What does The Life of a Rule conclude, in numbers, about “The Rule Lifecycle as a State Machine: Promotion, Demotion and Retirement”?
Answer: Chapter 5 checks an eight-state rule machine exhaustively, watches for wear with Page's cumulative sum, and uses an absorbing chain to show that a worn rule stays live 8.54 weeks on average.
Scope of the numbers: Marked boxes separate published evidence, runnable Python calculations (20 programs), classical sources and the author's hypotheses.
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: 8.54
| start: just promoted, working; wear-out hazard 1/52 per week | |||||
| h | weeks OK | WORN live | re-test OK | re-test WORN | P(cut while OK) |
| 2.0 | 40.6 | 4.95 | 6.18 | 5.42 | 0.1012 |
| 3.0 | 47.5 | 8.54 | 2.41 | 5.79 | 0.0395 |
| 4.0 | 50.2 | 12.04 | 0.95 | 5.93 | 0.0155 |
| h 3.0: expected life 64.3 weeks; P(cut OK) + P(cut WORN) = 1.0000 |
Show the full program
# Chapter 5. p53: weeks a worn-out rule keeps running, as an absorbing
# Markov chain (Kemeny and Snell's fundamental matrix N = (I - Q)^-1)
# Model, one step = one week, starting at promotion. Transient states:
# OK (promoted, still working), WORN (promoted, silently worn out),
# WOK (re-test after a false alarm), WWORN (re-test of a worn rule).
# Absorbing: CUT-OK (a working rule demoted) and CUT-WORN (a worn rule
# demoted). Wear-out hazard 1/52 a week (illustrative). Alarm chances
# are 1/ARL0 and 1/ARL1 from p52; a re-test ends each week with chance
# 600/E[N] and keeps with the exact batched P(KEEP) from p41 (m = 600):
# working 0.8714 at E[N] 4624, worn 0.0332 at E[N] 3495. Approximating
# the re-test length as geometric is the model's main simplification.
from fractions import Fraction as F
import math
WEAR = 1 / 52
CUSUM = {2.0: (43.9, 6.10), 3.0: (132.0, 9.02), 4.0: (354.2, 12.04)}
RE_OK, RE_WORN = (0.8714, 4624), (0.0332, 3495)
S = ["OK", "WORN", "WOK", "WWORN"]
def solve(mat):
n = len(mat)
aug = [[F(x) for x in row] + [F(int(i == j)) for j in range(n)]
for i, row in enumerate(mat)]
for c in range(n):
piv = next(r for r in range(c, n) if aug[r][c] != 0)
aug[c], aug[piv] = aug[piv], aug[c]
pv = aug[c][c]
aug[c] = [x / pv for x in aug[c]]
for r in range(n):
if r != c and aug[r][c] != 0:
f = aug[r][c]
aug[r] = [x - f * y for x, y in zip(aug[r], aug[c])]
return [[float(x) for x in row[n:]] for row in aug]
def chain(h):
arl0, arl1 = CUSUM[h]
a0, a1 = 1 / arl0, 1 / arl1
e_ok, e_worn = 600 / RE_OK[1], 600 / RE_WORN[1]
q = [[0.0] * 4 for _ in range(4)]
r = [[0.0] * 2 for _ in range(4)]
q[0][0] = (1 - WEAR) * (1 - a0)
q[0][1] = WEAR
q[0][2] = (1 - WEAR) * a0
q[1][1] = 1 - a1
q[1][3] = a1
q[2][2] = (1 - WEAR) * (1 - e_ok)
q[2][3] = WEAR
q[2][0] = (1 - WEAR) * e_ok * RE_OK[0]
r[2][0] = (1 - WEAR) * e_ok * (1 - RE_OK[0])
q[3][3] = 1 - e_worn
q[3][1] = e_worn * RE_WORN[0]
r[3][1] = e_worn * (1 - RE_WORN[0])
i_q = [[float(i == j) - q[i][j] for j in range(4)] for i in range(4)]
n = solve(i_q)
absorb = [math.fsum(n[0][k] * r[k][j] for k in range(4))
for j in range(2)]
return n[0], absorb
print("start: just promoted, working; wear-out hazard 1/52 per week")
print(" h weeks OK WORN live re-test OK re-test WORN"
" P(cut while OK)")
for h in sorted(CUSUM):
row, ab = chain(h)
print("%4.1f %8.1f %9.2f %10.2f %12.2f %15.4f"
% (h, row[0], row[1], row[2], row[3], ab[0]))
row, ab = chain(3.0)
print("h 3.0: expected life %.1f weeks; P(cut OK) + P(cut WORN) = %.4f"
% (math.fsum(row), math.fsum(ab)))Single source and citation
Fan, Y. (2026). Empirical Retirement of Live Rules: Wiring the Sequential Probability Ratio Test into Real Conversion Streams to Promote, Demote and Retire Copy Rules (Version 1.0.0), Chapter 5: The Rule Lifecycle as a State Machine: Promotion, Demotion and Retirement. 5D Library. DOI pending
DOI pending · monograph page goes live after publication · 5DL-B-263 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026263,
author = {Fan, Yafu},
title = {Empirical Retirement of Live Rules: Wiring the Sequential Probability Ratio Test into Real Conversion Streams to Promote, Demote and Retire Copy Rules},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}Other chapter answers in this book
- Chapter 1 · A Weekly Glance: How a Mediocre Copy Rule Gets Promoted by Mistake
- Chapter 2 · Wald's Two Straight Lines: Boundaries, the OC Function and the Mean Sample Size
- Chapter 3 · Exact Recursions: the Long Tail of the Sample Size and the Cost of a Cap
- Chapter 4 · Connecting Real Conversion Streams: Weekly Batches and Where the Baseline Comes From
- Chapter 6 · Many Rules in Parallel: Sharing Alpha, and How Many Promoted Rules Are Duds
- Chapter 7 · Classical Sources East and West: Checking by Trial in the Han Feizi, and Wald's Sequential Analysis
- Chapter 8 · Closing the Loop: Wiring the Weekly Radar's Verdicts into the Rule State Machine