|
< Earlier Kibitzing · PAGE 18 OF 22 ·
Later Kibitzing> |
| Sep-21-10 | | CCplayer: Cotoi, probably not, but if you have a fairly large sample and want to prove a point, then being able to select which part of the sample to use could make a huge difference. Point taken though about the detection methodology. I understand Tamars point of view too, and basically I agree - it is the naming and shaming that really matters here. |
|
| Sep-21-10 | | bartonlaos: This is my last bit. Chess.com mistakenly assumes that the opening database of a Titled player's pet repertoire extends to the same limit that is publicly accessed by Chess.com's analyzer. Chess.com provides a threshold of 35 moves for a game to be analyzed. Suppose that in one game the personal database extends 5 moves longer than the one used for Chess.com's analysis. Now consider that in this game, as with all games, not all moves played are equivalent. If there is a clear winning line, the strongest players will find it. If the rating of the Titled player's opponent is hundreds of points lower, then this makes it more likely that opponent errors will create clear winning lines for the Titled player to find. <Multiple moves 100% agreement no cheating.> Now suppose in this game, they also find a mate-in-8! Just these few considerations, should skew the results against Titled players who had never cheated. |
|
| Sep-21-10 | | bartonlaos: Tamar, they can ban whom they wish, but it's important to understand that such Titled players as Yelena could be unfairly banned because they constantly play much lower-rated opponents. |
|
| Sep-21-10 | | bartonlaos: I hope Yelena's reputation doesn't unfairly suffer too much from this. By the way - 'Cheater' 3 (Nosferatu) is FM Stefan Docx. |
|
| Sep-21-10 | | chesswizz1964: *bartonlaos*
I agree there is no way Yelena is using program. Her natural chessplaying beauty is easy to see. No program play chess so human as this game which is 1 of the 20 I find here: http://lousyatchess.blogspot.com/
http://www.chess.com/echess/game.ht...
[Event "NaScp with Miss - Board 2"]
[Site "Chess.com"]
[Date "2010.05.01"]
[White "kingkoy301"]
[Black "YelenaDembo"]
[Result "0-1"]
[WhiteElo "2503"]
[BlackElo "2767"]
[TimeControl "1 in 3 days"]
[Termination "YelenaDembo won by resignation"]
1.d4 Nf6 2.Nf3 g6 3.Nc3 d5 4.Bf4 Bg7 5.e3 O-O 6.h3 c5 7.dxc5 Nbd7 8.Be2 Ne4 9.Nxd5 e6 10.Nc7 e5 11.Nxa8 exf4 12.exf4 Bxb2 13.O-O Bxa1 14.Qxa1 Ndxc5 15.Rd1 Bd7 16.Ne5 Qc8 17.f3 Nf6 18.Qd4 Bf5 19.g4 Ne6 20.Qe3 Ng7 21.Qxa7 Be6 22.Nb6 Qc5+ 23.Kf1 Nfh5 24.gxh5 Nxh5 25.Qxb7 Nxf4 26.Ng4 Nxh3 27.Ke1 Bxg4 28.Nd5 Qf2+ 29.Kd2 Be6 30.Nf6+ Kg7 31.Qb2 Kh6 32.Ne4 Rd8+ 33.Kc1 Rxd1+ 34.Bxd1 Qe3+ 35.Kb1 Nf2 36.Nxf2 Qxf2 37.a4 Qd2 38.Qc1 Qxc1+ 39.Kxc1 Kg5 40.a5 h5 41.a6 h4 0-1 It is no in theory when 8...Ne4 is played says
http://www.365chess.com/opening.php
And I play on Rybka 2.3.2a on dual core then I only count very few #1 moves. Most are the exchanges. Yelena often pick moves which are worse than half pawn from best but she win because she is good human. Good human (WGM) can always beat a Rybka not even playing that many good moves! |
|
| Sep-21-10 | | bullyboy12708: Ok so how about then you look at all Yelena's games from chess.com. All games against players 2200+ with a minimum of 35 moves. Would you consider that "cherry picked"?? |
|
| Sep-21-10 | | cotoi: I analyzed the game posted by chesswiz using Firebird at depth 18, in multiline mode (3 lines). Excluding the opening theory (so beginning with move 8...Ne4), my results look like this: - move 13 was the second choice of the engine
- move 35 is not among the top 3 choices of the engine
- all the other moves are the first choice of Firebird. I don't remember to see such a matching between a human and an engine, even if the human player was called Fischer, Kasparov or Kramnik. Of course, that doesn't mean it is not possible that humans play like computers. |
|
| Sep-21-10 | | e4d4: <- move 13 was the second choice of the engine
- move 35 is not among the top 3 choices of the engine - all the other moves are the first choice of Firebird.> Sounds like a 3000 elo player |
|
| Sep-21-10 | | chesswizz1964: *cotoi*
ok so maybe my program not so good.
