Engines can't solve this puzzle (even not Stockfish 5)

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

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
vittyvirus
Posts: 645
Joined: Wed Jun 18, 2014 12:30 pm
Full name: Fahad Syed

Engines can't solve this puzzle (even not Stockfish 5)

Post by vittyvirus » Tue Jul 22, 2014 6:18 am

Mate in 18:
[d]1n6/Pp1p1p1p/1P1P1P1P/4K3/2p5/p7/rpp1P3/qkb1R3 w - - 0 1
Stockfish 5 gave up pretty early. As it is mate in 18, I thought engines would solve this reaching to 18*2 = 36 ply. But no! It's really difficult for them. I've let Houdini 4 Pro to analyse for more than a hour (36 plies), and Deep Rybka 4.1 for 40(!) plies. Here are the analysis:
FEN: 1n6/Pp1p1p1p/1P1P1P1P/4K3/2p5/p7/rpp1P3/qkb1R3 w - - 0 1

Houdini 4 Pro:
4/18 00:00 774 387k 0.00 a7xb8Q c4-c3 Qb8xb7
5/18 00:00 825 412k 0.00 a7xb8Q c4-c3 Qb8xb7
6/18 00:00 876 146k 0.00 a7xb8Q c4-c3 Qb8xb7
7/18 00:00 1k 33k 0.00 a7xb8Q c4-c3 Qb8xb7
8/18 00:00 1k 38k 0.00 a7xb8Q c4-c3 Qb8xb7
9/18 00:00 2k 46k 0.00 a7xb8Q c4-c3 Qb8xb7
10/18 00:00 2k 42k 0.00 a7xb8Q c4-c3 Qb8xb7
11/18 00:00 3k 63k 0.00 a7xb8Q c4-c3 Qb8xb7
12/18 00:00 6k 96k 0.00 a7xb8Q c4-c3 Qb8xb7
13/18 00:00 10k 167k 0.00 a7xb8Q c4-c3 Qb8xb7
14/18 00:00 20k 301k 0.00 a7xb8Q c4-c3 Qb8xb7
15/19 00:00 36k 497k 0.00 a7xb8Q c4-c3 Qb8xb7
16/19 00:00 63k 778k 0.00 a7xb8Q c4-c3 Qb8xb7
17/21 00:00 111k 1,196k 0.00 a7xb8Q c4-c3 Qb8xb7
18/27 00:00 190k 1,700k 0.00 a7xb8Q c4-c3 Qb8xb7
19/27 00:00 315k 2,216k 0.00 a7xb8Q c4-c3 Qb8xb7
20/27 00:00 486k 2,656k 0.00 a7xb8Q c4-c3 Qb8xb7
21/36 00:00 882k 3,253k 0.00 a7xb8Q c4-c3 Qb8xb7
22/37 00:00 1,475k 3,670k 0.00 a7xb8Q c4-c3 Qb8xb7
23/41 00:00 2,722k 4,026k 0.00 a7xb8Q c4-c3 Qb8xb7
24/43 00:01 5,012k 4,205k 0.00 a7xb8Q c4-c3 Qb8xb7
25/45 00:02 9,975k 4,303k 0.00 a7xb8Q c4-c3 Qb8xb7
26/45 00:04 17,932k 4,376k 0.00 a7xb8Q c4-c3 Qb8xb7
27/49 00:07 31,905k 4,335k 0.00 a7xb8Q c4-c3 Qb8xb7
28/51 00:13 59,662k 4,369k 0.00 a7xb8Q c4-c3 Qb8xb7
29/53 00:25 109,376k 4,342k 0.00 a7xb8Q c4-c3 Qb8xb7
30/59 00:48 212,787k 4,424k 0.00 a7xb8Q c4-c3 Qb8xb7
31/65 01:37 416,413k 4,250k 0.00 a7xb8Q c4-c3 Qb8xb7
32/70 04:38 1,157,995k 4,164k 0.00 a7xb8Q c4-c3 Qb8xb7
33/70 07:55 1,951,140k 4,104k 0.00 a7xb8Q c4-c3 Qb8xb7
34/70 15:15 3,694,830k 4,037k 0.00 a7xb8Q c4-c3 Qb8xb7
35/73 28:22 6,759,481k 3,969k 0.00 a7xb8Q c4-c3 Qb8xb7
36/78 1:01:42 14,749,519k 3,983k 0.00 a7xb8Q c4-c3 Qb8xb7

