Every pick, won or lost
Each pick is published before the game starts and marked won or lost from the final score when it ends. Nothing is edited afterwards and nothing is deleted, so the losses are on this page beside the wins. Every figure carries the number of picks it is based on.
This record is 49 days old and 12,531 picks long. That is a start, not proof: a handful of days cannot tell a good model from a lucky week. Read it, then come back as it grows.
Daily win rate
Since 15 Sep · every graded dayMonth by month
Every settled pick, and the best bets alone| Month | Win rate | Best bets | Picks |
|---|---|---|---|
| Nov 2026 | 60.9% | — | 616 |
| Oct 2026 | 61.3% | 73.9% 82 of 111 | 10,068 |
| Sep 2026 | 52.8% | 79% 49 of 62 | 1,847 |
By market
Every settled pick, split three waysBy confidence
Does the model’s own score predict anything?| Confidence | Picks | Won | Win rate |
|---|---|---|---|
| 86 and overany market, not only best tips | 303 | 220 | 72.6% |
| 80 to 85 | 2319 | 1729 | 74.6% |
| 70 to 79 | 2031 | 1307 | 64.4% |
| 60 to 69 | 3191 | 1909 | 59.8% |
| Under 60 | 4687 | 2354 | 50.2% |
A confidence score is the model saying how sure it is. If it means anything, the top band should win more often than the bottom one. This record is far too short to settle that either way, which is why each band prints the number of picks behind it — judge the figures against those counts, not on their own.
Day by day
The last two weeks, newest first| Day | Picks | Won | Win rate | Best bets |
|---|---|---|---|---|
| Mon, 2 Nov | 27 | 16 | 59.3% | — |
| Sun, 1 Nov | 589 | 359 | 61% | — |
| Sat, 31 Oct | 820 | 507 | 61.8% | 50% 1 of 2 |
| Fri, 30 Oct | 347 | 224 | 64.6% | 100% 1 of 1 |
| Thu, 29 Oct | 191 | 117 | 61.3% | 100% 1 of 1 |
| Wed, 28 Oct | 354 | 226 | 63.8% | 100% 1 of 1 |
| Tue, 27 Oct | 213 | 131 | 61.5% | — |
| Mon, 26 Oct | 146 | 95 | 65.1% | 0% 0 of 1 |
| Sun, 25 Oct | 629 | 389 | 61.8% | — |
| Sat, 24 Oct | 763 | 481 | 63% | 66.7% 2 of 3 |
| Fri, 23 Oct | 297 | 204 | 68.7% | — |
| Thu, 22 Oct | 245 | 150 | 61.2% | 100% 1 of 1 |
| Wed, 21 Oct | 336 | 225 | 67% | 100% 3 of 3 |
| Tue, 20 Oct | 174 | 104 | 59.8% | — |
The last few days link to their board, with every pick marked won or lost.
How the record is kept
A pick is one prediction on one market. The model publishes it before the game starts, with the odds it was priced against and a confidence score, and it is marked won or lost from the final score when the game ends. Ties and cancelled games settle no bet, so they are left out of both columns rather than counted as anything.
Nothing here is edited after a game starts and nothing is removed. That is the only claim this page makes, and it is the one worth checking: today’s board shows the picks before they are played, and this page shows the same picks after. The numbers on both come from the same place.
There is no profit figure here, and there will not be one. A staking return is a number about money nobody actually put down, a few long-priced winners move it further than a month of picks does, and it invites you to read a prediction site as an investment. What the model can honestly be judged on is how often it is right and how many times it has been asked — both are on this page.
Frequently asked questions
Every pick on this page was published before its game started. An account opens the ones that have not been played yet, on every game, before the game starts.An account opens every pick before the game starts.
Past results do not promise future ones: these are a model’s opinions, not certainties. Bet only what losing would not change your week. How we make our predictions