Solutions Stockfish should find 100 times faster

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

Moderator: Ras

Vinvin
Posts: 5287
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Solutions Stockfish should find 100 times faster

Post by Vinvin »

Let's make a list (you can add your positions too) :

1.Bd7! Mate in 5
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

rk6/p1r3p1/P3B1K1/1p2B3/8/8/8/8 w - -
1.Re5! going for the mate
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

r1b3kr/pp1n2Bp/2pb2q1/3p3N/3P4/2P2Q2/P1P3PP/4RRK1 w - - 0 1
1.Kg3! going to g5
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

2r2rk1/1bpR1p2/1pq1pQp1/p3P2p/P1PR3P/5N2/2P2PPK/8 w - - 0 1
1.Ba4+! closing the position
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

3B4/1r2p3/r2p1p2/bkp1P1p1/1p1P1PPp/p1P1K2P/PPB5/8 w - - 0 1
1.Rxf6!
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

1q1r3k/3P1pp1/ppBR1n1p/4Q2P/P4P2/8/5PK1/8 w - - 0 1
Qf6+! going for the mate !
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

r2q1r2/1b2bpkp/p3p1p1/2ppP1P1/7R/1PN1BQR1/1PP2P1P/4K3 w - - 0 1
1.Qe6!
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

5rk1/2p2Rpq/p2p2n1/2pPpQPp/4P2P/3P2N1/PP6/6K1 w - - 0 1
1.Rg4+! going for the mate
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

8/4K2p/6pr/5pk1/1pB2R2/p4PP1/8/8 w - - 0 1
1.Kd5!, freeing the way to trap the queen
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

1k3b1q/pP2p1p1/P1K1P1Pp/7P/2B5/8/8/8 w - - 0 1
1.Bf8+!, closing the king
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

6R1/8/2pB3k/2P4p/5p1q/5P2/4P1K1/8 w - - 0 1
1.Rxb3! freeing the way to invade
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

1r6/4k3/r2p2p1/2pR1p1p/2P1pP1P/pPK1P1P1/P7/1B6 b - - 0 1

8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

2b1r3/r2ppN2/8/1p1p1k2/pP1P4/2P3R1/PP3PP1/2K5 w - - 0 1
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

4q1kr/p6p/1prQPppB/4n3/4P3/2P5/PP2B2P/R5K1 w - - 0 1
1.Bg4! is draw, 1.h3 lose on 1...g4!.
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

8/8/8/5Bp1/7k/8/4pPKP/8 w - - 0 4
zullil
Posts: 6442
Joined: Tue Jan 09, 2007 12:31 am
Location: PA USA
Full name: Louis Zulli

Re: Solutions Stockfish should find 100 times faster

Post by zullil »

Vinvin wrote:Let's make a list (you can add your positions too) :

1.Bd7! Mate in 5
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

rk6/p1r3p1/P3B1K1/1p2B3/8/8/8/8 w - -
Can Stockfish solve this one at all? Here's a single-threaded (deterministic) search with a 1 GB hash table, using the most recent developmental version of Stockfish. Still not seeing at depth 52:

Code: Select all

info depth 52 seldepth 49 multipv 1 score cp 0 nodes 12269334342 nps 3613234 hashfull 999 tbhits 0 time 3395665 pv e6d5 b8c8 d5a8 c7c4 g6f5 c8d7 e5g7 d7c7 a8d5 c4h4 g7f6 h4a4 d5b7 c7d6 f6b2 a4c4 b7e4 c4a4 e4b7
With null move reductions disabled, the solution is found:

Code: Select all

info depth 28 seldepth 43 multipv 1 score mate 5 lowerbound nodes 112916725 nps 2928642 hashfull 435 tbhits 0 time 38556 pv e6d7
F. Bluemers
Posts: 880
Joined: Thu Mar 09, 2006 11:21 pm
Location: Nederland

Re: Solutions Stockfish should find 100 times faster

Post by F. Bluemers »

