SF doesn't see mate in 8

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

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Dann Corbit
Posts: 8836
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: SF doesn't see mate in 8

Post by Dann Corbit » Thu Jun 28, 2018 7:59 pm

Gave Komodo 10 minutes:

Code: Select all

 2                      
    Avoid move: 
    Best move (Komodo-12.1.1-64bit): Kf7-g7
    Not found in: 10:00
     21	00:02	  19,200,329	9,060,524	-0.51	c6 Bd3 d5 0-0 Bd6 h4 h5 c4 Bxf4 exf4 Re8 Nc3 Be6 c5 Nd7 Re1 Nf8 g3 Bd7 Rxe8 Qxe8 Qd2 Ne6 Re1 b6 cxb6
     22-	00:04	  37,387,385	8,479,308	-0.57	c6 c4
     22-	00:04	  40,107,603	8,423,721	-0.65	c6 c4
     22+	00:05	  45,718,150	8,359,035	-0.60	c6 c4 Bd6
     22	00:05	  46,536,912	8,355,948	-0.63	c6 h4 h5 c4 Bh6 g3 Na6 Nc3 Nc7 d5 d6 Be2 Bd7 Rc1 Rc8 0-0 Bxf4 exf4 Kg7 Bf3 Re8 Qd3 Na6
     23+	00:07	  64,418,165	8,220,331	-0.57	c6 c4 Bd6
     23	00:08	  71,475,877	8,216,096	-0.51	c6 c4 Bd6 g3 Bxf4 gxf4 d5 Qb3 dxc4 Bxc4+ Kg7 Rg1 Qe7 Nd2 b5 Be2 Be6 Qc2 Kh7 h4 h5 Qd1 Bf7 a4 bxa4 Qxa4 c5 Rb1
     24-	00:12	 101,286,555	8,446,402	-0.57	c6 d5
     24-	00:14	 128,435,103	8,607,185	-0.65	c6 d5
     24+	00:15	 137,558,249	8,615,210	-0.60	c6 d5 Bd6
     24-	00:16	 144,214,995	8,648,084	-0.67	c6 d5
     24	00:18	 160,289,634	8,655,858	-0.70	c6 d5 Bd6 Nc3 b5 g3 b4 Nce2 Be5 Bg2 Bxb2 Rb1 Be5 h4 h5 Rxb4 Qa5 Qd2 Na6 Rc4 Qxd2+ Kxd2 c5 Rb1 Rb8 Rxb8 Bxb8
     24+	00:23	 199,212,498	8,655,276	-0.57	Kg7 g3 Bd6
     24	00:23	 204,499,617	8,656,310	-0.66	Kg7 g3 Bd6 h4 Bxf4 gxf4 d5 Nc3 Nc6 Rg1 a6 Qf3 Ne7 0-0-0 b5 Bd3 Qd6 Kb1 Be6 Ne2 Rhb8 h5 g5 fxg5 fxg5
     25	00:24	 216,041,419	8,662,302	-0.66	Kg7 h4 Bd6 g3 Bxf4 gxf4 d5 Nc3 Be6 Qf3 a6 0-0-0 Nc6 Rg1 Qd6 Kb1 Ne7 Bd3 Rhb8 Qg2 Kf7 a3 b5 h5 g5 fxg5
     26	00:29	 259,479,060	8,654,292	-0.65	Kg7 g3 Bd6 h4 Bxf4 gxf4 d5 Nc3 Be6 Rg1 Nc6 Qf3 Ne7 Bd3 a6 0-0-0 b5 Kb1 Qd6 Ne2 Raf8 Qg3 Rhg8 h5 g5 a3 Bf7
     27-	00:39	 345,005,878	8,718,820	-0.71	Kg7 h4
     27+	00:50	 441,785,117	8,745,125	-0.67	Kg7 c4 Nc6
     27-	01:02	 541,663,750	8,704,846	-0.72	Kg7 h4
     27+	01:17	 673,695,767	8,677,574	-0.65	Kg7 c4 Bd6
