FOOTAnswers › Why do football predictions go wrong?
Question
Why do football predictions go wrong?
Short answer: Because football is a low-scoring, high-variance sport: one red card, deflection or early goal flips even a well-calibrated 70% favourite.
Variance is the point, not the bug
A 70% prediction fails 3 times in 10 by definition. Over a season of top-league matches there are hundreds of goal-scale random events; the best model in the world still lands roughly 50-55% of its 1X2 picks. Wrong-per-match is the expected state.
What specifically flips matches
Early goals (the leading team then defends), red cards (worth several tenths of expected goals), rotation before cup games, goalkeeper errors and VAR calls. Most are invisible before kick-off — which is why honest models publish probabilities, not promises.
What going wrong should look like
Graded misses in public, at roughly the advertised rate. If a service never shows losses, or wins far more often than its own numbers imply, the numbers are decoration. Ours are graded on every official pick.
FAQ
How often is a 90% favourite wrong?
About one time in ten — over a 38-game season that is several matches.
Do wrong predictions mean the model is bad?
Only if misses cluster far above the stated probabilities over many matches. That is what Brier scores measure.
What is the biggest single predictor of upsets?
No single signal exists — by definition an upset is what the model did not see coming.