Wednesday's racing produced a first-pick win rate of 19.4% from 36 races — seven winners from what was a fairly thin card. The any-top-3 model returned 22 winners (61.1%), which is in line with expectations and suggests the model was identifying the right dogs to be interested in; it was just the exact finishing positions within the frame that got shuffled.
Seven from 36 is below the model's typical rate, but small card days need to be read in context. Mid-week afternoon meetings with thin fields and unconventional grade mixes are where prediction models encounter the most noise. A single bad trip, a slow break, a pace shape that does not develop as expected — these things affect a six-dog field far more than they affect a deeper one.
The place rate was stronger at 50%, with 18 of the 36 top picks finishing in the top two. The model was rarely wildly wrong about who was competitive; it was just the final strides that did not go the right way often enough.
The second and third picks combined to produce eight more winners each, bringing the any-top-3 total to 22. Worth a look through the full results if you want to dig into specific races from yesterday.