What's the Glicko RD system about on chess.com?

[MJMcCready]

What's this about in the diagram below?

Captddure.PNG

[Harry Duff]

According to Wikipedia, a rating system intended as an improvement on the ELO system:

https://en.wikipedia.org/wiki/Glicko_ra ... 0deviation.

[Wadih Khoury]

It's some sort of confidence interval indicator.

[MJMcCready]

It strikes me as somewhat dubious that the score goes up the longer you are away from the game, based on the account that you are training or studying. If we were all professional players, I can understand that. Like many I can go many weeks without playing a single game and letting chess enter my thoughts at any point.

[Roger de Coverly]

It strikes me as somewhat dubious that the score goes up the longer you are away from the game, based on the account that you are training or studying.

It's an intrinsic feature of the Glicko method that the uncertainty about the reliability increases with absence and presumed lack of practice. That might even be true, but it's a flawed assumption when the rating system does not have a monopoly on a player's chess activity.

[Ian Thompson]

It strikes me as somewhat dubious that the score goes up the longer you are away from the game, based on the account that you are training or studying. If we were all professional players, I can understand that. Like many I can go many weeks without playing a single game and letting chess enter my thoughts at any point.

The number goes up to indicate increasing unreliability in your rating. It causes a bigger change in rating when you do play again after a long gap. Its effect is like having a standard Elo rating system and gradually increasing the k factor the longer you go without playing any games.

[MJMcCready]

ok, well I assumed that the longer you go without playing the rustier you become. If I don't play for a few months, I'm generally very off-form.

[Wadih Khoury]

ok, well I assumed that the longer you go without playing the rustier you become. If I don't play for a few months, I'm generally very off-form.

That is indeed what Glicko does.
It states that since you were off for a long time, either you got rusty or you were studying, hence your grade is more uncertain.

[MJMcCready]

But from what I understand, the longer you go without playing the more your rating is likely to increase because it reflects an unquantifiable increase in ability. But as stated, I usually worsen if I don't play regularly. I don't see how absence from the board should increase k-factor, shouldn't it limit it?

[Wadih Khoury]

I think the Glicko number you highlighted is not a rating. It's a variance number.
The lower it is, the more accurate your grade is.
One way to think about it, is that a glicko of 200 means your chess.com grade is irrelevant as you are assessed to be +/- 200 from your grade with x% confidence.

[Roger de Coverly]

But as stated, I usually worsen if I don't play regularly. I don't see how absence from the board should increase k-factor, shouldn't it limit it?

If they increase the K factor, you drop more points when you lose. It's a flawed idea when the rating system doesn't have a monopoly on play because you equally gain more points when you win. So someone playing exclusively on lichess for a few months and then returning to chess.com could obtain a higher rating for the same performances than someone continuously on chess.com.

[MJMcCready]

But as stated, I usually worsen if I don't play regularly. I don't see how absence from the board should increase k-factor, shouldn't it limit it?

If they increase the K factor, you drop more points when you lose. It's a flawed idea when the rating system doesn't have a monopoly on play because you equally gain more points when you win. So someone playing exclusively on lichess for a few months and then returning to chess.com could obtain a higher rating for the same performances than someone continuously on chess.com.

Yes it seems so. Bit strange.

[Michael Farthing]

But as stated, I usually worsen if I don't play regularly. I don't see how absence from the board should increase k-factor, shouldn't it limit it?

If they increase the K factor, you drop more points when you lose. It's a flawed idea when the rating system doesn't have a monopoly on play because you equally gain more points when you win. So someone playing exclusively on lichess for a few months and then returning to chess.com could obtain a higher rating for the same performances than someone continuously on chess.com.

Not really, Roger. The k factor reduces smoothing. If you actually have not changed in ability then your rating might become more jittery for a while but there should not be any trend present either upwards or downwards.

[Roger de Coverly]

If you actually have not changed in ability then your rating might become more jittery for a while but there should not be any trend present either upwards or downwards.

Jittery = incorrect rating. My point. It may not matter on chess.com, but would if applied to a national rating system because of tournaments with rating restrictions or rating qualifications.

[MJMcCready]

I don't know how it could possibly be accepted.