Author: Uri Blass
Date: 09:31:51 12/30/01
Go up one level in this thread
On December 30, 2001 at 11:58:02, Severi Salminen wrote: >Well, I have to say that now I have nothing to worry about. If I exclude horizon >nodes from qnodes and include them to normal nodes, it looks like my engine is >working very properly. Now the ratio qnodes/all_nodes is 21.8% which should be >enough. Again, it is only a matter of definition. Was it Sune who also reported >80% of nodes belong to qsearch, this might be the reason. probably it was not the case with sune because he also reported that his program with limited qsearch could see 1 ply deeper. It was not the case with his program. I still believe that limiting the qsearch may be productive but I agree that deciding to stop it after x plies is not the best idea. I do not want the qsearch to explode in rare cases and I believe that there must be a way to do it with no demage in the regular case and allowing less captures in later plies may be better. I can add that my result for depth 9 is 18,340,855 qnodes out of 157,635,152 nodes main line 1.exd5 Qxd5 2.Ne2 exd4 3.Nbc3 Qe6 4.Qxd4 Nc6 5.Qe3 I decided to stop it and not to search for depth 10 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.