Is Naum 4 a serious challenge for Rybka

[mclane]

2 things that N4 evaluates better are the value of bishop pair ... and also the value of 2 rooks vs. queen

this is because larry kaufman believes those 2 things are not that much important.

[M ANSARI]

Here is a game from many games played in my tourneys.

(12) Naum 4 - Rybka 3 [B92]
N4_60_1_7move–1 Home (15.1), 02.01.2009
[0.34;0.00]
Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz 3150 MHzW=20.1 ply; 4,049kN/s; 12,515 TBAs; HS-7moves.ctgB=17.7 ply; 202kN/s; 3 TBAs; HS-7moves.ctg 1.e4 B/0 0 c5 B/0 0 2.Nf3 B/0 0 d6 B/0 0 3.d4 B/0 0 cxd4 B/0 0 4.Nxd4 B/0 0 Nf6 B/0 0 5.Nc3 B/0 0 a6 B/0 0 6.Be3 B/0 0 e5 B/0 0 7.Nb3 B/0 0 Be6 B/0 0 8.Be2 0.34/19 1:41 Be7 0.00/18 1:02 9.0–0 0.28/20 1:56 0–0 0.01/18 1:25 10.Qd3 0.28/20 1:26 Nc6 -0.02/18 1:06 11.a3 0.27/21 1:32 Rc8 (h6) 0.00/18 1:06 12.Rfd1 (h3) 0.26/20 1:09 12...h6 -0.05/18 1:03 13.h3 0.21/21 2:31 Na5 0.00/18 28 14.Nxa5 0.16/20 30 Qxa5 0.00/16 0 15.a4 (Rac1) 0.16/21 1:16 15...Qb4 (Rc4) -0.04/18 1:00 16.Bd2 0.22/21 1:21 Bc4 0.00/19 1:18 17.Qf3 0.21/22 1:42 Bxe2 0.00/19 43 18.Qxe2 0.21/23 52 Qc4 (Bd8) 0.00/19 24 19.Qf3 0.26/20 28 Rfd8 (b6) 0.00/19 1:21 20.b3 (Be3) 0.19/19 1:09 20...Qe6 (Qc6) 0.05/18 47 21.Bc1 0.10/21 1:14 Rc6 (d5) 0.00/18 52 22.Ba3 0.28/21 1:11 Rdc8 (Nd7) 0.09/18 1:04 23.Bb4 0.23/21 1:03 Qd7 (b6) 0.11/17 1:27 24.a5 0.39/20 3:59 Qe8 (Qe6) 0.12/18 1:16 25.Rd3 (Ra2) 0.41/21 57 25...Qf8 (g6) 0.12/18 56 26.Qe2 (Rad1) 0.49/21 1:01 26...Qd8 (R8c7) 0.12/18 26 27.Ra2 0.48/21 53 Qc7 0.12/19 26 28.Rg3 0.48/21 52 Qd7 0.12/19 37 29.Re3 0.47/21 1:37 Qc7 (Bd8) 0.13/18 25 30.Rd3 (Qd3) 0.48/21 45 30...Bd8 0.13/18 47 31.Kh1 (Nd5) 0.48/22 1:30 31...Be7 0.13/19 57 32.Kh2 (Re3) 0.48/22 27 32...Bf8 (h5) 0.13/19 58 33.Qd1 (Rf3) 0.48/21 1:34 33...Qd7 (Be7) 0.13/17 31 34.Kg1 0.48/21 44 Qc7 0.28/18 35 35.Qd2 0.48/22 44 Qb8 (h5) 0.26/18 40 36.Qe2 0.49/21 1:03 Qc7 0.22/19 47 37.Rg3 (Rf3) 0.48/22 30 37...Be7 0.13/17 18 38.Qd2 (Rf3) 0.48/22 50 38...Bf8 (Kh8) 0.22/19 43 39.Re3 (Qd3) 0.47/21 35 39...Be7 0.15/19 1:38 40.Rf3 (Qd1) 0.47/21 20 40...Qd7 0.13/18 35 41.Qe2 (Qd3) 0.47/21 20 41...Bd8 0.13/19 50 42.Qd3 0.46/21 41 Qe7 (R6c7) 0.20/18 1:11 43.Nd1 0.48/20 1:12 b5 (Nh5) 0.19/18 51 44.axb6 0.51/20 27 Bxb6 0.23/19 49 45.c4 0.50/20 16 Bd4 0.22/18 29 46.Ne3 0.46/21 50 Bxe3 0.15/18 20 47.Rxe3 0.47/21 53 Rb8 0.20/19 44 48.Bd2 0.38/20 37 Rcb6 (h5) 0.13/18 20 49.Qc2 0.38/21 35 Nd7 0.13/19 37 50.Ba5 0.30/21 45 Rc6 (R6b7) 0.05/20 2:31 51.Qe2 (Qd3) 0.34/20 25 51...Nc5 0.05/20 42 52.Qd1 0.33/19 23 Kh7 (Qd7) 0.05/19 1:10 53.Ra3 (Rd2) 0.32/18 24 53...Rcc8 (Ne6) 0.05/17 1:00 54.Qc2 (Bc3) 0.32/18 24 54...Kg8 (Qb7) 0.02/17 21 55.Bc3 0.29/19 27 Qc7 (Qb7) 0.02/17 24 56.Bd2 0.25/19 28 Qc6 0.02/17 22 57.Qb1 (Kh2) 0.22/19 44 57...Rb6 (Rb7) 0.00/18 42 58.Ba5 0.12/18 16 Rbb8 (Rb7) 0.01/18 34 59.Kh2 (Be1) 0.17/19 17 59...Ne6 (Rb7) 0.01/17 21 60.Qa2 (Qd1) 0.19/19 46 60...Qb7 (Rb7) 0.02/17 38 61.Bd2 0.23/17 10 Rc6 0.03/17 12 62.Qa1 (Qc2) 0.20/18 18 62...Qb6 (Nd4) 0.02/17 18 63.Kg1 (Qd1) 0.23/18 11 63...Qd4 0.04/17 21 64.Bc3 (Qa2) 0.24/20 30 64...Qc5 0.03/17 35 65.Qa2 0.23/19 24 Qb6 (h5) 0.03/18 48 66.Rg3 (Ba5) 0.26/19 19 66...Kh8 (Nf4) 0.03/15 20 67.Qe2 (Bd2) 0.39/18 17 67...Qc5 (Qd8) 0.11/16 15 68.Ra5 0.39/18 6 Qa7 0.10/17 15 69.Bd2 0.32/18 8 Qe7 0.06/16 26 70.Qh5 0.33/19 13 Qf6 0.11/17 18 71.Qd1 (Rf3) 0.33/19 9 71...Nc5 0.02/17 18 72.Qc2 (Qb1) 0.33/19 13 72...Qh4 (Qe7) 0.02/17 36 73.Rg4

