关于「加倍阶梯的精确账:有限资本、尾部亏损与「亏了加倍」」,《鞅、停时与商业试错的停损界》给出的定量结论是什么?
结论:第二章精算加倍阶梯,并显示在好坏未知的项目上「亏了加倍」的期望是负 6.805。
数据口径:可选停时定理说,只要停止时刻不偷看未来且有上限,不赚不亏的事情在停下时的期望结果就等于起点;示意参数下,「第一次领先就停」的胜率是 0.824,期望却精确为零。 书中参数均为示意;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
| doubling ladder: k rungs need capital 2^k - 1 units | ||||||
| k capital | | P(bust) | E fair | | P(bust) | E | rounds | |
| | | p = 1/2 | | | p = 9/20 | per bust | ||
| 1 | 1 | | 0.500000 | 0 | | 0.550000 | -0.1000 | 1.8 |
| 2 | 3 | | 0.250000 | 0 | | 0.302500 | -0.2100 | 3.3 |
| 3 | 7 | | 0.125000 | 0 | | 0.166375 | -0.3310 | 6.0 |
| 4 | 15 | | 0.062500 | 0 | | 0.091506 | -0.4641 | 10.9 |
| 5 | 31 | | 0.031250 | 0 | | 0.050328 | -0.6105 | 19.9 |
| 6 | 63 | | 0.015625 | 0 | | 0.027681 | -0.7716 | 36.1 |
| 8 | 255 | | 0.003906 | 0 | | 0.008373 | -1.1436 | 119.4 |
| 10 | 1023 | | 0.000977 | 0 | | 0.002533 | -1.5937 | 394.8 |
| 12 | 4095 | | 0.000244 | 0 | | 0.000766 | -2.1384 | 1305.1 |
| fair, k = 10: P(bust) = 1/1024 exactly, E[one round] = 0 | ||||||
| fair: expected small wins before the first bust = 1023 |
核心公式
\[1+2+\cdots+2^{k-1}=2^k-1\]
查看完整验算程序
# Chapter 2. p21: The doubling ladder: exact bust probability and mean.
# Counter-example only. A "double after every loss" rule stakes 1, 2,
# 4, up to 2^(k-1) units until the first success. A capital of
# 2^k - 1 units pays for exactly k rungs. One round ends with +1 unit
# (some success within k tries) or with -(2^k - 1) units (k failures).
# p is the chance of success in one try: 1/2 is a fair trial, 9/20 a
# trial with a small edge against the player. Exact fractions; the
# values of p and k are illustrative.
from fractions import Fraction
def round_stats(p, k):
q = 1 - p
bust = q ** k
mean = (1 - bust) * 1 - bust * (2 ** k - 1)
return bust, mean
print("doubling ladder: k rungs need capital 2^k - 1 units")
print("%3s %7s | %10s %8s | %10s %9s %8s" % (
"k", "capital", "P(bust)", "E fair", "P(bust)", "E", "rounds"))
print("%3s %7s | %10s %8s | %10s %9s %8s" % (
"", "", "p = 1/2", "", "p = 9/20", "", "per bust"))
for k in (1, 2, 3, 4, 5, 6, 8, 10, 12):
bf, mf = round_stats(Fraction(1, 2), k)
bu, mu = round_stats(Fraction(9, 20), k)
print("%3d %7d | %10.6f %8s | %10.6f %9.4f %8.1f" % (
k, 2 ** k - 1, float(bf), str(mf), float(bu), float(mu),
float(1 / bu)))
b10, m10 = round_stats(Fraction(1, 2), 10)
print("fair, k = 10: P(bust) = %s exactly, E[one round] = %s" % (b10, m10))
print("fair: expected small wins before the first bust = %d" % (
(1 - b10) / b10))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《杜布鞅论可选停时定理、加倍追注谬误与商业试错见好就收停损界母典》第 2 章「加倍阶梯的精确账:有限资本、尾部亏损与「亏了加倍」」(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-249 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026249,
author = {Fan, Yafu},
title = {Doob's Martingales and the Optional Stopping Theorem: the Doubling Fallacy and Rigorous Stop-Loss Bounds for Business Trials},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}