I have been looking at a lot of N4 long time control games and somehow felt that although N4 had an excellent evaluation function, it seems incredibly poor with search. Especially when searching critical lines which have pawn pushes or pawn chain advances ... or simply critical lines. So I decided to do an 8 core N4 against a 1core R3 match with different contempt for R3. As you can see both R3's on single core can be very competitive with Octa N4, with a 0 contempt R3 single core actually coming out ahead. Now I am totally convinced that N4 search in some critical positions is so serious that it can probably be called a bug. Hopefully this can be corrected in a future N4 release or in N5. The amazing thing is that even with this quite obvious weakness, N4 can still come out as the strongest non Rybka engine.
I do think that N4 has a good future once a few of these weaknesses are fixed. I have chosen a few examples of the critical areas of search where N4 fails from some of the games. It might not be fair to fault N4 here since it seems that other engines also fail ... but this shows again what an incredible piece of efficient engineering Rybka is. I am a civil and architectural engineer and there is always admiration when someone is able to maximize strength to the critical areas of stress ... thus making your load bearing resources the most efficient possible. I think that is the best way to describe Rybka ... it puts the CPU resources to work where they count the most. I don't know what tricks Vas uses, but after looking at other R3 games on an antiquated Athlon 1200 manhandle the strongest commercial programs on Quadcores ... and here again N4 on Octa 4Ghz ... possibly the strongest commercial hardware readily available ... you cannot but be impressed.
N4_8core_R3_1core_15_cont_5_1-1 2009
1 Naum 4 +52/-39/=109 53.25% 106.5/200
2 Rybka 3 1-cpu +39/-52/=109 46.75% 93.5/200
N4_8core_R3_1core_0_cont_5_1-1 2009
1 Rybka 3 1-cpu _0_contempt +49/-47/=104 50.50% 101.0/200
2 Naum 4 +47/-49/=104 49.50% 99.0/200
CRITICAL POSITION 1:
[d]8/1p2b1pp/3p2k1/3P4/P3p1P1/8/1P1BK2P/8 b - g3 0 32
Here N4 plays Bg5? ... now a clip analysis for 5 minutes on 8 core N4 and 10 seconds 1 core R3. As you can see even with 8 cores and 5 minutes N4 cannot see that Bxg5 wins easily while single core R3 sees the win immediately
4180: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
8/1p4pp/3p2k1/3P2b1/P3p1P1/8/1P1BK2P/8 w - - 0 1
Analysis by Naum 4:
33.Be3 Be7 34.a5 Bf6 35.Bf4 Be7
+/= (0.29) Depth: 6/9 00:00:00 2kN
33.Be3 Kf6 34.Bd4+ Kg6 35.h3 Bf6 36.Ke3
+/= (0.27) Depth: 7/11 00:00:00 4kN
33.Be3 Kf6 34.Bd4+ Kg6 35.h4 Bxh4 36.Ke3 Kg5
+/= (0.29) Depth: 8/14 00:00:00 8kN
33.Be3 h5 34.gxh5+ Kxh5 35.a5 g6 36.Bxg5 Kxg5 37.Ke3
= (0.14) Depth: 9/16 00:00:00 25kN
33.Be3 h5 34.gxh5+ Kxh5 35.a5 Kg6 36.Bd4 Bf6 37.Ke3 Kf5
+/= (0.26) Depth: 10/17 00:00:00 37kN
33.Be3 Bxe3 34.Kxe3 Kg5 35.h3 h5 36.gxh5 Kf5 37.a5 Ke5 38.b4
= (0.17) Depth: 11/18 00:00:00 88kN
33.Bc3 Bf4 34.h4 h5 35.gxh5+ Kxh5 36.Bxg7 Kxh4 37.Bd4 Kg4 38.a5 Kf5
= (-0.02) Depth: 12/21 00:00:00 202kN
33.Bc3 Bf4 34.h4 h5 35.gxh5+ Kxh5 36.Bxg7 Kxh4 37.a5 Kg4 38.b4 Bg5 39.Bd4
= (0.19) Depth: 13/23 00:00:00 327kN
33.Bc3 Bf4 34.h4 h5 35.gxh5+ Kxh5 36.Bxg7 Kxh4 37.a5 Kg4 38.b4 Bg5 39.Bd4 Kf5
= (0.19) Depth: 14/17 00:00:00 358kN
33.Bc3 Bf4 34.h4 Bg3 35.Ke3 Bxh4 36.Kxe4 Bf6 37.Bxf6 Kxf6 38.Kf4 Kg6 39.a5 h5 40.gxh5+
+/= (0.30) Depth: 15/31 00:00:00 760kN, tb=2
33.Bc3 Bf4 34.h4 Bg3 35.Ke3 Bxh4 36.Kxe4 Bf6 37.Bxf6 Kxf6 38.Kf4 Kg6 39.a5 h5 40.b4 hxg4
+/= (0.28) Depth: 16/25 00:00:00 974kN, tb=2
33.Bc3 Bf4 34.h4 Bg3 35.Ke3 Bxh4 36.Kxe4 Bf6 37.Bxf6 Kxf6 38.a5 g6 39.Kf4 h5 40.g5+ Ke7 41.b4
+/= (0.38) Depth: 17/31 00:00:01 1901kN, tb=2
33.Bc3 Bf4 34.h4 Bg3 35.Ke3 Bxh4 36.Kxe4 Bf6 37.Bxf6 Kxf6 38.a5 Kg5 39.Kf3 h5 40.gxh5 Kf5 41.b4 Ke5
+/= (0.37) Depth: 18/37 00:00:01 2580kN, tb=2
33.Bc3 Bf6 34.Ke3 Kg5 35.h3 h5 36.Kxe4 hxg4 37.Bd2+ Kg6 38.hxg4 Bxb2 39.a5 Be5 40.Bf4 Bxf4 41.Kxf4 Kf6 42.g5+
+/= (0.30) Depth: 19/37 00:00:03 6737kN, tb=21
33.Bc3 Bf4 34.h4 Bg3 35.Ke3 Bxh4 36.Kxe4 Bf6 37.Bxf6 Kxf6 38.a5 Ke7 39.b4 Kd7 40.b5 Kc7 41.Kf4 Kc8 42.Ke3 Kd7
+/= (0.52) Depth: 20/60 00:00:10 27211kN, tb=2168
33.Bc3 Kf7 34.a5 g6 35.b4 Ke7 36.b5 Kd7 37.Bd2 Bf6 38.Be1 Be5 39.Ke3 Bxh2 40.Kxe4 Bg1 41.g5 Kc7 42.Bc3 Bc5 43.Bd4
+/= (0.69) Depth: 21/50 00:00:15 40277kN, tb=2906
33.Bc3 Kf7 34.a5 g6 35.b4 Ke7 36.b5 Kd7 37.Bd2 Bf6 38.Be1 Bg5 39.h4 Bf4 40.Bd2 Bg3 41.Ke3 Bxh4 42.Kxe4 Kc7 43.g5 Bf2
+/= (0.69) Depth: 22/45 00:00:21 51477kN, tb=3194
33.Bc3 Kf7 34.a5 g6 35.b4 Ke7 36.b5 Kd7 37.Bd2 Bf6 38.g5 Bd4 39.h3 Kc8 40.Bf4 Kd7 41.h4 Kc7 42.Be3 Be5 43.Bb6+ Kd7 44.Ke3
+/= (0.70) Depth: 23/45 00:00:34 76702kN, tb=7075
33.Bc3 Kf7 34.b4 Bf4 35.h4 g5 36.h5 Ke7 37.Bd4 Kf7 38.a5 Ke7 39.b5 Kd7 40.Bb6 Kc8 41.Be3 Kb8 42.Bd4 Kc7 43.a6 bxa6 44.bxa6 h6
+/- (0.81) Depth: 24/52 00:01:02 143mN, tb=11884
(, 13.01.2009)
4180: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
8/1p4pp/3p2k1/3P2b1/P3p1P1/8/1P1BK2P/8 w - - 0 1
