Most prediction apps advertise an accuracy number and never show their working. This page publishes ours in full - every resolved pre-match prediction, scored against the final result, with sample sizes, monthly history and the methodology behind it.
Covering matches from to . Last recalculated .
Overall accuracy
75.9%
5,199 resolved predictions
Avg Brier score
0.179
Lower is better · 0.25 = a coin flip
Matches covered
5,199
Pre-match predictions only
Every prediction carries a confidence rating. If the ratings are meaningful, high-confidence predictions should land more often than low-confidence ones. This is the table that shows whether that actually holds.
| Confidence | Predictions | Accuracy | Brier |
|---|---|---|---|
| High | 1,366 | 81.7% | 0.146 |
| Medium | 3,737 | 74.3% | 0.189 |
| Low | 96 | 57.3% | 0.251 |
A single lifetime average can hide a bad run. Each month is published as it lands, good or bad.
| Month | Predictions | Accuracy | Brier |
|---|---|---|---|
| December 2025 | 388 | 80.4% | 0.154 |
| January 2026 | 471 | 76.7% | 0.176 |
| February 2026 | 736 | 77.2% | 0.170 |
| March 2026 | 702 | 72.2% | 0.199 |
| April 2026 | 769 | 75.9% | 0.179 |
| May 2026 | 706 | 72.1% | 0.200 |
| June 2026 | 84 | 84.5% | 0.135 |
| July 2026 | 357 | 78.2% | 0.165 |
| August 2026 | 986 | 76.8% | 0.175 |
The competitions we have predicted most often. Coverage depth varies a lot, so the sample size column matters as much as the percentage.
| Competition | Predictions | Accuracy | Brier |
|---|---|---|---|
| Premier League | 497 | 76.3% | 0.177 |
| FA Cup | 416 | 76.0% | 0.181 |
| Championship | 322 | 75.5% | 0.182 |
| Europa Conference League | 306 | 76.8% | 0.175 |
| La Liga 2 | 280 | 75.4% | 0.186 |
| Serie B | 228 | 75.9% | 0.183 |
| La Liga | 225 | 77.8% | 0.172 |
| Serie A | 223 | 82.1% | 0.147 |
| Liga Portugal | 203 | 75.9% | 0.169 |
| Pro League | 192 | 78.7% | 0.164 |
| Super Lig | 184 | 80.4% | 0.158 |
| Ekstraklasa | 183 | 73.2% | 0.195 |
Only predictions published before kick-off are counted. Scorevance also produces live in-play predictions, and those are deliberately excluded: a prediction made at 85 minutes with a two-goal lead on the board is nearly always right, and counting it would inflate the headline number for free.
Where a fixture has more than one pre-match prediction, only the most recent one counts. A single match can never contribute more than once.
A prediction is only included once its match has finished and a final score is confirmed. Accuracy is the share of predictions whose selected outcome matched what actually happened. Postponed, abandoned and cancelled fixtures are not scored.
Hit rate alone rewards a model for hedging towards favourites. The Brier score measures calibration - how close the stated probability was to what happened - so a model claiming 90% confidence and being wrong is penalised more than one claiming 55%. Lower is better; 0.25 is what you would score by guessing at even odds.
Predictions are bucketed by when the match was played, not when the prediction was generated, so a Saturday fixture predicted on Thursday counts towards Saturday.
Any bucket with fewer than 30 resolved predictions shows its sample size but not a percentage. "3 out of 3 correct" is not a 100% accurate model, and publishing it as one would be misleading.
Every prediction on this page was published in the app before kick-off, with its probabilities and reasoning attached.
Get Scorevance - Free