Performance

Is the model actually calibrated?

A prediction engine is only credible if 80% selections win about 80% of the time. This page publishes that comparison directly, including the bands where the model is currently overconfident.

Loading settled results…

Settled bets
0
Win rate
0.0%
ROI
+0.0%
Profit
£0.00
Brier score
0.0000
Lower is better; 0.25 = coin flip
Avg closing line value
+0.0%

Calibration: predicted vs actual

Each band compares the average probability we published against how often those selections actually won.

50–59%predicted 0.0% · actual n/a · 0 bets
60–69%predicted 0.0% · actual n/a · 0 bets
70–74%predicted 0.0% · actual n/a · 0 bets
75–79%predicted 0.0% · actual n/a · 0 bets
80–84%predicted 0.0% · actual n/a · 0 bets
85–89%predicted 0.0% · actual n/a · 0 bets
90%+predicted 0.0% · actual n/a · 0 bets

The white marker is the predicted rate; the bar is the actual rate. Bands with no settled bets stay empty — nothing here is smoothed or filled in.

Cumulative profit

Sequential settled profit in units staked. The flat and declining stretches are real losing runs and are deliberately left visible.

Performance by competition

Competitions with at least three settled bets. Data reliability differs sharply between major and minor leagues, and so does the model's edge.

CompetitionBetsWin rateAvg oddsProfitROI

All figures are estimated model probabilities, not guarantees. No outcome is certain. Historical performance is measured separately from model confidence. 18+ only — gamble responsibly.