Author: David Rasmussen
Date: 12:59:14 01/15/02
Go up one level in this thread
On January 14, 2002 at 10:47:42, Robert Hyatt wrote: > >I suppose the "ultimate" would be to say "I am at ply=X, so I can't allow more >than "X" extensions (total) at any point in the tree up to ply=X. > >However, I think that would likely explode badly. I think that extensions are >"connected" so that parts of the tree get none, and some parts extend like mad >due to exposed kings and so forth. I really don't want to extend parts of the >tree where nothing serious is going on... > >Will be interesting to see how your test goes... The "ultimate" example I >gave is probably too far out in left field. And in fact, I didn't even like >the "two extensions max per two consecutive plies" idea. But you might find >a way to make it work better than I did.. I implemented it now. It scored consistently better in WAC, with node counts appx. the same as with a one ply extension limit per ply. Sometimes a little more, sometimes a little less. I don't know why, but it doesn't seem to explode in any of the tests I have done. If it does at some point, I will work with it some more. Actually, I am a little disappointed in that it seems to extend less very long lines, than previously. I want to be able to detect long tactical lines. I don't know whether I implemented it wrong, but it seems that I never extend more than the nominal depth of my iteration, so when searching 8 plies, I extend to 16 plies. I could understand that if I had set the limit to the nominal search depth. But I haven't. I have set it to the current ply, in the search node. I would think that some lines extend to, say, 15 plies in an 8 ply search. That means that I have extended 7 ply. But at ply 15, I am allowed to extend 15 plies, so further extensions would be allowed. That's weird. /David
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.