Last chance to win

[pferd]

I watched a game between Critter and Stockfish and here is a slightly modified position from it

[D]1r5k/p1q3p1/P2p1pPr/1PpPpP1P/2Pn2B1/4R3/Q4RK1/8 w - - 0 1


White has a last shot to the win game with b6! and gaining control over the open b file. I tried couple of engines for some minutes on 4 cores but without success. Only Houdini found it at d34. The others shuffle around and want to play something like Kf1 or Rd3.

Maybe you are luckier than me?

[MikeGL]

I watched a game between Critter and Stockfish and here is a slightly modified position from it

[D]1r5k/p1q3p1/P2p1pPr/1PpPpP1P/2Pn2B1/4R3/Q4RK1/8 w - - 0 1


White has a last shot to the win game with b6! and gaining control over the open b file. I tried couple of engines for some minutes on 4 cores but without success. Only Houdini found it at d34. The others shuffle around and want to play something like Kf1 or Rd3.

Maybe you are luckier than me?

What's the complete winning line?
Maybe white has some resources to hold a draw.

[AdminX]

I watched a game between Critter and Stockfish and here is a slightly modified position from it

[D]1r5k/p1q3p1/P2p1pPr/1PpPpP1P/2Pn2B1/4R3/Q4RK1/8 w - - 0 1


White has a last shot to the win game with b6! and gaining control over the open b file. I tried couple of engines for some minutes on 4 cores but without success. Only Houdini found it at d34. The others shuffle around and want to play something like Kf1 or Rd3.

Maybe you are luckier than me?

What's the complete winning line?
Maybe white has some resources to hold a draw.

1r5k/p1q3p1/P2p1pPr/1PpPpP1P/2Pn2B1/4R3/Q4RK1/8 w - - 0 1

Analysis by Houdini 6.03 Pro x64-pext (Regular):

1.b6 Qxb6 2.Rd3 Qb4 3.Rdd2 Nb3 4.Rc2 Qe1 5.Be2 Nd4 6.Rb2 Ra8 7.Qb1 Qxb1 8.Rxb1 Nc2 9.Rb7 Kg8 10.Kg3 Kf8 11.Rf7+ Kg8 12.Rd7 Kf8 13.Rxd6 Ke7 14.Re6+ Kf8 15.Rc6 Nb4 16.Rc7 Nxa6 17.Rf7+ Kg8 18.d6 Rd8 19.Rxa7 Nb4 20.Bf3 Kf8 21.Rf7+ Kg8 22.d7 Nd3 23.Ra2 e4 24.Bxe4 Ne5 25.Ra8 Nxf7 26.gxf7+ Kxf7 27.Rxd8 Ke7 28.Rg8 Kxd7 29.Rxg7+ Kd6 30.Bf3
White is clearly winning: +- (4.38 ++) Depth: 36/89 00:08:14 5389MN, tb=395968

[zullil]

I watched a game between Critter and Stockfish and here is a slightly modified position from it

[D]1r5k/p1q3p1/P2p1pPr/1PpPpP1P/2Pn2B1/4R3/Q4RK1/8 w - - 0 1


White has a last shot to the win game with b6! and gaining control over the open b file. I tried couple of engines for some minutes on 4 cores but without success. Only Houdini found it at d34. The others shuffle around and want to play something like Kf1 or Rd3.

Maybe you are luckier than me?

What's the complete winning line?
Maybe white has some resources to hold a draw.

1r5k/p1q3p1/P2p1pPr/1PpPpP1P/2Pn2B1/4R3/Q4RK1/8 w - - 0 1

Analysis by Houdini 6.03 Pro x64-pext (Regular):

