(UTC)
  ChessPub Forum › General › General Chess › Yelena Dembo on Chess.com
(Moderators: trw, RoleyPoley, proustiskeen)
‹ Previous Topic | Next Topic ›
Page Index Toggle Pages: 1 2 [3] 4 5 ... 12
Topic Tools
Very Hot Topic (More than 25 Replies) Yelena Dembo on Chess.com (Read 99678 times)
Göran
Senior Member
****
Offline


ChessPublishing is great!

Posts: 454
Location: Sweden
Joined: 02/13/08
Gender: Male
Re: Yelena Dembo on Chess.com
Reply #147 - 09/24/10 at 02:03:45
Post Tools
I have played on chess.com since a year and a half but now closed my account due to this very incident. I cannot support such a policy; publically announcing people to be known cheaters without valid proof. Shame on chess.com!

Drogo and Zygalsky are claiming that they are not accusing IM Dembo for cheating but in the next sentence they are saying –just look at our nice numbers, she is really playing like a chess-engine and just compare with the top ten players, nobody else has those nice numbers. That is accusing without standing for it – shame on them too. I think they are the real cheaters.

Drogo and Zygalsky would do less harm and themselves a favour if they attended a basic course in statistics instead of publicly accusing professional as well as amateur players for cheating. A highly respected member of this forum compared their actions with McCarthy. I think that is a very valid comparison and it really scares me. 

The only excuse I can see is that they are so full of numbers that they apparently don’t understand and as amateurs believe it is statistics.  But why should ignorance be an excuse?
  
What kind of proof is that?
Back to top
 
IP Logged
 
drogo
YaBB Newbies
*
Offline


I Love ChessPublishing!

Posts: 40
Joined: 09/21/10
Re: Yelena Dembo on Chess.com
Reply #146 - 09/24/10 at 01:08:01
Post Tools
Ametanoitos wrote on 09/23/10 at 23:29:45:
I examined 3 games of her and i got very different results by choosing different engines and different analysis time i gave to these engines. I saw some moves made by her that were obviously very human-like and matched the top 3 engine suggestion and some other moves very human-like that they did not! 



Which are the games? What engines did you use? What was the depth? What are the results? Sure there are human-like moves which match the engine suggestions, like forced recaptures.
  
Back to top
 
IP Logged
 
Seth_Xoma
God Member
*****
Offline


FM with 2 IM Norms - (2381)

Posts: 558
Location: Lansing
Joined: 11/05/04
Gender: Male
Re: Yelena Dembo on Chess.com
Reply #145 - 09/24/10 at 00:04:43
Post Tools
Ametanoitos wrote on 09/23/10 at 23:29:45:
This conversation entertained me to say the least but what you don't understand is that if i was Dembo i wouldn't be entertained at all! She is a very serious profesional that makes money from chess and an accusation of this kind should be much more carefully made. The whole thing about the 20 games carefully chosen to be examined is obviously wrong. Why 20 games and not 100? I examined 3 games of her and i got very different results by choosing different engines and different analysis time i gave to these engines. I saw some moves made by her that were obviously very human-like and matched the top 3 engine suggestion and some other moves very human-like that they did not! 

Again, the problem is that you can never be sure you can prove someone is a cheater and if you make that claim about a person who makes money from it you have to be able to prove it in a more reliable way. For example show 10 examples when an obvious human-move wasn't made and instead a "stupid-machine-move" was chosen by Dembo.


Absolutely. There's nothing wrong with suspending her account per se, but when chess.com accused her of flat-out cheating on their website with questionable evidence (not 100% verifiable), that could ruin her public image. This is where chess.com really dropped the ball, IMO.
  
Back to top
 
IP Logged
 
Ametanoitos
God Member
*****
Offline


The road to success is
under construction

Posts: 1429
Location: Patras
Joined: 01/04/05
Re: Yelena Dembo on Chess.com
Reply #144 - 09/23/10 at 23:29:45
Post Tools
This conversation entertained me to say the least but what you don't understand is that if i was Dembo i wouldn't be entertained at all! She is a very serious profesional that makes money from chess and an accusation of this kind should be much more carefully made. The whole thing about the 20 games carefully chosen to be examined is obviously wrong. Why 20 games and not 100? I examined 3 games of her and i got very different results by choosing different engines and different analysis time i gave to these engines. I saw some moves made by her that were obviously very human-like and matched the top 3 engine suggestion and some other moves very human-like that they did not! 

