关于「受众重叠就是相关:分散收益的折扣与有效人数」,《达人矩阵的组合方差》给出的定量结论是什么?
结论:第二章由受众重叠算出相关,二十人只相当于 4.17 个独立的人,分散折扣恰好等于相关。
数据口径:书中借用组合理论的方差、相关、夏普比率与均值—方差前沿,只借其算术,讨论的是营销预算的分配,不是证券投资;本书不构成任何个人化的理财或投资建议,不推荐任何金融产品。 书中的达人、回报、相关、成本与预算均为示意;全部计算由书中程序生成;所有合作都须如实标注广告,不讨论购买流量或虚假评价。
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: 4.17
| effective number of independent creators, N_eff | |
| N rho | 0.05 rho=0.10 rho=0.20 rho=0.30 rho=0.50 |
| 5 | 4.17 3.57 2.78 2.27 1.67 |
| 10 | 6.90 5.26 3.57 2.70 1.82 |
| 20 | 10.26 6.90 4.17 2.99 1.90 |
| 50 | 14.49 8.47 4.63 3.18 1.96 |
| 100 | 16.81 9.17 4.81 3.26 1.98 |
| 1000 | 19.63 9.91 4.98 3.33 2.00 |
| cap | 20.00 10.00 5.00 3.33 2.00 |
| variance of N | 20 relative to one creator, and the discount |
| rho | 0.05 relative variance 0.0975 independent 0.0500 discount 0.050 |
| rho | 0.10 relative variance 0.1450 independent 0.0500 discount 0.100 |
| rho | 0.20 relative variance 0.2400 independent 0.0500 discount 0.200 |
| rho | 0.30 relative variance 0.3350 independent 0.0500 discount 0.300 |
| rho | 0.50 relative variance 0.5250 independent 0.0500 discount 0.500 |
| rho | 0.2: mid-tier variance floor 0.128 vs head 1.44, ratio 0.089 |
| rho | 0.5: mid-tier variance floor 0.320 vs head 1.44, ratio 0.222 |
| rho | 0.8: mid-tier variance floor 0.512 vs head 1.44, ratio 0.356 |
| the floor reaches the head's variance only if one mid-tier | |
| creator's sd exceeds 2.6833 at rho | 0.20 (it is 0.80 here) |
查看完整验算程序
# Chapter 2. p22: how many independent bets do N correlated creators make?
# Model: N creators of equal variance, every pair correlated at rho.
# The equal-weight portfolio has the variance of N_eff independent ones,
# N_eff = N / (1 + (N - 1) * rho), which can never exceed 1 / rho.
# "Discount" = share of the independent-case variance reduction that
# correlation takes away. Pure arithmetic; rho values illustrative (示意).
import math
RHOS = (0.05, 0.10, 0.20, 0.30, 0.50)
print("effective number of independent creators, N_eff")
print(" N" + "".join(" rho=%.2f" % r for r in RHOS))
for n in (5, 10, 20, 50, 100, 1000):
print("%5d" % n + "".join("%10.2f" % (n / (1 + (n - 1) * r))
for r in RHOS))
print(" cap" + "".join("%10.2f" % (1 / r) for r in RHOS))
print()
print("variance of N=20 relative to one creator, and the discount")
for r in RHOS:
rel = (1 + 19 * r) / 20
ideal = 1 / 20
lost = (rel - ideal) / (1 - ideal)
print("rho=%.2f relative variance %.4f independent %.4f discount %.3f"
% (r, rel, ideal, lost))
print()
# 同等预算下顶流方差 1.44;N 名中腰部方差 0.64 * (1 + (N-1) rho) / N
for r in (0.2, 0.5, 0.8):
worst = 0.64 * r
print("rho=%.1f: mid-tier variance floor %.3f vs head 1.44, ratio %.3f"
% (r, worst, worst / 1.44))
print("the floor reaches the head's variance only if one mid-tier")
print("creator's sd exceeds %.4f at rho=0.20 (it is 0.80 here)"
% math.sqrt(1.44 / 0.2))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《达人矩阵头部集中度与中腰部组合套利、尾部方差夏普比率与百名垂直达人投资组合母典》第 2 章「受众重叠就是相关:分散收益的折扣与有效人数」(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-276 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026276,
author = {Fan, Yafu},
title = {Head Concentration versus Mid-Tier Portfolios in Creator Matrices: Tail Variance, Sharpe Ratios and the Allocation of a Budget across a Hundred Niche Creators},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}