One thing I'm curious about...
How would/could the accused go about trying to prove their innocence? 

When writing the FKG I remember analysing an old Estrin game with Fritz, and at one point both players followed a fritz line for something like (at least) 10 ply in a row - and at the time of the game fritz didn't exist and computers were hardly any good...

One of the big points of playing corr instead of otb is to have more time to make better decisions, so wouldn't one expect strong corr players to be closer to "perfection" than GMs playing OTB, even without computer aid? 

And even with computer aid, the correlation between choosing one of the top three moves from the computer must also be dependent on at what time you take the "top three" - in a complicated position and allowing it, say, 48 hours thinking time it would probably produce a very long list of different combinations of "top threes" during that time. While at another point in a game there may be only three reasonable moves to choose from.

This topic reminds me of a recent local football story, some trainer was complaining that the statistics clearly showed that the referees' decisions (freekicks & cards) are mainly in favour of the top teams when they play the teams at the bottom. He tried to make a big deal about this, until someone simply pointed out that the top teams usually also have the ball most of the time in those games...


Anyway, since it's practically impossible to prove cheating beyond any doubt, I guess the person accused would have a good case for a libel suit in the US. 

And it would be a wise decision to remove the name of the player accused by chess.com from this thread, as all is just hearsay and accusations based on statistics. After all the forum is high ranked on google on all chess topics, so it's unfair to expose anyone on so flimsy evidence.

These statistical analyses do remind me of Topalov's toilet-gate.
As far as I remember, Kramnik scored well above the norms used by the cheat-hunters.
At the time it was allready pointed out, that both
lengthy opening lines, opening preparation, exchanges and forced lines all contribute to an elevation of these scores. So you have to look at the games themselve to judge if computer assistence is likely or not.

Does it have any common sense to speculate about reasons for cheating? If yes, please tell me. It will make me think a little.

From experiencing the differences of CC with and without engines I really doubt you can proof anything by the method published in the topic in chessgames.com. But maybe here I will learn something too.

Most interesting is the public damage done to a person without a clear-cut proof and the behavior of the public and the chess.com site as well. A German proverb says: I can't gorge as much as I want to barf. 

Smyslov_Fan wrote on 09/20/10 at 04:41:33:


Leaving aside the question of whether this is to be believed at all, and further that of whether it was done with any rigor, the devil is in the term fairly consistent, which indicates that the entire process is subject to error.  It is, therefore, a statistical question.
Smyslov_Fan wrote on 09/20/10 at 04:41:33:
 

The same?  I posit on the contrary that choice of engine is a source of variation.

Smyslov_Fan wrote on 09/20/10 at 04:41:33:


Clearly this is not the maximal possible degree of matching, since you contradict this below.  So what is it, an average, a median, something somebody pulled out of thin air?  What is the population?  What is the sample?  What are the methods?  We are simply in the dark.

Smyslov_Fan wrote on 09/20/10 at 04:41:33:


Since the process is subject to error, the question is, given some reasonably specified statistical model and some estimate of the variability of human-machine matchup, with what degree of confidence can it be said that Dembo's matchup was different from that expected?

Somehow I doubt that sufficient rigor was applied to this question.  If the entire analysis and all the data were reported on the internet, one could review it.  In the mean time, scant credit can be given to its alleged results.


Smyslov_Fan wrote on 09/20/10 at 04:41:33:


Don't make me laugh.  Until I required quite recently, I made my living this way, you know?

Smyslov_Fan wrote on 09/20/10 at 04:41:33:


What tolerances?  What approach?  We simply don't know enough to evaluate the method used.

Smyslov_Fan wrote on 09/20/10 at 04:41:33:


Shame on you.

Smyslov_Fan wrote on 09/20/10 at 04:41:33:


Run the numbers yourself????  With what population? With what data? With what model?  If you had the expertise to "run the numbers," I expect you would have raised the questions yourself by now.

And of course it's not certain, since at best, it's a statistical result!!

Finally let me say that only an idiot would want to match a computer all the time, since computers remain rather stupid about many positions.  So the very notion that rate of matchup with some engine is likely to correlate with success is bogus.

TonyRo wrote on 09/20/10 at 13:34:26:


It does help (sometimes) 

TonyRo wrote on 09/20/10 at 12:46:48:


IF she has cheated I would guess that the reason probably won't be very rational (from a professional standpoint): perhaps because she felt that every one else was cheating, for the thrill,... .

It would be a bit like a millionaire shoplifting.