Author: Komputer Korner
Date: 13:25:16 05/18/98
Go up one level in this thread
Nimzo 3.5 had a feature whereby you could turn on a special annotating learn table whereby when analyzing back to front, it would store positions and evaluations it had analyzed in the higher move numbers which would speed up and thus deepen the search at the lower move numbers. This would seem to work better than the normal lower to higher mone number post game analyzing. On May 17, 1998 at 09:07:26, Marc Plum wrote: >I have been using the auto-annotation features in Fritz5, MChess Pro7.1, >and CM5500 on a number of games, some of my own tournament games, master >games, games played against the computer. Partly this is useful to >check for tactical resources, but I have mainly been interested in >insights into how the computer evaluates. > >When a human is annotating a game, he will use hindsight to correct his >initial impressions as he goes. For example if he thinks a combination >is unsound because of a particular defense, and then sees that defense >fail because of a resource that he overlooked, he will adjust his notes >accordingly. With CM5500 and MChess (which go through the move list >forward), the notes would originally show the combination as bad, then >abruptly switch to the opposite evaluation when the outcome becomes >clear. Fritz 5 works through the move list backward, and it might >almost seem that it is using hindsight of this kind, but I am not sure. >The examples that I have so far are inconclusive. > >How difficult is it to program hindsight into an annotation program?
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.