How can a programm evaluate human playing? The programm would have to be able to judge whether a move is "human" or not. But therefor it has to understand how a humen thinks, when he plays a position, in fact: which moves are natural.
And I believe that this is not possible for a computer programm.
Also players could have analized the moves at home with an engine and would be designated as cheaters.

This all seems for me like the attempt to accuse a student of cheating, who normally has bad grades, just because he wrote a good grade. You could argue, that his writing style was different but this can never be proven, maybe he was just well prepared or he advanced. The only thing you do is to accuse him unjustified.

But this is just what I think about this without knowing how this programm works...

I think it would work comparing players of different calibers making closer to ´best´ moves according to the computer programm. For example a 2600-player would have 90% of his choices in the top 3 of computer candidate moves, 80% in the top-2 and 70% would be ´first choice´. For a 2300-player these numbers would look like 80-70-60 etc. The percentages are arbitrary of course and might be different in practice. This is also one of the methods used to determine cheating online.

I really don't know what I am supposed to conclude from his published results. Hou Yifan is at the top of the list - is that supposed to mean she is the most likely cheater of the listed players? Assuming she is not, then her position on the list, indeed the entire list, seem rather meaningless.

---deleted---
http://www.chessvibes.com/terms

so whats the point?

Funny how they call this 'research'.

"...It cannot conclusively say whether a player is really cheating; all it can do is say that the move the player chose is strangely similar to those chosen by a chess engine..."

How is this different from Mamedyarov claiming that his opponent's moves matched Rybka's moves a lot?

And ChessVibes, please distinguish between mathematics and computer science. I can live with computer scientists making fools of themselves, but then also 'attribute' the results to computer science. So "Using computer science to detect cheating" would be more accurate (and less insulting to mathematicians).

Fascinating statistical analysis, but hardly a credible methodology for putting reputations at risk. This system clearly would "mark" players as cheaters who in fact are not. Back to the drawing board!

"Using math to detect cheating" is, in fact, cheating....

Good to hear that some people are actually doing serious research in this area.

I think that something tha these researchers are missing is the fact that a cheater doesnt have to cheat on every move he makes during a game. In fact a titled and experienced player knows what the critical points in a game are or when a lot of tactial calculation is needed so in a single game a lot of his moves doesnt have to follow an engine.

Programmers try to build evermore human thinking into their programs So in fact the rsesult would mean that they succeeded. And, by the way, who checks the computers that are used to find the cheaters? You could now already rig your computer so that it proves that Morphy, Eeuwe etc used computers to cheat.

Statistics is not a prooving science. It cannot proove anything.
It can only indicate the probability for a fact to be true or not.
Also, statistics is obliged to report the exact percentage for a fact to be true.
We have AGREED (arbitrary, of cource) to accept as "true", a fact that has a
percentage of probability higher than 95% (statistically significant).
Mostly so when it exceeds 99% (statistically very significant).
But be careful. They are only working decisions.
They are, in no way, considered to be proofing evidence.
And of cource not in a law court. They are not evidence is a legal way.
They only help scientists to make decisions on how to proceed.

In my opinion, the title of the article should be: "Using statistics to suspect cheating".
That would say all.