Model Check: Seven Days of Prediction Data

Across 986 races over the past seven days, the first pick has won 241 times for a 24.4% strike rate. The first pick has placed 428 times, a 43.4% place rate. Widen the net to any of the top three selections and 600 of 986 races produced a winner from the model's top trio — that's a 60.9% hit rate, and the place figure climbs to 88.6%.\n\nDay by day, the picture has been steady rather than spectacular. Last Friday was the standout session: 48 first-pick winners from 178 races, a 27% win rate on the biggest card of the week. Saturday dipped slightly to 31 from 97 (31.9%), though the smaller sample inflates that percentage. Sunday's 133-race card was the toughest for the model with 27 first-pick winners (20.3%), suggesting a day where upsets and trouble-in-running disrupted the form book more than usual.\n\nThe midweek cards have been consistent. Monday landed 30 from 107 (28%), Tuesday hit 35 from 158 (22.2%), and Wednesday delivered 34 from 169 (20.1%). That slight downward drift from Monday to Wednesday is worth noting but sits within normal variance — larger cards tend to include more competitive, harder-to-read races at open and graded level.\n\nThe key number for most users is the 60.9% top-three win rate. It means that if you're using the model to narrow your selections rather than picking outright winners, you're getting a runner in the frame in six out of ten races. That's the real utility. Picking exact winners in a sport with first-bend trouble, slow breaks, and six-dog fields at tight margins will always be hard. Identifying the right three to focus on is where the model earns its keep.\n\nNothing here suggests any systemic blind spot this week. The models are reading fields correctly across tracks and grades. If anything, the consistency across seven days, despite card sizes ranging from 97 to 178, shows the predictions are stable under different conditions.