Author: James T. Walker
Date: 06:22:27 03/19/99
Go up one level in this thread
On March 19, 1999 at 08:24:50, Ulrich Tuerke wrote: >On March 19, 1999 at 06:55:15, blass uri wrote: > >>Knowing the opponent is allowed so you can buy many computers >>play 1000 games against every posiblle opponent and remember the games that your >>program won and repeat them. > >Things are not that simple. You need 2 programs in order to reproduce games. >Nowadays programs have large books with lots of variations, where the moves are >selected by (weighted) chance. You need a lot of luck to reproduce some >variation in a match of 20 or 40 games as proceeded by the SSDF. > >On the other hand, computer chess is an attempt to simulate human chess. Usually >a human player does know the opponent, and he has time to prepare. Exactly the >same rules should apply to computer chess games. Sure, someone having time and >resources to prepare thoroughly can improve his chances. That's life. BTW, >wouldn't it be nice if Junior had known in some SSDF games that Comet is the >opponent. This way, some draws could have been avoided. In human games, a >grandmaster will neither repeat moves against an amateur to enforce a draw. >Games could become more human-like. > >Uli > >> >>I believe that a program can get more than 2700 ssdf rating list in this way. >> >>Uri Hello Ulrich, Right now it's not that hard to repeat games. I played a 52 game match between Fritz 5.32 and Junior 5.0 in which Junior 5 played the same two openings with white 19 times. Junior 5 won all 19 games. With auto232 Fritz 5.32 is not learning with the black pieces and Junior5 took advantage of the fact. Auto232 allowed Junior5 to learn with the white pieces though ! Jim Walker
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.