← All accuracy

Over/Under 2.5 accuracy

How often what we forecast on over/under 2.5 matches what actually happened, published unedited — the same numbers the overview shows, just this market alone.

In one sentence

Across 28,436 held-out matches, we are well calibrated on over/under 2.5, an average gap of -0.0pp between what we said and what happened. The chart below shows each probability band against how often it actually came in.

28,436 matches · model 20260904-0011

What we said, against what happened

What we said against what happened, for Over 2.5 goals0%100%100%0%We said 28%; it happened 43% of the time, across 42 matches.We said 37%; it happened 37% of the time, across 1137 matches.We said 46%; it happened 47% of the time, across 11414 matches.We said 55%; it happened 54% of the time, across 12290 matches.We said 63%; it happened 64% of the time, across 3317 matches.We said 72%; it happened 78% of the time, across 232 matches.We said 84%; it happened 100% of the time, across 4 matches.

Each dot is a tenth of the probability range: what we said, against how often it happened. On the dashed line we were exactly right; above it we were too cautious, below it too confident. The rule through each dot is one standard error, and the dot's area is how many matches it rests on — the sparse ones at the ends will always scatter.

Furthest out: when we said 84%, it happened 100% of the time, across 4 matches.

All 5 gate values for this market, unedited
GateValueResult
ece over under 2 50.0076pass
max reliability gap pp over under 2 5not evaluated
max reliability gap pp over under 2 5 measurable5.3342fail
reliability within sampling envelope over under 2 55.3342fail
brier over under 2 50.2447pass

Common questions

How accurate are your over/under predictions?

Our over 2.5 and under 2.5 forecasts are scored the same way as the match result: we publish what we said against what actually happened, and the summary states the verdict in plain language. The two-way markets are judged with Brier score rather than ranked probability score, because that is the honest metric for a two-outcome call.