关于「代购对账核销流程与证伪方案」,《截图不是钱》给出的定量结论是什么?

结论:第八章给出代购对账核销流程、蒸馏小工具「汇款凭证核销器」与预注册检验方案。书中以标注框区分已发表的证据、可运行的 Python 计算(20 段程序)、原典与作者的假说。

数据口径:书中订单、账号末码、金额、截图与哈希均由固定种子合成,参数为示意,全部计算由书中程序生成;本书只讨论核销方如何检出重复或改动过的凭证,不提供任何制作凭证的方法;属非个人化经营方法,不构成法律、财务或理财建议,任何置付与发货都须由主事者人工放行。 第一章用生日问题量化撞单:从 10 单到 100 单,期望撞单配对数增长 110.0 倍,撞单概率只增长 41.3 倍;示意目录的有效金额只有 19.3 种,5.2 单即有对半机会撞单。

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: 20

(1) copies needed: miss rate >= m0 rejected at 0.05
m0 k0 k=1 k=2
0.0559 93 124
0.02149 236 313
0.01299 473 628
(2) key ties AMT+L5+W30 at 1,000 orders a day, lam0.01065
days expected refute if ties >= P(tail) at that count
300.320 3 0.00429
900.959 5 0.00306
1801.917 7 0.00362
3653.887 10 0.00674
(3) registered daily pairs of equal amounts, s0.051690
n20: 9.82 pairs a day
n50: 63.32 pairs a day
n200: 1028.63 pairs a day
查看完整验算程序
# Chapter 8. p82: sample sizes and refutation lines for the protocol
# Model, three registered tests, all exact:
# (1) Duplicate detector. Known returned copies are fed in; if the
#     miss rate were at least m0, how many copies are needed so that
#     seeing at most k misses has probability <= 0.05 (binomial)?
# (2) Key ties. Chapter 2 predicts lam ties per day on the key
#     amount + last five + 30-minute window at 1,000 orders a day
#     (0.01065 from p21, with the p12 law). Over D days the count is
#     Poisson(D * lam); the smallest count c with P(N >= c) <= 0.01
#     refutes the model (more ties than the model allows).
# (3) Amount-only ties. The model also predicts C(n,2) * s pairs of
#     equal amounts a day; if the shop's own law is flatter or
#     steeper the observed mean will say so. We print the expected
#     daily pairs for the shop sizes of Chapter 1 as the registered
#     prediction (s = 0.051690 from p12).
import math


def binom_cdf(k, n, p):
    terms = [math.comb(n, i) * p ** i * (1 - p) ** (n - i)
             for i in range(k + 1)]
    return math.fsum(terms)


def poisson_tail(c, mu):
    head = [math.exp(-mu + i * math.log(mu) - math.lgamma(i + 1))
            for i in range(c)]
    return 1.0 - math.fsum(head)


print("(1) copies needed: miss rate >= m0 rejected at 0.05")
print("      m0   k=0   k=1   k=2")
for m0 in (0.05, 0.02, 0.01):
    row = "%8.2f" % m0
    for k in (0, 1, 2):
        n = k + 1
        while binom_cdf(k, n, m0) > 0.05:
            n += 1
        row += "  %4d" % n
    print(row)
LAM = 0.01065
print("(2) key ties AMT+L5+W30 at 1,000 orders a day, lam = %.5f" % LAM)
print("    days   expected   refute if ties >=   P(tail) at that count")
for d in (30, 90, 180, 365):
    mu = d * LAM
    c = 1
    while poisson_tail(c, mu) > 0.01:
        c += 1
    print("%8d  %9.3f  %18d  %22.5f" % (d, mu, c, poisson_tail(c, mu)))
S = 0.051690
print("(3) registered daily pairs of equal amounts, s = %.6f" % S)
for n in (20, 50, 200):
    print("    n = %4d: %10.2f pairs a day" % (n, n * (n - 1) / 2 * S))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《多模态汇款凭证与口语订单容错解析、转账末五码核销自动机与重复截图欺诈阻断母典》第 8 章「代购对账核销流程与证伪方案」(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-267 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026267,
  author    = {Fan, Yafu},
  title     = {Fault-Tolerant Parsing of Payment Receipts and Spoken Orders: Last-Five-Digit Write-Off Automata and Blocking Duplicate-Screenshot Fraud},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论