A balanced approach to imbalances

[hgm]

Oh, sorry, you said more than two. Well, if you are willing to consider Stockfish at 1 min per move a different engine from Stockfish at 1.5 sec/move, this is point number 3. That is 3 out of 3, so far.

Code: Select all

[Event "Computer Chess Game"]
[Site "MAKRO-PC"]
[Date "2013.10.24"]
[Round "-"]
[White "Stockfish 4 64 SSE4.2"]
[Black "QueeNy 0.16"]
[Result "0-1"]
[TimeControl "40/2400:40/240"]
[FEN "6k1/ppp2pp1/1nnnnn1p/8/8/7P/PPP2PP1/3QQ1K1 w - - 0 1"]
[SetUp "1"]

{--------------
. . . . . . k .
p p p . . p p .
. n n n n n . p
. . . . . . . .
. . . . . . . .
. . . . . . . P
P P P . . P P .
. . . Q Q . K .
white to play
--------------}
1. Qc1 {+3.05/25} Nbd5 {+5.98/14 5} 2. Qf1 {+3.07/28 1:14} Ne5 {+6.02/14 7}
3. Qce1 {+3.05/30 1:04} Nfd7 {+5.96/15 3} 4. Qa5 {+3.07/29 45} N7b6
{+5.93/15 4} 5. Qd2 {+3.01/30 2:16} c5 {+6.19/13 5} 6. Qdd1 {+3.01/23 1:03}
Nd4 {+6.32/13 4} 7. Qfe1 {+2.92/26 1:07} f6 {+5.83/14 4} 8. Qa5
{+3.35/26 56} Ned7 {+5.83/14 3} 9. b3 {+3.43/24 1:06} Ne4 {+5.77/14 5} 10.
Qxa7 {+4.16/27 1:13} Ndc3 {+5.53/15 6} 11. Qd3 {+4.16/29 1:50} Nd6
{+5.31/14 3} 12. h4 {+4.30/29 2:05} f5 {+5.73/13 4} 13. f3 {+3.53/24 1:09}
Kh7 {+5.81/14 6} 14. Kh1 {+3.27/27 1:54} Nbd5 {+6.00/15 4} 15. Qf1
{+2.82/27 1:40} Nxc2 {+7.00/15 4} 16. Qa8 {+2.60/27 55} N5b6 {+6.98/14 5}
17. Qa5 {+2.94/27 36} Ncd5 {+6.76/16 7} 18. Qf2 {+2.66/27 2:30} Nce3
{+6.79/17 7} 19. h5 {+2.94/26 41} f4 {+6.90/16 7} 20. Qg1 {+2.88/28 35} Kg8
{+6.85/15 4} 21. Qb1 {+2.80/28 38} Ne7 {+6.91/16 7} 22. Qd3 {+2.76/28 34}
N7f5 {+6.93/15 5} 23. a4 {+2.90/28 1:01} Kf7 {+7.38/15 5} 24. Qa7
{+2.52/25 43} Nbd5 {+7.36/15 7} 25. Kh2 {+2.34/27 49} Kf6 {+7.45/15 6} 26.
Qa8 {+2.24/27 43} Ne5 {+7.51/15 7} 27. Qd8+ {+2.24/26 48} Kf7 {+7.46/16 6}
28. Qb1 {+2.24/27 28} Ke6 {+7.44/15 7} 29. Qa1 {+3.23/25 1:29} Nc6
{+7.44/14 4} 30. Qg8+ {+2.86/26 33} Kd7 {+7.37/16 6} 31. a5 {+2.78/28 28}
Ncb4 {+7.32/14 4} 32. Kg1 {+2.64/26 2:17} Nf6 {+7.41/15 7} 33. Qf8
{+2.68/25 22} Ned5 {+7.35/14 6} 34. Qe5 {+2.44/23 24} Nd3 {+7.55/16 4} 35.
Qe2 {+2.30/24 7} N5b4 {+7.56/17 6} 36. Qd2 {+2.04/26 37} Nxh5 {+8.08/17 8}
37. Qd1 {+1.79/31 54} Nf6 {+8.04/16 8} 38. Qa1 {+1.79/34 32} Nfd5