8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

r1b3kr/pp1n2Bp/2pb2q1/3p3N/3P4/2P2Q2/P1P3PP/4RRK1 w - - 0 1

latest Dirty

Code: Select all

 20	+8.71 	1.95G  8:06.44	Bh6 b6 Re5 Ba6 Rg5 Rf8 Bxf8 Bxh2+ Kxh2 Bxf1 Ba3 Qxg5 Qf5 Qxf5 Nf6+ Nxf6 Bd6 Qxc2 Ba3 
 20	+7.47 	126.1M	0:52.57	c4 dxc4 Qe2 Ne5 Bxe5 Bg4 Qxg4 Qxg4 Nf6+ Kg7 Nxg4+ Bxe5 Rxe5 Rhg8 Re7+ Kh8 Ne5 Rg7 Rxg7 Kxg7 Rf7+ Kh6 Rxb7 Kg5 
 19	+7.31 	90.0M  	0:37.20	c4 dxc4 Qe2 Ne5 Bxe5 Bg4 Qxg4 Qxg4 Nf6+ Kg7 Nxg4+ Bxe5 Rxe5 Rhf8 Re7+ Kg6 Ne5+ Kh6 Rxb7 Rxf1+ Kxf1 Rf8+ Ke2 Kg5 Rxa7 
 19	+7.31 	73.0M  	0:30.78	c4 dxc4 Qe2 Ne5 Bxe5 Bg4 Qxg4 Qxg4 Nf6+ Kg7 Nxg4+ Bxe5 Rxe5 Rhf8 Re7+ Kg6 Ne5+ Kh6 Rxb7 Rxf1+ Kxf1 Rf8+ Ke2 Kg5 Rxa7 
 18	+7.38 	56.3M  	0:23.54	c4 dxc4 Qe2 Ne5 Bxe5 Bg4 Qxg4 Qxg4 Nf6+ Kg7 Nxg4+ Bxe5 Rxe5 Rhf8 Re7+ Kg6 Ne5+ Kg5 Rg7+ Kh6 Rxb7 
 18	+7.38 	48.0M  	0:20.35	c4 dxc4 Qe2 Ne5 Bxe5 Bg4 Qxg4 Qxg4 Nf6+ Kg7 Nxg4+ Bxe5 Rxe5 Rhf8 Re7+ Kg6 Ne5+ Kg5 Rg7+ Kh6 Rxb7 
 17	+7.38 	38.0M  	0:15.98	c4 dxc4 Qe2 Ne5 Bxe5 Bg4 Qxg4 Qxg4 Nf6+ Kg7 Nxg4+ Bxe5 Rxe5 Rhf8 Re7+ Kg6 Ne5+ Kg5 Rg7+ Kh6 Rxb7 Rxf1+ Kxf1 
 17	+7.38 	33.9M  	0:14.44	c4 dxc4 Qe2 Ne5 Bxe5 Bg4 Qxg4 Qxg4 Nf6+ Kg7 Nxg4+ Bxe5 Rxe5 Rhf8 Re7+ Kg6 Ne5+ Kg5 Rg7+ Kh6 Rxb7 Rxf1+ Kxf1 
 16	+6.38 	18.8M  	0:07.86	c4 b6 cxd5 Ba6 Re6 Bxf1 Rxg6 hxg6 Bxh8 gxh5 dxc6 Kxh8 cxd7 Rf8 Qxh5+ Kg7 Qg5+ Kh7 
 16	+6.38 	18.7M  	0:07.79	c4 b6 cxd5 Ba6 Re6 Bxf1 Rxg6 hxg6 Bxh8 gxh5 dxc6 Kxh8 cxd7 Rf8 Qxh5+ Kg7 Qg5+ Kh7 
 16	+5.59 	7.76M  	0:03.15	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Re7 Rb8 Rdd7 Bg6 Rxb7 Rxb7 
 15	+5.32 	6.85M  	0:02.75	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Re7 Rb8 Rdd7 Kg8 Rxb7 
 15	+5.32 	3.02M  	0:01.23	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Re7 Rb8 Rdd7 Kg8 Rxb7 
 14	+5.32 	2.59M  	0:01.04	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Re7 Rb8 Rdd7 Kg8 
 14	+5.32 	1.74M  	0:00.72	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Re7 Rb8 Rdd7 Kg8 
 13	+5.32 	1.74M  	0:00.72	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Re7 Rb8 Rdd7 Kg8 
 13	+5.32 	1.34M  	0:00.56	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Re7 Rb8 Rdd7 Kg8 
 12	+4.81 	937980	0:00.39	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Rd7 Rb8 
 12	+4.81 	472603	0:00.21	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Rd7 Rb8 
 11	+4.81 	472602	0:00.21	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Rd7 Rb8 
 11	+4.81 	395488	0:00.18	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Rd7 Rb8 
 10	+4.29 	261342	0:00.11	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Rd7 Re8 
 10	+4.29 	208985	0:00.09	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Rd7 Re8 
  9	+4.01 	158295	0:00.07	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 
  9	+4.01 	142769	0:00.06	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 
  8	+4.01 	127690	0:00.06	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 
too many check extensions,I guess :lol:
disabled check extension:

