How our football predictions work

Every upcoming fixture on this site carries a predicted score, the chance of a home win, a draw and an away win, and a short written preview. This page explains where each of those comes from, what they cannot know, and how often they have been right.

The predicted score

The score comes from a statistical model, not from anyone's opinion. For each fixture we take both teams' last 10 league matches and work out how many goals each side scores and concedes compared with the league's average for home and away teams this season. A team with only a few matches is pulled towards the league average, so a single big win does not swing its numbers.

That gives an expected number of goals for each side. A Poisson distribution — the standard way to model goal counts — turns those into a probability for every scoreline from 0–0 up to 6–6. Adding them up gives the home win, draw and away win percentages shown under each prediction. The predicted score is the most likely scoreline within the most likely result, so the score and the result always agree.

The written preview

An AI language model writes the short preview from the same numbers: each side's recent results, goals, clean sheets, home or away record and top scorers. It is told to use only those facts — no injuries, no team news, no history — and to end on the model's predicted score.

What the model does not know

It sees results and goals only. It knows nothing about injuries, suspensions, line-ups, a new manager, transfers, fixture congestion or what a match means to either side, and cup matches are not in its data. Treat every prediction as a starting point, not a verdict.

Track record

Every prediction stays on its match page after the final whistle, marked right or wrong. A prediction counts as the right result when it called the home win, draw or away win correctly.