Cheating in chess

[Roger Lancaster]

FIDE Fair Play Commission upon reviewing all the data submitted by FPP will decide, whether the case shall be referred to the Ethics and Disciplinary Commission for further sanctions.

I'm suggested that whether or not FIDE refers it, the player should do so. The damage is done, so she should act immediately and not wait for FPL.

Nick's absolutely right yet there's a strong tendency among chess players to whinge, on forums such as this, about how unfairly they've been treated by Lichess [and, in my experience, very often with prima facie justification] or Chess.com [where I've less experience] yet - when it comes to making a formal complaint to the chess world's governing body - they expect someone else to do it.

My endorsement of Nick comes from personal experience. In 2015, one of our club juniors was effectively penalised by the ECF and we took the matter to FIDE. After due consideration, the FIDE said - and I paraphrase - "The British junior championship in question wasn't a FIDE-recognised event so the tournament decisions aren't strictly a matter for us but, if it had been a FIDE-recognised event, there's no way we would have permitted this to happen". At that point, and only at that point, the ECF - whose then chief executive, Phil Ehr, had been adamant that its decision had been correct - conceded that it had been wrong all along.

More recently, the parent of another junior - accused of cheating by Lichess - contacted FIDE to point out that this was probably defamatory and particularly to be discouraged where young children were involved. He also pointed out that, since FIDE Laws were particularly insistent that players should not refer to others as cheats, it was inconsistent that platforms were permitted to do what players were not. I don't know who
spoke to whom but, within three weeks, Lichess had changed its wording to a "violation" of its terms.

However when, still more recently, I attempted to get together a petition to propose to FIDE that algorithms such as those used by Lichess and Chess.com [and, indeed, Ken Regan's methods] be subjected to rigorous independent audit, almost all those contacted elected to sit on their hands. At that point I'm afraid that I decided it wasn't my duty to assist those who weren't prepared to lift a finger to help themselves.

[NickFaulks]

Evidently Ponomariov has come out in her support.

https://chess24.com/en/read/news/cheati ... mpionships

My own attempts to post on chess24 have failed.

edit : now fixed, because I updated my dob?

[Roger de Coverly]

Naturally this leads to the conclusion that the processes used are therefore not free of flaws and innocent players may well be getting chucked out of tournaments and having their reputations trashed as a result of this.

I thought it was known that the statistical methods used can throw up false positives. There's also the point that inputs to the processes are the presumed unassisted strength of the player and the number of moves supposedly "book". Neither of these are error free.

[Matthew Turner]

Perhaps it is worth someone putting the games through PgnSpy and seeing what the statistical analysis actually says.

[NickFaulks]

However when, still more recently, I attempted to get together a petition to propose to FIDE that algorithms such as those used by Lichess and Chess.com [and, indeed, Ken Regan's methods] be subjected to rigorous independent audit, almost all those contacted elected to sit on their hands.

The argument put forward by the FPL Secretary at last year's FIDE Online Congress ( open to all but sadly not on record ) was that Ken Regan's statistical work is so far ahead of its time - he did actually namecheck Einstein by way of comparison - that nobody else would be capable of understanding it.

[Paolo Casaschi]

The argument put forward by the FPL Secretary at last year's FIDE Online Congress ( open to all but sadly not on record ) was that Ken Regan's statistical work is so far ahead of its time - he did actually namecheck Einstein by way of comparison - that nobody else would be capable of understanding it.

If you can't explain it simply, you don't understand it well enough.

[MartinCarpenter]

Naturally this leads to the conclusion that the processes used are therefore not free of flaws and innocent players may well be getting chucked out of tournaments and having their reputations trashed as a result of this.

I thought it was known that the statistical methods used can throw up false positives. There's also the point that inputs to the processes are the presumed unassisted strength of the player and the number of moves supposedly "book". Neither of these are error free.

Pure and simple a matter of how you apply them. Use them over enough games, and be prepared to let some marginal 'maybe cheaters' go - even perhaps 'probably cheaters but not sure' and you can more or less rule out false positives.

Used over a small number of games, and set to try and make sure you catch everyone who might be cheating and you'll get plenty of false positives.

[Wadih Khoury]

However when, still more recently, I attempted to get together a petition to propose to FIDE that algorithms such as those used by Lichess and Chess.com [and, indeed, Ken Regan's methods] be subjected to rigorous independent audit, almost all those contacted elected to sit on their hands.

