Bad misevaluation by Naum 4

[Jouni]

In tournament level test game against R3 Naum 4 suddenly thinks he is totally winning:

[d]r7/P7/1KP5/6pp/1n1B1bk1/1N2p3/8/R7 w - -

Analysis by Naum 4:

84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nd5 87.Bb6 e2 88.Re8 Be3 89.c7 e1Q 90.c8Q+
+- (3.90) Depth: 13/36 00:00:02 1174kN, tb=2
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nd5 87.Bb6 Bg3 88.c7 Nxc7 89.Bxc7 e2 90.Bxg3 Kxg3
+- (4.31) Depth: 14/43 00:00:09 3948kN, tb=152
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nc2 87.Re8 Kf3 88.c7 Bxc7 89.Kxc7 e2 90.Bc3 Kf2 91.Nc5
+- (4.34) Depth: 15/36 00:00:14 6004kN, tb=777
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nd5 87.Bb6 e2 88.Re8 Ne3 89.Ba5 Nc4 90.c7 Nxa5+ 91.Nxa5 Bxc7
+- (4.31) Depth: 16/36 00:00:24 11323kN, tb=1549
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nd5 87.Bb6 Nxb6 88.Kxb6 h4 89.Nd4 h3 90.Rh8 Be5 91.c7 Bxc7+ 92.Kxc7
+- (4.90) Depth: 17/45 00:02:06 72869kN, tb=16215
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nd5 87.Bb6 Nxb6 88.Kxb6 h4 89.Nd4 h3 90.Rh8 Be5 91.c7 Bxc7+ 92.Kxc7 Kg3
+- (4.90) Depth: 18/32 00:02:10 75367kN, tb=16259
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nd5 87.Rd8 e2 88.Bf2 Nc7 89.Nd4 Bh2 90.Nxe2 Ne6 91.Rd5 Kf3 92.Bg1 Bxg1 93.Nxg1+
+- (5.24) Depth: 19/55 00:06:04 210mN, tb=80817
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nxc6 87.Kxc6 h4 88.Rh8 h3 89.Nc1 Bg3 90.Ne2 Bh4 91.Be5 Kf3 92.Nc3 e2 93.Rf8+ Kg2
+- (4.18) Depth: 20/54 00:35:00 1250mN, tb=373498

Rybka evaluates position as draw. Quess who is right...

Jouni

[Dann Corbit]

In tournament level test game against R3 Naum 4 suddenly thinks he is totally winning:

[d]r7/P7/1KP5/6pp/1n1B1bk1/1N2p3/8/R7 w - -

Analysis by Naum 4:

84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nd5 87.Bb6 e2 88.Re8 Be3 89.c7 e1Q 90.c8Q+
+- (3.90) Depth: 13/36 00:00:02 1174kN, tb=2
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nd5 87.Bb6 Bg3 88.c7 Nxc7 89.Bxc7 e2 90.Bxg3 Kxg3
+- (4.31) Depth: 14/43 00:00:09 3948kN, tb=152
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nc2 87.Re8 Kf3 88.c7 Bxc7 89.Kxc7 e2 90.Bc3 Kf2 91.Nc5
+- (4.34) Depth: 15/36 00:00:14 6004kN, tb=777
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nd5 87.Bb6 e2 88.Re8 Ne3 89.Ba5 Nc4 90.c7 Nxa5+ 91.Nxa5 Bxc7
+- (4.31) Depth: 16/36 00:00:24 11323kN, tb=1549
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nd5 87.Bb6 Nxb6 88.Kxb6 h4 89.Nd4 h3 90.Rh8 Be5 91.c7 Bxc7+ 92.Kxc7
+- (4.90) Depth: 17/45 00:02:06 72869kN, tb=16215
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nd5 87.Bb6 Nxb6 88.Kxb6 h4 89.Nd4 h3 90.Rh8 Be5 91.c7 Bxc7+ 92.Kxc7 Kg3
+- (4.90) Depth: 18/32 00:02:10 75367kN, tb=16259
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nd5 87.Rd8 e2 88.Bf2 Nc7 89.Nd4 Bh2 90.Nxe2 Ne6 91.Rd5 Kf3 92.Bg1 Bxg1 93.Nxg1+
+- (5.24) Depth: 19/55 00:06:04 210mN, tb=80817
84.Kb7 Re8 85.a8Q Rxa8 86.Rxa8 Nxc6 87.Kxc6 h4 88.Rh8 h3 89.Nc1 Bg3 90.Ne2 Bh4 91.Be5 Kf3 92.Nc3 e2 93.Rf8+ Kg2
+- (4.18) Depth: 20/54 00:35:00 1250mN, tb=373498

Rybka evaluates position as draw. Quess who is right...

Jouni

I'm not sure if it is a draw or not, but it is certainly a board frought with peril.

