Football track record
Every prediction we publish is counted here, wins and misses alike. The last version before kick-off is the one that counts. Nothing is deleted.
Headline numbers
We started publishing recently: 5 matches published, none graded yet. Numbers appear here after the first final whistle.
Calibration
Every 1X2 probability we published, grouped in 10% bins. If we say 30%, it should happen about 30% of the time: well calibrated dots sit on the dashed line. Bigger dot = more matches.
| Predicted | Outcomes | Avg predicted | Happened |
|---|---|---|---|
| No graded outcomes yet. | |||
By league
| League | Matches | 1X2 accuracy | Brier | Market Brier | CLV |
|---|---|---|---|---|---|
| Premier League | 0 | not enough data yet | |||
| Championship | 0 | not enough data yet | |||
| League One | 0 | not enough data yet | |||
| League Two | 0 | not enough data yet | |||
| LaLiga | 0 | not enough data yet | |||
| Serie A | 0 | not enough data yet | |||
| Bundesliga | 0 | not enough data yet | |||
| Ligue 1 | 0 | not enough data yet | |||
League percentages appear from 30 graded matches; below that, chance dominates.
By market
| Market | Graded | Accuracy | Brier | Market accuracy | Market Brier |
|---|---|---|---|---|---|
| 1X2 | 0 | – | – | – | – |
| Over/under 2.5 | 0 | – | – | – | – |
| Both teams to score | 0 | – | – | – | – |
1X2 Brier is the 3-outcome version (0 = perfect, 2 = worst); over/under and BTTS use the yes/no version (0–1). Glossary
Fans vs model vs market
Fans vote on every match page before kick-off. The comparison appears once matches with at least 20 votes are played.
By month
| Month | Matches | 1X2 accuracy | Brier | Market Brier | CLV |
|---|---|---|---|---|---|
| No graded months yet. | |||||
How to check us
Each day's predictions are written to a JSON file. After the day ends (UTC), we seal it and publish its SHA-256 fingerprint. Each day's file includes the previous day's fingerprint, so changing any old prediction would break every fingerprint after it. You can download any day and check it yourself.
Limitations
- The sample is still small: 0 graded matches. Accuracy can swing by several points by chance alone; treat anything under a few hundred matches as early evidence.
- We have not yet shown that we beat the market. Until the Brier score and CLV are consistently better than the market's, assume we don't.
- Market prices come from one bookmaker feed; other books may differ.
- Accuracy alone is a weak measure: picking the favourite every time scores well. Brier score and calibration are the honest tests.