Analysis by Rybka 3 1-cpu:
33.Be3 Kf6 34.Bd4+
+/= (0.31) Depth: 6 00:00:00 0kN
33.Be3 h6 34.a5 Kf6
+/= (0.30) Depth: 7 00:00:00 1kN
33.Bxg5
+/= (0.51 !) Depth: 7 00:00:00 1kN
33.Bxg5
+/- (0.71 !) Depth: 7 00:00:00 1kN
33.Bxg5
+/- (1.11 !) Depth: 7 00:00:00 1kN
33.Bxg5 Kxg5 34.Ke3
+/- (1.31) Depth: 7 00:00:00 1kN
33.Bxg5
+- (1.51 !) Depth: 8 00:00:00 2kN
33.Bxg5
+- (1.71 !) Depth: 8 00:00:00 2kN
33.Bxg5
+- (2.11 !) Depth: 8 00:00:00 2kN
33.Bxg5
+- (2.91 !) Depth: 8 00:00:00 3kN
33.Bxg5 Kxg5 34.Ke3 h6 35.Kxe4 Kf6 36.a5 Ke7 37.b4 g6 38.b5
+- (3.12) Depth: 8 00:00:00 6kN
33.Bxg5 Kxg5[] 34.Ke3 h6 35.Kxe4 Kf6 36.a5 Ke7 37.b4 g6 38.b5
+- (3.12) Depth: 9 00:00:00 6kN
33.Bxg5 Kxg5[] 34.Ke3 h6 35.Kxe4 Kf6 36.a5 Ke7 37.b4 g6
+- (3.12) Depth: 10 00:00:00 8kN
33.Bxg5 Kxg5[] 34.Ke3 h6 35.Kxe4 Kf6 36.a5 Ke7 37.b4 g6 38.b5 Kd7 39.h4
+- (3.10) Depth: 11 00:00:00 11kN
33.Bxg5 Kxg5[] 34.Ke3 h6 35.Kxe4 Kf6 36.a5 Ke7 37.b4 g6 38.b5 Kd7 39.h4
+- (3.10) Depth: 12 00:00:00 32kN
33.Bxg5 Kxg5[] 34.Ke3[] Kf6 35.Kxe4 Ke7 36.b4 g6 37.a5 Kd7 38.Kf4 h6
+- (3.11) Depth: 13 00:00:00 65kN
33.Bxg5 Kxg5[] 34.Ke3[] Kf6 35.Kxe4 Ke7 36.a5 Ke8 37.Kf5 Kf7 38.g5
+- (3.13) Depth: 14 00:00:00 80kN
33.Bxg5 Kxg5[] 34.Ke3[] Kf6 35.Kxe4 Ke7 36.a5 g6 37.g5 Kd7 38.Kd4 Ke7 39.Kc4 Kd7 40.b4
+- (3.13) Depth: 15 00:00:01 108kN
33.Bxg5
+- (3.33 !) Depth: 16 00:00:01 134kN
33.Bxg5 Kxg5[] 34.Ke3[] Kf6 35.Kxe4[] Ke7 36.a5 g6 37.g5 Kd7 38.Kd4 Ke7 39.Kc4
+- (3.37) Depth: 16 00:00:01 177kN
33.Bxg5
+- (3.57 !) Depth: 17 00:00:02 228kN, tb=1
33.Bxg5
+- (3.77 !) Depth: 17 00:00:02 265kN, tb=2
33.Bxg5 Kxg5[] 34.Ke3[] Kf6 35.Kxe4[] Ke7 36.a5[] g6 37.g5[] Kd7 38.Kd4 Ke7 39.Kc4
+- (3.76) Depth: 17 00:00:03 317kN, tb=2
33.Bxg5
+- (3.96 !) Depth: 18 00:00:04 397kN, tb=7
33.Bxg5
+- (4.16 !) Depth: 18 00:00:04 447kN, tb=13
33.Bxg5
+- (4.56 !) Depth: 18 00:00:05 514kN, tb=20
(, 13.01.2009)
CRITICAL POSITION 2:
[d]8/8/8/P4Qp1/4p2p/1Bp3k1/3q4/5K2 w - - 0 101
In this position N4 is probably already lost but anyway plays Qe5+
[d]8/8/8/P3Q1p1/4p2p/1Bp3k1/3q4/5K2 b - - 0 101
Again a clip analysis on 8 core N4 @ 5 minutes vs. 1 core R3 @ 10 seconds ... as is fairly obvious N4 even with big hardware is oblivious to any danger
4229: Naum 4 - Rybka 3 1-cpu _0_contempt, N4_R3_0_cont_5_1-1 2009
8/8/8/P3Q1p1/4p2p/1Bp3k1/3q4/5K2 b - - 0 1
Analysis by Naum 4:
101...Kf3 102.Qf5+ Kg3
= (0.00) Depth: 6/10 00:00:00 1kN
101...Kf3 102.Qf5+ Kg3
= (0.00) Depth: 7/15 00:00:00 2kN
101...Kf3 102.Qf5+ Kg3
= (0.00) Depth: 8/15 00:00:00 3kN
101...Kf3 102.Qf5+ Kg3
= (0.00) Depth: 9/18 00:00:00 5kN
101...Kf3 102.Qf5+ Kg3
= (0.00) Depth: 10/17 00:00:00 9kN
101...Kf3 102.Qf5+ Kg3
= (0.00) Depth: 11/16 00:00:00 15kN
101...Kf3 102.Qf5+ Kg3
= (0.00) Depth: 12/17 00:00:00 26kN
101...Kf3 102.Qf5+ Kg3
= (0.00) Depth: 13/22 00:00:00 42kN
101...Kf3 102.Qf5+ Kg3
= (0.00) Depth: 14/23 00:00:00 85kN
101...Kf3 102.Qf5+ Kg3
= (0.00) Depth: 15/23 00:00:00 137kN
101...Kf3 102.Qf5+ Kg3
= (0.00) Depth: 16/24 00:00:00 226kN, tb=1
101...Kf3 102.Qf5+ Kg3
= (0.00) Depth: 17/25 00:00:00 356kN, tb=5
101...Kf3 102.Qf5+ Kg3
= (0.00) Depth: 18/29 00:00:00 633kN, tb=17
(, 13.01.2009)
4229: Naum 4 - Rybka 3 1-cpu _0_contempt, N4_R3_0_cont_5_1-1 2009
8/8/8/P3Q1p1/4p2p/1Bp3k1/3q4/5K2 b - - 0 1
Analysis by Rybka 3 1-cpu:
101...Qf4+
-+ (-7.69) Depth: 6 00:00:00 14kN
101...Qf4+
-+ (-10.99) Depth: 7 00:00:00 50kN
101...Qf4+
-+ (-11.16) Depth: 8 00:00:00 117kN
101...Qf4+
-+ (-11.62) Depth: 9 00:00:01 309kN
101...Qf4+
-+ (-13.65) Depth: 10 00:00:05 1245kN, tb=2
101...Qf4+
-+ (-14.09) Depth: 11 00:00:08 2137kN, tb=6
101...Qf4+
-+ (-15.19) Depth: 12 00:00:13 3288kN, tb=13
(, 13.01.2009)
N4 8core vs. R3 1core ... problems with N4 search
Moderators: hgm, Rebel, chrisw
-
M ANSARI
- Posts: 3707
- Joined: Thu Mar 16, 2006 7:10 pm
-
bob
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: N4 8core vs. R3 1core ... problems with N4 search
are you sure that N4 is not simply missing a key piece of knowledge dealing with white's candidate passed pawn on the queenside that is going to win the game when bishops come off? I don't think this is search, this is eval...
-
M ANSARI
- Posts: 3707
- Joined: Thu Mar 16, 2006 7:10 pm
Re: N4 8core vs. R3 1core ... problems with N4 search
Maybe you are right ... it is difficult to tell if it is evaluation or search, but on quite a few examples N4 will agree with R3 given enough time. Here are some more examples and I would appreciate your comments as I am curious also to know if it is a search problem or an evaluation problem.
Please note that I am doing a clip analysis of N4 on 8 cores 4Ghz while the clip analysis of R3 is a single core 4Ghz. Ofcourse if 8 cores of R3 would be used things would be even quicker for R3. But I want to highlight the fact that on some critical positions (not all positions) it seems to take way too long for N4 to see trouble.
Here Rxc6 looks very strong as it follows up with Qa4 ... but compare the times between 8 core N4 and 1 core R3.
