Elo rating system
The weak are meat, and the strong do eat.
— Old Georgie, Cloud Atlas, 2012
The Elo rating system is a formula by Arpad Elo used to rank chess players.
In a given pool of players, it assigns each one a number, such that someone rated 200 points above someone else should beat them roughly three quarters of the time. It's essentially like an automated, objective bookie, calculating the odds of who should win each game.
Automatically working out how good you are at an overall game such as chess is useful, as it allows you to be paired with opponents at roughly the same level. As your opponents should be neither too challenging nor too easy for you, the individual games you play should be neither frustrating nor boring.
The system's zero-sum. After each game, the victor's ranking increases the same amount as their opponent's decreases. In effect, they take the points from their opponent. Therefore, the arbitrary identical amount of points that each and every new player starts with[1] (1500 in most implementations, for legacy reasons, as this was roughly USCF players' average rating before Elo's rating system was implemented) is inherently the average (arithmetic mean) ranking.
If one lone player is especially strong, then as their ranking increases, all the other players' rankings will lower proportionally as their points are given to this lone warrior; conversely, if one lone player is especially weak, then as their ranking decreases, all the other players' rankings will proportionally rise as they (collectively, often only indirectly) take and pass on points from this easy prey.
It makes sense, then, that a player's ranking only applies within that single pool of players. After all, the people playing in one isolated group (say, on a particular server online) might on average be much stronger or weaker players than in another group, yet their average ranking will still be that arbitrary initial number. As the main pool is the entirety of FIDE, this is scarcely a problem.
Elo's since been superseded by the more complex and accurate rival algorithms Glicko and Glicko-2, although it still has its use, as its simplicity makes it easy to verify your rating was correctly calculated.
Such formulae have since been applied to other games and sports, though it's perhaps not surprising that the kind of person who comes up with formulae might themselves prefer a nice game of chess. They've even been used by dating apps, which is curious, as who's a good match for you isn't necessarily a good match for other people who are just as agreeable.
If nothing else, rating systems such as Elo and Glicko have let me appreciate the dynamic range of players' abilities. If you can beat someone 200 points below you about three quarters of the time, while someone 200 points above you can beat you about three quarters of the time, and so on, with over ten people in the chain between the weakest and strongest players out there... you may not be able to appreciate what exactly the best players are doing that's so impressive, but you can rest assured that it genuinely is — Elo's done the maths.
References
- "An identical initial rating is assigned each member of the group." "The Proposed USCF Rating System: Its Development, Theory, and Applications" Arpad Elo, Chess Life, Aug 1967, pp. 242—247
Chess: Chessmen | Elo rating system | Style of chessmen