I HAVE found a game from the 20 I believe ANY machine will not like from Yelena's moves. While there are some that it may like, many are not so good (sorry Yelena) and she sometimes loses nearly half a pawn!
If there are any strong players on chessgames.com with strong program, please clear WGM Dembo's name by at least looking at this game.Fritz.. no
Chess master... no
Rybka....no!
Yelena... YES!
http://www.chess.com/echess/game.ht...
[Event "10th Chess.com Tournament (2201+) - Round 1"]
[Site "Chess.com"]
[Date "2010.03.01"]
[White "YelenaDembo"]
[Black "wh47"]
[Result "1-0"]
[WhiteElo "2747"]
[BlackElo "2377"]
[TimeControl "1 in 3 days"]
[Termination "YelenaDembo won by resignation"]
1.e4 e5 2.Nf3 Nc6 3.Bb5 a6 4.Ba4 Nf6 5.O-O Be7 6.Re1 O-O 7.Bxc6 dxc6 8.Nxe5 Re8 9.d3 Bb4 10.c3 Rxe5
11.cxb4 Qd4 12.Qd2 Re8 13.Qc3 Rd8 14.Bf4 Be6 15.Bxc7 Ng4 16.Qxd4 Rxd4 17.f3 Rd7 18.Bf4 Nf6 19.Nc3 Rxd3 20.Red1 Rad8
21.b3 h5 22.Kf2 Kf8 23.Rxd3 Rxd3 24.Ne2 Ne8 25.Be3 Rd7 26.Nf4 Nf6 27.h4 Ke8 28.Bb6 g6 29.Ke3 Nh7 30.Nd3 f5
31.Ne5 Rg7 32.Rd1 fxe4 33.Rd8+ Ke7 34.Rb8 exf3 35.Rxb7+ Kf6 36.Bd8+ Kxe5 37.Rxg7 fxg2 38.Rxg6 Bd5 39.Kf2 Nf8 40.Rg5+ Kd4
41.Rxh5 Ne6 42.Bf6+ Kd3 43.Be5 Kc2 44.Rf5 Kb1 45.h5 Be4 46.Rf7 Ng5 47.Rg7 g1=Q+ 48.Kxg1 Nf3+ 49.Kf2 Nxe5 50.Ke3 Bc2
51.Kd4 Nf3+ 52.Kc3 1-0
This game no in theory at 11.cxb4
Please help Yelena out. |
|
| Sep-21-10 | | bullyboy12708: Yelena had 20 (yes TWENTY) games analysed using top 3 analysis so to examine just 1 game to try and clear her name is inadequate wouldn't you think? |
|
| Sep-21-10 | | cotoi: http://www.chess.com/echess/game.ht...
OK, so for the game posted above, here is what Firebird at depth 18 thinks: - outside top three choices: one move, 45
- third choice: 3 moves, 20, 27, 44
- second choice: 5 moves, 28, 31, 36, 39, 43
- the rest of 33 moves are the top choice of Firebird. I should also mention that let's say at move 30 (maybe earlier), it is clear who is the winner. Black's pieces are so uncoordinated that, although the material is even, his position is unplayable. |
|
| Sep-21-10 | | chesswizz1964: I don't believe.
