Initial position - 27 pieces - Mate in 22 moves?

[fastgm]

I was interested in the following position. The initial position with five pieces less for black. How many moves does white need to checkmate?

[d]2b1kb2/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1
To solve the position, I used Stockfish 11. In my experience, the most exact engine when it comes to mate problems.
After 60 hours on a 10-core with 32 GB hash, Stockfish found a checkmate in 22 moves.

Stockfish 11 64 POPCNT:
96/44 60:29:31 4.669.383.221.790 21.441.650 +M22 1.e4 g6 2.Nc3 c6 3.Qe2 h5 4.Qe3 b6 5.Qd4 f6 6.Qa4 a6 7.Qd4 b5 8.Qb6 d6 9.Qxc6+ Bd7 10.Qxa6 Kf7 11.Nd5 Be6 12.Nc7 Bd7 13.Bxb5 Bg4 14.Qa8 h4 15.h3 Bc8 16.Qxc8 f5 17.Qe8+ Kg7 18.Ne6+ Kf6 19.Nxf8 fxe4 20.Qxg6+ Ke5 21.Qg5+ Kd4 22.c3+

and all black moves after 1.e4:
62/31 2:21:16 158.317.142.203 18.676.162 -M15 1. ... f5 2.Qh5+ g6 3.Qxh7 d6 4.Qxg6+ Kd8 5.d4 f4 6.Bxf4 a6 7.Bg5 Bf5 8.Qf7 Bxe4 9.Qxf8+ Kd7 10.Qxe7+ Kc6 11.Qxe4+ Kb6 12.Bd8 a5 13.Qe8 a4 14.Qb5+ Ka7 15.Bxc7 Ka8 16.Qxa4+
62/33 2:21:16 158.317.142.203 18.676.162 -M16 1. ... f6 2.Qh5+ g6 3.Qxh7 d6 4.Qxg6+ Kd7 5.Nc3 c6 6.Nf3 a6 7.Nd4 Bh6 8.Qxh6 f5 9.Qg7 fxe4 10.Nf5 Kc7 11.Qxe7+ Bd7 12.Qxd6+ Kb6 13.Qb8 Bxf5 14.Na4+ Ka5 15.Qd8+ Kxa4 16.b3+ Kb4 17.Qb6+
62/33 2:21:16 158.317.142.203 18.676.162 -M16 1. ... h5 2.Qxh5 g6 3.Qh7 d5 4.exd5 Bg4 5.Bb5+ c6 6.dxc6 bxc6 7.Bxc6+ Kd8 8.Qxf7 Kc7 9.Qxf8 Kxc6 10.Qxe7 Bd7 11.Nf3 Bc8 12.Ne5+ Kb6 13.Nc3 Bd7 14.Qxd7 a6 15.Nc4+ Kc5 16.d4+ Kxc4 17.Qa4+
62/39 2:21:16 158.317.142.203 18.676.162 -M16 1. ... Kd8 2.Qh5 Ke8 3.Qxh7 d6 4.Nf3 Bg4 5.Ng5 Kd7 6.Qg8 Bh5 7.Qxf8 a6 8.Nxf7 Bxf7 9.Qxf7 Kd8 10.d4 Kd7 11.Nc3 c6 12.Bg5 Kc7 13.Qxe7+ Kb6 14.Qxd6 a5 15.Qb8 c5 16.Bd8+ Kc6 17.Qc7+
62/35 2:21:16 158.317.142.203 18.676.162 -M17 1. ... a5 2.Qh5 a4 3.Qxh7 a3 4.Nxa3 d5 5.exd5 Bg4 6.Bb5+ c6 7.dxc6 bxc6 8.Bxc6+ Bd7 9.Qf5 e6 10.Bxd7+ Ke7 11.Qc5+ Kxd7 12.Qxf8 f6 13.Qxg7+ Kd6 14.Nb5+ Kc6 15.Qc7+ Kxb5 16.Qd6 Ka4 17.Qc5 e5 18.b3+
62/37 2:21:16 158.317.142.203 18.676.162 -M18 1. ... c5 2.Qh5 c4 3.Qc5 Kd8 4.Qa5+ b6 5.Qxa7 e6 6.Qxb6+ Ke8 7.Bxc4 g5 8.Nc3 Be7 9.Qb8 Kd8 10.Na4 h6 11.Nb6 Bc5 12.Qxc8+ Ke7 13.Qxc5+ d6 14.Nc8+ Kf6 15.Nxd6 g4 16.Qc7 h5 17.Qxf7+ Ke5 18.Qxe6+ Kd4 19.Qd5+
62/37 2:21:16 158.317.142.203 18.676.162 -M18 1. ... b5 2.Qh5 c6 3.Qxh7 g5 4.Nf3 f6 5.Qg6+ Kd8 6.Qg8 Ke8 7.e5 a5 8.Bd3 f5 9.Nxg5 f4 10.Qf7+ Kd8 11.Qxf8+ Kc7 12.Qxe7 Kb7 13.Qd8 f3 14.Nf7 fxg2 15.Nd6+ Ka6 16.Rg1 a4 17.Qc7 Bb7 18.b4 axb3 19.axb3+
62/39 2:21:16 158.317.142.203 18.676.162 -M19 1. ... e5 2.Qh5 c6 3.Nf3 d6 4.Qxh7 Be7 5.Qg8+ Kd7 6.Qxg7 Kd8 7.Qg8+ Kd7 8.Qxf7 Kd8 9.Nc3 Bg4 10.Qg8+ Kc7 11.Qxg4 Bf6 12.Qg8 Bd8 13.Ng5 Bxg5 14.Qxg5 Kd7 15.Qg7+ Kc8 16.Ba6 bxa6 17.0-0 a5 18.f4 exf4 19.Rxf4 c5 20.Rf8+
