SF novelty 10 g4!

Discussion of computer chess matches and engine tournaments.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
beram
Posts: 1187
Joined: Wed Jan 06, 2010 2:11 pm

SF novelty 10 g4!

Post by beram » Wed Dec 10, 2014 6:48 pm

How long does it takes the fastest hardware to find the SF novelty 11 g4! with Houdini 4.
On my AMD 1090T system using 6 cores it took about 40 minutes to find it at depth 29

Stockfish 301114 - Komodo 1331, TCEC Season 7 - Stage 4 2014
[D] r1bq1rk1/pp2bpp1/2p1pn1p/8/3P4/3B1N2/PPP1QPPP/R1B1K2R w KQ - 0 1

Analysis by Houdini 4 x64 (AMD 1090T @3200Mhz, 6cpu, 4096 hash)

11.Bf4 b6 12.0-0 Bb7 13.Rad1 Qc8 14.Rfe1 c5 15.c3 cxd4 16.Nxd4 Rd8 17.Be5 Nd7 18.Bg3 Nf6 19.Bb1 Qc5 20.Be5 Rac8 21.Qd3 Qc4 22.Qxc4 Rxc4 23.Nb5 Rxd1 24.Rxd1 Rg4 25.Bg3 a6 26.Nd4 Rg5 27.f3 Rd5 28.c4 Rd8 29.Bc7 Rc8 30.Bxb6 Rxc4 31.b3 Rc3
= (0.22) Depth: 28/63 00:09:05 5395mN
11.Bf4 b6 12.0-0 Bb7 13.Rad1 Bd6 14.Be5 Qe7 15.Rd2 c5 16.dxc5 Bxc5 17.Rfd1 Rfd8 18.h3 Rac8 19.c3 a5 20.a3 Bd5 21.Bc2 Bd6 22.Bd4 Bc5 23.c4 Bxf3 24.Qxf3 Bxd4 25.Rxd4 Rxd4 26.Rxd4 e5
= (0.23) Depth: 29/73 00:17:00 10213mN
11.g4
+/= (0.30 !) Depth: 29/75 00:40:30 24709mN

(Bram Mourik, AMD X6 1090T 3200 Mhz (Fritzmark 23,1 6x) 10.12.2014)

ernest
Posts: 1863
Joined: Wed Mar 08, 2006 7:30 pm

Re: SF novelty 10 g4!

Post by ernest » Wed Dec 10, 2014 7:28 pm

beram wrote:How long does it takes the fastest hardware to find the SF novelty 11 g4! with Houdini 4.
As often (especially with Stockfish ?), the trick is to use MultiPV

On my old dualcore, after 1.5 minute, depth 25

Analysis by Stockfish 261114 syz, MultiPV=3:

1. +/= (0.29): 1.g4
2. = (0.23): 1.0-0 c5 2.Bf4 b6 3.dxc5 Bxc5 4.Rad1 Qe7 5.a3 Bb7 6.b4 Nd5 7.Be5 Bd6 8.c4 Nf4 9.Bh7+ Kh8 10.Bxf4 Bxf4 11.Be4 Rfd8 12.Bxb7 Qxb7 13.g3 Bc7 14.Rfe1 Rac8
3. = (0.22): 1.Bf4 b6 2.Rd1 Bb7 3.0-0 Qd5 4.c3 Qh5 5.Be5 c5 6.dxc5 Bxc5 7.Rfe1 Be7 8.h3 Rfd8 9.a3 Bd5 10.Bc2 Rac8 11.Qe3 Bxf3 12.Qxf3 Qxf3 13.gxf3

zullil
Posts: 5680
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: SF novelty 10 g4!