Again, the problem is that you can never be sure you can prove someone is a cheater and if you make that claim about a person who makes money from it you have to be able to prove it in a more reliable way. For example show 10 examples when an obvious human-move wasn't made and instead a "stupid-machine-move" was chosen by Dembo.
  
Back to top
 
IP Logged
 
Smyslov_Fan
God Member
Correspondence fan
*****
Offline


Progress depends on the
unreasonable man. ~GBS

Posts: 6902
Joined: 06/15/05
Re: Yelena Dembo on Chess.com
Reply #143 - 09/23/10 at 23:18:32
Post Tools
Yakov Estrin was equivalent to Fritz 7 of course!

Wink

(He was the 7th CC World Champion)
  
Back to top
 
IP Logged
 
Uruk
Senior Member
****
Offline



Posts: 351
Joined: 02/03/09
Re: Yelena Dembo on Chess.com
Reply #142 - 09/23/10 at 21:49:17
Post Tools
drogo wrote on 09/23/10 at 21:24:28:

Apparently Yelena Dembo plays more like Rybka and less like Stockfish.

A very interesting conclusion. I'm holding my breath until you reveal which version of Fritz Estrin is closest to.
  
Back to top
 
IP Logged
 
TalJechin
God Member
*****
Offline


There is no secret ingredient.

Posts: 2892
Location: Malmö
Joined: 08/12/04
Gender: Male
Re: Yelena Dembo on Chess.com
Reply #141 - 09/23/10 at 21:34:46
Post Tools
Zygalski wrote on 09/23/10 at 19:22:06:
TalJechin wrote on 09/23/10 at 19:13:14:
Zygalski wrote on 09/23/10 at 19:06:27:
TalJechin wrote on 09/23/10 at 18:54:34:
Sorry, a quick relapse - just curious if you've checked the games of e.g. the top 10 at ICCF? van Oosterom, Ulf Andersson etc. 

That is: can the magic box verify that they use computers and if so are their stats higher, lower or roughly the same as those branded cheaters?


As I said before, and as you can surely also now see, all I said here is that Dembo's choice of non-database moves in those chess.com games was far more engine-like than any of the benchmarks.
Also, you have to consider that the nearest match up rates to Dembo's are in data sets with a far lower sample size.
Dembo sustained massive engine match up in a reasonably large sample of moves.
Then, the games themselves.  Dembo moved daily in around 10-20 games in progress over quite some time.  Many of these games were vs chess.com 2200-2600 rateds so they weren't patzers.

I think this is why the FIDE 2300 games mod, when seeing the Dembo match up results, said it alone was "extremely convincing" [evidence of engine use].


You mean that you haven't tried your magic box on a group that's likely to use engines? Incredible...

Send me the pgn's of the objectively chosen CC games you want me to look at & I'll post (just the results! Smiley) in here.
20+ games from a single player, all high-quality opponents, each with 35+ total moves.  Most recent from the player that fulfil all criteria would be good.
Ok..
Is that fair enough?


Here's 20 recent games, longer than 35 moves by van Oosterom - see attachment.
  
Oosterom.pgn ( 19 KB | Downloads )
Back to top
 
IP Logged
 
drogo
YaBB Newbies
*
Offline


I Love ChessPublishing!

Posts: 40
Joined: 09/21/10
Re: Yelena Dembo on Chess.com
Reply #140 - 09/23/10 at 21:24:28
Post Tools
Uruk wrote on 09/23/10 at 20:15:40:


Zygalski wrote on 09/23/10 at 18:34:28:

Now Yelena Dembos' 20 chess.com games vs 2200+ rateds all with 35+ move results:

Deep Rybka 3 x64 Hash:256 Time:30s Depth:12-20ply
AMD Phenom x 4 2.30Ghz 4GB DDR2 RAM

YelenaDembo (Games: 20)
Top 1 Match: 530/723 ( 73.3% )
Top 2 Match: 638/723 ( 88.2% )
Top 3 Match: 676/723 ( 93.5% )
(...)

