Prediction performance
Predictions are locked before kickoff and never edited. This page grades every market against real results.
269Predictions locked
250Main tips graded
70%Main-tip hit rate
67%Average stated probability
Every market, graded
| Market | Graded | Stated chance | Actual hit rate | Actual − stated | Return per 100 staked* |
|---|---|---|---|---|---|
| Home win | 250 | 45.2% | 50.8% | +5.6 pts | +10.3 |
| Draw | 250 | 23% | 18.4% | −4.6 pts | −21.4 |
| Away win | 250 | 31.8% | 30.8% | −1.0 pts | −7.0 |
| Home or draw (1X) | 250 | 68.2% | 69.2% | +1.0 pts | +0.8 |
| Away or draw (X2) | 250 | 54.8% | 49.2% | −5.6 pts | −13.1 |
| Either team wins (12) | 250 | 77% | 81.6% | +4.6 pts | +5.9 |
| Both teams to score (GG) | 250 | 55.4% | 57.6% | +2.2 pts | +6.0 |
| Not both teams to score (NG) | 250 | 44.6% | 42.4% | −2.2 pts | −0.7 |
| Over 1.5 goals | 250 | 78.3% | 82.4% | +4.1 pts | +6.2 |
| Under 1.5 goals | 250 | 21.7% | 17.6% | −4.1 pts | −7.9 |
| Over 2.5 goals | 250 | 56% | 60.8% | +4.8 pts | +11.5 |
| Under 2.5 goals | 250 | 44% | 39.2% | −4.8 pts | −9.0 |
| Over 3.5 goals | 250 | 34.7% | 40.8% | +6.1 pts | +25.5 |
| Under 3.5 goals | 250 | 65.3% | 59.2% | −6.1 pts | −8.7 |
By odds band
| Fair odds | Graded | Stated chance | Actual hit rate | Actual − stated | Return per 100 staked* |
|---|---|---|---|---|---|
| Under 1.40 | 724 | 79.1% | 79.8% | +0.7 pts | +1.0 |
| 1.40 to 1.79 | 707 | 63.4% | 64.1% | +0.7 pts | +1.1 |
| 1.80 to 2.29 | 672 | 49.7% | 49.4% | −0.3 pts | −0.2 |
| 2.30 to 3.19 | 543 | 37.8% | 37.4% | −0.4 pts | −1.0 |
| 3.20 and above | 854 | 22.2% | 21.5% | −0.7 pts | −1.4 |
How to read this. "Stated chance" is what the model said before kickoff; "Actual hit rate" is how often that market really happened. A well-calibrated model has the two close together, and small samples are noisy, so judge it over hundreds of picks.
* Return per 100 staked assumes every pick was bet at the model's own fair odds (100 ÷ stated chance), before any bookmaker margin. Real bookmaker odds are lower, so real returns would be lower than shown. A result near zero means the model is honest about its own odds, not that it beats the bookmakers.