Author: Bruce Moreland
Date: 00:26:16 11/27/00
Go up one level in this thread
On November 27, 2000 at 02:50:03, Tony Werten wrote: >On November 26, 2000 at 11:13:31, Severi Salminen wrote: > >>Hi! >> >>I forgor to mention that I use the "stand pat" beta cutoffs in qsearch: >> >>if(static_eval>=beta) >> return static_eval); >> >>I'd like to know should I test any mate or stalemate situations before returning >>(in those cases the score is totally wrong). Or do I allways make the assumption >>that there is at least one legal non capture move available (which is not >>obviously tested)? >> >>And is there any idea to try nullmoves in qsearch? Has anyone tried that and >>what was the result? > >If you only search a few ply, it's usefull. But for nowadays engines it only >blows up the search. The article from advances in computerchess 5 is outdated. It always blows up if you don't try to stop it from blowing up. When people start building their first chess program, there seems to be a specific set of steps they go through. They get move-generator crazy for a while, then they start asking questions about how many positions can be reached from a 5-ply search from the root. Once they get going they ask about search explosions and get discouraged because they realize that another program is effectively 35 times faster than theirs is. That is the time of the "oh my god I am really stupid" realization that is so healthy for every programmer. If people survive beyond this initial realization, they realize that you have to do something about qsearch or you get stuck in ply 4 or so. I had these problems five years ago. The version of my program that finished 3rd in the 1995 WMCCC in Paderborn did something like 80% quiescent nodes. bruce
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.