Tweet
Recently, Felix Dumont posted the game Sokolov - Kovalyov from the 2013 Quebec Open round 5 here:
http://www.chesstalk.info/forum/show...0950#post70950
In this post, Felix comments that this game is "definitely a must-see if you are looking for an example of precise GM play."
A little earlier this year, in this post,
http://www.chesstalk.info/forum/show...9839#post69839
Felix says "If I caught someone playing like (Borislav Ivanov) while I am a TD, I would disqualify him on the spot. There's no need to do further investigations..."
Borislav Ivanov is a supposed chess cheat. The evidence against him is not that he was caught with any electronic device (despite some efforts, that has not yet happened). Instead, the evidence is superficial: based on statistical comparisons of computer engine moves versus Ivanov's moves, and declaring that there is too close a resemblance between the two. Dr. Kenneth Regan, an occasional poster here, has gone to great lengths to procure such evidence using his own computer analysis, and has apparently concluded Ivanov is guilty as charged. I remind readers that such work by Dr. Regan does serve a possible ulterior motive of enhancing Dr. Regan's professional status, given that he is a computer professor and researcher at a Buffalo university. It is even possible that Dr. Regan may be vying to be the judge and jury of last resort for chess federations everywhere, perhaps even for a fee. None of this is proven, but computer statistics may some day bear it out... LOL, "hoisted by his own petard".
Getting back to Felix Dumont: the first thing that is wrong here is that Felix is praising "precise GM play" while simultaneously saying that playing too precise should disqualify someone on the spot, no further investigations needed. Obviously there is a disconnect there.
I got to wondering with respect to the Sokolov - Kovalyov game, just HOW precise was it? I decided that the first weekend I had free I would run the moves through Stockfish 2.3.1, rated approximately 3200 ELO.
Well, folks, the verdict is in. I analyzed moves 9 through 53 for both White and Black (45 moves total for each side). Depending on where one draws the line on such superficial methods, Sokolov can be suspected of cheating. But the real shocker is that Felix' hero, Kovalyov, was MOST DEFINITELY cheating... if you are one of those who believes in these methods.
Now, keep in mind, one doesn't have to be emulating Houdini's top choices to be cheating. There are probably many dozens of computer engines all rated over the 3100 ELO level, high enough that if any one of them could be used by a third party to supply moves to a tournament player during a game, that player should easily win the game. This makes clear one of the major problems with catching such cheaters if they are really out there avoiding detection of their electronic devices: the sheer number of 3100+ ELO engines, and the fact they all differ by some degree on their choice of moves, makes the whole prospect a combinatorial nightmare. How do we possibly test each and every engine, given that the cheater may choose one less conspicuous than Houdini?
The engine I used, Stockfish 2.3.1 which is freely available, is rated around 3200 at normal tournament time controls. In my analysis, I had it analyze to at least 24 plies... and if that was too quick, I had it continue on to as high as 29 plies.
I can supply anyone who requests it the Excel spreadsheet that contains my results. But here is the summary:
Number of moves analyzed: 45 per side
List of Stockfish 2.3.1 Move Rankings Actually Played From Moves 9 to 53 Inclusive:
Sokolov (W): 3,1,1,1,1,2,2,2,1,1,1,2,2,4,1,2,1,1,2,2,1,8,1,1,1,1,1,1,1,1,1,4,1,1,1,1,1,2,1,8,1,1,1,12,1
Kovalyov (B): 1,1,2,2,1,1,1,12,1,2,2,3,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,1,1,3,1,1,1,1,1,1,2,1,1
For those unfamiliar with top-3 methodology for detecting engine use, top-1 is the percentage of positions where the 1st engine choice was played, top-2 is the percentage of positions where either the 1st or 2nd engine choice was played, etc.
Sokolov
Top 1 Match: 30/45 66.7%
Top 2 Match: 39/45 86.7%
Top 3 Match: 40/45 88.9%
Kovalyov
Top 1 Match: 35/45 77.8%
Top 2 Match: 42/45 93.3%
Top 3 Match: 44/45 97.8%
Kovalyov's numbers definitely lie in the cheating zone for those who believe in the integrity of such data. Only one of Kovalyov's moves was not in Stockfish's top 3 choices. Black's move 16 was the outlier, but even that move, Stockfish's 12th choice, was only ranked 0.40 points below Stockfish's top choice. Less than half a pawn. A perfect opportunity for a cheater to slip in such an outlier to try and avoid suspicion should anyone actually do this kind of analysis.
