Seven days of data, 1,181 races. The top-rated dog won 300 of them — a first-pick win rate of 25.4%. The top three picks between them won 63.4% of all races, and placed in 88.9%.
The week wasn't uniform. June 13 was the highest-volume day: 246 races, 66 first-pick winners, 26.8% strike rate. June 17 was the cleanest performance: 54 winners from 169 races, a 32% strike rate — the kind of day where the data lines up with the results across a range of tracks and grades. June 14 was the weakest, 28 wins from 119 races at 23.5%, and Friday 19 June came in at 19.2% — below the week's average on a high-volume evening of 219 races.
The variation is real and worth understanding. High-race-count days tend to include a wider spread of tracks, grade conditions and racing jurisdictions. On the Irish side especially, where grading conventions differ from UK standards and some tracks have thinner form data in the database, the model carries more uncertainty per race. The stronger days clustered around mid-week UK-heavy cards where the form data is deeper and more consistent.
The place rate tells a more stable story than the win rate. Top picks place in 43.7% of races across the week — roughly double what you'd get from a random selection in a 6-runner field. The model's clearest strength is identifying dogs within range of winning. Exact finishing positions are harder to pin down, but the selection radius is reliable. That's what the 88.9% any-top-3-place rate reflects: when the model's top three picks are all in the right area, the field tends to confirm it.