|
< Earlier Kibitzing · PAGE 14 OF 22 ·
Later Kibitzing> |
Sep-19-10
 | | tamar: To be fair to Yelena, chess.com rules are very strict, and include more than using computers. Their definition of cheating includes using multiple accounts, for example. |
|
| Sep-19-10 | | BobCrisp: It's clear from some of the associated members' comments that suspected computer cheating was the reason behind the ban. |
|
| Sep-19-10 | | bartonlaos: These are some of the numbers that chess.com used to help make their determination - Percentage agreement with Deep Rybka 3:
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% )
Top 4 Match: 698/723 ( 96.5% )
Here are the numbers from a normal correspondence player: 9th Correspondence Chess World Championship 1977-80** 1st Place:
Oim, Tonu O. (EST) (Games: 14)
Top 1 Match: 263/433 ( 60.7% )
Top 2 Match: 352/433 ( 81.3% )
Top 3 Match: 388/433 ( 89.6% )
Top 4 Match: 400/433 ( 92.4% )
---
The numbers look too close for me to make a determination, though I'm no statistician - especially considering improvements to the opening databases over the past 30 years. Certainly they don't seem so extraordinary that Chess.com needed to follow through with the steps they made. She's been one of my favorite players, so I'm probably biased, but I'm giving her the benefit of the doubt, and will wait for her response. |
|
| Sep-19-10 | | BobCrisp: Did they say how many games (and at what time controls?) were used for testing? |
|
| Sep-19-10 | | bartonlaos: This was all the information I was able to find in the time available. Maybe someone can post more details. The typical correspondence matches on chess.com are 1 move per 3 days. In any event, Chess.com is now re-routing Yelena's profile that displayed the CHEATER logo, to a 'disabled account' screen. So maybe they aren't too sure of their methods after all - makes me wonder if they can survive her counterattack. |
|
| Sep-19-10 | | Zygalski: But Yelena seems to have played nearly twice as many non-database moves as Correspondence Chess World Champion Oim in those samples & also got much higher %'s.
73.3% top 1 as opposed to Oim's 60.7% from half as many moves! |
|
| Sep-19-10 | | Zygalski: You seem to have also found the sample with the single highest rate.
For instance;
6th Correspondence Chess World Championships 1968-71 *Top 4 finishers
1st Place:
Rittner, H. (Games: 15)
Top 1 Match: 203/337 ( 60.2% )
Top 2 Match: 254/337 ( 75.4% )
Top 3 Match: 276/337 ( 81.9% )
Top 4 Match: 293/337 ( 86.9% )
2nd Place:
Zagorovsky, V. (Games: 15)
Top 1 Match: 201/378 ( 53.2% )
Top 2 Match: 267/378 ( 70.6% )
Top 3 Match: 311/378 ( 82.3% )
Top 4 Match: 328/378 ( 86.8% )
3rd Place:
Estrin, Y. (Games: 15)
Top 1 Match: 201/339 ( 59.3% )
Top 2 Match: 263/339 ( 77.6% )
Top 3 Match: 291/339 ( 85.8% )
Top 4 Match: 301/339 ( 88.8% )
4th Place:
Thiele, E. (Games: 15)
Top 1 Match: 302/511 ( 59.1% )
Top 2 Match: 386/511 ( 75.5% )
Top 3 Match: 417/511 ( 81.6% )
Top 4 Match: 443/511 ( 86.7% ) |
|
| Sep-19-10 | | Zygalski: 7th Correspondence Chess World Championships 1972-75
Top 3 finishers
1st place
Estrin (10 games)
Top 1 Match: 166/276 (60,1% )
Top 2 Match: 208/276 (75,4% )
Top 3 Match: 227/276 (81,2% )
2nd place
Boey (13 games)
Top 1 Match: 268/449 (59,7% )
Top 2 Match: 342/449 (76,2% )
Top 3 Match: 376/449 (83,7% )
3rd place
Zagorovsky V (10 games)
Top 1 Match: 153/252 (60,7% )
Top 2 Match: 190/252 (75,4% )
Top 3 Match: 208/252 (82,5% )
Top 3 Players
Top 1 Match: 587/977 ( 60.1% )
Top 2 Match: 740/977 ( 75.7% )
Top 3 Match: 811/977 ( 83.0% ) |
|
| Sep-19-10 | | Zygalski: Kramnik-Topalov 2006 WC
Kramnik
Top 1 Match: 306/552 (55,4% )
Top 2 Match: 417/552 (75,5% )
