Author: Uri Blass
Date: 07:56:19 09/29/01
Go up one level in this thread
On September 29, 2001 at 09:38:53, Georg v. Zimmermann wrote: >Hi, > >it is not a good idea to invent new meanings for words that are clearly defined. > >quiesce search and normal search are 2 very different things in most chess >programs. > >normal search is terminated when the depth + extensions is used up. > >quiesce search is terminated when there are no more good captures( programs use >different way to figure what is good ), promotions and checks for some. I understand. I knew that the search at depth 0 is called qsearch but I did not know that it is not considered as extensions by programmers. I tried to explain the problem to somebody who does not know much about computer chess so I used the word extensions and I think that this is the way that non programmers understand the word extensions. > >>These case may not happen in real games but I can imagine that cases when the >>program waste too much time on the qsearch can happen(it may be only 0.01 >>seconds for one qsearch but if you add all these times then it may be a problem) >> >>Uri > >I think what you are trying to say is that this is an extreme example of what >might happen in some normal positions. > >Well I am sure if you show Dr.Hyatt a position from a real game where Crafty >fails to reach the correct decission because it doesn't search deep enough and >the quiesce nodes searched are very very high he will consider limiting the >qsearch. > >Georg I do not know but it is my intuition. It may be interesting to see some statistics about the number of nodes and the number of plies in the qsearch. I suspect that the strategy to stop the qsearch after enough positions may be better. A possible idea is to limit the number of plies in the qsearch(it is possible to use a fixed number or to use also the number of positions that was searched in the qsearch). It is possible to use test positions to see if limiting the qsearch helps Crafty to solve test position faster. I do not expect big improvement but I expect at least a small improvement and my guess is that the possible average improvement is between 1% and 10% speed improvement Uri
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.