Author: Alexander Davies
Date: 16:15:59 11/22/00
Go up one level in this thread
On November 22, 2000 at 13:54:03, Dann Corbit wrote: >Any move leading directly to a sure checkmate is exactly as good as any other. >Computers do not care how long it takes. Very few chess playing programs find >the shortest mate routinely. Only chess solvers like Chest or Problemiste do >that. > >Any move that leads directly to a verified checkmate is a best move. A move that leads to the shortest mate is generally better than one that leads to a longer mate. If there is a human operator, there is a serious danger of the computer losing on time if the program plays into long winning sequence rather than playing a quick forced mate. Even if the game is played without a human operator, playing moves that lead to long mates can still be a problem. If Black gives queen odds, then 1.e4 e5 2.Ba6?? is still a bad move. If a program always plays a forced mate sequence that is two ply longer than the shortest available, then it will never win against best play. For example, in Dann Corbit's post "WAC revisted (again)", he wrote: [D]1r3r1k/3p4/1p1Nn1R1/4Pp1q/pP3P1p/P7/5Q1P/6RK w - - bm Qe2; id "WAC.243"; ... "It seems that Qe2 is best, but I wonder how b5 would turn out. Might it also be a winning move?" 1.b5, while winning, doesn't really do anything. The PV is "Pb4-b5 Rb8-d8 Qf2-e2", which confirms this opinion. It has happened that programs screw up completely winning positions by playing such "second-best" moves. The program sees that a good move or position can always be delayed until the next move. However, something like the 50-move rule can come up, forcing the program to sac material to stay in the game. I'm sure someone like Bob Hyatt can remember such games from computer tournaments. One day I intend to write a program (maybe in Scheme, for weirdo value) and if it plays moves like 1.b5 without ever playing the killer move then I'll be very frustrated!! Alex Davies FIDE 2336
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.