Author: Robert Hyatt
Date: 10:51:03 06/02/99
Go up one level in this thread
On June 02, 1999 at 12:45:58, Pat King wrote: >On June 01, 1999 at 19:49:12, James Swafford wrote: > >>On June 01, 1999 at 18:14:27, Pat King wrote: >> >>>I've implemented a pretty generic aspirated AB with check extension and >>>quiescent search. Of course, with the extensions come some variability in the >>>node count for a given fixed depth search. But I'm finding swings of a factor of >>>10 in nodes searched from move to move! Is this unusual? When I get my !@#^@#%$% >>>hash working, will the improved move ordering help this any? >>> >>>All input appreciated! >>> >>>pat >> >> >>If you have some liberal extensions, coupled with a quiescence >>search that can continue as long as there are captures (or checks?) >>to be made, then it makes sense to me that the node counts for >>tactical positions will much, much higher than for positions that >>are quieter in nature. >> >>Does the pv show a long line of capture moves (after the full >>width search)? >It's just a straight AB, no pv, but yes, the longer searches are in the more >complicated positions. I was just taken aback by the range. I had hoped to >establish some reasonable upper and lower bounds to help in time management. This is normal. In a fairly long standard time control game yesterday, I was watching Crafty do generally 12-13 ply searches... until it managed to expose the opponent's king. Then the depth dropped to 10 plies or so as the check/one-reply/etc extensions kicked in. trying to predict the 'depth' is nearly impossible if you have extensions, as it is hard to predict when they will get activated and when they won't, until kings get exposed...
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.