A chess rating is a single number that tries to answer one question: how many points would you expect to score against another rated player? Everything else about ratings follows from that question. Once you understand it, rating swings, differences between sites and "rating anxiety" all make much more sense.
Where Elo comes from
The system is named after Arpad Elo, a Hungarian-American physics professor and chess player. His method was adopted by the United States Chess Federation in 1960 and by FIDE, the international chess federation, in 1970. Most modern systems, including the Glicko family used by many online sites, are refinements of his idea rather than replacements.
The core idea: expected score
Elo assumes that each player's strength varies a little from game to game, and that the difference between two ratings predicts the result. The rating itself means nothing in isolation. A 1500 only means something compared with other players in the same pool.
The expected score for player A against player B is:
E = 1 / (1 + 10^((Rb - Ra) / 400))
You do not need to calculate this yourself, but the pattern is worth knowing:
- Equal ratings: each player is expected to score 0.5 (a win counts 1, a draw 0.5).
- 100 points stronger: expected score about 0.64.
- 200 points stronger: about 0.76.
- 400 points stronger: about 0.91.
So a 200-point gap does not mean the stronger player always wins. It means that over many games they should score about three points out of four. Upsets are built into the system.
How a rating changes after a game
After each game, your rating moves by:
change = K x (actual score - expected score)
K is a factor that controls how fast ratings move. A large K means big swings; a small K means a stable rating. FIDE, for example, uses a higher K for new players and juniors so their ratings can catch up with fast improvement, and a lower K for established players above a certain level.
A worked example with K = 20:
- You are rated 1600 and play someone rated 1800. Your expected score is about 0.24.
- If you win, you score 1. Change = 20 x (1 - 0.24) = about +15.
- If you draw, change = 20 x (0.5 - 0.24) = about +5.
- If you lose, change = 20 x (0 - 0.24) = about -5.
Your opponent's rating moves by the same amount in the opposite direction. That is why beating a stronger player gains a lot and losing to them costs little: the system was already expecting you to lose.
White: king on b8, pawn on b7, rook on c1. Black: king on d8, rook on a2. White to move.
Rating pools: why your numbers differ
A rating only compares you with the people you have played and the people they have played. That group is the rating pool. Different pools have different numbers, even for the same player.
Reasons your online rating, your federation rating and your friend's rating on another site do not match:
- Different starting points. One site may start new players at 1500, another lets you choose a starting level, a federation may calculate your first rating from your first set of games. The starting number shifts the whole pool over time.
- Different formulas. Glicko and Glicko-2 add a "rating deviation" that measures how certain the system is about you. New or inactive players move quickly; regular players move slowly. Classic Elo has no such term.
- Different players. An online pool contains many casual players who play thousands of fast games. A federation pool contains tournament players who travel to play slow games. The same skill lands at a different number in each.
- Different time controls. Bullet, blitz, rapid and classical are normally rated separately, because they reward different skills. Plenty of players are 300 points apart between their blitz and classical ratings.
The practical rule: only compare ratings within the same pool and the same time control. Saying "I am 1700 online, so I should be 1700 FIDE" is like converting a temperature without knowing whether it was Celsius or Fahrenheit.
What a rating change actually means
Ratings are noisy. A run of good or bad results moves your number even if your strength has not changed at all.
Some guidelines:
- Swings of 50 to 100 points in fast online chess are normal. A bad evening of blitz says more about tiredness than skill.
- Look at the trend, not the day. Your rating over the last two or three months is a far better guide than this week.
- Real improvement shows slowly. If you study well, expect your rating to climb in steps with flat periods between them. Flat periods are when new knowledge settles in.
- A peak is not your "true" rating. Everyone's peak includes some luck. Your average over a long period is closer to your real strength.
White: pawn on d5, king on e5, rook on h1. Black: king on e8, rook on a6. Black to move.
Common rating myths
- "Draws against weaker players are bad." Only slightly. A draw against someone 100 points below costs you a few points, not a disaster.
- "Avoiding games protects my rating." It does, but it also stops you improving. A rating you are afraid to use is not doing anything for you.
- "The system is rigged against me." Elo-type systems are zero-sum within a pool. If you are truly stronger than your rating, you will score above expectation and rise.
- "Ratings measure talent." They measure results against a pool, nothing more. Experience, opening knowledge, time management and nerves all go into the number.
How to use your rating
A rating is most useful as a feedback signal over months. Use it to check whether your training works, not to judge each game.
Better questions after a game are "Did I play well?" and "What did I learn?" A free game review answers both far more precisely than a +6 or -8 ever can. If you want rated games with a purpose, play at a time control you plan to improve in, join tournaments for serious games, or find a club where the same opponents become a natural benchmark for your progress.
The number will follow the chess. It rarely works the other way round.
Published 2026-11-01 · 22Softwares