You might be 1650 FIDE, 1400 in Chess.com rapid and 1850 in Lichess rapid, all at once. None of these numbers is wrong. They come from different rating systems applied to different groups of players, so they measure different things.
This post explains the two systems behind them, Elo and Glicko, how they work, where they differ, and who uses which.
| Organization | System | Where you play |
|---|---|---|
| FIDE | Elo | Over-the-board, worldwide |
| Chess.com | Glicko | Online |
| Lichess.org | Glicko-2 | Online |
What both systems have in common
Every modern chess rating system rests on the same idea: a rating is a prediction. The rating difference between two players says how many points each is expected to score against the other. After the game, your rating moves according to how your real result compared with that prediction:
- Score more than expected → your rating goes up.
- Score less than expected → your rating goes down.
- Beating a much stronger player is worth a lot. Beating a much weaker one is worth very little.
The two systems agree on how to predict results. They disagree on how far to move your rating after each result.
The Elo system
The Elo system is named after Arpad Elo, a Hungarian-American physics professor and chess player. The US Chess Federation adopted it in 1960, and FIDE in 1970. More than fifty years later it is still the official system for over-the-board chess, used by FIDE and most national federations.
How it works
Each player has one number: their rating. For every game:
Rating change = K × (score − expected score)
- score is 1 for a win, 0.5 for a draw, 0 for a loss.
- expected score comes from the rating difference. Equal players expect 0.50 each; a player 200 points higher expects about 0.76.
- K is a fixed multiplier that decides how fast ratings move.
Take two players rated 1500 who play each other. Each expects 0.50. If one wins with K = 20:
Winner: 20 × (1 − 0.50) = +10
Loser: 20 × (0 − 0.50) = −10
What one player gains, the other loses (as long as both use the same K).
The K-factor
K is the only way Elo can tell whether a rating is “settled”. Because the system has no measure of uncertainty, federations set K by rules: age, number of games played, rating level.
At FIDE, K is 40 for new players and juniors, 20 for most adults and 10 for anyone who has ever reached 2400. We covered all the details, including the full tournament calculation, in How Your FIDE Rating Is Calculated.
FIDE
FIDE runs three separate Elo lists (standard, rapid and blitz), published monthly. Everything played in a month is calculated from the same starting rating, and the change is applied on the next list. A few FIDE-specific rules: the 1400 rating floor, the 400-point cap on rating differences (for players below 2650), and the minimum of 5 games against rated opponents to get a first rating.
Where Elo falls short
Elo is simple, transparent and easy to check by hand. That is a big part of why FIDE still uses it. But it has a blind spot: it treats every rating as equally reliable.
Consider two players, both rated 1800:
- Ana has played 400 rated games over ten years, and plays every month.
- Bruno got his rating from 5 games, three years ago, and hasn’t played since.
Elo sees two identical 1800s. If you beat either of them, you gain exactly the same. But we clearly know much more about Ana’s strength than Bruno’s. Bruno might have improved a lot, or stopped studying. His 1800 is a guess.
Elo patches this with rules, like higher K for new players, but the rules are blunt. They don’t capture that Bruno’s rating became less reliable while he was away, and they don’t change how much your rating should move when you play him.
The Glicko system
Mark Glickman, a statistician (and chess player), created the Glicko system in 1995 to fix that blind spot. His idea: store not only a rating, but also how sure we are about it.
Rating + Rating Deviation
In Glicko, each player has two numbers:
- Rating (r): the best estimate of your strength, as in Elo.
- Rating Deviation (RD): how uncertain that estimate is. A low RD means the system is confident; a high RD means it isn’t.
Together they form a confidence interval. Roughly, the system is 95% sure your true strength is within r ± 2 × RD:
| Rating | RD | 95% range | Meaning |
|---|---|---|---|
| 1500 | 350 | 800 to 2200 | Brand-new player: could be anything |
| 1500 | 150 | 1200 to 1800 | Some games played |
| 1500 | 50 | 1400 to 1600 | Well-established, active player |
How RD changes
- Playing games lowers your RD. Each result is new evidence about your strength, so the system becomes more confident.
- Not playing raises your RD. Your strength may have changed while you were away, so the system becomes less confident, up to a maximum.
This is the key difference from Elo. Glicko automatically does what Elo tries to do with K-factor rules: new and returning players move fast, established active players move slowly.
What RD does to your rating changes
RD affects your rating in two ways.
1. Your own RD decides how much you move. Your RD works like an automatic, personal K-factor. Here’s what happens when three players, all rated 1500, each beat the same well-established 1500 opponent (RD 50):
| Your RD | Rating change | New RD |
|---|---|---|
| 350 (brand new) | +175 | 248 |
| 150 (some games) | +54 | 138 |
| 50 (established) | +7 | 50 |
Same result, same opponent, but the new player gains 25 times as much. The system knew almost nothing about them, so one win is strong evidence. For the established player, one win barely changes a picture built on hundreds of games.
For comparison, Elo with K = 20 would give +10 to all three.