62/39 2:21:16 158.317.142.203 18.676.162 -M19 1. ... c6 2.Qh5 d6 3.Qxh7 e5 4.Nf3 Be7 5.Qg8+ Kd7 6.Qxg7 Kd8 7.Qg8+ Kd7 8.Qxf7 Kd8 9.Nc3 Bg4 10.Qg8+ Kc7 11.Qxg4 Bf6 12.Qg8 Bd8 13.Ng5 Bxg5 14.Qxg5 Kd7 15.Qg7+ Kc8 16.Ba6 bxa6 17.0-0 a5 18.f4 exf4 19.Rxf4 c5 20.Rf8+
62/39 2:21:16 158.317.142.203 18.676.162 -M19 1. ... g5 2.Qh5 Bg7 3.Qxg5 Kf8 4.Qc5 c6 5.Qxa7 Ke8 6.Qa8 Kd8 7.Qa5+ b6 8.Qxb6+ Ke8 9.Qc7 e6 10.Qxc8+ Ke7 11.Nc3 f5 12.Nf3 Bf6 13.e5 Bg7 14.Na4 Bh6 15.Nc5 Kf7 16.Qxd7+ Kg8 17.Qe8+ Bf8 18.Nxe6 h5 19.Qxf8+ Kh7 20.Qg7+
62/39 2:21:16 158.317.142.203 18.676.162 -M19 1. ... b6 2.Qh5 h6 3.Qe5 d6 4.Qd5 a6 5.Qc6+ Kd8 6.Nc3 e6 7.Nb5 Bd7 8.Qxc7+ Ke8 9.Qb8+ Ke7 10.Qxd6+ Ke8 11.Nc7+ Kd8 12.Nxe6+ fxe6 13.Qxf8+ Kc7 14.b4 Bc6 15.Bb2 b5 16.Be5+ Kd7 17.Bd6 Bxe4 18.Qe7+ Kc6 19.Qc7+ Kd5 20.Qc5+
62/41 2:21:16 158.317.142.203 18.676.162 -M20 1. ... d6 2.Qh5 Bd7 3.Qxh7 Be6 4.Nf3 f6 5.Qh5+ g6 6.Qxg6+ Bf7 7.Qf5 e6 8.Qb5+ Ke7 9.Qxb7 Bh6 10.Qxc7+ Kf8 11.Qxd6+ Kg8 12.Nd4 Bf8 13.Qd7 f5 14.Nxe6 Bh6 15.Nd8 Bg6 16.Bc4+ Kh8 17.Nf7+ Kh7 18.Nxh6+ Kxh6 19.d4+ Kh5 20.Be2+ Kh4 21.Qd8+
62/43 2:21:16 158.317.142.203 18.676.162 -M21 1. ... e6 2.Qh5 h6 3.Qa5 Kd8 4.Qxa7 Bd6 5.Nc3 c6 6.Qd4 Kc7 7.Qxg7 f6 8.Qxf6 Bb4 9.Nf3 Bc5 10.Qxh6 Be7 11.Qg7 Kd8 12.Qh8+ Kc7 13.Nb5+ cxb5 14.Nd4 d6 15.Qe8 Bf6 16.Nxb5+ Kb8 17.Qf8 Be5 18.Qd8 d5 19.d4 Bd6 20.Qxd6+ Ka8 21.Qa3+ Kb8 22.Qa7+
62/43 2:21:16 158.317.142.203 18.676.162 -M21 1. ... d5 2.exd5 a6 3.Qf3 b5 4.Qf4 e5 5.Qxe5+ Kd8 6.Qh5 Ke8 7.Nf3 h6 8.Ne5 g6 9.Qe2 Bd6 10.Nxg6+ Kd7 11.Ne5+ Ke7 12.Qh5 Kd8 13.Nxf7+ Ke7 14.Ne5 Bxe5 15.Qxe5+ Kd8 16.Qf6+ Ke8 17.Bd3 Bg4 18.0-0 Kd7 19.Re1 Kc8 20.Re8+ Kb7 21.Qc6+ Ka7 22.Qxc7+
62/43 2:21:16 158.317.142.203 18.676.162 -M21 1. ... h6 2.Qh5 e6 3.Qa5 Kd8 4.Qxa7 Bd6 5.Nc3 c6 6.Qd4 Kc7 7.Qxg7 f6 8.Qxf6 Bb4 9.Nf3 Bc5 10.Qxh6 Be7 11.Qg7 Kd8 12.Qh8+ Kc7 13.Nb5+ cxb5 14.Nd4 d6 15.Qe8 Bf6 16.Nxb5+ Kb8 17.Qf8 Be5 18.Qd8 d5 19.d4 Bd6 20.Qxd6+ Ka8 21.Qa3+ Kb8 22.Qa7+
62/43 2:21:16 158.317.142.203 18.676.162 -M21 1. ... g6 2.Nc3 c6 3.Qe2 h5 4.Qe3 b6 5.Qd4 f6 6.Qa4 a6 7.Qd4 Bh6 8.Qxb6 Bf4 9.Nf3 a5 10.Qxa5 Bb7 11.Qa7 Bc7 12.Qxb7 Bd8 13.Nb5 h4 14.Qc8 Kf7 15.Qxd8 cxb5 16.Bxb5 g5 17.Qxd7 h3 18.Qe8+ Kg7 19.Qxe7+ Kg6 20.Be8+ Kh6 21.Qf7 hxg2 22.Qg6+
62/43 2:21:16 158.317.142.203 18.676.162 -M21 1. ... a6 2.Qh5 h6 3.Qc5 Kd8 4.Bc4 f6 5.Qh5 e6 6.Qf7 Bb4 7.Qg8+ Ke7 8.Qxg7+ Ke8 9.Qg8+ Bf8 10.Be2 d6 11.Bh5+ Kd7 12.Qxf8 c6 13.Qxf6 a5 14.d3 Kc7 15.Qe7+ Bd7 16.Be8 Kb6 17.Qxd7 d5 18.Be3+ Ka6 19.Qc7 d4 20.Bxd4 c5 21.Bxc5 b6 22.Qxb6+

