Rating difference between LiChess and Chess.com is about 100 points? (Chess.com Elo is lower) • page 1/6 • General Chess Discussion • lichess.org

What is the rating difference between LiChess and Chess.com at blitz? An informal survey found it's about 100 points, with Chess.com Elo being lower. This is for below about 2200 Elo. For titled and stronger players, the Elo is about the same. It reminds me of the Elo difference between the USCF and Fide, USCF Elo is "inflated" by about 100 points, meaning Fide Elo was lower by about 100 points.

I play on both sites and roughly the 100 point difference seems about right.

In social media sometimes people make fun of this 100 point difference, but frankly it's within the margin of error in statistics, as Elo for blitz is hard to pin down. You need a large sample size and from the informal survey I found the sample size was too small to really be confident in the 100 point difference. But I do feel Chess.com players are a bit stronger at the same Elo.

What is the rating difference between LiChess and Chess.com at blitz? An informal survey found it's about 100 points, with Chess.com Elo being lower. This is for below about 2200 Elo. For titled and stronger players, the Elo is about the same. It reminds me of the Elo difference between the USCF and Fide, USCF Elo is "inflated" by about 100 points, meaning Fide Elo was lower by about 100 points. I play on both sites and roughly the 100 point difference seems about right. In social media sometimes people make fun of this 100 point difference, but frankly it's within the margin of error in statistics, as Elo for blitz is hard to pin down. You need a large sample size and from the informal survey I found the sample size was too small to really be confident in the 100 point difference. But I do feel Chess.com players are a bit stronger at the same Elo.

kaotic99

#2

400 difference for me. I quit playing though.

Deadban

#3

It is a known fact that chesscom ratings are higher than Lichess, for the mere reason that the starting ratings and the rating systems themselves are different, which unfortunately is the less known thing in the big picture.

This survey that you have found leaves me speechless as it's basically impossible to determine what they claim to have calculated. Nonetheless please share it here so we can have a look at it.

It is a known fact that chesscom ratings are higher than Lichess, for the mere reason that the starting ratings and the rating systems themselves are different, which unfortunately is the less known thing in the big picture. This survey that you have found leaves me speechless as it's basically impossible to determine what they claim to have calculated. Nonetheless please share it here so we can have a look at it.

XsYyLaxa

#4

@Deadban said in #3:

It is a known fact that chesscom ratings are higher than Lichess, for the mere reason that the starting ratings and the rating systems themselves are different, which unfortunately is the less known thing in the big picture.

This survey that you have found leaves me speechless as it's basically impossible to determine what they claim to have calculated. Nonetheless please share it here so we can have a look at it.

Hi @Deadban - here is the survey: https://chessgoals.com/rating-comparison/

Also a thread on Chess.com says for lower rated players, LiChess is inflated but for higher rated players it's the opposite, with an equilibrium at about 2200

Here's a LiChess thread on this: https://lichess.org/forum/general-chess-discussion/chesscom-rating-vs-lichess-rating

bTW the rating difference does not bother me at all. If you ever play in a FIDE OTB tournament that's "open", you'll find that the unrated players are often masters and titled players who have never played in a FIDE international tournament so they are unrated. It happened to some players I know in Asia. They played as "unrated" in a FIDE open but they are National Masters in their home country. Needless to say most of us playing would not stand a chance to even make a 1500 Elo performance rating in such tournaments.

@Deadban said in #3: > It is a known fact that chesscom ratings are higher than Lichess, for the mere reason that the starting ratings and the rating systems themselves are different, which unfortunately is the less known thing in the big picture. > > This survey that you have found leaves me speechless as it's basically impossible to determine what they claim to have calculated. Nonetheless please share it here so we can have a look at it. Hi @Deadban - here is the survey: https://chessgoals.com/rating-comparison/ Also a thread on Chess.com says for lower rated players, LiChess is inflated but for higher rated players it's the opposite, with an equilibrium at about 2200 Here's a LiChess thread on this: https://lichess.org/forum/general-chess-discussion/chesscom-rating-vs-lichess-rating bTW the rating difference does not bother me at all. If you ever play in a FIDE OTB tournament that's "open", you'll find that the unrated players are often masters and titled players who have never played in a FIDE international tournament so they are unrated. It happened to some players I know in Asia. They played as "unrated" in a FIDE open but they are National Masters in their home country. Needless to say most of us playing would not stand a chance to even make a 1500 Elo performance rating in such tournaments.

