Forecasts
Some fixtures carry a small bar with three numbers: how likely the home side is to win, a draw, and an away win. This is where those come from, and how often they have been right.
The record
- Games forecast
- 22,903
- Result called right
- 67% vs 47% by guessing the usual
- Brier score
- 0.47 vs 0.60 (lower is better)
Scored on every one of the 42,632 decided games here where both sides already had 3 or more results — the same rule that decides whether a fixture shows a forecast at all. “Guessing the usual” always picks the most common outcome (46% home wins, 13% draws, 41% away). Ratings last rebuilt Nov 8.
How it works
Every team carries a rating that starts equal and moves with each result: beat a stronger side and it rises a lot, beat a weaker one and it barely moves, and the margin counts — a 5–0 says more than a 1–0, though not five times as much. Games are replayed in the order they were played, so a rating is always what the results up to that day support (K = 60).
Two ratings give the chance of each outcome. Level sides draw the most; a mismatch almost never does. Nothing about age or gender is modelled, because teams only ever meet their own group and the ratings separate on their own. A forecast is shown only when both sides have 3 or more results here — before that, a number would look like knowledge it is not.
It learns only from scores on this site, so it knows nothing about a team’s games elsewhere, injuries, or who is on the pitch that Saturday. Take it as “probably”, never as “certainly”. Ratings are rebuilt every night from all results; the events are where the results come from.