关于「实证闭环:五件套规则库的月度约束比扫描与预注册检验」,《规则叠加的悬崖》给出的定量结论是什么?

结论:第八章给出八步月度扫描、规则约束比死锁预警器与预注册检验:分级检验每组 28 个快照即有 0.851 的功效。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。

数据口径:书中规则册与数据均为示意;全部计算由书中程序生成;本书属非个人化经营方法,不构成法律、合规或经营决策意见。 第四章给一本 24 个开关的示意规则册做体检:裸约束比 0.958,载荷 0.475,死锁时约束比中位数只有 1.292,最小冲突核横跨三个规则族。

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: 0.851, 28

Test A: formulas per ratio for +/- 0.05385
fitted window exponent s0.791
predicted window at n200: 0.28
Test B: red vs green snapshots, one-sided alpha 0.05
p_red p_green n per group exact power null rejection
0.300.05 28 0.851 0.054
0.250.05 39 0.833 0.049
0.200.05 60 0.830 0.050
查看完整验算程序
# Chapter 8. p82: sample sizes for the pre-registered tests
# Test A (replication): formulas per ratio so that P(sat) is known to
# +/- 0.05 at 95 per cent; and the shrinking of the 10-90 window,
# fitted as width ~ n^-s on the p12 widths (2.48, 1.54, 0.80 at
# n = 12, 25, 50); s > 0 means a sharpening cliff.
# Test B (libraries): next-month deadlock share among red snapshots
# (h_2 >= 0.15) versus green ones (h_2 < 0.05); assumed shares 0.30
# and 0.05 are illustrative. One-sided alpha 0.05, power 0.80, normal
# approximation, then exact power of the same z-test by enumeration.

import math
from statistics import NormalDist

Z = NormalDist()
za, zb = Z.inv_cdf(0.95), Z.inv_cdf(0.80)

n_a = math.ceil((Z.inv_cdf(0.975) / 0.05) ** 2 * 0.25)
print("Test A: formulas per ratio for +/- 0.05: %d" % n_a)
ns, ws = [12, 25, 50], [2.48, 1.54, 0.80]
lx = [math.log(v) for v in ns]
ly = [math.log(v) for v in ws]
mx, my = math.fsum(lx) / 3, math.fsum(ly) / 3
sxy = math.fsum((a - mx) * (b - my) for a, b in zip(lx, ly))
sxx = math.fsum((a - mx) ** 2 for a in lx)
print("  fitted window exponent s = %.3f" % (-sxy / sxx))
print("  predicted window at n = 200: %.2f"
      % math.exp(my + sxy / sxx * (math.log(200) - mx)))


def z_test(x1, x2, n):
    p1, p2 = x1 / n, x2 / n
    p = (x1 + x2) / (2 * n)
    se = math.sqrt(2 * p * (1 - p) / n)
    return se > 0 and (p1 - p2) / se > za


def exact_power(p1, p2, n):
    f1 = [math.comb(n, k) * p1 ** k * (1 - p1) ** (n - k)
          for k in range(n + 1)]
    f2 = [math.comb(n, k) * p2 ** k * (1 - p2) ** (n - k)
          for k in range(n + 1)]
    return math.fsum(f1[i] * f2[j] for i in range(n + 1)
                     for j in range(n + 1) if z_test(i, j, n))


print("Test B: red vs green snapshots, one-sided alpha 0.05")
print("p_red  p_green  n per group  exact power  null rejection")
for p1, p2 in ((0.30, 0.05), (0.25, 0.05), (0.20, 0.05)):
    pb = (p1 + p2) / 2
    num = za * math.sqrt(2 * pb * (1 - pb)) + \
        zb * math.sqrt(p1 * (1 - p1) + p2 * (1 - p2))
    n = math.ceil((num / (p1 - p2)) ** 2)
    print("%5.2f  %7.2f  %11d  %11.3f  %14.3f"
          % (p1, p2, n, exact_power(p1, p2, n),
             exact_power(pb, pb, n)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《布尔可满足性相变阈值、促销合规与门禁叠加的约束比临界与全盘死锁预警母典》第 8 章「实证闭环:五件套规则库的月度约束比扫描与预注册检验」(版本 1.0.0). 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-252 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026252,
  author    = {Fan, Yafu},
  title     = {Boolean Satisfiability Phase Transitions: the Critical Constraint Ratio at Which Stacked Promotion, Compliance and Gate Rules Deadlock, and How to See It Coming},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论