What's my real rating?

A big cognitive hole in peoples' understanding of rating systems is that they believe a rating change represents "points" that can be accumulated.  Both Elo and Glicko are statistical models that represent locations on a slope/hill/curve:

Here's some "light reading", for those of you who have the time... 

http://www.glicko.net/research.html

Thanks Tim, I went through Glickman's papers. BTW I'm a statistician and once upon a time I operated what might have been the first net-based chess rating system for e-correspondence chess (on the CHESS-L BitNet list, for those who were alive at the time).

In A Comprehensive Guide to Chess Ratings, Glickman states "players who have not competed in tournaments for an extended amount of time have abilities that may be changing, so that their ratings become less reflective of their true average strength". But in this paper he only provided an intuitive argument for having a time-varying scale factor. In particular, I'd like to see how the time factor is estimated. I'm curious enough that I'll try to chase it down when I find time.

Pathfinder,

Isn't it complicated by running into time series problems?  Also as I understand it chess populations are not represented by a normal curve.  If we break them into sub groups those groups may be normal, but there do seem to be at least two underlying populations which the thread pointed out earlier about Jr ratings.  Also isn't the problem that 1. There is documented ratings inflation in ELO commonly (wrongly or rightly) attributed to changes brought about by Kasparov?  So that now the Super Elite GMs are 2600+ when at one time such ratings were even more rare.  Does the fact that the population is not constant complicate attempts to study and predict?  I'd love to see a regression analysis but you have to have players with ratings from many sites to generate data, and I think the explanation would only be representative of today. 

Also in comparing ratings you have to consider the older players [likie me] have for one reason or another lost a good part of their chess prowess.

Also, you have to consider that maybe there are considerly more or maybe there are considerly less of a population playing chess and that in general populations increase over time.

Also there is a myraid of chess aids now tht did not exist in the past.

For all purposes, the curve is as close to a normal distribution as the accuracy (and math) will allow.  Any perceived variation is due to other factors, such as players leaving competitive play, new players entering, results not entered on time, etc.  You also have to consider that USCF/FIDE (which don't use a strict Glicko) have rating "floors", which lock a champion player into a bracket so they no longer compete against inferior players (note: this explanation is greatly simplified).  In any case, inflation hasn't been caused by anyone (Kasparov or others) making changes to the system.  It's a normal and predicted consequence of "points" entering the system, in the form of over-valued newer players who consequently lose to established masters.

see:

http://en.wikipedia.org/wiki/Elo_rating_system#Ratings_inflation_and_deflation

Actually, I guess it would be more accurate to call it a logistic distribution.

http://en.wikipedia.org/wiki/Logistic_distribution

Don't try to compare your chess.com or other internet chess site ratings with USCF or FIDE ratings.  First off, most games played on chess websites are played at much shorter time controls and are not indicative of your OTB (over the board) 40/2 tournament rating.  You would probably have to play around 100 OTB tournament games, against all levels of players, to get an accurate assessment of your rating.  Remember that everyone starts on here at 1200 and has to work their way up.  So you could conceivably be a Fide Master OTB and lose to a Grandmaster who has a 1200 rating on here (assuming it's his first game on this website) and that would knock your rating way down, artificially.  So, if you want a semi-accurate rating, you'll have to play some OTB tournaments against a wide range of classes.  I wish I knew my "real" rating too.  I estimate it to be somewhere in the 1700-1800 range, but since I no longer play OTB tournaments, I guess I'll never know for sure.  My online ratings are always much lower than that though, as the faster time controls force me to rush my moves and therefore make more mistakes, while the 2-dimensional flat board messes with my depth perception and I often hang pieces due to that.  I am much better OTB, with longer time controls.

Curious, Why does everybody start at 1200--this does not make sense to me.

For one thing it will keep the higher rated players from playing.

It causes distortions in ratings also.

1200 was just picked as the arbitrary midpoint of the curve.  They could've picked any other number, say 0 or 200 or 2000, and then the ratings would be correspondingly higher or lower as the case may be.  There's no distortion based on the provisional rating of 1200; higher-rated players can just typically seek out players of similar rating if they feel a 1200 player is not worth playing against due to risk/reward.