4245: Naum 4 - Rybka 3 1-cpu _0_contempt, N4_R3_0_cont_5_1-1 2009
[d]2rr2k1/5p2/2B4p/P2P2p1/5qP1/1pR2P1P/1P6/1K2Q3 b - - 0 1
Analysis by Naum 4:
41...Qb4 42.Re3 Qc4 43.Qe2 Qb4 44.a6
+/- (1.01) Depth: 6/14 00:00:00 6kN
41...Qb4 42.Qe2 Qxa5 43.Rxb3 Rb8 44.Rxb8 Rxb8
= (0.14) Depth: 7/20 00:00:00 24kN
41...Qb4 42.Qd1 Qxa5 43.Qxb3 Kg7 44.Qc4 Qa7 45.f4
= (0.12) Depth: 8/22 00:00:00 33kN
41...Qb4 42.Qg1 Qxa5 43.Rxb3 Qd2 44.Qe3 Qg2 45.Rb7 Qxh3
= (0.02) Depth: 9/25 00:00:00 70kN
41...Qb4 42.Qe2 Qxa5 43.Rxb3 Rb8 44.Re3 Qc5 45.Re4 Rb3 46.Ka2
= (0.03) Depth: 10/28 00:00:00 253kN
41...Qb4 42.Qe3 Qxa5 43.Rxb3 Qa6 44.Qe1 Rb8 45.Re3 Qc4 46.Rc3 Qb4
= (0.05) Depth: 11/29 00:00:00 556kN
41...Qb4 42.Qe3 Qxa5 43.Rxb3 Qa6 44.Qe1 Rb8 45.Re3 Qc4 46.Rc3 Qb4 47.Qd2
= (0.05) Depth: 12/28 00:00:00 656kN
41...Qb4 42.Qe3 Qxa5 43.Rxb3 Qa6 44.Qe1 Rb8 45.Re3 Qc4 46.Qc1 Qa6 47.Qc5 Qf1+
= (0.04) Depth: 13/28 00:00:00 843kN
41...Qb4 42.Qe3 Qxa5 43.Rxb3 Qa6 44.Qe1 Rb8 45.Re3 Qc4 46.Rc3 Qb4 47.b3 Qd4 48.Qe3
= (0.03) Depth: 14/33 00:00:00 1365kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qd5 46.Qf3 Qd2 47.Qe4 Qh2 48.Rd3
= (-0.10) Depth: 14/36 00:00:01 4778kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Qd3+ 44.Kc1 Qxh3 45.d6 Rxd6 46.Qe8+ Kg7 47.Qe5+ Rf6 48.Qd4 Qg2
= (-0.19) Depth: 15/58 00:00:08 53139kN, tb=126
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Qd3+ 44.Kc1 Qxh3 45.d6 Rxd6 46.Qe8+ Kg7 47.Qe5+ Rf6 48.Qd4 Qg2 49.Kd1
= (-0.19) Depth: 16/47 00:00:09 56956kN, tb=130
41...Rxc6 42.Rxc6 Qa4 43.Qc3 Re8 44.d6 Qa2+ 45.Kc1 Qa1+ 46.Kd2 Qe1+ 47.Kd3 Re3+ 48.Kd4 Rxc3 49.Rxc3 Qd2+
-+ (-4.06) Depth: 17/62 00:01:38 772mN, tb=29281
(, 13.01.2009)
4245: Naum 4 - Rybka 3 1-cpu _0_contempt, N4_R3_0_cont_5_1-1 2009
[d]2rr2k1/5p2/2B4p/P2P2p1/5qP1/1pR2P1P/1P6/1K2Q3 b - - 0 1
Analysis by Rybka 3 1-cpu:
41...Rxc6
= (-0.20 !) Depth: 6 00:00:00 4kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Qxh3
=/+ (-0.30) Depth: 6 00:00:00 4kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Rxd5 44.Rxd5 Qxd5 45.Kc1
= (-0.07) Depth: 7 00:00:00 8kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Rxd5 44.Rxd5 Qxd5 45.Kc1 Qc4+ 46.Qc3
= (0.00) Depth: 8 00:00:00 12kN
41...Rxc6
= (-0.20 !) Depth: 9 00:00:00 33kN
41...Rxc6 42.Rxc6[] Qxf3[] 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qa6 46.Ra3 Qc4
=/+ (-0.39) Depth: 9 00:00:00 77kN
41...Rxc6 42.Rxc6[] Qxf3 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qa6 46.Ra3 Qc4
=/+ (-0.36) Depth: 10 00:00:00 95kN
41...Rxc6 42.Rxc6[] Qxf3 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qa6 46.Ra3 Qc4 47.Qb3
=/+ (-0.35) Depth: 11 00:00:01 141kN
41...Rxc6 42.Rxc6[] Qxf3 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qa6 46.Ra3 Qc4 47.Qb3 Qf4
=/+ (-0.32) Depth: 12 00:00:02 217kN
41...Rxc6
=/+ (-0.52 !) Depth: 13 00:00:04 509kN
41...Rxc6
-/+ (-0.72 !) Depth: 13 00:00:05 610kN
41...Rxc6
-/+ (-1.12 !) Depth: 13 00:00:07 845kN
41...Rxc6
-+ (-1.92 !) Depth: 13 00:00:09 1125kN
41...Rxc6 42.Rxc6[] Qa4 43.Qc3 Re8[] 44.d6 Qa2+[] 45.Kc1 Qa1+ 46.Kd2[] Qe1+ 47.Kd3 Re3+ 48.Kd4[] Rxc3[] 49.Rxc3 Qd2+ 50.Kc4 Qxd6 51.Rxb3 Qe5 52.Kb4 Qe2 53.Kc5 Kg7 54.Kb4
-+ (-2.34) Depth: 13 00:00:18 2112kN
(, 13.01.2009)
Here is another position ... although Bd3 is obvious, N4 takes a long time to see the forced critical line unlike even 1 core Rybka 3
4246: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
[d]2r3k1/p1qb4/4p3/1PNpPpb1/3Q2p1/4P1P1/2R2P2/5BK1 w - - 0 1
Analysis by Naum 4:
42.Bd3 Qd8 43.Qa4 Rc7 44.Qa5 Qc8
= (0.14) Depth: 6/15 00:00:00 4kN
42.Bd3 Qd8 43.Qa4 Rc7 44.Qa5 Qc8 45.Qb4
= (0.08) Depth: 7/16 00:00:00 6kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8
+/= (0.52) Depth: 8/19 00:00:00 16kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7
+/= (0.52) Depth: 9/33 00:00:00 62kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5
+/= (0.57) Depth: 10/31 00:00:00 145kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6
+/= (0.57) Depth: 11/29 00:00:00 242kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Qd6
+/= (0.59) Depth: 12/37 00:00:00 539kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7
+/= (0.58) Depth: 13/35 00:00:00 914kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Bd6
+/= (0.58) Depth: 14/41 00:00:00 1891kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Bd6 49.Qa6
+/= (0.58) Depth: 15/41 00:00:01 3950kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Bd6 49.Qa6 Be5
+/= (0.58) Depth: 16/41 00:00:01 6898kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Be5 49.Kg2 Kf8 50.Qa7
+/= (0.61) Depth: 17/43 00:00:03 14072kN, tb=2
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Be5 49.Kg2 Kf8 50.Bd1 d4
+/= (0.59) Depth: 18/49 00:00:06 29677kN, tb=5
42.Bd3 Kf7 43.Nxe6 Bxe6 44.Rxc7+ Rxc7 45.Kg2 Re7 46.Qc5 Bh6 47.Qc1 Kg6 48.Qd1 Kg7 49.Qh1 Bd7 50.Qa1 Be6 51.Qc3
+- (2.19) Depth: 19/62 00:03:33 1417mN, tb=48272
42.Bd3 Be7 43.Nxe6 Qb6 44.Rxc8+ Bxc8 45.Bxf5 Qxd4 46.Nxd4 Bxf5 47.Nxf5 Bg5 48.e6 Kf8 49.Kf1
+- (2.53) Depth: 20/48 00:03:44 1482mN, tb=48273
(, 13.01.2009)
4246: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
[d]2r3k1/p1qb4/4p3/1PNpPpb1/3Q2p1/4P1P1/2R2P2/5BK1 w - - 0 1
Analysis by Rybka 3 1-cpu:
42.Bd3 Qd8 43.Qa1
+/= (0.48) Depth: 6 00:00:00 2kN
42.Bd3
+/= (0.68 !) Depth: 7 00:00:00 5kN
42.Bd3
+/- (0.88 !) Depth: 7 00:00:00 8kN
42.Bd3
+/- (1.28 !) Depth: 7 00:00:00 12kN
42.Bd3 Qd8 43.Bxf5 exf5 44.Qxd5+
+/- (1.27) Depth: 7 00:00:00 13kN
42.Bd3 Qd8 43.Bxf5 exf5 44.Qxd5+
+/- (1.27) Depth: 8 00:00:00 15kN
42.Bd3 Qd8 43.Bxf5 exf5 44.Qxd5+ Kh8 45.Rd2 Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6
+/- (1.22) Depth: 9 00:00:00 36kN
42.Bd3
+- (1.42 !) Depth: 10 00:00:01 83kN
42.Bd3
+- (1.62 !) Depth: 10 00:00:01 119kN
42.Bd3
+- (2.02 !) Depth: 10 00:00:02 223kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7[] Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 10 00:00:02 247kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2 Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7[] Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 11 00:00:03 296kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2 Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7[] Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 12 00:00:04 408kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7[] Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 13 00:00:08 761kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7 Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 14 00:00:14 1373kN
42.Bd3
+- (2.20 !) Depth: 15 00:00:31 2877kN
42.Bd3
+- (2.40 !) Depth: 15 00:00:53 4786kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7[] Bf7 50.Rd7[] Qe6 51.Qc7[] Bxe3 52.Re7 Bxf2+ 53.Kxf2 Qd5 54.Qb8+ Kh7 55.Qe5
+- (2.40) Depth: 15 00:00:58 5395kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7[] Bf7 50.Rd7[] Qe6 51.Qc7[] Bxe3 52.Re7 Bxf2+ 53.Kxf2 Qd5 54.Qb8+ Kh7 55.Qe5