Even after hours of forward- and backward analysis, I haven't been able to find a faster mate.
Is there a faster mate? Especially after the probably strongest continuation 1.e4 g6?
Maybe with the maneuver h4 and Rh3 or a4 and Ra3?

[Jouni]

Crystal is definitely better mate finder than SF11. Try it with NNUE disabled.

[fastgm]

In my experience, Stockfish 11 is better than Crystal on longer checkmates.
But I will try the new version Crystal 3.1 with NNUE disabled.

[Jouni]

Forget Crystal. It's not good in these kind of positions! It's depth is as average 7 plies below normal SF and it always wants to open with 1.e3 .

[duncan]

I was interested in the following position. The initial position with five pieces less for black. How many moves does white need to checkmate?

[d]2b1kb2/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1
To solve the position, I used Stockfish 11. In my experience, the most exact engine when it comes to mate problems.
After 60 hours on a 10-core with 32 GB hash, Stockfish found a checkmate in 22 moves.

Stockfish 11 64 POPCNT:
96/44 60:29:31 4.669.383.221.790 21.441.650 +M22 1.e4 g6 2.Nc3 c6 3.Qe2 h5 4.Qe3 b6 5.Qd4 f6 6.Qa4 a6 7.Qd4 b5 8.Qb6 d6 9.Qxc6+ Bd7 10.Qxa6 Kf7 11.Nd5 Be6 12.Nc7 Bd7 13.Bxb5 Bg4 14.Qa8 h4 15.h3 Bc8 16.Qxc8 f5 17.Qe8+ Kg7 18.Ne6+ Kf6 19.Nxf8 fxe4 20.Qxg6+ Ke5 21.Qg5+ Kd4 22.c3+

