A mate in 3 that engines cannot solve ? WOW!

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

Moderators: hgm, Rebel, chrisw

bob
Posts: 20943
Joined: Mon Feb 27, 2006 7:30 pm
Location: Birmingham, AL

Re: A mate in 3 that engines cannot solve ? WOW!

Post by bob »

AdminX wrote:
Robert Flesher wrote:[d] 8/1n3Np1/1N4Q1/1bkP4/p1p2p2/P1P2R2/3P2PK/B2R4 w - - 0 0
Which engine can't solve it? :D

Analysis by Bright-0.4a3:

1.Rh1 Be8 2.Qb1 Bxf7 3.Qb4#
+- (#3) Depth: 6/21 00:00:00 58kN
+- (#3) Depth: 120/21 00:00:00 326kN


Analysis by Onno-1-1-1:

1.Rh1 Be8 2.Qb1 Bxf7 3.Qb4#
+- (#3) Depth: 127/127 00:00:01 267kN
+- (#3) Depth: 127/127 00:00:02 269kN


Analysis by Crafty 23.00:

1.Rxf4 Be8 2.Nxc4 Kxd5 3.Nb6+ Kc5 4.Rf5#
+- (31.32) Depth: 3/11 00:00:00 1kN
+- (#4) Depth: 9/19 00:00:00 2790kN

1.Rh1 Be8 2.Qb1 Bxf7 3.Qb4#
+- (#3) Depth: 9/21 00:00:02 1768kN
+- (#3) Depth: 10/21 00:00:04 17554kN


Analysis by Loop 13.6:

1.Rh1 Be8 2.Qb1 Bxf7 3.Qb4#
+- (#3) Depth: 6 00:00:00 32kN
+- (#3) Depth: 50 00:00:05 33762kN
Most will find the mate in 4 first, and then stop too quickly. The shorter mate has fewer checks and takes a deeper search to find.
User avatar
cms271828
Posts: 316
Joined: Wed Apr 12, 2006 10:47 pm

Re: A mate in 3 that engines cannot solve ? WOW!

Post by cms271828 »

Thanks, that doesn't answer what I was saying though, but I see I missed Qg1# if black plays Bb5, which then explains why the key move is Rh1, to allow the queen to slide across the rank.

So that explains everything except for the fact that my engine couldn't solve it in 3 moves, only 4.

I really don't get why, but seems other engines also cannot solve it in 3, is it down to null move or transposition table? I can see how it is caused from alpha-beta pruning :?


Ok, just editing comment, I ran it with null move turned off and it found it in 3, its cause black is limited with what he can move, and this causes zugzwangs, but my engine only turns null move off when a few pieces left which is what must be the problem.
Colin
Michel
Posts: 2273
Joined: Mon Sep 29, 2008 1:50 am

Re: A mate in 3 that engines cannot solve ? WOW!

Post by Michel »

GnuChess finds the mate in 3 with null move turned off.

But some engines find it with null move turned on. I am very curious how they do that.

The null move refutes Rh1 so when null move is turned on one should not be able to find the mate in three....
Martin Brown
Posts: 46
Joined: Sun Oct 18, 2009 12:07 pm

Re: A mate in 3 that engines cannot solve ? WOW!

Post by Martin Brown »

Robert Flesher wrote:[d] 8/1n3Np1/1N4Q1/1bkP4/p1p2p2/P1P2R2/3P2PK/B2R4 w - - 0 0
Rybka 2.32 32bit, Shredder10, HIARCS10 get it OK here
Fritz 8 sees Rb1, Rh1, Rxf4, Re1 all #4 Play Rh1 for it and it sees the right path but unwind that move and it is still blind to the mating solution.
Fritz5.32 solves it without difficulty.
Junior7 finds Nc8 and that is all
CometB27 gets it but claims #2 (fence post error)
CometB68 only sees Rxf4 and claims #3 (same fence post)

And the venerable Turing engine gets right it in under 5s!

Material gain from Rxf4 seems to blind the more aggressive pruning engines to what is a very passive looking but critical first move.
Martin Brown
Michel
Posts: 2273
Joined: Mon Sep 29, 2008 1:50 am

Re: A mate in 3 that engines cannot solve ? WOW!