FEN: 1n6/Pp1p1p1p/1P1P1P1P/4K3/2p5/p7/rpp1P3/qkb1R3 w - - 0 1

Stockfish 5 64 bit:
1/1 00:00 37 9k -10.40 a7xb8Q
2/2 00:00 71 18k -7.90 a7xb8Q c4-c3 Qb8xb7
3/3 00:00 122 31k -7.90 a7xb8Q c4-c3 Qb8xb7
4/4 00:00 174 35k 0.00 a7xb8Q c4-c3 Qb8xb7
5/4 00:00 226 45k 0.00 a7xb8Q c4-c3 Qb8xb7
6/4 00:00 278 46k 0.00 a7xb8Q c4-c3 Qb8xb7
7/4 00:00 330 55k 0.00 a7xb8Q c4-c3 Qb8xb7
8/4 00:00 387 55k 0.00 a7xb8Q c4-c3 Qb8xb7
9/4 00:00 446 64k 0.00 a7xb8Q c4-c3 Qb8xb7
10/4 00:00 595 74k 0.00 a7xb8Q c4-c3 Qb8xb7
11/4 00:00 661 73k 0.00 a7xb8Q c4-c3 Qb8xb7
12/4 00:00 744 74k 0.00 a7xb8Q c4-c3 Qb8xb7
13/4 00:00 837 76k 0.00 a7xb8Q c4-c3 Qb8xb7
14/4 00:00 921 84k 0.00 a7xb8Q c4-c3 Qb8xb7
15/4 00:00 1k 85k 0.00 a7xb8Q c4-c3 Qb8xb7
16/4 00:00 1k 91k 0.00 a7xb8Q c4-c3 Qb8xb7
17/4 00:00 1k 99k 0.00 a7xb8Q c4-c3 Qb8xb7
18/4 00:00 2k 104k 0.00 a7xb8Q c4-c3 Qb8xb7
19/4 00:00 2k 117k 0.00 a7xb8Q c4-c3 Qb8xb7
20/4 00:00 2k 123k 0.00 a7xb8Q c4-c3 Qb8xb7
21/4 00:00 2k 130k 0.00 a7xb8Q c4-c3 Qb8xb7
22/4 00:00 3k 144k 0.00 a7xb8Q c4-c3 Qb8xb7
23/4 00:00 3k 159k 0.00 a7xb8Q c4-c3 Qb8xb7
24/4 00:00 5k 184k 0.00 a7xb8Q c4-c3 Qb8xb7
25/4 00:00 6k 206k 0.00 a7xb8Q c4-c3 Qb8xb7
26/4 00:00 6k 222k 0.00 a7xb8Q c4-c3 Qb8xb7
27/4 00:00 7k 244k 0.00 a7xb8Q c4-c3 Qb8xb7
28/4 00:00 8k 254k 0.00 a7xb8Q c4-c3 Qb8xb7
29/4 00:00 10k 276k 0.00 a7xb8Q c4-c3 Qb8xb7
30/4 00:00 14k 307k 0.00 a7xb8Q c4-c3 Qb8xb7
31/4 00:00 16k 323k 0.00 a7xb8Q c4-c3 Qb8xb7
32/4 00:00 21k 358k 0.00 a7xb8Q c4-c3 Qb8xb7
33/4 00:00 23k 373k 0.00 a7xb8Q c4-c3 Qb8xb7
34/4 00:00 24k 386k 0.00 a7xb8Q c4-c3 Qb8xb7
35/4 00:00 25k 392k 0.00 a7xb8Q c4-c3 Qb8xb7
36/4 00:00 28k 420k 0.00 a7xb8Q c4-c3 Qb8xb7
37/4 00:00 29k 429k 0.00 a7xb8Q c4-c3 Qb8xb7
38/4 00:00 31k 445k 0.00 a7xb8Q c4-c3 Qb8xb7
39/4 00:00 32k 451k 0.00 a7xb8Q c4-c3 Qb8xb7
40/4 00:00 34k 474k 0.00 a7xb8Q c4-c3 Qb8xb7
41/4 00:00 36k 482k 0.00 a7xb8Q c4-c3 Qb8xb7
