关于「实战:一份三年独家供货合约,五个声音两两打架」,《多维神谕冲突消解与权重裁决》给出的定量结论是什么?

结论:第八章让一位虚构的代工厂老板面对一份三年独家供货合约:三票攻、两票守,月尺度上的合成恰好是 -0.5000,裁决为收缩;预先登记的那一项买方违约若被查实,记 -1.0,裁决变为加固防线。把权重放进合理的怀疑范围,对冲与收缩几乎各占一半,于是落成一份一年期、非独家、有银行保函的合约,顺利时少赚 RM 41,200,对方停单时少亏 RM 406,200;盲测账本要 240 次有结果的判定,才有 0.8795 的把握把合成与抛硬币分开。

数据口径:本书不重新起盘,只接收前面各卷交来的一张票,用对数几率把五张票按「问事类型 × 时间尺度」加权合成;读同一份出生资料的三个系统与读发问时刻的两个系统,各自按设计效应 1 + (n - 1)ρ 打折;合成值落到九级行动阶梯上,再配一段对客户说的合成话术。 全部数字由书中程序生成;本书不是投资、法律、医疗或心理意见。

English version

对照数据表

下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: -0.5000, -1.0

votes: bazi attack, ziwei attack, qimen defend, liuyao defend, west attack
head count3 attack, 2 defend -> a plain majority would say attack
scale birth raw/n/part moment raw/n/part L p verdict
natal +1.76/3/+0.8800-0.29/2/-0.2231 +0.6569 0.6586 probe
decade +1.64/3/+0.8200-0.48/2/-0.3692 +0.4508 0.6108 hedge
year +1.50/3/+0.7500-0.86/2/-0.6615 +0.0885 0.5221 hedge
month +1.20/3/+0.6000-1.43/2/-1.1000 -0.5000 0.3775 trim <- pre-registered
day +0.84/3/+0.4200-1.81/2/-1.3923 -0.9723 0.2744 trim
moment +0.76/3/+0.3800-1.90/2/-1.4615 -1.0815 0.2532 trim
month, unrounded float sum-0.5000000000000001; rounded to 4 places: -0.5000; line -0.5 -> trim
tie rule of the ladder: L = +0.5 opens probe, L-0.5 opens trim (no tie lands in the middle)
evidence gate at the month scale (votes alone L -0.5000, trim):
probe needs evidence >= +1.0000 (check: probe)
staged_attack needs evidence >= +1.7000 (check: staged_attack)
all_in needs evidence >= +2.5000 (check: all_in)
fortify needs evidence <-0.7000 (check: fortify)
retreat needs evidence <-1.5000 (check: retreat)
pre-registered fact: buyer already 60+ days overdue to any supplier, evidence -1.0
-> L -1.5000, fortify (the boss's own 0.70 is a hunch, not evidence: it stays 0)
查看完整验算程序
# Chapter 8. p80: The case. An invented contract manufacturer is asked to sign a three-year exclusive deal.
# Signals (stipulated for the case, each from the reading rules of an earlier volume): bazi attack,
# ziwei attack, qimen defend, liuyao defend, west attack. Question type: contract. The question as
# pre-registered on 2026-11-16 reads "sign by 30 November or not", so its time scale is month. The program
# prints the verdict at all six scales to show what a scale chosen after the fact would buy, the unrounded
# sum at the chosen scale, and what a plain head count of the votes would have said. Rule restated from
# reference/synthesis_rules.py, including its evidence gate: posterior(..., evidence=0.0) adds the log-odds
# of a hard business fact (signed order, cash received, judgment or regulatory decision in force, default
# that has happened) to the votes. The case pre-registers one such fact and its value before any reading.
import math

SYSTEMS = ["bazi", "ziwei", "qimen", "liuyao", "west"]
SCALES = ["natal", "decade", "year", "month", "day", "moment"]
VALUE = {"attack": 1, "defend": -1, "hold": 0}
WEIGHT = {"bazi": 0.6, "ziwei": 0.8, "qimen": 1.0, "liuyao": 0.9, "west": 0.6}   # the contract row
FIT = {
    "bazi":   {"natal": 1.0, "decade": 0.9, "year": 0.7, "month": 0.5, "day": 0.3, "moment": 0.2},
    "ziwei":  {"natal": 1.0, "decade": 1.0, "year": 0.9, "month": 0.6, "day": 0.3, "moment": 0.2},
    "qimen":  {"natal": 0.2, "decade": 0.3, "year": 0.5, "month": 0.8, "day": 1.0, "moment": 1.0},
    "liuyao": {"natal": 0.1, "decade": 0.2, "year": 0.4, "month": 0.7, "day": 0.9, "moment": 1.0},
    "west":   {"natal": 0.6, "decade": 0.5, "year": 0.6, "month": 0.7, "day": 0.7, "moment": 0.8},
}
GROUPS = {"birth": (["bazi", "ziwei", "west"], 0.5), "moment": (["qimen", "liuyao"], 0.3)}
CASE = {"bazi": "attack", "ziwei": "attack", "qimen": "defend", "liuyao": "defend", "west": "attack"}


