A Knight endgame

[Arpad Rusz]

Black to move in the following position. It is only a draw, or do you think Black wins this endgame?

[D]4n3/2p3k1/p1P5/p2P2Kp/P6P/5P1P/8/8 b - - 0 1

[Robert Flesher]

Black to move in the following position. It is only a draw, or do you think Black wins this endgame?

[D]4n3/2p3k1/p1P5/p2P2Kp/P6P/5P1P/8/8 b - - 0 1

Prior to any deep calculations, this looks like a draw or a win to me. I'll give it a looksee after work tommorrow. Thank-you for posting it!

[Arpad Rusz]

I have done some analysis: the following line is almost forced (and leads to draw):
1...Kf7 2.Kh6! Nd6 3.Kxh5 Nc4 4.Kg5 Nb2 5.d6! Nxa4 6.d7 Nc5 7.d8N+! Kf8 8.Nb7 Nb3 9.Nd6! Nd2 10.Nf5 a4 11.Nd4 a3 12.Nc2 a2 13.Kf6 Nb3 14.h5 a5 15.h6 Kg8 16.Ke7 Nd4 17.Na1 Nxc6+ 18.Kf6!=
Probably the unique way to save the game.

[Michel]

I have done some analysis: the following line is almost forced (and leads to draw):
1...Kf7 2.Kh6! Nd6 3.Kxh5 Nc4 4.Kg5 Nb2 5.d6! Nxa4 6.d7 Nc5 7.d8N+! Kf8 8.Nb7 Nb3 9.Nd6! Nd2 10.Nf5 a4 11.Nd4 a3 12.Nc2 a2 13.Kf6 Nb3 14.h5 a5 15.h6 Kg8 16.Ke7 Nd4 17.Na1 Nxc6+ 18.Kf6!=
Probably the unique way to save the game.

GnuChess prefers 3 ... Kf6

[Arpad Rusz]

A test position:
[D]8/2N2p1p/8/k7/8/7p/2P5/6K1 w - - 0 1

1.Nd5! is mate in 38 but I have seen no engine to show the win in a reasonable time (without using 6-men tablebases).
The win has a long 6-men phase, 5-men tablebases doesn't really helping here, so you can leave them on.

[Arpad Rusz]

GnuChess prefers 3 ... Kf6

I have analysed that move too: it also leads to draw.
3...Kf6 4.Kg4 Nc4 5.Kf4 Nb6 6.d6! cxd6 7.Ke4 Ke6 8.c7 Kd7 9.h5 Kxb7 10.Kf5! Nd7 11.Ke6 =

[Michel]

It seems you are right.... Very nice!


GnuChess plays 7 ... Nxa4 in your line and then needs a long time to see that this also leads to a draw.

[zullil]

A test position:
[D]8/2N2p1p/8/k7/8/7p/2P5/6K1 w - - 0 1

1.Nd5! is mate in 38 but I have seen no engine to show the win in a reasonable time (without using 6-men tablebases).
The win has a long 6-men phase, 5-men tablebases doesn't really helping here, so you can leave them on.

Stockfish-2.1.1 seems very happy here, after less than 30 minutes:

Code: Select all

info depth 42 seldepth 75 multipv 1 score cp 7683 nodes 7703378599 nps 5624554 time 1369598 pv c7d5 a5b5 d5e3 b5b4 e3d1 b4a5 g1h2 a5b4 h2h3 b4b5 h3h4 b5c5 h4g3 c5b6 g3f4 b6c6 f4e4 c6c5 e4d3 c5b6 d3d4 b6a5 c2c4 a5b6 d1f2 b6a5 c4c5 a5b5 f2d3 b5a5 d4d5 h7h5 c5c6 h5h4 d3f2 a5b4 d5d4 f7f5 c6c7 b4b3 c7c8q b3a2 c8f5