Over the past seven days, the model assessed 1,199 races. The top pick produced 304 winners — a first-pick win rate of 25.4%. In a six-runner race with no form signals or draw advantages, a random selection would win around 16.7% of the time. The model outperformed that baseline by roughly nine percentage points across 1,199 races and a full week of racing.
The place rate for the top pick was 46.6%, meaning the highest-rated dog ran in the first two in almost half of all races last week. Extend the net to any of the top three predictions and the win coverage rises to 63.8%, with 88.2% of races returning at least one top-three prediction in the places.
There was variance within the week, which is normal. June 8 was the toughest day: 30 wins from 160 races for an 18.8% strike rate, still above baseline but below the weekly average. The following day brought a partial recovery, and by June 13 the Sunday card delivered the week's best single-day result — 66 wins from 246 races at 26.8%, which is comfortably above target.
No prediction model runs perfectly across every card. Track conditions, interference, and compressed form fields all create races where even a strong analytical signal gets overturned. The question is whether the model adds value over time, and over 1,199 races last week the answer was clearly yes.
Full daily accuracy figures are on the Today's Results page, updated throughout each racing day.