The Short Answer
Quick answer: Chess.com uses the Glicko system, not plain Elo. Every rating comes with a rating deviation (RD), which shows how sure the system is about it. A high RD means bigger swings. In my sample of 390 rapid ratings, the median RD was 52 for players active this week and 120 for players idle over a year.
I pulled those ratings from Chess.com's public data on 9 November 2026.
Below is what RD does, how much one win is worth at each RD, and where a new rating starts.
What I Checked
Chess.com explains its system in words, not numbers. So I combined three sources.
- Chess.com's help center article on ratings, and the co-founder's older article on Glicko.
- Mark Glickman's paper on the Glicko system, which gives the formulas.
- Public stats for 420 Chess.com accounts, picked at random from the player lists of six countries.
The countries were the United States, India, the United Kingdom, Germany, Brazil and the Philippines. Of the 420 accounts, 390 had a rapid rating.
For each rating, I read the RD and the date of the last game. Chess.com's public API returns both.
Glicko, Not Elo
Elo gives each player one number. Glicko gives two: a rating and a rating deviation.
Chess.com's co-founder explained the reason in an article on the site. With Elo, "you have to assume that everyone's rating is just as sure as everyone else's rating." That isn't true. A player with 1,000 games has a far more reliable rating than someone on their first game.
Glicko fixes that. The RD tells the system how much to trust each rating, and it changes how far a rating moves.
Chess.com's help center says it plainly. "Chess.com uses the Glicko rating system."
It doesn't say which version in the pages I read. Lichess uses Glicko-2, a later version by the same author.
Rating Deviation in Plain Words
Think of RD as a margin of error. A low RD means the system is confident. A high RD means it's guessing.
Glickman's paper turns it into a range. Take your rating, then add and subtract twice the RD. The system is 95% confident your real strength sits inside that range.
| Rating | RD | 95% range |
|---|---|---|
| 1500 | 30 | 1440 to 1560 |
| 1500 | 60 | 1380 to 1620 |
| 1500 | 150 | 1200 to 1800 |
| 1500 | 350 | 800 to 2200 |
So two players can both show 1500 and mean very different things. One is a settled 1500. The other could be anywhere from a beginner to a club player.
Why Your Rating Moves More After a Break
RD doesn't only fall as you play. It also rises while you're away.
Chess.com's help center says, "If you haven't played in a while, your RD will increase, causing your rating to change more dramatically after your next game." The reason it gives is simple. Chess.com can't tell whether you've been practicing elsewhere or haven't played at all.
Glickman's paper puts a number on it, using his own example settings. A typical RD of 50 climbs back to 350 after about five years without games. That's the RD of a brand new player.
Chess.com doesn't publish its own settings. My sample shows how fast it happens there.
What My Sample Showed About Breaks
Here's the median rapid RD in my sample, grouped by how long ago each player's last rapid game was.
| Last rapid game | Players | Median RD |
|---|---|---|
| Within 7 days | 243 | 52 |
| 7 to 30 days ago | 44 | 75 |
| 30 to 90 days ago | 28 | 74 |
| 90 to 365 days ago | 42 | 86 |
| Over a year ago | 30 | 120 |
The pattern matches what Chess.com describes. Active players sit lowest, and players gone for a year sit more than twice as high.
The middle groups are small, so don't read much into the gap between 75 and 74. The direction is the part that holds.
How Many Games Until a Rating Settles
Games played matter even more than breaks. I looked only at players with a rapid game in the past 30 days, then grouped them by games played.
| Rapid games played | Players | Median RD |
|---|---|---|
| Under 20 | 59 | 145 |
| 20 to 99 | 78 | 68 |
| 100 to 999 | 117 | 39 |
| 1,000 or more | 36 | 34 |
The big drop comes early. Somewhere between 20 and 100 games, the median RD falls by half.
After 100 games, it barely moves. That's why a seasoned player's rating feels stuck. Each game nudges it only a little, because the system already trusts it.
How Much One Win Is Worth
To show the effect of RD, I ran Glickman's formulas for a single win. You're rated 1500 and you beat another 1500 with an RD of 50. Only your RD changes.
| Your RD | Points gained for the win |
|---|---|
| 50 | About 7 |
| 100 | About 26 |
| 200 | About 86 |
| 350 | About 175 |
That's the same win, worth 25 times more at RD 350 than at RD 50.
These are textbook numbers, not Chess.com's exact ones, since it doesn't publish its settings. But the shape is the same. New and returning players swing hard, and settled players move a few points at a time.
Your Opponent's RD Counts Too
The help center makes a second point that surprises people. Your opponent's RD changes your result as well.
If you lose to a much lower rated player with a high RD, "your rating won't change much since their rating may not be accurate." The same logic applies to wins.
Glickman's formulas show it. You're 1500 with an RD of 50, and you beat another 1500.
| Opponent's RD | Points you gain |
|---|---|
| 50 | About 7 |
| 200 | About 6 |
| 350 | About 5 |
The effect is smaller than your own RD's, but it's real. A win over a player nobody can place yet proves less.
The Rating Gap Still Matters
RD isn't the only factor. The rating gap works much like it does in Elo.
Here's Glickman's math with both players at an RD of 50, starting from 1500.
| Result | Points |
|---|---|
| Beat a 1700 | About +11 |
| Draw with a 1700 | About +4 |
| Beat a 1300 | About +3 |
| Lose to a 1300 | About −11 |
The help center describes the same rule in words. Beating a stronger player gains more, and losing to a weaker one costs more.
So a draw with a stronger player can still raise your rating. A win over a much weaker player barely counts.
Where a New Rating Starts
Chess.com doesn't start everyone from scratch. Its help center lists three cases.
- If you've played other live time controls, your new one starts from the closest in speed. Rapid uses blitz before bullet, for example.
- If you haven't played any live games, your rating comes from the skill level you chose when you signed up.
- Daily chess uses only that sign up skill level.
The help center adds that very high ratings "may be lowered" when they carry over, so you start at a more realistic level.
It doesn't list the number for each skill level. Players on its forums report figures, but I couldn't confirm them in an official page.
Daily Ratings Swing the Most
The help center says live and daily ratings "never carry over to each other." My sample shows another difference: daily RD runs far higher.
| Time control | Ratings in my sample | Median RD |
|---|---|---|
| Rapid | 390 | 63 |
| Blitz | 280 | 72 |
| Bullet | 220 | 111 |
| Daily | 180 | 257 |
Daily games take days or weeks each, so most players finish few of them. Fewer finished games means a less certain rating.
So if your daily rating jumps 100 points after one game, that's normal. It's the system still working out where you belong.
Chess.com vs Lichess Ratings
Both sites use the Glicko family, but the numbers aren't interchangeable. Each site has its own player pool and its own settings.
Lichess uses Glicko-2 and marks a rating as provisional with a question mark. Our guide to how Lichess ratings work covers its version.
Because the pools differ, the same player usually shows a different number on each site. Our Chess.com vs Lichess rating guide explains the usual gap.
Within Chess.com, each time control keeps its own rating and RD too. Our guide to rapid vs blitz ratings shows how far those drift apart.
What This Means for Your Games
A few habits follow from how Glicko works.
- Expect big swings in your first 20 games. They don't mean much yet.
- After a long break, expect your first games back to move your rating more than usual.
- Don't chase a few points after a settled run. At a low RD, each game moves you only a little.
- Judge progress over 50 games or more, not over one evening.
The fastest way up isn't a lucky streak. It's fewer blunders game after game.
Between rated games, play a few against our chess bot. No rating rides on them, so you can test a new opening without the swing.