Seven days of racing, 1,103 races tracked. Here is what the model returned.
The headline number for the week: 24.7% first-pick win rate. That is the percentage of races where the top-rated dog won. For context, a random selection across a typical six-runner field would win 16-17% of the time. The model is running around 8 percentage points clear of chance, which is the consistent signal the system has been built on.
First-pick place rate (finishing in the top two) was 41.9%, and the any-top-three win rate — meaning at least one of the three predicted runners finished first — was 62.2%. The any-top-three place rate was 88.1%, so in nearly 9 out of 10 races across the week, one of the model's three picks finished in the frame.
The range across individual days tells the real story. June 17 was the strongest: 54 winners from 169 races, a 32% first-pick win rate. June 19 was the toughest: 42 from 219, which is 19.2% — comfortably below average. That kind of variance is normal in greyhound racing. Individual days are heavily influenced by how many upsets occur, how much interference there is, and whether front-runners or closers dominate on a given evening.
What the weekly totals show is that these swings average out. 24.7% over more than a thousand races is not noise. The model does not predict every winner — nothing could — but it finds the right dog more often than the alternatives. If you want the full day-by-day breakdown, the results page has everything.