Author: Alessandro Damiani
Date: 03:06:09 06/27/98
Go up one level in this thread
On June 26, 1998 at 11:47:17, Ulrich Tuerke wrote: >On June 26, 1998 at 11:03:24, Inmann Werner wrote: > >>Today I did some "time waste counting" of my program. >> >>There I found something amazing for me (maybe not for you). >>From all nodes visited during search, >>quiet search takes more than 50% of all these nodes! >> >>Is this normal or is the program stupid? >> >>I don´t allow null-move in quiet search. In the test, I only produce beating >>moves, rochade and pawns conversions. >> >>I did the test with different positions. >> >>How much time do other programs spend in the quiet search? >> >>Werner > >I am surprised that you produce castling moves (rochade) in quies search. As far >as I know, this is quite unusual. Do you think that a position where a castling >move is possible is kind of unquiet ? Or what's the idea behind it ? > >Uli most programs evaluate threats to the king (kingsafety, bishops, rooks, ...). Therefore castling moves can change the evaluation drastically. Searching those moves in a quiescence search means that the programmer wants a positional quiescence, too. I read about it in a text from Reiner Feldmann (author of Zugzwang), which is on the www. Actually I use in Fortress only a material based quiescence search. It is designed to generate trees as small as possible. Alessandro
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.