Lichess and Chess.com Elo Comparison

Lichess and Chess.com are two of the most popular chess websites, yet their Elo systems are very different.

• Lichess uses the Glicko-2 rating system. New players generally start with a rating of 1500, although their rating may change significantly during their provisional period. Against an opponent with a similar rating, a normally established player may gain or lose roughly 5-6 or 6–7 rating points, depending on factors such as rating deviation and recent activity.

• Chess.com also uses a Glicko-based rating system, but its ratings are not directly comparable to Lichess ratings. New players typically begin with a different starting rating, and the number of points gained or lost depends on the players’ ratings, rating deviation, and whether their ratings are still provisional.

How do we compare them each other? Is there a mathematical formula to do that task? Let me know! :D

"Is there a mathematical formula to do that task"

No. And this question has been asked, and answered, dozens if not hundreds of times across the internet.

I use the ChessGoals Chess Rating Comparison tool, which I find the most useful.

1500=1200

The graphs were built from two public Kaggle datasets: a Chess.com dataset containing 60,000+ games (adityajha1504/chesscom-user-games-60000-games) and the Lichess game dataset (datasnaek/chess), which contains about 20,000 games from around 2017. The game records were cleaned to identify unique players and their ratings, with rated games used for the main comparison and provisional/starting-rating players excluded. Instead of assuming that individual players were the same across both platforms, the analysis compares the population percentiles of each rating system—for example, identifying what Lichess rating represents the same population percentile as a given Chess.com rating. The resulting percentile mappings were then used to create the distribution, percentile, rating-conversion, and rating-difference graphs, allowing us to visualize how the two rating scales compare across their represented player populations.

@Toscani said ^

The graphs were built from two public Kaggle datasets

Are the charts based on rapid games? The ratings for the different time controls are different.

The interesting thing, for me, is that the greatest difference is at low ratings - by the time you get to 2200 the ratings are much more similar.

my friend is 3000 blitz on cheese.com and 2000 fide ;-;

The higher the ratings climbs the less the ratings are different.
You can manual compare top rated players to prove that.
https://www.chessmonitor.com/players

The Problem with the datasets. They are incompatible in time controls and so the graphs are not comparable.
Thanks for pointing out the issue of time controls. I did not think of that.

LichessorgMixed (Bullet, Blitz, Rapid, Classical)Live ratings (No assistance permitted)
ChesscomDaily (2-3 days/move)Correspondence rating (Do they use opening books?)

The current two datasets cannot be meaningfully combined for a cross-platform rating comparison.
I would need to find similar time controlled datasets to get matching data from both platforms.
Then I could use the python scripts to reproduce so valid graphs.

@rbhoopathi04 said ^

my friend is 3000 blitz on cheese.com and 2000 fide ;-;

”cheese.com” ?

A player’s Chess.com ratings are generally lesser than Lichess