Computer Chess Club Archives


Search

Terms

Messages

Subject: computer learning

Author: walter irvin

Date: 11:07:33 07/29/00


computer learning should be the next leap .you can tweek your chess engines
until you are blue in the face , maybe get 20 or 30 elo .or wait til faster
hardware saves you .but if programmers want real elo gain 200 -300 then this is
the way .egtb has helped BUT alot of those positions the computer would have won
any way ,or the computer gets beat before the end game is reached .

the oppening is where the game is won or lost at the highest level .computers
SHOULD be way ahead of people in this area .a program should NEVER have to spend
time thinking on a position that it has played before AND never lost .also
anti-GM should be easy ?? create 5 different engine settings all good but with
vastly diff playing styles .when anti-gm is selected the program generates a
series of random nums between 1 and 5 .thats the order the diff styles or engine
settings would get played .you could also have the program return a random
number between 3 - 12 . that would be the number of moves each engine would get
to play before the next engine takes over .to me its stupid to play a GM a game
when:::play your program figure out exact moves to crunch you .this way it would
be very hard to reproduce a game to beat you .also anti-GM should only contain
oppenings that are wide oppen if possible .



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.