Here's my rating data from various chess organizations, for comparison:
FIDE otb/classical - 2225
USCF otb/classical - 2186
ICC 15-minute - hovers around 2100
ICC 5-minute - hovers around 2050
ChessCube overall - hovers around 2200
Lichess overall - hovers around 1650
Facebook chess - hovers around 1750
I've also goofed around on playchess.com, chess.com, FICS and the Russian site OK Chess, or whatever it's called, but probably not enough to have valid, established ratings at those sites.
I'd be interested in seeing other players report their rating comparisons in this forum to see what patterns develop.
This is not a debate about which site has the most accurate ratings. There is no "correct" or "incorrect" rating for a player. By this I mean the absolute rating numbers derived by applying a particular organization's rating formula aren't terribly relevant. All a rating indicates is a player's relative strength among all the players who play on a site or within an organization. Nevertheless, there are reasons for rating differences among chess sites and I'd be interested in knowing to what extent chess cheating -- using chess engine software to generate recommended moves played in a game -- affects ratings at various internet chess sites.
In my humble opinion, ICC and ChessCube have much nicer, more customizable, feature-rich interfaces than Lichess (which is to be expected since ICC and CC are paid-members-only sites whereas Lichess is free and open to the public) but for some reason I've been playing more on Lichess recently. I find its utter simplicity and easy accessibility attractive, I guess. All the cheating on Lichess is a turn-off though, a real headache sometimes.
Facebook's main chess interface is truly horrible in nearly all respects, but I've found myself playing there late at night too often, perhaps because the chess link from FB is ever lurking before my tired eyes. I've found the cheating on FB among higher rated players about as bad as on Lichess. Which is to say, very bad indeed.
The experienced eye can sense cheating pretty easily. Some moves or move-sequences are just too unusual or too precise to have been conceived by a human playing at rapid speed. When I play an opening with which I'm very familiar and comfortable, and I get methodically rolled off the board in 25-30 moves in a game that culminates with a clever 5-move combination, I wonder. When it happens two or three games in a row, I take a time-out to investigate. All too often, I find I've just lost 50 or more hard-earned rating points to Rybka or its kin disguised as a human. The offender's rating is promptly reset to 1200 by the dutiful Lichess staff of volunteers, but the offender reappears, sometimes within a matter of minutes, with a new user ID and the same nasty cheating habits.
Lichess relies primarily on users to snitch on one another as cheaters, and those caught in the act cannot effectively be banned from the site. Registration at Lichess requires no actual identification or verification of the person. Cheating on other sites, such as ICC, Playchess and ChessCube, is automatically monitored by software, and cheaters once caught are generally banned for life from playing at those sites. Thus, cheating is far more common at free sites such as Lichess, FICS and Facebook Chess.
Exactly WHY some people choose to play moves selected by a computer to win in no-stakes, long-distance chess games with people they don't know and can't see, against people who don't know and can't see THEM, is perplexing, but it's as common in internet chess as ants at a picnic and it's a frustrating phenomenon to all of us who just want to log on and play a few games of chess.
I have no training in medicine, but I would surmise that chess cheating must trigger a release of dopamine in the same pleasure center that makes the brains of adolescent-geek-hackers light up like a Las Vegas slot machine. The experience of winning by cheating must be like striking the enemy target's bulls-eye with a drone rocket from 10,000 miles away, but in a manner that requires zero skill. Cheap thrill.
I suspect, but do not know, that the greater prevalence of cheating on Lichess, as compared with that on ICC and CC for instance, contributes to relatively lower overall Lichess ratings. The Lichess administrator could probably confirm this, though he probably can't say exactly to what extent this is true.
I believe all the chess organizations listed above use some variant of the Elo rating system, but that doesn't imply that their ratings of chess players are or necessarily should be comparable. There's some correlation between the rate of cheating at a site and the average ratings there, I believe, but how strong that correlation is I don't know. The reasons for the sometimes significant differences in chess ratings from one site to another are myriad and complex.
To the person in an earlier post on Lichess's rating system who noticed that a player gained well over 50 points for a single win: That can happen under the traditional Elo system, but generally only if the player/account were new to the chess site/organization and thus had a provisional rating when the game was played. A player who has an established (nonprovisional) rating can win or lose no more than 32 points in a single game under the standard application of the Elo formula as I understand it.
The purpose of using magnified rating adjustments for provisionally rated players is to accelerate a new player's placement in the rating array at his or her actual playing strength while mitigating the gain or loss of rating points experienced by those playing against untested players. Once a new player has played a specified number of rated games, e.g. 24, his or her rating becomes nonprovisional and the regular Elo performance rating adjustment formula is applied thereafter.
Hope this brief explanation helps. The numbers 32 and 24 mentioned above are based on my memory of the factors used by the USCF in its application of the Elo system. The Lichess administrator may have selected different factors, but even if they are different, that alone should not create a material difference in the average rating of players at the Lichess site as compared with those at other sites.
A common example of an Elo variant adopted by some chess organizations is a maximum rating adjustment (a K-factor) of less than 32 for higher-rated players, such as masters rated over a certain threshold, thereby limiting the magnitude of such players' rating fluctuations. The thinking is that large rating fluctuations occurring within a short span of time may be more appropriate for lower-rated players than for a strong master. A gifted, young 1500 player may with practice become a legitimate 1800-strength player within a matter of a few months, but 2500 players don't make the jump to 2800 strength quite so quickly, for those relative handful of 2500's who ever will do so.