2. Your opponent’s RD decides how much the game counts. A result against a player with a high RD is weaker evidence, because we don’t really know how strong they were. If our established 1500 player (RD 50) beats a brand-new 1500 player (RD 350) instead, they gain only about +5, instead of +7.
This fixes the Ana-and-Bruno problem: in Glicko, Bruno’s RD would have grown during his three years away, so beating him would count for less than beating Ana, and Bruno’s own rating would adjust quickly once he returned.
A classic example
Glickman’s paper includes a worked example. A player rated 1500 with RD 200 plays three games in one rating period:
| Opponent | Opp. RD | Result |
|---|---|---|
| 1400 | 30 | Win |
| 1550 | 100 | Loss |
| 1700 | 300 | Loss |
The new rating is 1464, and the RD drops from 200 to 151: one win and two losses cost 36 points, and three games of evidence made the system more confident. Note that the loss to the 1700 player cost relatively little. That opponent’s RD of 300 means we’re not sure they are really 1700, so the game says less about our player.
Glicko-2: adding volatility
Around 2000, Glickman published Glicko-2, which adds a third number: volatility (σ). It measures how consistent your results are.
- A player whose results match their rating game after game has low volatility.
- A player on a big winning streak, or with erratic results (say, a fast-improving junior), gets higher volatility, and the system lets their rating move faster.
RD answers “how much do we know about you?” Volatility answers “how much does your strength seem to change?” Two players can have the same rating and RD, but the one with higher volatility will see bigger swings.
Chess.com
Chess.com uses the Glicko system (the original one, with rating and RD). Each time control (bullet, blitz, rapid, daily) has its own rating. When you sign up, you choose your experience level, and your starting rating depends on that choice. New accounts start with a high RD, so your first games cause large swings. If you stop playing for a while, your RD grows again, and your next results will move your rating more.
Lichess.org
Lichess uses Glicko-2, with a few adjustments of its own. Instead of grouping games into rating periods, Lichess updates your rating immediately after every game.
- Everyone starts at 1500 ± 1000, meaning the system is 95% sure you are somewhere between 500 and 2500 (an RD of 500).
- Your rating shows a question mark (?) while your RD is above 110. That is a provisional rating.
- To appear on leaderboards you need, among other conditions, at least 30 rated games and an RD below 75 in standard chess.
- About a year of inactivity makes a rating provisional again.
You can see your RD on Lichess: hover over your rating or look in your profile’s rating stats.
Elo vs Glicko side by side
| Elo | Glicko / Glicko-2 | |
|---|---|---|
| Numbers per player | 1 (rating) | 2 (rating + RD), or 3 with volatility |
| Uncertainty | Handled by K-factor rules | Built in (RD) |
| New players | Higher K by rule | High RD, automatically |
| Returning after a break | No change | RD grows, rating moves faster |
| Opponent’s reliability | Ignored | Games against uncertain players count less |
| Gains and losses | Symmetric with equal K | Usually not symmetric |
| Can you check it by hand? | Yes, with a table | Hard without a calculator |
| Updates | FIDE: monthly lists | Chess.com and Lichess: after every game |
| Used by | FIDE, most national federations | Chess.com, Lichess, many online games |
Why FIDE still uses Elo
If Glicko is more accurate, why hasn’t FIDE switched? A few reasons:
- Transparency. Any player or arbiter can check an Elo calculation with a table and a pencil. Titles, norms and qualifications depend on ratings, so being easy to verify matters.
- Monthly lists. Over-the-board chess is organized around tournaments and monthly lists, not an instant update after each game. The K-factor rules work well enough in that setting.
- Stability. Changing the system would affect every FIDE-rated player and decades of historical comparison.
Online platforms have the opposite situation: millions of games a day, players who come and go, and new accounts all the time. Glicko handles that much better, and nobody needs to check it by hand.
Why your ratings don’t match
Back to the opening question: why are your three ratings different?
- Different pools of players. A rating only means something relative to the other players in the same pool. The Lichess 1800s play other Lichess players, not FIDE players.
- Different starting points. FIDE, Chess.com and Lichess give new players different initial ratings. That shifts the whole scale.
- Different systems and settings. K values, RD limits, rating floors and update rules all shape the numbers over time.
- Different time controls. Online ratings are usually for faster games, which reward different skills from classical chess.
There is no official conversion formula. Online conversion tables exist, but treat them as rough estimates. The only meaningful comparison is within the same pool: your Lichess rating against other Lichess players, your FIDE rating against other FIDE players.
Summary
- Both systems predict results from the rating difference and adjust your rating by result − expected.
- Elo (FIDE) uses a single number and a K-factor set by rules. It’s simple and easy to verify.
- Glicko (Chess.com) adds the Rating Deviation, which measures how uncertain your rating is. It grows when you’re inactive and shrinks when you play.
- Glicko-2 (Lichess) also tracks volatility, so players whose strength changes quickly can adjust faster.
- Ratings from different organizations are not interchangeable. Compare yourself only within the same system.
Whatever the system, the path to a higher number is the same: play more, learn from your games, and let the math catch up.