Author: Tord Romstad
Date: 07:38:23 07/04/04
Go up one level in this thread
On July 04, 2004 at 10:21:24, Mathieu Pagé wrote: >On July 04, 2004 at 10:10:19, Tord Romstad wrote: > >>On July 04, 2004 at 09:46:31, Mathieu Pagé wrote: >> >>>Steffen is right I have no qsearch so I think that the flipping on the result is >>>normal. To avoid re-search problem in the real version on my engines I use a >>>iterative deepening of +2 ply. It work better than +1. >> >>Hi Mathieu, >> >>I suggest that you stop worrying about the branching factor until you have >>implemented a qsearch() function and a slightly less rudimentary evaluation >>function. A material-only eval without a qsearch will result in all sorts >>of weird search results, and I don't think you can make any interesting >>conclusions by observing the branching factor. >> >>Tord > >Hi Tord, > >You are probably right, I will follow your suggesstion. I just wanted to be sure >there was not a bug in my search. In general, this is the right approach. Keep everything as simple as possible until you are reasonably sure you don't have any serious bugs. However, a fixed-depth alpha beta search simply doesn't work well without a qsearch or a sophisticated swapoff function (which is much harder to write). Writing a basic qsearch function is rather easy and doesn't add a lot of complexity to your code. Tord
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.