Seven days of data, 1,218 races run, and the headline number is 25.6% — the first-pick win rate across the full week. In a typical six-dog race, a random selection would land around 17%. The model's consistent edge above that baseline is what the whole system is built to demonstrate, and 25.6% over more than 1,200 races is a meaningful result.
The week was not uniform. The best day was Monday the 9th of June, where 38 wins from 126 races gave a 30.2% first-pick strike rate — a notably strong performance across a smaller card. The weakest day was Sunday the 8th, where just 30 winners from 160 races produced an 18.8% rate that barely cleared the random baseline. On days like that, upsets cluster across multiple cards and the form dogs run into variables that no model can fully anticipate — interference, going changes, dogs simply running below their rating on the day.
What is encouraging is the speed of the recovery. Monday the 9th bounced back to 30.2%, and Tuesday the 10th held at 27.9%. The dip on the 8th looks like variance rather than a structural problem. By Friday the 12th, 54 wins from 222 races put the rate back at 24.3%, consistent with the week's average.
The any-top-3 numbers add important context. Over seven days, 63.6% of races were won by one of the model's first three picks. The place rate on those top three reached 87.8%, meaning in nearly nine out of ten races run this week, at least one of the top three selections was in the money. That is where the model's genuine value sits — it does not always identify the outright winner, but it consistently maps the likely participants in the finish far more accurately than chance would produce.
Larger cards — Friday's 222 races, Saturday the 6th's 253 — both came in between 24-25%, which reflects the statistical reality that bigger samples pull results toward the model's true average. Smaller cards allow more variance in either direction. The full daily breakdown is available on the results page.