Carlsen withdrawal after loss to Niemann

[AndrewGrant]

https://twitter.com/atl_kings/status/1568656197812891653

Would like to see the information concerning...
1. The ratings of the players he played against, ( at the time and currently )
2. Whether he knew the pairings ahead of time to prep,

You can make data say whatever you want if you try hard enough.

I would not be surprised to learn that you could also conclude that Hans is a significantly better player only when given the opportunity to prepare deeply against his opponent, and that he is unable to beat similar-aged players (who have very un-established ratings) decisively in Swiss events for example.

[M ANSARI]

Wow that is a pretty exhaustive and comprehensive effort! I have to admit it is very damning and most likely some more details will come out from Chess.com. Usually if you see one cockaroach in your kitchet, it means there is a 100 you haven't seen. For me by far the most damning evidence is the total inability to know what was going on in his game against Firouzja. I mean I am no 2700 GM and I could see the resource of queen check gives a tempo for the bishop to cover that attacking bishop on the diagonal. For a person who has immeresed for hours in the position and made some incredibly powerful tactical moves ... without having to check full analysis to its finality ... very very suspicious. Remember that this is someone who has claimed that these super GM's play terrible chess ... so obviously he should be better at analysis than all of them ... yet he analyzes at maybe a 2400 elo level. After that interview I got suspsiciou and decided to look at his Mamydyarov win ... and boy that was a suspicious game. There was a string of perfect moves while the engine was in analysis and he says "oh so I played this really well" like they were all based on intuition and no real deep calculation behind it! Yeah right. If anything his analysis shows that he has absolutely crappy intuition!

I a now sure that something is up and it is just a matter of finding out how or what methods he used. Maybe there needs to be a Hackathron to try and find the best way to circumvent the anti cheating methods used OTB at the moment. Remember a good cheater is not stupid enough to use all chess moves or push his rating to SF 15's 3800+ level. He is fine going to 2700 or 2750 level to get invited to lucrative competitions and can have an occasional win against a Super GM to grab headlines and get adulation. I think this is going to get much more interesting.

[M ANSARI]

I remember that Tour De France made a similar statement when the Armstrong controversy was happening. Armstrong was also tested and went through the anti-cheating system with no problems. At one point he was claiming that he was the most tested athlelete in the world and hadn't failed a test ever!

So for now legally Hans should be given the benefit of the doubt, just like with Armstrong, but in no way shape or form should the digging stop!

[AndrewGrant]

Both sides can make the data say what they want, as shown here.

TLDR:

[DrCliche]

Would like to see the information concerning...
1. The ratings of the players he played against, ( at the time and currently )

That was a common response people gave to that person's tweet, which is why I was intrigued and looked into it more deeply, with more detailed data and a more rigorous statistical analysis, which I already included in the very post you quoted.

But, to recap: the average strength of Niemann's opponents was a statistically insignificant predictor of his performance in USCF-rated classical tournaments over the ~two year time period from March 2019 through November 2020.

If you wish to drill down even further, I included link to all of the relevant cross tables. Each individual game, as well as the rating of each individual opponent, is given in those tables.

2. Whether he knew the pairings ahead of time to prep,

I very much doubt this data exists anywhere but in the mind of Hans Niemann. Maybe by contacting each tournament director you could find out what pairing procedure was used at each tournament, but I'm not going to do that. The regression I ran showed that 67% of the variation in Niemann's performance at USCF-rated classical tournaments over that time period is already explained by the broadcast status of the tournament.

Unless you believe broadcast status is VERY strongly and significantly correlated with Niemann's ability to prepare for his opponents, prep couldn't explain enough of the variation in Niemann's performance, and my results remain damning.

Assuming you've accounted for confounders, even if only 10% of the variation in performance remains explained by broadcast status, it would be incredibly concerning. There should be no correlation whatsoever, so the onus is on you to show me the confounders, and explain how the mere fact that a tournament was being broadcast over the internet could plausibly and honestly spike someone's performance rating by a couple hundred Elo.

(Also, the correlation of broadcast status and performance, and the estimate for its regression coefficient, are almost certainly lower bounds.)

You can make data say whatever you want if you try hard enough.

I wasn't trying to make the data say anything. I came into this drama with no horse in the race. If anything, I was slightly biased against Carlsen because my previous impression of him was that he was a bit of a jerk, a prima donna, and a sore loser. (And I had no idea who Niemann was.)

My opinion was somewhat shifted in favor of Carlsen by Niemann's interviews, because:

• Niemann admitted to multiple instances of past cheating, but essentially claimed the only times he'd ever cheated were the times he'd been caught, which is obviously an absurd lie.
• Niemann seemed completely out of his depth in post-game analysis, a sentiment echoed by a fair number of strong players, even super-GMs.
• Niemann's demeanor and "impassioned" defenses of himself set off warning bells in my mind for narcissistic personality disorder.

Nonetheless, I still remained fundamentally undecided, because I don't assign much weight to evaluations based on limited data, especially when I've never personally interacted with someone.

Anyway, I was sent that tweet and was intrigued by its claims. So I verified the correctness and completeness of the data, gathered more detailed data (like number of games and average rating of opponents) that I thought were reasonable proxies for potential confounding factors (like fatigue, mathematical caps on performance rating, or playing less seriously against low-rated opponents), and ran a bog standard linear regression. (I also carried out an informal, cursory check to see if I could identify any other players from those same tournaments who exhibited remotely similar patterns in their performances. I could not.)

