Niemann - Carlsen

[Pargat Perrer]

Kenneth and others argued against centipawns normally being a good metric in this way, once you're up, say a Queen in a position, the centipawn score becomes irrelevant as all kinds of moves win other than the machine-generated move. Keep in mind that no one is arguing that Hans is not a strong player in his own right. So even 2500-level move choices are probably good enough to survive, and if he finds an advantage in a critical position, then he can cruise along to a win with 2500-level moves without caring too much about the centipawn score.

Sid, maybe you can mention to Ken Regan that if this is indeed what is happening with Niemann, it should be easy to spot by the following method (quoting from my last post): do a move-by-move score assessment of each of his games and graph it .... you should see that he falls behind in every game, and then "miraculously" makes a single move to either get back to equality or even get ahead. This should be EASY to detect, and it should be happening like a repeating and regular cycle in EVERY ONE of his games.

Even if he is playing 2500 level moves against a 2700+ super-GM, he should keep falling behind, centipawn by centipawn so to speak, then "miraculously" make it all up with a critical juncture move where he utilized his Anal Beads Code Vibrations (LOL) cheating device.

Perhaps no one has thought to look for this signal? If there really is an undetectable cheating method, and a player is only using it at critical junctures, then every game he plays against a superior rated player should show this repeating pattern of falling behind ... them suddenly with one move getting back to equality or even moving ahead.

This is a pattern that can be found! I wonder if Ken Regan has thought of it....

If this repeating signal were found in all of Neimann's games of the recent past, for example those 6 tournaments you mentioned Sid, then I would think that is strong evidence of using this cheating strategy (even if the actual device is never found).

And if the pattern is NOT found, then I think it clears Neiman for good.

[Sid Belzberg]

TBH similar to what Pargat indicated, this seems unlikely to be a successful strategy to reach 2700 to me. If you're making generally inferior moves at that level, I would expect it to be very difficult to reach a "critical juncture" because you're just going to have the worse position. If he's not showing an unprecedently exceptional level of play, and just having results that beat expectation based on the quality of his play, to me, that's more likely an indication of not cheating than cheating.

It is possible that there's something about him (maybe his play style, maybe they just dislike him personally, maybe it's the cheating allegations that have other players seeing monsters under the bed) that cause people to play worse than normal against him; I'm sure we all remember "Fischer Fever" where normally elite GMs including unbeatable former World Champions like Petrosian suddenly and inexplicably started playing worse against him.

Kenneth and others argued against centipawns normally being a good metric in this way, once you're up, say a Queen in a position, the centipawn score becomes irrelevant as all kinds of moves win other than the machine-generated move. Keep in mind that no one is arguing that Hans is not a strong player in his own right. So even 2500-level move choices are probably good enough to survive, and if he finds an advantage in a critical position, then he can cruise along to a win with 2500-level moves without caring too much about the centipawn score.

[David Ottosen]

No, the interpretation of the result shows that Nieman's average centipawn loss and standard deviation during his ascent from 2500-2700 is entirely different from other GMs with similar rating increases during the 2500-2700 period. Because the general quality of his moves was lower than other GMs during their ascension, it could be interpreted that he received assistance at critical junctures hence the general quality of his moves was comparatively poor compared to his peers, but the odd critically important move pushed him over the top.

TBH similar to what Pargat indicated, this seems unlikely to be a successful strategy to reach 2700 to me. If you're making generally inferior moves at that level, I would expect it to be very difficult to reach a "critical juncture" because you're just going to have the worse position. If he's not showing an unprecedently exceptional level of play, and just having results that beat expectation based on the quality of his play, to me, that's more likely an indication of not cheating than cheating.

It is possible that there's something about him (maybe his play style, maybe they just dislike him personally, maybe it's the cheating allegations that have other players seeing monsters under the bed) that cause people to play worse than normal against him; I'm sure we all remember "Fischer Fever" where normally elite GMs including unbeatable former World Champions like Petrosian suddenly and inexplicably started playing worse against him.

[Pargat Perrer]

After watching the video on the computer analysis I'm starting to think he was cheating. The evidence seems convincing but the big question for me anyways is how did he cheat? I'm only asking in regards to the live games. That's what I can't wrap my head around.

in post #104 of this thread, Sid provided this link:

https://medium.com/@rafaelvleite82/h...l-bea9485055de

Here's what caught my eye in that article:

"Many people are trying to figure out how a player could be cheating in an Over The Board Tournament. Even Elon Musk tweeted about this, saying he could be using anal beads to communicate with someone and get the best moves by some kind of code vibrations."

Of course! Anal Beads Code Vibrations! ABCVs!!!!

