Author: Vincent Lejeune
Date: 06:13:30 04/25/02
Go up one level in this thread
On April 25, 2002 at 08:39:53, Vincent Lejeune wrote: >On April 25, 2002 at 08:14:25, Otello Gnaramori wrote: > >>On April 25, 2002 at 07:21:15, Sune Fischer wrote: >> >>>When/if computers learn to recognize these "trivial" drawn positions, then I >>>think they will become virtually unbeatable by humans. >> >>I don't think it will be a "mission impossible" task to implement that >>particular function (grab the never concept in fortress like positions) with a >>major effort from programmers and researchers in chess computer area. >>I think that to solve that problem will request a reasonable high amount of work >>by the "authorized personnel". > >This concept is a long termne planning it's a revolution in computer chess >because : > 1) it requiers a big amount of knowledge (calculate entry point in the >ennemy field of the king and other piece(s) > 2) it requiers a big amount of ressource (many things to view for one >position, may be fortress-hash-table possible ??? ) > >I think such evaluation would be possible only in restricted circumstances, e.g. >: endgame at low depth (below 7 or 8 plies) > Here's a picture of board the after move 51.Nc3 '1' is a square "control" by white and white can't force white to free the control of this square 'B' is a square where a black piece is blocked (very often as here : a pawn) :ABCDEFGH -+-------- 8: 7: 6: 5:11111B1 4: 1 1B 3:1 1 2:1 1 1: 1 1 So you can see there's a wall : a5,b5,c5,d5,e5,f5,g5,g4,h4 This 9 squares are sufficient to close definitively the position, and one condition needed to be a fortress : the white side can make moves without breaking the wall : here Kg1-f1-g1-f1 (or f2 after black play g3) One day there will be such knowledge in our chessprogram ! > > > >> >>w.b.r. >>Otello
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.