Quiet Move Journal · Vol. 013 · October 16, 2026
Ask a chess player how good they are and you'll probably get a number. 1200. 1850. 2400. It's such a natural part of the game now that it's easy to forget the number has an author, and that he taught physics in Milwaukee.
A broken system
In the 1950s, American chess ran on a rating system devised by Kenneth Harkness, introduced in his Official Blue Book and Encyclopedia of Chess in 1956. It rated players by the strength of their competition at each tournament, and that produced strange results. It was possible to gain points while losing every game in a tournament, and possible to lose points after winning every game.
In 1959 the United States Chess Federation asked one of its own to fix it: a physics teacher named Arpad Elo.
The physicist at the board
Elo was born Élő Árpád Imre in Hungary in 1903 and came to the United States with his parents in 1913. He studied physics at the University of Chicago and taught it at Marquette University in Milwaukee from 1926 until he retired in 1969.
He was also a strong player. By the 1930s he was the best in Milwaukee, and he won the Wisconsin state championship again and again across three decades. In 1935 he was elected president of the American Chess Federation, one of the two groups that merged into the USCF in 1939.
How it works
Elo's idea was simple to state. A player's performance varies from game to game, but around a true average. The gap between two players' ratings predicts how a game between them should go. A player rated 100 points higher is expected to score about 64 percent. At 200 points, about 76 percent. After each game, your rating moves by how much better or worse you did than expected. Beat someone you were supposed to beat and you gain a little. Beat someone far stronger and you gain a lot.
He made one practical choice that mattered as much as the math. He kept the numbers close to the ones players already knew, so an average player stayed around 1500, a strong club player around 2000, and a grandmaster around 2500. Nobody had to relearn what a number meant.
The USCF approved the system at its 1960 meeting in St. Louis. FIDE adopted it in 1970, and its first official list came out in July 1971. Bobby Fischer was on top at 2760, the only player in the world above 2700. Boris Spassky, the world champion, was second at 2690. A year later they met in Reykjavik.
Pencil, paper, calculator
The system was designed to be worked out with pencil and paper. When pocket calculators arrived, Elo compiled the official FIDE ratings on a Hewlett-Packard, and he kept running the numbers himself until the mid-1980s.
He was careful not to oversell them. "Ratings provide merely a comparison of performances, no more and no less," he wrote in Chess Life in 1962. Measuring a player's strength, he said, was like measuring the position of a cork "bobbing up and down on the surface of agitated water with a yard stick tied to a rope and which is swaying in the wind."
In 1978 he published The Rating of Chessplayers, Past and Present, which applied his method backward to the great players of history. Using five-year peak averages, he put Capablanca on top at 2725, just ahead of Lasker and Botvinnik.
One player, several numbers
A player who is strong at a slow game isn't always strong at a fast one, so ratings split by time control. FIDE keeps three: standard (what most players call classical), rapid and blitz. Under FIDE's rules a rapid game gives each player more than 10 minutes and less than 60, and a blitz game gives 10 minutes or less. The July 2012 list was the first to carry separate rapid and blitz ratings. Before that, only slow games counted.
Bullet, under three minutes a side, belongs to the internet. FIDE doesn't rate it, but the big sites do. Lichess sorts every game into bullet, blitz, rapid or classical by its clock and keeps a rating for each. Like Chess.com, it uses the Glicko system, which the statistician Mark Glickman invented in 1995 as an improvement on Elo's.
So the same player can be 1900 at rapid and 1600 at bullet, and both numbers are honest. One measures calculation. The other measures hands and nerve.
A rating for one tournament
A rating sums up years of games. A performance rating sums up one event. It answers a simple question: what rating would have produced exactly this result against these opponents? Play at your performance rating and your real rating doesn't move.
FIDE works it out from two things: the average rating of your opponents, and your percentage score. Score 50 percent and your performance equals the field's average. Score more and it climbs above it. Players at the board use a shortcut: count each win as the opponent's rating plus 400, each loss as minus 400, each draw as the opponent's rating, and average the lot.
A player who scores 2½ out of 3 against opponents rated 2400, 2500 and 2600 has a FIDE performance of 2773, far above any of them. That number matters, because performance is how titles are earned.
What the numbers unlock
Every FIDE title has a rating floor. The lower titles come from the rating alone, once a player has at least 30 rated games: Candidate Master at 2200 and FIDE Master at 2300, or Woman Candidate Master at 2000 and Woman FIDE Master at 2100.
The top titles need more than a number. An International Master must reach 2400 at some point, and a Grandmaster 2500 (2200 for Woman International Master, 2300 for Woman Grandmaster). They also need norms: tournament performances at title level against a strong, international field. A grandmaster norm means performing at 2600 or better against opponents averaging at least 2380, over at least nine games, and a player needs norms adding up to at least 27 games.
That link between ratings and titles turned Elo's statistics into something players fought over. At the 1974 Olympiad in Nice, the writer William Hartston watched Elo field a line of players asking for the rules to be bent in their favor. When the last of them had gone, Hartston later recalled, Elo turned to him and said, "I think I have created a monster."
Beyond the board
The monster didn't stay in chess. Elo built his method for two players and a board, but it works for anything where two sides meet and one wins. FIFA has used it for its women's world ranking and, since the 2018 World Cup, for the men's too. Tennis has an Elo-based rating, the European Go Federation adopted one, and online games from League of Legends to World of Warcraft have used it to decide who plays whom.
Plenty of people who meet his formula have never pushed a pawn. They just notice that their next opponent seems about as good as they are. That is the quiet genius of it. A fair game is one where you don't know who will win.
He died in 1992 at his home in Brookfield, Wisconsin. Somewhere tonight, a club player will lose eight points to a teenager and blame the teenager. They should blame the professor.