[d]1r5k/5pp1/p1rp3p/R1n1p3/2P1P1Rq/1P5P/2QB1PP1/6K1 b - - 0 73

Things are pretty even and looks like N4 is in no danger here ... slowly though R3 starts some counterplay taking advantage of white g rook that seems a little out of play

0.33/19 10 73...Qe7 (Qd8) -0.07/18 23 74.Ra3 0.14/18 18 Rcb6 -0.09/17 1:05 75.b4 0.04/18 29 Ne6 -0.10/17 17 76.Qd3 (Kh2) 0.04/18 17 76...Rc6 (Nd4) -0.10/16 11 77.Qd5 (Kh2) 0.01/17 6 77...Qc7 -0.23/15 23 78.Kh2

[d]1r5k/2q2pp1/p1rpn2p/3Qp3/1PP1P1R1/R6P/3B1PPK/8 b - - 0 78

N4 gives up its c pawn for black's a pawn ... still nothing serious happening

0.00/16 17 78...Rxc4 -0.15/15 9 79.Rxa6 0.00/18 5 Rd4 -0.17/16 7 80.Qa2 0.00/19 8 Kh7 (h5) -0.03/16 42 81.Rg3 0.00/16 6 Rb7 (Nf4) -0.04/16 30 82.Rga3 0.00/16 11 Nf4 -0.04/15 6 83.R3a4 (f3) 0.05/17 10 83...Qd7 (Qc4) -0.02/15 15 84.Bxf4 0.16/18 7 exf4 0.00/16 5 85.Qc2 0.16/18 6

[d]8/1r1q1ppk/R2p3p/8/RP1rPp2/7P/2Q2PPK/8 b - - 0 85

N4 likes its position here but ...

