************************
This is a posting connected with Harvey Friedman's posting http://www.cs.uky.edu/fmpc/archive.dir/0605010.html
"What if the top human players (who are really interested and have a good intuition for this kind of thing) are allowed to play a given chess computer for, say, a week (or a month)? I still believe, perhaps naively, that the top human player will find the weaknesses and kill the computer. Even if the computer has some "learning mechanism" built in."
Of course I am not a top player but I did some experiments. I played against Fritz 8 in the last months. Fritz 8 is nowadays pretty cheap since the currently available latest version is Fritz 10 which beated Kramnik last year. The modus was 4 minutes each and two seconds bonus per move:
I won 7
drew 86
and lost 705
My ELO computed by Fritz is 2522 which is better than my usual one. Fritz rated himself by 3000.
I found severe weaknesses of the computer but these only allowed for a draw. When I won a game the computer changed to a completely new opening. To be able to draw I played typically strange openings with white like 1. e4 e6 2. d3 Kc6 3. c3 d5 4. Kf3 dxe 5. dxe QxQ 6. KxQ This turns out to be useful to reach solid positions with white. But even drawing better positions turned out to be pretty difficult.
Andreas Weiermann