Author: Dave Gomboc
Date: 11:52:39 07/20/99
Go up one level in this thread
On July 20, 1999 at 14:04:29, Dan Andersson wrote: >A random function that inserts extensions on certain moves without any chess or >depth related information and still have a maximum, the function could quite >possibly be dependant on its previous states. For example, a no brainer: q is a >random variable [0..1], q<r P(r)=1 else P=0.5 where depth=depth+P(0.5) up to >maxply in an alpha beta search. The search will have the insurance level of >maxply, a mean level of 1.5 maxply and a maximum level of 2 maxply. The thing is >that varying the random function will get you some interesting properties, >sometimes a smoother transition between iterations or moves (especially if you >save hashtable), other times times tactical deathstrokes even against >faster-deeper searchers (that often in conjunction with risky use of insurance >level, i.e. assuming a mean-depth search or some such) but most often the >function chosen will just slow down the search and kill seach dead. > >Regards Dan Hehehe, well if you find the random extension functions that don't often slow down the search and kill it dead, let us know. :-) Dave
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.