Seven days, 1,181 races, and 262 winning first picks. That is a 22.2% win rate for the model's top-rated selection across the full week -- 5.5 percentage points above the 16.7% baseline you would expect from random selection across six-runner fields. For context, a sustained edge of that size over four figures of races is a meaningful signal, not statistical noise.
Saturday June 6th was the week's strongest day: 63 winners from 253 races (24.9%), with 117 first-pick places and 151 any-top-3 winners from 253 races (59.7%). Thursday June 5th was the toughest, producing 42 winners from 264 races (15.9%) -- the one day below the neutral baseline. That was a high-volume card, and underperformance on large days has a disproportionate effect on weekly averages. It is worth noting that 264 races in one day includes a wide range of grades and tracks, and some of the harder-to-predict race types naturally cluster on busy days.
The place rate for the week sits at 41.2% for first picks -- important for anyone running place-based staking systems. The any-top-3 win rate of 58.4% means the winner came from our top three predictions in nearly six of every ten races over the seven days.
What is working well is day-to-day consistency. No day in the past week posted a first-pick win rate below 15.9%, and four of the seven cleared 20%. The model's combination of speed ratings, suitability scores, and field-relative composite weighting is designed to produce steady output rather than volatile swings. Full day-by-day and model-by-model breakdown is on the [results page](/todays-results).