Baffling test position

[hgm]

Can your favorite engine handle this position?

[d] 1q1qk1q1/pppppppp/8/8/8/8/PPPPPPPP/NNN1KNNN w - - 0 1

Who do you think would have the advantage in this position? (I.e., which side would on the average win more often, when played between equally skilled players?)

It would be interesting to see games of many different engines, playing both sides (i.e. self-play), so that we can be sure the result is not due to the winning engine being stronger, but must be due to the imbalance of the initial position.

[David Dahlem]

Three queens against 6 knights? That's easy, black wins.

[Aaron Becker]

I ran a bunch of short time control games with Daydreamer playing against itself. The result was +13-78=9 for white.

[zullil]

Can your favorite engine handle this position?

[d] 1q1qk1q1/pppppppp/8/8/8/8/PPPPPPPP/NNN1KNNN w - - 0 1

Who do you think would have the advantage in this position? (I.e., which side would on the average win more often, when played between equally skilled players?)

It would be interesting to see games of many different engines, playing both sides (i.e. self-play), so that we can be sure the result is not due to the winning engine being stronger, but must be due to the imbalance of the initial position.

I couldn't resist watching Stockfish play itself from this starting position. Black won easily:

Code: Select all

[Event "?"]
[Site "?"]
[Date "2010.03.20"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "0-1"]
[SetUp "1"]
[FEN "1q1qk1q1/pppppppp/8/8/8/8/PPPPPPPP/NNN1KNNN w - - 0 1"]

1. Nab3 f5 2. d4 b6 3. Nd3 d6 4. Ne3 c5 5. c4 b5 6. N1d2 d5 7. cxd5 Qxh2 8. Kf1
Qxh1 9. dxc5 Qb8 10. d6 exd6 11. Nxf5 dxc5 12. Ng3 Qh6 13. Ndxc5 a5 14. Ngf3 a4
15. Nc1 Qc7 16. N5d3 Qhc6 17. Nfe1 Qxc1 18. Nxc1 Qxc1 19. Nf3 Qxa2 20. Ne5
Qcxb2 21. N5d3 Qd2 22. Kg1 b4 23. Ne4 Qxe2 24. Ng3 Qd1 25. Kh2 b3 26. Ne4 Qde2
27. Nd6+ Kf8 28. Nc4 a3 29. Nb6 b2 30. Nd7+ Ke8 31. N7e5 b1=Q 32. N1f3 Qbxd3
33. Nxd3 Qxd3 34. Ne5 Qe4 35. Kg3 Qxe5+ 36. f4 Qh5 37. f5 Qg5+ 38. Kh3 Qaxg2#
0-1

[bob]

Can your favorite engine handle this position?

[d] 1q1qk1q1/pppppppp/8/8/8/8/PPPPPPPP/NNN1KNNN w - - 0 1

Who do you think would have the advantage in this position? (I.e., which side would on the average win more often, when played between equally skilled players?)

It would be interesting to see games of many different engines, playing both sides (i.e. self-play), so that we can be sure the result is not due to the winning engine being stronger, but must be due to the imbalance of the initial position.

Crafty doesn't allow illegal positions. in this case, an impossible number of knights and queens, since you can't have more than 9 total queens+pawns, or more than 10 total knights + pawns.

[hgm]

Three queens against 6 knights? That's easy, black wins.

Actually, this is not so obvious. It depends a lot on the engine, some are better in it than others. As you can see from the DayDereamer result, there is a significant number of wins by white. And this despite the fact that the engines will have a bad conception of the piece values, thinking that it is still a good deal tot trade three Knights for one Queen, while in fact this is the only way to bungle the win.

My suspicion is that a properly programmed engine, using piecve values Q=9 and N=5, would always win for whate. In any case this would be true for 7 Knights against 3 Queens.:

[d] 1q1qk1q1/pppppppp/8/8/8/8/PPPPPPPP/NNNNKNNN w

[zullil]

My suspicion is that a properly programmed engine, using piecve values Q=9 and N=5, would always win for whate. In any case this would be true for 7 Knights against 3 Queens.:

[d] 1q1qk1q1/pppppppp/8/8/8/8/PPPPPPPP/NNNNKNNN w

N=5?

