Author: Robert Hyatt
Date: 08:57:15 05/27/02
Go up one level in this thread
On May 27, 2002 at 04:45:41, Sune Fischer wrote: >Currently I have a search depth limit of 60 plies. >However in this very simple (drawn) position my engine has no >problem reaching 60 plies in no time. > >[D]8/8/p1K5/P4k2/8/8/8/8 b - - 0 1 > >Actually it crashes when it goes beyond 60 plies in a search, has to do with the >PV array that has a limited length (and probably other stuff too). > >The reason it can reach this deep is probably because the position is solved as >drawn rather quick (and because of transpositions), but unlike a mate score that >I can detect, I can't *detect* a draw score in the same way. >Is there a way to now if an exact draw score has been returned? > >My outer iteration loop stops of course at the 60 plies, but the search is still >called and on top of that there is the qsearch. >I measured up to 71 plies reached, naturally it crashes. > >How do I end the search properly when these extreme deeps are reached? >Should I just stop iterating at ply 45, that would give the search 15 plies of >safety margin? > >Does other engines too have a maximum depth limit, if so how did you solve this >problem? > >Thanks, >-S. One thing you _must_ do is have a max depth cutoff in the Search() procedure. IE look at Crafty's check against "MAXDEPTH". If you don't do this, you can let extensions drive you beyond array dimensions in rare positions, and that will produce bad results (if not an outright crash). I limit my iteration max in Crafty to 60 plies, and I limit the max search depth to 60 plies as well, since arrays have finite dimensions.
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.