关于「尾部方差:偏态回报的分位数、预期亏空与幂律」,《达人矩阵的组合方差》给出的定量结论是什么?
结论:第三章改用偏态回报:均值为正的顶流方案有 0.523 的季度净亏,二十人组合亏掉一半以上预算的季度只有 0.0037,顶流是 0.2047。
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: 0.0037, 0.2047, 0.523
| head lognormal: mu -0.0458, sigma 0.7851; exact P(loss) 0.523 | |
| mid lognormal: mu 0.0411, sigma 0.5899; exact P(loss) 0.472 | |
| return correlation of two mid-tier creators | 0.223 |
| quantile p1 p5 p25 p50 p75 p95 p99 | |
| head | -0.846 -0.734 -0.436 -0.043 0.607 2.458 4.940 |
| one mid | -0.734 -0.603 -0.298 0.039 0.546 1.754 3.159 |
| 20 mids | -0.440 -0.303 -0.049 0.178 0.463 0.991 1.491 |
| mean sd P(loss) P(loss>50%) | |
| head | 0.292 1.177 0.523 0.2047 |
| one mid | 0.241 0.802 0.473 0.1047 |
| 20 mids | 0.240 0.406 0.303 0.0037 |
查看完整验算程序
# Chapter 3. p31: same means, different shapes: quantiles of skewed returns.
# Model: gross return per yuan G is lognormal, net return R = G - 1.
# Head: E[G] = 1.30, sd 1.20. Each mid-tier creator: E[G] = 1.24, sd 0.80,
# with log-returns sharing one factor at log-correlation 0.25. Portfolio:
# 20 mid-tier creators, equal budget. 40,000 simulated quarters; the head's
# loss probability is also computed exactly. Fixed seed; illustrative (示意).
import math
import random
import statistics
N, Q, RHO_LOG = 20, 40000, 0.25
def lognormal_params(mean, sd):
s2 = math.log(1.0 + (sd / mean) ** 2)
return math.log(mean) - s2 / 2.0, math.sqrt(s2)
mh, sh = lognormal_params(1.30, 1.20)
mm, sm = lognormal_params(1.24, 0.80)
a, b = math.sqrt(RHO_LOG), math.sqrt(1.0 - RHO_LOG)
rng = random.Random(20261113)
head, single, port, pair = [], [], [], []
for _ in range(Q):
f = rng.gauss(0.0, 1.0)
mids = [math.exp(mm + sm * (a * f + b * rng.gauss(0.0, 1.0))) - 1.0
for _ in range(N)]
single.append(mids[0])
pair.append(mids[1])
port.append(math.fsum(mids) / N)
head.append(math.exp(mh + sh * rng.gauss(0.0, 1.0)) - 1.0)
nd = statistics.NormalDist()
print("head lognormal: mu %.4f, sigma %.4f; exact P(loss) %.3f"
% (mh, sh, nd.cdf(-mh / sh)))
print("mid lognormal: mu %.4f, sigma %.4f; exact P(loss) %.3f"
% (mm, sm, nd.cdf(-mm / sm)))
print("return correlation of two mid-tier creators: %.3f"
% statistics.correlation(single, pair))
qs = (0.01, 0.05, 0.25, 0.50, 0.75, 0.95, 0.99)
print("quantile " + "".join("%7s" % ("p%d" % round(100 * q)) for q in qs))
for name, xs in (("head", head), ("one mid", single), ("20 mids", port)):
s = sorted(xs)
print("%-9s " % name + "".join("%7.3f" % s[int(q * Q)] for q in qs))
print(" mean sd P(loss) P(loss>50%)")
for name, xs in (("head", head), ("one mid", single), ("20 mids", port)):
print("%-9s %6.3f %6.3f %8.3f %10.4f"
% (name, statistics.fmean(xs), statistics.stdev(xs),
len([v for v in xs if v < 0]) / Q,
len([v for v in xs if v < -0.5]) / Q))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 3 章「尾部方差:偏态回报的分位数、预期亏空与幂律」(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}
}