+- (2.40) Depth: 16 00:01:16 7225kN
(, 13.01.2009)
Yet another example ... Rybka has just played Kg5 and sees a strong plus. N4 takes a long time to agree
4256: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
[d]8/8/4k1b1/1p1p1pK1/p1pP4/P1P2PP1/N7/8 b - - 0 1
Analysis by Naum 4:
80...Be8 81.Nb4 Bd7 82.Kf4 Kd6 83.g4
= (0.00) Depth: 6/8 00:00:00 1kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Nc5 Be6
= (0.07) Depth: 7/12 00:00:00 1kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Nc5 Be6 84.Kf4
= (0.12) Depth: 8/12 00:00:00 2kN
80...Be8 81.Nb4 Bd7 82.Nc2 Kd6 83.Ne3 b4 84.cxb4 Kc6
+/= (0.58) Depth: 9/18 00:00:00 12kN
80...Be8 81.Nb4 Bd7 82.Nc2 Bc6 83.f4 b4 84.Nxb4 Bb5 85.g4
+/- (0.96) Depth: 10/25 00:00:00 52kN
80...Be8 81.Nb4 Bd7 82.Nc2 Bc6 83.f4 b4 84.Nxb4 Bb5 85.g4 fxg4
+/- (0.96) Depth: 11/17 00:00:00 54kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Nc2 Bd7 85.Kg6 b4 86.Nxb4
+/- (0.81) Depth: 12/23 00:00:00 79kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.Nc2 Bd7 84.Ne3 b4 85.cxb4 c3 86.g4 fxg4
+/= (0.29) Depth: 13/31 00:00:00 118kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.Nc2 Bd7 84.Ne3 b4 85.cxb4 c3 86.g4 fxg4 87.Kxg4
+/= (0.29) Depth: 14/15 00:00:00 122kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.g4 fxg4 84.f5+ Kd7 85.Kxg4 Kd6 86.Kf4 Bb7 87.f6 Ke6
+/= (0.43) Depth: 15/35 00:00:00 195kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.g4 fxg4 84.f5+ Kd7 85.Kxg4 Kd6 86.Kf4 Bb7 87.f6 Ke6 88.Kg5
+/= (0.43) Depth: 16/25 00:00:00 238kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.g4 fxg4 84.f5+ Kd6 85.Kxg4 Bd7 86.Kf4 Be8 87.f6 Ke6 88.Kg5 Kd6
+/= (0.58) Depth: 17/32 00:00:00 370kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Kg7 Bc8 85.Kf7 Bd7 86.Kg6 Ke6 87.Kh5 Bc8 88.Kg5 Bd7 89.Na6
+/= (0.57) Depth: 18/35 00:00:01 851kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Kg7 Bc8 85.Kf7 Bd7 86.Kg6 Ke6 87.Kh5 Bc8 88.Kg6
+/= (0.57) Depth: 19/28 00:00:01 1001kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Kg7 Bc8 85.Kf7 Bd7 86.Kg6 Ke6 87.Kh5 Bc8 88.Kg5 Bd7 89.Na6 Ke7 90.Kg6
+/= (0.57) Depth: 20/37 00:00:01 1486kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Kg7 Bc8 85.Kf7 Bd7 86.Kg6 Ke6 87.Kh5 Be8+ 88.Kh6
+/= (0.57) Depth: 21/31 00:00:02 2104kN
80...Be8 81.Nb4 Bd7 82.g4 fxg4 83.fxg4 Be8 84.Kh6 Bf7 85.Na6 Kf6 86.g5+ Kf5 87.Nb4 Be6 88.g6 Kf6 89.Nc2 Ke7 90.g7 Bg8 91.Kg6
+- (1.93) Depth: 22/62 00:01:07 107mN, tb=10099
(, 13.01.2009)
4256: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
[d]8/8/4k1b1/1p1p1pK1/p1pP4/P1P2PP1/N7/8 b - - 0 1
Analysis by Rybka 3 1-cpu:
80...Be8 81.Nb4 Bd7 82.Nc2 Be8
+/- (1.01) Depth: 6 00:00:00 1kN
80...Be8 81.Nb4 Bd7 82.Kf4 Kd6 83.Nc2
+/- (0.83) Depth: 7 00:00:00 2kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Kf4 Bc8 84.Nb4
+/- (0.81) Depth: 8 00:00:00 2kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Kf4 Bc8 84.Nb4
+/- (0.81) Depth: 9 00:00:00 2kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Kf4 Bc8 84.Nb4
+/- (0.81) Depth: 10 00:00:00 3kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Kg6 Bc8 84.Nb4 Bd7
+/- (0.73) Depth: 11 00:00:00 5kN
80...Be8 81.Nb4 Bd7 82.Na6 Kf7 83.g4 fxg4 84.fxg4 Bc8 85.Nb4 Be6 86.Kf4
+/- (0.85) Depth: 12 00:00:00 12kN
80...Be8 81.Nb4 Bd7 82.Na6 Kf7 83.g4 fxg4 84.fxg4 Bc8 85.Nb4 Bb7 86.Kf5 Bc8+ 87.Ke5 Bxg4 88.Kxd5
+/- (0.88) Depth: 13 00:00:00 16kN
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Na6 Be8 86.Nc7+ Kd6 87.Nxe8+ Kc6 88.g5 b4 89.axb4 a3
+- (1.94) Depth: 14 00:00:02 238kN
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Na6 Be8 86.Nc7+ Kd6 87.Nxe8+ Kc6 88.g5 b4 89.axb4 a3
+- (1.94) Depth: 15 00:00:03 332kN
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Na6 Be8 86.Nc7+ Kd6 87.Nxe8+ Kc6 88.g5 b4 89.axb4 a3
+- (1.94) Depth: 16 00:00:04 368kN
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Na6 Be8 86.Nc7+ Kd6 87.Nxe8+ Kc6[] 88.g5 b4[] 89.axb4 a3 90.g6 a2 91.g7 a1Q[] 92.g8Q Qxc3 93.Qe6+ Kb5 94.Nd6+ Kxb4 95.Qxd5 Qc1+
+- (1.99) Depth: 17 00:00:06 647kN
80...Be8
+- (1.79 !) Depth: 18 00:00:09 935kN, tb=5
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Kg7 Be8 86.Na6
+- (1.79) Depth: 18 00:00:09 972kN, tb=5
80...Be8 81.Nb4 Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6 Bf7 85.Kg7 Be8 86.Na6[] Bd7 87.Kg6 Be8+ 88.Kh6 Bf7 89.g5
+- (1.99) Depth: 19 00:00:14 1394kN, tb=5
80...Be8 81.Nb4 Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6 Bf7 85.Kg7 Be8 86.Na6[] Bd7 87.Kg6 Be8+ 88.Kh6 Bf7 89.g5
+- (1.99) Depth: 20 00:00:15 1506kN, tb=5
80...Be8 81.Nb4 Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6 Bf7 85.Kg7 Be8 86.Na6 Bd7 87.Kg6[] Bc6 88.Kh6 Be8 89.Nc7+
+- (2.12) Depth: 21 00:00:26 2635kN, tb=5
(, 13.01.2009)
Please note that I am doing a clip analysis of N4 on 8 cores 4Ghz while the clip analysis of R3 is a single core 4Ghz. Ofcourse if 8 cores of R3 would be used things would be even quicker for R3. But I want to highlight the fact that on some critical positions (not all positions) it seems to take way too long for N4 to see trouble.
Here Rxc6 looks very strong as it follows up with Qa4 ... but compare the times between 8 core N4 and 1 core R3.
4245: Naum 4 - Rybka 3 1-cpu _0_contempt, N4_R3_0_cont_5_1-1 2009
[d]2rr2k1/5p2/2B4p/P2P2p1/5qP1/1pR2P1P/1P6/1K2Q3 b - - 0 1
Analysis by Naum 4:
41...Qb4 42.Re3 Qc4 43.Qe2 Qb4 44.a6
+/- (1.01) Depth: 6/14 00:00:00 6kN
41...Qb4 42.Qe2 Qxa5 43.Rxb3 Rb8 44.Rxb8 Rxb8
= (0.14) Depth: 7/20 00:00:00 24kN
41...Qb4 42.Qd1 Qxa5 43.Qxb3 Kg7 44.Qc4 Qa7 45.f4
= (0.12) Depth: 8/22 00:00:00 33kN
41...Qb4 42.Qg1 Qxa5 43.Rxb3 Qd2 44.Qe3 Qg2 45.Rb7 Qxh3
= (0.02) Depth: 9/25 00:00:00 70kN
41...Qb4 42.Qe2 Qxa5 43.Rxb3 Rb8 44.Re3 Qc5 45.Re4 Rb3 46.Ka2
= (0.03) Depth: 10/28 00:00:00 253kN
41...Qb4 42.Qe3 Qxa5 43.Rxb3 Qa6 44.Qe1 Rb8 45.Re3 Qc4 46.Rc3 Qb4
= (0.05) Depth: 11/29 00:00:00 556kN
41...Qb4 42.Qe3 Qxa5 43.Rxb3 Qa6 44.Qe1 Rb8 45.Re3 Qc4 46.Rc3 Qb4 47.Qd2
= (0.05) Depth: 12/28 00:00:00 656kN
41...Qb4 42.Qe3 Qxa5 43.Rxb3 Qa6 44.Qe1 Rb8 45.Re3 Qc4 46.Qc1 Qa6 47.Qc5 Qf1+
= (0.04) Depth: 13/28 00:00:00 843kN
41...Qb4 42.Qe3 Qxa5 43.Rxb3 Qa6 44.Qe1 Rb8 45.Re3 Qc4 46.Rc3 Qb4 47.b3 Qd4 48.Qe3
= (0.03) Depth: 14/33 00:00:00 1365kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qd5 46.Qf3 Qd2 47.Qe4 Qh2 48.Rd3
= (-0.10) Depth: 14/36 00:00:01 4778kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Qd3+ 44.Kc1 Qxh3 45.d6 Rxd6 46.Qe8+ Kg7 47.Qe5+ Rf6 48.Qd4 Qg2
= (-0.19) Depth: 15/58 00:00:08 53139kN, tb=126
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Qd3+ 44.Kc1 Qxh3 45.d6 Rxd6 46.Qe8+ Kg7 47.Qe5+ Rf6 48.Qd4 Qg2 49.Kd1
= (-0.19) Depth: 16/47 00:00:09 56956kN, tb=130
41...Rxc6 42.Rxc6 Qa4 43.Qc3 Re8 44.d6 Qa2+ 45.Kc1 Qa1+ 46.Kd2 Qe1+ 47.Kd3 Re3+ 48.Kd4 Rxc3 49.Rxc3 Qd2+
-+ (-4.06) Depth: 17/62 00:01:38 772mN, tb=29281
(, 13.01.2009)
