How the predictions work

Full-time gives win, draw and loss chances for football matches. This page says plainly how they are made, how well they have done, and where they are weak. Figures are from tests run in November 2026.

What this is, and is not

The percentages are statistical estimates, for information and entertainment. They are not tips, not betting advice and not a guarantee of anything. This site does not take bets, has no accounts and shows no adverts. 18+ only.

How a prediction is made

Each club gets an attacking and a defensive rating from the goals it has scored and conceded over roughly the last two years (in the domestic leagues, recent matches count more). For a match, the two ratings give the expected goals for each side. From those, the model works out the chance of every scoreline, and adds them up into home win, draw and away win. Nothing here is hand-picked or adjusted by opinion.

A 60% chance does not mean the team will win. It means that in about 6 of 10 matches rated like this one, that team wins.

How accurate it has been

To test it fairly, every past match was predicted using only the matches played before it. The same test was then repeated on matches played this season, which nothing in the model was tuned on.

2025/26, 16 leagues49%top pick right, 5,392 matches

2026/27 so far50%top pick right, about 1,400 matches

Always pick the home team43%for comparison

Model gave 60% or more74 to 79%of those picks were right

The model is clearly better than guessing each league’s usual home, draw and away rates. It rarely picks a draw, because a draw is almost never the single most likely result, even though about one match in four ends level.

Compared with the bookmakers

On 6,545 matches from the 16 domestic leagues (2025/26 and 2026/27 so far), the bookmakers’ average odds were a little better than this model: their top pick was right about 52% of the time, against about 49.5% for this model. The gap was similar in the five biggest leagues and in the smaller ones, and a little wider in Greece, the Süper Lig and the Championship. A mix of the two did not beat the bookmakers either. So this site does not claim to beat the bookmakers, and nobody should place bets because of it. Where bookmaker odds are available, the match page shows them next to the model so you can see where they disagree.

What it does not do well

Goals markets. Over / under goals and both teams to score are barely better than guessing the average rate. Treat them as weak signals.
Corners and cards are experimental and not yet proven.
Handicap lines (Asian handicap, draw no bet) are read off the same scoreline chances, so they add no new information about the teams.
New or unrated clubs (for example just promoted, or from small leagues in Europe) show as “Estimate” and are less reliable.
Team news. Line-ups are shown but not used by the model. Injury adjustments exist for the Premier League only and are not yet proven to help.
Football is unpredictable. Even the bookmakers get roughly half of their top picks wrong.

See for yourself

Every prediction the server saves before kickoff is listed on the public record page with its final result, including the misses. Nothing is changed after the match.