     27	01:19	 686,714,476	8,680,008	-0.64	Kg7 c4 Bd6 g3 Nc6 Nc3 b6 h4 Bb7 Bg2 h5 0-0 Na5 Qd3 Bxg2 Nxg2 Qe8 b3 Nb7 e4 fxe4 Nxe4 Qf7 Ne3 c6 a3 Rhe8 Rfe1 Bc7 b4 Nd6 Nxd6
     28-	01:27	 756,547,781	8,689,564	-0.70	Kg7 Qf3
     28	01:39	 862,417,848	8,698,057	-0.66	c6 d5 Bd6 Nc3 Bxf4 exf4 d6 Qd2 c5 Be2 Na6 0-0 Nc7 Nd1 Bd7 Ne3 Re8 Rae1 a5 c4 Kg7 h4 h5 g3 b5 a3 bxc4 Bxc4 Bb5 Bxb5
     29+	01:45	 916,447,758	8,719,771	-0.60	c6 d5 Bd6
     29-	01:47	 935,690,540	8,722,671	-0.64	c6 d5
     29-	02:01	1,060,188,114	8,745,053	-0.75	c6 d5
     29+	02:37	1,373,188,360	8,743,642	-0.59	Nc6 h4 h5
     29-	02:46	1,461,014,468	8,750,607	-0.66	Nc6 c4
     29	03:04	1,615,643,593	8,760,746	-0.71	Nc6 c4 b6 Be2 Bb4+ Nc3 Bxc3+ bxc3 Bb7 0-0 Qe7 Re1 Na5 h4 h5 c5 Kg7 Qd2 Rhe8 c4 Kh7 Rad1 Be4 f3 Bb7 Bd3 Qf7 Qc3 Rad8 Kh1 d6 cxd6
     30+	03:29	1,830,667,441	8,733,604	-0.65	Nc6 c4 Bd6
     30-	03:31	1,849,589,783	8,734,696	-0.69	Nc6 c4
     30+	03:43	1,951,284,495	8,726,140	-0.64	Nc6 c4 Bd6
     30-	03:44	1,961,752,455	8,724,647	-0.71	Nc6 c4
     30	03:49	2,004,183,700	8,724,689	-0.72	Nc6 c4 Bd6 Nc3 Ne7 Bd3 Bxf4 exf4 d5 0-0 Be6 Qb3 c6 c5 b6 Na4 Kg7 Rae1 Bf7 Re3 Re8 Rfe1 b5 Nc3 Qc7 a4 a6 Qc2 Ng8 g3 b4 Nd1 a5
     30+	03:57	2,074,248,639	8,719,744	-0.61	Kg7 h4 h5
     30	04:00	2,099,736,336	8,716,190	-0.65	Kg7 h4 h5 Qf3 Bd6 Nc3 c6 Bd3 Bxf4 Qxf4 d5 0-0 Be6 Ne2 Bf7 Qg3 Kh7 b4 Nd7 Nf4 Nb6 a4 Nc4 Bxc4 dxc4 Rab1 a6 Qf3 Qe7 b5 Qe4 Qxe4 fxe4
     31-	04:08	2,159,774,873	8,706,765	-0.71	Kg7 h4
     31+	04:32	2,370,710,795	8,694,480	-0.67	Kg7 h4 h5
     31	05:15	2,729,320,133	8,661,513	-0.69	Kg7 h4 h5 c4 b6 Qf3 Nc6 Bd3 Bd6 Nc3 Bb7 0-0-0 Qe8 Nb5 Rc8 Nxd6 cxd6 Kb1 Ba8 d5 Ne5 Qe2 Bb7 b3 Qe7 Qb2 Rhe8 Rhe1 Kg8 Be2 Kh7 g3 Kg8 a3 Qg7 Nd3 Nxd3 Bxd3
     32+	06:02	3,132,256,313	8,639,501	-0.63	Kg7 h4 h5
     32-	06:33	3,402,256,774	8,645,167	-0.67	Kg7 Qf3
     32+	06:59	3,619,189,375	8,628,831	-0.62	Kg7 h4 h5
     32-	07:05	3,672,418,269	8,623,550	-0.69	Kg7 h4
     32	09:52	5,105,358,565	8,610,801	-0.72	Kg7 c4 Nc6 Nc3 Bd6 g3 Rb8 h4 h5 Qd2 b6 0-0-0 Bb7 Rg1 Na5 Kb1 Bb4 d5 Qe8 Qc2 Ba6 Nb5 Bxb5 cxb5 Qe5 Rc1 Bd6 Bg2 Rhe8 Rgd1 Rbd8 a3 Kh7
   2018-06-28 12:58:07 PM, Time for this analysis: 00:10:00, Rated time: 20:00

