Author: Dann Corbit
Date: 10:51:42 11/22/00
Go up one level in this thread
On November 22, 2000 at 00:51:55, Sam Slutzky wrote: >On November 21, 2000 at 03:38:39, Dann Corbit wrote: >> >>In any case, b7 must be added. Any move that provably leads to checkmate is >>just as good or better than any other choice. >> >>The EPD test suite must be amended. If your program chooses b7, that is a >>correct choice. > >I'm not entirely in agreement on this. It would seem to be that the shortest >mate is actually the proper solution to the problem although, of course, any >mate will win. Likewise, in a mate-in-x problem, if a program doesn't find >mate, but still plays a clearly winning move, it has not really solved the >problem although it is certain to win. In an extreme case (as in some tablebase >cases), a longer mate may actually be a draw due to the fifty-move rule. It's >all semantics really, but I don't see that everyone needs to update their test >suites in these cases. > Most chess programs that are not also mate solvers do not find the shortest mate on a huge number of chess problems. Any move that leads directly and certainly to a certain checkmate is a best move, by any sense of the word "best" in my opinion. I suspect that with one hundred checkmate positions chosen at random found by chess playing programs, Chest by Heiner Marxen will find shorter mates with surprising frequency. I have used Chest to find shorter mates in many, many instances in the CAP database.
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.