A rating is a prediction.
The Elo system, designed by the physicist and chess master Arpad Elo and adopted by FIDE in 1970, doesn't try to measure how good you are in the abstract. It predicts how you will score against a particular opponent — and corrects itself every time the prediction is wrong.
Beat someone the system expected you to beat and you gain a little. Beat someone it expected you to lose to and you gain a lot. Over many games, your rating settles where its predictions come true.
A 400-point gap means a 91% expected score.
Ratings aren't a grade. They're a prediction — and they move by how wrong it was.Two lines of maths.
First, the expected score — the share of points you are predicted to take, from 0 to 1 — depends only on the rating difference:
E = 1 / (1 + 10(opponent − you) / 400)
Then, after the game, your rating moves by the difference between what happened (1 for a win, ½ for a draw, 0 for a loss) and what was expected, multiplied by the K-factor:
new rating = old rating + K × (result − E)
Move the sliders. Watch how the stakes change as the gap widens.
Over the boardElo
- ✕Same certainty after 5 games or 5,000
- ✕One number per player
- ✕Used for FIDE ratings
OnlineGlicko
- ✓Rating deviation tracks uncertainty
- ✓New players move fast, then settle
- ✓chess.com: Glicko · Lichess: Glicko-2
K: how fast your number moves.
The K-factor is the speed dial. A high K reacts quickly to new results — right for players whose strength is still changing. A low K keeps established ratings steady. FIDE uses three:
| K | Who it applies to (FIDE) | Win vs equal opponent |
|---|---|---|
| 40 | New players, until 30 rated games; and players under 18 rated below 2300 | +20 |
| 20 | Everyone else rated below 2400 | +10 |
| 10 | Players who have reached 2400 | +5 |
Online, it's Glicko.
Elo has one blind spot: it is as confident about a player with five games as one with five thousand. Mark Glickman's Glicko system fixes that by giving every rating a rating deviation — a measure of uncertainty.
- A new or long-inactive account has a high deviation, so its rating moves fast to find its level.
- An active, established account has a low deviation, so each game moves it only a little.
- chess.com uses Glicko; Lichess uses Glicko-2, which also tracks how consistent your results are.
That is why your first dozen games on a new site produce big swings, and why the swings calm down once you've played a while.
What your number means — and doesn't.
A rating only means something inside its own pool. chess.com, Lichess and FIDE each have different players, starting points and settings, so the same player has three different numbers. As a rule of thumb, Lichess ratings run higher than chess.com's, and both differ from FIDE. Even different time controls on one site are separate ratings.
With that caveat, here is roughly what the ranges look like:
Roughly what each range looks like.
Under 800
Learning how the pieces work together. Games are decided by pieces left undefended.
800 – 1200
Knows the basic tactics; still drops material regularly. Opening principles start to matter.
1200 – 1600
Solid club player. Sees most one- and two-move tactics; plans start to appear.
1600 – 2000
Strong club player. Real openings, real plans; games lost to deeper tactics and technique.
2000 – 2200
Expert. Few outright blunders; strong positional understanding.
2200 and up
Candidate Master 2200, FIDE Master 2300, International Master 2400 and Grandmaster 2500 — the last two with norms.
Learn to read a rating from the moves.
Strong players can watch a few moves of a game and guess both ratings within a couple of hundred points. They're reading the signals: how quickly pieces develop, whether pieces are left hanging, whether the moves follow a plan or react to threats one at a time.
It is a surprisingly good training exercise. To judge a game's level you have to notice exactly the things that separate one level from the next — and once you notice them in others' games, you start noticing them in your own.
The fastest way to raise your rating is still to make fewer mistakes than your opponent. Start with the ten-second blunder check.
Questions, answered.
What is a good chess rating?
It depends on the pool, but as a rough guide on most sites: under 800 is a beginner, 1200–1600 is a solid club player, 1800+ is strong, and 2000+ is expert level. FIDE's titles start at 2200 for Candidate Master.
Why is my Lichess rating higher than my chess.com rating?
Each site is its own pool of players with its own starting point and settings, so the numbers aren't on the same scale. Lichess ratings generally run higher than chess.com's for the same player, and neither maps exactly onto FIDE.
How many points do I gain for a win?
It depends on the rating gap and the K-factor. Against an equal opponent with K = 20 a win is worth 10 points; beating someone 200 points higher is worth about 15; beating someone 200 lower, about 5.
What is the K-factor in chess?
The multiplier that sets how fast a rating moves. In FIDE ratings it is 40 for new players and juniors under 2300, 20 for most adults, and 10 once a player has reached 2400.
What is the difference between Elo and Glicko?
Glicko adds a rating deviation — a measure of how sure the system is about your rating. New or inactive players have a high deviation, so their ratings move quickly; established players move slowly. chess.com uses Glicko and Lichess uses Glicko-2.
