Originally posted by DeepThought
I think there could be dangers in this, I'm not convinced you should have posted your figures, A lot of the games you analysed were against engines, which could cause statistical problems. You need some way of randomizing your choice of games to analyse...
I disagree.
I think that the games need to be against strong opposition and/or engine users.
1) Weak opposition chosen randomly would cause the strong player to play obvious 1st/2nd choice moves when the weak player plays... err weak moves. This could skew results, causing an innocent stronger player to come under suspicion when they really shouldn't be.
Strong players will put up more resistance, create positions of greater balance (therefore many moves with slight variances in score - so engine match-up stats importance are increased in value) and not least the games will last longer so crucially you get more moves out of book.
2) Games vs engine users at least give you some sort of frame of reference in a game. If banned engine has say 32/36 top 3 moves out of book and our suspect has 34/36 then over time his performance against the engines is a factor.
If he is consistently picking Fritz's top 3 moves more than other users banned for engine use then you can draw a rather simple conclusion.
ps
I know it's not proof. As I say it is merely statistical evidence.