Author: Paul Richards
Date: 17:42:59 05/05/99
Go up one level in this thread
On May 05, 1999 at 19:29:47, Francis Monkman wrote: > >On May 05, 1999 at 18:55:03, Paul Richards wrote: > >>Strange thing though, as is so often the case, once you supply >>the correct move, it sees the continuations immediately at a much >>shallower depth. So then why the hell doesn't it see it by 15 ply? >>It really does seem as though correspondence analysis would find >>these thing faster. In other words, at 15 ply it doesn't see f5 >>as best, yet as soon as you supply f5, it sees it winning within >>10 or 11 ply. It doesn't get confused by any drawing chances >>either, it easily finds a clear win. > >Exactly!! I've been noticing this 'phenomenon' for years, had new >examples in the last few days. Haven't we all? Is there a fault in the >search/pruning algorithm(s)? There usually seems to be a drawing possibility somewhere in the search that throws a monkey wrench into the works. Crafty 16.6 does particularly badly with this position. You can give it several moves of the solution and it still doesn't see past drawing possibilities. I would love to see a technical explanation for this sort of thing. These solutions are well within the search horizon, but yet they don't get found. In any case if this sort of thing could be addressed effectively the programs would be much stronger.
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.