Every prediction, graded in the open
For every settled match we compare the model's most-likely outcome to what actually happened. Prediction accuracy and calibration — not ROI, not tips. An analysis you can't check is just a tip.
We publish the leagues the model is validated on. A few competitions (including Brazil's Série A) are still in calibration and aren't scored here yet — the real season starts when the big European leagues kick off.
Across 150 settled football predictions, Momus Modal's model called the correct 1X2 outcome 56% of the time, versus 54.5% for the market's own favourite on the same fixtures, and its pick avoided defeat (double chance) 78% of the time, with a Brier calibration score of 0.597. This is prediction accuracy measured in the open — not ROI, not tips.
Accuracy by window
Confidence vs outcomes
When the model says it's X% confident, does it hit X% of the time? A small gap means the probabilities are honest.
Reliability by predicted outcome
League evidence
How we compare
Most prediction sites lead with a single, round headline win-rate — usually blended across easy markets (like over-0.5-goals), quoted as marketing, and rarely graded in the open. We publish the hard one instead: 1X2 accuracy on the actual result, with the full sample and method, updated live. Judge us on that.
Cross-site numbers aren't apples-to-apples — different markets, leagues, samples and grading rules — so read this as context, not a scoreboard. Competitor claims cited from their public pages (nerdytips.com), July 2026.
Slice this record your way — by league, outcome, model confidence, divergence from the market or price bracket — and see the model's accuracy on exactly those fixtures.
Predictions graded on the 1X2 result. Club-competition fixtures only — World Cup and internationals run on a separate model and aren't counted here. Statistical analysis for information and education only.