关于「集齐十二款要多少盒:调和数、方差与完整分布」,《盲盒收集与福袋概率公示》给出的定量结论是什么?

结论:等概率的 12 款,集齐的期望盒数是 37.2385,标准差 13.72,中位数 35,九成分位 55。全部期望中有 0.790 花在后 6 款上。十万次固定种子的模拟与公式相符。这些都是数学事实;买家是否普遍低估它们,则是一条等待调查的假说。下一章的问题更尖锐:一个标称 1/144 的隐藏款,买满 144 盒,到底有多大把握?

数据口径:书中参数为示意;全部计算由书中程序生成;本书是非个人化法律或理财建议,法条以上架前人工复核为准。 第一章用调和数、精确分布与固定种子模拟算等概率集齐;第二章用几何分布算隐藏款等待,并陈复合事件判断偏差的相反证据,说明保底只改变等待(保底 144 次时平均 91.21 次)而不改变标称概率;第三章用容斥公式与泊松化积分算不等概率,12 款加 1/144 隐藏款的全套期望是 149.07 盒;第四章用双杯问题算合买与重复款,以及开盒止损;第五章算福袋的期望、方差与低于售价的概率(示意福袋为 0.8280)以及小概率加权。

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: 0.790, 12, 13.72, 37.2385

n12: E[T] = 12 * H_12 = 86021 / 2310
= 37.2385 boxes
sd13.72, ratio E[T]/n = 3.10
last design alone takes on average 12 boxes
n E[T] sd E[T]/n n ln n + g n + 1/2
614.70 6.24 2.45 14.71
821.74 8.72 2.72 21.75
1029.29 11.21 2.93 29.30
1237.24 13.72 3.10 37.25
1654.09 18.75 3.38 54.10
2490.62 28.87 3.78 90.63
mean wait for design 1, 6, 11, 121.00 1.71 6.00 12.00
share of E[T] spent on the last 6 designs0.790
查看完整验算程序
# Chapter 1. p11: expected boxes and spread to complete n equal designs.
# Model: n regular designs, each box shows one design with probability
# 1/n, independently. The number of boxes T to see all n designs is a
# sum of geometric waits with means n/(n-k), k = 0 to n - 1, so
#   E[T] = n * H_n,   Var[T] = n^2 * H2_n - n * H_n,
# with H_n the sum of 1/k and H2_n the sum of 1/k^2 for k = 1 to n
# (Feller 1968; exact identities, no parameters to estimate). The last
# column is the asymptotic form with Euler's constant g = 0.5772.
# Series sizes 6 to 24 are illustrative; 12 is the common shelf size.
import math
from fractions import Fraction


def harmonic(n):
    return sum_fr(Fraction(1, k) for k in range(1, n + 1))


def sum_fr(xs):
    total = Fraction(0)
    for x in xs:
        total += x
    return total


def moments(n):
    h1 = math.fsum(1.0 / k for k in range(1, n + 1))
    h2 = math.fsum(1.0 / (k * k) for k in range(1, n + 1))
    mean = n * h1
    var = n * n * h2 - n * h1
    return mean, math.sqrt(var)


e12 = 12 * harmonic(12)
print("n = 12: E[T] = 12 * H_12 = {} / {}".format(
    e12.numerator, e12.denominator))
print("        = {:.4f} boxes".format(float(e12)))
m12, s12 = moments(12)
print("        sd = {:.2f}, ratio E[T]/n = {:.2f}".format(s12, m12 / 12))
print("last design alone takes on average {} boxes".format(12))
print()
print(" n    E[T]     sd    E[T]/n   n ln n + g n + 1/2")
for n in (6, 8, 10, 12, 16, 24):
    m, s = moments(n)
    approx = n * math.log(n) + 0.5772156649 * n + 0.5
    print("{:2d}  {:7.2f}  {:5.2f}  {:6.2f}   {:7.2f}".format(
        n, m, s, m / n, approx))
print()
waits = [12.0 / (12 - k) for k in range(12)]
print("mean wait for design 1, 6, 11, 12: {:.2f} {:.2f} {:.2f} {:.2f}"
      .format(waits[0], waits[5], waits[10], waits[11]))
half = math.fsum(waits[6:])
print("share of E[T] spent on the last 6 designs: {:.3f}".format(
    half / m12))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《盲盒收集券问题、隐藏款几何等待时间与福袋概率公示合规母典》第 1 章「集齐十二款要多少盒:调和数、方差与完整分布」(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-210 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026210,
  author    = {Fan, Yafu},
  title     = {Blind-Box Coupon Collecting, Geometric Waits for Hidden Variants and Odds Disclosure for Lucky Bags},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论