Must be more than single move outside the top 3. There are over 40 moves when game leaves theory. 33 first choice moves? Get a proper chess program and Yelena's innocence would be clear. |
|
| Sep-21-10 | | BabalooMoon: <cotoi> The game above follows Korchut-Glogowski (2005 email) until Black's 15th move. There are only 37 remaining moves to evaluate. <bartonlaos> <The serious flaw is a lack of proper normal controls> I couldn't agree more, I applaud your efforts. |
|
| Sep-21-10 | | cotoi: Ok, so let's update the results:
- outside top three choices: one move, 45
- third choice: 3 moves, 20, 27, 44
- second choice: 4 moves, 28, 31, 36, 39, 43
- the rest of 29 moves are the top choice of Firebird. In percentages:
- top choice: 78.3 %
- top 2 choices: 86.4 %
- top 3 choices: 97.3 %
Has anything changed? |
|
| Sep-21-10 | | bartonlaos: <BabalooMoon> Thanks! cotoi,
The game cited by chesswizz1964 is an example of why statisticians say each move should not be equivalently measured. Because it is filled with a lot of forcing moves or clear winning lines above candidates, and consistent simplification. It also contains a "human-like" exchange to a winning endgame: Black to move:
click for larger view
37...Qd2
So you must study the game not the tally of moves. As an aside, her opponent <kingkoy301>, a member in good-standing at Chess.com, played with virtually 100% Top-choice agreement. His exception is 32. Ne4. Did he cheat? Of course not, just study the game and you'll see it for yourself. Even though it may appear that someone played with engine-like precision, it does not mean they used a computer. This single game shows how erroneous it is to take engine-fidelity as any indication of 'cheating' without considering the Chess itself. It demonstrates why the Top-3 method is meaningless to serve as justification for defamation. There is no one outside the awesome team at Chess.com, including yourself, who has proper justification to accuse Yelena in this manner. |
|
| Sep-21-10 | | bartonlaos: (And Chess.com withdrew its accusation!) |
|
| Sep-21-10 | | cotoi: First I do not accuse Yelena of anything. I just analyzed some games. As for the game mentioned above, it was clear that, for the last couple of moves Yelena has a winning position and tries to simplify. So yes, some of her moves were logical. In general, if you make a good move, it doesn't mean you cheat. Also, when a candidate move stands clearly above the other candidate moves, it should not be surprising that a strong player chooses it. However Yelena appears to match an engine's choice much more often than Kasparov. But, as you say, maybe Kasparov didn't study his games enough. Or maybe it's just a coincidence which only happened in two games. |
|
| Sep-21-10 | | bartonlaos: Or maybe it's just that Kasparov wasn't playing cc games against an opponent hundreds of points below his own. The stronger the opponent, the greater their ability to find good moves, making it harder for you to find the clear winning lines. Conversely, the weaker is your opponent, the easier it is to maintain fidelity. As for analyzing or interpreting what the percentages mean?? Without the proper controls any percentage you get from analyzing those games is arbitrary. |
|
| Sep-21-10 | | Fezzik: Chess.com did not withdraw its accusation. On legal advice, they removed the comment that WGM Dembo had cheated. Dembo's account remains closed at chess.com, and chess.com has said that it stands by its decision. (And no, I don't work for chess.com.) |
|
| Sep-21-10 | | bartonlaos: Two separate issues. Chess.com is standing by their decision to ban her from the site, which is their own business and can be made any way they so decide. As Tamar wrote, they are using a broad-brush stroke. This method contains major flaws, but they must accept the flaws even though they won't hold up in court. Nevertheless, these flaws are serious enough for them to withdraw the accusation - her name does not appear among the list of cheaters recently published on their website. |
|
| Sep-21-10 | | cotoi: True, Kasparov didn't play CC games but it looks that none of the CC world champions from the past didn't know to play CC games. Or maybe they didn't study their games enough, even though they took years to complete a single game. None of them achieved such a high agreement with a chess engine. I think that the percentages are clear for any human with normal intelligence. For example, 78.3% of all the moves Yelena made in this game http://www.chess.com/echess/game.ht...
excluding the opening moves, appear to be the moves preferred by Firebird at ply depth 18. There are many examples when very strong players like super-GMs get to play against much weaker opponents (for instance open tournaments). However, as far as I know, nobody matched an engine so consistently. |
|
| Sep-21-10 | | bullyboy12708: What about analysing this game then?
