A pawn endgame

[Arpad Rusz]

[D]8/2pk4/8/4Kp1p/5P1P/6P1/8/8 w - - 0 1

White wins?

[Michel]

It seems like it, but the pv move looks a bit crazy though.

[lucasart]

It seems like it, but the pv move looks a bit crazy though.

Indeed! The correct moves are inhuman.

It seems that the (only?) winning move is Kf6!

Now:
1/ If black answers c6, the correct move is to go back to e5!
2/ If black plays c5, the correct move is Kxf5

Both options amount to the same.

[zullil]

[D]8/2pk4/8/4Kp1p/5P1P/6P1/8/8 w - - 0 1

White wins?

Stockfish seems to think so:

Code: Select all

info depth 73 seldepth 121 score cp 8461 nodes 65190107675 nps 11181767 time 5830036 multipv 1 pv e5f6 c7c6 f6e5 d7c7 e5f5 c6c5 f5g6 c5c4 f4f5 c4c3 f5f6 c3c2 f6f7 c2c1q f7f8q c1e1 f8c5 c7b7 c5b5 b7a7 b5g5 e1e2 g6h6 e2e8 g5g6 e8d8 h6h5 d8d2 g3g4 d2e3 g6f7 a7b8 f7g8 b8a7 g8g7 a7b8 g4g5 e3e4 g7g6 e4c4 g6d6 b8a8 d6e5 c4d3 e5e8 a8b7 e8f7 b7b8 f7f6 d3h7 f6h6 h7f7 h6g6 f7d7 g6b6 b8a8 b6f6 d7h7 f6h6 h7f7 h6g6 f7c4 g6f5 c4g8 h5h6 g8h8 f5h7 h8f8 h7g7 f8e8 h4h5 a8b8 g7h7 b8a8 h7g7

[Michel]

The fact that Kf6 is the best move is not just some silly opposition thing but it rather has to do with avoiding perpetual check in the resulting queen endgame. Looks very subtle from a human perpective.

[Eelco de Groot]

Here is Scid Serpent at depth 45:

8/2pk4/8/4Kp1p/5P1P/6P1/8/8 w - -

Engine: Scid Serpent 20140216_009 MP (Q6700, 4 threads, 512 MB)
by Tord Romstad, Marco Costalba and Joona Kiiski

46/79 11:02 +11.05 1.Kf6 c6 2.Ke5 c5 3.Kxf5 Kc7 4.Kg6 c4
5.f5 c3 6.f6 c2 7.f7 c1Q 8.f8Q Kb7
9.Kxh5 Qc2 10.g4 Kb6 11.Qf5 Qd1
12.Qf6+ Kc5 13.Kh6 Qe2 14.h5 (2.848.592.835) 4298

best move: Ke5-f6 time: 11:03.394 min n/s: 4.298.101 nodes: 2.848.592.835

Kxf5 draws, only Kxf6 wins:


8/2pk4/8/4Kp1p/5P1P/6P1/8/8 w - -

Engine: Scid Serpent 20140216_009 MP (512 MB)
by Tord Romstad, Marco Costalba and Joona Kiiski

42 3:12 +11.05 1.Kf6 c6 2.Ke5 c5 3.Kxf5 Kc7 4.Kg6 c4
5.f5 c3 6.f6 c2 7.f7 c1Q 8.f8Q Kb7
9.Kxh5 Qc2 10.g4 Kb6 11.Qf5 Qd1
12.Qf6+ Kc5 13.Kh6 Qe2 14.h5 (655.734.403) 3404

42 3:12 0.00 1.Kxf5 c5 2.Ke5 Kc6 3.f5 c4 4.f6 Kd7
5.Kd4 Ke6 6.Kxc4 Kxf6 7.Kd3 Ke5
8.Ke3 Kf5 9.Kf2 Ke6 10.Kf1 Kd5
11.Kg2 Kd4 12.Kg1 Ke5 13.Kh2 Kf5
14.Kg1 (655.734.403) 3404

Eelco

[peter]

Hi Arpad!

White wins?

Yes!

[Event "?"]
[Site "?"]
[Date "2014.03.10"]
[Round "?"]
[White "Arpad"]
[Black "Rusz"]
[Result "1-0"]
[SetUp "1"]
[FEN "8/2pk4/8/4Kp1p/5P1P/6P1/8/8 w - -"]

1. Kf6 (1. Kxf5 $2 c5 {R}) 1... c6 (1... c5 2. Kxf5 {#38})
2. Ke5 (2. Kxf5 $2 c5 {R}) 2... c5 3. Kxf5 {#38} 1-0

[tpetzke]

iCE gets it with depth 28 already, which is probably not a good sign as it means it prunes not enough. Anyway it takes more than 5 minutes and is beyond its normal search horizon.

Code: Select all

26/40 0:14   2.56  1. Kxf5 c5 2. Ke5 Kc6 3. f5 c4 4. f6 Kd7 5. Kd4 Ke6
                   6. Kxc4 Kxf6 7. Kd4 Kf5 8. Ke3 Ke5 9. Kf3 Kf5 10. Kf2 Kf6
                   11. Kg2 Kg6 12. Kh2 Kf5  (21.051.180)
27/43 0:23   2.56  1. Kxf5 c5 2. Ke5 Kc6 3. f5 c4 4. f6 Kd7 5. Kd4 Ke6
                   6. Kxc4 Kxf6 7. Kd4 Kf5 8. Ke3 Ke5 9. Kf3 Kf5 10. Kf2 Kf6
                   11. Kg2 Kg6 12. Kh2 Kf5 13. Kh3 Kg6 14. Kg2
                   (31.712.171)
28/50 4:58   5.12  1. Kf6 c5 2. Kxf5 c4 3. Ke4 Kc6 4. Kd4  (415.193.507)