def posterior(signals, scale, evidence=0.0):
    total, parts = float(evidence), {}
    for g, (members, rho) in GROUPS.items():
        raw = sum(WEIGHT[s] * FIT[s][scale] * VALUE[signals[s]] for s in members)
        n = sum(1 for s in members if signals[s] != "hold")
        part = raw / (1 + (n - 1) * rho) if n > 1 else raw
        parts[g] = (raw, n, part)
        total += part
    L = round(total, 4)
    v = list(signals.values())
    if L >= 2.0:
        act = "all_in"
    elif L >= 1.2:
        act = "staged_attack"
    elif L >= 0.5:
        act = "probe"
    elif L > -0.5:
        if "attack" in v and "defend" in v:
            act = "hedge"
        else:
            act = "hold" if v.count("hold") >= 3 else "trigger_wait"
    elif L > -1.2:
        act = "trim"
    elif L > -2.0:
        act = "fortify"
    else:
        act = "retreat"
    return total, L, parts, act


if __name__ == "__main__":
    v = list(CASE.values())
    print("votes: " + ", ".join("{} {}".format(s, CASE[s]) for s in SYSTEMS))
    print("head count: {} attack, {} defend -> a plain majority would say attack".format(
        v.count("attack"), v.count("defend")))
    print(" scale    birth raw/n/part         moment raw/n/part        L        p       verdict")
    for t in SCALES:
        total, L, parts, act = posterior(CASE, t)
        b, m = parts["birth"], parts["moment"]
        print(" {:<7}  {:+.2f}/{}/{:+.4f}          {:+.2f}/{}/{:+.4f}          {:+.4f}  {:.4f}  {}{}".format(
            t, b[0], b[1], b[2], m[0], m[1], m[2], L, 1 / (1 + math.exp(-L)), act,
            "   <- pre-registered" if t == "month" else ""))
    total, L, parts, act = posterior(CASE, "month")
    print("month, unrounded float sum: {!r}; rounded to 4 places: {:+.4f}; line -0.5 -> {}".format(
        total, L, act))
    print("tie rule of the ladder: L = +0.5 opens probe, L = -0.5 opens trim (no tie lands in the middle)")
    print("evidence gate at the month scale (votes alone L {:+.4f}, {}):".format(L, act))
    for rung, line in (("probe", 0.5), ("staged_attack", 1.2), ("all_in", 2.0)):
        e = round(line - L, 4)
        check = posterior(CASE, "month", e)[3]
        print("   {:<14} needs evidence >= {:+.4f}  (check: {})".format(rung, e, check))
    for rung, line in (("fortify", -1.2), ("retreat", -2.0)):
        e = round(line - L, 4)
        check = posterior(CASE, "month", e)[3]
        print("   {:<14} needs evidence <= {:+.4f}  (check: {})".format(rung, e, check))
    e = -1.0
    total, L2, parts, act2 = posterior(CASE, "month", e)
    print("pre-registered fact: buyer already 60+ days overdue to any supplier, evidence {:+.1f}".format(e))
    print("   -> L {:+.4f}, {} (the boss's own 0.70 is a hunch, not evidence: it stays 0)".format(L2, act2))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《多维神谕冲突消解、权重裁决与总参谋行动处方母典》第 8 章「实战:一份三年独家供货合约,五个声音两两打架」(版本 0.1.0). 5D Library. DOI 待发布

DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-132 · v0.1.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026132,
  author    = {Fan, Yafu},
  title     = {The Multi-Oracle Conflict-Resolution and Chief-of-Staff Action Codex},
  year      = {2026},
  publisher = {5D Library},
  edition   = {0.1.0},
  note      = {DOI pending}
}

本书其他章节结论