Author: Geoff
Date: 06:08:33 04/26/03
Hi
I have been experimenting with a chess engine, very similar to TSCP 1.73 at the
moment. I came across this test position which should be a mate for white in 5
{D} 1b6/kPp5/p1P5/R5Rr/P1N1P3/8/6p1/6Kb w - - 0 1
Solution is 1.g5-c5 h5-g5 2.c5-f5 g5-h5 3.f5-g5 h5xg5 4.a5xg5 a6-a5 5.g5xa5#
This chess engine uses iterative deepening and a standard alpha Beta Search
using +/- infinity. According to my understanding it should spot the mate at Ply
10. An alpha beta search with a +/- infinity search window should not miss a
mate as the pruning is guarenteed to be perfect ?
Is my understanding wrong here or must this be a bug ?
The thing that really puzzled me is that I tried version 1.81 where as far as I
can see the eval and search code is the same, but V 1.81 spots the mate at ply
10.
Thanks for any thoughts on this
regards Geoff
1.73 Misses the checkmate ??
ply nodes score pv
1 28 980 g5h5
2 95 115 g5g2 h1g2
3 676 640 g5g2 h5a5 c4a5
4 2032 576 a5a6 a7a6 g5h5 b8a7
5 11687 960 c4d6 c7d6 g5h5 a7b6 h5h1
6 41864 571 g5e5 h5g5 a5a6 a7a6 e5g5 b8a7
7 306957 880 g5f5 h5g5 c4d6 c7d6 f5g5 a7b6 g5g2
8 932164 576 g5f5 h5g5 f5e5 g5h5 a5a6 a7a6 e5h5 b8a7
9 7280073 786 g5f5 h5g5 a5a6 a7a6 f5g5 b8a7 g1h2 g2g1q g5g1
10 14031564 845 g5c5 h5g5 e4e5 g5e5 c4e5 a7b6 c5d5 b8a7 a5c5 a7b8
e4e5 is the wrong move !!! grrrrrrr
1.81 Spots the checkmate
ply nodes score pv
1 28 980 g5h5
2 95 115 g5g2 h1g2
3 676 640 g5g2 h5a5 c4a5
4 2032 576 a5a6 a7a6 g5h5 b8a7
5 11687 960 c4d6 c7d6 g5h5 a7b6 h5h1
6 41876 571 g5e5 h5g5 a5a6 a7a6 e5g5 b8a7
7 307185 880 g5f5 h5g5 c4d6 c7d6 f5g5 a7b6 g5g2
8 935877 576 g5f5 h5g5 f5e5 g5h5 a5a6 a7a6 e5h5 b8a7
9 7330801 786 g5f5 h5g5 a5a6 a7a6 f5g5 b8a7 g1h2 g2g1q g5g1
10 13846050 9991 g5c5 h5g5 c5f5 g5h5 f5g5 h5g5 a5g5 a6a5 g5a5
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.