Top 3 Match: 461/552 (83,5% )
Topalov
Top 1 Match: 309/555 (55,7% )
Top 2 Match: 417/555 (75,1% )
Top 3 Match: 447/555 (80,5% ) |
|
| Sep-19-10 | | Zygalski: Kasparov-Karpov 1985 WC
Karpov
Top 1 Match: 318/576 (55.2% )
Top 2 Match: 436/576 (75.7% )
Top 3 Match: 483/576 (83.9% )
Kasparov
Top 1 Match: 311/583 (53.4% )
Top 2 Match: 411/583 (70.5% )
Top 3 Match: 477/583 (81.8% ) |
|
| Sep-19-10 | | Zygalski: And compare to Yelena's:
Percentage agreement with Deep Rybka 3:
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% )
Top 4 Match: 698/723 ( 96.5% ) |
|
| Sep-19-10 | | Zygalski: Surely she should be World Champion by now & not 40th best female player if she matches like that unassisted with Deep Rybka 3 when not playing the games on chess.com? |
|
| Sep-19-10 | | bartonlaos: Zygalski, you are missing a couple of things that matter. The first thing that matters is whether the numbers for 20 games of Yelena Dembo are extraordinarily different from 20 games of the highest known 'normal' value. That's why I didn't list the others. The second thing that matters is the process by which these 20 games were picked. If chess.com is correct to assume the 20 games reflect her entire record, then her rating should be at least 3500. It isn't. If she wasn't cheating, then her correspondence rating should reflect her normal OTB rating. It does (cc 2500+ vs. otb 2500-). This suggests extreme bias in handpicking the 20 games used for analysis. The solution is to run the numbers on every single game in Yelena Dembo's database. This will obviously show that her agreement with Rybka 3 is normal when compared to other correspondence players of her 2500+ rating. If she files a countersuit, chess.com would be forced to explain their methods - these aren't published methods accepted by the peer-reviewed process - these 'methods' are the result of a bunch of hacks trying to catch cheaters. The methods have errors and can be exposed - that is, if Yelena Dembo wants the free press, or the chess.com website, depending on damages. |
|
| Sep-19-10 | | Zygalski: I understand that the 20 chess.com games were the 20 that she most recently played vs 2200+ rated chess.com players all of which had 35 or more total moves.
Also, Dembo's otb was analysed:
"I picked the last 20 games >= 35 moves from chessgames.com. All against high-rated opposition (lowest ELO in the set is 2222, highest is 2655)" thinking that perhaps Dembo's natural style of unassisted play is rather more engine-like than should be expected. The results of the analysis were:
Dembo OTB games:
Stockfish 1.8, 512MB hash, min/max ply=12/30, 40s/ply, 2GHz Core Duo:
Yelena Dembo (Games: 20)
{ Top 1 Match: 422/929 ( 45.4% )
{ Top 2 Match: 615/929 ( 66.2% )
{ Top 3 Match: 717/929 ( 77.2% )
{ Top 4 Match: 777/929 ( 83.6% )
So not nearly so engine-like otb it seems. |
|
| Sep-19-10 | | Zygalski: I do accept that from a practical point of view, chess.com have changed the WGM Yelena Dembo profile from being "account closed for cheating" to "account closed". A costly & lengthy legal threat would not suit a website! But let me ask you a question, do you think chess.com would remove a WGM unless they thought they had good reason to? |
|
| Sep-19-10 | | Zygalski: FYI here is the analysis of the 20 the most recently completed chess.com games vs 2200+ rateds, all with 35 or more total moves. These games were analysed by 4 different members on 4 different systems. The analysis excludes moves from a 4m game database: 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% )
Top 4 Match: 698/723 ( 96.5% )
Houdini 1.03a_1CPU @ 45 secs/move 416 MB Hash
Intel 2,4 gHz 1GB Ram
YelenaDembo Totals
