Author: Inmann Werner
Date: 14:01:58 02/15/00
Go up one level in this thread
On February 15, 2000 at 11:49:36, Andrew Dados wrote: >On February 15, 2000 at 03:49:30, Inmann Werner wrote: > >>On February 14, 2000 at 15:10:57, Andrew Dados wrote: >> >>> Let us consider a node at ply 0, where eval>alpha. We call qsearch, returned >>>score is < alpha. Now, at next iteration that position will also fail low >>>(assuming same alpha), because there will be no other moves for failing low side >>>then qsearch moves again... >>>So: we can store that score in HT with (depth + 1) instead of (depth) due to >>>above... or am I missing something? (I do that trick for a while, no side >>>effects found yet except for lower node count). >>> >>>(The above assumes same set of moves is being considered on all qsearch plys) >>> >>>-Andrew- >> >>Seems like an interesting thing, but I do not understand full, what you mean. > >>1) You mean distance 0 , not ply 0 or? (last normal search node?) >yes - last search node before qsearch > >>2) At distance 0, the eval>alpha comes from the evaluation function? >yes > >>3) After qsearch, the eval < alpha comes back? >yes > >>4) then you store at distance 0 in Hash wit 1 distance increased? >yes > >> >>???? >Somehow I managed to be vague in that post... > >Andrew I tried to implement it, and nothing at all happened. There was no position where this happened. After thinking about it... In my prog, the quiet starts with the eval of distance 0. The side is on move and does only "piece sacking" moves. If no one is good, the eval at distance 0 is my eval. If in Check, this search is no qsearch at all, but extended full. Maybe my Qsearch is a little small, but for me, Qsearch is only to find a quiet position....(makes it somehow fast) Hope I lost nothing, little bit complicated.... Werner
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.