Cheating in chess

[Matthew Turner]

If you click on the link it will take you to 1 hour 52 minutes in. If you watch for 30 minutes or so you will have a good understanding of the case against Maghsoodloo.

Is it possible to paraphrase it in a paragraph or two?


No, unfortunately, you have to look at evidence and think about things in order to understand how they work.

[Matthew Turner]

Quite so.

Nobody has responded to the evidence on the Maghsoodloo case, so I'll repost in case you missed it
https://www.twitch.tv/videos/868118939?t=01h52m00s

Thanks Matthew I didn't go through it before so it was interesting to see what led to the decision.

I've now had a quick look and I'm not going through all of it because I can see that its basically the same argument but I have to say the evidence is not convincing to me.

The evidence starts with a claim that smart cheaters are using Houdini now instead of Stockfish, though no evidence.

There are claims that Maghsoodloo played like a God and made no mistakes during this tournament. I guess that's based on Ken Regan's software which we don't understand and can't explain. We just have to take it on faith that the anti-cheating software is infallible.

Another claim is that Maghsoodloo was using the newest version of Houdini yet he makes a simple mistake in the featured game where he plays ...Rxb2+ missing a simple win of ...Nxe3 when White can't play Rxe3 because of ...Rd1 mate. This is explained by saying that Maghsoodloo was in time pressure when it was more important to play moves fast than accurately. We're told that at this point Maghsoodloo had 51 seconds plus increments and his opponent 42 seconds plus increments. I'm not really convinced by that as that's plenty of time to glance at Houdini if that's what he was doing. It of course may be true but all I see is further assumptions without evidence.

Further moves such as ...Kf8 and ...Ng5 were thought suspicious but the Russian translator actually said he could have played ...Kf8 himself ie getting the king off an open file. ...Ng5 was thought suspicious because it opened up the diagonal for the bishop at g7. Really? So what? Maybe he just attacked the h2 pawn quickly realised f7 was under threat and protected it with ...Ng5 even though White can now play h4. I'm not saying he wasn't cheating but its not evidence that he was.

Who is qualified to query Maghsoodloo's moves anyway? For that I'd want some hard evidence from a camera during play for example.

Chris,
I would stick at it and look at some more of the video. I think you are right to challenge what is being said, but I'll just correct a factual inaccuracy. None of this is based on the Regan Software, the analysis is carried out on the freely available PGNspy

[Kevin Thurlow]

"If you click on the link it will take you to 1 hour 52 minutes in. If you watch for 30 minutes or so you will have a good understanding of the case against Maghsoodloo."

I have now watched the first 20 minutes and it's certainly interesting (and Naroditsky is always a good presenter). I thought ....Kf8 was a normal move, but you do need to look at all the moves!

[Alex McFarlane]

Not sure where to start here.

Prof Regan never claimed that his software identified cheats. It just gave an indication of the players to be watched in otb.
Arguably, it is more reliable than the platforms own methods and is therefore now being used for a purpose that was not intended.

When I do arbiter courses I point out that Regan's software produces results in three bands: (a) no concerns; (b) some concern but probably nothing; (c) alarm bell time. In the course I usually say at this level I often joke that this is 'catch the cheat'. However, I then go on to say that there is a chance of a false positive.
With online chess the Regan software is being used in a way that was never intended. Having said that, it becomes a question of how much trust you give to the Regan test? What was designed as an aid to detection has become the jury.

As a result of online play my (c) above has been divided into two groups; c1) remains as get further proof but c2), in the absence of anything better, has become "on the balance of probabilities this player is guilty of cheating".

How tournament organisers use this information differs. Some will ban immediately others will wait on a cumulation of such results.
Both methods have their pluses and minuses which have been covered elsewhere in this thread.

There is no simple solution to this problem, not even cameras. One of the alleged signs on a camera is staring at something away from the board. Whilst this is a good indication there are also players who stare into space when concentrating!
Other than catching them there is no definite proof. A smart cheat will not leave definite proof. The arbiter is therefore reduced to considering probabilities.

As stated elsewhere, Regan's is not the only software that can be used but many consider it the best.

[Joseph Conlon]

Commenting on the interview with Andy Howie - I was also perturbed by the statement that you cannot tell cheating by one game. Maybe it was exaggeration for effect, but it’s clearly wrong from a statistical viewpoint (for pretty much any reasonable choice of ‘how unlikely is something before you call it cheating’). More games are just more moves, and a single long game with many - say 70 - moves provides 50 or so choices to be made. Even at a binary level, and splitting everything into right/wrong, exponentiating 2 to the power of 50 gives a gargantuan number.

(Intelligent cheating may be hard to detect, but if someone is following an engine first choice every move, a single game already has plenty of data)

There’s a tempting rebuttal which is that everyone can play really well one off, by chance. To me a useful illustration against this is the following:

1. Play against Stockfish on a really simple ending, K vs K +R or K vs K + B + N or K vs K + B + B (give Stockfish the material).
2. Even in this, just imagine the likelihood of a player reproducing the Stockfish moves when mating. There are loads of ways to mate - the combinatorial choice is enormous and the chance of picking the same ones as the engine is tiny.

