Hans is clearly cheating. Comparing his past games against what an engine would do is pretty damning. Chess engines are far superior to players and the best players in the world top out in the high 70s percent correlations (Magnus averages around 70%).
Hans has a string of games at 100% correlation[0], meaning he's playing perfect games. Past players who achieved this later went on to admit to cheating[1]. Magnus knows this because he owns part of chess.com and presumably sees the data.
Magnus has a lot riding on his statement. He wouldn't make it unless he was sure.
It's cherry-picked games and it doesn't compare to the "engine correlation" of other high ranked players against similar opponents. I would not rely on it as evidence that Hans is "clearly cheating."
She says that the ROI probability calculation is wrong, which is the last part of her video and a separate topic.
The part about correlations has not been retracted AFAIK. I agree that there's a need for a baseline though, there is one example on her recent twitter feed but more samples are needed to get a better picture.
No it's not. Just the probability calculation had a few flaws.
The string of 100% games, esp. in the few tournament games he needed a win, is decisive. Note that he played a lot of games in each tournament on the 70% level, to fool statistics.
When I was younger, I spent many, many hours playing one particular video game. I became a “known cheater” at the game, despite never actually cheating (I’d cheated at other games in my early teens, but had since given up that lifestyle).
I can recall several players on discussion boards analysing my statistics and explaining how I was clearly cheating because it was impossible for a human to play like me. Humans, they said, just weren’t that accurate.
One cheat-detection algorithm even “caught” me one day, and I was promptly banned from that server. Confused about what had happened, I sought out the server documentation online so I could see what they had used to “detect” me. My crime, it turns out, was scoring too many kills per second.
I keep this in mind whenever I see another person accused of something similar. Sometimes people have just put in more effort and study than we choose to comprehend.
In one of my high school computer classes we'd often set up UT2003 LAN games. I was once called a cheater by someone sitting behind me who, at any point, could have directly observed my every move and corresponding input. I was baffled and amused at the same time. People that lose games have plenty of incentive to claim their opponent is cheating.
On the other hand, as my skill level increased in FPS games it became more and more obvious when one of my opponents was cheating. So IMO you can trust Magnus' ability to accurately estimate the chance of Hans cheating, but you can't trust his motivations for making the claim.
Aside: When you combine high skill levels with cheats that were designed first and foremost to avoid detection, it becomes almost impossible to do detect them. For example in FPS games "aim-bots" are crude compared to "hit-scanners" that simply auto-trigger when your crosshair happens to pass over a valid target. Combine a hit scanner with a player who already has top tier accuracy, and you get super-human accuracy. Let the player enable and disable the hit scanner in real time and they control exactly how accurate they are without any conspicuous appearances. You'd have to (externally) record and sync the monitor's output with a camera that monitors their mouse movements, and even then you'd need EXTREMELY accurate timing - most likely a capture rate higher than the monitor refresh rate.
Chess has a rating system, and we know what the best humans rate out at compared to the best chess engines, and we know the likelihood of lower ranked players beating higher ranked ones, just depending on how much difference there is, and where in the ratings the two players are. Chess is not some video game one person perfected playing.
The parent said they were accused of cheating at a game because they played it enough to perfect it, which isn't something humans can achieve with chess. Even the best chess engines don't play perfect chess, they just play at such a high level that no human can beat them.
At any rate, a 19 year old ranking a couple hundred points below the world champion for the past decade isn't going to have enough practice to use that as an excuse. So no, it's not the same thing.
Outlier players exist in every game, probably because of normal distribution.
Federer for tennis, Carlsen for chess, Serral for SC2 and dorkwood for his game.
The only way to certify their level is through strictly controlled competitions. eSports and chess do the same.
My most pro game experience was Battlefield 2 and I'd go on a public 64 player server and score most kills with only the knife. Thing is: My team mates and other pros were capable of doing the same. No pro would accuse us of cheating, just the average Joe's.
I saw something about a runner being disqualified because they moved 9 hundredths of a second after the gun. Ten hundredths is permitted but nine is not.
It’s not impossible to believe someone could be just a sliver faster.
Ten hundredth is what’s impossible plus a generous safety margin. If you are under it’s definitely a false start. If you are just over, it was most likely a false start but you were lucky this time.
It is extremely unlikely that a human being can hear a sound and move their foot within 100ms. Based on what we know of the brain and nervous system, it's most likely that anyone who reacts under 120ms has jumped the gun.[1]
In case anyone is curious, the starting times for the first five runners at the 110mH event at the 2022 World Championships were (in milliseconds) 99, 108, 109, 124, and 126.
While I personally disagree with the comment I am replying to regarding reaction times under 120ms, the issue at this World Championships was more likely that the reaction timing was wrong ...
Agreed. His ELO improvement alone is evidence enough imo.
His rating plateaued around 2300 from end of 2015 to mid 2018. Then in the last 2 years, I believe his improvement from 2400 to 2700 is the most rapid in history.
If you compare his rating with other young players like Duda, Firouzja, Gukesh - then his rating increase looks very unique.
I'm not aware of any chess prodigies that have followed a trajectory quite like this. So maybe Hans is an unusual talent. Or maybe, he's receiving computer assistance.
Daniel Rensch and others have said that Magnus has not seen chesscom cheating algorithms or lists. It had been a rumor in the chess world for a while that Hans has cheated before
He hasn't been officially shown but that doesn't mean that someone in the know hasn't leaked it to him. He's deeply connected and respected in the chess world.
Further, even if he didn't see it people notice when GMs get bans as their accounts turn inactive (which they had in this case).
Hans has a string of games at 100% correlation[0], meaning he's playing perfect games. Past players who achieved this later went on to admit to cheating[1]. Magnus knows this because he owns part of chess.com and presumably sees the data.
Magnus has a lot riding on his statement. He wouldn't make it unless he was sure.
[0] https://www.youtube.com/watch?v=jfPzUgzrOcQ
[1] https://en.wikipedia.org/wiki/S%C3%A9bastien_Feller