Author: Andrew Williams
Date: 15:49:47 08/18/99
Go up one level in this thread
On August 18, 1999 at 17:57:01, Scott Gasch wrote: >Hi there, > >By converting from standard negamax to a negascout search I made a dent in the >move branching factor and the program is searching deeper. I also got razoring >working. > >However razoring is kinda stupid without extensions and whenever I add >extensions to the search routine the program slows _way_ down. I think the most >common extension is a check extension that is what I am implementing now. So if >side to move is in check I am calling the recursive search routine with the >depth unadjusted... This is 1000% slower than no check extensions :( > >So I tried a thing that I don't understand completely but I see Crafty do and >have heard about before -- now my depth for one ply is 16 and I try fractional >extensions for check. This still slows the program down a lot. > >I originally thought this might be because of repeated positions and branches of >the tree going back and forth until the 50 move draw rule cut them off. However >I just added a draw by repetition check that seems to work and still when I turn >on extensions it is s l o w. > >Does anyone have any thoughts about the pitfalls of extension that I might be >falling into? > >Thanks again for the help. > >Scott Hi Scott, Can you give an example position, together with your "before" and "after" search info? This helps a lot in diagnosis and helps promote discussion. Cheers Andrew
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.