How to Catch a Chess Cheater

But this would be misleading... 

I analyzed a game yesterday and at some point the computer gave me +0.80 if I play "the best move", and just around +0.15 if I play the second-best move. 

The second-best was easy to see and rather intuitive. It was based on a calculation that was adequate to my playing strength, and indeed, that was what I play. 

However, the best move was based on letting him grab a free pawn, while threatening my queen. 

Then you need to see that you need to actually let him take the queen

The you need to see that after he takes your queen, you take the exchange...

And you need to evaluate the position based on 3100 level understanding to see that you're actually winning. Even five moves deeper into the position, you can go astray by taking a piece rather than making a subtle bishop retreat - down a queen still, mind you - and suddenly the positions are equal again - equal to Houdini, doesn't mean that you would personally survive this...

So if I was just told that I was +0.80 but without being shown that crazy variant, I would feel simply frustrated that I can't make the position "feel" like +0.80 (which I really can't... that variation was totally crazy and deeply invisible for someone 1900-2000). 

Such cases are not rare or special, by the way... 

did you ever try telling a much weaker player, when he's playing, that his position is much better, when it really is, but saying nothing more?

You did notice that it doesn't really give him confidence, didn't you?! :-) that's because he really has no idea what to do with this information. The only thing you do is embarrass him, by showing him that he doesn't know how, and that you do. 

I agree, when I look up openings in correspondence chess I will frequently go for the "lame" slight advantage rather than the "big but incomprehensible" type of advantage. I put understandability (for me) above any engine evaluation...

That's the safe way to go about it...

Another option, equally valid, would be to temporarily crush your rating by taking exactly the incomprehensible route, be mercilessly crushed by your opponent after ten moves - then look it all over with your favorite engine. Now you understand it all, and ready to spring it on your next unsuspecting opponent - provided you can remember all the shades, details and nuances...

The IPR tool offers many insights.  However, its use to label extreme excellence as cheating diminishes us all.  This “cure” is worse than the disease.

Not really...

The brilliant guy who plays like a machine and has consistently performances of 3149 and above, if he's really a genuine, should be able to talk about chess and teach us stuff, to show his prowess in chess problems, in blitz chess, over time, to be open and communicative, to know a lot about chess in general - explain endgames, ideas, strategy... he should have a huge and profound insight about the game, hugely eclipsing anybody who ever played chess before him - as he would obviously simply dwarf players such as Bobby Fischer, Jose Raul Capablanca, Emanuel Lasker, Gary Kasparov, Anatoly Karpov, Vladimir Kramnik, Mikhail Tal, Magnus Carlsen or Alexander Alekhine - to name just a few people, who would be immediately relegated to relative patzer status by reason of his very appearance...

Frankly, I can't wait to meet this guy. 

Solskytz, you won’t have to worry about meeting “this guy”.  He would be labeled a cheater and dismissed from chess.

<Hicetnunc> Right!

<SpinWatch> /sarcasm

1.  “If” there is a press-conference room.  Rising stars don’t always start in major events with full amenities.  

2.  “If” the judges like the new star enough to admit him or her to the pressroom.   They may prefer the “other winner”.

3.  “If” he or she wasn’t disqualified before ever getting to the “champion’s inquisition”. 

4.  If his or her moves were fed into a computer for timely analysis before the end of the tournament. 

5.  If the new champion isn’t too temperamental to stand cross examination.   For example, is anyone confident that all of our previous world champions could have or would have stood-up to a rigorous cross-examination?

Imagine somebody trying to register Rubinstein or Fischer.

SpinWatch, the IPR is not used to catch a cheater, anyway. It's the z-score that catches a cheater. The IPR is merely a fun metric that is not very reliable. So, you're correct about its reliability, but you're incorrect about its importance. I am going to point you to the comments at the USCF website for more information, because that is where the discussion has been the most technical and useful. I apologize if you finished the profile coming away with these misconceptions. This is my shortcoming as the writer of the piece.