Homer Simpson: "Mmmmmmm .... anal beads code vibrations .... "

[Pargat Perrer]

.... hence the general quality of his moves was comparatively poor compared to his peers, but the odd critically important move pushed him over the top.

Maybe the top rated players on this board can answer this question: Is what Sid just described a viable way to win consistently and get to 2700 (assuming your cheating method goes undiscovered and uninvestigated)?

I have to think it would NOT be viable ... because for example you are playing let's say 4 consecutive moves that are slightly inferior to those of your super-GM opponent's 4 moves and then on the 5th move you use your cheating device and find the absolutely best move at the critical juncture .... but you are playing this move from an INFERIOR POSITION, are you not? That means you are constantly having to come back from inferior positions, game after game, time after time.

Can that work to get one to a 2700 rating? Again, assuming the cheating method works and you correctly assess when you should use it, i.e. you correctly determine the "critical junctures".

EDIT: if what Sid proposes is indeed what has been going on with Neimann, it should show up .... do a move-by-move score assessment of each of his games and graph it .... you should see that he falls behind in every game, and then "miraculously" makes a single move to either get back to equality or even get ahead. This should be EASY to detect, and it should be happening like a repeating and regular cycle in EVERY ONE of his games.

[Sid Belzberg]

I guess what I mean is this; say a 2300 has an average centipawn loss of 40, while a 2700 has an average centipawn loss of 10. In the normal progression you describe, you see something like a player go from 40 to 35 to 30 to 20 to 15 to 10 (or whatever) as they progress from 2300 to 2700. My question is whether Niemann was a normal 2300, happily losing 40 centipawn per game, and then overnight switched to losing 10 (bypassing the normal progression) or whether he suddenly improved to losing 25 per game, but saw an improvement in his results that seems unusual for this level of play; ie, does his centipawn loss rate indicate that he's a real 2700, or does his centipawn loss rate indicate he's a 2500 who through a statistically improbable run of bad play by his opponents has reached 2700?

No, the interpretation of the result shows that Nieman's average centipawn loss and standard deviation during his ascent from 2500-2700 is entirely different from other GMs with similar rating increases during the 2500-2700 period. Because the general quality of his moves was lower than other GMs during their ascension, it could be interpreted that he received assistance at critical junctures hence the general quality of his moves was comparatively poor compared to his peers, but the odd critically important move pushed him over the top.

[Sheldon Ab]

After watching the video on the computer analysis I'm starting to think he was cheating. The evidence seems convincing but the big question for me anyways is how did he cheat? I'm only asking in regards to the live games. That's what I can't wrap my head around.

[Peter McKillop]

.....
I have to say I don't buy the "we need evidence beyond a reasonable doubt" standard. ......

Here's another log for the fire:

https://en.chessbase.com/post/wall-s...likely-cheated

[David Ottosen]

Every player shows a linear correlation between their rating and their centipawn loss score as well as the overall consistency of each move, the higher the rating the lower the centipawn score and the lower the standard deviation. So far this analysis has held true for every player that has ascended from 2500-2700. Hans Neiman is the only player that shows no improvement in the overall consistency of his moves (ie tends to choose the best move the machine offers or very close to what the machine offers as second best) , this is the centipawn loss that for every other player during their 2500-2700 ascent gets lower and lower but not with Hans Neiman,
This would indicate perhaps that in critical positions he chooses the best move but most of the time has lower quality moves using centipawn loss as a barometer. Given that this pattern is unique does not look good. Apparently, chess.com will have a big announcement this week so we will see what happens.

I guess what I mean is this; say a 2300 has an average centipawn loss of 40, while a 2700 has an average centipawn loss of 10. In the normal progression you describe, you see something like a player go from 40 to 35 to 30 to 20 to 15 to 10 (or whatever) as they progress from 2300 to 2700. My question is whether Niemann was a normal 2300, happily losing 40 centipawn per game, and then overnight switched to losing 10 (bypassing the normal progression) or whether he suddenly improved to losing 25 per game, but saw an improvement in his results that seems unusual for this level of play; ie, does his centipawn loss rate indicate that he's a real 2700, or does his centipawn loss rate indicate he's a 2500 who through a statistically improbable run of bad play by his opponents has reached 2700?

[Tom O'Donnell]

Suppose you are playing a game against a known cheater. Your opponent makes what looks like a double-edged sacrifice. Would the thought not cross your mind "I wonder if this guy is using an engine right now?". I think it would for many people even if there is no evidence their opponent is using an engine.

One possibility is that Niemann is using this uncertainty to his benefit and is strategically gambling more over the board.

