http://www.ficgs.com/Match-FICGS-vs...

Dirk Ghysens (2006-09-18 10:33:56)
Not all, Henri

"...Unfortunately Yelena Dembo is no longer playing at Gameknot; they threw her out."

Those stupid chess playing sites! Who's gonna ban her next?

I'm 100 percent sure she's not gonna be banned at CG...

Ken Regan, the guy cited as the originator of some of the methods, makes an important point: "The main statistical principle which these pages show has been misunderstood by the chess world is that <a move that is given a clear standout evaluation by a program is much more likely to be found by a strong human player. And a match to any engine on such a move is much less statistically significant than one on a move given slight but sure preference over many close alternatives.>"

It's been my experience that when I play much lower rated players, say 800 points, which is likely the OTB rating difference between IM Dembo and her opponents on chess.com, that I quickly get a large material advantage and the game takes on a very forcing character. There are usually only a small number of moves worth considering, i.e. how to win another piece or checkmate.

So the "top 3" method should break down here, in particular when compared to ICCF type matches between evenly matched opponents.

I think others have pointed this out with a game that Dembo had vs a "2000" opponent. To get a better flavor, I think you have to look at some opponents rated 1600 or so. You might find some of those in big open tournaments she has played.

2844, 2608, 2724, 2661, 2616, 2534, 2455, 2719, 2646, 2596, 2525, 2436, 2686, 2738, 2797

So the highest in the set was rated 2844 & the lowest 2436. The average opponent rating for Yelena's last 15 games was 2639.

<Zygalski>

You've just MISREPRESENTED THE BATCH ANALYSIS, and by doing so, have admitted your wrongdoing. Muhu-haha!

In your list, 14 were NEVER used in the batch set of games that were analyzed against her. Why did you try to misrepresent your analysis? It's because you know in your heart and mind that you misapplied the method giving her an abnormally high-efficiency for later analyzers. When looking through your batch set, instead of seeing a strong move, they decided that an engine made it. They had been convinced of your flawed results before they ever started:

Guilty before proven innocent.

<Zygalski> With your last misrepresentation of the batch analysis, players will easily conclude that you should no longer be trusted.

The official statement from Chess.com is that her account is no longer active, see this same thread Sep 20. They don't have a case against her. That's why Yelena isn't taking them for an expensive joy-ride through the international court system.

But now <Zygalski>, who <"first selected Yelena for Batch Analysis">, is the one who continues to independently wage a campaign against her on chess-related sites throughout the internet, misrepresenting data if needed to convince the others of his position.

http://ratings.fide.com/card.phtml?...

I will continue to post in these threads whilst you yourself mis-represent match up methodology, twisting it to suit your personal fetish.

As has been said many times before, the methodology is designed for analysis of large batches of objectively chosen games. Individual games can have very high match rates.
For instance:
V Palciauskas vs J Muhana, 1978
Yields the following match rates for the (then) non-database moves:

White: Palciauskas, V
Top 1 Match: 30/37 (81,1% )
Top 2 Match: 34/37 (91,9% )
Top 3 Match: 34/37 (91,9% )

Black: Muhana, J
Top 1 Match: 22/37 (59,5% )
Top 2 Match: 30/37 (81,1% )
Top 3 Match: 31/37 (83,8% )

Yet Victor Palciauskas, winner of the 10th CC World Championships, could only achieve the following cumulative match rates from all games with 20+ non-database moves:

Palciauskas, V (11 games)
Top 1 Match: 209/355 (58,9% )
Top 2 Match: 268/355 (75,5% )
Top 3 Match: 291/355 (82,0% )

So whilst Yelena may have the odd high match rate game, when the method is applied properly to her OTB, the following rates resulted:

Last 20 games >= 35 moves from chessgames.com. All against high-rated opposition (lowest ELO in the set is 2222, highest is 2655)

<Xeroxx>
The alleged cheating & the closed Dembo account refer to actions staff took at chess.com in CC games where engine use if forbidden.

You can always analyse the games yourself.
Here's a link to them:

<The real question is, what the heck is she doing playing these patzers?>>

<Xeroxx>

Here is the full set of opponent ratings in the 20 Dembo chess.com games:

2797, 2503, 2304, 2377, 2367, 2309, 2371, 2475, 2371, 2528, 2209, 2355, 2238, 2228, 2266, 2216, 2299, 2257, 2198, 2189.

So the lowest in the set was 2189, the highest 2797. The average opponent rating from the set is 2342.85.

These are all top 1 percentile chess.com highest rateds, so not "extremely low-rated opponents". Another inaccuracy on your part.

Finally, again <without personal attacks> if you can manage, please show me an inaccuracy by Dembo in any of these chess.com games where she drops a mere 0.25 or more of a pawn at depth 20. You, or anyone else have yet to provide any such evidence which is surprising, considering even the best GM's regularly drop 0.50 or more in their games.

