Author: Michael Neish
Date: 21:35:03 01/06/00
Hi, I asked a question recently on move ordering, which Bob Hyatt was kind enough to answer. In short, my progam needs about a million nodes to reach ply 6 in a certain position, whereas other very strong programs such as Crafty and HIARCS can reach the same depth in less than 100,000 nodes. At the moment I'm not using any enhancements such as hash tables (though these will come soon). Bob, I guess I'm addressing this corollary to you, although of course this is an open discussion. I did a bit of detective work and I found that in the position I'm looking at, about 75% of the nodes searched are captures in the quiescence search routine. So if there was any way to limit these then the overall branching ratio would be improved quite a bit. You mentioned the Swap() routine in Crafty which simply trades off pieces attacking a particular square. Can this be used to good effect in the quiescence search routine? If so, how do you handle the fact that in a sequence of captures one capture may occur on one square and the next one on another square? I don't think the actual position is relevant to this discussion, but I'll be happy to provide it on request. Cheers, Mike.
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.