4245: Naum 4 - Rybka 3 1-cpu _0_contempt, N4_R3_0_cont_5_1-1 2009
[d]2rr2k1/5p2/2B4p/P2P2p1/5qP1/1pR2P1P/1P6/1K2Q3 b - - 0 1
Analysis by Rybka 3 1-cpu:
41...Rxc6
= (-0.20 !) Depth: 6 00:00:00 4kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Qxh3
=/+ (-0.30) Depth: 6 00:00:00 4kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Rxd5 44.Rxd5 Qxd5 45.Kc1
= (-0.07) Depth: 7 00:00:00 8kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Rxd5 44.Rxd5 Qxd5 45.Kc1 Qc4+ 46.Qc3
= (0.00) Depth: 8 00:00:00 12kN
41...Rxc6
= (-0.20 !) Depth: 9 00:00:00 33kN
41...Rxc6 42.Rxc6[] Qxf3[] 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qa6 46.Ra3 Qc4
=/+ (-0.39) Depth: 9 00:00:00 77kN
41...Rxc6 42.Rxc6[] Qxf3 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qa6 46.Ra3 Qc4
=/+ (-0.36) Depth: 10 00:00:00 95kN
41...Rxc6 42.Rxc6[] Qxf3 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qa6 46.Ra3 Qc4 47.Qb3
=/+ (-0.35) Depth: 11 00:00:01 141kN
41...Rxc6 42.Rxc6[] Qxf3 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qa6 46.Ra3 Qc4 47.Qb3 Qf4
=/+ (-0.32) Depth: 12 00:00:02 217kN
41...Rxc6
=/+ (-0.52 !) Depth: 13 00:00:04 509kN
41...Rxc6
-/+ (-0.72 !) Depth: 13 00:00:05 610kN
41...Rxc6
-/+ (-1.12 !) Depth: 13 00:00:07 845kN
41...Rxc6
-+ (-1.92 !) Depth: 13 00:00:09 1125kN
41...Rxc6 42.Rxc6[] Qa4 43.Qc3 Re8[] 44.d6 Qa2+[] 45.Kc1 Qa1+ 46.Kd2[] Qe1+ 47.Kd3 Re3+ 48.Kd4[] Rxc3[] 49.Rxc3 Qd2+ 50.Kc4 Qxd6 51.Rxb3 Qe5 52.Kb4 Qe2 53.Kc5 Kg7 54.Kb4
-+ (-2.34) Depth: 13 00:00:18 2112kN
(, 13.01.2009)
Here is another position ... although Bd3 is obvious, N4 takes a long time to see the forced critical line unlike even 1 core Rybka 3
4246: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
[d]2r3k1/p1qb4/4p3/1PNpPpb1/3Q2p1/4P1P1/2R2P2/5BK1 w - - 0 1
Analysis by Naum 4:
42.Bd3 Qd8 43.Qa4 Rc7 44.Qa5 Qc8
= (0.14) Depth: 6/15 00:00:00 4kN
42.Bd3 Qd8 43.Qa4 Rc7 44.Qa5 Qc8 45.Qb4
= (0.08) Depth: 7/16 00:00:00 6kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8
+/= (0.52) Depth: 8/19 00:00:00 16kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7
+/= (0.52) Depth: 9/33 00:00:00 62kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5
+/= (0.57) Depth: 10/31 00:00:00 145kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6
+/= (0.57) Depth: 11/29 00:00:00 242kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Qd6
+/= (0.59) Depth: 12/37 00:00:00 539kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7
+/= (0.58) Depth: 13/35 00:00:00 914kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Bd6
+/= (0.58) Depth: 14/41 00:00:00 1891kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Bd6 49.Qa6
+/= (0.58) Depth: 15/41 00:00:01 3950kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Bd6 49.Qa6 Be5
+/= (0.58) Depth: 16/41 00:00:01 6898kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Be5 49.Kg2 Kf8 50.Qa7
+/= (0.61) Depth: 17/43 00:00:03 14072kN, tb=2
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Be5 49.Kg2 Kf8 50.Bd1 d4
+/= (0.59) Depth: 18/49 00:00:06 29677kN, tb=5
42.Bd3 Kf7 43.Nxe6 Bxe6 44.Rxc7+ Rxc7 45.Kg2 Re7 46.Qc5 Bh6 47.Qc1 Kg6 48.Qd1 Kg7 49.Qh1 Bd7 50.Qa1 Be6 51.Qc3
+- (2.19) Depth: 19/62 00:03:33 1417mN, tb=48272
42.Bd3 Be7 43.Nxe6 Qb6 44.Rxc8+ Bxc8 45.Bxf5 Qxd4 46.Nxd4 Bxf5 47.Nxf5 Bg5 48.e6 Kf8 49.Kf1
+- (2.53) Depth: 20/48 00:03:44 1482mN, tb=48273
(, 13.01.2009)
4246: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
[d]2r3k1/p1qb4/4p3/1PNpPpb1/3Q2p1/4P1P1/2R2P2/5BK1 w - - 0 1
Analysis by Rybka 3 1-cpu:
42.Bd3 Qd8 43.Qa1
+/= (0.48) Depth: 6 00:00:00 2kN
42.Bd3
+/= (0.68 !) Depth: 7 00:00:00 5kN
42.Bd3
+/- (0.88 !) Depth: 7 00:00:00 8kN
42.Bd3
+/- (1.28 !) Depth: 7 00:00:00 12kN
42.Bd3 Qd8 43.Bxf5 exf5 44.Qxd5+
+/- (1.27) Depth: 7 00:00:00 13kN
42.Bd3 Qd8 43.Bxf5 exf5 44.Qxd5+
+/- (1.27) Depth: 8 00:00:00 15kN
42.Bd3 Qd8 43.Bxf5 exf5 44.Qxd5+ Kh8 45.Rd2 Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6
+/- (1.22) Depth: 9 00:00:00 36kN
42.Bd3
+- (1.42 !) Depth: 10 00:00:01 83kN
42.Bd3
+- (1.62 !) Depth: 10 00:00:01 119kN
42.Bd3
+- (2.02 !) Depth: 10 00:00:02 223kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7[] Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 10 00:00:02 247kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2 Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7[] Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 11 00:00:03 296kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2 Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7[] Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 12 00:00:04 408kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7[] Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 13 00:00:08 761kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7 Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 14 00:00:14 1373kN
42.Bd3
+- (2.20 !) Depth: 15 00:00:31 2877kN
42.Bd3
+- (2.40 !) Depth: 15 00:00:53 4786kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7[] Bf7 50.Rd7[] Qe6 51.Qc7[] Bxe3 52.Re7 Bxf2+ 53.Kxf2 Qd5 54.Qb8+ Kh7 55.Qe5
+- (2.40) Depth: 15 00:00:58 5395kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7[] Bf7 50.Rd7[] Qe6 51.Qc7[] Bxe3 52.Re7 Bxf2+ 53.Kxf2 Qd5 54.Qb8+ Kh7 55.Qe5
+- (2.40) Depth: 16 00:01:16 7225kN
(, 13.01.2009)
Yet another example ... Rybka has just played Kg5 and sees a strong plus. N4 takes a long time to agree
4256: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
[d]8/8/4k1b1/1p1p1pK1/p1pP4/P1P2PP1/N7/8 b - - 0 1
Analysis by Naum 4:
80...Be8 81.Nb4 Bd7 82.Kf4 Kd6 83.g4
= (0.00) Depth: 6/8 00:00:00 1kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Nc5 Be6
= (0.07) Depth: 7/12 00:00:00 1kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Nc5 Be6 84.Kf4
= (0.12) Depth: 8/12 00:00:00 2kN
80...Be8 81.Nb4 Bd7 82.Nc2 Kd6 83.Ne3 b4 84.cxb4 Kc6
+/= (0.58) Depth: 9/18 00:00:00 12kN
80...Be8 81.Nb4 Bd7 82.Nc2 Bc6 83.f4 b4 84.Nxb4 Bb5 85.g4
+/- (0.96) Depth: 10/25 00:00:00 52kN
80...Be8 81.Nb4 Bd7 82.Nc2 Bc6 83.f4 b4 84.Nxb4 Bb5 85.g4 fxg4
+/- (0.96) Depth: 11/17 00:00:00 54kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Nc2 Bd7 85.Kg6 b4 86.Nxb4
+/- (0.81) Depth: 12/23 00:00:00 79kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.Nc2 Bd7 84.Ne3 b4 85.cxb4 c3 86.g4 fxg4
+/= (0.29) Depth: 13/31 00:00:00 118kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.Nc2 Bd7 84.Ne3 b4 85.cxb4 c3 86.g4 fxg4 87.Kxg4
+/= (0.29) Depth: 14/15 00:00:00 122kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.g4 fxg4 84.f5+ Kd7 85.Kxg4 Kd6 86.Kf4 Bb7 87.f6 Ke6
+/= (0.43) Depth: 15/35 00:00:00 195kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.g4 fxg4 84.f5+ Kd7 85.Kxg4 Kd6 86.Kf4 Bb7 87.f6 Ke6 88.Kg5
+/= (0.43) Depth: 16/25 00:00:00 238kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.g4 fxg4 84.f5+ Kd6 85.Kxg4 Bd7 86.Kf4 Be8 87.f6 Ke6 88.Kg5 Kd6
