There are claims Niemann’s online games show evidence of cheating. I haven’t read that much, and there was some confusion in the thread I read. Obviously important if true.

_________________
Occupy Babel!

There is usually a lot of really bad analysis floating around for cheating cases but there are exceptions. If you have links to the actual reports then I'd actually be interested, especially if it is not bad arguments and bad statistics.

In "top" go tournaments, there tends to be a duty to play all one's games, or all those for the time for which one has registered, especially because tournament result tiebreakers can depend on the opponents' tournament results or otherwise influence the fair competition among participants. E.g., this is so for EGF tournaments. Basically, a player has to play against anybody against whom he is paired. Cheating suspicion cannot make an exception because it is not the player who decides whether an opponent might have cheated but the arbitration bodies decide. Resigning prematurely for reasons unrelated to the board position and not having allowed justification (such as sudden illness) also tends to be prohibited as unsportsmanlike behaviour.

There might have been an exception in European go tournaments when quite a few players declared their intention to boycott playing against (Bela-)Russians or tournament participation not due to cheating suspicions but due to ethical considerations related to war; this exception seems to have been mostly dissolved by prohibiting participation of (Bela-)Russians in EGF tournaments. As to cheating suspicion, however, I am not aware of any such exception.

Apparently, some top chess tournaments do not have tournament rules preventing (all) unjustified dropping of rounds or premature resignation. Cheating allegations aside, maybe chess should learn from go and establish similar tournament rules?

Refusing to play particular opponents in tournaments is said to damage their reputation. However, this does not per se involve defamation. Statements besides tournaments of suspecting that certain players might cheat (again) are not exactly defamation, either. However, many perceive it as such nevertheless. Players making (too many) unproved allegations can damage their own reputation.

There is a theory that Magnus, as he has been becoming more involved in Chess.com, might recently have learned about more (real or alleged) cheating instances of Niemann, became more concerned about cheating possibility due to his hard loss against Niemann and therefore now refuses to play against him in tournaments. (Niemann's alleged or misinterpreted sandbagging or weak play in friendly beach fast games against Magnus might have added to the suspicisions.) However, this theory might be replaced by the simpler theory that Magnus simply does not want to play against Niemann in tournaments any more because of suspecting him to be a notorious cheater.

Improved anti-cheating measures can be considered but, eventually, what can one do against body implantations if a player has medically approved metal implantations or heart pacemakers? Passive modern body scan machines like those at airports? Tournament organisers digging rest rooms? Hopelessly ambiguous, over- or misinterpreted AI anti-cheating tools? Promotion of ethical behaviour among a game's world-wide players community?

On https://www.fide.com/news/1999 the FIDE President mentions "the anti-cheating protocols" of FIDE. What are those?

EDIT:

In https://www.youtube.com/watch?v=jfPzUgzrOcQ&t=702s Yosha Iglésias uses some tool to talk about accuracy / correlation between a game's moves and AI moves and the press mentions some 100% occasions for Niemann. What are those accuracy / correlation numbers? The percentage of identical moves themselves? Some derived alleged likelood of cheating?

I think what is meant by "the anti-cheating protocols" is the whole body of rules and regulations. This includes

• Anti-cheating regulations
• Anti-cheating protection measures
• Complaint forms

There are also other relevant documents:

• The ethics and disciplinary commission procedural rules
• Ethics and disciplinary code
• and an interpretation guide for the above two but that might be out of date.

Fide has and ethics and disciplinary commission (EDC) and a fair play commission (FPL). Cheating and improper accusations of cheating are the domain of the FPL, I am not sure about other similar transgressions but I think the EDC is the default and that the FPL is about anti-cheating in a restricted sense. For example hiding your opponents queen is, I assume, not the kind of cheating the FPL is concerned with.


I think the Sinquefield cup had a rule that it was not allowed to drop out after round 4. I heard that in the official live commentary but I have not idea if it is an actual rule or simply a convenient explanation. Magnus dropped out in round 4 and might not have strictly violate any rule. The argument that this violates the spirit of the competition is I think valid but players have a lot of leeway before commissions start sanctioning them. Possibly, Magnus is heading down a road where he will risk sanctions but it seems likely that this will blow over.

I'll have to watch that vide with the "most incriminating evidence" later, maybe I'll buy popcorn

There is some chessbase documentation online but it is rudimentary. As far as I can tell the "correlation" being talked about is a ratio of coincidence between the game moves and any move from a large committee of computer programs. That is, it is basically recall but measuring if the game moves match any of the computer moves.

