Seven days of racing, 1,211 races, 315 first-pick winners. The weekly first-pick win rate lands at 26% — in line with long-run expectations and above the baseline you would get from picking at random in a six-runner field.
The week had a notable range. June 8 was the flat day: 30 first-pick winners from 160 races, a 18.75% win rate that sat well below the weekly average. No obvious single cause stands out from the data — it was the kind of day where the model's top selections simply got beaten by second and third picks more often than usual. June 13 (Saturday) was the best, with 66 winners from 246 races (26.8%) on the week's largest programme. The model tends to perform better on bigger cards, which makes intuitive sense: the volume of clear-form races dilutes the low-confidence selections.
Place rates are where the week's performance looks strongest. The first-pick place rate — covering wins and seconds — runs at 47.6% across all 1,211 races. That means nearly half of all top selections at least hit the frame. Zoom out to the top-three selections and the any-top-3 place rate reaches 88.7%, meaning nine in ten races had one of the model's three picks in the money. That is a useful benchmark for anyone running place-based approaches.
The gap between the first-pick win rate (26%) and the any-top-3 win rate (64.3%) is the most revealing number. Roughly 38% of races are being won by the model's second or third pick rather than the top selection. That tells you the ratings are doing a good job identifying the competitive runners in a race, even when the exact finishing order is hard to call. For win-only systems, 26% is the number to measure against. For anyone spreading across the top three, this week's data makes a reasonable case for that approach.