+/= (0.58) Depth: 17/32 00:00:00 370kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Kg7 Bc8 85.Kf7 Bd7 86.Kg6 Ke6 87.Kh5 Bc8 88.Kg5 Bd7 89.Na6
+/= (0.57) Depth: 18/35 00:00:01 851kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Kg7 Bc8 85.Kf7 Bd7 86.Kg6 Ke6 87.Kh5 Bc8 88.Kg6
+/= (0.57) Depth: 19/28 00:00:01 1001kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Kg7 Bc8 85.Kf7 Bd7 86.Kg6 Ke6 87.Kh5 Bc8 88.Kg5 Bd7 89.Na6 Ke7 90.Kg6
+/= (0.57) Depth: 20/37 00:00:01 1486kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Kg7 Bc8 85.Kf7 Bd7 86.Kg6 Ke6 87.Kh5 Be8+ 88.Kh6
+/= (0.57) Depth: 21/31 00:00:02 2104kN
80...Be8 81.Nb4 Bd7 82.g4 fxg4 83.fxg4 Be8 84.Kh6 Bf7 85.Na6 Kf6 86.g5+ Kf5 87.Nb4 Be6 88.g6 Kf6 89.Nc2 Ke7 90.g7 Bg8 91.Kg6
+- (1.93) Depth: 22/62 00:01:07 107mN, tb=10099
(, 13.01.2009)
4256: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
[d]8/8/4k1b1/1p1p1pK1/p1pP4/P1P2PP1/N7/8 b - - 0 1
Analysis by Rybka 3 1-cpu:
80...Be8 81.Nb4 Bd7 82.Nc2 Be8
+/- (1.01) Depth: 6 00:00:00 1kN
80...Be8 81.Nb4 Bd7 82.Kf4 Kd6 83.Nc2
+/- (0.83) Depth: 7 00:00:00 2kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Kf4 Bc8 84.Nb4
+/- (0.81) Depth: 8 00:00:00 2kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Kf4 Bc8 84.Nb4
+/- (0.81) Depth: 9 00:00:00 2kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Kf4 Bc8 84.Nb4
+/- (0.81) Depth: 10 00:00:00 3kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Kg6 Bc8 84.Nb4 Bd7
+/- (0.73) Depth: 11 00:00:00 5kN
80...Be8 81.Nb4 Bd7 82.Na6 Kf7 83.g4 fxg4 84.fxg4 Bc8 85.Nb4 Be6 86.Kf4
+/- (0.85) Depth: 12 00:00:00 12kN
80...Be8 81.Nb4 Bd7 82.Na6 Kf7 83.g4 fxg4 84.fxg4 Bc8 85.Nb4 Bb7 86.Kf5 Bc8+ 87.Ke5 Bxg4 88.Kxd5
+/- (0.88) Depth: 13 00:00:00 16kN
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Na6 Be8 86.Nc7+ Kd6 87.Nxe8+ Kc6 88.g5 b4 89.axb4 a3
+- (1.94) Depth: 14 00:00:02 238kN
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Na6 Be8 86.Nc7+ Kd6 87.Nxe8+ Kc6 88.g5 b4 89.axb4 a3
+- (1.94) Depth: 15 00:00:03 332kN
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Na6 Be8 86.Nc7+ Kd6 87.Nxe8+ Kc6 88.g5 b4 89.axb4 a3
+- (1.94) Depth: 16 00:00:04 368kN
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Na6 Be8 86.Nc7+ Kd6 87.Nxe8+ Kc6[] 88.g5 b4[] 89.axb4 a3 90.g6 a2 91.g7 a1Q[] 92.g8Q Qxc3 93.Qe6+ Kb5 94.Nd6+ Kxb4 95.Qxd5 Qc1+
+- (1.99) Depth: 17 00:00:06 647kN
80...Be8
+- (1.79 !) Depth: 18 00:00:09 935kN, tb=5
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Kg7 Be8 86.Na6
+- (1.79) Depth: 18 00:00:09 972kN, tb=5
80...Be8 81.Nb4 Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6 Bf7 85.Kg7 Be8 86.Na6[] Bd7 87.Kg6 Be8+ 88.Kh6 Bf7 89.g5
+- (1.99) Depth: 19 00:00:14 1394kN, tb=5
80...Be8 81.Nb4 Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6 Bf7 85.Kg7 Be8 86.Na6[] Bd7 87.Kg6 Be8+ 88.Kh6 Bf7 89.g5
+- (1.99) Depth: 20 00:00:15 1506kN, tb=5
80...Be8 81.Nb4 Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6 Bf7 85.Kg7 Be8 86.Na6 Bd7 87.Kg6[] Bc6 88.Kh6 Be8 89.Nc7+
+- (2.12) Depth: 21 00:00:26 2635kN, tb=5
(, 13.01.2009)
-
Martin
Re: N4 8core vs. R3 1core ... problems with N4 search
The draw score in position 2 (after Qe5) is very strange. Anatoli (much weaker than Naum) instantly sees that White is totally lost.
Looks like Naum has a serious bug here. Possibly a hashing bug in combination with parallel search.
Do you get the same result if you let Naum run on one core?
Looks like Naum has a serious bug here. Possibly a hashing bug in combination with parallel search.
Do you get the same result if you let Naum run on one core?
-
M ANSARI
- Posts: 3707
- Joined: Thu Mar 16, 2006 7:10 pm
Re: N4 8core vs. R3 1core ... problems with N4 search
I don't have N4 loaded on my computer now but I wouldn't be surprised if some of these positions fail due to some sort of bug. In position 2 I was also surprised by the node search result but assumed that this was due to simply not finding a better move after 5 minutes.
-
bob
- Posts: 20943
- Joined: Mon Feb 27, 2006 7:30 pm
- Location: Birmingham, AL
Re: N4 8core vs. R3 1core ... problems with N4 search
Hard to say here. Crafty finds Rxc6 instantly and sticks with it. Whether this is a tactical oversight by N4 or whether it has some evaluation term that makes something else look better I can't guess...M ANSARI wrote:Maybe you are right ... it is difficult to tell if it is evaluation or search, but on quite a few examples N4 will agree with R3 given enough time. Here are some more examples and I would appreciate your comments as I am curious also to know if it is a search problem or an evaluation problem.
Please note that I am doing a clip analysis of N4 on 8 cores 4Ghz while the clip analysis of R3 is a single core 4Ghz. Ofcourse if 8 cores of R3 would be used things would be even quicker for R3. But I want to highlight the fact that on some critical positions (not all positions) it seems to take way too long for N4 to see trouble.
Here Rxc6 looks very strong as it follows up with Qa4 ... but compare the times between 8 core N4 and 1 core R3.
4245: Naum 4 - Rybka 3 1-cpu _0_contempt, N4_R3_0_cont_5_1-1 2009
[d]2rr2k1/5p2/2B4p/P2P2p1/5qP1/1pR2P1P/1P6/1K2Q3 b - - 0 1
Analysis by Naum 4:
41...Qb4 42.Re3 Qc4 43.Qe2 Qb4 44.a6
+/- (1.01) Depth: 6/14 00:00:00 6kN
41...Qb4 42.Qe2 Qxa5 43.Rxb3 Rb8 44.Rxb8 Rxb8
= (0.14) Depth: 7/20 00:00:00 24kN
41...Qb4 42.Qd1 Qxa5 43.Qxb3 Kg7 44.Qc4 Qa7 45.f4
= (0.12) Depth: 8/22 00:00:00 33kN
41...Qb4 42.Qg1 Qxa5 43.Rxb3 Qd2 44.Qe3 Qg2 45.Rb7 Qxh3
= (0.02) Depth: 9/25 00:00:00 70kN
41...Qb4 42.Qe2 Qxa5 43.Rxb3 Rb8 44.Re3 Qc5 45.Re4 Rb3 46.Ka2
= (0.03) Depth: 10/28 00:00:00 253kN
41...Qb4 42.Qe3 Qxa5 43.Rxb3 Qa6 44.Qe1 Rb8 45.Re3 Qc4 46.Rc3 Qb4
= (0.05) Depth: 11/29 00:00:00 556kN
41...Qb4 42.Qe3 Qxa5 43.Rxb3 Qa6 44.Qe1 Rb8 45.Re3 Qc4 46.Rc3 Qb4 47.Qd2
= (0.05) Depth: 12/28 00:00:00 656kN
41...Qb4 42.Qe3 Qxa5 43.Rxb3 Qa6 44.Qe1 Rb8 45.Re3 Qc4 46.Qc1 Qa6 47.Qc5 Qf1+
= (0.04) Depth: 13/28 00:00:00 843kN
41...Qb4 42.Qe3 Qxa5 43.Rxb3 Qa6 44.Qe1 Rb8 45.Re3 Qc4 46.Rc3 Qb4 47.b3 Qd4 48.Qe3
= (0.03) Depth: 14/33 00:00:00 1365kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qd5 46.Qf3 Qd2 47.Qe4 Qh2 48.Rd3
= (-0.10) Depth: 14/36 00:00:01 4778kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Qd3+ 44.Kc1 Qxh3 45.d6 Rxd6 46.Qe8+ Kg7 47.Qe5+ Rf6 48.Qd4 Qg2
= (-0.19) Depth: 15/58 00:00:08 53139kN, tb=126
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Qd3+ 44.Kc1 Qxh3 45.d6 Rxd6 46.Qe8+ Kg7 47.Qe5+ Rf6 48.Qd4 Qg2 49.Kd1
= (-0.19) Depth: 16/47 00:00:09 56956kN, tb=130
41...Rxc6 42.Rxc6 Qa4 43.Qc3 Re8 44.d6 Qa2+ 45.Kc1 Qa1+ 46.Kd2 Qe1+ 47.Kd3 Re3+ 48.Kd4 Rxc3 49.Rxc3 Qd2+
-+ (-4.06) Depth: 17/62 00:01:38 772mN, tb=29281
If it misses an outright mate, that has to be a search issue, otherwise much of this is just positional judgement.