and all black moves after 1.e4:
62/31 2:21:16 158.317.142.203 18.676.162 -M15 1. ... f5 2.Qh5+ g6 3.Qxh7 d6 4.Qxg6+ Kd8 5.d4 f4 6.Bxf4 a6 7.Bg5 Bf5 8.Qf7 Bxe4 9.Qxf8+ Kd7 10.Qxe7+ Kc6 11.Qxe4+ Kb6 12.Bd8 a5 13.Qe8 a4 14.Qb5+ Ka7 15.Bxc7 Ka8 16.Qxa4+
62/33 2:21:16 158.317.142.203 18.676.162 -M16 1. ... f6 2.Qh5+ g6 3.Qxh7 d6 4.Qxg6+ Kd7 5.Nc3 c6 6.Nf3 a6 7.Nd4 Bh6 8.Qxh6 f5 9.Qg7 fxe4 10.Nf5 Kc7 11.Qxe7+ Bd7 12.Qxd6+ Kb6 13.Qb8 Bxf5 14.Na4+ Ka5 15.Qd8+ Kxa4 16.b3+ Kb4 17.Qb6+
62/33 2:21:16 158.317.142.203 18.676.162 -M16 1. ... h5 2.Qxh5 g6 3.Qh7 d5 4.exd5 Bg4 5.Bb5+ c6 6.dxc6 bxc6 7.Bxc6+ Kd8 8.Qxf7 Kc7 9.Qxf8 Kxc6 10.Qxe7 Bd7 11.Nf3 Bc8 12.Ne5+ Kb6 13.Nc3 Bd7 14.Qxd7 a6 15.Nc4+ Kc5 16.d4+ Kxc4 17.Qa4+
62/39 2:21:16 158.317.142.203 18.676.162 -M16 1. ... Kd8 2.Qh5 Ke8 3.Qxh7 d6 4.Nf3 Bg4 5.Ng5 Kd7 6.Qg8 Bh5 7.Qxf8 a6 8.Nxf7 Bxf7 9.Qxf7 Kd8 10.d4 Kd7 11.Nc3 c6 12.Bg5 Kc7 13.Qxe7+ Kb6 14.Qxd6 a5 15.Qb8 c5 16.Bd8+ Kc6 17.Qc7+
62/35 2:21:16 158.317.142.203 18.676.162 -M17 1. ... a5 2.Qh5 a4 3.Qxh7 a3 4.Nxa3 d5 5.exd5 Bg4 6.Bb5+ c6 7.dxc6 bxc6 8.Bxc6+ Bd7 9.Qf5 e6 10.Bxd7+ Ke7 11.Qc5+ Kxd7 12.Qxf8 f6 13.Qxg7+ Kd6 14.Nb5+ Kc6 15.Qc7+ Kxb5 16.Qd6 Ka4 17.Qc5 e5 18.b3+
62/37 2:21:16 158.317.142.203 18.676.162 -M18 1. ... c5 2.Qh5 c4 3.Qc5 Kd8 4.Qa5+ b6 5.Qxa7 e6 6.Qxb6+ Ke8 7.Bxc4 g5 8.Nc3 Be7 9.Qb8 Kd8 10.Na4 h6 11.Nb6 Bc5 12.Qxc8+ Ke7 13.Qxc5+ d6 14.Nc8+ Kf6 15.Nxd6 g4 16.Qc7 h5 17.Qxf7+ Ke5 18.Qxe6+ Kd4 19.Qd5+
62/37 2:21:16 158.317.142.203 18.676.162 -M18 1. ... b5 2.Qh5 c6 3.Qxh7 g5 4.Nf3 f6 5.Qg6+ Kd8 6.Qg8 Ke8 7.e5 a5 8.Bd3 f5 9.Nxg5 f4 10.Qf7+ Kd8 11.Qxf8+ Kc7 12.Qxe7 Kb7 13.Qd8 f3 14.Nf7 fxg2 15.Nd6+ Ka6 16.Rg1 a4 17.Qc7 Bb7 18.b4 axb3 19.axb3+
62/39 2:21:16 158.317.142.203 18.676.162 -M19 1. ... e5 2.Qh5 c6 3.Nf3 d6 4.Qxh7 Be7 5.Qg8+ Kd7 6.Qxg7 Kd8 7.Qg8+ Kd7 8.Qxf7 Kd8 9.Nc3 Bg4 10.Qg8+ Kc7 11.Qxg4 Bf6 12.Qg8 Bd8 13.Ng5 Bxg5 14.Qxg5 Kd7 15.Qg7+ Kc8 16.Ba6 bxa6 17.0-0 a5 18.f4 exf4 19.Rxf4 c5 20.Rf8+
62/39 2:21:16 158.317.142.203 18.676.162 -M19 1. ... c6 2.Qh5 d6 3.Qxh7 e5 4.Nf3 Be7 5.Qg8+ Kd7 6.Qxg7 Kd8 7.Qg8+ Kd7 8.Qxf7 Kd8 9.Nc3 Bg4 10.Qg8+ Kc7 11.Qxg4 Bf6 12.Qg8 Bd8 13.Ng5 Bxg5 14.Qxg5 Kd7 15.Qg7+ Kc8 16.Ba6 bxa6 17.0-0 a5 18.f4 exf4 19.Rxf4 c5 20.Rf8+
62/39 2:21:16 158.317.142.203 18.676.162 -M19 1. ... g5 2.Qh5 Bg7 3.Qxg5 Kf8 4.Qc5 c6 5.Qxa7 Ke8 6.Qa8 Kd8 7.Qa5+ b6 8.Qxb6+ Ke8 9.Qc7 e6 10.Qxc8+ Ke7 11.Nc3 f5 12.Nf3 Bf6 13.e5 Bg7 14.Na4 Bh6 15.Nc5 Kf7 16.Qxd7+ Kg8 17.Qe8+ Bf8 18.Nxe6 h5 19.Qxf8+ Kh7 20.Qg7+
62/39 2:21:16 158.317.142.203 18.676.162 -M19 1. ... b6 2.Qh5 h6 3.Qe5 d6 4.Qd5 a6 5.Qc6+ Kd8 6.Nc3 e6 7.Nb5 Bd7 8.Qxc7+ Ke8 9.Qb8+ Ke7 10.Qxd6+ Ke8 11.Nc7+ Kd8 12.Nxe6+ fxe6 13.Qxf8+ Kc7 14.b4 Bc6 15.Bb2 b5 16.Be5+ Kd7 17.Bd6 Bxe4 18.Qe7+ Kc6 19.Qc7+ Kd5 20.Qc5+
62/41 2:21:16 158.317.142.203 18.676.162 -M20 1. ... d6 2.Qh5 Bd7 3.Qxh7 Be6 4.Nf3 f6 5.Qh5+ g6 6.Qxg6+ Bf7 7.Qf5 e6 8.Qb5+ Ke7 9.Qxb7 Bh6 10.Qxc7+ Kf8 11.Qxd6+ Kg8 12.Nd4 Bf8 13.Qd7 f5 14.Nxe6 Bh6 15.Nd8 Bg6 16.Bc4+ Kh8 17.Nf7+ Kh7 18.Nxh6+ Kxh6 19.d4+ Kh5 20.Be2+ Kh4 21.Qd8+
62/43 2:21:16 158.317.142.203 18.676.162 -M21 1. ... e6 2.Qh5 h6 3.Qa5 Kd8 4.Qxa7 Bd6 5.Nc3 c6 6.Qd4 Kc7 7.Qxg7 f6 8.Qxf6 Bb4 9.Nf3 Bc5 10.Qxh6 Be7 11.Qg7 Kd8 12.Qh8+ Kc7 13.Nb5+ cxb5 14.Nd4 d6 15.Qe8 Bf6 16.Nxb5+ Kb8 17.Qf8 Be5 18.Qd8 d5 19.d4 Bd6 20.Qxd6+ Ka8 21.Qa3+ Kb8 22.Qa7+
62/43 2:21:16 158.317.142.203 18.676.162 -M21 1. ... d5 2.exd5 a6 3.Qf3 b5 4.Qf4 e5 5.Qxe5+ Kd8 6.Qh5 Ke8 7.Nf3 h6 8.Ne5 g6 9.Qe2 Bd6 10.Nxg6+ Kd7 11.Ne5+ Ke7 12.Qh5 Kd8 13.Nxf7+ Ke7 14.Ne5 Bxe5 15.Qxe5+ Kd8 16.Qf6+ Ke8 17.Bd3 Bg4 18.0-0 Kd7 19.Re1 Kc8 20.Re8+ Kb7 21.Qc6+ Ka7 22.Qxc7+
62/43 2:21:16 158.317.142.203 18.676.162 -M21 1. ... h6 2.Qh5 e6 3.Qa5 Kd8 4.Qxa7 Bd6 5.Nc3 c6 6.Qd4 Kc7 7.Qxg7 f6 8.Qxf6 Bb4 9.Nf3 Bc5 10.Qxh6 Be7 11.Qg7 Kd8 12.Qh8+ Kc7 13.Nb5+ cxb5 14.Nd4 d6 15.Qe8 Bf6 16.Nxb5+ Kb8 17.Qf8 Be5 18.Qd8 d5 19.d4 Bd6 20.Qxd6+ Ka8 21.Qa3+ Kb8 22.Qa7+
62/43 2:21:16 158.317.142.203 18.676.162 -M21 1. ... g6 2.Nc3 c6 3.Qe2 h5 4.Qe3 b6 5.Qd4 f6 6.Qa4 a6 7.Qd4 Bh6 8.Qxb6 Bf4 9.Nf3 a5 10.Qxa5 Bb7 11.Qa7 Bc7 12.Qxb7 Bd8 13.Nb5 h4 14.Qc8 Kf7 15.Qxd8 cxb5 16.Bxb5 g5 17.Qxd7 h3 18.Qe8+ Kg7 19.Qxe7+ Kg6 20.Be8+ Kh6 21.Qf7 hxg2 22.Qg6+
62/43 2:21:16 158.317.142.203 18.676.162 -M21 1. ... a6 2.Qh5 h6 3.Qc5 Kd8 4.Bc4 f6 5.Qh5 e6 6.Qf7 Bb4 7.Qg8+ Ke7 8.Qxg7+ Ke8 9.Qg8+ Bf8 10.Be2 d6 11.Bh5+ Kd7 12.Qxf8 c6 13.Qxf6 a5 14.d3 Kc7 15.Qe7+ Bd7 16.Be8 Kb6 17.Qxd7 d5 18.Be3+ Ka6 19.Qc7 d4 20.Bxd4 c5 21.Bxc5 b6 22.Qxb6+