0 of 2 matching moves
2018-06-28 12:58:08 PM, Total time: 12:20:04 AM
Rated time: 20:00 = 1200 Seconds
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

tpoppins
Posts: 676
Joined: Tue Nov 24, 2015 8:11 pm
Location: upstate

Re: SF doesn't see mate in 8

Post by tpoppins » Thu Jun 28, 2018 9:18 pm

Holy spam, Dann! I knew you'd like this one.

It looks like this thread could produce two related test positions:

Image
rnbq1b1r/pppp1k2/5ppp/5p2/3P1N2/4P3/PPP2PPP/RN1QKB1R b KQ - 0 7
am 8...Bd6?

BTW, how long does it take SF dev to reject 8...Bd6 and pick something else?

Image
rnbq3r/pppp1k2/3b1ppp/5p2/3P1N2/4P3/PPP2PPP/RN1QKB1R w KQ - 4 8
bm 9.Bc4!

I just went through the engines I have with the second position. A lot of them like 9.Bc4 but just a few can follow through.
Out of 300+ only these know how to turn Drawfish into tasty hardtack in ten moves at 40/40:

The Magnificent Baker's Dozen

BlackMamba 2.0 (3060 on the 40/40 list)
Bright 0.5c (2780)
Chiron 4 (3110)
Critter 1.6a (3100)
Deep Fritz 14 (3020)
Gandalf 7 (2630)
Hakkapeliitta TCEC v2 (2990)
Loop 13.6 AKA Loop 2007 (2790)
Naum 4.6 (3010)
Rybka 4 (3080)
Sting SF 9.9 (3010)
Strelka 5.5 (3100)
Wildcat 8 (2630)

and these are the notable failures:

Houdini 6 (both regular and Tactical)
Komodo 12.1.1 (both regular and MCTS)
Matefinder 9
Scorpio 2.8 MCTS
Leela Chess 0.10 w323

An honorable mention for Taltos rev118 (a mere 2400 Elo). It opted for 9.Nxg6?!! True, Drawfish gradually overcame it, but you gotta admire the spirit!

Tirsa Poppins
CCRL

User avatar
JVMerlino
Posts: 956
Joined: Wed Mar 08, 2006 9:15 pm
Location: San Francisco, California

Re: SF doesn't see mate in 8

Post by JVMerlino » Fri Jun 29, 2018 12:17 am

The King 3.12 (from CM9000) sees 8.Bc4+ and 9.Nxg6 in just 7 seconds on my crappy machine.

Code: Select all

Time	Depth	Score	Positions	Moves
Time	Depth	Score	Positions	Moves
0:00	1/3	0.86	819		8.Bc4+ Kg7 9.O-O Nc6 10.Nc3
0:00	1/4	0.79	2794		8.Bc4+ Kg7 9.O-O b6 10.Nd2 Bb7
0:00	1/5	0.77	13621		8.Bc4+ Kg7 9.Nc3 c6 10.O-O b5
0:00	1/6	0.60	32724		8.Bc4+ Kg7 9.O-O c6 10.Nd2 b5 11.Be2 Bb7
0:00	1/7	0.60	117767		8.Bc4+ Kg7 9.Nc3 c6 10.Qf3 Bxf4	11.Qxf4 d5
0:00	1/8	0.69	322726		8.Bc4+ Kg7 9.O-O c6 10.Nd2 b5 11.Bd3 Na6 12.c4 Nb4
0:02	1/9	0.67	869145		8.Bc4+ Kg7 9.Nc3 Bxf4 10.exf4 Re8+ 11.Be2 d5 12.O-O Nc6 13.Bb5 Qd6
0:07	1/10	1.75	2939503		8.Bc4+ Kg7 9.Nxg6 b5 10.Bd5 c6 11.Nh4 Bb4+ 12.c3 d6 13.Qh5 cxd5	14.Qg6+ Kf8 15.cxb4
0:09	2/11	1.92	4098233		8.Bc4+ Kg7 9.Nxg6 b5 10.Bd5 c6 11.Nh4 Bb4+ 12.c3 d6 13.Qh5 cxd5	14.Qg6+ Kf8 15.cxb4 Nc6 16.Nc3 Nxb4 17.Nxb5
0:17	3/12	2.03	8191035		8.Bc4+ Kg7 9.Nxg6 b5 10.Bd5 c6 11.Nh4 Bb4+ 12.c3 d6 13.cxb4 cxd5 14.Nc3 f4 15.Qh5 fxe3 16.Qg6+ Kf8 17.fxe3
0:37	4/13	2.09	18642325	8.Bc4+ Kg7 9.Nxg6 b5 10.Bd5 c6 11.Nh4 Bb4+ 12.c3 d6 13.Qf3 Qe8 14.Bb3 Ba5 15.Nxf5+ Bxf5 16.Qxf5



It also wants to play 7...Kg7 in the am Bd6 position.

jm

peter
Posts: 1514
Joined: Sat Feb 16, 2008 6:38 am

Re: SF doesn't see mate in 8

Post by peter » Fri Jun 29, 2018 4:55 am

tpoppins wrote:
Thu Jun 28, 2018 9:18 pm
Holy spam, Dann! I knew you'd like this one.

