Author: Anthony Cozzie
Date: 10:51:28 02/21/04
Go up one level in this thread
On February 21, 2004 at 08:44:45, Michel Langeveld wrote: >On February 21, 2004 at 08:37:50, Uri Blass wrote: > >>On February 21, 2004 at 08:23:59, Andrew Wagner wrote: >> >>>The position is [D]r2q1rk1/ppp4p/2n2b2/3bppp1/Q7/P2PBNP1/1P2PPBP/R4RK1 w - - 0 1 >>> >>>The idea of course is to find Bxg5 Bxg5 Nxg5 Bxg2 (Qxg5 Bxd5) Qh4!(threatening >>>mate while defending the knight) 1-0 >>> >>>and I'm sure most engines would find that very easily. The question is a.) Is >>>this a move quiescence should find, and how? and b.) How do you keep from >>>pruning this move via SEE or MVVLA if you have them implemented? >> >>This move is not a move that qsearch should find because qsearch does not search >>nothing like that. >> >>Qh4 is not a capture or check and qsearch has not enough depth to find threats >>of mate in 1. >> >>Even after Qh4 programs should not evaluate the position as better for white by >>qsearch because black is a piece up for a pawn and the black king is not in >>check so by stand pat programs return beta and programs need to search at least >>one ply more in the main search to find the mate(it may be even more than one >>ply more because after Bxf1 your program may prune Qxh7# as a capture that is >>not good enough to get the score above beta and even after Bxf1 Qxh7# if your >>program does not evaluate mates it may not see the mate and need another ply to >>see that the king is captured). >> >>Note that in the case of movei I think that one ply after Qh4 is enough because >>it prunes no captures in the first ply of the qsearch and finding if the >>position is mate is part of the evaluation. >> >>Uri > >I spoke with Andrew about this in ICC and said something similiair. >Not a normal task for QSearch to pick this up. > >But on the other hand if you think about it if an engine is capable to find Qh4 >in a cheap way then this engine is in on the right path to become a tactical >monster. > >[D]r2q1rk1/ppp4p/2n5/4ppN1/Q7/P2P2P1/1P2PPbP/R4RK1 w - - >In this position a queen on h7 is mate. All moves that makes it possible to get >a queen on h7 are interesting. With this reasoning Qh4 can be found. I don't think *anyone* does _mate threats_ in qsearch, let alone mate threats at all depths in qsearch. This is an 8 ply combination for Zappa. Seeing tactics of this magnitude is simply not the job for the q-search. If you bulked up your q-search enough to find this, you would get 6 ply brute-force at 40/2. anthony
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.