Code: Select all

23	+18.06 	174.5M	1:05.95	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Qg4 Bf4 Rxf4 Re1+ Kf2 Nf6 Nxf6+ Kf7 
 23	+18.06 	163.7M	1:02.51	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Qg4 Bf4 Rxf4 Re1+ Kf2 Nf6 Nxf6+ Kf7 
 22	+19.86 	144.4M	0:55.37	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Qg4 Bf4 Rxf4 Re1+ Kf2 Nf6 Nxf6+ Kf7 
 22	+19.86 	137.5M	0:53.18	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Qg4 Bf4 Rxf4 Re1+ Kf2 Nf6 Nxf6+ Kf7 
 21	+15.54 	80.3M  	0:31.17	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Qg4 Bf4 Rxf4 Kh7 Qf5+ Kg8 Be5 Rg7 Nxg7 
 21	+15.54 	76.3M  	0:29.87	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Qg4 Bf4 Rxf4 Kh7 Qf5+ Kg8 Be5 Rg7 Nxg7 
 20	+6.42 	27.8M  	0:10.85	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Bxh6 Ba3 Bg5 Rh7 Qe2 Bf8 Qe6+ Kh8 
 20	+6.42 	24.1M  	0:09.63	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Bxh6 Ba3 Bg5 Rh7 Qe2 Bf8 Qe6+ Kh8 
 19	+6.42 	18.7M  	0:07.39	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Bxh6 Ba3 Bg5 Rh7 Qe2 Bf8 Qe6+ 
 19	+6.42 	17.0M  	0:06.79	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Bxh6 Ba3 Bg5 Rh7 Qe2 Bf8 Qe6+ 
 18	+6.42 	14.4M  	0:05.75	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Bxh6 Ba3 Bg5 Rh7 Qe2 Bf8 Qe6+ 
 18	+6.42 	13.1M  	0:05.32	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Bxh6 Ba3 Bg5 Rh7 Qe2 Bf8 Qe6+ 
 17	+6.22 	9.72M  	0:03.84	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Bxh6 Kh7 Bg5 Re4 Qf5+ 
 17	+6.22 	9.22M  	0:03.66	Re5 h6 Rf5 Kh7 Rf7 Re8 Be5+ Kg8 Rg7+ Qxg7 Bxg7 Re7 Bxh6 Kh7 Bg5 Re4 Qf5+ 
 17	+4.98 	3.57M  	0:01.39	Qe2 Nf8 Bxf8 Bg4 Nf6+ Kxf8 Nxg4+ Kg7 Rf6 Bxh2+ Kxh2 Qh5+ Kg3 Rae8 Qf1 Rhf8 Re7+ 
 16	+5.10 	2.97M  	0:01.15	Qe2 Nf8 Bxf8 Bg4 Nf6+ Kxf8 Nxg4+ Kg7 Rf6 Bxh2+ Kxh2 Qh5+ Kg3 Rae8 Qf2 h6 
 16	+5.10 	2.34M  	0:00.95	Qe2 Nf8 Bxf8 Bg4 Nf6+ Kxf8 Nxg4+ Kg7 Rf6 Bxh2+ Kxh2 Qh5+ Kg3 Rae8 Qf2 h6 
 15	+5.35 	2.16M  	0:00.86	Qe2 Nf8 Bxf8 Bg4 Nf6+ Kxf8 Nxg4+ Kg7 Rf6 Bxh2+ Kxh2 Qh5+ Kg3 Rae8 Qf1 
 15	+5.35 	1.93M  	0:00.78	Qe2 Nf8 Bxf8 Bg4 Nf6+ Kxf8 Nxg4+ Kg7 Rf6 Bxh2+ Kxh2 Qh5+ Kg3 Rae8 Qf1 
 14	+4.91 	920784	0:00.37	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Rd7 b6 Rde7 Kg8 
 14	+4.91 	709540	0:00.30	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Rd7 b6 Rde7 Kg8 
 13	+4.91 	709539	0:00.30	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Rd7 b6 Rde7 Kg8 
 13	+4.91 	584408	0:00.25	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Rd7 b6 Rde7 Kg8 
 12	+4.81 	378901	0:00.16	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Rd7 Rb8 
 12	+4.81 	278830	0:00.13	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Rd7 Rb8 
 11	+4.69 	218891	0:00.10	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Rd7 
 11	+4.69 	185313	0:00.09	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Bxh8 Kxh8 Rd7 
 10	+4.29 	155474	0:00.07	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Rd7 Re8 
 10	+4.29 	140816	0:00.07	Qe2 Nf6 Rxf6 Bg4 Rxg6 Bxe2 Rxd6 Bxh5 Rd7 Re8 
Vinvin
Posts: 5287
Joined: Thu Mar 09, 2006 9:40 am
Full name: Vincent Lejeune