Post by zullil » Wed Dec 10, 2014 7:40 pm

ernest wrote:
beram wrote:How long does it takes the fastest hardware to find the SF novelty 11 g4! with Houdini 4.
As often (especially with Stockfish ?), the trick is to use MultiPV

On my old dualcore, after 1.5 minute, depth 25

Analysis by Stockfish 261114 syz, MultiPV=3:

1. +/= (0.29): 1.g4
2. = (0.23): 1.0-0 c5 2.Bf4 b6 3.dxc5 Bxc5 4.Rad1 Qe7 5.a3 Bb7 6.b4 Nd5 7.Be5 Bd6 8.c4 Nf4 9.Bh7+ Kh8 10.Bxf4 Bxf4 11.Be4 Rfd8 12.Bxb7 Qxb7 13.g3 Bc7 14.Rfe1 Rac8
3. = (0.22): 1.Bf4 b6 2.Rd1 Bb7 3.0-0 Qd5 4.c3 Qh5 5.Be5 c5 6.dxc5 Bxc5 7.Rfe1 Be7 8.h3 Rfd8 9.a3 Bd5 10.Bc2 Rac8 11.Qe3 Bxf3 12.Qxf3 Qxf3 13.gxf3
And here SF (16 threads, 8 GB hash, MultiPV=1) finds it at depth 24 in 12 seconds:

Code: Select all

info depth 24 seldepth 34 multipv 1 score cp 27 lowerbound nodes 63746556 nps 5209328 time 12237 pv g2g4
info depth 24 currmove g2g4 currmovenumber 1
info depth 24 currmove e1g1 currmovenumber 2
info depth 24 currmove a1b1 currmovenumber 21
info depth 24 currmove c1h6 currmovenumber 41
info depth 24 seldepth 34 multipv 1 score cp 20 upperbound nodes 70755213 nps 5541171 time 12769 pv g2g4 d8b6
info depth 24 currmove g2g4 currmovenumber 1
info depth 24 seldepth 34 multipv 1 score cp 28 lowerbound nodes 78776864 nps 5749296 time 13702 pv g2g4
info depth 24 currmove g2g4 currmovenumber 1
info depth 24 currmove b2b4 currmovenumber 14
info depth 24 seldepth 34 multipv 1 score cp 20 nodes 81252314 nps 5761758 time 14102 pv g2g4 d8b6 c2c3 c6c5 d4c5 b6c5 g4g5 h6g5 c1g5 f8d8 h1g1 c8d7 g5h6 e7f8 h6e3 c5h5 g1g5 h5h3 f3e5 f8d6 e5d7 d8d7 e1c1 d6h2 c1b1 d7d5 g5d5 f6d5 d3e4 d5e3

ernest
Posts: 1863
Joined: Wed Mar 08, 2006 7:30 pm

Re: SF novelty 10 g4!

Post by ernest » Wed Dec 10, 2014 8:34 pm

...and indeed, with the usual MultiPV=1 my old dualcore did not take that long
(but still, g4! was found at depth 31 instead of depth 25 with MultiPV=3)

Analysis by Stockfish 261114 syz:

1.0-0 Bd7 2.Ne5 c5 3.Be3 cxd4
= (0.18) Depth: 6/6 00:00:00 4kN
1.0-0 Bd7 2.Ne5 c5 3.dxc5 Bxc5 4.Qf3
+/= (0.48) Depth: 7/7 00:00:00 7kN
1.0-0 c5 2.dxc5 Bxc5 3.Ne5 Bd7 4.Be3 Bd6 5.Nxd7
+/= (0.26) Depth: 8/8 00:00:00 14kN
1.0-0 c5 2.dxc5 Bxc5 3.Be3 Bd6 4.c3 Bd7 5.Rad1
= (0.23) Depth: 9/12 00:00:00 31kN
1.c3 c5 2.dxc5 Bxc5 3.0-0 Bd6 4.Be3 Bd7 5.Rad1 Nd5
= (0.13) Depth: 10/12 00:00:00 57kN
1.0-0 b6 2.Ne5 Bb7 3.Qe3 Nd5 4.Qg3 Bh4 5.Qf3 Nb4 6.Bd2 Nxd3
+/= (0.32) Depth: 11/14 00:00:00 119kN
1.0-0 b6 2.Ne5 Bb7 3.Qe3 c5 4.dxc5 Bxc5 5.Qg3 Nh5 6.Qg4 Qd5
= (0.22) Depth: 12/14 00:00:00 172kN
1.0-0 c5 2.Bf4 b6 3.dxc5 bxc5 4.Rad1 Qb6 5.b3 Bb7 6.c4 Rad8 7.Ne5
+/= (0.31) Depth: 13/19 00:00:00 255kN
1.0-0 b6 2.Ne5 Bb7 3.Rd1 Kh8 4.Bc4 Bd6 5.Bf4 Qc7 6.c3 Rad8 7.Qe3 Nd5 8.Bxd5 exd5 9.Qd3
= (0.18) Depth: 14/20 00:00:00 551kN
1.0-0 b6 2.Ne5 Qxd4 3.Nxc6 Qd7 4.Nxe7+ Qxe7 5.Bf4 Bb7 6.c4 Qc5 7.Rad1 Rac8 8.b3 Rcd8 9.Be3
= (0.19) Depth: 15/23 00:00:00 945kN
1.0-0 b6 2.Ne5 Qxd4 3.Nxc6 Qd7 4.Nxe7+ Qxe7 5.Bf4 Bb7 6.c4 Qc5 7.Rad1 Rac8 8.b3 Rfd8 9.Be3 Qe7 10.f3 Qc7 11.Qb2
= (0.22) Depth: 16/23 00:00:00 1190kN
1.0-0 b6 2.Ne5 Qxd4 3.Nxc6 Qd7 4.Nxe7+ Qxe7 5.Bf4 Nd5 6.Be5 Bb7 7.a3 Rfd8 8.Rad1 Nf6 9.c3 Rac8 10.Bc2
+/= (0.28) Depth: 17/25 00:00:01 1499kN
1.0-0 b6 2.Ne5 Qxd4 3.Nxc6 Qd7 4.Nxe7+ Qxe7 5.Bf4 Rd8 6.c4 Bb7 7.Rad1 Rac8 8.b3 Qc5 9.Be3 Qc7 10.f3 Rd7 11.Qb2
= (0.24) Depth: 18/25 00:00:01 1692kN
1.0-0 c5 2.dxc5 Bxc5 3.Be3 Qb6 4.Bxc5 Qxc5 5.Rad1 b6 6.Qe5 Qxe5 7.Nxe5 Bb7 8.f3 Rfd8 9.Rfe1 Rac8 10.a3 a6 11.c3 a5
+/= (0.26) Depth: 19/25 00:00:03 4497kN
1.0-0 c5
= (0.20 --) Depth: 20/25 00:00:03 5210kN
1.0-0 c5 2.dxc5 Bxc5 3.Bf4 b6 4.Rad1 Qe7 5.a3 a5 6.c3 Rd8 7.Be5 Bb7 8.Nd4 Bd6 9.Rfe1 Rac8 10.f3 Bxe5 11.Qxe5 Rc5
= (0.17) Depth: 20/26 00:00:04 6473kN
1.0-0
= (0.23 !) Depth: 21/27 00:00:04 6780kN
1.0-0
+/= (0.29 !) Depth: 21/27 00:00:05 7347kN
