关于「六类问事的权重表:五个声音各管哪一段」,《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:第二章定权重,改用等权会让 8,748 种裁决里的 1334 种换档。

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: 1334, 8748

rung counts per question over 1458 cases (243 patterns x 6 scales)
question all_in staged probe hedge hold trim fortif retrea trigge
money0 63 349 562 66 349 63 0 6
contract0 94 314 574 64 314 94 0 4
partner0 75 335 586 50 335 75 0 2
love0 61 333 626 44 333 61 0 0
launch0 76 324 590 64 324 76 0 4
relocate0 57 346 582 64 346 57 0 6
cases that change rung when one weight moves by -0.1 / +0.1 (out of 1458)
question bazi ziwei qimen liuyao west busiest
money34/48 46/50 64/74 68/50 50/62 qimen
contract52/48 44/50 86/58 64/50 48/72 qimen
partner56/36 58/58 50/84 80/38 62/76 west
love50/42 58/52 76/64 60/72 62/52 qimen
launch38/40 54/38 82/58 56/52 48/42 qimen
relocate40/34 46/42 62/74 46/50 48/58 qimen
equal-weight counterfactual: every cell set to the table mean 0.773
money242 of 1458 cases change rung (0.166)
contract256 of 1458 cases change rung (0.176)
partner182 of 1458 cases change rung (0.125)
love254 of 1458 cases change rung (0.174)
launch200 of 1458 cases change rung (0.137)
relocate200 of 1458 cases change rung (0.137)
all questions1334 of 8748 (0.152)
monotonicity audit: one hold turned into attack; cases where L falls instead of rising
630 of 14580 added attack votes lower L (0.043)
largest drop 0.1767: money decade west added, L 1.49901.3223
查看完整验算程序
# Chapter 2. p23: Which weights carry the verdicts: rung counts, +/-0.1 nudges, equal weights, monotonicity.
# Every question type is run over all 3^5 = 243 vote patterns at all 6 scales (1458 cases per question).
# A nudge moves one cell of WEIGHT by +0.1 or -0.1 and counts the cases whose rung changes. The equal-weight
# counterfactual sets every cell to the table mean (Dawes 1979: unit weights are hard to beat).
# Monotonicity audit: turn one hold into an attack; does L go DOWN? With the design-effect divisor
# 1 + (n - 1) rho, a weak agreeing vote can pull a group's total below what the stronger votes gave alone.
import itertools

SYSTEMS = ["bazi", "ziwei", "qimen", "liuyao", "west"]
QTYPES = ["money", "contract", "partner", "love", "launch", "relocate"]
SCALES = ["natal", "decade", "year", "month", "day", "moment"]
ACTIONS = ["all_in", "staged_attack", "probe", "hedge", "hold", "trim", "fortify", "retreat", "trigger_wait"]
VALUE = {"attack": 1, "defend": -1, "hold": 0}
WEIGHT = {
    "money":    {"bazi": 0.9, "ziwei": 1.0, "qimen": 0.8, "liuyao": 0.7, "west": 0.5},
    "contract": {"bazi": 0.6, "ziwei": 0.8, "qimen": 1.0, "liuyao": 0.9, "west": 0.6},
    "partner":  {"bazi": 0.8, "ziwei": 1.0, "qimen": 0.7, "liuyao": 0.8, "west": 0.7},
    "love":     {"bazi": 0.7, "ziwei": 1.0, "qimen": 0.5, "liuyao": 0.8, "west": 0.9},
    "launch":   {"bazi": 0.7, "ziwei": 0.8, "qimen": 1.0, "liuyao": 0.7, "west": 0.6},
    "relocate": {"bazi": 0.8, "ziwei": 0.9, "qimen": 0.9, "liuyao": 0.6, "west": 0.5},
}
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)}
THRESH = (2.0, 1.2, 0.5)
PATTERNS = [dict(zip(SYSTEMS, c)) for c in itertools.product(["attack", "defend", "hold"], repeat=5)]


