trying to prove mate against 2 knights and 8 pawns

[peter]

Pity not to be kept in hash with correct DTM-outpu further then till 3rd move of Black's.

But at least till that it works.
After 1.e4 a6 2.d4 Nf6 3.e5 Nd5:

SlowChess Blitz 2.9 avx2 (32768 MB hash, 31 threads)
FEN: 1n2k3/1ppppppp/p7/3nP3/3P4/8/PPP2PPP/RNBQKBNR w HA - 1 4

[37] M15 00:05.5 95.36M Qh5 b5 Qxh7 d6 Qg8+ Kd7 Qxb8 Nb4 Bxb5+ axb5 Qxb5+ Kd8 Qxb4 dxe5 dxe5 Kd7 Nc3 f6 Nd5 e6 exf6 exd5 fxg7 Kc6 g8=Q Kd7 Qf7+ Kc6 Qe8#

[Uri Blass]

Bad replies against 1.e4

1...f6
2.Qh5+ is mate in at most 16 moves(means 17 from the root position)
possible main line start with
2.Qh5+ Kf8 3.Bc4 e6 4.Qc5+ Ne7 5.Qxc7 Nbc6 6.Qxd7 Ne5 7.Qxe6 Nxc4 8.b3 b5 9.bxc4 b4 10.Bb2

1...f5
2.Qh5+ is mate in at most 16 moves
possible line start with
2.Qh5+ Kd8 3.Qxh7 Nf6 4.Qh8+ Ne8 5.Bc4 d5 6.Bxd5 Nd7 7.d4 fxe4 8.Bg5 Nf6 9.Bxf6 gxf6 10. Bf7

[peter]

1...g6 seems somewhat logical against Qh5, but might not be longer than 18 moves neither.
Eg.
1.e4 g6 2.d4 d6 3.Nc3 h5 4.d5 c5:

SlowChess Blitz 2.9 avx2 (32768 MB hash, 31 threads)
FEN: 1n2k1n1/pp2pp2/3p2p1/2pP3p/4P3/2N5/PPP2PPP/R1BQKBNR w HA c6 0 5

[35] M14 00:00.6 10.78M e5 Kf8 exd6 exd6 Bf4 Kg7 Bxd6 Nd7 Qe2 Ngf6 Qe7 h4 Be5 Nxe5 Qxe5 a6 Ne4 h3 Qxf6+ Kf8 Ng5 Ke8 Qxf7+ Kd8 Ne6+ Kc8 Qe8#

[peter]

1...g6 seems somewhat logical against Qh5, but might not be longer than 18 moves neither.
Eg.
1.e4 g6 2.d4 d6 3.Nc3 h5 4.d5 c5:

SlowChess Blitz 2.9 avx2 (32768 MB hash, 31 threads)
FEN: 1n2k1n1/pp2pp2/3p2p1/2pP3p/4P3/2N5/PPP2PPP/R1BQKBNR w HA c6 0 5

[35] M14 00:00.6 10.78M e5 Kf8 exd6 exd6 Bf4 Kg7 Bxd6 Nd7 Qe2 Ngf6 Qe7 h4 Be5 Nxe5 Qxe5 a6 Ne4 h3 Qxf6+ Kf8 Ng5 Ke8 Qxf7+ Kd8 Ne6+ Kc8 Qe8#

Another one not quite unplausible line, stll #18:

1.e4 g6 2.d4 a6 3.d5 f6:


[fen]1n2k1n1/1pppp2p/p4pp1/3P4/4P3/8/PPP2PPP/RNBQKBNR w KQ -[/fen]


Engine: SlowChess Blitz 2.9 avx20 (32768 MB)
by Jonathan Kreuzer

37/43 0:04 +M15 4.d6 cxd6 5.Qd5 e6 6.Qxd6 Nc6 7.Nc3 b6
8.Be3 Nge7 9.Bh6 g5 10.O-O-O Kf7
11.Qxd7 Ne5 12.Qd8 Kg6 13.Qxe7 Kxh6
14.h4 Nd3+ 15.cxd3 Kh5 16.Qf7+ Kh6
17.Qxf6+ (71.806.867) 17137

