Problem: when an extra bishop harms.

[peter]

Code: Select all

7n/5p1b/5Pp1/5ppk/4p1r1/5QPK/6P1/8 w - -
Engine: The Huntsman 1 (32768 MB)
von the Stockfish developers (see AUTHORS f

75     11:35  +M22     1.Dxg4+ fxg4+ 2.Kh2 e3 3.Kg1 Kh6 
                       4.Kf1 Lg8 5.Ke2 Lh7 6.Kxe3 Lg8 7.Ke4 Lh7 
                       8.Ke5 Lg8 9.Kd6 Kh5 10.Kd7 Lh7 
                       11.Ke7 Kh6 12.Kf8 Kh5 13.Kg7 Lg8 
                       14.Kxh8 (63.918.828.531) 91887 

74     11:35  +M29     1.De2 Lg8 2.Kh2 e3 3.Dxe3 Te4 4.g4+ Txg4 
                       5.Df3 Kh6 6.g3 Ta4 7.Dh1 Kh5 8.Kg2+ Th4 
                       9.gxh4 gxh4 10.Dd1+ Kh6 11.Dd8 g5 
                       12.Dxg8 Sg6 13.Dxf7 g4 14.Dg7+ (63.918.828.531) 91887 

74     11:35  +M29     1.Dd1 e3 2.Kh2 Lg8 3.De2 Lh7 4.Dxe3 Te4 
                       5.g4+ Txg4 6.Df3 Kh6 7.g3 Ta4 8.Dh1 Kh5 
                       9.Kg2+ Th4 10.gxh4 gxh4 11.Dd1+ Kh6 
                       12.Dd8 Kg5 13.Dxh8 Lg8 14.Dxg8 (63.918.828.531) 91887 

74     11:35  +M36     1.De3 Th4+ 2.gxh4 gxh4 3.g3 hxg3 
                       4.De2+ Kh6 5.Kxg3 g5 6.Dh2+ Kg6 
                       7.Dh1 Kxf6 8.Da1+ Kg6 9.Dxh8 f4+ 
                       10.Kh3 e3 11.Kg4 e2 12.De8 Kf6 
                       13.Dd8+ Ke6 14.Db6+ (63.918.828.531) 91887 

Hmm, maybe I need to take a closer look at Matefish again.

If you look at Matefish once in a while again, latest sources are still from May 2024, aren't they? Meanwhile, I wouldn't even need ready to use compile anymore, would make one of my own, just want to be sure, you wouldn't refresh the repo just a few days after I downloaded now from here
https://github.com/joergoster/Stockfish ... s/tag/v3.2
Better yet to wait a little longer still?
Sorry for my insistence

[Joerg Oster]

Code: Select all

7n/5p1b/5Pp1/5ppk/4p1r1/5QPK/6P1/8 w - -
Engine: The Huntsman 1 (32768 MB)
von the Stockfish developers (see AUTHORS f

75     11:35  +M22     1.Dxg4+ fxg4+ 2.Kh2 e3 3.Kg1 Kh6 
                       4.Kf1 Lg8 5.Ke2 Lh7 6.Kxe3 Lg8 7.Ke4 Lh7 
                       8.Ke5 Lg8 9.Kd6 Kh5 10.Kd7 Lh7 
                       11.Ke7 Kh6 12.Kf8 Kh5 13.Kg7 Lg8 
                       14.Kxh8 (63.918.828.531) 91887 

74     11:35  +M29     1.De2 Lg8 2.Kh2 e3 3.Dxe3 Te4 4.g4+ Txg4 
                       5.Df3 Kh6 6.g3 Ta4 7.Dh1 Kh5 8.Kg2+ Th4 
                       9.gxh4 gxh4 10.Dd1+ Kh6 11.Dd8 g5 
                       12.Dxg8 Sg6 13.Dxf7 g4 14.Dg7+ (63.918.828.531) 91887 

74     11:35  +M29     1.Dd1 e3 2.Kh2 Lg8 3.De2 Lh7 4.Dxe3 Te4 
                       5.g4+ Txg4 6.Df3 Kh6 7.g3 Ta4 8.Dh1 Kh5 
                       9.Kg2+ Th4 10.gxh4 gxh4 11.Dd1+ Kh6 
                       12.Dd8 Kg5 13.Dxh8 Lg8 14.Dxg8 (63.918.828.531) 91887 

74     11:35  +M36     1.De3 Th4+ 2.gxh4 gxh4 3.g3 hxg3 
                       4.De2+ Kh6 5.Kxg3 g5 6.Dh2+ Kg6 
                       7.Dh1 Kxf6 8.Da1+ Kg6 9.Dxh8 f4+ 
                       10.Kh3 e3 11.Kg4 e2 12.De8 Kf6 
                       13.Dd8+ Ke6 14.Db6+ (63.918.828.531) 91887 

Hmm, maybe I need to take a closer look at Matefish again.

