Author: David Blackman
Date: 04:35:35 06/14/99
Go up one level in this thread
On June 14, 1999 at 00:32:46, Will Singleton wrote: >>On June 13, 1999 at 21:10:57, Pat King wrote: >> >>>Before implementing null move, I had a simple aspirated AB search, and would >>>research with a window of (Beta, Infinity) for fail highs and (-Inifinity Alpha) >>>for fail lows. And life was good. With null move, however, it's not uncommon >Of course, if you don't use fail-soft, then it never occurs. I think you can >get away with that by using pvs. > >Will I've seen it in programs that don't do fail-soft. It's too late at night for me to remember exactly why it happens, or analyse it properly, but i'm pretty sure fail-soft doesn't come into the equation. I suspect a big transposition table in combination with no fail-soft might hide the effect in some cases, but it would still actually be happening up the tree somewhere. All you need is a selective search mechanism that changes it's idea of "interesting moves" when the window changes. Actually it has allow some moves when the window is low that it rejects when the window is higher. Null move does that. In the null move case, this only matters when you have zugzwang (very rare except in certain endings) or where reducing the search depth by a few plies hides something that matters more than a tempo does (moderately common during bizarre tactics). So if it happens every single move, you probably have a bug, but if it's just occasionally, and most often during sharp tactics or endgames, then it's a known effect.
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.