It looks like this thread could produce two related test positions:

Image
rnbq1b1r/pppp1k2/5ppp/5p2/3P1N2/4P3/PPP2PPP/RN1QKB1R b KQ - 0 7
am 8...Bd6?

BTW, how long does it take SF dev to reject 8...Bd6 and pick something else?

Image
rnbq3r/pppp1k2/3b1ppp/5p2/3P1N2/4P3/PPP2PPP/RN1QKB1R w KQ - 4 8
bm 9.Bc4!
Astonishing indeed!

CFish after 5' with 24 threads at the first one:

730: Fritz14,D - Wasp,3, 7...Ld6?
rnbq1b1r/pppp1k2/5ppp/5p2/3P1N2/4P3/PPP2PPP/RN1QKB1R b KQ - 0 1

Analysis by Cfish 270618 64 POP N:

7...Bd6 8.Nd3 Nc6 9.c4 Nb4 10.Nc1 Kg7 11.Nc3 b6 12.Be2 c6 13.0-0 Na6 14.Nd3 Qe7 15.Bf3 Bb7 16.Ne2 Qf7 17.Qb3 Nc7 18.g3 Rab8 19.Rfd1 Rhe8 20.Rac1 Be7 21.Qc2 d5 22.c5 bxc5 23.Nxc5 Bxc5 24.Qxc5 Ne6
+/- (0.87) Depth: 35/46 00:04:48 5505MN

and having played 7...Bd6 (?) (having the first computing in hash still):

rnbq3r/pppp1k2/3b1ppp/5p2/3P1N2/4P3/PPP2PPP/RN1QKB1R w KQ - 0 1

Analysis by Cfish 270618 64 POP N:

8.Nd3 Na6 9.Be2 c6 10.c4 Nb4 11.Nc1 Na6 12.0-0 Kg7 13.Nd3 b6 14.h4 Qe8 15.Nd2 Bb7 16.g3 Rb8 17.Bf3 Qf7 18.a3 Rhe8 19.Rc1 Nc7 20.Re1 Be7 21.Qc2 Ne6 22.Kg2 a5 23.Nf4 h5 24.Nd3 Bd6 25.b4 Be7
+/- (0.91) Depth: 35/43 00:04:27 5346MN

Awesome!
:)
Peter Martan

User avatar
MikeB
Posts: 2616
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: SF doesn't see mate in 8

Post by MikeB » Fri Jun 29, 2018 5:30 am

Nice position - thx for sharing.

Here current McBrain (not released) on a lowly Raspberry Pi Model 3B+(DGT-PI Housing) sees the mate in 8 at depth 32 and 765M nodes.
( Actually used this position to test Pi throttling - it ran for about ~20 minutes and maintained 1200 Mhz while using all 4 cores - the hardware falls in between iPad 4 and iPad 5 - Geekbench result:http://browser.geekbench.com/geekbench2/2659839),

The below analysis was captured on Mac Pro which was ssh-ed into the Pi via a WIFI connection.

