Is lichess rating too linient?

@JasonNewst
The stupidity of things said these threads is just too much. It just makes me want to weep. Man, I just said in my above post that they are relative.
Can you not read it? Here I will quote it for you again.
"Of course they are relative. You need opponents to compare against. Who can't live with the fact they are relative?"
Yes ratings are relative! What has that got to do with it? Speed is relative and we measure speed don't we? But 100 people can measure speed in 100 different units and create a lot of confusion completely unnecessarily. That's exactly what Lichess is doing here by intentionally making them inconsistent with other organizations. I ask you why create confusion, how does that benefit anyone?

"Rating does not work like that at all, I mean mathematically it would require lots of manipulation"
Lol, "lots of manipulations" Have you ever seen the equations for ratings calculations? Unless you have a degree in mathematics don't even bother trying to understand them. True the conversion difference is not a constant number so you can't just subtract 200 for example. But it's only a LITTLE more complicated than that. You adjust a different amount for each bracket that can be determined by existing data sets. So for 15xx you subract 250. For 17xx you subtract 100 etc. You can make it more granulated than that but that's the concept right there. It's infinitely less complicated than the equations they use to come up with the ratings in the first place. I would volunteer to write the algorithms if Lichess would ask me. I could do it in 2- 5 minutes.

"There's really no meaningful correlation here other than stronger players tend to have higher ratings compared to others in the same pool."
That's because Lichess CHOOSES to have them not correlate when they could. That's the choice that I'm criticizing here.

@sheckley666
"Why should lichess go all this trouble?"
As I mentioned in my above post, it's no trouble at all. I, a lay person, could write the algorithms in 5 minutes.

"From where to get reliable data for the calibration?"
Lots of sources this for example ,
imgur.com/a/nWy4x
and this,
docs.google.com/spreadsheets/d/1GcVG_slcTUJfi0sQMeaxes_jwYFBrfWaR1ejSSOH-SM/htmlview

"And why compare playing short time controls on a flat monitor with long time controls on a 3d board?"
In case you haven't noticed Lichess has different ratings for different time controls.
Yes your setup will affect it with factors like 3d boards. That's fair enough. But that is a problem that exists regardless of whether we make the ratings correlate with other sites or not. . For the people that I'm arguing with who say that we should only view ratings as a way to compare players in the same pool, that argument is just as much of an issue. Another player on this same website you are on who has a rating 100 points lower than yours might actually be a better chess player than you , just that his internet is wonky. There are even confounding factors in IRL OTB games. So we can either choose to accept that limitation or do away with the idea of ratings all together. But it's hypocritical to apply that argument to one way of using ratings and ignore it for another. And it should be noted that this kind of variance is greatest in short time controls but diminishes with longer time controls. It's not going to matter much what kind of board you use or if your internet lags a little in classical games.

@ThighBald

Your analogy with speed is false but revealing. That's my point, You seem to think there is a "frame of reference" that's equal for everyone. Why are we able to calculate the speed the same way all over the world ? Because we all use the ground as the frame of reference.

A car in Canada goes 70 kph, another car in Japan goes 60 kph, so obviously the first car is faster. We understand this because everyone is measuring compared to the ground, that's our common frame of reference.
But what if I said a car in Canada is going 70 kph faster than the slowest car in the race??? If we take the slowest car in a race as frame of reference we can still measure speeds compared to one another and rank the cars in that race but we cannot make comparisons across races.

Unfortunately in chess we don't have a 'common frame of reference' we can only compare players against each other. There's no absolute value for chess. All rating systems compare players against one another in their active pool of players.

chessgoals.com/rating-comparison/

This is an interesting list. At lower echelons lichess blitz ratings correspond better with FIDE ratings but in the mid range chess.com corresponds better with FIDE while lichess shows higher equivalent ratings. However for ratings above 1900 chess.com passes lichess and shows more inflated ratings.

As you can see it's not really that simple!

@ThighBald
1 The algorithm is indeed the smallest problem
2 I don't know how much effort was put in these graphs and tables. Looks like quite an amount. And you have to repeat this work permanently because - of course - the rating pools may very well develop differently.
3 Yes, lichess has different ratings for different time controls, as has FIDE. Unfortunately, they do not match very well.

@ThighBald
Do you even look at your graphs? according to the graph you posted lichess blitz ratings correspond reasonably well with Fide ratings, what else do you want from them!? lol

Why should lichess ratings have any relationship at all with FIDE ratings or USCF or other sites? I think lichess should set the average or median player at a rating of say, 4000. Then the lowest might be about 1000, and the highest around 15,000. That would make the point that it has absolutely nothing to do with FIDE ratings.
Or set lichess median or starting rating at 40. Then anything over 100 would be something to brag about.
As others have said, there is no such thing as a natural measure of chess skill. There is only a comparative measure within a particular given pool, in a particular type of competition, based on an artificially created scale.

Learn to live with a little ambiguity.

And regarding speed and the ground as reference, Einstein pointed out that even those measures are relative.

I agree with @sparowe14
It is confusing that most servers and associations use rating systems that look similar.

@ThighBald
"""There's really no meaningful correlation here other than stronger players tend to have higher ratings compared to others in the same pool."
That's because Lichess CHOOSES to have them not correlate when they could. That's the choice that I'm criticizing here.
""
I dont have the math concepts in order they do correlate(so people who say they dont are outright wrong) but even if lichess would make some fance formulas make tha actual numberat same range they would not correlate any more or less. Following two vectors have correlation of 1 i.e maximum. for such cases simple conversion formula would make exact predictions
1 2 3 4 5
3 6 9 12 15

But correlation israther low to understand what is the real problem have glimpse on following diagramm
imgur.com/a/nWy4x
if prior to rapid/classic split so lets look at the blitz. take FIDE 2000 vertical and obselve where cloud get thick it is about 1800 and where it gets not thick and thats about 2200.

So we can estimate that 400 point span.

@sparowe14 , I'll suggest one step better. Set lichess ratings to a unity of 1.0 . All weaker players' ratings approach 1.0 but do not ever reach it. My blitz rating would be a fantastic looking 0.0001234.

Nice.

Ok guys let's be clear, each platform uses a different system: Lichess uses glicko2; Chess.com uses Glicko; Fide uses Elo.

This argument makes little sense because discussing about changing one system for another for no reason (besides they are doing it) isn't smart.

It would be much more interesting to discuss the perks and cons of those systems (especially since they are officialy documented). And then maybe why or why not lichess should change.