Author: James T. Walker
Date: 13:40:02 03/25/02
Go up one level in this thread
On March 25, 2002 at 15:20:44, jerry young wrote: >Why would one want theri prog to learn anything from 1 minute games ? Such fast >games can create "anti" learning, fooling the program into finding a baad move >statiscally good because in a fast game it is followed by an even worse move. >The older learning programs had a minimum of 20 seconds per move before ani >impact of learning on the books was deemed valid. Hello Jerry, The idea was not to see if learning from 100 games made a difference but to see if previous learning from 1700 games had any effect on this 100 games. Jim
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.