Even after hours of forward- and backward analysis, I haven't been able to find a faster mate.
Is there a faster mate? Especially after the probably strongest continuation 1.e4 g6?
Maybe with the maneuver h4 and Rh3 or a4 and Ra3?

I am weak at chess and gave stockfish 3 sec /move, used stockfish help and was mate in 16. I assume it did not play very well.


[pgn]
[Event "Computer chess game"]
[Site "PC"]
[Date "2021.01.26"]
[Round "?"]
[White "user"]
[Black "Stockfish_20092110_x64_avx2"]
[Result "1-0"]
[BlackElo "1000"]
[Time "22:03:27"]
[WhiteElo "2400"]
[TimeControl "0+3"]
[SetUp "1"]
[FEN "2b1kb2/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1"]
[Termination "normal"]
[PlyCount "31"]
[WhiteType "human"]
[BlackType "program"]

1. e4 a6 {(a7-a6 Qd1-h5) -29.18/26 3} 2. Nc3 e6 {(e7-e6 Qd1-g4) -28.99/24
3} 3. Nf3 g6 {(g7-g6 Qd1-e2) -28.69/23 3} 4. d4 d6 {(d7-d6 Qd1-d3)
-28.79/21 3} 5. d5 Bg7 {(Bf8-g7) -28.52/22 3} 6. dxe6 Bxe6 {(Bc8xe6 Nc3-d5)
-28.49/20 3} 7. e5 Bg4 {(Be6-g4 Nc3-d5 Bg7xe5 Nd5xc7+ Ke8-d8 Nf3xe5 Bg4xd1
Ne5xf7+ Kd8xc7 Ke1xd1 b7-b5 Bf1-d3 Kc7-c6 Nf7-g5 h7-h6 Ng5-f7 h6-h5 Bc1-g5
a6-a5 Bd3-e4+ d6-d5 Be4xg6) -29.63/18 3} 8. exd6 cxd6 {(c7xd6 Nc3-e4)
-33.75/17 3} 9. h3 Bxf3 {(Bg4xf3 Qd1xf3) -29.77/17 3} 10. Qxf3 f5 {(f7-f5
Nc3-e2) -31.40/16 1} 11. Qxb7 Bf6 {(Bg7-f6 Bc1-h6 Bf6-e7 Bf1-c4 d6-d5
Bc4xd5 Be7-h4 Bd5-c6+ Ke8-d8 Qb7-d7+) -M5/63 3} 12. Bh6 f4 {(f5-f4 Nc3-d5
Bf6-h4 O-O-O Ke8-d8 Nd5-b6 Bh4xf2 Qb7-d7+) -M5/63} 13. Nd5 Bh4 {(Bf6-h4
O-O-O Ke8-d8 Nd5-b6 Bh4xf2 Qb7-d7+) -M5/63} 14. O-O-O f3 {(f4-f3 Rd1-e1+
Ke8-d8 Qb7-c7+) -M5/63} 15. Re1+ Be7 {(Bh4-e7 Qb7xe7+) -M5/63} 16. Qxe7#
1-0
[/pgn]

[Uri Blass]

I know that stockfish is designed to see wrong things because of hash collisions so I would like to know if it can see wrong mates(mates that are shorter than the shortest possible mate or see mates when there is no possible mate because of hash collisions).

I am not expert in stockfish's code and I would like to know if hash collisions can in thoery generate this situation or not.
Does somebody has a proof that it can generate this situation or alternatively a proof that it is in theory impossible?