{+8.08/16 13} 39. Qa4+ {+1.49/30 29} Ke6 {+8.03/16 7} 40. Kh2 {+1.49/31 32}
Nc3 {+8.11/16 11} 41. Qa1 {+1.43/34 1:11} Ncb5 {+8.08/16 6} 42. Kg1
{+1.39/33 1:22} Nbd4 {+8.15/15 4} 43. Kh1 {+1.19/33 1:55} Ndc2 {+8.16/14 4}
44. Qb1 {+1.13/36 2:18} Nce3 {+8.18/15 6} 45. Kg1 {+1.13/36 1:17} Nbc2
{+8.32/16 12} 46. Qa2 {+0.00/27 4:27} Nc1 {+8.44/15 6} 47. Qb1
{+0.00/30 59} Ncd4 {+8.88/14 4} 48. Qg8+ {+0.00/27 59} Ke7 {+8.92/15 7} 49.
a6 {-1.61/26 2:46} bxa6 {+9.06/14 3} 50. Qa1 {-3.09/26 1:36} Ncxb3
{+9.15/15 6} 51. Qxa6 {-5.25/26 1:19} c4 {+9.15/15 3} 52. Qb8
{-6.02/27 1:24} Kd7 {+9.00/14 7} 53. Qg8 {-1.45/28 55} Nc6 {+9.57/15 6} 54.
Kh1 {-6.30/27 41} Nbd4 {+10.17/16 5} 55. Qa1 {-6.54/29 39} N4b5
{+10.26/15 3} 56. Qa2 {-6.70/30 1:25} c3 {+11.91/16 5} 57. Qge6+
{-7.95/30 1:34} Kc7 {+11.92/17 5} 58. Qeb3 {-10.18/31 3:53} Nfd4
{+12.12/16 5} 59. Qg8 {-7.79/26 29} c2 {+12.94/16 6} 60. Qb2
{-9.21/29 1:09} Kb6 {+14.75/15 3} 61. Qf8 {-9.19/28 23} Ndc4 {+15.60/16 4}
62. Qxc2 {-10.38/29 41} Ndxc2 {+15.87/18 7} 63. Qxf4 {-11.03/28 49} Nbd4
{+16.03/18 7} 64. Qg3 {-11.61/28 20} Ndf5 {+17.15/18 4} 65. Qf2
{-14.54/29 33} Nb2 {+18.04/19 6} 66. g4 {-89.04/25 21} Nd3 {+18.29/19 4}
67. Qh2 {-102.87/30 33} Nfd4 {+19.47/19 4} 68. g5 {-149.48/33 1:34} hxg5
{+19.56/18 5} 69. Qg1 {-103.03/23 58} g4 {+22.26/20 5} 70. fxg4
{-99.70/21 25} Nf3 {+22.34/21 6} 71. Qg3 {-99.84/21 5} Nce1 {+22.61/22 8}
72. g5 {-99.86/28 6} Nf5 {+319.92/23 7} 73. Qxf3 {-99.88/37 5} Nxf3
{+319.94/25 5} 74. g6 {-99.90/76 6} Nce5 {+319.95/26 5} 75. Kg2
{-99.92/83 5} N3h4+ {+319.96/26 5} 76. Kh1 {-99.94/100 0.2} Kb7
{+319.97/31 7} 77. Kg1 {-99.96/100 0.2} Nef3+ {+319.98/33 9} 78. Kf1
{-99.98/100 0.2} Ng3# {+319.99/36 7}
{Xboard adjudication: Checkmate} 0-1

With the Queens QueeNy of course loses very easily: it just keeps trading Queens for two Knights until it has only Pawns against Knight + Pawns. The Knight value was tuned to win with 7 Knights vs 3 Queens, or with 3 Queens against 6 Knights. It was not tuned to win with 2 Queens against 5 Knights. (Which should only be possible if the opponent is very stupid anyway, and would definitely require an eval term that discourages Q for 2N when only a single Q and 3N are left.)

