p72 显示气味对三十年前记忆的优势依赖于回忆者的年龄。 (https://5dlibrary.com/answers/zh/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c07/) - Ch.8 The Evidence Loop: Field Tools and Checker Specifications — Chapter 8 gives the distilled Sound-and-Scent Signature Alignment and Congruence Calculator, checker specifications grouped by the four kits, a ledger and a pre-registered test needing 276 people per group. Scent and music in a space must be announced in advance, with a way to leave and care for people sensitive to smell, and no design aimed at minors. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c08/) 第8章 实证闭环:实战工具与检查器规格 — 第八章给出蒸馏小工具「声香签名对齐与跨感官一致度试算器」、按四件套分组的检查器规格、台账与预注册检验,主检验每组 276 人。空间投放香氛与音乐须事先公示、可选择离开、顾及气味敏感人群,不针对未成年人。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/acoustic-olfactory-cross-modal-coupling-proustian-retrieval-state-anchors/c08/) ## 5DL-B-393 · Dual-Coding Concrete Lexicons, Multi-Sensory Mental Imagery Pinning and Textual Brand-Persona Crystallisation 中文书名:具象词汇双编码认知、多感官意象钉扎与文字固化品牌形象母典 DOI: pending Cite: Fan, Y. (2026). Dual-Coding Concrete Lexicons, Multi-Sensory Mental Imagery Pinning and Textual Brand-Persona Crystallisation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Mixed presentation inflates the ratio while pure presentation and rich context shrink it, so that under illustrative parameters the same words give a ratio of 1.68 in pure lists and 3.29 in mixed lists. Parameters are illustrative and all calculations are generated by the book's programs; this is general business guidance, not legal advice, and it promises no sales or income. Abstract: A sentence circulates in the copywriting trade: concrete, touchable nouns and action verbs are remembered several times better than abstract adjectives, and copy that is concrete enough pins a brand image in the reader's mind for good. This book refutes the inflated part of that sentence and keeps the true part. The concreteness effect is real, supported by Paivio's dual coding theory and by experiments with prose; but the multiple depends on design. Mixed presentation inflates the ratio while pure presentation and rich context shrink it, so that under illustrative parameters the same words give a ratio of 1.68 in pure lists and 3.29 in mixed lists. Permanence fails: forgetting curves keep falling, and spaced repetition raises a half-life rather than abolishing decay. The book therefore corrects its blueprint: the multiple becomes a design-dependent range of ratios, pinning becomes retention, and because concrete wording raises how credible a statement seems rather than whether it is true, invented experiences, reviews and figures are the first red line. Parameters are illustrative and all calculations are generated by the book's programs; this is general business guidance, not legal advice, and it promises no sales or income. Chapter 1 scores texts with a small illustrative lexicon and compares mixed lists, pure lists and context. Chapter 2 develops the additive dual-coding model, in which the gain from the image code equals p_i times the failure probability of the verbal code, and shows that the margin shrinks sharply when the two routes are correlated. Chapter 3 computes a sensory dose: under illustrative parameters short copy should carry 2 to 4 senses, and sensory writing describes experience, never benefit. Chapter 4 gives a persona stability formula combining signature features, scene detail and consistency of tone, with the best synthetic card at 0.910, and shows an invented detail falling below plain wording after 390 visitors. Chapter 5 compares massed and spaced exposure: four exposures at equal intervals leave retention of 0.8175 on day 30, four massed exposures only 0.0856. Chapter 6 simulates the von Restorff isolation effect and finds a best contrast with a median of 0.43 and a band from 0.25 to 0.70. Chapter 7 reads Liu Xie's following the thing and Aristotle's putting before the eyes and writes them as fit and activity. Chapter 8 gives the image-pinning ledger, the distilled Concreteness and Image-Pinning Scorer, checker specifications grouped by the four kits and pre-registered tests whose main comparison needs 163 readers per group. All sample copy uses synthetic goods and people, and no real brand is reviewed. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:文案行业流传一句话:具体可触摸的名词与动作动词比抽象形容词好记好几倍,写得够具体就能把品牌形象永久钉进读者脑海。本书证伪这句话的夸张部分,保留它的真实部分。具体性效应真实存在,帕维奥的双编码理论与散文层面的实验都支持它;可「倍数」取决于设计,混合呈现放大比值,纯呈现与丰富语境缩小比值,示意参数下同一批词的纯表比值 1.68、混合表比值 3.29。「永久」并不成立,遗忘曲线持续下降,间隔重复抬高的是半衰期。本书据此做了蓝图修正:倍数写成设计依赖的比值区间,钉入写成保持率,具体措辞提高的是「显得可信」而不是「真实」,所以虚构亲历、虚构评价与虚构数字是第一红线。书中参数为示意,全部计算由书中程序生成;本书是非个人化经营建议,不构成法律意见,不承诺任何销量或收入。 第一章用作者自编的示意小词表给文本打具体度分,并比较混合表、纯表与语境。第二章写双编码加法模型,意象码增量等于 p_i 乘以言语码失败概率,两条通路相关时边际大幅缩水。第三章算感官通道剂量,示意参数下短文案的可行区间是 2 到 4 种感官,感官描写只写体验、不写功效。第四章给出形象稳定度公式,标志特征、场景细节与声调一致性三项合成,示意卡片最高 0.910;虚构细节在 390 位访客之后即跌破朴素措辞。 第五章比较集中与分散曝光,等间隔四次第 30 天保持率 0.8175,集中四次只有 0.0856。第六章模拟冯·雷斯托夫孤立效应,最优反差中位 0.43,区间 0.25 到 0.70。第七章考辨刘勰《文心雕龙》的随物宛转与亚里士多德《修辞学》的置于眼前,并把它们写成拟合度与活动度。第八章给出形象钉扎台账、蒸馏小工具「具象度与形象钉扎评分器」、按四件套分组的检查器规格与预注册检验,主检验每组需要 163 名读者。全书示意文案只用合成商品与合成人物,不点名评价任何真实品牌。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The True Size of the Concreteness Effect: the Multiple Is a Product of Design — The adjective pile hits 12 lexicon words, with mean concreteness 1.17, an adjective share of 0.58 and an action-verb share of 0.00. It covers only 1 sense channel, so its entropy is 0.000. The concrete text hits 15 words, with mean concreteness 4.33, a noun share of 0.60 and an action-verb share of 0.27; all five senses appear, and its entropy of 2.258 bits is close to the ceiling of 2.322. The ratio of mean concreteness between the two texts is 3.71. (https://5dlibrary.com/answers/en/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c01/) 第1章 具体性效应的真实大小:倍数是设计的产物 — 形容词堆砌那段,命中 12 个词,平均具体度 1.17,形容词占 0.58,动作动词占 0.00,感官通道只覆盖 1 种,熵为 0.000。具体那段命中 15 个词,平均具体度 4.33,名词占 0.60,动作动词占 0.27,五种感官全部出现,熵 2.258,接近上限 2.322。两段平均具体度之比是 3.71。 (https://5dlibrary.com/answers/zh/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c01/) - Ch.2 The Additive Dual-Coding Model: How Much the Image Code Adds, and When It Adds Nothing — The additive dual-coding model yields three working conclusions. The gain from the image code equals p_i × (1 − p_v), so it shrinks as the verbal code grows stronger. If both routes depend on the same attention, the margin of the image code shrinks sharply, by more than half under the illustrative parameters. The returns to concreteness are concentrated at the low end: going from empty to concrete is well worth it, going from concrete to more concrete is hardly worth it at all. (https://5dlibrary.com/answers/en/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c02/) 第2章 双编码加法模型:意象码能加多少,何时加了也白加 — 双编码加法模型给出三条可操作结论。意象码的增量等于 p_i × (1 − p_v),言语码越强,增量越小。两条通路若同时受注意力影响,意象码边际会大幅缩水,示意参数下损失过半。具体度收益集中在低端,从空洞到具体最值,从具体到更具体几乎不值。 (https://5dlibrary.com/answers/zh/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c02/) - Ch.3 The Sensory Dose: How Many Senses to Write In, and How the Senses Stay Faithful to the Facts — Chapter 3 computes a sensory dose: under illustrative parameters short copy should carry 2 to 4 senses, and sensory writing describes experience, never benefit. (https://5dlibrary.com/answers/en/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c03/) 第3章 多感官剂量:写进几种感官,感官怎样守住事实 — 第三章算感官通道剂量,示意参数下短文案的可行区间是 2 到 4 种感官,感官描写只写体验、不写功效。 (https://5dlibrary.com/answers/zh/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c03/) - Ch.4 Sculpting in Words: a Stability Formula for Personas and the Red Line of Truthfulness — Chapter 4 gives a persona stability formula combining signature features, scene detail and consistency of tone, with the best synthetic card at 0.910, and shows an invented detail falling below plain wording after 390 visitors. (https://5dlibrary.com/answers/en/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c04/) 第4章 文字塑像:形象稳定度公式与真实性红线 — 第四章给出形象稳定度公式,标志特征、场景细节与声调一致性三项合成,示意卡片最高 0.910;虚构细节在 390 位访客之后即跌破朴素措辞。 (https://5dlibrary.com/answers/zh/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c04/) - Ch.5 The Retention Behind "Pinned for Good": Forgetting Curves, Spaced Repetition and Whether a Signature Feature Should Change — Chapter 5 compares massed and spaced exposure: four exposures at equal intervals leave retention of 0.8175 on day 30, four massed exposures only 0.0856. (https://5dlibrary.com/answers/en/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c05/) 第5章 「永久钉入」的保持率:遗忘曲线、间隔重复与标志特征该不该换 — 第五章比较集中与分散曝光,等间隔四次第 30 天保持率 0.8175,集中四次只有 0.0856。 (https://5dlibrary.com/answers/zh/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c05/) - Ch.6 The Isolation Effect and the Best Contrast: How a Signature Feature Stands Out from Its Background — Chapter 6 simulates the von Restorff isolation effect and finds a best contrast with a median of 0.43 and a band from 0.25 to 0.70. (https://5dlibrary.com/answers/en/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c06/) 第6章 孤立效应与最佳反差:标志特征怎样从背景里跳出来 — 第六章模拟冯·雷斯托夫孤立效应,最优反差中位 0.43,区间 0.25 到 0.70。 (https://5dlibrary.com/answers/zh/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c06/) - Ch.7 A Critical Reading of the Classics: Liu Xie's Following the Thing and Aristotle's Putting Before the Eyes — None of the four clauses of the static version contains an action; its activity is 0.00 and its overall concreteness 2.38. All four clauses of the active version contain one; its activity is 1.00, its overall concreteness 4.86, and the concreteness of its first 8 words 5.00. The late-start version has activity 0.67 and overall concreteness 3.43, which does not look bad. But the concreteness of its first 8 words is only 1.00. A skimming reader sees only that "the value is clear and the experience is warm" and is offered no picture at all. (https://5dlibrary.com/answers/en/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c07/) 第7章 原典考辨:《文心雕龙》的随物宛转与亚里士多德的置于眼前 — 静态版四个小句没有一个动作,活动度 0.00,全文具体度 2.38。活动版四个小句全有动作,活动度 1.00,全文具体度 4.86,前 8 个词具体度 5.00。晚起版活动度 0.67,全文具体度 3.43,看起来不差;可前 8 个词具体度只有 1.00,扫读的读者只看到「价值很清楚、体验很温暖」,什么画面也没有。 (https://5dlibrary.com/answers/zh/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c07/) - Ch.8 Closing the Loop: the Image-Pinning Ledger, Field Tools and Checker Specifications — Chapter 8 gives the image-pinning ledger, the distilled Concreteness and Image-Pinning Scorer, checker specifications grouped by the four kits and pre-registered tests whose main comparison needs 163 readers per group. All sample copy uses synthetic goods and people, and no real brand is reviewed. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c08/) 第8章 实证闭环:形象钉扎台账、实战工具与检查器规格 — 第八章给出形象钉扎台账、蒸馏小工具「具象度与形象钉扎评分器」、按四件套分组的检查器规格与预注册检验,主检验每组需要 163 名读者。全书示意文案只用合成商品与合成人物,不点名评价任何真实品牌。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/dual-coding-concrete-lexicon-multisensory-imagery-pinning-brand-persona-crystallisation/c08/) ## 5DL-B-394 · Stripping Statistical Fingerprints of Machine Prose, Sentence-Length Burstiness Reconstruction and Master-Level Stylistic Cadence 中文书名:机器生成文本统计指纹剥离、句长突发律动重构与宗师级文气呼吸感母典 DOI: pending Cite: Fan, Y. (2026). Stripping Statistical Fingerprints of Machine Prose, Sentence-Length Burstiness Reconstruction and Master-Level Stylistic Cadence (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All examples are synthetic text written by the author or public-domain text, parameters are illustrative, and every calculation is generated by the book's programs. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: A popular claim among writers holds that machine prose has three tells, sentences of uniform length, an addiction to stock transitions and an epidemic of lists of three, and that rewriting to the rhythm of the masters removes its "plastic taste" for good, with a detector that no longer recognises the text as the test of success. This book refutes that claim and recasts it as testable craft. Sentence length has been a fingerprint of style for a long time: Yule and Williams showed more than eighty years ago that its distribution is right-skewed and roughly lognormal, and telling two writers apart by mean sentence length alone takes hundreds or thousands of sentences. Blueprint corrections: the burstiness index of Goh and Barabási, applied to sentence lengths, is only a rescaled coefficient of variation, so rhythm must be sought in order; detectors of machine text are unreliable and markedly biased against non-native writers, with a published average false-positive rate of 61.3 % on TOEFL essays, and at low base rates the positive predictive value of a flag can be as low as 0.140, so the book aims at no detector score and forbids using any score to accuse anyone; lists of three are only a hypothesis; and "master-level" is a slogan, since the coefficient of variation of the Gettysburg Address, 0.712, falls to 0.308 without its last sentence, while Liu Xie's sentence on four- and six-character lines has a clause-length coefficient of variation of only 0.096. All examples are synthetic text written by the author or public-domain text, parameters are illustrative, and every calculation is generated by the book's programs. Chapters 1 to 6 cover the sentence-length fingerprint and the lognormal, burstiness and measures of order, the Bayesian arithmetic and bias of detectors, Poisson tests of stock-phrase density and population-level excess vocabulary, runs of triads and compression redundancy, and a punctuation-only dynamic programme with a ledger of true detail: five punctuation changes raise a synthetic draft's coefficient of variation from 0.123 to 0.546 without moving a word. Chapter 7 reads Han Yu, Liu Xie, Aristotle's Rhetoric and Longinus and ends with a distillation spec for four kits. Chapter 8 gives the distilled Cadence and Stock-Phrase Density Checker, a checker specification grouped by the four kits, a cadence ledger and pre-registered tests; a paired blind test needs 153 judgments at a choice rate of 0.60. This is a book of writing craft: every 5D Library volume discloses AI assistance, and its tools serve the reader's experience and must not be used to obscure machine involvement where disclosure is required; concrete details must be true. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:写作圈流行一种说法:机器文本有三处破绽,句长太平均、过渡套话成瘾、三点排比泛滥,照着名家节律改一遍就能「彻底根除 AI 塑料味」,改到检测器认不出就算过关。本书证伪这种说法,并把它改写成可检验的写作工艺。句长分布作为文体指纹由来已久,尤尔与威廉斯在八十多年前就指出它右偏、近似对数正态;单凭平均句长区分两位作者需要成百上千个句子。蓝图修正:戈与巴拉巴西的突发度用在句长本身只是变异系数的换算,节律要从顺序里找;机器文本检测器不可靠,对非母语写作者偏差明显,已发表研究报告托福作文的平均误判率达 61.3 %,基础比例低时被标记稿件的阳性预测值可以只有 0.140,所以本书不以任何检测分数为目标,也不得用分数指控任何人;三元排比只是作者假说;「宗师级」只是口号,葛底斯堡演说的句长变异系数 0.712 去掉最后一句就降到 0.308,刘勰论四字六字那一句的分句变异系数只有 0.096。书中示例均为作者自写的合成文字或公有领域文本,参数为示意,全部计算由书中程序生成。 第一至六章依次讲句长指纹与对数正态、突发度与顺序量、检测器的贝叶斯账与偏差、套话密度的泊松检验与群体超额词汇、三元连发与压缩冗余、只动句读的动态规划与真实细节账本:合成草稿只改 5 处句读,变异系数就从 0.123 升到 0.546,而字词一个不动。第七章考辨韩愈《答李翊书》、刘勰《文心雕龙·章句》、亚里士多德《修辞学》与朗吉努斯《论崇高》,并给出四件套蒸馏规格。第八章给出蒸馏小工具「文气节律与套话密度检查器」、按四件套分组的检查规格表、文气节律台账与预注册检验,配对盲测在选择率 0.60 时需要 153 次判断。本书是写作工艺书:5D Library 书库各卷如实标注 AI 辅助,本书工具用于提高读者体验,不得用于在须披露的场合隐去机器参与;具身细节必须真实。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Sentence Length Is an Old Fingerprint: from Yule and Williams to Delta — Chapters 1 to 6 cover the sentence-length fingerprint and the lognormal, burstiness and measures of order, the Bayesian arithmetic and bias of detectors, Poisson tests of stock-phrase density and population-level excess vocabulary, runs of triads and compression redundancy, and a punctuation-only dynamic programme with a ledger of true detail: five punctuation changes raise a synthetic draft's coefficient of variation from 0.123 to 0.546 without moving a word. (https://5dlibrary.com/answers/en/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c01/) 第1章 句长是旧指纹:从尤尔、威廉斯到 Delta — 第一至六章依次讲句长指纹与对数正态、突发度与顺序量、检测器的贝叶斯账与偏差、套话密度的泊松检验与群体超额词汇、三元连发与压缩冗余、只动句读的动态规划与真实细节账本:合成草稿只改 5 处句读,变异系数就从 0.123 升到 0.546,而字词一个不动。 (https://5dlibrary.com/answers/zh/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c01/) - Ch.2 Burstiness, Autocorrelation and the Share of Short Sentences: Rhythm Lives in Order — The burstiness index was built for the gaps between events; applied to sentence lengths it is merely the coefficient of variation on another scale. Program p23 showed that one set of lengths in three orders leaves the coefficient of variation and B untouched while the autocorrelation moves from 0.703 to -0.292 and the flat pairs from 0.818 to none. Half of the sense of breathing lies in the distribution and half in the order. Program p22 warned that in a draft of ten sentences the sampling noise of these numbers is so large that they carry almost no information; a judgement needs at least twenty sentences, and measures of order need many more. (https://5dlibrary.com/answers/en/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c02/) 第2章 突发度、自相关与短句占比:节律藏在顺序里 — 突发度指数原本刻画事件间隔,用在句长上只是变异系数的另一种刻度。p23 证明,同一组长度换三种顺序,变异系数与 B 纹丝不动,自相关却从 0.703 变到 -0.292,平板对从 0.818 变到零。呼吸感一半在分布,一半在顺序。p22 则提醒我们,十句话的稿子里,这些数字的抽样噪声大到几乎没有信息,判定至少要二十句,顺序量要更多。 (https://5dlibrary.com/answers/zh/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c02/) - Ch.3 Why a Detector Score Is Not Evidence: False Positives, Bias and Near-Chance Human Judgment — Detectors are unreliable; published studies say so again and again. Program p31 showed that even with respectable ranking ability, most flagged drafts can be innocent when the base rate is low. Program p32 showed that a single threshold can multiply the false-positive rate more than tenfold for writers with a narrower vocabulary. Program p33 showed that human judgment needs hundreds of trials before it can be told apart from chance. (https://5dlibrary.com/answers/en/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c03/) 第3章 检测器为什么不能当证据:误报、偏差与人眼近随机 — 检测器不可靠,这是已发表研究反复给出的结论。p31 说明,即便排序能力看上去不错,基础比例低时被标记的稿子大半是冤枉的;p32 说明,同一个阈值落在词汇面较窄的写作者身上,误报率可以翻十倍以上;p33 说明,人眼判别要几百次才能和随机区分开。 (https://5dlibrary.com/answers/zh/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c03/) - Ch.4 Stock-Phrase Density and Excess Vocabulary: Poisson Tails and Population Lower Bounds — The trouble with stock phrases lies in density, not in any single word. Program p41 gave the algorithm for hits per thousand units; p42 showed that a short draft at three times the baseline often falls short of statistical significance, so the threshold for a red mark must move with length. Program p43 re-enacted the excess-vocabulary reasoning of Kobak and colleagues: ratios suit rare words, gaps suit common ones, and a gap gives a lower bound for a population that cannot be broken back down into any single draft. (https://5dlibrary.com/answers/en/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c04/) 第4章 套话密度与超额词汇:泊松尾概率与群体下界 — 套话的问题在密度,不在某个词。p41 给出每千字频次的算法,p42 说明短稿子的密度即使是基线三倍,也常常达不到统计显著,判红门槛必须随长度变化。p43 重演了科巴克等人的超额词汇思路:比值适合稀有词,差值适合常用词,差值给出的是群体下界,拆不回任何一篇稿子。 (https://5dlibrary.com/answers/zh/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c04/) - Ch.5 Lists of Three and Redundancy: Regular-Expression Counts, Run Lengths and Compression Ratios — The list of three is neither a machine invention nor a fault in itself. Program p51 split it into two quantities, frequency per thousand units and longest run; a single list easily inflates the frequency of a short text, while run length is closer to what readers feel as mechanical. Program p52 widened the view to repetition at the level of the sentence frame: a document whose sentences all differ on the surface can still have a high trigram repetition rate. Compression ratio and repetition rate are gauges of redundancy, not of authorship, and they mean something only for drafts of sufficient length. (https://5dlibrary.com/answers/en/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c05/) 第5章 三元排比与冗余:正则计数、连发长度与压缩比 — 三元排比不是机器的发明,也不是毛病本身。p51 把它拆成两个量:每千字频次与最长连发;短文本的频次很容易被一处列举拉高,连发长度更接近读者对「机械」的感受。p52 把视野放宽到句框层面的重复:句子字面全不相同,词三元组重复率仍可以很高;压缩比与重复率都是冗余指标,不是作者指标,而且只对足够长的稿子有意义。 (https://5dlibrary.com/answers/zh/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c05/) - Ch.6 Rebuilding Rhythm: a Dynamic Programme That Moves Only Punctuation, and a Ledger for Concrete Detail — The first constraint on revision is that facts stay put. Program p61 turned that constraint into a checkable quantity with a hand-written edit distance: a revision counts as purely rhythmic only if its content distance is zero; if not, the words that entered or left must be checked. Program p62 used dynamic programming to find the revision with the fewest changes under that constraint; raising the draft's coefficient of variation from 0.123 to 0.546 took only 5 changes, and the same run exposed an objective that pushes contrast to its limit. Program p63 opened a ledger of details: numbers and proper names need a record, and sensory words go to a person. (https://5dlibrary.com/answers/en/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c06/) 第6章 节律重构:只动句读的动态规划与具身细节账本 — 改稿的第一条约束是事实不动。p61 用手写的编辑距离把这条约束变成可检验的量:内容距离为零,才算纯节律修改;不为零,就要核对进出的字词。p62 用动态规划在这条约束下找改动最少的方案,草稿的变异系数从 0.123 升到 0.546 只需 5 处改动;同时暴露出目标函数会把对比推到极致的问题。p63 给细节建账,数字与专名必须有记录,感官词交人工确认。 (https://5dlibrary.com/answers/zh/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c06/) - Ch.7 Classical Sources: Han Yu on Breath and Words, Liu Xie on Fours and Sixes, Aristotle on Rhythm, Longinus on Harmony — Han Yu's length of words, Liu Xie's fours and sixes, Aristotle's neither metre nor formlessness, and Longinus's harmony on a par with thought all say one thing: rhythm is part of the content. Program p71 showed that Liu Xie's sentence on construction is so even that its coefficient of variation is only 0.096, while Han Yu's sentence is a sharp contrast of two characters and twelve. Program p72 showed that the high coefficient of variation of a famous text can come from one long sentence, and that its sampling band is too wide to serve as a standard. The classics give a direction, not numbers; the numbers must come from the ledger. (https://5dlibrary.com/answers/en/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c07/) 第7章 原典考辨:韩愈论气与言、刘勰论四六、亚里士多德论节奏、朗吉努斯论和谐 — 韩愈的「言之短长」、刘勰的「四字六字」、亚里士多德的「既非格律亦非无节奏」、朗吉努斯的「和谐与思想并重」,都在说同一件事:节律是内容的一部分。p71 显示,刘勰论句法的那一句整齐到变异系数只有 0.096,韩愈那一句则是两字与十二字的强烈对比;p72 显示,一篇名作的高变异系数可以只来自一个长句,抽样区间宽到不能当标准用。原典给了方向,没有给数值;数值要从台账里来。 (https://5dlibrary.com/answers/zh/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c07/) - Ch.8 Field Tools and Checker Specifications: a Cadence Ledger, Pre-Registered Tests and Conditions of Refutation — Chapter 8 gives the distilled Cadence and Stock-Phrase Density Checker, a checker specification grouped by the four kits, a cadence ledger and pre-registered tests; a paired blind test needs 153 judgments at a choice rate of 0.60. This is a book of writing craft: every 5D Library volume discloses AI assistance, and its tools serve the reader's experience and must not be used to obscure machine involvement where disclosure is required; concrete details must be true. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c08/) 第8章 实战工具与检查器规格:文气节律台账、预注册检验与推翻条件 — 第八章给出蒸馏小工具「文气节律与套话密度检查器」、按四件套分组的检查规格表、文气节律台账与预注册检验,配对盲测在选择率 0.60 时需要 153 次判断。本书是写作工艺书:5D Library 书库各卷如实标注 AI 辅助,本书工具用于提高读者体验,不得用于在须披露的场合隐去机器参与;具身细节必须真实。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/machine-prose-fingerprint-stripping-sentence-burstiness-master-cadence/c08/) ## 5DL-B-395 · High-Agency Active Inference, Intentional Perturbation Probes and Hidden Market-Truth Exploration 中文书名:高能动性主动推理、意念强度先手扰动与商业隐藏事实勘探母典 DOI: pending Cite: Fan, Y. (2026). High-Agency Active Inference, Intentional Perturbation Probes and Hidden Market-Truth Exploration (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative, calculations come from the book's programs, and all data are synthetic. This is general business analysis, not personalised legal, tax or investment advice; it promises no income. Abstract: A popular saying among online operators holds that agency and the strength of one's intention decide which reality one gets to see: the bystander can never tell confounded illusions from the truth, and only fast strikes that stir the market make real willingness to pay and a counterpart's hand appear. This book refutes two exaggerations in that saying. First, intention by itself does not change outside reality; a meta-analysis of intention and random number generators found only a tiny effect that publication bias could explain. "Intention strength" here names commitment only: probes per week, days before the first probe, and the cost cap per probe. Second, the bystander is not forever blind. The corrected blueprint: with unobserved confounding and no instrument, natural experiment or discontinuity, observational data cannot identify a causal effect; when the back-door criterion holds or a credible instrument exists, they can. Parameters are illustrative, calculations come from the book's programs, and all data are synthetic. This is general business analysis, not personalised legal, tax or investment advice; it promises no income. Chapter 1 measures the bias floor of a passive log: with illustrative parameters the floor is +0.150, it does not fall as the log grows, and a randomised probe at a real, payable price beats an infinitely long log with only 27 visitors. Chapter 2 sets out what back-door adjustment, instruments and published thresholds each require. Chapter 3 compares probe prices by Lindley information: the most informative price is 45, the most lucrative 35. Chapter 4 compares the regret of Thompson sampling with fixed habits: striking fast means starting early, not stopping early. Chapter 5 derives a stopping rule from the value of information, which halts the probe at 375 visitors; probes per week have a ceiling beyond which perturbation is excessive; and delay before the first probe has a price, with 90-day regret of 773.6 when starting at once and 5,804.9 when never starting. Chapter 6 covers the BDM mechanism, hypothetical bias between stated and real purchases, and inference of a counterpart's floor from public rate cards and public open calls only, with a posterior interval from 400 to 480. Chapter 7 examines Lu You, Wang Yangming, Bacon and Bernard. Chapter 8 gives the first-move probe ledger, the distilled Intentional Perturbation Probe Designer and Information-Gain Calculator, checker specifications grouped by the four kits and a pre-registered refutation plan. Every probe is purchasable and honestly delivered, with no invented scarcity, no resetting countdown, no posing as a client and no pricing by protected characteristics. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:经营圈流行一种说法:行动力与意念强度决定你能看见什么现实,旁观者永远分不清混杂假象,只有快速出手搅动市场,真实支付意愿与对手底牌才会显影。本书证伪其中两处夸大。第一,意念本身不改变外部现实,检验人的意图能否影响随机数发生器的元分析只找到可由发表偏差解释的极小效应;本书只把「意念强度」当作承诺强度的操作化名称:每周探针数、首个探针前的等待天数、单探针成本上限。第二,旁观并非永远看不清。核对文献后的蓝图修正是:存在未观测混杂、又没有工具变量、自然实验或断点时,观测数据不能识别因果效应;满足后门准则或有可信工具变量时,观测数据同样能识别。书中参数为示意,全部计算由书中程序生成,数据全部是合成数据。本书是非个人化经营建议,不构成法律、税务或投资意见,不向任何人承诺收入。 第一章量出旁观日志的偏差地板:示意参数下地板为 +0.150,不随日志变长而下降,只需 27 位随机访客,真实可买的探针就胜过无限长的日志。第二章逐一演示后门调整、工具变量与公开门槛的识别条件,以及控制了价格之后的变量会怎样制造偏差。第三章用林德利信息比较探针价格:信息最多的价格是 45,收入最高的价格是 35。第四章比较汤普森抽样与几种固定习惯的遗憾,说明快速出击的正确含义是尽早开始收集证据,而不是尽早停止。 第五章用信息价值给出停止规则,探针停在 375 位访客;每周探针数有上限,超过信息峰值即过度扰动;首探延迟有价格,立刻开始的 90 天遗憾是 773.6,一直不动是 5,804.9。第六章讲 BDM 机制、陈述偏好与真实购买之间的假设性偏差,以及只用公开报价与公开邀约推断合作方底价,后验区间从 400 到 480。第七章考辨陆游、王阳明、培根与贝尔纳,章末给出四件套蒸馏规格。第八章给出先手探针台账、蒸馏小工具「先手扰动探针设计与信息增益试算器」、按四件套分组的检查器规格与预注册证伪方案。所有探针真实可买、如实交付、预售如实标注,零虚假稀缺、零会重置的倒计时,不冒充客户询价,不按受保护特征差别定价。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Bias Floor of Confounding: Why a Longer Passive Log Is Wrong with More Confidence — Chapter 1 measures the bias floor of a passive log: with illustrative parameters the floor is +0.150, it does not fall as the log grows, and a randomised probe at a real, payable price beats an infinitely long log with only 27 visitors. (https://5dlibrary.com/answers/en/high-agency-active-inference-perturbation-probes-hidden-market-truth/c01/) 第1章 混杂的偏差地板:旁观日志为什么越长越稳地错 — 第一章量出旁观日志的偏差地板:示意参数下地板为 +0.150,不随日志变长而下降,只需 27 位随机访客,真实可买的探针就胜过无限长的日志。 (https://5dlibrary.com/answers/zh/high-agency-active-inference-perturbation-probes-hidden-market-truth/c01/) - Ch.2 Identifiability: What the Back Door, the Instrument and the Published Threshold Each Require — Case B is the more common situation. When the proxy's accuracy a is 0.5 it carries no information about intent, and the stratified contrast is still +0.0700, with the full bias of +0.1500 intact. At a = 0.7 the adjusted value is +0.0537 and the bias +0.1337. At a = 0.9 the adjusted value is -0.0098 and the bias +0.0702, still large. Only when a equals 1.0, that is when the proxy is intent itself, does the bias vanish. The more the proxy resembles the real confounder, the less bias remains, but as long as it is not identical the bias does not disappear. The channels, devices and time slots in a typical log are, for the most part, shadows of intent. (https://5dlibrary.com/answers/en/high-agency-active-inference-perturbation-probes-hidden-market-truth/c02/) 第2章 可识别性:后门、工具变量与公开门槛各需要什么条件 — 情形 B 是更常见的处境。代理变量的准确度 a 为 0.5 时,它与意向毫无关系,分层后的对比值仍是 +0.0700,偏差 +0.1500 原封不动。a 升到 0.7,调整后 +0.0537,偏差 +0.1337;a 到 0.9,调整后 -0.0098,偏差 +0.0702,仍然很大。只有 a 等于 1.0,也就是代理变量完全等于意向时,偏差才归零。代理越像本体,剩下的偏差越小,可是只要不完全一样,偏差就不会消失。日志里那些「渠道」「设备」「时段」,大多只是意向的影子。 (https://5dlibrary.com/answers/zh/high-agency-active-inference-perturbation-probes-hidden-market-truth/c02/) - Ch.3 Expected Information Gain: Put the Probe Where It Teaches the Most — Chapter 3 compares probe prices by Lindley information: the most informative price is 45, the most lucrative 35. (https://5dlibrary.com/answers/en/high-agency-active-inference-perturbation-probes-hidden-market-truth/c03/) 第3章 期望信息增益:探针价格该设在学得最多的地方 — 第三章用林德利信息比较探针价格:信息最多的价格是 45,收入最高的价格是 35。 (https://5dlibrary.com/answers/zh/high-agency-active-inference-perturbation-probes-hidden-market-truth/c03/) - Ch.4 Exploration and Exploitation: The Regret Ledger of Thompson Sampling and Fixed Habits — Under Thompson sampling, each doubling of the horizon adds a roughly constant amount of regret: 233.8, 296.7, 280.4 and 260.4. That is what logarithmic growth looks like, of the same order as the Lai-Robbins bound. Exploring for 200 visitors adds ever larger amounts: 173.2, 320.8, 564.7 and 1,120.2, because the stuck runs keep losing money in proportion to time. Exploring for 800 visitors loses heavily at first, reaching 1,594.4 after 1,000 visitors; afterwards it adds little, but at 8,000 visitors its total of 1,919.7 is still higher than the 1,604.7 of Thompson sampling. (https://5dlibrary.com/answers/en/high-agency-active-inference-perturbation-probes-hidden-market-truth/c04/) 第4章 探索与利用:汤普森抽样与几种固定习惯的遗憾账 — 汤普森抽样每翻一倍视野,新增遗憾大致持平:233.8、296.7、280.4、260.4。这正是对数增长的样子,与赖与罗宾斯的下界同阶。先试 200 位的做法,新增遗憾越来越大:173.2、320.8、564.7、1,120.2,因为卡住的那些运行一直在按时间亏钱。先试 800 位的做法,前期亏得重,访客数到 1,000 时累计 1,594.4;之后新增很少,但在 8,000 时累计 1,919.7,仍高于汤普森抽样的 1,604.7。 (https://5dlibrary.com/answers/zh/high-agency-active-inference-perturbation-probes-hidden-market-truth/c04/) - Ch.5 Commitment Strength Made Operational: Probes per Week, Delay to the First Probe and the Stopping Rule — Chapter 5 derives a stopping rule from the value of information, which halts the probe at 375 visitors; probes per week have a ceiling beyond which perturbation is excessive; and delay before the first probe has a price, with 90-day regret of 773.6 when starting at once and 5,804.9 when never starting. (https://5dlibrary.com/answers/en/high-agency-active-inference-perturbation-probes-hidden-market-truth/c05/) 第5章 承诺强度的操作化:每周探针数、首探延迟与停止规则 — 第五章用信息价值给出停止规则,探针停在 375 位访客;每周探针数有上限,超过信息峰值即过度扰动;首探延迟有价格,立刻开始的 90 天遗憾是 773.6,一直不动是 5,804.9。 (https://5dlibrary.com/answers/zh/high-agency-active-inference-perturbation-probes-hidden-market-truth/c05/) - Ch.6 Bringing Willingness to Pay into View, and a Counterpart's Floor: BDM, Hypothetical Bias and Inference from Public Information — Chapter 6 covers the BDM mechanism, hypothetical bias between stated and real purchases, and inference of a counterpart's floor from public rate cards and public open calls only, with a posterior interval from 400 to 480. (https://5dlibrary.com/answers/en/high-agency-active-inference-perturbation-probes-hidden-market-truth/c06/) 第6章 支付意愿显影与合作方底价:BDM 机制、假设性偏差与公开信息推断 — 第六章讲 BDM 机制、陈述偏好与真实购买之间的假设性偏差,以及只用公开报价与公开邀约推断合作方底价,后验区间从 400 到 480。 (https://5dlibrary.com/answers/zh/high-agency-active-inference-perturbation-probes-hidden-market-truth/c06/) - Ch.7 The Classics Examined: Lu You, Wang Yangming, Bacon and Bernard on Doing and on Provoked Observation — The log left to its own course has 10,000 visitors, and its slope has a standard error of 0.00800 by formula and 0.00833 in simulation. The vexed probe has only 400 visitors, yet its standard error is only 0.00490, with 0.00483 in simulation; at 1,600 visitors it falls to 0.00245, with 0.00251 in simulation. One probe visitor is worth 66.7 visitors left to their own course; to match a probe of 400, a log left to its own course would need 26,667 visitors. (https://5dlibrary.com/answers/en/high-agency-active-inference-perturbation-probes-hidden-market-truth/c07/) 第7章 原典考辨:陆游、王阳明、培根与贝尔纳论躬行与激发的观察 — 自行其道的日志有 10,000 位访客,斜率标准误按公式是 0.00800,模拟值 0.00833;扰动过的探针只有 400 位访客,标准误却只有 0.00490,模拟值 0.00483;1,600 位时降到 0.00245,模拟值 0.00251。一位探针访客相当于 66.7 位自行其道的访客;要追上 400 位的探针,自行其道的日志需要 26,667 位访客。 (https://5dlibrary.com/answers/zh/high-agency-active-inference-perturbation-probes-hidden-market-truth/c07/) - Ch.8 Field Tools and Checker Specifications: The First-Move Probe Ledger, Pre-Registered Tests and the Plan for Refutation — Chapter 8 gives the first-move probe ledger, the distilled Intentional Perturbation Probe Designer and Information-Gain Calculator, checker specifications grouped by the four kits and a pre-registered refutation plan. Every probe is purchasable and honestly delivered, with no invented scarcity, no resetting countdown, no posing as a client and no pricing by protected characteristics. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/high-agency-active-inference-perturbation-probes-hidden-market-truth/c08/) 第8章 实战工具与检查器规格:先手探针台账、预注册检验与证伪方案 — 第八章给出先手探针台账、蒸馏小工具「先手扰动探针设计与信息增益试算器」、按四件套分组的检查器规格与预注册证伪方案。所有探针真实可买、如实交付、预售如实标注,零虚假稀缺、零会重置的倒计时,不冒充客户询价,不按受保护特征差别定价。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/high-agency-active-inference-perturbation-probes-hidden-market-truth/c08/) ## 5DL-B-396 · Shattering the Illusion of Explanatory Depth, Epistemic Calibration and the Razor Separating True Knowledge from Assumed Familiarity 中文书名:解释深度错觉破除、认知边界校准与真知假知剥离剃刀母典 DOI: pending Cite: Fan, Y. (2026). Shattering the Illusion of Explanatory Depth, Epistemic Calibration and the Razor Separating True Knowledge from Assumed Familiarity (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative, all data are synthetic, and every calculation is generated by the book's programs. Chapter 1 measures, in synthetic data, a drop of 1.283 points in self-rated understanding of devices against an interval for facts from -0.263 to 0.063 that includes zero. Abstract: Two sentences circulate in business conversation. One says that the deadliest thing is "thinking you know", and that writing out the mechanism makes false knowledge fall away; the other says that the less people know, the more confident they are. This book turns the first into claims that data can refute and removes the second from the decision maker's toolkit. Checking the sources forced four corrections. The illusion of explanatory depth is strongest for causal mechanisms in devices and natural phenomena and weak or absent for facts, procedures and narratives, so the razor is reserved for mechanism claims and claims of fact are checked by retrieval. Writing out mechanisms reliably lowers self-rated understanding, but its effect on attitudes did not reappear in preregistered replications, so any change in decisions is treated as a small effect with stated conditions for refutation. The Dunning-Kruger quartile chart can be produced by noise plus a sense of being above average, so the book uses calibration curves and reliability tables instead. And "fatal" and "stripping" are replaced by intervals of loss. Parameters are illustrative, all data are synthetic, and every calculation is generated by the book's programs. This is a self-check manual offering general, non-personalised business guidance; it judges no one's intelligence, guarantees no decision and promises no return. Chapter 1 measures, in synthetic data, a drop of 1.283 points in self-rated understanding of devices against an interval for facts from -0.263 to 0.063 that includes zero. Chapter 2 covers the Brier score and Murphy's decomposition, the real hit rate of ninety per cent ranges (0.589 when ranges are half as wide as they should be) and the statistical artefact in the quartile chart. Chapter 3 turns the four-step razor (causal chain, order of magnitude, failure boundary, counterexample) into a scoring rule. Chapter 4 ranks an assumption list by expected loss, sensitivity and cost of testing, and computes the value of information and the least-cost order of tests. Chapter 5 quantifies reasons against, the premortem and the false progress from editing confidences after the fact. Chapter 6 samples wide illustrative parameters: routing between razor and retrieval cuts net loss by a median of 22.7 %, with 8.2 % at the fifth percentile and 41.2 % at the ninety-fifth, against a median of 5.1 % when the razor is applied to everything. Chapter 7 reads the Xunzi on dispelling obsession, chapter seventy-one of the Laozi and Descartes's four rules, and gives the distillation spec for the four kits. Chapter 8 sets out field tools and checker specifications: the false-knowledge register, the distilled Four-Step Razor and Pre-Decision Assumption-List Scorer, checkers grouped by kit, and a pre-registered test with its conditions for refutation; testing the drop in self-rating needs 88 raters, and testing a small Brier improvement needs 1,047 people per group. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:商业圈有两句流行说法:最致命的是「我以为我知道」,只要逼自己写出机制,假知就会现形;以及「越无知越自信」。本书把第一句改写成可被数据推翻的命题,并把第二句从决策工具箱里拿掉。核对原文后,本书做了四处蓝图修正:解释深度错觉在装置与自然现象的因果机制上最强,在事实、程序、叙事类知识上弱或不显著,所以剃刀只用于机制类判断,事实类判断改用检索核对;写出机制会稳健降低自评理解度,可对态度的作用在预注册复现中没有重现,本书只把「改变决策」写成小效应并写明推翻条件;达克效应那张四分位图可由噪声加「略高于平均」生成,本书改用校准曲线与可靠度表;「致命」「剥离」换成损失区间。书中参数为示意,数据全部合成,全部计算由书中程序生成。本书是决策自检工具书,属于非个人化经营建议,不评判任何人的智力,不保证任何决策正确,也不承诺任何收益。 第一章在合成数据里量出装置类自评下降 1.283 分,事实类下降区间从 -0.263 到 0.063,跨过零,并给出剃刀与检索的分流规则。第二章讲 Brier 分数与墨菲三分解、九成区间的真实命中率(区间只有应有宽度一半时命中率 0.589),以及四分位图的统计伪迹。第三章把四步剃刀(因果链条、数值量级、失效边界、反例)写成评分器,并检验费米拆分与链条乘积概率。第四章按期望损失、灵敏度与验证成本给假设清单排序,计算信息价值与最优验证次序。第五章量化反向理由、事前验尸与结果揭晓后改写原始确信度造成的假进步。 第六章在宽示意参数下抽样,剃刀与检索分流的净损失减少中位 22.7 %,五分位 8.2 %,九十五分位 41.2 %;全用剃刀中位只有 5.1 %。第七章考辨《荀子·解蔽》、《老子》第七十一章与笛卡尔《谈谈方法》的四条规则,并给出四件套蒸馏规格。第八章给出实战工具与检查器规格:假知剥离台账、蒸馏小工具「四步剃刀与决策前假设清单评分器」、按四件套分组的检查器、预注册检验与推翻条件;检验自评下降需 88 名评分者,检验 Brier 小幅改善每组需 1,047 人。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Boundary of the Illusion of Explanatory Depth: Strongest for Mechanisms, Weakest for Facts — Chapter 1 measures, in synthetic data, a drop of 1.283 points in self-rated understanding of devices against an interval for facts from -0.263 to 0.063 that includes zero. (https://5dlibrary.com/answers/en/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c01/) 第1章 解释深度错觉的边界:机制类最强,事实类最弱 — 第一章在合成数据里量出装置类自评下降 1.283 分,事实类下降区间从 -0.263 到 0.063,跨过零,并给出剃刀与检索的分流规则。 (https://5dlibrary.com/answers/zh/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c01/) - Ch.2 Calibration, Not Confidence: the Brier Score, Murphy's Decomposition and the Statistical Artefact in the Dunning-Kruger Chart — Chapter 2 covers the Brier score and Murphy's decomposition, the real hit rate of ninety per cent ranges (0.589 when ranges are half as wide as they should be) and the statistical artefact in the quartile chart. (https://5dlibrary.com/answers/en/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c02/) 第2章 看校准,不看自信:Brier 分数、Murphy 分解与达克效应图的统计伪迹 — 第二章讲 Brier 分数与墨菲三分解、九成区间的真实命中率(区间只有应有宽度一半时命中率 0.589),以及四分位图的统计伪迹。 (https://5dlibrary.com/answers/zh/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c02/) - Ch.3 The Four-Step Razor: Causal Chain, Order of Magnitude, Failure Boundary and Counterexample — The four-step razor scores every claim on four components, chain, magnitude, boundary and counterexample, and a weighted total of 0.60 passes. In the synthetic data the two claims held with the most confidence are exactly the two that fail. Fermi decomposition helps only when factor errors are small and independent; otherwise it can do worse than a gut estimate. Longer chains make weaker conclusions, and testing should begin with the weakest link. (https://5dlibrary.com/answers/en/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c03/) 第3章 四步剃刀:因果链条、数值量级、失效边界与反例 — 四步剃刀给每条判断打四个分:链条、量级、边界、反例,加权后 0.60 及格。合成数据里,确信度最高的两条判断恰好都不及格。费米拆分只在因子误差小且互不相关时有用,否则可能比直觉更差。链条越长,结论越弱,验证应当先挑最弱的环。 (https://5dlibrary.com/answers/zh/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c03/) - Ch.4 Scoring the Assumption List and the Value of Information: Which Assumption to Test First — Now the value of information. At a prior of 0.400 the current decision is to hold; EVPI is 24.00 and the pilot's EVSI is 8.40, about 0.350 of the value of perfect information. At a prior of 0.550 EVSI rises to 19.05. Exactly at the break-even point of 0.625, EVSI reaches its maximum of 24.38, which is 0.650 of the value of perfect information. Past the break-even point the current decision is already to launch, and EVSI falls: 17.70 at a prior of 0.700, 8.80 at 0.800, and 0.00 at 0.900. (https://5dlibrary.com/answers/en/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c04/) 第4章 假设清单评分与信息价值:先验证哪一条 — 再看信息价值。先验为 0.400 时,当下不上线,EVPI 是 24.00,试投的 EVSI 是 8.40,约为完全信息价值的 0.350。先验升到 0.550,EVSI 升到 19.05。恰好在打平点 0.625,EVSI 达到最大的 24.38,占完全信息价值的 0.650。过了打平点,当下已经会上线,EVSI 开始下降:先验 0.700 时是 17.70,0.800 时是 8.80,0.900 时降到 0.00。 (https://5dlibrary.com/answers/zh/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c04/) - Ch.5 Reasons Against, the Premortem and Sealing the Ledger against Hindsight — Writing a reason against can shrink overconfidence, but it can improve only a thin layer of the Brier score; in this chapter's illustrative model, at most 0.00931. A premortem's coverage of failure causes saturates with the number of writers, five to seven are enough, and what matters most is that the writers bring different backgrounds. Rewriting even a tenth of the original confidences after the outcome creates paper progress far larger than any real improvement, so the ledger must be sealed. (https://5dlibrary.com/answers/en/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c05/) 第5章 反向理由、事前验尸与后见之明的封存 — 写一条反对理由,可以缩小过度自信;可它能改善的只是 Brier 分数里很薄的一层,在本章示意模型里,最多 0.00931。事前验尸让失败原因的覆盖率随人数饱和,五到七人足够,关键在于写的人背景不同。结果揭晓后哪怕只改写一成原始确信度,账面就会出现远大于真实改善的假进步,台账必须封存。 (https://5dlibrary.com/answers/zh/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c05/) - Ch.6 Turning "Fatal" and "Stripping" into Intervals: a Band of Decision Loss, a Small Effect and the Conditions for Refutation — Chapter 6 samples wide illustrative parameters: routing between razor and retrieval cuts net loss by a median of 22.7 %, with 8.2 % at the fifth percentile and 41.2 % at the ninety-fifth, against a median of 5.1 % when the razor is applied to everything. (https://5dlibrary.com/answers/en/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c06/) 第6章 把「致命」与「剥离」换成区间:决策损失带、小效应与推翻条件 — 第六章在宽示意参数下抽样,剃刀与检索分流的净损失减少中位 22.7 %,五分位 8.2 %,九十五分位 41.2 %;全用剃刀中位只有 5.1 %。 (https://5dlibrary.com/answers/zh/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c06/) - Ch.7 Reading the Classics: the Xunzi on Dispelling Obsession, Chapter Seventy-One of the Laozi and Descartes's Four Rules — The true overall repeat rate is 0.3250. The reader of one corner has an error of 0.2494 with a sample of 20, and with a sample of 1,280 the error has fallen only to 0.2244; it barely moves. The stratified reader's error is 0.0984 at 20 and falls to 0.0124 at 1,280. The last line gives the bias floor for the reader of one corner: 0.2250, exactly the gap between the loyal customers' repeat rate and the overall rate. (https://5dlibrary.com/answers/en/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c07/) 第7章 原典考辨:《荀子·解蔽》、《老子》第七十一章与笛卡尔的四条规则 — 整体复购率是 0.3250。只读老客的人,样本 20 时误差 0.2494;样本增到 1,280,误差只降到 0.2244,几乎不动。分层抽样的人,样本 20 时误差 0.0984,到 1,280 降到 0.0124。最后一行给出一隅读者的偏差下限:0.2250,正是老客复购率与整体复购率之差。 (https://5dlibrary.com/answers/zh/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c07/) - Ch.8 Field Tools and Checker Specifications: the False-Knowledge Register, the Four-Step Razor Scorer and a Pre-Registered Test — Chapter 8 sets out field tools and checker specifications: the false-knowledge register, the distilled Four-Step Razor and Pre-Decision Assumption-List Scorer, checkers grouped by kit, and a pre-registered test with its conditions for refutation; testing the drop in self-rating needs 88 raters, and testing a small Brier improvement needs 1,047 people per group. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c08/) 第8章 实战工具与检查器规格:假知剥离台账、四步剃刀评分器与预注册检验 — 第八章给出实战工具与检查器规格:假知剥离台账、蒸馏小工具「四步剃刀与决策前假设清单评分器」、按四件套分组的检查器、预注册检验与推翻条件;检验自评下降需 88 名评分者,检验 Brier 小幅改善每组需 1,047 人。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/explanatory-depth-illusion-epistemic-calibration-true-vs-assumed-knowledge-razor/c08/) ## 5DL-B-397 · Payment-Acquirer Risk Step-Functions, Multi-Gateway Hot Failover and Rolling-Reserve Survival for Digital Fleets 中文书名:数字商品收单风控阶跃模型、多网关热备切换与防冻卡滚动准备金母典 DOI: pending Cite: Fan, Y. (2026). Payment-Acquirer Risk Step-Functions, Multi-Gateway Hot Failover and Rolling-Reserve Survival for Digital Fleets (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations come from the book's programs. This is general risk analysis, not legal, financing or investment advice, and promises no income. Abstract: Sellers of digital goods often say that in a surge the danger is traffic you cannot handle, and that more payment channels thin chargebacks out. This book refutes both sentences. In a surge the lethal risk is the acquirer's risk step: a dispute ratio over a line, a jump in takings that triggers review, a payout freeze or rolling reserve, and cash that runs dry within days. The second sentence is not risk control at all but a prohibited practice: spreading traffic over merchant accounts or gateways to lower a ratio breaks card-network rules and acquirer contracts, and ends in termination and a listing on industry termination files. The book makes five corrections to its blueprint: 0.35 % to 0.65 % is only the author's internal warning band, and the names and thresholds of card-network monitoring programmes change, so the live official pages govern; hot failover is not zero failure, because a common-cause share caps the gain from redundancy; dispute ratios computed by arrival date understate the truth in a surge and overstate it in a decline; blanket refunds cost more than targeted ones; and a payout freeze hurts cash far more than a rolling reserve. Parameters are illustrative; all calculations come from the book's programs. This is general risk analysis, not legal, financing or investment advice, and promises no income. Chapter 1 uses binomial noise and Beta posteriors to show that the step sits on the measured ratio. Chapter 2 shows that late-arriving disputes push an arrival-date ratio as low as 35.5 % of the true rate during a surge and up to 1.2622 % after a twenty-day campaign, and pulls the estimate back with a reporting-delay adjustment. Chapter 3 detects jumps in takings with CUSUM, EWMA and Bayesian online changepoints; at one false alarm a year a one-standard-deviation rise is seen in 8.5 days on average. Chapter 4 derives the threshold for refunding before a dispute, 0.443 with the illustrative parameters, and compares strict, blanket and targeted policies. Chapter 5 computes the availability ceiling of two acquiring channels with a common-cause share, and uses same-key retries, status queries and a one-success order ledger to bring 1,640 double charges per hundred thousand payments down to none left open at the end of the day. Chapter 6 finds that a ten per cent reserve held ninety days needs a buffer of only 2,900, whereas a thirty-day freeze needs 93,000, and that even a buffer of 50,000 is used up in a share of 0.0175 of years. Chapter 7 reads the Zuozhuan, the Doctrine of the Mean, Joseph's granaries in Genesis and Bagehot's Lombard Street on reserves. Chapter 8 gives the acquirer-risk ledger, the distilled Acquirer Risk Red-Line and Rolling-Reserve Survival Calculator, a checker table grouped by the four kits and pre-registered tests; the refund-desk test needs 63,900 payments per arm. Boxes separate published evidence, runnable Python (22 programs), classical sources and the author's hypotheses. 摘要:做数字商品的人常说:流水暴涨时最怕流量接不住;通道多开几个,拒付摊薄了就安全。本书证伪这两句话。流水暴涨时,致命的是收单机构的风控台阶:拒付比率越线、交易额跃升触发人工审查、结算被冻结或被扣滚动准备金,现金在几天之内断血。后一句更不是风控手段,而是违规做法:以多商户号或多网关分流来压低比率,违反卡组织规则与收单协议,后果是终止合作与列入行业终止名单。本书的蓝图修正有五条:0.35 % 至 0.65 % 只是作者设定的内部预警带,卡组织监测计划的名称与门槛会变动,以官方现行页面为准;热备不等于零故障,共因份额封住了冗余的收益;按到达日计算拒付比率,暴涨时低估、回落时高估;无条件全退的代价高于定向退款;结算冻结比滚动准备金更伤现金。书中参数为示意,全部计算由书中程序生成。本书是非个人化的经营与风险分析,不构成法律、融资或投资建议,不向任何人承诺收入。 第一章用二项分布与贝塔后验说明台阶长在测得比率上,并对照卡组织口径。第二章证明拒付晚到使按到达日计算的比率在暴涨期最低只剩真实比率的 35.5 %,一次二十天的活动之后又冲到 1.2622 %,并用报告延迟调整把估计拉回真实值。第三章用累积和、指数加权平均与贝叶斯在线变点检测交易额跃升,一年一次误报时,单位标准差的抬升平均 8.5 天被看见。第四章求出争议前退款的阈值,示意参数下为 0.443,并比较严格、全退与定向三种政策。 第五章算出两条收单通道在共因份额下的可用性上界,并用同键重试、状态查询与一单一成的订单账,把每十万笔支付中 1,640 笔重复扣款降到日终零未处理。第六章算出一成、九十天的滚动准备金只需 2,900 的缓冲,而冻结三十天需要 93,000,缓冲留足 50,000 时仍有 0.0175 的年份会用光。第七章考辨《左传》「居安思危」、《中庸》「凡事豫则立」、《创世记》约瑟积谷与白芝浩《伦巴第街》的准备金之论。第八章给出收单风控台账、蒸馏小工具「收单风控红线与滚动准备金生存试算器」、按四件套分组的检查器规格表与预注册检验,退款台检验每组需要 63,900 笔交易。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Steps: Noise in the Chargeback Ratio, the Beta Posterior and Card-Network Definitions — Start with Part A. Take a shop whose true ratio is 0.30 %, less than half of the red line. With 500 transactions a month, the probability that the measured ratio crosses the red line is 0.0654, which works out to a false red light roughly once every fifteen months. With 2,000 transactions a month that probability falls to 0.0087. For a shop whose true ratio is 0.40 %, the red light comes on in a fraction 0.1425 of months at 500 transactions, and in 0.0634 of months at 2,000. The ratio of a small shop is noisy by nature. A single chargeback, landing in a thin month, is enough to tip it over the line. (https://5dlibrary.com/answers/en/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c01/) 第1章 台阶:拒付比率的噪声、贝塔后验与卡组织口径 — 先看 A 段。真实比率 0.30 %,比红线低一半还多。月交易 500 笔时,测得比率越过红线的概率是 0.0654,每十五个月左右就有一次假红灯;月交易 2,000 笔时降到 0.0087。真实比率 0.40 % 的店,月交易 500 笔时有 0.1425 的月份亮红灯,2,000 笔时是 0.0634。小店的比率天生嘈杂,一笔拒付就能把它推过线。 (https://5dlibrary.com/answers/zh/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c01/) - Ch.2 Lag: Cohort Measures, Arrival-Date Measures and Underestimation During Surges — Chapter 2 shows that late-arriving disputes push an arrival-date ratio as low as 35.5 % of the true rate during a surge and up to 1.2622 % after a twenty-day campaign, and pulls the estimate back with a reporting-delay adjustment. (https://5dlibrary.com/answers/en/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c02/) 第2章 滞后:同期群口径、到达日口径与暴涨时的低估 — 第二章证明拒付晚到使按到达日计算的比率在暴涨期最低只剩真实比率的 35.5 %,一次二十天的活动之后又冲到 1.2622 %,并用报告延迟调整把估计拉回真实值。 (https://5dlibrary.com/answers/zh/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c02/) - Ch.3 Jumps: Cumulative Sums, Exponentially Weighted Averages and Bayesian Online Change Points — Chapter 3 detects jumps in takings with CUSUM, EWMA and Bayesian online changepoints; at one false alarm a year a one-standard-deviation rise is seen in 8.5 days on average. (https://5dlibrary.com/answers/en/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c03/) 第3章 跃升:累积和、指数加权平均与贝叶斯在线变点 — 第三章用累积和、指数加权平均与贝叶斯在线变点检测交易额跃升,一年一次误报时,单位标准差的抬升平均 8.5 天被看见。 (https://5dlibrary.com/answers/zh/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c03/) - Ch.4 Refunds: The Optimal Interception Threshold Before a Dispute — Chapter 4 derives the threshold for refunding before a dispute, 0.443 with the illustrative parameters, and compares strict, blanket and targeted policies. (https://5dlibrary.com/answers/en/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c04/) 第4章 退款:争议发生之前的最优拦截阈值 — 第四章求出争议前退款的阈值,示意参数下为 0.443,并比较严格、全退与定向三种政策。 (https://5dlibrary.com/answers/zh/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c04/) - Ch.5 Hot Failover: Common-Cause Failure, Timeout Budgets and Idempotency — Chapter 5 computes the availability ceiling of two acquiring channels with a common-cause share, and uses same-key retries, status queries and a one-success order ledger to bring 1,640 double charges per hundred thousand payments down to none left open at the end of the day. (https://5dlibrary.com/answers/en/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c05/) 第5章 热备:共因失效、超时预算与幂等 — 第五章算出两条收单通道在共因份额下的可用性上界,并用同键重试、状态查询与一单一成的订单账,把每十万笔支付中 1,640 笔重复扣款降到日终零未处理。 (https://5dlibrary.com/answers/zh/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c05/) - Ch.6 Reserves: Rolling Holdbacks, Settlement Freezes and the Minimum Buffer That Keeps Cash Flowing — Chapter 6 finds that a ten per cent reserve held ninety days needs a buffer of only 2,900, whereas a thirty-day freeze needs 93,000, and that even a buffer of 50,000 is used up in a share of 0.0175 of years. (https://5dlibrary.com/answers/en/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c06/) 第6章 准备金:滚动扣留、结算冻结与现金不断血的最低缓冲 — 第六章算出一成、九十天的滚动准备金只需 2,900 的缓冲,而冻结三十天需要 93,000,缓冲留足 50,000 时仍有 0.0175 的年份会用光。 (https://5dlibrary.com/answers/zh/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c06/) - Ch.7 Reading the Classics: Peace and Peril in the Zuo Zhuan, Forethought in the Doctrine of the Mean, Joseph's Granaries and Bagehot's Reserve — The table is worth reading row by row, because the spread is wide. A shop with a daily net inflow of 150 that saves all of it needs 333 days to fill the buffer, and the probability that a shock arrives first is 0.2114. If the same shop saves only a quarter of its inflow, it needs 1,333 days, and the probability that the shock wins the race rises to 0.6132. A shop with a daily net inflow of 475 that saves everything fills the buffer in 105 days, with a probability of 0.0722 that a shock arrives first; saving half, it takes 211 days, and the probability is 0.1393. A shop with a daily net inflow of 900 that saves everything needs only 56 days, and the probability of an early shock is just 0.0388. (https://5dlibrary.com/answers/en/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c07/) 第7章 原典考辨:《左传》居安思危、《中庸》豫则立、约瑟积谷与白芝浩的准备金 — 每天净流入 150、全部存下时,要 333 天才攒够,这期间冲击先到的概率是 0.2114;只存四分之一,要 1,333 天,冲击先到的概率是 0.6132。每天净流入 475、全部存下,105 天攒够,冲击先到的概率 0.0722;存一半,211 天,概率 0.1393。每天净流入 900、全部存下,56 天攒够,冲击先到的概率只有 0.0388。 (https://5dlibrary.com/answers/zh/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c07/) - Ch.8 The Empirical Loop: the Acquirer-Risk Ledger, Field Tools and Checker Specifications — Chapter 8 gives the acquirer-risk ledger, the distilled Acquirer Risk Red-Line and Rolling-Reserve Survival Calculator, a checker table grouped by the four kits and pre-registered tests; the refund-desk test needs 63,900 payments per arm. Boxes separate published evidence, runnable Python (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c08/) 第8章 实证闭环:收单风控台账、实战工具与检查器规格 — 第八章给出收单风控台账、蒸馏小工具「收单风控红线与滚动准备金生存试算器」、按四件套分组的检查器规格表与预注册检验,退款台检验每组需要 63,900 笔交易。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/payment-acquirer-risk-step-functions-multi-gateway-hot-failover-rolling-reserve-survival/c08/) ## 5DL-B-398 · Exit-Trajectory Micro-Downsells, Zero-Friction Lead Magnet Capture and Thank-You-Page Ten-Second Monetisation 中文书名:离场轨迹微降单挽留、零门槛诱饵捕获与付款成功页黄金十秒连带母典 DOI: pending Cite: Fan, Y. (2026). Exit-Trajectory Micro-Downsells, Zero-Friction Lead Magnet Capture and Thank-You-Page Ten-Second Monetisation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Cursor tracks do carry information about intent, but a model returns probabilities: on synthetic desktop data the AUC is 0.809, falling to 0.618 for touch sessions that lose the cursor signals. Parameters are illustrative, all calculations are generated by the book's programs, and all prices are unitless illustrative units. Abstract: A popular saying among small online sellers holds that pure profit leaks out at both ends of the funnel: detect a visitor's intention to leave within milliseconds and an ultra-cheap ice-breaker downsell wins her back; catch the buyer in the ten seconds of peak dopamine after payment and a one-click add-on on the thank-you page doubles the order value. This book splits the saying into three claims that data can refute and corrects its blueprint against the literature. Cursor tracks do carry information about intent, but a model returns probabilities: on synthetic desktop data the AUC is 0.809, falling to 0.618 for touch sessions that lose the cursor signals. A downsell trains buyers to pretend to leave, the Coase conjecture and strategic waiting at work; the profit-maximising price ratio for a single pop-up is about 0.432, but once cannibalisation, strategic waiting and the downward drift of the reference price are counted the long-run optimum is about 0.67, with a price floor of 0.33 under a cap of one pop-up per 30 days. Dopamine encodes reward prediction errors, and no study shows a ten-second window after payment; what can be measured on a thank-you page is the half-life of add-on acceptance, estimated at 6.50 seconds under illustrative parameters, with 0.830 of acceptances in the first 10 seconds. The median daily order-value multiplier is only 1.028, so "doubling" is a quantile that is almost never reached, and the book promises no income. Parameters are illustrative, all calculations are generated by the book's programs, and all prices are unitless illustrative units. This is general business analysis, not legal, tax or financial advice. Chapter 1 builds the exit-risk function and maps the trade-off between recall, precision and false pops. Chapter 2 sets out the three accounts of a downsell and derives the frequency cap. Chapter 3 follows the reference price down to a price floor. Chapter 4 covers the zero-friction lead magnet: the friction curve of forms, the diversion effect of a zero price, one-click unsubscribe and the principles of the Malaysian and Taiwanese data protection laws. Chapter 5 rewrites the "ten seconds of dopamine" with the product-limit estimator and a temporal-difference model. Chapter 6 computes the multiplier distribution of first-party one-click add-ons, with pre-ticked boxes and third-party data passes as counter-examples. Chapter 7 reads Bai Gui in the Records of the Grand Historian, the Han Feizi, Smith's Wealth of Nations and Franklin's The Way to Wealth, and turns them into clauses for the four kits. Chapter 8 gives the field tools and checker specifications, the two-end ledger and the pre-registered tests. Dark patterns appear only as counter-examples. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:线上小生意里流行一句说法:漏斗两端最容易漏掉纯利润,访客关页前毫秒级识别离场意图、弹出超低价破冰降单就能挽回流失,买家付款后多巴胺峰值的十秒内在感谢页一键加购,客单价就能倍增。本书把这句话拆成三个可被数据推翻的命题,并按文献做了蓝图修正。光标轨迹确实带有意图信息,但模型输出的是概率,合成桌面数据上的 AUC 为 0.809,触屏会话失去鼠标信号后降到 0.618。降单会训练买家「作势离开」,即科斯猜想与策略性等待;单次弹层的最优价格比约 0.432,计入蚕食、策略性等待与参照价下移后,长期最优约 0.67,频控每 30 天一次时价格下限 0.33。多巴胺编码的是奖赏预测误差,没有文献证明付款后存在十秒窗口;感谢页上可测的是加购接受率的半衰期,示意参数下估计为 6.50 秒,前 10 秒集中了 0.830 的接受。日客单价倍数中位数只有 1.028,「倍增」只是几乎碰不到的分位数,本书不作任何收益担保。书中参数为示意,全部计算由书中程序生成,价格一律为无币种示意单位。本书是非个人化经营分析,不是法律、税务或理财建议。 第一章建离场风险函数,比较逻辑回归与离散时间风险模型,求召回率、精确率与误弹率的权衡。第二章算降单的挽回、蚕食与学习三笔账,求频控上限。第三章算参照价下移与价格下限。第四章写零门槛诱饵:表单摩擦曲线、零价格的分流效应、一键退订与马来西亚、台湾个人资料保护法的原则层级要求。第五章用乘积极限法与时间差分模型改写「多巴胺十秒」。第六章算第一方一键加购的倍数分布,并以预勾选与第三方数据转递为反例。第七章考辨《史记·货殖列传》白圭、《韩非子·说难》、斯密《国富论》与富兰克林《致富之路》,并落成四件套条款。第八章给出实战工具与检查器规格、两端变现台账与预注册检验。暗黑模式在全书只作反例。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Exit-Risk Function: What Cursor Tracks Predict, and What Is Left on a Touch Screen — The cut-off decides who sees the pop-up. With a cut-off of 0.3, the share of leavers who are caught, the recall, is 0.517; the share of pop-ups that land on people who were really leaving, the precision, is 0.495; and the share of stayers who are interrupted by mistake, the false-pop rate, is 0.114. Raise the cut-off to 0.5 and precision climbs to 0.667 while recall falls to 0.260 and the false-pop rate to 0.028. Raise it again to 0.7 and precision reaches 0.792, but recall is down to 0.107 and the false-pop rate to 0.006. (https://5dlibrary.com/answers/en/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c01/) 第1章 离场风险函数:光标轨迹能预测什么,触屏上还剩什么 — 截断点决定弹层打给谁。截断点取 0.3 时,弹层打中要走的人的比例,也就是召回率,是 0.517;弹出去的层里真正要走的比例,也就是精确率,是 0.495;不走的人里被误弹的比例是 0.114。截断点抬到 0.5,精确率升到 0.667,召回率跌到 0.260,误弹率降到 0.028。再抬到 0.7,精确率 0.792,召回率只剩 0.107,误弹率 0.006。 (https://5dlibrary.com/answers/zh/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c01/) - Ch.2 The Economics of a Downsell: Cannibalisation, Strategic Waiting and the Frequency Cap — A downsell moves three accounts: recovery, cannibalisation and learning. The first two decide the sweet spot and the floor of the price ratio; the third decides the frequency cap. Under the illustrative parameters, the best ratio for a digital product is about 0.432, a cap of one keeps the steady state positive, and a cap of two turns it negative; physical goods should not be price-downsold at all. The Coase conjecture explains why the more often a downsell is honoured, the less of the full price survives. (https://5dlibrary.com/answers/en/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c02/) 第2章 降单经济学:蚕食、策略性等待与频控上限 — 降单的账有三笔:挽回、蚕食与学习。前两笔决定价格比的甜区与下限,第三笔决定频控。示意参数下,数字品的最优价格比约 0.432,频控一次稳态仍为正,频控两次转负;实物品类根本不该做价格降单。科斯猜想说明,降单兑现得越勤,原价越守不住。 (https://5dlibrary.com/answers/zh/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c02/) - Ch.3 The Reference Price Moves Down: How a Downsell Rewrites the Ruler in the Buyer's Head — The fourth account of a downsell is the reference price. A downsell moves the ruler in the buyer's head down, full-price conversion falls with it, losses weigh more than gains, and the harm takes one to two months to fade after the downsell stops. Putting the value of the pop-up, strategic waiting and the reference price on one grid, the best solution under the illustrative parameters is a cap of one with a price ratio of about 0.67 and a floor of 0.33; the more often the pop-up appears, the shallower the downsell must be. (https://5dlibrary.com/answers/en/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c03/) 第3章 参照价格下移:降单怎样改写买家心里那把尺 — 降单的第四笔账是参照价:它把买家心里那把尺往下拨,原价转化随之下降,而且损失比收益分量重,停掉降单之后还要一两个月才能恢复。把弹层价值、策略性等待与参照价放进同一张网格,示意参数下最优解是频控一次、价格比约 0.67,下限 0.33;弹得越勤,只能降得越浅。 (https://5dlibrary.com/answers/zh/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c03/) - Ch.4 The Zero-Friction Lead Magnet: the Friction Curve of a Form, the Zero-Price Effect and One-Click Unsubscribe — With email alone, the form yields 420 leads with a first-order probability of 0.040, or 16.80 first orders per 1,000 visitors, and stores 420 personal fields. Add a name and the leads fall to 374 while the first-order probability rises to 0.045; first orders are 16.75, practically unchanged, but the stored fields jump to 748. Add a phone number and the leads drop to 307, with 15.20 first orders. Ask for all six fields and first orders come to 14.58, with 1,368 stored fields. (https://5dlibrary.com/answers/en/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c04/) 第4章 零门槛诱饵:表单摩擦曲线、零价格效应与一键退订 — 只收邮箱时,420 条线索,首单概率 0.040,每 1,000 个访客 16.80 个首单,存下 420 条个人字段。加上名字,线索降到 374,首单概率升到 0.045,首单 16.75,几乎持平,存下的字段却翻到 748。再加电话,线索掉到 307,首单 15.20。六个字段全收时,首单 14.58,存下的字段 1,368 条。 (https://5dlibrary.com/answers/zh/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c04/) - Ch.5 Attention Decay on the Thank-You Page: No Ten-Second Dopamine Window, Only a Half-Life — None of the three components of the "ten-second dopamine peak" can be taken at face value. Dopamine encodes prediction errors; when a successful payment arrives as expected the error is close to zero, and the signal has already appeared at the click on "pay". Ten seconds is not a physiological constant but a quantile of the distribution of acceptance times; under the illustrative parameters the first 10 seconds hold 0.830 of acceptances, but that number depends entirely on the half-life of the shop's own page. What can be written down as a rule is a curve of attention decay: the offer appears with the thank-you content, does not interrupt and does not hurry. (https://5dlibrary.com/answers/en/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c05/) 第5章 感谢页的注意衰减:没有多巴胺十秒,只有一条半衰期 — 「多巴胺峰值十秒」三个成分,一个都不能照搬。多巴胺编码的是预测误差,付款成功如期出现时误差接近零,信号早在点击付款那一刻就出现了。十秒不是生理常数,而是接受时间分布上的一个分位数,示意参数下前 10 秒集中了 0.830 的接受,但这个数完全取决于自家页面的半衰期。真正可以写成规则的,是一条注意衰减曲线:提示与感谢内容同时出现,不打断,不催促。 (https://5dlibrary.com/answers/zh/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c05/) - Ch.6 The First-Party One-Click Add-On: the Distribution of Order-Value Multipliers and the Net Accounts of Three Designs — Under the illustrative parameters "doubling" is all but unreachable: the median daily multiplier is about 1.028, and even the optimistic scenario gives only 1.084. Of the three designs, the pre-ticked box looks best in the short-run ledger but is flagged red for its chargeback ratio and its manipulative nature; the third-party data pass has a negative net and is prohibited by law; only the first-party explicit click is legitimate. Mental accounting puts a post-payment add-on at a natural disadvantage, which is repaired by folding it into the same order and showing the new total; and add-on prices must be open before the click. (https://5dlibrary.com/answers/en/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c06/) 第6章 第一方一键加购:客单价倍数的分布,以及三种设计的净账 — 「倍增」在示意参数下几乎碰不到:日倍数中位数约 1.028,乐观情景也只有 1.084。三种设计里,预勾选短期账面最好看,却因拒付比超线与操纵属性判红;第三方数据转递净额为负且受法律禁止;只有第一方明示点击是正当做法。心理账户让付款后加购天然吃亏,补救是并入同一订单、显示新总额;附加价格必须在点击前公开。 (https://5dlibrary.com/answers/zh/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c06/) - Ch.7 Reading the Classics: Bai Gui's Taking and Giving, Han Fei on Persuasion, Smith on Exchange and Franklin's Small Leak — Offering the entry product at exit nets 250.7 per 1,000 offers shown. Offering a complement at exit makes no sense when the visitor is not even buying the main product, and the net is -105.0. Offering a complement on the thank-you page nets 492.5. Offering the entry product on the thank-you page has an acceptance of only 0.020 and may make a buyer who has just paid full price feel cheated and ask for a refund, so the net is only 92.0. Following Bai Gui's pairing, the two cells together come to 743.2; swapping the offers gives a combined -13.0. (https://5dlibrary.com/answers/en/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c07/) 第7章 原典考辨:白圭取与、韩非说难、斯密交换与富兰克林小漏 — 离场时给入门品,每 1,000 次展示净得 250.7;离场时给互补品,访客连主商品都不买,互补品无从谈起,净额 -105.0。感谢页给互补品,净得 492.5;感谢页给入门品,接受率只有 0.020,还可能让刚付了原价的买家觉得吃亏、要求退款,净额只有 92.0。按白圭的配法,两格合计 743.2;把提议换过来,合计 -13.0。 (https://5dlibrary.com/answers/zh/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c07/) - Ch.8 Field Tools and Checker Specifications: the Two-End Ledger, Four-Kit Criteria and Pre-Registered Refutation — Chapter 8 gives the field tools and checker specifications, the two-end ledger and the pre-registered tests. Dark patterns appear only as counter-examples. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c08/) 第8章 实战工具与检查器规格:两端变现台账、四件套判据与预注册证伪 — 第八章给出实战工具与检查器规格、两端变现台账与预注册检验。暗黑模式在全书只作反例。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/exit-trajectory-micro-downsells-zero-friction-lead-magnet-thank-you-page-ten-second-monetisation/c08/) ## 5DL-B-399 · Pareto Whale Auto-Detection, High-Net-Worth Self-Selection Signals and Ten-X VIP Tier Escalation 中文书名:帕累托鲸鱼客群自动显影、高净值信号自筛选与十倍客单尊享层跃迁母典 DOI: pending Cite: Fan, Y. (2026). Pareto Whale Auto-Detection, High-Net-Worth Self-Selection Signals and Ten-X VIP Tier Escalation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs. This is general business analysis, not personalised legal, tax or financial advice, and it promises no income to anyone. Abstract: A popular story among low-price automated shops says that revenue follows a power law, that the top 2 % of super-high-net-worth customers bring in a disproportionate share of profit, that a few implicit probes will detect these whales automatically, and that a private VIP lane will then close orders ten to fifty times the usual size. This book tests the story sentence by sentence. After checking the original literature it corrects its blueprint: a power law must be tested first, and a purely lognormal synthetic book passes the power-law goodness-of-fit test with a p-value of 0.430, while only the likelihood ratio test, with z equal to -1.271, reports that the data cannot decide; the profit share of the top 2 % ranges from 0.121 to 0.768 with the shape of the tail and can only be reported as an interval; "automatic detection" with the BG/NBD model recognises next year's top 2 % with a hit rate of 0.625, not 1; the decoy effect can weaken or reverse under natural presentation; and the "implicit probe" becomes an entry that every visitor can see and choose. Parameters are illustrative and all calculations are generated by the book's programs. This is general business analysis, not personalised legal, tax or financial advice, and it promises no income to anyone. Chapter 1 separates power law from lognormal with the Hill estimator, the Clauset procedure and Vuong's test. Chapter 2 gives interval estimates of head shares from closed forms and the bootstrap and shows the risk of extrapolating with the wrong tail. Chapter 3 forecasts individual value with BG/NBD and rules that the score sets only the order of follow-up, never prices, never hides an entry and is never joined to outside data. Chapter 4 uses the menu screening of Mussa and Rosen to generate tier price gaps of about 9.57 and 51.66 times, shows that the unconstrained menu leaves 0.575 of buyers worse off, and finds that locking the base tier keeps 0.984 of the profit with no buyer worse off. Chapter 5 computes the expected profit, break-even false positive rate and Bayesian posterior of a public VIP entry, with a weekly no-degradation test on base-tier conversion. Chapter 6 shows with a multinomial logit that the direction of decoy and compromise effects depends on a parameter to be estimated, sizes a VIP consultant desk with Erlang C, one consultant serving 23 members and four serving 176, and shows that the VIP tier's cost multiple rises faster than its price multiple. Chapter 7 reads the Han Feizi, the Shiji chapter on money-makers, Dupuit on third-class carriages and Pareto on the income curve, and turns the conclusions into clauses for the four kits. Chapter 8 gives the distilled Whale Tail Test and VIP Opt-in Menu Calculator, checker specifications grouped by the four kits, a weekly whale ledger and pre-registered tests, the most expensive needing 14,579 customers. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:平价自动化站点流行一种说法:收入服从幂律,前 2 % 的超级高净值客户贡献不成比例的利润,埋几个隐式探针就能把这些鲸鱼自动显影出来,再开一条尊享通道成交十倍至五十倍的客单。本书逐句检验这套说法。核对原始文献之后做了蓝图修正:幂律必须先检验,一本纯对数正态的合成账簿通过了幂律拟合优度检验,p 值 0.430,只有似然比检验给出 z 为 -1.271、无法裁决;前 2 % 的利润份额随尾部形状从 0.121 变到 0.768,只能按区间报;「自动显影」用 BG/NBD 模型认出下一年前 2 % 客户的命中率是 0.625,不是 1;诱饵效应在自然呈现下可能减弱或反转;「隐式探针」改写为对所有访客公开、可自选的入口。书中参数为示意,全部计算由书中程序生成。本书是非个人化经营分析,不构成法律、税务或理财建议,不向任何人承诺收入。 第一章用希尔估计、克劳塞特流程与武昂检验区分幂律与对数正态。第二章用闭式公式与自助法给出头部份额区间,并说明按错误尾巴外推的风险。第三章用 BG/NBD 模型预测个人价值,规定预测分数只决定跟进先后,不用于定价、不隐藏入口、不拼接站外资料。第四章用穆萨与罗森的菜单甄别生成约 9.57 倍与 51.66 倍的档位价差,并指出不受约束的菜单让 0.575 的买家比现状更差;锁住基础档后利润仍保住 0.984,没有买家变差。 第五章算出公开尊享入口的期望利润、盈亏平衡假阳性率与贝叶斯后验,并给出普通档转化的周度不劣化检验。第六章用多项 logit 说明诱饵与折中效应的方向取决于要估计的参数,用爱尔朗 C 算出尊享顾问的容量,1 位顾问承接 23 位会员,4 位承接 176 位,并说明尊享档成本倍数比价格倍数涨得更快。第七章考辨《韩非子·五蠹》、《史记·货殖列传》、杜普伊论三等车厢与帕累托论收入曲线,并把结论落成四件套条款。第八章给出蒸馏小工具「鲸鱼尾部检验与尊享档自选菜单试算器」、按四件套分组的检查器规格、每周鲸鱼显影台账与预注册检验,最贵的一项需要 14,579 位客户。高净值信号只用买家在本站自愿做出的行为,零敏感资料、零外购财富数据、零按受保护特征分流或差别定价。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Power Law or Lognormal: Test the Tail Before Talking About Whales — Program p12 tries 60 candidate thresholds between the 0.50 and 0.98 quantiles and draws 200 synthetic books for each real book. The result is not what the long-tail chart would lead anyone to expect. For book LN the chosen threshold is 134.4, with 427 customers in the tail, an exponent of 2.574 and a distance of 0.0300. Its p-value is 0.430, comfortably above the bar, so the power law is not rejected. For book PT the chosen threshold is 125.4, not far from the true 120, with 329 customers in the tail and an exponent of 2.659, close to the true value of 1 plus 1.6. Yet its p-value is only 0.100, sitting exactly on the bar. (https://5dlibrary.com/answers/en/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c01/) 第1章 幂律还是对数正态:先检验尾巴,再谈鲸鱼 — 程序 p12 在 0.50 到 0.98 分位数之间取 60 个候选起点,每本账簿造 200 本合成账簿。结果出人意料。账簿 LN 选出的起点是 134.4,尾部 427 位客户,指数 2.574,距离 0.0300,p 值 0.430,远高于门槛,幂律没有被拒绝。账簿 PT 选出的起点是 125.4,离真实起点 120 不远,尾部 329 位客户,指数 2.659,与真值 1 加 1.6 很接近,p 值却只有 0.100,恰好压在门槛上。 (https://5dlibrary.com/answers/zh/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c01/) - Ch.2 The Profit Share of the Top 2 %: Closed Forms, the Bootstrap and the Danger of Extrapolation — The share of the top 2 % is decided by the shape of the tail, and the 80/20 rule is the special case of a Pareto index of 1.161. The profit share always exceeds the revenue share, by an amount that depends on the per-head service cost. The heavier the tail, the wider the bootstrap interval. Extrapolating a lognormal as a power law overstates the one-in-a-thousand point 1.80 times and the one-in-ten-thousand point 3.54 times. (https://5dlibrary.com/answers/en/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c02/) 第2章 前 2 % 的利润份额:闭式区间、自助法与外推之险 — 前 2 % 客户的份额由尾部形状决定,二八定律只是帕累托指数 1.161 时的特例。利润份额总高于收入份额,但高出多少取决于按人头计的服务成本。尾巴越重,自助区间越宽;把对数正态当幂律外推,千分之一分位处会高估 1.80 倍,万分之一分位处高估 3.54 倍。 (https://5dlibrary.com/answers/zh/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c02/) - Ch.3 Who Will Be Next Year's Whale: BG/NBD and the Prediction of High Value — BG/NBD models the purchase rate and silent attrition separately and estimates its parameters from three summaries: the number of repeat purchases, the time of the last purchase and the length of observation. On the synthetic book, two of the four parameters are estimated closely and two drift because the likelihood surface is flat. At the level of the whole book the conditional forecast differs from the simulation by about four per cent, and the hit rate for next year's top 2 % is 0.625, above the 0.450 of the spend ranking. (https://5dlibrary.com/answers/en/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c03/) 第3章 谁会是明年的鲸鱼:BG/NBD 与高价值预测 — BG/NBD 把购买速率与悄悄流失分开建模,只用复购次数、最后购买时间与观察期三项摘要就能估计参数。合成账簿上,四个参数里两个估得很准,两个受似然面平坦所累偏差稍大;条件预测在全书层面与模拟结果相差约四个百分点;认出明年前 2 % 的命中率是 0.625,高于按消费额排名的 0.450。 (https://5dlibrary.com/answers/zh/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c03/) - Ch.4 Menu Screening: Three Mussa-Rosen Tiers and the No-Degradation Constraint on the Base Tier — Chapter 4 uses the menu screening of Mussa and Rosen to generate tier price gaps of about 9.57 and 51.66 times, shows that the unconstrained menu leaves 0.575 of buyers worse off, and finds that locking the base tier keeps 0.984 of the profit with no buyer worse off. (https://5dlibrary.com/answers/en/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c04/) 第4章 菜单甄别:Mussa–Rosen 三档与基础档不劣化约束 — 第四章用穆萨与罗森的菜单甄别生成约 9.57 倍与 51.66 倍的档位价差,并指出不受约束的菜单让 0.575 的买家比现状更差;锁住基础档后利润仍保住 0.984,没有买家变差。 (https://5dlibrary.com/answers/zh/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c04/) - Ch.5 Opt-in Entries: High-Value Signals on a Public Menu and Their Expected Profit — Once the "implicit probe" is rewritten as a public entry, the signal comes from the buyer's voluntary choice rather than from covert observation. The entry's worth depends on the ratio of its true and false positive rates; under the illustrative parameters, a true positive rate of 0.4 tolerates a false positive rate of up to 0.087. Combining the opt-in signal with purchase history lifts the posterior from 0.02 to as high as 0.954, and anyone who actively states an interest is worth a reply. The weekly test on base-tier conversion turns the no-degradation constraint into a red, yellow or green flag every week. (https://5dlibrary.com/answers/en/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c05/) 第5章 自选入口:公开菜单上的高净值信号与期望利润 — 「隐式探针」改写为公开入口之后,信号来自买家的自愿选择,而不是暗中观测。入口的价值取决于真阳性率与假阳性率之比;在示意参数下,真阳性率 0.4 时假阳性率可以容忍到 0.087。自选信号与购买历史合起来,后验概率从 0.02 一路升到 0.954,而且任何主动表达意向的人都值得回复。普通档转化的周度检验把不劣化约束变成每周一次的红黄绿判定。 (https://5dlibrary.com/answers/zh/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c05/) - Ch.6 Displaying the Tiers and Sizing the VIP Layer: Decoys, Compromise and Queues — Chapter 6 shows with a multinomial logit that the direction of decoy and compromise effects depends on a parameter to be estimated, sizes a VIP consultant desk with Erlang C, one consultant serving 23 members and four serving 176, and shows that the VIP tier's cost multiple rises faster than its price multiple. (https://5dlibrary.com/answers/en/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c06/) 第6章 档位陈列与尊享层容量:诱饵、折中与排队 — 第六章用多项 logit 说明诱饵与折中效应的方向取决于要估计的参数,用爱尔朗 C 算出尊享顾问的容量,1 位顾问承接 23 位会员,4 位承接 176 位,并说明尊享档成本倍数比价格倍数涨得更快。 (https://5dlibrary.com/answers/zh/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c06/) - Ch.7 Classical Sources: the Han Feizi, the Shiji, Dupuit and Pareto — When the poor make up 0.5 of travellers, removing the roof raises profit from 3.825 to 3.950, a gain of 0.125. At 0.6 the gain is 0.030; at 0.7 removing the roof loses 0.065; at 0.8 it loses 0.160. The last line gives the threshold: removing the roof pays while the share of poor travellers is below 0.632. Without the roof the third-class fare falls to 1.600 and the second-class fare rises to 5.100. The middling traveller's information rent falls from 1.50 to 0.90, and that difference is exactly what the company collects. (https://5dlibrary.com/answers/en/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c07/) 第7章 原典考辨:《韩非子》《史记》与杜普伊、帕累托 — 穷人占 0.5 时,拆掉顶棚让利润从 3.825 升到 3.950,多出 0.125;占 0.6 时多出 0.030;占 0.7 时反而少了 0.065;占 0.8 时少了 0.160。程序最后一行给出临界点:穷人占比低于 0.632 时,拆顶棚有利可图。拆掉顶棚后,三等票降到 1.600,二等票升到 5.100;中等旅客的信息租金从 1.50 降到 0.90,被公司收走的正是这部分。 (https://5dlibrary.com/answers/zh/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c07/) - Ch.8 Field Tools and Checker Specifications: the Whale Ledger, the Four Kits and a Pre-registered Refutation — Chapter 8 gives the distilled Whale Tail Test and VIP Opt-in Menu Calculator, checker specifications grouped by the four kits, a weekly whale ledger and pre-registered tests, the most expensive needing 14,579 customers. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c08/) 第8章 实战工具与检查器规格:鲸鱼显影台账、四件套与预注册证伪 — 第八章给出蒸馏小工具「鲸鱼尾部检验与尊享档自选菜单试算器」、按四件套分组的检查器规格、每周鲸鱼显影台账与预注册检验,最贵的一项需要 14,579 位客户。高净值信号只用买家在本站自愿做出的行为,零敏感资料、零外购财富数据、零按受保护特征分流或差别定价。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/pareto-whale-auto-detection-high-net-worth-self-selection-ten-x-vip-tier-escalation/c08/) ## 5DL-B-400 · Ad-Creative Fatigue Half-Lives, Orthogonal Storyboard Recombination and Automated Permutation Rendering 中文书名:投放素材频次疲劳半衰期、分镜基因正交重组与百条广告全自动混剪母典 DOI: pending Cite: Fan, Y. (2026). Ad-Creative Fatigue Half-Lives, Orthogonal Storyboard Recombination and Automated Permutation Rendering (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations come from the book's programs. This is general business analysis, not individual advice, and promises no advertising result. Abstract: A popular story in paid-traffic marketing says that creatives decay fast during scale-up because repeated exposure makes the click-through rate fall with an estimable half-life, and that splitting a short video into four modules (hook, pain point, proof, call to action) and combining them with an orthogonal array lets a dozen shots be synthesised automatically into a hundred and more low-correlation test creatives, breaking the production bottleneck. Testing each link against the literature and by program, the book corrects that blueprint. Fatigue is not a monotone exponential: repetition effects rise before they fall, paused creatives partly recover, and dashboard declines also carry auction drift. Orthogonal arrays estimate main effects only and alias interactions, so "low correlation" is a property of the design matrix. Advertising experiments have very little power, and testing a hundred creatives at once requires false discovery control. Platforms allocate impressions by predicted performance, which biases creative tests. Parameters are illustrative; all calculations come from the book's programs. This is general business analysis, not individual advice, and promises no advertising result. Chapter 1 writes three fatigue curves as formulas, fits a half-life with a bootstrap interval to a synthetic log and corrects for the environment with fresh viewers; under illustrative parameters environmental drift accounts for 0.211 of the aggregate decline. Chapter 2 estimates, with a Naik-type model, a wear-out half-life of 8.42 days on air and a recovery half-time of 3.92 days off air, and derives a frequency cap and a rotation size. Chapter 3 builds L4, L8, L9 and L16 from scratch; the contrast variance is 0.6667 under L9 against 2.0000 one factor at a time, but a hook-by-call-to-action synergy lands wholly on the pain-point module. Chapter 4 separates design distance, correlated performance and the shared fatigue of shared shots: a pool recombined from 12 shots is as fresh to viewers as two or three new creatives, with an average relative click-through rate of 0.828. Chapter 5 computes sample sizes and the false discovery rate: detecting a ten per cent lift needs 163,095 impressions per creative. Chapter 6 shows that under adaptive platform allocation the reported relative difference between two identical creatives misses the truth by 0.348 on average, and that Thompson sampling lowers regret from 317.8 clicks under an equal split to a median of 187.1, at the price of biased estimates of the differences. Chapter 7 reads the Sunzi, Liu Xie, Leibniz and Ecclesiastes. Chapter 8 gives the fatigue-and-remix ledger, a rendering and naming pipeline, the distilled Creative Fatigue Half-Life and Orthogonal Remix Scheduler, checker specifications grouped by the four kits and pre-registered tests. Boxes separate published evidence, runnable Python calculations (22 programs), classical sources and hypotheses. 摘要:买量圈流行的说法是:扩量期素材衰退得快,因为观众重复曝光使点击率按可估计的半衰期衰减;把短视频拆成开头钩子、痛点、证据、行动召唤四个模块,用正交表组合,十几个镜头就能自动合成上百条低相关的测试素材,突破产能瓶颈。本书逐条核对文献并用程序检验,做了蓝图修正:疲劳不是单调指数线,重复效应先升后降,停投后会部分恢复,后台下滑还混着竞价环境;正交表只估主效应,交互作用被混叠,「低相关」只是设计矩阵的性质;广告实验的统计功效极低,上百条素材同时比较必须控制错误发现率;平台按预测表现分配曝光,素材测试因此有偏。书中参数为示意,全部计算由书中程序生成。本书是非个人化经营分析,不构成个人化经营、法律或理财建议,不承诺任何投放结果。 第一章把三种疲劳曲线写成公式,从合成日志拟合半衰期与自助法区间,并用新观众校正环境,示意参数下环境漂移占汇总下滑的 0.211。第二章用纳伊克型模型估出磨损半衰期 8.42 个在播日、恢复半程 3.92 个停播日,并推出频次上限与轮换条数。第三章从零构造 L4、L8、L9、L16,L9 的对比方差为 0.6667,逐个改因素为 2.0000,但钩子与召唤的配合会全部落到痛点模块上。第四章区分设计距离、表现相关与共享镜头的疲劳:12 个镜头重组出的素材池,对观众只相当于两到三条全新素材,平均相对点击率 0.828。 第五章计算样本量与错误发现率:测出一成的点击率提升,每条素材需要 163,095 次曝光。第六章说明平台自适应分配下,两条相同素材报表上的相对差异平均偏离真值 0.348;汤普森抽样把遗憾从均分的 317.8 次点击降到中位数 187.1 次,代价是差异估计有偏。第七章考辨《孙子·兵势》五声之变、刘勰《文心雕龙·通变》、莱布尼茨《论组合术》与《传道书》。第八章给出素材疲劳与混剪台账、渲染命名流水线、蒸馏小工具「素材疲劳半衰期与正交混剪排程试算器」、按四件套分组的检查器规格与预注册检验。素材只用自有或已授权镜头,零伪造证言与前后对比,AI 合成画面按平台现行规则如实标注。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Fatigue Curve Is Not an Exponential: the Two-Factor Process of Repetition and the Estimation of Half-Lives — Chapter 1 writes three fatigue curves as formulas, fits a half-life with a bootstrap interval to a synthetic log and corrects for the environment with fresh viewers; under illustrative parameters environmental drift accounts for 0.211 of the aggregate decline. (https://5dlibrary.com/answers/en/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c01/) 第1章 疲劳曲线不是一条指数线:重复曝光的两因素过程与半衰期估计 — 第一章把三种疲劳曲线写成公式,从合成日志拟合半衰期与自助法区间,并用新观众校正环境,示意参数下环境漂移占汇总下滑的 0.211。 (https://5dlibrary.com/answers/zh/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c01/) - Ch.2 Pausing Brings Recovery: Naik-Type Dynamics, Frequency Caps and the Size of a Rotation — Chapter 2 estimates, with a Naik-type model, a wear-out half-life of 8.42 days on air and a recovery half-time of 3.92 days off air, and derives a frequency cap and a rotation size. (https://5dlibrary.com/answers/en/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c02/) 第2章 停投会恢复:纳伊克型动态、频次上限与轮换条数 — 第二章用纳伊克型模型估出磨损半衰期 8.42 个在播日、恢复半程 3.92 个停播日,并推出频次上限与轮换条数。 (https://5dlibrary.com/answers/zh/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c02/) - Ch.3 What an Orthogonal Array Can Estimate: Building L4, L8, L9 and L16, Main-Effect Efficiency and the Structure of Aliasing — Chapter 3 builds L4, L8, L9 and L16 from scratch; the contrast variance is 0.6667 under L9 against 2.0000 one factor at a time, but a hook-by-call-to-action synergy lands wholly on the pain-point module. (https://5dlibrary.com/answers/en/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c03/) 第3章 正交表能估什么:L4、L8、L9、L16 的构造、主效应效率与混叠结构 — 第三章从零构造 L4、L8、L9、L16,L9 的对比方差为 0.6667,逐个改因素为 2.0000,但钩子与召唤的配合会全部落到痛点模块上。 (https://5dlibrary.com/answers/zh/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c03/) - Ch.4 How Different Are Recombined Creatives? Hamming Distance, Correlated Performance and the Shared Fatigue of Shared Shots — Chapter 4 separates design distance, correlated performance and the shared fatigue of shared shots: a pool recombined from 12 shots is as fresh to viewers as two or three new creatives, with an average relative click-through rate of 0.828. (https://5dlibrary.com/answers/en/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c04/) 第4章 重组素材有多不同:汉明距离、表现相关与共享镜头的疲劳 — 第四章区分设计距离、表现相关与共享镜头的疲劳:12 个镜头重组出的素材池,对观众只相当于两到三条全新素材,平均相对点击率 0.828。 (https://5dlibrary.com/answers/zh/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c04/) - Ch.5 How Many Winners Can a Hundred Creatives Reveal? Sample Size, the False Discovery Rate and the Signal-to-Noise Ratio of Advertising Returns — Chapter 5 computes sample sizes and the false discovery rate: detecting a ten per cent lift needs 163,095 impressions per creative. (https://5dlibrary.com/answers/en/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c05/) 第5章 一百条素材能找出几条赢家:样本量、错误发现率与广告回报的信噪比 — 第五章计算样本量与错误发现率:测出一成的点击率提升,每条素材需要 163,095 次曝光。 (https://5dlibrary.com/answers/zh/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c05/) - Ch.6 The Platform Will Not Run Your Experiment: the Bias of Allocating by Prediction and Thompson-Sampling Races — Chapter 6 shows that under adaptive platform allocation the reported relative difference between two identical creatives misses the truth by 0.348 on average, and that Thompson sampling lowers regret from 317.8 clicks under an equal split to a median of 187.1, at the price of biased estimates of the differences. (https://5dlibrary.com/answers/en/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c06/) 第6章 平台不会替你做随机实验:按预测分配曝光的偏差与汤普森抽样赛马 — 第六章说明平台自适应分配下,两条相同素材报表上的相对差异平均偏离真值 0.348;汤普森抽样把遗憾从均分的 317.8 次点击降到中位数 187.1 次,代价是差异估计有偏。 (https://5dlibrary.com/answers/zh/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c06/) - Ch.7 Classical Sources: the Changes of Five Notes, the Art of Combination and Nothing New Under the Sun — Leibniz's count is correct. The question is what it is used to count. The non-empty subsets of 12 shots number 4,095, but a valid creative must take exactly one shot from each of the four modules, and there are only 81 such combinations, 0.0198 of all the subsets. If the order of the four modules may also vary, there are 1,944; multiplied by two cut lengths, 3,888. These numbers show that "a hundred creatives" or even "a thousand" is easy to reach by counting. What is hard is that most combinations are structurally invalid, and many of the valid ones repeat one another in the viewer's perception. (https://5dlibrary.com/answers/en/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c07/) 第7章 原典考辨:五声之变、组合之术与日光之下 — 莱布尼茨的计数对,问题在于用它数什么。12 个镜头的全部非空子集有 4,095 种,但一条合格的素材,必须从四个模块各取恰好一个镜头,这样的组合只有 81 种,占全部子集的 0.0198。若四个模块的顺序也可以变,是 1,944 种;再乘以两种剪辑长度,是 3,888 种。这些数字说明,「上百条」甚至「上千条」在计数上都不难达到;难的是,绝大多数组合在结构上无效,或在知觉上重复。 (https://5dlibrary.com/answers/zh/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c07/) - Ch.8 Closing the Loop: the Fatigue-and-Remix Ledger, Field Tools and Checker Specifications — Chapter 8 gives the fatigue-and-remix ledger, a rendering and naming pipeline, the distilled Creative Fatigue Half-Life and Orthogonal Remix Scheduler, checker specifications grouped by the four kits and pre-registered tests. Boxes separate published evidence, runnable Python calculations (22 programs), classical sources and hypotheses. (https://5dlibrary.com/answers/en/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c08/) 第8章 实证闭环:素材疲劳与混剪台账、实战工具与检查器规格 — 第八章给出素材疲劳与混剪台账、渲染命名流水线、蒸馏小工具「素材疲劳半衰期与正交混剪排程试算器」、按四件套分组的检查器规格与预注册检验。素材只用自有或已授权镜头,零伪造证言与前后对比,AI 合成画面按平台现行规则如实标注。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/ad-creative-fatigue-half-lives-orthogonal-storyboard-recombination-automated-permutation-rendering/c08/) ## 5DL-B-401 · Zero-Ticket Self-Repairing Order Interfaces, Self-Service Correction State Machines and Pre-Emptive Support Elimination 中文书名:日入百单零客服自愈界面、错填自助修正状态机与售后工单源头消灭母典 DOI: pending Cite: Fan, Y. (2026). Zero-Ticket Self-Repairing Order Interfaces, Self-Service Correction State Machines and Pre-Emptive Support Elimination (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations come from the book's programs; the book is general business analysis, not individual or legal advice. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: Once a small shop passes a hundred orders a day, a popular saying holds that a good interface brings customer service to zero and removes nine tenths of after-sale messages at the source. This book turns the slogan into testable claims and finds that concentration is real but nine tenths is out of reach. The elimination rate equals the sum, over self-serviceable classes, of share times self-service completion, less the enquiries that self-service creates. In a synthetic month the self-serviceable classes hold only 0.869 of tickets, since refunds, payments and miscellaneous requests must stay with people; once uncertainty in completion and new enquiries is sampled, the ninety per cent interval for elimination runs from 0.468 to 0.601, and reaching 0.9 would need a uniform completion rate of 1.086, which is arithmetically impossible. "Zero-ticket" is therefore restated as a target interval for falling human intervention. Parameters are illustrative and all calculations come from the book's programs; the book is general business analysis, not individual or legal advice. Chapter 1 clusters tickets with a home-made tokenizer, Jaccard similarity and single-link merging, reaching purity 0.888 at threshold 0.30, and shows that clusters become root causes only when a person names them. Chapter 2 gives the ceiling and its interval. Chapter 3 writes delivery and correction as a state machine, enumerates 234 reachable configurations, proves there is no deadlock and no endless resending with a person one step away, and shows an idempotency key removing the 30.3 duplicate e-mails per thousand requests caused by retries. Chapter 4 compares three attempt limits; merging repeats within twenty minutes nearly ends harm to genuine users, and verifying the holder with receipt evidence keeps hijacks near 0.0182 a year. Chapter 5 finds that one self-service recompute within 24 hours catches 0.981 of genuine errors while keeping precision stamps and never filling in a birth time, eliminating 0.754 of wrong-input tickets. Chapter 6 shows chases per hundred orders falling from 41.0 to 16.5 as visible milestones go from none to two, and uses Erlang C to show that a small shop saves busy time rather than staff. Chapter 7 examines the Han Feizi on the dike breached through an ant hole, the Laozi on acting before a thing exists, and Babbage's 1832 account of the division of mental labour: fixing a flaw in week twenty-six costs 27.1 times a check before launch. Chapter 8 gives the ledger, the distilled Ticket Root-Cause and Self-Service State Machine Checker, checker specifications grouped by four kits and a pre-registered test needing 1,089 orders per arm to detect a fall of three tenths. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:小店日入百单之后,流行说法是:界面做好,客服归零,九成以上售后消息可以在源头消灭。本书把这句口号拆成可检验的命题,结论是集中属实、九成不可达。消灭率等于各可自助类别份额乘自助完成率之和,再扣掉自助流程自己引出的新咨询。在合成的一个月里,可自助类别合计只占 0.869,退款、付款与杂项必须留给人;把完成率与新咨询的不确定性抽样进去,消灭率九成区间为 0.468 到 0.601,要达到 0.9 需要统一完成率 1.086,算术上做不到。「零客服」因此写成人工介入率下降的目标区间。书中参数为示意,全部计算由书中程序生成;本书是非个人化经营建议,不构成法律意见。 第一章用自写分词、Jaccard 相似度与单链聚类给工单分簇,阈值 0.30 时纯度 0.888,并说明簇必须由人命名才成为根因。第二章给出消灭率上界与区间。第三章把交付与修正流程写成状态机,穷举 234 个可达配置,证明无死锁、无无限重发,人工一步可达;幂等键消除网络重试造成的每千次 30.3 封多发邮件。第四章比较三种次数锁,合并二十分钟内重复请求后真人误伤几乎归零;换绑先用收据凭证验证持有人,每年劫持约 0.0182 次。第五章算出 24 小时一次的自助重算接住 0.981 的真错填,并守住本命站精度钢印与不自补时刻的规定,消灭错填类人工工单的 0.754。第六章显示可见进度节点从零到两个,每百单催问从 41.0 次降到 16.5 次,并用 Erlang C 说明小店省下的是忙碌时间而非值守人数。 第七章考辨《韩非子·喻老》「千丈之堤,以蝼蚁之穴溃」、《老子》「为之于未有」与巴贝奇 1832 年论脑力分工:拖到第二十六周再修漏洞,代价是上线前检查的 27.1 倍。第八章给出台账字段表、蒸馏小工具「工单根因聚类与自助修正状态机检查器」、按四件套分组的检查器规格与预注册检验:日入百单时检出三成下降每组需 1,089 单。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者假说。 Chapter answers: - Ch.1 Root-Cause Clustering of Tickets: What a Month of After-Sale Messages Is Made Of — Chapter 1 clusters tickets with a home-made tokenizer, Jaccard similarity and single-link merging, reaching purity 0.888 at threshold 0.30, and shows that clusters become root causes only when a person names them. (https://5dlibrary.com/answers/en/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c01/) 第1章 工单根因聚类:一个月的售后消息由什么构成 — 第一章用自写分词、Jaccard 相似度与单链聚类给工单分簇,阈值 0.30 时纯度 0.888,并说明簇必须由人命名才成为根因。 (https://5dlibrary.com/answers/zh/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c01/) - Ch.2 The Ceiling on Elimination: Class Share Times Self-Service Completion — The elimination rate equals the sum of share times completion across self-serviceable classes, minus the enquiries that self-service creates. In this example the self-serviceable share is 0.869, which is the ceiling; once the uncertainty in completion and in new enquiries is sampled, the ninety per cent interval runs from 0.468 to 0.601. Nine tenths is out of reach because refunds, payments and miscellaneous requests must stay with people, and they account for 0.131. (https://5dlibrary.com/answers/en/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c02/) 第2章 消灭率的上界:类别份额乘自助完成率 — 消灭率等于各可自助类别份额乘完成率之和,减去自助流程新引出的咨询。本例中可自助份额是 0.869,这是天花板;把完成率与新咨询的不确定性都抽样进去,九成区间是 0.468 到 0.601。九成之所以不可达,是因为退款、付款与杂项必须留给人,而这一块占了 0.131。 (https://5dlibrary.com/answers/zh/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c02/) - Ch.3 The Self-Service Correction State Machine: Guards, Idempotency and Exhaustive Checking — Chapter 3 writes delivery and correction as a state machine, enumerates 234 reachable configurations, proves there is no deadlock and no endless resending with a person one step away, and shows an idempotency key removing the 30.3 duplicate e-mails per thousand requests caused by retries. (https://5dlibrary.com/answers/en/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c03/) 第3章 自助修正状态机:守卫、幂等与穷举验证 — 第三章把交付与修正流程写成状态机,穷举 234 个可达配置,证明无死锁、无无限重发,人工一步可达;幂等键消除网络重试造成的每千次 30.3 封多发邮件。 (https://5dlibrary.com/answers/zh/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c03/) - Ch.4 Abuse Limits and Rebind Verification: Stopping Scripts Without Stopping People — Chapter 4 compares three attempt limits; merging repeats within twenty minutes nearly ends harm to genuine users, and verifying the holder with receipt evidence keeps hijacks near 0.0182 a year. (https://5dlibrary.com/answers/en/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c04/) 第4章 防滥用次数锁与换绑验证:拦住脚本,不拦真人 — 第四章比较三种次数锁,合并二十分钟内重复请求后真人误伤几乎归零;换绑先用收据凭证验证持有人,每年劫持约 0.0182 次。 (https://5dlibrary.com/answers/zh/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c04/) - Ch.5 One Timed Self-Service Recompute: Window Length, Precision Stamps and Misuse — Chapter 5 finds that one self-service recompute within 24 hours catches 0.981 of genuine errors while keeping precision stamps and never filling in a birth time, eliminating 0.754 of wrong-input tickets. (https://5dlibrary.com/answers/en/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c05/) 第5章 限时一次自助重算:窗口长度、精度钢印与误用 — 第五章算出 24 小时一次的自助重算接住 0.981 的真错填,并守住本命站精度钢印与不自补时刻的规定,消灭错填类人工工单的 0.754。 (https://5dlibrary.com/answers/zh/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c05/) - Ch.6 Progress Transparency and the Residual Human Desk: a Chase Model and Erlang C — Chapter 6 shows chases per hundred orders falling from 41.0 to 16.5 as visible milestones go from none to two, and uses Erlang C to show that a small shop saves busy time rather than staff. (https://5dlibrary.com/answers/en/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c06/) 第6章 进度透明与剩余人工值守:催单模型与 Erlang C — 第六章显示可见进度节点从零到两个,每百单催问从 41.0 次降到 16.5 次,并用 Erlang C 说明小店省下的是忙碌时间而非值守人数。 (https://5dlibrary.com/answers/zh/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c06/) - Ch.7 Classical Sources East and West: the Ant Hole and the Division of Mental Labour — Chapter 7 examines the Han Feizi on the dike breached through an ant hole, the Laozi on acting before a thing exists, and Babbage's 1832 account of the division of mental labour: fixing a flaw in week twenty-six costs 27.1 times a check before launch. (https://5dlibrary.com/answers/en/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c07/) 第7章 东西方原典考辨:蝼蚁之穴与脑力分工 — 第七章考辨《韩非子·喻老》「千丈之堤,以蝼蚁之穴溃」、《老子》「为之于未有」与巴贝奇 1832 年论脑力分工:拖到第二十六周再修漏洞,代价是上线前检查的 27.1 倍。 (https://5dlibrary.com/answers/zh/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c07/) - Ch.8 The Empirical Loop: Field Tools and Checker Specifications — Chapter 8 gives the ledger, the distilled Ticket Root-Cause and Self-Service State Machine Checker, checker specifications grouped by four kits and a pre-registered test needing 1,089 orders per arm to detect a fall of three tenths. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c08/) 第8章 实证闭环:实战工具与检查器规格 — 第八章给出台账字段表、蒸馏小工具「工单根因聚类与自助修正状态机检查器」、按四件套分组的检查器规格与预注册检验:日入百单时检出三成下降每组需 1,089 单。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/zero-ticket-self-repairing-order-interfaces-self-service-correction-state-machines/c08/) ## 5DL-B-402 · Multi-Jurisdiction Digital Revenue Clearing, Cross-Border Tax-Base Compliance Boundaries and Audit-Ready Automated Ledgers 中文书名:跨国数字业务多主体清分、跨境税基合规边界与一键审计就绪账本母典 DOI: pending Cite: Fan, Y. (2026). Multi-Jurisdiction Digital Revenue Clearing, Cross-Border Tax-Base Compliance Boundaries and Audit-Ready Automated Ledgers (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: The book is a handbook of accounting and compliance routines and is not personal tax advice. Parameters are illustrative, data synthetic, and all calculations come from the book's programs. Abstract: Among small teams that sell digital goods across borders, a popular belief holds that the danger in the books is omission: record every payment and nothing can go badly wrong. This book finds the belief incomplete. The larger risks come from three habits that omit nothing: recording revenue at the net amount a platform pays out, recording foreign revenue at the rate of the day it is converted, and sharing costs among related entities informally. The book is a handbook of accounting and compliance routines and is not personal tax advice. Rates, thresholds and registration duties are stated as general rules with source and year and listed for human review; no tax conclusion is given and no arrangement for reducing a tax base is described. Parameters are illustrative, data synthetic, and all calculations come from the book's programs. The blueprint is corrected: gross or net turns on control, not on how a platform pays; Taiwan's registration threshold for cross-border electronic services changed in 2025; no Vietnamese rate is stated because the sources conflict; and "ten minutes", "one click" and "audit-ready" are restated as quantiles with refutation conditions. Chapter 1 scores seven synthetic deals under the control judgement of IFRS 15; the cash-landed rule errs on 3 of them and overstates the total by 0.2763. Chapter 2 compares transaction-date and conversion-day recording under IAS 21; the mean absolute revenue gap is 0.0602, mostly from the shift between periods. In Chapter 3, multilateral netting settles 64 internal invoices among five entities in 4 transfers and moves 0.2518 of the gross amount. Chapter 4 shows that one entity's share of a cost pool moves from 0.2100 by usage to 0.7001 by revenue, so keys must be written in advance and follow benefit. In Chapter 5 four-way reconciliation matches 0.9500 of orders automatically, a tolerance of 0.03 clears exactly the benign residues, and cross-entity duplicates are found only by original document number. Chapter 6 finds that filing conversion slips on the day raises median evidence-chain completeness from 0.8525 to 0.9425; with an exception rate of 0.005 cleared as it arises, the month-end close finishes within ten minutes in 0.9035 of months, against 0.0145 when exceptions wait for month end. Chapter 7 examines the four-pillar form of the History of the Ming and Pacioli's double entry of 1494: the pillars catch differences from the count but not duplicates across books, and the trial balance detects omissions, duplicates and wrong-entity postings at a rate of 0.0000. Chapter 8 sets out field tools and checker specifications, the month-end ledger, a pre-registered test with its refutation condition, and the runnable Multi-Entity Clearing and Month-End Audit-Ready Reconciler. Boxes separate published evidence, runnable Python calculations (21 programs), early sources and hypotheses. 摘要:跨国做数字生意的小团队里流行一种说法:账务风险在于漏记,只要每笔钱都记上就不会出事。本书核对准则与文献后认为,这一说法并不完整:更大的风险来自三个不漏记的习惯,即按平台到账净额记收入、按结汇那天的汇率记收入、关联主体之间随手分摊费用。