Computer Chess Club Archives


Search

Terms

Messages

Subject: Evaluation feedback to q-search

Author: Hugh Cumper

Date: 14:35:49 08/24/00


In some positions a turbulent position may not be detected by the usual q-search
criteria, basically captures and checks, but is detected by the evaluation
function. In such cases why not have a feedback loop from evaluation to search?
For instance in Crafty's evaluation function there is a test for a trapped
bishop, if white has a bishop on a7 and black pawns on b7 or b6 and c7 then the
bishop is deemed trapped and a fixed penalty incurred. In reality the severity
of the trapped bishop problem depends greatly on other aspects of the position.
Why not have the evaluation function report back that the position is not
quiescent after all and further searching is required?



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.