Model
What is expected goals — and how our model uses it
A plain-language introduction to modelled goal expectation, why it beats raw shot counts, and how it feeds match probabilities.
Why raw scorelines mislead
A 1–0 win can hide fifteen clear chances or one lucky deflection. Modelled goal expectation replaces the final score with the quality of chances a team creates, which is far more stable from match to match — and therefore more useful for prediction.
Key idea: chance quality predicts future results better than past results do.
From chances to probabilities
The model converts each team's expected goals into a full score distribution, then sums the scorelines belonging to each outcome. That is why the predicted score and the 1/X/2 probabilities always agree.
| Model input | Example | Weight |
|---|---|---|
| Finished-match results | last 2 seasons | high |
| Home advantage | per league | medium |
| Recent form | last 5 matches | medium |
What the model ignores
Injuries, lineups, motivation and odds are deliberately excluded: they are either unreliable at scale or would make the model circular. That is also why expert commentary exists as a separate, clearly-labelled layer.
FAQ
Is expected goals the same as the predicted score?
No. Expected goals are continuous model values; the predicted score is the most probable whole-number scoreline derived from them.
Why don't you publish odds-based probabilities?
Model probabilities are calculated independently and do not use bookmaker odds — that keeps the public record honest.
Sources: public match data 2024–2026 · Matchform model documentation v1.5