Code: Select all

[Event "Computer Chess Game"]
[Site "MAKRO-PC"]
[Date "2013.10.24"]
[Round "-"]
[White "QueeNy 0.16"]
[Black "Stockfish 4 64 SSE4.2"]
[Result "0-1"]
[TimeControl "40/300"]
[FEN "6k1/ppp2pp1/1nnnnn1p/8/8/7P/PPP2PP1/3QQ1K1 w - - 0 1"]
[SetUp "1"]

{--------------
. . . . . . k .
p p p . . p p .
. n n n n n . p
. . . . . . . .
. . . . . . . .
. . . . . . . P
P P P . . P P .
. . . Q Q . K .
white to play
--------------}
1. b3 {-5.92/14} Nbd5 {-2.48/20 7} 2. f3 {-5.88/15 8} Nf5 {-1.67/21 5} 3.
c3 {-5.79/14 5} Nef4 {-0.66/22 7} 4. Qdd2 {-5.62/15 5} Nfe3 {-0.74/22 4} 5.
Qexe3 {-5.13/16 5} Nxe3 {+0.26/19 2.3} 6. Qxe3 {-5.24/18 7} Ne6
{+0.26/24 13} 7. b4 {-5.17/16 5} a6 {+0.26/23 7} 8. g3 {-5.19/16 4} Ne8
{+0.38/22 6} 9. a4 {-5.08/16 5} Nd6 {+0.52/22 7} 10. Kf2 {-5.31/17 8} Kf8
{+0.56/22 6} 11. Qe2 {-5.27/17 8} a5 {+0.56/27 6} 12. b5 {-5.37/18 4} Nb8
{+1.35/23 8} 13. Qe3 {-5.69/18 5} Nd7 {+1.55/26 9} 14. Qa7 {-5.68/20 6} b6
{+1.53/27 18} 15. Qa8+ {-6.02/21 6} Ke7 {+1.53/27 9} 16. Qd5 {-6.04/20 9}
Ndc5 {+1.53/27 9} 17. Qa2 {-6.12/19 4} Nd3+ {+1.45/27 9} 18. Kg1
{-6.05/17 6} Ne5 {+1.49/27 6} 19. f4 {-5.87/18 6} Nec4 {+1.83/22 5} 20. Qe2
{-6.02/18 5} Kf8 {+1.93/27 6} 21. Kg2 {-6.00/18 4} Nc5 {+2.40/23 7} 22. Qa2
{-6.35/20 6} Nce4 {+2.48/25 8} 23. Qa1 {-6.51/19 5} Ne3+ {+2.78/21 6} 24.
Kg1 {-7.00/20 14} Nxg3 {+3.69/23 10} 25. Kf2 {-6.94/19 7} Nef5 {+3.85/22 7}
26. Kf3 {-7.47/18 5} Nge4 {+4.06/23 8} 27. Qe1 {-7.59/19 5} Ne7
{+4.06/22 6} 28. Qxe4 {-7.38/21 5} Nxe4 {+9.23/21 2.6} 29. Kxe4
{-7.99/21 13} Ke8 {+9.67/22 2.5} 30. Ke5 {-8.62/21 8} Ng6+ {+13.45/21 2.5}
31. Ke4 {-8.26/20 9} Nf8 {+14.62/22 2.1} 32. c4 {-8.53/19 6} Ne6
{+17.55/23 5} 33. Ke5 {-9.49/21 14} Nc5 {+20.80/23 4} 34. Kd4 {-9.78/19 11}
Nxa4 {+22.84/22 3} 35. f5 {-11.17/20 18} Nc5 {+29.65/22 3} 36. Ke5
{-12.39/19 9} a4 {+98.53/22 8} 37. Kd4 {-14.27/20 9} Ke7 {+146.75/24 32}
38. Kc3 {-14.36/19 6} Kf6 {+99.71/22 10} 39. h4 {-16.48/19 5} Kxf5
{+99.73/20 3} 40. Kd4 {-17.59/19 8} a3 {+99.79/19 8} 41. Kc3 {-19.80/19 9}
Ke5 {+99.81/19 4} 42. Kc2 {-20.50/20 6} f5 {+99.83/20 4} 43. Kb1
{-21.75/19 5} f4 {+99.85/23 3} 44. Ka2 {-319.93/20 5} f3 {+99.89/63 28} 45.
Kxa3 {-319.94/23 5} f2 {+99.91/76 1:13} 46. Ka2 {-319.95/25 5} f1=Q
{+99.93/100 0.2} 47. Kb2 {-319.97/33 5} Qe2+ {+99.95/100 0.2} 48. Kb1
{-319.98/100 0.2} Nd3 {+99.97/100 0.2} 49. h5 {-319.99/100 0.1} Qb2#
{+99.99/100 0.2}
{Xboard adjudication: Checkmate} 0-1

