Author: Tony Werten
Date: 22:57:25 05/30/05
Go up one level in this thread
On May 30, 2005 at 14:30:20, Dieter Buerssner wrote: >On May 29, 2005 at 12:44:10, Gian-Carlo Pascutto wrote: > >>You can just do the SE searches after the nullmove, no problem there? > >This is an interesting point. >I think, the biggest disadvantage of nullmove in general is not the Zugzwang >problem, but hiding tactics. My favorite example (without a really concrete >position) is a situation with bKg8, bPf7, g6, h7. White has a pawn or a bishop >on f6. Assume white has to do some queen walk, say qd3-qh3-qh6-qg7#. After each >queen move, black null move can fail high; certainly it can faĆl high more than >once during such a walk. This would eat something like 6 plies (or more) from >your depth, making rather easy tactics very deep. From SE, one would hope, that >it helps here. When you delay it after the null move, it will help zero in such >a situation. Correct. But by delaying the singularity test after the nullmove, you reduce the amount of work significantly. When we're on a fail high node, and nullmove fails low, we know that we actually need to do a move to fail high. Now you only have to test how many of those there are. The only assumption that is taken is that when a position fails high on a nullmove there will be more than 1 move that will also fail high. This is only a bit stronger than normal nullmove, wich assumes that on a nullmove failhigh there will be at least 1 move that will also fail high. BTW I have never tested it but a conformation search after nullmove (wich could do a singularity test) might help here. Then again, it will cost more as well. Tony > >Regards, >Dieter
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.