关于「威尔逊下界:3 条五星为什么排不过 200 条 4.6 星」,《星级自选择与贝叶斯排序》给出的定量结论是什么?

结论:第三章用威尔逊下界给出排序:3 条五星为 0.4385,200 条 4.6 星为 0.8859;教科书 Wald 区间在小样本高好评率处覆盖率可低到 0.049。

数据口径:书中的顾客、商品与参数均为示意模型;全部计算由书中程序生成;本书是非个人化的经营参考,不是针对具体店铺的经营建议,也不是法律意见,书中提到的法规以上架前人工复核为准;本书反对一切刷单、购买评论与筛选式求评。 第一章用「购买偏差加少报偏差」的两段筛子解释评论星级的 J 型分布,示意参数下购买偏差使均值上移 +0.532 星。

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: 0.4385, 0.8859, 200, 4.6

listingnmeanposLB z95LB z90
A: 3 x five35.0001.0000.43850.5258
B: 200 @ 4.62004.6000.9300.88590.8942
C: 12 x five125.0001.0000.75750.8160
D: 40 @ 4.8404.8000.9500.83500.8597
an all-positive listing needs n >= 30 to pass B at z95
an all-positive listing needs n >= 23 to pass B at z90
closed form for all-positive: LB = n / (n + z^2), z^2 = 3.8415
查看完整验算程序
# Chapter 3. p31: Wilson's lower bound puts 200 reviews at 4.6 above 3 fives.
# Wilson (1927) score interval for a binomial share p from k of n:
#   (p + z^2/2n - z sqrt(p(1-p)/n + z^2/4n^2)) / (1 + z^2/n).
# A review counts as positive when it gives 4 or 5 stars. Listing B is a
# constructed example whose 200 reviews average exactly 4.6.
import math

Z95 = 1.959964      # two-sided 95 percent
Z90 = 1.644854      # one-sided 95 percent, also used by some shops


def wilson_lb(k, n, z=Z95):
    if n == 0:
        return 0.0
    p = k / n
    a = p + z * z / (2 * n)
    b = z * math.sqrt(p * (1 - p) / n + z * z / (4 * n * n))
    return (a - b) / (1 + z * z / n)


def mean_star(c):
    return math.fsum((i + 1) * x for i, x in enumerate(c)) / math.fsum(c)


LISTINGS = (
    ("A: 3 x five", (0, 0, 0, 0, 3)),
    ("B: 200 @ 4.6", (5, 2, 7, 40, 146)),
    ("C: 12 x five", (0, 0, 0, 0, 12)),
    ("D: 40 @ 4.8", (0, 1, 1, 3, 35)),
)
print("%-14s %4s %6s %6s %8s %8s" % (
    "listing", "n", "mean", "pos", "LB z95", "LB z90"))
for name, c in LISTINGS:
    n = int(math.fsum(c))
    k = c[3] + c[4]
    print("%-14s %4d %6.3f %6.3f %8.4f %8.4f" % (
        name, n, mean_star(c), k / n, wilson_lb(k, n),
        wilson_lb(k, n, Z90)))
target = wilson_lb(186, 200)
need = 1
while wilson_lb(need, need) <= target:
    need += 1
print("an all-positive listing needs n >= %d to pass B at z95" % need)
need90 = 1
while wilson_lb(need90, need90, Z90) <= wilson_lb(186, 200, Z90):
    need90 += 1
print("an all-positive listing needs n >= %d to pass B at z90" % need90)
print("closed form for all-positive: LB = n / (n + z^2), z^2 = %.4f"
      % (Z95 * Z95))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《评论星级J型自选择偏差、贝叶斯平均排序与九品中正评级通胀母典》第 3 章「威尔逊下界:3 条五星为什么排不过 200 条 4.6 星」(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-212 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026212,
  author    = {Fan, Yafu},
  title     = {J-Shaped Self-Selection in Star Ratings, Bayesian-Average Ranking and Rating Inflation from the Nine-Rank System to Online Marketplaces},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论