85...f3!

[d]8/1r1q1ppk/R2p3p/8/RP1rP3/5p1P/2Q2PPK/8 w - - 0 86

R3 sacs a pawn for tactical compensation ... but really it is not clear yet what the compensation is but this move is full of venom if not handled correctly

-0.06/15 9 86.gxf3 (Qc3) 0.10/17 5 86...Rc7 (Qe6) -0.21/16 19 87.Qb2 (Qe2) 0.00/17 9 87...Rcc4 -0.21/14 5 88.Qb1

[d]8/3q1ppk/R2p3p/8/RPrrP3/5P1P/5P1K/1Q6 b - - 0 88

N4 is oblivious to any danger ... but Rybka already sees something with evaluation increasing

0.00/18 6 88...Qe7 (Qe6) -0.43/15 11 89.R6a5

[d]8/4qppk/3p3p/R7/RPrrP3/5P1P/5P1K/1Q6 b - - 0 89

Again N4 search sees nothing, yet it might be already totally lost ... in the meantime R3 search smells blood and evaluation jumps to –1.54

Here is N4 analysis on Octa 4ghz @ 5 minutes

12: Naum 4 - Rybka 3, N4_60_1_7move-1 2008
8/4qppk/3p3p/R7/RPrrP3/5P1P/5P1K/1Q6 b - - 0 1

Analysis by Naum 4:

89...Qh4 90.Qe1 Rc3 91.Kg2 Rc2 92.Qe3
= (0.18) Depth: 6/20 00:00:00 20kN
89...Qh4 90.Qg1 Rd2 91.Ra2 Rxb4 92.Rxd2 Qf4+
= (0.06) Depth: 7/18 00:00:00 31kN
89...Qh4 90.Qg1 Rd2 91.Ra2 Rxb4 92.Qg3 Qxg3+ 93.Kxg3
= (0.08) Depth: 8/19 00:00:00 39kN
89...Qh4 90.Qg1 Rd3 91.Ra3 Rxa3 92.Rxa3 Rxb4 93.Qg3 Qe7
= (0.15) Depth: 9/19 00:00:00 67kN
89...Qh4 90.Qg1 Rd3 91.Ra3 Rxa3 92.Rxa3 Rxb4 93.Qg3 Qxg3+ 94.Kxg3
= (0.11) Depth: 10/21 00:00:00 99kN
89...Qh4 90.Qg1 Rd2 91.Ra2 Rxb4 92.Qg3 Rxa2 93.Rxa2 Qe7 94.Qf4 g5
= (0.09) Depth: 11/25 00:00:00 329kN
89...Qh4 90.Qg1 Rd2 91.Ra2 Rxb4 92.Qg3 Rxa2 93.Rxa2 Qe7 94.Ra6 Rd4 95.Ra5
= (0.10) Depth: 12/28 00:00:00 500kN
89...Qh4 90.Qg1 Rd2 91.Ra2 Rxb4 92.Qg3 Rxa2 93.Rxa2 Qe7 94.Qg4 g5 95.Kg2 Kg7
= (0.07) Depth: 13/27 00:00:00 770kN
89...Qh4 90.Qg1 Rd2 91.Ra2 Rxb4 92.Rxd2 Qf4+ 93.Qg3 Qxd2 94.Ra7 Kg8 95.Kg2 g5 96.h4
= (0.06) Depth: 14/35 00:00:00 1386kN
89...Qh4 90.Qg1 d5 91.exd5 Rd2 92.Ra2 Rcc2 93.Rxc2 Rxc2 94.Kg2 Qg5+ 95.Kh2 Qf4+ 96.Kg2 Qg5+
= (0.00) Depth: 15/41 00:00:01 3643kN, tb=1
89...Qh4 90.Qg1 d5 91.Qg3 Qxg3+ 92.Kxg3 dxe4 93.Ra7 Rxb4 94.Rxb4 Rxb4 95.fxe4 Rxe4 96.f3 Re1 97.Rxf7
= (0.00) Depth: 16/43 00:00:02 10251kN, tb=70
89...Qh4 90.Qg1 d5 91.Qg3 Qxg3+ 92.Kxg3 dxe4 93.Ra7 Rxb4 94.Rxb4 Rxb4 95.fxe4 Rxe4 96.f3 Re1 97.Rxf7 Kg6
= (0.00) Depth: 17/36 00:00:03 12012kN, tb=70
89...Rc3 90.Rd5 Qh4 91.Kg2 Rd2 92.Rxd2 Qg5+ 93.Kh1 Qxd2 94.Ra1 Qxf2 95.Qf1 Qxf3+ 96.Qxf3 Rxf3 97.Kg2 Rb3
-/+ (-0.98) Depth: 17/62 00:01:09 449mN, tb=2606
89...Rc3 90.Rd5 Qh4 91.Kg2 Rd2 92.Rxd2 Qg5+ 93.Kh1 Qxd2 94.Ra1 Qxf2 95.Qf1 Qxf3+ 96.Qxf3 Rxf3 97.Kg2 Rb3 98.Ra4
-/+ (-0.98) Depth: 18/49 00:01:20 508mN, tb=2620
89...Rc3 90.Rd5 Qh4 91.Kg2 Rd2 92.Rxd2 Qg5+ 93.Kh1 Qxd2 94.Ra1 Rxf3 95.Qe1 Qxe1+ 96.Rxe1 Rxh3+ 97.Kg2 Rb3 98.Rc1 Rxb4
-+ (-1.60) Depth: 19/62 00:04:17 1485mN, tb=10113
(, 03.01.2009)