So QueeNy is probably the worst engine to play the Queens in this position. Not only is it tactically some 600 Elo weaker than its opponent, it has a totally wrong evaluation for it as well.

But in any case, the Knights won again! By now you should notice a pattern in this...

[rjgibert]

[d]6k1/ppp2pp1/1nnnnn1p/8/8/7P/PPP2PP1/3QQ1K1 w - - 0 1

Somehow Black has managed to promote 2 pawns into 3 knights. What a cheater!

But more "seriously," if you want to create a "realistic" position, the knights should be located on the White side of the board where they were promoted rather than on the Black side of the board.

[hgm]

I am afraid that 'realistic' is a concept that cannot really be applied, no matter what. If you assume that this position is the result of a FIDE game. But there is of course no need to assume that. They are more likely just early middle-game positions from the game tree of the Chess variant 'Charge of the Light Brigade'.

I was supprised to find that Fruit and Stockfish seemed to handle these positions without problems. I remember that many engines would simply crash, or ignore some of the pieces when you loaded them with positions where Knights + Pawns > 10.

The matter of the elephantiasis effect is not entirely theoretical, however. It is for instance quite heavily felt in Capablanca Chess, where each side has 3 super-pieces with a value of approximately a Queen in the opening array. There it explains why trading a Queen for a Rook + Bishop is favorable for the side losing the Queen when the other two super-pieces (Chancellor and Archbishop) are still present, while it is a very bad trade when they are not (as in orthodox Chess, where you would need at least an extra Pawn, and preferably two next to the R + B to break even). By removing R+B by the Queen sac your remaining C and A will suffer les depreciation in compensation, while removal of the Queen does nothing for the effective value of the opponent C and A. (On the contrary; the C and A devaluated the Queens, as they are enough less valuable (about 0.5 Pawn) to want to avoid trading them for the latter, and yours now still devaluate the opponent's Queen, while you have handomely gotten rid of your own devaluated Queen!)

[hgm]

No one can convince me that this position is not won for white: it simply is, but maybe I should handle the white side myself.

Well, I wish you good luck. You know where to find the engine. This time I am willing to make a bet: You won't score a single half-point out of 10 games against QueeNy, when you allow the latter 10 sec/move (arbitrary time for you).

2 queens worse than 5 knights: it absolutely makes no sense, it is almost the same as 2 queens against 4 normal minor pieces and a little bit more, which is obvious is a big advantage to the queen side.

It makes no sense to you. In the framework of the elephantiasis effect it makes of course perfect sense. Your intuition about these unnatural ('fairy') positions seems completely wrong, and it seems everything you say is only founded in intuition. My claims, on the other hand, are based on methodical analysis with engines, where I played many different positions hundreds of times with various engines, and even went so far as to adapt an engine to make it follow what proved to be the best strategy (in the 7-vs-3 and 6-vs-3 case). Intuition sucks...