[Dann Corbit]

There does seem to be some problem with the given pv. I would say that it is not trustworthy.
Here is analysis of the pv positions, from the position before mate, backwards to the 4th move:

Code: Select all

4k3/1Q2b2p/p2p2pB/3N4/8/5p1P/PPP2PP1/2K1RB1R w - - acd 999; acs 0; bm Qxe7#; ce 32766; c3 "Qxe7#"; dm 1; pm Qxe7#; pv Qxe7#; 
4k3/1Q5p/p2p2pB/3N4/7b/5p1P/PPP2PP1/2K1RB1R b - - acd 999; acs 0; bm Kd8; ce -32765; c3 "Be7"; dm -1; pm Kd8; pv Kd8 Qc7#; 
4k3/1Q5p/p2p2pB/3N4/7b/5p1P/PPP2PP1/2KR1B1R w - - acd 999; acs 0; bm Re1+; ce 32764; c3 "Re1+"; dm 2; pm Re1+; pv Re1+ Be7 Qxe7#; 
4k3/1Q5p/p2p2pB/3N4/5p1b/7P/PPP2PP1/2KR1B1R b - - acd 51; acs 0; bm Kd8; ce -32763; c3 "f3"; dm -2; pm f3; pv Kd8 Nb6 Be7 Qb8+; 
4k3/1Q5p/p2p2pB/3N4/5p1b/7P/PPP2PP1/R3KB1R w KQ - acd 999; acs 0; bm Bxa6 Kd2 O-O-O; ce 32762; c3 "O-O-O"; dm 3; pm Bxa6 Kd2 O-O-O; pv Bxa6 Bxf2+ Kxf2 g5 Qe7#; 
4k3/1Q5p/p2p1bpB/3N4/5p2/7P/PPP2PP1/R3KB1R b KQ - acd 999; acs 0; bm Bc3+; ce -32761; c3 "Bh4"; dm -3; pm Bc3+; pv Bc3+ bxc3 Kd8 Bg5+ Ke8 Qe7#; 
4k3/1Q5p/p2p1bpB/8/5p2/2N4P/PPP2PP1/R3KB1R w KQ - acd 999; acs 0; bm Bxa6 Nd5 O-O-O; ce 32760; c3 "Nd5"; dm 4; pm Bxa6 Nd5 O-O-O; pv Nd5 Bc3+ bxc3 Kd8 Bg5+ Ke8 Qe7#; 
4k3/1Q5p/p2p1bpB/5p2/8/2N4P/PPP2PP1/R3KB1R b KQ - acd 999; acs 0; bm Bxc3+; ce -32759; c3 "f4"; dm -4; pm Bxc3+; pv Bxc3+ bxc3 g5 Bd3 g4 Bxf5 g3 Qd7#; 
4k3/1Q5p/p2p1bp1/5p2/8/2N4P/PPP2PP1/R1B1KB1R w KQ - acd 999; acs 1; bm Bh6; ce 32758; c3 "Bh6"; dm 5; pm Bh6; pv Bh6 Bxc3+ bxc3 g5 Bd3 g4 Bxf5 g3 Qd7#; 
4k3/1Q4bp/p2p2p1/5p2/8/2N4P/PPP2PP1/R1B1KB1R b KQ - acd 51; acs 0; bm Kf8; ce -32757; c3 "Bf6"; dm -5; pm Bf6; pv Kf8 Bc4 Bxc3+ bxc3 d5 Bxd5 Ke8 Bg5 a5 Qe7+; 
4k3/1p4bp/p2p2p1/5p2/8/2N2Q1P/PPP2PP1/R1B1KB1R w KQ - acd 999; acs 11; bm Qxb7; ce 32756; c3 "Qxb7"; dm 6; pm Qxb7; pv Qxb7 Bh8 Nd5 Kf8 Bh6+ Kg8 Qc8+ Kf7 Qd7+ Kg8 Qe8#; 
4k3/1p3pbp/p2p2p1/8/8/2N2Q1P/PPP2PP1/R1B1KB1R b KQ - acd 60; acs 2; bm Bxc3+; ce -32755; c3 "f5"; dm -6; pm f5; pv Bxc3+ bxc3 g5 Bc4 d5 Qxd5 Ke7 Bxg5+ f6 O-O-O fxg5 Qd8+; 
4k3/1p3pbp/p2p2p1/8/8/2N2b1P/PPP2PP1/R1BQKB1R w KQ - acd 999; acs 1084; bm Qxd6 Qxf3 gxf3; ce 32752; c3 "Qxf3"; dm 8; pm Qxd6 Qxf3 gxf3; pv gxf3 Bxc3+ bxc3 Ke7 Bc4 h6 Bf4 d5 Qxd5 b5 Qe5+ Kd7 Rd1+ Kc6 Qc7#; 
4k3/1p3pbp/p2p2p1/8/6b1/2N2N1P/PPP2PP1/R1BQKB1R b KQ - acd 51; acs 1; bm Kd7; ce -32751; c3 "Bxf3"; dm -8; pm Bxf3; pv Kd7 hxg4 h5 Qd5 Bxc3+ bxc3 Kc7 Bf4 Kd7 Qxd6+ Ke8 Qe5+ Kf8 Bh6+ Kg8 Qg7+; 