1.0-0 b6 2.Rd1 Bb7 3.Ne5 Kh8 4.Be3 c5 5.dxc5 Bxc5 6.Ba6 Qc7 7.Bxb7 Qxb7 8.Bxc5 bxc5 9.b3 Rad8 10.a4 Rd5 11.Rxd5
+/= (0.34) Depth: 21/27 00:00:05 8556kN
1.0-0 c5
+/= (0.28 --) Depth: 22/27 00:00:07 10699kN
1.0-0 c5
= (0.22 --) Depth: 22/27 00:00:08 11926kN
1.0-0 c5 2.dxc5 Bxc5 3.Bf4 b6 4.Rad1 Qe7 5.a3 Bb7 6.b4 Nd5 7.Be5 Bd6 8.c4 Nf4 9.Bxf4 Bxf4 10.g3 Bd6 11.Be4 Rfd8 12.Nd4 Qd7 13.Bxb7
= (0.16) Depth: 22/32 00:00:09 13966kN
1.0-0
= (0.22 !) Depth: 23/32 00:00:10 15919kN
1.0-0 c5 2.dxc5 Bxc5 3.Bf4 b6 4.Rad1 Qe7 5.a3 Bb7 6.b4 Nd5 7.Be5 Bd6 8.c4 Nf4 9.Bxf4 Bxf4 10.g3 Bd6 11.Be4 Rfd8 12.Nd4 Kh8 13.Bxb7
= (0.24) Depth: 23/32 00:00:11 17376kN
1.0-0 c5
= (0.18 --) Depth: 24/32 00:00:13 20031kN
1.0-0
= (0.24 !) Depth: 24/33 00:00:16 25069kN
1.0-0 c5
= (0.18 --) Depth: 24/33 00:00:22 32609kN
1.0-0
= (0.25 !) Depth: 24/33 00:00:22 33299kN
1.0-0 c5 2.Bf4 b6 3.Rad1 cxd4 4.Bc4 Bb7 5.Nxd4 Qc8 6.c3 Rd8 7.Rfe1 a5 8.Bd3 Qc5 9.a4 Rac8 10.Be5 Rd7 11.Bg3 Rxd4 12.cxd4 Qd5 13.f3 Qxd4+ 14.Qf2
+/= (0.27) Depth: 24/33 00:00:24 35268kN
1.0-0 c5
= (0.20 --) Depth: 25/33 00:00:25 37248kN
1.Bf4 b6 2.Rd1 Bb7 3.0-0 Qc8 4.c3 c5 5.Rfe1 Rd8 6.Be5 a6 7.h3 b5 8.dxc5 Qxc5 9.Bd4 Qc6 10.a3 Rac8 11.Qe5 Bd6 12.Qe3 Be7 13.Bc2 Kh8 14.Bb6
= (0.19) Depth: 25/33 00:00:36 52804kN
1.Bf4
= (0.25 !) Depth: 26/33 00:00:44 65125kN
1.Bf4 b6
= (0.19 --) Depth: 26/34 00:00:50 73431kN
1.Bf4
= (0.25 !) Depth: 26/34 00:00:51 75572kN
1.Bf4 b6 2.Rd1 Bb7 3.0-0 Qd5 4.c3 Qh5 5.Be5 Rad8 6.Rfe1 c5 7.dxc5 Bxc5 8.h3 Be7 9.Nd4 Qxe2 10.Bxe2 Rc8 11.Bf3 Bxf3 12.Nxf3 Rfd8 13.Rxd8+ Rxd8 14.Nd4 Rc8 15.f3 Nd7
= (0.22) Depth: 26/34 00:00:53 78943kN
1.Bf4 b6 2.Rd1 Bb7 3.0-0 Qd5 4.a3 Qh5 5.Be5 Rad8 6.Rfe1 c5 7.dxc5 Bxc5 8.c3 Ng4 9.Bg3 Nf6 10.h3 Rd7 11.Bc2 Rxd1 12.Rxd1 Rc8 13.b4 Be7 14.Qe3 Nd5
= (0.17) Depth: 27/34 00:01:03 92629kN
1.0-0
= (0.23 !) Depth: 28/41 00:01:24 124mN
1.0-0
