关于「双账本:Brier 分数、校准分解与不让步的硬边界」,《效用、真实与双账本》给出的定量结论是什么?

结论:第六章建立双账本:行动框架可按效用编码,概率账以 Brier 分数校准,硬边界账不让步;框架改写概率账时,示意破产比例从 0.0925 升到 0.7055。

数据口径:书中参数与数据均为示意或合成,全部计算由书中程序生成;本书不是医疗建议,不提供任何临床心理服务,不是个人投资建议,不承诺任何收益;功能性信念只用于自己的行动框架,不用于欺骗他人。 第六章建立双账本:行动框架可按效用编码,概率账以 Brier 分数校准,硬边界账不让步;框架改写概率账时,示意破产比例从 0.0925 升到 0.7055。

English version

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: 0.0925, 0.7055

policymedian growthruin sharemean stake
A0.027290.17350.1666
B0.035620.09250.1666
C0.015920.70550.3666
D0.034670.09450.1833
查看完整验算程序
# Chapter 6. p62: two ledgers kept apart against one ledger rewritten
# 300 sequential ventures per path, 2,000 paths (fixed seed, common
# random numbers). Venture j has a calibrated success chance q_j drawn
# from [0.40, 0.60]; success returns 1.5 times the stake, failure loses
# it. A motivational frame raises the real chance by LIFT = 0.02 (the
# small self-fulfilling channel of Chapter 4). Policies:
#   A  no frame, stake = Kelly fraction of q
#   B  frame on, stake = Kelly of the calibrated q (two ledgers)
#   C  frame on, ledger rewritten to q + 0.12, stake = Kelly of that
#   D  as C but half the stake
# Reported: median growth of log wealth per venture, the share of paths
# that ever fall below 0.2 of the start, and the mean stake.
import math
import random
import statistics

ODDS, LIFT, INFL, N, PATHS = 1.5, 0.02, 0.12, 300, 2000
rng = random.Random(42762)
QS = [[rng.uniform(0.40, 0.60) for _ in range(N)] for _ in range(PATHS)]
US = [[rng.random() for _ in range(N)] for _ in range(PATHS)]


def kelly(p):
    return min(0.99, max(0.0, (p * (ODDS + 1) - 1) / ODDS))


POLICIES = {"A": (0.0, 0.0, 1.0), "B": (LIFT, 0.0, 1.0),
            "C": (LIFT, INFL, 1.0), "D": (LIFT, INFL, 0.5)}
print("policy  median growth  ruin share  mean stake")
for name in "ABCD":
    lift, infl, scale = POLICIES[name]
    growth, ruined, stakes = [], 0, []
    for qs, us in zip(QS, US):
        logs, hit = [], False
        for q, u in zip(qs, us):
            f = scale * kelly(min(0.99, q + infl))
            stakes.append(f)
            win = u < q + lift
            logs.append(math.log(1 + ODDS * f) if win
                        else math.log(1 - f))
            if not hit and math.fsum(logs) < math.log(0.2):
                hit = True
        growth.append(math.fsum(logs) / N)
        ruined += hit
    print("%6s  %13.5f  %10.4f  %10.4f" % (
        name, statistics.median(growth), ruined / PATHS,
        statistics.fmean(stakes)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《效用优先于字面真实:演化适存度博弈、功能性信念重塑与物理硬边界母典》第 6 章「双账本:Brier 分数、校准分解与不让步的硬边界」(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-427 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026427,
  author    = {Fan, Yafu},
  title     = {When Utility Outperforms Literal Truth: Evolutionary Fitness-Payoff Games, Functional Belief Engineering and Physical Guardrails},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论