Post by Michel »

Material gain from Rxf4 seems to blind the more aggressive pruning engines to what is a very passive looking but critical first move.
No I think the issue is that after Rh1 the null move fails high.

I still wonder how the engines that find the mate in three handle this.
User avatar
rvida
Posts: 481
Joined: Thu Apr 16, 2009 12:00 pm
Location: Slovakia, EU

Re: A mate in 3 that engines cannot solve ? WOW!

Post by rvida »

Indeed, it seems to be a null move issue. Critter 0.42 finds it in later iterations thanks to null move verification search.

Code: Select all

FEN: 8/1n3Np1/1N4Q1/1bkP4/p1p2p2/P1P2R2/3P2PK/B2R4 w - - 0 1 

Critter 0.42:
  2/3+	00:00	         298	0	+29,94	Rf3xf4
  2/3	00:00	         454	0	+30,52	Rf3xf4 Bb5e8
  2/6+	00:00	         776	0	+31,03	Rd1e1
  2/6	00:00	         812	0	+31,61	Rd1e1 Bb5e8
  3/5	00:00	       1.109	0	+31,61	Rd1e1 Bb5e8 Re1xe8
  4/6+	00:00	       1.787	0	+32,05	Rd1e1
  4/6+	00:00	       1.852	0	+33,08	Rd1e1
  4/6	00:00	       2.187	0	+33,20	Rd1e1 Bb5e8 Re1xe8 Nb7a5
  5/13+	00:00	       3.173	0	+33,65	Rd1e1
  5/13+	00:00	       3.555	0	+34,93	Rd1e1
  5/13+	00:00	       3.705	0	+37,08	Rd1e1
  5/14	00:00	       5.721	0	+M5	Rd1e1 Bb5e8 Re1xe8 Nb7d6 Qg6xd6+ Kc5b5 Qd6c6+ Kb5a6
  6/12	00:00	       9.979	665.266	+M5	Rd1e1 Bb5e8 Re1xe8 Nb7d6 Qg6xd6+ Kc5b5 Qd6c6+ Kb5a6 Re8a8+ Ka6xb6
  7/12	00:00	      17.798	1.186.533	+M5	Rd1e1 Bb5e8 Re1xe8 Nb7d6 Qg6xd6+ Kc5b5 Qd6c6+ Kb5a6 Re8a8+ Ka6a7
  7/12	00:00	      27.199	906.633	+M4	d2d4+ c4xd3/ep Rf3xf4 Bb5e8 c3c4 Be8xf7 Nb6xa4+
  8/9	00:00	      39.031	848.500	+M4	d2d4+ c4xd3/ep Rf3xf4 Bb5e8 c3c4 Be8xf7 Nb6xa4+
  9/9	00:00	      59.442	771.974	+M4	d2d4+ c4xd3/ep Rf3xf4 Bb5e8 c3c4 Be8xf7 Nb6xa4+
 10/9	00:00	      96.393	688.521	+M4	d2d4+ c4xd3/ep Rf3xf4 Bb5e8 c3c4 Be8xf7 Nb6xa4+
 11/9	00:00	     151.019	692.747	+M4	d2d4+ c4xd3/ep Rf3xf4 Bb5e8 c3c4 Be8xf7 Nb6xa4+
 12/9	00:00	     228.646	732.839	+M4	d2d4+ c4xd3/ep Rf3xf4 Bb5e8 c3c4 Be8xf7 Nb6xa4+
 13/9	00:00	     366.310	734.088	+M4	d2d4+ c4xd3/ep Rf3xf4 Bb5e8 c3c4 Be8xf7 Nb6xa4+
 14/9	00:00	     572.339	764.137	+M4	d2d4+ c4xd3/ep Rf3xf4 Bb5e8 c3c4 Be8xf7 Nb6xa4+
 14/9	00:01	     786.778	775.150	+M3	Rd1h1 Bb5e8 Qg6b1 Be8xf7 Qb1b4+
 15/7	00:01	     805.102	781.652	+M3	Rd1h1 Bb5e8 Qg6b1 Be8xf7 Qb1b4+