4k3/1p3pbp/p2p2p1/8/6b1/2N2N2/PPP2PPP/R1BQKB1R w KQ - acd 999; acs 84; bm Qxd6; ce 32754; c3 "h3"; dm 7; pm Qxd6; pv Qxd6 Bf8 Qb8+ Ke7 Nd5+ Ke6 Qe5+ Kd7 Qc7+ Ke6 Nf4+ Kf6 Qe5#; 
4k3/1pp2pbp/p2P2p1/8/6b1/2N2N2/PPP2PPP/R1BQKB1R b KQ - acd 60; acs 1; bm c5; ce -32753; c3 "cxd6"; dm -7; pm cxd6; pv c5 d7+ Bxd7 Qd6 Bxc3+ bxc3 Kd8 Bf4 b5 O-O-O Ke8 Bh6 f6 Qxd7+; 
4k3/1pp2pbp/p2p2p1/4P3/6b1/2N2N2/PPP2PPP/R1BQKB1R w KQ - acd 999; acs 149004; bm exd6; ce 32750; c3 "exd6"; dm 9; pm exd6; pv exd6 c6 d7+ Bxd7 Bg5 h6 Qd6 Bxc3+ bxc3 hxg5 Rd1 g4 Ne5 g3 Qxd7+ Kf8 Qxf7#; 
4k3/1pp2pbp/p2pb1p1/4P3/8/2N2N2/PPP2PPP/R1BQKB1R b KQ - acd 60; acs 46; bm Kd7; ce -32745; c3 "Bg4"; dm -11; pm Bg4; pv Kd7 exd6 c6 Na4 Bd5 Nb6+ Kxd6 Bf4+ Ke7 Nxd5+ cxd5 Qxd5 Bxb2 Bc4 Bc3+ Kd1 b5 Qxf7+ Kd8 Be6 Bxa1 Bc7+; 
4k3/1pp2pbp/p2pb1p1/8/4P3/2N2N2/PPP2PPP/R1BQKB1R w KQ - acd 58; acs 171; bm Nd5; ce 32744; c3 "e5"; dm 12; pm e5; pv Nd5 Kd7 Nxc7 Kxc7 Bf4 Bf8 Qd4 Be7 Qc5+ Kd7 Bxd6 Bxd6 O-O-O b5 Qxd6+ Kc8 Qe7 Kb8 Rd8+ Bc8 Qd7 Ka8 Rxc8+; 
2b1k3/1pp2pbp/p2pP1p1/8/4P3/2N2N2/PPP2PPP/R1BQKB1R b KQ - acd 55; acs 104; bm fxe6; ce -32743; c3 "Bxe6"; dm -12; pm Bxe6; pv fxe6 Qd3 Bf8 Qd4 h6 Qh8 Kf7 Bxh6 Bxh6 Qxh6 Bd7 Qh7+ Ke8 Ne5 dxe5 O-O-O Bc6 Qxc7 Bd5 exd5 Kf8 dxe6 b6 Rd8+; 
2b1k3/1pp2pbp/p2pp1p1/3P4/4P3/2N2N2/PPP2PPP/R1BQKB1R w KQ - acd 51; acs 903; bm Qd3; ce 32740; c3 "dxe6"; dm 14; pm dxe6; pv Qd3 e5 Qc4 Kd7 Ng5 Bf6 Nxf7 g5 Be2 h5 Bxh5 Be7 Bxg5 Bxg5 Bg4+ Ke8 Qxc7 Bd2+ Kxd2 Kf8 Nh6 Be6 dxe6 Ke8 e7 b5 Bh5+; 
2b1kb2/1pp2p1p/p2pp1p1/3P4/4P3/2N2N2/PPP2PPP/R1BQKB1R b KQ - acd 51; acs 1248; bm e5; ce -32737; c3 "Bg7"; dm -15; pm Bg7; pv e5 Qd3 h6 Qc4 Kd8 Ng5 Bd7 Nxf7+ Kc8 Be3 b5 Qb4 Be8 Qa5 Bxf7 Nxb5 axb5 Qa8+ Kd7 Qxf8 Bxd5 Bxb5+ c6 exd5 Kc7 Bxc6 h5 Ba7 g5 Qb8+; 
2b1kb2/1pp2p1p/p2pp1p1/8/3PP3/2N2N2/PPP2PPP/R1BQKB1R w KQ - acd 51; acs 3165; bm Qd3; ce 32728; c3 "d5"; dm 20; pm d5; pv Qd3 c6 Qc4 Be7 Qb3 g5 Nxg5 b5 Nxh7 Kd7 Ne2 a5 Qg3 a4 Qg7 a3 Qxf7 axb2 Nf6+ Kd8 Qe8+ Kc7 Qxe7+ Kb8 Ne8 bxa1Q Qc7+ Ka8 Qxc8+ Ka7 Qc7+ Ka6 Nxd6 Qb2 Qxc6+ Ka5 Bd2+ Qc3 Qxb5+; 
2b1kb2/1ppp1p1p/p3p1p1/8/3PP3/2N2N2/PPP2PPP/R1BQKB1R b KQ - acd 51; acs 11063; bm Bg7; ce -32735; c3 "d6"; dm -16; pm d6; pv Bg7 Qd2 d6 Qf4 f6 Qh4 h5 e5 Ke7 exf6+ Bxf6 Qe4 Kd7 Qxg6 Be7 Qf7 b5 d5 c6 dxe6+ Kd8 Be3 Bxe6 Qxe6 Ke8 Ng5 h4 Bb6 c5 Qc8+ Bd8 Qxd8+; 
2b1kb2/1ppp1p1p/p3p1p1/8/4P3/2N2N2/PPPP1PPP/R1BQKB1R w KQ - acd 59; acs 16640; bm Qe2; ce 32726; c3 "d4"; dm 21; pm d4; pv Qe2 Kd8 Qe3 f6 Qf4 Be7 Qh6 g5 Qxh7 c6 Qg7 g4 Qxg4 b6 h4 f5 exf5 a5 Qg8+ Kc7 Qe8 Bf6 Ne4 Bd8 d4 Bf6 Nxf6 b5 Bf4+ e5 Qxe5+ d6 Qxd6+ Kb6 Qb8+ Bb7 Nd7+ Ka6 Bc7 a4 Nc5+; 