def logodds(sig, w, t):
    total = 0.0
    for members, rho in GROUPS.values():
        raw = sum(w[s] * FIT[s][t] * VALUE[sig[s]] for s in members)
        n = sum(1 for s in members if sig[s] != "hold")
        total += raw / (1 + (n - 1) * rho) if n > 1 else raw
    return round(total, 4)


def rung(sig, L):
    v = [sig[s] for s in SYSTEMS]
    big, mid, small = THRESH
    if L >= big:
        return "all_in"
    if L >= mid:
        return "staged_attack"
    if L >= small:
        return "probe"
    if L > -small:
        if "attack" in v and "defend" in v:
            return "hedge"
        return "hold" if v.count("hold") >= 3 else "trigger_wait"
    if L > -mid:
        return "trim"
    return "fortify" if L > -big else "retreat"


def verdicts(q, w):
    return [rung(p, logodds(p, w, t)) for t in SCALES for p in PATTERNS]


if __name__ == "__main__":
    base = {q: verdicts(q, WEIGHT[q]) for q in QTYPES}
    print("   rung counts per question over 1458 cases (243 patterns x 6 scales)")
    print("   question  " + "".join("{:>7s}".format(a[:6]) for a in ACTIONS))
    for q in QTYPES:
        print("   {:9s} ".format(q) + "".join("{:7d}".format(base[q].count(a)) for a in ACTIONS))
    print()
    print("   cases that change rung when one weight moves by -0.1 / +0.1 (out of 1458)")
    print("   question  " + "".join("{:>12s}".format(s) for s in SYSTEMS) + "   busiest")
    for q in QTYPES:
        cells, worst = [], (0, "")
        for s in SYSTEMS:
            out = []
            for d in (-0.1, 0.1):
                w = dict(WEIGHT[q])
                w[s] = round(w[s] + d, 1)
                n = sum(a != b for a, b in zip(base[q], verdicts(q, w)))
                out.append(n)
            worst = max(worst, (sum(out), s))
            cells.append("{:4d}/{:<4d}".format(*out))
        print("   {:9s} ".format(q) + "".join("{:>12s}".format(c) for c in cells) + "   " + worst[1])
    mean = round(sum(WEIGHT[q][s] for q in QTYPES for s in SYSTEMS) / 30, 3)
    flat = {s: mean for s in SYSTEMS}
    print()
    print("   equal-weight counterfactual: every cell set to the table mean {:.3f}".format(mean))
    total = 0
    for q in QTYPES:
        n = sum(a != b for a, b in zip(base[q], verdicts(q, flat)))
        total += n
        print("   {:9s} {:4d} of 1458 cases change rung ({:.3f})".format(q, n, n / 1458))
    print("   all questions: {} of {} ({:.3f})".format(total, 1458 * 6, total / (1458 * 6)))
    print()
    print("   monotonicity audit: one hold turned into attack; cases where L falls instead of rising")
    falls, tries, worst = 0, 0, (0.0, "")
    for q in QTYPES:
        for t in SCALES:
            for p in PATTERNS:
                for s in SYSTEMS:
                    if p[s] != "hold":
                        continue
                    p2 = dict(p)
                    p2[s] = "attack"
                    a, b = logodds(p, WEIGHT[q], t), logodds(p2, WEIGHT[q], t)
                    tries += 1
                    if b < a - 1e-9:
                        falls += 1
                        worst = max(worst, (round(a - b, 4), "{} {} {} added, L {:.4f} -> {:.4f}".format(
                            q, t, s, a, b)))
    print("   {} of {} added attack votes lower L ({:.3f})".format(falls, tries, falls / tries))
    print("   largest drop {:.4f}: {}".format(worst[0], worst[1]))

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 2 章「六类问事的权重表:五个声音各管哪一段」(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