Carlsen withdrawal after loss to Niemann

[Uri Blass]

22.Nd6+ seems to be another not best move in this game when only Stockfish7 choose based on their analysis and same for 25.Rc1.

I do not understand how they choose the engines and if it is only at least one engine out of many engines suggest the move of the players then I expect clearly bigger accuracy than 70% by Magnus Carlsen.

[AndrewGrant]

Why is it that not one of these titled players that makes a video of them looking at Hans game makes the effort to compare this to other players? I don't think this person cares at all, they are just farming the click-bait while the click-bait is good. And the 100% metric to me is pretty flimsy, if the claim is "If you hit 100% then you must be cheating". Because 1. Many games, 2. the metric of 100% is dictated by the engine, the depth, the hash, the other config, the version, was it cleared per position, ...; there are a billion factors.

Find a near perfect game from a player, and then try every Stockfish, Komodo, Ethereal, and Leela commit in the last year, see if one matches. If not, tinker with the depth. Then do the other settings. You will find a solution that suggests cheating. All it takes is some hacking of the sample size and a good selection of games with fairly forced positions.

Yes, what I think is needed now is for someone (whether the same person or another who knows the details) to run all the classical games of Magnus Carlsen for year 2019 (his best year probably) thru the same metrics. If none of them show more than the 98% that was supposedly the best (achieved by a cheater) prior to Niemann, while Niemann showed several at 100%, that would be pretty convincing. It would not only indicate cheating, it would indicate "stupid cheating"; a clever cheater would at least either make occasional inferior moves or use an engine like Lc0 that is sufficiently unlike Stockfish to avoid a high match percentage; the assumption would be that cheat-detectors would be unlikely to check with Lc0 for several reasons. It should be relatively easy for a neutral party to confirm or refute these charges by comparing Carlsen 2019 with Niemann during the period in which cheating is alleged.

I wonder how did they get the 100%

I checked the first game when 100% is written and I do not see 100% same moves

...

Your observation is exactly what I am talking about.

[dkappe]

22.Nd6+ seems to be another not best move in this game when only Stockfish7 choose based on their analysis and same for 25.Rc1.

I do not understand how they choose the engines and if it is only at least one engine out of many engines suggest the move of the players then I expect clearly bigger accuracy than 70% by Magnus Carlsen.

My hypothesis is that if you take every engine on the CCRL and check them against a GM game, your chances of matching at least one engine at least once every move is unity.

[MonteCarlo]

I've never used Chessbase, and was almost considering buying it just to do the comparison legwork (absurd as such a purchase would be); so many people in the comments to that video were making claims about how often that 100% metric would come up in other players' games, but no one had actually done any digging.

I'm glad I decided to wait, as things seem to be trickling in: https://imgur.com/a/KOesEyY was posted in a reddit thread, for example.

I suspect there will be a lot more of these found, but I suppose we'll see.

At the end of the day this seems like it'll be just like the broadcast/non-broadcast stuff: raw data that's wrong and/or misunderstood combined with a complete lack of rigor in analysis.

Humans being humans, I suppose

[supersharp77]

Body Language Expert On Hans Niemann.......

https://www.youtube.com/watch?v=OK9ZkoSQNFs

[Chessqueen]

Body Language Expert On Hans Niemann.......

https://www.youtube.com/watch?v=OK9ZkoSQNFs

It will be Christmas, and you guys still talking about the same subject, give it up guys! None of you will get anything solve about this accusation of cheating, so why waste your time

[lkaufman]

22.Nd6+ seems to be another not best move in this game when only Stockfish7 choose based on their analysis and same for 25.Rc1.

I do not understand how they choose the engines and if it is only at least one engine out of many engines suggest the move of the players then I expect clearly bigger accuracy than 70% by Magnus Carlsen.

My hypothesis is that if you take every engine on the CCRL and check them against a GM game, your chances of matching at least one engine at least once every move is unity.

