Does Glicko Rating System Work Well With CC?

i'm not opposed to tweaking the RD settings some. i am currently at RD 81. i play a lot of chess here (obviously). what should my RD be? what would be more fair in your opinion?

here is my current rating graph of 1 year:

does that fluctuate too much? i dunno. i know that my actual SKILL doesn't change that much, but i know that the amount of effort i put into my games, and the amount of stress i feel, does fluctuate that much. plus, i know i have gotten a lot better with all of the study and play that i do.

personally i am not a ratings watcher. i don't care about my rating as anything more than a rough indicator. perhaps that is reflected in our loose RD. maybe we should tigthen it up. but with ratings you can never win :) people want their ratings to go up faster, and down more slowly. right now we get some complaints of people's ratings dropping too much, but we never get complaints about them rising :) if we change it, we might start getting more complaints. i dunno.

we've also complemplated shaving 100 points off of everyone's rating and making the starting rating 1000. but if we do that, and we lower the RD... how many games before an 1800 rated player is accurately rated?

Well, if you shaved everyone's rating by 100 points, the 1800 would be as accurately rated as they were before the shave, because the ratings are relative. If you dropped the starting rating by 200 at the same time, that would merely mess up the old player's ratings relative to the new player's ratings. I have no idea how long that would take to balance out. It would depend on the rate of new players coming in and the number of games per rating period. I could try to run some simulations on that.

I haven't thought about the ratings deviations as much, so I'm not sure of the effects of lowering everyone's RD. I have been wondering about the length of the ratings period compared to the typical rate of play here. If the ratings period is a day, but I'm only finishing a game every two or three days, my RD is going to keep floating up because of the periods where I don't have a completed game. If the rating period is a week then that's not a factor. 

erik, I think the problem is not people whining about their ratings going down, but quite simply the question of accuracy.

It's great having tournaments for 1400-1600 players, etc, but if the ratings are unnecessarily inaccurate, this is undermined.

You're right that effort varies greatly, and this suggests a higher RD than would otherwise be best. My guess is that it is too high even for that.

Can I suggest an experiment. Calculate what everybody's ratings would be today if the RD increased, say, 5 times more slowly. And see if those ratings more accurately predict match results than the current ratings do.

I'm beginning to change my mind on this topic.

My last 10 games:

Average opponent rating: 1927. Results: 8 wins, 2 draws

Starting rating = 2468. Current rating = 2424

I lost 103 points from my 2 draws (against opponents rated 2015 and 1813), but because they were my first results in a while, my RD had increased significantly. My wins all came in quick succession, so my RD kept falling, making each win worth less. It seems a bit tough that with this record I've actually dropped 44 ratings points overall. It looks like I'm going to draw another couple of games against 2000-2100 players, so my rating is going to take another hammering.

Conclusion: I think the constant for calculating the RD should be reduced for turn-based chess. I'll no doubt change my mind when the current system works in my favour...

@jonnyjupiter: A 2468 in a ten game match with a 1927 has an expected score of 9.5. Therefore 8 wins and 2 draws is a slight underperformance (just based on the average), and a slight ratings drop is reasonable.

ok - looking for INTELLIGENT replies to this topic:

http://www.chess.com/forum/view/community/math-people-only-changes-to-how-much-ratings-change

basically, i think that a target RD for someone who plays a LOT of chess on Chess.com should be 30-50. i play quite a bit and i'm 81.

come give your opinion!

Thanks for that - it does help me to understand it. I did feel that I was playing poorly for one of the draws, so if I had won that my rating would have stayed pretty much at a level.

Fair enough then - I underperformed and thus my rating dropped appropriately. Hey ho.

I play very much, can anyone please tell me what's my RD? I guess about 70.

Here u go mate! :-)

What a good guess from me!

Hahaha Yes, very good guess indeed!

Maybe it would be even better if asked a bit earlier :)

Yeah anytime u want just ask. :-)

I don't need it too often.

I'm kinda angry at the worst losee because my opponent kept moving at very little time left, sometimes a few minutes in a 3-day game to prolong not-losing (I was a queen up) and I timed out when I had to go on a tourney. Bad luck.

That does suck. I also hate losing or drawing a game that i should have won. it shows that my technique still needs work i believe. Or playing so badly you wonder if u should give up the game hahaha.

But losing by blundering and by timeout is something different.

Oh yeah! That's a league of its own. Almost enough to want to kill yourself i reckon.

http://www.chess.com/echess/profile/haddad56

What's his RD? May be the lowest on this site.

I didn't read through all four pages, but I'd say we really need Glicko. People of all abilities come and start playing, a 1200 could be a complete beginner, or a master-level player. You don't get this sort of unpredictability of newcomers in OTB play. We need a system that assigns an accurate rating quickly.

 47..

AWARDCHESS is 46 i believe.