42/4 00:00 38k 499k 0.00 a7xb8Q c4-c3 Qb8xb7
43/4 00:00 40k 516k 0.00 a7xb8Q c4-c3 Qb8xb7
44/4 00:00 42k 535k 0.00 a7xb8Q c4-c3 Qb8xb7
45/4 00:00 44k 558k 0.00 a7xb8Q c4-c3 Qb8xb7
46/4 00:00 48k 584k 0.00 a7xb8Q c4-c3 Qb8xb7
47/4 00:00 51k 600k 0.00 a7xb8Q c4-c3 Qb8xb7
48/4 00:00 54k 617k 0.00 a7xb8Q c4-c3 Qb8xb7
49/4 00:00 56k 633k 0.00 a7xb8Q c4-c3 Qb8xb7
50/4 00:00 58k 654k 0.00 a7xb8Q c4-c3 Qb8xb7
51/4 00:00 60k 672k 0.00 a7xb8Q c4-c3 Qb8xb7
52/4 00:00 64k 698k 0.00 a7xb8Q c4-c3 Qb8xb7
53/4 00:00 67k 719k 0.00 a7xb8Q c4-c3 Qb8xb7
54/4 00:00 71k 742k 0.00 a7xb8Q c4-c3 Qb8xb7
55/4 00:00 74k 770k 0.00 a7xb8Q c4-c3 Qb8xb7
56/4 00:00 80k 797k 0.00 a7xb8Q c4-c3 Qb8xb7
57/4 00:00 88k 837k 0.00 a7xb8Q c4-c3 Qb8xb7
58/4 00:00 93k 863k 0.00 a7xb8Q c4-c3 Qb8xb7
59/4 00:00 97k 889k 0.00 a7xb8Q c4-c3 Qb8xb7
60/4 00:00 102k 908k 0.00 a7xb8Q c4-c3 Qb8xb7
61/4 00:00 107k 936k 0.00 a7xb8Q c4-c3 Qb8xb7
62/4 00:00 118k 970k 0.00 a7xb8Q c4-c3 Qb8xb7
63/4 00:00 124k 994k 0.00 a7xb8Q c4-c3 Qb8xb7
64/4 00:00 130k 1,022k 0.00 a7xb8Q c4-c3 Qb8xb7
65/4 00:00 136k 1,043k 0.00 a7xb8Q c4-c3 Qb8xb7
66/4 00:00 143k 1,078k 0.00 a7xb8Q c4-c3 Qb8xb7
67/4 00:00 151k 1,108k 0.00 a7xb8Q c4-c3 Qb8xb7
68/4 00:00 163k 1,139k 0.00 a7xb8Q c4-c3 Qb8xb7
69/4 00:00 169k 1,168k 0.00 a7xb8Q c4-c3 Qb8xb7
70/4 00:00 185k 1,206k 0.00 a7xb8Q c4-c3 Qb8xb7
71/4 00:00 194k 1,226k 0.00 a7xb8Q c4-c3 Qb8xb7
72/4 00:00 203k 1,259k 0.00 a7xb8Q c4-c3 Qb8xb7
73/4 00:00 215k 1,301k 0.00 a7xb8Q c4-c3 Qb8xb7
74/4 00:00 225k 1,330k 0.00 a7xb8Q c4-c3 Qb8xb7
75/4 00:00 235k 1,367k 0.00 a7xb8Q c4-c3 Qb8xb7
76/4 00:00 248k 1,403k 0.00 a7xb8Q c4-c3 Qb8xb7
77/4 00:00 261k 1,444k 0.00 a7xb8Q c4-c3 Qb8xb7
78/4 00:00 278k 1,495k 0.00 a7xb8Q c4-c3 Qb8xb7
79/4 00:00 291k 1,525k 0.00 a7xb8Q c4-c3 Qb8xb7
80/4 00:00 309k 1,577k 0.00 a7xb8Q c4-c3 Qb8xb7
81/4 00:00 359k 1,676k 0.00 a7xb8Q c4-c3 Qb8xb7
82/4 00:00 403k 1,743k 0.00 a7xb8Q c4-c3 Qb8xb7