[fastgm]

[d]2b1kb2/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1
This is currently the critical variant that should be analyzed:

1. e4 g6 2. Nc3 c6 3. Qe2 h5 4. Qe3 b6 5. Qd4 f6 6. Qa4 a6 7. Qd4 Bh6 8. Qxb6 Bf4 9. Nf3 a5 10. Qxa5 Bb7 11. Qa7 Bc8 12. Qa8 Kd8 13. Ba6 e6 14. Qxc8+ Ke7 15. d4 Bxc1 16. Rxc1 e5 17. dxe5 fxe5 18. Nxe5 Kf6 19. Qxd7 Kg5 20. Qxc6 h4 21. Qxg6+ Kf4 22. Qf6#

[peter]

[d]2b1kb2/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1
This is currently the critical variant that should be analyzed:

1. e4 g6 2. Nc3 c6 3. Qe2 h5 4. Qe3 b6 5. Qd4 f6 6. Qa4 a6 7. Qd4 Bh6 8. Qxb6 Bf4 9. Nf3 a5 10. Qxa5 Bb7 11. Qa7 Bc8 12. Qa8 Kd8 13. Ba6 e6 14. Qxc8+ Ke7 15. d4 Bxc1 16. Rxc1 e5 17. dxe5 fxe5 18. Nxe5 Kf6 19. Qxd7 Kg5 20. Qxc6 h4 21. Qxg6+ Kf4 22. Qf6#

As told here

https://forum.computerschach.de/cgi-bin ... #pid139302

I'd say, it's rather this one:

2b1kb2/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1

1. e4 g6 2. Qe2 d6 3. e5 e6 4. Nf3 Be7 5. Qc4 Kd7 6. d4 a6 7. Ng5 b5 8. Qb3 h6
9. Nc3 hxg5 10. Nxb5 axb5 11. Qf3 d5 12. Qxf7 Kd8 13. h4 gxh4 14. Bxb5 c6 15.
Bxc6 Kc7 16. Qxe7+ Kxc6 17. Rxh4 Bd7 18. Qc5+ Kb7 19. Rh7 Kb8 20. Rxd7 g5 21. Qf8#

Can't prove it by output back to the (rather strange, at the edge to be reached by legal play) given starting position, my old hardware didn't make it further than to the position after first 3 moves of the line above

2b1kb2/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1

1. e4 g6 2. Qe2 d6 3. e5 e6

2b1kb2/ppp2p1p/3pp1p1/4P3/8/8/PPPPQPPP/RNB1KBNR w KQ - 0 1

Analysis by Cfish 110720 64 POPCNT NUMA:

4.Sf3 h6 5.Dc4 Kd7 6.Df4 f5 7.exf6 a6 8.Da4+ Kd8 9.De4 Kd7 10.Dxg6 e5 11.Df7+ Kc6 12.Dxf8 Le6 13.De8+ Ld7 14.Sxe5+ dxe5 15.Dxe5 b5 16.d4 Kb6 17.Lf4 a5 18.Dxc7+ Ka6 19.Dxd7 Kb6 20.Dxb5+ Ka7 21.Da6#
+- (#18) Tiefe: 65/35 00:02:20 3789MN

Sorry for the german abbrevations, didn't want to stop the running process for changing.

BTW, I don't think my Xeon X5670 will make it to get the #21 into output at starting position, too little hash, too many sidelines to be corrected, if engine on it's own doesn't find the very shortest line on its own.

But one thing you're right about, Andreas, SF since NNUE has lost a little of it's mate- search- speed due to NNUE- code, even without usage of net.
That's why I took one of the latest Cfish compiles before NNUE- came up.

Good candidates for analysing this aberrant mate are SlowChess 2.5 (very good specialised mate search), HekaStockfish with find mate and Xiphos, the latest one very thorough as for cutting detours on its own, just a little too litlle time to depth for this one "problem").

Somebody else with better hardware may "prove" this DTM, maybe I yet wouldn let any longer then 21 get to be surely proven by output only as for my perosnal pov, could well be even shorter, 20 seems as probable to me as 21.

Just don't think output at the starting position conclusive at all to be got with the current software- hardware combinations regards

[fastgm]

[d]2b1kb2/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1
This is currently the critical variant that should be analyzed:

1. e4 g6 2. Nc3 c6 3. Qe2 h5 4. Qe3 b6 5. Qd4 f6 6. Qa4 a6 7. Qd4 Bh6 8. Qxb6 Bf4 9. Nf3 a5 10. Qxa5 Bb7 11. Qa7 Bc8 12. Qa8 Kd8 13. Ba6 e6 14. Qxc8+ Ke7 15. d4 Bxc1 16. Rxc1 e5 17. dxe5 fxe5 18. Nxe5 Kf6 19. Qxd7 Kg5 20. Qxc6 h4 21. Qxg6+ Kf4 22. Qf6#

With 11.d4! it is one move shorter, so currently mate in 21 moves. Maybe there are still improvements for black before move 10.

1. e4 g6 2. Nc3 c6 3. Qe2 h5 4. Qe3 b6 5. Qd4 f6 6. Qa4 a6 7. Qd4 Bh6 8. Qxb6 Bf4 9. Nf3 a5 10. Qxa5 Bb7 11. d4 g5 12. Bxf4 gxf4 13. Qc7 e6 14. Qxb7 Ke7 15. Ne5 fxe5 16. dxe5 Ke8 17. O-O-O Kf7 18. Rxd7+ Kg6 19. Qc8 h4 20. Qg8+ Kh6 21. Rh7#