I'm wondering why the Glicko Rating System was selected for this site, a correspondence chess site?  I'm also wondering how many others are finding it impossible to play enough games in order to get a low RD and therefore are forever doomed to high fluctuations in rating?

i don't like the rating system used here. If I lose to a person rated the same as me, I lose 20 points, where as if I win to somebody rated the same as me, I get 10 points added. Within the last week, I have lost more than 150 rating points!!

What is the Glicko rating system, and what does RD stand for?  Pardon the ignorance.

Google: Glicko

 The Glicko rating system is a method of determining chess ratings.  It uses a factor called RD (rating deviation) to attempt to determine your playing strength based on how active you are.  The more games you play in a given time period, the lower your RD is and therefore, the less your ratings will fluctuate.  My question is whether this system is right for correspondence chess? 

What Glicko himself had to say:

"The system does not conserve rating points - and with good reason! Suppose two players both have ratings of 1700, except one has not played in awhile and the other playing constantly. In the former case, the player's rating is not a reliable measure while in the latter case the rating is a fairly reliable measure. Let's say the player with the uncertain rating defeats the player with the precisely measured rating. Then I would claim that the player with the imprecisely measured rating should have his rating increase a fair amount (because we have learned something informative from defeating a player with a precisely measured ability) and the player with the precise rating should have his rating decrease by a very small amount (because losing to a player with an imprecise rating contains little information). That's the intuitive gist of my extension to the Elo system."

Basically, the more you play, the smaller the fluctuation in your rating.

That would make sense if one game was an accurate sample of your playing strength.  The fact is we haven't leaned anything (or only very little) informative in the situation described because the sample size is so small it's statistically insignificant.  Lets make coin flip analogy, Flipper A has a long history of flipping, and it is known that he wins a coin flip around %50 of the time.  Well here comes Flipper B.  He's new to the flipping scene and we don't know much about him.  if he wins a flip against Flipper A is it reasonable to expect him to win %100 of future flips?  Of course not!  He only fliped one flip!  We don't have enuogh information to make that judgement.  If he continues to win all flips THEN we would have something to base such a claim on.  This is probably the stupidist thing I've ever writen.

Not many comments.  Nobody has any thoughts?

I think fluctuations show that the rating system is very accurate.  Human beings do not perfrom in as such consistant a fashion as we would wish to belive about ourselves.

I think the Glick system may be more accurate than the current ELO system as employed by the USCF in which they have created artificial floors inflating the entire rating tableau upwards.

I am not arguing it is a panacea, only that it actually seems to be more accurate than people may wish and that there is no real alternative that has proven to be better, more reliable, or more accurate.

Also, I do not think you will get too many comments as most people do not take the time to understand it or learn how it works.  There are always legions of strongly un-informed opinions, but even here, I think it is rather confusing to most people and I admit i do not have my own brain wrapped around it entirely.

Then again, and most people willingly confirm this, I am an idiot.

True.  My question, though, is whether the Glicko Rating System is appropriate for correspondence chess given each game takes much longer to complete and unless you play many games all at once, you will always have a high RD and therefore never an accurate rating!

That's because you don't understand the system.  I doubt anyone on the entire site has both the same rating and RD as you, and someone with the same rating but different RD is, for the sake of determining rating adjustments, NOT Rated the same.

Ok.  I will try to address this, but first let me caveat:   Live chess here is not the same as correspondence chess.  Similar, yes.  It is turn-based/day per move, and unused time does not accumulate. 

What do you consider a high RD?  Mine right now is 50 or so.   I play between 30 and 48 games at a time.  Some take longer than others to complete.  Each day I finish between 3 and 12 games, and in 2 months have completed approx. 440 games for an average of roughly 7 games/day. 

Is my rating here a bit lower than reality, substantiated by my "real" rating (FIDE 2200) and my ICCF (2387)?  It certainly fluctuates more here than in real world, and rightly so! 

I will of course be interested to see what the numbers are at 1,000 games and so on.

I do not see why is is not good for CC play, or on-line play, or even otb, blitz, or whatever type of play one would choose.  It seems a most flexible system that tailors itself (better than any other) to the actual numbers created by each individual while realizing a margin or error for both human error, time, personal improvement, etc.

What am I really missing>?

Is the biggest bone of contention the changes of rating?  Are we so used to a paleolithic system that updates ratings once every 2 months while running 2 to 4 months behind even in those updates? 

So, I am saying, yes, I think it is not only fine, but most suitable for CC play.