+/= (0.29 !) Depth: 28/41 00:01:41 150mN
1.0-0 c5
= (0.23 --) Depth: 28/41 00:01:57 174mN
1.0-0 c5 2.Be3 b6 3.dxc5 Bxc5 4.Bxc5 bxc5 5.Ne5 Qc7 6.Rad1 Rb8 7.c4 Nd7 8.Nxd7 Bxd7 9.b3 Bc6 10.Rfe1 Rfd8 11.Qe5 Qb6 12.f3 Rd4 13.Bc2 Rbd8 14.Rxd4 cxd4
= (0.24) Depth: 28/41 00:02:00 178mN
1.0-0 c5 2.Be3 b6 3.dxc5 Bxc5 4.Bxc5 bxc5 5.Ne5 Qc7 6.Rfe1 Bb7 7.Rad1 Rad8 8.f3 Rd5 9.Bc4 Rxd1 10.Rxd1 Rd8 11.Rxd8+ Qxd8 12.Nxf7 Qd4+ 13.Kf1 Bd5 14.Bxd5 exd5 15.c3 Qf4 16.Ne5 Qxh2 17.Qe3 Qh1+ 18.Qg1 Qh5
= (0.20) Depth: 29/41 00:02:20 207mN
1.0-0
+/= (0.27 !) Depth: 30/41 00:02:44 243mN
1.0-0 c5
= (0.20 --) Depth: 30/41 00:03:30 311mN
1.0-0
+/= (0.27 !) Depth: 30/41 00:03:52 345mN
1.0-0 c5
= (0.19 --) Depth: 30/41 00:04:34 407mN
1.0-0 c5 2.Bf4 b6 3.Rad1 cxd4 4.Nxd4 Bb7 5.Bc4 Qc8 6.Rfe1 Rd8 7.c3 Nd5 8.Be5 Qc5 9.Bd3 Nf6 10.a3 Rac8 11.Ba6 Bxa6 12.Qxa6 Rd7 13.Qe2 Rcd8 14.Qd3 Ng4 15.Qg3 Bf6 16.Bf4
= (0.23) Depth: 30/41 00:04:46 424mN
1.g4
+/= (0.29 !) Depth: 31/45 00:09:46 837mN
1.g4
+/= (0.35 !) Depth: 31/45 00:12:52 1083mN
1.g4
+/= (0.44 !) Depth: 31/45 00:13:46 1160mN
1.g4
+/= (0.58 !) Depth: 31/45 00:16:59 1422mN
1.g4 e5 2.dxe5 Re8 3.h3 Nd7 4.Be3 Bf8 5.0-0-0 Qa5 6.Kb1 Nxe5 7.Bd2 Qd5 8.Nxe5 Qxe5 9.Qxe5 Rxe5 10.f4 Re8 11.Rhe1 Be6 12.c4 Rad8 13.Bc3 Bc8 14.Kc2 Bd6 15.f5 Bg3 16.Rxe8+ Rxe8 17.Bd4 b6 18.c5 b5
+/= (0.51) Depth: 31/45 00:20:42 1725mN
1.g4
+/= (0.57 !) Depth: 32/45 00:21:50 1822mN, tb=1
1.g4 e5
+/= (0.51 --) Depth: 32/45 00:22:13 1856mN, tb=1
1.g4
+/= (0.57 !) Depth: 32/45 00:22:35 1887mN, tb=1
1.g4 e5 2.dxe5 Re8 3.h3 Nd7 4.Be3 Bf8 5.0-0-0 Qa5 6.Kb1 Nxe5 7.Bd2 Qd5 8.Nxe5 Qxe5 9.Qxe5 Rxe5 10.f4 Re8 11.f5 Bd7 12.Rhe1 Be7 13.Bf4 Bg5 14.Bxg5 hxg5 15.Be4 Rad8 16.c3 Kh7 17.f6+ g6
+/= (0.57) Depth: 32/45 00:22:51 1908mN, tb=1
1.g4
+/= (0.63 !) Depth: 33/45 00:28:00 2336mN, tb=1
1.g4
+/= (0.69 !) Depth: 33/45 00:30:43 2559mN, tb=1

