Yesterday's 146 races produced a first-pick win rate of 20.5% — 30 winners from 30 attempts — which falls a few points below the seven-day running average of 26.4%. Not a bad day, but not the model at its sharpest either.
The broader picture is more encouraging. Among all three model selections, at least one of the top three landed the winner in 92 of 146 races — a 63% coverage rate that shows good discrimination across the card even when the top pick was getting turned over more often than usual. The any-top-three place rate ran at 93.8%, meaning in nearly all races, at least one of the three selections ran competitively.
Days that dip below 22% for the first pick tend to cluster when racing produces more interference-affected results or when sprint grades at smaller tracks produce unpredictable early positions. With 146 races spread across venues from Hove to Waterford, that kind of variance is built into the system. The model does not have a bad day in the way a single punter might — it has days where randomness bites harder than usual, and yesterday was one of those.
The seven-day context is reassuring. From 1,184 races this week, the first-pick win rate sits at 26.4%, and the top-three hit rate at 64.4%. Yesterday was a minor correction in an otherwise solid week, not a signal that anything has changed in how the model reads form.