How chess ratings work: Elo, Glicko and what your number means

Play a few rated games online and you get a number: 800, 1200, 1650. Win and it goes up, lose and it goes down. But where does it come from, why did you gain 7 points for one win and 15 for another, and is your 1200 the same as your friend’s 1200 on a different site?

Here’s how chess ratings actually work, starting with the system that started it all.

What a rating measures

A chess rating is a prediction tool. The difference between two players’ ratings predicts how the game between them is likely to go. A big gap means the higher-rated player is expected to score most of the points. A small gap means it’s close to a coin flip.

After each game, the system compares what was expected with what actually happened, and adjusts both players’ ratings. Beat someone you were expected to beat, and you gain a little. Beat someone much stronger, and you gain a lot.

That’s the whole idea. Everything else is details.

The Elo system

The Elo rating system is named after Arpad Elo, a Hungarian-American physics professor and keen chess player. The United States Chess Federation adopted his system in 1960, and FIDE, the international chess federation, adopted it in 1970. It’s still the basis of official over-the-board ratings today.

Notice it’s written Elo, not “ELO”. It’s a person’s name, not an acronym.

The expected score formula

Elo’s key formula gives your expected score against an opponent, where a win counts as 1, a draw as ½ and a loss as 0:

E = 1 / (1 + 10^{(R_opp − R_you) / 400})

You don’t need to memorize it. What matters is what it says: the expected score depends only on the rating difference, and every 400 points of difference multiplies the odds by 10.

So a player rated 200 points higher is expected to score about 76% against you over many games. Not 100%. Upsets are built into the system.

Updating the rating

After a game, your new rating is:

New rating = old rating + K × (actual score − expected score)

K is the “K-factor”: how fast ratings move. A bigger K means bigger swings. Federations use a larger K for new players, so their rating finds its level quickly, and a smaller K for established players, so their rating stays stable.

A worked example

Say you’re rated 1500 and you play someone rated 1700, with K = 20.

1. Your expected score: E = 1 / (1 + 10^200/400) = 1 / (1 + 3.16) ≈ 0.24.
2. If you win (score 1): 20 × (1 − 0.24) ≈ +15 points.
3. If you draw (score ½): 20 × (0.5 − 0.24) ≈ +5 points.
4. If you lose (score 0): 20 × (0 − 0.24) ≈ −5 points.

Your opponent’s rating moves by the same amounts in the opposite direction. That’s why losing to a much stronger player costs very little, while beating them is a big boost.

Glicko and Glicko-2: Elo with a confidence meter

Elo has one weakness: it treats every rating as equally reliable. But if someone has played 3 games, we know much less about their strength than if they’ve played 3,000. And if someone hasn’t played for a year, their old rating might be out of date.

The Glicko system, created by statistician Mark Glickman, fixes this by giving every player two numbers:

• Rating: the best estimate of your strength, on the same kind of scale as Elo.
• Rating deviation (RD): how uncertain that estimate is.

When your RD is high, your rating moves fast, because the system is still figuring you out. As you play more, your RD shrinks and your rating settles down. If you stop playing for a while, your RD slowly grows again.

RD also affects your opponents. Beating someone whose rating is very uncertain tells the system less than beating someone with a well-established rating.

Glicko-2, a later version, adds a third number called volatility, which tracks how consistent your results are. Many online platforms use Glicko or Glicko-2 rather than classic Elo. Lichess, for example, uses Glicko-2.

Provisional ratings

When you first join a site, your rating is provisional: the system doesn’t know you yet, so your RD is high and your rating can jump by large amounts after each game. Some sites mark provisional ratings with a question mark. After enough games, it settles and starts moving in smaller steps. So if your rating swung wildly at first, that’s normal.

Why your rating is different on different sites

Your rating on one site isn’t directly comparable to your rating on another, or to an official federation rating. A few reasons:

• Different starting points. Sites start new players at different numbers.
• Different player pools. A rating only means something relative to the people you play. Two separate groups of players can drift apart over time.
• Different systems and settings. Elo vs Glicko, different K-factors, different rating floors.
• Different time controls. Most sites keep separate ratings for bullet, blitz, rapid and classical, because people perform differently at different speeds.

So the only fair comparison is within the same pool: your blitz rating on one site vs other players’ blitz ratings on the same site.

What the numbers mean

Rating scales vary, but on the official FIDE scale there are some well-known landmarks. FIDE titles come with rating requirements:

• Candidate Master (CM): 2200
• FIDE Master (FM): 2300
• International Master (IM): 2400, plus norm results in strong events
• Grandmaster (GM): 2500, plus norm results in strong events

The highest FIDE rating ever reached is 2882, by Magnus Carlsen in 2014.

For most of us, those numbers are far away, and that’s fine. A rating is most useful as a measure of your own progress: are you better than you were three months ago?

Puzzle ratings work the same way

On Chessicous, every puzzle has its own rating, and so do you. Each attempt counts like a game against the puzzle: solve it and you “win”, miss it and you “lose”. We use a Glicko-style update, so your puzzle rating moves quickly at first and settles down as you solve more. Try one now:

Then head to the puzzle trainer to start your own rated streak, or work through themes like skewers and forks. The bots on the play page also show an approximate rating, so you can pick an opponent that fits.

How to raise your rating

A rating goes up when you win more games than expected. The fastest way to do that isn’t learning more openings. It’s making fewer mistakes:

• Stop hanging pieces. Before every move, check what your opponent’s last move attacks, and whether your move leaves anything undefended. Our list of common beginner mistakes covers the usual suspects.
• Do tactics regularly. Most games below master level are decided by tactics. A few puzzles a day add up.
• Review your losses. Run your games through game review and look for the one move where things went wrong.
• Play longer games. You learn more from a 15-minute game where you have time to think than from ten 1-minute games.
• Be patient. Ratings go up in steps and plateaus, not straight lines. A losing streak is normal.

If you want a structured routine, try our 30-day improvement plan.