It could also be that his opponents are genuinely unsure about what is going on and are simply playing worse as a result. Their attention is divided between playing good chess and wondering if they are being cheated at that moment. Even if there is zero evidence they are being cheated.

I have to say I don't buy the "we need evidence beyond a reasonable doubt" standard. Civil cases for millions of dollars are determined by preponderance of evidence. Maybe split the difference and say if it's 75+% someone cheated, they get the hook. I guess the big question is what exactly constitutes evidence, vs conjecture/confirmation bias.

[Sid Belzberg]

Sid, those articles I cited---my own---are absolutely relevant as put in my replies to Nate Solon. I categorically disavow any support that might be implied for the sources you queried me about on FB Messenger.

I apologize, Ken, I only read the article and not the comments you made that clearly referred to centipawn loss analysis. As mentioned originally when you showed me this article, Hans Neiman's Graph of his rating versus centipawn loss during his ascent (in particular when he ascended from 2500-2700) is entirely different than all other grandmasters that achieved these lofty heights..
Your thoughts on this are what I was seeking, as I mentioned in this forum and on your messenger, and still am. As I mentioned, the code and methodologies, and data sources are entirely disclosed and reproducible.

I can assure you that I am not alone in the chess world or the data science world in this request I posed.

https://en.chessbase.com/post/statis...f-hans-niemann

[Kenneth Regan]

Sid, those articles I cited---my own---are absolutely relevant as put in my replies to Nate Solon. I categorically disavow any support that might be implied for the sources you queried me about on FB Messenger.

[Patrick Kirby]

Every player shows a linear correlation between their rating and their centipawn loss score as well as the overall consistency of each move, the higher the rating the lower the centipawn score and the lower the standard deviation. So far this analysis has held true for every player that has ascended from 2500-2700. Hans Neiman is the only player that shows no improvement in the overall consistency of his moves (ie tends to choose the best move the machine offers or very close to what the machine offers as second best) , this is the centipawn loss that for every other player during their 2500-2700 ascent gets lower and lower but not with Hans Neiman,
This would indicate perhaps that in critical positions he chooses the best move but most of the time has lower quality moves using centipawn loss as a barometer. Given that this pattern is unique does not look good. Apparently, chess.com will have a big announcement this week so we will see what happens.

But as I mentioned in my last post, wouldn't systematically cheating in critical positions still create a signal in the data?

Assuming that "critical" positions are ones:
1. Where human players are more likely to make mistakes; and
2. Where the difference between the best move and the second best move is significant;

It seems like cheating in these positions would still significantly increase average move quality compared to non-cheating players, even if it was only a few moves a game.

[Sid Belzberg]

I did not realize Sid Belzberg would echo my comments on Facebook Messager---a private medium---on this forum. I did say I would check into various Internet stuff today, and in fact Nate Solon today has presented most of what I was going to say: https://zwischenzug.substack.com/p/d...-niemann-cheat See also my comment just placed there.

We discussed centipawn loss analysis that, understandably, you implied you were too busy to look at at the moment but would take a closer look at it when you had time. The articles you mention are not related to what we discussed.

[Sid Belzberg]

Just want to be clear if I understand; basically he goes from having a normal 2300 level centipawn average score to suddenly having a 2700 level centipawn score basically overnight, and maintains it for the years since it started? As opposed to the more normal 2300 level score, then slowly getting better?

I ask because my understanding of the first game with Carlsen was that his game score wasn't actually that spectacular and his win was more Carlsen playing well below his usual level. So the question is more "did he jump from 2300 level scores to 2700 level scores overnight and saw the associated rise" or "did he jump from 2300 to 2500/2600 and for some reason, his style of play induces more bad play by his opponents".

Every player shows a linear correlation between their rating and their centipawn loss score as well as the overall consistency of each move, the higher the rating the lower the centipawn score and the lower the standard deviation. So far this analysis has held true for every player that has ascended from 2500-2700. Hans Neiman is the only player that shows no improvement in the overall consistency of his moves (ie tends to choose the best move the machine offers or very close to what the machine offers as second best) , this is the centipawn loss that for every other player during their 2500-2700 ascent gets lower and lower but not with Hans Neiman,
This would indicate perhaps that in critical positions he chooses the best move but most of the time has lower quality moves using centipawn loss as a barometer. Given that this pattern is unique does not look good. Apparently, chess.com will have a big announcement this week so we will see what happens.

As one poster put it,

"These correlations for all the players other than Hans is insanely high, so he sticks out like a sore thumb and the likelihood of randomly playing one good move in a game and it not be cheating seems... Small."