Code: Select all

info depth 32 seldepth 50 multipv 1 score cp -62 upperbound nodes 732953184 nps 642767 hashfull 999 tbhits 0 time 1140309 pv c4d3 b8c6
info depth 32 currmove c4d3 currmovenumber 1
info depth 32 seldepth 50 multipv 1 score cp -50 lowerbound nodes 740863191 nps 642699 hashfull 999 tbhits 0 time 1152737 pv c4d3
info depth 32 currmove c4d3 currmovenumber 1
info depth 32 currmove d1f3 currmovenumber 2
info depth 32 currmove b1d2 currmovenumber 3
info depth 32 currmove d1g4 currmovenumber 4
info depth 32 currmove e1g1 currmovenumber 5
info depth 32 currmove c2c3 currmovenumber 6
info depth 32 currmove e1f1 currmovenumber 7
info depth 32 currmove b1c3 currmovenumber 8
info depth 32 currmove e1d2 currmovenumber 9
info depth 32 currmove e3e4 currmovenumber 10
info depth 32 currmove d4d5 currmovenumber 11
info depth 32 currmove d1h5 currmovenumber 12
info depth 32 seldepth 50 multipv 1 score mate 8 lowerbound nodes 764774602 nps 644778 hashfull 999 tbhits 0 time 1186105 pv d1h5
info depth 32 currmove d1h5 currmovenumber 1
info depth 32 currmove c4d3 currmovenumber 2
info depth 32 currmove h2h3 currmovenumber 3
info depth 32 currmove e1e2 currmovenumber 4
info depth 32 currmove c2c3 currmovenumber 5
info depth 32 currmove b1c3 currmovenumber 6
info depth 32 currmove b1d2 currmovenumber 7
info depth 32 currmove b2b3 currmovenumber 8
info depth 32 currmove d1f3 currmovenumber 9
info depth 32 currmove h1g1 currmovenumber 10
info depth 32 currmove d1g4 currmovenumber 11
info depth 32 currmove f2f4 currmovenumber 12
info depth 32 currmove g2g3 currmovenumber 13
info depth 32 currmove c4f7 currmovenumber 14
info depth 32 currmove e1d2 currmovenumber 15
info depth 32 currmove e3e4 currmovenumber 16
info depth 32 currmove c4e6 currmovenumber 17
info depth 32 currmove f2f3 currmovenumber 18
info depth 32 currmove e1g1 currmovenumber 19
info depth 32 currmove a2a3 currmovenumber 20
info depth 32 currmove c4b3 currmovenumber 21
info depth 32 currmove g2g4 currmovenumber 22
info depth 32 currmove b2b4 currmovenumber 23
info depth 32 currmove d1d3 currmovenumber 24
info depth 32 currmove e1f1 currmovenumber 25
info depth 32 currmove c4e2 currmovenumber 26
info depth 32 currmove h2h4 currmovenumber 27
info depth 32 currmove c4d5 currmovenumber 28
info depth 32 currmove d1e2 currmovenumber 29
info depth 32 currmove d4d5 currmovenumber 30
info depth 32 currmove c4f1 currmovenumber 31
info depth 32 currmove c4b5 currmovenumber 32
info depth 32 currmove h1f1 currmovenumber 33
info depth 32 currmove d1c1 currmovenumber 34
info depth 32 currmove a2a4 currmovenumber 35
info depth 32 currmove b1a3 currmovenumber 36
info depth 32 currmove c4g8 currmovenumber 37
info depth 32 currmove d1d2 currmovenumber 38
info depth 32 currmove c4a6 currmovenumber 39
info depth 32 seldepth 50 multipv 1 score mate 8 nodes 764792955 nps 644778 hashfull 999 tbhits 0 time 1186132 pv d1h5 g6h5 c4f7 h5h4 b1d2 d6b4 c2c3 f5f4 d2f3 h4g4 e3e4 b4c3 b2c3 h6h5 h2h3

peter
Posts: 1514
Joined: Sat Feb 16, 2008 6:38 am

Re: SF doesn't see mate in 8

Post by peter » Fri Jun 29, 2018 6:12 am

Dann Corbit wrote:
Thu Jun 28, 2018 7:59 pm
Gave Komodo 10 minutes:
Here's SF MatefFinder with Study Mode off (Default):

730: Fritz14,D - Wasp,3, 7...Ld6?
rnbq1b1r/pppp1k2/5ppp/5p2/3P1N2/4P3/PPP2PPP/RN1QKB1R b KQ - 0 1

Analysis by MateFinder 280618 64:

7...Bd6 8.Nd3 Kg7 9.c4 b6 10.Nc3 Bb7
+/- (1.03) Depth: 7/7 00:00:18 103kN
...
7...Bd6 8.Qf3 Nc6 9.Nc3 Bxf4 10.Qxf4 d5 11.Nb5 Nb4 12.0-0-0 c6 13.Nd6+ Ke7 14.Nxc8+ Qxc8 15.Kb1 h5 16.c4 Qb8 17.Qf3 Qd6 18.a3 Na6 19.Bd3 Nc7 20.cxd5 cxd5 21.h3 Kf7 22.g4 fxg4 23.hxg4 h4 24.e4
+/- (0.79) Depth: 29/38 00:01:17 1029MN
7...Bd6 8.Bc4+
+/- (0.87 --) Depth: 30/31 00:01:26 1184MN
7...Bd6 8.Bc4+
+/- (0.94 --) Depth: 30/31 00:01:44 1490MN
7...Kg7
+/- (0.87 ++) Depth: 30/38 00:02:10 1918MN
7...Kg7 8.h4 Bd6 9.Qf3 Bxf4 10.Qxf4 d6 11.Nc3 Nc6 12.0-0-0 Ne7 13.d5 Bd7 14.Qb4 Rb8 15.Ne2 b5 16.Nf4 Ra8 17.Qd4 Qb8 18.c4 bxc4 19.Bxc4 Qb6 20.Qc3 Rab8 21.a3 a5 22.Kb1
+/- (0.83) Depth: 30/38 00:02:23 2140MN
...
7...c6 8.c4 Bd6 9.Nd3 Na6 10.g3 h5 11.Nc3 b6 12.h4 Kg7 13.Bg2 Qg8 14.b3 Ba3 15.Qd2 Bb7 16.Nf4 Bd6 17.0-0 Qf7 18.Nd3 Ba3 19.Rad1 Rae8 20.d5 Bd6 21.Ne2 cxd5 22.Ndf4 Bb4 23.Bxd5 Bxd2 24.Bxf7 Kxf7 25.Rxd2 Nc5
+/- (0.92) Depth: 33/40 00:06:57 6811MN
7...h5
+/- (0.85 ++) Depth: 34/41 00:14:03 14056MN

