Chess ratings explained

What a chess rating is

A rating is a running summary of your results against other rated players, nothing more. It does not measure you in isolation; it places you inside the pool of players whose games produced it. Put the same player in a pool of strong club players or in a pool of casual online players and the number can come out very differently.

Every rating carries uncertainty, and the Glicko systems track it as a number of its own, the rating deviation. Mark Glickman, who designed those systems, suggests reading a rating as an interval: the rating plus or minus twice its deviation gives a 95 percent confidence interval. A 1850 rating with a deviation of 50 therefore spans 1750 to 1950. The number on your profile is the middle of a range, not a precise reading.

That is why ratings move. Each result is new evidence: beat someone you were expected to lose to and the number rises a lot, lose to someone you were expected to beat and it falls a lot. When the system knows little about you, because you are new or returning after a break, it lets each result count for more.

How the update works in the Elo family

The Elo family of systems follows one idea. Before a game, the gap between the two ratings gives an expected score for each side: a large rating edge predicts a score close to a full point for the stronger player, and equal ratings predict half a point each. After the game, your rating changes by how far your actual score landed from that expectation.

The size of the change is scaled by a number called the K factor. A high K makes ratings jump after each result, which suits players whose true level is still unknown. A low K makes ratings steady, which suits established players.

Elo is old and mathematically simple. Lichess notes that FIDE used it back in 1971, when it was the only choice. Glickman's Glicko added a rating deviation, and Glicko-2 added a volatility measure for how erratic a player's results have been. Glickman says both systems are in the public domain.

FIDE ratings

FIDE, the world chess federation, publishes its rating lists under the Rating Regulations in its Handbook. The rules below come from that chapter, each figure with its clause.

A first rating is computed as if the player had also played 2 (B02 8.2.2) extra games against opponents rated 1800 (B02 8.2.2), scoring half a point in each. A first rating computed this way cannot exceed 2200 (B02 8.2.3). To appear on a list at all you need games against rated opponents, at least 5 (B02 7.1.4), and they may come from several tournaments. Games for a first rating can be pooled across a window of 26 months (B02 7.1.4). The lowest published rating is 1400 (B02 7.1.4).

For players rated under 2650 (B02 8.3.1), a rating difference counts for at most 400 (B02 8.3.1), however large the real gap. Players at 2650 (B02 8.3.1) or above use the full difference, a rule the Handbook prints with its own effective date, 1 October 2025. A rating change is rounded to the nearest whole number, with halves rounded away from zero (B02 8.3.4).

A player with no rated play for 1 year (B02 7.2.2) is treated as inactive. The list closes 3 days (B02 7.1.3) before it is published, so results that arrive after the closing date wait for a later list.

The K factor depends on where you stand. The Handbook lists four cases, plus a cap on the whole rating period, and does not say which case wins when two of them apply:

• A new player has a K of 40 (B02 8.3.3) until 30 games (B02 8.3.3) are completed.
• A player rated under 2400 (B02 8.3.3) has a K of 20 (B02 8.3.3).
• Once a rating has reached 2400 (B02 8.3.3), K is 10 (B02 8.3.3) and stays there even if the rating later drops.
• A junior has a K of 40 (B02 8.3.3) until the end of the year in which they turn 18 (B02 8.3.3), while rated under 2300 (B02 8.3.3).
• K times the number of games on one list for a rating period may be at most 700 (B02 8.3.3).

Lichess ratings

Lichess is a free online chess site, and its own FAQ says its ratings come from the Glicko-2 method. A new account starts at 1500 with an uncertainty of 1000, which means the system is 95 percent sure the true level lies between 500 and 2500. That wide range is why a new player's rating can move several hundred points at a time.

A question mark beside a Lichess rating marks it as provisional, which Lichess defines as a rating deviation above 110. The question mark can return after about a year without playing, because the uncertainty grows while you are away.

Lichess keeps a separate rating for each game category, so each kind of game you play has its own number. To appear on a leaderboard you need at least 30 rated games in a category, a game in the last week and a low deviation. Lichess also says its calculation takes piece color into account, awarding more points for results with Black.

US Chess ratings

The US Chess Federation describes its system in a specification by Mark Glickman and Thomas Doan, dated 6 April 2026. It defines six separate rating systems: Blitz, Quick Chess and Regular, each over the board and online. Events are assigned by time control: Blitz for main time plus increment from 5 to 10 minutes, Quick Chess above 10 and below 30, and Regular at 30 or more. Over-the-board Quick ratings also apply up to 65 minutes, which overlaps the Regular range and gives some games a dual rating.

Ratings are recalculated for each rated event rather than game by game, and official ratings are shown rounded to a whole number. A rating counts as established once it rests on more than 25 games. A player with no rating receives a starting estimate from age or from other ratings the player holds; for an unrated adult with no other rating, the age-based start rises with age up to 1300 for adults older than 26.

US Chess also uses floors. No rating can fall below 100, and an over-the-board player's personal floor can rise to at most 150 through wins, draws and events played. Established players earn higher floors in steps of 100 from 1200 to 2100, set 200 points below their highest established rating, rounded down to a step.

Chess.com ratings

Chess.com's help center says every player has a rating meant to measure skill and match players against similar opponents. The site uses the Glicko system, with a rating deviation as its measure of how confident it is in a rating. The size of a change depends on the rating gap to the opponent and on how confident the site is in both players' ratings, so new and returning players, who carry a larger deviation, see bigger moves.

Live Chess keeps separate Bullet, Blitz and Rapid ratings, and the help center says they do not affect one another. A live game is assigned to a rating type by estimated length, counting the base time plus the increment over an assumed 40 moves, so a 2 minute game with 12 second increments counts as Rapid. Chess.com says large gaps between a player's Rapid, Blitz and Bullet ratings are normal and the site cannot adjust them to match.

Why the numbers differ, and why we do not convert them

Two ratings from different systems are built from different ingredients. The pools of players differ, the formulas differ, the time controls that feed each rating differ, and new players start in different places. Lichess says directly that its ratings cannot be compared with ratings from other organizations or sites, because the systems, implementations and player pools differ.

People still ask for a conversion, such as what a particular online number is worth in an official rating. We do not publish one, because we have no sourced conversion for comparing players across these systems and an invented one would mislead you. The table below sets out what each system says about itself; where the source we read was silent, the cell says so.

How to use your rating

Compare yourself only inside one system, and ideally inside one time control. Lichess itself advises treating ratings as relative figures inside one pool of players, useful for estimating who is likely to win, draw or lose.

Watch the trend across many games rather than any one result. While a rating is still provisional on Lichess, or not yet established in US Chess (which takes more than 25 games), a few results can swing it a long way, so read little into it until then. After that, a steady climb or a long flat stretch tells you more than any single jump.

Use the band you are in, not the exact number, to choose what to study. Lower down, the quickest gains tend to come from tactics and a blunder check before every move, because games there are often decided by a piece left hanging. Further up, more games turn on opening choices, endgame technique and converting an advantage. Look at the games you lose to players near your own level and ask which of these keeps appearing.

If you want a second pair of eyes on that question, the first lesson is a paid first lesson in which George Li, International Master, reviews your games and builds a plan. For the arithmetic of your own results, try the rating calculator and the performance rating tool.