An AI race predictor for UK racing — published in the open
Our in-house XGBoost + LightGBM model forecasts a win probability for every runner in every UK race, and we compare it to the market. Then we publish what actually happened — including the losing stretches. The honest finding: it does not reliably beat efficient bookmaker prices. This is research, not a tipping service.
Our in-house model lost 16.8% ROI on the pre-registered Oct-Nov 2024 backtest window.
This page publishes what it predicts and tracks every result. We do this because nobody else does — the methodology is open, the losses are visible, the analysis is honest. The model output is presented as a comparison to the market, not as a recommendation to back, lay, or stake on any runner.
Read the full methodology in our in-house AI horse-racing model write-up. Track the running ledger on the Stablebet track record page.
Gambling can be addictive. Please bet responsibly. Free, confidential support from GamCare, GamStop and BeGambleAware. See our responsible-gambling page for more.
Races today
2
model run each morning
ROI to SP
—
all-time, flat £1 win
Top-pick strike
—
ledger accruing
Today's predictions
Every UK race today — the model's win probability next to the market's, with where they disagree.
OpenTrack record
Cumulative ROI, calibration and strike rate over time. Every pick matched to its result, losses included.
OpenHow it works
The full methodology and backtest history — what the model is, and why it doesn't beat efficient prices.
OpenDoes AI horse racing prediction work? The honest answer
Does AI horse racing prediction actually work?
Not in the sense most sites imply. Our model is roughly twice as accurate as random at picking the winner, but it does not reliably beat efficient bookmaker prices — every filter that looked profitable on one window lost money on a pre-registered, untouched window. The market already prices in most of the same signal.
Can an AI model beat the bookies?
On UK racing, our evidence says no, not consistently. The overround is brutal and the market is efficient. If a model genuinely beat the bookies, the people running it would bet it themselves rather than publish it. We publish ours precisely because it's research, not an edge for sale.
Is this a tipping service?
No. We don't sell picks, run a subscription, or tell you what to back. The AI Race Predictor outputs probabilities and compares them to the market so you can interrogate the model. The track record exists so the claims are checkable, not taken on trust.
The full write-up — methodology, features, and every backtest window including the losing ones — is in our in-house AI horse-racing model guide.