I've worked with the Elo system as a player and otb TD for about 40 years now, but I am by no means an expert on the system and all its variants. If I've misstated something important please correct me.
chez
FIDE otb/classical - 2225
USCF otb/classical - 2186
ICC 15-minute - hovers around 2100
ICC 5-minute - hovers around 2050
ChessCube overall - hovers around 2200
Lichess overall - hovers around 1650
Facebook chess - hovers around 1750
I've also goofed around on playchess.com, chess.com, FICS and the Russian site OK Chess, or whatever it's called, but probably not enough to have valid, established ratings at those sites.
I'd be interested in seeing other players report their rating comparisons in this forum to see what patterns develop.
This is not a debate about which site has the most accurate ratings. There is no "correct" or "incorrect" rating for a player. By this I mean the absolute rating numbers derived by applying a particular organization's rating formula aren't terribly relevant. All a rating indicates is a player's relative strength among all the players who play on a site or within an organization. Nevertheless, there are reasons for rating differences among chess sites and I'd be interested in knowing to what extent chess cheating -- using chess engine software to generate recommended moves played in a game -- affects ratings at various internet chess sites.
In my humble opinion, ICC and ChessCube have much nicer, more customizable, feature-rich interfaces than Lichess (which is to be expected since ICC and CC are paid-members-only sites whereas Lichess is free and open to the public) but for some reason I've been playing more on Lichess recently. I find its utter simplicity and easy accessibility attractive, I guess. All the cheating on Lichess is a turn-off though, a real headache sometimes.
Facebook's main chess interface is truly horrible in nearly all respects, but I've found myself playing there late at night too often, perhaps because the chess link from FB is ever lurking before my tired eyes. I've found the cheating on FB among higher rated players about as bad as on Lichess. Which is to say, very bad indeed.
The experienced eye can sense cheating pretty easily. Some moves or move-sequences are just too unusual or too precise to have been conceived by a human playing at rapid speed. When I play an opening with which I'm very familiar and comfortable, and I get methodically rolled off the board in 25-30 moves in a game that culminates with a clever 5-move combination, I wonder. When it happens two or three games in a row, I take a time-out to investigate. All too often, I find I've just lost 50 or more hard-earned rating points to Rybka or its kin disguised as a human. The offender's rating is promptly reset to 1200 by the dutiful Lichess staff of volunteers, but the offender reappears, sometimes within a matter of minutes, with a new user ID and the same nasty cheating habits.
Lichess relies primarily on users to snitch on one another as cheaters, and those caught in the act cannot effectively be banned from the site. Registration at Lichess requires no actual identification or verification of the person. Cheating on other sites, such as ICC, Playchess and ChessCube, is automatically monitored by software, and cheaters once caught are generally banned for life from playing at those sites. Thus, cheating is far more common at free sites such as Lichess, FICS and Facebook Chess.
Exactly WHY some people choose to play moves selected by a computer to win in no-stakes, long-distance chess games with people they don't know and can't see, against people who don't know and can't see THEM, is perplexing, but it's as common in internet chess as ants at a picnic and it's a frustrating phenomenon to all of us who just want to log on and play a few games of chess.
I have no training in medicine, but I would surmise that chess cheating must trigger a release of dopamine in the same pleasure center that makes the brains of adolescent-geek-hackers light up like a Las Vegas slot machine. The experience of winning by cheating must be like striking the enemy target's bulls-eye with a drone rocket from 10,000 miles away, but in a manner that requires zero skill. Cheap thrill.
I suspect, but do not know, that the greater prevalence of cheating on Lichess, as compared with that on ICC and CC for instance, contributes to relatively lower overall Lichess ratings. The Lichess administrator could probably confirm this, though he probably can't say exactly to what extent this is true.
I believe all the chess organizations listed above use some variant of the Elo rating system, but that doesn't imply that their ratings of chess players are or necessarily should be comparable. There's some correlation between the rate of cheating at a site and the average ratings there, I believe, but how strong that correlation is I don't know. The reasons for the sometimes significant differences in chess ratings from one site to another are myriad and complex.
To the person in an earlier post on Lichess's rating system who noticed that a player gained well over 50 points for a single win: That can happen under the traditional Elo system, but generally only if the player/account were new to the chess site/organization and thus had a provisional rating when the game was played. A player who has an established (nonprovisional) rating can win or lose no more than 32 points in a single game under the standard application of the Elo formula as I understand it.
The purpose of using magnified rating adjustments for provisionally rated players is to accelerate a new player's placement in the rating array at his or her actual playing strength while mitigating the gain or loss of rating points experienced by those playing against untested players. Once a new player has played a specified number of rated games, e.g. 24, his or her rating becomes nonprovisional and the regular Elo performance rating adjustment formula is applied thereafter.
Hope this brief explanation helps. The numbers 32 and 24 mentioned above are based on my memory of the factors used by the USCF in its application of the Elo system. The Lichess administrator may have selected different factors, but even if they are different, that alone should not create a material difference in the average rating of players at the Lichess site as compared with those at other sites.
A common example of an Elo variant adopted by some chess organizations is a maximum rating adjustment (a K-factor) of less than 32 for higher-rated players, such as masters rated over a certain threshold, thereby limiting the magnitude of such players' rating fluctuations. The thinking is that large rating fluctuations occurring within a short span of time may be more appropriate for lower-rated players than for a strong master. A gifted, young 1500 player may with practice become a legitimate 1800-strength player within a matter of a few months, but 2500 players don't make the jump to 2800 strength quite so quickly, for those relative handful of 2500's who ever will do so.
I've worked with the Elo system as a player and otb TD for about 40 years now, but I am by no means an expert on the system and all its variants. If I've misstated something important please correct me.
chez
