关于「必胜坏牌组的统计检验与平衡补丁的回归检查」,《卡牌费用回归与循环克制》给出的定量结论是什么?

结论:第六章计算检出坏牌组所需的对局数与补丁效果的预测区间,改一费仍有 0.407 的概率不够。

数据口径:本书用一款自拟的示意卡牌游戏「衡枢」把这些问题写成可以运行的算术。 书中卡池、胜率与价格参数均为示意;全部计算由书中程序生成;本书不是法律、财务或投资建议,也不推荐任何购买。

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: 0.407

patches 18: dw = +0.0056 + 0.0354 * dr
slope se 0.0012, intercept se 0.0022, R^2 0.982
through the origin: slope 0.0354
mean residual, nerfs +0.0018, buffs -0.0018
Durbin-Watson 2.25 (near 2: no drift across patches)
Y07 to cost 4: predicted dw -0.0601, 95% PI -0.0810 to -0.0392
needed to bring Surge to 0.5 vs the Nash field: -0.0578
chance the patch is not enough (normal approx): 0.407
查看完整验算程序
# Chapter 6. p62: Regression diagnostics for balance patches.
# An invented patch history (illustrative, seed 20261112): 18 past card
# changes, each moving a deck's key card by dr stat points against the
# p11 cost line, with the deck's field win rate then moving by dw. True
# model used to make the data: dw = 0.036 * dr + noise (sd 0.012).
# We fit dw on dr through the origin and with an intercept, check the
# residuals, and predict the patch that moves Y07 from cost 3 to cost 4
# (dr = -1.858, one cost step on the p11 line), with a 95% prediction
# interval from the t quantile for 16 degrees of freedom.
import math
import random

rng = random.Random(20261112)
DR = [-3, -2.5, -2, -2, -1.5, -1, -1, -0.5, 0.5, 1, 1, 1.5, 2, 2,
      2.5, 3, -1.5, 1.5]
DW = [0.036 * r + rng.gauss(0, 0.012) for r in DR]
n = len(DR)
T16 = 2.120

mx = math.fsum(DR) / n
my = math.fsum(DW) / n
sxx = math.fsum((x - mx) ** 2 for x in DR)
sxy = math.fsum((x - mx) * (y - my) for x, y in zip(DR, DW))
b = sxy / sxx
a = my - b * mx
res = [y - a - b * x for x, y in zip(DR, DW)]
s = math.sqrt(math.fsum(e * e for e in res) / (n - 2))
se_b = s / math.sqrt(sxx)
se_a = s * math.sqrt(1 / n + mx * mx / sxx)
syy = math.fsum((y - my) ** 2 for y in DW)
print("patches {}: dw = {:+.4f} + {:.4f} * dr".format(n, a, b))
print("slope se {:.4f}, intercept se {:.4f}, R^2 {:.3f}".format(
    se_b, se_a, 1 - math.fsum(e * e for e in res) / syy))
b0 = math.fsum(x * y for x, y in zip(DR, DW)) / math.fsum(
    x * x for x in DR)
print("through the origin: slope {:.4f}".format(b0))
lo = [e for x, e in zip(DR, res) if x < 0]
hi = [e for x, e in zip(DR, res) if x > 0]
print("mean residual, nerfs {:+.4f}, buffs {:+.4f}".format(
    math.fsum(lo) / len(lo), math.fsum(hi) / len(hi)))
dw_ = math.fsum((res[i] - res[i - 1]) ** 2 for i in range(1, n))
print("Durbin-Watson {:.2f} (near 2: no drift across patches)".format(
    dw_ / math.fsum(e * e for e in res)))
x0 = -1.858
pred = a + b * x0
half = T16 * s * math.sqrt(1 + 1 / n + (x0 - mx) ** 2 / sxx)
print("Y07 to cost 4: predicted dw {:+.4f}, 95% PI {:+.4f} to {:+.4f}"
      .format(pred, pred - half, pred + half))
need = 0.5578 - 0.5
print("needed to bring Surge to 0.5 vs the Nash field: {:+.4f}".format(
    -need))
sd = half / T16
print("chance the patch is not enough (normal approx): {:.3f}".format(
    0.5 * math.erfc((-need - pred) / (sd * math.sqrt(2)))))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《非对称策略卡牌费用收益回归、牌库构筑超几何卡手率边界与优势策略循环克制》第 6 章「必胜坏牌组的统计检验与平衡补丁的回归检查」(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-179 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026179,
  author    = {Fan, Yafu},
  title     = {Cost-Benefit Regression for Asymmetric Strategy Cards, Hypergeometric Brick-Rate Bounds in Deck Building and Cyclic Counters to Dominant Strategies},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论