Author: Uri Blass
Date: 12:07:40 12/29/01
Go up one level in this thread
On December 29, 2001 at 13:59:25, Ed Schröder wrote: >On December 28, 2001 at 08:14:02, Uri Blass wrote: > >>if you extend the qsearch too much you may get >>often irrelevant positions that you analyze. > >So? > >What you get back is an accurate score of the leaf, the only thing that counts. > >Limiting QS to 6,7... plies is a bad thing because it will return an unreliable >score too many times. > >Ed I think that if the result of the qsearch is based on selective search of more than 7 plies forward then you are going to be wrong often even when you do not limit the qsearch so the question is if you lose more than you gain from it. Limiting the number of moves in the qsearch in the last plies may be a good idea and I agree that deciding about a limit of 7 plies is not the best idea but deciding about no limit for the qsearch espacially if you let checks in the qsearch based on another post can cause the qsearch tree to explode and it is better to use evaluation that is less correct than not being able to search deep. I may consider also having checks in the qsearch but it is clear that I need to prevent the qsearch to explode so basically the qsearch should include less moves in later plies and you can finish with only one move in the last ply that is also a way to prevent the qsearch to explode. Christophe explained in another post that he does SEE by this way and it may be a good idea to finish the qsearch by this way when the only question is how many plies you need to search before you finish the qsearch in that way. Uri
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.