But maybe the answer is in the strength of Queeny: is it really stronger than Stockfish and much stronger than Fruit? This might explain things.

Tactically it is highly inferior: the engine that it was derived of (before I retuned its evaluation to make it unsuitable for playing normal games) would not even score 1% against Fruit or Stockfish.

What is even harder proof that this cannot be the explanation is that QueeNy with the Knights can beat itself, even when the Queens get 10 times as much thinking time. Surely the engine that can thing 10 times longer, and as a consequence searches 2-3 ply deeper, must have the advantage. The reason it loses anyway is that the position is hopeless.

Whether the strategy it uses for playing the Queens is good or not is a bit of a moot point, btw. In a badly lost position there just isn't such thing as a good strategy. When you are playing Bishops in KBBKRRR, is it 'good' to win the exchange? Even if you do it twice, you are left in KKR, a certain loss. It might delay the mate, because KRK takes many moves, but that in no way reduces the certainty of the loss. And it might even speed up the loss, sometimes that is difficult to predict. Does that mean a Bishop is worth more than a Rook? I don't think so.

As I do not run engine-engine games myself nowadays, and can not check your results (but I do not doubt you are earnest), probably the only way to make things straight would be to win a couple of games in a row against Queeny.

Yes, that woud be the day!

I won't hold my breath, though! And we did not even get to the point that it really was about: 7 Knights versus 3 Queens. You could try this one:

[d]3nkn2/2pnpnp1/3nnn2/8/8/8/3PPP2/1Q1QK1Q1 w - - 0 1
Very bad for black! All its pawns are isolated, while white has a nice King shield of connected Pawns!

[Lyudmil Tsvetkov]

No one can convince me that this position is not won for white: it simply is, but maybe I should handle the white side myself.

Well, I wish you good luck. You know where to find the engine. This time I am willing to make a bet: You won't score a single half-point out of 10 games against QueeNy, when you allow the latter 10 sec/move (arbitrary time for you).

2 queens worse than 5 knights: it absolutely makes no sense, it is almost the same as 2 queens against 4 normal minor pieces and a little bit more, which is obvious is a big advantage to the queen side.

It makes no sense to you. In the framework of the elephantiasis effect it makes of course perfect sense. Your intuition about these unnatural ('fairy') positions seems completely wrong, and it seems everything you say is only founded in intuition. My claims, on the other hand, are based on methodical analysis with engines, where I played many different positions hundreds of times with various engines, and even went so far as to adapt an engine to make it follow what proved to be the best strategy (in the 7-vs-3 and 6-vs-3 case). Intuition sucks...

But maybe the answer is in the strength of Queeny: is it really stronger than Stockfish and much stronger than Fruit? This might explain things.

Tactically it is highly inferior: the engine that it was derived of (before I retuned its evaluation to make it unsuitable for playing normal games) would not even score 1% against Fruit or Stockfish.

What is even harder proof that this cannot be the explanation is that QueeNy with the Knights can beat itself, even when the Queens get 10 times as much thinking time. Surely the engine that can thing 10 times longer, and as a consequence searches 2-3 ply deeper, must have the advantage. The reason it loses anyway is that the position is hopeless.

Whether the strategy it uses for playing the Queens is good or not is a bit of a moot point, btw. In a badly lost position there just isn't such thing as a good strategy. When you are playing Bishops in KBBKRRR, is it 'good' to win the exchange? Even if you do it twice, you are left in KKR, a certain loss. It might delay the mate, because KRK takes many moves, but that in no way reduces the certainty of the loss. And it might even speed up the loss, sometimes that is difficult to predict. Does that mean a Bishop is worth more than a Rook? I don't think so.

As I do not run engine-engine games myself nowadays, and can not check your results (but I do not doubt you are earnest), probably the only way to make things straight would be to win a couple of games in a row against Queeny.