83/4 00:00 488k 1,893k 0.00 a7xb8Q c4-c3 Qb8xb7
84/4 00:00 506k 1,915k 0.00 a7xb8Q c4-c3 Qb8xb7
85/4 00:00 602k 2,034k 0.00 a7xb8Q c4-c3 Qb8xb7
86/4 00:00 624k 2,065k 0.00 a7xb8Q c4-c3 Qb8xb7
87/4 00:00 645k 2,075k 0.00 a7xb8Q c4-c3 Qb8xb7
88/4 00:00 681k 2,095k 0.00 a7xb8Q c4-c3 Qb8xb7
89/4 00:00 753k 2,195k 0.00 a7xb8Q c4-c3 Qb8xb7
90/4 00:00 783k 2,229k 0.00 a7xb8Q c4-c3 Qb8xb7
91/4 00:00 838k 2,301k 0.00 a7xb8Q c4-c3 Qb8xb7
92/4 00:00 866k 2,329k 0.00 a7xb8Q c4-c3 Qb8xb7
93/4 00:00 909k 2,381k 0.00 a7xb8Q c4-c3 Qb8xb7
94/4 00:00 938k 2,412k 0.00 a7xb8Q c4-c3 Qb8xb7
95/4 00:00 975k 2,444k 0.00 a7xb8Q c4-c3 Qb8xb7
96/4 00:00 1,021k 2,491k 0.00 a7xb8Q c4-c3 Qb8xb7
97/4 00:00 1,147k 2,589k 0.00 a7xb8Q c4-c3 Qb8xb7
98/4 00:00 1,223k 2,596k 0.00 a7xb8Q c4-c3 Qb8xb7
99/4 00:00 1,384k 2,661k 0.00 a7xb8Q c4-c3 Qb8xb7
100/4 00:00 1,470k 2,723k 0.00 a7xb8Q c4-c3 Qb8xb7
101/4 00:00 1,521k 2,755k 0.00 a7xb8Q c4-c3 Qb8xb7
102/4 00:00 1,584k 2,788k 0.00 a7xb8Q c4-c3 Qb8xb7
103/4 00:00 1,616k 2,801k 0.00 a7xb8Q c4-c3 Qb8xb7
104/4 00:00 1,651k 2,817k 0.00 a7xb8Q c4-c3 Qb8xb7
105/4 00:00 1,698k 2,850k 0.00 a7xb8Q c4-c3 Qb8xb7
106/4 00:00 1,727k 2,868k 0.00 a7xb8Q c4-c3 Qb8xb7
107/4 00:00 1,772k 2,895k 0.00 a7xb8Q c4-c3 Qb8xb7
108/4 00:00 1,816k 2,924k 0.00 a7xb8Q c4-c3 Qb8xb7
109/4 00:00 1,862k 2,950k 0.00 a7xb8Q c4-c3 Qb8xb7
110/4 00:00 2,048k 2,990k 0.00 a7xb8Q c4-c3 Qb8xb7
111/4 00:00 2,191k 3,055k 0.00 a7xb8Q c4-c3 Qb8xb7
112/4 00:00 2,238k 3,074k 0.00 a7xb8Q c4-c3 Qb8xb7
113/4 00:00 2,369k 3,121k 0.00 a7xb8Q c4-c3 Qb8xb7
114/4 00:00 2,456k 3,137k 0.00 a7xb8Q c4-c3 Qb8xb7
115/4 00:00 2,579k 3,184k 0.00 a7xb8Q c4-c3 Qb8xb7
116/4 00:00 2,790k 3,248k 0.00 a7xb8Q c4-c3 Qb8xb7
117/4 00:00 2,942k 3,305k 0.00 a7xb8Q c4-c3 Qb8xb7
118/4 00:00 3,001k 3,324k 0.00 a7xb8Q c4-c3 Qb8xb7
119/4 00:00 3,269k 3,391k 0.00 a7xb8Q c4-c3 Qb8xb7
120/4 00:01 4,308k 3,552k 0.00 a7xb8Q c4-c3 Qb8xb7