In any case, here's another 3-min-per-side game using Stockfish. Even with seven knights, Black wins. Of course, this is just one game!

Code: Select all

[Event "?"]
[Site "?"]
[Date "2010.03.21"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "0-1"]
[SetUp "1"]
[FEN "1q1qk1q1/pppppppp/8/8/8/8/PPPPPPPP/NNNNKNNN w - - 0 1"]

1. Nab3 b6 2. Nhg3 c5 3. Nf3 c4 4. Nbd4 e5 5. Nb5 a6 6. N5c3 d5 7. Nfe3 d4 8.
Nxc4 dxc3 9. Nbxc3 f6 10. Nce3 Kf8 11. Ncd5 Qe6 12. Nd3 a5 13. N1c3 Qc6 14.
Ngf5 Qce8 15. Ne4 Qa4 16. Nec3 Qc6 17. Ne4 Qba8 18. Nfd6 Qa7 19. c4 a4 20. Nef5
Qa5 21. Ndxe5 fxe5 22. e3 g6 23. Neg5 Qcd7 24. Nb7 Qb8 25. Nxa5 bxa5 26. Nfe7
Qxb2 27. Ke2 Kg7 28. Ne4 Qg4 29. Ndc3 Qxg2 30. Ned5 Qc1 31. Ng3 e4 32. Ng5 h5
33. h4 Qh2 34. c5 Qxh4 35. Ng3xe4 Qg4+ 36. f3 Qd7 37. Nf4 Qa7 38. Nfe6+ Kh8 39.
Nd6 Kg8 40. Nge4 Qa6+ 41. Ncb5 Qa1 42. Nd4 h4 43. c6 h3 44. c7 h2 45. Nf2 Qxa2
46. f4 Kg7 47. Ne8+ Kh7 48. Nd6 Qg8 49. c8=Q Qaxc8 50. Nxc8 Qxc8 51. Na3 Qc1
52. Ndb5 Kg7 53. d4 Qb2+ 54. Kf3 Qc1 55. e4 h1=Q+ 56. Nxh1 Qxh1+ 57. Ke3 Qh3+
58. Kd2 Qg2+ 59. Ke3 Qg3+ 60. Ke2 Qxf4 61. Nc3 Qc1 62. Ncb5 Kf6 63. Kd3 g5 64.
e5+ Ke6 65. d5+ Kxd5 66. e6 Kxe6 67. Nc2 g4 68. Nc3 Qf1+ 69. Ne2 a3 70. Ne3
Qb1+ 71. Kd2 a2 72. Nd4+ Ke5 73. Ndc2 g3 74. Ng4+ Kf4 75. Nge3 Ke4 76. Ng2 a1=Q
77. Nxa1 Qxa1 78. Kc2 Qa2+ 79. Kc3 Qxg2 80. Kc4 Qc2+ 81. Kb5 g2 82. Ka6 g1=Q
83. Kb7 Qgb1+ 84. Ka6 Qcb2 85. Kxa5 Q2a1# 0-1

[bob]

My suspicion is that a properly programmed engine, using piecve values Q=9 and N=5, would always win for whate. In any case this would be true for 7 Knights against 3 Queens.:

[d] 1q1qk1q1/pppppppp/8/8/8/8/PPPPPPPP/NNNNKNNN w

N=5?

In any case, here's another 3-min-per-side game using Stockfish. Even with seven knights, Black wins. Of course, this is just one game!

Code: Select all

[Event "?"]
[Site "?"]
[Date "2010.03.21"]
[Round "?"]
[White "?"]
[Black "?"]
[Result "0-1"]
[SetUp "1"]
[FEN "1q1qk1q1/pppppppp/8/8/8/8/PPPPPPPP/NNNNKNNN w - - 0 1"]