And after GUI (DF14) restart with Study Mode on:

rnbq1b1r/pppp1k2/5ppp/5p2/3P1N2/4P3/PPP2PPP/RN1QKB1R b KQ - 0 1

Analysis by MateFinder 280618 64:

7...Bd6 8.Bc4+ Kg7 9.Nc3 Bxf4 10.exf4 d5 11.Nxd5
+/- (0.95) Depth: 7/8 00:00:13 72kN
...
7...Bd6 8.Qf3 Bxf4 9.Qxf4 Nc6 10.Nc3 Ne7 11.d5 d6 12.0-0-0 Bd7 13.Kb1 a6 14.Bd3 Kg7 15.Rhe1 g5 16.Qd4 Qb8 17.Ne2 Qa7 18.Qxa7 Rxa7 19.Nd4 Raa8 20.g3
+/- (0.76) Depth: 25/32 00:00:28 277MN
7...Bd6 8.Qf3
+/- (0.83 --) Depth: 26/33 00:00:31 324MN
7...Bd6 8.Qf3
+/- (0.91 --) Depth: 26/34 00:00:38 438MN
7...Kg7
+/- (0.83 ++) Depth: 26/39 00:01:14 1057MN
7...Kg7 8.h4 Bd6 9.Qf3 Bxf4 10.Qxf4 d6 11.Nc3 Nc6 12.0-0-0 Bd7 13.Qf3 Rb8 14.Nd5 Ne7 15.Nf4 Qg8 16.Kb1 b5 17.d5 Qf7 18.Be2 a5 19.e4 fxe4 20.Qxe4 c6 21.dxc6 Bxc6 22.Qd3 Nf5 23.Qa3
+/- (0.83) Depth: 26/39 00:01:21 1174MN
7...Kg7 8.h4
+/- (0.90 --) Depth: 27/40 00:01:21 1184MN
7...c6
+/- (0.83 ++) Depth: 27/40 00:01:55 1760MN
7...Kg7 8.Nc3 Bd6 9.Qf3 Nc6 10.0-0-0 Bxf4 11.Qxf4 d6 12.Kb1 Ne7 13.h4 h5 14.Bd3 Bd7 15.Ne2 Nd5 16.Qf3 Qg8 17.c4 Nb4 18.Nf4 b5 19.Be2 bxc4 20.d5 Nd3 21.Bxd3 cxd3 22.Rxd3
+/- (0.87) Depth: 27/40 00:02:02 1893MN
7...c6 8.c4 Bd6 9.Nd3 Na6 10.Nc3 Kg7 11.g3 b6 12.Bg2 Qe7 13.0-0 Qf7 14.Qb3 Nc7 15.Ne2 Be7 16.Qc2 Bb7 17.Rfc1 Rhb8 18.a3 d5 19.c5 bxc5 20.dxc5 Ne6 21.b4 Ba6 22.Nd4 Nxd4
+/- (0.87) Depth: 28/40 00:02:21 2203MN
...
7...Kg7
+/- (0.90 ++) Depth: 33/42 00:06:24 6315MN
7...Kg7 8.Nc3 Bd6 9.Qf3 Bxf4 10.Qxf4 d6 11.0-0-0 Nc6 12.h4 Bd7 13.Bd3 Nb4 14.Bc4 Nc6 15.Kb1 Ne7 16.Bd3 Rb8 17.Rhe1 b5 18.Qf3 a5 19.Ne2 Qg8 20.e4 fxe4 21.Qxe4 Qf7 22.Nf4 Rhe8 23.d5 f5 24.Qd4+ Kg8 25.Qa7 Nxd5 26.Nxd5 Qxd5
+/- (0.90) Depth: 33/50 00:06:38 6546MN


Both with 24 threads of a 12x3GHz Intel CPU, 32G hash.

