Engine mating bug with won position

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

stevenaaus
Posts: 608
Joined: Wed Oct 13, 2010 9:44 am
Location: Australia

Engine mating bug with won position

Post by stevenaaus »

[D]4Q1K1/7R/8/2k5/7P/8/8/8 b - - 0 79
Robbo family seem to have a bug with this position.
Ivanhoe-Beta version 999947c reports

Code: Select all

.....
12 [M4]  79.... Kd5 80.Rd7+ Kc4 81.Qc8+ Kb3 82.Rb7+ Ka3 83.Qa8#  (0.79)
13 [M4]  79.... Kd5 80.Rd7+ Kc4 81.Qc8+ Kb3 82.Rb7+ Ka2 83.Qa8#  (1.56)
14 [M9]  79.... Kd5 80.Rd7+ Kc4 81.Qc8+ Kb3 82.Rb7+ Ka2 83.Qd8 Ka3 84.Qb8 Ka2 85.Qe5 Ka3 86.Qe6 Ka4 87.Qc8 Ka3 88.Qa8#  (2.98)
15 [M11]  79.... Kd5 80.Rd7+ Kc4 81.Qc8+ Kb3 82.Rb7+ Ka2 83.Qd8 Ka3 84.Qb8 Ka4 85.Rc7 Ka5 86.Qc8 Kb6 87.Rb7+ Ka6 88.Qb8 Ka5 89.Qe8 Ka6 90.Qb5#  (5.22)
Something funny is happening. Komodo32 2.03 JA (Is there a 32 bit Linux Komodo 3 yet ?!?) doesn't like it either.
Top
Jim Collins
Posts: 60
Joined: Sat Mar 11, 2006 6:11 pm

Re: Engine mating bug with won position

Post by Jim Collins »

Ivanhoe version B57beta

New game
4Q1K1/7R/8/2k5/7P/8/8/8 b - - 0 1

Analysis by Ivanhoe-B57beta:

79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc2 82.Qc8+ Kb2 83.Rb7+ Ka2 84.Qa8#
+- (#799 --) Depth: 15 00:00:00 603kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc2 82.Qc8+ Kb3 83.Rb7+ Ka4 84.Qd8 Ka3 85.Qd3+ Ka4 86.Qb5+ Ka3 87.Qc6 Ka2 88.Qc2+ Ka1 89.Ra7#
+- (#6) Depth: 15 00:00:00 1018kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc2 82.Qc8+ Kb3 83.Rb7+ Ka4 84.Qd8 Ka3 85.Qd3+ Ka4 86.Qb5+ Ka3 87.Qc6 Ka2 88.Qc2+ Ka1 89.Ra7#
+- (#6) Depth: 15 00:00:00 1044kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc2 82.Qc8+ Kb3 83.Rb7+ Ka4 84.Qd8 Ka3 85.Qd3+ Ka4 86.Qb5+ Ka3 87.Qc6 Ka2 88.Qc2+ Ka1 89.Ra7#
+- (#6) Depth: 16 00:00:00 1704kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc2 82.Qc8+ Kb3 83.Rb7+ Ka4 84.Qd8 Ka3 85.Qd3+ Ka4 86.Qb5+ Ka3 87.Qc6 Ka2 88.Qc2+ Ka1 89.Ra7#
+- (#6) Depth: 16 00:00:00 1717kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc2 82.Qc8+ Kb3 83.Rb7+ Ka4 84.Qd8 Ka3 85.Qd3+ Ka4 86.Qb5+ Ka3 87.Qc6 Ka2 88.Qc2+ Ka1 89.Ra7#
+- (#6) Depth: 17 00:00:00 3130kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc2 82.Qc8+ Kb3 83.Rb7+ Ka4 84.Qd8 Ka3 85.Qd3+ Ka4 86.Qb5+ Ka3 87.Qc6 Ka2 88.Qc2+ Ka1 89.Ra7#
+- (#6) Depth: 17 00:00:00 3151kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qa8 Kc5 83.Qc8+ Kb5 84.Rb7+ Ka5 85.Qc3+ Ka6 86.Qc6+ Ka5 87.Qh6 Ka4 88.Qf4+ Ka3 89.Qc1+ Ka4 90.Qc2+ Ka5 91.Qc3+
+- (#6) Depth: 18 00:00:01 6437kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qa8 Kc5 83.Qc8+ Kb5 84.Rb7+ Ka5 85.Qc3+ Ka6 86.Qc6+ Ka5 87.Qh6 Ka4 88.Qf4+ Ka3 89.Qc1+ Ka4 90.Qc2+ Ka5 91.Qc3+
+- (#6) Depth: 18 00:00:01 6440kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qb8 Kc3 83.Qc8+ Kb2 84.Rb7+ Ka2 85.Qc2+ Ka1 86.Ra7#
+- (#6) Depth: 19 00:00:01 11321kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qb8 Kc3 83.Qc8+ Kb2 84.Rb7+ Ka2 85.Qc2+ Ka1 86.Ra7#
+- (#6) Depth: 19 00:00:01 11376kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qc8+ Kb5 83.Rb7+ Ka5 84.Qa8#
+- (#5) Depth: 20 00:00:03 19034kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qc8+ Kb5 83.Rb7+ Ka5 84.Qa8#
+- (#5) Depth: 20 00:00:03 19105kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qc8+ Kb5 83.Rb7+ Ka5 84.Ra7+ Kb6 85.Rb7+ Ka5
+- (#5) Depth: 21 00:00:05 32195kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qc8+ Kb5 83.Rb7+ Ka5 84.Ra7+ Kb6 85.Rb7+ Ka5
+- (#5) Depth: 21 00:00:05 32198kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qc8+ Kb5 83.Rb7+ Ka5 84.Qa8#
+- (#5) Depth: 22 00:00:08 53335kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qc8+ Kb5 83.Rb7+ Ka5 84.Qa8#
+- (#5) Depth: 22 00:00:08 53338kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qc8+ Kb5 83.Rb7+ Ka5 84.Rc7 Ka4 85.Rh7 Ka5 86.Ra7+ Kb6 87.Rb7+ Ka5
+- (#5) Depth: 23 00:00:14 88498kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qc8+ Kb5 83.Rb7+ Ka5 84.Rc7 Ka4 85.Rh7 Ka5 86.Ra7+ Kb6 87.Rb7+ Ka5
+- (#5) Depth: 23 00:00:14 88499kN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qc8+ Kb5 83.Rb7+ Ka5 84.Rc7 Ka4 85.Rh7 Ka5 86.Qa8+ Kb6 87.Qd5 Ka6 88.Qc6+ Ka5 89.Qc8 Ka4 90.Rb7 Ka5
+- (#5) Depth: 24 00:00:22 139mN
79...Kc4 80.Rc7+ Kd3 81.Rd7+ Kc4 82.Qc8+ Kb5 83.Rb7+ Ka5 84.Rc7 Ka4 85.Rh7 Ka5 86.Qa8+ Kb6 87.Qd5 Ka6 88.Qc6+ Ka5 89.Qc8 Ka4 90.Rb7 Ka5
+- (#5) Depth: 24 00:00:22 13