Yes, the solution is simple enough. Take the best version of Stockfish that was available at the time of the alleged cheating. If he was cheating so grossly as is suggested, he would probably have used this version (or one very close to it). Specify some precise rules (hardware, time per move, number of threads, mpv 2 (for example). Measure the percentage of matches in the suspect events. Then do exactly the same for Magnus Carlsen for the year 2019. Compare the results (not just mean, but the top 5%, 10%, etc., since cheating, if any, might not happen in every game). If Hans matches Stockfish significantly better than Carlsen for the better games, compared by identical criteria, then we can be almost certain of cheating. Unfortunately, this can't totally clear Hans if he matches significantly less, since he could have been using Lc0 (or even Dragon), but it would put the ball back into the court of the accusers to provide evidence. I could do it myself, but I might be considered to be biased due to the connection between Komodo and chess.com, so I'd like to see it done by someone with no particular reason to care about the result.

[CornfedForever]

I could do it myself, but I might be considered to be biased due to the connection between Komodo and chess.com, so I'd like to see it done by someone with no particular reason to care about the result.

Well, you could still do it yourself (even the OTB games) if only to 'confirm'/'deny' claims or just to feel comfortable about the analysis of others.

In any case, I think the whole idea of matching 'best moves' is misguided. I have to side with Ken Regan's approach over these off the cuff ideas we keep seeing. We should never forget that chess games are 'lost' more often than they are 'won' by good play. I think that's really what happened between Hans and Magnus at the Sinquefield Cup.

I know I have won many a tournament where engines will say I did not play all that well. But, in the (lets say, 5) games I played my opponents (higher rated than me or not) made enough mistakes (large or small) during our specific games for which I found enough to capitalize on. That's the way cookies crumble sometimes.
'
Also, we should be careful with our assumptions. Why on Earth would he need to pick the 'best Stockfish' (or any engine) version at the time? You can go back 10 -20 yrs or more and find versions of all sorts of engines which would be just as good for the desired purpose of cheating at chess.

Just remember, I say that firmly believing in the 2nd sentence above.

Thinking of that and as an aside, my last OTB tourney was miserable. 1.5/4 !! I haven't scored less than 50% in...30 yrs? Ironically the game I won was judged by a 2400+ player to be the 'brilliancy prize' of the tournament's Open section. Was it truly 'brilliant'. I actually do not think so. It was a nice game where I let my young opponent weaken himself and, in time pressure, I (mostly) intuitively sacrificed an exchange which led to a win. It was all because of (mostly little - no 'dropped material') mistakes by my opponent. See sentence 2 above.

[M ANSARI]

I think that this is just another metric to look at. I always believed that there must be data or information in the games played if analyzed statistically en masse. Taking one game at a time is probably not useful unless the cheating is very primitive. Of course you have to weed out quick draws that are based on preparation and look at games that were actually played after book lines ... but maybe not! Maybe memorization of book lines in an unusualIy high fashion might also be unusual. I would not use that as the only metric but it would be something if there is one person that has a score that sticks out when compared to other players in the same category.

I am now pretty sure that it was Chess.com guys, after they bought out MC companies ... gave MC incriminating evidence of cheating by Hans and Dlugy. A few weeks before that Hans and MC seemed to be on amicable terms playing chess on the beach, but something changed after Chess.com bought out the MC group of companies.



On another note ... interesting that Bobby Fischer had the highest rating at 72% ... higher than MC and Kasparov! All way before engines were even a factor in chess preparation. That just shows what a genius he was!!!

[MonteCarlo]

I'm not at all sure that Fischer had the highest anything in this metric (if it even matters).

His % in that 20 game streak was listed in that slide, but given that there's at least one old MC game at 100% (listed in my previous post), whatever that number means, and that same slide claims 98% by Feller is the highest on record, I'd take that with a shaker of salt.

They list that number for that Fischer win streak, but nothing in that presentation gives me good reason to trust that his is the highest in a 20 game sequence.

It's tempting to draw all sorts of interesting conclusions when people present like this, but I'll retain a healthy skepticism until there's good data presented and not just random claims from someone who's clearly done insufficient fact-checking.

Cheers!