Here is the link to her chess.com games.

http://www.chess.com/echess/profile...

Simply load any version of a 3000+ Elo rated engine up, preferably on a multi-core modern pc & look for the top 3 or 4 lines & allow the engine to reach depth of 20 & compare to the played moves.

http://www.chess.com/echess/game.ht...

[Event "Let's Play!"]
[Site "Chess.com"]
[Date "2010.02.01"]
[White "BoboXXL"]
[Black "YelenaDembo"]
[Result "0-1"]
[WhiteElo "2608"]
[BlackElo "2541"]
[TimeControl "1 in 10 days"]
[Termination "YelenaDembo won by resignation"]

1.e4 e5 2.Nf3 Nc6 3.Bc4 Nf6 4.Ng5 d5 5.exd5 Na5 6.Bb5+ c6 7.dxc6 bxc6 8.Bd3 h6 9.Ne4 Nd5 10.O-O Nf4 11.Be2 Be7 12.d3 Rb8 13.Re1 O-O 14.a3 c5 15.Bf3 Qc7 16.b3 Rd8 17.Nec3 Be6 18.Bxf4 exf4 19.Bg4 Bxg4 20.Qxg4 Bf6 21.Ra2 c4 22.Ne4 Be5 23.b4 cxd3 24.cxd3 Nb3 25.Qd1 Nd4 26.Nbd2 a5 27.Nc5 Ra8 28.Nc4 Bf6 29.Ne4 axb4 30.Nxf6+ gxf6 31.Qh5 Re8 32.Rxe8+ Rxe8 33.h3 b3 34.Rb2 Kg7 35.Qd5 Re1+ 36.Kh2 f3+ 37.g3 Qc8 38.Qh5 f5 39.Qh4 Qc5 40.Qf4 Re2 41.Rd2 b2 42.Rxb2 Nc2 43.d4 Qxc4 44.Qxf3 Nxd4 45.Qb7 Re1 46.Rd2 Qf1 47.h4 Qg1+ 48.Kh3 Re4 49.Qxe4 fxe4 0-1

4 x AMD Phenom 2.30 GHz 4GB RAM
Analysis Deep Rybka 3 x64 Hash:640 Time:60s Depth: Max 20 ply

Result for non-database moves:

White: BoboXXL
{ Top 1 Match: 19/39 ( 48.7% )
{ Top 2 Match: 30/39 ( 76.9% )
{ Top 3 Match: 31/39 ( 79.5% )
{ Top 4 Match: 33/39 ( 84.6% )

<These are all top 1 percentile chess.com highest rateds, so not "extremely low-rated opponents". Another inaccuracy on your part.> Zygalski, it's called Chess.com's pool of online correspondence players. In that pool any patzer can have a high rating. It also exists for large OTB pools, such as USCF ratings. These ratings have relative strengths that are are slightly different than the FIDE rating. A USCF rating is about 100 points lower than a FIDE rating. And a Chess.com rating is about 100 points lower than a USCF rating. So it doesn't matter whether they are top 1% of a tight chesspool, all that matters is their relative strength. Her opponents were incredibly weak, by hundreds of points, relative to her own FIDE rating and especially considering her tournament experience.

So you've finally understood that. Good. It was a major error, and by doing that you manipulated the downstream interpretation of her results. Guilty before innocent. But instead of apologizing for your wrongdoing, you decide to stop talking about the batch that was analyzed, and instead cherry-pick a brand new game that wasn't analyzed to try to rescue your sunken position. It's not going to work.

That game you listed was played in the Two Knights Defense, an opening that is as familiar to any master as a milk is to a child. Yelena Dembo has probably had that opening mastered for at least 20 years, being a rated chessplayer since she was 3! In this game, Dembo followed Shcherbakov vs. Khismatullin Moscow. 2010.2.7. The results for black were equally devastating.

Top 1: 70%
Top 2: 75%
Top 3: 90%
Top 4: 100%
Houdini depth 20: http://www.freeimagehosting.net/upl...

That's from an entire game in a few hours, over the board, typical of the forcing nature of this opening, and also of Yelena's style. Why you are so exasperated about the fact that Yelena can do about 10-20% better given 2 to 6 days per move, free access to Chessbase, and an interactive digital board with which to practice lines and variations is beyond anyone's comprehension.

I've been running some tests on games where opponents are mismatched by more than 600 points. I have to go to some other chess sites for that, as I haven't found many on chessgames.com.

The new rybka aquarium scripter makes running these sorts of analyses much easier.

In a small sample of 20 otb games, the average match rate (using the same methodology of the articles) for the higher rated player out beyond move 12 is over 90%.

Small sample. Not conclusive. But interesting enough to run a bigger test.

Heck, I might get a published paper on this illustrating the limits of automated cheat detection.