@Lachesis I also agree with him. He has a very good point. It is indeed true that the Lichess rapid ratings are not as accurate as those no chess.com compared to USCF OTB Rapid ratings. For instance, my lichess rapid rating is 1670, peaking at 1718. However, my chess.com rapid rating is 1554. My USCF quick rating is 1350, which indeed proves that chess.com is way more accurate than lichess. Very great point, @Lachesis
@justthere said in #1:
> ... why ? ...
If you look into 'the Gaussian distribution' you will get a better feel for how it all works ...
> ... why ? ...
If you look into 'the Gaussian distribution' you will get a better feel for how it all works ...
You can't compare Lichess, chesscom, FIDE, USCF ratings. See lichess.org/page/rating-systems
Let's put it this way, your rating here is measured in, let's say, oranges. On chesscom in apples and USCF's rating in bananas.
Does it make sense to say that your 3 oranges are bigger than 2 bananas or 2 apples? Not really, because oranges, bananas and apples are different. Does it make sense to say that 2 apples on chesscom are the same as 2 USCF bananas. Nope, even though the number of apples and bananas is the same.
Let's put it this way, your rating here is measured in, let's say, oranges. On chesscom in apples and USCF's rating in bananas.
Does it make sense to say that your 3 oranges are bigger than 2 bananas or 2 apples? Not really, because oranges, bananas and apples are different. Does it make sense to say that 2 apples on chesscom are the same as 2 USCF bananas. Nope, even though the number of apples and bananas is the same.
@Lachesis said in #14:
> My question to bufferunderrun is why chess.com rapid ratings closely compares to those of the US Federation, considering that chess.com are oranges compared to Federation bananas.
My guess is that they have someone on staff, constantly tweaking the rating parameters, trying to achieve the illusion of apples.
I think lichess should change to something that looks nothing like apples at all. Maybe a fraction, with the median at zero, GMs around 1.0, and novices negative. That way, no one will be tempted to compare.
> My question to bufferunderrun is why chess.com rapid ratings closely compares to those of the US Federation, considering that chess.com are oranges compared to Federation bananas.
My guess is that they have someone on staff, constantly tweaking the rating parameters, trying to achieve the illusion of apples.
I think lichess should change to something that looks nothing like apples at all. Maybe a fraction, with the median at zero, GMs around 1.0, and novices negative. That way, no one will be tempted to compare.
I think chesscom uses Glicko1 with modified initial rating and USCF is modified Elo, if I'm not mistaken.
So different systems, they won't give you the same numbers. What you likely see is coincidental amount of apples and bananas being the same and then think that they are correlated.
So different systems, they won't give you the same numbers. What you likely see is coincidental amount of apples and bananas being the same and then think that they are correlated.
OK, sure, certainly ... but populations are always 'normally distributed' so that's why the 'canonical mean' is 1500 ...
@Chess_Galaxy1078 said in #9:
> EDIT: Please mind any spelling mistakes. I wrote this in a hurry.
Thank God you had at least enough time for explaining that!
> EDIT: Please mind any spelling mistakes. I wrote this in a hurry.
Thank God you had at least enough time for explaining that!
LOL
the goal of ratings on lichess is to predict winning chances for chess games between two players. it is not to predict uscf or any other otb ratings. the same btw is true for chesscom ratings and the otb ratings themselves.
thus, you would call ratings "accurate" when they accurately predict those winning chances. and *not* if they closely follow some other rating system. and again, the same is true for otb ratings: uscf ratings are "accurate" when they are doing well in predicting winning chances, *not* when they closely follow fide ratings.
thus, you would call ratings "accurate" when they accurately predict those winning chances. and *not* if they closely follow some other rating system. and again, the same is true for otb ratings: uscf ratings are "accurate" when they are doing well in predicting winning chances, *not* when they closely follow fide ratings.
This topic has been archived and can no longer be replied to.