NM MrPushwood

#5

Although at least then it WOULD be Elo! (my apologies to Pyrokar).

Toscani

edited

#6

Wow, a hybrid speciation graft of 100 rating point difference. The 100 is a cross-breed of a rating.

bufferunderrun

#7

As far as I understand, chesscom uses Glicko-1and not Elo. Please see https://lichess.org/page/rating-systems

XsYyLaxa

#8

@bufferunderrun said in #7:

As far as I understand, chesscom uses Glicko-1and not Elo. Please see lichess.org/page/rating-systems

Interesting. Though it seems that both Elo and Glicko-1 would converge to the same number if you had two populations of the same people playing the same number of games each. Seems that Glicko-1 will allow for leeway for players that are more active and they can rise faster, and it's not zero sum in that points won by one side are exactly offset by the points lost by the other side.

@bufferunderrun said in #7: > As far as I understand, chesscom uses Glicko-1and not Elo. Please see lichess.org/page/rating-systems Interesting. Though it seems that both Elo and Glicko-1 would converge to the same number if you had two populations of the same people playing the same number of games each. Seems that Glicko-1 will allow for leeway for players that are more active and they can rise faster, and it's not zero sum in that points won by one side are exactly offset by the points lost by the other side.

petri999

#9

@XsYyLaxa said in #8:

Interesting. Though it seems that both Elo and Glicko-1 would converge to the same number if you had two populations of the same people playing the same number of games each. Seems that Glicko-1 will allow for leeway for players that are more active and they can rise faster, and it's not zero sum in that points won by one side are exactly offset by the points lost by the other side.
Wherher algoritm is Glicko-1 or two has no bearing on abosolute values. Glicko two is better at tracking changes but difference is really small.

Difference comes from starting rating which is in lichess 1500 and in chess.com mostly 1200. Elo will result in same differences between two players eventually but actual level will settle to differently due to totally different initial value computation.

@XsYyLaxa said in #8: > Interesting. Though it seems that both Elo and Glicko-1 would converge to the same number if you had two populations of the same people playing the same number of games each. Seems that Glicko-1 will allow for leeway for players that are more active and they can rise faster, and it's not zero sum in that points won by one side are exactly offset by the points lost by the other side. Wherher algoritm is Glicko-1 or two has no bearing on abosolute values. Glicko two is better at tracking changes but difference is really small. Difference comes from starting rating which is in lichess 1500 and in chess.com mostly 1200. Elo will result in same differences between two players eventually but actual level will settle to differently due to totally different initial value computation.

XsYyLaxa

#10

@petri999 said in #9:

Wherher algoritm is Glicko-1 or two has no bearing on abosolute values. Glicko two is better at tracking changes but difference is really small.

Difference comes from starting rating which is in lichess 1500 and in chess.com mostly 1200. Elo will result in same differences between two players eventually but actual level will settle to differently due to totally different initial value computation.

That makes sense, but how do you account for the apparent convergence of Elo above 2200, as rumor has it, between Chess . com and LiChess? Unless that's unfounded speculation, or, perhaps the initial conditions matter little for those small number of players that reach levels above 2200?

@petri999 said in #9: > Wherher algoritm is Glicko-1 or two has no bearing on abosolute values. Glicko two is better at tracking changes but difference is really small. > > Difference comes from starting rating which is in lichess 1500 and in chess.com mostly 1200. Elo will result in same differences between two players eventually but actual level will settle to differently due to totally different initial value computation. That makes sense, but how do you account for the apparent convergence of Elo above 2200, as rumor has it, between Chess . com and LiChess? Unless that's unfounded speculation, or, perhaps the initial conditions matter little for those small number of players that reach levels above 2200?