Yes, that woud be the day!

I won't hold my breath, though! And we did not even get to the point that it really was about: 7 Knights versus 3 Queens. You could try this one:

[d]3nkn2/2pnpnp1/3nnn2/8/8/8/3PPP2/1Q1QK1Q1 w - - 0 1
Very bad for black! All its pawns are isolated, while white has a nice King shield of connected Pawns!

Intuition does not suck, actually intuition, judging by the Latin word it is derived from, means insight, careful concentration, higher knowledge. People who have intuition by definition know more than people who do not have it, because intuition is based on fact and is the product of either genetic, naturally conditioned factors (like astrological influences), or hard labour.

I think the engines simply play bad with such an unnatural imbalance for which they have no knowledge. I analysed a bit the position with the 2 queens and 5 knights (where the knights are still excellently placed), and did not in any way find a winning continuation for black, I do not know if white could win. My main line is as follows:
1. b3 Nbd5 2.g3 a5 3. c4 Ne7 4.Qe3 Nef5 5. Qc3 Ned4 6.a3 Nde4 7.Qe1 g5 8.b4 ab4 9.ab4 b5 10.cb5 Nb5 11.Qc2 Ncd4 12.Qce4 Ne4 13.Qe4 Kg7 14.Kf1, and you tell me what happens here, I do not think black can win.

[d]8/2p2pk1/7p/1n3np1/1P1nQ3/6PP/5P2/5K2 b - - 0 14

But again, at the beginning of the game the black knights were excellently placed in distinction to the white queens.

If there is some point about the elephantiasis effect, it is that the extremely large number of pieces of lower power (absolutely unrealistic with any game situation) by definition defend each other well, so that it is difficult to attack them, while they can attack the enemy while being defended. But this is all about defending bonus points, nothing more. For example 7N vs 3 Q is better for the knights than 5 knights vs 2Q, for the simple reason that the knights defend each other better by definition when in a large quantity. Similarly, 5N vs 2Q is better for the knights than 3N vs 1Q, as you can judge by the positions and the necessity of the queen side to exchange. With a low number of minors, the rule already does not hold true, as the minors already do not defend each other, although theoretically they should still be able to interdict queen activity. It is as simple as that, it is all about mutual defence.

Regarding the new position you posted, I think the black king shelter, with 7 knights in it, is much safer than the white one, so black is better even in terms of king safety. But I will again say it is a draw at most.

[Lyudmil Tsvetkov]

And let us have a closer look at those 3 positions featuring the imbalance of Qs vs Ns:

[d]8/2p2pk1/7p/1n3np1/1P1nQ3/6PP/5P2/5K2 b - - 0 14

Here I count 10 mobile squares for the Q and 8 for the Ns, with both sides almost optimally placed, i.e. equally

[d]6k1/ppp2pp1/1nnnnn1p/8/8/7P/PPP2PP1/3QQ1K1 w - - 0 1
Here I count 12 mobile squares for the Qs, and 26 for the Ns, with the knights placed much more centrally

[d]2n2nk1/n1p1nnp1/1n3p1n/8/1P1PQ3/1Q3P2/3Q4/6K1 w - - 0 1
Here I count 32 mobile squares for the Qs, and 15 for the Ns, this time with the queens placed much more centrally.

Obviously, mobility with such piece configuration is not a function of the number of pieces of lower power, but rather, as usual, of the centralisation of pieces. I.e., the theory that pieces of lower power interdict the access to squares of stronger pieces, is simply not true. However, the fact that in all cases the knights are extremely well defended, is apparent. There is no discussion about this, it is obvious. In games chess engines might make mistakes, but it is difficult to make such when counting squares. So that elephantiasis, to the point where it is valid for situations with extremely large amount of pieces, is due to mutual piece defence, and not interdiction of squares to enemy pieces.

[Uri Blass]

The question is if 7 knights are strong or maybe 3 queens are weak.
I wonder if white is winning in the following positions when white has no queens and computers evaluates white as better based on common material values.

