Author: Robert Hyatt
Date: 19:28:48 07/04/02
Go up one level in this thread
On July 04, 2002 at 19:15:08, Bas Hamstra wrote: >>You are making a classic mistake that nobody directly involved in AI >>research would make: Selective search is _not_ better than a full-width >>search with good search extension rules. Selective search and a full-width >>search with good extension rules are absolutely identical. I don't see >>why anybody claims that if they don't do a selective search, they can't be >>as good as someone that does. That is simply _wrong_. > >I have seen this before, but I am not convinced this is true. Let's take the >same program with and without nullmove. One has a branching factor of around 4.5 >and the other around 2.5. The one with without nullmove sees tactically more, so >we must somehow compensate with extensions. Which will make the BF difference >even worse. Somehow there is something wrong here. > >Are you saying pruning >= Beta nullmove-nodes is the same as extending < Beta >nodes? Same search behaviour, an any time control? > >Best regards, >Bas. What is null-move doing? Simply reducing the search depth along pathways where the depth would be wasted. What does selective extensions do? Extend the depth along pathways where the depth is important. Yes, the two will produce _identical_ results... If they are done with the goal of producing identical results...
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.