Program that calculates rating by analyzing your games?

Is there a program that can analyze a database of your games and then approximate your rating?

Im positive it does exist. I have heard of them before. But I'm not sure how available to the general public it is.

Are you talking about Elostat or BayesElo?  Fritz can also generate a list of Elo ratings for an archive of games based solely on the results (i.e. 1-0, 1/2-1/2).

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

"... further assumption is necessary, because chess performance in the above sense is still not measurable. One cannot look at a sequence of moves and say, "That performance is 2039." Performance can only be inferred from wins, draws and losses. Therefore, if a player wins a game, he is assumed to have performed at a higher level than his opponent for that game. Conversely if he loses, he is assumed to have performed at a lower level. If the game is a draw, the two players are assumed to have performed at nearly the same level.".

I think that might be what bobobbob is asking about, i.e. a program that takes the results of games, and generates ratings for all the players in the pool.

Here is a note about ELOstat which is a program that is packaged with Arena (Chess engine interface)

ELOStat - Version 1.3

by Dr. Frank Schubert, 01/2005

Note:

This program is freeware and may be used without restrictions for all non-commercial, i.e. private or scientific purposes. If you want to publish rating lists produced by ELOStat on private websites or in scientific magazines you should mention the name of the program and the version number. It is also possible to use ELOStat for commercial purposes. If you want to publish rating lists produced by ELOStat on commercial websites or in magazines, it must be guaranteed that the database on which the rating list is based is made public if there is any need for it. The program has been validated extensively but correctness of the results produced by ELOStat cannot be quaranteed.

Description of the program:

ELOStat produces a comprehensive statistical evaluation of PGN databases and generates an Elo rating list similar to the swedish SSDF list. It should be used primarily for computer chess games but of course it can also be applied to ‘human‘ chess databases. The PGN files should have been produced by a commercial chess program or database program. PGN files generated ‘manually‘ by a text editor can cause problems if they don’t comply strictly with PGN conventions.

Chess.com does not provide ratings for vote chess/team match groups, so I often use an Elo program to generate approximate ratings to help assign teams to divisions according to strength or for seeding teams in the early rounds of swiss tournaments.  I thought that might be what bobobbob was looking for.  Obviously, you cannot expect the ratings generated for one pool to line up with those from another pool, but I thought he might already know this.

To clarify, I was looking for a program that will say, "You blundered 1% of the time so your rating is 2000" or something similar. 

To quote tonydal: "lol"

Your rating is based by your wins/losses/draws relative to to other players in the same pool, it is all relative... based on odds not % of blunders.  Any program that claimed to spit out a rating based on this would be completely useless because the only rating that matters is the one relative to other players.

Trying to predict chess rating from one game would be like trying to predict your facemash score from one hotornot score. Not gonna happen

No.

An easy reason (but not the only reason) is that ratings are not absolute values, they're relative to the pool of players you play in.  Even within the same organization such as FIDE or even chess.com.  Two players of the same strength that regularly plays on chess.com live chess 12 hours apart from each other will probably be facing different players and have different ratings.  Not 100s of points, but this is just for example.

I realize that these are not programs, but Chess.com's Tactics Trainer and Chess Mentor give you a rating estimate I believe.  Along the same lines there is this thread which points to a 10 problem quiz and the Bratko-Kopec test.

1. Nc3 Nh6 2. Nb1 Ng4 3. Nh3 Nf6 4. Ng1 Ng8 1/2-1/2

Estimated rating result:

White: 3200

Black: 3400

There is definitely no correlation between % of blunders and ratings.  Blunders are often the result of calculating several moves a head and getting an idee fixe.  A high rated player is more likely to make blunders than a low rated player looking one move at a time.  Rubinstein one of the all time great players was very prone to making blunders.

A high rated player is more likely to make blunders than a low rated player looking one move at a time.

:p

Definitely?

That's what happens when you talk about stuff that you don't have a clue about.

But it's true that a higher rated player typically will be moaning more about his so-called blunders than a lower rated player who is simply amazed by what his opponent's pieces are able to do without finding any fault with himself

Computer ratings have been correlated highly with search depth. 

Computer ratings are about as predictive versus humans as they are

with computers. Thus one can rate a human by playing  or considering

play at various search depths.

However, I am still not sure how Shredder for example computes its player

ratings, since it is not based on win-loss, which I think would be more

appropriate. 

A fellow called nimh on chessgames.com has been fooling around with some statistics along those lines.  Here is one of his papers.