Author: Francis Monkman
Date: 15:28:49 05/19/99
This is another excellent example of a 'simple' idea, that seems to prove hard for programs to find. r4rk1/1nqb1pbn/3p2pB/pppP3p/4P3/1PP2NNP/P1BQ2P1/R4RK1 w - - [Smejkal-Medina, Amsterdam II 1971] Looking at the position, the idea is, clear the b1-h7 diagonal, then piece for piece and the queen to g5 giving check etc. e5! dxe5 Nxh5 gxh5 Bxh7 Kxh7 Bxg7 Kxg7 Qg5+ (9-ply, and surely there's enough 'interest' here to grab an extension?) ...Kh7 Qxh5+ Kg7 (for obvious but misplaced reasons, programs prefer ...Kg8) Ng5 1-0 None of the engines I tried (Fritz, Rebel, Genius, CSTal II) found e5 within 5 mins. I left LGG 2.0 looking at it while I took a break, and it had found it after 16 mins. (So not to say the others won't.) And another from 1971, Smyslov-Adorjan, Amsterdam I r1b2k1r/pp2p2p/5pp1/1n1NB3/8/1P4P1/P1P4P/2KR3R w - - I'm sure I waited hours to see if Genius 3 would get this, when I first played through it. Also a simple idea -- White will exchange two pieces for a rook and two pawns (already a material advantage) after which, a human chess player should see at a glance that his two rooks' mobility, and the pressure they can put on Black's king and undeveloped queenside mean he can 'mop up' the remaining kingside pawns without too much trouble: Nxf6! exf6 Rd8+ Kg7 Bxf6+ Kxf6 Rxh8 Kg7 Rd8 Nc7 Re1 1-0 Again, none found it within the 5 minutes I allowed (you see I'm getting impatient!), except, yes you guessed, LGG 2.0 immediately plays this, and for 'all the right reasons'. I'm not trying to plug LGG too hard, really. I've watched it do some pretty terrible things too! But for some of these 'difficult' moves, it really does seem to have a consistent edge over the rest. Francis
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.