Daily results
Each day's model picks graded on the 90-minute score, with correct and wrong counted side by side.
| Day | Date | Correct | Wrong | Accuracy |
|---|---|---|---|---|
| Tue | 17 Nov 2026 | 83 | 48 | 63% |
| Mon | 16 Nov 2026 | 56 | 15 | 79% |
| Sun | 15 Nov 2026 | 314 | 146 | 68% |
| Sat | 14 Nov 2026 | 355 | 174 | 67% |
| Fri | 13 Nov 2026 | 88 | 45 | 66% |
| Thu | 12 Nov 2026 | 47 | 26 | 64% |
| Wed | 11 Nov 2026 | 64 | 38 | 63% |
| Tue | 10 Nov 2026 | 60 | 26 | 70% |
| Mon | 9 Nov 2026 | 47 | 21 | 69% |
| Sun | 8 Nov 2026 | 527 | 233 | 69% |
| Sat | 7 Nov 2026 | 549 | 251 | 69% |
| Fri | 6 Nov 2026 | 165 | 76 | 68% |
| Thu | 5 Nov 2026 | 65 | 30 | 68% |
| Wed | 4 Nov 2026 | 90 | 46 | 66% |
| Tue | 3 Nov 2026 | 94 | 61 | 61% |
| Mon | 2 Nov 2026 | 59 | 28 | 68% |
| Sun | 1 Nov 2026 | 263 | 145 | 64% |
| Sat | 31 Oct 2026 | 423 | 263 | 62% |
| Fri | 30 Oct 2026 | 148 | 86 | 63% |
| Thu | 29 Oct 2026 | 101 | 57 | 64% |
| Wed | 28 Oct 2026 | 166 | 88 | 65% |
| Tue | 27 Oct 2026 | 155 | 80 | 66% |
| Mon | 26 Oct 2026 | 57 | 30 | 66% |
| Sun | 25 Oct 2026 | 326 | 178 | 65% |
| Sat | 24 Oct 2026 | 510 | 276 | 65% |
| Fri | 23 Oct 2026 | 166 | 106 | 61% |
| Thu | 22 Oct 2026 | 59 | 49 | 55% |
| Wed | 21 Oct 2026 | 82 | 49 | 63% |
| Tue | 20 Oct 2026 | 88 | 42 | 68% |
| Mon | 19 Oct 2026 | 86 | 51 | 63% |
Grading rules
Picks are graded on the score at the end of normal time, stoppage time included; extra time and penalties are ignored and postponed or abandoned games are void. Value picks are counted separately because they are priced longer and judged on profit rather than accuracy – see the leaderboard. Yesterday's full list with final scores: yesterday's predictions.
Results: every pick graded on the 90-minute score
This page lists finished matches with the model's pre-match pick and its outcome. Nothing is removed or edited after kick-off: each selection appears exactly as it was published, followed by the final score and a tick or a cross. It is the raw material for every record and accuracy figure on the site.
How grading works
Each pick is settled on the score at the end of normal time, including stoppage time. Extra time and penalty shoot-outs in cup ties are ignored, which is how almost every bookmaker settles standard markets too. Postponed or abandoned matches are marked void and excluded from hit rates. Draw no bet and whole-line Asian handicaps can also produce a void when the stake is returned.
Reading a day's results
A day's hit rate is a small sample. If the model's picks average 65% probability, a perfectly calibrated set of sixty picks will produce anywhere between about 33 and 45 winners on most days, simply through chance. Use the daily numbers as a snapshot and the longer records for judgement:
- The track record summarises thirty and ninety days, with profit at level stakes.
- The calibration page groups picks by stated probability and compares that with how often they landed.
- The leaderboard ranks markets and leagues by return.
What a loss tells you
One losing pick tells you almost nothing about the model. A 70% selection loses three times in ten by design. What matters is whether, across hundreds of 70% selections, the winners come close to seventy in every hundred – and whether the prices taken were long enough to make that hit rate profitable.
Expected goals versus outcome
Each finished match keeps its pre-match expected goals. Comparing them with the final score is instructive. A side projected at 2.3 goals that scores none has either faced an excellent goalkeeper, wasted good chances or been less dominant than expected. Over a few rounds, the ratings learn from such results; over a single match, they mostly reflect variance.
Using the archive
Pick a date to see that day's full graded list. If you are testing a strategy – say, backing only picks with positive expected value in one league – the archive lets you check how it would have performed. Remember that past performance never guarantees future results, especially over short periods.
Related pages
The yesterday board shows the previous day in the same format as live predictions, and the weekly model report in insights summarises each week. Over 18s only.
A simple review routine
Once a week, filter the results for the markets you bet on, note the hit rate and average odds, and compare the hit rate with the break-even figure (1 ÷ average odds). Doing this regularly turns the results archive into a feedback loop rather than a scoreboard, and it helps you notice a weak market before it costs too much.
It also helps to separate results by price band. Picks at odds below 1.50 behave very differently from picks above 2.50: the first group wins often and loses rarely but expensively, the second loses often and wins less often but generously. Mixing them in one hit rate hides more than it reveals.