Author: Matthew Hull
Date: 12:13:11 08/26/03
Go up one level in this thread
On August 26, 2003 at 14:10:11, chandler yergin wrote: >A question from a Novice here, so please bear with me. >If, a Chess Program is "out of book", then the static weights given to the >program in the quiescent position determines the evaluation? >This may vary from program to program right? >Which is why different programs often disagree even in the PV & secondary line? >What comes to mind is the Urusov Opening, where very few games have been played. >Statistically, the White wins are not credible; for example, Steinitz & lasker >both fell into an Opening trap or combination, so if the program picks a move >statistically, it would not be the best. >I'm just trying to understand more about how the program works when on it's own >so to speak. >Any info would be helpful, >Thank You, >CY You are correct. A badly composed opening book can land a program in a bad position from which it's evaluation function may not be able to deliver it. For instance, if a program plays closed positions badly, then having an opening library full of those kinds of destinations will scupper that program's chances of winning against opponents good at closed postitions. Using human play statistics alone to compose opening libraries not sound as you have correctly pointed out. That's why commercial programmers hire out this kind of work, especially for championship events, tuning the book (and keeping it secret) to suit the strengths of the program in question. In this way, and with a little luck, a mediocre engine might lift it's ELO performance in an event. MH
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.