Crafty chooses the move instantly, but it takes about 30 seconds for it to fail high to +3 from +1.5...
(, 13.01.2009)
4245: Naum 4 - Rybka 3 1-cpu _0_contempt, N4_R3_0_cont_5_1-1 2009
[d]2rr2k1/5p2/2B4p/P2P2p1/5qP1/1pR2P1P/1P6/1K2Q3 b - - 0 1
Analysis by Rybka 3 1-cpu:
41...Rxc6
= (-0.20 !) Depth: 6 00:00:00 4kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Qxh3
=/+ (-0.30) Depth: 6 00:00:00 4kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Rxd5 44.Rxd5 Qxd5 45.Kc1
= (-0.07) Depth: 7 00:00:00 8kN
41...Rxc6 42.Rxc6 Qxf3 43.Rc5 Rxd5 44.Rxd5 Qxd5 45.Kc1 Qc4+ 46.Qc3
= (0.00) Depth: 8 00:00:00 12kN
41...Rxc6
= (-0.20 !) Depth: 9 00:00:00 33kN
41...Rxc6 42.Rxc6[] Qxf3[] 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qa6 46.Ra3 Qc4
=/+ (-0.39) Depth: 9 00:00:00 77kN
41...Rxc6 42.Rxc6[] Qxf3 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qa6 46.Ra3 Qc4
=/+ (-0.36) Depth: 10 00:00:00 95kN
41...Rxc6 42.Rxc6[] Qxf3 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qa6 46.Ra3 Qc4 47.Qb3
=/+ (-0.35) Depth: 11 00:00:01 141kN
41...Rxc6 42.Rxc6[] Qxf3 43.Rc3 Qxd5 44.Qe3 Qxa5 45.Rxb3 Qa6 46.Ra3 Qc4 47.Qb3 Qf4
=/+ (-0.32) Depth: 12 00:00:02 217kN
41...Rxc6
=/+ (-0.52 !) Depth: 13 00:00:04 509kN
41...Rxc6
-/+ (-0.72 !) Depth: 13 00:00:05 610kN
41...Rxc6
-/+ (-1.12 !) Depth: 13 00:00:07 845kN
41...Rxc6
-+ (-1.92 !) Depth: 13 00:00:09 1125kN
41...Rxc6 42.Rxc6[] Qa4 43.Qc3 Re8[] 44.d6 Qa2+[] 45.Kc1 Qa1+ 46.Kd2[] Qe1+ 47.Kd3 Re3+ 48.Kd4[] Rxc3[] 49.Rxc3 Qd2+ 50.Kc4 Qxd6 51.Rxb3 Qe5 52.Kb4 Qe2 53.Kc5 Kg7 54.Kb4
-+ (-2.34) Depth: 13 00:00:18 2112kN
(, 13.01.2009)
Here is another position ... although Bd3 is obvious, N4 takes a long time to see the forced critical line unlike even 1 core Rybka 3
4246: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
[d]2r3k1/p1qb4/4p3/1PNpPpb1/3Q2p1/4P1P1/2R2P2/5BK1 w - - 0 1
Analysis by Naum 4:
42.Bd3 Qd8 43.Qa4 Rc7 44.Qa5 Qc8
= (0.14) Depth: 6/15 00:00:00 4kN
42.Bd3 Qd8 43.Qa4 Rc7 44.Qa5 Qc8 45.Qb4
= (0.08) Depth: 7/16 00:00:00 6kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8
+/= (0.52) Depth: 8/19 00:00:00 16kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7
+/= (0.52) Depth: 9/33 00:00:00 62kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5
+/= (0.57) Depth: 10/31 00:00:00 145kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6
+/= (0.57) Depth: 11/29 00:00:00 242kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Qd6
+/= (0.59) Depth: 12/37 00:00:00 539kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7
+/= (0.58) Depth: 13/35 00:00:00 914kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Bd6
+/= (0.58) Depth: 14/41 00:00:00 1891kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Bd6 49.Qa6
+/= (0.58) Depth: 15/41 00:00:01 3950kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Bd6 49.Qa6 Be5
+/= (0.58) Depth: 16/41 00:00:01 6898kN
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Be5 49.Kg2 Kf8 50.Qa7
+/= (0.61) Depth: 17/43 00:00:03 14072kN, tb=2
42.Bd3 Qd8 43.Nxd7 Rxc2 44.Nf6+ Bxf6 45.Bxc2 Bh8 46.Qxa7 Bxe5 47.b6 Bb8 48.Qb7 Be5 49.Kg2 Kf8 50.Bd1 d4
+/= (0.59) Depth: 18/49 00:00:06 29677kN, tb=5
42.Bd3 Kf7 43.Nxe6 Bxe6 44.Rxc7+ Rxc7 45.Kg2 Re7 46.Qc5 Bh6 47.Qc1 Kg6 48.Qd1 Kg7 49.Qh1 Bd7 50.Qa1 Be6 51.Qc3
+- (2.19) Depth: 19/62 00:03:33 1417mN, tb=48272
42.Bd3 Be7 43.Nxe6 Qb6 44.Rxc8+ Bxc8 45.Bxf5 Qxd4 46.Nxd4 Bxf5 47.Nxf5 Bg5 48.e6 Kf8 49.Kf1
+- (2.53) Depth: 20/48 00:03:44 1482mN, tb=48273
(, 13.01.2009)
4246: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
[d]2r3k1/p1qb4/4p3/1PNpPpb1/3Q2p1/4P1P1/2R2P2/5BK1 w - - 0 1
Analysis by Rybka 3 1-cpu:
42.Bd3 Qd8 43.Qa1
+/= (0.48) Depth: 6 00:00:00 2kN
42.Bd3
+/= (0.68 !) Depth: 7 00:00:00 5kN
42.Bd3
+/- (0.88 !) Depth: 7 00:00:00 8kN
42.Bd3
+/- (1.28 !) Depth: 7 00:00:00 12kN
42.Bd3 Qd8 43.Bxf5 exf5 44.Qxd5+
+/- (1.27) Depth: 7 00:00:00 13kN
42.Bd3 Qd8 43.Bxf5 exf5 44.Qxd5+
+/- (1.27) Depth: 8 00:00:00 15kN
42.Bd3 Qd8 43.Bxf5 exf5 44.Qxd5+ Kh8 45.Rd2 Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6
+/- (1.22) Depth: 9 00:00:00 36kN
42.Bd3
+- (1.42 !) Depth: 10 00:00:01 83kN
42.Bd3
+- (1.62 !) Depth: 10 00:00:01 119kN
42.Bd3
+- (2.02 !) Depth: 10 00:00:02 223kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7[] Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 10 00:00:02 247kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2 Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7[] Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 11 00:00:03 296kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2 Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7[] Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 12 00:00:04 408kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7[] Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 13 00:00:08 761kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7 Bf7 50.Rd7 Bd8 51.Rd1 Bb3 52.Qd7 Qxd7 53.Rxd7 Bb6 54.Rd6 Bc7
+- (2.00) Depth: 14 00:00:14 1373kN
42.Bd3
+- (2.20 !) Depth: 15 00:00:31 2877kN
42.Bd3
+- (2.40 !) Depth: 15 00:00:53 4786kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7[] Bf7 50.Rd7[] Qe6 51.Qc7[] Bxe3 52.Re7 Bxf2+ 53.Kxf2 Qd5 54.Qb8+ Kh7 55.Qe5
+- (2.40) Depth: 15 00:00:58 5395kN
42.Bd3 Qd8 43.Bxf5[] exf5 44.Qxd5+[] Kh8 45.Rd2[] Rxc5 46.Qxc5 Qe8 47.e6 Bxe6 48.Rd6[] Kg8 49.Qxa7[] Bf7 50.Rd7[] Qe6 51.Qc7[] Bxe3 52.Re7 Bxf2+ 53.Kxf2 Qd5 54.Qb8+ Kh7 55.Qe5
+- (2.40) Depth: 16 00:01:16 7225kN
This appears to be an evaluation issue as well. Crafty finds that black is better instantly and very quickly settles on Be8 as the best move also...