And here is R3 in same position at 5 minutes

12: Naum 4 - Rybka 3, N4_60_1_7move-1 2008
8/4qppk/3p3p/R7/RPrrP3/5P1P/5P1K/1Q6 b - - 0 1

Analysis by Rybka 3:

89...Qh4 90.Qg1 Rxb4
= (-0.08) Depth: 6 00:00:00 5kN
89...Qh4 90.Qg1 Rxb4 91.Rxb4 Rxb4 92.Qg3
= (-0.15) Depth: 7 00:00:00 6kN
89...Qh4 90.Qg1 Rxb4 91.Rxb4 Rxb4 92.Qg3 Qf6 93.Rf5
= (-0.13) Depth: 8 00:00:00 12kN
89...Qh4 90.Qg1 Rxb4 91.Rxb4 Rxb4 92.Qg3 Qf6 93.Rf5 Qe7
= (-0.13) Depth: 9 00:00:00 19kN
89...Qh4 90.Qg1 Rxb4 91.Rxb4 Rxb4 92.Qg3 Qf6 93.Rf5 Qe7
= (-0.13) Depth: 10 00:00:00 21kN
89...Qh4 90.Qg1 Rxb4 91.Rxb4 Rxb4 92.Qg3 Qf6 93.Rf5 Qe7 94.Rd5
= (-0.13) Depth: 11 00:00:00 28kN
89...Qh4 90.Qg1 Rc2 91.Ra2 Rxa2 92.Rxa2 Qf6 93.Qg3 Rxb4 94.Rd2 Rb6 95.Kg2 Kg8
= (-0.12) Depth: 12 00:00:00 84kN
89...g6 90.Qb3 Qh4 91.Ra2 d5 92.Rxd5
= (-0.19) Depth: 12 00:00:01 378kN
89...Rc3
=/+ (-0.33 !) Depth: 12 00:00:01 407kN
89...Rc3
=/+ (-0.53 !) Depth: 12 00:00:01 535kN
89...Rc3
-/+ (-0.93 !) Depth: 12 00:00:02 1136kN
89...Rc3 90.Ra2 Rdd3 91.Qg1 Rxf3 92.Qg4 Rxh3+ 93.Qxh3 Rxh3+ 94.Kxh3 Qxe4 95.b5 Qb7 96.Rb2 h5 97.Kh2 Qb6 98.Ra6
-/+ (-1.06) Depth: 12 00:00:08 3611kN
89...Rc3 90.Ra2 Rdd3 91.Qg1 Rxf3 92.Qg4 Rxh3+ 93.Qxh3 Rxh3+ 94.Kxh3 Qxe4 95.b5 Qb7 96.Rb2 h5 97.Kh2 Qb6 98.Ra6
-/+ (-1.16) Depth: 13 00:00:09 4004kN
89...Rc3
-/+ (-1.36 !) Depth: 14 00:00:13 6195kN
89...Rc3 90.Ra2 Rdd3 91.Qg1 Rxf3 92.Qg4 Rxh3+ 93.Qxh3 Rxh3+ 94.Kxh3 Qxe4 95.b5 Qb7 96.R2a3 d5 97.Rb3 Qb6 98.Raa3
-/+ (-1.36) Depth: 14 00:00:13 6426kN
89...Rc3 90.Ra2 Rdd3 91.Qg1 Rxf3 92.Qg4[] Rxh3+ 93.Qxh3 Rxh3+ 94.Kxh3 Qxe4 95.b5 Qb7 96.Kh2 Qb6 97.Kg2 h5 98.R2a3 d5 99.R5a4 Qxb5 100.Rd4 g5 101.Rad3 h4 102.Rxd5 Qc6 103.Kh2
