Author: Robert Hyatt
Date: 06:41:48 07/21/99
Go up one level in this thread
On July 21, 1999 at 01:59:17, Vincent Diepeveen wrote: >On July 21, 1999 at 01:22:41, Robert Hyatt wrote: > >>On July 20, 1999 at 18:10:59, José de Jesús García Ruvalcaba wrote: >> >>> Could somebody please explain in simple (but accurate) words the essentials of >>>conspiracy numbers search? >>> I know I can read the research papers from the Zugzwang team, but first I need >>>something less dense. >>>José. >> >> >>The basic idea is to detect deep forcing lines. The idea is this. If you >>reach a position P in the tree, and from that position you search a set of >>moves leading to new positions... how many of those successor moves actually >>affect the current score? >> >>IE if the score is a result of only one move, then we consider this a forcing >>position where there is no real chance to vary... The term 'conspiracy' comes >>from the idea of "how many positions below this position have to 'conspire' >>together to change the score at this position?" If changing the score of one >>move changes this position, it is forcing. If changing the score of many moves >>doesn't affect this one, it is a 'quiet' position... >> >>There are other details, but that is the gist of why it might be useful... >>Because at every node where the conspiracy number is low (only a few scores >>have to change to change this position) we can search deeper since this is a >>forcing position. In positions where it takes a lot of changes below this >>node to change the score at this node, we don't extend because it is not a >>forcing position. >> >>When you think about it, it is a form of singular extension, a form of check >>extension, a form of one-reply extension, etc... > >Not at all. It sees more than that. It sees variations very >*little* possibilities, which SE never would detect. "It is a form of" does not mean "it is identical to". CNS will definitely extend a move that SE would extend. It can also extend lines SE would not extend, _if_ you can afford the slow-down.
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.