Instead of 4...f6 ...d6
1.e4 g6 2.d4 a6 3.d5 d6 4.Nc3 Nf6 5.e6 dxe6 6.Nf3 e4 7.Ne5 Nfd7:


[fen]1n2k3/1ppnpp1p/p5p1/3PN3/4p3/2N5/PPP2PPP/R1BQKB1R w KQ -[/fen]

Engine: SlowChess Blitz 2.9 avx20 (32768 MB)
by Jonathan Kreuzer

29/37 0:13 +M11 8.Nxd7 Nxd7 9.Nxe4 Kd8 10.d6 cxd6
11.Bf4 f6 12.Nxd6 e6 13.Nf7+ Kc8
14.Qd6 e5 15.O-O-O exf4 16.Qxd7+ Kb8
17.Nd8 Ka7 18.Qxb7+ (257.556.575) 18582

Still #18 all in all but not as easily and quickly to be kept in hash at Backward of course, especially because of 7...e3 instead of ...Nfd7 (?!)

[Uri Blass]

If you want to prove mate in 18 then first step should be proving mate in at most 18 moves against the bad lines

After 1.e4 d5 I did not find mate in 18 but only mate in 19 that means mate in 18 after 1.e4 d5.
My line is 1.e4 d5 2.Qg4 Nd7 3.Qxg7 Ngf6 4.exd5 Kd8 5.Qxf7 Ne5 6.Qf8+ Ne8 7.Bb5 c6 8.d6 Ng6 9.dxe7+ Kd7 10.Be2 h5 and here 11.Qf5+ is mate in 9


I guess that there is a mate in at most 18 moves maybe with a different line for white but I doubt if there is a mate in 18 if black is going to play the best defence.

[peter]

If you want to prove mate in 18 then first step should be proving mate in at most 18 moves against the bad lines

After 1.e4 d5 I did not find mate in 18 but only mate in 19 that means mate in 18 after 1.e4 d5.
My line is 1.e4 d5 2.Qg4 Nd7 3.Qxg7 Ngf6 4.exd5 Kd8 5.Qxf7 Ne5 6.Qf8+ Ne8 7.Bb5 c6 8.d6 Ng6 9.dxe7+ Kd7 10.Be2 h5 and here 11.Qf5+ is mate in 9


I guess that there is a mate in at most 18 moves maybe with a different line for white but I doubt if there is a mate in 18 if black is going to play the best defence.

I'd rather try 2.exd4 after 1.e4 d5 (!?, I'd doubt that to be one of the best defences yet also, even if I so far didn't find shorter than 18 moves long lines neither. Haven't much hardware- time free at the moment)

1. e2-e4 d7-d5 2. e4xd5 Ng8-f6 3. Qd1-f3 Ke8-d8 4. d5-d6 Nb8-c6 5. d6xe7+ Nc6xe7 6. d2-d4 Ne7-c8 :

SlowChess Blitz 2.9 avx2 (32768 MB hash, 2 threads)
FEN: 2nk4/ppp2ppp/5n2/8/3P4/5Q2/PPP2PPP/RNB1KBNR w HA - 1 7

[31] M12 00:09.7 8446K Bg5 Nd6 Bxf6+ gxf6 Qxf6+ Kc8 Nf3 a6 Nc3 h5 Qh8+ Kd7 Ne5+ Ke6 Bc4+ Kf5 Nd5 Nxc4 Qxh5+ Ke6 Qxf7+ Kd6 Qd7#

Edit: after 6...Nc6 (instead of ..Nc8)

1. e2-e4 d7-d5 2. e4xd5 Ng8-f6 3. Qd1-f3 Ke8-d8 4. d5-d6 Nb8-c6 5. d6xe7+ Nc6xe7 6. d2-d4 Ne7-c6:

SlowChess Blitz 2.9 avx2 (32768 MB hash, 31 threads)
FEN: 3k4/ppp2ppp/2n2n2/8/3P4/5Q2/PPP2PPP/RNB1KBNR w HA - 1 7