If you look at Matefish once in a while again, latest sources are still from May 2024, aren't they? Meanwhile, I wouldn't even need ready to use compile anymore, would make one of my own, just want to be sure, you wouldn't refresh the repo just a few days after I downloaded now from here
https://github.com/joergoster/Stockfish ... s/tag/v3.2
Better yet to wait a little longer still?
Sorry for my insistence

No problem.
The latest sources are from January 16th.
You can find them here: https://github.com/joergoster/Stockfish/tree/matefish2
Click on the green 'Code' button and then choose 'Download zip'.

I don't think I will release anything new in the near future.

[peter]

The latest sources are from January 16th.
You can find them here: https://github.com/joergoster/Stockfish/tree/matefish2

Thanks for the prompt answer and the link to the yet somewhat newer sources, regards

[Ajedrecista]

Hello:

[...]

But it does confirm a Mate in 22:

Code: Select all

info string Success! Mate in 22 found!
info time 11374 multipv 1 depth 43 seldepth 43 nodes 42112165 nps 3702493 hashfull 43 tbhits 0 score mate 22 pv h3h2 h7g8 h2g1 g8h7 g1f2 e4e3 f2e3 h7g8 e3e2 g8h7 e2d3 h7g8 d3e4 g8h7 e4e5 h7g8 e5d6 g8h7 d6e7 h7g8 e7f8 g8h7 f8g7 h7g8 g7g8 h5h6 g8h8 h6h5 h8g8 h5h6 g8f7 h6h5 f7e6 h5h6 f6f7 h6h5 f7f8r h5h6 f8g8 h6h7 e6f7 h7h6 g8h8
bestmove h3h2 ponder h7g8

Pushing the pawn immediately after Kh3-h2 doesn't seem to be the best play for Black.

@Jörg: I wrote in my PGN that both 2.- ..., e3, 2.- ..., Bg8 and 2.- ..., Kh6 lose in the same number of moves, supposing optimal play by both sides, while you said that Pushing the pawn immediately after Kh3-h2 doesn't seem to be the best play for Black, this is, playing 2.- ..., e3. I let Chest 5.2 to be the referee, under normal search mode and 'defender moves' option enabled:

[d]7n/5p1b/5Pp1/6pk/4p1p1/6P1/6PK/8 b - - 1 2

Code: Select all

FEN: 7n/5p1b/5Pp1/6pk/4p1p1/6P1/6PK/8 b - - 1 2

ChestUCI:
FEN: 7n/5p1b/5Pp1/6pk/4p1p1/6P1/6PK/8 b - -   (4+8)
Stellungs-Analyse:  S:3/3
Suche nach Matt in -20 ...  (Hash=256MB)
Suche abgeschlossen ...  (Zeit=69.33s)
Matt in -20 gefunden !  (3 Lösungen in 01:09)
 20/20	01:09	  52.819.122	761.894	-M20	Kh5h6 Kh2g1 e4e3 Kg1f1 Bh7g8 Kf1e2 Bg8h7 Ke2xe3 Bh7g8 Ke3d3 Bg8h7 Kd3d4 Bh7g8 Kd4d5 Bg8h7 Kd5d6 Bh7g8 Kd6e7 Bg8h7 Ke7f8 Kh6h5 Kf8g7 Bh7g8 Kg7xg8 Kh5h6 Kg8xh8 Kh6h5 Kh8g8 Kh5h6 Kg8xf7 Kh6h7 Kf7e7 Kh7h8 f6f7 Kh8h7 Ke7f6 Kh7h8 f7f8Q+ Kh8h7 Qf8g7+
 20/20	01:09	  52.819.122	761.894	-M20	Bh7g8 Kh2g1 e4e3 Kg1f1 Bg8h7 Kf1e2 Bh7g8 Ke2xe3 Bg8h7 Ke3d3 Bh7g8 Kd3d4 Bg8h7 Kd4d5 Bh7g8 Kd5d6 Bg8h7 Kd6e7 Bh7g8 Ke7f8 Bg8h7 Kf8g7 Bh7g8 Kg7xg8 Kh5h6 Kg8xh8 Kh6h5 Kh8g8 Kh5h6 Kg8xf7 Kh6h7 Kf7e7 Kh7h8 f6f7 Kh8h7 Ke7f6 Kh7h8 f7f8Q+ Kh8h7 Qf8g7+
 20/20	01:09	  52.819.122	761.894	-M20	e4e3 Kh2g1 e3e2 Kg1f2 e2e1Q+ Kf2xe1 Bh7g8 Ke1e2 Bg8h7 Ke2e3 Bh7g8 Ke3e4 Bg8h7 Ke4e5 Bh7g8 Ke5d6 Bg8h7 Kd6e7 Bh7g8 Ke7f8 Bg8h7 Kf8g7 Bh7g8 Kg7xg8 Kh5h6 Kg8xh8 Kh6h5 Kh8g8 Kh5h6 Kg8xf7 Kh6h7 Kf7e7 Kh7h8 f6f7 Kh8h7 Ke7f6 Kh7h8 f7f8Q+ Kh8h7 Qf8g7+