FEN: 1n6/Pp1p1p1p/1P1P1P1P/4K3/2p5/p7/rpp1P3/qkb1R3 w - - 0 1

Deep Rybka 4.1 SSE42 x64:
1 00:00 36 2k 0.00 a7xb8Q c4-c3 Qb8xb7
2 00:00 55 3k 0.00 a7xb8Q c4-c3 Qb8xb7
3 00:00 74 5k 0.00 a7xb8Q c4-c3 Qb8xb7
4 00:00 90 6k 0.00 a7xb8Q c4-c3 Qb8xb7
5 00:00 112 7k 0.00 a7xb8Q c4-c3 Qb8xb7
6 00:00 136 9k 0.00 a7xb8Q c4-c3 Qb8xb7
7 00:00 176 11k 0.00 a7xb8Q c4-c3 Qb8xb7
8 00:00 243 8k 0.00 a7xb8Q c4-c3 Qb8xb7
9 00:00 334 10k 0.00 a7xb8Q c4-c3 Qb8xb7
10 00:00 466 15k 0.00 a7xb8Q c4-c3 Qb8xb7
11 00:00 804 25k 0.00 a7xb8Q c4-c3 Qb8xb7
12 00:00 1k 37k 0.00 a7xb8Q c4-c3 Qb8xb7
13 00:00 2k 53k 0.00 a7xb8Q c4-c3 Qb8xb7
14 00:00 2k 70k 0.00 a7xb8Q c4-c3 Qb8xb7
15 00:00 3k 97k 0.00 a7xb8Q c4-c3 Qb8xb7
16 00:00 4k 127k 0.00 a7xb8Q c4-c3 Qb8xb7
17 00:00 5k 111k 0.00 a7xb8Q c4-c3 Qb8xb7
18 00:00 8k 161k 0.00 a7xb8Q c4-c3 Qb8xb7
19 00:00 10k 218k 0.00 a7xb8Q c4-c3 Qb8xb7
20 00:00 14k 224k 0.00 a7xb8Q c4-c3 Qb8xb7
21 00:00 20k 255k 0.00 a7xb8Q c4-c3 Qb8xb7
22 00:00 26k 280k 0.00 a7xb8Q c4-c3 Qb8xb7
23 00:00 33k 304k 0.00 a7xb8Q c4-c3 Qb8xb7
24 00:00 42k 336k 0.00 a7xb8Q c4-c3 Qb8xb7
25 00:00 56k 355k 0.00 a7xb8Q c4-c3 Qb8xb7
26 00:00 76k 372k 0.00 a7xb8Q c4-c3 Qb8xb7
27 00:00 106k 397k 0.00 a7xb8Q c4-c3 Qb8xb7
28 00:00 268k 418k 0.00 a7xb8Q c4-c3 Qb8xb7
29 00:01 487k 421k 0.00 a7xb8Q c4-c3 Qb8xb7
30 00:01 644k 425k 0.00 a7xb8Q c4-c3 Qb8xb7
31 00:02 1,278k 452k 0.00 a7xb8Q c4-c3 Qb8xb7
32 00:03 1,777k 449k 0.00 a7xb8Q c4-c3 Qb8xb7
33 00:05 2,436k 448k 0.00 a7xb8Q c4-c3 Qb8xb7
34 00:07 3,312k 441k 0.00 a7xb8Q c4-c3 Qb8xb7
35 00:10 4,633k 439k 0.00 a7xb8Q c4-c3 Qb8xb7
36 00:15 6,940k 437k 0.00 a7xb8Q c4-c3 Qb8xb7
37 00:25 11,067k 430k 0.00 a7xb8Q c4-c3 Qb8xb7
38 00:38 16,266k 426k 0.00 a7xb8Q c4-c3 Qb8xb7
39 01:00 25,631k 424k 0.00 a7xb8Q c4-c3 Qb8xb7
40 01:40 42,213k 419k 0.00 a7xb8Q c4-c3 Qb8xb7
40 01:51 45,833k 410k 0.00 a7xb8Q c4-c3 Qb8xb7

