Lichess versus chess.com ratings/percentiles

I noticed that on chess.com my rating is much lower (for example in 10 min games) but on the other hand even with the lower rating my percentile on chess.com is far higher than on lichess.
To be specific:
- lichess 10 min:
rating: 1200-1300
percentile: 25%
- chess.com 10 min:
rating: 900-950
percentile: 47%
Why this differences and how do you explain the apparent contradiction of having so much higher percentiles with so much lower ratings?

Some servers are overrated like lichess and chess24 (+ some hundreds), others underrated like chess.com and playchess meanwhile (- some hundreds).

You cannot really compare because it depends on many factors:

-starting values
-different distribution/population
-dealing with cheaters
-formulas
-player's pool
-double accounts
-dead accounts

and so on.

PS: I have a span of 500 between different servers...

hi,

i guess @Sarg0n has said everything, infact the most important is the starting elo value, assuming on lichess the gaussian distribution of elo is centered on 1500 vs 1200 ( ? ) on chess.com

Ok, interesting remarks but what about percentiles? Why percentiles on chess.com are so much higher than on lichess? It seems like lichess is played by much expert/stronger people - on average -than chess.com. Maybe it is because one is less likely to know about lichess if he is not really into chess, while chess.com has a more widespread popularity? Actually if i google-search "chess online" the first result I see is lichess.

It might be because they count differently. Lichess only considers players who have been active in the last week. I'm listed on chess.com's leaderboard at #20216 (out of over 4 million!) even though I've only played three games, and those were in August.

If your rating is 900-950 and you claim the percentile puts you in the top 47% at chesscom, best check again. That can not be accurate. You are saying approx. 1/2 the players are under 1000. If that is true, then statistics reported are including possibly all those players who played a few games and quit; complete beginners who never played again.
You've got to be careful when evaluating statistical information and know how it was come by, what data sets were used/not used.

The playing pool is important along with the number of games. Even 100's of games are not enough of a sample. Players at lichess start with a higher rating in general, but players at chesscom have an option to start at 1800. Much of the statistical results given include players who never established a true rating by playing a representative amount of games vs all players. Example: a great deal of players only play friends or remain in a small group and only play each other. Another factor is the % of program users. lichess does a far better job of it, identifying and flagging abusers. Cheating is generally agreed as far, far more prevalent at chesscom at all ratings, making the achievement of a higher rating more difficult.

I reached approximately the same ratings at chesscom as I maintain here. My rating would after better than average play vs honest players reach it's peak, but could easily plummet with a run of bad luck after getting paired with program users, that chesscom takes forever to identify (and seems unable to identify most) and responds with a slap on the wrist. If you're a paying member, they simply ask you not to do it again and make an apology! Because they survive on paid membership, they are not about to start deleting accounts and get involved with refund issues.

I wouldn't trust a single statistic that chesscom provides. Every stat is hand manipulated to shed a better light on their performance and keeping their advertisers satisfied.
The biggest joke is the 20M+ members. A great % of "members" have created over dozens of accounts. In not much time their membership will surpass the earths total population!

Totally true.
If we look at the statistics of chess.com the great majority have an elo between 900 and 1200 and in lichess it is between 1500-1700.

In fact, if I look at my statistics. In quick games it says:

Lchess: Better than 72% of Rapid players.
Chess.com: 87.72

Summary that there are more strong users in lichess than in chess.com

In lichess it is very easy to see how many active users there are in each "ELO" range (Between 2000 and 2100 ELO in Blitz there are 4519 players) but in chess.com you can not see how many active users there are in that elo range.
Chess.com only shows the number of registered users in that range of elo, but there is no way of knowing how many historical users do not play. (Users on chess.com with elo 2000: 16000 registered users)

But of those 16,000 registered users would have to see how many are actually active and playing