1.b6 Qxb6 2.Rd3 Qb4 3.Rdd2 Nb3 4.Rc2 Qe1 5.Be2 Nd4 6.Rb2 Ra8 7.Qb1 Qxb1 8.Rxb1 Nc2 9.Rb7 Kg8 10.Kg3 Kf8 11.Rf7+ Kg8 12.Rd7 Kf8 13.Rxd6 Ke7 14.Re6+ Kf8 15.Rc6 Nb4 16.Rc7 Nxa6 17.Rf7+ Kg8 18.d6 Rd8 19.Rxa7 Nb4 20.Bf3 Kf8 21.Rf7+ Kg8 22.d7 Nd3 23.Ra2 e4 24.Bxe4 Ne5 25.Ra8 Nxf7 26.gxf7+ Kxf7 27.Rxd8 Ke7 28.Rg8 Kxd7 29.Rxg7+ Kd6 30.Bf3
White is clearly winning: +- (4.38 ++) Depth: 36/89 00:08:14 5389MN, tb=395968

Latest Cfish agrees, but it takes a long time to find b6:

Code: Select all

info depth 68 seldepth 70 multipv 1 score cp 255 nodes 691220236471 nps 47036935 hashfull 999 tbhits 0 time 14695265 pv e3d3 c7b6 d3d1 h8g8 f2b2 b8e8 d1e1 g8f8 e1a1 f8e7 a2a5 e8b8 a5b6 b8b6 g2f2 e7f8 f2e3 h6h8 a1h1 h8h6 b2f2 b6b8 h1b1 b8b6 e3e4 f8e7 f2h2 e7d8 h2g2 d8e7 g2g3 e7d7 g3h3 d7e7 b1b2 e7d7 h3h2 d7c7 e4e3 b6b8 h2g2 b8e8 e3e4 e8b8 b2b1 h6h8 g2g1 h8h6 g1h1 b8h8 h1h2 h8b8 b1a1 b8a8 h2f2 a8h8 a1h1 c7d7 f2b2 d7c7 b2h2 h8e8 h1e1 e8b8 e4e3 b8h8 e3d3 h8e8 e1b1 e8b8
info depth 69 currmove e3d3 currmovenumber 1
info depth 69 currmove f2b2 currmovenumber 2
info depth 69 currmove b5b6 currmovenumber 3
info depth 69 seldepth 84 multipv 1 score cp 262 lowerbound nodes 925267851555 nps 47071806 hashfull 999 tbhits 0 time 19656519 pv b5b6
info depth 69 currmove b5b6 currmovenumber 1
info depth 69 seldepth 84 multipv 1 score cp 270 lowerbound nodes 1014306355933 nps 47092004 hashfull 999 tbhits 0 time 21538823 pv b5b6
info depth 69 currmove b5b6 currmovenumber 1
info depth 69 seldepth 84 multipv 1 score cp 281 lowerbound nodes 1089651322405 nps 47135586 hashfull 999 tbhits 0 time 23117381 pv b5b6
info depth 69 currmove b5b6 currmovenumber 1
info depth 69 seldepth 84 multipv 1 score cp 297 lowerbound nodes 1180862147900 nps 47232660 hashfull 999 tbhits 0 time 25000966 pv b5b6
info depth 69 currmove b5b6 currmovenumber 1
info depth 69 seldepth 84 multipv 1 score cp 318 lowerbound nodes 1240513987451 nps 47262661 hashfull 999 tbhits 0 time 26247231 pv b5b6
info depth 69 currmove b5b6 currmovenumber 1
info depth 69 seldepth 84 multipv 1 score cp 347 lowerbound nodes 1302645999256 nps 47302661 hashfull 999 tbhits 0 time 27538535 pv b5b6
info depth 69 currmove b5b6 currmovenumber 1
info depth 69 seldepth 84 multipv 1 score cp 386 lowerbound nodes 1389031641967 nps 47365136 hashfull 999 tbhits 0 time 29326035 pv b5b6
info depth 69 currmove b5b6 currmovenumber 1
info depth 69 seldepth 84 multipv 1 score cp 436 lowerbound nodes 1501288903860 nps 47442108 hashfull 999 tbhits 0 time 31644650 pv b5b6

[MikeGL]

I watched a game between Critter and Stockfish and here is a slightly modified position from it

[D]1r5k/p1q3p1/P2p1pPr/1PpPpP1P/2Pn2B1/4R3/Q4RK1/8 w - - 0 1


White has a last shot to the win game with b6! and gaining control over the open b file. I tried couple of engines for some minutes on 4 cores but without success. Only Houdini found it at d34. The others shuffle around and want to play something like Kf1 or Rd3.