It seems that we found a blind spot with this family of positions. Thank you everyone for the numerous insights!

------------

Code: Select all

7n/5p1b/5Pp1/5ppk/4p1r1/5QPK/6P1/8 w - -
Engine: The Huntsman 1 (32768 MB)
von the Stockfish developers (see AUTHORS f

75     11:35  +M22     1.Dxg4+ fxg4+ 2.Kh2 e3 3.Kg1 Kh6 
                       4.Kf1 Lg8 5.Ke2 Lh7 6.Kxe3 Lg8 7.Ke4 Lh7 
                       8.Ke5 Lg8 9.Kd6 Kh5 10.Kd7 Lh7 
                       11.Ke7 Kh6 12.Kf8 Kh5 13.Kg7 Lg8 
                       14.Kxh8 (63.918.828.531) 91887 

74     11:35  +M29     1.De2 Lg8 2.Kh2 e3 3.Dxe3 Te4 4.g4+ Txg4 
                       5.Df3 Kh6 6.g3 Ta4 7.Dh1 Kh5 8.Kg2+ Th4 
                       9.gxh4 gxh4 10.Dd1+ Kh6 11.Dd8 g5 
                       12.Dxg8 Sg6 13.Dxf7 g4 14.Dg7+ (63.918.828.531) 91887 

74     11:35  +M29     1.Dd1 e3 2.Kh2 Lg8 3.De2 Lh7 4.Dxe3 Te4 
                       5.g4+ Txg4 6.Df3 Kh6 7.g3 Ta4 8.Dh1 Kh5 
                       9.Kg2+ Th4 10.gxh4 gxh4 11.Dd1+ Kh6 
                       12.Dd8 Kg5 13.Dxh8 Lg8 14.Dxg8 (63.918.828.531) 91887 

74     11:35  +M36     1.De3 Th4+ 2.gxh4 gxh4 3.g3 hxg3 
                       4.De2+ Kh6 5.Kxg3 g5 6.Dh2+ Kg6 
                       7.Dh1 Kxf6 8.Da1+ Kg6 9.Dxh8 f4+ 
                       10.Kh3 e3 11.Kg4 e2 12.De8 Kf6 
                       13.Dd8+ Ke6 14.Db6+ (63.918.828.531) 91887 

[...]

@Peter: so Huntsman seems to agree that to sac the queen ASAP is the fastest win after a very deep search of depth 75, way higher than the expected shortest win in 43 plies. I do not know if Hunstman already already saw that #22 many depths before.

Regards from Spain.

Ajedrecista.

[peter]

@Peter: so Huntsman seems to agree that to sac the queen ASAP is the fastest win after a very deep search of depth 75, way higher than the expected shortest win in 43 plies. I do not know if Hunstman already already saw that #22 many depths before.

It did, don't know at which one first, because the pasted output was copied after some Forward- Backward. You probably can easily reproduce it that way, the 30 threads and 32G hash I used, wouldn't have been necessary with MultiPV=1, as you see, even MultiPV=4 gets the shown depth in reasonable time with this one hardware, so I guess 8 good threads and 8G should be fully enough to reproduce single primary output backwarding already given (probably) best line, regards