Author: Uri Blass
Date: 01:33:02 12/30/01
Go up one level in this thread
On December 30, 2001 at 04:08:33, Severi Salminen wrote: >>It means that you increase qnodes even if the qsearch does not find relevant >>nodes to search so there is a mistake. > >Again we can see how NPS means nothing between different programs :) > >Shouldn't this be true: > >node=position in the search tree >qnode=in my program every node seen from qsearch() is a qnode. So every node >from which we would generate only captures and promotions (and whatever more) is >a qnode. >move=...a move :) > >So if I do a 1 ply search from initial position, I visit 1+20 nodes=21 nodes. >The 1 is the initial position and there are 20 legal moves from there. If I do a >2 ply search then the number is 1+20+400=421 nodes. If we do a 1 ply search and >assume (not true in initial position of course) all those 400 nodes at ply=2 are >captures. Then I should have 21 nodes and 400 qnodes even if I only return >statical evaluation from those nodes. > >I think you are counting something different. So you don't increase node counter >if you return eval>=beta in qsearch before generating any moves? IMHO that is >not right as you really visited the node. You are now counting moves, not nodes. >Could you try the other way so we could compare more easily? Put nodes++ after >calling qsearch() and qnodes++ in the beginning of qsearch() and print them. It >would be interesting. > >Severi I do not see the point of it. The point is to see the number of nodes that I can save if I have the same order of moves and could use SEE to get the same result as qsearch. The nodes in the beginning of the qsearch are not nodes that I could avoid by using SEE. Uri
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.