Author: Gerd Isenberg
Date: 15:43:01 01/30/02
Go up one level in this thread
On January 30, 2002 at 17:02:28, Miguel A. Ballicora wrote: >>I do not know what a confirmatory search is (I guess I can imagine, something >like YACE) but I think of 3 possible ways: >1) reduce depth, rather than pruning (Gaviota) >2) double nullmove (Diep) >3) apply only nullmove when depth is < x plies. x could be something like 7 etc. >This one I think is Crafty, according to some posts in the past (IIRC). >When you search so deep, there is one point where moves close to the root are >not nullmoved anymore. Solution to nullmove problems are just delayed. > >Regards, >Miguel > To 3) x is constantly 2. When i get a nullmove fail high and some other depth related conditions met, i use a reduced search to confirm the fail high or to detect zugzwang. When this confirmation search (same depth reduction as nullmove search: depth-R-1, but nice to play with another R) goes less beta, doing nothing is the best. Zugzwang is detected and the "normal" search continues, instead of pruning. Zugzwang detection is also stored in hash tables and may disable nullmove in further iterations. I imagine, that it's most similar with Vincent's double nullmove. But the control flow and the search depth of the second nullmove search are different. Gerd
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.