Seven days of racing, 1,174 races, and a first-pick win rate of 26.2% across the board. That is the headline number from the past week of predictions, and for context: in a six-runner field, a random pick would win roughly 16.7% of races. The model's 26.2% represents a consistent edge across a large and varied sample.
The week was not uniform, though. Tuesday June 17th produced the strongest single-day performance: 54 winners from 169 races (32%), the kind of day where the model's pace-and-suitability weighting aligned cleanly with how races actually ran. Sunday June 21st matched it almost exactly at 32.5% from 120 races, which is encouraging given that Sunday cards tend to feature a broader mix of tracks and grades.
The two weaker days were Thursday June 18th (20.5% from 146 races) and Friday June 19th (19.2% from 219 races). The 219-race Friday card was the largest single-day sample of the week, so a below-average hit rate on a big card is the kind of variance that pulls the weekly aggregate down noticeably. Across a larger volume of races, individual upsets carry more weight in the aggregate.
The any-top-3 figures tell the broader story. Across the week, one of the three predicted finishers won in 63% of races, and 88.5% of races saw a predicted runner place in the first two. Those numbers hold steady even on the weaker days, which suggests the model's ranking order is generally sound even when the top pick gets beaten. The place rate in particular -- 43.4% for the first pick alone -- reflects a model that is identifying the competitive frontrunners consistently, even when the winner comes from slightly lower in the predicted order.