Maybe you are luckier than me?

What's the complete winning line?
Maybe white has some resources to hold a draw.

1r5k/p1q3p1/P2p1pPr/1PpPpP1P/2Pn2B1/4R3/Q4RK1/8 w - - 0 1

Analysis by Houdini 6.03 Pro x64-pext (Regular):

1.b6 Qxb6 2.Rd3 Qb4 3.Rdd2 Nb3 4.Rc2 Qe1 5.Be2 Nd4 6.Rb2 Ra8 7.Qb1 Qxb1 8.Rxb1 Nc2 9.Rb7 Kg8 10.Kg3 Kf8 11.Rf7+ Kg8 12.Rd7 Kf8 13.Rxd6 Ke7 14.Re6+ Kf8 15.Rc6 Nb4 16.Rc7 Nxa6 17.Rf7+ Kg8 18.d6 Rd8 19.Rxa7 Nb4 20.Bf3 Kf8 21.Rf7+ Kg8 22.d7 Nd3 23.Ra2 e4 24.Bxe4 Ne5 25.Ra8 Nxf7 26.gxf7+ Kxf7 27.Rxd8 Ke7 28.Rg8 Kxd7 29.Rxg7+ Kd6 30.Bf3
White is clearly winning: +- (4.38 ++) Depth: 36/89 00:08:14 5389MN, tb=395968

Thanks for the winning line.
Looks cool and counter-intuitive when a bad B (hemmed by own pawns)
defeats a well placed N of black.
Activity of all 6-major pieces (3w+3b) trumps N vs B eval in this case it seems.

[pferd]

After a long search I came up with another engine beating this problem.

LazyGull solved it after a long 3 hour search:

Code: Select all

dep	score	nodes	time	(not shown:  tbhits	knps	seldep)
 32	+0.81 	122.5G	3:01:24	b6 Qxb6 Rd3 Qb4 Rdd2 Nb3 Rc2 Qe1 Be2 Nd4 Rb2 Rf8 Qb1 Qxb1 Rxb1 Nxe2 Rxe2 Rxh5 Rb7 Rg5+ Kh1 Rh5+ Rh2 Rxh2+ Kxh2 Ra8 Kg3 Kg8 Kf3 Kf8 Rf7+ Kg8 Rd7 Kf8 Ke4 Kg8 Rxd6 
 32	+0.73 	43.3G	1:05:11	Rf1 Rb6 Qb1 Qc8 Bd1 Kg8 Bc2 Kf8 Rh1 Rb8 Ra3 Ke7 Be4 Qc7 Kf2 Rd8 Qd1 Qd7 Qd3 Rb8 Qb1 Qd8 Ke1 Qc7 Kd2 Qd8 Rhh3 Qd7 Ke1 Qc8 Kd2 Qd7 
 31	+0.73 	22.0G	33:16.72	Rf1 Rb6 Qb1 Qc8 Bd1 Kg8 Bc2 Kf8 Rh1 Rb8 Ra3 Ke7 Be4 Qc7 Kf2 Rd8 Qd1 Qd7 Qd3 Rb8 Qb1 Qd8 Ke1 Qc7 Kd2 Qd8 Rhh3 Qd7 Ke1 Qc8 Kd2 Qd7 
 30	+0.73 	17.4G	26:26.13	Rf1 Rb6 Qb1 Qc8 Bd1 Kg8 Bc2 Kf8 Rh1 Rb8 Ra3 Ke7 Be4 Qc7 Kf2 Rd8 Qd1 Qd7 Qd3 Rb8 Qb1 Qd8 Ke1 Qc7 Kd2 Qd8 
 29	+0.73 	6.72G	9:59.10	Rf1 Rb6 Qb1 Qc8 Bd1 Kg8 Bc2 Kf8 Rh1 Rb8 Ra3 Ke7 Be4 Qc7 Kf2 Rd8 Qd1 Qd7 Qd3 Rb8 Qb1 Qc7 Ke1 Qc8 Kd2 Qd7 Qd3 Rc8 Qb1 Rch8 b6 Rxh5 Rxh5 
 28	+0.73 	4.03G	6:00.85	Rf1 Rb6 Qb1 Qc8 Bd1 Kg8 Bc2 Kf8 Rh1 Rb8 Ra3 Ke7 Be4 Qc7 Kf2 Rd8 Qd1 Qd7 Qd3 Rb8 Qb1 Qc7 Ke1 Qc8 Kd2 Qd7 Rhh3 Qc7 
 27	+0.73 	2.96G	4:24.13	Rf1 Rb6 Qb1 Qc8 Bd1 Kg8 Bc2 Kf8 Rh1 Rb8 Ra3 Ke7 Be4 Qc7 Kf2 Rd8 Qd1 Qd7 Qd3 Rb8 Qb1 Qc7 Ke1 Qc8 Kd2 Qd7 Rhh3 Rhh8 h6 gxh6 Rh5 
 26	+0.69 	2.76G	4:06.20	Rf1 Kg8 b6 Qxb6 Ree1 Qb2+ Qxb2 Rxb2+ Kg1 Rb4 Rb1 Nc2 Rxb4 Nxb4 Kg2 Rh8 Ra1 Nc2 Rb1 Nb4 Kf3 Kf8 Ke4 Ke7 Rb3 Kd8 Rh3 Rh6 Be2 Kc7 
 25	+0.65 	236.8M	0:21.33	Rf1 Kg8 Rfe1 Kf8 Ra3 Ke7 Qd2 Rhh8 Kf2 Kf8 Rae3 Rh6 Kg2 Ke7 Kh2 Rhh8 Rh3 Rh6 Rd3 Rhh8 Kg2 Rbd8 Rh3 Rh6 Rd3 Rb8 Rh3 Kf8 Rhe3 
 24	+0.65 	113.7M	0:10.31	Rf1 Kg8 Rfe1 Kf8 Ra3 Ke7 Qd2 Rhh8 Kf2 Kf8 Rae3 Rh6 Kg2 Ke7 Kh2 Rhh8 Rh3 Rh6 Rd3 Rhh8 Kg2 Rbd8 Rh3 Rh6 Rd3 Rb8 Rh3 Kf8 Rhe3 