Re: Solutions Stockfish should find 100 times faster

Post by Vinvin »

zullil wrote:
Vinvin wrote:Let's make a list (you can add your positions too) :

1.Bd7! Mate in 5
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

rk6/p1r3p1/P3B1K1/1p2B3/8/8/8/8 w - -
Can Stockfish solve this one at all? Here's a single-threaded (deterministic) search with a 1 GB hash table, using the most recent developmental version of Stockfish. Still not seeing at depth 52:

Code: Select all

info depth 52 seldepth 49 multipv 1 score cp 0 nodes 12269334342 nps 3613234 hashfull 999 tbhits 0 time 3395665 pv e6d5 b8c8 d5a8 c7c4 g6f5 c8d7 e5g7 d7c7 a8d5 c4h4 g7f6 h4a4 d5b7 c7d6 f6b2 a4c4 b7e4 c4a4 e4b7
With null move reductions disabled, the solution is found:

Code: Select all

info depth 28 seldepth 43 multipv 1 score mate 5 lowerbound nodes 112916725 nps 2928642 hashfull 435 tbhits 0 time 38556 pv e6d7
A lot engines have problem with this solution.
A friend rated 1600 Elo found the solution in a couple of minutes...
BBauer
Posts: 658
Joined: Wed Mar 08, 2006 8:58 pm

Re: Solutions Stockfish should find 100 times faster

Post by BBauer »

Good question.

I changed R in search.cpp from
Depth R = ((823 + 67 * depth) / 256 + std::min((eval - beta) / PawnValueMg, 3)) * ONE_PLY;
to
Depth R = depth<12 ? (2 + depth/2 ) * ONE_PLY : (5 + depth/6) * ONE_PLY;

which is faster and finds the right move on a high depth.