[JustinHorton]

Although if you're using Stockfish how often is the game going to go on for seventy moves

[Joseph Conlon]

Although if you're using Stockfish how often is the game going to go on for seventy moves

When both sides are, quite often.....

[Matthew Turner]

Chris Rice wrote,
"Further moves such as ...Kf8 and ...Ng5 were thought suspicious but the Russian translator actually said he could have played ...Kf8 himself ie getting the king off an open file. ...Ng5 was thought suspicious because it opened up the diagonal for the bishop at g7. Really? So what? Maybe he just attacked the h2 pawn quickly realised f7 was under threat and protected it with ...Ng5 even though White can now play h4. I'm not saying he wasn't cheating but its not evidence that he was."

So, I think the point here is that even for a very strong player it is very very hard at this time control to come up with the idea of re-routing the Knight from the central square e5 to g5. I agree it is possible for even a relatively weak player to play Nf3 (to open up the Bishop) and then realise that f7 is too weak and retreat the Knight to g5 (either because you think it is the best move or because you don't want to admit your 'mistake' and return it to e5). So this sort of thing happens, but it cannot keep happening innocently.

[Kevin Thurlow]

"One of the alleged signs on a camera is staring at something away from the board."

Yes, it means nothing - e.g. Ivanchuk frequently stares at the ceiling and I don't suppose he has a chess set stuck there...

Even I sometimes stare into space then look at the board in the hope that I will suddenly spot something. A friend (who saw me do this) suggested that I am visualising the position on a giant imaginary demonstration board. Playing online, I'm always looking round, sometimes in boredom, sometimes wondering what the dog is doing.

[Alex McFarlane]

Commenting on the interview with Andy Howie - I was also perturbed by the statement that you cannot tell cheating by one game. Maybe it was exaggeration for effect, but it’s clearly wrong from a statistical viewpoint (for pretty much any reasonable choice of ‘how unlikely is something before you call it cheating’).

This statement shows that the procedure used is not fully understood. The early moves are not included in any analysis as they could be 'theory'. 'Forced' moves should not be considered either. This means that in a 40 move game there might be very few moves which can be used to judge if cheating took place. This is why 'smart cheating' is so difficult to spot where the computer is only used for advice once or twice in a game.

One game may alert you to a player who requires further watching but it is seldom enough to provide sufficient proof.

[John McKenna]

I take all the above criticism of Regan's method with a pinch of salt. What qualifies anyone here to do anything other than quibble and nibble away at the edges of the computer science and statistics that Regan is a practising expert in

That is essentially the argument put forward by Yuri Garrett, the Secretary of FIDE's Fair Play Commission, at their recent open Zoom meeting. I wish I could direct you to the anticipated Youtube recording of this meeting, but there isn't one.

My experience of CAS judges leads me to suspect that they will not react well to being told that they must just believe the methods because they are obviously not clever enough to understand them.

Thanks, Nick, those kinds of teething troubles to do with the introduction of statistical (and also genetic) expert evidence into legal systems has preceded the ones under discussion here that are to do with chess.

And the "show me the science" argument has been advanced before, too. In the field of computing systems - such as neural nets, etc. - there are black boxes, which I know you are naturally suspicious of. However, I don't think you can just reject them out of hand. Otherwise you risk becoming a modern luddite.

By the way, I thought the Regan article - Predicting Chess and Horses - that Chris R linked to some pages ago would have piqued your interest - at least as far as how it applied to predicting the winners at the gee-gees. Guess not, but if it had surely you would've judged it by results and not how it produced them.

I've never tried to say that Regan's s/w system was perfectly fit as an all-purpose cheat-catcher, jury and judge. And I tend to agree with Joe Conlon's and Alex McFarlane's, above, characterisations of the state of play.

Particularly Alex M's -

As stated elsewhere, Regan's is not the only software that can be used but many consider it the best.

[Nick Grey]

John - accusations and false are disruptive and memory lasts for decades. I was called a cheat at your club for not ticking my score sheet . 2 of your players lost because of the noise and all after losing a rook in a game of chess.

Great points and reasons why I and others refuse to play graded chess on Lichess.

[MJMcCready]

John - accusations and false are disruptive and memory lasts for decades. I was called a cheat at your club for not ticking my score sheet . 2 of your players lost because of the noise and all after losing a rook in a game of chess.

Great points and reasons why I and others refuse to play graded chess on Lichess.

Sometimes accusations of cheating is a form of cheating in itself although some might try and get themselves off the hook and call it gamesmanship.

[Nick Grey]

"One of the alleged signs on a camera is staring at something away from the board."

Happened to a friend in an exam. Something was a 3yo sister. He re sat.

[Nick Grey]

Sometimes accusations of cheating is a form of cheating in itself. HE thought it was a rule.