Author: Tord Romstad
Date: 06:10:10 02/11/06
Go up one level in this thread
On February 11, 2006 at 09:04:39, Joachim Rang wrote: >the above mate is indeed incorrect. However as Maarten has pointed out the >initial position which you posted is indeed a mate in 14 (checked with a couple >of engines in backward analysis). In the second position after Nf6+ gxf6 it is a >mate in 6, the question is, why Glaurung prunes away the better answer Kf8? Yes, that is the question. The bug is clearly related to the parallel search. When using only one thread, Glaurung's output for this position is correct: 2 +34 25 134 Rd3 Qb2 3 +40 25 605 Rd3 Qb2 Rg3 4 +43 25 1404 Rd3 Qb2 Rg3 a5 5 +54 26 5445 Rd3 Qa5 Rg3 Bf5 Nf6+ Kh8 5 +68 27 10050 Re3 Qb2 Rg3 Qxe5 Nf6+ Kf8 Nxh7+ Kg8 6 +51 28 13924 Re3 Qb2 Rg3 Kh8 Rg5 g6 7 +40 33 34431 Re3 Qb2 Rg3 Kh8 Rg5 f6 Rh5 7 +223 36 43602 Nf6+ gxf6 exf6 Rf8 Rd8 c4 Rxc8 Kh8 Qxc4 8 +232 42 69603 Nf6+ gxf6 exf6 Rf8 Rd8 c4 Rxc8 Kh8 Qxc4 a5 9 +798 104 315156 Nf6+ Kf8 Qxh7 Qa4 e6 Ke7 Nxe8 Qxe8 exf7+ Kxf7 Rxe8 Kxe8 10 +864 151 504582 Nf6+ gxf6 exf6 Rf8 Qh4 Qd3 Rxd3 Bf5 Qg5+ Bg6 Qxc5 Bxd3 Qxa7 Bxc2 Qxa6 11 +1045 468 1711329 Nf6+ Kf8 Qxh7 Ke7 Nxe8 Qa4 Nd6 Be6 Qxg7 Qf4 Qg8 Rd7 12 +1385 1299 4765653 Nf6+ Kf8 Qxh7 Ke7 Qxg7 Rd8 Ng8+ Ke8 e6 Rxd1 exf7+ Kd7 f8=Q+ Kc6 Qgf6+ Kb7 Rxd1 Kb8 Qb6+ Rb7 Qfd6+ Ka8 Qxa6+ Qxa6 Qxa6+ Kb8 13 +1542 2640 9856685 Nf6+ Kf8 Qxh7 Qa4 e6 Ke7 exf7+ Kxf7 Nxe8 Qxe8 Rxe8 Kxe8 Qg8+ Ke7 Qxg7+ Ke6 Qxa7 c4 Kh2 14 +1881 6996 26085072 Nf6+ Kf8 Qxh7 Ke7 Qxg7 Rd8 Ng8+ Ke8 e6 Rxd1 exf7+ Kd7 f8=Q+ Kc6 Qgf6+ Kb7 Rxd1 Qa5 Rb1+ Qb4 Rxb4+ cxb4 Qxb4+ Ka8 Qc6+ Bb7 15 +2037 12078 45883926 Nf6+ Kf8 Qxh7 Ke7 Qxg7 Rd8 Ng8+ Ke8 e6 Rxd1 exf7+ Kd7 f8=Q+ Kc6 Qgf6+ Kb5 Rxd1 Rd7 Rb1+ Qb4 Rxb4+ cxb4 Qxc8 Rd1+ Kh2 a5 16 +2603 35031 132489734 Nf6+ Kf8 Qxh7 Ke7 Qxg7 Rd8 Ng8+ Ke8 e6 Rxd1 exf7+ Kd7 f8=Q+ Kc6 Qgf6+ Kb5 Rxd1 Rd7 Rb1+ Qb4 Rxb4+ cxb4 Qxc8 Rd6 Qxd6 Having found such a simple and reproducible problem, it shouldn't be too difficult to find and fix the bug. Thanks to everyone for the input! Tord
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.