Over the last seven days the model has been tested across 859 races, and the headline numbers hold up well against the standard we'd expect. The top selection won 25.6% of the time and placed in 43.1% of races outright. Widen it to any of the top three picks and the figures jump to a 65.7% win rate and an 88.2% place rate — meaning the model had a runner in the frame in nearly nine out of every ten races run this week.
Day to day, there's been some natural variance, which is exactly what you'd expect from a model working across hundreds of independent races rather than a sign of anything structurally wrong. July 11th was the strongest day for the top pick, winning 38 of 142 races (26.8%), while July 14th was the softest, with the top selection landing 24 of 112 (21.4%). That's a real gap, but it sits well within the range of ordinary week-to-week fluctuation for a strike rate hovering around a quarter of all races — greyhound racing has enough genuine unpredictability, interference, and traffic in running that no model lands on the exact winner every time.
What's more reassuring is how consistent the any-top-3 figures have stayed across the same stretch, holding in the 60-75% win range and 80%+ place range on every single day in the sample, including the softer July 14th. That consistency is the real story this week — even on the days the outright top pick underperformed, the model was still doing its core job of narrowing each field down to the right handful of contenders. Nothing in this week's numbers points to a systematic issue with the ratings themselves.