Do you know any very strong player with insight made public? Or website that generate insights?

There is a common trope that says that a the difference between a 2700 player and a 2300 player is much bigger than the difference between a 1400 player and a 1000 player. I personally think it's horseshit and I want to prove things either way.

I, and the people I have in mind who perpetuate the trope, are not talking about the amount of work or genius it takes to go from one rating to another, thy are genuinely saying the 1000 player will win 1 time out of ten against a 1400, but the 2300 will simply never ever win against a 2700. I on the other hand, put my thrust in the Elo rating system, and a 400 point difference is a 400 point difference, the weak player is expected to win 8 out of 100 times either way, or they wouldn't be at the rating they are. Mind you, I (and the rating system) counts draws as half wins, so the 2300 player only really need to draw 16 games out of 100.

Now there is a simple way to test who is wrong. Look very strong player and see whether, on average, they win points against much weaker opponent. Any rating system isn't perfect, and tend to lose in precision as the difference between player grow, so it might very well be the case that 2700 seldom ever lose to 2300. But if that is the case, since they'd win more than the system expected, they should win rating when playing against much weaker player. If they do, I was wrong, if they don't, I am right.

Which is why I am looking for profiles with public insight available, or even just a screenshot of the preset "do I win more rating points against weaker or stronger opponent". High ranking preferred, but any welcome. If for example everybody gains a bit more Elo against weaker player, it's also valuable information, it'd mean it's not just a high Elo thing. High Elo or high Glicko, the interpretations should be the same.

Alternatively, if someone know of an app which can generate this kind of insight, I am also interested, because for some reason, while the default setting for insight is private, games played are public and anybody can mine the games of anyone. Fortunately, I a competent enough to bake my own script and generate that insight myself, but since it's already been invented, I'd rather borrow someone's wheel rather than make my own.

You can just go here lichess.org/player and see for yourself.

> I on the other hand, put my thrust in the Elo rating system, and a 400 point difference is a 400 point difference

This is known to not be the case in chess. The actual and theoretical win rates don't match. For example see this : old.reddit.com/r/chess/comments/ugu1e2/a_fat_tails_flaw_of_the_elo_system/
which seems to be exactly the data you are looking for.

Mine’s public, but not that many tournament games.
From the few I have, I tend to agree that it’s easier to gain rating against higher opponents as you only need 1 titanic blunder every 10 games to draw even (which will basically always happen) so it’s almost impossible to lose rating if you catch a decent start or the field out-rates you by enough.

@f4xel said in #1:
> There is a common trope that says that a the difference between a 2700 player and a 2300 player is much bigger than the difference between a 1400 player and a 1000 player. I personally think it's horseshit...

Regardless, it's most likely true.

Air thin at the top. That’s why I stay low rated, saves me allot of money for asthma spray.

@f4xel said in #1:
> There is a common trope that says that a the difference between a 2700 player and a 2300 player is much bigger than the difference between a 1400 player and a 1000 player
Only in terms of competition and not on terms of strength, if you know what I mean!

An objective way of looking at this... do the 2700 rated players risk more or less often than the 1400 rated players who are playing lower level of competition. If people are dodging more... there is a reason. Super GM's rarely play in events without their peers for a reason.

Now the trope is true, a 2700 probably knows a heck a lot more than 2300 vs the lower rated opponents. After all, 1000 rated the moves are so bad they almost constitute randomness and 1400's will still actively blunder away free material without any real force being applied. But that is different than the practical nature of who is more likely to get a good result against the higher rated player.

@f4xel said in #1:

> Alternatively, if someone know of an app which can generate this kind of insight, I am also interested,

You could use scid vs pc, then load a large database (e.g., caissabase) and filter games where white player is 2700+ and black is in the range 2300-2400, and see the statistics. (For example, first see how many games are there, then filter the ones in which black draws, then filter the ones where black wins. I think it should not take too much time.) Then do the same when white player in the range 2300-2400 and black is 2700+, and see the statistics. Then do the same for 1400-1500 vs 1000-1100.

Only one difficulty I see is that it may be difficult to get data of under 1400 FIDE rated players. I don't think caissabase has games of 1000-1500 rated players. But you could do all the above steps for lichess database. Even though lichess and FIDE ratings are different, you may get some idea as to whether a difference of 400 at 2300+ is different from the difference of 400 at lower levels.