How accurate are our racing predictions?

Every UK and Ireland runner gets a win chance from our model. Here is the honest test of it: when we say a horse has an X% chance, does it win about X% of the time? The chart below is measured forward, on ratings stamped before each race ran, so nothing is fitted after the fact. It is the one number that matters to us, and we show it in full.

The reliability chart

Each dot is a band of our ratings. The closer a dot sits to the dotted line, the more honestly that rating reflects reality: a 20% call that wins 20% of the time lands right on it. Measured forward only, on scores stamped before each race, so nothing is tuned after the result.

Loading the live record…

Explore the full 1.9 million-run history

Window

Code

Class

Distance

Field size

Loading the record…

What this is, and what it isn't

This is calibration: do higher-rated horses win more often than lower-rated ones? They do, consistently, and that is the whole point of the score. It is not a betting system. Across most filters the level-stakes return is still negative, because the market is efficient and hard to beat. We would rather show you that honestly than sell you a tip. Figures are settled at Starting Price.