1. Nab3 b6 2. Nhg3 c5 3. Nf3 c4 4. Nbd4 e5 5. Nb5 a6 6. N5c3 d5 7. Nfe3 d4 8.
Nxc4 dxc3 9. Nbxc3 f6 10. Nce3 Kf8 11. Ncd5 Qe6 12. Nd3 a5 13. N1c3 Qc6 14.
Ngf5 Qce8 15. Ne4 Qa4 16. Nec3 Qc6 17. Ne4 Qba8 18. Nfd6 Qa7 19. c4 a4 20. Nef5
Qa5 21. Ndxe5 fxe5 22. e3 g6 23. Neg5 Qcd7 24. Nb7 Qb8 25. Nxa5 bxa5 26. Nfe7
Qxb2 27. Ke2 Kg7 28. Ne4 Qg4 29. Ndc3 Qxg2 30. Ned5 Qc1 31. Ng3 e4 32. Ng5 h5
33. h4 Qh2 34. c5 Qxh4 35. Ng3xe4 Qg4+ 36. f3 Qd7 37. Nf4 Qa7 38. Nfe6+ Kh8 39.
Nd6 Kg8 40. Nge4 Qa6+ 41. Ncb5 Qa1 42. Nd4 h4 43. c6 h3 44. c7 h2 45. Nf2 Qxa2
46. f4 Kg7 47. Ne8+ Kh7 48. Nd6 Qg8 49. c8=Q Qaxc8 50. Nxc8 Qxc8 51. Na3 Qc1
52. Ndb5 Kg7 53. d4 Qb2+ 54. Kf3 Qc1 55. e4 h1=Q+ 56. Nxh1 Qxh1+ 57. Ke3 Qh3+
58. Kd2 Qg2+ 59. Ke3 Qg3+ 60. Ke2 Qxf4 61. Nc3 Qc1 62. Ncb5 Kf6 63. Kd3 g5 64.
e5+ Ke6 65. d5+ Kxd5 66. e6 Kxe6 67. Nc2 g4 68. Nc3 Qf1+ 69. Ne2 a3 70. Ne3
Qb1+ 71. Kd2 a2 72. Nd4+ Ke5 73. Ndc2 g3 74. Ng4+ Kf4 75. Nge3 Ke4 76. Ng2 a1=Q
77. Nxa1 Qxa1 78. Kc2 Qa2+ 79. Kc3 Qxg2 80. Kc4 Qc2+ 81. Kb5 g2 82. Ka6 g1=Q
83. Kb7 Qgb1+ 84. Ka6 Qcb2 85. Kxa5 Q2a1# 0-1

With reasonable play, black has to win this. Only hope for white is to reach a position where no matter what happens, white forks a black queen and king, and then it might turn into a forced draw since material will be equal. I'd still bet on the mobility of 2 queens vs 6 knights any day. It will be hard to hold every pawn, and all the knights, as the queen is a "fork-master" herself...

The knight is a very poor piece due to lack of mobility, when compared to a facefull of queens.

[hgm]

When I play my new engine (about as strong as Joker now, so around 24oo Elo) as white against Stockfish, the latter was crushed:

Code: Select all

[Event "Computer Chess Game"]
[Site "SCHAAK_PC"]
[Date "2010.03.21"]
[Round "-"]
[White "NewEng 0.16"]
[Black "Stockfish 1.6 JA"]
[Result "*"]
[TimeControl "40/300"]
[FEN "1q1qk1q1/pppppppp/8/8/8/8/PPPPPPPP/NNNNKNNN w - - 0 1"]
[SetUp "1"]