-+ (-1.43) Depth: 15 00:00:24 11836kN
89...Rc3 90.Ra2 Rdd3 91.Qg1 Rxf3 92.Qg4[] Rxh3+ 93.Qxh3 Rxh3+ 94.Kxh3 Qxe4 95.b5 Qb7 96.Kh2 Qb6 97.Kg2 h5 98.R2a3 d5 99.R5a4 Qxb5 100.Rd4 g5 101.Rad3 h4 102.Rxd5 Qc6 103.Kh2 Qc4 104.Rd7
-+ (-1.55) Depth: 16 00:00:43 21756kN
89...Rc3
-+ (-1.75 !) Depth: 17 00:01:40 51969kN
89...Rc3 90.Qg1 Rxf3 91.Ra3 Rxa3 92.Rxa3 Rxb4 93.Qg3 Qxe4 94.Qf3 Qe5+ 95.Kg2 Rf4 96.Qd3+ g6 97.Qe3 Qd5+ 98.Kg3 Qg5+ 99.Kh2
-+ (-1.78) Depth: 17 00:02:43 85385kN
89...Rc3 90.Qg1 Rxf3 91.Ra3 Rxa3 92.Rxa3 Rxb4 93.Qg3 Rxe4 94.Qd3[] g6 95.Qd2 Qe5+ 96.Kg2 Rd4 97.Qe3 Kg7 98.Ra7 Rc4 99.Qf3 Qg5+ 100.Qg3 Qd5+ 101.Qf3 Qxf3+ 102.Kxf3
-+ (-1.85) Depth: 18 00:03:58 126mN
(, 03.01.2009)

It takes R3 simply 2 seconds to to reach an evaluation that takes N4 96 seconds... N4 the clip analysis shows that it takes too long to see that Rc3! is a killer move

89...Rc3!

[d]8/4qppk/3p3p/R7/RP1rP3/2r2P1P/5P1K/1Q6 w - - 0 90

(Qh4) -1.54/15 1:20 90.Ra3 (Ra2) 0.07/15 9

90...Qh4!

[d]8/5ppk/3p3p/R7/1P1rP2q/R1r2P1P/5P1K/1Q6 w - - 0 91


R3 had seen this temporary rook sac all the way through ... as for N4 it is only now that it sees it is lost ... but now it is way too late

-1.64/13 0 91.Rxc3 -4.95/14 7 Rd2! -1.64/13 0 92.Qf1 (Kh1) -5.78/15 15 92...Rxf2+ -5.30/12 8 93.Qg2 -5.78/16 5 Qf4+ -1.64/12 0 94.Kg1 -5.79/16 8 Rxg2+ -5.66/13 12 95.Kxg2 -4.99/6 0 Qd2+ -5.63/15 7 96.Kg3 -5.79/20 6 Qxc3 -5.67/15 10 97.h4 (Rf5) -5.79/18 5 97...Qxb4 -6.00/14 14 98.Ra1 (Rf5) -6.63/18 7 98...Kg6 (Qb8) -6.73/12 14 99.Rd1 (Rc1) -6.63/20 9 99...f5 (Qc5) -6.89/13 11 100.exf5+ (Kf4) -6.64/19 10 100...Kxf5 -7.20/13 6 0–1