Now, let it be known that I am not personally accusing Kovalyov of cheating in this game. But Felix Dumont, Nigel Hanrahan, Tom O'Donnell and Kenneth Regan all need to respond to this data with the very action that they would expect of others: a formal accusation against Kovalyov for cheating, a demand that he refund his prize monies for the event and that he be suspended for some significant period. In Felix's own words, "There's no need to do further investigations."
Ok, gang: time to put up or shut up. If organizers everywhere are to start prosecuting these cases, you can start right now with one of your own.
Personally, I think Kovalyov just played a very computer-like game. Perhaps he does it a lot. And I know that none of the above named Gang of Four are going to accuse Kovalyov, even if they do their own analysis and come up with the same result. They just don't have the backbone to stand up to their own convictions (excuse the pun).
When it comes to this whole matter of computer cheating in chess, the first $64,000 question is: how precise is too precise? Or in other words, where is the line drawn between cheating and just playing an amazing computer-like game of "precise GM play"?
This is the question that the likes of Felix Dumont, Nigel Hanrahan, Tom O'Donnell, and even Dr. Regan will not and presumably cannot answer. Or they might say something like "we need more than just one game of evidence." But keep in mind, a cheater may not be cheating in every game of an event. They may save their cheating until the penultimate or the last round, when the money is on the line. This kind of cheating could be going on right now under everyone's nose. If Kovalyov did cheat, he chose a prime opponent to do it against. As long as it MIGHT be happening in events around the world right now, using any one of over 100 computer engines, picking and choosing key games to cheat in, chess is stained.
And the next $64,000 questions is: who gets to look at the numbers and decide each individual case?
P.S. I will in the weeks to come do a similar analysis using the same engine and settings of one of Ivanov's supposed cheating games, and when done, I will post the numbers. Or if someone else has that engine and the time, go ahead and post your results.
http://www.chesstalk.info/forum/show...0950#post70950
In this post, Felix comments that this game is "definitely a must-see if you are looking for an example of precise GM play."
A little earlier this year, in this post,
http://www.chesstalk.info/forum/show...9839#post69839
Felix says "If I caught someone playing like (Borislav Ivanov) while I am a TD, I would disqualify him on the spot. There's no need to do further investigations..."
Borislav Ivanov is a supposed chess cheat. The evidence against him is not that he was caught with any electronic device (despite some efforts, that has not yet happened). Instead, the evidence is superficial: based on statistical comparisons of computer engine moves versus Ivanov's moves, and declaring that there is too close a resemblance between the two. Dr. Kenneth Regan, an occasional poster here, has gone to great lengths to procure such evidence using his own computer analysis, and has apparently concluded Ivanov is guilty as charged. I remind readers that such work by Dr. Regan does serve a possible ulterior motive of enhancing Dr. Regan's professional status, given that he is a computer professor and researcher at a Buffalo university. It is even possible that Dr. Regan may be vying to be the judge and jury of last resort for chess federations everywhere, perhaps even for a fee. None of this is proven, but computer statistics may some day bear it out... LOL, "hoisted by his own petard".
Getting back to Felix Dumont: the first thing that is wrong here is that Felix is praising "precise GM play" while simultaneously saying that playing too precise should disqualify someone on the spot, no further investigations needed. Obviously there is a disconnect there.
I got to wondering with respect to the Sokolov - Kovalyov game, just HOW precise was it? I decided that the first weekend I had free I would run the moves through Stockfish 2.3.1, rated approximately 3200 ELO.
Well, folks, the verdict is in. I analyzed moves 9 through 53 for both White and Black (45 moves total for each side). Depending on where one draws the line on such superficial methods, Sokolov can be suspected of cheating. But the real shocker is that Felix' hero, Kovalyov, was MOST DEFINITELY cheating... if you are one of those who believes in these methods.
