lichess.org
Donate

ELO Rating Implementation

Currently LIchess uses the glicko-2 rating system (en.lichess.org/qa/6/how-does-the-rating-system-work-on-here). Although I don't have a real life rating, I noticed based on playing people OTB at my chess club that the ratings here are about 400 points higher than the actual ELO rating system for blitz and rapid games. I came to this conclusion based on playing a few 1300 rated players who I struggle to beat OTB and can defeat me. Also, somebody on my friend's list has a rating that is about 1650 in blitz on lichess but his USCF rating is in the 1300's.

I was wondering if it is possible to implement an ELO rating rating system in addition to the already glicko-2 system to figure out approximately what one's real OTB rating would be.

I noticed that chess.com is starting to have ELO ratings on their website. Chess.com has something where you can join the USCF group and get an online USCF rating. Also, FIDE Online Arena has official online ratings, but in order to get these you have to pay money, which I am not inclined to do.

How would lichess implement an ELO rating system? If one already knows their real life OTB rating from either FIDE or USCF then they could provide this to the lichess team and get that ELO rating online. The people that don't have a real life rating would be considered unrated until they play a pool of people with an ELO rating.

I realize that online ratings don't reflect the ELO OTB ratings but was wondering if it is possible at all to make an ELO system online to know apporximately what one's ELO rating would be. I think the ELO rating would only work for blitz, rapid, and classical since in real life I don't think there is bullet ELO rating.

What are the problems with implementing such a system? I read a post by Clarkey about how ratings work on lichess and he said the problems were the following: "The main issue with Elo was that ratings moved too rapidly. Against a player of equal strength, it was possible to gain or lose 100 points in less than a handful of games. There were other issues as well, such as most players (of vastly different strengths) being bulked together near 1200. This was incredibly problematic." I don't really understand how these are major problems associated with the ELO system. The first point that you can gain or lose 100 points in less than a handful of games is not a downside with the ELO system if that's how it works. Also, I don't understand how people would be bulked with similar rating of 1200. Lichess has a large community of chess players probably with real life ELO ratings. The majority of chess players on lichess have glicko-2 ratings between 1600-2000 in blitz anyways. Also, people can already start out with their ELO ratings online if they provide proof of that rating from either USCF or FIDE. The only problem that I see is perhaps more people cheating since the ratings are approximate to real life ELO ratings. For instance an unrated person that wants to face an ELO rated person could use an engine since he/she doesn't yet have the official ELO rating yet until the game is finished.

What does everyone else think? Would people be interested in an ELO rating system? Personally, I would want to know my ELO rating without traveling to tournaments and paying money.

the fact that there's no simple correlation from lichess ratings to OTB ratings has nothing to do with the system being ELO or Glicko-2 or whatever. it has to do with the player pool being different and the games being played under different conditions.
Elo is a special case of Glicko-2 which ignores the fact that players learn over time.
I agree with the above responses. Doing this would be pointless, switching to elo wouldn't change how inaccurate the ratings are compared to OTB. If you want an OTB rating then go play in a tournament.
You could always try elometer.net and see what it thinks your likely ELO is.
Well, it's not easy to maintain ONE system in a "equal" way. The fourth figure after the point/comma in the calc formula can have drastic consequences in a "high frequence (elo) trade" with millions of games per day. (I'm sorry, I hope u understand what I mean)

In my personal opinion I do not think it's the algorithm but rather the starting value is much too high. The German median is 1550 btw.

This topic has been archived and can no longer be replied to.