Rating inflation in Lichess!

Absolute value is meaningless in sense that it does not tell your strength in a another pool. 1500 you are median lichess player inthat time control pool. Chess.com starting rating is 1200 (or at least was) making 1200 an average chess.com player. We do not have information how good players are compared on different pools so we cannot really say if 1200 c.com is about same as 1500 here. Could be so and can change in future. and as pointed above even comparison between different lichess time controls is not one-to-one nor it is intended to be

Agreed. What I meant was that I can imagine a situation when it would be actually desirable to make such transformation, whether it would be a simple shift or something more complicated. In particular if actual rating distribution was significantly different from the assumption the algorithm design is based on.

IMHO the problem here is the perception, namely a combination of OP being one of strong players and bullet rating distribution having higher standard deviation than slower time controls. But that may be just a natural consequence of bullet skills being spread wider.

My comp can barely run Chess .com but this site runs perfectly & is much faster when making moves.

The size of the pool compared is much smaller, so ratings will be always be higher on other sites...but there are less bodies stuck in 'rating hell' here :)

@Panagrellus said in #8:

Why? The absolute value of ratings is meaningless, so what would be the purpose of this proposed operation?

(Note the difference to the "compression operation" FIDE did recently, which dealt with an imbalance between weaker and stronger players).

Making the ratings closer to FIDE would be convenient so that you don't have to specify when talking about ratings.

@susiety said in #14:

Making the ratings closer to FIDE would be convenient so that you don't have to specify when talking about ratings.
It would be convenient but I don't think it's possible. For various reasons, there will always be people with similar FIDE rating but very different lichess ratings - and also people with similar lichess ratings but very different FIDE ratings.

Also, I suspect that when you ask "What is your FIDE rating?" without saying which one, most people will reply with their standard (classical) rating but when you ask the same about lichess rating, you will most often get blitz.

@susiety said in #14:

Making the ratings closer to FIDE would be convenient so that you don't have to specify when talking about ratings.

Every rating only has meaning compared to the other players in the same pool. Even if the very same system was used (say, FIDE switching to Glicko-2 with a starting value of 1500, just as on this site), the rating numbers would still not be comparable, because ratings are established relative to the other players - and arguably players in FIDE OTB tournaments are, on average, stronger than the average Lichess player.

@mkubecek said in #3:

Yes, that's r something like 1550; But it should be compared with the past (for example, 5 and 10 years ago). My opinion is mostly based on my 8-year playing experience on this site. I remember that a player with +3000 rating was very rare. But now, with this trend, we should see +3300 in the future. Subtracting something around 200 is just my estimate.

@petri999 said in #4:

I think this system has worked in other variants and time controls, but it seems that it has accumulated and inflated in bullet and to some extent blitz. I feel this even in my rating and other players that I have known for a long time. Overrating has actually occurred.

@nadjarostowa said in #7:

Thank you for your comment. In this regard, it was completely correct and logical, and this difference is related to the skill of the person; But my point is that, for example, when I was playing about 8 years ago, I rarely saw +3000 bullet rating ; But now I feel that such a rating is not attractive anymore. Of course, it is clear that one of the reasons was the arrival of a large number of GMs, IMs and top players, but I am unlikely to know that it is the only reason for this increase in ratings. In addition, most of these high-level players have much higher bullet ratings than Blitz, which is significant.

For a shift like that to be appropriate, the whole distribution would have to be shifted relative to the mean value (1500). Looking at the distribution graph, that doesn't seem to be the case.

Having more players in the right tail (assuming it's really more in the terms of percentiles, not just number of players) doesn't necessarily mean something is wrong. It can be result either of stronger players playing bullet on lichess than before - or of an influx of weaker players, i.e. average player being weaker than in the "good old times". Or it can be a combination of both. But the point is either would be actually correct and trying to "fix" it would actually do more harm than good. Shifting the ratings would break the mean value assumption and "squeezing" them towards 1500 would break the relation between rating difference and expected mean result.

As others already pointed out, only ratings within the same pool at a given moment can be compared meaningfully. But a rating of e.g. 2000 in 2014 and 2024 do not necessarily represent the same playing strength because the pool may have changed a lot.