关于「季度再平衡闭环与证伪方案」,《达人矩阵的组合方差》给出的定量结论是什么?
结论:第八章给出季度再平衡流程、达人组合方差计算器与预注册检验:方差比 0.50 时每组需要 53 个波次,计入顶流峰度后要 435 个。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: 0.50, 435, 53
| z(0.95) + z(0.80) | 2.4865 |
| ratio R n normal n with kurtosis A 31.8, B 4.0 | |
| 0.75 | 300 2523 |
| 0.50 | 53 435 |
| 0.25 | 14 109 |
| R | 0.50, n=53 per arm: power normal 0.809, lognormal arm A 0.531 |
| R | 0.25, n=14 per arm: power normal 0.794, lognormal arm A 0.526 |
| waves of two weeks: n | 53 per arm alternating is 212 weeks |
| 4 matched product lines in parallel | 14 waves, 28 weeks |
| 6 matched product lines in parallel | 9 waves, 18 weeks |
| 8 matched product lines in parallel | 7 waves, 14 weeks |
查看完整验算程序
# Chapter 8. p82: how many campaign waves a pre-registered test needs.
# Hypothesis: the variance of net return per yuan under a spread mid-tier
# plan (arm B) is at most R times that under a head-concentrated plan
# (arm A). Statistic: L = ln(sA^2 / sB^2), one-sided 5%, power 80%.
# For normal data var(ln s^2) is about 2 / (n - 1), so
# n = 1 + 4 z^2 / (ln R)^2, z = z(0.95) + z(0.80).
# With kurtosis k the variance is about (k - 1) / n per arm. Simulated
# power (4,000 runs, seeded) checks the normal n with normal data and
# with a lognormal arm A (sigma 0.7851, kurtosis 31.771 from p42).
# Ratios and kurtoses are illustrative (示意).
import math
import random
import statistics
nd = statistics.NormalDist()
z = nd.inv_cdf(0.95) + nd.inv_cdf(0.80)
KA, KB = 31.771, 4.0
print("z(0.95) + z(0.80) = %.4f" % z)
print(" ratio R n normal n with kurtosis A %.1f, B %.1f" % (KA, KB))
plans = {}
for r in (0.75, 0.5, 0.25):
nn = math.ceil(1 + 4 * z * z / math.log(r) ** 2)
nk = math.ceil(z * z * (KA - 1 + KB - 1) / math.log(r) ** 2)
plans[r] = nn
print("%9.2f %10d %12d" % (r, nn, nk))
rng = random.Random(20261113)
S = 0.7851
SCALE = 1.0 / math.sqrt((math.exp(S * S) - 1) * math.exp(S * S))
def power(n, r, lognormal_a):
crit = nd.inv_cdf(0.95) * math.sqrt(4.0 / (n - 1))
hits = 0
for _ in range(4000):
if lognormal_a:
a = [SCALE * math.exp(S * rng.gauss(0.0, 1.0)) for _ in range(n)]
else:
a = [rng.gauss(0.0, 1.0) for _ in range(n)]
b = [math.sqrt(r) * rng.gauss(0.0, 1.0) for _ in range(n)]
lr = math.log(statistics.variance(a) / statistics.variance(b))
if lr > crit:
hits += 1
return hits / 4000
for r in (0.5, 0.25):
n = plans[r]
print("R=%.2f, n=%d per arm: power normal %.3f, lognormal arm A %.3f"
% (r, n, power(n, r, False), power(n, r, True)))
print("waves of two weeks: n=%d per arm alternating is %d weeks"
% (plans[0.5], 2 * 2 * plans[0.5]))
for lines in (4, 6, 8):
w = math.ceil(plans[0.5] / lines)
print(" %d matched product lines in parallel: %d waves, %d weeks"
% (lines, w, 2 * w))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 8 章「季度再平衡闭环与证伪方案」(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-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}
}