Kind regards
Benhard
F. Bluemers
Posts: 880
Joined: Thu Mar 09, 2006 11:21 pm
Location: Nederland

Re: Solutions Stockfish should find 100 times faster

Post by F. Bluemers »

zullil wrote:
Vinvin wrote:Let's make a list (you can add your positions too) :

1.Bd7! Mate in 5
8 
7 
6 
5 
4 
3 
2 
1 
abcdefgh

rk6/p1r3p1/P3B1K1/1p2B3/8/8/8/8 w - -
Can Stockfish solve this one at all? Here's a single-threaded (deterministic) search with a 1 GB hash table, using the most recent developmental version of Stockfish. Still not seeing at depth 52:

Code: Select all

info depth 52 seldepth 49 multipv 1 score cp 0 nodes 12269334342 nps 3613234 hashfull 999 tbhits 0 time 3395665 pv e6d5 b8c8 d5a8 c7c4 g6f5 c8d7 e5g7 d7c7 a8d5 c4h4 g7f6 h4a4 d5b7 c7d6 f6b2 a4c4 b7e4 c4a4 e4b7
With null move reductions disabled, the solution is found:

Code: Select all

info depth 28 seldepth 43 multipv 1 score mate 5 lowerbound nodes 112916725 nps 2928642 hashfull 435 tbhits 0 time 38556 pv e6d7

