关于「东西方原典考辨:《淮南子》塞翁失马与高尔顿 1886 年的「向平庸回归」」,《均值回归与封面诅咒》给出的定量结论是什么?

结论:第七章对读《淮南子》塞翁失马与高尔顿 1886 年的论文,并算出短记录中「福后又福」比例的偏差,四次记录时为 0.4048。

数据口径:书中门店、达人、顾客与报告的参数均为示意;全部计算由书中程序生成;本书是非个人化的经营分析,不构成经营、法律或理财建议,涉及运程报告与原典之处只分析结构与概率,不作任何个人占断。 书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。

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: 0.4048

(A) P(four turns alternate strictly)
switch prob 0.50.1250
switch prob 0.60.2160
switch prob 0.80.5120
(B) mean share of gain-after-gain, fair independent turns
n4: 0.4048 (14 sequences counted)
n6: 0.4161 (62 sequences counted)
n10: 0.4454 (1022 sequences counted)
n16: 0.4667 (65534 sequences counted)
n4 exactly: 17/42
(C) turns needed to detect q0.6 vs 0.5: 153
查看完整验算程序
# Chapter 7. p71: The old man's horse as a test of serial independence.
# The tale has four turns: loss, gain, loss, gain. (A) If each turn is an
# independent fair coin, how likely is a strictly alternating run of four;
# and if loss and gain truly "beget each other" with switch probability q?
# (B) The streak-selection bias of Miller and Sanjurjo (2018): in short
# independent sequences, the average share of gains that follow a gain is
# below 1/2. Exact enumeration of all 2^n sequences. (C) How many
# recorded turns are needed to tell q = 0.6 from 0.5 (one-sided 0.05,
# power 0.8). Exact arithmetic; q illustrative.
# 塞翁失马:四次转折的交替概率、连串选择偏差与检验所需样本量。
import itertools
import math
from fractions import Fraction


def alternate_prob(q):
    return 2 * Fraction(1, 2) * q ** 3


def streak_share(n):
    vals = []
    for seq in itertools.product((0, 1), repeat=n):
        follow = [seq[i + 1] for i in range(n - 1) if seq[i] == 1]
        if follow:
            vals.append(Fraction(follow.count(1), len(follow)))
    return Fraction(math.fsum(float(v) for v in vals)) / len(vals), vals


if __name__ == "__main__":
    print("(A) P(four turns alternate strictly)")
    for q in (Fraction(1, 2), Fraction(3, 5), Fraction(4, 5)):
        p = alternate_prob(q)
        print("    switch prob {:.1f}: {:.4f}".format(float(q), float(p)))
    print("(B) mean share of gain-after-gain, fair independent turns")
    for n in (4, 6, 10, 16):
        m, vals = streak_share(n)
        print("    n = {:>2}: {:.4f}   ({} sequences counted)"
              .format(n, float(m), len(vals)))
    exact4 = Fraction(0)
    _, v4 = streak_share(4)
    for v in v4:
        exact4 += v
    print("    n = 4 exactly: {}".format(exact4 / len(v4)))
    z_a, z_b = 1.6449, 0.8416
    n = ((z_a * 0.5 + z_b * math.sqrt(0.6 * 0.4)) / 0.1) ** 2
    print("(C) turns needed to detect q = 0.6 vs 0.5: {}".format(
        math.ceil(n)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《均值回归与封面诅咒、运程报告后验校准与塞翁失马序列独立性母典》第 7 章「东西方原典考辨:《淮南子》塞翁失马与高尔顿 1886 年的「向平庸回归」」(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-202 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026202,
  author    = {Fan, Yafu},
  title     = {Regression to the Mean and the Cover Jinx: Posterior Calibration of Fortune Reports and Serial Independence in the Tale of the Lost Horse},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论