Using an ultra-high ELO chess engine to score each possible move, then reversing through the players moves and seeing how often it would have been a positive move (one that shifts the balance of the game in your favor) - or perfect move (not sure which). It is extremely rare to make 100% perfect moves in a game, let alone a series of games. Typical gameplay for high level chess player doesn't peak over 72-75% for a given series of N games. Niemann has several tournaments over this and several games with 100% perfect moves. The inconsistency is also a concern since he goes from mid-60's to 78/79 in a span of one tournament.

His games against Magnus were exceedingly high.

It's also worth pointing out that a player's odds of making the perfect move are inverse to their opponent's ELO: as the level of play rises, finding the right play becomes exponentially harder. The data suggests he's sometimes playing other grandmasters as good as those grandmasters would play a rando on Lichess.

It's Elo, not an acronym.

GP is not pulling this out of their butt, they are summarizing the video like GGP requested.

Also, your anecdote doesn't prove anything.

If someone wins the powerball the first time they buy a ticket it’s still a rare event.

yeah - if I pulled a random willing powerball ticket out of my massive pile of powerball tickets, that would be a really, really rare event. It would make me believe that it isn't such a rare thing, for sure.

FM Yosha puts forward a fairly convincing argument about odds and engine correlation, but another commenter rightly pointed out that these statistics are not seen as incriminating in and of themselves. Unfortunately, even when the preponderance of evidence seems to be against a player - best example is Sebastian Feller (https://en.wikipedia.org/wiki/S%C3%A9bastien_Feller) playing with superhuman accuracy at crucial moments, and whose team captain later admitted to helping him cheat - they can still cast enough doubt to be allowed to continue playing at the highest level.

Here is a blunder that Feller played on move 13 just over a month ago (https://new.chess24.com/wall/news/grandmaster-blunders-mate-...) - this same guy managed to draw against Magnus Carlsen in 2008, in a game where Carlsen also found the moves/mannerisms of his opponent highly unusual.

It's been talked to death. The consensus is you cannot just cherry-pick some games and claim he's cheating.

Everyone has games that are perfect. Everyone. Not just GMs or Super GMs. I have at least a few perfect games and I'm half the rating Hans is.

The games analyzed also have crazy blunders by his opponents.

Perfect when compared to the moves the top chess engines would make? Hikaru says he only scored 100% one time, and 70% is more typical for a GM, yet everyone does it?

First, it isn't THE top moves. It is ONE of the top moves. Huge difference.

Where does Hikaru say he only has a 100% correlation game one time? I've seen lots of examples of other players having such games.

the gist seems to be that he has unrealistically high correlation with game-engine recommendations, often all the way up to 100%, but only when playing "tough" opponents, and far lower / realistic correlation scores (around 50%) in other games.

for reference, magnus carlsen's correlation score at his peak averages around 70% (according to the video)

Ai written summary https://www.summarize.tech/youtu.be/jfPzUgzrOcQ

Cool! is this something you are working on?

Yes! I built this for exactly this use case :)

This is super cool! I tried it on a video I recorded a while ago that I completely forgot about and was like wait wow the summary came away with points that I'd be glad a viewer got.

One thing is it was a very long and rambling video and probably didn't do a great job of motivating examples rather than just getting bogged down in them for a while, so the summary doesn't really say how the examples support the central claim, but that may be the fault of the video honestly lol...

Also a few basic errors like writing "medium" where I'm pretty sure I said or at least meant "median" and in one case, I'd have to go back and watch this to be sure, but it seems like the summary says something is better in B than in A when I was saying it's better in A than B. The summary definitely touches the right content but I'm not sure it's correct.

Also funnily I have a tendency to sprinkle the word "like" liberally(for better or worse) and the summary copies some of the sentences verbatim, starting with "Like..."

(completely off the topic of cheating in chess, sorry...)

He played several games with 100% correlation with what chess engines considered to be the best move, and also played in 5 consecutive tournaments with such a high fraction of engine-preferred moves that his performance rivals the best players in history at the pinnacle of their careers.

Yes, and so does.... everyone else that is 2600+. And lots of people who aren't.

That's simply a lie.

It isn't a lie. Every person over 2600 will have games that 100% correlate with the engine using these methods.

I have games, as a 1400-1600 that are perfect games.

It would be nice to get an ELI5 on this too. I used to play chess and have an understanding of the significance... but I don't think I can fully appreciate it as well as someone with a solid background in both.