All pawns are passed and deadly.

Three pieces per side, including a major piece and neither king with any protection.

I can see easily where we can get zugzwang positions here.

I guess that neither engine can tell who is going to win. I know I can't.

My Rybka 3 does not see a draw that I can discern:

Code: Select all

Analysis from Q:\epd\sub\kb7.epd   
3/7/2009 1:11:15 AM Level: 2880 Seconds
Analyzing engine: Rybka 3

1) Kb7;                 
    Searching move: Kb6-b7
    Best move (Rybka 3): Bd4-c5
    Not found in: 48:00
      2	00:00	         358	21.564	+2.12	Kb6b7
      3	00:00	       1.182	71.198	+1.55	Kb6b7
      4	00:00	       2.057	123.904	+1.57	Kb6b7
      5	00:00	       6.947	215.567	+1.18	Kb6b7 Ra8g8
      6+	00:00	       9.892	211.029	+1.38	Kb6b7
      6+	00:00	      17.757	230.166	+1.58	Kb6b7
      6+	00:00	      20.799	224.191	+1.98	Kb6b7
      6	00:00	      31.690	257.544	+1.64	Kb6b7 Nb4xc6 Kb7xa8 h5h4
      7	00:00	      41.006	242.717	+1.33	Kb6b7 Nb4xc6 Kb7xa8 h5h4 Ra1g1+ Kg4f3
      8+	00:00	      53.389	232.639	+1.53	Kb6b7
      8	00:01	     155.823	255.300	+0.88	Bd4c5 Nb4c2 Ra1g1+
      9	00:01	     159.068	260.617	+0.78	Bd4c5 Nb4d5+ Kb6b7 Ra8g8
     10	00:01	     365.607	275.483	+0.59	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 Ra1g1+ Kg4f3 a7a8Q Rh8xa8 Kb7xa8 h5h4 Rg1h1 Kf3g2
     11	00:01	     419.556	272.432	+0.57	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 Ra1g1+ Kg4f3 Nb3d4+ Kf3f2 Rg1a1 Rh8h7+ Kb7c8 Rh7h8+ Kc8d7 Rh8a8 Ra1a2+
     12	00:02	     676.134	284.337	+0.57	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 Ra1g1+ Kg4f3 Nb3d4+ Kf3f2 Rg1a1 Rh8h7+ Kb7c8 Rh7h8+ Kc8d7 Rh8a8 Ra1a2+
     13+	00:04	   1.270.144	282.560	+0.77	Bd4c5
     13+	00:05	   1.324.788	277.760	+0.97	Bd4c5
     13	00:05	   1.349.358	278.464	+0.96	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 a7a8Q Rh8xa8 Kb7xa8 h5h4 Nb3d4 h4h3 Ra1g1+ Kg4h5 Ka8b7 h3h2 Rg1h1
     14+	00:08	   1.981.907	265.464	+1.16	Bd4c5
     14	00:11	   3.013.293	271.286	+0.57	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 a7a8Q Rh8xa8 Kb7xa8 h5h4 Nb3d4 h4h3 Ra1g1+ Kg4h5 Rg1h1 Kh5g4 Ka8b7 Bf4e5 Kb7c8 Kg4g3 Rh1g1+ Kg3f2
     15+	00:15	   4.208.169	279.561	+0.77	Bd4c5
     15	00:16	   4.504.353	279.170	+0.58	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 Ra1g1+ Kg4f3 Nb3d4+ Kf3f2 Rg1a1 Rh8h7+ Kb7c8 Rh7h8+ Kc8d7 Rh8a8 Ra1a2+ Kf2f1 Nd4c2 h5h4 Nc2xe3+ Nd5xe3 c6c7 Bf4xc7 Kd7xc7 Ne3g4 Ra2a1+ Kf1g2 Ra1a4 Ng4e5 Bc5d4
     16	00:32	   8.465.811	271.058	+0.73	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 a7a8Q Rh8xa8 Kb7xa8 h5h4 Nb3d4 h4h3 Ra1g1+ Kg4h5 Rg1h1 Kh5g4 Ka8b7 Bf4e5 Kb7c8 Kg4g3 Rh1g1+ Kg3f4 Kc8d7 Kf4e4 Nd4e6 Nd5f6+ Kd7d8 e3e2 Ne6xg5+
     17+	00:52	  13.171.142	259.699	+0.93	Bd4c5
     17+	00:58	  14.468.442	256.122	+1.13	Bd4c5
     17	01:09	  16.762.354	249.156	+1.19	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 a7a8Q Rh8xa8 Kb7xa8 h5h4 Nb3d4 h4h3 Ra1g1+ Kg4h5 Rg1e1 Kh5g4 Nd4c2 h3h2 Nc2xe3+ Bf4xe3 Bc5xe3 Kg4g3 Be3xg5 Kg3g2 Ka8b7 h2h1Q Re1xh1 Kg2xh1 Bg5e3 Kh1g2 Be3c5 Kg2f3 Kb7c8
