Seven days of data, 933 races. Here is what the models produced.
The headline figure is a 24.5% first-pick win rate across the full week. In a six-runner race where the theoretical baseline is around 16.7%, a rate of 24.5% represents a meaningful edge — the models are identifying winners at roughly one and a half times the baseline rate. That number is consistent across the week rather than being inflated by a single bumper day.
Place rates tell a similar story. The first pick placed — running in the first three — in 43.6% of all races. And when you look at any of the model's top three picks, 61.1% of races saw at least one of them win. Those three picks placed in a remarkable 88.9% of all races across the 933-race sample.
Looking at the individual days, Saturday May 30th was the strongest session with 48 first-pick wins from 178 races (27.0%), while Wednesday May 28th was the toughest at 23 wins from 131 (17.6%). That kind of variance is completely normal in greyhound racing — midweek cards often feature more mixed-grade racing where upsets are more frequent, and the lower sample sizes on individual days amplify the swings.
Tuesday June 2nd, yesterday's card, came in at 28.0% first-pick wins from 107 races, which was one of the better single-day returns of the week. The any-top-3 place rate yesterday was 95.3% — the model included the finishing dog in its top three for all but five races on the card.
The broader picture across the seven days is one of steady, reliable performance without dramatic outliers in either direction. A 24.5% first-pick win rate is genuinely useful if punters are getting prices above the implied probability. At 24.5%, the implied fair price is roughly 4.08 — meaning any pick that returns more than 4.08 in decimal odds is generating value if the model is correctly calibrated over a large enough sample.
All daily results going back further are available to review.