Top 1 Match: 498/729 - 68,3%
Top 2 Match: 625/729 - 85,7%
Top 3 Match: 666/729 - 91,4%
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% )
{ Top 4 Match: 681/723 ( 94.2% )
Core2Duo 2.00GHz 1GB RAM
[Analysis "Houdini 1.03a w32 2_CPU Hash:256 Time:45s Depth:16-30ply"] w/MegaBase+MegaCorr+Tablebase cutoff YelenaDembo (Games: 20)
{ Top 1 Match: 493/694 ( 71.0% )
{ Top 2 Match: 609/694 ( 87.8% )
{ Top 3 Match: 644/694 ( 92.8% )
{ Top 4 Match: 660/694 ( 95.1% ) |
|
| Sep-19-10 | | dpruess: bartonlaos, the people working at chess.com are not hacks; they are a really awesome team, that does fantastic work. also, i don't know of us ever falsely banning a single player. |
|
| Sep-19-10 | | Jim Bartle: What means of defense does a player have to defend him- or herself after being banned or accused cheated? |
|
| Sep-19-10 | | Zygalski: Strange the most powerful system returned the highest match %'s as well. With Deep Rybka 3 too. |
|
Sep-19-10
 | | smurph: She is entitled to her good name.
Have you the analysis of similarly strong players playing 2200. |
|
| Sep-19-10 | | BobCrisp: A list of other players accused of computer cheating: Von Neumann
Clemens Allwermann
Arkadi Naiditsch
Veselin Topalov
Vladimir Kramnik |
|
| Sep-19-10 | | bartonlaos: dpruess,
I know they are an awesome team. But if the chess.com workers are not professionally-trained statisticians following published methods to make this determination then Chess.com is opened to a lawsuit. This is because an awesome team of professionally-trained statisticians working for Yelena Dembo will be able to find serious holes in your methods. Your defense must show that such methods are accepted by the peer-reviewed process. If they aren't published, they can't be defended and Yelena Dembo would win. So the bottom line is whether Chess.com can defend its methods in court. If they can't be defended, then Chess.com needs to consider restoring Yelena Dembo's status. Zygalski - OTB numbers are irrelevant. |
|
| Sep-19-10 | | SugarDom: Well, Yelena never paid for her account. So what's the liability if chess.com does not allow to play in the site? That's their discretion... |
|
Sep-19-10
 | | Eric Schiller: Anyone can sue for any reason.But chess.com can make any regulations they want as long as they are applied without discrimination against minorities. They are under no obligation to demonstrate fairness. They can terminate accounts for any breach of standards. It is different if the make a public accusation of cheating as that gets into defamation. That said, unless the alleged breach takes place in a competition with real prizes it isn't clear that draconian action is needed unless warnings have been issued and ignored. Personally, I find absence of errors more suspicious than agreeing with computer analysis. Humans make mistakes,even the best. Computers don't blunder. |
|
| Sep-20-10 | | Zygalski: Exactly Eric.
Dembo's record before her chess.com account was closed was 140 wins, 15 draws & not a single loss, despite virtually all games being against players in the 2100-2700 rating range. Some of those players will no doubt have been using "help" too. The 20 games she played online which were analysed by those members are (unfortunately for Yelena) chess.com's property, so anyone can find them, analyse them & draw their own conclusions. Yelena's own chess.com rating was over 2800 & she was the #21 highest rated player. With her record she would no doubt have gone further up the ranking list, too. User: bartonlaos
You really don't find any of this strange at all, no? |
|
 |
|
|
< Earlier Kibitzing · PAGE 14 OF 22 ·
Later Kibitzing> |
Kibitzing