So, in order for someone to have a moderate RD such as 50 in your case, they must play between 30 and 50 games simultaneously.  For a high rated/experienced chess player, this is probably a manageable number of games.  But, for beginners like myself, who have to labor to find a good move in many situations, playing that many games simultaneously results in very poor moves and many losses.  Losses that would not have likely occurred if fewer simulataneous games were being played.

So, for someone who can only play ten or fewer games simultaneously, under this rating system they are forever doomed to an RD around 100 and wil never achieve an "accurate" rating!  However, the admins of this site could reduce the requirement of games played per time period in order to achieve an RD of 30.  I hope the admins have not just blindly applied the Glicko Rating System to their website without fully understanding its ramifications.

I think a RD of 100 is accurate for anyone under 2000.  Almost Always. 

But, that is 40 years of personal, anecdotal chess speaking.  I think for most people you just described, it is probably @ 150.  It is very common, and there are many more actual circumstances, of players 1200 to upset 1600 than it is for 2200 to upset 2600.  In scholsatic chess, people 400 beating 1400 happens with such alarming frequency, if people reacted the way they did here, everyone would be banned! j/k  

There are several reasons for this (K factors, etc.) but also people at lower end of knowledge and ability can improve much faster (because they have more room to do so!) and as such there ratings are more "suspect" to begin with.

No matter what number they assign or how it is derived, no one can take the real knowledge and skill you posess away from you. 

Like myself included, most of us are not as good as we want to be, or really are.

You are right that it's hard for the glicko system to give you an accurate rating if you don't play a large amount of games.

However, the traditional ELO method would be even less accurate.

Isn't an RD of 100 inaccurate by definition?  We are saying that if someone's "true" rating of 1550 was somehow absolutely known that we are satisfied by calling it anywhere between 1450 and 1650?  That doesn't seem so accurate!  However, if the RD was 30, for example, I could live with calling that person's rating between 1520 and 1580.

RoundTower wrote:

 I don't think this is true.  Gaining/losing 30-50 points per game because of an RD over 100 seems to be a very poor way to calculate a CC player's rating if they play fewer than 30-50 simultaneous games.  The ELO method would be more accurate in this instance.

You can't just say "Well, since people don't play as many games as otherwise, all of these ratings are going to be considered more accurate than they are."

I would say, no it is not. (RD of 100 being inaccurate by definition)

What I am saying is there is no such thing as a "true" rating to begin with. 

That entire concept to me is merely theoretical and the Glicko system is the latest and best attempt to deal with reality.

I am also certain some bright, young mathematically inlcined mind will improve upon it!

The smaller any sample is, the less certain anyone can be of the results derived from such sample.  I feel comfortable claiming that can be successfully applied to donuts as well as chess ratings. 

Again, I wonder, what is the genesis of your real beef here, and do you have viable solution(s)?  As your initial statement that this is a CC site is not wholly accurate (as I described above), I am now confused to what the real issue has become?

The short answer= "NO". Nor is it satisfactory for LIVE CHESS imo. My rating here is the worst anywhere by hundreds of points. And it feels a little ridiculous to go up and down from 1400 to 1700 here and back down again in a day or two of time. I'll win 20 or 30 or games vs players at my rating and get maybe 5 points per game. Then lose two games to equally rated players and have up to 45 points taken from me for EACH of those losses. It is simply ridiculous. Had I not played till tired and lost my last two (or my first and last game of the day), would I REALLY be 90 points STRONGER as a player?! No way.

And it does not jibe with ELO on other sites. And is not a particularly good feeling to know you are a Master in Corr with 6 other orgs; and being 1950 to 2100 in LIVE & Blitz everywhere else, but fighting a battle up to 1700 here, or 1600 after a bad day, only to lose it back as soon as risking that loss that comes when tired or the board starts freezing or stealing your last 6 seconds off the clock.

No one game should ever be That significant in any valid rating system. Elo is much better. And there are ways to utilize Provisional Ratings, Bonus Points, and "K"-factor changes that can make it as quick to give a True Established Rating as any valid system and can adjust out any over inflation or deflation as well, using the tools available in ELO.

I do like the fact here that there are three separate ratings for LIVE, based on different Time Controls used. That is a positive. But about the Only Positive of that entire rating quagmire. imho ( & even moreso, in my 'unhumble' opinion )

}8-)

PS.. in the end it comes down to this: What sensible rating system can take two losses vs 20 wins against players all of same rating area as you, and judge that the two results are Equal ... leaving you at the same rating as you started with before that days effort? It is nonsensical mathmatically and statistically.