TonyRo wrote on 09/20/10 at 11:19:16:


Yes but that reward (if there is any, her IM title would her a lot more reward then being at the top of a casual corr. site)  wouldn't weigh up to the possible consequences of getting caught cheating. It doesn't really make sense for a chess professional to risk her/his reputation to win a few meaningless corr. games. Weird... .

I don't understand why she should bother to 'cheat', unless there were large cash prizes involved it doesn't make any sense to risk tarnishing her reputation just to play a better correspondence game.

using an engine in correspondence chess is cheating?
personally, if I would be a professional player I'd never play against normal hobby-chess players  without an engine in correspondence, because I know most of them would use one against me, only to be able to state "I won against a GM in correspondence" (of course they will forget to add that they were helped by an engine).

Actually, chess.com's methodology is rather interesting. 

They ran stats on the best correspondence players of pre-computer generations, such as Estrin and company. Then they took a look at how the moves, after they left the database, matched up to engines. They focused only on games against other 2200 players. They found a fairly consistent threshhold. 

It doesn't matter which +2900 engine they use, the results will be the same. 

This is what they found:

The best a human can do in terms of matching a computer's play, is:
 
Top move:     65%
Top 2 moves  75%
Top 3 moves  85%

Dembo's stats were:

Top move:  73.3
Top 2          88.2
Top 3          93.5

By chess.com's own admission, Kramnik, Anand and Topalov matched the threshhold in their recent matches.  Dembo's scores were substantially higher. Can that be explained by new high standards of correspondence play?


I think it's an open question.

But for those scoffing at correlations as proof, that's what statistics studies, correlations. The likelihood of a human matching a computer 100% approaches 0, so their approach makes sense. Perhaps the tolerances are a bit low though.

It's very likely that Dembo cheated. But I don't know if it's a certainty. I'd have to run the numbers myself, which doesn't interest me much. 


What does interest and excite me is that it may be possible to rule out engine play in the future, which obviates one of Markovich's main justifications for using engines in correspondence chess!

Ooooooh, correlation!.  Awesome proof, that is.

Listing cheats, important work.

GMTonyKosten wrote on 09/19/10 at 22:40:22:

Correspondence games, it seems.

Someone at chessgames.com posted the % correlation of her last 20 games.

http://www.chessgames.com/perl/chessplayer?pid=36823

Hi, 

Some of you already know that WGM Yelena Dembo was unceremoniously booted from chess.com for cheating recently. The site listed her as a cheat.  She has now threatened to sue the site. 

Chess.com has decided to back away from calling Dembo a cheat in order to avoid a lawsuit. But they clearly believe they are 100% correct.