What does A Screenshot Is Not a Payment conclude, in numbers, about “The Last Five Digits and a Time Window: How Far Collisions Fall, and What Will Not Fall”?
Answer: Chapter 2 shows that at a thousand orders a day, amount plus last five digits plus a thirty-minute window collides once in 93.9 days on average, and at five thousand once in 3.8 days; repeat orders by the same buyer still tie on the key 1.220 times a day. The finite-state transducer of Chapter 3 fails nowhere in an exhaustive round trip of 509,498 texts.
Scope of the numbers: Orders, account tails, amounts, screenshots and hashes are synthesised from fixed seeds, parameters are illustrative, and every calculation is generated by the book's programs. 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: 3.8, 93.9
| collision mass of amounts s | 0.051690 |
| window factor q | 30 min 0.04123, 10 min 0.01384 |
| expected tied pairs per day | |
| key n | 20 n=50 n=200 n=1000 n=5000 |
| AMT | 9.82104 63.31984 1028.6 25819.0 645991.6 |
| AMT+L4 | 0.00098 0.00633 0.10286 2.58190 64.59916 |
| AMT+L5 | 0.00010 0.00063 0.01029 0.25819 6.45992 |
| AMT+L5+W30 | 0.00000 0.00003 0.00042 0.01065 0.26636 |
| AMT+L5+W10 | 0.00000 0.00001 0.00014 0.00357 0.08941 |
| days per expected tie, key AMT+L5+W30 and AMT+L5 | |
| n | 20 246945.7 days 10182.2 days |
| n | 50 38301.8 days 1579.3 days |
| n | 200 2357.8 days 97.2 days |
| n | 1000 93.9 days 3.9 days |
| n | 5000 3.8 days 0.2 days |
Show the full program
# Chapter 2. p21: adding the last digits and a time window to the key
# Model: the amount law of p12 (collision mass s, recomputed here);
# the payer's last k account digits are uniform on 10^k values and
# independent of the amount; payment times are uniform over a
# 1,440-minute day, so two payments fall within w minutes of each
# other with probability q = (2wT - w^2) / T^2. A tie on the key
# needs all parts to agree, so E[tied pairs per day] is
# C(n,2) * s * 10^-k * q. Independence is an assumption that Chapter
# 2 relaxes for repeat buyers (p22). Daily volumes of 20, 50, 200,
# 1,000 and 5,000 orders are illustrative.
import math
from collections import defaultdict
PRICES = [59 + 40 * k for k in range(30)]
W = [1.0 / r for r in range(1, 31)]
POP = [w / math.fsum(W) for w in W]
MIX = {1: 0.70, 2: 0.25, 3: 0.05}
T = 1440.0
def collision_mass():
parts = defaultdict(list)
for k, pk in MIX.items():
def walk(depth, total, prob):
if depth == k:
parts[total + (15 if total < 299 else 0)].append(prob)
return
for i, p in enumerate(POP):
walk(depth + 1, total + PRICES[i], prob * p)
walk(0, 0, pk)
law = [math.fsum(v) for v in parts.values()]
return math.fsum(p * p for p in law)
def q_window(w):
return (2 * w * T - w * w) / (T * T)
s = collision_mass()
KEYS = [("AMT", 1.0), ("AMT+L4", 1e-4), ("AMT+L5", 1e-5),
("AMT+L5+W30", 1e-5 * q_window(30)),
("AMT+L5+W10", 1e-5 * q_window(10))]
VOL = (20, 50, 200, 1000, 5000)
print("collision mass of amounts s = %.6f" % s)
print("window factor q: 30 min %.5f, 10 min %.5f"
% (q_window(30), q_window(10)))
print("expected tied pairs per day")
print("key " + "".join("%11s" % ("n=%d" % n) for n in VOL))
for name, f in KEYS:
row = ""
for n in VOL:
lam = n * (n - 1) / 2.0 * s * f
row += "%11.5f" % lam if lam < 100 else "%11.1f" % lam
print("%-14s%s" % (name, row))
print("days per expected tie, key AMT+L5+W30 and AMT+L5")
for n in VOL:
a = n * (n - 1) / 2.0 * s * KEYS[3][1]
b = n * (n - 1) / 2.0 * s * KEYS[2][1]
print(" n=%5d %12.1f days %10.1f days" % (n, 1 / a, 1 / b))Single source and citation
Fan, Y. (2026). Fault-Tolerant Parsing of Payment Receipts and Spoken Orders: Last-Five-Digit Write-Off Automata and Blocking Duplicate-Screenshot Fraud (Version 1.0.0), Chapter 2: The Last Five Digits and a Time Window: How Far Collisions Fall, and What Will Not Fall. 5D Library. DOI pending
DOI pending · monograph page goes live after publication · 5DL-B-267 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026267,
author = {Fan, Yafu},
title = {Fault-Tolerant Parsing of Payment Receipts and Spoken Orders: Last-Five-Digit Write-Off Automata and Blocking Duplicate-Screenshot Fraud},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}Other chapter answers in this book
- Chapter 1 · Matching on Amount Alone Must Collide: the Birthday Problem of Same-Day, Same-Amount Payments
- Chapter 3 · A Finite Automaton for Spoken and Written Chinese Amounts, Simplified and Traditional
- Chapter 4 · The Write-Off Automaton: a Screenshot Is a Claim, a Credit Is a Fact
- Chapter 5 · Average Hash and Hamming Distance: Detecting Repeated Screenshots, and Where Detection Is Blind
- Chapter 6 · The Receipt Fingerprint Library: the Birthday Problem Among Hashes, and Pigeonhole Search
- Chapter 7 · Classics East and West: Half-Seal Tallies in the Collected Statutes of the Ming and Galton's Finger Prints
- Chapter 8 · The Buying Agent's Reconciliation Loop and a Protocol for Refutation