Elo and power ratings, explained: how models measure team strength

An Elo rating boils a team's strength down to a single number that rises when it wins and falls when it loses. Born in chess and now used across sports, it treats every match as an exchange: beat a stronger side and you gain a lot of points; lose to a weaker one and you shed them. Over time the number settles at a level that reflects how good a team actually is, independent of the fixture list.

Before a match, the gap between two teams' ratings implies an expected result. If a favourite wins as expected, only a few points change hands; if an underdog pulls off an upset, a big chunk moves, because the result was surprising. The size of each swing is tuned by a factor that controls how reactive the system is — high enough to track real change, low enough to ignore one fluky afternoon.

A league table rewards accumulated points but is blind to who you played. Sitting third having faced the easiest run is not the same as sitting third after the hardest, yet the table cannot tell them apart. Elo can, because it weights every result by the opponent's strength. That makes it a cleaner measure of current quality — and it updates instantly after each game rather than waiting for the season to average out.

Modern versions go beyond win-or-lose. Many add a margin-of-victory adjustment so a 4-0 counts for more than a 1-0, and build in home advantage by giving the host a rating boost before kick-off. Some blend in expected goals so a team is rated on chances created rather than just results. These refinements make power ratings sharper at predicting the next match, which is their real job.

Elo-style power ratings sit under FIFA's world ranking, countless club rankings and most serious prediction models — including the team-strength input behind data-driven previews like those on SportMetrics. When you read that one side is a clear favourite, a rating like this is usually doing the quiet work of turning strength into a probability.