[Event "10th Chess.com Tournament (2201+) - Round 1"]
[Site "Chess.com"]
[Date "2010.03.01"]
[White "YelenaDembo"]
[Black "saksipotku"]
[Result "1-0"]
[WhiteElo "2723"]
[BlackElo "2371"]
[TimeControl "1 in 3 days"]
[Termination "YelenaDembo won by resignation"]
1.e4 e5 2.Nf3 Nf6 3.d4 Nxe4 4.Bd3 d5 5.dxe5 Be7 6.O-O O-O 7.Nc3 Nxc3 8.bxc3 Nc6 9.Re1 f6 10.Rb1 fxe5
11.Nxe5 Nxe5 12.Rxe5 Bd6 13.Qh5 h6 14.Re1 Qf6 15.Qxd5+ Kh8 16.Be3 Qe5 17.Qxe5 Bxe5 18.Bc5 Bd6 19.Bxd6 cxd6 20.Re7 Rb8
21.h3 Rd8 22.Rb4 Bd7 23.Rd4 Bc6 24.Rg4 Rg8 25.Rg6 Rge8 26.Rc7 Rbc8 27.Rcxg7 Rg8 28.Rxg8+ Rxg8 29.Rxh6+ Kg7 30.Rxd6 Re8
31.f4 Re1+ 32.Kf2 Rd1 33.g4 Rh1 34.Rxc6 bxc6 35.Bf1 Rh2+ 36.Bg2 1-0 |
|
| Sep-21-10 | | bullyboy12708: Or this game?
[Event "Burning European Pawns - Board 4"]
[Site "Chess.com"]
[Date "2010.02.08"]
[White "YelenaDembo"]
[Black "Ambrosia"]
[Result "1-0"]
[WhiteElo "2697"]
[BlackElo "2528"]
[TimeControl "1 in 3 days"]
[Termination "YelenaDembo won by resignation"]
1.e4 d6 2.d4 Nf6 3.Nc3 g6 4.f4 Bg7 5.Nf3 O-O 6.Bd3 Na6 7.e5 Ng4 8.h3 Nh6 9.Be3 c5 10.Qe2 Nf5
11.Bxf5 Bxf5 12.O-O-O Rc8 13.g4 Bd7 14.Kb1 Re8 15.f5 Qa5 16.fxg6 hxg6 17.a3 cxd4 18.Bxd4 Be6 19.Rhe1 dxe5 20.Nxe5 Nb8
21.Qf2 Nc6 22.Nxc6 Rxc6 23.Bxg7 Kxg7 24.Ne2 Qb6 25.Qg3 Rd6 26.Nf4 Red8 27.Rxd6 Rxd6 28.h4 Qd4 29.Nxe6+ Rxe6 30.Rxe6 Qd1+
31.Ka2 Qd5+ 32.Qb3 fxe6 33.Qxd5 exd5 34.Kb3 e5 35.c4 1-0 |
|
| Sep-21-10 | | bartonlaos: cotoi,
There is a very good reason not to accept percentages alone. They don't relate to anything. If you want something to relate, then why don't you report both her and her opponent's Firebird fidelity for that game after the opening database terminus at move 15. It would be interesting to see how <wh47>, a 'normal' opponent that is rated hundreds of points below her performs. The bottom line is that there is nothing to relate to Dembo's numbers. Without a proper set used for the normal control, there is no way to understand whether her fidelity is extraordinarily significant. All other sets are irrelevant, even Kasparov's. ---
PS - Here is the previous game I had mentioned, regarding the apparent total fidelity of <kinkoy301>:
http://www.chess.com/echess/game.ht... |
|
| Sep-21-10 | | Jim Bartle: What's wrong with that last game? I don't see anything odd about a high-rated player playing it? |
|
 |
|
|
< Earlier Kibitzing · PAGE 18 OF 22 ·
Later Kibitzing> |
Kibitzing