本书是会计与合规流程的工具书,非个人税务建议;书中税率、门槛与登记要求只写法规通则口径,注明官方来源与年份并列入须人工复核清单,不给任何税额结论,不写任何压低税基的安排。书中参数为示意,数据为合成;全部计算由书中程序生成。核对原文后,本书把蓝图修正为:总额还是净额取决于控制权,不取决于打款方式;台湾跨境电子劳务的登记门槛已在 2025 年调整;越南的具体税率因资料互相矛盾不写;「十分钟」「一键」「审计就绪」改写为分位数与推翻条件。 第一章按 IFRS 15 的控制判断给七类合成交易打分,「到账即收入」有 3 类出错,误差有正有负,合计高估 0.2763。第二章按 IAS 21 比较交易日入账与结汇日入账,平均绝对收入偏差 0.0602,主要来自期间错位。第三章五个主体 64 张内部发票经多边轧差只需 4 笔转账,移动金额降到 0.2518。第四章说明同一成本池按收入比例与按使用量分摊,同一主体的承担比例从 0.2100 变到 0.7001,分摊键必须事前写下并跟随受益。第五章四方对账的自动匹配率为 0.9500,容差 0.03 恰好清掉良性尾差,跨主体重复只能靠原始单据号找出。 第六章显示结汇凭证当天归档可把证据链完整度中位数从 0.8525 提到 0.9425;例外率 0.005 且随到随清时,月末十分钟内关账的比例为 0.9035,攒到月末一次处理只有 0.0145。第七章考辨《明史·食货志》的四柱式与帕乔利 1494 年的复式账簿:四柱查得出账实不符,却查不出跨册重复;试算平衡对漏记、重复与记错主体的检出率为 0.0000。第八章给出实战工具与检查器规格、月末关账台账、预注册检验与推翻条件,以及可运行的「多主体清分与月末审计就绪对账器」。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Gross or Net: Judging the Principal by Control, Not by How the Platform Pays — Chapter 1 scores seven synthetic deals under the control judgement of IFRS 15; the cash-landed rule errs on 3 of them and overstates the total by 0.2763. (https://5dlibrary.com/answers/en/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c01/) 第1章 总额还是净额:按控制权判主体,而不是按平台怎么打款 — 第一章按 IFRS 15 的控制判断给七类合成交易打分,「到账即收入」有 3 类出错,误差有正有负,合计高估 0.2763。 (https://5dlibrary.com/answers/zh/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c01/) - Ch.2 When the Currency Turns: Transaction-Date Rates, Closing-Rate Remeasurement and Decimal Rounding — Chapter 2 compares transaction-date and conversion-day recording under IAS 21; the mean absolute revenue gap is 0.0602, mostly from the shift between periods. In Chapter 3, multilateral netting settles 64 internal invoices among five entities in 4 transfers and moves 0.2518 of the gross amount. (https://5dlibrary.com/answers/en/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c02/) 第2章 结汇时点:交易日汇率、期末重估与十进制舍入 — 第二章按 IAS 21 比较交易日入账与结汇日入账,平均绝对收入偏差 0.0602,主要来自期间错位。 (https://5dlibrary.com/answers/zh/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c02/) - Ch.3 Clearing Among Entities: Bilateral and Multilateral Netting and Open Balances — Section B lists the net positions. The destiny entity is a net receiver of +17,449.23, the library entity of +11,350.45 and the shared-service entity of +22,272.25; the marketing entity is a net payer of -34,708.64 and the purchasing entity of -16,363.29. Section C checks that the five positions add to exactly 0.00. That is a necessary condition for correct netting: whenever the total is not zero, some invoice has been recorded on one side only. Section D gives the ratio: after multilateral netting the amount actually moved is 0.2518 of what full payment would move. (https://5dlibrary.com/answers/en/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c03/) 第3章 多主体清分:双边轧差、多边轧差与在途余额 — 模拟假设内部往来每天平均产生 6,761.68 个示意单位,即上一个合成月份的总额平均摊到三十天;每次结算要两天才到账。A 段给出四种周期。每天结算,一个月要 120.00 笔转账,平均在途余额按利特尔公式为 20,285.05,逐日模拟为 20,251.24。每周结算,转账降到 17.14 笔,在途余额升到约 40,570.09。半月结算,8.00 笔,在途约 67,616.82。每月结算,4.00 笔,在途约 118,329.44,逐日模拟为 117,653.27。模拟值略低于公式值,是因为模拟从零开始,头几天还没有在途的钱,再加上模拟天数不能被每个周期整除。 (https://5dlibrary.com/answers/zh/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c03/) - Ch.4 Shared Costs Between Related Entities: the Arm's Length Principle, Allocation Keys and Contemporaneous Documentation — Chapter 4 shows that one entity's share of a cost pool moves from 0.2100 by usage to 0.7001 by revenue, so keys must be written in advance and follow benefit. In Chapter 5 four-way reconciliation matches 0.9500 of orders automatically, a tolerance of 0.03 clears exactly the benign residues, and cross-entity duplicates are found only by original document number. (https://5dlibrary.com/answers/en/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c04/) 第4章 关联分摊:正常交易原则、分摊键与同期文档 — 第四章说明同一成本池按收入比例与按使用量分摊,同一主体的承担比例从 0.2100 变到 0.7001,分摊键必须事前写下并跟随受益。 (https://5dlibrary.com/answers/zh/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c04/) - Ch.5 Four-Way Reconciliation: Auto-Match Rate, Tolerance and Cross-Entity Duplicates — Section A shows that 79 of the 2,000 lines carry genuine faults. With a tolerance of 0.00 the automatic match rate is 0.8985, the queue holds 203 lines, and no genuine fault is let through. A tolerance of 0.01 raises the rate to 0.9200 with 160 in the queue. At 0.03 the rate is 0.9605 and the queue holds 79 lines, exactly the genuine faults: every benign residue is cleared and no fault is swallowed. At 0.10 nothing changes. At 0.50 the rate rises to 0.9630 and the queue falls to 74, but 5 genuine faults have been swallowed; at 1.00 the rate is 0.9680, the queue 64, and the swallowed faults rise to 15. (https://5dlibrary.com/answers/en/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c05/) 第5章 四方对账:自动匹配率、容差与跨主体重复入账 — A 段显示,这 2,000 条里有 79 条真实错误。容差为 0.00 时,自动匹配率 0.8985,队列 203 条,没有任何真错误被放过。容差放到 0.01,匹配率 0.9200,队列 160 条。放到 0.03,匹配率 0.9605,队列 79 条,恰好只剩真错误,良性尾差全部清掉,没有一条真错误被吞掉。再放到 0.10,结果不变。放到 0.50,匹配率 0.9630,队列 74 条,却有 5 条真错误被吞掉;放到 1.00,匹配率 0.9680,队列 64 条,被吞掉的真错误升到 15 条。 (https://5dlibrary.com/answers/zh/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c05/) - Ch.6 Evidence Chains, First-Digit Leads and Month-End Close Time — Chapter 6 finds that filing conversion slips on the day raises median evidence-chain completeness from 0.8525 to 0.9425; with an exception rate of 0.005 cleared as it arises, the month-end close finishes within ten minutes in 0.9035 of months, against 0.0145 when exceptions wait for month end. (https://5dlibrary.com/answers/en/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c06/) 第6章 证据链、首位数线索与月末关账耗时 — 第六章显示结汇凭证当天归档可把证据链完整度中位数从 0.8525 提到 0.9425;例外率 0.005 且随到随清时,月末十分钟内关账的比例为 0.9035,攒到月末一次处理只有 0.0145。 (https://5dlibrary.com/answers/zh/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c06/) - Ch.7 Four Pillars and Double Entry: the Ming Registers and Pacioli's Ledger Examined — Chapter 7 examines the four-pillar form of the History of the Ming and Pacioli's double entry of 1494: the pillars catch differences from the count but not duplicates across books, and the trial balance detects omissions, duplicates and wrong-entity postings at a rate of 0.0000. (https://5dlibrary.com/answers/en/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c07/) 第7章 四柱与复式:《明史·食货志》与帕乔利账簿考辨 — 第七章考辨《明史·食货志》的四柱式与帕乔利 1494 年的复式账簿:四柱查得出账实不符,却查不出跨册重复;试算平衡对漏记、重复与记错主体的检出率为 0.0000。 (https://5dlibrary.com/answers/zh/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c07/) - Ch.8 Field Tools and Checker Specifications: the Month-End Ledger, the Four Kits and a Pre-Registered Test — Chapter 8 sets out field tools and checker specifications, the month-end ledger, a pre-registered test with its refutation condition, and the runnable Multi-Entity Clearing and Month-End Audit-Ready Reconciler. Boxes separate published evidence, runnable Python calculations (21 programs), early sources and hypotheses. (https://5dlibrary.com/answers/en/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c08/) 第8章 实战工具与检查器规格:月末关账台账、四件套与预注册检验 — 第八章给出实战工具与检查器规格、月末关账台账、预注册检验与推翻条件,以及可运行的「多主体清分与月末审计就绪对账器」。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/multi-jurisdiction-digital-revenue-clearing-cross-border-tax-base-audit-ready-ledgers/c08/) ## 5DL-B-403 · Antifragile Reputation Judo, Evidence-Backed Smear Neutralisation and Platform-Independent Audience Sovereignty 中文书名:商业声誉反脆弱柔道、恶意差评证据链反制与去平台依赖私域主权母典 DOI: pending Cite: Fan, Y. (2026). Antifragile Reputation Judo, Evidence-Backed Smear Neutralisation and Platform-Independent Audience Sovereignty (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is general business guidance, not legal advice. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: A popular belief among merchants holds that bad reviews should be buried, deleted or ignored, and that a shop is safe as long as its platform account stays in good standing. This book breaks both beliefs into propositions that data could overturn. After checking the sources it corrects its commissioning blueprint: the study reporting a rise of 0.12 stars after managers began replying attributes the rise to a change in who writes reviews and did not measure bystanders' trust; the blemishing and pratfall effects hold only when a positive impression comes first and the negative item is light; a hash chain proves that a record existed at a moment, not that it is true; and several titles and an author order are corrected. Its parameters are illustrative; all calculations are generated by its programs; it is general business guidance, not legal advice. Chapter 1 shows how under-reporting bends ratings into a J shape, leaving the page average about 0.410 stars above the true experience. Chapter 2 writes the bystander's ledger as a Bayesian update: a bystander with a prior of 0.80 falls to 0.516 after one extreme review and rises to 0.901 after a calm reply with checkable evidence; with a calm tone at least 0.575 of a reply's claims must be checkable to reverse the review, and a hostile tone never does. Chapter 3 covers whom to answer, how and how fast, with a red-line check for reply copy. Chapter 4 shows that clues of malice are only clues: with all three clues present and a base rate of 0.05, the posterior probability of malice is only 0.341. Chapter 5 keeps delivery evidence in a SHA-256 hash chain with daily Merkle roots: tampering that recomputes every later hash is caught by the chain alone at a rate of 0.000 and at 1.000 with daily external anchors, and personal data never enter the chain. Chapter 6 compares two audience strategies under Poisson shocks and Markov flows; the 99th percentile of the twenty-four-month revenue loss falls from 0.793 on the platform alone to 0.437 with a steadily built owned list. Chapter 7 examines the Laozi on water, the Sunzi on being unbeatable and Nietzsche's maxim, defines antifragility through Jensen's inequality and uses survivorship bias to separate Nietzsche from Taleb; the single-shock safe line for the list share is 0.444. Chapter 8 gives the distilled Public Reply Reversal and Delivery Hash-Chain Checker, checker specifications grouped by the four house kits, the ledger fields and a pre-registered plan: detecting a shift of 0.12 stars needs 1,598 reviews per arm. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:店家圈子的流行说法是:差评要压、要删、要沉默;粉丝都在平台上,账号在,生意就稳。本书把这两句话拆成可以被数据推翻的命题。核对原文后,本书修正立项蓝图:管理者回应后评分升高 0.12 星的研究,归因于评论者的选择变化,并未直接测量旁观者信任;瑕疵效应与出丑效应只在先有正面印象、负面信息轻微时成立;哈希链只证明记录在某一时刻已经存在,不证明内容为真;另更正若干文献题目与作者顺序。书中参数为示意;全部计算由书中程序生成;本书是非个人化经营建议,不是法律意见。 第一章说明少报偏差把评价压成 J 形,页面平均分比真实体验高出约 0.410 星。第二章用贝叶斯更新写旁观者的账:先验 0.80 的旁观者读完一条极端差评降到 0.516,读到冷静且附可核验证据的回应后升到 0.901;冷静语气下可核验比例至少 0.575 才能反转差评,敌意语气无解。第三章讨论回应谁、怎么回、多快回,并给出回应文案的红线检查。第四章说明恶意线索只能作线索:三条线索全亮、基础比率 0.05 时,恶意后验只有 0.341。 第五章用 SHA-256 哈希链与默克尔日根保存交付证据:改记录并重算后续哈希的篡改,只靠链条检出率为 0.000,加每日外部锚点后为 1.000;个人资料一律不入链。第六章用泊松冲击与马尔可夫流转比较两种受众策略,二十四个月收入损失的第 99 百分位从只靠平台的 0.793 降到持续沉淀自有名录的 0.437。第七章考辨《老子》论水、《孙子》论不可胜与尼采格言,用詹森不等式定义反脆弱,并以幸存者偏差区分尼采与塔勒布;单次冲击的名录安全线为 0.444。第八章给出蒸馏小工具「差评公开回应反转条件与交付哈希证据链检查器」、按四件套分组的检查器规格、台账字段与预注册方案:检出 0.12 星的变化每组需要 1,598 条评价。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 J-Shaped Ratings and the Weight of One Bad Review: How Reviews Move Sales and Why the Average Misleads — Chapter 1 shows how under-reporting bends ratings into a J shape, leaving the page average about 0.410 stars above the true experience. (https://5dlibrary.com/answers/en/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c01/) 第1章 J 形评论与一条差评的分量:评价如何影响成交,平均分为何失真 — 第一章说明少报偏差把评价压成 J 形,页面平均分比真实体验高出约 0.410 星。 (https://5dlibrary.com/answers/zh/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c01/) - Ch.2 The Bystander's Bayesian Ledger: One Extreme Review and One Public Reply — Chapter 2 writes the bystander's ledger as a Bayesian update: a bystander with a prior of 0.80 falls to 0.516 after one extreme review and rises to 0.901 after a calm reply with checkable evidence; with a calm tone at least 0.575 of a reply's claims must be checkable to reverse the review, and a hostile tone never does. (https://5dlibrary.com/answers/en/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c02/) 第2章 旁观者的贝叶斯账:一条极端差评加一条公开回应 — 第二章用贝叶斯更新写旁观者的账:先验 0.80 的旁观者读完一条极端差评降到 0.516,读到冷静且附可核验证据的回应后升到 0.901;冷静语气下可核验比例至少 0.575 才能反转差评,敌意语气无解。 (https://5dlibrary.com/answers/zh/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c02/) - Ch.3 Whom to Answer, How and How Fast: Reversal Conditions for Management Replies and the Red Lines of Reply Copy — Part B is the point of the exercise. The true effect is +0.12 stars, yet when a shop has only 30 reviews on each side the observed difference has a standard deviation of 0.297 and a chance of 0.316 of being negative. Nearly a third of shops would conclude that replying made their ratings worse. With 100 reviews on each side, the chance of seeing the wrong sign is still 0.225; with 300 it is 0.104; only at 1,000 does it fall to 0.013. (https://5dlibrary.com/answers/en/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c03/) 第3章 回应谁、怎么回、多快回:管理者回应的反转条件与文案红线 — B 段才是重点。真实效果是 +0.12 星,可是一家店前后各只有 30 条评价时,观察到的均值差标准差是 0.297,有 0.316 的机会看到的是负数,也就是说,近三分之一的店会以为回应让评分变差了。前后各 100 条时,看错方向的机会仍有 0.225;300 条时 0.104;到 1,000 条才降到 0.013。 (https://5dlibrary.com/answers/zh/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c03/) - Ch.4 Malicious Reviews Leave Clues, Not Verdicts: Text Overlap, Bunching in Time and Base Rates — Chapter 4 shows that clues of malice are only clues: with all three clues present and a base rate of 0.05, the posterior probability of malice is only 0.341. (https://5dlibrary.com/answers/en/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c04/) 第4章 恶意差评只有线索没有判决:文本重叠、时间聚集与基础比率 — 第四章说明恶意线索只能作线索:三条线索全亮、基础比率 0.05 时,恶意后验只有 0.341。 (https://5dlibrary.com/answers/zh/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c04/) - Ch.5 Hash Chains of Delivery Evidence and Daily Merkle Roots: Proving When a Record Existed, Not That It Is True — Chapter 5 keeps delivery evidence in a SHA-256 hash chain with daily Merkle roots: tampering that recomputes every later hash is caught by the chain alone at a rate of 0.000 and at 1.000 with daily external anchors, and personal data never enter the chain. (https://5dlibrary.com/answers/en/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c05/) 第5章 交付证据哈希链与默克尔日根:证明存在时刻,不证明内容为真 — 第五章用 SHA-256 哈希链与默克尔日根保存交付证据:改记录并重算后续哈希的篡改,只靠链条检出率为 0.000,加每日外部锚点后为 1.000;个人资料一律不入链。 (https://5dlibrary.com/answers/zh/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c05/) - Ch.6 Platform Shocks and the Owned List: Audience Stocks and Loss Quantiles under Poisson Shocks — Chapter 6 compares two audience strategies under Poisson shocks and Markov flows; the 99th percentile of the twenty-four-month revenue loss falls from 0.793 on the platform alone to 0.437 with a steadily built owned list. (https://5dlibrary.com/answers/en/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c06/) 第6章 平台突变与自有名录:泊松冲击下的受众存量与损失分位数 — 第六章用泊松冲击与马尔可夫流转比较两种受众策略,二十四个月收入损失的第 99 百分位从只靠平台的 0.793 降到持续沉淀自有名录的 0.437。 (https://5dlibrary.com/answers/zh/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c06/) - Ch.7 Water, Invincibility and the Hammer: Laozi, Sunzi, Nietzsche and Taleb on Antifragility — Chapter 7 examines the Laozi on water, the Sunzi on being unbeatable and Nietzsche's maxim, defines antifragility through Jensen's inequality and uses survivorship bias to separate Nietzsche from Taleb; the single-shock safe line for the list share is 0.444. (https://5dlibrary.com/answers/en/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c07/) 第7章 水、不可胜与锤子:老子、孙子、尼采与塔勒布的反脆弱考辨 — 第七章考辨《老子》论水、《孙子》论不可胜与尼采格言,用詹森不等式定义反脆弱,并以幸存者偏差区分尼采与塔勒布;单次冲击的名录安全线为 0.444。 (https://5dlibrary.com/answers/zh/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c07/) - Ch.8 Field Tools and Checker Specifications: the Reputation and Audience Ledger, Pre-Registered Tests and the Conditions for Refutation — Chapter 8 gives the distilled Public Reply Reversal and Delivery Hash-Chain Checker, checker specifications grouped by the four house kits, the ledger fields and a pre-registered plan: detecting a shift of 0.12 stars needs 1,598 reviews per arm. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c08/) 第8章 实战工具与检查器规格:声誉与受众主权台账、预注册检验与推翻条件 — 第八章给出蒸馏小工具「差评公开回应反转条件与交付哈希证据链检查器」、按四件套分组的检查器规格、台账字段与预注册方案:检出 0.12 星的变化每组需要 1,598 条评价。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/antifragile-reputation-judo-evidence-hash-chains-platform-independent-audience/c08/) ## 5DL-B-404 · Transitioning from Sprint to Sovereign Cruise: The Three-Hour Daily Decision Cap and Sustainable Creator Vitality 中文书名:冲刺期向巡航态平滑切换、每日三小时高杠杆拍板制与创作者永续生命力母典 DOI: pending Cite: Fan, Y. (2026). Transitioning from Sprint to Sovereign Cruise: The Three-Hour Daily Decision Cap and Sustainable Creator Vitality (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; energy and sleep are personal tracking trends only; the book is not medical advice and not personalised business or investment advice. Chapter 1 uses a concave hours-output curve and a parameter grid to show that the best daily duration falls in a narrow band, with 0.734 of the illustrative grid between two and four hours. Abstract: A popular belief among founders holds that once a cash-flow turnaround is won, one must keep sprinting around the clock to protect it. This book tests that belief and finds that it does not survive a change of conditions: once automation coverage, exception rate, cash runway and recovery time cross an admission line, further sprinting yields no positive marginal output, and narrowing the owner's top-leverage effort to a limited window, illustratively three hours, does not make long-run output worse. After checking the sources, the book corrects its blueprint in several places: factory data on working hours give only the direction of a concave curve; research on deliberate practice measured no universal four-hour ceiling; the parole-ruling pattern can arise from scheduling, and ego depletion came close to zero in multi-laboratory preregistered replications, so the case for the three-hour cap rests instead on queueing theory and switching costs; the fitness-fatigue model is cited from verifiable papers of 1976 and 1990, and the linear model is shown not to imply that sprinting is harmful. Its parameters are illustrative; all calculations are generated by its programs; energy and sleep are personal tracking trends only; the book is not medical advice and not personalised business or investment advice. Chapter 1 uses a concave hours-output curve and a parameter grid to show that the best daily duration falls in a narrow band, with 0.734 of the illustrative grid between two and four hours. Chapter 2 judges sprint, transition and cruise with Schmitt double thresholds on four gauges; the band cuts switching by a factor of 5.0. Chapter 3 treats the founder as a single server: raising utilisation from eighty to ninety-five per cent multiplies waiting by 4.75, and a single three-hour window lengthens the longest creative block from 3.34 to 5.68 hours. Chapter 4 bounds the evidence on decision fatigue and selects window items with a heap by value per minute, covering 0.682 of value under a heavy tail against 0.220 for first come, first served. Chapter 5 sets checker pass lines and a watch on reversal rates. Chapter 6 uses the fitness-fatigue impulse-response model to show that whether sprinting pays depends on an overreach coefficient each person must measure. Chapter 7 examines the line on tension and release in the Book of Rites and Seneca's On the Shortness of Life, and turns the conclusions into clauses for the four kits. Chapter 8 gives the cruise ledger, the distilled Sprint-to-Cruise Admission and Three-Hour Decision Queue, checker specifications and a preregistered single-case ABAB test whose power for a medium effect with fourteen-day phases is 0.758. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. 摘要:创业圈流行一种说法:现金流翻身仗打赢之后,要继续全天候冲刺才能守住战果。本书检验这一说法,结论是它在条件变化之后并不成立:当自动化覆盖率、例外率、现金跑道与事故恢复时间越过准入线,继续冲刺的边际产出不为正,把本人最高杠杆投入收敛到一个有限窗口(示意三小时)不会让长期产出更差。核对原文后,本书作了几处蓝图修正:工厂工时数据只给出凹形的方向;刻意练习研究没有测出普遍的四小时上限;假释裁决研究的图样可由排程造成,自我损耗在多实验室预注册复现中接近零效应,因此三小时制的理由改为排队论与注意力切换成本;体能—疲劳模型改引可核验的 1976 年与 1990 年论文,并说明线性模型本身推不出冲刺有害。书中参数为示意;全部计算由书中程序生成;精力与睡眠只作自我追踪趋势;本书不是医疗建议,也不是个人化经营或投资建议。 第一章用凹形工时曲线与参数网格说明最优日时长集中在一个不宽的区间,示意网格中 0.734 的组合落在两到四小时。第二章用四项指标的施密特双阈值判定冲刺、过渡与巡航,缓冲带把切换次数降为单一阈值的五分之一,倍数 5.0。第三章把创始人当作单服务台,利用率从八成升到九成五时等待放大 4.75 倍;单一三小时窗口把最长整块创作时间从 3.34 小时提到 5.68 小时。第四章说明决策疲劳证据须立界,并用堆排序按每分钟价值选题,重尾情形下覆盖价值 0.682,先到先办只有 0.220。 第五章给出检查器放行线与推翻率监控。第六章用体能—疲劳脉冲响应模型说明,冲刺是否划算取决于一个须自己测量的过度系数。第七章考辨《礼记·杂记下》「一张一弛」与塞涅卡《论生命之短暂》,并把结论落成四件套条款。第八章给出巡航台账、蒸馏小工具「冲刺转巡航准入与每日三小时拍板队列器」、检查器规格与单案例 ABAB 预注册检验,每阶段 14 天时中等效应的检验功效为 0.758。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Concave Curve of Hours and Output: When a Sprint Starts to Lose Money — Chapter 1 uses a concave hours-output curve and a parameter grid to show that the best daily duration falls in a narrow band, with 0.734 of the illustrative grid between two and four hours. (https://5dlibrary.com/answers/en/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c01/) 第1章 工时与产出的凹形:冲刺从哪一刻开始赔本 — 第一章用凹形工时曲线与参数网格说明最优日时长集中在一个不宽的区间,示意网格中 0.734 的组合落在两到四小时。 (https://5dlibrary.com/answers/zh/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c01/) - Ch.2 The Admission Line: Four Gauges and a Schmitt Double Threshold — Chapter 2 judges sprint, transition and cruise with Schmitt double thresholds on four gauges; the band cuts switching by a factor of 5.0. (https://5dlibrary.com/answers/en/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c02/) 第2章 准入线:四项指标与施密特双阈值 — 第二章用四项指标的施密特双阈值判定冲刺、过渡与巡航,缓冲带把切换次数降为单一阈值的五分之一,倍数 5.0。 (https://5dlibrary.com/answers/zh/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c02/) - Ch.3 The Founder as a Single Server: Queueing and the Cost of Switching — Chapter 3 treats the founder as a single server: raising utilisation from eighty to ninety-five per cent multiplies waiting by 4.75, and a single three-hour window lengthens the longest creative block from 3.34 to 5.68 hours. (https://5dlibrary.com/answers/en/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c03/) 第3章 创始人是一台单服务台:排队论与切换成本 — 第三章把创始人当作单服务台,利用率从八成升到九成五时等待放大 4.75 倍;单一三小时窗口把最长整块创作时间从 3.34 小时提到 5.68 小时。 (https://5dlibrary.com/answers/zh/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c03/) - Ch.4 Bounding Decision Fatigue and Filling the Three Hours: Heap-Based Top-k Selection — Chapter 4 bounds the evidence on decision fatigue and selects window items with a heap by value per minute, covering 0.682 of value under a heavy tail against 0.220 for first come, first served. (https://5dlibrary.com/answers/en/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c04/) 第4章 决策疲劳立界与三小时选题:堆排序取前 k 项 — 第四章说明决策疲劳证据须立界,并用堆排序按每分钟价值选题,重尾情形下覆盖价值 0.682,先到先办只有 0.220。 (https://5dlibrary.com/answers/zh/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c04/) - Ch.5 Delegation Rules: Checker Pass Lines, the Cost of Wrong Passes and Watching the Reversal Rate — Section A assumes an exception rate of 0.05. With 50 items a day at 5 minutes each, coverage of 0.648 suffices; at 8 minutes it must be 0.800; at 12 minutes, 0.884. With 100 items a day the three columns require 0.851, 0.926 and 0.968. With 200 items a day, 5 minutes each requires 0.952 and 8 minutes requires 0.989, while 12 minutes has no solution. At 400 items a day or more, none of the three columns has a solution. Section B fixes 200 items a day at 8 minutes and shows the role of the exception rate: with no exceptions coverage must be 0.940; with an exception rate of 0.05 it must be 0.989; at 0.10 or more no level of coverage fits. Section C turns the cost of exceptions into minutes: under this setting, each additional percentage point of exceptions costs the owner 15.2 minutes a day. (https://5dlibrary.com/answers/en/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c05/) 第5章 委托规则:检查器放行线、误放行代价与推翻率监控 — A 段把放行线写成风险额的分位数。放行线为零,也就是全部交本人,本人每天要花 960 分钟,总代价 48.0。放行线在中位数,也就是风险额 1.00 时,本人花 480 分钟,误放行损失 4.27,总代价 28.3。放行线在第七十百分位,风险额 1.89,本人 288 分钟,总代价 23.7。放行线在第八十百分位,风险额 2.76,本人 192 分钟,误放行损失 13.37,总代价 23.0,是全表最低。再往上,第八十五百分位总代价 23.4,第九十百分位 24.6,第九十五百分位 27.3,第九十九百分位 32.8;全部放行时,误放行损失升到 37.05,总代价 37.0。B 段确认最低点在第八十百分位。C 段补充一个容易忽略的事实:放行线在第九十百分位时,自动通道里的事项件数占九成,风险额却只占总额的 0.534。 (https://5dlibrary.com/answers/zh/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c05/) - Ch.6 The Long Rhythm: Fitness-Fatigue Impulse Response and the Limits of Self-Tracking — The numbers in the table are sobering. With 14 days of records, a slide of 0.5 points is detected with probability only 0.099, a slide of 1.0 with 0.175 and a slide of 1.5 with 0.302. With 28 days the three figures are 0.116, 0.243 and 0.422. With 42 days they are 0.150, 0.343 and 0.534. With 56 days, 0.192, 0.360 and 0.634. With 84 days, 0.181, 0.471 and 0.792. The small dip in the column for a slide of 0.5 between 56 and 84 days is simulation noise, not a pattern. (https://5dlibrary.com/answers/en/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c06/) 第6章 长期节律:体能—疲劳脉冲响应与自我追踪的边界 — 表里的数字很冷静。记 14 天时,下滑 0.5 分只有 0.099 的概率被看出,下滑 1.0 分是 0.175,下滑 1.5 分是 0.302。记 28 天,三档分别是 0.116、0.243、0.422。记 42 天是 0.150、0.343、0.534。记 56 天是 0.192、0.360、0.634。记 84 天是 0.181、0.471、0.792。下滑 0.5 分那一列在 56 天与 84 天之间略有回落,是模拟误差,不是规律。 (https://5dlibrary.com/answers/zh/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c06/) - Ch.7 Classical Sources East and West: Tension and Release, and the Shortness of Life — All tension, seven high days out of seven, gives a mean performance of 67.5 with a minimum that falls to -83.8 and a total load of 1,820. Six on and one off gives a mean of 75.4, a minimum of -51.7 and a total load of 1,612. Five on and two off gives a mean of 79.3, the highest in the table, with a minimum of -37.0 and a total load of 1,404. Four on and three off gives 79.0, a minimum of -26.9 and a load of 1,196. Two on and two off gives 77.0, a minimum of -20.6 and a load of 1,100. One on and one off gives 73.3, a minimum of -22.7 and a load of 1,092. All release gives a mean of only 37.5, a minimum of 6.9 and a load of 364. The last column shows performance per unit of load: 0.037 for all tension, 0.056 for five on and two off, 0.070 for two on and two off, and 0.103 for all release. (https://5dlibrary.com/answers/en/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c07/) 第7章 东西方原典考辨:一张一弛与生命之短 — 只张不弛,即七天全高负荷,平均表现 67.5,最低跌到 -83.8,总负荷 1,820。张六弛一,平均 75.4,最低 -51.7,总负荷 1,612。张五弛二,平均 79.3,是全表最高,最低 -37.0,总负荷 1,404。张四弛三,平均 79.0,最低 -26.9,总负荷 1,196。张二弛二,平均 77.0,最低 -20.6,总负荷 1,100。张一弛一,平均 73.3,最低 -22.7,总负荷 1,092。只弛不张,平均只有 37.5,最低 6.9,总负荷 364。最后一列是每单位负荷换来的表现:只张不弛 0.037,张五弛二 0.056,张二弛二 0.070,只弛不张 0.103。 (https://5dlibrary.com/answers/zh/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c07/) - Ch.8 Closing the Loop: The Cruise Ledger, Field Tools and Checker Specifications — Chapter 8 gives the cruise ledger, the distilled Sprint-to-Cruise Admission and Three-Hour Decision Queue, checker specifications and a preregistered single-case ABAB test whose power for a medium effect with fourteen-day phases is 0.758. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c08/) 第8章 实证闭环:巡航台账、实战工具与检查器规格 — 第八章给出巡航台账、蒸馏小工具「冲刺转巡航准入与每日三小时拍板队列器」、检查器规格与单案例 ABAB 预注册检验,每阶段 14 天时中等效应的检验功效为 0.758。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/sprint-to-cruise-transition-three-hour-decision-cap-creator-vitality/c08/) ## 5DL-B-405 · Autonomous Nightly Funnel Micro-Surgery, Bandit Traffic Allocation and Zero-Touch Conversion Self-Improvement 中文书名:夜间自治漏斗微创调优、多臂赌博机流量自动倾斜与零人工转化率自举母典 DOI: pending Cite: Fan, Y. (2026). Autonomous Nightly Funnel Micro-Surgery, Bandit Traffic Allocation and Zero-Touch Conversion Self-Improvement (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its 23 programs; it is general business analysis, not business, legal or financial advice, and it promises no rise in conversion. Abstract: The popular belief holds that a website no longer needs anyone watching it: in the small hours a program finds the leakiest step of the funnel and generates candidate changes, and by day Thompson sampling tilts traffic towards the best version and deletes the weak ones, so that conversion climbs by itself. This book examines the four hurdles such a machine must clear and makes four corrections to its blueprint. First, a bandit is designed to pay little regret, not to produce trustworthy effect sizes: the naive means of adaptive data come out 7.0% too low and must be corrected by inverse probability weighting with allocation probabilities recorded batch by batch; a naive rule that looks every night and deletes whatever is behind wrongly deletes 0.241 of equally good versions, and a three-part rule of minimum exposure, posterior win probability and a time-uniform confidence sequence cuts this to 0.001. Second, conversion is not stationary, and known cycles must be handled by grouping; posteriors split by type of day raise the share of traffic reaching the better version of the day to 0.889. Third, the goal is net profit, not conversion: Thompson sampling rewarded on the pay rate sends 0.623 of the traffic to a discounted version, against 0.064 when the reward is net contribution per visitor. Fourth, "zero touch" is rewritten as an upper bound of 99 minutes of human review a week and a list of actions that always need a human signature. Chapter 1 checks a six-step funnel and writes the compliance gate as seven executable predicates; Chapter 2 compares the even split, explore-then-commit, UCB1 and Thompson sampling; Chapter 3 covers the reward function and the break-even bar for discounts; Chapter 4 weekday effects, abrupt changes and the cost of forgetting; Chapter 5 elimination rules; Chapter 6 the negative bias of adaptive data, sample ratio mismatch (a logging fault alone can manufacture a lift of 3.1%) and guardrail rollback. Chapter 7 examines the Great Commentary on the Book of Changes on change at the limit, the Sunzi on weak and strong points, and the random allocation of Peirce and Jastrow in 1885, and gives the distillation spec for the four kits. Chapter 8 gives the nightly funnel ledger, the calculator, checkers grouped by the four kits and a pre-registered test: under the base prior the median cumulative gain in net contribution over half a year is about +0.082, with a probability of 0.093 of gaining nothing. Its parameters are illustrative; all calculations are generated by its 23 programs; it is general business analysis, not business, legal or financial advice, and it promises no rise in conversion. 摘要:流行说法认为,网站不再需要人工盯盘:凌晨自动找出漏斗流失最大的环节、生成候选改动,白天用汤普森采样把流量倾斜给最好的版本并自动淘汰差的版本,转化率就会自己往上爬。本书逐一检查这台机器要越过的四道坎,并据文献做了四处蓝图修正。第一,赌博机为少付后悔而设计,不为得出可信效应量而设计:自适应数据的朴素均值偏低 7.0%,须用逐批记录的分配概率做逆概率加权;每晚看一眼、差了就删的朴素规则会误删 0.241 的同等好版本,改用最小曝光、后验胜率与时间均匀置信序列三重规则后降到 0.001。第二,转化率不平稳,已知周期要分组,按日型分组的后验把流向当天较好版本的份额提到 0.889。第三,目标是净利而非转化率:以付款率为奖励的汤普森采样把 0.623 的流量送给打折版本,以每访客净贡献为奖励时打折版本只得 0.064。第四,「零人工」改写为每周人工复核上界 99 分钟与永远须人工签字的例外清单。 第一章做六步漏斗体检并把合规门禁写成七条可执行谓词;第二章比较平均分、先试后定、UCB1 与汤普森采样;第三章讲奖励函数与降价保本门槛;第四章讲星期效应、突变与遗忘的代价;第五章讲淘汰规则;第六章讲自适应数据负偏、样本比例失衡(日志故障可凭空造出 3.1% 的提升)与护栏回滚;第七章考辨《周易·系辞下》「穷则变,变则通,通则久」、《孙子·虚实》与皮尔士和贾斯特罗 1885 年的随机分配,并给出四件套蒸馏规格;第八章给出夜间漏斗台账、试算工具、按四件套分组的检查器与预注册检验:基准先验下半年累计净贡献提升中位数约 +0.082,仍有 0.093 的概率一无所获。书中参数为示意;全部计算由书中 23 段程序生成;本书是非个人化的经营分析,不构成经营、法律或财务建议,不宣称任何转化率提升的保证。 Chapter answers: - Ch.1 The Funnel Check-up: Six-Step Beta Posteriors, Nightly Change Priorities and a Compliant Candidate Pool — Night B uses one night's small counts: 1,650 people enter, followed by 1,010, 445, 190, 112 and 77. Now the sample step has P(top) of only 0.714 and the first screen 0.286. The sample step's priority interval runs from 3,511 to 5,140 and the first screen's from 3,270 to 4,968, with medians of 4,305 and 4,067; they really cannot be told apart. The rule returns HOLD: change nothing tonight, add another night of data and recompute tomorrow. (https://5dlibrary.com/answers/en/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c01/) 第1章 漏斗体检:六步 Beta 后验、夜间改动优先级与合规候选池 — 单晚的 B 夜换了一组小样本:进来 1,650 人,往后依次是 1,010、445、190、112、77 人。这时试读的 P(top) 只有 0.714,首屏占 0.286。试读的优先级区间从 3,511 到 5,140,首屏从 3,270 到 4,968,中位数 4,305 对 4,067,确实分不开。规则给出 HOLD:今晚不改,明晚再合并一天的数据重算。 (https://5dlibrary.com/answers/zh/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c01/) - Ch.2 Thompson Sampling and Regret: from Two Small Samples in 1933 to the Even Split — Chapter 2 compares the even split, explore-then-commit, UCB1 and Thompson sampling; (https://5dlibrary.com/answers/en/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c02/) 第2章 汤普森采样与后悔:从 1933 年的两份小样本到对照均分 — 第二章比较平均分、先试后定、UCB1 与汤普森采样; (https://5dlibrary.com/answers/zh/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c02/) - Ch.3 The Reward Function: Net Contribution per Visitor, Not the Conversion Rate — With the pay rate as reward, the realised pay rate is 0.0550 and net contribution per visitor 1.7140; the discount version takes 0.623 of the traffic and the trust version only 0.124. With net contribution as reward the pay rate is 0.0505, a little lower, but net contribution per visitor is 1.9453, well above; the trust version receives 0.615 of the traffic and the discount version only 0.064. (https://5dlibrary.com/answers/en/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c03/) 第3章 奖励函数:每访客净贡献,而不是转化率 — 以付款率为奖励时,实现的付款率是 0.0550,每访客净贡献是 1.7140;流量份额里打折版拿走 0.623,信任版只有 0.124。以净贡献为奖励时,付款率是 0.0505,略低;每访客净贡献却是 1.9453,高出一截;信任版拿到 0.615 的流量,打折版只剩 0.064。 (https://5dlibrary.com/answers/zh/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c03/) - Ch.4 Non-stationary Traffic: Weekday Effects, Discounted Memory and Sliding Windows — A per-day discount of 0.9 gives B 0.345 in the first week, 0.707 in the second, and settles at 0.843 and 0.847 in the third and fourth, for a total regret of 105.5 orders. A discount of 0.7 reacts faster in the first week, at 0.544, but then hovers around 0.754, 0.763 and 0.730, with a total regret of 129.7 orders: forgetting too fast means paying for exploration all the time. The seven-day window does best: 0.568 in the first week, 0.901 in the third, and a total regret of 87.0 orders. (https://5dlibrary.com/answers/en/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c04/) 第4章 流量不平稳:星期效应、折扣遗忘与滑动窗口 — 每日折扣 0.9 在第一周给 B 0.345,第二周升到 0.707,第三、四周稳定在 0.843 与 0.847,总后悔 105.5 单。折扣 0.7 第一周反应更快,给到 0.544,但之后只在 0.754、0.763、0.730 附近徘徊,总后悔 129.7 单。忘得太快,平时也要多付学费。七天窗口表现最好:第一周 0.568,第三周 0.901,总后悔 87.0 单。 (https://5dlibrary.com/answers/zh/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c04/) - Ch.5 Elimination Rules: Minimum Exposure, Posterior Win Probability and Time-uniform Confidence Sequences — Take the four lines at test level 0.05 first. The false-deletion rate lies between 0.003 and 0.006, and the deletion rate of the good version is always 0.000. The share of weak versions deleted by night 28 lies between 0.139 and 0.143, rising to between 0.532 and 0.577 by night 56. Raising the minimum exposure from 0 to 8,000 makes almost no difference. (https://5dlibrary.com/answers/en/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c05/) 第5章 淘汰规则:最小曝光、后验胜率与时间均匀置信序列 — 先看检验水平 0.05 这四行。误删率在 0.003 到 0.006 之间,好版本遭删率始终是 0.000。差版本第 28 晚遭删比例在 0.139 到 0.143 之间,第 56 晚升到 0.532 至 0.577。最小曝光量从 0 加到 8,000,几乎不影响结果。 (https://5dlibrary.com/answers/zh/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c05/) - Ch.6 Trustworthy Data: Negative Bias in Adaptive Data, Sample Ratio Mismatch and Guardrail Rollback — Chapter 6 the negative bias of adaptive data, sample ratio mismatch (a logging fault alone can manufacture a lift of 3.1%) and guardrail rollback. (https://5dlibrary.com/answers/en/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c06/) 第6章 数据可信度:自适应数据的负偏、样本比例失衡与护栏回滚 — 第六章讲自适应数据负偏、样本比例失衡(日志故障可凭空造出 3.1% 的提升)与护栏回滚; (https://5dlibrary.com/answers/zh/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c06/) - Ch.7 Classical Sources East and West: "At the Limit, Change" in the Great Commentary, the Sunzi on Weak and Strong Points, and Random Allocation in Peirce and Jastrow — Chapter 7 examines the Great Commentary on the Book of Changes on change at the limit, the Sunzi on weak and strong points, and the random allocation of Peirce and Jastrow in 1885, and gives the distillation spec for the four kits. (https://5dlibrary.com/answers/en/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c07/) 第7章 东西方原典考辨:《系辞下》「穷则变」、《孙子·虚实》与皮尔士和贾斯特罗的随机分配 — 第七章考辨《周易·系辞下》「穷则变,变则通,通则久」、《孙子·虚实》与皮尔士和贾斯特罗 1885 年的随机分配,并给出四件套蒸馏规格; (https://5dlibrary.com/answers/zh/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c07/) - Ch.8 The Nightly Funnel Ledger: Field Tools and Checker Specifications, a Pre-registered Test and the Conditions of Refutation — Chapter 8 gives the nightly funnel ledger, the calculator, checkers grouped by the four kits and a pre-registered test: under the base prior the median cumulative gain in net contribution over half a year is about +0.082, with a probability of 0.093 of gaining nothing. Its parameters are illustrative; all calculations are generated by its 23 programs; it is general business analysis, not business, legal or financial advice, and it promises no rise in conversion. (https://5dlibrary.com/answers/en/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c08/) 第8章 夜间漏斗台账:实战工具与检查器规格、预注册检验与推翻条件 — 第八章给出夜间漏斗台账、试算工具、按四件套分组的检查器与预注册检验:基准先验下半年累计净贡献提升中位数约 +0.082,仍有 0.093 的概率一无所获。书中参数为示意;全部计算由书中 23 段程序生成;本书是非个人化的经营分析,不构成经营、法律或财务建议,不宣称任何转化率提升的保证。 (https://5dlibrary.com/answers/zh/autonomous-nightly-funnel-micro-surgery-thompson-bandit-traffic-zero-touch-conversion/c08/) ## 5DL-B-406 · Axiom-Driven Business Wargaming, Multi-Agent Adversarial Stress-Testing and Pre-Launch Zero-Tuition Sandboxes 中文书名:四百卷公理驱动商业兵棋推演、五方敌对代理压力测试与零学费试错沙盘母典 DOI: pending Cite: Fan, Y. (2026). Axiom-Driven Business Wargaming, Multi-Agent Adversarial Stress-Testing and Pre-Launch Zero-Tuition Sandboxes (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate evidence, runnable Python calculations (22 programs), classical sources and hypotheses. Abstract: A popular belief holds that a plan is safe once it has passed a test against each kind of risk in turn. This book refutes it with a round-based sandbox. Acting alone, five adversarial agents (a price-sensitive buyer, an arbitrage speculator, a compliance auditor, an imitating competitor and a liquidity shock) raise the illustrative plan's probability of ruin to at most 0.0010; acting together, they ruin it with probability 0.2165, with a 95 % interval from 0.1990 to 0.2351. After checking the literature the book corrects its blueprint: "exhaustive" testing is impossible and is rewritten as Latin hypercube coverage, global sensitivity and a list of unvisited regions; "zero tuition" is rewritten as moving part of the tuition onto paper, checked against a backfilled sample; the loss-aversion coefficient of 2.25 comes from Tversky and Kahneman's paper of 1992, not from the paper of 1979; and "driven by the axioms of four hundred volumes" becomes an axiom table of 19 conclusions drawn from 17 published volumes between B-001 and B-332, each with its source volume and range of validity. Parameters are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not investment, legal or financial advice, and promises no income; the arbitrage agent is used only to find and close loopholes in one's own rules. Chapter 1 builds the axiom table and the engine and reproduces zero-intelligence traders: a single budget rule lifts efficiency from 0.8929 to 0.9730. Chapter 2 covers reference points and loss aversion: the demand curve is 1.339 times as steep above the reference price as below it, and one promotion costs 522.1 points of sixteen-round margin. Chapter 3 lets the arbitrage agent read only our own rule book; closing every loophole lowers joint ruin to 0.0670. Chapter 4 shows that under capacity commitment the former monopolist keeps 0.4444 of its profit, and that resolving violations lowers ruin to 0.0920. Chapter 5 covers gambler's ruin, presale runs and frozen channels; doubling opening cash lowers ruin to 0.0080. Chapter 6 decomposes the joint attack: the five marginal damages sum to 0.5160, a forty-point Latin hypercube covers 0.8653 of two-dimensional cells, and opening cash has a first-order sensitivity index of 0.5437. Chapter 7 examines the temple calculation of the Sunzi and Clausewitz's friction: under correlated error the temple calculation has a ceiling of 0.8800, and a forty-step plan finishes on time with probability only 0.0842. Chapter 8 gives the field tools and checker specifications, the planning ledger and a pre-registered test: after backfill the unhardened plan's ruin is rewritten as 0.6525 and the hardened plan's as 0.0040, and testing the core claim needs 62 to 199 plans per group. Marked boxes separate evidence, runnable Python calculations (22 programs), classical sources and hypotheses. 摘要:立项会上常听到的流行说法是:把新方案分别对着每类风险各压一遍,全部过关就安全。本书用一台回合制沙盘证伪它:五类敌对代理,即价格敏感买家、套利投机者、合规审查官、模仿竞争者与流动性冲击,单独出手时最坏只让示意方案的破产概率升到 0.0010;五类同时出手,破产概率是 0.2165,九成五区间从 0.1990 到 0.2351。核对文献之后,本书作了蓝图修正:「穷举」不可能,改写成拉丁超立方覆盖率、全局敏感度与未覆盖区域清单;「零学费」改写成「把一部分学费提前到纸面」,并用回填小样本推翻或保留结论;损失厌恶系数 2.25 出自特沃斯基与卡尼曼 1992 年论文而非 1979 年论文;「四百卷公理驱动」写成一张公理表,从 B-001 至 B-332 已出版卷抽取 19 条结论、来自 17 卷,每条注明来源卷号与适用范围。本卷在出版局编号为 5DL-B-406(指挥官原令编号 5DL-B-398)。书中参数为示意,单位是无币种的点;全部计算由书中程序生成;本书是非个人化经营分析,不构成投资、法律或财务建议,也不构成任何收益承诺;套利代理只用于找出并堵上自家规则漏洞。 第一章建立公理表与回合制引擎,并复现零智能交易者:预算约束一条规则把效率从 0.8929 抬到 0.9730。第二章写参考点与损失厌恶,需求曲线在参考价处上下斜率之比为 1.339,一次八折促销让十六回合毛利少 522.1 点。第三章让套利代理只读自家规则书,堵上全部漏洞后联合破产概率降到 0.0670。第四章说明产能先定时原独家经营者保留 0.4444 的利润,整改违规使破产概率降到 0.0920。第五章写赌徒破产、预售挤兑与渠道冻结,期初现金加倍使破产概率降到 0.0080。 第六章分解联合进攻,五个边际破坏之和为 0.5160;四十点拉丁超立方的两维覆盖率为 0.8653,期初现金的一阶敏感度指数为 0.5437。第七章考辨《孙子·计篇》的庙算与克劳塞维茨的摩擦:相关错误下庙算的上限是 0.8800,四十步计划按时完成的概率只有 0.0842。第八章给出实战工具与检查器规格、沙盘立项台账与预注册检验:回填之后未加固方案的破产概率改写为 0.6525,加固方案为 0.0040;核心命题的检验每组需要 62 到 199 个方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Axiom Table and the Round-Based Engine: Nineteen Conclusions, Zero-Intelligence Agents and a Sandbox That Can Go Bankrupt — Chapter 1 builds the axiom table and the engine and reproduces zero-intelligence traders: a single budget rule lifts efficiency from 0.8929 to 0.9730. (https://5dlibrary.com/answers/en/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c01/) 第1章 公理表与回合制引擎:十九条结论、零智能代理与一台会破产的沙盘 — 第一章建立公理表与回合制引擎,并复现零智能交易者:预算约束一条规则把效率从 0.8929 抬到 0.9730。 (https://5dlibrary.com/answers/zh/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c01/) - Ch.2 The Price-Sensitive Buyer: Reference Points, Loss Aversion and the Hangover after a Promotion — Chapter 2 covers reference points and loss aversion: the demand curve is 1.339 times as steep above the reference price as below it, and one promotion costs 522.1 points of sixteen-round margin. (https://5dlibrary.com/answers/en/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c02/) 第2章 价格敏感买家:参考点、损失厌恶与促销之后的宿醉 — 第二章写参考点与损失厌恶,需求曲线在参考价处上下斜率之比为 1.339,一次八折促销让十六回合毛利少 522.1 点。 (https://5dlibrary.com/answers/zh/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c02/) - Ch.3 The Arbitrage Speculator: a Red Team That Reads Only Our Own Rule Book, a Breach List and the Order of Repairs — Chapter 3 lets the arbitrage agent read only our own rule book; closing every loophole lowers joint ruin to 0.0670. (https://5dlibrary.com/answers/en/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c03/) 第3章 套利投机者:只读自家规则书的红队、破绽清单与堵漏顺序 — 第三章让套利代理只读自家规则书,堵上全部漏洞后联合破产概率降到 0.0670。 (https://5dlibrary.com/answers/zh/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c03/) - Ch.4 The Imitating Competitor and the Compliance Auditor: Capacity Commitment, Entry Delay and Random Checks against a Clause List — Chapter 4 shows that under capacity commitment the former monopolist keeps 0.4444 of its profit, and that resolving violations lowers ruin to 0.0920. (https://5dlibrary.com/answers/en/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c04/) 第4章 模仿竞争者与合规审查官:产能承诺、进入延迟与条款清单的随机抽查 — 第四章说明产能先定时原独家经营者保留 0.4444 的利润,整改违规使破产概率降到 0.0920。 (https://5dlibrary.com/answers/zh/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c04/) - Ch.5 The Liquidity Shock: Gambler's Ruin, Refund Runs and Frozen Channels — Chapter 5 covers gambler's ruin, presale runs and frozen channels; doubling opening cash lowers ruin to 0.0080. (https://5dlibrary.com/answers/en/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c05/) 第5章 流动性冲击:赌徒破产、退款挤兑与渠道冻结 — 第五章写赌徒破产、预售挤兑与渠道冻结,期初现金加倍使破产概率降到 0.0080。 (https://5dlibrary.com/answers/zh/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c05/) - Ch.6 The Joint Attack, Coverage and Global Sensitivity: Rewriting "Exhaustive" as Three Numbers That Can Be Checked — Chapter 6 decomposes the joint attack: the five marginal damages sum to 0.5160, a forty-point Latin hypercube covers 0.8653 of two-dimensional cells, and opening cash has a first-order sensitivity index of 0.5437. (https://5dlibrary.com/answers/en/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c06/) 第6章 联合进攻、覆盖率与全局敏感度:把「穷举」改写成可以核对的三个数 — 第六章分解联合进攻,五个边际破坏之和为 0.5160;四十点拉丁超立方的两维覆盖率为 0.8653,期初现金的一阶敏感度指数为 0.5437。 (https://5dlibrary.com/answers/zh/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c06/) - Ch.7 Eastern and Western Classics: the Temple Calculation of the Sunzi and Clausewitz's Friction — Chapter 7 examines the temple calculation of the Sunzi and Clausewitz's friction: under correlated error the temple calculation has a ceiling of 0.8800, and a forty-step plan finishes on time with probability only 0.0842. (https://5dlibrary.com/answers/en/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c07/) 第7章 东西方原典考辨:《孙子·计篇》的庙算与克劳塞维茨的摩擦 — 第七章考辨《孙子·计篇》的庙算与克劳塞维茨的摩擦:相关错误下庙算的上限是 0.8800,四十步计划按时完成的概率只有 0.0842。 (https://5dlibrary.com/answers/zh/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c07/) - Ch.8 Field Tools and Checker Specifications: the Sandbox Planning Ledger, Backfill Calibration and a Pre-Registered Test — Chapter 8 gives the field tools and checker specifications, the planning ledger and a pre-registered test: after backfill the unhardened plan's ruin is rewritten as 0.6525 and the hardened plan's as 0.0040, and testing the core claim needs 62 to 199 plans per group. Marked boxes separate evidence, runnable Python calculations (22 programs), classical sources and hypotheses. (https://5dlibrary.com/answers/en/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c08/) 第8章 实战工具与检查器规格:沙盘立项台账、回填校准与预注册检验 — 第八章给出实战工具与检查器规格、沙盘立项台账与预注册检验:回填之后未加固方案的破产概率改写为 0.6525,加固方案为 0.0040;核心命题的检验每组需要 62 到 199 个方案。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/axiom-driven-business-wargaming-five-adversarial-agents-zero-tuition-sandbox/c08/) ## 5DL-B-407 · Commander Intent Compilation, Automated DAG Task Decomposition and Zero-Courier Multi-Agent Dispatch 中文书名:总司令自然语言意图编译、有向无环任务自动拆解与零传令兵自治调度母典 DOI: pending Cite: Fan, Y. (2026). Commander Intent Compilation, Automated DAG Task Decomposition and Zero-Courier Multi-Agent Dispatch (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative and all calculations are generated by its programs. It names no server, account, key or deployment command, and it is general business analysis, not personalised advice. Abstract: The popular belief holds that once a high-level goal is handed to a clever enough orchestrator, it decomposes the goal into a task graph by itself, sub-agents claim the tasks and retry their failures by themselves, the commander looks only at results, and nobody need carry messages between windows again. This book falsifies the reading of that belief as "zero human". In an illustrative simulation of batches of four volumes, an intent compiler plus dispatcher needs on average 9.951 human interventions per batch, 0.830 fewer than the 58.4 hand-carried messages of the old relay; but irreversible and outward actions must be signed by a person, and signatures alone are 0.137 of the old count, a floor that never reaches zero. Its blueprint is corrected: zero courier is not zero human; every retry needs an idempotency key, backoff with jitter, a retry budget and a circuit breaker; more resources do not always mean more speed; and the self-correction of language-model agents is limited evidence. Its parameters are illustrative and all calculations are generated by its programs. It names no server, account, key or deployment command, and it is general business analysis, not personalised advice. Chapter 1 reads intents with a five-slot rule compiler that returns any order with an empty slot instead of guessing, and decomposes one order into a fifteen-node task graph with signature nodes. Chapter 2 detects cycles with Kahn's sort and computes float by the critical path method; a deterministic plan of 17.4 hours is met with probability only 0.0213 under Monte Carlo. Chapter 3 covers contract-net claiming and Graham's list scheduling and reproduces a timing anomaly: three executors finish in 12, four need 15. Chapter 4 shows a global locking order driving simulated deadlock rates of 0.1795 and 0.8655 to zero, and recovers orphaned locks with leases and fencing tokens. Chapter 5 shows blind retries causing 532 duplicates in ten thousand operations, and retries every second sending 147,693 calls during a thirty-second outage; none of the four guard rails can be dropped. Chapter 6 sorts exceptions into four classes and checks tasks in flight with Little's law. Chapter 7 examines "governing many is like governing few; it is a matter of division into units" from the chapter on momentum in Sunzi's Art of War, and Smith's pin factory in The Wealth of Nations, and gives the distillation spec for the four kits. Chapter 8 gives the dispatch ledger, the Intent Compilation and Task-DAG Dispatch Checker, a checker specification table grouped by the four kits, and a pre-registered test: non-inferiority of the error rate needs 1,188 agent tasks per arm, and bounding the violation rate below 0.01 needs 299 consecutive clean batches. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:流行说法认为,只要把高层目标交给一台足够聪明的编排器,它就会自动拆成任务图,子智能体自动领单、失败自动重试,统帅只看结果,从此不再需要人在窗口之间传话。本书证伪其中「零人工」这一层含义。在四卷一批的示意模拟里,改用意图编译器加调度器之后,每批人工介入均值为 9.951 次,比人工传话的 58.4 次下降 0.830;但不可逆与对外动作必须由人签字,仅签字一项就占旧口径的 0.137,是永远降不到零的地板。核对原文之后,本书做了蓝图修正:零传令兵不等于零人工;重试必须同时具备幂等键、退避加抖动、重试预算与熔断;加资源未必提速;语言模型智能体的自我纠错只是有限证据。书中参数为示意,全部计算由书中程序生成;全书只写调度原理与公开流程形态,不含服务器、端口、账户、密钥与任何部署命令;本书是非个人化经营分析,不构成经营、法律或运维建议。 第一章用五槽规则编译器读意图,缺槽即退回追问,绝不猜测,并把一句命令拆成带签字节点的十五节点任务图。第二章用卡恩排序查环、用关键路径法算浮动,确定性计划 17.4 小时在蒙特卡洛中按时完成的概率只有 0.0213。第三章讲合同网领单与格雷厄姆列表调度,复现时序异常:三个执行者用时 12,加到四个反而要 15。第四章说明全序加锁把模拟中的死锁率从 0.1795 与 0.8655 压到零,并用租约与防护令牌回收孤儿锁。第五章说明盲目重试在一万次操作里造成 532 次重复,每秒重试在一场三十秒故障里发出 147,693 次调用,四道护栏缺一不可。 第六章把例外分为四类,并用利特尔定律核对在途任务。第七章考辨《孙子兵法·势篇》「凡治众如治寡,分数是也」与斯密《国富论》的制针工场,并给出四件套蒸馏规格。第八章给出调度台账、意图编译与任务有向图调度检查器、按四件套分组的检查器规格表与预注册检验:差错率的非劣效检验每组需要 1,188 个智能体任务,而把违规率上界压到 0.01 以下需要连续 299 个干净批次。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Intent Compilation: a Five-Slot Rule Compiler, Asking for Missing Slots and Hierarchical Decomposition — But as soon as a tenth of the empty slots govern irreversible steps, the picture reverses. In the third row, with the same wrong-guess rate of 0.3, the mean cost of guessing rises to 7.871 and the worst single case is 400.0; the break-even point drops to 0.0265, and few habits are that accurate. The fifth row is more persuasive still. Even with a wrong-guess rate as low as 0.05, the mean cost of 1.319 is twice the cost of asking, and the worst case still exceeds two hundred. (https://5dlibrary.com/answers/en/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c01/) 第1章 意图编译:五槽规则编译器、缺槽追问与分层拆解 — 但只要一成的空槽管着不可逆动作,局面立刻翻转。