Stockfish 1.8, 512MB hash, min/max ply=12/30, 40s/ply, 2GHz Core Duo:

{ YelenaDembo (Games: 20) }
{ Top 1 Match: 484/723 ( 66.9% )  
{ Top 2 Match: 608/723 ( 84.1% ) 
{ Top 3 Match: 655/723 ( 90.6% )


Sorry, can't you see something fishy here ? (There are several good answers.)

Science is hard.


Apparently Yelena Dembo plays more like Rybka and less like Stockfish. That's what I would call the style of a strong player Smiley

Science is not that hard. But being an ignorant idiot is much easier.
  
Back to top
 
IP Logged
 
Uruk
Senior Member
****
Offline



Posts: 351
Joined: 02/03/09
Re: Yelena Dembo on Chess.com
Reply #139 - 09/23/10 at 21:00:15
Post Tools
Ok, this shows matchup rates are heavily engine-dependent, so you're comparing apples and oranges.
Also, I doubt Wch finals is akin to 2200+ opposition.

Thus your analysis is utter crap, I'm afraid.
  
Back to top
 
IP Logged
 
Uruk
Senior Member
****
Offline



Posts: 351
Joined: 02/03/09
Re: Yelena Dembo on Chess.com
Reply #138 - 09/23/10 at 20:15:40
Post Tools
Zygalski wrote on 09/23/10 at 18:19:04:

7th CC World Championship 1972-1975 (...)

Fritz 11 @ 30 seconds per move
4x AMD Phenom 2.30Ghz 
4GB DDR2 RAM
Hash Table 512MB
Database used www.chesslive.de (...)

Top 3 Players
Top 1 Match: 587/977 ( 60.1% )
Top 2 Match: 740/977 ( 75.7% )
Top 3 Match: 811/977 ( 83.0% ) 


Zygalski wrote on 09/23/10 at 18:34:28:

Now Yelena Dembos' 20 chess.com games vs 2200+ rateds all with 35+ move results:

Deep Rybka 3 x64 Hash:256 Time:30s Depth:12-20ply
AMD Phenom x 4 2.30Ghz 4GB DDR2 RAM

YelenaDembo (Games: 20)
Top 1 Match: 530/723 ( 73.3% )
Top 2 Match: 638/723 ( 88.2% )
Top 3 Match: 676/723 ( 93.5% )
(...)

Stockfish 1.8, 512MB hash, min/max ply=12/30, 40s/ply, 2GHz Core Duo:

{ YelenaDembo (Games: 20) }
{ Top 1 Match: 484/723 ( 66.9% )  
{ Top 2 Match: 608/723 ( 84.1% ) 
{ Top 3 Match: 655/723 ( 90.6% )


Sorry, can't you see something fishy here ? (There are several good answers.)

Science is hard.
  
Back to top
 
IP Logged
 
drogo
YaBB Newbies
*
Offline


I Love ChessPublishing!

Posts: 40
Joined: 09/21/10
Re: Yelena Dembo on Chess.com
Reply #137 - 09/23/10 at 19:45:25
Post Tools
Markovich wrote on 09/23/10 at 16:55:40:


Start with a sample set that includes all available corr games and only games played pre-computer by players of some minimum strength.  Do not include any OTB games.  Calculate matchup percentages for both players.  Be careful to apply an objective criterion of what constitutes a move elligible for consideration, and apply the same criterion to each move.  Be careful to use the same engine and engine settings across all games.  

From the games the test set, select a large number of sets of 20 games (say 2,000), not necessarily played by the same player, and in each randomly select Black's or White's performance.  Sample randomly and with replacement.  


So, you formulate here a statistical hypothesis. I guess that you have some empirical evidence to justify why randomness is needed. Why do you think that a randomly chosen CC player it's likely to have higher matchup rates than the top CC players (whose games were already tested)? Do you have an example of such a player?