Code: Select all

26	    #5 	38.4M  	0:12.37	Bd7 b4 Kf7 b3 Ke7 g5 Kd8 g4 Bxc7# 
 25	    #5 	38.4M  	0:12.37	Bd7 b4 Kf7 b3 Ke7 g5 Kd8 g4 Bxc7# 
 25	    #5 	38.4M  	0:12.36	Bd7 b4 Kf7 b3 Ke7 g5 Kd8 g4 Bxc7# 
 24	+4.69 	38.1M  	0:12.29	Bd7 b4 Kxg7 b3 Kf8 b2 Bf5 b1=Q Bxb1 Kc8 Ke8 Kb8 Bd3 Kc8 Be4 Kb8 Bxc7+ Kxc7+ Bxa8 Kb6 Bb7 Kc5 Ke7 Kb5 Kf6 Kb6 Kf5 Kb5 Kf4 
 24	+4.69 	38.0M  	0:12.28	Bd7 b4 Kxg7 b3 Kf8 b2 Bf5 b1=Q Bxb1 Kc8 Ke8 Kb8 Bd3 Kc8 Be4 Kb8 Bxc7+ Kxc7+ Bxa8 Kb6 Bb7 Kc5 Ke7 Kb5 Kf6 Kb6 Kf5 Kb5 Kf4 
 23	+4.69 	38.0M  	0:12.28	Bd7 b4 Kxg7 b3 Kf8 b2 Bf5 b1=Q Bxb1 Kc8 Ke8 Kb8 Bd3 Kc8 Be4 Kb8 Bxc7+ Kxc7+ Bxa8 Kb6 Bb7 Kc5 Ke7 Kb5 Kf6 Kb6 Kf5 Kb5 Kf4 
 23	+4.69 	38.0M  	0:12.28	Bd7 b4 Kxg7 b3 Kf8 b2 Bf5 b1=Q Bxb1 Kc8 Ke8 Kb8 Bd3 Kc8 Be4 Kb8 Bxc7+ Kxc7+ Bxa8 Kb6 Bb7 Kc5 Ke7 Kb5 Kf6 Kb6 Kf5 Kb5 Kf4 
 23	  0.00 	29.9M  	0:09.77	Bd5 Kc8 Bxa8 Rc4 Bb7+ Kd7 Kf7 b4 Bd5 Rc6 Bxc6+ Kxc6 Bd4 g6 Ke6 b3 Be5 g5 Kf5 g4 Kxg4 Kb6 Kf5 Kxa6 Ke4 Kb6 Bb8 a5 Be5 b2 Bxb2 Kc7 
 22	+0.13 	18.0M  	0:05.87	Bd5 Kc8 Bxa8 Rc4 Bb7+ Kd7 Kf7 b4 Bd5 Rc6 Bxc6+ Kxc6 Kg6 Kb6 Kf5 b3 Ke4 Kxa6 Kd3 Kb5 Bd4 
 22	+0.13 	17.6M  	0:05.74	Bd5 Kc8 Bxa8 Rc4 Bb7+ Kd7 Kf7 b4 Bd5 Rc6 Bxc6+ Kxc6 Kg6 Kb6 Kf5 b3 Ke4 Kxa6 Kd3 Kb5 Bd4 
 21	+0.13 	13.6M  	0:04.45	Bd5 Kc8 Bxa8 Rc4 Bb7+ Kd7 Kf5 g6+ Kf6 b4 Bd5 Rc6+ Bxc6+ Kxc6 Bb2 Kb5 Ke5 Kxa6 Bd4 Kb5 Ke4 b3 Kd3 a5 Be5 
 21	+0.13 	12.9M  	0:04.29	Bd5 Kc8 Bxa8 Rc4 Bb7+ Kd7 Kf5 g6+ Kf6 b4 Bd5 Rc6+ Bxc6+ Kxc6 Bb2 Kb5 Ke5 Kxa6 Bd4 Kb5 Ke4 b3 Kd3 a5 Be5 
 20	+0.39 	8.69M  	0:02.85	Bd5 Kc8 Bxa8 Rc4 Bb7+ Kd7 Kf5 b4 Bd5 Rc6 Bxc6+ Kxc6 Ke4 Kb5 Kd3 Kxa6 Kc4 b3 Kxb3 
 20	+0.39 	8.39M  	0:02.78	Bd5 Kc8 Bxa8 Rc4 Bb7+ Kd7 Kf5 b4 Bd5 Rc6 Bxc6+ Kxc6 Ke4 Kb5 Kd3 Kxa6 Kc4 b3 Kxb3 
 19	+0.66 	3.91M  	0:01.29	Bd5 Kc8 Bxa8 Kd7 Bxc7 Kxc7 Bd5 Kb6 Kf5 b4 Bc4 b3 Kg6 b2 Bd3 Ka5 Kg5 b1=Q Bxb1 Kxa6 
 19	+0.66 	3.80M  	0:01.26	Bd5 Kc8 Bxa8 Kd7 Bxc7 Kxc7 Bd5 Kb6 Kf5 b4 Bc4 b3 Kg6 b2 Bd3 Ka5 Kg5 b1=Q Bxb1 Kxa6 
 18	+0.67 	2.25M  	0:00.75	Bd5 b4 Bxa8 Kc8 Bxc7 Kxc7 Be4 Kb6 Bd3 b3 Kxg7 b2 Kf6 b1=Q Bxb1 Kxa6 Be4 Kb5 
 18	+0.67 	2.16M  	0:00.73	Bd5 b4 Bxa8 Kc8 Bxc7 Kxc7 Be4 Kb6 Bd3 b3 Kxg7 b2 Kf6 b1=Q Bxb1 Kxa6 Be4 Kb5 
 17	+0.52 	359955	0:00.12	Bd5 b4 Bxa8 Kxa8 Bxc7 b3 Be5 b2 Bxb2 Kb8 Be5+ Kc8 Kxg7 Kd7 Kf6 Kc6 Bc7 Kxc7 Ke5 Kc6 
 17	+0.52 	302932	0:00.11	Bd5 b4 Bxa8 Kxa8 Bxc7 b3 Be5 b2 Bxb2 Kb8 Be5+ Kc8 Kxg7 Kd7 Kf6 Kc6 Bc7 Kxc7 Ke5 Kc6 
 16	+0.52 	275337	0:00.10	Bd5 b4 Bxa8 Kxa8 Bxc7 b3 Be5 b2 Bxb2 Kb8 Be5+ Kc8 Kxg7 Kd7 Kf6 Kc6 Bc7 Kxc7 Ke5 Kc6 
 16	+0.52 	239083	0:00.09	Bd5 b4 Bxa8 Kxa8 Bxc7 b3 Be5 b2 Bxb2 Kb8 Be5+ Kc8 Kxg7 Kd7 Kf6 Kc6 Bc7 Kxc7 Ke5 Kc6 
 15	+0.27 	186081	0:00.07	Bd5 b4 Bxc7+ Kxc7 Bxa8 Kb6 Bd5 Kxa6 Kf5 Kb5 Ke4 g5 Kd4 b3 Bxb3 a5 
 15	+0.27 	121121	0:00.05	Bd5 b4 Bxc7+ Kxc7 Bxa8 Kb6 Bd5 Kxa6 Kf5 Kb5 Ke4 g5 Kd4 b3 Bxb3 a5 
 14	+0.28 	90486  	0:00.04	Bd5 b4 Bxc7+ Kxc7 Bxa8 Kb6 Bd5 Kxa6 Kf5 Kb5 Ke4 g5 Kd4 b3 Bxb3 
 14	+0.28 	75563  	0:00.03	Bd5 b4 Bxc7+ Kxc7 Bxa8 Kb6 Bd5 Kxa6 Kf5 Kb5 Ke4 g5 Kd4 b3 Bxb3 
laptop Intel(R) Core(TM) i7-2670QM CPU @ 2.20GHz

