Team Ratings Match Predictor
Every club has an attack rating and a defence rating, learned from four years of results. Pick a home side and an away side to see what the ratings expect.
- Home win
- Draw
- Away win
Goals
Likeliest scores
The ratings
| Club | Attack | Defence |
|---|
How to read it
- Attack 1.00 is a typical Premiership attack. Celtic's 1.70 means they score about 70% more than a typical side would against the same opponents.
- Defence works the other way: lower is better. A defence of 0.80 lets in 20% fewer goals than typical; 1.40 lets in 40% more.
- Expected goals for the home side = its attack × the away side's defence × home advantage. The away side's is its attack × the home side's defence.
- Expected goals become chances by assuming goals arrive at a steady rate, as in the Poisson Match Predictor, with a small correction for 0–0, 1–0, 0–1 and 1–1.
- The table is sorted by attack minus defence, so the strongest all-round sides come first.
How it was built
The ratings come from every Premiership match in the four years to the end of 2025/26, recent matches counting more: a match counts half as much after 231 days. The fit takes turns finding the attacks that best explain the goals scored against these defences, then the defences that best explain the goals conceded against these attacks, until they settle. Refitted before every matchday of the five test seasons, this method scored a log loss of 0.945, the first model in the series to beat logistic regression.
Where it goes wrong
- The ratings stand still. They're frozen at the end of 2025/26 and refreshed each summer. Signings, sales, new managers and form since then aren't in them.
- Promoted sides are rated from their last Premiership seasons, or not at all if they have none in the last four years.
- No team news. Injuries, suspensions and a rotated side before a European tie are invisible to it; the bookmakers know.
How it was built and tested: Every team its own attack and defence. Dixon-Coles from scratch
Results data from football-data.co.uk. These figures come from a statistical model and are for analysis and education. Football remains uncertain and model predictions will frequently be wrong.