关于「计然与杜皮特:贵贱反转与分级收费原典考辨」,《熟客唤醒与凑单闪购》给出的定量结论是什么?
结论:第七章考辨《史记·货殖列传》计然的贵贱反转与杜皮特 1844 年的分级桥费。
数据口径:示意参数下,面额 RM 6 的券只在间隔比 0.85 到 11.37 之间净值为正,刚买客与久睡客都是负值;不设对照组的核销记账会把增量高估 12.2 倍。 书中参数为示意,不对应 HOMIE-X 任何真实客户、订单或供货商,书中价格一律为示意;全部计算由书中程序生成。
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
PHI 0.90: half-life of a premium 6.58 weeks sell-dear 1.0326 plain 1.0140 edge +0.0186 (se 0.0015) PHI 1.00: no half-life (random walk) sell-dear 1.0297 plain 1.0346 edge -0.0049 (se 0.0022) C. OLS estimate of PHI from 52 weekly points (1,000 runs) true 0.90: mean 0.867, 5th 0.719, 95th 0.961 true 1.00: mean 0.968, 5th 0.848, 95th 1.025
查看完整验算程序
# Chapter 7. p71: Ji Ran's "dear at the top turns cheap" as a testable
# mean-reversion claim. The resale premium x_t of a hot item (log scale)
# follows x_t = PHI x_{t-1} + 0.10 e_t weekly. PHI = 0.90 is reversion,
# PHI = 1.00 a random walk (no reversion). A lot of 12 units over 26
# weeks. Rule "sell dear like dung": sell 4 units in any week with
# x_t > +0.15, otherwise 1 a week; plain rule: 1 unit a week, the rest in
# week 26. Score: mean realised exp(x) per unit. 2,000 paths per world,
# seed 20261120. Part C: OLS estimate of PHI from 52 weekly points.
import math
import random
import statistics
rng = random.Random(20261120)
LOT, WEEKS, SIG, HOT = 12, 26, 0.10, 0.15
def path(phi, n):
x, out = 0.0, []
for _ in range(n):
x = phi * x + SIG * rng.gauss(0.0, 1.0)
out.append(x)
return out
def realise(xs, rule):
left, got = LOT, []
for t, x in enumerate(xs):
if left == 0:
break
k = 1
if rule == "dear" and x > HOT:
k = 4
if t == len(xs) - 1:
k = left
k = min(k, left)
got.extend([math.exp(x)] * k)
left -= k
return math.fsum(got) / len(got)
if __name__ == "__main__":
for phi in (0.90, 1.00):
if phi < 1.0:
hl = math.log(0.5) / math.log(phi)
print("PHI {:.2f}: half-life of a premium {:.2f} weeks".format(
phi, hl))
else:
print("PHI 1.00: no half-life (random walk)")
dear, plain = [], []
for _ in range(2000):
xs = path(phi, WEEKS)
dear.append(realise(xs, "dear"))
plain.append(realise(xs, "plain"))
diff = [a - b for a, b in zip(dear, plain)]
print(" sell-dear {:.4f} plain {:.4f} edge {:+.4f} (se {:.4f})"
.format(statistics.fmean(dear), statistics.fmean(plain),
statistics.fmean(diff),
statistics.stdev(diff) / math.sqrt(len(diff))))
print("C. OLS estimate of PHI from 52 weekly points (1,000 runs)")
for phi in (0.90, 1.00):
est = []
for _ in range(1000):
xs = path(phi, 53)
num = math.fsum(a * b for a, b in zip(xs[:-1], xs[1:]))
den = math.fsum(a * a for a in xs[:-1])
est.append(num / den)
est.sort()
print(" true {:.2f}: mean {:.3f}, 5th {:.3f}, 95th {:.3f}".format(
phi, statistics.fmean(est), est[49], est[949]))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《私域代购熟客复购唤醒、满额凑单毛利配比与限时清仓脉冲闪购母典》第 7 章「计然与杜皮特:贵贱反转与分级收费原典考辨」(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-353 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026353,
author = {Fan, Yafu},
title = {Repeat-Buyer Reactivation in Private-Channel Commerce, Threshold Bundle Margin Design and Flash-Clearance Dynamics},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}