Best
zullil
Posts: 6442
Joined: Tue Jan 09, 2007 12:31 am
Location: PA USA
Full name: Louis Zulli

Re: Solutions Stockfish should find 100 times faster

Post by zullil »

BBauer wrote:Good question.

I changed R in search.cpp from
Depth R = ((823 + 67 * depth) / 256 + std::min((eval - beta) / PawnValueMg, 3)) * ONE_PLY;
to
Depth R = depth<12 ? (2 + depth/2 ) * ONE_PLY : (5 + depth/6) * ONE_PLY;

which is faster and finds the right move on a high depth.

Kind regards
Benhard
Unaltered Stockfish does eventually find the mate. With 16 threads and 8 GB hash:

Code: Select all

info depth 52 seldepth 55 multipv 1 score mate 5 nodes 28672781815 nps 38108327 hashfull 983 tbhits 0 time 752402 pv e6d7 b5b4 g6f7 b4b3 f7e8 b3b2 e8d8 b2b1q e5c7
BBauer
Posts: 658
Joined: Wed Mar 08, 2006 8:58 pm

Re: Solutions Stockfish should find 100 times faster

Post by BBauer »

Unaltered SF with 16 threads may solve positions which you cannot solve with 1 thread. At least that is my feeling.
Kind regards
Bernhard
User avatar
hgm
Posts: 28354
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: Solutions Stockfish should find 100 times faster

Post by hgm »

Vinvin wrote:A lot engines have problem with this solution.
A friend rated 1600 Elo found the solution in a couple of minutes...
Well, Fairy-Max finds it in 9 sec:

