We throw the term "Elo" around constantly. We play Guess the Elo, we grind our Elo on Chess.com, and we complain when we lose Elo to a lower-rated player. But have you ever actually wondered where the number comes from? Is an 1800-rated player simply "twice as good" as a 900-rated player? (Spoiler: no.)
Let's break down how the Elo rating system actually functions, and why predicting a player's rating is so difficult.
Who Invented Elo?
The system is named after its creator, Arpad Elo, a Hungarian-American physics professor and chess master. Before his system was adopted in the 1960s, chess rankings were chaotic and largely based on subjective tournament results. Elo wanted a mathematical way to predict the outcome of a match based on past performance.
The core concept of the Elo system is that performance is a probability curve. A player isn't always playing at their exact rating level; some days they play like a genius, and other days they hang their queen on move six. Elo calculates the expected outcome of a match between two players based on the difference in their ratings.
How Points are Exchanged
Elo is a zero-sum system. If you win points, your opponent loses the exact same amount of points. The number of points exchanged depends entirely on the rating difference.
If you (rated 1500) play another 1500, the system expects a 50/50 outcome. If you win, you might gain 8 points, and they lose 8 points. However, if you (rated 1500) play a Grandmaster (rated 2500), the system expects you to lose 99.9% of the time. If the Grandmaster wins, they might only gain 1 point, because the system already knew they were going to win. But if you somehow beat the GM, you might gain 25 points in a single match, because you defied the mathematical expectation.
Why 1800s Sometimes Look Like 800s
This is the entire premise of Guess the Elo. If an 1800 is mathematically supposed to be an excellent player, why do we sometimes see them blunder mates in one?
Because Elo isn't a measure of your *maximum* potential; it's a measure of your *average* performance over time. A 1500-rated player might have the tactical vision of a 2000, but the endgame technique of a 1000. In one game, they might crush their opponent with a brilliant sacrifice, making them look like a master. In the next game, they might blunder in a drawn endgame, making them look like a beginner. When you're guessing an Elo based on a single game, you're only seeing one data point on their probability curve.
The Online Inflation (and Deflation) Problem
It's also worth noting that an Elo of 1500 on Lichess is not the same as an Elo of 1500 on Chess.com, and neither is the same as a FIDE over-the-board rating. Different platforms start new players at different default ratings (often 1200 or 1500), which shifts the entire mathematical bell curve.
So the next time you guess 600 and the player is actually 1400, don't feel too bad. You just caught them on a very bad day.