JJJ
Posts: 1287
Joined: Sat Apr 19, 2014 11:47 am

Re: SF novelty 10 g4!

Post by JJJ » Wed Dec 10, 2014 8:36 pm

FEN: r1bq1rk1/pp2bpp1/2p1pn1p/8/3P4/3B1N2/PPP1QPPP/R1B1K2R w KQ - 0 1

Stockfish_14111222_x64_modern:

28/38+ 00:53 292 455k 5 499k +0,30 g2-g4
28/38+ 01:19 430 926k 5 406k +0,36 g2-g4
28/38- 01:25 459 335k 5 402k +0,30 g2-g4 Nf6-d5

Stockfish_14111222_x64_modern:
Stockfish_14111222_x64_modern

53 secondes with the best stockfish on 4 core.

beram
Posts: 1187
Joined: Wed Jan 06, 2010 2:11 pm

Re: SF novelty 10 g4!

Post by beram » Wed Dec 10, 2014 9:07 pm

zullil wrote:
ernest wrote:
beram wrote:How long does it takes the fastest hardware to find the SF novelty 11 g4! with Houdini 4.
As often (especially with Stockfish ?), the trick is to use MultiPV

On my old dualcore, after 1.5 minute, depth 25

Analysis by Stockfish 261114 syz, MultiPV=3:

1. +/= (0.29): 1.g4
2. = (0.23): 1.0-0 c5 2.Bf4 b6 3.dxc5 Bxc5 4.Rad1 Qe7 5.a3 Bb7 6.b4 Nd5 7.Be5 Bd6 8.c4 Nf4 9.Bh7+ Kh8 10.Bxf4 Bxf4 11.Be4 Rfd8 12.Bxb7 Qxb7 13.g3 Bc7 14.Rfe1 Rac8
3. = (0.22): 1.Bf4 b6 2.Rd1 Bb7 3.0-0 Qd5 4.c3 Qh5 5.Be5 c5 6.dxc5 Bxc5 7.Rfe1 Be7 8.h3 Rfd8 9.a3 Bd5 10.Bc2 Rac8 11.Qe3 Bxf3 12.Qxf3 Qxf3 13.gxf3
And here SF (16 threads, 8 GB hash, MultiPV=1) finds it at depth 24 in 12 seconds:

Code: Select all

info depth 24 seldepth 34 multipv 1 score cp 27 lowerbound nodes 63746556 nps 5209328 time 12237 pv g2g4
info depth 24 currmove g2g4 currmovenumber 1
info depth 24 currmove e1g1 currmovenumber 2
info depth 24 currmove a1b1 currmovenumber 21
info depth 24 currmove c1h6 currmovenumber 41
info depth 24 seldepth 34 multipv 1 score cp 20 upperbound nodes 70755213 nps 5541171 time 12769 pv g2g4 d8b6
info depth 24 currmove g2g4 currmovenumber 1
info depth 24 seldepth 34 multipv 1 score cp 28 lowerbound nodes 78776864 nps 5749296 time 13702 pv g2g4
info depth 24 currmove g2g4 currmovenumber 1
info depth 24 currmove b2b4 currmovenumber 14
info depth 24 seldepth 34 multipv 1 score cp 20 nodes 81252314 nps 5761758 time 14102 pv g2g4 d8b6 c2c3 c6c5 d4c5 b6c5 g4g5 h6g5 c1g5 f8d8 h1g1 c8d7 g5h6 e7f8 h6e3 c5h5 g1g5 h5h3 f3e5 f8d6 e5d7 d8d7 e1c1 d6h2 c1b1 d7d5 g5d5 f6d5 d3e4 d5e3
And how long for your system with Houdini 4 ?
btw with Komodo 8 it took me with same system 4cpu 1hour 50 minutes depth 31

48: Stockfish 301114 - Komodo 1331, TCEC Season 7 - Stage 4 2014
r1bq1rk1/pp2bpp1/2p1pn1p/8/3P4/3B1N2/PPP1QPPP/R1B1K2R w KQ - 0 1

Analysis by Komodo 8 64-bit 4cpu:

11.Bf4 b6 12.0-0 Bb7 13.Rad1 Qc8 14.c4 Rd8 15.Bg3 Bf8 16.Rd2 Bb4 17.Rc2 Bf8 18.Rd1 c5 19.dxc5 Qxc5 20.Rcd2 Be7 21.Ne5 Qc7 22.a3 a5 23.h3 Rac8 24.Bf4 Qc5 25.Kh2 Qc7 26.Bc2 Rxd2 27.Rxd2 Bd6 28.f3
= (0.23) Depth: 30 00:37:37 9304mN
11.Bf4 b6 12.0-0 Bb7 13.Rad1 Qc8 14.c3 c5 15.Rfe1 Qc6 16.dxc5 Qxc5 17.Be5 Rad8 18.Bd4 Qc6 19.a3 Bc5 20.Be5 Nd7 21.Bb5 Qc8 22.Bf4 Nf6 23.Bd3 a5 24.h3 Bd5 25.Be5 Qb7 26.b4 Bd6 27.Bxf6
= (0.22) Depth: 31 00:56:03 14034mN
11.g4
+/= (0.35 !) Depth: 31 01:50:23 28882mN


(Bram Mourik, AMD X6 1090T 3200 Mhz (Fritzmark 23,1 6x) 10.12.2014)

zullil
Posts: 5680
Joined: Mon Jan 08, 2007 11:31 pm
Location: PA USA
Full name: Louis Zulli

Re: SF novelty 10 g4!