The results of that analysis are so overwhelming and definitive that I have a hard time imagining any other interpretation of the data.

I would not be surprised to learn that you could also conclude that Hans is a significantly better player only when given the opportunity to prepare deeply against his opponent, and that he is unable to beat similar-aged players (who have very un-established ratings) decisively in Swiss events for example.

I dunno, maybe? But like I said, unless his capacity for deep preparation is very very very strongly and positively correlated with the broadcast status of a tournament (which sounds facially absurd to me, maybe you can explain how this could be the case), then it doesn't really matter. 67% of the variation in Niemann's performance is explained by broadcast status, and the regression coefficient for that status has a vanishingly small p-value: 0.00088.

I'm quite certain I don't need to tell you this, but for those who aren't familiar, the usual significance threshold is 0.05, and a lower number means a stronger result. P-value isn't everything, and "p-hacking" is a concern in some contexts, but this is not one of those contexts. All I did was run a standard analysis of an unbroken block of two years of the dude's recent games. Even if the original tweeter "cherry picked" the interval ... it's two solid years of games and a p-value almost two orders of magnitude smaller than it needs to be. Not sure what else needs to be said. The ball is absolutely not in my court.

[M ANSARI]

Watch this clip which for me I had to turn transcript or subtitles in english to understand (there is a youtube function for that). A very very interesting interview that might bring more light to the issue. In the transcript Magnus says that the scariest thing is that someone can get help by just even suggestions of what the best continuation is ... nothing else ... it would make even not so strong players invincible. It is an interesting watch even though is in Norwegian.

[DrCliche]

Both sides can make the data say what they want, as shown here.

The tweet you quote is simply factually wrong, as far as I can tell, given the time I allotted for my initial analysis. For example, your tweeter erroneously includes "quick" (rapid) events because he isn't aware that USCF "quick" events also change one's "regular" (classical) rating. You would be wise to note that I already pointed out this potential misunderstanding in my original post.

So not only did both you and your tweeter fail to adequately research and understand how and why USCF classical ratings change, you didn't even bother to read my post where I relayed exactly that information, and explained why I excluded some seemingly classical events from my analysis. (They were, in fact, "quick" events, not "regular" events, and Niemann's "regular" rating was changed by those events because that's just how USCF ratings work.)

As for the broadcast status of those events, though I was unable to definitively confirm all of the information in the original tweet, neither did I find any information contradicting it. I checked the official websites for every single one of the tallied tournaments, but I didn't perform any sort of exhaustive google search trying to find links or references to live broadcasts. I also already explained this in my original post.

It's possible that both the original tweet's broadcast status claims and my brief checks of those claims have mistakes, but I would certainly need to see more than a random tweet from a random person providing literally no proof or citations whatsoever, who also stupidly and erroneously claims that "quick" (rapid) events were incorrectly excluded from the analysis.

[M ANSARI]

You take the bishop and promote. Is there a point?

... this is maybe 700 ELO level if it was a puzzle on LIchess or Chess.com

[AndrewGrant]

Both sides can make the data say what they want, as shown here.

The tweet you quote is simply factually wrong, as far as I can tell, given the time I allotted for my initial analysis. For example, your tweeter erroneously includes "quick" (rapid) events because he isn't aware that USCF "quick" events also change one's "regular" (classical) rating. You would be wise to note that I already pointed out this potential misunderstanding in my original post.

So not only did both you and your tweeter fail to adequately research and understand how and why USCF classical ratings change, you didn't even bother to read my post where I relayed exactly that information, and explained why I excluded some seemingly classical events from my analysis. (They were, in fact, "quick" events, not "regular" events, and Niemann's "regular" rating was changed by those events because that's just how USCF ratings work.)

As for the broadcast status of those events, though I was unable to definitively confirm all of the information in the original tweet, neither did I find any information contradicting it. I checked the official websites for every single one of the tallied tournaments, but I didn't perform any sort of exhaustive google search trying to find links or references to live broadcasts. I also already explained this in my original post.

My point is to say, why should I take your analysis as true and correct, and not this guy's analysis as true and correct? Is removing quick events proper? Does that fact that you could not verify a few of the events have a significant impact? Were the few events pointed out marked correctly or incorrectly, can you answer that? Should your analysis take into account what I mentioned before about seeded events and Swiss events? Should your analysis take into account the increased volatility of ratings of young players?

No easy answer to those questions, and so I am not inclined to take anyone's spreadsheets as proof of anything more than one's ability to find data that supports their view. The only people who can offer compelling arguments here are 1. Carlsen, who has refused to speak, and 2. Chesscom, who has claimed to send their information to Hans. Hans has not refuted this, so we can assume that Chesscom did indeed send Hans their information, and that Hans would prefer it not be public. I'm happy to hop on the online cheating bandwagon as a result, but OTB Carlsen must speak.

It's possible that both the original tweet's broadcast status claims and my brief checks of those claims have mistakes, but I would certainly need to see more than a random tweet from a random person providing literally no proof or citations whatsoever, who also stupidly and erroneously claims that "quick" (rapid) events were incorrectly excluded from the analysis.

As far as the people outside this forum are concerned, you are also a random tweet from a random person with little to no citations.

[chrisw]

Did nobody consider that this whole affair isn’t a set up? Chess.com and Magnus are now a corporate entity with massively increased profile and a valuation in proportion to public interest in chess in general. Venture capital backers of chess.com must be rubbing their hands with glee. Breaking into mainstream with a Magnus-Hans televised grudge match posing as World Chess Championship by-passing FIDE, anyone?