As you see, all engines get this a draw, although it not.

[Event "Babic 1996"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Der brillante Schachzug 100"]
[Black "Matt in 18"]
[Result "*"]
[Annotator "Walter Eigenmann"]
[SetUp "1"]
[FEN "1n6/Pp1p1p1p/1P1P1P1P/4K3/2p5/p7/rpp1P3/qkb1R3 w - - 0 1"]
[PlyCount "35"]
[Source "Walter Eigenman"]
[SourceDate "2011.11.01"]

1. axb8=N $3 (1. axb8=Q $2 c3 2. Qxb7 $11 (2. Kd4 $11)) (1. a8=Q $2 Nc6+ 2. Kd5
Nb4+ $11) 1... c3 2. Na6 bxa6 3. b7 a5 4. b8=N a4 5. Nc6 dxc6 6. d7 c5 7. d8=N
c4 8. Ne6 fxe6 9. Kd6 e5 10. f7 e4 11. f8=N e3 12. Ng6 hxg6 13. h7 g5 14. h8=N
g4 15. Nf7 g3 16. Ng5 g2 17. Ne4 g1=Q 18. Nxc3# *

[pgn][Event "Babic 1996"]
[Site "?"]
[Date "????.??.??"]
[Round "?"]
[White "Der brillante Schachzug 100"]
[Black "Matt in 18"]
[Result "*"]
[Annotator "Walter Eigenmann"]
[SetUp "1"]
[FEN "1n6/Pp1p1p1p/1P1P1P1P/4K3/2p5/p7/rpp1P3/qkb1R3 w - - 0 1"]
[PlyCount "35"]
[Source "Walter Eigenman"]
[SourceDate "2011.11.01"]

1. axb8=N $3 (1. axb8=Q $2 c3 2. Qxb7 $11 (2. Kd4 $11)) (1. a8=Q $2 Nc6+ 2. Kd5
Nb4+ $11) 1... c3 2. Na6 bxa6 3. b7 a5 4. b8=N a4 5. Nc6 dxc6 6. d7 c5 7. d8=N
c4 8. Ne6 fxe6 9. Kd6 e5 10. f7 e4 11. f8=N e3 12. Ng6 hxg6 13. h7 g5 14. h8=N
g4 15. Nf7 g3 16. Ng5 g2 17. Ne4 g1=Q 18. Nxc3# *
[/pgn]

lech
Posts: 1063
Joined: Sun Feb 14, 2010 9:02 pm

Re: Engines can't solve this puzzle (even not Stockfish 5)

Post by lech » Tue Jul 22, 2014 1:03 pm

Maybe the palliatives like null move search = off or Monte Carlo can help you.
The next version of Sting (not published yet) gets it quikly. I wrote about it (http://talkchess.com/forum/viewtopic.php?t=52829). It is one of the most difficult positions to solve.
Maybe, I can't be friendly, but let me be useful.

