Author: Gerd Isenberg
Date: 11:22:37 11/21/03
Go up one level in this thread
On November 21, 2003 at 12:11:04, Georg v. Zimmermann wrote: >Hi, > >the following is a position I found in one of Sunsetter's search tree leef >nodes. > >[D]r2qkb1r/ppp1pppp/2n5/4pb2/Q1P5/N7/PP3PPP/R1B1KBNR w KQkq - 0 9 > > >Of course it sees that black is lost, but due to the eval() still being bad >beyond reason, even though I spend a couple of days on it now, it doesnt see >that its already 1-0. > >One of the reasons for that is that it thinks the Qa4 of white is worse than the >Qd8 of black, which is of course absurd. White queen is active, can not be >attacked. Black queen is passive, and doesnt know where to go after Rd1 soon. >Right now I just use a piece square table for the queen with some big penalties >to avoid it move around with the queen too much in early opening. > Hi Georg, my "too early" queen term makes the otherwise better white queen about equal with the black one. The "too early" penalty is dependent on the number of own and opponent knights and bishops on their initial squares and their ability to move. There is also the pinned knight and the black kings bishop, which compensate the "too early" queen a bit. Another idea is to amplify the material difference (specially with pawns vs. piece) if positional evaluation score of the lower material side is worse some threshold too. >How does one achieve both ? >- avoid too much queen movement in the early middlegame >- not make it think the white queen here is bad > > >Georg Queen movement without real threats should not happen since other pieces are developed. So it's a matter of tuning and weights. Gerd
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.