Where the model and the market disagree
Each day at 09:00 BST our in-house ensemble model (XGBoost classifier + LightGBM LambdaRank + sentence-transformers commentary RAG) runs over every UK race card, producing a per-runner win probability. The tables below compare that probability to the market's implied probability after the overround is removed.
This is research output, not advice. We publish the model's thinking so readers can interrogate it — not so anyone can copy the numbers and bet them.
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.
Ascot 16:20 · Gold Cup
Group 1 · 2m4f · Going Good
| Horse | SP | Model prob | Market prob | Disagreement |
|---|---|---|---|---|
| Scandinavia R Moore / A O'Brien | 3.00 | 29.8% | 32.3% | Market 2.5pp shorter |
| Trawlerman R Havlin / J & T Gosden | 4.33 | 24.1% | 22.4% | Market 1.7pp longer |
| Sweet William (TBC) / J & T Gosden | 5.50 | 17.1% | 17.6% | In line with market |
| Rahiebb W Buick / W Haggas | 8.00 | 12.2% | 12.1% | In line with market |
| Illinois (Coolmore 2nd) / A O'Brien | 13.00 | 8.4% | 7.4% | Market 1.0pp longer |
| Caballo De Mar (TBC) / G Scott | 17.00 | 4.6% | 5.7% | Market 1.1pp shorter |
| Ethical Diamond (TBC) / W Mullins | 21.00 | 3.8% | 2.5% | Market 1.3pp longer |
Ascot 15:00 · Norfolk Stakes
Group 2 · 5f · Going Good
| Horse | SP | Model prob | Market prob | Disagreement |
|---|---|---|---|---|
| Sample Juvenile A R Moore / A O'Brien | 2.75 | 34.1% | 35.2% | Market 1.1pp shorter |
| Sample Juvenile B W Buick / C Appleby | 5.00 | 21.8% | 19.4% | Market 2.4pp longer |
| Sample Juvenile C T Marquand / W Haggas | 7.00 | 14.2% | 13.8% | In line with market |
How to read this page
- Model prob — what our model thinks the horse's win probability is, per-race softmaxed so the column sums to 100% across each race.
- Market prob — the market's implied probability after the bookmaker overround has been removed.
- Disagreement — where the model and market diverge, expressed as percentage points. “Market 2.5pp shorter” means the market is pricing the horse at a probability 2.5pp higher than the model thinks. We deliberately do not show this as a buy or sell signal.
For the full per-race deep-dive (5/5 trends scorecard, head-to-head match-ups, where-the-value-sits write-ups) see the relevant race preview. For the running ledger of past model predictions and outcomes, see /our-track-record/.