User avatar
hgm
Posts: 23795
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Engines can't solve this puzzle (even not Stockfish 5)

Post by hgm » Tue Jul 22, 2014 2:21 pm

The problem here is that there is one very long branch in the tree that leads to mate, while all other branches from the root liberate all the black pieces, which makes the tree explode. So engines that more or less try to use the same depth on all branches (give or take a few reductions / extensions) would never get to the end of that long branch to see the mate.

What you need to solve this kind of problem is not conventional iterative deepening, but an engine that works from a node budget, which it will evenly divide over all his daughter nodes. The sub-tree with low branching ratio will be searched far deeper than the others, with the same number of nodes, so that the mate will be seen with a quite low node budget.

User avatar
RolandoFurioso
Posts: 55
Joined: Sat Feb 22, 2014 6:29 pm
Location: Frankfurt

Re: Engines can't solve this puzzle (even not Stockfish 5)

Post by RolandoFurioso » Tue Jul 22, 2014 4:01 pm

Of course, much depends upon move ordering and extensions/reductions employed. I've just checked it with my own engine Fischerle 0.9.50 64 Bit. Employing just the standard (match play rather than chess problem solving) settings and allocating 256 MB hash, Fischerle finds the solution (mate in 18) in around 270 seconds, searching slightly less than 99 million positions:

L1: -9897 ♙a7x♞b8♕ ♟c4-c3 ♕b8x♟b7
L2: -10457
L2: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L3: -240
L3: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L4: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L5: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L6: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L7: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L8: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L9: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L10: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L11: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L12: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L13: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L14: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L15: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L16: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L17: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L18: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L19: 0 ♙a7x♞b8♕ ♟c4-c3 ♔e5-e4 PATT
L20: 0 ♙a7x♞b8♕ ♟c4-c3 ♕b8x♟b7 PATT
L21: 0 ♙a7x♞b8♕ ♟c4-c3 ♕b8x♟b7 PATT
L22: 240
HV: 999965 ♙a7x♞b8♘ ♟c4-c3 ♘b8-a6 ♟b7x♘a6 ♙b6-b7 ♟a6-a5 ♙b7-b8♘ ♟a5-a4 ♘b8-c6 ♟d7x♘c6 ♙d6-d7 ♟c6-c5 ♙d7-d8♘ ♟c5-c4 ♘d8-e6 ♟f7x♘e6 ♔e5-d6 ♟e6-e5 ♙f6-f7 ♟e5-e4 ♙f7-f8♘ ♟e4-e3 ♘f8-g6 ♟h7x♘g6 ♙h6-h7 ♟g6-g5 ♙h7-h8♘ ♟g5-g4 ♘h8-g6 ♟g4-g3 ♘g6-f4 ♟g3-g2 ♘f4-d5 ♟g2-g1♛ ♘d5x♟c3 MATT

User avatar
RolandoFurioso
Posts: 55
Joined: Sat Feb 22, 2014 6:29 pm
Location: Frankfurt

Re: Engines can't solve this puzzle (even not Stockfish 5)

Post by RolandoFurioso » Tue Jul 22, 2014 4:11 pm

lech wrote:Maybe the palliatives like null move search = off or Monte Carlo can help you.
The next version of Sting (not published yet) gets it quikly. I wrote about it (http://talkchess.com/forum/viewtopic.php?t=52829). It is one of the most difficult positions to solve.
Of course, these are the most natural things to try. In my own experiments, however, switching off null move pruning actually disabled my engine (Fischerle) to find the solution M18, as it turned out to be no longer able to quickly reach the nominal search depth required (given the implemented extensions / reductions) to find the mate.

