What does Brush Smoothness and Fractal Dimension conclude, in numbers, about “Testing Beauty: Judge Reliability, Statistical Power and Forking Paths”?

Answer: Chapter 6 computes judge reliability, statistical power and forking paths: picking the best of 16 windows lifts the false-positive rate to 0.3725.

Scope of the numbers: Trajectories, paper, ink and scoring parameters are illustrative; all calculations are generated by the book's programs; the book is general analysis, not business, design or medical advice. Marked boxes separate published evidence, runnable Python calculations (21 programs), classical sources and the author's hypotheses.

中文版

Data table

The numbers below are printed verbatim by the chapter's Python program and can be reproduced by running it. standard-library Python onlyprinted values that match the answer: 0.3725, 16

windows mc = 0.5c = 0.9independent 1-(0.95)^m
10.05300.04800.0500
20.08620.07270.0975
40.15250.09020.1855
80.24800.11250.3366
160.37250.14400.5599
Show the full program
# Chapter 6. p63: choosing the box window after seeing the data
# Model: two groups of 20 images with NO true difference. An analyst
# measures D with m box windows (plus thresholds and crops) and
# reports the window with the smallest p value. The m estimates of
# one image share a common part: corr(D_i, D_j) = c. Each window gets
# a two-sample z test on the standardised group means; we count how
# often the best of m windows has p < 0.05. m and c are illustrative;
# 4000 replicates per row, seed fixed.
import math
import random
from statistics import NormalDist

N, REPS = 20, 4000
Z = NormalDist()


def false_alarm(m, c, rng):
    hits = 0
    a, b = math.sqrt(c), math.sqrt(1 - c)
    for _ in range(REPS):
        means = []
        for _g in range(2):
            com = [rng.gauss(0, 1) for _ in range(N)]
            cols = []
            for _w in range(m):
                vals = [a * x + b * rng.gauss(0, 1) for x in com]
                cols.append(math.fsum(vals) / N)
            means.append(cols)
        best = max(abs(means[0][w] - means[1][w]) for w in range(m))
        p = 2 * (1 - Z.cdf(best / math.sqrt(2.0 / N)))
        hits += p < 0.05
    return hits / REPS


if __name__ == "__main__":
    rng = random.Random(20261113)
    print("windows m   c = 0.5   c = 0.9   independent 1-(0.95)^m")
    for m in (1, 2, 4, 8, 16):
        r5 = false_alarm(m, 0.5, rng)
        r9 = false_alarm(m, 0.9, rng)
        print("{:>8d}    {:.4f}    {:.4f}    {:.4f}".format(
            m, r5, r9, 1 - 0.95 ** m))

Single source and citation

Fan, Y. (2026). Minimum-Jerk Brush Trajectories, Capillary Ink Flow in Flying-White Strokes and the Fractal Dimension of Stroke Beauty (Version 1.0.0), Chapter 6: Testing Beauty: Judge Reliability, Statistical Power and Forking Paths. 5D Library. DOI pending

DOI pending · monograph page goes live after publication · 5DL-B-229 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026229,
  author    = {Fan, Yafu},
  title     = {Minimum-Jerk Brush Trajectories, Capillary Ink Flow in Flying-White Strokes and the Fractal Dimension of Stroke Beauty},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

Other chapter answers in this book