Model Leaderboard

One published pre-race forecast per eligible settled race, including abstentions. Wins mean 1st; places mean 1st–3rd, including wins.

Some verified pre-race reports are temporarily unavailable.

How We Track Performance

Each race uses its latest complete, immutable forecast recorded before the earliest scheduled start. Reruns count once; incomplete or late forecasts are excluded. The published top pick is scored even when the model abstained. Actual off-time is not independently verified.

ROI figures are calculated using flat $10 units on each top-rated runner, and every settled selection is included — including losing months. The model's top pick does not beat the starting-price favourite over the published record, and that comparison is on the track-record page rather than buried here.

A selection is settled only once the race is run and graded. Races still to jump are shown but excluded from the day's strike rate and ROI, and a scratching is a void — the stake comes back, so it is neither a win nor a loss. Where the feed publishes no starting price we settle at the market price recorded when the rating was made, which runs marginally below SP on winners; selections with no price at all are counted in the strike rate but left out of the ROI, and the priced sample size is shown beside it.

View full methodology →

See which races to skip and which have a clear model winner

Skip or play