Seven days, 948 races, and a dataset large enough to draw meaningful conclusions. Here's how the prediction models performed over the past week.
The first pick won 230 of 948 races for a 24.3% win rate and placed in 407, giving a place rate of 42.9%. Those are solid, workable numbers. A near one-in-four win rate on the top selection means the model is consistently identifying the strongest contender in the field, and when it doesn't win, it's placing more often than not.
The second pick contributed 181 winners (19.1%) with 389 places, and the third pick added 160 winners (16.9%) with 332 places. The drop-off between selections is gradual rather than a cliff edge, which tells you the model isn't just finding one obvious favourite — it's ranking the competitive dogs in a meaningful order.
Across all three picks combined, 571 of 948 races produced at least one winner from the top three — that's a 60.2% any-top-3 win rate. The place coverage is more striking still: 839 out of 948 races, or 88.5%, saw at least one of the top three finish placed. For anyone building forecast or combination bets, that coverage figure is the one that matters most.
Day by day, the week had its peaks and troughs. Friday was the strongest individual performance with 48 first-pick winners from 178 races (27.0%). Wednesday was tougher at 23 from 131 (17.6%) — a below-average session that dragged the weekly numbers down slightly. That kind of variance is normal and expected. No model wins every day, and the honest read is that a 17.6% day over 131 races represents a genuinely difficult card rather than a model failure.
The seven-day view is encouraging. The fundamentals are holding up, the place rates are strong, and the any-top-3 coverage gives users a consistent platform to work from.