For example, if someone plays a 20 move game every move could easily coincide with moves from any of 10 computer programs if we don't require that it is the same program every time or that the computers are not simply predicting 10 different moves in the more difficult positions.

Besides that the statistical argument doesn't seem to lead anywhere there is a problem with what is being argued from the beginning. Namely that Hans would be cheating in about one-in-ten games, having little regard to winning prize money or the fact that he would be a strong favorite in many of the games anyway. I say "from the beginning" because the argument is presented without looking at the games when he played terribly poorly in light of this statistic.

I didn't bother to read any of the material so apologies for that. However, did they do the logical next step and check the correlation for the rest of the top 20 players (including Magnus) to see how they would fare?

_________________
Dave Sigaty
"Short-lived are both the praiser and the praised, and rememberer and the remembered..."
- Marcus Aurelius; Meditations, VIII 21

Not "they" but some individuals have been digging in some games and found something like the following correlation numbers (which I try to recall from memory, I recall partly wrongly and show you only to give you a very rough idea of how these numbers are):

100% Hans Niemann in a few(?) of his best-correlated games. (I do not say that this would indicate cheating. I am just reporting numbers.)

Various (including high) numbers for Hans Niemann checked by a supposed expert statistician on chess cheating: no evidence for cheating found. (Magnus Carlsen has (subjectively?) interpreted for himself similar but unpublished (internal Chess.com?) numbers for Hans Niemann as clear cheating evidence. Maybe he thinks that each 100% value would be evidence beyond any doubt and regardless of statistical verification? Most of the press jumps on 100% and copies that without knowing what such numbers mean. Most Youtubers and most of their viewers throw around such numbers and have no idea whatsoever on statistics.)

100% in one game and 98% in one game of a proved cheater in these two games.

82% in one game a (Super-(?))GM checks for himself as his subjectively best game ever.

72% Bobby Fischer in a famous 20 games winning series.

70% Magnus Carlsen in some of his top games

69% Garry Kasparov in some of his top games

67~68%(?) Super-GM

65~66%(?) GM

Such numbers vary. At different times, different engines might be used. Chess.com and Chessbase numbers differ greatly.

My opinion on a player having relatively many 100% games: most likely, this means a) he has studied with AI a lot and applies his acquired knowledge in AI playing style or b) he cheats. Statistics cannot distinguish these two possibilities if correlations are derived from several engines.

Magnus Carlsen also studies a lot with AI and doesn't seem able to approach 100% if I believe the video. It doesn't formally prove anything but it's suspicious.

I would expect that picking up techniques by studying AI would not result in perfect emulation of AI for 30+ consecutive moves.

The ironic thing is if you only use AI to cheat a few times per game, detection can be avoided. This is going to be the real challenge going forward to building circumstantial cases - cheaters are going to learn to be better cheaters.

I thought this was a convincing criticism of Iglésias' analysis https://www.reddit.com/r/chess/comments ... video_with
Basically it seems that her methodology was deeply flawed

Playing strength and ability to learn in a particular manner need not be strongly correlated. A player can learn from AI but nevertheless decide to limit its impact on his playing style. A player might use different AIs or AI program versions than those used for correlation numbers. There are just too many parameters to draw definite conclusions.

It is also important to remember that Hans has been playing a lot of tournaments, he is said to be travelling to every tournament he can, this has two consequences: his opponents are not very high rated (compared to Hans) and there are a lot more games in total than is typical.

The statistic being used in the video is simply bonkers. Of course one can ask what other players score on this statistic but it is not that better players score higher. It depends too much on the opponent (i.e. making an early mistake, blundering into easier lines, do they play topical lines or avoid theory, ...) for it to be realistic to base suspicion on this statistic. Many of the replies in this thread are still treating it as a useful statistic when it is clearly not.

There must be lots of games where a grandmaster plays in a tournament (online or OTB) against someone rated 300 points lower. If that population of games also contains a fair number of near engine-perfect games by the grandmaster, it would support kvasir's theory.

People do appear to find these one-hundred games for lot of players when they look for them. Did you looked for them?

My comment was based on a Hacker News thread the day I posted—there were several criticisms of the one video, but also what looked like other stats.

It sounds like you’re better informed than I am. I don’t yet have an opinion, I was just sharing to note that there’s an attempt to get evidence beyond gut feelings.

_________________
Occupy Babel!

... what can one do against body implantations ...

Finally our cyborg overlords have arrived.

What can we do?

Stop playing for wins and start playing for joy.

In the best case, playing for the joy of our opponent.

Just here to report that non-AI cheating is alive and well.