 23	+0.65 	60.2M  	0:05.57	Rf1 Kg8 Rfe1 Kf8 Ra3 Ke7 Qd2 Rhh8 Kf2 Kf8 Rae3 Rh6 Kg2 Ke7 Kh2 Rhh8 Rh3 Rh6 Rd3 Rhh8 Kg2 Rbd8 Rh3 Rh6 Rd3 Rb8 Rh3 Kf8 Rhe3 
 22	+0.65 	39.1M  	0:03.72	Rf1 Kg8 Rfe1 Kf8 Ra3 Ke7 Qd2 Rhh8 Kf2 Kf8 Rae3 Rh6 Kg2 Ke7 Kh2 Rhh8 Rh3 Rh6 Kg2 Kf8 Rhe3 
 21	+0.65 	20.4M  	0:02.09	Rf1 Kg8 Rfe1 Kf8 Ra3 Ke7 Qd2 Rhh8 Kf2 Kf8 Rae3 Rh6 Kg2 Ke7 Ra3 
 20	+0.65 	10.9M  	0:01.26	Rf1 Kg8 Rfe1 Kf8 Ra3 Ke7 Qd2 Rhh8 Kf2 Kf8 Rae3 Rh6 Kg2 Ke7 Ra3 
 19	+0.65 	6.95M  	0:00.91	Rf1 Kg8 Rfe1 Kf8 Ra3 Ke7 Qd2 Rhh8 Kf2 Kf8 Rae3 Rh6 Ra3 Rh8 
 18	+0.65 	3.35M  	0:00.56	Rf1 Kg8 Rfe1 Kf8 Ra3 Ke7 Qd2 Rhh8 Kf2 Kf8 Rae3 Rh6 Ra3 Rh8 
 17	+0.65 	2.06M  	0:00.41	Rf1 Kg8 Rfe1 Kf8 Ra3 Ke7 Qd2 Qd7 Kh2 Rbh8 Rh3 Qc7 Kg2 Rb8 Rhe3 Rhh8 Rh3 Rh6 
 16	+0.65 	1.28M  	0:00.29	Rf1 Kg8 Rfe1 Kf8 Kf2 Ke7 Qd2 Rhh8 Rh3 Rh6 Kg2 Kf8 Kf2 Ke7 
 15	+0.65 	844169	0:00.21	Rf1 Kg8 Rfe1 Kf8 Kf2 Ke7 Qd2 Rhh8 Rh3 Rh6 Kg2 Kf8 Kf2 Ke7 
 14	+0.65 	478666	0:00.13	Rf1 Kg8 Rfe1 Kf8 Kf2 Ke7 Qd2 Rhh8 Rh3 Rh6 Kg2 Kf8 Kf2 Qd7 Kg2 Ke7 Rd3 
 13	+0.65 	436892	0:00.12	Rf1 Kg8 Rfe1 Kf8 Kf2 Ke7 Qd2 Rhh8 Rh3 Rh6 Kg2 Kf8 Kf2 Qd7 Kg2 Ke7 Rd3 
 12	+0.65 	348809	0:00.10	Rf1 Kg8 Rfe1 Kf8 Kf2 Ke7 Qd2 Rhh8 Rh3 Rh6 Kg2 Rd8 Ra3 Kf8 Kf2 Ke7 Qa5 
 11	+0.63 	224787	0:00.06	Rf1 Kg8 Qd2 Kf8 Kf2 Ke7 Rfe1 Rhh8 Rh3 Rh6 Kg2 Rd8 Rhe3 Rhh8 
 10	+0.64 	170735	0:00.05	Rf1 Kg8 Qd2 Kf8 Kf2 Ke7 Rfe1 Kd7 Ra3 Ke7 Qa5 
 10	+0.63 	155256	0:00.05	Qd2 Kg8 Rf1 Kf8 Kf2 Ke7 Rfe1 Kd7 Ra3 Ke7 Kg2 Rd8 
  9	+0.66 	113885	0:00.04	Qd2 Kg8 Rf1 Kf8 Kf2 Qd7 Kg1 Re8 Qg2 Ke7 Kh2 
  8	+0.66 	103828	0:00.03	Qd2 Kg8 Rf1 Kf8 Kf2 Qd7 Kg1 Re8 Qg2 Ke7 
  8	+0.65 	65264  	0:00.02	Rf1 Kg8 Qd2 Kf8 Kh2 Rh8 Qf2 Ke7 
  8	+0.75 	32832  	0:00.01	Rf1 Kg8 Qd2 Kf8 Rd1 Ke7 Rde1 Rhh8 Rh3 Rh6 
  7	+0.75 	22233  	0:00.01	Rf1 Kg8 Qd2 Kf8 Rd1 Ke7 Rde1 Rhh8 

