Every number on this site comes from one small, transparent model. There is no insider information, no betting market behind it and no hand-tuning of individual games. This article explains exactly how it works, so you can judge for yourself how much weight any figure deserves.
1. A single power rating per team
Each of the 48 teams carries one number — its power rating. It starts from a pre-tournament estimate built on recent international results and squad strength, scaled so that the gaps between teams behave like chess Elo: a 100-point edge is a meaningful favourite, a 300-point edge is a near-certain winner on paper.
The rating is deliberately simple. It does not read injuries, suspensions, line-ups, weather, travel or motivation. It only knows who has played whom, and how those games finished.
2. Turning two ratings into a match
For any fixture we take the two ratings and convert the gap into an expected-goals figure for each side. Those expected goals feed a Poisson model — the standard statistical way to turn an average scoring rate into the probability of each exact scoreline (0–0, 1–0, 2–1 and so on).
From that grid of scoreline probabilities we read off three things you see on every match page: the most likely reference scoreline, the split between a home win, draw and away win, and each team's expected goals. The reference scoreline is the single most probable result — not a prediction that it will happen, just the centre of a wide range.
3. Learning after every game
When a real result comes in, both teams' ratings are nudged by an Elo update: beat a stronger side and you gain points, lose to a weaker one and you drop them, with the size of the move depending on how surprising the result was. Every projection further down the bracket is then recomputed from the new ratings. This is why the standings, scoreline references and knockout path on this site shift as the tournament unfolds.
4. How to read it — and how not to
The model is a clean baseline: a way to ask "what does recent form alone say about this matchup?" It is useful precisely because it is dumb to context — it never talks itself into a narrative. But that is also its limit. A late injury to a key player, a dead-rubber final group game, or a tactical mismatch can move a real match well away from the model's centre.
Treat the figures here as a reference point for discussion, not a forecast and never as betting advice. The most interesting matches are usually the ones where you think the model is wrong — and now you know exactly what it is and is not looking at.