Komodo 3.1

M - Komodo64 3.1, 9999"/1 Microsoft
4Q1K1/7R/8/2k5/7P/8/8/8 b - - 0 1

Analysis by Komodo64 3.1:


+- (#6) Depth: 9 00:00:00 57kN
79...Kc4 80.Qe4+ Kc5 81.Rh5+ Kd6 82.Qd5+ Kc7 83.Rh7+ Kb6 84.Rb7+ Ka6 85.Qb5#
+- (#6) Depth: 9 00:00:00 58kN

(, Microsoft 06.09.2011)
Top
zullil
Posts: 6442
Joined: Tue Jan 09, 2007 12:31 am
Location: PA USA
Full name: Louis Zulli

Re: Engine mating bug with won position

Post by zullil »

Jim Collins wrote: Komodo 3.1

M - Komodo64 3.1, 9999"/1 Microsoft
4Q1K1/7R/8/2k5/7P/8/8/8 b - - 0 1

Analysis by Komodo64 3.1:


+- (#6) Depth: 9 00:00:00 57kN
79...Kc4 80.Qe4+ Kc5 81.Rh5+ Kd6 82.Qd5+ Kc7 83.Rh7+ Kb6 84.Rb7+ Ka6 85.Qb5#
+- (#6) Depth: 9 00:00:00 58kN

(, Microsoft 06.09.2011)
Does Komodo eventually say that after Black's best move White mates in four?
Top
Jim Collins
Posts: 60
Joined: Sat Mar 11, 2006 6:11 pm

Re: Engine mating bug with won position

Post by Jim Collins »

Hi Louis. No it doesn't. It shows mate in 5.
Top
zullil
Posts: 6442
Joined: Tue Jan 09, 2007 12:31 am
Location: PA USA
Full name: Louis Zulli

Re: Engine mating bug with won position

Post by zullil »

Jim Collins wrote:Hi Louis. No it doesn't. It shows mate in 5.
Seems a bit weird. Here's what Critter gives, without TBs:

Code: Select all

15/13  00:00    218104  487928    -M4 Kc4 Rd7 Kb4 Qc8 Ka5 Rb7 Ka6 Qa8#
Top
Michel
Posts: 2272
Joined: Mon Sep 29, 2008 1:50 am

Re: Engine mating bug with won position

Post by Michel »

Code: Select all

Hi Louis. No it doesn't. It shows mate in 5.
This is probably due to the engine trusting mate scores in the hash table.

Because of LMR the hash table may contain mate scores which are not minimal. If such mate scores are subsequently trusted the engine will no longer find the shortest mate.
Top

Return to “Computer Chess Club: General Topics”