[MikeB]

I watched a game between Critter and Stockfish and here is a slightly modified position from it

[D]1r5k/p1q3p1/P2p1pPr/1PpPpP1P/2Pn2B1/4R3/Q4RK1/8 w - - 0 1


White has a last shot to the win game with b6! and gaining control over the open b file. I tried couple of engines for some minutes on 4 cores but without success. Only Houdini found it at d34. The others shuffle around and want to play something like Kf1 or Rd3.

Maybe you are luckier than me?

Good position - but the rook a h6 appears to have been placed there artificially and you end up with a position that never would have been arrived from a real game.

So I modified the position, replacing the h6 rook with a pawn and the premise is still the same - last chance for white to win is to play b6.


[d]1r5k/p1q3p1/P2p1pPp/1PpPpP1P/2Pn2B1/4R3/Q4RK1/8 w - - 0 1

now SF-McB 4.1 sees it quite easily on my machine.

dep score nodes time (not shown: tbhits knps seldep)
44 +12.57! 12.6G 8:15.12 b6!
43 +11.40! 8.26G 5:36.47 b6!
43 +9.98! 5.16G 3:41.50 b6!
43 +9.98! 5.16G 3:41.50 b6!
43 +8.85! 2.88G 2:06.89 b6!
43 +7.97! 1.71G 1:11.10 b6!
43 +7.28! 1.14G 0:46.82 b6!
42 +6.59 661.9M 0:26.91 Qd2 Kg8 Bd1 Qd7 Bc2 Qc8 Kh2 Kf8 Be4 Qd8 Re1 Rb6 Qc3 Ke7 Re3 Rb8 Qe1 Kf8 Ra3 Ke7 Qe3 Kd7 Rf1 Ke8 Qf2 Kd7 Qd2 Qc8 Kh3 Ke7 Qd3 Qd8 Raa1 Qc8 Bg2 Qd8 Rae1 Qa5 Kh2 Kd7 Ra1 Qc7
41 +6.59 432.1M 0:17.42 Qd2 Kg8 Bd1 Qd7 Bc2 Qc8 Kh2 Kf8 Be4 Qd8 Re1 Rb6 Qc3 Ke7 Re3 Rb8 Qe1 Kf8 Ra3 Ke7 Rf1 Ke8 Kg2 Qc8 Rf2 Ke7 Qa5 Kd7 Rb2 Rb6 Kh2 Ke7 Rf2 Rb8 Rf1 Kd7 Qd2 Qd8 Kh3 Ke7 Re3 Qc8
40 +6.59 291.3M 0:11.74 Qd2 Kg8 Bd1 Qd7 Bc2 Kf8 Kh3 Ke7 Be4 Qd8 Kh2 Kf8 Re1 Rb6 Qc3 Ke7 Re3 Rb8 Qe1 Kf8 Ra3 Ke7 Rf1 Ke8 Kg2 Qc8 Re3 Qd8 Kh2 Kd7 Qc3 Qc8 Ree1 Ke7 Qd2 Qd7 Qd3 Qd8 Qa3 Qc7
39 +6.59 201.7M 0:08.18 Qd2 Kg8 Bd1 Qd7 Bc2 Kf8 Kh3 Ke7 Be4 Qd8 Kh2 Kf8 Re1 Qc8 Qa2 Rb6 Re3 Ke7 Ra3 Rb8 Kg2 Qc7 Kg3 Kd7 Qd2 Qc8 Kh3 Ke7 Kh4 Qd8 Kg3 Ke8 Kh3 Kd7 Kg4 Qc8 Kg3 Ke7 Re3
38 +6.59 121.8M 0:05.00 Qd2 Kg8 Bd1 Kf8 Bc2 Qc8 Be4 Ke7 Qa5 Ke8 Kh3 Kf8 Qd2 Ke8 Ra3 Ke7 Kh4 Qd8 Qe3 Qc8 Kh3 Qd8 Ra1 Kf8 Rff1 Ke8 Kh4 Kd7 Rf2 Ke7 Kg4 Kd7 Ra3 Qc8 Kh3 Ke7 Qc3 Qd8 Qe1

[Ovyron]

now SF-McB 4.1 sees it quite easily on my machine.

Yeah, I have done those tricks in the past. I.e. an engine shows some equalish score for a position, and since material is equal, it's suggesting some passive moves. But you know the opponent's Bishop is not better than a pawn, so you edit the FEN and force the engine to analyze the fake position with the pawn instead of the Bishop, and suddenly, the scores rightly shoot up and it suggests moves that work in the original position, and basically, no engine would have suggested those moves in the first position.

I wonder if this can be done automatically? Like, a detector for the engine that can be turned on where it sees there's these big pieces stuck that are no better than a pawn and treating them as such will help the engine find the best move?

This may not be practical for games, but when you have to edit FENs to trick the engine into seeing wins there ought the be a potential improvement somewhere.