Author: Bertil Eklund
Date: 12:14:27 11/18/98
Go up one level in this thread
On November 18, 1998 at 03:32:57, blass uri wrote: >I found that fritz5 won 5 times the same game against Rebel8 in the ssdf games. > >I ask if Fritz5 is best in repeating wins or there are better programs in >repeating. > >The learning that help fritz5 is repeating winning lines with big probability >without remembering the winning game as theory. > >It is more logical to remember all the winning game as theory against humans but >against deterministic programs that cannot learn it can be counterproductive >because the computer opponent cannot use the permanent brain in this case and >can do another move and the game will not be the same. > >Uri Gandalf3 won the same game against Rebel9 4 times. Gandalf3 won 7 and draw one against McP6 in McP6´s special bishopvariation!1.e4-e5 2.Lc4 McP6 tried new variations each time but it was to late. Junior won 4 times with 1.c4-c6. The problem with Rebel´s booklearning is in a line where it is no alternative moves it plays it over and over again. Rebel is a static program that always play the same game. This never happens with programs that always play different i.e. Nimzo98 Mchess Genius5 and Fritz have learners and plays different next game with the same opening. Rebel10 with its huge book doesn´t seem to have this problem, I have done some testing with Rebel10 against other top-programs and my guess is that should fight for place 1 to 3.But of course we are not allowed to tell how good this program really are. Bertil
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.