Chess ratings are a method to explain a player’s skill level, and also to determine the expected score against any given opponent. In this post, we explain lichess ratings, chess.com ratings, FIDE ratings, and USCF ratings. With the current landscape in 2020, online chess has definitely increased in popularity as over-the-board events were canceled for most of the year. Chess ratings are highly correlative between the different systems though, so improving one’s online chess skills should transfer to over-the-board play. This plot shows the correlation between USCF Regular Ratings (y-axis) and Chess.com Blitz Ratings (x-axis).
Modern chess rating systems have been around now since the mid-1900s and have been implemented by both online chess sites and over-the-board (OTB) chess federations. In 1948, the West German Chess Federation used the Ingo system, which is possibly the first documented chess rating system. The first system used by the United States Chess Federation (USCF) was the Harkness system, from 1950-1960. Professor Mark Glickman’s A Comprehensive Guide to Chess Ratings has more information on the history of chess ratings.
In chess, there is always a winner and a loser, or the two players split the point. This is completely balanced and often referred to as a zero-sum game. Let’s take a fictional character Bill and say we want to know how good he is at chess. Bill can beat his friends but will be attending a chess club for the first time. Bill shows up to the club and he beats James 7-3 (seven games to three), but loses to Mark 2-8. Mark beats James 10-0 in their ten-game match. It’s clear the rankings for these players should be:
- Mark (18/20, 90% score)
- Bill (9/20, 45% score)
- James (3/20, 15% score)
Now extrapolate that out to a region that has 300 active players. We’d like to know how strong each player is, what the expected score is between any two players, and how can we split these players up into different sections if we are to run a large event. That’s where ratings come into play, and we will start with one of the most popular systems, the Elo system.
Elo System (USCF Ratings, FIDE Ratings)
One of the most famous rating systems, the Elo system, was developed by Arpad Elo and implemented by the USCF in 1960. Elo’s system made some assumptions to help keep the computations manageable, but this is not an issue with the power of computers nowadays. Since Elo ratings will depend on the pool of players and their average rating, ratings between systems are not always comparable. The rating comparison page attempts to map these two systems, which indeed are quite similar.
Establishing a Rating
The USCF, using the Elo system, has a special rating formula for players that have played less than 26 rated games. The formula is a way to estimate one’s rating by calculating a performance rating for each of the first 26 games, and then averaging across the total games played. After game 26, a player moves into the established rating category. For each game, a player wins their performance rating is the opponent’s rating plus 400. Draws count as the opponent’s rating, and losses count as the opponent’s rating minus 400. Here’s an example five-round event:
|Result||Opponent Rating||Performance Rating|
The player’s performance rating through five games is the sum of the performance rating for each game, divided by five. In this case, the provisional rating would be 1290 with five games played.
Once a player has completed 26 rated games, they have what’s called an established rating. Back in the 1990s, I used to help my local club compute club ladder ratings using a table similar to this one for established players. Note that this is an approximation and that computers will use decimal points for the ratings.
For two players with the same rating (Rating Difference 0-24), the winning player’s rating increases 16 points and the losing player’s rating decreases 16 points. For a draw, both players will stay the same rating. The simple math tells us these players should score about 50% against one another.
Let’s assume now that Player A is rated 100 points above Player B. If Player A wins, she gains 12 rating points and Player B loses 12 rating points. If Player B wins though, the change is 20 points for each player. A draw increases Player B’s rating by 4 points and decreases Player A’s rating by 4 points. Again these are approximations, but this makes sense when we look at a rating change by the expected score table. Player A is expected to score about 64% against an opponent rated 100 points lower.
The k-factor determines the maximum points gained or lost in a single game. USCF uses a formula to determine the k-factor and FIDE uses a lookup table based on player ratings. Typically as players play more games or have higher ratings the k-factors decrease. This allows these more established players to have ratings that are also more stable. It also helps account for the matches between a high and low rated player, putting fewer points at risk for the higher player.
Elo rating systems do a pretty good job for the most part, but there are some potential issues that Professor Mark Glickman in particular has attempted to address with his rating systems.
- Rating Inflation/Deflation
- Stability of Ratings
Glicko System (Chess.com Ratings)
Fast forward to 1995 when Dr. Mark Glickman from Harvard University developed a system to improve on the commonly used Elo system. The new Glicko System improved on the Elo system by factoring in player rating reliability. The Rating Deviation as it was coined, or RD, was a value that represented the confidence in each player’s rating. The more games someone played, the more confident we can be in their current rating. Lower RD means we have more confidence, higher RD means we are less sure of the player’s ability.
An example Professor Glickman gives in his paper nicely describes a potential issue with the Elo system. Let’s pretend Player A has played 30 games and is an established 1700 player. Player B is rated 1700 but has played over 1000 games. If one of these players was truly 1500 strength or 1900 strength, would we expect it to be Player A or Player B? It seems like Player A is a reasonable guess considering their rating is based on a small sample of only 30 games.
Now let’s say Player A beats Player B in one game. The Elo system would say that Player A gains 16 points and Player B loses 16 points. We would actually like to see Player A gain more than 16 points and Player B to lose less than 16 points based on the reliability of their ratings. The RD values for each player account for this.
Glicko-2 System (Lichess Ratings)
If we haven’t lost you on the math yet of the Elo and Glicko systems, you’ll be in for a treat when you research the Glicko-2 system! Also developed by none other than Professor Mark Glickman, this system adds an extra component to the Glicko system. Glicko-2 uses a rating and RD, just like Glicko. The third component is a volatility value that measures a player’s consistency. Please read up on the linked article above if you want more details on how it works mathematically. The Glicko-2 system works well when players have a lot of games, and the system has been adopted by Lichess.
Chess Rating Levels
What is a chess rating? ChessGoals lists the following categories for players based on chess.com blitz ratings:
- Novice (<800)
- Beginner (800-1099)
- Intermediate (1100-1399)
- Intermediate2 (1400-1699)
- Advanced (1700-1999)
- Expert (2000-2299)
Typically a chess master is rated 2200, and will be approximately 2300 chess.com blitz. Our rating comparison page can help serve as a guide across different rating platforms.
Chess.com Blitz Ratings
The average chess.com blitz rating is 913, and the majority of players fall between 400 and 1400. Typically ratings start at 1200 on chess.com and quickly adjust to one’s playing level with the Glicko system. Chess.com ratings tend to follow closely to FIDE and USCF ratings. I wrote a chess.com blog post titled Are Chess.com Ratings the Most Accurate? with additional information.
Lichess Blitz Ratings
Lichess ratings start at 1500 and tend to be higher than Chess.com ratings. Here’s a look at their rating distributions:
How to Become a Chess Master
The primary objective for the ChessGoals project is to help you reach chess master. We use data from hundreds of players just like you looking to work their way up the chess ranks. Here are a few steps you can take right now to get you on the right track:
- Subscribe to emails for a more detailed and downloadable plan
- Find a free ChessGoals study plan
- Join the ChessGoals Club on chess.com
- Sign up for a chess.com premium membership and taking advantage of their vast resources. We wrote a previous post titled Is a Chess.com Premium Membership Worth It?
- Start playing on chess.com or lichess! One of the most important factors to chess improvement is playing hours of chess each week. We wrote a post comparing Lichess vs Chess.com with more details.