Elo measures performance against expectation
An Elo rating is not a tally of wins. It is an estimate of playing strength, continually adjusted by comparing what happened with what the rating system predicted. Before a match, the rating gap implies an expected score: a higher-rated player is expected to score more, while a lower-rated player is expected to score less. A draw counts as a partial score.
Afterward, the system compares the actual result with that expectation. In simplified form, the change is a sensitivity factor multiplied by the difference between actual and expected score. The sensitivity factor controls how quickly a rating can move; the comparison determines the direction and relative size of the change.
Why the same win can mean different things
Suppose a favored player wins. That result is close to the system’s forecast, so it provides little new evidence about the player’s strength. The rating may still rise, but only slightly. If the underdog wins, the result is far above expectation, and the adjustment is larger. A draw can also help the underdog’s rating if the system expected a decisive loss.
This makes Elo a kind of running forecast: each result nudges the estimate toward the performance it reveals. One upset can move a rating substantially, but it does not prove that the winner has permanently become stronger. The system is responding to evidence from a match, not awarding a lasting bonus for drama.
The practical consequence is that a player’s rating reflects not just how often they win, but whom they beat—and how surprising those results were.
