Mate in 7 problem - Blue Marlin fails

[Karim]

[SetUp "1"] [FEN "4R1K1/1p6/N1pp1k1p/2p4P/1b3p1P/p2Pp3/2BPP3/3N4 w - - 0 1"] { 4R1K1/1p6/N1pp1k1p/2p4P/1b3p1P/p2Pp3/2BPP3/3N4 w - - 0 1 } *
Just checked the problem with blue marlin 14.1 arm7 with nnue network embedded. waited till it reached (D=25 534/Kn/s) using (analyze this) app, best move was Nxb4 , but failed to find mate in 7th move. Comparing to crystal 180420 without nnue ? Crystal found mate in 7th move in just (D=20 576 kn/s)
FEN: 4R1K1/1p6/N1pp1k1p/2p4P/1b3p1P/p2Pp3/2BPP3/3N4 w - - 0 1

[Peter Grayson]

[SetUp "1"] [FEN "4R1K1/1p6/N1pp1k1p/2p4P/1b3p1P/p2Pp3/2BPP3/3N4 w - - 0 1"] { 4R1K1/1p6/N1pp1k1p/2p4P/1b3p1P/p2Pp3/2BPP3/3N4 w - - 0 1 } *
Just checked the problem with blue marlin 14.1 arm7 with nnue network embedded. waited till it reached (D=25 534/Kn/s) using (analyze this) app, best move was Nxb4 , but failed to find mate in 7th move. Comparing to crystal 180420 without nnue ? Crystal found mate in 7th move in just (D=20 576 kn/s)
FEN: 4R1K1/1p6/N1pp1k1p/2p4P/1b3p1P/p2Pp3/2BPP3/3N4 w - - 0 1

An interesting mating position that highlights the movement of today's engines from precision to "a winning move will do" or even perhaps missing a key move.

A Mate in 7 that poses no difficulty for many of the "Old School" engines but is completely missed in a reasonable time by today's latest Stockfish 270821.

Running on my i7-10700 machine it made interesting reading ...

Shredder 7: 1.d4 bxa6 2.d5 cxd5 3.Ba4 Bxd2 4.Bd7
Be1 5.Bg4 Bxh4 6.Re6+ Kg5 7.Rg6# #7/14

Quickest solver by time was Zappa Mexico II x64

Zappa Mexico II x64: 1.d4 bxa6 2.d5
cxd5 3.Ba4 Ba5 4.Nc3 Bxc3 5.Bd7 exd2 6.Bg4 d1=Q 7.Re6# #7/21

Deep Shredder 10.1 x64: 1.d4 bxa6 2.d5 cxd5 3.Ba4 Bxd2 4.Bd7 Be1 5.Bg4 Bxh4 6.Re6+ Kg5 7.Rg6# #7/20

The native Chessbase Fritz engines shut off the CPU once mate was found but unfortunately it was premature!

Deep Fritz 8: 1.d4 bxa6 2.d5 cxd5 3.dxe3 Be1 4.exf4 d4 5.e4 Bxh4 6.Bb3 c4 7.Bxc4 #9/15

Deep Fritz 10: 1.d4 bxa6 2.dxe3 Be1 3.d5 cxd5 4.exf4 d4 5.e4 Bxh4 6.Bb3 c4 7.Bxc4 d5 8.Bxd5 a2 9.Re6# #9/16

The Deep Junior Engine also followed suit!

Deep Junior 10: 1.d4 bxa6 2.dxe3 Be1 3.d5 cxd5 4.exf4 d4 5.e4 Bxh4 6.Bb3 c4 7.Bxc4 d5 8.Bxd5 a2 9.Re6# #9/21

Stockfish 8 announced #33 with Nxb4, dropped to #31 and then after a little while longer announced #7

Stockfish 8 64 BMI2: 1.d4 bxa6 2.d5 cxd5 3.Ba4 Bxd2 4.Bd7 Be1 5.Bg4 Bxh4 6.Re6+ Kg5 7.Rg6# #7/53

Houdini 6.03 Pro x64-pext initially found #31 for Nxb4 but had a problem with successive iterations and ply depths increasing the mate depth to #38 and then after some time announced #7 and with the ply depth counter racing forward it had hit 83 ply depth before I could stop it.

Houdini 6.03 Pro x64-pext: 1.d4 bxa6 2.d5 cxd5 3.Ba4 a2 4.Nc3 Bxc3 5.Bd7 exd2 6.Bh3 a1=Q 7.Re6# #7/83

Komodo 11.3.1 64-bit was hopeless and I gave up after 10 minutes with the output not changing for 6 minutes of those.

Komodo 11.3.1 64-bit: 1.Nxb4 Kf5 2.dxe3 Kg4 3.Re4 b5 4.Rxf4+ Kh3 5.Kh7 c4 6.dxc4 bxc4 7.Kxh6 c5 8.Na2 d5 9.Kg5 d4 10.Be4 19.57/24

Stockfish 9 found #23 with Nxb6 but did not find 1.d4

Stockfish 9 64 BMI2: 1.Nxb4 cxb4 2.dxe3 a2 3.Rf8+ Ke7 4.Ra8 Kf6 5.Rxa2 Kf5 6.Ra7 Kg4 7.Rxb7 Kxh5 8.Rxb4 Kxh4 9.Rxf4+ Kg3 10.Rf6 Kg2 11.Rxh6 d5 12.Rxc6 Kf1 13.Rc5 Ke1 14.Rxd5 Kf1 15.d4 Kg1 16.Rc5 Kf1 17.d5 Ke1 18.d6 Kd2 19.d7 Kc1 20.Rd5 Kxc2 21.d8=Q Kb3 22.Qb6+ Ka4 23.Ra5# #23/46

The difference using a position solver was highlighted very quickly. It did run on a few ply before I could stop it but solved in seconds rather than minutes ...

Crystal 190121: 1.d4 bxa6 2.d5 cxd5 3.Ba4 a2 4.Nc3 Bxc3 5.Bd7 c4 6.Bg4 exd2 7.Re6# #7/26.

Compared to the latest Stockfish ...

Stockfish 270821: 1.Nxb4 cxb4 2.dxe3 a2 3.Rf8+ Ke7 4.Ra8 Kf6 5.e4 f3 6.e3 b3 7.Bxb3 Ke7 8.Rxa2 Kd8 9.Rf2 Kc7 10.Rxf3 Kb6 11.Rf7 Ka6 12.Kg7 b5 13.Rf6 Ka5 14.Rxd6 Kb4 15.Ba2 Kc5 16.Rxh6 b4 17.Bf7 Kb5 18.Be8 Ka5 19.Rxc6 Kb5 20.Rc4+ Ka5 21.Bc6 b3 22.Ra4+ Kb6 66.02/50

The drift towards Lc0 may have benefits but as exhibited here, there are certainly draw backs and I wonder if that contributes to the lack of wins in today's matches that now seems to require the use of unbalanced openings to achieve any sort of result in longer time control games?

Peter