Author: Thorsten Czub
Date: 08:48:29 08/23/98
Go up one level in this thread
On August 23, 1998 at 11:39:33, Jay Scott wrote: >What you're saying is that Fritz's evaluator has large systematic >errors. But as long as it's OK at ranking positions in the search >tree, it can still play good moves. In the absence of an immediate >draw, only the relative evaluations of positions matter, not the >absolute evaluations, and systematic errors don't hurt. > > Jay This is also my idea. Maybe there are many many positions where fritz has no criteria to evaluate at the root because it has no idea about the position. But fritz strength is not the right evaluation but deep search. Maybe fritz does not decide with EVALS which moves to put in the tree and which not, maybe it uses different things than score to decide about the ranking/move-order. I think this happens when I see these draw-areas. Than fritz decides about moves without having a knowledge or evaluation base. Sometimes it works, often not. I can only say from experience that fritz loses/draws when i see these behaviour, or vice versa, this behaviour often leads into senseless draws or stupid moves resulting in lost games. This is all i want to say. I thought this may be interesting, because I have not yet seen fritz lose a game without this behaviour.
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.