Yet another lichess vs. FIDE rating comparison

Hi everyone. I just wanted to share some analysis of ratings correlations that I posted on reddit. Rather than repeat it all here I'll just post the link.

www.reddit.com/r/lichess/comments/mdvbkp/yet_another_lichess_vs_fide_rating_comparison/

I think the important part again that estimater are far of no matter what. with 95 certainty lichess 2000 is between 1500 and 2100 FIDE. as anyonce can check from your plot



An that is blitz which has best correlation. You said bein surprised about it. I allways assumed that this is the case just because it played most and in particular strong player do not play much rapid and even less classic on lichesss. statistical meaurement needs data points to work.

Classical ratin seem close to random in predictive power

ethanlebowitz.github.io/RatingConverter/index.html

Saw this nice website that calculates your estimated elo with your lichess rating info.

@Aadithya2002 yes it calculates but it based to almost dataset as in picture above. So what ever estimate is at best +-200 points. So it is not really of any use. You get just good estimate with ruler from pics above and since the dataset used for those pics used only verifiable FIDE rating it is more accurate - which is no accurate at all - just best possible

@Aadithya2002 , your website is a joke.
They estimate me as 1686.

So. I randomly add a top 10 GM Mamedyarov (GM Azerichessss) and his rating is estimated as 2380 LMAO.

@drmrboss
It also estimates Magnus Carlsen as ~2500 heh. Not sure where that site is getting it's calculation from

Guys, this is not my website.. I appreciate the effort of the creator..

Also, thanks for your observations that the estimations are not very accurate.

@Steven-ODonoghue the calculatin in that site is getting information from tabulating lichess ratings vs self proclaimed FIDE rating and then doing multivariate linear regression from those data points. And id click "about the data" it will take you to page explaining how they come to data
ethanlebowitz.github.io/RatingConverter/about.html
what is different on article by @dryguy is that he made effort to get real FIDE rating by correlating names and FIDE-id's which is tad more accurate. Just that even best model lichess-blitz explain about 36% of variability of FIDE rating. Which you would also find out from the reddit post linked

There is nothing wron with the estimate on the link. It is best effort. It just possible to make better estimates as is evident from scattergram i posted above.
So people wanting know they FIDE really just have attend tournaments to find out. Which is most likely impossible. Very few countries allow tournament at the moment