Should have given the two settings some more tries to see some SMP- variance, but I guess the difference will be somewhat reproducible.

Nullmove could matter, LMRs probably do so too, mate-finding code at the first and hardest one avoid move position, I'm not sure about neither.
Will try now with komodo and its corresponding UCI- options.

Edit: Quick testing didn't show help of switching off nullmove neither of LMRs (only) as for komodo and 7...Bd6 within 5 minutes with empty hash started.
So it seems find-mate-code could yet be MateFinders trick here only too.
:)
Peter Martan

peter
Posts: 1514
Joined: Sat Feb 16, 2008 6:38 am

Re: SF doesn't see mate in 8

Post by peter » Fri Jun 29, 2018 6:44 am

peter wrote:
Fri Jun 29, 2018 6:12 am
Edit: Quick testing didn't show help of switching off nullmove neither of LMRs (only) as for komodo and 7...Bd6 within 5 minutes with empty hash started.
So it seems find-mate-code could yet be MateFinders trick here only too.
:)
Edit edit:
Or yet not, here's CorChess (less nullmove and less LMRs than SF dev.?, CorChess doesn't use find-mate-code, does it?)

rnbq1b1r/pppp1k2/5ppp/5p2/3P1N2/4P3/PPP2PPP/RN1QKB1R b KQ - 0 1

Analysis by CorChess 2.4 180618 64 POPCNT:

7...d5 8.c4 Bd6 9.Nxd5 c6 10.Ndc3 Bb4
+/- (0.74) Depth: 7/7 00:00:00 29kN
7...Kg7 8.Bd3 d5 9.c4 Bd6 10.Nxd5 c6 11.Ndc3
+/- (0.94) Depth: 8/8 00:00:00 49kN
7...Kg7 8.Bd3 d5 9.c4 dxc4 10.Bxc4 Bd6 11.0-0 c6
+/- (0.95) Depth: 9/10 00:00:00 81kN
7...Bd6 8.g3 c6 9.h4 Bxf4 10.gxf4 h5 11.c4 d6 12.Nc3 Nd7
+/- (0.87) Depth: 10/12 00:00:00 243kN
...
7...Bd6 8.Nd3 b6 9.c4 Na6 10.Nc3 Kg7 11.Be2 c6 12.0-0 Qf8 13.Re1 Nc7 14.a3 Ne6 15.b4 Bb7 16.f4 Qf7 17.Rb1 Ba6 18.b5 cxb5 19.Nxb5 Bxb5 20.cxb5 Bxa3 21.Ra1 Bd6 22.Bf3 Nc7 23.Bxa8 Rxa8 24.Qa4 Qd5
+/= (0.68) Depth: 33/41 00:01:56 2161MN
7...Bd6 8.Bc4+
+/- (0.81 --) Depth: 34/45 00:03:14 3574MN
7...c6 8.d5 Bd6 9.Nc3 Bxf4 10.exf4 d6 11.h4 h5 12.Rh3 a6 13.Re3 Re8 14.Qe2 Rxe3 15.Qxe3 c5 16.a4 Ra7 17.0-0-0 b6 18.Bc4 Re7 19.Qd2 Re8 20.Re1 Rxe1+ 21.Qxe1 Qe7 22.Qh1 Bb7 23.g3 a5 24.Qf1 Nd7 25.Bb5 Bc8 26.Kb1 Kg7 27.Bc6
+/- (0.76) Depth: 34/48 00:06:39 7351MN
...
7...c6 8.c4 Bd6 9.Nd3 Na6 10.h4 h5 11.g3 Kg7 12.Bg2 Nb4 13.Nc1 b6 14.0-0 Na6 15.Re1 Qg8 16.Nd2 Qf7 17.a3 Bb7 18.Nd3 Nc7 19.Qc2 Be7 20.a4 d5 21.a5 Rad8 22.axb6 axb6 23.Ra7 Rb8 24.c5 bxc5 25.Nxc5
+/- (0.82) Depth: 35/45 00:09:50 10886MN

