Champions League ยท Track record
How the recorded top picks for Champions League have done. Each pick is recorded when the message goes out and settled after its match. Bets count two ways: as % of bankroll at the stake the message suggested, and flat, as 1 unit on every bet. Predictions only count towards the hit rate, since they came with no stake.
No picks recorded yet for Champions League.
Every prediction
The picks above are the few matches the model felt strongest about. This is all of them: the model's home/draw/away chances for every match, recorded shortly before kickoff with the books' lines beside them, and scored the same way once the match is over.
| Matches | Log-loss | Brier | RPS | Called right | Model's log-loss on the same matches | Gap (model โ it) | ||
|---|---|---|---|---|---|---|---|---|
| Model | 25 | 0.9349 | 0.5464 | 0.2238 | 68% | |||
| DraftKings | 25 | 0.7338 | 0.4056 | 0.16 | 80% | 0.9349 | +0.2011 (+0.0828 to +0.3168) | no clear difference |
| Kalshi | 25 | 0.7377 | 0.4064 | 0.1615 | 80% | 0.9349 | +0.1971 (+0.0715 to +0.3180) | no clear difference |
| Polymarket | 25 | 0.7373 | 0.4066 | 0.1615 | 80% | 0.9349 | +0.1976 (+0.0774 to +0.3168) | no clear difference |
Does the injury discount help?
On the 18 matches where a missing player changed the prediction, the model scored 0.9456 with the discount and 0.9239 without it (gap +0.0217, -0.0152 to +0.0616): no clear difference yet.
Do the percentages come true?
| Chance given | Times | Average given | Happened |
|---|---|---|---|
| 0โ20% | 12 | 13.2% | 16.7% |
| 20โ30% | 29 | 24.6% | 13.8% |
| 30โ40% | 14 | 36% | 35.7% |
| 40โ50% | 8 | 44.8% | 75% |
| 50โ60% | 6 | 54.4% | 66.7% |
| 60โ70% | 3 | 65.8% | 66.7% |
| 70โ100% | 3 | 80.1% | 66.7% |