Post by zullil » Wed Dec 10, 2014 9:27 pm

beram wrote: And how long for your system with Houdini 4 ?
No idea---I don't use Houdini. (This is why I replied to Ernest's post and not to your original one.) I'm sure someone will supply Houdini results soon.

Werewolf
Posts: 1198
Joined: Thu Sep 18, 2008 8:24 pm

Re: SF novelty 10 g4!

Post by Werewolf » Thu Dec 11, 2014 4:09 pm

beram wrote:How long does it takes the fastest hardware to find the SF novelty 11 g4! with Houdini 4.
On my AMD 1090T system using 6 cores it took about 40 minutes to find it at depth 29
It's not possible to answer your question accurately without running the test many, many times.

When chess engines are run on more than one core you get something called indeterminancy. That means you get randomness in the search. If you ran this position 10 times on your AMD machine you'd get 10 different results. On one attempt it might solve it in 5 minutes, on another attempt it might take over an hour.

My machine is a dual xeon E5-2687W (16 physical cores total @ 3.4GHz.).
It runs at 30 Million NPS and yet failed to find g4 in 20 minutes.

But I only did the test once.

jdart
Posts: 3835
Joined: Fri Mar 10, 2006 4:23 am
Location: http://www.arasanchess.org

Re: SF novelty 10 g4!

Post by jdart » Thu Dec 11, 2014 5:01 pm

The position before g4 seems to be novel in itself: I don't find it in my databases. Can you post the game?

--Jon

ernest
Posts: 1863
Joined: Wed Mar 08, 2006 7:30 pm

Re: SF novelty 10 g4!

Post by ernest » Thu Dec 11, 2014 5:11 pm

Werewolf wrote:On one attempt it might solve it in 5 minutes, on another attempt it might take over an hour.
Right, Carl !
But I found that very often, at least with Stockfish, using MultiPV=more than 1, for instance 3, you get the solution at a lower depth, and the spread (like your 5min, 1hour) is much reduced.

By the way, my Houdini 4 with MultiPV=3 has now been running for 2 hours and starting depth 30

Analysis by Houdini 4 w32 Syz:

1. = (0.22): 1.Bf4 Nd5 2.Bd2 b6 3.0-0 Bb7 4.Rad1 Nf6 5.Bf4 Qc8 6.Be5 c5 7.Rfe1 cxd4 8.Bxd4 Qc7 9.Be5 Qc5 10.c3 Rad8 11.h3 Rd5 12.Bd4 Qc7 13.a3 Rfd8 14.b4 Qc6 15.Qe3 R5d7 16.Bc2 Rd6 17.Qd3
2. = (0.17): 1.0-0 Qc7 2.g3 Bd6 3.a4 Rd8 4.a5 c5 5.dxc5 Bxc5 6.Bf4 Bd6 7.Be5 Bxe5 8.Qxe5 Qe7 9.Rfd1 Bd7 10.Qe3 Be8 11.Ne5 Nd5 12.Qe2 Rac8 13.a6 b6 14.c3 f5 15.Bc2 Nf6 16.Bb3 Ne4 17.Kg2
3. = (0.11): 1.c3 c5 2.0-0 b6 3.dxc5 Bxc5 4.Bf4 Qe7 5.Rad1 Bb7 6.Rd2 Rad8 7.Rfd1 a6 8.Bb1 Rxd2 9.Qxd2 a5 10.Bd6 Bxd6 11.Qxd6 Qxd6 12.Rxd6 Bxf3 13.gxf3 Rb8 14.f4 Nh5 15.f5 exf5 16.Bxf5 Nf4 17.f3 g6 18.Be4 f5 19.Bc2 Kg7 20.Kf2

no trace of g4!...

Post Reply