@crtex said in #40:
> They say in the last part of the blog that past 2000 FIDE there is no accurate way to estimate online classical to FIDE, because there aren't enough high rated online classical players. 2400 classical on lichess is only 20 players with low deviation. 2300 classical is only 70 people.
> That's why I don't really like this estimation system. There aren't enough sample points.
i understand i didnt take that into account before.
Interesting. Lichess indeed has very few classical players with such a high rating, see https://adjva4.dpdns.org/player/top/200/classical.
Note that these don't count if you live in India
Salam menim adim musadi
Ehrjtjrth eşq mgnyfz.Zf cennete sjafjrhrbsdsaf djdeyaf
@NoseKnowsAll What about the CFC chart?
@NoseKnowsAll Very interesting results!
If understand correctly the app that you are working on is this:
https://dojoratingbandconverter.netlify.app/
So it takes Lichess account info, chess.com account info, and UCSF ID as inputs.
Have you considered passing more explanatory variables to the model? For example:
Lichess blitz
Lichess rapid
Lichess classical
chess.com blitz
chess.com rapid
uscf rating
and fide rating is the target variable.
I think this might be help since many players especially stronger ones don't have classical rating (on lichess), and some people might have it but the rating might be outdated. And since for chess.com there is not classical anyway and you have to use rapid, maybe taking lichess rapid rating into account would make sense.
Adding more explanatory variables would make the model more complex, and it looks like there is not a ton of training data, so there could be overfitting concerns, but I still wanted to share the idea.
Interesting. Not sure what it all means.
@Akavall No, I'm not working on that app. I've helped the ChessDojo Training Program (https://www.chessdojo.club/) revamp their internal cohorts. I want to say that your link was a third-party application developed before the website automatically told people what cohort they should be in. Things have advanced now - the website automatically sorts you into the correct cohort based on your lichess, chesscom, USCF, FIDE, ACF, ECF, etc.
There are no plans to integrate blitz and rapid ratings into this model because the Dojo is all about improving at classical chess. It asks you to work on your classical ratings.
If only chess platforms could agree on implementing a letter category system for their own chess ratings, players could then more easily understand their skill levels. Finding common ground in the lettering category would be a great start.
Great work!
At this point however, I do wonder whether comparing online blitz to otb classical would be more reliable. In speaking to fellow chess club members, it seems their blitz rating is a fairly reliable indicator of their classical strength, as surprising as that may sound.
Meanwhile, I must also link this which at least for me, seems to map quite well (although it has no control for rating deviations): https://chessratingscomparison.com/Graphs