第三行,同样的猜错率 0.3,猜测的平均成本升到 7.871,最坏一单是 400.0;盈亏点跌到 0.0265,几乎没有哪种惯例能猜得这么准。第五行更有说服力:即使猜错率低到 0.05,平均成本 1.319 仍然是追问的两倍,最坏一单仍然超过两百。 (https://5dlibrary.com/answers/zh/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c01/) - Ch.2 The Directed Acyclic Task Graph: Kahn's Topological Sort, the Critical Path and Monte Carlo Durations — Chapter 2 detects cycles with Kahn's sort and computes float by the critical path method; a deterministic plan of 17.4 hours is met with probability only 0.0213 under Monte Carlo. (https://5dlibrary.com/answers/en/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c02/) 第2章 任务有向无环图:卡恩拓扑排序、关键路径与蒙特卡洛工期 — 第二章用卡恩排序查环、用关键路径法算浮动,确定性计划 17.4 小时在蒙特卡洛中按时完成的概率只有 0.0213。 (https://5dlibrary.com/answers/zh/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c02/) - Ch.3 Multi-Agent Claiming: Contract-Net Bidding, Graham's List Scheduling and Timing Anomalies — Chapter 3 covers contract-net claiming and Graham's list scheduling and reproduces a timing anomaly: three executors finish in 12, four need 15. (https://5dlibrary.com/answers/en/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c03/) 第3章 多智能体领单:合同网出价、格雷厄姆列表调度与时序异常 — 第三章讲合同网领单与格雷厄姆列表调度,复现时序异常:三个执行者用时 12,加到四个反而要 15。 (https://5dlibrary.com/answers/zh/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c03/) - Ch.4 Resource Locks: a Global Order against Deadlock, Leases for Orphaned Locks and Fencing Tokens — Chapter 4 shows a global locking order driving simulated deadlock rates of 0.1795 and 0.8655 to zero, and recovers orphaned locks with leases and fencing tokens. (https://5dlibrary.com/answers/en/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c04/) 第4章 资源锁:全序加锁防死锁、租约回收孤儿锁与防护令牌 — 第四章说明全序加锁把模拟中的死锁率从 0.1795 与 0.8655 压到零,并用租约与防护令牌回收孤儿锁。 (https://5dlibrary.com/answers/zh/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c04/) - Ch.5 Automatic Recovery after Failure: Idempotency Keys, Exponential Backoff with Jitter, Retry Budgets and Circuit Breakers — Chapter 5 shows blind retries causing 532 duplicates in ten thousand operations, and retries every second sending 147,693 calls during a thirty-second outage; none of the four guard rails can be dropped. (https://5dlibrary.com/answers/en/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c05/) 第5章 失败后的自动恢复:幂等键、指数退避加抖动、重试预算与熔断 — 第五章说明盲目重试在一万次操作里造成 532 次重复,每秒重试在一场三十秒故障里发出 147,693 次调用,四道护栏缺一不可。 (https://5dlibrary.com/answers/zh/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c05/) - Ch.6 Exception Escalation: Human Interventions, Budget Overrun, Deadline Probability and Little's Law — Exceptions fall into four classes (irreversible actions, budget overruns, three consecutive acceptance failures and lock-wait timeouts), plus a catch-all for unclassified errors. Each report carries five slots of context, and its deadline comes only from a real time limit. In the simulation the mean is 9.951 interventions per batch, 0.830 below the relay, but signatures are a floor and never reach zero. Budget and deadline are two independent risks and must be set and reported separately, and Little's law lets the ledger find tasks that are lost or stuck. (https://5dlibrary.com/answers/en/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c06/) 第6章 例外呈报:人工介入次数、预算超支、截止概率与利特尔定律 — 例外分四类:不可逆动作、预算超支、验收三连败、锁等待超时,另有兜底的未分类错误;呈报自带五槽上下文,截止只来自真实时限。模拟中每批介入均值 9.951 次,比人工传话下降 0.830,但签字是地板,永远降不到零。预算与截止是两个独立的风险,要分开定、分开报;利特尔定律让台账能查出丢失与卡住的任务。 (https://5dlibrary.com/answers/zh/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c06/) - Ch.7 Eastern and Western Sources Examined: Division into Units in Sunzi's Chapter on Momentum and Smith's Pin Factory — A flat organisation reaches its peak output of 13.00 at twenty-five people. At fifty people only 1.00 remains, and beyond that output is negative: communication consumes all the time there is. Three hundred people in a flat organisation have 44,850 pairwise channels. A tree with units of five keeps output proportional to headcount throughout, 270.00 for three hundred people; the price is 76 unit heads and four hops from the top to the bottom. (https://5dlibrary.com/answers/en/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c07/) 第7章 东西方原典考辨:《孙子兵法·势篇》「分数」与斯密制针工场 — 平铺的组织,人数到二十五时产出达到顶点 13.00,到五十人时只剩 1.00,再往上就是负数:沟通把全部时间吃光了。三百人平铺,两两渠道有 44,850 条。按五人编队的树,产出始终与人数成正比,三百人时是 270.00;代价是要设 76 位组长,命令从顶层传到底层要经过四跳。 (https://5dlibrary.com/answers/zh/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c07/) - Ch.8 Field Tools and Checker Specifications: the Dispatch Ledger, the Intent Compilation and Task-DAG Dispatch Checker, and a Pre-Registered Test — Chapter 8 gives the dispatch ledger, the Intent Compilation and Task-DAG Dispatch Checker, a checker specification table grouped by the four kits, and a pre-registered test: non-inferiority of the error rate needs 1,188 agent tasks per arm, and bounding the violation rate below 0.01 needs 299 consecutive clean batches. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c08/) 第8章 实战工具与检查器规格:调度台账、意图编译与任务有向图调度检查器、预注册检验 — 第八章给出调度台账、意图编译与任务有向图调度检查器、按四件套分组的检查器规格表与预注册检验:差错率的非劣效检验每组需要 1,188 个智能体任务,而把违规率上界压到 0.01 以下需要连续 299 个干净批次。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/commander-intent-compilation-dag-task-decomposition-zero-courier-agent-dispatch/c08/) ## 5DL-B-408 · The Four-Hundred-Volume Holographic Knowledge Commonwealth: Self-Sustaining Cashflow-Compute Loops and Unbounded Creative Sovereignty 中文书名:四百卷全息知识共同体、现金流与算力自持闭环及无限创作自由态母典 DOI: pending Cite: Fan, Y. (2026). The Four-Hundred-Volume Holographic Knowledge Commonwealth: Self-Sustaining Cashflow-Compute Loops and Unbounded Creative Sovereignty (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: The library is an archipelago, not a hologram; a synthetic model shows that if half of new volumes each make one bridging citation, most new volumes join one giant component. Under an illustrative baseline g is 0.018, the critical scale is 55.6 and the feasible reinvestment band runs from 0.294 to 0.529; only 0.231 of the phase diagram meets both the growth and the personal-draw conditions. Abstract: A popular saying holds that four hundred volumes have fused into a holographic network whose business organs bring net cash flow that buys compute that breeds original theory, in a loop that feeds itself and sets its author free. This book breaks the slogan into parameters and programs and tests each link. It is 5DL-B-408 (originally ordered as 5DL-B-400). Counted honestly, there are 408 numbers, 382 to 383 delivered volumes and 26 still being written. A network built from the core keywords in the public master index has only 87 links among three hundred and thirty-two volumes, an isolated share of 0.699, a clustering coefficient of 0.639 and a zone modularity of 0.285. Keywords are reused at a rate of only 0.067; sparseness comes from the high rate of new words, not from missing cumulative advantage. The library is an archipelago, not a hologram; a synthetic model shows that if half of new volumes each make one bridging citation, most new volumes join one giant component. Only a share of 0.160 of volumes is wired to a paid route. On the loop, the library has 769,420 potential keyword pairs, of which a share of 0.00257 has been realised; growth is limited by processing capacity rather than ideas, and reinvestment makes the ideas added over sixty months 1.77 times those under a fixed budget. Self-sustenance is written as an inequality: the reinvestment rate times the return after platform fees and refunds must exceed depreciation plus maintenance leakage, and capital must exceed a critical scale. Under an illustrative baseline g is 0.018, the critical scale is 55.6 and the feasible reinvestment band runs from 0.294 to 0.529; only 0.231 of the phase diagram meets both the growth and the personal-draw conditions. Monte Carlo gives a thirty-six-month ruin rate of 0.0002 and a tenth-percentile runway of 12.2 months; ruin is U-shaped in the reinvestment rate, reaching 0.2380 at 0.80. Scarcity research is reported as a small, disputed effect. "Freedom" is written only as conditions on resources and time; automation is eaten by the maintenance of new organs, and at a retirement rate of 0.01 an automation share of 0.880 is needed to reach an illustrative twenty free hours a week. Chapter 7 examines "depending on nothing" in the Zhuangzi, "daily renewal" in the Great Learning, Keynes's 1930 essay on leisure and Book X of Aristotle's Nicomachean Ethics, and ends with a distillation spec for the four kits. Chapter 8 gives the self-sustenance ledger, the distilled Cash-Compute Self-Sustenance Equation and Knowledge-Network Growth Calculator, checker specifications grouped by kit and a pre-registered plan of refutation. Parameters are illustrative and cash is in abstract units; no real revenue, transaction, account or server appears. All calculations come from 21 standard-library programs in the book. It is general business analysis, not business, legal, tax or investment advice, and promises no income and no psychological or bodily effect. 摘要:书库编号越过四百之后,流行说法是:四百卷已连成全息网络,网络长出商业器官,器官带来经常性净现金流,现金流买来算力,算力催生原创理论,回路自己养活自己,创作者从此进入无限自由的创作状态。本书把这句口号拆成参数与程序逐环检验。本卷在出版局编号为 5DL-B-408(指挥官原令编号 5DL-B-400)。如实计数:编号 408,交付 382 到 383,在写 26。用公开总索引登记的核心关键词构造卷间网络,三百三十二卷之间只有 87 条边,孤立卷占 0.699,聚类系数 0.639,学科战区模块度 0.285;关键词复用率只有 0.067,稀疏的主因是新词率太高,而不是缺少累积优势。书库是群岛,不是全息网;合成模型显示,半数新卷各做一次桥接引用,就能让大部分新卷接入同一巨连通块。蒸馏漏斗里接上付费路由的卷只占 0.160。 回路方面,书库有 769,420 对潜在关键词组合,已实现的只占 0.00257,增长受处理能力而非点子约束,再投入使六十个月新增想法达到固定预算的 1.77 倍。自持写成不等式:再投入率乘以扣除平台费与退款之后的回报率,须大于折旧率加维护漏损率,且资本须高于临界规模。示意基准下 g 为 0.018,临界规模 55.6,再投入率可行带 0.294 到 0.529;相图里同时满足增长与个人报酬条件的格子只占 0.231。蒙特卡洛给出三十六个月破产率 0.0002、跑道第十百分位 12.2 个月;破产率对再投入率呈 U 形,再投入率 0.80 时为 0.2380。稀缺心态文献只作小效应与争议并陈;「自由态」只写成资源与时间条件,自动化会被新器官的维护反弹吃掉,退役率 0.01 时须有 0.880 的自动化才达到每周二十个自由小时的示意门槛。 第七章考辨《庄子·逍遥游》的「无待」、《礼记·大学》的「日新」、凯恩斯 1930 年论闲暇的短文与亚里士多德《尼各马可伦理学》第十卷,章末给出四件套蒸馏规格。第八章给出共同体自持台账、蒸馏小工具「现金流算力自持方程与知识网络增长试算器」、按四件套分组的检查器规格与预注册证伪方案。书中参数为示意,现金流全部为抽象单位,不写任何真实收入、成交、账户或服务器;全部计算由书中 21 段标准库程序生成。本书是非个人化的经营分析,不构成经营、法律、税务或投资建议,不承诺任何收入,也不承诺任何心理或身体效果。 Chapter answers: - Ch.1 Four Hundred Volumes, Counted Honestly: Numbered, Delivered and Still Being Written — Of the three hundred and thirty-two volumes, 158 have some distillation target, a share of 0.476 with a 95 per cent Wilson interval from 0.423 to 0.530. A small tool has been built for 100, a share of 0.301. A rule in one of the five kits names 61 of them as its source, a share of 0.184. And 53 are wired to a paid route, a share of 0.160, with an interval from 0.124 to 0.203. There are 174 volumes, a share of 0.524, with no target at all. (https://5dlibrary.com/answers/en/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c01/) 第1章 四百卷如实计数:编号、已交付与在写 — 三百三十二卷里,有任何蒸馏去向的是 158 卷,占 0.476,百分之九十五的威尔逊区间是 0.423 到 0.530;建成小工具的是 100 卷,占 0.301;被五件套某条规则点名为出处的是 61 卷,占 0.184;接上付费路由的是 53 卷,占 0.160,区间 0.124 到 0.203。没有任何去向的有 174 卷,占 0.524。 (https://5dlibrary.com/answers/zh/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c01/) - Ch.2 A Sparse Network of Axioms: Degree Distribution, Clustering, Connected Pieces and Modularity — Lower the rate of new words to 0.800 and differences begin to show. Under cumulative advantage the number of distinct words falls to 1,064.4, the most-used word appears 9.4 times, and 8.6 words appear five times or more. Lower the rate to 0.500 and cumulative advantage produces clear hubs: the most-used word appears 58.2 times on average and 44.9 words appear five times or more, while under uniform reuse the most-used word appears only 10.7 times. The tail exponent of Simon's model says the same thing. At a new-word rate of 0.933 the exponent is 15.93, so steep that there is practically no tail. At 0.800 it is 6.00. Only at 0.500 does it fall to 3.00, close to the order of magnitude usual for citation networks. (https://5dlibrary.com/answers/en/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c02/) 第2章 稀疏的公理网络:度分布、聚类、连通块与模块度 — 把新词率降到 0.800,差别开始出现:累积优势下不同词降到 1,064.4 个,最常用词出现 9.4 次,而有 8.6 个词出现五次以上。降到 0.500,累积优势给出明显的枢纽,最常用词平均出现 58.2 次,五次以上的词有 44.9 个;均匀复用下最常用词只出现 10.7 次。西蒙模型的尾部指数也说明同一件事:新词率 0.933 时指数是 15.93,尾部陡到几乎没有尾;0.800 时是 6.00;0.500 时才降到 3.00,接近引文网络常见的量级。 (https://5dlibrary.com/answers/zh/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c02/) - Ch.3 Hub Volumes, Isolated Volumes and Governance Rules for a Knowledge Commons — The synthetic library uses 1,250 keywords, close to the real 1,241 reported by program p12 in Chapter 1. It has 90 links against 87, an isolated share of 0.651 against 0.699, and a largest connected piece of 19 volumes against 14. Two measures are further off. Transitivity is 0.396, below the real 0.639, and the modularity of the zone partition is 0.397, above the real 0.285. (https://5dlibrary.com/answers/en/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c03/) 第3章 枢纽卷、孤立卷与知识共同体的治理规则 — 合成书库用掉 1,250 个关键词,与真实的 1,241(第一章程序 p12)很接近。边数 90 对 87,孤立比例 0.651 对 0.699,最大连通块 19 卷对 14 卷。差得较远的是两项:传递性 0.396 低于真实的 0.639,战区模块度 0.397 高于真实的 0.285。 (https://5dlibrary.com/answers/zh/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c03/) - Ch.4 Recombinant Innovation: Combinatorial Explosion, the Processing Bottleneck and the Atypical Tail — In the first case, the library as it stands, the stock starts at 1,241 ideas and there are 200 attempts each month. The stock of ideas rises from 1,241.0 to 1,289.0, then 1,337.0, 1,385.0 and 1,433.0, and after sixty months reaches 1,481.0. It grows by a fixed 4.0 ideas a month, which is a straight line. The untried combinations are so many that they can never be used up, and the constraint comes entirely from how many attempts can be made each month. (https://5dlibrary.com/answers/en/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c04/) 第4章 重组创新:组合爆炸、处理能力瓶颈与非典型尾部 — 第一种,书库现状,起点 1,241 个想法,每月固定尝试 200 次。想法存量从 1,241.0 依次升到 1,289.0、1,337.0、1,385.0、1,433.0,六十个月后是 1,481.0。每月固定增加 4.0 个,这是一条直线。待试组合多到永远用不完,约束完全来自每月能尝试多少次。 (https://5dlibrary.com/answers/zh/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c04/) - Ch.5 The Cash-and-Compute Loop: the Self-Sustaining Inequality, the Phase Diagram and the Critical Scale — The fixed cost gives the loop an extra, unstable point of balance. Under the baseline the critical scale is 55.6 points. Starting from 40.0 points, capital has fallen to 36.3 after twelve months, 26.0 after thirty-six and 10.2 after sixty, on its way to zero. Starting from 55.0 points, just below the critical scale, capital drifts slowly down to 53.9 after sixty months. Starting from 56.0 points, just above it, capital drifts slowly up to 56.9. From 80.0 points it reaches 126.8 after sixty months. From 100.0 points it passes through 110.6, 140.0 and 185.2. From 160.0 points it reaches 360.2 after sixty months. (https://5dlibrary.com/answers/en/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c05/) 第5章 现金流与算力回路:自持不等式、相图与临界规模 — 固定成本让回路多了一个不稳定的平衡点。基准下临界规模是 55.6 点。起始资本 40.0 点时,十二个月后剩 36.3,三十六个月后 26.0,六十个月后 10.2,一路走向零。起始资本 55.0 点,只比临界规模低一点点,六十个月后 53.9,缓慢下滑。起始资本 56.0 点,只比临界规模高一点点,六十个月后 56.9,缓慢上升。起始 80.0 点,六十个月后 126.8;起始 100.0 点,依次是 110.6、140.0 与 185.2;起始 160.0 点,六十个月后 360.2。 (https://5dlibrary.com/answers/zh/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c05/) - Ch.6 Runway, Ruin and the Kelly Analogy: Resource Conditions for Autonomous Time — The second part applies the same idea to the reinvestment rate, running two thousand paths for each rate. At a reinvestment rate of 0.25, g is negative, the median capital shrinks to 42.6 points and the ruin rate is 0.0070. At 0.30, g is only +0.0010, the median capital is 63.1 and the ruin rate is 0.0025. At 0.40 and 0.50 the ruin rate is 0.0000 in both cases, with median capital of 130.9 and 253.0. At 0.60 the ruin rate comes back to 0.0010, at 0.70 it is 0.0090, and at 0.80 it jumps to 0.2380, while median capital reaches 1,531.4 points. (https://5dlibrary.com/answers/en/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c06/) 第6章 跑道、破产与凯利类比:自主时间的资源条件 — 第二部分把同样的思路用到再投入率上,每个再投入率跑两千条路径。再投入率 0.25 时 g 为负,资本中位数缩到 42.6 点,破产率 0.0070。再投入率 0.30 时 g 只有 +0.0010,资本中位数 63.1,破产率 0.0025。再投入率 0.40 与 0.50 时破产率都是 0.0000,资本中位数 130.9 与 253.0。再投入率 0.60 时破产率回到 0.0010,0.70 时 0.0090,0.80 时跳到 0.2380,而资本中位数高达 1,531.4 点。 (https://5dlibrary.com/answers/zh/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c06/) - Ch.7 Eastern and Western Sources: "Depending on Nothing", "Daily Renewal", and Keynes and Aristotle on Leisure — Chapter 7 examines "depending on nothing" in the Zhuangzi, "daily renewal" in the Great Learning, Keynes's 1930 essay on leisure and Book X of Aristotle's Nicomachean Ethics, and ends with a distillation spec for the four kits. (https://5dlibrary.com/answers/en/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c07/) 第7章 东西方原典考辨:「无待」「日新」与凯恩斯、亚里士多德论闲暇 — 第七章考辨《庄子·逍遥游》的「无待」、《礼记·大学》的「日新」、凯恩斯 1930 年论闲暇的短文与亚里士多德《尼各马可伦理学》第十卷,章末给出四件套蒸馏规格。 (https://5dlibrary.com/answers/zh/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c07/) - Ch.8 Field Tools and Checker Specifications: the Self-Sustenance Ledger, the Calculator and a Pre-Registered Refutation — Chapter 8 gives the self-sustenance ledger, the distilled Cash-Compute Self-Sustenance Equation and Knowledge-Network Growth Calculator, checker specifications grouped by kit and a pre-registered plan of refutation. Parameters are illustrative and cash is in abstract units; no real revenue, transaction, account or server appears. All calculations come from 21 standard-library programs in the book. It is general business analysis, not business, legal, tax or investment advice, and promises no income and no psychological or bodily effect. (https://5dlibrary.com/answers/en/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c08/) 第8章 实战工具与检查器规格:自持台账、试算器与预注册证伪 — 第八章给出共同体自持台账、蒸馏小工具「现金流算力自持方程与知识网络增长试算器」、按四件套分组的检查器规格与预注册证伪方案。书中参数为示意,现金流全部为抽象单位,不写任何真实收入、成交、账户或服务器;全部计算由书中 21 段标准库程序生成。本书是非个人化的经营分析,不构成经营、法律、税务或投资建议,不承诺任何收入,也不承诺任何心理或身体效果。 (https://5dlibrary.com/answers/zh/four-hundred-volume-holographic-knowledge-commonwealth-cashflow-compute-loop/c08/) ## 5DL-B-409 · Ambient Multimodal Cognitive Adjutants: Low-Latency Audio-Visual Memory Streams, Proactive Intent Anticipation and Real-Time Decision Support 中文书名:全天候多模态情境感知智能副官、低延迟视听记忆流与主动预判辅助母典 DOI: pending Cite: Fan, Y. (2026). Ambient Multimodal Cognitive Adjutants: Low-Latency Audio-Visual Memory Streams, Proactive Intent Anticipation and Real-Time Decision Support (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and data synthetic; all calculations are generated by the book's 23 standard-library programs. The book is not individual business advice or legal advice; capture is limited to devices and scenes the wearer controls, bystander content is not kept by default or is redacted at once, and no effect on hearing or the body is claimed. Abstract: Four popular beliefs surround personal AI adjutants: speak up early and often; record everything; the better the prediction, the more the adjutant should act ahead of the person; and confirm every risky action with a dialog. This book falsifies each in turn and rewrites it as a proposition a ledger can overturn. The net value of proactive prompts is bounded by a hit-rate threshold: 0.417 for a definition hint on a chart desk but only 0.073 for a gate red-line reminder. An overconfident adjutant that prompts every time loses 4,610 seconds a week, while a calibrated one that waits for task boundaries gains 5,623. The endpoint wait of voice activity detection is a floor under latency; in an illustrative architecture end-to-end latency has a median of 1,083 ms and a 95th percentile of 1,540 ms, with only 0.323 of requests within one second, so "zero latency", "before you speak" and "cinematic" are rewritten as percentiles with conditions of refutation. Speculative prefetch fetches read-only data only, and sending, paying, deleting and deploying wait for a person; prefetch has a break-even hit rate, and on a busy server even one prefetched item raises the mean wait from 76.9 to 85.9 ms. The rolling memory window has an optimal length that shrinks once bystander exposure is priced; deleting only originals leaves 2,883 derived copies after a thousand requests. Confirmation is tiered by risk: confirming everything lets high-risk errors through at 0.753, tiering at 0.233. Chapter 7 examines "seeing the first sign and acting" from the Book of Changes, "preparation makes things stand" from the Doctrine of the Mean, Augustine's palaces of memory and Bush's memex, and gives the distillation spec for the four kits. Chapter 8 gives the event ledger, the Prompt Threshold and Prefetch Latency Budgeter, a checker table grouped by the four kits and a pre-registered test needing 1,110 events per arm. Parameters are illustrative and data synthetic; all calculations are generated by the book's 23 standard-library programs. The book is not individual business advice or legal advice; capture is limited to devices and scenes the wearer controls, bystander content is not kept by default or is redacted at once, and no effect on hearing or the body is claimed. 摘要:关于贴身智能副官,流行说法有四条:插话越早越勤越好;一切都录下来最保险;预测越准越该替人提前动手;凡有风险一律弹窗确认。本书逐条证伪,并把它们改写成可以用台账推翻的命题。主动提示的净价值受命中率阈值约束,带盘参谋的术语补充提示阈值为 0.417,门禁红线提醒只有 0.073;过度自信的副官一律提示,一周净亏 4,610 秒,校准良好并等到任务边界再说,一周净赚 5,623 秒。语音端点等待是延迟的地板,示意架构下端到端延迟中位数 1,083 毫秒、95 分位 1,540 毫秒,一秒内完成的只有 0.323,「零延迟」「人类开口前」「电影级」都被改写为分位数与推翻条件。 投机预取只拉只读数据,外发、付款、删除、部署一律等人确认;预取有盈亏平衡命中率,服务器忙时预取一项也会让用户平均等待从 76.9 毫秒升到 85.9 毫秒。滑动记忆窗存在最优长度,计入旁人暴露后更短;只删原件,一千次删除后留下 2,883 份派生副本。确认按风险分级,一律弹窗时高危错误放行率 0.753,分级后为 0.233。第七章考辨《周易·系辞下》「见几而作」、《礼记·中庸》「豫则立」、奥古斯丁的记忆殿堂与 Bush 的 memex,并给出四件套蒸馏规格;第八章给出副官事件台账、蒸馏小工具「主动提示阈值与投机预取延迟预算器」、按四件套分组的检查器规格表与预注册检验,主检验每组需要 1,110 条事件。 书中参数为示意,数据为合成;全部计算由书中 23 段标准库程序生成。本书是非个人化经营建议,不构成法律意见;采集只限佩戴者本人可控的设备与场景,旁人内容默认不留存或即时脱敏;不涉及任何听觉或身体功效。 Chapter answers: - Ch.1 The Hit-Rate Threshold: When Speaking Up Pays — Section A has four rows, one for each kind of prompt the book returns to. The follow-up hint on the chart desk, where the user has asked about chart structure and the adjutant wants to add a definition, is worth only 20 seconds against an interruption of 6, and its threshold is 0.417; at a boundary it drops to 0.267. Handing over context at the moment the commander switches fronts is worth far more, and its threshold is only 0.188, or 0.113 at a boundary. The order-group digest has a threshold of 0.371, or 0.234 at a boundary. The gate red-line reminder carries the largest benefit, and its threshold falls to 0.073, or 0.048 at a boundary. (https://5dlibrary.com/answers/en/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c01/) 第1章 命中率阈值:主动插话何时净为正 — A 段四行,对应本书反复回到的四种提示。带盘参谋的追问提示(用户问完命盘结构,副官想补一句术语解释)收益只有 20 秒,打断 6 秒,阈值是 0.417;换到边界处说,降到 0.267。跨战线切换时的上下文递送收益大得多,阈值只有 0.188,边界处 0.113。订单群摘要的阈值是 0.371,边界处 0.234。门禁红线提醒收益最大,阈值低到 0.073,边界处 0.048。 (https://5dlibrary.com/answers/zh/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c01/) - Ch.2 Voice Activity Detection: the Endpoint Wait as a Latency Floor — The track holds 498,384 samples, cut into 3,114 frames. The noise floor is estimated at -32.7 dB. There are 1,442 background frames, sitting at the floor (-0.2 dB) with a zero-crossing rate of 0.143. The 1,073 voiced frames sit on average 12.1 dB above the floor with a zero-crossing rate of only 0.060. The 160 fricative frames sit just 3.9 dB above the floor on average, yet their zero-crossing rate reaches 0.468. The 420 frames of within-utterance pauses look almost exactly like the background (-0.2 dB and 0.150). The 19 knock frames are 7.7 dB above the floor, with a zero-crossing rate of 0.351. (https://5dlibrary.com/answers/en/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c02/) 第2章 语音活动检测:端点等待是延迟的地板 — 整条音轨有 498,384 个采样点,切成 3,114 帧。噪声底估计为 -32.7 分贝。底噪帧共 1,442 帧,能量在底噪附近(-0.2 分贝),过零率 0.143。浊音帧 1,073 帧,平均高出底噪 12.1 分贝,过零率只有 0.060。擦音帧 160 帧,平均只高出 3.9 分贝,过零率却高达 0.468。句中停顿 420 帧,特征与底噪几乎一样(-0.2 分贝、0.150)。敲击声 19 帧,能量高出 7.7 分贝,过零率 0.351。 (https://5dlibrary.com/answers/zh/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c02/) - Ch.3 End-to-End Latency Budgets: Percentiles, Not Slogans — Section A is the itemised bill for the baseline architecture. The largest item is generation of the first token, with a median of 379 milliseconds, a 95th percentile of 797 and a 99th percentile of 1,073; it also has the longest tail. Next comes the endpoint wait, with a median of 260 milliseconds and a 95th percentile of 328, the floor described in Chapter 2, which barely moves. Recognition has a median of 160 milliseconds. Retrieval has a median of only 70, but its 95th percentile jumps to 219, the widest spread of any stage. The total has a median of 1,083 milliseconds, a 95th percentile of 1,540 and a 99th percentile of 1,820. (https://5dlibrary.com/answers/en/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c03/) 第3章 端到端延迟预算:分位数而非口号 — 解析结果一目了然:至少碰上一个慢分片的概率是 $1 - 0.99^k$。$k$ 为 1 时只有 0.010,$k$ 为 10 时 0.096,$k$ 为 50 时 0.395,$k$ 为 100 时 0.634。对应的 99 分位从 53 毫秒涨到 564、718、942,再到 1,177 与 1,303 毫秒。扇出到一百处时,连中位数都变成了 315 毫秒,因为超过一半的请求都碰上了慢分片。 (https://5dlibrary.com/answers/zh/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c03/) - Ch.4 Speculative Prefetch: Read-Only, Break-Even and Server Load — Without prefetch the median is 967 milliseconds, the 95th percentile 1,416 and the 99th percentile 1,697, with 0.564 of requests finished within one second. At a hit rate of 0.30 the median falls to 891 but the 95th percentile only to 1,352. At 0.60 the median is 809 and the 95th percentile 1,263. At 0.90 the median is 744, the 95th percentile 1,085 and the share within one second 0.912. At 0.99 the 95th percentile is 1,008 and the 99th percentile 1,194. (https://5dlibrary.com/answers/en/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c04/) 第4章 投机预取:只读、盈亏平衡与服务器负载 — 不预取时,中位数 967 毫秒,95 分位 1,416,99 分位 1,697,一秒内完成 0.564。命中率 0.30 时,中位数降到 891,95 分位只降到 1,352。命中率 0.60 时,中位数 809,95 分位 1,263。命中率 0.90 时,中位数 744,95 分位 1,085,一秒内完成 0.912。命中率 0.99 时,95 分位 1,008,99 分位 1,194。 (https://5dlibrary.com/answers/zh/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c04/) - Ch.5 The Rolling Memory Window: Forgetting Curves, Bystander Exposure and One-Tap Deletion — With no price on bystanders ($k = 0$) the optimal window remains 50 minutes and holds 12.50 bystander-minutes at any moment; with on-device redaction the same 50-minute window holds only 1.10. At a price of 0.50 the optimal window without redaction shrinks to 29 minutes and utility falls to 0.542, while with redaction the window stays at 48 minutes, utility is 0.559, and exposure is only 1.08. At a price of 2.00 the window without redaction is just 14 minutes, with utility 0.475 and a net value of 0.359; with redaction the window is 43 minutes and the net value 0.524. (https://5dlibrary.com/answers/en/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c05/) 第5章 滑动记忆窗:遗忘曲线、旁人暴露与一键删除 — 不计旁人代价时($k = 0$),最优窗口仍是 50 分钟,此刻持有 12.50 个旁人分钟;开启就地脱敏后,同样 50 分钟窗口只持有 1.10 个旁人分钟。代价系数升到 0.50,不脱敏的最优窗口缩到 29 分钟,效用降到 0.542;开启脱敏后最优窗口仍有 48 分钟,效用 0.559,暴露只有 1.08。系数升到 2.00,不脱敏的窗口只剩 14 分钟,效用 0.475,净值 0.359;脱敏后窗口 43 分钟,净值 0.524。 (https://5dlibrary.com/answers/zh/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c05/) - Ch.6 Risk-Tiered Confirmation: Fatigue and the Undo Window — Holding for 5 seconds catches almost nothing (0.014) and has a negative net value. Ten seconds catch 0.062, 30 seconds 0.290, 60 seconds 0.493, 120 seconds 0.648 and 300 seconds 0.734, but by five minutes the cost of delay exceeds the gain and the net value turns negative. The best hold is 74 seconds, catching 0.548 for a net value of 9.04 per thousand sends. The 60-second undo window used for reversible actions in program p62, with a catch rate of 0.49, comes from this calculation. (https://5dlibrary.com/answers/en/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c06/) 第6章 风险分级确认:确认疲劳与撤销窗口 — 暂存 5 秒几乎拦不住什么(0.014),净值为负。10 秒拦下 0.062,30 秒拦下 0.290,60 秒拦下 0.493,120 秒拦下 0.648,300 秒拦下 0.734,但五分钟的延误代价已经超过收益,净值转负。最优暂存是 74 秒,拦下 0.548,每千条净值 9.04。程序 p62 里可撤销动作用的 60 秒撤销窗口,拦截率 0.49,正来自这里。 (https://5dlibrary.com/answers/zh/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c06/) - Ch.7 Seeing the First Sign and the Palaces of Memory: an East-West Source Critique — If only 50 minutes are kept, there are 300 segments: an unindexed search inspects 150 on average, a time index 101, and a trail of curated notes 92. Keep one day, 5,760 segments, and the unindexed search inspects 2,880, the time index 1,195, while the trail still needs only 96. Keep seven days, 40,320 segments, and the unindexed search inspects 20,160, the time index 8,109. Keep a year, 2,102,400 segments, and the unindexed search inspects 1,051,200 on average and the time index 420,529, while the trail of notes needs only 105. (https://5dlibrary.com/answers/en/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c07/) 第7章 见几而作与记忆殿堂:东西方原典考辨 — 只保留 50 分钟时,共 300 段,无索引平均要看 150 段,按时间索引要看 101 段,按笔记线索只要 92 段。保留一天,5,760 段,无索引要看 2,880 段,时间索引 1,195 段,笔记线索仍是 96 段。保留七天,40,320 段,无索引 20,160 段,时间索引 8,109 段。保留一年,2,102,400 段,无索引平均要看 1,051,200 段,时间索引 420,529 段,笔记线索只要 105 段。 (https://5dlibrary.com/answers/zh/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c07/) - Ch.8 The Adjutant Event Ledger: Field Tools and Checker Specifications — Chapter 8 gives the event ledger, the Prompt Threshold and Prefetch Latency Budgeter, a checker table grouped by the four kits and a pre-registered test needing 1,110 events per arm. (https://5dlibrary.com/answers/en/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c08/) 第8章 副官事件台账、实战工具与检查器规格 — 第八章给出副官事件台账、蒸馏小工具「主动提示阈值与投机预取延迟预算器」、按四件套分组的检查器规格表与预注册检验,主检验每组需要 1,110 条事件。 (https://5dlibrary.com/answers/zh/ambient-multimodal-cognitive-adjutants-audio-visual-memory-streams-proactive-intent-anticipation/c08/) ## 5DL-B-410 · Bicameral AI Architecture: Fusing Long-Context Holographic Strategy Models with Local Deterministic Execution Agents 中文书名:双脑异构大模型协同架构:全息战略参谋与本地确定性执行引擎融合母典 DOI: pending Cite: Fan, Y. (2026). Bicameral AI Architecture: Fusing Long-Context Holographic Strategy Models with Local Deterministic Execution Agents (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; "bicameral" is an engineering metaphor and no model is claimed to have awareness; the book is not personalised business or engineering advice. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. Abstract: Three popular beliefs promise fewer errors from large language models: use a bigger and stronger model; let two models vote and release only what they agree on; let the model reflect on and correct itself. This book refutes their shared premise, that errors are independent. After checking the literature it states its thesis as follows: the gain from combining two heterogeneous models comes from the lack of correlation between their errors, so heterogeneity is the crux and duplication is not; a long-context strategy model deliberates, while a local deterministic execution engine uses gates, tests and receipts to turn errors into intercepted errors; and correction must be driven by executable test receipts. Its parameters are illustrative; all calculations are generated by its programs; "bicameral" is an engineering metaphor and no model is claimed to have awareness; the book is not personalised business or engineering advice. Chapter 1 uses a Gaussian copula to show that, at illustrative error rates, the second model's gain is 0.0880 at a latent correlation of 0 and only 0.0243 at 0.9, and that a weaker but different reviewer leaks only 0.498 of what a seemingly powerful look-alike leaks. Chapter 2 shows that binary correlations on a ledger fall far below the latent correlation, and proves by enumeration that diversity and voting accuracy are not monotonic: majority accuracy can range from 0.4 to 0.9. Chapter 3 shows that a common-factor correlation of only 0.1 makes a three-stage pipeline leak 1.87 times the independent product, and 8.65 times at 0.5. Chapter 4 counts ambiguity with a five-slot intent packet and its entropy, and rewrites "zero ambiguity" as an audit bound: 299 clean packets in a row are needed before the ambiguity rate can be stated below 0.010. Chapter 5 builds state-sync packets with hash chains, version vectors and idempotency keys; when 10,000 instructions are delivered 849 extra times, the keys keep side effects at exactly 10,000. Chapter 6 gives the stable interval of the correction gain and shows how it narrows when receipts arrive a round late; self-reflection without an outside signal pulls accuracy towards 0.6250, receipt-driven revision towards 0.9600. Chapter 7 examines temple reckoning in Sunzi, "breadth with precision" in the Doctrine of the Mean and Clausewitz on friction, and ends with a distillation spec for four business kits. Chapter 8 gives the bicameral ledger, the distilled Bicameral Error-Orthogonality Gain and State-Sync Packet Checker, a checker specification table grouped by kit, and a pre-registered test: lowering the leak from 0.030 to 0.015 needs 1,534 tasks per arm at power 0.80. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. 摘要:谈到让大模型少犯错,流行说法有三种:换一个更大更强的模型;让两个模型投票,一致才放行;让模型自己反思纠错。本书证伪它们共同的前提:错误彼此独立。核对文献后,本书把立项命题写成:两个异构模型融合的增益来自误差不相关,异构是要害,双份不是;长上下文战略模型负责推演,本地确定性执行引擎用门禁、测试与回执把错误转成被拦下的错误;纠偏必须以可执行的检验回执为信号。书中参数为示意;全部计算由书中程序生成;「双脑」只是工程隐喻,不宣称任何模型有意识;本书是非个人化的经营与工程建议。 第一章用高斯 copula 说明,示意错误率下第二个模型的增益在潜在相关为 0 时是 0.0880,相关 0.9 时只剩 0.0243;较弱但不同的复核者,漏网只有看似强大的相似复核者的 0.498。第二章说明台账上的二值相关远低于潜在相关,并以穷举证明多样性与投票精度并不单调,多数决精度可从 0.4 到 0.9。第三章说明三段流水线只要有 0.1 的公共因子相关,漏网就是独立乘积的 1.87 倍,相关 0.5 时是 8.65 倍。第四章用五槽位意图包与歧义熵计数,并把「零歧义」改写为审计上界:零发现须连续审计 299 份,才能说歧义率低于 0.010。 第五章给出状态同步包的哈希链、版本向量与幂等键,10,000 条指令重复送达 849 次时,幂等键使副作用恰为 10,000 次。第六章给出纠偏增益的稳定区间,并说明回执晚到一轮时区间收窄;无外部信号的自我反思把精度拉向 0.6250,回执驱动的修订拉向 0.9600。第七章考辨《孙子兵法》的庙算、《礼记·中庸》的「致广大而尽精微」与克劳塞维茨论摩擦,章末给出四件套蒸馏规格。第八章给出双脑协同台账、蒸馏小工具「双脑误差正交增益与状态同步包校验器」、按四件套分组的检查器规格表与预注册检验:漏网率从 0.030 降到 0.015 的对照,功效 0.80 时每组需要 1,534 个任务。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Correlated Errors: Where the Second Model's Gain Comes From — Chapter 1 uses a Gaussian copula to show that, at illustrative error rates, the second model's gain is 0.0880 at a latent correlation of 0 and only 0.0243 at 0.9, and that a weaker but different reviewer leaks only 0.498 of what a seemingly powerful look-alike leaks. (https://5dlibrary.com/answers/en/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c01/) 第1章 误差相关:第二个模型的增益从哪里来 — 第一章用高斯 copula 说明,示意错误率下第二个模型的增益在潜在相关为 0 时是 0.0880,相关 0.9 时只剩 0.0243;较弱但不同的复核者,漏网只有看似强大的相似复核者的 0.498。 (https://5dlibrary.com/answers/zh/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c01/) - Ch.2 Diversity Measures and the Ambiguity Decomposition: How Different Is Different Enough — Chapter 2 shows that binary correlations on a ledger fall far below the latent correlation, and proves by enumeration that diversity and voting accuracy are not monotonic: majority accuracy can range from 0.4 to 0.9. (https://5dlibrary.com/answers/en/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c02/) 第2章 多样性度量与歧义分解:异构到底要多异 — 第二章说明台账上的二值相关远低于潜在相关,并以穷举证明多样性与投票精度并不单调,多数决精度可从 0.4 到 0.9。 (https://5dlibrary.com/answers/zh/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c02/) - Ch.3 The Three-Stage Pipeline: Turning Errors into Intercepted Errors — Chapter 3 shows that a common-factor correlation of only 0.1 makes a three-stage pipeline leak 1.87 times the independent product, and 8.65 times at 0.5. (https://5dlibrary.com/answers/en/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c03/) 第3章 三段流水线:把错误变成被拦下的错误 — 第三章说明三段流水线只要有 0.1 的公共因子相关,漏网就是独立乘积的 1.87 倍,相关 0.5 时是 8.65 倍。 (https://5dlibrary.com/answers/zh/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c03/) - Ch.4 The Intent Translation Protocol: Slots, Ambiguity Entropy and Upper Bounds — Chapter 4 counts ambiguity with a five-slot intent packet and its entropy, and rewrites "zero ambiguity" as an audit bound: 299 clean packets in a row are needed before the ambiguity rate can be stated below 0.010. (https://5dlibrary.com/answers/en/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c04/) 第4章 意图转译协议:槽位、歧义熵与上界 — 第四章用五槽位意图包与歧义熵计数,并把「零歧义」改写为审计上界:零发现须连续审计 299 份,才能说歧义率低于 0.010。 (https://5dlibrary.com/answers/zh/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c04/) - Ch.5 State-Sync Packets: Hash Chains, Version Vectors and Idempotency Keys — Chapter 5 builds state-sync packets with hash chains, version vectors and idempotency keys; when 10,000 instructions are delivered 849 extra times, the keys keep side effects at exactly 10,000. (https://5dlibrary.com/answers/en/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c05/) 第5章 状态同步包:哈希链、版本向量与幂等键 — 第五章给出状态同步包的哈希链、版本向量与幂等键,10,000 条指令重复送达 849 次时,幂等键使副作用恰为 10,000 次。 (https://5dlibrary.com/answers/zh/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c05/) - Ch.6 Two-Way Correction: Gain, Noise and the Limits of Self-Correction — Chapter 6 gives the stable interval of the correction gain and shows how it narrows when receipts arrive a round late; self-reflection without an outside signal pulls accuracy towards 0.6250, receipt-driven revision towards 0.9600. (https://5dlibrary.com/answers/en/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c06/) 第6章 双向纠偏:增益、噪声与自我纠错的边界 — 第六章给出纠偏增益的稳定区间,并说明回执晚到一轮时区间收窄;无外部信号的自我反思把精度拉向 0.6250,回执驱动的修订拉向 0.9600。 (https://5dlibrary.com/answers/zh/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c06/) - Ch.7 Eastern and Western Classical Sources: Temple Reckoning, Breadth with Precision, and Friction — Block A is the arithmetic behind the remark that the simplest thing is difficult. A per-step success rate of 0.98 sounds reassuring. After twenty steps the whole order succeeds with probability only 0.6676, and after forty only 0.4457. Lower the per-step rate to 0.95 and twenty steps leave 0.3585, so that most orders fail even though every individual step works nineteen times in twenty. No single step is the culprit; the product is. Block B adds a receipt at every step. With a per-step rate of 0.98 and twenty steps, the order now succeeds with probability 0.9538; at forty steps, 0.9098. Block C converts the comparison into failures per 1,000 orders. At twenty steps and 0.98 per step, an operation without receipts loses 332.4 orders in every thousand; with receipts it loses 46.2. (https://5dlibrary.com/answers/en/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c07/) 第7章 东西方原典考辨:庙算、致广大与摩擦 — A 段:复核彼此独立时,五次复核发现缺陷的概率是 0.8319,十次 0.9718,二十次 0.9992,「多算」果然胜。复核相关 0.6 时,十次只有 0.6982,二十次 0.7817;相关 0.9 时,二十次也只有 0.5272。B 段算出要把发现概率提到 0.90 需要几次:独立时 7 次,相关 0.3 时 15 次,相关 0.6 时 96 次。C 段把对比写得更直白:五次不同的复核得 0.8319,二十次相似的复核只得 0.7817。 (https://5dlibrary.com/answers/zh/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c07/) - Ch.8 The Empirical Loop: Field Tools and Checker Specifications — Chapter 8 gives the bicameral ledger, the distilled Bicameral Error-Orthogonality Gain and State-Sync Packet Checker, a checker specification table grouped by kit, and a pre-registered test: lowering the leak from 0.030 to 0.015 needs 1,534 tasks per arm at power 0.80. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c08/) 第8章 实证闭环:实战工具与检查器规格 — 第八章给出双脑协同台账、蒸馏小工具「双脑误差正交增益与状态同步包校验器」、按四件套分组的检查器规格表与预注册检验:漏网率从 0.030 降到 0.015 的对照,功效 0.80 时每组需要 1,534 个任务。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/bicameral-ai-architecture-holographic-strategy-models-local-deterministic-execution-agents/c08/) ## 5DL-B-411 · Topological Coupling of Concurrent Multi-Core Operations: Lock-Free Blackboard Signalling and Cross-Node Emergent Gains 中文书名:多核并发操作拓扑耦合、共享黑板无锁通信与跨节点涌现增益母典 DOI: pending Cite: Fan, Y. (2026). Topological Coupling of Concurrent Multi-Core Operations: Lock-Free Blackboard Signalling and Cross-Node Emergent Gains (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative, data are synthetic, and every figure is generated by the book's programs; the book is not individual business advice and promises no speed-up. Chapter 1 shows where the peak comes from, near 6.78 windows under illustrative parameters. Abstract: Two beliefs circulate wherever work runs in parallel windows: the more windows, the faster; and every node should talk to every other. This book refutes the half of each that is repeated as a law. Concurrent throughput is bounded jointly by the serial fraction and by coherence overhead, so under Gunther's universal scalability law it has a peak, beyond which extra nodes slow the whole down; coupling between nodes has an optimum, and neither full isolation nor full connection is it; and "lock-free", "conflict-free" and "emergent" are rewritten as quantities that can be measured, given intervals and overturned. After checking the sources, the book corrects the blueprint: the scalability law was first published in 1993 as a super-serial capacity model and named later; Mason and colleagues concluded in 2008 that the best network depends on the problem, not that small worlds always win; and "lock-free" means only that reads take no lock and writes only append, with the append itself atomic. Parameters are illustrative, data are synthetic, and every figure is generated by the book's programs; the book is not individual business advice and promises no speed-up. Chapter 1 shows where the peak comes from, near 6.78 windows under illustrative parameters. Chapter 2 fits σ and κ by linearised least squares; a stratified bootstrap gives a peak interval from 6.37 to 7.80 and shows that 0.962 of resamples put eight windows past the peak. Chapter 3 simulates exclusive-create locks in memory: overwriting loses 0.907 of updates with sixteen writers, exclusive creation never lets two windows hold one number, and ordered requests or single holding rule out deadlock. Chapter 4 explains why an append-only board needs no lock on the read path, with a best subscription share near 0.7 under illustrative parameters. Chapter 5 compares four topologies on NK landscapes: full connection starts fastest, but by round 80 the ring reaches 0.9280 against 0.9150; semi-isolated subgroups peak at 0.9259 with moderate cross-group links. Chapter 6 writes emergence as cross-node reuse and a net-gain interval on a hypergraph incidence matrix; a reading lag of one day removes 0.944 of the repeated pitfalls seen under isolation. Chapter 7 examines Sunzi on division and numbers, the Analects on harmony without sameness, Leibniz's windowless monads and Smith's division of labour, and turns them into clauses for four kits. Chapter 8 gives the Concurrency Coupling and Blackboard Reuse Calculator, a checker table, the ledger rhythm and a pre-registered test needing 156 records per arm. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:多窗口并行作业里流行两句话:窗口开得越多越快;节点之间最好全部互通。本书证伪两句话被说成定律的那一半。并发吞吐同时受串行比例与相干开销约束,按 Gunther 的通用可扩展性定律存在峰值,过了峰值加节点反而变慢;节点之间的耦合度存在最优值,全隔离与全连通都不是最好;「无锁」「零冲突」「涌现」都改写成可测量、给区间、可被推翻的量。核对原文后,本书修正了立项蓝图的几处说法:通用可扩展性定律最初在 1993 年以超串行容量模型发表,名称是后来固定的;Mason 等人 2008 年的结论是最优网络取决于问题,不是小世界一律最好;「无锁」只指读路径不取锁、写路径只追加,追加本身仍须原子。书中参数为示意,数据为合成,全部计算由书中程序生成;本书是非个人化经营建议,不提供任何加速比承诺。 第一章说明峰值从何而来,示意参数下峰值在 6.78 个窗口附近。第二章用线性化最小二乘拟合 σ 与 κ,分层自助法给出峰值区间 6.37 到 7.80,并显示八个窗口有 0.962 的重抽落在峰值右侧。第三章在内存里模拟独占新建锁:覆盖写的丢失比例在十六个写者时达到 0.907,独占新建没有一个号被两个窗口同时持有,按序申请或一次只持一把锁即可消除死锁。第四章说明只追加黑板为什么读路径不必取锁,示意参数下最优订阅比例约 0.7。 第五章在 NK 景观上比较四种拓扑:全连通起步最快,到第 80 轮环形以 0.9280 超过全连通的 0.9150;半隔离子群在适度跨组连线时最高,为 0.9259。第六章用超图关联矩阵把涌现写成跨节点复用量与净增益区间,读取延迟缩到一天可消除隔离状态下 0.944 的重复踩坑。第七章考辨孙子「分数」、论语「和而不同」、莱布尼茨「单子无窗」与斯密分工,并落成四件套条款。第八章给出并发耦合度与黑板复用增益试算器、检查器规格表、台账节奏与预注册检验,检验每臂需 156 次记录。书中以标注框区分已发表的证据、可运行的计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Serial Fraction and Coherence Overhead: Why Throughput Has a Peak — Chapter 1 shows where the peak comes from, near 6.78 windows under illustrative parameters. (https://5dlibrary.com/answers/en/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c01/) 第1章 串行比例与相干开销:吞吐为什么有峰值 — 第一章说明峰值从何而来,示意参数下峰值在 6.78 个窗口附近。 (https://5dlibrary.com/answers/zh/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c01/) - Ch.2 Fitting σ and κ from Throughput Records: Least Squares, Bootstrap Intervals and Design — Chapter 2 fits σ and κ by linearised least squares; a stratified bootstrap gives a peak interval from 6.37 to 7.80 and shows that 0.962 of resamples put eight windows past the peak. (https://5dlibrary.com/answers/en/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c02/) 第2章 从吞吐记录拟合 σ 与 κ:最小二乘、自助区间与观测设计 — 第二章用线性化最小二乘拟合 σ 与 κ,分层自助法给出峰值区间 6.37 到 7.80,并显示八个窗口有 0.962 的重抽落在峰值右侧。 (https://5dlibrary.com/answers/zh/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c02/) - Ch.3 Exclusive-Create Locks: Conflict Rates, the Cost of Moving On, and Lost Updates — Chapter 3 simulates exclusive-create locks in memory: overwriting loses 0.907 of updates with sixteen writers, exclusive creation never lets two windows hold one number, and ordered requests or single holding rule out deadlock. (https://5dlibrary.com/answers/en/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c03/) 第3章 独占新建锁:冲突率、顺延成本与丢失更新 — 第三章在内存里模拟独占新建锁:覆盖写的丢失比例在十六个写者时达到 0.907,独占新建没有一个号被两个窗口同时持有,按序申请或一次只持一把锁即可消除死锁。 (https://5dlibrary.com/answers/zh/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c03/) - Ch.4 The Append-Only Blackboard: Lock-Free Reads, Hash De-duplication and the Cost of Subscribing — Chapter 4 explains why an append-only board needs no lock on the read path, with a best subscription share near 0.7 under illustrative parameters. (https://5dlibrary.com/answers/en/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c04/) 第4章 只追加黑板:读路径无锁、哈希去重与订阅成本 — 第四章说明只追加黑板为什么读路径不必取锁,示意参数下最优订阅比例约 0.7。 (https://5dlibrary.com/answers/zh/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c04/) - Ch.5 Coupling Topology and Search: Isolation, Ring, Small World and Full Connection on NK Landscapes — Chapter 5 compares four topologies on NK landscapes: full connection starts fastest, but by round 80 the ring reaches 0.9280 against 0.9150; semi-isolated subgroups peak at 0.9259 with moderate cross-group links. (https://5dlibrary.com/answers/en/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c05/) 第5章 耦合拓扑与探索:NK 景观上的隔离、环形、小世界与全连通 — 第五章在 NK 景观上比较四种拓扑:全连通起步最快,到第 80 轮环形以 0.9280 超过全连通的 0.9150;半隔离子群在适度跨组连线时最高,为 0.9259。 (https://5dlibrary.com/answers/zh/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c05/) - Ch.6 Emergence as Measurable Reuse: Hypergraph Incidence and the Net Gain of Coupling — Chapter 6 writes emergence as cross-node reuse and a net-gain interval on a hypergraph incidence matrix; a reading lag of one day removes 0.944 of the repeated pitfalls seen under isolation. (https://5dlibrary.com/answers/en/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c06/) 第6章 把涌现写成可测复用量:超图关联矩阵与净耦合增益 — 第六章用超图关联矩阵把涌现写成跨节点复用量与净增益区间,读取延迟缩到一天可消除隔离状态下 0.944 的重复踩坑。 (https://5dlibrary.com/answers/zh/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c06/) - Ch.7 Eastern and Western Sources Examined: Sunzi on Division, the Analects on Harmony, Windowless Monads and Smith's Division of Labour — With seven nodes the full mesh has 21 channels; the best arrangement is 4 groups, which brings the count down to 9. With one thousand and twenty-four nodes the full mesh has 523,776 channels; split into 79 groups of 13, the structure needs only 9,210, more than fifty times fewer. Part B gives the approximate rule: the best group size is about $(2n)^{1/3}$, which comes to 5.0 for sixty-four nodes, 8.0 for two hundred and fifty-six, and 12.7 for one thousand and twenty-four, in agreement with the search. (https://5dlibrary.com/answers/en/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c07/) 第7章 东西方原典考辨:孙子「分数」、论语「和而不同」、单子无窗与斯密分工 — 七个节点时,全网状 21 条通道,最好分成 4 组,通道降到 9 条。一千零二十四个节点时,全网状 523,776 条,按每组 13 人分成 79 组,只需 9,210 条,少了五十多倍。B 段给出近似规律:最好的组规模约为 $(2n)^{1/3}$,六十四个节点时 5.0,二百五十六个时 8.0,一千零二十四个时 12.7,与搜索结果吻合。 (https://5dlibrary.com/answers/zh/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c07/) - Ch.8 Field Tools and Checker Specifications: the Concurrency Ledger, Four-Kit Clauses and a Pre-Registered Test — Chapter 8 gives the Concurrency Coupling and Blackboard Reuse Calculator, a checker table, the ledger rhythm and a pre-registered test needing 156 records per arm. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c08/) 第8章 实战工具与检查器规格:并发协同台账、四件套条款与预注册检验 — 第八章给出并发耦合度与黑板复用增益试算器、检查器规格表、台账节奏与预注册检验,检验每臂需 156 次记录。书中以标注框区分已发表的证据、可运行的计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/topological-coupling-concurrent-multi-core-lock-free-blackboard-cross-node-emergent-gains/c08/) ## 5DL-B-412 · Impedance Matching for High-Dimensional Systems: Lossless Downward Compatibility without Core Frequency Degradation 中文书名:高维内核向下兼容阻抗匹配、多层级语义降压转换与高频纯粹度守恒母典 DOI: pending Cite: Fan, Y. (2026). Impedance Matching for High-Dimensional Systems: Lossless Downward Compatibility without Core Frequency Degradation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. Abstract: A popular saying holds that reaching a mass market requires making content shallow, lowering the core to buy reach; its mirror image holds that depth owes no explanation and those who get it will get it. This book refutes the premise the two share, that a message can only be shipped whole or rewritten whole. Borrowing the structure of radio-frequency impedance matching, it treats the core as the source, the reader's momentary cognitive bandwidth as the load, and one-line summaries, direct-answer pages and course lessons as the matching network, and argues that most of the time one should add network rather than alter the source, and that the number of layers has an optimum. The circuit relations serve only as a structural analogy and a parameterised tool, not as a law of human cognition; parameters are illustrative; all calculations are generated by the book's programs; the book is not individual business advice, and its natal-site examples concern only the layered presentation of the natal system, involve no divination and make no medical or bodily claim. Chapters 1 and 2 cover reflection coefficients, standing-wave ratios, multi-section matching and the Bode–Fano limit, and compute where the analogy fails. Chapter 3 compares three strategies per ten thousand illustrative readers: publishing the core alone reaches 1,411; a shallow rewrite reaches 7,906 but produces 949 units of misstatement and leaves no original core; four layers reach 8,864 and deliver 4,170 core units. With a concentrated audience the ratio of layering to rewriting is only 0.97, a correction to the blueprint. Chapter 4 shows that three layers are optimal at baseline. Chapter 5 shows a nine-item routing probe placing 0.679 of readers correctly, and a readability formula giving a passage, its reversal and a shuffled version the same score of 90.1. Chapter 6 restates "lowering the core collapses the brand" as a conditional proposition drawn from the brand-extension literature: rewriting the core leaves belief at 0.206 after twelve periods, while marked but sloppy layers keep belief at 0.994 even though each reader meets 4.80 misstatements, so brand belief is not the standard of compliance. Chapter 7 examines the Analects on "those above the middle", the Great Commentary's "the easy is easily known", Plato's Phaedrus and Aristotle's Rhetoric; in an illustrative model a popular audience follows best, with probability 0.172, when six steps are stated; the chapter ends with a distillation spec for the four kits. Chapter 8 gives the field tools and checker specifications: the Audience Reflection Coefficient and Layered Interface Matching Calculator, criteria grouped by the four kits, a two-level content ledger and pre-registered tests, with 1,209 readers per arm for the upgrade-rate test. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. 摘要:流行说法认为,要做大众市场就必须把内容做浅、降低内核换取覆盖;反过来也有人认为,高深无须解释,懂的人自然懂。本书证伪两者共同的前提:信号只能整份送出或整份改写。借射频阻抗匹配的结构,本书把内核当作信号源、把读者此刻的认知带宽当作负载、把一句话摘要、直答页与学堂课时当作匹配网络,主张多数时候应加网络而不是改源,且层数存在最优值。射频关系只作结构类比与参数化工具,不是人类认知的定律;书中参数为示意;全部计算由书中程序生成;本书是非个人化经营建议,命理站示例只涉及本命体系的分层呈现,不涉及占卜,不作任何医疗或身体功效宣称。 第一、二章讲反射系数、驻波比、多节匹配与 Bode–Fano 极限,并算出类比在哪里失效。第三章在每一万名示意读者上比较三种策略:只发内核送达 1,411 人;改写浅版送达 7,906 人,却有 949 个错误陈述单位且原版内核不复存在;四层接口送达 8,864 人、送出 4,170 个内核单位。受众集中时分层与改写之比仅 0.97,这是一项蓝图修正。第四章说明基线参数下最优为三层。第五章说明九题分流小测分对 0.679,可读性公式给一段话、其反义改写与乱序版本同样的 90.1 分。第六章按品牌延伸文献把「降内核导致品牌坍塌」改写为条件命题:改写内核十二期后信念降至 0.206,而标层但粗糙的接口信念仍有 0.994,每位读者却读到 4.80 条错误陈述,所以品牌信念不是合规标准。 第七章考辨《论语·雍也》「中人以上」、《周易·系辞上》「易则易知」、柏拉图《斐德罗篇》与亚里士多德《修辞学》,示意模型里大众听众写出六步时最易跟上,概率 0.172;章末给出四件套蒸馏规格。第八章给出实战工具与检查器规格:蒸馏小工具「受众阻抗反射系数与分层接口匹配试算器」、按四件套分组的判据表、双层内容台账与预注册检验,升层率检验每组需 1,209 名读者。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Reflection Coefficient: a Structural Analogy for Audience Mismatch — In the first four rows the source is 50 ohms. With a 100-ohm load the mismatch delivers 0.889; the turns ratio is 1.414, the quarter-wave section is 70.7 ohms, and after matching the load receives 1.000. With a 200-ohm load the mismatch delivers 0.640, the turns ratio is 2.000 and the quarter-wave section is 100.0 ohms. With an 800-ohm load the mismatch delivers only 0.221; the turns ratio is 4.000 and the quarter-wave section 200.0 ohms. With a 12.5-ohm load the turns ratio is 0.500 and the quarter-wave section 25.0 ohms. In all four cases matching brings delivery to 1.000 without altering a single word of the source. (https://5dlibrary.com/answers/en/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c01/) 第1章 反射系数:受众失配的结构类比 — 表的前四行,源阻抗都是 50 欧姆。负载 100 欧姆时,不匹配送达 0.889,匝比 1.414,四分之一波长段 70.7 欧姆,匹配后送达 1.000。负载 200 欧姆时,不匹配送达 0.640,匝比 2.000,四分之一波长段 100.0 欧姆。负载 800 欧姆时,不匹配只送达 0.221,匝比 4.000,四分之一波长段 200.0 欧姆。负载 12.5 欧姆时,匝比 0.500,四分之一波长段 25.0 欧姆。四种负载在匹配后都送达 1.000,信号源一个字也没改。 (https://5dlibrary.com/answers/zh/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c01/) - Ch.2 Multi-Section Matching and the Bode–Fano Limit: Bandwidth against Depth of Match — In the nominal design the worst reflection falls steadily with the number of sections: 0.1725 with one, 0.0408 with two, 0.0107 with three, 0.0028 with four and 0.0010 with five. With an error limit of 0.02, three sections are the most robust, failing the threshold in only 0.007 of builds; four and five sections do worse, failing in 0.033 and 0.050, and the mean worst reflection climbs back from 0.021 to 0.026. With an error limit of 0.08 the pattern is starker. Two sections fail in 0.695 of builds, three in 0.655, four in 0.752 and five in 0.843, while the mean worst reflection rises from 0.073 to 0.103. (https://5dlibrary.com/answers/en/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c02/) 第2章 多节匹配与 Bode–Fano 极限:带宽与匹配深度的权衡 — 名义设计里,最差反射随节数一路下降:一节 0.1725,两节 0.0408,三节 0.0107,四节 0.0028,五节 0.0010。误差上限为 0.02 时,三节最稳,超标比例只有 0.007;四节、五节反而升到 0.033 与 0.050,平均最差反射也从 0.021 回升到 0.026。误差上限为 0.08 时,情况更明显:两节超标 0.695,三节 0.655,四节 0.752,五节 0.843,平均最差反射从 0.073 一路升到 0.103。 (https://5dlibrary.com/answers/zh/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c02/) - Ch.3 The Distribution of Reader Bandwidth: One Deep Text, One Rewritten Text, or Layers — Chapter 3 compares three strategies per ten thousand illustrative readers: publishing the core alone reaches 1,411; a shallow rewrite reaches 7,906 but produces 949 units of misstatement and leaves no original core; four layers reach 8,864 and deliver 4,170 core units. With a concentrated audience the ratio of layering to rewriting is only 0.97, a correction to the blueprint. (https://5dlibrary.com/answers/en/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c03/) 第3章 读者带宽分布:单层高深、单层降级与分层接口 — 第三章在每一万名示意读者上比较三种策略:只发内核送达 1,411 人;改写浅版送达 7,906 人,却有 949 个错误陈述单位且原版内核不复存在;四层接口送达 8,864 人、送出 4,170 个内核单位。受众集中时分层与改写之比仅 0.97,这是一项蓝图修正。 (https://5dlibrary.com/answers/zh/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c03/) - Ch.4 The Optimal Number of Layers: Gains in Reach, Cost of Upkeep and Distortion between Layers — With the core alone, 1,411 readers are reached and the net value is 1,411. With two layers, 3,427 readers are reached, 2,551 correct units are delivered against 67 wrong ones, and the net value is 2,166. With three layers, reach is 6,065, correct units 3,260, wrong units 145, and the net value 2,470, the highest in the table. With four layers, reach climbs to 8,269 and correct units to 3,379, wrong units rise to 165, and the net value falls to 2,298. Five layers reach 9,453 readers for a net value of 2,096; six reach 9,874 for 1,849; seven reach almost everyone, 9,977 readers, and the net value is down to 1,567. (https://5dlibrary.com/answers/en/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c04/) 第4章 最优层数:覆盖增益、维护成本与层间失真 — 只有内核一层时,送达 1,411 人,净值 1,411。加到两层,送达 3,427 人,送出正确单位 2,551,错误 67,净值 2,166。三层时送达 6,065 人,正确单位 3,260,错误 145,净值 2,470,是全表最高。四层时送达升到 8,269 人,正确单位 3,379,错误 165,净值却降到 2,298。五层时送达 9,453 人,净值 2,096;六层时送达 9,874 人,净值 1,849;七层时几乎人人送达,9,977 人,净值只剩 1,567。 (https://5dlibrary.com/answers/zh/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c04/) - Ch.5 Expertise Reversal, Routing Readers and the Limits of Readability — Chapter 5 shows a nine-item routing probe placing 0.679 of readers correctly, and a readability formula giving a passage, its reversal and a shuffled version the same score of 90.1. (https://5dlibrary.com/answers/en/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c05/) 第5章 专长逆转、读者分流与可读性的边界 — 第五章说明九题分流小测分对 0.679,可读性公式给一段话、其反义改写与乱序版本同样的 90.1 分。 (https://5dlibrary.com/answers/zh/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c05/) - Ch.6 Bayesian Updating of Brand Belief: When Dilution Happens — Chapter 6 restates "lowering the core collapses the brand" as a conditional proposition drawn from the brand-extension literature: rewriting the core leaves belief at 0.206 after twelve periods, while marked but sloppy layers keep belief at 0.994 even though each reader meets 4.80 misstatements, so brand belief is not the standard of compliance. (https://5dlibrary.com/answers/en/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c06/) 第6章 品牌信念的贝叶斯更新:稀释在什么条件下发生 — 第六章按品牌延伸文献把「降内核导致品牌坍塌」改写为条件命题:改写内核十二期后信念降至 0.206,而标层但粗糙的接口信念仍有 0.994,每位读者却读到 4.80 条错误陈述,所以品牌信念不是合规标准。 (https://5dlibrary.com/answers/zh/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c06/) - Ch.7 Classical Sources: "Those above the Middle", Ease and Simplicity, the Phaedrus and the Rhetoric — Chapter 7 examines the Analects on "those above the middle", the Great Commentary's "the easy is easily known", Plato's Phaedrus and Aristotle's Rhetoric; in an illustrative model a popular audience follows best, with probability 0.172, when six steps are stated; the chapter ends with a distillation spec for the four kits. (https://5dlibrary.com/answers/en/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c07/) 第7章 原典考辨:「中人以上」、「易简」、斐德罗与修辞学 — 第七章考辨《论语·雍也》「中人以上」、《周易·系辞上》「易则易知」、柏拉图《斐德罗篇》与亚里士多德《修辞学》,示意模型里大众听众写出六步时最易跟上,概率 0.172;章末给出四件套蒸馏规格。 (https://5dlibrary.com/answers/zh/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c07/) - Ch.8 The Audience Reflection Coefficient and Layered Interface Matching Calculator: Field Tools and Checker Specifications — Chapter 8 gives the field tools and checker specifications: the Audience Reflection Coefficient and Layered Interface Matching Calculator, criteria grouped by the four kits, a two-level content ledger and pre-registered tests, with 1,209 readers per arm for the upgrade-rate test. Marked boxes separate published evidence, runnable Python calculations (23 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c08/) 第8章 受众阻抗反射系数与分层接口匹配试算器:实战工具与检查器规格 — 第八章给出实战工具与检查器规格:蒸馏小工具「受众阻抗反射系数与分层接口匹配试算器」、按四件套分组的判据表、双层内容台账与预注册检验,升层率检验每组需 1,209 名读者。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/impedance-matching-high-dimensional-systems-lossless-downward-compatibility-core-frequency/c08/) ## 5DL-B-413 · Critical Ignition Conditions for Business Flywheels, Loop-Gain Friction Accounting and Cross-Vertical Fractal Replication 中文书名:商业飞轮自激临界条件、环路增益损耗核算与跨赛道分形复制母典 DOI: pending Cite: Fan, Y. (2026). Critical Ignition Conditions for Business Flywheels, Loop-Gain Friction Accounting and Cross-Vertical Fractal Replication (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative and all calculations come from its programs. It is not individual business advice, promises no revenue, growth or payback, and discusses only single-tier referral rewards paid to the direct referrer. Abstract: A popular saying holds that a flywheel, once turning, cannot be stopped: if the product of conversion rates and multipliers around the loop exceeds one, the business runs on its own, and once it works it can be replicated across verticals without limit. This book rewrites the saying as testable claims. After checking the sources, it corrects its blueprint (a gain above one if and only if the loop runs on its own; unlimited replication by the minimal unit): a gain above one is only necessary; a supercritical branching process still stalls with positive probability; positive feedback holds only before saturation; and replication is a loss account multiplied ring by ring. Its parameters are illustrative and all calculations come from its programs. It is not individual business advice, promises no revenue, growth or payback, and discusses only single-tier referral rewards paid to the direct referrer. Chapter 1 accounts for the loop gain with fractions and turns the product into a logarithmic friction account in which every ring has an elasticity of one. Chapter 2 gives ring-by-ring leakage and Pareto concentration from a synthetic quarter of flows and friction events. Chapter 3 computes ignition probabilities with the Galton-Watson process: with Poisson referrals and a gain of 1.125 a single seed stalls with probability 0.7863 and a nine-in-ten chance needs 10 independent seed groups; with concentrated referrals (k of 0.5) the stall probability rises to 0.9250 and 30 groups are needed. Chapter 4 uses the Jury test to show that with data four weeks old the weekly correction can be at most 0.3473. Chapter 5 shows that a flywheel with a gain of 1.125 pushes itself only below a penetration of 0.1111 and ends at 0.2249. Chapter 6 keeps the loss account of replication: if each of five rings keeps nine tenths, the mother must exceed 1.6935 before a copy can push itself. Chapter 7 examines the Sunzi on momentum, chapter 42 of the Laozi and Maxwell's "On Governors" of 1868, and distils clauses for the four kits. Chapter 8 gives the flywheel ledger, the distilled Flywheel Loop-Gain and Ignition-Probability Calculator, checker specifications for the four kits and pre-registered tests: with a true gain of 1.125, 5,643 buyers must complete a turn before the lower bound clears one with a four-in-five chance. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:商业圈流行「飞轮转起来就停不下来」:环路上转化率与放大倍数的乘积过了一,就能脱手自转,跑通后还能跨赛道无限复制。本书把这句话改写成可检验的命题。核对文献之后,立项蓝图「当且仅当增益大于一即可自转、按最小单元即可无限复制」被修正为:增益大于一只是必要条件;超临界分支过程仍有正的熄火概率;正反馈只在饱和之前成立;复制是逐环相乘的折损账。书中参数为示意;全部计算由书中程序生成;本书是非个人化经营建议,不承诺任何收益、增长或回本,推荐激励只讨论单层、只给直接推荐人的安排。 第一章用分数核算环路增益,把乘积变成对数摩擦账,说明每一环的弹性都是一。第二章用合成季度流量与摩擦事件给出逐环漏损与帕累托集中。第三章用 Galton–Watson 分支过程计算点火概率:增益 1.125 的泊松推荐下单个种子熄火概率为 0.7863,九成把握需 10 组独立种子;推荐高度集中(k 为 0.5)时熄火概率升到 0.9250,需 30 组。第四章用 Jury 判据说明延迟 4 周时每周纠偏力度上限只有 0.3473。第五章说明增益 1.125 的飞轮只在渗透率低于 0.1111 时自推,最终覆盖 0.2249。 第六章核算复制折损:五环各保留九成时母飞轮增益须高于 1.6935,复制品才可能自推。第七章考辨《孙子兵法·势篇》、《老子》第四十二章与 Maxwell 1868 年的《论调速器》,并给出四件套蒸馏规格。第八章给出飞轮环路台账、蒸馏小工具「飞轮环路增益与点火概率核算器」、四件套检查器规格与预注册检验:真实增益 1.125 时须有 5,643 个买家走完一整圈,区间下界才有八成把握越过一。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Loop Gain: the Product, the Logarithmic Friction Account and the Weakest Ring — With 100 buyers per turn, the delta interval runs from 0.606 to 2.089 and the bootstrap interval from 0.480 to 1.840, and the estimate falls below one with probability 0.405. A flywheel whose true gain is 1.125 therefore looks as if it were shrinking in four quarters out of ten when the sample is this small. With 400 buyers per turn the interval narrows to 0.826 to 1.533 (0.797 to 1.478 by the bootstrap), but it still contains one, and the chance of an estimate below one is 0.234. Only at 1,600 buyers per turn does the lower bound approach one (0.964, or 0.961 by the bootstrap), and only at 6,400 buyers does it reach 1.041 and clear one with room to spare. (https://5dlibrary.com/answers/en/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c01/) 第1章 环路增益:乘积、对数摩擦账与最弱环 — 每圈 100 个买家时,delta 区间是 0.606 到 2.089,自助区间是 0.480 到 1.840,估计值低于一的概率是 0.405。也就是说,一个真实增益 1.125 的飞轮,在样本这么小的时候,有四成的季度看起来像在萎缩。每圈 400 个买家,区间收窄到 0.826 至 1.533(自助法 0.797 至 1.478),仍然盖住一,低于一的概率 0.234。要到每圈 1,600 个买家,区间下界才接近一(0.964,自助法 0.961);到 6,400 个买家,下界 1.041,才算稳稳越过。 (https://5dlibrary.com/answers/zh/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c01/) - Ch.2 The Friction Funnel: Ring-by-Ring Leakage in One Quarter, Pareto Concentration and Friction in Time — Together the 640 events destroy 380.7 loop units. First comes the snag at the purchase ring: 180 events, a loss of 108.0 and a cumulative share of 0.284. Second come refunds at the purchase ring: 72 events, a loss of 72.0, cumulative share 0.473. Third is churn at the feedback ring, with a loss of 61.6 and cumulative share 0.635. Fourth is manual handling at the ring from feedback to answer pages, with as many as 144 events but a loss of only 43.2 and a cumulative share of 0.748. Fifth is the snag at the ring from visits to the monograph page, with a loss of 19.8 and a cumulative share of 0.800. Of twelve classes of event, the top five account for four fifths of the loss. (https://5dlibrary.com/answers/en/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c02/) 第2章 摩擦漏斗:一个季度的逐环漏损、帕累托集中与时间摩擦 — 640 起事件合计损失 380.7 个环路单位。排第一的是购买环的卡点,180 起,损失 108.0,累计份额 0.284;第二是购买环的退款,72 起,损失 72.0,累计 0.473;第三是反馈环的流失,损失 61.6,累计 0.635;第四是反馈到直答页一环的人工介入,事件多达 144 起,损失却只有 43.2,累计 0.748;第五是访问到专著页的卡点,损失 19.8,累计 0.800。十二类事件中,前五类就占了八成损失。 (https://5dlibrary.com/answers/zh/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c02/) - Ch.3 Ignition Probability: Why a Supercritical Flywheel Can Still Stall — Chapter 3 computes ignition probabilities with the Galton-Watson process: with Poisson referrals and a gain of 1.125 a single seed stalls with probability 0.7863 and a nine-in-ten chance needs 10 independent seed groups; with concentrated referrals (k of 0.5) the stall probability rises to 0.9250 and 30 groups are needed. (https://5dlibrary.com/answers/en/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c03/) 第3章 点火概率:超临界飞轮为什么仍会熄火 — 第三章用 Galton–Watson 分支过程计算点火概率:增益 1.125 的泊松推荐下单个种子熄火概率为 0.7863,九成把握需 10 组独立种子;推荐高度集中(k 为 0.5)时熄火概率升到 0.9250,需 30 组。 (https://5dlibrary.com/answers/zh/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c03/) - Ch.4 Delay and Oscillation: a Gain above One Is Not Stability — Chapter 4 uses the Jury test to show that with data four weeks old the weekly correction can be at most 0.3473. (https://5dlibrary.com/answers/en/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c04/) 第4章 延迟与振荡:增益大于一不等于稳定 — 第四章用 Jury 判据说明延迟 4 周时每周纠偏力度上限只有 0.3473。 (https://5dlibrary.com/answers/zh/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c04/) - Ch.5 The Saturation Window: between the Ignition Threshold and the Ceiling — Chapter 5 shows that a flywheel with a gain of 1.125 pushes itself only below a penetration of 0.1111 and ends at 0.2249. (https://5dlibrary.com/answers/en/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c05/) 第5章 饱和窗口:点火阈值与饱和上限之间 — 第五章说明增益 1.125 的飞轮只在渗透率低于 0.1111 时自推,最终覆盖 0.2249。 (https://5dlibrary.com/answers/zh/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c05/) - Ch.6 Cross-Vertical Fractal Replication: Isomorphism Scores, Transfer Loss and the Minimal Isomorphic Unit — Chapter 6 keeps the loss account of replication: if each of five rings keeps nine tenths, the mother must exceed 1.6935 before a copy can push itself. (https://5dlibrary.com/answers/en/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c06/) 第6章 跨赛道分形复制:同构评分、迁移折损与最小同构单元 — 第六章核算复制折损:五环各保留九成时母飞轮增益须高于 1.6935,复制品才可能自推。 (https://5dlibrary.com/answers/zh/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c06/) - Ch.7 Eastern and Western Sources Examined: Rushing Water, Three Begetting the Myriad, and Maxwell's "On Governors" — Chapter 7 examines the Sunzi on momentum, chapter 42 of the Laozi and Maxwell's "On Governors" of 1868, and distils clauses for the four kits. (https://5dlibrary.com/answers/en/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c07/) 第7章 东西方原典考辨:激水漂石、三生万物与 Maxwell《论调速器》 — 第七章考辨《孙子兵法·势篇》、《老子》第四十二章与 Maxwell 1868 年的《论调速器》,并给出四件套蒸馏规格。 (https://5dlibrary.com/answers/zh/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c07/) - Ch.8 Closing the Empirical Loop: the Flywheel Ledger, Field Tools and Checker Specifications — Chapter 8 gives the flywheel ledger, the distilled Flywheel Loop-Gain and Ignition-Probability Calculator, checker specifications for the four kits and pre-registered tests: with a true gain of 1.125, 5,643 buyers must complete a turn before the lower bound clears one with a four-in-five chance. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c08/) 第8章 实证闭环:飞轮环路台账、实战工具与检查器规格 — 第八章给出飞轮环路台账、蒸馏小工具「飞轮环路增益与点火概率核算器」、四件套检查器规格与预注册检验:真实增益 1.125 时须有 5,643 个买家走完一整圈,区间下界才有八成把握越过一。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/business-flywheel-critical-ignition-loop-gain-friction-accounting-fractal-replication/c08/) ## 5DL-B-414 · Formalising Fuzzy Founder Wishes: Constraint Inversion, Falsifiable Spec Synthesis and Executable Revenue System Compilation 中文书名:创始人模糊愿望形式化编译、约束反演与可执行商业系统生成母典 DOI: pending Cite: Fan, Y. (2026). Formalising Fuzzy Founder Wishes: Constraint Inversion, Falsifiable Spec Synthesis and Executable Revenue System Compilation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Amounts are in illustrative units with no currency, parameters are illustrative, and all calculations are generated by the book's programs. The book is non-personalised business guidance, not personal financial or investment advice; the psychological literature is reported only as associations with their conditions, and the lower end of an interval is never a promise of income. Abstract: A popular belief among founders holds that the more clearly a vision is written and the more vividly it is imagined, the more likely it is to come true; the commissioning blueprint went further and held that one emotional sentence could be compiled, with zero loss and in four steps, into an asset that passes acceptance tests. This book refutes the first claim and corrects the second. Positive fantasy goes with less subsequent effort, while mental contrasting and implementation intentions go with action, so the first step of compilation is to write obstacles and counterexamples, not the vision; and formalisation must always rule between readings, so "zero loss" is rewritten as an auditable ambiguity loss. Amounts are in illustrative units with no currency, parameters are illustrative, and all calculations are generated by the book's programs. The book is non-personalised business guidance, not personal financial or investment advice; the psychological literature is reported only as associations with their conditions, and the lower end of an interval is never a promise of income. Chapter 1 shows in a toy model that, with the same average effort, mental contrasting raises successes per hundred units of effort from 50.7 to 65.8. Chapter 2 breaks wishes into five slots: six fictional wishes have 99 readings, ruling seven vague words books 10.34 bits of loss, and a panel of three readers exposes all the ambiguity in one round with probability only 0.121. Chapter 3 back-substitutes through the revenue identity: a net of 3,000 needs at least 201 orders and 10,048 visits, and elimination under capacity and budget constraints bounds the net at 2,412.0, so the wish contradicts itself. Chapter 4 shows that a plan that reaches the target exactly at the means does so with probability only 0.427. Chapter 5's property tester finds a counterexample to a discount rule on the 4th random case, and delta debugging reduces a contradiction among twelve constraints to 3 in 33 calls. Chapter 6 chooses the minimum acceptable prototype with the Kano classification and a knapsack. Chapter 7 examines "prepared, it stands" from the Doctrine of the Mean, the man of Zheng who bought shoes in the Han Feizi, and Descartes' second rule; buying shoes by the measurement fits only 0.585 of the time; and it gives the distillation spec for the four kits. Chapter 8 gives the distilled Wish Constraint Inversion and Counterexample Compiler, a table of checker specifications grouped by the four kits, the wish ledger and a pre-registered test whose first verdict needs 44 wishes. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:创业圈流行一种说法:愿景写得越清楚、想象得越鲜活,事情越容易成;立项蓝图更进一步,认为一句感性愿景可以经四步零损耗编译为可验收资产。本书证伪前一半,修正后一半。正面幻想与较低的后续投入相关,心理对照与执行意图才与行动相关,所以编译的第一步是写障碍与反例,不是写愿景;形式化必然要在多种读法之间裁决,「零损耗」改写为可审计的歧义损耗。书中金额为无币种示意单位,参数为示意,全部计算由书中程序生成;本书是非个人化经营建议,不作个人财务或投资建议,心理学文献只写关联与适用条件,区间下界不是任何收入承诺。 第一章用玩具模型说明,同样的平均投入下,心理对照让每百单位投入的成功数从 50.7 升到 65.8。第二章把愿望拆成五槽,六句虚构愿望共 99 种读法,七个含混词全部裁决记 10.34 比特损耗,三人小组一次暴露全部歧义的概率只有 0.121。第三章沿收入恒等式回代,净额 3,000 需至少 201 单、10,048 次访问;消元在产能与预算约束下求得净额上界 2,412.0,判定该愿望自相矛盾。第四章说明按均值恰好达标的计划,达标概率只有 0.427。 第五章的性质测试在第 4 个随机用例找到折扣规则的反例,增量调试用 33 次判定把十二条约束的矛盾缩到 3 条。第六章用 Kano 分类与背包动态规划选最小可验收原型。第七章考辨《礼记·中庸》「豫则立」、《韩非子》郑人买履与笛卡尔第二规则,按尺码买鞋的合脚率只有 0.585,并给出四件套蒸馏规格。第八章给出蒸馏小工具「愿望约束反演与反例编译器」、按四件套分组的检查器规格表、愿望台账与预注册检验:44 条愿望即可做第一次判决。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Obstacles First: Positive Fantasy, Mental Contrasting and the First Step of Compilation — Chapter 1 shows in a toy model that, with the same average effort, mental contrasting raises successes per hundred units of effort from 50.7 to 65.8. (https://5dlibrary.com/answers/en/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c01/) 第1章 先写障碍:正面幻想、心理对照与编译的第一步 — 第一章用玩具模型说明,同样的平均投入下,心理对照让每百单位投入的成功数从 50.7 升到 65.8。 (https://5dlibrary.com/answers/zh/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c01/) - Ch.2 Five Slots: Counting Readings, Recording Rulings and Separating the Optative from the Indicative — Chapter 2 breaks wishes into five slots: six fictional wishes have 99 readings, ruling seven vague words books 10.34 bits of loss, and a panel of three readers exposes all the ambiguity in one round with probability only 0.121. (https://5dlibrary.com/answers/en/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c02/) 第2章 五槽拆解:歧义计数、裁决记录与祈使陈述之分 — 第二章把愿望拆成五槽,六句虚构愿望共 99 种读法,七个含混词全部裁决记 10.34 比特损耗,三人小组一次暴露全部歧义的概率只有 0.121。 (https://5dlibrary.com/answers/zh/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c02/) - Ch.3 Constraint Inversion: Weakest Preconditions, Elimination and Arc Consistency — Chapter 3 back-substitutes through the revenue identity: a net of 3,000 needs at least 201 orders and 10,048 visits, and elimination under capacity and budget constraints bounds the net at 2,412.0, so the wish contradicts itself. (https://5dlibrary.com/answers/en/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c03/) 第3章 约束反演:最弱前置条件、消元与弧相容 — 第三章沿收入恒等式回代,净额 3,000 需至少 201 单、10,048 次访问;消元在产能与预算约束下求得净额上界 2,412.0,判定该愿望自相矛盾。 (https://5dlibrary.com/answers/zh/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c03/) - Ch.4 Break-Even Floors and the Probability of Reaching a Target: Intervals, Not Points — Chapter 4 shows that a plan that reaches the target exactly at the means does so with probability only 0.427. (https://5dlibrary.com/answers/en/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c04/) 第4章 盈亏下界与达成概率:区间而不是点 — 第四章说明按均值恰好达标的计划,达标概率只有 0.427。 (https://5dlibrary.com/answers/zh/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c04/) - Ch.5 Counterexamples First: Property-Based Testing, Shrinking and the Minimal Contradictory Subset — Chapter 5's property tester finds a counterexample to a discount rule on the 4th random case, and delta debugging reduces a contradiction among twelve constraints to 3 in 33 calls. (https://5dlibrary.com/answers/en/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c05/) 第5章 反例先行:性质测试、收缩与最小矛盾子集 — 第五章的性质测试在第 4 个随机用例找到折扣规则的反例,增量调试用 33 次判定把十二条约束的矛盾缩到 3 条。 (https://5dlibrary.com/answers/zh/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c05/) - Ch.6 The Minimum Runnable Prototype: Kano Classes and the Knapsack — With 15 days left in the budget, knapsack dynamic programming chooses a combination worth 37, one point more than the greedy method. The budget curve shows the marginal value falling to 1.00 a day after 28 days, which sets the upper limit of the minimum prototype. Among 40 Kano questionnaires, the reviews feature, set up with one-dimensional quality in the lead, was classified as indifferent, which shows that a classification must state its lead and its sample size. "The later, the dearer" is not a law; the reason for small prototypes is that they let the business learn early. (https://5dlibrary.com/answers/en/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c06/) 第6章 最小可运行原型:Kano 分类与背包取舍 — 背包动态规划在 15 天剩余预算里选出价值 37 的组合,比贪心多一分;预算曲线在 28 天后边际价值跌到每天 1.00,最小原型的上限由此算出。Kano 问卷 40 份答卷里,设定为一元居首的书评功能被判成了无差异,说明分类本身要写明票差与样本量。「越晚改越贵」不当定律,原型做小的理由是学得早。 (https://5dlibrary.com/answers/zh/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c06/) - Ch.7 The Classics Examined: Preparedness, the Man of Zheng and Descartes' Second Rule — Chapter 7 examines "prepared, it stands" from the Doctrine of the Mean, the man of Zheng who bought shoes in the Han Feizi, and Descartes' second rule; buying shoes by the measurement fits only 0.585 of the time; and it gives the distillation spec for the four kits. (https://5dlibrary.com/answers/en/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c07/) 第7章 原典考辨:豫则立、郑人买履与笛卡尔第二规则 — 第七章考辨《礼记·中庸》「豫则立」、《韩非子》郑人买履与笛卡尔第二规则,按尺码买鞋的合脚率只有 0.585,并给出四件套蒸馏规格。 (https://5dlibrary.com/answers/zh/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c07/) - Ch.8 Field Tools and Checker Specifications: The Wish Ledger, Red Lines and a Pre-Registered Test — Chapter 8 gives the distilled Wish Constraint Inversion and Counterexample Compiler, a table of checker specifications grouped by the four kits, the wish ledger and a pre-registered test whose first verdict needs 44 wishes. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c08/) 第8章 实战工具与检查器规格:愿望台账、红线与预注册检验 — 第八章给出蒸馏小工具「愿望约束反演与反例编译器」、按四件套分组的检查器规格表、愿望台账与预注册检验:44 条愿望即可做第一次判决。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/formalising-fuzzy-founder-wishes-constraint-inversion-falsifiable-spec-synthesis/c08/) ## 5DL-B-415 · Seven-Day Cashflow Ignition from Large Digital Asset Catalogs: Firepower Concentration Ratios and Saturation Monetisation Sprints 中文书名:重装数字资产库首周现金流点火、火力收敛系数与饱和变现突击母典 DOI: pending Cite: Fan, Y. (2026). Seven-Day Cashflow Ignition from Large Digital Asset Catalogs: Firepower Concentration Ratios and Saturation Monetisation Sprints (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative and prices are in illustrative units without currency; all calculations are generated by its programs; it is not individual business advice and promises no income. Chapter 2 defines the firepower concentration ratio; in an illustrative catalog pushing only the top three earns 0.805 of the optimum. Abstract: Owners who hold hundreds of backlist monographs, a few software tools and some paid add-ons often hear a popular saying: spreading forces makes returns shrink by Lanchester's square law, so pick three knives, concentrate for seven days and break through the daily cash target. This book refutes the causal chain inside that saying. The square law needs aimed fire, real attrition and a zero-sum ending, and marketing has none of them; advertising response is mostly concave. Any benefit of concentration comes only from three mechanisms: fixed costs, switching costs and a minimum effective exposure threshold. After checking the literature, the book corrects "shrinking with the square" into testable parameter boundaries and rewrites "break through in seven days" as achievement probabilities and quantiles. Its parameters are illustrative and prices are in illustrative units without currency; all calculations are generated by its programs; it is not individual business advice and promises no income. Chapter 1 gives numerical and closed-form solutions of Lanchester's equations and shows that in an attraction model of share concentration loses to an even split even under steep response. Chapter 2 defines the firepower concentration ratio; in an illustrative catalog pushing only the top three earns 0.805 of the optimum. Chapter 3 uses eight combinations of the mechanisms: with all three absent, three or fewer assets is optimal in a share of 0.000 of catalogs, and forcing the top three earns 0.793 of the optimum. Chapter 4 takes stock of three hundred synthetic assets, picks knives by exact search under a preparation-time cap and sets promotion intensity as elasticity times margin. Chapter 5 shows that committing after day one picks the best knife with probability 0.550. Chapter 6 schedules by dynamic programming and simulates achievement: reaching the daily target on all seven days has probability 0.021, on at least three days 0.605. Chapter 7 examines the passages on concentration in Sunzi's Xu Shi chapter and Clausewitz's On War and their limits. Chapter 8 gives the runnable Catalog Firepower Concentration and Seven-Day Cash Scheduler, checker specifications grouped by the four kits, the sprint ledger and pre-registered tests, the first needing 3,826 visitors per group. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:手握几百卷存量专著、若干软件工具与付费加购的经营者,常听到一句流行说法:兵力分散会让收益按兰彻斯特平方律折损,挑出三把尖刀集中七天,就能打穿单日现金流目标。本书证伪其中的因果链:平方律要求瞄准射击、真实损耗与零和结局,营销里一个都不满足;广告响应多为凹形。集中火力的好处只来自固定成本、切换成本与最低有效曝光门槛三种机制。按立项蓝图核对文献后,本书把「按平方折损」修正为可检验的参数边界,把「七天打穿」改写为达成概率与分位数。书中参数为示意,价格用无币种示意单位;全部计算由书中程序生成;本书是非个人化经营建议,不承诺任何收入。 第一章写出兰彻斯特方程的数值解与解析解,说明在吸引力份额模型里即使响应很陡,集中仍不如分散。第二章定义火力收敛系数,示意书目里只推前三件仅得最优的 0.805。第三章用八种机制组合说明,三者全无时三件以内为最优的比例为 0.000,强推前三件只得最优的 0.793。第四章盘点三百件合成存量,在准备工时约束下精确挑刀,并按推广弹性乘毛利定推广强度。第五章说明七天窗口里第一天后全押只有 0.550 的机会押对。第六章用动态规划排程、用蒙特卡洛给出达成概率:七天每天过日目标的概率仅 0.021,至少三天过线为 0.605。第七章考辨《孙子兵法·虚实篇》与克劳塞维茨《战争论》关于集中兵力的原文及其边界。第八章给出可运行的「存量资产火力收敛与七日现金流排程器」、按四件套分组的检查器规格、七日突击台账与预注册检验,检验一每组需 3,826 位访客。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Why the Square Law Does Not Move into Marketing — Section A is the first key table of the book. With p at 2.0, an enemy who stays united leaves me 60.00; split into two groups, it leaves me 82.46; into four, 91.65; into eight, 95.92. With p at 1.0, however the enemy divides, I keep 20.00 and not a unit more. Section B isolates the gain from facing the enemy in two groups rather than one: 22.46 units when p is 2.00, 19.24 when p is 1.50, 12.24 when p is 1.25, and nothing at all when p is 1.00. (https://5dlibrary.com/answers/en/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c01/) 第1章 平方律为什么搬不进营销 — A 段是全书第一张关键表。p 等于 2.0 时,对手合兵一处,我方剩 60.00;对手分两股,我方剩 82.46;分四股剩 91.65;分八股剩 95.92。p 等于 1.0 时,无论对手怎样分兵,我方都只剩 20.00。B 段把分成两股带来的好处单列出来:p 为 2.00 时多保住 22.46 个单位,p 为 1.50 时多保住 19.24,p 为 1.25 时多保住 12.24,p 为 1.00 时一点好处都没有。 (https://5dlibrary.com/answers/zh/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c01/) - Ch.2 Four Response Shapes and the Firepower Concentration Ratio — Chapter 2 defines the firepower concentration ratio; in an illustrative catalog pushing only the top three earns 0.805 of the optimum. (https://5dlibrary.com/answers/en/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c02/) 第2章 响应曲线的四种形状与火力收敛系数 — 第二章定义火力收敛系数,示意书目里只推前三件仅得最优的 0.805。 (https://5dlibrary.com/answers/zh/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c02/) - Ch.3 Three Sources of a Return to Concentration: Fixed Costs, Switching Costs and Exposure Thresholds — Chapter 3 uses eight combinations of the mechanisms: with all three absent, three or fewer assets is optimal in a share of 0.000 of catalogs, and forcing the top three earns 0.793 of the optimum. (https://5dlibrary.com/answers/en/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c03/) 第3章 集中火力的三个来源:固定成本、切换成本与曝光门槛 — 第三章用八种机制组合说明,三者全无时三件以内为最优的比例为 0.000,强推前三件只得最优的 0.793。 (https://5dlibrary.com/answers/zh/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c03/) - Ch.4 Taking Stock of Three Hundred Assets and Choosing the Knives — Section A: at an elasticity of 0.10, the tool with a margin of 0.955 should spend 0.0955 of its revenue on promotion, the book with a margin of 0.868 should spend 0.0868, and the KDP edition with a margin of 0.400 only 0.0400. Section B converts these ratios into weekly budgets: at the illustrative revenues, the three knives receive 74.04, 64.32 and 63.51, and the KDP edition only 12.00. Section C: with equal revenue and elasticity, the promotion intensity of a thick margin relative to a thin one is 2.387. (https://5dlibrary.com/answers/en/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c04/) 第4章 三百件存量的盘点与尖刀筛选 — A 段:弹性为 0.10 时,毛利 0.955 的工具,推广支出占收入的比例应为 0.0955;毛利 0.868 的书为 0.0868;毛利 0.400 的 KDP 版本只有 0.0400。B 段换成每周预算:以示意收入计,三把刀分别是 74.04、64.32、63.51,KDP 版本只有 12.00。C 段:收入与弹性相同时,厚毛利与薄毛利的推广强度之比是 2.387。 (https://5dlibrary.com/answers/zh/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c04/) - Ch.5 Trial Firing and Culling: Which Knives to Try First, and When to Change — Chapter 5 shows that committing after day one picks the best knife with probability 0.550. (https://5dlibrary.com/answers/en/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c05/) 第5章 试投淘汰:先试哪几把刀,什么时候换 — 第五章说明七天窗口里第一天后全押只有 0.550 的机会押对。 (https://5dlibrary.com/answers/zh/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c05/) - Ch.6 The Seven-Day Schedule, the Probability of Reaching the Target, and Stop-Loss and Add-On Rules — Chapter 6 schedules by dynamic programming and simulates achievement: reaching the daily target on all seven days has probability 0.021, on at least three days 0.605. (https://5dlibrary.com/answers/en/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c06/) 第6章 七日排程、达成概率与止损加码 — 第六章用动态规划排程、用蒙特卡洛给出达成概率:七天每天过日目标的概率仅 0.021,至少三天过线为 0.605。 (https://5dlibrary.com/answers/zh/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c06/) - Ch.7 Sunzi and Clausewitz: The Classical Sources on Concentration Examined — Section B turns to the market reading. The other side spends 1 in each of ten markets of equal size. Should we put all 10 into one market, or 1 into each? With g at 0.5, 1.0 and 2.0, putting everything in one place wins 0.760, 0.909 and 0.990 of a market; putting 1 into each wins a steady 5.000 markets. Section C adds a threshold. With a threshold m of 0.5, spreading still wins 5.000; with the threshold raised to 2.0, a stake of 1 cannot reach it anywhere, spreading earns nothing, and concentration still earns 0.909. Section D turns concentration into rotation: putting 10 into one market each period gives a cumulative 9.091 market-periods over ten periods, while spreading 1 into each market every period gives 50.000 market-periods, provided the threshold is no higher than 1, and 0.000 otherwise. (https://5dlibrary.com/answers/en/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c07/) 第7章 孙子与克劳塞维茨:集中兵力的原典考辨 — B 段换成市场的读法:对方在十个同样大小的市场各投 1,我方把 10 全压进一个市场,还是每个市场各投 1?g 为 0.5、1.0、2.0 时,全压一处分别拿到 0.760、0.909、0.990 个市场;每处各投 1,稳拿 5.000 个市场。C 段加入门槛:门槛 m 为 0.5 时,分散仍得 5.000;门槛升到 2.0,各投 1 都够不到,分散归零,集中仍有 0.909。D 段把集中改成轮流:每期把 10 压进一个市场,十期下来累计 9.091 个市场期;每期分散各投 1,十期累计 50.000 个市场期,前提是门槛不高于 1,否则为 0.000。 (https://5dlibrary.com/answers/zh/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c07/) - Ch.8 Field Tools and Checker Specifications: The Ledger, the Red Flags and the Pre-Registered Tests — Chapter 8 gives the runnable Catalog Firepower Concentration and Seven-Day Cash Scheduler, checker specifications grouped by the four kits, the sprint ledger and pre-registered tests, the first needing 3,826 visitors per group. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c08/) 第8章 实战工具与检查器规格:台账、红灯与预注册检验 — 第八章给出可运行的「存量资产火力收敛与七日现金流排程器」、按四件套分组的检查器规格、七日突击台账与预注册检验,检验一每组需 3,826 位访客。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/seven-day-cashflow-ignition-digital-asset-catalogs-firepower-concentration-sprints/c08/) ## 5DL-B-416 · Pocket Command of Multi-Container Fleets: Mobile Remote Control, Exception-Only Push Filtering and Away-From-Keyboard Continuity 中文书名:移动终端单手遥控多容器集群、异常中断推送滤波与离席零停摆指挥母典 DOI: pending Cite: Fan, Y. (2026). Pocket Command of Multi-Container Fleets: Mobile Remote Control, Exception-Only Push Filtering and Away-From-Keyboard Continuity (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book offers non-personalised operating guidance, promises no availability figure, and was written without connecting to any machine. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: Founders who step away from the desk tend to hold three popular beliefs: more pushes mean more safety; a phone in the hand is a fleet in the hand; enough self-repair scripts deliver zero downtime. This book rewrites the three as claims that data can refute. Its spine is that pushing only exceptions is not laziness but a consequence of signal-to-noise arithmetic. After checking the sources, it makes five corrections to the commissioning blueprint: the belief that more pushes are safer is refuted; waking the phone only for out-of-bounds anomalies becomes a push budget; zero downtime becomes an availability target with an error budget; recovery through lost links becomes an automatic repair share that must be measured; and one-thumb control becomes upper bounds on bytes and steps per day. Its parameters are illustrative; all calculations are generated by its programs; the book offers non-personalised operating guidance, promises no availability figure, and was written without connecting to any machine. Chapter 1 uses Bayes' rule to show that with some nine thousand checks a day, a base rate of 0.0005 and a false-alarm rate of 0.001, a push is genuine with probability only 0.3312; once alarm response is modelled, the share of real faults acted on at the best threshold is 0.807. Chapter 2 compares fixed thresholds, cumulative sums and weighted averages and hangs alerts on the error budget: a target of 0.999 leaves only 43.2 minutes of full outage a month. Chapter 3's three-stage grouping cuts pushes from 214.2 a day to 1.30, a reduction of 0.9939. Chapter 4 shows that the phi accrual detector falls far short of its nominal false-alarm rate on a long-tailed link, 0.0001 a day nominal against 51.00 observed at a threshold of 8, so that reliable verdicts need confirmation from several vantage points. Chapter 5 uses backoff, jitter, circuit breakers and a Markov model to show that with human repair from the phone averaging forty-five minutes, at least 0.948 of failures must be repaired automatically to fit the budget. Chapter 6 shows that backlog during a lost link is bounded, that 1,540 bytes a day keep the phone current, and that dangerous operations need a second confirmation, deny by default, carry leases and are written to a hash chain. Chapter 7 examines lines from the Records of the Grand Historian and from Sunzi together with Ashby's law of requisite variety; a split division of labour gives a mean outage of 7.8 minutes against 39.4 for all-remote handling. Chapter 8 gives the field tools and checker specifications, the away-from-desk ledger and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:离开桌面的创始人常信三句流行说法:推送越多越安全;手机在手,集群就在手;自愈脚本够多就能零停摆。本书把三句话改写成可以被数据推翻的命题。主轴是:只推异常不是偷懒,而是信噪比的必然。核对原文后,本书对立项蓝图做了五处修正:推送越多越安全被证伪;「只在异常越界时唤醒手机」改写为推送预算;「零停摆」改写为可用性目标与错误预算;「断网自愈」改写为必须实测的自动修复份额;「单手遥控」改写为每日字节与操作步数上界。书中参数为示意;全部计算由书中程序生成;本书是非个人化经营建议,不承诺任何可用性数字,只写书、不连接任何机器。 第一章用贝叶斯公式说明,每天九千多次检查、基础率 0.0005、误报率 0.001 时,单条推送为真的概率只有 0.3312;加入告警响应后,最优阈值下被处理的真故障份额为 0.807。第二章比较固定阈值、累积和与指数加权,并把告警挂在错误预算上:目标 0.999 时每月只有 43.2 分钟全停预算。第三章三级归并把推送从每天 214.2 条降到 1.30 条,降幅 0.9939。第四章说明 phi 累积检测器在长尾链路上远达不到名义误报率:阈值为 8 时名义每天 0.0001 次,实测 51.00 次,可靠判定要靠多观测点确认。 第五章用退避、抖动、熔断与马尔可夫模型算出,人在手机端平均四十五分钟修好时,自动修复份额至少要 0.948 才进得了预算。第六章说明断网积压有上界,手机每天只需 1,540 字节就能跟上状态,危险操作二次确认、默认拒绝,审批带租约,裁决写进哈希链。第七章考辨《史记》「运筹策帷帐之中」、《孙子·谋攻》「将能而君不御」与阿什比必要多样性定律,分工时平均停机 7.8 分钟,全部远程时 39.4 分钟。第八章给出实战工具与检查器规格、离席指挥台账与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Base Rates: the More Pushes, the Less Credible Each One — Chapter 1 uses Bayes' rule to show that with some nine thousand checks a day, a base rate of 0.0005 and a false-alarm rate of 0.001, a push is genuine with probability only 0.3312; once alarm response is modelled, the share of real faults acted on at the best threshold is 0.807. (https://5dlibrary.com/answers/en/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c01/) 第1章 基础率:推送越多,越不可信 — 第一章用贝叶斯公式说明,每天九千多次检查、基础率 0.0005、误报率 0.001 时,单条推送为真的概率只有 0.3312;加入告警响应后,最优阈值下被处理的真故障份额为 0.807。 (https://5dlibrary.com/answers/zh/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c01/) - Ch.2 Detectors: Cumulative Sums, Weighted Averages and the Error Budget — Chapter 2 compares fixed thresholds, cumulative sums and weighted averages and hangs alerts on the error budget: a target of 0.999 leaves only 43.2 minutes of full outage a month. (https://5dlibrary.com/answers/en/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c02/) 第2章 检测器:累积和、指数加权与错误预算 — 第二章比较固定阈值、累积和与指数加权,并把告警挂在错误预算上:目标 0.999 时每月只有 43.2 分钟全停预算。 (https://5dlibrary.com/answers/zh/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c02/) - Ch.3 Grouping and Interruption: One Event, One Push — Chapter 3's three-stage grouping cuts pushes from 214.2 a day to 1.30, a reduction of 0.9939. (https://5dlibrary.com/answers/en/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c03/) 第3章 归并与打断:一件事只推一条 — 第三章三级归并把推送从每天 214.2 条降到 1.30 条,降幅 0.9939。 (https://5dlibrary.com/answers/zh/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c03/) - Ch.4 Heartbeats and Probes: Late Is Not Dead — Chapter 4 shows that the phi accrual detector falls far short of its nominal false-alarm rate on a long-tailed link, 0.0001 a day nominal against 51.00 observed at a threshold of 8, so that reliable verdicts need confirmation from several vantage points. (https://5dlibrary.com/answers/en/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c04/) 第4章 心跳与探活:迟到不等于倒下 — 第四章说明 phi 累积检测器在长尾链路上远达不到名义误报率:阈值为 8 时名义每天 0.0001 次,实测 51.00 次,可靠判定要靠多观测点确认。 (https://5dlibrary.com/answers/zh/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c04/) - Ch.5 Automatic Repair: Backoff, Circuit Breakers and the Availability Ledger — Chapter 5 uses backoff, jitter, circuit breakers and a Markov model to show that with human repair from the phone averaging forty-five minutes, at least 0.948 of failures must be repaired automatically to fit the budget. (https://5dlibrary.com/answers/en/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c05/) 第5章 自愈:退避、熔断与可用性账本 — 第五章用退避、抖动、熔断与马尔可夫模型算出,人在手机端平均四十五分钟修好时,自动修复份额至少要 0.948 才进得了预算。 (https://5dlibrary.com/answers/zh/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c05/) - Ch.6 Lost Links, Thin Links and Permissions: What the Phone May and May Not Do — Chapter 6 shows that backlog during a lost link is bounded, that 1,540 bytes a day keep the phone current, and that dangerous operations need a second confirmation, deny by default, carry leases and are written to a hash chain. (https://5dlibrary.com/answers/en/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c06/) 第6章 断网、窄带与权限:手机能做什么,不能做什么 — 第六章说明断网积压有上界,手机每天只需 1,540 字节就能跟上状态,危险操作二次确认、默认拒绝,审批带租约,裁决写进哈希链。 (https://5dlibrary.com/answers/zh/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c06/) - Ch.7 The Tent, the Field and Requisite Variety: Eastern and Western Sources Examined — Chapter 7 examines lines from the Records of the Grand Historian and from Sunzi together with Ashby's law of requisite variety; a split division of labour gives a mean outage of 7.8 minutes against 39.4 for all-remote handling. (https://5dlibrary.com/answers/en/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c07/) 第7章 帷帐、阃外与必要多样性:东西方原典考辨 — 第七章考辨《史记》「运筹策帷帐之中」、《孙子·谋攻》「将能而君不御」与阿什比必要多样性定律,分工时平均停机 7.8 分钟,全部远程时 39.4 分钟。 (https://5dlibrary.com/answers/zh/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c07/) - Ch.8 Field Tools and Checker Specifications: the Away-From-Desk Ledger and a Pre-Registered Test — Chapter 8 gives the field tools and checker specifications, the away-from-desk ledger and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c08/) 第8章 实战工具与检查器规格:离席指挥台账与预注册检验 — 第八章给出实战工具与检查器规格、离席指挥台账与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/pocket-command-multi-container-fleets-exception-only-push-filtering-afk-continuity/c08/) ## 5DL-B-417 · Divergent Pathways of Contemplative Insight: Top-Down Analytical Decoupling versus Embodied Relational Attunement 中文书名:两性深度觉察路径分岔:抽象逻辑解构与具身关系照见的认知神经母典 DOI: pending Cite: Fan, Y. (2026). Divergent Pathways of Contemplative Insight: Top-Down Analytical Decoupling versus Embodied Relational Attunement (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative or synthetic; all calculations are generated by its programs; it is not medical advice, does not replace qualified professionals, and promises no awakening or attainment of any kind. Chapter 4 maps three families of meditation onto analytic, embodied and mixed routes; assignment by profile adds 0.0694 of gain over assignment by sex, and the illustrative dose optimum is about 185 minutes a week. Abstract: A saying circulates in contemplative-course circles: men suit top-down conceptual analysis and detachment, women suit bottom-up bodily feeling and relational mirroring. This book keeps what is true in it, that routes of contemplative practice really do fork and that different training modules produce different changes, and falsifies where it places the fork. After checking the sources, it corrects the brief's "two roads of the sexes" to distributions of individual preference and starting point on two routes, with sex as a weak prior and assignment based on a measurable profile together with the learner's own choice. Its parameters are illustrative or synthetic; all calculations are generated by its programs; it is not medical advice, does not replace qualified professionals, and promises no awakening or attainment of any kind. Chapter 1 shows that a small effect means a large overlap: at the upper edge of the small band two groups still overlap by 0.861, and under illustrative parameters sex-based assignment misplaces 0.426 of learners against 0.147 for a profile of reliability 0.80. Chapter 2 separates interoceptive accuracy, sensibility and awareness, and shows how heartbeat counting scores reward guessing. Chapter 3 shows that the gap in empathy lives mostly in self-report: with a true gap of 0.10 and a role-expectation shift of 0.30, the gap on self-report is 13.2 times the gap on physiology. Chapter 4 maps three families of meditation onto analytic, embodied and mixed routes; assignment by profile adds 0.0694 of gain over assignment by sex, and the illustrative dose optimum is about 185 minutes a week. Chapter 5 shows that rhythms can only be found within persons; with about 84 days of records a rhythm of amplitude 0.50 is detected 0.817 of the time. Chapter 6 gives Beta-posterior monitoring and stopping thresholds for difficult experiences, and shows that weekly peeking pushes false declarations from 0.057 to 0.179. Chapter 7 examines the fasting of the mind in the Zhuangzi, Descartes's Second Meditation and James on emotion, and turns them into clauses for four business kits. Chapter 8 gives the course ledger, the distilled Awareness Route Profiler and Difficult-Experience Monitor, checker specifications and a pre-registered test; the profile-by-route interaction needs about 1,400 learners. Marked boxes separate published evidence, runnable Python calculations (25 programs), classical sources and the author's hypotheses. 摘要:觉察课程圈子里流行一句话:男人适合自上而下的概念拆解与抽离,女人适合自下而上的身体感受与关系镜映。本书只保留其中真实的部分:觉察路径确实分岔,不同练习模块练出不同变化;被证伪的是分岔的位置。核对文献后,本书把立项蓝图的「两性路径」修正为个体在两条路径上的偏好与起点分布,性别只是弱先验,分轨依据是可测的个人画像加本人自选。书中参数为示意或合成;全部计算由书中程序生成;本书不构成医疗建议,不替代合格专业人员,不对开悟或任何境界作保证。 第一章说明小效应意味着大重叠:小效应上沿两组重叠系数仍有 0.861;示意参数下按性别分轨错配 0.426,按信度 0.80 的画像只有 0.147。第二章拆开内感受准确度、敏感度与觉知,并说明心搏计数分数会奖励猜测。第三章说明共情差距多在自评通道:真实差距 0.10、角色预期偏移 0.30 时,自评与生理两种通道的差距之比 13.2 倍。第四章把冥想三族落到分析、具身与混合三条路径,按画像分轨比按性别多出 0.0694 的增益,示意剂量最优约每周 185 分钟。 第五章说明节律只能在个人之内找,记约 84 天时振幅 0.50 的节律检出率 0.817。第六章给出困难体验的 Beta 后验监测与停止阈值,并说明每周偷看会把误报从 0.057 推到 0.179。第七章考辨《庄子》心斋、笛卡儿第二沉思与詹姆斯情绪论,并落成四件套条款。第八章给出课程台账、蒸馏小工具「觉察路径画像分轨与困难体验监测器」、检查器规格表与预登记检验,画像与路径的交互约需 1,400 名学员。书中以标注框区分已发表的证据、可运行的 Python 计算(25 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Overlapping Distributions: Effect Sizes, the Overlap Coefficient and the Mismatch Rate of Sex-Based Tracks — Chapter 1 shows that a small effect means a large overlap: at the upper edge of the small band two groups still overlap by 0.861, and under illustrative parameters sex-based assignment misplaces 0.426 of learners against 0.147 for a profile of reliability 0.80. (https://5dlibrary.com/answers/en/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c01/) 第1章 重叠的分布:效应量、重叠系数与按性别分轨的错配率 — 第一章说明小效应意味着大重叠:小效应上沿两组重叠系数仍有 0.861;示意参数下按性别分轨错配 0.426,按信度 0.80 的画像只有 0.147。 (https://5dlibrary.com/answers/zh/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c01/) - Ch.2 Three Dimensions of Interoception: Why Accuracy, Sensibility and Awareness Cannot Stand In for One Another — Part B of p22 works through several cases. An observed correlation of 0.20 between two measures with reliabilities of 0.60 and 0.70 corrects to 0.309. If both reliabilities are 0.80, it corrects to 0.250. An observed 0.10 with reliabilities of 0.50 and 0.70 corrects to 0.169. An observed 0.30 with both reliabilities at 0.90 corrects to 0.333. (https://5dlibrary.com/answers/en/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c02/) 第2章 内感受三维:准确度、敏感度与觉知为何不能互相代替 — p22 的 B 段算了几组数:观测相关 0.20,两把尺子信度 0.60 与 0.70,校正后 0.309;信度都到 0.80,校正后 0.250;观测 0.10、信度 0.50 与 0.70,校正后 0.169;观测 0.30、信度都是 0.90,校正后 0.333。 (https://5dlibrary.com/answers/zh/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c02/) - Ch.3 The Measurement Gap in Empathy: Self-Report, Behaviour, Physiology and Reverse Inference from Brain Signals — Chapter 3 shows that the gap in empathy lives mostly in self-report: with a true gap of 0.10 and a role-expectation shift of 0.30, the gap on self-report is 13.2 times the gap on physiology. (https://5dlibrary.com/answers/en/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c03/) 第3章 共情的测量鸿沟:自评、行为、生理与脑信号的反向推断 — 第三章说明共情差距多在自评通道:真实差距 0.10、角色预期偏移 0.30 时,自评与生理两种通道的差距之比 13.2 倍。 (https://5dlibrary.com/answers/zh/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c03/) - Ch.4 Three Families of Meditation and Two Routes: Analytic Decoupling, Embodied Relational Attunement and the Mixed Route — Chapter 4 maps three families of meditation onto analytic, embodied and mixed routes; assignment by profile adds 0.0694 of gain over assignment by sex, and the illustrative dose optimum is about 185 minutes a week. (https://5dlibrary.com/answers/en/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c04/) 第4章 冥想三族与两条路径:分析式解构、具身关系式照见与混合路径 — 第四章把冥想三族落到分析、具身与混合三条路径,按画像分轨比按性别多出 0.0694 的增益,示意剂量最优约每周 185 分钟。 (https://5dlibrary.com/answers/zh/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c04/) - Ch.5 Rhythms Within the Person: Why Group Averages Erase Cycles, and How an Individual Can Detect and Use One — Chapter 5 shows that rhythms can only be found within persons; with about 84 days of records a rhythm of amplitude 0.50 is detected 0.817 of the time. (https://5dlibrary.com/answers/en/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c05/) 第5章 个体内节律:组平均为何抹平周期,个人又怎样检出与利用 — 第五章说明节律只能在个人之内找,记约 84 天时振幅 0.50 的节律检出率 0.817。 (https://5dlibrary.com/answers/zh/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c05/) - Ch.6 Difficult Experiences, Pausing and Referral: Bayesian Monitoring, Stopping Thresholds and the Trap of Peeking — Chapter 6 gives Beta-posterior monitoring and stopping thresholds for difficult experiences, and shows that weekly peeking pushes false declarations from 0.057 to 0.179. (https://5dlibrary.com/answers/en/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c06/) 第6章 困难体验、暂停与转介:贝叶斯监测、停止阈值与偷看数据的陷阱 — 第六章给出困难体验的 Beta 后验监测与停止阈值,并说明每周偷看会把误报从 0.057 推到 0.179。 (https://5dlibrary.com/answers/zh/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c06/) - Ch.7 Eastern and Western Classics Examined: the Fasting of the Mind in the Zhuangzi, Descartes's Second Meditation and James on Emotion — With a flagging rate of 0.50 and a false-flagging rate of 0.05, one round lowers the share of false beliefs from 0.3000 to 0.1840 while keeping 0.950 of the true ones. After six rounds the false share is only 0.0090, but only 0.735 of the true beliefs remain. If the false-flagging rate rises to 0.15, it takes eight rounds to push the false share below 0.01, and only 0.272 of the true beliefs survive. Part C shows the point Descartes was looking for: a belief that no round of doubt can flag survives all eight rounds intact, at 1.000. (https://5dlibrary.com/answers/en/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c07/) 第7章 东西方原典考辨:《庄子》心斋、笛卡儿第二沉思与詹姆斯情绪论 — 标出率 0.50、误标率 0.05 时,一轮之后假信念占比从 0.3000 降到 0.1840,真信念保留 0.950;六轮之后假信念只剩 0.0090,可真信念也只剩 0.735。误标率升到 0.15,要八轮才能把假信念占比压到 0.01 以下,真信念只剩 0.272。C 段是笛卡儿要找的那个点:一个任何一轮都标不出来的信念,八轮之后仍然完整保留,1.000。 (https://5dlibrary.com/answers/zh/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c07/) - Ch.8 Closing the Loop: a Ledger for Awareness Courses, Field Tools and Checker Specifications — Chapter 8 gives the course ledger, the distilled Awareness Route Profiler and Difficult-Experience Monitor, checker specifications and a pre-registered test; the profile-by-route interaction needs about 1,400 learners. Marked boxes separate published evidence, runnable Python calculations (25 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c08/) 第8章 实证闭环:觉察课程台账、实战工具与检查器规格 — 第八章给出课程台账、蒸馏小工具「觉察路径画像分轨与困难体验监测器」、检查器规格表与预登记检验,画像与路径的交互约需 1,400 名学员。书中以标注框区分已发表的证据、可运行的 Python 计算(25 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/divergent-contemplative-insight-pathways-analytical-decoupling-embodied-relational-attunement/c08/) ## 5DL-B-418 · Biophilic Botanical Fractals and Geometric Symmetry in Interfaces: Subconscious Stress Reduction and First-Impression Trust Lift 中文书名:亲生命自然分形与几何对称:数字界面潜意识降压与首屏信任度工程母典 DOI: pending Cite: Fan, Y. (2026). Biophilic Botanical Fractals and Geometric Symmetry in Interfaces: Subconscious Stress Reduction and First-Impression Trust Lift (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations come from its programs; it is neither medical nor personalised business advice. Chapter 3 finds, in an illustrative model, that a first screen of a few seconds retains at most 0.125 of a ten-minute laboratory effect, and that testing 24 markers at once gives a 0.708 chance of at least one false alarm. Abstract: A popular claim holds that natural plant fractals combined with precise geometry on a page lower visitors' subconscious stress and markedly lift first-screen trust and conversion; that a fractal dimension of 1.3 to 1.5 is the comfort zone; that rotational symmetry signals craft; and that a golden ratio governs the mix. The book splits the claim into six testable parts and corrects the blueprint against the sources: moderate dimensions are preferred, but in small laboratory studies of static images, and the peak drifts; "stress reduction" means only laboratory markers such as skin conductance and EEG alpha, and no health effect is claimed; "a marked lift in conversion" becomes a mediation-chain hypothesis with conditions of refutation; and the "golden ratio" becomes a best interval fitted on one's own data. Parameters are illustrative; all calculations come from its programs; it is neither medical nor personalised business advice. Chapter 1 estimates dimension by box counting and shows how scale range and grey threshold move the estimate. Chapter 2 uses the bootstrap to show that 24 participants pin the peak only to an interval about 0.098 wide, and that the pooled peak moves to 1.55 when half the test images are exact fractals. Chapter 3 finds, in an illustrative model, that a first screen of a few seconds retains at most 0.125 of a ten-minute laboratory effect, and that testing 24 markers at once gives a 0.708 chance of at least one false alarm. Chapter 4 shows that symmetry lowers the cost of description, yet threefold symmetry has an orientation entropy of 3.606 bits, higher than a random pattern. Chapter 5 computes contrast with the WCAG formula: mid-grey secondary text reaches only 4.39:1 on a plain page; the near-optimal interval of the illustrative mix is 0.30 to 0.40 and excludes 0.618; and the pooled peak of a bimodal population has nothing to do with 1.618. Chapter 6 reads one synthetic A/B test: a relative lift of 0.150, a p-value of 0.107 and a posterior probability of 0.946 that B beats A still do not justify declaring a winner; peeking twenty times pushes the false-alarm rate to 0.253; and a price change of 0.086 units cancels the gain in first impression. Chapter 7 examines Ji Cheng's Yuanye ("made by people, as if opened by heaven"), the Great Commentary on the Book of Changes, Vitruvius on symmetry and Mandelbrot's coastline, and shows that under modular symmetry the golden ratio is the hardest ratio to make commensurable, missing by 0.084 of a module on a side of 12 modules. Chapter 8 gives the distilled Biophilic Fractal Ratio and First-Screen Trust Experiment Calculator, checker specifications, ledger fields and a pre-registered test; at a baseline of 0.03 and a relative lift of 15 percent, each arm needs 24,193 visitors. Boxes separate published evidence, runnable Python (23 programs), classical sources and hypotheses. 摘要:设计圈流行一种说法:把花草树木的自然分形纹理与精确几何图形融进页面,访客的潜意识压力就会下降,首屏信任与商业转化随之显著提升;分形维数 1.3 至 1.5 是舒适区,旋转对称是低熵的工艺信号,有机与几何之间存在黄金配比。本书把这句话拆成六个可检验部件,逐一按文献立界。核对原文后,本书写下蓝图修正:中等维数受偏爱,但证据来自小样本实验室静态图片,峰位随刺激类型与人群漂移;「降压」只是皮肤电导、脑电 alpha 一类实验室压力标志物的差异,不是血压读数,本书不宣称任何健康功效;「显著提升转化」改写为带推翻条件的中介链假说;「黄金配比」改写为用自家数据拟合的最优区间。书中参数为示意;全部计算由书中程序生成;本书不是医疗建议,也不是个人化经营建议。 第一章用标准库盒计数估分形维数,并说明尺度范围与灰度阈值都会移动估计值。第二章用自助法说明 24 名被试只能把峰位钉在约 0.098 宽的区间里,测试图片中精确分形占一半时合并峰位移到 1.55。第三章按示意模型算出几秒首屏最多保留实验室十分钟效应的 0.125,同时检验 24 项指标时至少一项误报的概率达 0.708。第四章说明对称降低描述成本,但三阶对称的朝向熵 3.606 比特反而高于随机图案。第五章按 WCAG 公式实算对比度,中灰辅助文字在纯色页面上只有 4.39:1,示意配比的近优区间为 0.30 至 0.40,不含 0.618,双峰人群的合并峰与 1.618 无关。 第六章读一次合成 A/B 检验:相对提升 0.150、p 值 0.107、B 优于 A 的后验概率 0.946,仍不能宣布胜出;反复偷看二十次把误报率推到 0.253;价格变动 0.086 个单位即可抵消首印象增益。第七章考辨计成《园冶》「虽由人作,宛自天开」、《周易·系辞上》、维特鲁威论对称与曼德博的海岸线,说明按模数理解对称时黄金比例最难公度,在 12 模数边长上差 0.084 个模数。第八章给出蒸馏小工具「亲生命分形配比与首屏信任实验核算器」、按四件套分组的检查器规格、台账字段与预注册检验,基线 0.03、相对提升百分之十五时每臂需要 24,193 人。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Box Counting: Turning Natural Texture into a Dimension — The results follow a systematic pattern: the higher the theoretical dimension, the further below it the estimate falls. At H equal to 0.9 the theoretical value is 1.10 and the mean estimate is 1.144, slightly too high. At H equal to 0.5 the theory says 1.50 but the mean estimate is only 1.409. At H equal to 0.3 the theory says 1.70 and the mean estimate is 1.554. The cause is the finite number of points. The rougher the curve, the finer the boxes needed to capture all its wiggles, and four thousand points cannot support such fine levels. The standard deviation across twenty curves lies between 0.024 and 0.041, which shows that different random seeds under the same parameter also produce visible differences. (https://5dlibrary.com/answers/en/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c01/) 第1章 盒计数:把自然纹理量成一个维数 — 结果有一个系统规律:理论维数越高,估计值低得越多。H 为 0.9 时理论值 1.10,估计均值 1.144,反而略高;H 为 0.5 时理论值 1.50,估计均值只有 1.409;H 为 0.3 时理论值 1.70,估计均值 1.554。原因在于有限点数:曲线越粗糙,越需要更细的盒子才能把起伏数全,而四千多个点撑不起那么细的层级。二十条曲线之间的标准差在 0.024 到 0.041 之间,说明同一参数下不同随机种子也会带来可见差异。 (https://5dlibrary.com/answers/zh/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c01/) - Ch.2 The Preference Band: Evidence, Sample Size and Drift for Moderate Fractal Dimension — Chapter 2 uses the bootstrap to show that 24 participants pin the peak only to an interval about 0.098 wide, and that the pooled peak moves to 1.55 when half the test images are exact fractals. (https://5dlibrary.com/answers/en/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c02/) 第2章 偏好带:中等分形维数的证据、样本与漂移 — 第二章用自助法说明 24 名被试只能把峰位钉在约 0.098 宽的区间里,测试图片中精确分形占一半时合并峰位移到 1.55。 (https://5dlibrary.com/answers/zh/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c02/) - Ch.3 What "Subconscious Stress Reduction" Measures: Laboratory Stress Markers — Chapter 3 finds, in an illustrative model, that a first screen of a few seconds retains at most 0.125 of a ten-minute laboratory effect, and that testing 24 markers at once gives a 0.708 chance of at least one false alarm. (https://5dlibrary.com/answers/en/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c03/) 第3章 「潜意识降压」量的是什么:实验室压力标志物 — 第三章按示意模型算出几秒首屏最多保留实验室十分钟效应的 0.125,同时检验 24 项指标时至少一项误报的概率达 0.708。 (https://5dlibrary.com/answers/zh/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c03/) - Ch.4 Symmetry Order and Entropy: Which Entropy Is Low? — Chapter 4 shows that symmetry lowers the cost of description, yet threefold symmetry has an orientation entropy of 3.606 bits, higher than a random pattern. (https://5dlibrary.com/answers/en/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c04/) 第4章 对称阶与熵:低熵指的是哪一种熵 — 第四章说明对称降低描述成本,但三阶对称的朝向熵 3.606 比特反而高于随机图案。 (https://5dlibrary.com/answers/zh/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c04/) - Ch.5 The Ratio: Best Intervals for the Organic Share, Contrast Gates and the Golden Ratio — Chapter 5 computes contrast with the WCAG formula: mid-grey secondary text reaches only 4.39:1 on a plain page; the near-optimal interval of the illustrative mix is 0.30 to 0.40 and excludes 0.618; and the pooled peak of a bimodal population has nothing to do with 1.618. (https://5dlibrary.com/answers/en/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c05/) 第5章 配比:有机占比的最优区间、对比度门禁与黄金比例之辨 — 第五章按 WCAG 公式实算对比度,中灰辅助文字在纯色页面上只有 4.39:1,示意配比的近优区间为 0.30 至 0.40,不含 0.618,双峰人群的合并峰与 1.618 无关。 (https://5dlibrary.com/answers/zh/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c05/) - Ch.6 First-Screen Trust: Fifty Milliseconds, A/B Tests and the Mediation Chain — Chapter 6 reads one synthetic A/B test: a relative lift of 0.150, a p-value of 0.107 and a posterior probability of 0.946 that B beats A still do not justify declaring a winner; peeking twenty times pushes the false-alarm rate to 0.253; and a price change of 0.086 units cancels the gain in first impression. (https://5dlibrary.com/answers/en/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c06/) 第6章 首屏信任:五十毫秒、A/B 检验与中介链 — 第六章读一次合成 A/B 检验:相对提升 0.150、p 值 0.107、B 优于 A 的后验概率 0.946,仍不能宣布胜出;反复偷看二十次把误报率推到 0.253;价格变动 0.086 个单位即可抵消首印象增益。 (https://5dlibrary.com/answers/zh/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c06/) - Ch.7 Classical Sources East and West: the Yuanye, the Great Commentary, Vitruvius and Mandelbrot — Chapter 7 examines Ji Cheng's Yuanye ("made by people, as if opened by heaven"), the Great Commentary on the Book of Changes, Vitruvius on symmetry and Mandelbrot's coastline, and shows that under modular symmetry the golden ratio is the hardest ratio to make commensurable, missing by 0.084 of a module on a side of 12 modules. (https://5dlibrary.com/answers/en/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c07/) 第7章 东西方原典考辨:《园冶》《系辞》、维特鲁威与曼德博 — 第七章考辨计成《园冶》「虽由人作,宛自天开」、《周易·系辞上》、维特鲁威论对称与曼德博的海岸线,说明按模数理解对称时黄金比例最难公度,在 12 模数边长上差 0.084 个模数。 (https://5dlibrary.com/answers/zh/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c07/) - Ch.8 Field Tools and Checker Specifications: the Interface Trust Ledger and a Pre-Registered Test — Chapter 8 gives the distilled Biophilic Fractal Ratio and First-Screen Trust Experiment Calculator, checker specifications, ledger fields and a pre-registered test; at a baseline of 0.03 and a relative lift of 15 percent, each arm needs 24,193 visitors. Boxes separate published evidence, runnable Python (23 programs), classical sources and hypotheses. (https://5dlibrary.com/answers/en/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c08/) 第8章 实战工具与检查器规格:界面信任实验台账与预注册检验 — 第八章给出蒸馏小工具「亲生命分形配比与首屏信任实验核算器」、按四件套分组的检查器规格、台账字段与预注册检验,基线 0.03、相对提升百分之十五时每臂需要 24,193 人。书中以标注框区分已发表的证据、可运行的 Python 计算(23 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/biophilic-botanical-fractals-geometric-symmetry-interface-stress-reduction-first-impression-trust/c08/) ## 5DL-B-419 · Three-Second Thin-Slicing Trust: Visual-Auditory-Olfactory Congruence and Instant Authority Signalling 中文书名:三秒薄切片信任捕获:视听嗅三通道同频一致性与瞬间权威建立母典 DOI: pending Cite: Fan, Y. (2026). Three-Second Thin-Slicing Trust: Visual-Auditory-Olfactory Congruence and Instant Authority Signalling (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and every number is printed by a program. Chapter 1 separates reliability from validity: under illustrative parameters a three-second judgement reaches 0.612 of its ceiling validity and agrees with the long-run evaluation in 0.579 of cases. Abstract: A popular saying holds that trust is fixed in three seconds, multiplied when picture, voice and scent agree, and capped by the weakest channel. This book tests it against the original sources: thin-slice judgements are fast and consensual but not thereby accurate; integration reduces variance rather than multiplying trust; superadditivity belongs mainly to single neurons; a low voice is a weak, context-bound prior; and scent effects depend on congruence, while many people find fragrance troublesome. Parameters are illustrative and every number is printed by a program. The book covers only making genuine, deliverable signals clear; it teaches no faking of credentials, claims no wellbeing effect for any scent, and is non-personalised business guidance. Chapter 1 separates reliability from validity: under illustrative parameters a three-second judgement reaches 0.612 of its ceiling validity and agrees with the long-run evaluation in 0.579 of cases. Chapter 2 shows that with three equally precise channels the standard deviation falls at most to 0.577, and that three channels combine to a discrimination of 1.044, only 1.305 times the best single channel. Chapter 3 uses common-source causal inference to show that once the gap exceeds 3.99 visual standard deviations the observer splits off the jarring channel, and that noise in one channel raises the false-split rate for honest scenes from 0.0235 to 0.0984. Chapter 4 places additive, multiplicative and weakest-link rules on one power-mean dial; the payoff from fixing the weakest channel relative to a strong one is 2.8 times under the geometric mean and 304.8 times at p of -4. Chapter 5 estimates pitch by autocorrelation and shows that jitter can be measured from three seconds while rate variability needs longer recordings; when the effect of a low voice varies between contexts, the chance of a negative effect in the next context can reach 0.120. Chapter 6 quantifies first-screen clarity and scent intensity: planning for a troubled share of 0.347, an announcement plus an off switch raises average acceptance from 0.452 to 0.733, and a single shop-wide scent breaks even at a fitting share of 0.455. Chapter 7 examines the Analects on fine words, Mencius on the pupil of the eye and Aristotle on persuasion through character, and gives distillation clauses for the four kits. Chapter 8 presents the distilled Three-Channel Congruence Trust Combiner and Weakest-Link Calculator, checker specifications grouped by the four kits, a three-second trust ledger and a pre-registered test: a main effect of 0.03 needs 1,047 viewers per cell and 8,376 across the factorial design, and peeking raises the false-alarm rate from 0.0540 to 0.1940. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:门店与短视频圈子流传一句话:信任在前三秒定型,画面、嗓音、气味三路同频时信任成倍放大,任一路违和就触发防备,整体取决于最短板。本书逐条检验这句流行说法,并按原始文献写下蓝图修正:薄切片判断快而一致,却不等于准确;最大似然整合带来的是方差缩减,不是乘法放大;超加性主要见于单个神经元,行为层面多为加性或亚加性;低音高只是依语境而变的弱先验;环境香氛的效应依赖一致性,而且相当比例的人对香氛不适。书中参数为示意;全部计算由书中程序生成;本书只写如何让真实、可兑现的信号在前三秒清晰一致,不写任何伪造资质或假权威的手法,不宣称任何气味的健康功效,是非个人化经营建议。 第一章区分信度与效度,示意参数下三秒判断只达到效度上限的 0.612,与长期评价的一致率为 0.579。第二章说明三路等精度时标准差最多缩到 0.577,三路判别力合并为 1.044,只是最强单路的 1.305 倍。第三章用共同源因果推断说明,差距超过 3.99 个视觉标准差时观者把违和的一路拆开,而一路噪声变大会让诚实场景的误拆率从 0.0235 升到 0.0984。第四章用幂平均把加性、乘性与短板放在一条刻度上,补短板相对补强项的收益在几何平均下是 2.8 倍,在 p 为 -4 时是 304.8 倍。 第五章用自相关估计基频,说明抖动三秒可测、语速变异要更长录音,低音高效应在语境间差异较大时,下一个语境出现负效应的概率可达 0.120。第六章量化首屏净度与气味剂量,不适比例按 0.347 规划时,告知加可关闭把平均接受度从 0.452 提高到 0.733,统一香氛的盈亏点在品类一致比例 0.455。第七章考辨《论语》「巧言令色」、《孟子》论眸子与亚里士多德的品格说服,并给出四件套蒸馏规格。第八章给出蒸馏小工具「三通道一致性信任合成与短板核算器」、按四件套分组的检查器规格、三秒信任台账与预注册检验:主效应 0.03 每格需要 1,047 人,三因子共 8,376 人,中途偷看会把误报率从 0.0540 推到 0.1940。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Reliability and Validity of Thin Slices: How Steady, and How Right, Is a Three-Second Judgement? — Chapter 1 separates reliability from validity: under illustrative parameters a three-second judgement reaches 0.612 of its ceiling validity and agrees with the long-run evaluation in 0.579 of cases. (https://5dlibrary.com/answers/en/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c01/) 第1章 薄切片的信度与效度:三秒判断有多稳,又有多准 — 第一章区分信度与效度,示意参数下三秒判断只达到效度上限的 0.612,与长期评价的一致率为 0.579。 (https://5dlibrary.com/answers/zh/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c01/) - Ch.2 Maximum-Likelihood Integration: Several Channels Buy Precision, Not Multiplication — Chapter 2 shows that with three equally precise channels the standard deviation falls at most to 0.577, and that three channels combine to a discrimination of 1.044, only 1.305 times the best single channel. (https://5dlibrary.com/answers/en/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c02/) 第2章 最大似然整合:多通道带来的是精度提升,不是乘法放大 — 第二章说明三路等精度时标准差最多缩到 0.577,三路判别力合并为 1.044,只是最强单路的 1.305 倍。 (https://5dlibrary.com/answers/zh/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c02/) - Ch.3 Conflict and Causal Inference: Why an Observer Splits a Jarring Channel Instead of Merging It — Chapter 3 uses common-source causal inference to show that once the gap exceeds 3.99 visual standard deviations the observer splits off the jarring channel, and that noise in one channel raises the false-split rate for honest scenes from 0.0235 to 0.0984. (https://5dlibrary.com/answers/en/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c03/) 第3章 违和与因果推断:一路信号失真时,观者为何拆开而不是合并 — 第三章用共同源因果推断说明,差距超过 3.99 个视觉标准差时观者把违和的一路拆开,而一路噪声变大会让诚实场景的误拆率从 0.0235 升到 0.0984。 (https://5dlibrary.com/answers/zh/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c03/) - Ch.4 Four Combination Rules: Fitting Additive, Multiplicative, Weakest-Link and Soft-Minimum Models — Chapter 4 places additive, multiplicative and weakest-link rules on one power-mean dial; the payoff from fixing the weakest channel relative to a strong one is 2.8 times under the geometric mean and 304.8 times at p of -4. (https://5dlibrary.com/answers/en/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c04/) 第4章 四种合成规则:加性、乘性、短板与软最小值的拟合比较 — 第四章用幂平均把加性、乘性与短板放在一条刻度上,补短板相对补强项的收益在几何平均下是 2.8 倍,在 p 为 -4 时是 304.8 倍。 (https://5dlibrary.com/answers/zh/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c04/) - Ch.5 Steadiness of the Voice: Autocorrelation Pitch, Jitter and the Variability of Speaking Rate — Chapter 5 estimates pitch by autocorrelation and shows that jitter can be measured from three seconds while rate variability needs longer recordings; when the effect of a low voice varies between contexts, the chance of a negative effect in the next context can reach 0.120. (https://5dlibrary.com/answers/en/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c05/) 第5章 嗓音稳定度:自相关基频、抖动与语速变异系数 — 第五章用自相关估计基频,说明抖动三秒可测、语速变异要更长录音,低音高效应在语境间差异较大时,下一个语境出现负效应的概率可达 0.120。 (https://5dlibrary.com/answers/zh/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c05/) - Ch.6 Visual Clarity and Ambient Scent: Layout Entropy, the Inverted U of Intensity and Congruence — Chapter 6 quantifies first-screen clarity and scent intensity: planning for a troubled share of 0.347, an announcement plus an off switch raises average acceptance from 0.452 to 0.733, and a single shop-wide scent breaks even at a fitting share of 0.455. (https://5dlibrary.com/answers/en/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c06/) 第6章 视觉净度与空间气味:构图熵、剂量倒 U 与一致性 — 第六章量化首屏净度与气味剂量,不适比例按 0.347 规划时,告知加可关闭把平均接受度从 0.452 提高到 0.733,统一香氛的盈亏点在品类一致比例 0.455。 (https://5dlibrary.com/answers/zh/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c06/) - Ch.7 Eastern and Western Classics: "Fine Words and an Ingratiating Face", the Pupil of the Eye, and Aristotle on Persuasion through Character — Part A shows that with a prior of one half, if fine words have a likelihood ratio of 0.5 (the compensating case), the posterior probability that the person is reliable falls to 0.333 after the signal is seen; with a prior of 0.8, it falls to 0.667. If the likelihood ratio is 1 (neutral), the posterior does not move. A verifiable delivery record, with a likelihood ratio of 3, raises the posterior to 0.750 or 0.923; two independent records, with a ratio of 9, raise it to 0.900 and 0.973. Fine words together with one record give a ratio of 0.5 times 3, which is 1.5, and a posterior of 0.600. (https://5dlibrary.com/answers/en/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c07/) 第7章 东西方原典考辨:「巧言令色」「眸子」与亚里士多德的品格说服 — A 段显示,先验为一半时,若巧言令色的似然比是 0.5(补偿型),看到它之后,对方可靠的后验降到 0.333;先验为 0.8 时降到 0.667。若似然比是 1(中性),后验不变。一条可核验的交付记录,似然比取 3,后验升到 0.750 或 0.923;两条独立记录,似然比 9,后验升到 0.900 与 0.973。巧言令色加一条记录,似然比 0.5 乘 3 等于 1.5,后验 0.600。 (https://5dlibrary.com/answers/zh/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c07/) - Ch.8 Field Tools and Checker Specifications: A Three-Second Trust Ledger, a Combination Calculator and a Pre-Registered Test — Chapter 8 presents the distilled Three-Channel Congruence Trust Combiner and Weakest-Link Calculator, checker specifications grouped by the four kits, a three-second trust ledger and a pre-registered test: a main effect of 0.03 needs 1,047 viewers per cell and 8,376 across the factorial design, and peeking raises the false-alarm rate from 0.0540 to 0.1940. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c08/) 第8章 实战工具与检查器规格:三秒信任台账、合成规则核算器与预注册检验 — 第八章给出蒸馏小工具「三通道一致性信任合成与短板核算器」、按四件套分组的检查器规格、三秒信任台账与预注册检验:主效应 0.03 每格需要 1,047 人,三因子共 8,376 人,中途偷看会把误报率从 0.0540 推到 0.1940。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/three-second-thin-slicing-trust-visual-auditory-olfactory-congruence-instant-authority/c08/) ## 5DL-B-420 · Collapsing Idea-to-Reality Latency: Feedback-Bandwidth Limits, Zero-Friction Toolchains and High-Velocity Execution Physics 中文书名:从意念构想到现实显化的时延坍缩:控制环路带宽极限与超高速执行物理母典 DOI: pending Cite: Fan, Y. (2026). Collapsing Idea-to-Reality Latency: Feedback-Bandwidth Limits, Zero-Friction Toolchains and High-Velocity Execution Physics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative and all calculations are generated by the book's programs; it offers general, not individual, business guidance, promises no launch date or income, and every production deployment is carried out only after a person confirms it. Chapter 2 splits latency into four drags, decision, start-up, feedback and hand-off, with an illustrative median of 99.9 hours and P90 of 199.8 hours, the tail driven mainly by feedback and decision. Abstract: Three popular sayings govern how people try to turn ideas into results faster: things are slow because effort is weak, the fuller the schedule the more efficient the team, and checks can be dropped when time is short. This book tests each with control engineering, queueing theory and decision models, and all three are refuted. After checking the sources it corrects the blueprint's slogan of compressing "a flash of inspiration to launch and payment" into hours, rewriting it as a conditional claim: only with a ready template, a payment flow already live and tested, fully automated gates, demand checked in advance and no hand-offs can end-to-end latency percentiles fall into hours, and then 0.992 of ideas finish within 24 hours. "Manifestation" is a figure of speech for a verifiable delivery chain. Parameters are illustrative and all calculations are generated by the book's programs; it offers general, not individual, business guidance, promises no launch date or income, and every production deployment is carried out only after a person confirms it. Chapter 1 shows that delay sets the ceiling on force: multiplying the gain by 2.5 with the delay fixed stretches settling from 6.69 to 40.63 days. Chapter 2 splits latency into four drags, decision, start-up, feedback and hand-off, with an illustrative median of 99.9 hours and P90 of 199.8 hours, the tail driven mainly by feedback and decision. Chapter 3 shows that at utilisation 0.90 the queueing wait is already 9.00 service times. Chapter 4 gives the U-shaped cost of batch size, with the cost per change falling from 4.19 to 1.20 once delivery overhead drops, and uses the Smith predictor to show the gains and risks of rehearsing in a model. Chapter 5 weighs hand-offs, switching and hesitation: with verbal hand-offs only 0.478 of ideas pass a chain of eight without rework, and the best study time is 1.00 hour for a reversible decision against 16.00 hours for an irreversible one. Chapter 6 shows that gates should be sped up, never skipped: at a fault rate of 0.05 and an 8-hour rollback, expected hours are 0.220 with the gate and 0.400 without, and ordering checks by hit rate over run time brings the first red light from 12.11 to 3.73 minutes. Chapter 7 examines Sunzi on clumsy speed, Han Feizi on the ant hole and Maxwell's 1868 paper on governors: a frequent-feedback group hits its target by day 10 in 0.969 of trials against 0.377 for private polishing, and a quicker linkage lifts the critical governing power from 1.50 to 4.50. Chapter 8 gives the latency ledger, the distilled Idea-to-Delivery Latency Decomposer and Loop-Bandwidth Calculator, checker specifications for four toolkits and a pre-registered test needing 79 ideas per period. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. 摘要:关于「怎样更快把构想变成结果」,流行说法有三句:慢是因为力度不够,排得越满越高效,赶时间时可以先省掉检查。本书用控制工程、排队论与决策模型逐一检验,三句都被证伪。核对文献后,本书把立项蓝图「把灵光一闪到上线收钱压缩到小时级」修正为条件命题:只有模板现成、支付流程已上线并验证、门禁全自动、需求已验证且无交接时,端到端时延分位数才可能落到小时级,此时 24 小时内完成的份额为 0.992;「显化」只作书名修辞,指构想经可验证的步骤变成可交付结果。书中参数为示意,全部计算由书中程序生成;本书是非个人化经营建议,不承诺任何上线时间或收入,生产部署一律人确认后执行。 第一章说明延迟决定力度上限:延迟不变时把增益乘 2.5 倍,落定时间从 6.69 天拉长到 40.63 天。第二章把端到端时延拆成决策、启动、反馈与交接四种阻力,示意中位数 99.9 小时、九成分位 199.8 小时,尾部主要由反馈与决策决定。第三章说明利用率 0.90 时排队等待已是服务时间的 9.00 倍。第四章给出批量的 U 形成本,交付开销下降后单件成本从 4.19 降到 1.20,并用史密斯预估器说明「先在模型里预演」的收益与风险。 第五章称量交接、切换与犹豫:口头交接下八人链只有 0.478 的构想不需返工;可逆决策的最优研究时间为 1.00 小时,不可逆决策为 16.00 小时。第六章说明门禁只提速不跳开:缺陷率 0.05、回滚 8 小时时,有门禁的期望工时 0.220,无门禁 0.400;按命中率除以耗时排序,缺陷改动见到红灯的时间从 12.11 分钟降到 3.73 分钟。第七章考辨孙子「拙速」、韩非「蝼蚁之穴」与麦克斯韦 1868 年论调速器,频繁反馈组到第 10 天命中份额 0.969,闭门打磨组 0.377;连杆滞后缩短后临界调节力度从 1.50 升到 4.50。第八章给出时延台账、蒸馏小工具「构想到交付时延分解与环路带宽核算器」、按四件套分组的检查器规格与预注册检验,每期需 79 个构想。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Loop Bandwidth: With the Delay Unchanged, More Gain Only Buys Oscillation — Chapter 1 shows that delay sets the ceiling on force: multiplying the gain by 2.5 with the delay fixed stretches settling from 6.69 to 40.63 days. (https://5dlibrary.com/answers/en/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c01/) 第1章 环路带宽:延迟不变时,加大增益只会振荡 — 第一章说明延迟决定力度上限:延迟不变时把增益乘 2.5 倍,落定时间从 6.69 天拉长到 40.63 天。 (https://5dlibrary.com/answers/zh/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c01/) - Ch.2 Decomposing End-to-End Latency: Four Drags and the Slowest Tenth — Chapter 2 splits latency into four drags, decision, start-up, feedback and hand-off, with an illustrative median of 99.9 hours and P90 of 199.8 hours, the tail driven mainly by feedback and decision. (https://5dlibrary.com/answers/en/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c02/) 第2章 端到端时延分解:四种阻力与最慢那一成 — 第二章把端到端时延拆成决策、启动、反馈与交接四种阻力,示意中位数 99.9 小时、九成分位 199.8 小时,尾部主要由反馈与决策决定。 (https://5dlibrary.com/answers/zh/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c02/) - Ch.3 Queues: Why Full Utilisation Is Not Efficiency — Chapter 3 shows that at utilisation 0.90 the queueing wait is already 9.00 service times. (https://5dlibrary.com/answers/en/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c03/) 第3章 排队:满负荷为什么不是高效 — 第三章说明利用率 0.90 时排队等待已是服务时间的 9.00 倍。 (https://5dlibrary.com/answers/zh/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c03/) - Ch.4 Batches and Rehearsal: the U-Shaped Cost and the Smith Predictor — Chapter 4 gives the U-shaped cost of batch size, with the cost per change falling from 4.19 to 1.20 once delivery overhead drops, and uses the Smith predictor to show the gains and risks of rehearsing in a model. (https://5dlibrary.com/answers/en/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c04/) 第4章 批量与预演:U 形成本与史密斯预估器 — 第四章给出批量的 U 形成本,交付开销下降后单件成本从 4.19 降到 1.20,并用史密斯预估器说明「先在模型里预演」的收益与风险。 (https://5dlibrary.com/answers/zh/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c04/) - Ch.5 Hand-offs, Switching and Hesitation: Three Losses Along the Chain — Chapter 5 weighs hand-offs, switching and hesitation: with verbal hand-offs only 0.478 of ideas pass a chain of eight without rework, and the best study time is 1.00 hour for a reversible decision against 16.00 hours for an irreversible one. (https://5dlibrary.com/answers/en/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c05/) 第5章 交接、切换与犹豫:链条上的三种损耗 — 第五章称量交接、切换与犹豫:口头交接下八人链只有 0.478 的构想不需返工;可逆决策的最优研究时间为 1.00 小时,不可逆决策为 16.00 小时。 (https://5dlibrary.com/answers/zh/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c05/) - Ch.6 Toolchain Friction and Gates: Why Speed Never Skips a Gate — Chapter 6 shows that gates should be sped up, never skipped: at a fault rate of 0.05 and an 8-hour rollback, expected hours are 0.220 with the gate and 0.400 without, and ordering checks by hit rate over run time brings the first red light from 12.11 to 3.73 minutes. (https://5dlibrary.com/answers/en/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c06/) 第6章 工具链摩擦与门禁:提速为什么不跳门禁 — 第六章说明门禁只提速不跳开:缺陷率 0.05、回滚 8 小时时,有门禁的期望工时 0.220,无门禁 0.400;按命中率除以耗时排序,缺陷改动见到红灯的时间从 12.11 分钟降到 3.73 分钟。 (https://5dlibrary.com/answers/zh/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c06/) - Ch.7 Classical Sources East and West: Clumsy Speed, the Ant Hole and Maxwell on Governors — Chapter 7 examines Sunzi on clumsy speed, Han Feizi on the ant hole and Maxwell's 1868 paper on governors: a frequent-feedback group hits its target by day 10 in 0.969 of trials against 0.377 for private polishing, and a quicker linkage lifts the critical governing power from 1.50 to 4.50. (https://5dlibrary.com/answers/en/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c07/) 第7章 东西方原典考辨:拙速、蝼蚁之穴与麦克斯韦论调速器 — 第七章考辨孙子「拙速」、韩非「蝼蚁之穴」与麦克斯韦 1868 年论调速器,频繁反馈组到第 10 天命中份额 0.969,闭门打磨组 0.377;连杆滞后缩短后临界调节力度从 1.50 升到 4.50。 (https://5dlibrary.com/answers/zh/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c07/) - Ch.8 Closing the Loop: the Latency Ledger, Field Tools and Checker Specifications — Chapter 8 gives the latency ledger, the distilled Idea-to-Delivery Latency Decomposer and Loop-Bandwidth Calculator, checker specifications for four toolkits and a pre-registered test needing 79 ideas per period. Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c08/) 第8章 实证闭环:时延台账、实战工具与检查器规格 — 第八章给出时延台账、蒸馏小工具「构想到交付时延分解与环路带宽核算器」、按四件套分组的检查器规格与预注册检验,每期需 79 个构想。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/collapsing-idea-to-reality-latency-feedback-bandwidth-zero-friction-toolchains-execution-physics/c08/) ## 5DL-B-421 · Dual-Mating Multi-Objective Conflicts, Parental-Investment Asymmetry and the Pareto Frontier of Partner Selection 中文书名:双轨择偶多目标冲突、亲代投资不对称与伴侣筛选帕累托边界母典 DOI: pending Cite: Fan, Y. (2026). Dual-Mating Multi-Objective Conflicts, Parental-Investment Asymmetry and the Pareto Frontier of Partner Selection (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its populations, traits and parameters are synthetic or illustrative, and all calculations are generated by its programs. Marked boxes separate evidence, hypotheses and classical sources; the book contains 22 programs. Abstract: A popular claim holds that anyone who wants both a high-tension signal and certain, exclusive commitment is contradicting himself or herself, because the two traits exclude each other by nature. This book splits the claim into propositions of fact, mechanism and norm and writes each as a model that can be computed and refuted by data. Throughout it speaks of roles, such as the high-fixed-investment party and the party asked to invest, and it reports differences between the sexes only as effect sizes with their overlap. Its populations, traits and parameters are synthetic or illustrative, and all calculations are generated by its programs. It is not relationship counselling, psychological counselling or clinical judgement, and it offers no fate-based derivation. Chapter 1 uses optimal stopping to show that choosiness is a parameter of roles: as fixed investment rises from zero to 4.0, the acceptance threshold rises from 0.8000 to 0.9106, and no sex appears in the formula; at an effect size of d = 1.00 two distributions still overlap by 0.617. Chapter 2 is the core of the book: starting from two independent traits and keeping only the top 10 % by overall appeal, the correlation becomes −0.708, and market clearing produces the same trade-off. Chapter 3 writes the "impossible triangle" as a feasible probability under joint thresholds: requiring the top 20 % on three traits leaves a feasible share of 0.00920, the first value below one in a hundred; relaxing one threshold multiplies the feasible probability by 5.57, while meeting twice as many people only doubles it. Chapter 4 shows that weighted sums cannot reach the inward-bending middle of a Pareto front. Chapter 5 separates necessities from luxuries with a budget constraint, shows the limited predictive power of stated preferences, and notes that two meta-analyses of ovulatory-cycle preference shifts reached opposite conclusions and a large study found no shift, so the "dual track" survives only as a role-general, falsifiable hypothesis about situations. Chapter 6 covers the Nash and Kalai-Smorodinsky solutions and the outside option principle; the party asked to invest uses only truthful disclosure, outside options, its own floor and the right to leave, and in stable matching the single share of the all-max type reaches 0.780. Chapter 7 reads the ode "Mang", the chapter on the meaning of the marriage rite and Darwin's account of 1871 side by side and gives the distillation spec for the four kits. Chapter 8 gives the calculator, the checker specifications, a weekly ledger and preregistered tests. Marked boxes separate evidence, hypotheses and classical sources; the book contains 22 programs. 摘要:流行说法认为,同时索取高张力信号与高确定排他投入的人自相矛盾,因为这两种特质天生互斥。本书把这一说法拆成事实、机制与规范三个命题,逐一写成可以计算、可以被数据推翻的模型。全书以「高固定投资方」「被索取方」等角色指称,两性差异只写效应量与重叠;书中人群、特质与参数均为合成或示意,全部计算由书中程序生成;本书不是婚恋咨询、心理咨询或临床判断,也不作任何命理推导。 第一章用最优停止说明挑剔是角色参数:固定投资从零升到 4.0,接受阈值从 0.8000 升到 0.9106,公式中没有性别;效应量 d = 1.00 时两个分布仍重叠 0.617。第二章是全书核心:从两个独立特质出发,只保留总吸引力前 10 % 的人,相关系数就变成 −0.708,市场出清同样制造此消彼长。第三章把「不可能三角」写成联合阈值下的可行概率:三项都要求前 20 % 时可行比例为 0.00920,首次跌破百分之一;放宽一项门槛使可行概率乘以 5.57,多见一倍的人只乘以二。第四章说明加权和找不到帕累托前沿向内凹陷的中段。 第五章用预算约束区分必需品与奢侈品,说明陈述偏好预测力有限,并指出排卵周期偏好转移的两篇元分析结论相反、大样本研究未见转移,因此「双轨」只作角色通用、可以证伪的情境假说。第六章讲纳什解、卡莱-斯莫罗金斯基解与外部选项原则,被索取方只用如实披露、外部选项、底线与退出权;稳定匹配中「全都要型」单身比例达 0.780。第七章对读《诗经·卫风·氓》、《礼记·昏义》与达尔文 1871 年的论述,并给出四件套蒸馏规格。第八章给出试算工具、检查器规格、周度台账与预注册检验。书中以标注框区分证据、假说与原典,共含 22 段程序。 Chapter answers: - Ch.1 Roles, Not Sexes: Asymmetric Parental Investment, Acceptance Thresholds and Overlapping Effect Sizes — Chapter 1 uses optimal stopping to show that choosiness is a parameter of roles: as fixed investment rises from zero to 4.0, the acceptance threshold rises from 0.8000 to 0.9106, and no sex appears in the formula; at an effect size of d = 1.00 two distributions still overlap by 0.617. (https://5dlibrary.com/answers/en/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c01/) 第1章 角色而非性别:亲代投资不对称、接受阈值与效应量重叠 — 第一章用最优停止说明挑剔是角色参数:固定投资从零升到 4.0,接受阈值从 0.8000 升到 0.9106,公式中没有性别;效应量 d = 1.00 时两个分布仍重叠 0.617。 (https://5dlibrary.com/answers/zh/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c01/) - Ch.2 Berkson Selection: Where the Observed Trade-off Comes From — Chapter 2 is the core of the book: starting from two independent traits and keeping only the top 10 % by overall appeal, the correlation becomes −0.708, and market clearing produces the same trade-off. (https://5dlibrary.com/answers/en/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c02/) 第2章 伯克森筛选:观察到的此消彼长从哪里来 — 第二章是全书核心:从两个独立特质出发,只保留总吸引力前 10 % 的人,相关系数就变成 −0.708,市场出清同样制造此消彼长。 (https://5dlibrary.com/answers/zh/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c02/) - Ch.3 The Collapse of the Feasible Set: The "Impossible Triangle" as a Joint-Threshold Probability — Chapter 3 writes the "impossible triangle" as a feasible probability under joint thresholds: requiring the top 20 % on three traits leaves a feasible share of 0.00920, the first value below one in a hundred; relaxing one threshold multiplies the feasible probability by 5.57, while meeting twice as many people only doubles it. (https://5dlibrary.com/answers/en/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c03/) 第3章 可行集塌缩:把「不可能三角」写成联合阈值概率 — 第三章把「不可能三角」写成联合阈值下的可行概率:三项都要求前 20 % 时可行比例为 0.00920,首次跌破百分之一;放宽一项门槛使可行概率乘以 5.57,多见一倍的人只乘以二。 (https://5dlibrary.com/answers/zh/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c03/) - Ch.4 The Pareto Front: Why Weighted Scoring Misses the Part That Bends Inward — With no floors all 300 people pass, 9 of them on the front. With both floors at 0.5, 43 people remain, a share of 0.143, 3 of them on the front. With both at 0.6, 14 remain, a share of 0.047. With the floors at 0.7 and 0.7, only 2 people remain, a share of 0.007, and only 1 of them is on the front. With both floors at 0.8 nobody remains. (https://5dlibrary.com/answers/en/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c04/) 第4章 帕累托前沿:加权打分为什么找不到凹进去的那一段 — 没有底线时 300 人全部通过,其中 9 位在前沿上。两项底线都设 0.5,剩下 43 人,占 0.143,其中 3 位在前沿上。都设 0.6,剩 14 人,占 0.047。底线升到 0.7 与 0.7,只剩 2 人,占 0.007,前沿上只剩 1 位。两项都设 0.8 时,无人剩下。 (https://5dlibrary.com/answers/zh/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c04/) - Ch.5 Budget Constraints and Stated Preferences: Necessities, Luxuries and a Test of the "Dual-Track" Hypothesis — The program simulates 80 small studies of 20 to 60 participants each. Each study has 4 measures it could report and reports only the one with the largest effect. When the true shift is zero, the pooled effect over all studies is 0.162, and over the "published" studies alone it is 0.189. With a true shift of 0.05 the two values are 0.216 and 0.247; with 0.10 they are 0.272 and 0.325. Conversely, a single preregistered study of 584 participants has power 0.998 to detect a shift of 0.20, 0.676 for a shift of 0.10, and only 0.227 for a shift of 0.05. (https://5dlibrary.com/answers/en/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c05/) 第5章 预算约束与陈述偏好:必需品、奢侈品与「双轨」假说的检验 — 程序模拟 80 项小样本研究,每项样本 20 到 60 人,每项有 4 种可报告的测量,只报告其中效应最大的那一种。当真实转移量为零时,全部研究的汇总效应是 0.162,只看「发表」的研究是 0.189;真实转移为 0.05 时,两者分别是 0.216 与 0.247;为 0.10 时是 0.272 与 0.325。反过来,一项预先登记、样本 584 人的研究,对 0.20 的转移有 0.998 的检验功效,对 0.10 有 0.676,对 0.05 只有 0.227。 (https://5dlibrary.com/answers/zh/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c05/) - Ch.6 Bargaining Boundaries and Stable Matching: Outside Options, Threat Points and the Single Rate of "Wanting It All" — Chapter 6 covers the Nash and Kalai-Smorodinsky solutions and the outside option principle; the party asked to invest uses only truthful disclosure, outside options, its own floor and the right to leave, and in stable matching the single share of the all-max type reaches 0.780. (https://5dlibrary.com/answers/en/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c06/) 第6章 议价边界与稳定匹配:外部选项、威胁点与「全都要」的单身率 — 第六章讲纳什解、卡莱-斯莫罗金斯基解与外部选项原则,被索取方只用如实披露、外部选项、底线与退出权;稳定匹配中「全都要型」单身比例达 0.780。 (https://5dlibrary.com/answers/zh/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c06/) - Ch.7 Classical Sources East and West: Exit Costs in "Mang", the Multi-Objective Contract of the "Meaning of the Marriage Rite" and Darwin in 1871 — Chapter 7 reads the ode "Mang", the chapter on the meaning of the marriage rite and Darwin's account of 1871 side by side and gives the distillation spec for the four kits. (https://5dlibrary.com/answers/en/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c07/) 第7章 东西方原典考辨:《氓》的退出成本、《昏义》的多目标契约与达尔文 1871 — 第七章对读《诗经·卫风·氓》、《礼记·昏义》与达尔文 1871 年的论述,并给出四件套蒸馏规格。 (https://5dlibrary.com/answers/zh/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c07/) - Ch.8 Field Tools and Checker Specifications: Calculator, Checkers, Ledger and Preregistered Tests — Chapter 8 gives the calculator, the checker specifications, a weekly ledger and preregistered tests. Marked boxes separate evidence, hypotheses and classical sources; the book contains 22 programs. (https://5dlibrary.com/answers/en/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c08/) 第8章 实战工具与检查器规格:试算器、检查器、台账与预注册检验 — 第八章给出试算工具、检查器规格、周度台账与预注册检验。书中以标注框区分证据、假说与原典,共含 22 段程序。 (https://5dlibrary.com/answers/zh/dual-mating-multi-objective-conflict-parental-investment-pareto-frontier/c08/) ## 5DL-B-422 · Dynamic Time-Inconsistency, Asymmetric Freeroll Reneging and Collateralised Anti-Hold-Up Contracts 中文书名:动态时间不一致博弈、收益私有化亏损转嫁与反掀桌质押契约母典 DOI: pending Cite: Fan, Y. (2026). Dynamic Time-Inconsistency, Asymmetric Freeroll Reneging and Collateralised Anti-Hold-Up Contracts (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book is not legal advice and not individual financial or business advice, every example is synthetic, and all prices are unit-free. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. Abstract: The popular view of promises holds that a verbal promise will certainly be broken once payoffs turn against the promisor, that more collateral is always safer, and that keeping gains while rewriting the rules after losses is a flaw of character. This book rewrites all three as propositions that data could refute. After checking the sources it corrects the founding blueprint: Barro and Gordon's paper on reputation appeared in the Journal of Monetary Economics; Strotz's paper is dated 1956 by common practice; the experimental paper on self-imposed deadlines listed in the blueprint has been retracted and is replaced by field experiments on commitment savings and commitment contracts; and symmetric sunk collateral becomes collateral sized to the temptation to breach and held by a third party or released by a self-executing rule. Its parameters are illustrative; all calculations are generated by its programs; the book is not legal advice and not individual financial or business advice, every example is synthetic, and all prices are unit-free. Chapter 1 uses the toy model of Kydland and Prescott to show that breach is most tempting in the period it happens, while lost credibility costs more in every period after; under grim trigger, reputation keeps the rule only when the discount factor exceeds 0.333. Chapter 2 reproduces preference reversal and bounds a forfeitable deposit from below and above. Chapter 3 holds the core arithmetic: in a one-off relationship the probability of breach falls with collateral plus relationship value, and keeping it at 0.10 needs collateral of 7.63; in a repeated relationship with common payoffs the grim-trigger bar is 0.500 and the risk-dominance bar 0.600; with a public end date the largest bond equals the last-round temptation. Chapter 4 writes limited liability as a call option, whose holder gains 10.94 as volatility rises, and shows that a loss-averse partner refuses to join once the probability of rules being rewritten after the fact exceeds 0.461. Chapter 5 gives the under-investment caused by hold-up and the bounds on collateral, and shows that collateral the counterparty can seize unilaterally creates reverse hold-up: an honest promisor posts at most 7.75 under unilateral holding but up to 150.00 under third-party escrow. Chapter 6 compares escrow milestones with paying first or last, and shows that when monitoring errs, expulsion at the first offence wrongly expels 0.506 of steady members while graduated sanctions keep compliance at 0.963. Chapter 7 examines the Zhou-Zheng hostage exchange in the Zuo Commentary, the pole at the south gate in the Records of the Grand Historian, Hobbes's Leviathan and the Odyssey, and gives the distillation spec for the four kits. Chapter 8 gives the commitment ledger, the distilled Anti-Hold-Up Collateral and Reneging Threshold Calculator, checker specification tables and a pre-registered test needing 199 deals per group. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:关于承诺,流行说法是:口头承诺在利益倒挂时必然违约,质押越多越安全,「赢了归我、输了改规则」是某类人的品格缺陷。本书把这三种说法改写成可以被数据推翻的命题。核对文献后,本书修正了立项蓝图:巴罗与戈登讨论声誉的论文刊于《货币经济学杂志》;斯特罗茨论文按通行引法记为 1956 年;蓝图所列自设截止的实验论文已撤稿,改用承诺储蓄与承诺合同的田野实验;「对称沉没质押」改为按违约诱惑定额、由第三方托管或规则自动执行的质押。书中参数为示意;全部计算由书中程序生成;本书不是法律意见,也不是个人化财务或经营建议,所有例子都是合成情境,价格一律为无币种示意单位。 第一章用基德兰德与普雷斯科特的玩具模型说明,背约那一期最诱人,而可信度一旦丧失,每期都要多付代价;冷酷触发下贴现因子高于 0.333 才能靠声誉守住规则。第二章复现偏好反转,并给出可没收定金的上下界。第三章是核心算术:单次关系里背约概率随质押与关系价值之和下降,要把背约概率压到 0.10 需质押 7.63;重复关系里常用收益下冷酷触发门槛 0.500,风险占优门槛 0.600;有公开终点时最大保证金等于末轮诱惑。第四章把有限责任写成看涨期权,波动上升使持有方多得 10.94,事后改规则的概率超过 0.461 时损失厌恶的一方拒绝合作。 第五章给出敲竹杠下的投资不足与质押上下界,并说明可被单方没收的质押会制造反向敲竹杠:诚实一方在单方持有下最多只肯交 7.75,第三方托管下可到 150.00。第六章比较托管里程碑与先付后付,并说明监察有误时一次性除名会误除 0.506 的守约者,分级制裁合规率仍有 0.963。第七章考辨《左传》周郑交质、《史记》徙木立信、霍布斯《利维坦》与《奥德赛》,并给出四件套蒸馏规格。第八章给出承诺履约台账、蒸馏小工具「反掀桌质押与违约阈值试算器」、检查器规格表与预注册检验,每组需要 199 笔合作。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Rules and Discretion: Why Optimal Plans Undo Themselves — Chapter 1 uses the toy model of Kydland and Prescott to show that breach is most tempting in the period it happens, while lost credibility costs more in every period after; under grim trigger, reputation keeps the rule only when the discount factor exceeds 0.333. (https://5dlibrary.com/answers/en/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c01/) 第1章 规则与相机抉择:最优计划为什么会自我推翻 — 第一章用基德兰德与普雷斯科特的玩具模型说明,背约那一期最诱人,而可信度一旦丧失,每期都要多付代价;冷酷触发下贴现因子高于 0.333 才能靠声誉守住规则。 (https://5dlibrary.com/answers/zh/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c01/) - Ch.2 Present Bias: Preference Reversal and Self-Imposed Deposits — Present bias makes one person choose differently at different dates. Hyperbolic discounting reverses the preference at 5.00 days; quasi-hyperbolic discounting reverses it the moment the smaller reward lands today. Sophisticated agents, who know they will delay, lose far less than naive ones; a forfeitable deposit can restrain the present self, but it has both a lower and an upper bound, and once the accident probability exceeds 0.667 no deposit is worth taking. (https://5dlibrary.com/answers/en/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c02/) 第2章 现时偏向:个人层面的偏好反转与自设质押 — 现时偏向让同一个人在不同时点给出相反的选择。双曲贴现在 5.00 天处反转偏好,准双曲贴现只要小奖励落到今天就反转。老练者知道自己会拖,损失远小于天真者;可没收定金能约束当下的自己,但定金有下界也有上界,意外概率超过 0.667 时任何定金都不划算。 (https://5dlibrary.com/answers/zh/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c02/) - Ch.3 The Reneging Threshold: Incentive Arithmetic and the Critical Discount Factor — Chapter 3 holds the core arithmetic: in a one-off relationship the probability of breach falls with collateral plus relationship value, and keeping it at 0.10 needs collateral of 7.63; in a repeated relationship with common payoffs the grim-trigger bar is 0.500 and the risk-dominance bar 0.600; with a public end date the largest bond equals the last-round temptation. (https://5dlibrary.com/answers/en/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c03/) 第3章 违约阈值:激励相容破裂的算术与临界贴现因子 — 第三章是核心算术:单次关系里背约概率随质押与关系价值之和下降,要把背约概率压到 0.10 需质押 7.63;重复关系里常用收益下冷酷触发门槛 0.500,风险占优门槛 0.600;有公开终点时最大保证金等于末轮诱惑。 (https://5dlibrary.com/answers/zh/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c03/) - Ch.4 Freerolls: Convex Payoffs under Limited Liability and Ex Post Rule Rewriting — Chapter 4 writes limited liability as a call option, whose holder gains 10.94 as volatility rises, and shows that a loss-averse partner refuses to join once the probability of rules being rewritten after the fact exceeds 0.461. (https://5dlibrary.com/answers/en/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c04/) 第4章 收益私有化亏损转嫁:有限责任的凸性报酬与事后改规则 — 第四章把有限责任写成看涨期权,波动上升使持有方多得 10.94,事后改规则的概率超过 0.461 时损失厌恶的一方拒绝合作。 (https://5dlibrary.com/answers/zh/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c04/) - Ch.5 Hold-Up: Specific Investment, Hostages and the Bounds on Collateral — Chapter 5 gives the under-investment caused by hold-up and the bounds on collateral, and shows that collateral the counterparty can seize unilaterally creates reverse hold-up: an honest promisor posts at most 7.75 under unilateral holding but up to 150.00 under third-party escrow. (https://5dlibrary.com/answers/en/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c05/) 第5章 掀桌即敲竹杠:专用性投资、人质与质押额的上下界 — 第五章给出敲竹杠下的投资不足与质押上下界,并说明可被单方没收的质押会制造反向敲竹杠:诚实一方在单方持有下最多只肯交 7.75,第三方托管下可到 150.00。 (https://5dlibrary.com/answers/zh/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c05/) - Ch.6 Escrow, Automatic Execution and Graduated Sanctions — Chapter 6 compares escrow milestones with paying first or last, and shows that when monitoring errs, expulsion at the first offence wrongly expels 0.506 of steady members while graduated sanctions keep compliance at 0.963. (https://5dlibrary.com/answers/en/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c06/) 第6章 托管、自动执行与分级制裁 — 第六章比较托管里程碑与先付后付,并说明监察有误时一次性除名会误除 0.506 的守约者,分级制裁合规率仍有 0.963。 (https://5dlibrary.com/answers/zh/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c06/) - Ch.7 Classical Sources East and West: the Zhou-Zheng Hostage Exchange, the Pole at the South Gate and Leviathan's Sword — Each of the four sources grasps one face of the commitment problem, and each leaves something out. Hostages without inner good faith do limited work and lose force quickly when outside options improve; a small payment builds reputation, but its signal fails once the probability of imitation exceeds 0.286; covenants without enforcement are only words, yet the enforcer need not be a sword; and the key to self-binding is to restrict action while keeping information. (https://5dlibrary.com/answers/en/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c07/) 第7章 东西方原典考辨:周郑交质、徙木立信与利维坦之剑 — 四处原典都抓住了承诺问题的一个侧面,又都说得不完整。人质在没有内在之信时作用有限,并随外部选项改善迅速失效;小额兑现能建立声誉,但模仿概率超过 0.286 时信号失效;没有执行的信约只是言辞,而执行力量不必只有刀剑;自我约束的关键是限制行动、保留信息。 (https://5dlibrary.com/answers/zh/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c07/) - Ch.8 Closing the Loop: The Commitment Ledger, Field Tools and Checker Specifications — Chapter 8 gives the commitment ledger, the distilled Anti-Hold-Up Collateral and Reneging Threshold Calculator, checker specification tables and a pre-registered test needing 199 deals per group. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c08/) 第8章 实证闭环:承诺履约台账、实战工具与检查器规格 — 第八章给出承诺履约台账、蒸馏小工具「反掀桌质押与违约阈值试算器」、检查器规格表与预注册检验,每组需要 199 笔合作。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/dynamic-time-inconsistency-freeroll-reneging-collateralised-anti-hold-up-contracts/c08/) ## 5DL-B-423 · Unbundling Paternal Selection from Romantic Commitment: Deferred-Promise Options, Avoidant Autonomy and Reciprocal Stopping Rules 中文书名:亲职基因优选与排他恋爱解耦、远期空头期权钩子与对等底线停时母典 DOI: pending Cite: Fan, Y. (2026). Unbundling Paternal Selection from Romantic Commitment: Deferred-Promise Options, Avoidant Autonomy and Reciprocal Stopping Rules (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations are generated by the book's programs; the book is not counselling, relationship advice or clinical judgement, and makes no divinatory inference. Marked boxes separate evidence, runnable calculations (21 programs), classical sources and the author's hypotheses. Abstract: A popular belief holds that once someone has said "one day we will build a family", the relationship is settled; another holds that anyone who talks about the distant future but declines exclusivity now will inevitably default. This book refutes the premise both share, that the value of a deferred promise can be read from its wording. That value depends on who holds the choice, whether a premium was paid and how uncertain the future is. "Parental selection" here means the criteria either party uses to screen a future co-parent; the book is gender-neutral and refers to roles. After checking the literature it corrects its blueprint: the option model shows a structural incentive, not evidence of motive; contact with back burners is unrelated to commitment among people in committed relationships and is therefore a weak signal; adult attachment is a continuous dimension, so no one can be labelled avoidant from a single behaviour; cheap talk loses information as interests diverge rather than carrying none; and the thirty-seven per cent rule of the secretary problem fails when recall, non-identical options and costly observation enter. Parameters are illustrative; all calculations are generated by the book's programs; the book is not counselling, relationship advice or clinical judgement, and makes no divinatory inference. Chapter 1 uses a binomial tree to show that when volatility rises from 0.10 to 0.60 the value of a zero-premium distant title to its holder grows 4.44 times. Chapter 2 gives the separating condition for costly signals and the partition equilibria of cheap talk. Chapter 3 shows that at a base rate of 0.16 a label based on one behaviour is wrong 0.686 of the time. Chapter 4 derives sequential stopping boundaries, against 15.2 weeks for a fixed-length test. Chapter 5 shows why the thirty-seven per cent rule fails and replaces it with a reservation threshold set by the cost of observation. Chapter 6 measures the loss from continuing on sunk costs, 2.02 per decision on average, and explains lock-in with the investment-model meta-analysis. Chapter 7 examines the ode "Mang" of the Book of Songs, You Ruo's saying in the Analects that words close to what is right can be made good, and Aristotle's story of Thales hiring olive presses, and gives the distillation spec for the four kits. Chapter 8 gives the reciprocal floor rule (at a threshold of 0.60 and three weeks, false alarms of 0.081 in reciprocal relationships and an average trigger in week 3.64 in one-sided ones), the distilled Deferred-Promise Option and Reciprocal Stopping Checker, checker specifications, ledger fields and a pre-registered test. Every rule binds only the reader's own investment, communication and exit; the other party may decline to commit. Marked boxes separate evidence, runnable calculations (21 programs), classical sources and the author's hypotheses. 摘要:流行说法认为,只要对方说过「以后一定和你成家」,关系就已有着落;另一种说法认为,凡是只谈远期、不肯当下排他的人必然违约。本书证伪两者共同的前提:远期承诺的价值可以从措辞读出。远期承诺的价值取决于谁持有选择权、是否付过权利金、未来有多不确定。书中「亲职选择」指任一方对未来共同养育伙伴的筛选标准,全书性别中立,以角色指称。核对文献后,本书修正立项蓝图:期权模型只给出结构激励而非动机证据;与后备关系保持联系在已有固定关系者中与承诺水平无关,所以是弱信号;成人依恋是连续维度,不能凭单一行为贴「回避型」标签;廉价话语的信息量随利益分歧递减而非一概为零;秘书问题的三成七规则在可回头、非同分布、观察有成本时不适用。书中参数为示意;全部计算由书中程序生成;本书不是心理咨询、婚恋咨询或临床判断,也不做任何命理推导。 第一章用二叉树说明,波动率从 0.10 升到 0.60,零权利金远期头衔对持有方的价值升为 4.44 倍。第二章给出有成本信号的分离条件与廉价话语的分段均衡。第三章说明,在基础率 0.16 下凭单一行为贴标签,错误率为 0.686。第四章用序贯概率比检验给出停止边界,固定周数检验需 15.2 周。第五章说明三成七规则的失效,并以观察成本决定的保留阈值代替。第六章量化按沉没成本继续的损失,平均每次 2.02,并用投资模型元分析解释锁定。 第七章考辨《诗经·卫风·氓》「信誓旦旦,不思其反」、《论语·学而》有子「信近于义,言可复也」与亚里士多德《政治学》泰勒斯预租榨油坊的故事,并给出四件套蒸馏规格。第八章给出对等底线规则(阈值 0.60、连续三周时对等关系误报 0.081,偏向关系平均第 3.64 周触发)、蒸馏小工具「远期承诺期权估值与对等停时判别器」、检查器规格、台账字段与预注册检验。一切规则只约束读者自己的投入、沟通与退出,对方有不承诺的自主权。书中以标注框区分证据、可运行计算(21 段程序)、原典与作者假说。 Chapter answers: - Ch.1 A Deferred Promise Is an Option: Who Holds It, Who Writes It, Who Was Paid the Premium — Chapter 1 uses a binomial tree to show that when volatility rises from 0.10 to 0.60 the value of a zero-premium distant title to its holder grows 4.44 times. (https://5dlibrary.com/answers/en/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c01/) 第1章 远期承诺是一张期权:谁持有,谁卖出,谁收了权利金 — 第一章用二叉树说明,波动率从 0.10 升到 0.60,零权利金远期头衔对持有方的价值升为 4.44 倍。 (https://5dlibrary.com/answers/zh/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c01/) - Ch.2 Costly Signals and Cheap Talk: How Much Information Is Left in a Costless Title — With a bias of 0.005, talk can sustain 10 intervals and removes 0.980 of the prior variance. With a bias of 0.01 it sustains 7 intervals and removes 0.960. With 0.05 it sustains 3 intervals and removes 0.809. At 0.10 only 2 intervals remain, removing 0.630, and at 0.20 there are still 2 intervals but they remove only 0.270. Once the bias reaches 0.25 or more, there is a single interval: the receiver's loss stays at 0.08333 and the variance removed by talk is 0.000. (https://5dlibrary.com/answers/en/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c02/) 第2章 有成本的信号与廉价话语:零成本头衔还剩多少信息 — 偏差为 0.005 时,话语能切出 10 段,消除先验方差的 0.980;偏差 0.01 时切 7 段,消除 0.960;偏差 0.05 时切 3 段,消除 0.809;偏差 0.10 时只剩 2 段,消除 0.630;偏差 0.20 时仍是 2 段,只消除 0.270。偏差达到 0.25 或以上,段数是 1,接收方损失停在 0.08333,话语消除的方差是 0.000。 (https://5dlibrary.com/answers/zh/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c02/) - Ch.3 Base Rates and Dimensions: The Error Rate of Labelling from a Single Behaviour — Chapter 3 shows that at a base rate of 0.16 a label based on one behaviour is wrong 0.686 of the time. (https://5dlibrary.com/answers/en/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c03/) 第3章 基础率与维度:凭单一行为贴标签的误判率 — 第三章说明,在基础率 0.16 下凭单一行为贴标签,错误率为 0.686。 (https://5dlibrary.com/answers/zh/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c03/) - Ch.4 The Sequential Probability Ratio Test: How Many Weeks of Records Are Enough to Stop — Chapter 4 derives sequential stopping boundaries, against 15.2 weeks for a fixed-length test. (https://5dlibrary.com/answers/en/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c04/) 第4章 序贯概率比检验:要看多少周的记录才足以停下 — 第四章用序贯概率比检验给出停止边界,固定周数检验需 15.2 周。 (https://5dlibrary.com/answers/zh/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c04/) - Ch.5 Why the Secretary Problem's Thirty-Seven Per Cent Rule Fails Here — The secretary problem's thirty-seven per cent rule depends on six assumptions, and the situations of this book violate almost all of them. Once the payoff is made continuous, observation is given a cost, recall is allowed or a trend is introduced, the optimal share to pass over ranges from 0.05 to 0.60. A better-suited replacement is a reservation threshold set by the cost of observation: the higher the cost, the lower the threshold and the earlier one stops. (https://5dlibrary.com/answers/en/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c05/) 第5章 秘书问题的三成七规则为什么在这里失效 — 秘书问题的三成七规则依赖六条假设,本书的局面几乎每条都不满足。改成连续收益、加入观察成本、允许回头或引入趋势后,最优放过比例从 0.05 到 0.60 不等。更合适的替代是由观察成本决定的保留阈值:成本越高,门槛越低,停得越早。 (https://5dlibrary.com/answers/zh/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c05/) - Ch.6 Sunk Costs and the Investment Model: Why Stopping Gets Harder the More One Has Invested — Chapter 6 measures the loss from continuing on sunk costs, 2.02 per decision on average, and explains lock-in with the investment-model meta-analysis. (https://5dlibrary.com/answers/en/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c06/) 第6章 沉没成本与投资模型:为什么投入越多越难止损 — 第六章量化按沉没成本继续的损失,平均每次 2.02,并用投资模型元分析解释锁定。 (https://5dlibrary.com/answers/zh/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c06/) - Ch.7 Classical Sources: the Distant Oath of "Mang", You Ruo on Words That Can Be Made Good, and Thales's Olive Presses — Of the 8 promises with both a date and a cost, 7 were kept, giving a posterior mean of 0.800 and a ninety per cent interval from 0.571 to 0.959. Of the 7 promises with a date only, 4 were kept: posterior mean 0.556, interval 0.289 to 0.807. There are only 3 promises with a cost but no date, of which 2 were kept: posterior mean 0.600, with a very wide interval from 0.249 to 0.902. Of the 12 distant titles with neither a date nor a cost, only 2 were kept: posterior mean 0.214, interval 0.066 to 0.410. (https://5dlibrary.com/answers/en/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c07/) 第7章 原典考辨:《氓》之远期誓言、有子论「言可复」与泰勒斯预租榨油坊 — 带日期且附成本的 8 条承诺兑现了 7 条,后验均值 0.800,九成区间 0.571 到 0.959。只带日期的 7 条兑现 4 条,后验均值 0.556,区间 0.289 到 0.807。附成本无日期的只有 3 条,兑现 2 条,后验均值 0.600,区间很宽,从 0.249 到 0.902。既无日期也无成本的远期头衔 12 条,只兑现 2 条,后验均值 0.214,区间 0.066 到 0.410。 (https://5dlibrary.com/answers/zh/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c07/) - Ch.8 Reciprocal Floor Stopping: The Empirical Loop, Field Tools and Checker Specifications — Chapter 8 gives the reciprocal floor rule (at a threshold of 0.60 and three weeks, false alarms of 0.081 in reciprocal relationships and an average trigger in week 3.64 in one-sided ones), the distilled Deferred-Promise Option and Reciprocal Stopping Checker, checker specifications, ledger fields and a pre-registered test. Every rule binds only the reader's own investment, communication and exit; the other party may decline to commit. Marked boxes separate evidence, runnable calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c08/) 第8章 对等底线停时:实证闭环、实战工具与检查器规格 — 第八章给出对等底线规则(阈值 0.60、连续三周时对等关系误报 0.081,偏向关系平均第 3.64 周触发)、蒸馏小工具「远期承诺期权估值与对等停时判别器」、检查器规格、台账字段与预注册检验。一切规则只约束读者自己的投入、沟通与退出,对方有不承诺的自主权。书中以标注框区分证据、可运行计算(21 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/unbundling-parental-selection-romantic-commitment-deferred-promise-options-stopping-rules/c08/) ## 5DL-B-424 · Asymmetric Intimacy Thresholds: Male Marginal-Cost Accounting versus Female Contextual-Tension Arousal in Mating Economics 中文书名:亲密决策阈值不对称:男性边际成本核算与女性情境张力唤醒的演化经济学母典 DOI: pending Cite: Fan, Y. (2026). Asymmetric Intimacy Thresholds: Male Marginal-Cost Accounting versus Female Contextual-Tension Arousal in Mating Economics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative or synthetic; all calculations are generated by the book's programs; the book is not counselling, relationship advice or clinical judgement, and not individual business advice. Chapter 3 shows in a synthetic decomposition that perceived safety, experience and stigma carry 0.842 of the raw gap. Abstract: A popular claim holds that the sexes approach intimacy with entirely different decision functions: one counts opportunity cost and risk, the other ignores low cost and responds only to scarcity, unpredictability and tension. This book keeps that title and tests it. The title is a stylisation of population averages: every person is a mixture of a marginal-cost function and a contextual-elements function, and the two sexes overlap heavily; the asymmetry comes mainly from different cost terms. Two corrections to the blueprint follow: Conley (2011) showed that the acceptance gap shrinks substantially once the proposer's perceived safety, expected experience and stigma are controlled, so "indifferent to low cost" becomes "dominated by different cost terms"; and the evidence for "scarcity, unpredictability and high tension" is thin and is treated only critically. Consent is the floor: arousal, liking and single cues are not consent, which must be explicit, current and revocable. Parameters are illustrative or synthetic; all calculations are generated by the book's programs; the book is not counselling, relationship advice or clinical judgement, and not individual business advice. Chapter 1 converts effect sizes into overlap: at the meta-analytic sex-drive effect of 0.69 the two distributions share an area of 0.730, 0.313 of random pairs point the other way, and guessing a person from a sex label is right only 0.635 of the time. Chapter 2 uses one threshold formula to reproduce "a price cut moves only one side": halving time and resources raises one composition of cost from 0.198 to 0.500 and the other only from 0.083 to 0.119, while halving safety and stigma lifts the latter to 0.354. Chapter 3 shows in a synthetic decomposition that perceived safety, experience and stigma carry 0.842 of the raw gap. In Chapter 4 a chain of states keeps an exit at every step; even with a perfect context the path of declining keeps 0.509; safety is a gate, and with safety at 0.3 or less the gated score is at most 0.300. Chapter 5 examines the evidence on uncertainty and misattributed arousal. Chapter 6 uses signal detection to show that a single friendly cue at a base rate of 0.10 yields a posterior of only 0.163, that stacking correlated cues cannot reach the bar for action. Chapter 7 examines "food, drink, man and woman" in the Book of Rites, Gaozi's saying in the Mencius and Darwin's phrase of 1871, "with the rarest exception", which would need an effect size of 3.29. Chapter 8 gives the distilled Dual-Function Threshold and Context-Safety Calculator, checker specifications grouped by four kits, ledgers and a pre-registered test (2,046 per cell in the conservative plan); an illustrative model shows invented urgency leaving two kinds of buyer at 0.888 and 0.777 of the honest page's net sales after twelve months. The book contains 22 programs. 摘要:两性网络话语里流行一种说法:男女推进亲密关系时用的是截然不同的决策函数,一方只算机会成本与风险,另一方对低成本无感,只有稀缺、不可预测与高情绪张力才能跨过阈值。本书逐字保留这一书名,把它放上检验台。结论是:书名是总体平均的程式化,每个人都是「边际成本核算」与「情境要素」两类函数的混合,两性分布高度重叠;阈值不对称主要来自成本项构成不同,双方都在算账。核对原文后,本书写明两处蓝图修正:Conley 2011 显示控制提议者的安全感、预期体验与污名代价后,接受率差异大幅缩小,「对低成本无感」应改写为「主导成本项不同」;「稀缺、不可预测与高张力」证据薄弱,只作批判性评述。同意是全书底线:唤起、好感与任何单一情境线索都不等于同意,同意须明确、当下、可随时撤回。书中参数均为示意或合成;全部计算由书中程序生成;本书不是心理咨询、婚恋咨询或临床判断,也不是个人化经营建议。 第一章由效应量算重叠率:元分析报告的性驱力效应 0.69 下,两条分布共有面积 0.730,随机配对中 0.313 方向相反;用性别标签猜一个人只对 0.635。第二章用同一条阈值公式复现「降价只打动一方」:时间与资源减半使一种成本构成的接受概率从 0.198 升到 0.500,另一种只从 0.083 升到 0.119,而安全与污名减半使后者升到 0.354。第三章在合成拆解里,感知安全、体验与污名承担原始差距的 0.842,并说明零接受率会把优势比放大到失真。第四章写双重控制与回应式意愿,状态链在每一步都留着拒绝出口,情境满分时拒绝路径仍保有 0.509;安全是闸门,安全不超过 0.3 时闸门分最高只有 0.300。 第五章审查「不确定提升好感」与唤醒误归因研究的样本、重复与伦理问题,说明小样本显著结果会成倍高估效应。第六章用信号检测说明单一友好线索在基础率 0.10 下后验只有 0.163,相关线索叠加也够不着行动门槛,只有明确、可撤回的同意能跨过。第七章考辨《礼记·礼运》「饮食男女」、告子「食色性也」与达尔文 1871 年「罕有例外」一句,后者需要效应量 3.29,并对照 Bateman 实验的重复检验。第八章给出蒸馏小工具「亲密决策双函数阈值与情境安全试算器」、按四件套分组的检查器规格、台账与预注册检验(保守方案每格 2,046 人);示意模型显示虚假紧迫十二个月后使两类买家净成交只剩如实页面的 0.888 与 0.777。全书含 22 段程序。 Chapter answers: - Ch.1 The Title Is a Stylisation: Mixtures of Two Functions and Overlapping Distributions — Chapter 1 converts effect sizes into overlap: at the meta-analytic sex-drive effect of 0.69 the two distributions share an area of 0.730, 0.313 of random pairs point the other way, and guessing a person from a sex label is right only 0.635 of the time. (https://5dlibrary.com/answers/en/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c01/) 第1章 书名是程式化写法:双函数混合与两性分布的重叠 — 第一章由效应量算重叠率:元分析报告的性驱力效应 0.69 下,两条分布共有面积 0.730,随机配对中 0.313 方向相反;用性别标签猜一个人只对 0.635。 (https://5dlibrary.com/answers/zh/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c01/) - Ch.2 Marginal-Cost Accounting: A Threshold Model, Opportunity Cost and the Limits of Sexual Economics — Chapter 2 uses one threshold formula to reproduce "a price cut moves only one side": halving time and resources raises one composition of cost from 0.198 to 0.500 and the other only from 0.083 to 0.119, while halving safety and stigma lifts the latter to 0.354. (https://5dlibrary.com/answers/en/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c02/) 第2章 边际成本核算:阈值模型、机会成本与性经济学理论的边界 — 第二章用同一条阈值公式复现「降价只打动一方」:时间与资源减半使一种成本构成的接受概率从 0.198 升到 0.500,另一种只从 0.083 升到 0.119,而安全与污名减半使后者升到 0.354。 (https://5dlibrary.com/answers/zh/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c02/) - Ch.3 Blueprint Correction One: What the Acceptance Gap Contains — Chapter 3 shows in a synthetic decomposition that perceived safety, experience and stigma carry 0.842 of the raw gap. In Chapter 4 a chain of states keeps an exit at every step; even with a perfect context the path of declining keeps 0.509; safety is a gate, and with safety at 0.3 or less the gated score is at most 0.300. (https://5dlibrary.com/answers/en/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c03/) 第3章 蓝图修正一:接受率差距里装着安全、体验与污名成本 — 第三章在合成拆解里,感知安全、体验与污名承担原始差距的 0.842,并说明零接受率会把优势比放大到失真。 (https://5dlibrary.com/answers/zh/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c03/) - Ch.4 The Contextual-Elements Function: Dual Control, Responsive Willingness and Safety as a Gate — Part A lists the mean willingness of the two groups in four contexts, together with two measures of explanatory power: the share of the variance in willingness explained by the inhibition weight and the share explained by the group label. When concern is as low as 0.2, mean willingness in the two groups is 0.686 and 0.679, almost identical; the inhibition weight explains only 0.043 and the group label only 0.0008. When concern rises to 1.0, the groups move to 0.497 and 0.460; when it rises to 2.0, they move to 0.305 and 0.258, the explanatory power of the inhibition weight climbs to 0.461, and that of the group label is only 0.0126. With a stronger cue of 3.0, the groups rise to 0.491 and 0.434. (https://5dlibrary.com/answers/en/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c04/) 第4章 情境要素函数:双重控制、回应式意愿与安全闸门 — A 段列出四种情境下两组平均意愿,以及两种解释力:抑制权重能解释意愿方差的份额,与组标签能解释的份额。顾虑低到 0.2 时,两组平均意愿是 0.686 与 0.679,几乎相同;抑制权重只解释 0.043,组标签只解释 0.0008。顾虑升到 1.0,两组变为 0.497 与 0.460;顾虑升到 2.0,变为 0.305 与 0.258,抑制权重的解释力升到 0.461,组标签只有 0.0126。线索加强到 3.0 时,两组升到 0.491 与 0.434。 (https://5dlibrary.com/answers/zh/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c04/) - Ch.5 Blueprint Correction Two: Why the Evidence for "Unpredictability and High Tension" Is Thin — With a true effect of 0.2 and 16 people per group, power is only 0.079, and the average published effect is 0.852, an overestimate by a factor of 4.26. With 50 per group, power is 0.165, the average is 0.496 and the overestimate is 2.48; with 200 per group, power is 0.513 and the overestimate falls to 1.39. When the true effect is zero, a share of 0.024 of studies with 16 per group still come out "significant", with an average reported effect of 0.828. With a true effect of 0.5 and 16 per group, power is 0.291 and the overestimate is 1.84; with 64 per group, power is 0.804 and the overestimate is only 1.12. (https://5dlibrary.com/answers/en/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c05/) 第5章 蓝图修正二:「不可预测与高张力」的证据为何薄弱 — 真实效应为 0.2、每组 16 人时,功效只有 0.079,发表结果的平均效应是 0.852,高估 4.26 倍。每组 50 人时,功效 0.165,平均 0.496,高估 2.48 倍;每组 200 人时,功效 0.513,高估降到 1.39 倍。真实效应为零时,每组 16 人仍有 0.024 的研究会「显著」,平均报告效应 0.828。真实效应为 0.5、每组 16 人时,功效 0.291,高估 1.84 倍;每组 64 人时,功效 0.804,高估只剩 1.12 倍。 (https://5dlibrary.com/answers/zh/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c05/) - Ch.6 Signal Detection: Single Situational Cues, Base Rates and the Bar for Consent — Chapter 6 uses signal detection to show that a single friendly cue at a base rate of 0.10 yields a posterior of only 0.163, that stacking correlated cues cannot reach the bar for action. (https://5dlibrary.com/answers/en/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c06/) 第6章 信号检测:单一情境线索、基础率与同意的门槛 — 第六章用信号检测说明单一友好线索在基础率 0.10 下后验只有 0.163,相关线索叠加也够不着行动门槛,只有明确、可撤回的同意能跨过。 (https://5dlibrary.com/answers/zh/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c06/) - Ch.7 Classical Sources East and West: "Food, Drink, Man and Woman", the Saying of Gaozi and Darwin's "Rarest Exception" — Chapter 7 examines "food, drink, man and woman" in the Book of Rites, Gaozi's saying in the Mencius and Darwin's phrase of 1871, "with the rarest exception", which would need an effect size of 3.29. (https://5dlibrary.com/answers/en/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c07/) 第7章 东西方原典考辨:「饮食男女」、告子之言与达尔文的「罕有例外」 — 第七章考辨《礼记·礼运》「饮食男女」、告子「食色性也」与达尔文 1871 年「罕有例外」一句,后者需要效应量 3.29,并对照 Bateman 实验的重复检验。 (https://5dlibrary.com/answers/zh/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c07/) - Ch.8 Closing the Empirical Loop: Field Tools and Checker Specifications — Chapter 8 gives the distilled Dual-Function Threshold and Context-Safety Calculator, checker specifications grouped by four kits, ledgers and a pre-registered test (2,046 per cell in the conservative plan); an illustrative model shows invented urgency leaving two kinds of buyer at 0.888 and 0.777 of the honest page's net sales after twelve months. The book contains 22 programs. (https://5dlibrary.com/answers/en/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c08/) 第8章 实证闭环:实战工具与检查器规格 — 第八章给出蒸馏小工具「亲密决策双函数阈值与情境安全试算器」、按四件套分组的检查器规格、台账与预注册检验(保守方案每格 2,046 人);示意模型显示虚假紧迫十二个月后使两类买家净成交只剩如实页面的 0.888 与 0.777。全书含 22 段程序。 (https://5dlibrary.com/answers/zh/asymmetric-intimacy-thresholds-marginal-cost-accounting-contextual-arousal-mating-economics/c08/) ## 5DL-B-425 · Shorting Low-Implied-Volatility Windows in Macro Downturns: Delta-Vega Dual Convexity, Bear-Rally Compression and Thirty-Day Hit-Rate Bounds 中文书名:宏观下行期低隐含波动率做空窗口、双希腊值凸性共振与三十日胜率边界母典 DOI: pending Cite: Fan, Y. (2026). Shorting Low-Implied-Volatility Windows in Macro Downturns: Delta-Vega Dual Convexity, Bear-Rally Compression and Thirty-Day Hit-Rate Bounds (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: "Shorting" here means only bearish exposure with bounded loss, buying puts and put debit spreads; all prices are in index points of a synthetic index, and no historical market data are used. The book is financial education, not personal investment advice, with no promise of returns; it gives no trade signals and recommends no contract, institution or trading strategy, and options can lose their entire premium. Abstract: When macro data weaken, the index rallies and implied volatility is pressed low, three popular beliefs circulate: a long put is sure to profit within thirty days; low implied volatility always explodes soon; a fall plus rising volatility pays twice over. This book tests each. "Shorting" here means only bearish exposure with bounded loss, buying puts and put debit spreads; all prices are in index points of a synthetic index, and no historical market data are used. The book is financial education, not personal investment advice, with no promise of returns; it gives no trade signals and recommends no contract, institution or trading strategy, and options can lose their entire premium. Its parameters are illustrative; all calculations are generated by its programs. Chapter 1 uses Black-Scholes Greeks and the variance risk premium to separate expectation from hit rate: even if the index falls by a tenth a year, an at-the-money put held to expiry wins with probability only 0.4358. Chapter 2 lays Heston-type stochastic variance over three-state regime switching, and the thirty-day hit rate becomes a band from 0.1630 to 0.3628. Chapter 3 shows that the half-life of volatility mean reversion is measured in weeks and that its timing is a wide distribution. Chapter 4 performs a six-term Taylor decomposition and compares Derman's three skew rules and spot-volatility correlation, showing that the second half of dual convexity depends on the market regime. Chapter 5 compares tenors of 30, 45 and 60 days and put debit spreads. Chapter 6 uses the Kelly criterion to cap a single position: in the base world the Kelly fraction is only 0.0291, and estimation error argues for a quarter of it. Chapter 7 examines "when the dear reaches its extreme it turns cheap" from the Records of the Grand Historian, Bachelier's principle that the speculator's expectation is zero, Kelly's own words on information rate, and a saying wrongly credited to Keynes; from a low start, the probabilities of reaching the median first and of sinking lower first are 0.5095 and 0.4586. Chapter 8 gives the field tools and checker specifications: a calculator, checks grouped by four kits, ledger fields and a pre-registered sequential test; with a break-even hit rate of 0.2941, a fixed-sample test needs 62 trades. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:宏观转弱、指数反弹、隐含波动率被压到低位时,常见三条流行说法:买入看跌期权三十日内必然获利;低隐含波动率一定很快爆发;下跌加波动率上升就是双倍收益。本书逐条检验。书中的「做空」只指损失有上限的看跌敞口,即买入看跌期权与看跌借方价差;全部价格以合成指数的指数点表示,不使用任何历史行情。本书是金融教育读物,非个人投资建议,不承诺任何收益,不给买卖信号,不推荐任何合约、机构或交易策略;期权可能损失全部权利金。书中参数为示意;全部计算由书中程序生成。 第一章用 Black–Scholes 希腊值与方差风险溢价说明期望与胜率是两回事:假设指数每年下跌一成,平值看跌期权持有到期的胜率也只有 0.4358。第二章在三状态体制切换上叠加 Heston 型随机方差,三十日胜率是一条 0.1630 到 0.3628 的带。第三章说明波动率均值回归的半衰期以周计,回归时点是很宽的分布。第四章做六项泰勒分解,比较 Derman 的三种偏斜规则与现货波动率相关,说明双重凸性的第二重取决于市场体制。第五章比较 30、45、60 日期限与看跌价差。第六章用凯利准则给单笔仓位定上限,基准世界的凯利比例只有 0.0291,估计误差逼人取四分之一。 第七章考辨《史记·货殖列传》「贵上极则反贱」、巴舍利耶「投机者的数学期望为零」、凯利关于信息率的原话,以及一句被误归于凯恩斯的格言;从低位出发,先回到中位数与先跌得更低的概率是 0.5095 对 0.4586。第八章给出实战工具与检查器规格:试算器「低波动窗口看跌头寸双希腊值情景与凯利仓位上限试算器」、按四件套分组的检查项、台账字段与预注册序贯检验,盈亏平衡胜率 0.2941 时,固定样本检验需要 62 笔交易。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 The Variance Risk Premium: Why a Long Put Loses on Average — Chapter 1 uses Black-Scholes Greeks and the variance risk premium to separate expectation from hit rate: even if the index falls by a tenth a year, an at-the-money put held to expiry wins with probability only 0.4358. (https://5dlibrary.com/answers/en/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c01/) 第1章 方差风险溢价:看跌多头为什么平均是负的 — 第一章用 Black–Scholes 希腊值与方差风险溢价说明期望与胜率是两回事:假设指数每年下跌一成,平值看跌期权持有到期的胜率也只有 0.4358。 (https://5dlibrary.com/answers/zh/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c01/) - Ch.2 Bear-Rally Regimes: the Thirty-Day Hit Rate Is a Band, Not a Number — Chapter 2 lays Heston-type stochastic variance over three-state regime switching, and the thirty-day hit rate becomes a band from 0.1630 to 0.3628. (https://5dlibrary.com/answers/en/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c02/) 第2章 熊市反弹体制:三十日胜率是一条带而不是一个数 — 第二章在三状态体制切换上叠加 Heston 型随机方差,三十日胜率是一条 0.1630 到 0.3628 的带。 (https://5dlibrary.com/answers/zh/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c02/) - Ch.3 Clustering and Half-Life: Low Implied Volatility Does Not Explode on Schedule — Program p31 shows that as GARCH(1,1) persistence rises from 0.980 to 0.995, the half-life stretches from 34.3 to 138.3 trading days, and that starting from 0.10, expected volatility first returns to 0.15 after 26.7 to 107.5 trading days. Program p32 shows that a Heston-type variance starting at 0.11 returns to 0.20 within 30 days with probability 0.1896 to 0.2595, and is still no higher than 0.13 on day 30 with probability 0.4812 to 0.6146. The direction of mean reversion is certain; its timing is a wide distribution. (https://5dlibrary.com/answers/en/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c03/) 第3章 波动率聚集与半衰期:低隐含波动率不会按日历爆发 — 程序 p31 显示,GARCH(1,1) 持续性从 0.980 升到 0.995,半衰期从 34.3 个交易日拉长到 138.3 个交易日;从 0.10 出发,预期波动率第一次回到 0.15 需要 26.7 到 107.5 个交易日。程序 p32 显示,Heston 型方差从 0.11 出发,30 天内回到 0.20 的概率是 0.1896 到 0.2595,第 30 天仍不高于 0.13 的概率是 0.4812 到 0.6146。均值回归的方向确定,时点是一个很宽的分布。 (https://5dlibrary.com/answers/zh/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c03/) - Ch.4 Delta-Vega Dual Convexity: Taylor Decomposition, Skew Rules and Spot-Volatility Correlation — Program p41 splits the result of a put into six terms. In scenario A, delta, gamma, vega and vanna pull together and the full revaluation shows a gain of 70.12 points. In scenario B the Taylor approximation overstates the gain by about twelve points, so large scenarios must be fully revalued. The short-squeeze rally of scenario C loses -17.84 points as direction and volatility strike together, and scenario E, a fall with falling volatility, gains only 50.12 points. Program p42 revalues the put under Derman's three rules: for the same 200-point fall, sticky delta and the sticky implied tree differ by twenty points. Program p43 shows that with a correlation of -0.7 the vega part averages +11.43 points on falling paths, while with zero correlation it becomes -1.75 points. Dual convexity is real, but its second half depends on the market's regime and is not built into the option. (https://5dlibrary.com/answers/en/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c04/) 第4章 双希腊值凸性共振:泰勒分解、偏斜规则与现货波动率相关 — 程序 p41 把看跌期权损益拆成六项:情景 A 中 delta、gamma、vega、vanna 同向,全价盈利 70.12 点;情景 B 中泰勒近似高估约十二点,大情景须全价重估;逼空反弹情景 C 亏损 -17.84 点,方向与波动率双杀;下跌但波动率下降的情景 E 只赚 50.12 点。程序 p42 按 Derman 的三种规则重估,同一次 200 点下跌,粘性 delta 与粘性隐含树之间相差二十点。程序 p43 显示,相关系数 -0.7 时下跌路径上 vega 部分平均 +11.43 点,相关为零时变成 -1.75 点。双重凸性真实存在,但它的第二重取决于市场体制,不是期权自带的。 (https://5dlibrary.com/answers/zh/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c04/) - Ch.5 Choosing a Tenor: the Time-Value Accounts of 30-, 45- and 60-Day Contracts — Chapter 5 compares tenors of 30, 45 and 60 days and put debit spreads. (https://5dlibrary.com/answers/en/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c05/) 第5章 期限选择:三十、四十五与六十日合约的时间价值账 — 第五章比较 30、45、60 日期限与看跌价差。 (https://5dlibrary.com/answers/zh/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c05/) - Ch.6 A Kelly Cap on a Single Position: Skewed Payoffs, Estimation Error and Fractional Kelly — Chapter 6 uses the Kelly criterion to cap a single position: in the base world the Kelly fraction is only 0.0291, and estimation error argues for a quarter of it. (https://5dlibrary.com/answers/en/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c06/) 第6章 单笔凯利上限:偏斜收益、估计误差与分数凯利 — 第六章用凯利准则给单笔仓位定上限,基准世界的凯利比例只有 0.0291,估计误差逼人取四分之一。 (https://5dlibrary.com/answers/zh/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c06/) - Ch.7 Classical Sources East and West: Dearness Turning Cheap, the Speculator's Zero Expectation, and a Saying about Solvency — Chapter 7 examines "when the dear reaches its extreme it turns cheap" from the Records of the Grand Historian, Bachelier's principle that the speculator's expectation is zero, Kelly's own words on information rate, and a saying wrongly credited to Keynes; from a low start, the probabilities of reaching the median first and of sinking lower first are 0.5095 and 0.4586. (https://5dlibrary.com/answers/en/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c07/) 第7章 东西方原典考辨:贵上极则反贱、投机者的零期望与偿付能力引语 — 第七章考辨《史记·货殖列传》「贵上极则反贱」、巴舍利耶「投机者的数学期望为零」、凯利关于信息率的原话,以及一句被误归于凯恩斯的格言;从低位出发,先回到中位数与先跌得更低的概率是 0.5095 对 0.4586。 (https://5dlibrary.com/answers/zh/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c07/) - Ch.8 Field Tools and Checker Specifications: a Ledger for Puts Bought in Low-Volatility Windows, and a Pre-Registered Test — Chapter 8 gives the field tools and checker specifications: a calculator, checks grouped by four kits, ledger fields and a pre-registered sequential test; with a break-even hit rate of 0.2941, a fixed-sample test needs 62 trades. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c08/) 第8章 实战工具与检查器规格:低波动窗口看跌头寸台账与预注册检验 — 第八章给出实战工具与检查器规格:试算器「低波动窗口看跌头寸双希腊值情景与凯利仓位上限试算器」、按四件套分组的检查项、台账字段与预注册序贯检验,盈亏平衡胜率 0.2941 时,固定样本检验需要 62 笔交易。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/short-low-iv-windows-macro-downturn-delta-vega-dual-convexity-thirty-day-hit-rate-bounds/c08/) ## 5DL-B-426 · Optimal Early Profit-Taking in Options: Harvesting Volatility Spikes, Twenty-One-Day Decay Cliffs and the Expiration Hold Trap 中文书名:期权提前止盈最优停时、恐慌脉冲波动率收割与持有至到期回吐陷阱母典 DOI: pending Cite: Fan, Y. (2026). Optimal Early Profit-Taking in Options: Harvesting Volatility Spikes, Twenty-One-Day Decay Cliffs and the Expiration Hold Trap (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All prices are in index points, every path is synthetic and the parameters are illustrative; all calculations are generated by the book's programs. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. Abstract: Should an option be held to expiry? Three beliefs circulate among traders: hold to expiry to collect it all once the direction is right; close a spread at fifty to sixty-five percent of its maximum profit and leave with 21 days to go, which significantly improves the annualized Sharpe ratio; and take profit as early as possible. By optional stopping, if quotes are fair and the holder agrees with them, every closing rule has the same expectation; differences come only from a gap between forecast and quote, from costs and from turnover. All prices are in index points, every path is synthetic and the parameters are illustrative; all calculations are generated by the book's programs. This is a book of financial education: it is not personal investment advice, it makes no promise of returns, and it recommends no contract or trading strategy; an option can lose its entire premium and carries risks of early assignment, illiquidity and slippage, and the investor suitability rules of each jurisdiction apply. Chapter 1 shows that after a sharp fall the time value in a long put's quote is a price set at a high volatility, so that with a forecast realized volatility of 0.20 holding to expiry hands back 1.096 points, while a debit spread that has fallen through its short strike has a negative volatility exposure and holding has the advantage. Chapter 2 shows that decay follows a square-root law with no cliff, and that "21 days" is a threshold on a risk ratio. Chapter 3 shows that on the first day after a spike the volatility give-back makes up 0.751 of the expected loss. Chapter 4 solves for the optimal closing boundary on a trinomial lattice; it varies with the state and the structure and is not a fixed share, and a stand-alone 50% take-profit ranks last under both quoting assumptions. Chapter 5 tests the practitioners' rule with sixty rules in four scenarios: for a buyer of debit spreads, holding to expiry ranks first by annualized Sharpe ratio; for a seller, a 50% take-profit with a 21-day stop raises the Sharpe ratio from 0.338 to 0.558 at zero cost and is overturned at a cost of 0.05 points; and even if the difference is real, a single trader would need 73.1 years of records to separate it from zero. Chapter 6 discusses give-back, pin risk and the disposition effect: of the trades that once reached half their maximum profit, 0.206 end in a loss when held to expiry. Chapter 7 examines "to hold and fill it is not as good as to stop" from the Laozi, Fan Li in the Shiji, Wald's sequential analysis and a maxim of doubtful origin. Chapter 8 gives a calculator, checker specifications, a closing-decision ledger and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. 摘要:期权该不该拿到期?交易者之间流传三条说法:看对方向就拿到期才吃得满;价差赚到五成到六成五就平、剩 21 天强制离场能显著提升年化夏普;止盈越早越好。本书把三条说法写成可计算、可检验的命题。核心论证来自最优停止:若报价公允且持有人同意这个定价,任何平仓规则的期望都相同,拿到期并不多吃;差别只能来自预测与报价之差、交易成本与资金周转。书中价格一律以指数点表示,路径全部合成,参数为示意;全部计算由书中程序生成。本书是金融教育读物,非个人投资建议,不承诺任何收益,不推荐任何合约或交易策略;期权可能损失全部权利金,并有提前指派、流动性与滑点风险,须遵守所在法域的投资者适当性规则。 第一章说明急跌之后,看跌多头报价里的时间价值是市场按高波动率开出的价钱,预测实现波动率 0.20 时持有到期要让渡 1.096 点;已跌过空头腿的借方价差则波动率敞口为负,持有反而占优。第二章说明衰减曲线遵循平方根定律、没有悬崖,「21 日」是风险比值门槛。第三章说明脉冲之后第一天的预期流失中,波动率回吐占 0.751。第四章在三叉格点上解出最优平仓边界,它随状态与结构变化,不是固定比例;单独的 50% 止盈在两种报价假设下都排在末位。 第五章用六十条规则、四种情景检验从业者经验法则:对借方价差的买方,按年化夏普排第一的都是拿到期;对卖方,50% 止盈加 21 日止损在零成本时把夏普从 0.338 提到 0.558,成本 0.05 点即告翻转;即使差别真实,单个交易者也需要 73.1 年记录才能把它与零分开。第六章讨论回吐、钉住风险与处置效应:曾赚到最大收益一半的交易中有 0.206 持有到期后亏损。第七章考辨《老子》「持而盈之,不如其已」、《史记》范蠡、瓦尔德的序贯分析与一句出处可疑的格言。第八章给出试算工具、检查器规格、平仓决策台账与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 "Hold to Expiry to Collect It All" Is a Bookkeeping Error: Continuation Value against Closing Value — Chapter 1 shows that after a sharp fall the time value in a long put's quote is a price set at a high volatility, so that with a forecast realized volatility of 0.20 holding to expiry hands back 1.096 points, while a debit spread that has fallen through its short strike has a negative volatility exposure and holding has the advantage. (https://5dlibrary.com/answers/en/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c01/) 第1章 「拿到期才吃得满」是一笔错账:继续价值与立即平仓价值 — 第一章说明急跌之后,看跌多头报价里的时间价值是市场按高波动率开出的价钱,预测实现波动率 0.20 时持有到期要让渡 1.096 点;已跌过空头腿的借方价差则波动率敞口为负,持有反而占优。 (https://5dlibrary.com/answers/zh/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c01/) - Ch.2 There Is No Cliff on the Decay Curve: Where Time Value Speeds Up, and by How Much — Chapter 2 shows that decay follows a square-root law with no cliff, and that "21 days" is a threshold on a risk ratio. (https://5dlibrary.com/answers/en/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c02/) 第2章 衰减曲线上没有悬崖:时间价值在何处加速、加速多少 — 第二章说明衰减曲线遵循平方根定律、没有悬崖,「21 日」是风险比值门槛。 (https://5dlibrary.com/answers/zh/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c02/) - Ch.3 Panic Spikes: Jumps, Mean Reversion and the Half-Life of the Volatility Give-Back — Chapter 3 shows that on the first day after a spike the volatility give-back makes up 0.751 of the expected loss. (https://5dlibrary.com/answers/en/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c03/) 第3章 恐慌脉冲:跳跃、均值回归与波动率回吐的半衰期 — 第三章说明脉冲之后第一天的预期流失中,波动率回吐占 0.751。 (https://5dlibrary.com/answers/zh/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c03/) - Ch.4 The Optimal Closing Boundary Is Not a Fixed Share: the Snell Envelope on a Lattice — Chapter 4 solves for the optimal closing boundary on a trinomial lattice; it varies with the state and the structure and is not a fixed share, and a stand-alone 50% take-profit ranks last under both quoting assumptions. (https://5dlibrary.com/answers/en/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c04/) 第4章 最优平仓边界不是一个固定比例:格点上的斯奈尔包络 — 第四章在三叉格点上解出最优平仓边界,它随状态与结构变化,不是固定比例;单独的 50% 止盈在两种报价假设下都排在末位。 (https://5dlibrary.com/answers/zh/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c04/) - Ch.5 Sixty Rules in Four Scenarios: a Simulated Test of Profit-Taking Shares and Time Stops — Chapter 5 tests the practitioners' rule with sixty rules in four scenarios: for a buyer of debit spreads, holding to expiry ranks first by annualized Sharpe ratio; for a seller, a 50% take-profit with a 21-day stop raises the Sharpe ratio from 0.338 to 0.558 at zero cost and is overturned at a cost of 0.05 points; and even if the difference is real, a single trader would need 73.1 years of records to separate it from zero. (https://5dlibrary.com/answers/en/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c05/) 第5章 六十条规则、四种情景:止盈比例与时间止损的模拟检验 — 第五章用六十条规则、四种情景检验从业者经验法则:对借方价差的买方,按年化夏普排第一的都是拿到期;对卖方,50% 止盈加 21 日止损在零成本时把夏普从 0.338 提到 0.558,成本 0.05 点即告翻转;即使差别真实,单个交易者也需要 73.1 年记录才能把它与零分开。 (https://5dlibrary.com/answers/zh/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c05/) - Ch.6 Three Traps of Holding to Expiry: Give-Back, Pin Risk and the Disposition Effect — Chapter 6 discusses give-back, pin risk and the disposition effect: of the trades that once reached half their maximum profit, 0.206 end in a loss when held to expiry. (https://5dlibrary.com/answers/en/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c06/) 第6章 持有至到期的三个陷阱:回吐、钉住风险与处置效应 — 第六章讨论回吐、钉住风险与处置效应:曾赚到最大收益一半的交易中有 0.206 持有到期后亏损。 (https://5dlibrary.com/answers/zh/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c06/) - Ch.7 "To Hold and Fill It Is Not as Good as to Stop": Stopping and Leaving in Classical Sources of East and West — With a type I error of 0.05 and a type II error of 0.20, the upper constant is 16.000 and the lower 0.2105. For an effect of 0.10 standard deviations a fixed-sample test needs 619 paired trades; the sequential test needs on average about 268 under the null hypothesis and about 381 under the alternative, and the simulation gives 295 and 444, with an actual type I error of 0.045 and a power of 0.790. For an effect of 0.20 the fixed sample is 155 trades and the sequential averages are 67 and 95; for 0.40 they are 39, 17 and 24, with a power of 0.848. The sequential test saves between a third and three fifths of the sample on average, at the price of a sample size that is not fixed in advance. The pre-registered plan of Chapter 8 uses it. (https://5dlibrary.com/answers/en/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c07/) 第7章 「持而盈之,不如其已」:东西方原典里的停止与离场 — 取第一类错误 0.05、第二类错误 0.20,上界常数是 16.000,下界是 0.2105。效应量为 0.10 个标准差时,固定样本检验需要 619 笔配对交易;序贯检验在原假设下的平均样本量约 268 笔,在备择假设下约 381 笔,模拟分别得到 295 笔与 444 笔,实际第一类错误 0.045,检验力 0.790。效应量 0.20 时,固定样本 155 笔,序贯平均 67 笔与 95 笔;0.40 时分别是 39 笔、17 笔与 24 笔,检验力 0.848。序贯检验平均省下三到六成的样本,代价是单次检验的样本量不固定。第八章的预注册方案会采用它。 (https://5dlibrary.com/answers/zh/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c07/) - Ch.8 Field Tools and Checker Specifications: the Closing-Decision Ledger, the Calculator and a Pre-Registered Test — Chapter 8 gives a calculator, checker specifications, a closing-decision ledger and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c08/) 第8章 实战工具与检查器规格:平仓决策台账、试算器与预注册检验 — 第八章给出试算工具、检查器规格、平仓决策台账与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(21 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/options-early-profit-taking-optimal-stopping-volatility-spike-harvest-decay-cliff-hold-trap/c08/) ## 5DL-B-427 · When Utility Outperforms Literal Truth: Evolutionary Fitness-Payoff Games, Functional Belief Engineering and Physical Guardrails 中文书名:效用优先于字面真实:演化适存度博弈、功能性信念重塑与物理硬边界母典 DOI: pending Cite: Fan, Y. (2026). When Utility Outperforms Literal Truth: Evolutionary Fitness-Payoff Games, Functional Belief Engineering and Physical Guardrails (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters and data are illustrative or synthetic, and every calculation is generated by its programs. The book is not medical advice, offers no clinical psychological service, is not personal investment advice and makes no promise of returns; functional beliefs are for one's own action frame and never for deceiving others. Abstract: Three popular claims surround the question of whether beliefs should be true or useful: that perception closer to objective reality always confers a survival advantage, that confidence brings success, and that a strong enough belief breaks through any limit. This book writes each claim as a bounded mathematical model and uses fixed-seed, standard-library Python programs to compute where it holds and where it fails. Its parameters and data are illustrative or synthetic, and every calculation is generated by its programs. The book is not medical advice, offers no clinical psychological service, is not personal investment advice and makes no promise of returns; functional beliefs are for one's own action frame and never for deceiving others. Chapters 1 and 2 rebuild "fitness beats truth" with replicator dynamics and the Moran process: with one perceptual bit and a single-peaked payoff, the share of the fitness-coded strategy rises from its starting value to 0.7712 by time 60. Yet once the number of goals rises from one to two, truth coding does at least as well on 0.703 of random goal sets, and when the payoff model is wrong or a cliff is present, the code that resolves the boundary truthfully wins. This is the book's correction of its blueprint: utility beats truth only inside a stated, bounded model. Chapter 3 uses a contest model of the Johnson and Fowler type to show that a stable upward self-bias appears only when the value-to-cost ratio exceeds about 0.75, and often harms the group. Chapter 4 gives the ceiling, peak and reversal point of the self-fulfilling gain. Chapter 5 connects overconfidence to the Kelly criterion: for an even-money stake with a true win probability of 0.55, overestimating that probability by 0.05 drives long-run growth to -0.000138. Chapter 6 builds the dual ledger, in which the action frame may be coded for utility, the probability ledger is calibrated with the Brier score and the hard-boundary ledger does not yield; when the frame rewrites the probability ledger, the illustrative ruin share rises from 0.0925 to 0.7055. Chapter 7 examines the Xunzi, the Han Feizi, James and Bacon. Chapter 8 gives field tools and checker specifications, ledger fields and a pre-registered test. Marked boxes separate published evidence, runnable calculations (22 programs), classical sources and the author's hypotheses. 摘要:关于信念应该真实还是有用,流行着三种说法:知觉越接近客观真实越有生存优势;自信就会成功;信念足够强就能突破一切限制。本书把每一种说法写成有界的数学模型,用固定种子的标准库 Python 程序算出它成立与失效的条件。书中参数与数据均为示意或合成,全部计算由书中程序生成;本书不是医疗建议,不提供任何临床心理服务,不是个人投资建议,不承诺任何收益;功能性信念只用于自己的行动框架,不用于欺骗他人。 第一、二章用复制子动力学与 Moran 过程复现「适存胜过真实」:一比特知觉、单峰收益下,适存编码策略的份额从起点到第 60 个时间单位升到 0.7712;但目标一旦从一个增加到两个,真实编码就在 0.703 的随机目标组上不输,收益模型设错或存在悬崖时,能如实刻画边界的编码胜出。这是本书的蓝图修正:效用胜过真实只在写明的有界模型内成立。第三章用 Johnson 与 Fowler 型竞争模型显示,稳定的自我高估只在价值代价比高于约 0.75 时出现,且常让群体受损。第四章给出自我实现增益的天花板、峰值与反转点。 第五章把过度自信接上凯利公式:真实胜率 0.55 的等额押注,高估胜率 0.05 就使长期增长率跌到 -0.000138。第六章建立双账本:行动框架可按效用编码,概率账以 Brier 分数校准,硬边界账不让步;框架改写概率账时,示意破产比例从 0.0925 升到 0.7055。第七章考辨《荀子·天论》、《韩非子·五蠹》、James 与培根。第八章给出实战工具与检查器规格、台账字段与预注册检验。全书以标注框区分已发表证据、可运行计算(22 段程序)、原典与作者假说。 Chapter answers: - Ch.1 Fitness Beats Truth: a Replicator Contest between One-Bit Perceptions — Chapters 1 and 2 rebuild "fitness beats truth" with replicator dynamics and the Moran process: with one perceptual bit and a single-peaked payoff, the share of the fitness-coded strategy rises from its starting value to 0.7712 by time 60. Yet once the number of goals rises from one to two, truth coding does at least as well on 0.703 of random goal sets, and when the payoff model is wrong or a cliff is present, the code that resolves the boundary truthfully wins. This is the book's correction of its blueprint: utility beats truth only inside a stated, bounded model. (https://5dlibrary.com/answers/en/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c01/) 第1章 适存胜过真实:一比特知觉的复制子竞赛 — 第一、二章用复制子动力学与 Moran 过程复现「适存胜过真实」:一比特知觉、单峰收益下,适存编码策略的份额从起点到第 60 个时间单位升到 0.7712;但目标一旦从一个增加到两个,真实编码就在 0.703 的随机目标组上不输,收益模型设错或存在悬崖时,能如实刻画边界的编码胜出。这是本书的蓝图修正:效用胜过真实只在写明的有界模型内成立。 (https://5dlibrary.com/answers/zh/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c01/) - Ch.2 Truth Strikes Back: Shifting Goals, Wrong Models and Cliff-Edge Costs — Three programs have drawn the boundary of "fitness beats truth". Raise the number of goals from one to two and the truth code already does at least as well on 0.703 of the random goal sets; shift the payoff model by 0.15 and the truth code overtakes; put a cliff in the range and the winner is the code that resolves the edge of the cliff truthfully. Fitness beats truth is a conditional theorem, not a maxim for living. (https://5dlibrary.com/answers/en/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c02/) 第2章 真实反超:目标多变、模型设错与悬崖代价 — 本章用三段程序给适存胜过真实划了边界。目标从一个增加到两个,真实编码就在 0.703 的随机目标组上不输;收益模型偏 0.15,真实编码反超;悬崖存在时,赢家是能把悬崖边界如实刻画出来的那种编码。适存胜过真实是一个有条件的定理,而不是一条处世格言。 (https://5dlibrary.com/answers/zh/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c02/) - Ch.3 When Overconfidence Evolves: the Value-to-Cost Ratio and Noisy Reading of Rivals — Chapter 3 uses a contest model of the Johnson and Fowler type to show that a stable upward self-bias appears only when the value-to-cost ratio exceeds about 0.75, and often harms the group. (https://5dlibrary.com/answers/en/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c03/) 第3章 过度自信的演化条件:价值代价比与读数噪声 — 第三章用 Johnson 与 Fowler 型竞争模型显示,稳定的自我高估只在价值代价比高于约 0.75 时出现,且常让群体受损。 (https://5dlibrary.com/answers/zh/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c03/) - Ch.4 The Ceiling on Self-Fulfilment: How Far Confidence Can Lift the Odds — With GAMMA at 0.02, the best inflation is only 0.10 and the largest gain is 0.0094, far below the ceiling of 0.20 for this value of GAMMA. Once b reaches 0.5 the result is already -0.0776, and at 1 it is -0.2863. With GAMMA at 0.05 the best inflation is 0.25 and the largest gain 0.0537; at b equal to 0.5 a gain of 0.0224 remains, while at b equal to 1 the result is already -0.1363. (https://5dlibrary.com/answers/en/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c04/) 第4章 自我实现增益的上限:信心能抬高多少成功率 — GAMMA 为 0.02 时,最优膨胀只有 0.10,最大增益 0.0094,远低于这一 GAMMA 下的天花板 0.20。b 一旦到 0.5,结果已经是 -0.0776;到 1,是 -0.2863。GAMMA 为 0.05 时,最优膨胀是 0.25,最大增益 0.0537;b 为 0.5 时还剩 0.0224,b 为 1 时已经是 -0.1363。 (https://5dlibrary.com/answers/zh/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c04/) - Ch.5 Overconfidence Meets the Kelly Criterion: the Growth Cost of an Inflated Win Rate and the Probability of Ruin — Chapter 5 connects overconfidence to the Kelly criterion: for an even-money stake with a true win probability of 0.55, overestimating that probability by 0.05 drives long-run growth to -0.000138. (https://5dlibrary.com/answers/en/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c05/) 第5章 过度自信遇上凯利公式:高估胜率的增长代价与破产概率 — 第五章把过度自信接上凯利公式:真实胜率 0.55 的等额押注,高估胜率 0.05 就使长期增长率跌到 -0.000138。 (https://5dlibrary.com/answers/zh/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c05/) - Ch.6 The Dual Ledger: the Brier Score, Its Calibration Decomposition and Boundaries That Do Not Yield — Chapter 6 builds the dual ledger, in which the action frame may be coded for utility, the probability ledger is calibrated with the Brier score and the hard-boundary ledger does not yield; when the frame rewrites the probability ledger, the illustrative ruin share rises from 0.0925 to 0.7055. (https://5dlibrary.com/answers/en/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c06/) 第6章 双账本:Brier 分数、校准分解与不让步的硬边界 — 第六章建立双账本:行动框架可按效用编码,概率账以 Brier 分数校准,硬边界账不让步;框架改写概率账时,示意破产比例从 0.0925 升到 0.7055。 (https://5dlibrary.com/answers/zh/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c06/) - Ch.7 Classical Sources: the Constant Course of Heaven, the Hare and the Stump, the Expedient and Obedience to Nature — Over just 3 decisions, C is ahead of B on 0.528 of paths, its mean wealth is 1.185 times B's and its median wealth 1.02 times: in the short run the single ledger seems the more expedient. Over 10 decisions, C leads on only 0.409 of paths, yet its mean wealth is 1.689 times B's while its median has fallen to 0.7993. Over 30, the share of paths on which C leads is 0.339, the mean ratio 5.931 and the median ratio 0.5479. Over 100, the share is 0.229, the mean ratio 44.62 and the median ratio 0.1737. Over 300 decisions, C leads on only 0.097 of paths, yet its mean wealth is 864 times B's while its median wealth is a mere 0.002656 times. (https://5dlibrary.com/answers/en/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c07/) 第7章 原典考辨:天行有常、守株待兔、有利之物与服从自然 — 只看 3 次决策,C 在 0.528 的路径上领先 B,平均财富是 B 的 1.185 倍,中位财富是 1.02 倍:短期看,单账本似乎更有利。看 10 次,C 只在 0.409 的路径上领先,平均财富却是 1.689 倍,中位财富已降到 0.7993。看 30 次,领先比例 0.339,平均财富 5.931 倍,中位 0.5479 倍。看 100 次,领先比例 0.229,平均财富 44.62 倍,中位 0.1737 倍。看 300 次,C 只在 0.097 的路径上领先,平均财富却是 B 的 864 倍,中位财富只剩 0.002656 倍。 (https://5dlibrary.com/answers/zh/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c07/) - Ch.8 Field Tools and Checker Specifications: the Dual-Ledger Records, Red-Flag Rules and a Pre-Registered Test — Chapter 8 gives field tools and checker specifications, ledger fields and a pre-registered test. Marked boxes separate published evidence, runnable calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c08/) 第8章 实战工具与检查器规格:双账本台账、判红规则与预注册检验 — 第八章给出实战工具与检查器规格、台账字段与预注册检验。全书以标注框区分已发表证据、可运行计算(22 段程序)、原典与作者假说。 (https://5dlibrary.com/answers/zh/utility-over-literal-truth-fitness-payoff-games-functional-belief-engineering-physical-guardrails/c08/) ## 5DL-B-428 · Performative Speech Acts, Schelling Focal-Point Framing and Verbal Reality-Collapse in High-Stakes Negotiation 中文书名:施为性言语行为、谢林聚点定框与高维语言坍缩商业现实母典 DOI: pending Cite: Fan, Y. (2026). Performative Speech Acts, Schelling Focal-Point Framing and Verbal Reality-Collapse in High-Stakes Negotiation (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All dialogues are synthetic, all parameters illustrative, and every calculation is generated by the book's programs. The book is general business guidance, not individual advice and not legal advice; commissions appear only at a single level, paid to the direct referrer. Abstract: Three beliefs circulate wherever people bargain: that talk merely describes facts, that which of several possible endings a negotiation reaches is a matter of chance, and that opening first is a mistake while the most certain voice wins. This book refutes or bounds each in turn. It writes only about honest commitments, open framing and structured options that can be delivered, and contains no deception, threats or covert influence techniques. The "verbal collapse" of the title is a metaphor only, meaning the fall in the Shannon entropy of the counterpart's expectations once a frame is stated and the selection of one equilibrium among several; it carries no meaning from quantum physics. All dialogues are synthetic, all parameters illustrative, and every calculation is generated by the book's programs. The book is general business guidance, not individual advice and not legal advice; commissions appear only at a single level, paid to the direct referrer. Chapter 1 uses the performative theory of Austin and Searle to break an offer into six felicity conditions: in an illustrative model a structured offer lifts the felicitous share from 0.305 to 0.734, and a fully credible offer lowers the entropy of the counterpart's expectations by 0.918 bits. Chapter 2 discusses Schelling's focal points, reports the published fall in coordination from 0.82 to 0.52 found by Crawford and colleagues, and shows in a model how a sliver of payoff asymmetry drops coordination from 0.889 to 0.497. Chapter 3 estimates the anchor weight of first offers, finds precise, grounded offers ahead of round ones by 0.172, and draws a red line at a breakdown probability of 0.236. Chapter 4 uses the partition equilibria of Crawford and Sobel and the tests of Farrell and Rabin to show when costless words are believed; Chapter 5 treats audience costs and keeping records, where one broken promise costs a twenty-promise record 0.074 of its lower bound. Chapter 6 shows that confidence is an asset only when calibrated: saying "certain" and being wrong drops the listener's belief that you are calibrated to 0.085, while saying "probably" and being wrong leaves it at 0.833; in a full offer, words carry 0.914 of the decision information, so "7-38-55" is no law of communication. Chapter 7 examines Confucius on correct names, Xunzi on names and realities, Aristotle on character and Austin on performatives, and sets out the distillation spec for the four kits. Chapter 8 gives the Opening Frame Anchor-Weight and Focal-Point Coordination Calculator, checker tables grouped by the four kits, the negotiation framing ledger and pre-registered tests. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:谈判桌上流传三条说法:说话只是描述事实;多个可能结局里落到哪一个全凭运气;先开价吃亏,声音越笃定越能赢。本书逐条证伪或立界。书中只写诚实承诺、公开定框与可兑现的结构化选项,不写任何欺骗、威胁或隐蔽影响手法;书名中的「语言坍缩」只作比喻,指对方期望分布的香农熵在定框后下降以及多重均衡中有一个被选中,不含任何量子物理含义。书中对话全部合成,参数为示意,全部计算由书中程序生成;本书是非个人化经营建议,不是法律意见;分润只写单层、只给直接推荐人。 第一章用奥斯汀与塞尔的施为句理论,把报价拆成六个适切条件:示意模型里结构化报价把适切概率从 0.305 提到 0.734,一份完全可信的报价让对方期望熵下降 0.918 比特。第二章讨论谢林聚点,复述克劳福德等人公布的协调率从 0.82 降到 0.52,示意模型里一点点收益不对称让协调率从 0.889 跌到 0.497。第三章估计首报价锚定权重,精确有依据的报价比整数报价高 0.172,并在破裂概率 0.236 处划出红线。第四章用克劳福德与索贝尔的分区均衡和法雷尔与拉宾的检验,说明不花钱的话何时可信;第五章讨论观众成本与守约记录,一次失约让二十次记录的下界损失 0.074。 第六章说明笃定只在校准时是资产:说「肯定」而出错,听者认为你是校准型的概率跌到 0.085,说「大概」而出错仍为 0.833;在完整报价里,言语承载 0.914 的决策信息,「7-38-55」不是沟通法则。第七章考辨孔子正名、荀子名实、亚里士多德论品格与奥斯汀论施为,并给出四件套蒸馏规格。第八章给出开局定框锚定权重与聚点协调概率试算器、按四件套分组的检查器表、谈判定框台账与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Performative Utterances and Felicity Conditions: When Saying Is Doing — Chapter 1 uses the performative theory of Austin and Searle to break an offer into six felicity conditions: in an illustrative model a structured offer lifts the felicitous share from 0.305 to 0.734, and a fully credible offer lowers the entropy of the counterpart's expectations by 0.918 bits. (https://5dlibrary.com/answers/en/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c01/) 第1章 施为句与适切条件:话语何时就是行为 — 第一章用奥斯汀与塞尔的施为句理论,把报价拆成六个适切条件:示意模型里结构化报价把适切概率从 0.305 提到 0.734,一份完全可信的报价让对方期望熵下降 0.918 比特。 (https://5dlibrary.com/answers/zh/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c01/) - Ch.2 Schelling Focal Points: How Minds Meet among Many Equilibria, and When They Stop Meeting — Chapter 2 discusses Schelling's focal points, reports the published fall in coordination from 0.82 to 0.52 found by Crawford and colleagues, and shows in a model how a sliver of payoff asymmetry drops coordination from 0.889 to 0.497. (https://5dlibrary.com/answers/en/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c02/) 第2章 谢林聚点:多重均衡里怎样想到一处,又何时失灵 — 第二章讨论谢林聚点,复述克劳福德等人公布的协调率从 0.82 降到 0.52,示意模型里一点点收益不对称让协调率从 0.889 跌到 0.497。 (https://5dlibrary.com/answers/zh/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c02/) - Ch.3 First Offers as Anchors: How Heavy the First Number Is, and How High Is Too High — Chapter 3 estimates the anchor weight of first offers, finds precise, grounded offers ahead of round ones by 0.172, and draws a red line at a breakdown probability of 0.236. (https://5dlibrary.com/answers/en/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c03/) 第3章 首报价锚定:第一个数字有多重,开多高会破裂 — 第三章估计首报价锚定权重,精确有依据的报价比整数报价高 0.172,并在破裂概率 0.236 处划出红线。 (https://5dlibrary.com/answers/zh/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c03/) - Ch.4 Cheap Talk: When Costless Words Are Still Believed — Costless words are not worthless. When the conflict of interest is small, they remove most uncertainty, 0.980 of it at b = 1/200; when the bias reaches 1/4, they transmit nothing. Whether a statement is credible depends on whether it passes the self-signalling and self-committing tests; rewording it does nothing, and restructuring payoffs does everything. An honest promise can select the high-trust equilibrium between two, but it cannot create an equilibrium that does not exist. (https://5dlibrary.com/answers/en/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c04/) 第4章 廉价磋商:说了不花钱的话,何时还有人信 — 不花钱的话并非全不算数。利益偏差小时,它能去掉大部分不确定性,b 为 1/200 时传递 0.980;偏差达到 1/4,传递归零。一句话可信与否,看它能否通过自我信号与自我承诺两条检验;改措辞没用,改收益结构才有用。诚实的承诺能在两个均衡之间选出高信任的那个,但不能凭空造出一个均衡。 (https://5dlibrary.com/answers/zh/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c04/) - Ch.5 Credible Commitment: Audience Costs, Keeping Records and Repeated Relationships — Chapter 5 treats audience costs and keeping records, where one broken promise costs a twenty-promise record 0.074 of its lower bound. (https://5dlibrary.com/answers/en/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c05/) 第5章 承诺可信度:观众成本、守约记录与重复关系 — 第五章讨论观众成本与守约记录,一次失约让二十次记录的下界损失 0.074。 (https://5dlibrary.com/answers/zh/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c05/) - Ch.6 Voice, Confidence and Calibration: When Certainty Is an Asset and When It Is a Liability — Chapter 6 shows that confidence is an asset only when calibrated: saying "certain" and being wrong drops the listener's belief that you are calibrated to 0.085, while saying "probably" and being wrong leaves it at 0.833; in a full offer, words carry 0.914 of the decision information, so "7-38-55" is no law of communication. (https://5dlibrary.com/answers/en/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c06/) 第6章 声音、笃定与校准:笃定何时是资产,何时是负债 — 第六章说明笃定只在校准时是资产:说「肯定」而出错,听者认为你是校准型的概率跌到 0.085,说「大概」而出错仍为 0.833;在完整报价里,言语承载 0.914 的决策信息,「7-38-55」不是沟通法则。 (https://5dlibrary.com/answers/zh/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c06/) - Ch.7 Classical Sources East and West: Correct Names, Names and Realities, Character and Performatives — Confucius's "if names are not correct, speech does not flow" is a multiplicative chain in probability; in the illustrative model, defining every term lifts the probability of flowing speech from 0.0095 to 0.8330. Xunzi's "names fixed, realities distinguished" is the precondition of credible signals. Aristotle's character is a prior and evidence a likelihood ratio, and the two add in log odds. Austin supplies the dimension of procedure. The ancients were right about names, realities and character; they did not reach coordination failure, the ceiling on information or the calibration penalty. (https://5dlibrary.com/answers/en/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c07/) 第7章 东西方原典考辨:正名、名实、品格与施为 — 孔子的「名不正,则言不顺」在概率上是乘法链,示意模型里全部术语定义后言顺概率从 0.0095 升到 0.8330;荀子的「名定而实辨」是可信信号的前提;亚里士多德的品格是先验,证据是似然比,二者在对数赔率上相加;奥斯汀补上了程序这一维。古人说对了名实与品格,没有说到协调失灵、信息上限与校准惩罚。 (https://5dlibrary.com/answers/zh/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c07/) - Ch.8 Field Tools and Checker Specifications: The Framing Ledger, the Calculator and Pre-Registered Tests — Chapter 8 gives the Opening Frame Anchor-Weight and Focal-Point Coordination Calculator, checker tables grouped by the four kits, the negotiation framing ledger and pre-registered tests. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c08/) 第8章 实战工具与检查器规格:谈判定框台账、试算器与预注册检验 — 第八章给出开局定框锚定权重与聚点协调概率试算器、按四件套分组的检查器表、谈判定框台账与预注册检验。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/performative-speech-acts-schelling-focal-point-framing-verbal-reality-collapse-negotiation/c08/) ## 5DL-B-429 · Reconciling Universal Mission and Intimate Attachment: Concentric Time-Energy Allocation, Role-Conflict Damping and Non-Zero-Sum Ethics 中文书名:普世志业大爱与私密关系小爱的相容边界:同心圆预算分配与非零和伦理母典 DOI: pending Cite: Fan, Y. (2026). Reconciling Universal Mission and Intimate Attachment: Concentric Time-Energy Allocation, Role-Conflict Damping and Non-Zero-Sum Ethics (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Parameters are illustrative; all calculations are generated by the book's programs; the book does not replace psychological, relationship or family counselling and is not personalised advice. Chapter 2 uses social-brain layers and decay dynamics to compute a core floor quota of 14.38 hours a week under illustrative parameters. Abstract: A popular belief holds that serving the wider world and caring for those closest to us compete for one budget of time, and that extreme impartial love and narrow possessiveness must both collapse. This book rewrites those slogans as propositions that data can overturn. After checking the literature it corrects its commissioning blueprint: the time budget is a hard constraint, but spillover between roles has both signs; Goode's scarcity view, the expansion views of Sieber and Marks, and the enrichment theory of Greenhaus and Powell can all hold at once, and zero-sum is the special case in which conflict and enrichment cancel. Hamilton's rule is used only as an evolutionary description, never as a moral rule. Parameters are illustrative; all calculations are generated by the book's programs; the book does not replace psychological, relationship or family counselling and is not personalised advice. In the net-effect model of Chapter 1, pure conflict lowers total output by 0.219 against the zero-sum baseline and pure enrichment raises it by 0.223. Chapter 2 uses social-brain layers and decay dynamics to compute a core floor quota of 14.38 hours a week under illustrative parameters. Chapter 3 turns "must collapse" into probabilities with Monte Carlo simulation: with 150 relationships weighted equally the core is lost with probability 1.000, at a median of week 10; with few people and many hours the probability falls to 0.000; core exclusivity zeroes the outer circle while each extra core hour buys on average only 0.313. Chapter 4 locks the floor first, fills the remainder by concave returns, and uses one-to-many works as leverage, with a break-even rate of 1.045 uses per work per week; raising the friends' floor by simulated quantile lowers the core loss probability to 0.015 while reaching 2,378 touches a year. Chapter 5 models role conflict as a damped oscillator; when enrichment minus conflict reaches 0.70 the system is critically damped and returns to plan fastest. Chapter 6 shows that weighted sums find only 132 of 601 points on an illustrative frontier and miss the stretch where the core just crosses its threshold, which the epsilon-constraint method recovers. Chapter 7 examines the Mohist-Confucian dispute recorded in the Mencius, the Mozi's chapter on impartial care and Aristotle on the number of friends, compressing impartiality and graded extension into one parameter whose largest core-protecting value is 0.740, and closes with a distillation spec for the four kits. Chapter 8 gives a weekly ledger, red-yellow-green rules, a configurator tool and a pre-registered test with 36 people per arm, refuting the central proposition if the upper confidence bound of the difference falls below 0.05. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:流行说法认为,心怀天下与善待至亲争夺同一份时间,极端博爱与狭隘占有都必然崩溃。本书把这些口号改写成可以用数据推翻的命题。核对文献后,本书修正立项蓝图:时间预算是硬约束,角色之间的溢出却有正有负,古德的稀缺论、西伯与马克斯的扩张论、格林豪斯与鲍威尔的增益论可以同时成立,零和只是冲突与增益恰好抵消的特例;汉密尔顿法则只作演化描述,不作规范性指令。书中参数为示意;全部计算由书中程序生成;本书不替代心理咨询、婚恋咨询或家庭辅导,也不是个人化建议。 第一章的净效应模型里,纯冲突让总产出比零和基线低 0.219,纯增益让它高 0.223。第二章用社会脑层级与衰减动力学算出核心底座定额,示意参数下每周 14.38 小时。第三章用蒙特卡罗把「必然崩溃」写成概率:150 段关系等权时核心层失守概率为 1.000,中位在第 10 周;等权对象少、时间多时失守概率可降到 0.000;核心独占则让外圈归零,核心层每多一小时平均只换来 0.313。第四章先锁定额、再按凹回报注水,并以一对多作品为杠杆,盈亏线是每件作品每周 1.045 次使用;把挚友定额按模拟分位数上调后,核心失守概率降到 0.015,一年触达 2,378 次。 第五章把角色冲突写成受阻尼振荡,增益减冲突达到 0.70 时系统临界阻尼、最快回到计划。第六章证明加权求和只找到示意前沿 601 点中的 132 点,看不见核心层刚好越过阈值那一段,而先设下限的 ε 约束法能找回它。第七章考辨《孟子》所记墨儒之争、《墨子·兼爱中》与亚里士多德论友爱人数,把兼爱与推恩压成一个参数,能守住核心的最大递减系数为 0.740,章末给出四件套蒸馏规格。第八章给出周台账、红黄绿规则、配置器小工具与每组 36 人的预注册检验,差值置信区间上界低于 0.05 即推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Zero-Sum Is Not the Default: the Net Effect of Scarcity, Expansion and Enrichment — Zero-sum is not the default. The time constraint is hard, but spillover has two directions and two signs. In the illustrative model with a 60-hour budget, pure conflict lowers total output by 0.219 and pure enrichment raises it by 0.223; when the two are equal the system returns exactly to the zero-sum baseline, and swapping directions can flip the sign of the net effect. Hamilton's rule describes selection; it does not assign duties. (https://5dlibrary.com/answers/en/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c01/) 第1章 零和不是默认:稀缺、扩张与增益的净效应 — 零和不是默认。时间约束是硬的,溢出却有两个方向、两种符号。在 60 小时预算的示意模型里,纯冲突让总产出降低 0.219,纯增益让它提高 0.223,两者相等时恰好回到零和基线;方向互换就能改变净效应的符号。汉密尔顿法则描述选择,不规定义务。 (https://5dlibrary.com/answers/zh/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c01/) - Ch.2 Concentric Circles: Layers, Decay and the Core Floor Quota — Chapter 2 uses social-brain layers and decay dynamics to compute a core floor quota of 14.38 hours a week under illustrative parameters. (https://5dlibrary.com/answers/en/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c02/) 第2章 同心圆:层级、衰减与核心底座定额 — 第二章用社会脑层级与衰减动力学算出核心底座定额,示意参数下每周 14.38 小时。 (https://5dlibrary.com/answers/zh/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c02/) - Ch.3 Two Ways to Lose: the Collapse Probability of Equal Weight and the Price of Core Exclusivity — Chapter 3 turns "must collapse" into probabilities with Monte Carlo simulation: with 150 relationships weighted equally the core is lost with probability 1.000, at a median of week 10; with few people and many hours the probability falls to 0.000; core exclusivity zeroes the outer circle while each extra core hour buys on average only 0.313. (https://5dlibrary.com/answers/en/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c03/) 第3章 两种失守:等权博爱的崩溃概率与核心独占的代价 — 第三章用蒙特卡罗把「必然崩溃」写成概率:150 段关系等权时核心层失守概率为 1.000,中位在第 10 周;等权对象少、时间多时失守概率可降到 0.000;核心独占则让外圈归零,核心层每多一小时平均只换来 0.313。 (https://5dlibrary.com/answers/zh/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c03/) - Ch.4 Floor Quota plus Outer Leverage: Concave Water-Filling and One-to-Many Reach — Chapter 4 locks the floor first, fills the remainder by concave returns, and uses one-to-many works as leverage, with a break-even rate of 1.045 uses per work per week; raising the friends' floor by simulated quantile lowers the core loss probability to 0.015 while reaching 2,378 touches a year. (https://5dlibrary.com/answers/en/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c04/) 第4章 底座定额加外圈杠杆:凹回报注水与一对多触达 — 第四章先锁定额、再按凹回报注水,并以一对多作品为杠杆,盈亏线是每件作品每周 1.045 次使用;把挚友定额按模拟分位数上调后,核心失守概率降到 0.015,一年触达 2,378 次。 (https://5dlibrary.com/answers/zh/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c04/) - Ch.5 Damping Role Conflict: How Many Swings after a Shock, and How Long to Settle — Chapter 5 models role conflict as a damped oscillator; when enrichment minus conflict reaches 0.70 the system is critically damped and returns to plan fastest. (https://5dlibrary.com/answers/en/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c05/) 第5章 角色冲突的阻尼:冲击之后摆几次、多久落定 — 第五章把角色冲突写成受阻尼振荡,增益减冲突达到 0.70 时系统临界阻尼、最快回到计划。 (https://5dlibrary.com/answers/zh/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c05/) - Ch.6 The Pareto Frontier and the Blind Spot of Weighted Sums: Three Weightings Compared — Chapter 6 shows that weighted sums find only 132 of 601 points on an illustrative frontier and miss the stretch where the core just crosses its threshold, which the epsilon-constraint method recovers. (https://5dlibrary.com/answers/en/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c06/) 第6章 帕累托前沿与加权和盲区:三种权重比较 — 第六章证明加权求和只找到示意前沿 601 点中的 132 点,看不见核心层刚好越过阈值那一段,而先设下限的 ε 约束法能找回它。 (https://5dlibrary.com/answers/zh/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c06/) - Ch.7 The Mohist-Confucian Dispute and the Limit on Friends: Eastern and Western Sources Examined — Chapter 7 examines the Mohist-Confucian dispute recorded in the Mencius, the Mozi's chapter on impartial care and Aristotle on the number of friends, compressing impartiality and graded extension into one parameter whose largest core-protecting value is 0.740, and closes with a distillation spec for the four kits. (https://5dlibrary.com/answers/en/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c07/) 第7章 墨儒之争与友爱人数上限:东西方原典考辨 — 第七章考辨《孟子》所记墨儒之争、《墨子·兼爱中》与亚里士多德论友爱人数,把兼爱与推恩压成一个参数,能守住核心的最大递减系数为 0.740,章末给出四件套蒸馏规格。 (https://5dlibrary.com/answers/zh/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c07/) - Ch.8 Closing the Empirical Loop: Field Tools and Checker Specifications for the Concentric Allocator — Chapter 8 gives a weekly ledger, red-yellow-green rules, a configurator tool and a pre-registered test with 36 people per arm, refuting the central proposition if the upper confidence bound of the difference falls below 0.05. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c08/) 第8章 实证闭环:同心圆配置器的实战工具与检查器规格 — 第八章给出周台账、红黄绿规则、配置器小工具与每组 36 人的预注册检验,差值置信区间上界低于 0.05 即推翻核心命题。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/concentric-mission-intimacy-allocation-role-conflict-damping-non-zero-sum-ethics/c08/) ## 5DL-B-430 · Mental Simulation versus Behavioural Enactment in Intimacy: Dopaminergic Fantasy Loops, Oxytocin Bonding Costs and Commitment Asymmetries 中文书名:心智模拟亲密与具身行为兑现的神经分野:多巴胺投射幻象与真实契约代价母典 DOI: pending Cite: Fan, Y. (2026). Mental Simulation versus Behavioural Enactment in Intimacy: Dopaminergic Fantasy Loops, Oxytocin Bonding Costs and Commitment Asymmetries (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; it is not psychological, clinical or relationship counselling, and not personalised business advice. Chapter 4 audits the oxytocin evidence: typical samples have a power of about 0.165, and significant results overstate effects by a factor of 2.62 on average, so "behaviour triggers oxytocin bonding" stands only as a falsifiable hypothesis. Abstract: A popular account holds that intimacy imagined in the mind is a frictionless projection that forms a "supernormal dopamine loop", that real behaviour triggers oxytocin bonding, and that the sexes are born with different jealousy weights for two kinds of deviation. Checking each claim against the original literature, this book finds that all of them must be rewritten. "Intimacy" here means only decisions about emotional attachment, commitment and a shared life; the book contains no description of bodies, uses no clinical language, and treats truthful communication and informed consent as the floor. Its parameters are illustrative; all calculations are generated by its programs; it is not psychological, clinical or relationship counselling, and not personalised business advice. Chapter 1 uses learning rules to show that a fully predictable imagining stops producing positive errors once learned; the real problem is the absence of negative errors, so an illustrative estimate driven by simulation ends 0.600 above the truth and is never corrected. Chapter 2 separates wanting from liking and explains suppression rebound and imagined templates that overshoot every real person. Chapter 3 shows that positive fantasy reduces effort, while implementation intentions have a meta-analytic effect size of 0.65, which on a probit reading lifts a completion rate of 0.35 to 0.604. Chapter 4 audits the oxytocin evidence: typical samples have a power of about 0.165, and significant results overstate effects by a factor of 2.62 on average, so "behaviour triggers oxytocin bonding" stands only as a falsifiable hypothesis. Chapter 5 gives an asymmetry that needs no hormone: an observable broken promise enters reputation, and under common parameters takes 5 to 25 kept promises to repair. Chapter 6 writes the sex difference in jealousy as an effect size: the pooled continuous effect is 0.258, the distributions overlap by 0.897, and the best guess of sex from a rating is right with probability 0.551, so it cannot be used to infer anything about an individual. Chapter 7 examines Wang Yangming, the Great Learning, the Gospel of Matthew and Aristotle. Chapter 8 delivers the Simulation-Enactment Gap Checker, checker specifications grouped by the four kits, a ledger rhythm and pre-registered sample sizes. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:流行说法认为,脑内想象的亲密是零摩擦的投射,会形成「超常多巴胺回路」,而真实行为会触发催产素联结,两性对两类背离的嫉妒天生不同。本书逐条对照原始文献,结论是这些说法都须改写。书中「亲密」只指情感依附、承诺与共同生活的决策,全书零身体描写、非临床,如实沟通与知情同意是底线。书中参数为示意;全部计算由书中程序生成;本书不是心理、临床或婚恋咨询,也不是个人化经营建议。 第一章用学习规则说明,完全可预期的想象学会后不再产生正向误差,真正的问题是缺少负向误差:示意轨迹被模拟抬高到比真实值高出 0.600,且永不下修。第二章区分想要与喜欢,解释压抑反弹与想象模板越过真实对象。第三章说明积极幻想削减投入,执行意图的元分析效应量为 0.65,按概率单位读法可把 0.35 的完成率提到 0.604。第四章检视催产素证据:典型样本的功效约 0.165,显著结果的效应量平均被放大 2.62 倍,因此「行为触发催产素联结」只作可证伪假说。 第五章给出不依赖激素的不对称:可观察的违诺进入声誉,常见参数下需要 5 到 25 次守约修复。第六章把嫉妒的性别差异写成效应量:连续量表的合并效应 0.258,两组分布重叠 0.897,凭评分猜性别的最好准确率只有 0.551,不能用于推断个人。第七章考辨王阳明、《大学》、《马太福音》与亚里士多德。第八章给出「心智模拟与行为兑现缺口判别器」、按四件套分组的检查器规格、台账节奏与预注册样本量。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Prediction Error: Why Mental Simulation Is Never Corrected Downwards — Chapter 1 uses learning rules to show that a fully predictable imagining stops producing positive errors once learned; the real problem is the absence of negative errors, so an illustrative estimate driven by simulation ends 0.600 above the truth and is never corrected. (https://5dlibrary.com/answers/en/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c01/) 第1章 预测误差:心智模拟为何得不到下修 — 第一章用学习规则说明,完全可预期的想象学会后不再产生正向误差,真正的问题是缺少负向误差:示意轨迹被模拟抬高到比真实值高出 0.600,且永不下修。 (https://5dlibrary.com/answers/zh/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c01/) - Ch.2 Wanting and Liking: Why Two Channels Can Come Apart — In the first period the suppression group's intrusion rate climbs from 0.80 to 1.63, still below the expression group's rate of about 2.00. When the second period begins and both groups may think freely, the suppression group starts at 4.85 while the expression group stands at 1.90. Over the five minutes of the second period the suppression group totals 23.79 intrusions and the expression group 9.32, a ratio of 2.55. If the monitor's priming factor is set to zero, the ratio falls to 1.05. In this model, most of the rebound comes from the monitor repeatedly calling up the very target it is guarding against. (https://5dlibrary.com/answers/en/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c02/) 第2章 想要与喜欢:两条通道为何会分开 — 压抑组在第一阶段的闯入率从 0.80 爬到 1.63,仍低于表达组的 2.00 附近。进入第二阶段、两组都自由表达时,压抑组一开始就是 4.85,表达组只有 1.90。第二阶段五分钟合计,压抑组 23.79,表达组 9.32,比值 2.55。若把监控的抬升系数设为零,比值只剩 1.05。在这个模型里,反弹的大部分来自监控本身对目标的反复唤起。 (https://5dlibrary.com/answers/zh/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c02/) - Ch.3 The Cost of Positive Fantasy and the Exit through Implementation Intentions — Chapter 3 shows that positive fantasy reduces effort, while implementation intentions have a meta-analytic effect size of 0.65, which on a probit reading lifts a completion rate of 0.35 to 0.604. (https://5dlibrary.com/answers/en/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c03/) 第3章 积极幻想的代价与执行意图的出口 — 第三章说明积极幻想削减投入,执行意图的元分析效应量为 0.65,按概率单位读法可把 0.35 的完成率提到 0.604。 (https://5dlibrary.com/answers/zh/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c03/) - Ch.4 Oxytocin Bonding: How Strong Is the Chain of Evidence? — Chapter 4 audits the oxytocin evidence: typical samples have a power of about 0.165, and significant results overstate effects by a factor of 2.62 on average, so "behaviour triggers oxytocin bonding" stands only as a falsifiable hypothesis. (https://5dlibrary.com/answers/en/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c04/) 第4章 催产素联结:一条证据链有多牢 — 第四章检视催产素证据:典型样本的功效约 0.165,显著结果的效应量平均被放大 2.62 倍,因此「行为触发催产素联结」只作可证伪假说。 (https://5dlibrary.com/answers/zh/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c04/) - Ch.5 Reputation and Commitment: What One Observable Broken Promise Costs — Chapter 5 gives an asymmetry that needs no hormone: an observable broken promise enters reputation, and under common parameters takes 5 to 25 kept promises to repair. (https://5dlibrary.com/answers/en/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c05/) 第5章 声誉与承诺:单次可观察的违诺值多少 — 第五章给出不依赖激素的不对称:可观察的违诺进入声誉,常见参数下需要 5 到 25 次守约修复。 (https://5dlibrary.com/answers/zh/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c05/) - Ch.6 Sex Differences in Jealousy: Effect Sizes, Overlap and Measurement — Chapter 6 writes the sex difference in jealousy as an effect size: the pooled continuous effect is 0.258, the distributions overlap by 0.897, and the best guess of sex from a rating is right with probability 0.551, so it cannot be used to infer anything about an individual. (https://5dlibrary.com/answers/en/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c06/) 第6章 嫉妒的性别差异:效应量、重叠率与测量方式 — 第六章把嫉妒的性别差异写成效应量:连续量表的合并效应 0.258,两组分布重叠 0.897,凭评分猜性别的最好准确率只有 0.551,不能用于推断个人。 (https://5dlibrary.com/answers/zh/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c06/) - Ch.7 Eastern and Western Classical Sources: Knowing and Doing, Within and Without, the Thought in the Heart — A disposition that shows itself with probability 0.05 on each occasion has a showing probability of only 0.050 in one occasion, 0.226 in five and 0.642 in twenty; a disposition with probability 0.50 per occasion shows itself with probability 0.969 in five. As for precision, if the true rate of keeping promises is seven in ten, the 95 % interval after 5 observations is 0.598 wide, after 10 it is 0.500, after 20 it is 0.376, after 50 it is 0.248, and after 100 it is still 0.178. For what is within to "take form without" clearly takes a great many real acts. (https://5dlibrary.com/answers/en/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c07/) 第7章 东西方原典考辨:知行、内外与心中之念 — 一种每次以 0.05 概率显现的倾向,一次机会里只有 0.050 的显现概率,五次机会 0.226,二十次 0.642;每次概率为 0.50 的倾向,五次机会就有 0.969。估计精度方面,若真实守约率为七成,观察 5 次后的 95 % 区间宽 0.598,10 次后 0.500,20 次后 0.376,50 次后 0.248,100 次后仍有 0.178。所谓「形于外」,需要很多次真实行为才能「形」得清楚。 (https://5dlibrary.com/answers/zh/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c07/) - Ch.8 Closing the Loop: Field Tools and Checker Specifications — Chapter 8 delivers the Simulation-Enactment Gap Checker, checker specifications grouped by the four kits, a ledger rhythm and pre-registered sample sizes. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c08/) 第8章 实证闭环:实战工具与检查器规格 — 第八章给出「心智模拟与行为兑现缺口判别器」、按四件套分组的检查器规格、台账节奏与预注册样本量。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/mental-simulation-versus-enactment-intimacy-dopaminergic-fantasy-oxytocin-commitment-costs/c08/) ## 5DL-B-431 · Step-Down Communication Protocols for High-Clock-Speed Minds: Cognitive-Bandwidth Matching, Inference-Gap Bridging and Zero-Friction Alignment 中文书名:高转速心智降速通信协议:跨认知带宽阻抗匹配、跳步补全与零挫败协同母典 DOI: pending Cite: Fan, Y. (2026). Step-Down Communication Protocols for High-Clock-Speed Minds: Cognitive-Bandwidth Matching, Inference-Gap Bridging and Zero-Friction Alignment (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: Its parameters are illustrative; all calculations are generated by its programs; the book offers no psychological assessment or clinical judgement, names no real person or case, and is general, not personalised, advice on business communication. Chapter 8 gives the step-down ledger, a field checker tool with its specifications, and a pre-registered test with 198 records per arm; in the synthetic ledger green records have a rework rate of 0.115 against 0.520 for red ones. Abstract: A popular belief holds that people who think fast skip steps because they are cleverer, that receivers who cannot keep up are slow, and that the fix is to slow everything down and explain everything in more detail. This book refutes the second half of that belief. After checking the literature it corrects its commissioning blueprint: "high clock speed" means only that the same chain of inference occupies fewer chunks for the sender, a gap in chunk compression and domain expertise rather than a ranking of intelligence; Shannon's channel-capacity formula serves only as a structural analogy, and the testable quantity is the loss rate given by a model of step-by-step reconstruction; spelling out every step to a receiver in the same domain triggers expertise reversal, so the best number of bridged steps must be computed from the receiver's dictionary; and "lossless" or "zero friction" can only be stated as an upper bound on residual misunderstanding after a given number of read-back rounds. Its parameters are illustrative; all calculations are generated by its programs; the book offers no psychological assessment or clinical judgement, names no real person or case, and is general, not personalised, advice on business communication. Chapter 1 shows with chess chunks that fast and slow reverse with the domain. Chapter 2 builds the reconstruction model: with each skipped premise completed independently with probability 0.95, reconstruction alone loses 0.143 of three-step jumps, and working-memory overflow raises the loss to 0.518. Chapter 3 computes the best number of bridged steps: novices are best served by bridging everything, a practised receiver by two steps, and full bridging raises the latter's failure rate from 0.344 to 0.401; probing first beats the best fixed policy only at a classification accuracy of 0.81 or more. Chapter 4 treats read-back as a repetition code and shows that a ledger with no residual misunderstanding in sixty sessions still has an upper bound of 0.0487. Chapter 5 shows that one instruction per message is a conditional rule, best only when the overhead of a message lies below a threshold, 0.4877 in the illustration. Chapter 6 computes the curse of knowledge: anchoring on one's own dictionary makes the bridging decision jump, with an excess cost of 3.734, and disdain towards the person who asks raises the equilibrium miss rate from 0.2940 to 0.3600. Chapter 7 examines "do not skip the joints" in the Xue Ji, "no opening for one who is not eager" in the Analects, Aristotle on the shortened syllogism and Descartes' seventh Rule; teaching in order keeps 8.76 units in an unbroken run on average against 2.28 for mixed teaching. Chapter 8 gives the step-down ledger, a field checker tool with its specifications, and a pre-registered test with 198 records per arm; in the synthetic ledger green records have a rework rate of 0.115 against 0.520 for red ones. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:协作里流行一种说法:思维快的人说话跳步是因为更聪明,接收方跟不上是因为慢,解法是一律放慢、讲得越细越好。本书证伪这一说法的后一半。核对文献后,本书把立项蓝图修正为:「高转速」只指同一条推理链在发送方那里占的组块更少,是组块压缩率与领域专长的差,不是智力排序;香农信道容量公式只作结构类比,可检验的是逐步重建模型给出的丢包率;对同领域接收方补满步骤会触发专长逆转,最优补步数要按接收方字典去算;「无损」「零挫败」只能写成给定回授轮数下残余误解率的上界。书中参数为示意;全部计算由书中程序生成;本书不作心理评估或临床判断,不写真实人物与案例,是非个人化的经营与沟通建议。 第一章用象棋组块与四本字典说明快慢随领域翻转。第二章建立逐步重建模型:每个被跳过的前提独立补全概率取 0.95 时,只看重建,跳三步丢失 0.143,计入工作记忆溢出升到 0.518。第三章求最优补步数:新手补满最好,熟手补两步最好,补满使失败率从 0.344 升到 0.401;先问一句再补步,只有判对水平的准确率达到 0.81 以上才胜过固定策略。第四章把回授写成重复码,比较逐字复述与类比复述,并说明台账六十次会话零残余误解时,上界仍是 0.0487。 第五章说明单次单指令是条件命题:消息开销低于门槛才最优,示意门槛为 0.4877。第六章计算知识的诅咒:锚定在自己字典上会让补步决策跳变,额外成本达 3.734;贬低提问者使均衡误解率从 0.2940 升到 0.3600。第七章考辨《礼记·学记》「不陵节」、《论语·述而》「不愤不启」、亚里士多德论省略推论与笛卡尔第七规则,按节施教平均连续掌握 8.76 个单元,杂施只有 2.28 个。第八章给出降速通信台账、实战工具「跨带宽降速通信与跳步补全校验器」与检查器规格、每组 198 条记录的预注册检验,合成台账中绿灯记录返工率 0.115、红灯 0.520。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 High Clock Speed Is Not an Intelligence Ranking: Chunk Compression and Domain Expertise — This chapter rewrote high clock speed, replacing a ranking of intelligence with a gap in chunk compression. The number of slots is broadly the same, about four; the difference lies in how many steps each slot carries. The evidence from the chessboard and the model in p12 both show that the same slots with different chunk sizes are enough to create a fourfold difference in recall, and that the difference vanishes on random positions. Program p11 adds that the compression gap reverses with the domain, and that one of the sender's chunks unpacks on average into 2.67 of the receiver's. (https://5dlibrary.com/answers/en/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c01/) 第1章 高转速不是智力排序:组块压缩与领域专长差 — 本章把「高转速」从智力等级改写为组块压缩率差:槽位数大致相同,约四个;差别在每个槽位装多少步。棋盘证据与 p12 的模型都表明,同样的槽位、不同的组块大小,足以产生四倍的复原差,而随机局面上差别消失。p11 进一步显示,压缩率差依领域翻转,发送方的一块落到接收方那里平均要拆成 2.67 块。 (https://5dlibrary.com/answers/zh/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c01/) - Ch.2 A Step-by-Step Reconstruction Model: Skipped Steps, Loss Rates and Working-Memory Overflow — Chapter 2 builds the reconstruction model: with each skipped premise completed independently with probability 0.95, reconstruction alone loses 0.143 of three-step jumps, and working-memory overflow raises the loss to 0.518. (https://5dlibrary.com/answers/en/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c02/) 第2章 逐步重建模型:跳步、丢包率与工作记忆溢出 — 第二章建立逐步重建模型:每个被跳过的前提独立补全概率取 0.95 时,只看重建,跳三步丢失 0.143,计入工作记忆溢出升到 0.518。 (https://5dlibrary.com/answers/zh/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c02/) - Ch.3 Expertise Reversal: The Best Number of Bridging Steps Depends on the Receiver's Chunk Dictionary — Chapter 3 computes the best number of bridged steps: novices are best served by bridging everything, a practised receiver by two steps, and full bridging raises the latter's failure rate from 0.344 to 0.401; probing first beats the best fixed policy only at a classification accuracy of 0.81 or more. (https://5dlibrary.com/answers/en/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c03/) 第3章 专长逆转:最优补步数由接收方组块字典决定 — 第三章求最优补步数:新手补满最好,熟手补两步最好,补满使失败率从 0.344 升到 0.401;先问一句再补步,只有判对水平的准确率达到 0.81 以上才胜过固定策略。 (https://5dlibrary.com/answers/zh/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c03/) - Ch.4 Read-Back Checks: Repetition Codes, Analogy Restatement and an Upper Bound on Residual Misunderstanding — Chapter 4 treats read-back as a repetition code and shows that a ledger with no residual misunderstanding in sixty sessions still has an upper bound of 0.0487. (https://5dlibrary.com/answers/en/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c04/) 第4章 回授校验:重复码、类比复述与残余误解率上界 — 第四章把回授写成重复码,比较逐字复述与类比复述,并说明台账六十次会话零残余误解时,上界仍是 0.0487。 (https://5dlibrary.com/answers/zh/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c04/) - Ch.5 Shannon's Formula as Analogy Only: Channel Capacity, the Central Bottleneck and One Instruction per Message — Chapter 5 shows that one instruction per message is a conditional rule, best only when the overhead of a message lies below a threshold, 0.4877 in the illustration. (https://5dlibrary.com/answers/en/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c05/) 第5章 香农公式只作类比:信道容量、中枢瓶颈与单次单指令 — 第五章说明单次单指令是条件命题:消息开销低于门槛才最优,示意门槛为 0.4877。 (https://5dlibrary.com/answers/zh/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c05/) - Ch.6 The Curse of Knowledge and the Conversation Game: Conclusion First, the Price of Asking and Total Coordination Time — Chapter 6 computes the curse of knowledge: anchoring on one's own dictionary makes the bridging decision jump, with an excess cost of 3.734, and disdain towards the person who asks raises the equilibrium miss rate from 0.2940 to 0.3600. (https://5dlibrary.com/answers/en/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c06/) 第6章 知识的诅咒与会话博弈:结论先行、提问成本与协同总耗时 — 第六章计算知识的诅咒:锚定在自己字典上会让补步决策跳变,额外成本达 3.734;贬低提问者使均衡误解率从 0.2940 升到 0.3600。 (https://5dlibrary.com/answers/zh/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c06/) - Ch.7 Classical Sources East and West: the Xue Ji on Not Skipping Joints, the Analects on Readiness, Aristotle's Shortened Syllogism and Descartes' Seventh Rule — Chapter 7 examines "do not skip the joints" in the Xue Ji, "no opening for one who is not eager" in the Analects, Aristotle on the shortened syllogism and Descartes' seventh Rule; teaching in order keeps 8.76 units in an unbroken run on average against 2.28 for mixed teaching. (https://5dlibrary.com/answers/en/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c07/) 第7章 东西方原典考辨:《学记》「不陵节」、《论语》「不愤不启」、亚里士多德省略推论与笛卡尔第七规则 — 第七章考辨《礼记·学记》「不陵节」、《论语·述而》「不愤不启」、亚里士多德论省略推论与笛卡尔第七规则,按节施教平均连续掌握 8.76 个单元,杂施只有 2.28 个。 (https://5dlibrary.com/answers/zh/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c07/) - Ch.8 The Empirical Loop: A Step-Down Ledger, Field Tools and Checker Specifications — Chapter 8 gives the step-down ledger, a field checker tool with its specifications, and a pre-registered test with 198 records per arm; in the synthetic ledger green records have a rework rate of 0.115 against 0.520 for red ones. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c08/) 第8章 实证闭环:降速通信台账与实战工具与检查器规格 — 第八章给出降速通信台账、实战工具「跨带宽降速通信与跳步补全校验器」与检查器规格、每组 198 条记录的预注册检验,合成台账中绿灯记录返工率 0.115、红灯 0.520。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/step-down-communication-cognitive-bandwidth-matching-inference-gap-bridging/c08/) ## 5DL-B-432 · Zero Explanation-Tax Dynamics: Abandoning Validation-Seeking, Full-Power Reality Rendering and Proof-by-Artifact Gravity 中文书名:零解释税生存动力学:放弃低维认同求证、全功率自我显化与作品引力场母典 DOI: pending Cite: Fan, Y. (2026). Zero Explanation-Tax Dynamics: Abandoning Validation-Seeking, Full-Power Reality Rendering and Proof-by-Artifact Gravity (Version 1.0.0). 5D Library. DOI pending Book page: after publication Note: Illustrative model calculation, not a live quote. / 模型示意算例,非实时报价。 Scope: All parameters are illustrative and all calculations are generated by the book's programs. The book is general business information, not individual advice, psychological counselling or clinical judgement; it makes no promise of income, and rendering an idea means only turning it into a deliverable, verifiable artifact. Abstract: Two popular sayings hold either that anyone who explains clearly enough will in the end be understood by everyone, or, conversely, that projecting high-dimensional information onto a low dimension must lose information, so being understood by everyone is mathematically impossible. This book rewrites both as claims that data can refute. After checking the sources, it corrects its blueprint in four places: projection loss is set by the eigenvalue spectrum and can be small when information lies near a low-dimensional subspace; what holds strictly is that the intersection of many receivers' subspaces shrinks as an audience widens, stated as a bound and a probability; explaining is an investment with high-overlap audiences and a tax only with low-overlap ones; and the gravity of artifacts is preferential attachment plus homophilous screening plus noise, reported as quantiles and never as determinism. All parameters are illustrative and all calculations are generated by the book's programs. The book is general business information, not individual advice, psychological counselling or clinical judgement; it makes no promise of income, and rendering an idea means only turning it into a deliverable, verifiable artifact. Chapter 1 uses the theorem of Eckart and Young to show that a steep-spectrum work keeping 6 dimensions loses only 0.0446 of its energy, and that three retellings leave 0.4277 bits of each bit. Chapter 2 proves that the energy all receivers take in without loss is no more than the worst receiver gets, and gives the geometric decay of the common dimension with audience size. Chapter 3 finds the break-even threshold for explaining, 0.464 under illustrative parameters; explaining only above it has a net value of 320.0, explaining to everyone a net value of minus 203.8. Chapter 4 connects the spotlight effect to the decision to explain: of the explanations given under doubled perception, 0.736 are unnecessary. Chapter 5 compares artifacts with arguments as signals: a receiver judging from 4 verifiable samples reaches an accuracy of 0.770, which takes 66 self-descriptions to match. The gravity model of Chapter 6 shows the 90th and 10th percentiles of ally counts for creators of equal quality differing by a factor of 8.43, with social influence amplifying both inequality and unpredictability. Chapter 7 examines the Analects on remaining unvexed when not known, the Zhuangzi on not being spurred by the world's praise, Plato's cave and section 13 of Epictetus's Handbook, and finds that being unmoved by praise and blame wins only when an audience's mismatch exceeds about 0.245. Chapter 8 gives the explanation-tax and artifact ledger, the distilled Explanation-Tax Accountant and Artifact-Gravity Discriminator, checker specifications grouped by four business kits, and a pre-registered test needing 294 people per group. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. 摘要:流行说法有两句:只要解释得足够清楚,所有人终会理解;或者反过来,高维信息向低维投影必然损失,所以「被所有人理解」在数学上无解。本书把两句都改写成可以用数据推翻的命题。核对文献后,本书对立项蓝图作了四处修正:投影损失由特征值谱决定,信息贴近低维子空间时损失可以很小;严格成立的是多位接收者子空间的交集随受众变宽而收缩,写成上界与概率;解释对高重叠受众是投资、对低重叠受众才是税;「作品引力」是带随机项的优先连接加同质筛选,只给分位数,不写决定论。书中参数一律为示意,全部计算由书中程序生成;本书是非个人化经营建议,不是心理咨询或临床判断,不作任何收益承诺,「显化」只指把构想做成可交付、可核验的作品。 第一章用 Eckart 与 Young 定理说明陡谱作品留 6 维只丢 0.0446 的能量,转述链三次后每比特只剩 0.4277。第二章证明全体接收者共同零损失接收的能量不超过最差者所得,并给出共同维数随人数的几何衰减。第三章求出解释的盈亏阈值,示意参数下为 0.464,只对阈值以上受众解释的净值是 320.0,对所有人解释是负 203.8。第四章把聚光灯效应接到解释决策上,按两倍感知给出的解释里有 0.736 是多余的。 第五章比较作品与辩护两类信号:看 4 件可核验样本的判断准确率是 0.770,用自述达到同样准确率需要 66 句。第六章的引力模型显示,质量相同的创作者盟友数第九十与第十百分位相差 8.43 倍,社会影响同时放大不平等与不可预测性。第七章考辨《论语》「人不知而不慍」、《庄子·逍遥游》「举世誉之而不加劝」、柏拉图洞穴喻与爱比克泰德《手册》第十三节,算出「不为毁誉所动」在受众错位大于约 0.245 时才占优。第八章给出解释税与作品台账、蒸馏小工具「解释税核算与作品引力判别器」、按四件套分组的检查器规格与预注册检验,核心假设每组需 294 人。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 Chapter answers: - Ch.1 Projection Loss: the Real Bill for Rank Truncation, Retelling Chains and Random Projection — Chapter 1 uses the theorem of Eckart and Young to show that a steep-spectrum work keeping 6 dimensions loses only 0.0446 of its energy, and that three retellings leave 0.4277 bits of each bit. (https://5dlibrary.com/answers/en/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c01/) 第1章 投影损失:秩截断、转述链与随机投影的真实账单 — 第一章用 Eckart 与 Young 定理说明陡谱作品留 6 维只丢 0.0446 的能量,转述链三次后每比特只剩 0.4277。 (https://5dlibrary.com/answers/zh/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c01/) - Ch.2 The Intersection of Many Receivers: a Bound and a Probability for "Understood by Everyone" — In general position, each additional receiver removes $d-k$ dimensions from the intersection, and with $k = 10$ the sixth receiver empties the general-position part. Aligned audiences are much more favourable, yet the expected common dimension still falls geometrically with size: the probability of sharing at least one dimension drops below one half at the sixth member of a broad audience, the twentieth of a narrow one and the thirty-sixth of a mixed audience with a firm core. The energy of any message that all receivers can take in without loss is no more than what the worst receiver gets. (https://5dlibrary.com/answers/en/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c02/) 第2章 多接收者交集:「被所有人理解」的上界与概率 — 一般位置下,每多一位接收者,交集就少 $d-k$ 维,$k = 10$ 时第 6 位接收者让一般位置部分归零。对齐的受众乐观得多,但期望共同维数仍随人数几何递减:宽受众第 6 人、窄受众第 20 人、有牢固核心的混合受众第 36 人,「至少共享一维」的概率跌破一半。任何消息让全体接收者零损失接收的能量,不超过最差接收者所得。 (https://5dlibrary.com/answers/zh/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c02/) - Ch.3 Tax or Investment: the Net Value of Explaining and the Overlap Threshold — Chapter 3 finds the break-even threshold for explaining, 0.464 under illustrative parameters; explaining only above it has a net value of 320.0, explaining to everyone a net value of minus 203.8. (https://5dlibrary.com/answers/en/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c03/) 第3章 税还是投资:解释的净现值与重叠度阈值 — 第三章求出解释的盈亏阈值,示意参数下为 0.464,只对阈值以上受众解释的净值是 320.0,对所有人解释是负 203.8。 (https://5dlibrary.com/answers/zh/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c03/) - Ch.4 Where the Urge to Explain Comes From: Self-Verification, the Sociometer and the Spotlight — Chapter 4 connects the spotlight effect to the decision to explain: of the explanations given under doubled perception, 0.736 are unnecessary. (https://5dlibrary.com/answers/en/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c04/) 第4章 解释冲动从哪里来:自我验证、社会计量器与聚光灯 — 第四章把聚光灯效应接到解释决策上,按两倍感知给出的解释里有 0.736 是多余的。 (https://5dlibrary.com/answers/zh/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c04/) - Ch.5 Artifacts and Arguments: Costly Signals, the Handicap Principle and Cheap Talk — Chapter 5 compares artifacts with arguments as signals: a receiver judging from 4 verifiable samples reaches an accuracy of 0.770, which takes 66 self-descriptions to match. The gravity model of Chapter 6 shows the 90th and 10th percentiles of ally counts for creators of equal quality differing by a factor of 8.43, with social influence amplifying both inequality and unpredictability. (https://5dlibrary.com/answers/en/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c05/) 第5章 作品与辩护:有成本信号、累赘原理与廉价话语 — 第五章比较作品与辩护两类信号:看 4 件可核验样本的判断准确率是 0.770,用自述达到同样准确率需要 66 句。 (https://5dlibrary.com/answers/zh/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c05/) - Ch.6 The Gravity of Artifacts: Preferential Attachment, Homophilous Screening and an Irreducible Random Term — The gravity of artifacts consists of two mechanisms. Homophilous screening decides who comes, raising the share of nearby tastes from a random 0.263 to as much as 0.849. Cumulative advantage decides how many come, and spreads the ally counts of equal-quality creators by a ratio of 8.43. Social influence amplifies inequality and unpredictability together; at influence 0.9 unpredictability is more than twice the independent level. Moving explaining time into artifacts raises the median ally count from 125 to 263 with a low-overlap audience, but does not necessarily win with a high-overlap one. (https://5dlibrary.com/answers/en/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c06/) 第6章 作品引力:优先连接、同质筛选与不可消除的随机项 — 作品引力由两个机制构成:同质筛选决定来的是谁,口味相近者占比从随机的 0.263 升到 0.849;累积优势决定来多少,同时把质量相同者的盟友数拉开到 8.43 倍。社会影响放大不平等,也放大不可预测性:影响 0.9 时不可预测性是独立条件的两倍有余。把解释时间改投作品,在低重叠受众下盟友中位数从 125 升到 263,在高重叠受众下却未必占优。 (https://5dlibrary.com/answers/zh/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c06/) - Ch.7 Classical Sources East and West: Unvexed, Unspurred, the Shadows of the Cave, and Content to Be Thought Foolish — Chapter 7 examines the Analects on remaining unvexed when not known, the Zhuangzi on not being spurred by the world's praise, Plato's cave and section 13 of Epictetus's Handbook, and finds that being unmoved by praise and blame wins only when an audience's mismatch exceeds about 0.245. (https://5dlibrary.com/answers/en/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c07/) 第7章 东西方原典考辨:不慍、不加勸、洞穴之影与甘于被看作无知 — 第七章考辨《论语》「人不知而不慍」、《庄子·逍遥游》「举世誉之而不加劝」、柏拉图洞穴喻与爱比克泰德《手册》第十三节,算出「不为毁誉所动」在受众错位大于约 0.245 时才占优。 (https://5dlibrary.com/answers/zh/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c07/) - Ch.8 The Empirical Loop: Field Tools and Checker Specifications — Chapter 8 gives the explanation-tax and artifact ledger, the distilled Explanation-Tax Accountant and Artifact-Gravity Discriminator, checker specifications grouped by four business kits, and a pre-registered test needing 294 people per group. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses. (https://5dlibrary.com/answers/en/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c08/) 第8章 实证闭环:实战工具与检查器规格 — 第八章给出解释税与作品台账、蒸馏小工具「解释税核算与作品引力判别器」、按四件套分组的检查器规格与预注册检验,核心假设每组需 294 人。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。 (https://5dlibrary.com/answers/zh/zero-explanation-tax-validation-abandonment-proof-by-artifact-gravity/c08/)