User avatar
SMIRF
Posts: 91
Joined: Wed Mar 26, 2014 3:29 pm
Location: Buettelborn/Hessen/Germany
Contact:

Re: Engines can't solve this puzzle (even not Stockfish 5)

Post by SMIRF » Tue Jul 22, 2014 11:11 pm

My weak private engine FullChess GUI-64 - SMIRF Engine (Ver. MS-180a)
surprisingly seems to have an idea:
  • 00:36.5 (18.04=) +M~??? 8.axb8=N (axb8=Q) c3 9.Na6 bxa6 10.b7 a5 11.b8=N a4 12.Nc6 dxc6 13.d7 c5 14.d8=N c4 15.Ne6 fxe6
    00:34.4 (18.01=) +0.000 8.axb8=Q (axb8=R) c3 9.Qxb7
    00:14.7 (17.01=) +0.000 8.axb8=Q (axb8=R) c3 9.Qxb7
    00:08.4 (16.01=) +0.000 8.axb8=Q (axb8=R) c3 9.Qxb7
    00:02.8 (15.01=) +0.000 8.axb8=Q (axb8=R) c3 9.Qxb7
    00:01.2 (14.01=) +0.000 8.axb8=Q (axb8=R) c3 9.Qxb7
    00:00.6 (13.01=) +0.000 8.axb8=Q (axb8=R) c3 9.Qxb7
    00:00.2 (12.01=) +0.000 8.axb8=Q (axb8=R) c3 9.Qxb7
    00:00.1 (11.01=) +0.000 8.axb8=Q (axb8=R) c3 9.Qxb7
    00:00.0 (10.01=) +0.000 8.axb8=Q (axb8=R) c3 9.Qxb7

User avatar
vittyvirus
Posts: 645
Joined: Wed Jun 18, 2014 12:30 pm
Full name: Fahad Syed

Re: Engines can't solve this puzzle (even not Stockfish 5)

Post by vittyvirus » Wed Jul 23, 2014 6:02 am

lech wrote:Maybe the palliatives like null move search = off or Monte Carlo can help you.
The next version of Sting (not published yet) gets it quikly. I wrote about it (http://talkchess.com/forum/viewtopic.php?t=52829). It is one of the most difficult positions to solve.
What would be its elo? and where from can I get the latest sting version?

User avatar
vittyvirus
Posts: 645
Joined: Wed Jun 18, 2014 12:30 pm
Full name: Fahad Syed

Re: Engines can't solve this puzzle (even not Stockfish 5)

Post by vittyvirus » Wed Jul 23, 2014 6:05 am

I wonder why does Stockfish completely give up at this position? And why don't engines find the mate in 18 at 36 ply?

lech
Posts: 1063
Joined: Sun Feb 14, 2010 9:02 pm

Re: Engines can't solve this puzzle (even not Stockfish 5)

Post by lech » Wed Jul 23, 2014 7:30 am

vittyvirus wrote:
lech wrote:Maybe the palliatives like null move search = off or Monte Carlo can help you.
The next version of Sting (not published yet) gets it quikly. I wrote about it (http://talkchess.com/forum/viewtopic.php?t=52829). It is one of the most difficult positions to solve.
What would be its elo? and where from can I get the latest sting version?
As I earlier wrote I can't see ELO difference, and it will be published soon or soooooon. 8-)
Maybe, I can't be friendly, but let me be useful.

User avatar
hgm
Posts: 23795
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Engines can't solve this puzzle (even not Stockfish 5)

Post by hgm » Wed Jul 23, 2014 8:15 am

vittyvirus wrote:I wonder why does Stockfish completely give up at this position? And why don't engines find the mate in 18 at 36 ply?
Because when strong engines such as Stockfish say they searched 36 ply, it only means they searched one selected branch 36 ply. Almost all other branchess in the tree will have been searched to only 9 ply or so. If the mate is in there (which it will be, if the branch leading to it will not in some way achieve something very good in these first 9 ply), it is still very far beyond the horizon indeed.

Post Reply