Sting SF Black Hole 11

[lech]

Now Sting starts with the UCI: OldSting option as the default value.
To use the Black Hole idea, set the UCI: BlackHole option to true.
This is a safe choice. The "Black Hole" idea requires knowledge of it, otherwise it is impossible to
correctly interpret the returned data.
Moreover, Sting's Black Hole idea may not work correctly in time management mode (play).

download: https://www.mediafire.com/file/oicgues2 ... 1.zip/file

[peter]

Thanks for the new version!

Sting is one of the very few if not the only one engine to solve this Salai- study

https://www.yacpdb.org/#search/NEsxazFc ... LzEvMA==/1

with single thread (3.5GHz) in 1 minute:

Code: Select all

Sting Black Hole 11 (based on Stockfish 2.1.1) 64bit by Marek Kwiatkowski
position fen 4K1k1/8/1p5p/1Pp3b1/8/1P3P2/P1B2P2/8 w - -
setoption name hash value 1024
go infinite
...
info depth 30 multipv 1 score cp 157 lowerbound nodes 180697475 nps 2927128 time 61732 pv f3f4 g5f4 e8d7 f4d2 d7c6 d2a5 b3b4 c5b4 c2b3 g8g7 c6d5 g7f6 d5e4 h6h5 e4f4 f6g6 f4e5 h5h4 e5f4 g6h5 f2f3 h5h6 f4g4
...
stop
info depth 30 multipv 1 score cp 264 nodes 307145564 nps 2872049 time 106943 pv f3f4 g5f4 e8d7 f4d2 d7c6 d2a5 b3b4 c5b4 c2b3 g8g7 c6d5 g7f6 d5e4 h6h5 e4f4 f6g6 f4e5 h5h4 e5f4 g6h5 f2f3 h5h6 f4g4 h4h3 g4h3 h6g5 h3g3 g5f5 f3f4 f5e4 g3g4 e4d4 g4g5
info nodes 307145564 nps 2871485 time 106964
bestmove f3f4 ponder g5f4

Notice the major dual with 1.Bd1, which yet isn't any easier to be found than main solution 1.f4 is, regards

[lucario6607]

What is the black hole thing? and is this nnue or hce?

[Dann Corbit]

It uses a classical, hand written evaluation.
If you want to make sense of engine names, we have fruit, stockfish, clover, integral, halogen, Beowulf, and booot to ponder too.
I would not read too much into the name. My best guess: It sucks all of his time into an inescapable singularity. He has tried to escape 11 times without success.
Sting tends to be really good for certain classes of problems.

[acase]

Thanks for the new version!

Sting is one of the very few if not the only one engine to solve this Salai- study

https://www.yacpdb.org/#search/NEsxazFc ... LzEvMA==/1

with single thread (3.5GHz) in 1 minute:

Code: Select all

Sting Black Hole 11 (based on Stockfish 2.1.1) 64bit by Marek Kwiatkowski
position fen 4K1k1/8/1p5p/1Pp3b1/8/1P3P2/P1B2P2/8 w - -
setoption name hash value 1024
go infinite
...
info depth 30 multipv 1 score cp 157 lowerbound nodes 180697475 nps 2927128 time 61732 pv f3f4 g5f4 e8d7 f4d2 d7c6 d2a5 b3b4 c5b4 c2b3 g8g7 c6d5 g7f6 d5e4 h6h5 e4f4 f6g6 f4e5 h5h4 e5f4 g6h5 f2f3 h5h6 f4g4
...
stop
info depth 30 multipv 1 score cp 264 nodes 307145564 nps 2872049 time 106943 pv f3f4 g5f4 e8d7 f4d2 d7c6 d2a5 b3b4 c5b4 c2b3 g8g7 c6d5 g7f6 d5e4 h6h5 e4f4 f6g6 f4e5 h5h4 e5f4 g6h5 f2f3 h5h6 f4g4 h4h3 g4h3 h6g5 h3g3 g5f5 f3f4 f5e4 g3g4 e4d4 g4g5
info nodes 307145564 nps 2871485 time 106964
bestmove f3f4 ponder g5f4

Notice the major dual with 1.Bd1, which yet isn't any easier to be found than main solution 1.f4 is, regards

ShashChess 31.1 using the "High Petrosian" parameter on my computer (dual Xeon 6150 with 32GB hash table) finds 1.f4!:


FEN: 4K1k1/8/1p5p/1Pp3b1/8/1P3P2/P1B2P2/8 w - - 0 1

ShashChess31.1-x86-64-bmi2:
NNUE evaluation using nn-dabb1ed23026.nnue enabled

55/51 01:44 5,683,866k 54,556k +1.18 1.Ke8-d7 Bg5-d2 2.Kd7-e6 Kg8-g7 3.Bc2-d3 h6-h5 4.Ke6-f5 h5-h4 5.Kf5-g4 Kg7-f6 6.Kg4xh4 Bd2-f4 7.Kh4-g4 Kf6-e5 8.a2-a4 Bf4-d2 9.Bd3-e4 Bd2-c3 10.Be4-c6 Ke5-f6 11.f3-f4 Kf6-g6 12.Bc6-e4+ Kg6-f6 13.Be4-d5 Bc3-e1 14.Kg4-f3 Kf6-f5 15.Bd5-e4+ Kf5-e6 16.Kf3-e2 Be1-a5 17.Ke2-d3 Ke6-f6 18.Kd3-c4 Kf6-e6 19.Be4-h7 Ke6-f6 20.Kc4-d5 Ba5-d2 21.Bh7-e4 Bd2-b4 22.Be4-d3 Bb4-a5 23.Kd5-e4 Ba5-b4 24.Bd3-f1 Bb4-c3 25.Bf1-h3 Bc3-a5 26.Ke4-d5
56/47+ 01:48 5,912,971k 54,410k +1.24 1.f3-f4
56/47+ 01:48 5,915,124k 54,409k +1.29 1.f3-f4
56/72+ 09:08 30,886,984k 56,310k +7.42 1.f3-f4

I'll have to see if it can reproduce it though.

[peter]

ShashChess 31.1 using the "High Petrosian" parameter on my computer (dual Xeon 6150 with 32GB hash table) finds 1.f4!:

I know ShashChess to be good at this one too, I just wanted to point out Sting solving single threaded rather quickly as well.
Here's latest ShashChess version with 30 threads of the 16x3.5GHz CPU and 32G hash:

Code: Select all

4K1k1/8/1p5p/1Pp3b1/8/1P3P2/P1B2P2/8 w - -

Engine: ShashChess40 (32768 MB)
von A. Manzo, F. Ferraguti, K. Kiniama and

1/02   0:00   +2.13    1.Kd7 (286) 26 
...
44/70  0:44   +0.66    1.Kd7 Ld2 2.Ke6 Kg7 3.Ld3 Le1 4.f4 h5 
                       5.a4 Lc3 6.Kf5 h4 7.Kg4 Le1 8.Kxh4 Lxf2+ 
                       9.Kg4 Le1 10.Kf5 Ld2 11.Ke6 Lxf4 
                       12.a5 bxa5 13.b6 Lg3 14.Ke7 (1.074.107.526) 24411 
                       TB:611.635 
45/70  0:46   +0.71++  1.f4 (1.160.299.651) 24727 
                       TB:657.495 
...
               
46/33  1:20   +6.19++  1.f4 (1.832.293.960) 22722 
                       TB:2.264.040 

Two more trials took 1x1:01 (finding 1.Bd1 before f4 this time) and 1x24", Sting with 30 threads at 3 trials: 13", 14" and 19".
ShashChess with single thread and 1G hash:

Code: Select all

4K1k1/8/1p5p/1Pp3b1/8/1P3P2/P1B2P2/8 w - -

Engine: ShashChess40 (1024 MB)
von A. Manzo, F. Ferraguti, K. Kiniama and

21/56  0:01   +0.98    1.a4 Ld2 2.Ke7 Kg7 3.Ld3 Lb4 4.Ke6 h5 
                       5.f4 h4 6.Lf1 Kf8 7.f5 La5 8.f6 Ke8 
                       9.Kf5 Kf8 10.Kg4 Le1 11.Kxh4 Lxf2+ 
                       12.Kg5 Le1 13.Kf5 La5 14.Lg2 (1.218.697) 752 
                       TB:1.554 
22/42  0:02   +1.57    1.Kd7 Ld2 2.Kc6 La5 3.f4 Kg7 4.Ld1 Kg6 
                       5.Lg4 h5 6.f5+ Kh6 7.Le2 h4 8.Lg4 Kg5 
                       9.Lh3 Kf6 10.a3 Kf7 11.b4 Lxb4 
                       12.axb4 c4 13.Lg2 (2.117.244) 835 
                       TB:2.326 
...
64/93  7:28   +0.89++  1.Kd7 (676.352.491) 1507 TB:967.006 
64/93  7:46   +1.22++  1.f4 (703.396.909) 1506 TB:1.026.670 
...
64/93  8:16   +5.47++  1.f4 (734.978.916) 1481 TB:1.286.662