Author: Howard Exner
Date: 00:29:04 08/15/98
Go up one level in this thread
On August 14, 1998 at 02:39:43, Frank Schneider wrote: >Hi all! > >There are many testsuites where a best move has to be found and >I'm satisfied with the results Gromit gets. However, when playing >real games most losses are caused by making bad moves (and not by not >playing good moves). > >Is there a testsuite with positions where programs have find out that >a good-looking move is bad? > >Is there a testsuite where the best move does not win but is the only >move to assure a draw? > >Frank Here are a few positions where moves that appear good should be avoided. In every case playing the am (avoid move) will result in either missing a win or turning a draw into a loss. 8/7p/R7/7k/8/p5K1/8/r7 b - - am a3a2; Q6n/8/8/8/8/6K1/pk6/8 w - - am Qxh8; 8/1Q5n/8/8/8/2p5/7K/2k5 w - - am Qxh7; 1kb5/8/p1p3p1/Pp1p1p1p/1P1P3P/5PP1/2PN4/5K2 w - - am=f3f4; 8/5p2/p3p1p1/1pbkP2p/5P1P/1P4P1/P3KB2/8 w - - am Bxc5; 1r6/5k2/p4p1K/5R2/7P/8/6P1/8 w - - am Kh5; 8/6p1/1p5p/1R2k3/4p3/1P2P3/1K4PP/3r4 b - - am Rd5; 8/8/6p1/3Pkp2/4P3/2K5/6P1/n7 w - - am e4f5; bm d6; 8/8/6p1/3PkP2/8/2K5/6P1/n7 b - - am=g6f5; bm Kxf5; The first position will be the hardest for computers. The winning plan is to move the black king towards the a6 pawn. The next two are routine draws for those familiar with the draws of the rook and bishop pawns. I believe the ChessMaster programs avoid capturing the Knight in these two examples immediately. So de Koning must have included code for these. On todays faster hardware many programs will probably just calculate a win by avoiding the knight captures even without the knowledge code. The last two problems I now give Rebel 8 a tough time. Try number nine before number 8 as number 9 is one ply after.
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.