Author: Vincent Diepeveen
Date: 09:41:55 02/24/99
Go up one level in this thread
On February 24, 1999 at 03:43:14, Harald Faber wrote: >On February 23, 1999 at 11:55:42, blass uri wrote: > >>It seems that they cooked Fritz5.32 to win the nunn test games. >>This proves that the nunn test does not prove nothing about the strength of >>chess engines because it is a known test. >> >>I suggest another test based on the positions after the 10 move of black in the >>last category 18+ tournament. >>Uri > >Me and Peter Stahlhacke are constructing an alternative Nunn-test with 152 >positions all over the ECO codes. They are the most popular opening lines played >by >11,000 computer games and we effort to take equal positions where both sides >have chances. >It would be great if some others would do the same with human games out of the >ChessbaseBigbase or so. Volunteers? The more positions from mainlines, the more impact a good book will have. If i remember well the original idea was to play *without* book, and that's not the case here. all positions from Nunn are in the book, so the best book still wins there. If you turn however book *really* off in so many positions, then i'm sure f5.32 will get crushed instead of crushing others. The problem of a known test set is that the programmers have that too. The best test of a program is to see how it performs at the world championships, regrettably only 7 games next world champ, but those games are the games which count. Crafty has well prepared with a 16 processor system or something, diep is well prepared with a 4 processor system (thanks Bob!), K6-3 seems very fast, so does the PIII-500. Let's see what happens in paderborn!
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.