Hacker News new | ask | show | jobs
by _fizz_buzz_ 920 days ago
I am not sure if that is generally true. It’s true for lichess and chess.com (compared to ELO) It really has only to do with the formula they use. They have different formulas. I think e.g. lichess has a formula so that the median is centered around 1500. This is different from how ELO is calculated. I would assume if they all had the same way to calculate the rating, it should be somewhat close.
1 comments

Yes and no - they use different systems, but even if they used same system ratings would be different.

This is since ratings only measure player's performance relative to other players in the same poll. A lot more people (and a lot more begineers) play online than in FIDE rated tournaments. So playing poll in tournaments is stronger than online playing poll, so even if we assume same rating system in all cases, players will have lower ratings in tournament play than online.

What complicates matters further is that playing polls aren't totally unified - in FIDE Elo ratings it's possible to see regional differences (indians for example are in general underrated), since playing poll is segmented by distance (due to travel costs not a lot of players play internationally). Additionally COVID19 made a total mess of ratings, since there was not enough events last few years, so majority of younglings are underrated due to not playing enough rated games, while improving as fast as the previous generation.

FIDE is currently deciding on rating reform, which will be implemented in january, trying to handle this situation.

What complicates matters even more is that FIDE Elo algorithm is not optimized for accuracy, but for calculations by hand (try calculating glicko2 rating change without computer!), so glicko2 more accurately predicts player performance.

And finally, time controls differ a lot between online and tournament play. Online even 5minutes feels slow, while in tournaments 90min+30sec/move is one of shorter time controls. Performance between slow and fast play is in general correlated, but this correlation is weaker at higher ratings (since both sides of the game have some non-transferable skills, so blitz specialists for example exists).