Hacker News Reader

Comment by tromp | Hacker News Reader

Applying sound statistical methods to thousands of Niemann games, cheating expert Ken Regan finds no evidence of cheating [1].

> The verdict

> Dr. Regan analyzed all of Hans Niemann's games over the last two years, including online games, such as played on Chess.com and their events, and his conclusion is there is no reason whatsoever to suspect him of cheating. The wide range of results in a bell curve, with some good and some bad, is actually a sign of a healthy distribution of results. Many of the so-called points of suspicion are in fact quite normal and suspicion is really the result of faulty analysis by zealous amateurs. Even online his play has been quite devoid of anything unusual.

[1] https://en.chessbase.com/post/is-hans-niemann-cheating-world...

View on HN

This is a strong point in Hans's favor, and Regan is very reputable, but he's also going against the Chess.com team (and Hans himself!) in claiming that he finds no evidence of Hans cheating *online* either.

On the contrary, I don't find it significant at all. I had not even heard of Dr. Regan. He seems to have achieved the IM Title decades ago, today he is probably simply a weak player.

The best expert on opining whether Niemann has been cheating or not is the player who understands the game of chess more deeply than anyone else. He understands it from the human point of view (for example: which variations can a player of X or Y level calculate accurately, which ones might seem to work but are in practice too complicated/too risky to play, which moves are absolutely unbelievable from a human because one has to find ten super moves in a row starting from a non-forcing position) as well as from the engine's point of view, due to his experience in analysing with computers. He also has recent experience in playing Niemann in a tournament.

I trust that Carlsen knows what he is doing.

> He seems to have achieved the IM Title decades ago, today he is probably simply a weak player.

Wow.

Regan has been an active and respected member of scientific community, a lecturer and book author for a long time. If anything that adds and not takes away from his credibility. He's also the most prominent expert on cheating detection that FIDE has ever worked with and he's an authority on the matter.

Oh, and he's rated at around 2400 ELO. But I guess it's very easy to just slander him like that. Go back to reddit.

I will quote you Fabiano Caruana, who said the following on September 22, after my previous comment:

“I would take Regan’s analysis with the large grain of salt, and the reason why is not because I have any insight into his algorithm or his methods, but because I know of a case of, a very high profile case, where with absolute certainty I can say that someone was cheating in an important event. And the person was investigated and was also exonerated based on Regan’s analysis. And I am certain that there was cheating. There is no doubt in my mind that this person was cheating and they got away with it.“

So this is one more example of an academic, who is so far away from reality and practice, that real experts just ignore his opinion.

Edit: Regan is rated 2372, he was born 1959 and his last tournament was in October 2012.

In a similar way that the best chess engine developer is not the best chess player there's little reason to think that the best person to detect chess engine use is the best player.

Indeed there is every reason. The hardest part is to understand the game of chess. As i wrote, it includes having the entire tree of variations for every position in your mind while investigating. It is absolutely not a matter of simply comparing one-dimensional engine evaluations.

No, the hardest part is likely to understand statistics, have data on cheaters and non-cheaters, analyze the differences, develop and test your methods on validation data, iterate on your process etc.

>includes having the entire tree of variations for every position in your mind while investigating

Too bad even Magnus can't do that and the closest you can get is by writing software to do it for you.

I do not even understand what it is that you think Carlsen cannot do. But every master can do it for one position at a time. Then he just analyses a number of positions, from Niemann's games and reaches a conclusion.

You did not describe the algorithm or the data which would help writing the software you are talking about. As I said, for me if this software would be based on comparing the engine's best move with the human's move and give such metrics as average centipawn loss, it would not be nearly adequate for this purpose.

And I insist that all the things you described as the hardest part are only poor substitutes for understanding the game of chess.

I think the important bit is the "last two years". Niemann has admitted cheating like 5-6 years ago.

2-3 years ago* when he was 16, he is now 19.

I remember reading when he was 12 and 14

When he was 16 - that 3 years ago at max.

His analysis is on games in the last two years, and Hans' admitted cheating online is before that.

Quoting Dr. Regan:

""What I'm saying, as justifying my not needing to take the time to individually into tournaments to see which were broadcast and which were not, is that if there is any bias in my data, then it's towards broadcast games (i.e more of it is analyzed due to availability) and yet I show something entirely normal."

This is a simple logic error. The guy might be reputable but haven't really thought this through.

Quick explanation of the mistake if it's not immediately obvious: it's possible to cheat in a way that his statistical method won't catch but the difference in performance between broadcasted and not broadcasted tournaments will highlight it. His statement assumes that his statistical model will catch every cheater which is obviously nonsense. His methods might be very good in a sense that if they show cheating we can be sure cheating in fact occurred but it can't be good enough to find every cheater, if anything because of limitation of input: just the chess moves. For example he wouldn't catch a stronger GM communicating moves to Hans (it's still very human and normal, just stronger) while comparing performance in broadcasted and not broadcasted tournaments would (if someone communicated the moves to Hans in broadcasted ones).

It's very hard to take the guy seriously after that, especially when he is very arrogant about "there is no evidence" statement while it would be more prudent to say his limited statistical model didn't find any.

Even his online play where he admitted to cheating?

The online cheating occurred more than 2 years ago. The analysis covered the past 2 years.

Not that hard to reconcile......

"Dr. Regan analyzed all of Hans Niemann's games over the last two years."

How can one even cheat in chess? When I was a kid we used to distract the opponent and move some figures as he didn't watch. Apparently this is not the case but I can hardly imagine another way.

There are these things called chess engines - current best is called Stockfish - which calculate the best possible move in any given situation.

To cheat in chess you input the current state of the game and you get the best possible move out.

One method I can think of immediately (when playing face-to-face in the same room)

You carry some device which lets you input data, computes the best move, and actuates that response back to you.

Let's say that this device is simply a very small microcontroller which takes in input via some pressure sensor, and outputs via some actuator that creates vibration. You've loaded some chess program onto that microcontroller.

You hide the device somewhere on you, tape the input pressure sensor to your thigh, and paint / cover up the sensor and wire with material similar to your skin. Same with the output actuator.

When it is your turn, you input the current state of the board via the pressure sensor. Input sequence is sent to the program, which calculates best possible move. The output actuator vibrates the sequence of what piece, and where to move.

Probably one of the older tricks in the book, as far as device-guided heating goes.

Someone built this and posted it on HN a while ago.

Here's one option - https://news.ycombinator.com/item?id=32729105

Reply on news.ycombinator.com