I was sure I saw some graphs of this, hence they do exist, I just can't find them right now. Perhaps one needs a premium membership?
To your other point: I'm against all intervention in ratings. Different rating pools ---> different ratings, period. I'd rather have an accurate (with chess.com) rating of 354 or 9654, than something comparable with OTB federations, but with a 300 point margain of error due to interventions. They don't allow me to track my progress.
The way I see it, properly done interventions would not only help you compare your game skill more accurately to other sites, but would also *improve* the trackability of your genuine progress, as they would counteract any inflation or deflation due to any influx of new players. I think that most players wanting to track their genuine progress would want to see where they stood in comparision to other players in general, and not just on their particular site.
Also, regarding the engine idea, it seems to me that this would be a pretty solid way of measuring such things, as the top moves of top-rated engines are by definition pretty good, hence the win-rate/ratings of said engines. Not that one wants to try to emulate an engine per se, but if one is making higher-level choices as per engine analysis the odds are pretty good that he or she is indeed progressing in his or her understanding/skill in chess.
Alas, I am getting tired and am repeating myself, I will check this thread in a day or so...
I can't figure out how to get to this page from the site... and is anyone knows how, plaese speak up. Lucky me though, I bookmarked it a few months back...
http://www.chess.com/echess/players.html
Ooh, thanks-those are the graphs I meant.
Thanks wowiezowie, thats what I was looking for.
Anyone have a graph for various (offline) chess federations?
I made a funny observation when looking at the graph wowiezowie posted, it seems the current number one (Alias GMArionian) has only played 4 games, which he has won all off. All of those was against the same opponent, Armenoids. Armenoids in return has scored all his wins against Ampadre. Ampadre in return has scored all his wins against H_nersisyan, which funnely enough has a diamond membership and seems to be the genuine account this player uses, while all of his others accounts are for the sole purpose of getting high ranked accounts through cheating.
FIDE/USCF/yahoo.com use the standard (contrarians might use the word "first" or "skeptical") system devised by the great statistician Arpad Elo Jr. This is known as the ELO system, and to boil it down to two phrases, both competitors fight over 32 points. The split of the 32 and the result of the game depends on their pre-existing ELO ratings, so a cut (200 points) between a Master (2400) and an Expert (2200) would be that the Master wins 3 out of 4, the Expert winning the fourth game, or similarly the Master winning two and drawing two in a four-game match.
The iniatiating poster's goals are far too lofty. When Zsuzsa (now Susan) Polgar's female rating got out of hand in chess-crazy but Soviet-dominated Hungary, all her rated opponents were given an extra hundred points, but not her. Imagine the personal outrage.
The Glicko System, used by this site as well as ICC, is a superset of the ELO system and whose workings exist in the always-opaque. Nevertheless, the less you play, the more the next game counts, so a win in ELO would net you no rating points whereas taking your time in Glicko will give you a few with a win. Neither system protects against schemes such as selective pairing (a favorite with the Soviet Union in 1962, as all three participants admitted) nor Mr Fifty Percent.
To understand the Glicko system guidelines, you need some basic statistics as a pre-requisite. At any rate, the man is named Glickmann, and his academic and lecture papers are easily found on the web.
I made a funny observation when looking at the graph wowiezowie posted, it seems the current number one (Alias GMArionian) has only played 4 games, which he has won all off. All of those was against the same opponent, Armenoids. Armenoids in return has scored all his wins against Ampadre. Ampadre in return has scored all his wins against H_nersisyan, which funnely enough has a diamond membership and seems to be the genuine account this player uses, while all of his others accounts are for the sole purpose of getting high ranked accounts through cheating.
Cheating? You don't say!
I made a funny observation when looking at the graph wowiezowie posted, it seems the current number one (Alias GMArionian) has only played 4 games, which he has won all off. All of those was against the same opponent, Armenoids. Armenoids in return has scored all his wins against Ampadre. Ampadre in return has scored all his wins against H_nersisyan, which funnely enough has a diamond membership and seems to be the genuine account this player uses, while all of his others accounts are for the sole purpose of getting high ranked accounts through cheating.
Cheating? You don't say!
Glickmann addresses the problem of "selective pairing" but offers no resolution. I personally feel that, locally, selective pairing is what happens when there's not a sufficiently personally-disengaged tournament going on.
You can assert cheating, but yet you can't articulate a coherent anti-selective pairing tournament format. You are not alone, yet you lack an alternative. This sociological situation defies physics by sucking and blowing simultaneously, but may be overcome as it is not a matter of physics.
Actually it's trivial to solve selective pairing. You just do so at the cost of never allowing someone to pick their opponent, which is a desired feature both here and in the real world.
Any ideas on what causes the shape and the skew of the ratings curve, hopefully from mathematician/statistician types? I've been curious about it for a long time. Should the distribution be this skewed or should it be more symmetrical than it is?
http://www.mark-weeks.com/aboutcom/aa03a25.htm
"This is a ratings distribution for USCF Non-Scholastic Members. (Source: USCF 2002 Ratings Distribution)."
The skew of the USCF distribution is quite different from the chess.com one.
Looks like a Poisson distribution...
Selective pairing has never been disallowed by any rating system I know of, and would be counter to this site's mission statement of being "the Facebook of chess". Also, the cost would be intolerable in f2f tournaments, particularly in rural areas. Even in the late 1980s in NYC I encountered the same opponent more than once.
I am wondering what the deviation curve looks like regarding chess ratings, both at this site and at official (offline) chess federations. At this site I am wondering about both online (correspondence) and live chess.
Also, has anyone run enough (a representative sample) of games through a top computer engine to figure out what the curve might look like from an objective viewpoint? I realize complete objectivity is impossible here, but I would think that some reasonable measures could be deduced by comparing peoples moves to top moves as determined by a top level (2800+) rated engine.
I am thinking adjusting a sites ratings to this measure would be a good exercise in order to adjust for inflation or deflation of ratings as compared to other sites and/or federations, which would make comparisons of a persons ratings on different sites much fairer and useful for all involved. It would also be nice to compare the results to chess players historically, both at master and non-master levels.
Thanks in advance to all who repsond intelligently to this question.