Python
159 posts
How Date Inference Works, and Why Ambiguity Is the Problem
Guessing a strptime format from examples is a scoring problem with ties. The dangerous outputs are the wrong guesses that still parse cleanly.
Bradley-Terry: Maximum Likelihood Ratings on an Elo Scale
Bradley-Terry fits every comparison at once instead of nudging after each game. Adding it to elote, and the four design choices that made it usable next to Elo.
Five Backtest Invariants Worth Enforcing In Code
A backtest simulates information availability, and every way it leaks the future flatters your returns. Five invariants, each with the test that pins it.
Same App, Ten Years Apart
Pedal Wrencher in 2015 and in 2026: identical product, zero shared code. What a decade of stack turnover actually bought, and what it didn't.
PyGeoHash v3.3.2: Reliability at the Edges
PyGeoHash v3.3.2 continues the reliability run with antimeridian fixes and stricter validation, plus native ARM64 wheels and smoke-tested artifacts.
Cache Keys Must Come From the Resolved Call
A memoization decorator that reads kwargs cannot see positional arguments, so distinct calls collide. How to build keys from the bound signature instead.
The Kelly Criterion With Real Payoffs and Real Fees
The Kelly formula everyone quotes assumes you lose your entire stake. With real loss multipliers and fees the sizing changes. Deriving the general form.
The Billing Bug You Can't Have
How emit's prepaid credit system makes overdrawing impossible under concurrency, using an append-only ledger and one carefully shaped SQL update.
Prove the Test Can Fail: Mutate the Fix
A regression test is unverified until you have watched it fail against the unfixed code. The procedure, and three shapes that survive a naive check.
Stars, Views, and Clones Across 63 Repos
I ran stargazers 0.2.0 against my own GitHub account. The repos with the most stars and the repos with the most clones turned out to be different repos.