[d]nnnnknnn/pppppppp/8/8/8/8/PPPPPPPP/RNB1KBNR w - - 1 1

[d]nnnnknnn/pppppppp/8/8/8/8/PPPPPPPP/RR2KRRR w - - 1 1

[lucasart]

I modified the starting position:
* turn all black pieces into knights
* changed white rooks into queens and removed its minor pieces

It was quite a fascinating game to watch:
[pgn]
[White "Discocheck"]
[Black "Discocheck"]
[Result "1/2-1/2"]
[FEN "nnnnknnn/pppppppp/8/8/8/8/PPPPPPPP/Q2QK2Q w - - 0 1"]

1. b3 Nfe6 2. g4 Kf8 3. h4 Nab6 4. a4 c5 5. a5 Na8 6. f4 Nbc6 7. f5 Nec7 8.
e4 e6 9. Qg1 d6 10. c3 Nc8e7 11. Qf3 exf5 12. gxf5 Nf6 13. d4 Nce8 14. Kd1
cxd4 15. cxd4 Nac7 16. d5 Ne5 17. Qe2 a6 18. Kc2 Nfg4 19. Kc1 f6 20. Kb2 g6
21. fxg6 hxg6 22. h5 gxh5 23. Qh1 f5 24. Qxh5 Nhf7 25. exf5 Nef6 26. Qhh1
Nf2 27. Qh4 Nfd3+ 28. Kb1 Nexd5 29. Qg3 Nb5 30. Qexd3 Nxd3 31. Qxd3 Ne5 32.
Qh3 Ndc6 33. Qb2 Nxa5 34. Qh8+ Ke7 35. Qc8 Ne4 36. Qh2 Nbc3+ 37. Ka1 Nxb3+
38. Kb2 Nbc5 39. Qh7+ Nf7 40. Qhg8 Kf6 41. Qce8 Kxf5 42. Qgxf7+ Ndf6 43.
Qg6+ Kf4 44. Qh6+ Kf3 45. Qeg6 Nd1+ 46. Kb1 Ndf2 47. Qc1 Ncd3 48. Qc7 Ndc5
49. Qe7 Kf4 50. Qh6+ Kf5 51. Ka1 b5 52. Qe3 Nfd3 53. Qh3+ Kg5 54. Qf7 Kf4
55. Qe7 Kg5 56. Qg2+ Kf5 57. Qf1+ Kg6 58. Qg2+ Kf5 59. Qf1+ Kg6 60. Qg1+
Kf5 61. Qh2 b4 62. Qh3+ Kg5 63. Qf7 b3 64. Qe3+ Kf5 65. Qc4 a5 66. Qh3+ Kg6
67. Qg2+ Kf5 68. Qf1+ Ke5 69. Qe2 a4 70. Kb1 a3 71. Qh2+ Kf5 72. Qh3+ Kg6
73. Qg2+ Kf5 74. Qf3+ Kg6 75. Qg2+ Kf5 76. Qf3+ Kg6 77. Qcxd3 Nxd3 78. Qxd3
Kg5 79. Ka1 Nc5 80. Qxd6 Nfe4 81. Qb8 Kg4 82. Qb4 a2 83. Qe1 Kf4 84. Qh4+
Ke3 85. Qh3+ Kd4 86. Qh8+ Kc4 87. Qh4 Kd4 88. Qh8+ Kd3 89. Qb2 Ke3 90. Qh2
Nc3 91. Qe5+ Kd3 92. Qf5+ N3e4 93. Qe5 Nc3 94. Qg3+ Kc4 95. Qh4+ N3e4 96.
Qg4 Kc3 97. Qg2 Kc4 98. Qe2+ Kd4 99. Qb2+ Nc3 100. Qf2+ Kd3 101. Qf1+ Kd4
102. Qf7 Nd5 103. Qg7+ Ke4 104. Qg6+ Kd4 105. Qg4+ Kd3 106. Qg6+ Kc4 107.
Qg4+ Kd3 108. Qd1+ Kc4 109. Qg4+ Kb5 110. Qe2+ Kb4 111. Qe8 Nd3 112. Qe4+
Kc3 113. Qg4 Nc5 114. Qg8 Nd3 115. Qg4 Nb6 116. Qg7+ Kc4 117. Qc7+ Kb5 118.
Qf7 Nc5 119. Qe8+ Kc4 120. Qe2+ Kd4 121. Qg4+ Kd5 122. Qg8+ Kc6 123. Qg5
Kb5 124. Qg4 Nd5 125. Qh5 Nc3 126. Qh4 N3e4 127. Qd8 Kc4 128. Qe8 Nc3 129.
Qg8+ Kd4 130. Kb2 Nd3+ 131. Kxb3 a1=Q 132. Qc4+ Ke3 133. Qxc3 Qa8 134. Kc4
Qa4+ 135. Kd5 Qb5+ 136. Kd6 Qb8+ 137. Ke7 Qg8 138. Kd7 Qb8 139. Ke7 Qb6
140. Kd7 Qb7+ 141. Kd6 Qa6+ 142. Kc7 Qa7+ 143. Kd6 Qb7 144. Ke6 Qg2 145.
