The Chess.com Theory Disproven

@kajalmaya said in #2:

• your average opponent rating on that site is 1267 while your rating is 1483; what's going on?

he starts with very few rating points there and has to work his way up (no provisional ratings on chess.com).

@Rookitiki said in #11:

he starts with very few rating points there and has to work his way up (no provisional ratings on chess.com).

But my point is this: he has 250+ games on the site, his rating is 1487 and average opponent rating is 1267. This means a lot of his games are against much lower rated players. This suggests that his rating 1487 is not reliable. I would trust it if his average opponents were roughly equal in rating and he scored roughly 50% against them. This is my expectation if I want to believe that someone's rating is not farmed and is stable, although I wouldn't expect it for people on the extreme of the distribution (like Magnus).

Although, such claims are difficult to prove, a numerous amount of signs point to this true. When I used to play standard, my chess rating on lichess was about 1300 whereas my chess.com rating was 900. I think some other chess players have encountered the same effect, but not enough to run a test. On lichess, chess.com is called mud.com and people are discouraged to play there. On chess.com, they filter out the word lichess and other chess websites to stop advertisement and... Yes customers going to other sites. So there is not a lot of average and low rated chess players that play on both. There is more professional chess players switching from sites since they are expected to play a lot of chess, and also why not, but I don't think a lot of them would be smiley about their games being used in a rating survey, and not on the front cover of chess vogue (disclaimer Incase some gets angry: I'm only joking, I would want that anyway).

Anyway, bla bla bla, you get it, it's a messy subject!

you get a ton of elo from chess.com though (but you lose a lot)

i have to revert my point about the different rating systems, since i noticed even tho i have some bigger samples as the account in question, my average opponent is always at least 100 points lower, sometimes even around 250.
the time controls with most samplesize (and least improvement) show the least deviation from the current rating.

guess its just a natural occurence / math:

• either you improve and thus the avrg opponent rating will be below the current rating. the larger deviation will be a sign of improvement. the average will always stay behind as long as you keep progressing
• or you get stuck and the difference becomes smaller, but it will take forever.

there is also the option of you getting worse, that would be the only way to get close to the average opponent rating with less than tens of thousands of games i think (in quick match at least - check @german11, his avrg opponent is 700 points higher through tournaments).

@kajalmaya said in #12:

But my point is this: he has 250+ games on the site, his rating is 1487 and average opponent rating is 1267. This means a lot of his games are against much lower rated players. This suggests that his rating 1487 is not reliable. I would trust it if his average opponents were roughly equal in rating and he scored roughly 50% against them. This is my expectation if I want to believe that someone's rating is not farmed and is stable, although I wouldn't expect it for people on the extreme of the distribution (like Magnus).

I think the point is that his average is dragged down by all the very low rated people he faced in the beginning of the climb. As long as he had a decent number of games at his own level at the end of the climb it should be fine.