关于「事后对上:宽松匹配、可选停止与重复计数」,《出生时刻误差与交界翻转》给出的定量结论是什么?

结论:第六章计算零假设下的事后对上:题目毫无区分力时,问到对上为止,40 题之内的确认概率是 0.9958。

数据口径:本书只分析排盘规则的结构、误差传播与概率,不做任何个人占断,不宣称任何体系有预测效力;书中参数为示意,全部计算由书中程序生成。 第一章说明,示意模型中凭记忆报告的时刻有 0.2782 恰好落在时辰分界上,不取整时只有 0.0083。

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: 0.9958, 40

assumed q = 0.8, odds factor 4.0000, net count needed s = 3
Noptional stoppingone look at N
40.25000.1250
80.57810.2891
120.76270.3877
200.92490.5034
400.99580.6358
assumed q = 0.7, odds factor 2.3333, net count needed s = 4
Noptional stoppingone look at N
40.12500.1250
80.35940.2891
120.53320.3877
200.75220.5034
400.94910.6358
查看完整验算程序
# Chapter 6. p62: stop when it fits, and a useless bank "confirms".
# Null model: every answer is a fair coin under both candidates A and
# B. The reader, however, scores each item as if it had P(yes | A) = q
# and P(yes | B) = 1 - q, with the side that "yes" favours set by the
# item, so each answer multiplies the odds by L = q / (1 - q) or 1 / L
# with equal chance. Starting from even odds, a candidate is declared
# confirmed once its posterior reaches 0.95, i.e. once the net count
# of favouring answers reaches s = ceil(log 19 / log L). Exact dynamic
# programming over the net count. We compare stopping as soon as the
# threshold is hit (optional stopping) with one look after all N.
import math


def thresholds(q):
    lr = q / (1 - q)
    return math.ceil(math.log(19.0) / math.log(lr) - 1e-12)


def optional(N, s):
    # 净计数的随机游走,触到 +s 或 -s 即停止并宣布对上
    dist = {0: 1.0}
    hit = []
    for _ in range(N):
        nxt = {}
        for x, p in dist.items():
            for y in (x - 1, x + 1):
                if abs(y) >= s:
                    hit.append(0.5 * p)
                else:
                    nxt.setdefault(y, []).append(0.5 * p)
        dist = {y: math.fsum(v) for y, v in sorted(nxt.items())}
    return math.fsum(hit)


def one_look(N, s):
    terms = [math.comb(N, j) * 0.5 ** N for j in range(N + 1)
             if abs(2 * j - N) >= s]
    return math.fsum(terms)


for q in (0.8, 0.7):
    s = thresholds(q)
    print("assumed q = %.1f, odds factor %.4f, net count needed s = %d"
          % (q, q / (1 - q), s))
    print("   N   optional stopping   one look at N")
    for N in (4, 8, 12, 20, 40):
        print("%4d %19.4f %15.4f" % (N, optional(N, s), one_look(N, s)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《出生时刻记忆误差传播与鲁棒定盘、交界时辰盘面翻转概率与最少前事反推题数母典》第 6 章「事后对上:宽松匹配、可选停止与重复计数」(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-284 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026284,
  author    = {Fan, Yafu},
  title     = {Error Propagation from Remembered Birth Times and Robust Chart Fixing: Flip Probabilities at Double-Hour Boundaries and the Fewest Past-Event Questions},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论