{--------------
. q . q k . q .
p p p p p p p p
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
P P P P P P P P
N N N N K N N N
white to play
--------------}
1. Nhg3 {+3.67/13 4} c5 {+9.65/15 16} 2. e4 {+3.82/13 7} e6 {+9.69/14 18}
3. Ng1e2 {+3.93/9} g6 {+10.10/14 28} 4. Nbc3 {+3.86/13 7} f5 {+10.22/13 15}
5. f4 {+3.81/15 8} b5 {+10.42/13 4} 6. exf5 {+3.91/13 4} gxf5
{+10.62/13 13} 7. Nab3 {+3.91/15 7} b4 {+10.46/14 15} 8. Nb1 {+4.06/10} h5
{+10.62/14 13} 9. Nfe3 {+4.09/15 10} h4 {+10.58/13 7} 10. Ngf1 {+4.23/15 6}
d6 {+10.22/13 11} 11. d4 {+4.77/14 5} cxd4 {+9.77/14 18} 12. Nbxd4
{+5.15/14 5} Kd7 {+9.69/13 16} 13. Nf2 {+5.40/13 4} Qf6 {+9.61/12 8} 14.
Nbd2 {+5.66/13 9} a5 {+9.21/13 9} 15. N2f3 {+5.78/13 4} a4 {+8.92/11 5} 16.
Nh3 {+5.95/13 5} Qge8 {+8.76/12 10} 17. Nd3 {+6.61/13 5} b3 {+7.43/12 9}
18. axb3 {+7.10/13 3} axb3 {+6.94/12 4} 19. Nxb3 {+7.51/12 4} Qa8
{+6.70/14 8} 20. Nhg5 {+7.87/12 5} h3 {+6.38/12 3} 21. Nxh3 {+8.17/13 9}
Kc8 {+5.81/12 5} 22. Nbd4 {+8.77/14 8} Qeh8 {+5.93/13 12} 23. b3
{+9.17/13 5} Qhd8 {+5.45/12 8} 24. Nfg3 {+8.83/13 5} Qa1+ {+5.37/11 4} 25.
Kf2 {+9.99/15 8} e5 {+4.24/12 3} 26. fxe5 {+11.34/16 9} dxe5 {+3.71/12 3}
27. Ndxe5 {+11.60/15 9} Qfa6 {+3.59/11 1.0} 28. Nhf4 {+12.07/14 12} Qc7
{+2.42/11 1.5} 29. Nfe6 {+12.37/14 5} f4 {-1.33/12 2.3} 30. N2xf4
{+13.05/14 6} Qe7 {-1.29/11 5} 31. Ngf5 {+16.25/14 6} Qb4 {-8.96/11 2.6}
32. Ndc6 {+19.65/13 5} Qba3 {-10.26/11 0.8} 33. N3c4 {+21.63/13 5} Qa7+
{-16.28/11 1.3} 34. N3d4 {+25.95/15 20} Kb7 {-17.53/12 1.3} 35. Nd5
{+26.71/14 9} Ka8 {-18.34/10 1.0} 36. b4 {+35.09/14 10} Qg1+ {-25.14/10}
37. Kxg1 {+319.91/17 5} Qc1+ {-99.78/13} 38. Kf2 {+319.92/17 16} Qxc2+
{-25.24/9} 39. Kg3 {+319.93/17 9}
*

For some reason Stockfish stopped playing here, and was flagged.

[hgm]

Another one, this time at longer time control (40 moves / 15 min). Again, Stockfish, playing the Queens, is crushed.

Note that Stockfish evaluates the initial position at +9.57, so we have the unique opportunity here to see Stockfish lose from a +9.57!

Code: Select all

[Event "Computer Chess Game"]
[Site "SCHAAK_PC"]
[Date "2010.03.21"]
[Round "-"]
[White "NewEng 0.16"]
[Black "Stockfish 1.6 JA"]
[Result "1-0"]
[TimeControl "40/900"]
[FEN "1q1qk1q1/pppppppp/8/8/8/8/PPPPPPPP/NNNNKNNN w - - 0 1"]
[SetUp "1"]

