A rating updated by the gap between actual and expected results, so a win against a stronger opponent is worth more than one against a weaker.
Devised by Elo for chess and adopted by the US federation around 1960 and internationally in 1970, the system converts a rating difference into an expected score and moves both players by the surprise. It has since been applied to football, esports and matchmaking, and its logic is Bayesian in effect if not in derivation.
In practice
- A player who beats a much stronger opponent gains far more than for a routine win, and the system needs no judgement to decide that.
- A matchmaking system that pairs equal ratings produces close games, which is the property it was not designed for.
- Further reading
- Wikipedia
- Subject
- Sport & Performance
- Essay
- The best-measured contests there are