Author: David Dahlem
Date: 08:48:35 09/08/04
Go up one level in this thread
On September 08, 2004 at 11:00:00, Alessandro Scotti wrote: >On September 08, 2004 at 08:49:33, David Dahlem wrote: > >>I tried several of the mate in 33 positions using my table method, and in all >>cases, it was able to mate against tablebases within the 50 move rule, usually >>in 46-48 moves. > >I have tried several of them too but then I found a "mate in 30" position that >broke the algo. That's why I don't feel confident enough with "manual" tests. A >position that is a longer mate for ETGB may not be the most difficult for an >algorithm and fixing it even with a small change invalidates all previous tests. >Besides, within the path to mate, the losing kings has often different moves to >make and if we don't try them all we're just testing one single path. (If OTOH >we test all branches then eventually we should also come to the "mate in 30" >position above... correct?) >That's why I was thinking of automatic tuning/testing... or using ETGBs, which >at this point would be like admitting defeat though... This position is the only one i have found so far that defeats my table arrays. 7k/3n4/b7/8/8/8/8/4K3 b - id "Mate in 28"; I'm currently testing and trying to tweak the array values. If most of the positions are solved against tablebases, then it should do ok against non-tablebase engines. :-) Regards Dave
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.