That is what I mean when I say that Rybka has much faster search ... it searches the critical lines quicker and deeper. The tricks that Vas uses to make Rybka very quickly understand which lines to search are what makes Rybka 3 a phenomenally fast searcher. A chess engine is a careful mixture of search and knowledge ... an evaluation function that is inefficient and too laden with knowledge code can become a slow or inefficient searcher ... you need the correct mix to come out with the best solution. At the moment Rybka dominates because it has the best mix ... at least with today's hardware.

[Jouni]

Ok I trust You! BTW in 10 testsuites I tried Rybka won Naum in all. But
still N4 is the only engine, which gives interesting matches aganst R3 - and
even won one match 11-9 with predefined starting positions (HISTORY):

Code: Select all

TEST  blitz

                          
1   Naum 4                 ½½½½1½0½1½½½½1½0½1½½  11.0/20
2   Rybka 3 1-cpu 32-bit  ½½½½0½1½0½½½½0½1½0½½    9.0/20

Jouni

[Tomcass]

I have very clear that the answer to the title of this post is a big YES! when we talk about long time controls.

Tomorrow I will post the final results of mi test of 200 games at 60 minutes in a Quad 6700 between both engines. The victory of Naum4 has been impressive.

My feeling is that at long time controls and fast hardware, Naum4 is slightly better than Rybka3.

Regards from Barcelona.

Tom.

[Uri Blass]

I have very clear that the answer to the title of this post is a big YES! when we talk about long time controls.

Tomorrow I will post the final results of mi test of 200 games at 60 minutes in a Quad 6700 between both engines. The victory of Naum4 has been impressive.

My feeling is that at long time controls and fast hardware, Naum4 is slightly better than Rybka3.

Regards from Barcelona.

Tom.

Some comments:

1)I read that it is better to use contempt=0 for rybka for long time control.
I read that the default contempt=15 is better only for blitz or against significantly weaker opponents

2)I guess that it is better to use the same book for both sides for a good comparison.

I know that based on CCRL Rybka is clearly better at 40/40 even when the default contempt=15 is used
http://computerchess.org.uk/ccrl/4040/r ... t_all.html

only rybka 64 bit is used but the difference between 64 bits and 32 bits is clearly smaller than the difference between 4 cpu and 1 cpu and even Rybka 1 cpu has similiar rating to Naum 4 cpu.

Uri

[geots]

I have very clear that the answer to the title of this post is a big YES! when we talk about long time controls.

Tomorrow I will post the final results of mi test of 200 games at 60 minutes in a Quad 6700 between both engines. The victory of Naum4 has been impressive.

My feeling is that at long time controls and fast hardware, Naum4 is slightly better than Rybka3.

Regards from Barcelona.

Tom.

Some comments:

1)I read that it is better to use contempt=0 for rybka for long time control.
I read that the default contempt=15 is better only for blitz or against significantly weaker opponents

2)I guess that it is better to use the same book for both sides for a good comparison.

I know that based on CCRL Rybka is clearly better at 40/40 even when the default contempt=15 is used
http://computerchess.org.uk/ccrl/4040/r ... t_all.html

only rybka 64 bit is used but the difference between 64 bits and 32 bits is clearly smaller than the difference between 4 cpu and 1 cpu and even Rybka 1 cpu has similiar rating to Naum 4 cpu.

Uri

To be a little more blunt, at longer time controls Rybka 3 will destroy Naum 4. Actually, the discussion is really not worthy of using up the bandwidth.

[Laskos]

I have very clear that the answer to the title of this post is a big YES! when we talk about long time controls.

Tomorrow I will post the final results of mi test of 200 games at 60 minutes in a Quad 6700 between both engines. The victory of Naum4 has been impressive.