No, you don't. You didn't do any analysis, you don't hae alternative methods, you just sit comfortably in an armchair and call the others "idiots". You know something: look in the mirror (if you have one in your house).
  
Back to top
 
IP Logged
 
TalJechin
God Member
*****
Offline


There is no secret ingredient.

Posts: 2892
Location: Malmö
Joined: 08/12/04
Gender: Male
Re: Yelena Dembo on Chess.com
Reply #136 - 09/23/10 at 19:13:14
Post Tools
Zygalski wrote on 09/23/10 at 19:06:27:
TalJechin wrote on 09/23/10 at 18:54:34:
Sorry, a quick relapse - just curious if you've checked the games of e.g. the top 10 at ICCF? van Oosterom, Ulf Andersson etc. 

That is: can the magic box verify that they use computers and if so are their stats higher, lower or roughly the same as those branded cheaters?


As I said before, and as you can surely also now see, all I said here is that Dembo's choice of non-database moves in those chess.com games was far more engine-like than any of the benchmarks.
Also, you have to consider that the nearest match up rates to Dembo's are in data sets with a far lower sample size.
Dembo sustained massive engine match up in a reasonably large sample of moves.
Then, the games themselves.  Dembo moved daily in around 10-20 games in progress over quite some time.  Many of these games were vs chess.com 2200-2600 rateds so they weren't patzers.

I think this is why the FIDE 2300 games mod, when seeing the Dembo match up results, said it alone was "extremely convincing" [evidence of engine use].


You mean that you haven't tried your magic box on a group that's likely to use engines? Incredible...
  
Back to top
 
IP Logged
 
Markovich
God Member
*****
Offline



Posts: 6099
Location: Columbus, Ohio
Joined: 09/17/04
Re: Yelena Dembo on Chess.com
Reply #135 - 09/23/10 at 19:09:56
Post Tools
I'm done posting to this thread; I've repeated myself too much already.

Zygalski, you don't need to report such detailed results; I frankly do not doubt that whatever you did, you did without gross error.  Reporting the method in detail doesn't mean reporting every last detail.  The problem is just to step back and treat the whole thing as a statistical problem.  Note, you still have yet to say, mechanically, how did you derive your thresholds?
  
The Great Oz has spoken!
Back to top
 
IP Logged
 
Schaakhamster
God Member
*****
Offline


I Love ChessPublishing!

Posts: 650
Joined: 05/13/08
Re: Yelena Dembo on Chess.com
Reply #134 - 09/23/10 at 18:56:41
Post Tools
TalJechin wrote on 09/23/10 at 18:54:34:
Sorry, a quick relapse - just curious if you've checked the games of e.g. the top 10 at ICCF? van Oosterom, Ulf Andersson etc. 

That is: can the magic box verify that they use computers and if so are their stats higher, lower or roughly the same as those branded cheaters?


That would be interesting. Perhaps most relevant would be the last few years.  

edit: interesting would also be to exclude obvious moves (recapturing when your opponent has take your queen) or forcing moves where only one move is legal or logical.
  
Back to top
 
IP Logged
 
TalJechin
God Member
*****
Offline


There is no secret ingredient.

Posts: 2892
Location: Malmö
Joined: 08/12/04
Gender: Male
Re: Yelena Dembo on Chess.com
Reply #133 - 09/23/10 at 18:54:34
Post Tools
Sorry, a quick relapse - just curious if you've checked the games of e.g. the top 10 at ICCF? van Oosterom, Ulf Andersson etc. 

That is: can the magic box verify that they use computers and if so are their stats higher, lower or roughly the same as those branded cheaters?
  
Back to top
 
IP Logged
 
Page Index Toggle Pages: 1 2 [3] 4 5 ... 12
Topic Tools
‹ Previous Topic | Next Topic ›
Bookmarks: del.icio.us Digg Facebook Google Google+ Linked in reddit StumbleUpon Twitter Yahoo
 
  « Board Index ‹ Board Top  

ChessPub Forum » Powered by YaBB 2.6.12!
YaBB Forum Software © 2000-2026. All Rights Reserved.

Valid RSS Valid XHTML Valid CSS Powered by Perl Source Forge