Now, keep in mind, one doesn't have to be emulating Houdini's top choices to be cheating. There are probably many dozens of computer engines all rated over the 3100 ELO level, high enough that if any one of them could be used by a third party to supply moves to a tournament player during a game, that player should easily win the game. This makes clear one of the major problems with catching such cheaters if they are really out there avoiding detection of their electronic devices: the sheer number of 3100+ ELO engines, and the fact they all differ by some degree on their choice of moves, makes the whole prospect a combinatorial nightmare. How do we possibly test each and every engine, given that the cheater may choose one less conspicuous than Houdini?
The engine I used, Stockfish 2.3.1 which is freely available, is rated around 3200 at normal tournament time controls. In my analysis, I had it analyze to at least 24 plies... and if that was too quick, I had it continue on to as high as 29 plies.
I can supply anyone who requests it the Excel spreadsheet that contains my results. But here is the summary:
Number of moves analyzed: 45 per side
List of Stockfish 2.3.1 Move Rankings Actually Played From Moves 9 to 53 Inclusive:
Sokolov (W): 3,1,1,1,1,2,2,2,1,1,1,2,2,4,1,2,1,1,2,2,1,8,1,1,1,1,1,1,1,1,1,4,1,1,1,1,1,2,1,8,1,1,1,12,1
Kovalyov (B): 1,1,2,2,1,1,1,12,1,2,2,3,2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,2,1,1,3,1,1,1,1,1,1,2,1,1
For those unfamiliar with top-3 methodology for detecting engine use, top-1 is the percentage of positions where the 1st engine choice was played, top-2 is the percentage of positions where either the 1st or 2nd engine choice was played, etc.
Sokolov
Top 1 Match: 30/45 66.7%
Top 2 Match: 39/45 86.7%
Top 3 Match: 40/45 88.9%
Kovalyov
Top 1 Match: 35/45 77.8%
Top 2 Match: 42/45 93.3%
Top 3 Match: 44/45 97.8%
Kovalyov's numbers definitely lie in the cheating zone for those who believe in the integrity of such data. Only one of Kovalyov's moves was not in Stockfish's top 3 choices. Black's move 16 was the outlier, but even that move, Stockfish's 12th choice, was only ranked 0.40 points below Stockfish's top choice. Less than half a pawn. A perfect opportunity for a cheater to slip in such an outlier to try and avoid suspicion should anyone actually do this kind of analysis.
Now, let it be known that I am not personally accusing Kovalyov of cheating in this game. But Felix Dumont, Nigel Hanrahan, Tom O'Donnell and Kenneth Regan all need to respond to this data with the very action that they would expect of others: a formal accusation against Kovalyov for cheating, a demand that he refund his prize monies for the event and that he be suspended for some significant period. In Felix's own words, "There's no need to do further investigations."
Ok, gang: time to put up or shut up. If organizers everywhere are to start prosecuting these cases, you can start right now with one of your own.
Personally, I think Kovalyov just played a very computer-like game. Perhaps he does it a lot. And I know that none of the above named Gang of Four are going to accuse Kovalyov, even if they do their own analysis and come up with the same result. They just don't have the backbone to stand up to their own convictions (excuse the pun).
When it comes to this whole matter of computer cheating in chess, the first $64,000 question is: how precise is too precise? Or in other words, where is the line drawn between cheating and just playing an amazing computer-like game of "precise GM play"?
This is the question that the likes of Felix Dumont, Nigel Hanrahan, Tom O'Donnell, and even Dr. Regan will not and presumably cannot answer. Or they might say something like "we need more than just one game of evidence." But keep in mind, a cheater may not be cheating in every game of an event. They may save their cheating until the penultimate or the last round, when the money is on the line. This kind of cheating could be going on right now under everyone's nose. If Kovalyov did cheat, he chose a prime opponent to do it against. As long as it MIGHT be happening in events around the world right now, using any one of over 100 computer engines, picking and choosing key games to cheat in, chess is stained.
And the next $64,000 questions is: who gets to look at the numbers and decide each individual case?
P.S. I will in the weeks to come do a similar analysis using the same engine and settings of one of Ivanov's supposed cheating games, and when done, I will post the numbers. Or if someone else has that engine and the time, go ahead and post your results.
Comment