My feeling is that at long time controls and fast hardware, Naum4 is slightly better than Rybka3.

Regards from Barcelona.

Tom.

Some comments:

1)I read that it is better to use contempt=0 for rybka for long time control.
I read that the default contempt=15 is better only for blitz or against significantly weaker opponents

2)I guess that it is better to use the same book for both sides for a good comparison.

I know that based on CCRL Rybka is clearly better at 40/40 even when the default contempt=15 is used
http://computerchess.org.uk/ccrl/4040/r ... t_all.html

only rybka 64 bit is used but the difference between 64 bits and 32 bits is clearly smaller than the difference between 4 cpu and 1 cpu and even Rybka 1 cpu has similiar rating to Naum 4 cpu.

Uri

I would go for fixed positions, say 100, black and white alternatively. I always do my tests on fixed positions, I do not want to introduce book's variance in the problem (I do not even know how to estimate this variance). Bob very wisely is using several thousands of fixed positions black/white.

Contempt factor should be set to 0 probably, Naum is too strong on long time controls.

Nevertheless, I am very curious about Tomcass results because it seems that N4 is improving very much on R3 in long matches/fast hardware conditions.

Kai

[Uri Blass]

I have very clear that the answer to the title of this post is a big YES! when we talk about long time controls.

Tomorrow I will post the final results of mi test of 200 games at 60 minutes in a Quad 6700 between both engines. The victory of Naum4 has been impressive.

My feeling is that at long time controls and fast hardware, Naum4 is slightly better than Rybka3.

Regards from Barcelona.

Tom.

Some comments:

1)I read that it is better to use contempt=0 for rybka for long time control.
I read that the default contempt=15 is better only for blitz or against significantly weaker opponents

2)I guess that it is better to use the same book for both sides for a good comparison.

I know that based on CCRL Rybka is clearly better at 40/40 even when the default contempt=15 is used
http://computerchess.org.uk/ccrl/4040/r ... t_all.html

only rybka 64 bit is used but the difference between 64 bits and 32 bits is clearly smaller than the difference between 4 cpu and 1 cpu and even Rybka 1 cpu has similiar rating to Naum 4 cpu.

Uri

To be a little more blunt, at longer time controls Rybka 3 will destroy Naum 4. Actually, the discussion is really not worthy of using up the bandwidth.

Rybka3 will certainly destroy Naum4 in CCRL conditions or CEGT conditions.

The question here is about different conditions when one of them seem to be long book line and another one is x minutes per game time control and not x minutes/40 moves.

Uri

[geots]

I have very clear that the answer to the title of this post is a big YES! when we talk about long time controls.

Tomorrow I will post the final results of mi test of 200 games at 60 minutes in a Quad 6700 between both engines. The victory of Naum4 has been impressive.

My feeling is that at long time controls and fast hardware, Naum4 is slightly better than Rybka3.

Regards from Barcelona.

Tom.

Some comments:

1)I read that it is better to use contempt=0 for rybka for long time control.
I read that the default contempt=15 is better only for blitz or against significantly weaker opponents

2)I guess that it is better to use the same book for both sides for a good comparison.

I know that based on CCRL Rybka is clearly better at 40/40 even when the default contempt=15 is used
http://computerchess.org.uk/ccrl/4040/r ... t_all.html

only rybka 64 bit is used but the difference between 64 bits and 32 bits is clearly smaller than the difference between 4 cpu and 1 cpu and even Rybka 1 cpu has similiar rating to Naum 4 cpu.

Uri

To be a little more blunt, at longer time controls Rybka 3 will destroy Naum 4. Actually, the discussion is really not worthy of using up the bandwidth.

Rybka3 will certainly destroy Naum4 in CCRL conditions or CEGT conditions.

The question here is about different conditions when one of them seem to be long book line and another one is x minutes per game time control and not x minutes/40 moves.

Uri

Ah, i see. Thanks, Uri.

[Laskos]

Rybka3 will certainly destroy Naum4 in CCRL conditions or CEGT conditions.


Uri

Are you sure that R3 will destroy N4 on a quad 120/40? Let's see. I don't like they are using books though.

Kai