Kd6 Qg3+ 146. Kd7 Qf4 147. Ke7 Qe4+ 148. Kd6 Qf4+ 149. Kd7 Qa4+ 150. Kc7
Qf4+ 151. Kd7 Qh2 152. Ke7 Qh6 153. Qf6 Qh7+ 154. Kd6 Qh2+ 155. Kc6 Qc2+
156. Kb6 Qc5+ 157. Kb7 Qd5+ 158. Qc6 Qe5 159. Qc4 Nf4 160. Qb3+ Ke4 161.
Qc2+ Kf3 162. Qb3+ Ke2 163. Qc2+ Kf3 164. Qb3+ Kg4 165. Kc6 Qe8+ 166. Kc5
Kg5 167. Qc4 Qd7 168. Kb4 Nd5+ 169. Kb3 Qh3+ 170. Kb2 Qh8+ 171. Ka2 Qe5
172. Qc1+ Ne3 173. Qg1+ Kf5 174. Qf2+ Kg4 175. Qe2+ Kg5 176. Qb2 Qc7 177.
Qb5+ Nf5 178. Qb2 Kh4 179. Qb3 Nd6 180. Qb2 Nf5 181. Qb3 Nd6 182. Qb2 Nf5
1/2-1/2
[/pgn]
Although it ended in a draw, it is clear that black had the better side of it. The problem however, is that in order to win, black must find a way to shield his king from constant perpetual check threats, while pushing his passed pawns to promotion.

I don't know if black could have won this, and which move was the mistake that lead to a draw instead of a win, if there is any such single mistake here.

[Ryan Benitez]

Black should be able to win this. The power of having many pieces and weakness of having only major pieces is that value of material you are willing to lose in an exchange. When 3 minors come swarming for your king you trade anything you have have to stop the attack. At l least with a rook on board you could force a rook for 2 minor exchange. Another factor is how likely your pieces are to helping a pawn promote. A passer should never be valued higher than the weakest piece that can be exchanged for it unless that piece is clearly out of range to do so.

[Ralph Stoesser]

Black should be able to win this. The power of having many pieces and weakness of having only major pieces is that value of material you are willing to lose in an exchange. When 3 minors come swarming for your king you trade anything you have have to stop the attack. At l least with a rook on board you could force a rook for 2 minor exchange.
...

I don't know how Fruit handles king safety, but Stockfish's evaluation can't see the king from the side with the queens ever being in danger, because it doesn't evaluate king attacking chances for a side that doesn't has a queen.
I'm not sure if that's a big part of the problem why SF grossly misevaluates those positions though. It might need to calculate knight swarm attacking chances to be able to compensate for the material imbalance.