关于「实践:读一张时盘、一次起卦,与一份预先登记的商业预测检验」,《转盘奇门百格与六爻商业直断》给出的定量结论是什么?
结论:第八章用一张不含出生资料的样盘时盘与一次商业起卦示范查法,并给出一份预先登记的商业预测检验:以 Brier 分数与基础率比较,方向命中率从 0.50 提高到 0.60 需要 158 次预测。书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。
数据口径:全部计算由书中程序生成;本书不是投资、法律、职业、心理或婚姻建议。 书中以标注框区分已发表的证据、可运行的 Python 计算、原典与作者的假说。
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, 0.60, 158
| (1) event: receivable in full by the agreed date | |
| outcome yes: chart forecast 0.30 Brier 0.4900 log 1.2040; base rate 0.60 Brier 0.1600 log 0.5108 | |
| outcome no : chart forecast 0.30 Brier 0.0900 log 0.3567; base rate 0.60 Brier 0.3600 log 0.9163 | |
| (2) always the base rate p: mean Brier p(1-p) | |
| p | 0.50: 0.2500 |
| p | 0.60: 0.2400 |
| p | 0.70: 0.2100 |
| (3) stated q, true hit frequency h: mean Brier | |
| h q | 0.70 q=0.80 q=0.90 q=h |
| 0.55 | 0.2700 0.3100 0.3700 0.2475 |
| 0.60 | 0.2500 0.2800 0.3300 0.2400 |
| 0.65 | 0.2300 0.2500 0.2900 0.2275 |
| 0.70 | 0.2100 0.2200 0.2500 0.2100 |
| (4) direction calls, exact binomial, alpha 0.05, power 0.80 | |
| base-rate hit 0.50, true hit 0.55: N | 620 |
| base-rate hit 0.50, true hit 0.60: N | 158 |
| base-rate hit 0.60, true hit 0.65: N | 585 |
| base-rate hit 0.60, true hit 0.70: N | 143 |
查看完整验算程序
# Chapter 8. p07: Scoring pre-registered business forecasts, and how many.
# Each forecast is a probability q, written down before the outcome, for
# a precisely defined yes/no event ("at least 95 per cent of the invoice
# received by the agreed date"). Brier score (q - y)^2 and log score
# -ln(prob given to what happened); lower is better for both. The rival
# is the base rate: the share of such events that came true in the
# firm's own past records. (1) The sample cast's forecast against a base
# rate of 0.60. (2) What an always-base-rate forecaster scores on
# average, p(1 - p). (3) A reader who states q for the side the chart
# favours and is right with frequency h scores h(1-q)^2 + (1-h)q^2.
# (4) Direction calls: smallest N for which a one-sided exact binomial
# test at alpha 0.05 has power 0.80 against a true hit rate p1 when the
# base-rate rule hits p0 (scan from below the normal-theory N).
from math import exp, lgamma, log, sqrt
from statistics import NormalDist
def tail(n, c, p):
lp, lq = log(p), log(1 - p)
return sum(exp(lgamma(n + 1) - lgamma(x + 1) - lgamma(n - x + 1)
+ x * lp + (n - x) * lq) for x in range(c, n + 1))
def n_needed(p0, p1, alpha=0.05):
z = NormalDist().inv_cdf
na = ((z(1 - alpha) * sqrt(p0 * (1 - p0)) + z(0.80) *
sqrt(p1 * (1 - p1))) / (p1 - p0)) ** 2
n = max(10, int(0.8 * na))
while True:
c = int(n * p0)
while tail(n, c, p0) > alpha:
c += 1
if tail(n, c, p1) >= 0.80:
return n
n += 1
if __name__ == "__main__":
q, base = 0.30, 0.60
print(" (1) event: receivable in full by the agreed date")
for y in (1, 0):
print(" outcome {}: chart forecast {:.2f} Brier {:.4f} log {:.4f};"
" base rate {:.2f} Brier {:.4f} log {:.4f}".format(
"yes" if y else "no ", q, (q - y) ** 2,
-log(q if y else 1 - q), base, (base - y) ** 2,
-log(base if y else 1 - base)))
print(" (2) always the base rate p: mean Brier p(1-p)")
for p in (0.50, 0.60, 0.70):
print(" p = {:.2f}: {:.4f}".format(p, p * (1 - p)))
print(" (3) stated q, true hit frequency h: mean Brier")
print(" h q=0.70 q=0.80 q=0.90 q=h")
for h in (0.55, 0.60, 0.65, 0.70):
print(" {:.2f} ".format(h) + " ".join(
"{:.4f}".format(h * (1 - s) ** 2 + (1 - h) * s ** 2)
for s in (0.70, 0.80, 0.90, h)))
print(" (4) direction calls, exact binomial, alpha 0.05, power 0.80")
for p0, p1 in ((0.50, 0.55), (0.50, 0.60), (0.60, 0.65), (0.60, 0.70)):
print(" base-rate hit {:.2f}, true hit {:.2f}: N = {}".format(
p0, p1, n_needed(p0, p1)))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《转盘奇门拆补法与六爻纳甲商业断事母典》第 8 章「实践:读一张时盘、一次起卦,与一份预先登记的商业预测检验」(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-121 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026121,
author = {Fan, Yafu},
title = {The Rotating Qimen Chaibu and Liuyao Najia Codex of Business Verdicts},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}