BTW Fritz15 (Rybka 4.2) is very good at this.
Started with empty hash it doesn't show 7...Bd6 (?) in output at all, to be sure, it's for the right reason, give it more than 20 MV- lines and you'll see the white winning eval and 8.Bc4! at once too for 7...Bd6 at order-nr.19 or so.
Peter Martan

peter
Posts: 1514
Joined: Sat Feb 16, 2008 6:38 am

Re: SF doesn't see mate in 8

Post by peter » Fri Jun 29, 2018 11:26 am

peter wrote:
Fri Jun 29, 2018 6:12 am
Edit: Quick testing didn't show help of switching off nullmove neither of LMRs (only) as for komodo and 7...Bd6 within 5 minutes with empty hash started.
So it seems find-mate-code could yet be MateFinders trick here only too.
:)
As for komodo it seems to be a matter of King Safety only. Here set to 200 (max.):


Hash table size is now 32768 meg
Threads now set to 24
Found 510 tablebases (up to 6-man).
Probing tablebases up to 6-man.
Engine: Komodo 12.1.1 64-bitN (32768 MB)
by Don Dailey, Larry Kaufman, Mark Lefler
20.00 0:01 +0.78 1...d5 2.Qf3 c6 3.c4 Bd6 4.Nc3 Bxf4 5.Qxf4 Be6 6.cxd5 cxd5 7.Bd3 Nc6 8.O-O Kg7 9.Rac1 Rc8 10.a3 Qc7 11.Ne2 Qxf4 12.Nxf4 Bf7 13.Rc5 b6 (20.969.747) 15450
...
23.01 0:06 +0.80-- 1...d5 2.Nd2 (92.249.792) 13318
23.06 0:09 +0.68++ 1...Bd6 2.Nxg6 Kxg6 (125.461.183) 13141
23.01 0:11 +0.60++ 1...Bd6 2.Nxg6 Kxg6 (151.983.050) 13077
23.01 0:13 +0.65-- 1...Bd6 2.Qf3 (170.999.966) 13033
...
24.01 0:26 +0.90-- 1...Bd6 2.Qf3 (345.277.335) 13081
24.02 0:29 +0.89 1...d5 2.Nd2 Bd6 3.Nxd5 c6 4.Nc3 Kg7 5.Bc4 Qe7 6.O-O b5 7.Bd3 Be6 8.a4 b4 9.Ne2 Nd7 10.c4 c5 11.a5 Bf7 12.Ng3 Rad8 13.Bxf5 Bxg3 14.Bxd7 Rxd7 (379.480.457) 13057
...
26.01 0:43 +0.90-- 1...d5 2.Qf3 (564.010.438) 12972
26.02 0:47 +0.78++ 1...Bd6 2.Qf3 Bxf4 (620.017.160) 12979
...
27.01 0:54 +0.73 1...Bd6 2.Qf3 Bxf4 3.Qxf4 d5 4.c4 c6 5.cxd5 cxd5 6.Bd3 Nc6 7.O-O Nb4 8.Be2 Be6 9.Nc3 Kg7 10.Rac1 Nc6 11.Bd3 Rc8 12.Ne2 Bf7 13.Ng3 Ne7 14.h4 h5 (709.744.184) 12984
28.01 0:55 +0.79-- 1...Bd6 2.Bc4+ (719.990.047) 12989
...
28.02 1:15 +0.91++ 1...d5 2.h4 Kg7 (972.188.059) 12950
...
30.01 4:31 +1.05 1...d5 2.h4 h5 3.c4 Bd6 4.Qf3 c6 5.Nc3 Bxf4 6.Qxf4 Be6 7.cxd5 cxd5 8.Bd3 Nc6 9.Rc1 Nb4 10.Bb1 Rc8 11.O-O Qc7 12.Qf3 Qd6 13.Ne2 Nc6 14.Nf4 Ne7 (3.611.673.187) 13300
31.01 4:37 +0.99++ 1...d5 2.h4 h5 (3.694.849.566) 13299
best move: d7-d5 time: 4:38.103 min n/s: 13.299.477 nodes: 3.698.607.091
Peter Martan

peter
Posts: 1514
Joined: Sat Feb 16, 2008 6:38 am

Re: SF doesn't see mate in 8

Post by peter » Fri Jun 29, 2018 2:56 pm

peter wrote:
Fri Jun 29, 2018 6:44 am
here's CorChess
Pity, this one I could not reproduce.
Maybe it was big SMP- fortune, maybe I had hash- remains from other engines tested before.
Peter Martan

Henrik Dinesen
Posts: 874
Joined: Wed Mar 08, 2006 8:52 pm
Location: Denmark

Re: SF doesn't see mate in 8

Post by Henrik Dinesen » Fri Jun 29, 2018 11:53 pm

It's obviously a problem for several strong engines, but then again it points to the many holes in the Dutch defence.
Henrik

Post Reply