Your network blocks the Lichess assets!

lichess.org
Donate

Rating inflation in Lichess!

I feel that there is an inflation in the bullet rating in Lichess. I think about 200 bullet rating points should be reduced from everyone. Is there another way?

I feel that there is an inflation in the bullet rating in Lichess. I think about 200 bullet rating points should be reduced from everyone. Is there another way?

There would be a terrible fuss if 200 points were removed from everyone's rating at a stroke. :-)

It can be done subtly though over a period of time if it is necessary.

There would be a terrible fuss if 200 points were removed from everyone's rating at a stroke. :-) It can be done subtly though over a period of time if it is necessary.

The median of bullet ratings seems to be ~1540 at the moment. I can't estimate the mean value as easily but I seriously doubt it would be as high as 1700. Where does this constant of 200 come from?

The median of bullet ratings seems to be ~1540 at the moment. I can't estimate the mean value as easily but I seriously doubt it would be as high as 1700. Where does this constant of 200 come from?

Lichess rating have slow pull toward mean on 1500. According to commit in version control it should pull medina to 1500 in about 6 months

So there can be inflation/deflation if player pool changes significantly. Though seem highly unlikely. Sudden burst of lot weaker players enter the pool and median corrector would start pulling rating up to keep it in 1500. I mean that amount new bullet player... can happen but I do doubt

Lichess rating have slow pull toward mean on 1500. According to commit in version control it should pull medina to 1500 in about 6 months So there can be inflation/deflation if player pool changes significantly. Though seem highly unlikely. Sudden burst of lot weaker players enter the pool and median corrector would start pulling rating up to keep it in 1500. I mean that amount new bullet player... can happen but I do doubt

Why doesn't lichess balance mean puzzle rating to 1500?

Why doesn't lichess balance mean puzzle rating to 1500?

Even if it does, I suspect the main problem is that in puzzles, you "play" against puzzles, not other users, therefore the "pool" consists of both users ad puzzles. And as majority of puzzles are easy, it's likely that puzzle ratings tend to be lower and user ratings higher.

Even if it does, I suspect the main problem is that in puzzles, you "play" against puzzles, not other users, therefore the "pool" consists of both users ad puzzles. And as majority of puzzles are easy, it's likely that puzzle ratings tend to be lower and user ratings higher.

@dreamsmorning48 said in #1:

I feel that there is an inflation in the bullet rating in Lichess. I think about 200 bullet rating points should be reduced from everyone. Is there another way?

I am not sure. For most people I check, their bullet rating is lower than blitz or rapid.

For some (and you, obviously), it is quite the opposite. I can only speculate on the reasons, like very decent mouse skills, maybe very fast reaction times, and effective chess play tailored for very short time controls. (While most players - myself included - would mostly play the same stuff across all time controls.)

With your rather high rating, your exposure to similar players might be higher than on average.

@dreamsmorning48 said in #1: > I feel that there is an inflation in the bullet rating in Lichess. I think about 200 bullet rating points should be reduced from everyone. Is there another way? I am not sure. For most people I check, their bullet rating is lower than blitz or rapid. For some (and you, obviously), it is quite the opposite. I can only speculate on the reasons, like very decent mouse skills, maybe very fast reaction times, and effective chess play tailored for very short time controls. (While most players - myself included - would mostly play the same stuff across all time controls.) With your rather high rating, your exposure to similar players might be higher than on average.

@dreamsmorning48 said in #1:

I think about 200 bullet rating points should be reduced from everyone

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).

@dreamsmorning48 said in #1: > I think about 200 bullet rating points should be reduced from everyone 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).

@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).

This is a very good point. Many people think of those ratings as an absolute measure. But by design they are not;they can only predict expected outcomes based on the rating differences between the players.

Hence, it is even a long stretch to compare players over different times (like all this "we had so many players above 2700 at such and such time").

But the human mind is what it is, and likes to interprets things in absolute terms, giving meaning to different rating levels on different platforms, or different time controls. Sure, it would be nice to have the same rating on all platforms to make comparison easy...

But comparing for example bullet rating to standard rating: what would it even mean if you have the same rating? Do you play equally strong? What does that mean? Chess quality and making the same moves? Obviously not... and considering that people are good at one thing but not necessarily at another, and the pool of players may also different, there seems to be no way to bring them into any kind of symc.

@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). This is a very good point. Many people think of those ratings as an absolute measure. But by design they are not;they can only predict expected outcomes based on the rating differences between the players. Hence, it is even a long stretch to compare players over different times (like all this "we had so many players above 2700 at such and such time"). But the human mind is what it is, and likes to interprets things in absolute terms, giving meaning to different rating levels on different platforms, or different time controls. Sure, it would be nice to have the same rating on all platforms to make comparison easy... But comparing for example bullet rating to standard rating: what would it even mean if you have the same rating? Do you play equally strong? What does *that* mean? Chess quality and making the same moves? Obviously not... and considering that people are good at one thing but not necessarily at another, and the pool of players may also different, there seems to be no way to bring them into any kind of symc.

@Panagrellus said in #8:

The absolute value of ratings is meaningless
That's not completely true for Glicko-2 which has a fixed initial rating value (1500 on lichess), also used for initial normalization in the calculations, and ratings should be centered around it. That being said, being off by 40-50 points of rating - which is the case currently in bullet and rapid - isn't so much that it would require an immediate action.

And as I said in #3, I definitely don't think that the bullet ratings as a whole are off by as much as 200.

@Panagrellus said in #8: > The absolute value of ratings is meaningless That's not completely true for Glicko-2 which has a fixed initial rating value (1500 on lichess), also used for initial normalization in the calculations, and ratings should be centered around it. That being said, being off by 40-50 points of rating - which is the case currently in bullet and rapid - isn't so much that it would require an immediate action. And as I said in #3, I definitely don't think that the bullet ratings as a whole are off by as much as 200.

This topic has been archived and can no longer be replied to.