TrapMetricsThe analyst behind TrapMetrics: the person who built the rating model, runs it every morning, and writes up what it finds. The work is five years of the official UK and Irish record, 1.5 million runs, turned into a calibrated win chance per dog and tested in public against every settled race. The pieces here are written under a pen name; every number in them can be checked on the evidence page, which is the only credential that matters.
What the writer knows
- Probability models for greyhound racing: conditional logit and gradient boosting on point-in-time form
- Calibration and forward testing at Betfair closing prices, with the losing numbers published
- Going measured from race times rather than the official allowance; track geometry from split times
- System back-testing that refuses to flatter: holdout, walk-forward, bootstrap intervals, no Goal Seek
- The published expert forecasts, Timeform and the Racing Post, measured against the model and the market
How the articles are made
Every measurement uses only what was knowable before the off. Every figure carries its race count and its window. Anything chosen by searching the past says so. The market beats every rating, ours included, and that sentence appears on every page where it matters.