An audit would be nice, but then I equally think that the argument "then cheaters will know how to fool it" is valid.

More and more cheats now understand that they need to vary their move times. Previously, many cheats would take a somehow fixed amount of time, irrespective of the difficulty of the move (they were just waiting for the stockfish analysis to be stable enough in a few seconds). That is a big contributor to Lichess's cheat detection. But since Lichess talked about this point, determined cheats know how to avoid this.
Similarly, good cheats now avoid removing focus of their window, using other devices to cheat, alternating best moves with somehow inaccurate moves.

The more information cheaters have, the better they cheat.

It is a horrible situation I must say, where false positives are devastating and deeply unfair, while false negatives ruin all forms chess (which is a competitive sport at heart) and reward corrupt behaviours. I am not sure what the solution should look like.

[NickFaulks]

An audit would be nice, but then I equally think that the argument "then cheaters will know how to fool it" is valid.

It was only proposed that a statistician should be permitted to look at the statistics.

[Kevin Thurlow]

"It is a horrible situation I must say, where false positives are devastating and deeply unfair, while false negatives ruin all forms chess (which is a competitive sport at heart) and reward corrupt behaviours. I am not sure what the solution should look like."

Well, yes. It doesn't help either that organizers frequently refuse to discuss the issue, saying they cannot reveal if anyone has been accused of cheating, or what the organization has done about accusations, as this might get them sued. So I'm left thinking, "Oh - they've got someone cheating but they've swept it under the carpet, either because it's one of their mates, or because they think (wrongly) that admitting there are cheats reflects badly on the organization." I assume they're allowing the cheats to carry on, so I rather lose interest.

I've played the odd cheat at online blitz, but don't care about that, but playing a normal length game does seem a waste of time (although I have played one in the last year).

[Matthew Turner]

Ponomariov has linked to this
https://www.chess.com/analysis/game/pgn/5qqxsjiH7t

As a starter the accuracy is given as 95 for Osmak.
To give some reference here are my scores from the British Rapid (which had the same 10/5 time control) with the result

98.6 (draw)
98 (win)
96.3 (draw)
98.4 (win)
98.4 (win)
80 (draw)
84.8 (win)

[Thomas Rendle]

I've only taken a quick look, but does seem surprising on the surface. If this does end up coming to court (doubtful?) you'd hope they'd have more evidence than just the moves. It's very possible that they do and are quite correctly choosing not to share at this stage.

[Matt Bridgeman]

However when, still more recently, I attempted to get together a petition to propose to FIDE that algorithms such as those used by Lichess and Chess.com [and, indeed, Ken Regan's methods] be subjected to rigorous independent audit, almost all those contacted elected to sit on their hands.

An audit would be nice, but then I equally think that the argument "then cheaters will know how to fool it" is valid.

More and more cheats now understand that they need to vary their move times. Previously, many cheats would take a somehow fixed amount of time, irrespective of the difficulty of the move (they were just waiting for the stockfish analysis to be stable enough in a few seconds). That is a big contributor to Lichess's cheat detection. But since Lichess talked about this point, determined cheats know how to avoid this.
Similarly, good cheats now avoid removing focus of their window, using other devices to cheat, alternating best moves with somehow inaccurate moves.

The more information cheaters have, the better they cheat.

It is a horrible situation I must say, where false positives are devastating and deeply unfair, while false negatives ruin all forms chess (which is a competitive sport at heart) and reward corrupt behaviours. I am not sure what the solution should look like.

The pain free way casinos generally deal with card counters these days is just to say in a friendly manner, ‘sorry you’re too good for this game. We invite you to go and play a different one in the casino.’

[NickFaulks]

The pain free way casinos generally deal with card counters these days is just to say in a friendly manner, ‘sorry you’re too good for this game. We invite you to go and play a different one in the casino.’

I've always felt that bridge clubs should do the same. The ability to remember what has gone gives my opponents an unfair advantage.

[Roger de Coverly]

Ponomariov has linked to this

The linked game starts 1. d4 e6 2. c4 Nf6 3. Nf3 Nf6 4. Nc3 a6. The first seven half moves have a "book" flag, but not the eighth. If they are starting the engine detection at 4. .. a6 they will get misleading results.