OneArrow, I enjoyed your article about Dr. Regan’s work on IPR and z-score.  The IPR rating and z-score could be used by individuals to point out relative strengths and weaknesses in their game.  This information could assist players in determining an optimum path of future study. 

I am a retired civil engineer and currently tutor math at a local community college, so I understand “z” in statistics.  Being 5 standard deviations off normal is rare and the Wikipedia gives this event odds of 1 in 1,744,278.  While these odds are rare, they are hardly the magnitude of lottery winners.  In the article one computer assistance cheater had a z-score of 5.09 which translated to odds of one chance in five million.  Again, this is hardly lottery magnitude.  While such odds might seem high, their exact determination are likely redundant since beating higher level completion usually requires the lower ranked player to make far better moves than he or she normally does. 

Since many train with computers, the observation that a segment of their play closely matches computer’s choices isn’t surprising.

Long-odds and moves-matched are circumstantial evidence and do not constitute proof of cheating.  While long-odds and moves-matched may contribute to the preponderance of evidence, there needs to be other strong evidence to justify cheating accusations.

One of the charms of chess is that we may, just for a while, play at an exalted level.  We might just once in our chess lives – win the chess lottery and have our dream tournament.

Stewards of the game should avoid even the appearance of dashing these hopes and dreams. 

Spinwatch, there are people producing upsets, and winning chess tournaments, without their play being outrageously above what they usually do... or matching Houdini more often than not... Following your "logic", I guess you think Borislav Ivanov was a fair and honest player...

It looks like you have difficulties accepting statistical methods as a reliable way to detect cheating... It's interesting to note that your account here has been created on June 28th 2014 (3 days ago).

Would you mind telling us what your previous account on chess.com was and the reason for its closure by chess.com staff ?

Besides, according to chess.com rules, these discussions about cheating should be held in this dedicated group, rather than in the general forums :

http://www.chess.com/groups/view/cheating-forum

(you need to click the [join button])

you don't catch a chess cheater. its like catching a lepricon if that's how you spell it

OneArrow: you can´t get physical evidence since you can´t touch the players. 

But then if one player refuses to be registered, you would put that in the "behavioral signals" column?

Exactly. The Anti-cheating Committee's Rule proposal is available online (Google it, I don't know the link offhand), and the proposal is scheduled to be ratified at the Olympiad in Tromso. At the moment, there are no official FIDE rules or regulations about how to deal with alleged cheaters. Each TD or arbiter deals with each specific situation how he sees fit. Generally what happens is that suspicious behavior (physical or behavioral evidence) gets noticed, and then Regan gets a call to process, via his statistical methods, the alleged offender's games. More times than not the alleged cheater is acquited without incident. We hear only about the non-acquitals. Ivanov, for instance, had a z-score above 5. Note, too, that in the article Regan says he accepts multiple "high" z-scores within a short period of time as also suspicious. Nevertheless, the FIDE/ACP anti-cheating proposal up for ratification in Tromso is the "official" protocol at this point.

I started this article with the intention to write about artificial intelligence, how AI relates to human cognition, and to explore the mathematical complexity of chess. I wound up focusing the article on cheating, because this is Regan's expertise, but such a focus was not my original intention. At least I got to sneak in a bit of the AI/human cognition discussion.... But nobody seems to want to talk about that!

...

Hicetnunc, I don’t have an informed opinion about Borislav Ivanov being a cheater.  I leave that to the officials who have all the facts.

It looks like you have difficulties accepting statistical methods as a reliable way to detect cheating.

I see an unknown tennis player ranked 144^th in the world just defeated the number 1 rank.  Does that mean he cheated?  I don’t think that rare performance alone is enough to prove cheating.

It's interesting to note that your account here has been created on June 28th 2014 (3 days ago).

Would you mind telling us what your previous account on chess.com was and the reason for its closure by chess.com staff ?

I wasn’t aware I had a previous account here. 

Besides, according to chess.com rules, these discussions about cheating should be held in this dedicated group, rather than in the general forums :

The quality of ideas in a comment determines its worth, not how long the account has been in place.  Am I to be branded a “statistical deviant”, labeled a cheater, and be banished from this discussion?