Stigma wrote on 06/07/13 at 14:57:30:


My gut feeling is that a reasonably strong, "chess cultured" amateur (2200-2300 FIDE) could play like a 2500-2600 GM without much suspicion.  Obviously if they have an established, stable rating, and suddenly start performing at GM level that's very suspicious, but you know what I mean.

Even in this case I suspect that there will be some areas of suspicion, for instance not knowing the standard techniques in complicated but essentially theoretical endgaems (Rook, f, and h-pawn vs. Rook, for instance).  But these scenarios will be few and far between.   

I think that a very determined cheat, especially with an accomplice, will succeed.  In some sense this isn't exactly new, though.  Cheating has occurred with accomplices in various forms forever, such as a 2300-rated coach helping his junior players (which I saw first hand and fell victim to in California once).  It's just that now you don't even need a very strong accomplice to help you--just some software. 

The only real solution I see are "wireless-free" playing areas, which are very easy to implement from a technological perspective, but nearly impossible from a legal one (as far as I know).  Perhaps laws will change as cell phones and other devices become a problem in other areas, like university exams and the like.

Smyslov_Fan wrote on 06/07/13 at 15:08:57:

It's sad if that's all we can do, though? "Human-like" centaur play done skillfully may be virtually indistinguishable from the play of the young players you mention. Which is why I maintain that the only long-term solution is physical ways to detect cheating that attempts to catch any cheater, not just those with highly suspicious match-up-rates.

ErictheRed wrote on 06/07/13 at 03:23:08:

Allright, yes, I already agree with you it's clear Ivanov is cheating. What interested me in my last few posts was how someone could have done it more subtly to avoid detection while still gaining in playing strength.

Maybe you would need a human involved to avoid "looking" to much like a computer then. Do you have any thoughts on the accomplice/centaur scenarios?

http://pogonina.com/index.php?option=com_content&task=view&id=1763&Itemid=1&&lan...

The Sad Case of Borislav Ivanov
by GM Daniel Gormally
Thursday, 06 June 2013

@ErictheRed:

First, some of the suggested scenarios involve an accomplice. A reasonably strong human player can look at a list that reads

31.Qh1 +3.20
31.Nxg5 +1.80
31.Be3 +0.00
...
and conclude quite quickly "I would never even consider Qh1", discard it as a "computer move", and send the second-best but also winning Nxg5 instead. He could even be on the lookout for positions where the difference between the engine's top x moves (5, 7 or whatever) is especially small, and choose a lower one that looks humanly logical in those cases. In your candidate move list, the differences in evaluation are small enough that they may not matter much in a human game.

In short, the accomplice could simply be a "centaur" (human player with computer support) instructed to give higher priority to human judgement than he normally would if trying to maximize playing strength.

Second, shouldn't it be possible to get engines closer to human play (at some cost to playing strength, obviously)? Human biases are well-known and books have been written about them (Afek/Neiman: Invisible Chess Moves). So you could assign lower values to backward bishop moves, moves that put more than two pieces en prise, wrecked pawn structures, moves that leave most of one's pieces undefended, material compared to piece activity (since humans are worse at consolidating against an initiative); higher value to king safety and healthy pawn structures etc.

MartinC wrote on 06/06/13 at 19:01:33:


You forgot /immoral. But when we see how eager Ivanov is to tarnish the reputation of both our favorite game and his fellow players, he could well be some sort of psychopath. We're also lucky he's not been "smart" enough to cheat in a less detectable way, i.e. by only following the computer in some positions, increasing performance more gradually, or using some custom-built engine that mimics strong human play. But here I go giving the wrong people ideas again... 

These more subtle strategies are the reason we need some sort of technical, physical way to detect cheating in every tournamnent were real money is at stake, and we need it fast. So it's good that ErictheRed and other knowledge people are starting to suggest such technical solutions. If we have to depend on suspicious, consistently "computer-like" play and only search those who are already under suspicion, more subtle cheaters will get a free pass.

I want to reiterate for the umpteenth time in this thread that it is not that Mr. Ivanov plays good chess that he's under suspicion, rather, that his particular brand of good chess correlates so highly to the output of an engine.

Going all the way back to the first great tournament result of his (I haven't bothered to follow all of his later games), I remember that he played the first or second choice (and never deviated by more than something like '0.06' from the first choice) of Houdini running on my laptop for 50+ moves in that very boring endgame where he blundered at the very end.  I think that the likelihood of any human player doing that is vanishingly small.  For instance, his GM opponent was able to hold a draw throughout (right up until the last blunder when it was supposed that Mr. Ivanov accidentally moved the wrong piece to the correct square) by playing "human" moves that Houdini was far less keen on.  There were certainly other ways of playing (and other orders of moves) that were just as good; it's incredibly unlikely that Mr. Ivanov came up with Houdini's way of playing entirely without assistance.  And that's just one game, there were many others like that...anyhow...

As an electronics engineer, I'd say that the camera/earpiece proposition is not entirely implausible, but would take a fair bit of work to get up and running properly.  Of course, a junior or senior engineering student at a university could do something like this as part of a project at school (or on their own time); it's possible.  Still, you need a power source, cabling, transmitter, and receiver.  It should be entirely trivial to detect this method, so long as you have the authority to use a hand-held metal detector or give a pat down.

If he's cheating, then catching him should be relatively easy, if the arbiters have the authority to search him mid-game.  Perhaps if there is money at stake in a tournament, the case becomes the jurisdiction of Bulgarian law enforcement officers, and a proper search could be done during one of his excellent performances?  A wireless jammer (though usually illegal to be used without a license) would also do the trick.  He may be using cell phone frequencies, or perhaps something higher, like 2.4 GHz. 

Is there enough evidence to warrant that?  Would Bulgarian police become involved in this sort of thing?  I don't know. 

Actually, an antenna and spectrum analyzer would detect his transmissions trivially--but spectrum analyzers cost a lot of money!  It would be nice if chess tournaments could be held in some kind of wireless-free room, but again, you need a license to use a "jammer," and locking everyone inside a Faraday cage isn't going to happen...

No, not that scary. Still far too big really.

When it gets scary for the existence of chess is when you can't expect it to find anything even if you search someone.

Receivers in shoes are much safer than ear pieces anyway.

Its capturing/processing the board which isn't that trivial.