Author: Scott Gasch
Date: 17:47:14 09/06/04
Go up one level in this thread
On September 06, 2004 at 15:07:00, Stuart Cracraft wrote: >[D] 5n2/B3K3/2p2Np1/4k3/7P/3bN1P1/2Prn1P1/1q6 w - - 0 0 > >Anyone can get this besides Ed? > >Please post your search times. This is an interesting position. I saw it on Ed's "Programmer Stuff" page a while back and I read his explaination. To paraphrase, Ed usually only considers checks in the first few plies of the qsearch. However when a side in check only has one or two legal evasions he considers checks more deeply in that line. When I added the same kind of rule to my engine it solved this position instantly. However I found some degenerate cases where this rule caused the search tree to explode... there were tons of long check, reply, check, reply lines in qsearch. My solution was to keep the spirit of the idea but tune it back a little. The result? My engine no longer finds this mate in 30 in a reasonable amount of time. I always wondered how people who allow checks very deep in the qsearch avoid having it over-fire. It's been so long that I can't remember what position I found as a counter-example to this rule but it was an ECM test position and around #150 I think. Scott
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.