{--------------
. q . q k . q .
p p p p p p p p
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
P P P P P P P P
N N N N K N N N
white to play
--------------}
1. Nhg3 {+3.77/15 19} c5 {+9.57/16 29} 2. d3 {+3.87/15 23} d5 {+9.85/16 46}
3. e4 {+3.89/15 36} dxe4 {+10.02/15 52} 4. dxe4 {+4.11/14 14} h5
{+9.81/16 1:23} 5. Nbd2 {+4.46/14 23} g6 {+9.57/15 53} 6. Nab3
{+4.73/15 22} Qdc8 {+9.53/15 41} 7. c4 {+5.08/15 16} b6 {+9.17/15 40} 8.
Nc3 {+5.24/15 15} Qd7 {+8.96/16 1:03} 9. Ngf3 {+5.47/16 17} a5
{+8.80/16 41} 10. a4 {+5.52/15 12} f6 {+8.76/16 34} 11. h4 {+5.61/15 17}
Kd8 {+8.28/17 55} 12. Ne3 {+5.90/16 15} g5 {+8.48/16 24} 13. hxg5
{+6.31/16 14} fxg5 {+7.35/16 26} 14. Ned5 {+6.04/16 12} g4 {+6.98/16 45}
15. Nh4 {+5.92/17 24} Qdb7 {+6.78/15 21} 16. Nd3 {+5.99/17 23} Qg5
{+6.66/15 12} 17. Nhf5 {+5.74/17 23} e6 {+8.84/14 7} 18. N5f4 {+5.86/16 12}
Qf7 {+7.63/15 23} 19. Ne3 {+5.80/17 23} h4 {+8.40/13 10} 20. Nge2
{+5.98/16 15} e5 {+7.87/14 18} 21. Nfd5 {+6.37/16 28} h3 {+7.75/14 19} 22.
gxh3 {+6.87/16 13} gxh3 {+9.05/10} 23. Ndf1 {+7.18/16 13} Kc8 {+6.98/13 9}
24. Nh2 {+7.55/15 25} Qe6 {+7.03/14 5} 25. Nd2 {+7.75/15 26} Qg7
{+3.63/15 13} 26. Nf5 {+8.21/16 28} Qg2 {+5.37/14 9} 27. Ndf3 {+9.37/16 25}
Kb7 {+4.60/13 9} 28. Neg3 {+9.98/16 35} Qh8 {-0.36/15 16} 29. Nde3
{+10.69/15 22} Ka7 {-1.21/14 11} 30. Ke2 {+11.61/16 29} Qeg8 {-0.76/12 5}
31. Ndxe5 {+11.23/15 28} Qgh7 {-2.18/14 10} 32. Nhg4 {+11.72/16 22} Kb7
{-3.47/14 12} 33. Nxg2 {+12.76/16 17} hxg2 {-1.93/8} 34. Nd6+
{+12.40/15 42} Ka8 {-3.67/14 10} 35. Nd5 {+12.46/15 14} Q7g7 {-4.88/12 5}
36. Ke3 {+12.41/14 18} Qh1 {-1.49/13 3} 37. Ne2 {+12.35/15 17} Qb1
{-1.97/13 5} 38. Nxb6+ {+12.68/15 17} Ka7 {-2.38/13 3} 39. Nd5
{+12.67/16 50} Qxb2 {-1.25/11 2.8} 40. Nb5+ {+12.81/15 12} Ka8 {-1.45/12 5}
41. Ndf4 {+13.10/16 35} Qa1 {-1.77/16 45} 42. Nxg2 {+13.68/17 22} Qb7
{-4.24/17 57} 43. Nec3 {+14.49/17 17} Qc1+ {-7.23/17 1:23} 44. Nd2
{+14.65/17 13} Qg7 {-8.96/17 39} 45. Nd5 {+14.90/16 20} Qg5+ {-8.08/16 51}
46. Ngf4 {+14.94/16 36} Qa1 {-9.61/17 35} 47. f3 {+15.09/15 16} Qh4
{-8.68/16 1:09} 48. Nfd3 {+14.77/14 23} Qg1+ {-10.78/17 35} 49. Ke2
{+15.18/16 17} Qhh1 {-11.35/17 31} 50. Ngf2 {+15.38/16 13} Qhh2
{-14.22/17 58} 51. Nd7 {+17.88/15 24} Kb7 {-15.59/17 20} 52. e5
{+18.07/16 26} Kc8 {-16.94/15 13} 53. e6 {+20.27/16 25} Qg7 {-20.12/15 12}
54. Ne4 {+21.13/16 21} Qxd7 {-27.71/17 22} 55. Nb6+ {+23.27/16 17} Kd8
{-69.05/18 50} 56. Nxd7 {+23.52/17 13} Ke7 {-88.40/19 43} 57. N3xc5
{+28.41/20 26} Qh8 {-64.72/17 38} 58. Nec3 {+319.91/20 20} Kd8
{-99.78/13 6} 59. Nd5 {+319.92/20 12} Qb2+ {-99.80/11} 60. Ke1
{+319.93/20 19} Qb4+ {-99.84/11} 61. Nxb4 {+319.95/28 12} axb4 {-99.86/9}
62. Ng4 {+319.96/31 15} Ke7 {-99.94/9} 63. Nge5 {+319.97/100} Ke8
{-99.96/6} 64. Nc6 {+319.98/100} b3 {-99.98/6} 65. Nc7# {+319.99/100}
{White mates} 1-0