关于「有噪问答:说谎者游戏、属相的年界误差与费诺不等式」,《铁板神数的比特账》给出的定量结论是什么?

结论:第六章计算属相的年界误差,约 0.0911 的人按公历自报属相会出错,并用说谎者游戏与费诺不等式给出下界。

数据口径:书中参数为示意;全部计算由书中程序生成;本书分析结构与概率,不对任何个人作任何占断,也不构成法律或理财建议。 第五章说明一句条文的期望信息量至多 0.5307 比特,按示意规模每分钟有 19.01 人出生。

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: 0.0911

1 Jan to lunar new year, mean 33.3 days: share 0.0911
lunar new year vs lichun, mean gap 8.4 days: share 0.0231
eps bits per animal answer
0.00013.583
0.053.126
0.09112.830
0.152.456
bits collected P(error) floor, 120 fen120000 verses
3.00.421 0.763
5.00.131 0.644
6.6650.000 0.546
8.0180.000 0.466
12.00.000 0.230
查看完整验算程序
# Chapter 6. p62: Noisy animals and Fano's inequality. Part 1: the
# zodiac year turns at the lunar new year (or at lichun), not on 1
# January, so a client who answers by calendar year can be wrong. Dates
# for 2020 to 2026 are listed below (lunar new year; lichun). Part 2: a
# 12-answer question whose answer is wrong with probability eps (spread
# over the other 11) carries at most log2 12 - h(eps) - eps log2 11
# bits. Part 3: Fano: P(error) >= (log2 N - I - 1) / log2 N when I bits
# about N equally likely cells have been collected.
import datetime as dt
import math

NEWYEAR = {2020: (1, 25), 2021: (2, 12), 2022: (2, 1), 2023: (1, 22),
           2024: (2, 10), 2025: (1, 29), 2026: (2, 17)}
LICHUN = {2020: 4, 2021: 3, 2022: 4, 2023: 4, 2024: 4, 2025: 3,
          2026: 4}                                   # day of February
win, gap = [], []
for y in sorted(NEWYEAR):
    m, d = NEWYEAR[y]
    ny = dt.date(y, m, d)
    lc = dt.date(y, 2, LICHUN[y])
    win.append((ny - dt.date(y, 1, 1)).days)
    gap.append(abs((ny - lc).days))
yl = 365.2425
w, g = math.fsum(win) / len(win), math.fsum(gap) / len(gap)
print("  1 Jan to lunar new year, mean {:.1f} days: share {:.4f}".format(
    w, w / yl))
print("  lunar new year vs lichun, mean gap {:.1f} days: share {:.4f}"
      .format(g, g / yl))


def h2(p):
    return -(p * math.log2(p) + (1 - p) * math.log2(1 - p))


print("  eps     bits per animal answer")
for eps in (0.0001, 0.05, 0.0911, 0.15):
    c = math.log2(12) - h2(eps) - eps * math.log2(11)
    print("  {:<6}  {:.3f}".format(eps, c))
print("  bits collected   P(error) floor, 120 fen   120000 verses")
for info in (3.0, 5.0, 6.665, 8.018, 12.0):
    f1 = max(0.0, (math.log2(120) - info - 1) / math.log2(120))
    f2 = max(0.0, (math.log2(120000) - info - 1) / math.log2(120000))
    print("  {:>14}   {:>23.3f}   {:>13.3f}".format(info, f1, f2))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《铁板神数考刻分决策树、条文库仿射编码与信息熵套取下界》第 6 章「有噪问答:说谎者游戏、属相的年界误差与费诺不等式」(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-157 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026157,
  author    = {Fan, Yafu},
  title     = {Tieban Shenshu Ke-Fen Decision Trees, Affine Coding of the Verse Bank and Entropy Lower Bounds on Elicitation},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论