(, 13.01.2009)
Yet another example ... Rybka has just played Kg5 and sees a strong plus. N4 takes a long time to agree
4256: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
[d]8/8/4k1b1/1p1p1pK1/p1pP4/P1P2PP1/N7/8 b - - 0 1
Analysis by Naum 4:
80...Be8 81.Nb4 Bd7 82.Kf4 Kd6 83.g4
= (0.00) Depth: 6/8 00:00:00 1kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Nc5 Be6
= (0.07) Depth: 7/12 00:00:00 1kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Nc5 Be6 84.Kf4
= (0.12) Depth: 8/12 00:00:00 2kN
80...Be8 81.Nb4 Bd7 82.Nc2 Kd6 83.Ne3 b4 84.cxb4 Kc6
+/= (0.58) Depth: 9/18 00:00:00 12kN
80...Be8 81.Nb4 Bd7 82.Nc2 Bc6 83.f4 b4 84.Nxb4 Bb5 85.g4
+/- (0.96) Depth: 10/25 00:00:00 52kN
80...Be8 81.Nb4 Bd7 82.Nc2 Bc6 83.f4 b4 84.Nxb4 Bb5 85.g4 fxg4
+/- (0.96) Depth: 11/17 00:00:00 54kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Nc2 Bd7 85.Kg6 b4 86.Nxb4
+/- (0.81) Depth: 12/23 00:00:00 79kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.Nc2 Bd7 84.Ne3 b4 85.cxb4 c3 86.g4 fxg4
+/= (0.29) Depth: 13/31 00:00:00 118kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.Nc2 Bd7 84.Ne3 b4 85.cxb4 c3 86.g4 fxg4 87.Kxg4
+/= (0.29) Depth: 14/15 00:00:00 122kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.g4 fxg4 84.f5+ Kd7 85.Kxg4 Kd6 86.Kf4 Bb7 87.f6 Ke6
+/= (0.43) Depth: 15/35 00:00:00 195kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.g4 fxg4 84.f5+ Kd7 85.Kxg4 Kd6 86.Kf4 Bb7 87.f6 Ke6 88.Kg5
+/= (0.43) Depth: 16/25 00:00:00 238kN
80...Be8 81.Nb4 Bd7 82.f4 Bc8 83.g4 fxg4 84.f5+ Kd6 85.Kxg4 Bd7 86.Kf4 Be8 87.f6 Ke6 88.Kg5 Kd6
+/= (0.58) Depth: 17/32 00:00:00 370kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Kg7 Bc8 85.Kf7 Bd7 86.Kg6 Ke6 87.Kh5 Bc8 88.Kg5 Bd7 89.Na6
+/= (0.57) Depth: 18/35 00:00:01 851kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Kg7 Bc8 85.Kf7 Bd7 86.Kg6 Ke6 87.Kh5 Bc8 88.Kg6
+/= (0.57) Depth: 19/28 00:00:01 1001kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Kg7 Bc8 85.Kf7 Bd7 86.Kg6 Ke6 87.Kh5 Bc8 88.Kg5 Bd7 89.Na6 Ke7 90.Kg6
+/= (0.57) Depth: 20/37 00:00:01 1486kN
80...Be8 81.Nb4 Bd7 82.f4 Kd6 83.Kf6 Be6 84.Kg7 Bc8 85.Kf7 Bd7 86.Kg6 Ke6 87.Kh5 Be8+ 88.Kh6
+/= (0.57) Depth: 21/31 00:00:02 2104kN
80...Be8 81.Nb4 Bd7 82.g4 fxg4 83.fxg4 Be8 84.Kh6 Bf7 85.Na6 Kf6 86.g5+ Kf5 87.Nb4 Be6 88.g6 Kf6 89.Nc2 Ke7 90.g7 Bg8 91.Kg6
+- (1.93) Depth: 22/62 00:01:07 107mN, tb=10099
(, 13.01.2009)
4256: Rybka 3 1-cpu _0_contempt - Naum 4, N4_R3_0_cont_5_1-1 2009
[d]8/8/4k1b1/1p1p1pK1/p1pP4/P1P2PP1/N7/8 b - - 0 1
Analysis by Rybka 3 1-cpu:
80...Be8 81.Nb4 Bd7 82.Nc2 Be8
+/- (1.01) Depth: 6 00:00:00 1kN
80...Be8 81.Nb4 Bd7 82.Kf4 Kd6 83.Nc2
+/- (0.83) Depth: 7 00:00:00 2kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Kf4 Bc8 84.Nb4
+/- (0.81) Depth: 8 00:00:00 2kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Kf4 Bc8 84.Nb4
+/- (0.81) Depth: 9 00:00:00 2kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Kf4 Bc8 84.Nb4
+/- (0.81) Depth: 10 00:00:00 3kN
80...Be8 81.Nb4 Bd7 82.Na6 Kd6 83.Kg6 Bc8 84.Nb4 Bd7
+/- (0.73) Depth: 11 00:00:00 5kN
80...Be8 81.Nb4 Bd7 82.Na6 Kf7 83.g4 fxg4 84.fxg4 Bc8 85.Nb4 Be6 86.Kf4
+/- (0.85) Depth: 12 00:00:00 12kN
80...Be8 81.Nb4 Bd7 82.Na6 Kf7 83.g4 fxg4 84.fxg4 Bc8 85.Nb4 Bb7 86.Kf5 Bc8+ 87.Ke5 Bxg4 88.Kxd5
+/- (0.88) Depth: 13 00:00:00 16kN
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Na6 Be8 86.Nc7+ Kd6 87.Nxe8+ Kc6 88.g5 b4 89.axb4 a3
+- (1.94) Depth: 14 00:00:02 238kN
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Na6 Be8 86.Nc7+ Kd6 87.Nxe8+ Kc6 88.g5 b4 89.axb4 a3
+- (1.94) Depth: 15 00:00:03 332kN
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Na6 Be8 86.Nc7+ Kd6 87.Nxe8+ Kc6 88.g5 b4 89.axb4 a3
+- (1.94) Depth: 16 00:00:04 368kN
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Na6 Be8 86.Nc7+ Kd6 87.Nxe8+ Kc6[] 88.g5 b4[] 89.axb4 a3 90.g6 a2 91.g7 a1Q[] 92.g8Q Qxc3 93.Qe6+ Kb5 94.Nd6+ Kxb4 95.Qxd5 Qc1+
+- (1.99) Depth: 17 00:00:06 647kN
80...Be8
+- (1.79 !) Depth: 18 00:00:09 935kN, tb=5
80...Be8 81.Nb4[] Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6[] Bf7 85.Kg7 Be8 86.Na6
+- (1.79) Depth: 18 00:00:09 972kN, tb=5
80...Be8 81.Nb4 Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6 Bf7 85.Kg7 Be8 86.Na6[] Bd7 87.Kg6 Be8+ 88.Kh6 Bf7 89.g5
+- (1.99) Depth: 19 00:00:14 1394kN, tb=5
80...Be8 81.Nb4 Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6 Bf7 85.Kg7 Be8 86.Na6[] Bd7 87.Kg6 Be8+ 88.Kh6 Bf7 89.g5
+- (1.99) Depth: 20 00:00:15 1506kN, tb=5
80...Be8 81.Nb4 Bd7 82.g4[] fxg4 83.fxg4 Be8 84.Kh6 Bf7 85.Kg7 Be8 86.Na6 Bd7 87.Kg6[] Bc6 88.Kh6 Be8 89.Nc7+
+- (2.12) Depth: 21 00:00:26 2635kN, tb=5
(, 13.01.2009)
-
M ANSARI
- Posts: 3707
- Joined: Thu Mar 16, 2006 7:10 pm
Re: N4 8core vs. R3 1core ... problems with N4 search
I have to admit since I have no idea how to program an engine and have never even looked at source code ... so I will take your word for it. But I think Rybka has code that focuses its search and thus makes it by far the most efficient searcher. I am not sure if that goes down as "evaluation" or "search efficiency". For me I think of evaluation as a "static evaluation" where no tactical shots are present within a few ply. I guess evaluation could also include a "probing evaluation" which tells an engine which line could be dangerous or critical and thus it will give priority in search for those lines evaluated as critical.
IMHO N4 has an extremely strong "static" evaluation. I can see that in several thousand games I have looked at. This is also evident in its excellent scores against Rybka and non Rybka engines. For example it evaluates a bishop pair very high ... and although that evaluation seems primitive in a sense by being high in all positions ... I think in open positions where bishops are not impeded that its evaluation is actually more accurate than Rybka. Many of N4 wins on equal hardware against R3 is due to that very fact. I also believe that N4 would win many more of those games if it were able to search on par to Rybka. I am analyzing a few such games and will post on it in another thread.
IMHO N4 has an extremely strong "static" evaluation. I can see that in several thousand games I have looked at. This is also evident in its excellent scores against Rybka and non Rybka engines. For example it evaluates a bishop pair very high ... and although that evaluation seems primitive in a sense by being high in all positions ... I think in open positions where bishops are not impeded that its evaluation is actually more accurate than Rybka. Many of N4 wins on equal hardware against R3 is due to that very fact. I also believe that N4 would win many more of those games if it were able to search on par to Rybka. I am analyzing a few such games and will post on it in another thread.