Code: Select all

dep	score	nodes	time	(not shown:  tbhits	knps	seldep)
 21	+79.95 	10.5M  	0:09.20	e6d7 b5b4 g6f7 g7g5 f7e7 g5g4 e7d8 b4b3 e5c7
 21	+1.93 	9.99M  	0:08.79	e6d5 b8c8 e5c7 c8c7 d5a8 c7b6 a8b7 b5b4 b7d5 b6a6 d5c4 a6b6 c4b3 b6b5 b3d1 a7a5 d1e2 b5c5 e2d1 c5b5 d1e2
 20	+2.07 	3.22M  	0:02.85	e6d5 b8c8 e5c7 c8c7 d5a8 c7b6 a8b7 b5b4 b7d5 b6a6 g6g7 a6b5 d5b3 a7a5 b3e6 b4b3 e6b3 a5a4 b3d5 a4a3
 19	+2.10 	2.09M  	0:01.82	e6d5 b8c8 e5c7 c8c7 d5a8 c7b6 a8b7 b5b4 b7d5 b6a6 g6g7 a6b6 d5e6 a7a6 e6d5 a6a5 d5b3 b6c6 b3e6
 18	+2.13 	1.28M  	0:01.10	e6d5 b8c8 e5c7 c8c7 d5a8 c7b6 a8b7 b5b4 b7d5 b6a6 g6g7 a6b6 d5e6 a7a6 e6d5 a6a5 d5b3
 17	+2.07 	839649	0:00.73	e6d5 b8c8 e5c7 c8c7 d5a8 c7b6 a8b7 b5b4 b7d5 b6a6 g6g7 a6b6 d5e6 a7a5 e6c4
 16	+2.01 	519661	0:00.45	e6d5 b8c8 e5c7 c8c7 d5a8 c7b6 a8b7 b5b4 b7d5 b6a6 g6g7 a6b6 d5c4
 15	+2.05 	342069	0:00.31	e6d5 b8c8 e5c7 c8c7 d5a8 c7b6 a8b7 b5b4 b7d5 b6a6 d5c4 a6b6 g6g7 a7a5 c4d5 a5a4
 14	+2.04 	121019	0:00.12	e6d5 b5b4 d5a8 b4b3 a8d5 b3b2 e5b2 c7e7 b2g7 e7e3 g7f6 e3g3 f6g5 g3g4 d5e6
 13	+2.13 	80293  	0:00.07	e6d5 b5b4 d5a8 b4b3 a8d5 b3b2 e5b2 c7e7 b2g7 e7e3 g7f6 e3g3 f6g5 g3h3
 12	+2.19 	54756  	0:00.06	e6d5 b5b4 d5a8 b4b3 a8d5 b3b2 e5b2 c7e7 b2g7 e7e3 g7d4 e3g3
 11	+2.25 	36660  	0:00.03	e6d5 b5b4 d5a8 b4b3 a8d5 b3b2 e5b2 c7e7 b2g7 e7e3 g7d4
 10	+1.81 	22153  	0:00.01	e6d5 b5b4 d5a8 b4b3 a8f3 b8c8 f3g4 c8d8 e5c7 d8c7 g6g7 b3b2
  9	+2.12 	10898  	0:00.01	e6d5 b5b4 d5a8 b4b3 a8f3 b3b2 e5b2 c7d7 b2g7
  8	+2.12 	8641    	0:00.01	e6d5 b5b4 d5a8 b4b3 a8f3 b3b2 e5b2 c7d7 b2g7
  7	+1.34 	3727    	0:00.00	e6d5 b5b4 d5a8 b4b3 e5c7 b8c7 a8d5 b3b2 g6g7
  6	+1.73 	1274    	0:00.00	e6d5 b5b4 d5a8 b4b3 e5c7 b8c7
  5	+1.74 	725      	0:00.00	e6d5 b5b4 e5c7 b8c7 d5a8 b4b3 g6g7
  4	+1.85 	517      	0:00.00	e6d5 b5b4 e5c7 b8c7 d5a8
  3	+0.61 	199      	0:00.00	e6d5 b5b4 e5c7
  2	+0.61 	117      	0:00.00	e6d5 b5b4 e5c7
  2	 -1.98 	20        	0:00.00	e5c7 b8c7 g6g7
  1	 -1.98 	11        	0:00.00	e5c7
  0	#
Weird that engines find this so difficult. It is not a null-move problem, there is no zugzwang involved.
Dann Corbit
Posts: 12777
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: Solutions Stockfish should find 100 times faster

Post by Dann Corbit »

Stockfish has ultra-aggressive null move reductions.
This causes Stockfish to miss some tactics.
It also causes Stockfish to outsearch most engines by 5 or more plies.

I think overall it is a big win.

On the other hand, if we can pinpoint the exact conditions where the aggressive reductions should be throttled, it may make Stockfish play even better.