Author: Dann Corbit
Date: 13:58:17 06/15/04
Go up one level in this thread
On June 15, 2004 at 16:54:33, David Dahlem wrote: >On June 15, 2004 at 16:31:22, Dann Corbit wrote: > >>On June 15, 2004 at 16:01:50, David Dahlem wrote: >> >>>On June 15, 2004 at 15:56:27, Dann Corbit wrote: >>> >>>>On June 15, 2004 at 15:45:08, David Dahlem wrote: >>>> >>>>>On June 15, 2004 at 15:33:41, David Dahlem wrote: >>>>> >>>>>>One of the problems with the current method of testing engines with test suites >>>>>>(e.g. WM-Test) is the problem of proving that the proposed solution move is >>>>>>actually the best move, especially with positions of a positional nature. >>>>>>Perhaps a new method would avoid this problem, namely a suite of mate positions, >>>>>>with known, more easily proven solutions? Time to solution could be the criteria >>>>>>by which engines are evaluated. >>>>>> >>>>>>Just an idea. Any thoughts? Would this work? >>>>>> >>>>>>Regards >>>>>>Dave >>>>> >>>>>This method would also avoid the problem of engines finding the solution for the >>>>>wrong reason. :-) >>>> >>>>The engine that performs best at these problems will be an engine like Chest, >>>>which cannot play chess but can solve mates. >>>> >>>>When the game is in doubt, is when almost all of the moves are played. If we >>>>tune for the very end, the engines will play well in that phase. But it is a >>>>tiny fraction of the game. >>>> >>>>There are many checkmate test suites. >>>>BWTC springs to mind, as an example. >>> >>>Not all checkmate problems are from endgames, many checkmates occur in the >>>middle game, even in openings. :-) >> >>The ones that occur in the openings are only going to happen to the weakest >>players. But they might make an interesting suite anyway. >> >>I was a bit surprised to see that there are 5363 certain mates within the first >>5 plies of the game of chess (ignoring transpositions). >>ftp://cap.connx.com/pub/chess-engines/new-approach/qmates.epd > >These are all "mate in 3" or less, certainly too easy for chess engines. >Something like "mate in 7" or more would be a more valid test for chess engines. It isn't really intended as a test suite. It is just a list of mates that are close to the origin. It might be useful for the raw beginner, but certainly not for Shredder and the like. There may be more distant mates in the first 5 plies, but it will take a very long time to find them.
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.