     18	01:18	  19.000.800	249.289	+1.19	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 a7a8Q Rh8xa8 Kb7xa8 h5h4 Nb3d4 h4h3 Ra1g1+ Kg4h5 Rg1e1 Kh5g4 Nd4c2 h3h2 Nc2xe3+ Bf4xe3 Bc5xe3 Kg4g3 Be3xg5 Kg3g2 Ka8b7 h2h1Q Re1xh1 Kg2xh1 Bg5e3 Kh1g2 Be3c5 Kg2f3 Kb7c8
     19	01:46	  26.073.803	252.356	+1.00	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 a7a8Q Rh8xa8 Kb7xa8 h5h4 Nb3d4 h4h3 Ra1g1+ Kg4h5 Rg1e1 Kh5g4 Nd4c2 h3h2 Nc2xe3+ Bf4xe3 Bc5xe3 Kg4g3 Be3xg5 Kg3g2 Ka8b7 h2h1Q Re1xh1 Kg2xh1 Bg5e3 Kh1g2 Be3c5 Kg2f3 Kb7c8
     20	03:01	  46.961.886	265.579	+1.00	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 a7a8Q Rh8xa8 Kb7xa8 h5h4 Nb3d4 h4h3 Ra1g1+ Kg4h5 Rg1e1 Kh5g4 Nd4c2 h3h2 Nc2xe3+ Bf4xe3 Bc5xe3 Kg4g3 Be3xg5 Kg3g2 Ka8b7 h2h1Q Re1xh1 Kg2xh1 Bg5e3 Kh1g2 Be3c5 Kg2f3 Kb7c8
     21	05:19	  82.347.817	264.332	+1.15	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 a7a8Q Rh8xa8 Kb7xa8 h5h4 Nb3d4 h4h3 Ra1g1+ Kg4h5 Rg1e1 Kh5g4 Nd4c2 h3h2 Nc2xe3+ Bf4xe3 Bc5xe3 Kg4g3 Be3xg5 Kg3g2 Ka8b7 h2h1Q Re1xh1 Kg2xh1 Bg5e3 Kh1g2 Be3c5 Kg2f3 Kb7c8
     22	09:13	 145.455.643	269.186	+1.15	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 a7a8Q Rh8xa8 Kb7xa8 h5h4 Nb3d4 h4h3 Ra1g1+ Kg4h5 Rg1e1 Kh5g4 Nd4c2 h3h2 Nc2xe3+ Bf4xe3 Bc5xe3 Kg4g3 Be3xg5 Kg3g2 Ka8b7 h2h1Q Re1xh1 Kg2xh1 Bg5e3 Kh1g2 Be3c5 Kg2f3 Kb7c8
     23	21:37	 361.683.791	285.531	+1.12	Bd4c5 Nb4d5+ Kb6b7 Ra8h8 a7a8Q Rh8xa8 Kb7xa8 h5h4 Nb3d4 h4h3 Ra1g1+ Kg4h5 Rg1e1 Kh5g4 Nd4c2 h3h2 Nc2xe3+ Bf4xe3 Bc5xe3 Kg4g3 Be3xg5 Kg3g2 Ka8b7 h2h1Q Re1xh1 Kg2xh1 Bg5h6 Kh1g2 Bh6f8 Kg2f3 Kb7c8
   3/7/2009 1:33:43 AM, Time for this analysis: 00:22:25, Rated time: 48:00

0 of 1 matching moves
3/7/2009 1:33:44 AM, Total time: 12:22:29 AM
Rated time: 48:00 = 2880 Seconds

[AdminX]

Hi Jouni,

Very interesting position, it has study like qualities! Naum appears to be way off base. As you walk Naum thru some of the best lines you can see Naum's evals start to drop. The position appears to lead to a draw.

[M ANSARI]

In such a position it is search that counts and not evaluation. If N4 did fail in this position it most likely failed due to inability to search deep enough.

[Graham Banks]

In such a position it is search that counts and not evaluation. If N4 did fail in this position it most likely failed due to inability to search deep enough.

Despite all its reported failings, Naum 4 is still the second strongest engine, and by a reasonable margin too. It also has a nice playing style in my opinion.
Alex has done a good job.

PS - this is just a general statement and not a jab at anybody.

[M ANSARI]

I agree also ... I like how N4 handles bishops, especially a bishop pair. With regards to search, that was not meant to denegrate N4 in any way, but more to show that some positions in chess simply need deep search ... something that is more related to hardware used and time available rather than an engine defect.

I will say however that Rybka 3 is a phenomenal searcher ... by far the fastest searcher of any engine out there, and by a long gap.