[31] M12 00:09.1 159.68M Bg5 Kc8 Bxf6 gxf6 Qxf6 a6 Nc3 Kb8 Qxf7 Nd8 Qd7 Ne6 Qxe6 c6 Na4 Ka7 Nc5 Kb6 Qe7 Ka5 Qxb7 h6 b4#

Edit edit: after 6...h6 instead of ...Nc6

1. e2-e4 d7-d5 2. e4xd5 Ng8-f6 3. Qd1-f3 Ke8-d8 4. d5-d6 Nb8-c6 5. d6xe7+ Nc6xe7 6. d2-d4 h7-h6:

SlowChess Blitz 2.9 avx2 (32768 MB hash, 31 threads)
FEN: 3k4/ppp1npp1/5n1p/8/3P4/5Q2/PPP2PPP/RNB1KBNR w HA - 0 7

[31] M12 00:04.3 75.92M Qxb7 g5 Qb8+ Kd7 Nf3 Ng6 Ne5+ Nxe5 dxe5 Ne8 Bb5+ c6 Qb7+ Nc7 Qxc6+ Kc8 Nc3 Kb8 Ne4 Nxb5 Nc5 Nd6 Nd7#
Having had some ponder- time there then you can at least go backwards for one full move to keep the correct DTM (18 all in all) in hash and output, edit- time will be over soon now.

[peter]

edit- time will be over soon now.

So it was, just one more output, already two moves more backwards:

1. e2-e4 d7-d5 2. e4xd5 Ng8-f6 3. Qd1-f3 Ke8-d8 4. d5-d6 Nb8-c6

SlowChess Blitz 2.9 avx2 (32768 MB hash, 31 threads)
FEN: 3k4/ppp1pppp/2nP1n2/8/8/5Q2/PPPP1PPP/RNB1KBNR w HA - 1 5

[35] M14 00:13.5 240.73M dxe7+ Kxe7 Bb5 Kd7 Bxc6+ bxc6 d4 Ke7 Qxc6 Kf8 Qxc7 h6 Qd8+ Ne8 Nf3 f6 O-O Kf7 Nh4 Nd6 Re1 Ne4 Rxe4 f5 Re7+ Kf6 Qf8#

I'm outa here now for a while, Uri, anyhow there won't be any prove for a shortest DTM as long as you can't keep at least any DTM in hash and output backwards to starting position, if single lines of your own interactive trials aren't proving enough for you, you'll have to go on giving your best ones and I'll try to find detours within them as soon as I have hardware- time free for it.

One thing I guess you will have seen already: engine has to be shown again and again pawn- moves and moves of light figures preparing sacs, these two kind of moves engines don't get into mate search of their own at that position (kind of non- practical anyhow yet too, so "normal" mate search doesn't work as well as with in practical game playing normally arising ones, and then there is this very big search tree exploding after first 2 to 3 moves, mate search with longer DsTM always depend on forced moves and lines, here are too litle of them at the earlier follow- up positions, that makes it so hard for engines, of course for human player as well, but here kind of strategy obviously helps to get pawns and light figures forward more then the engine would do so on its own. I'm somewhat more used to dealing with this kind of puzzle because of my long and many trials with the other one Strangmüller- position with Bishops instead of Knights.

[peter]

Just one more

After 1.e4 d5(?!) 2.e5 could well be even shorter than exd5 (!?)

1. e2-e4 d7-d5 2. e4-e5 d5-d4 3. Qd1-g4 Nb8-c6

SlowChess Blitz 2.9 avx2 (32768 MB hash, 31 threads)
FEN: 4k1n1/ppp1pppp/2n5/4P3/3p2Q1/8/PPPP1PPP/RNB1KBNR w HA - 2 4

[31] M12 00:19.1 353.37M Bb5 Kd8 Qxg7 Kd7 Qxg8 a6 Bxc6+ Kxc6 Qc8 d3 cxd3 Kd5 Nf3 Kc6 Nc3 h6 Na4 Kb5 Qxc7 Kxa4 Qb6 h5 b3#

#15 all in all

But here's 3...Nc6 and 2...d4 are to be doubted yet still, the pawn- move I simply tried to make White's d4 impossible.

Edit: backwards to postion after 2...d4:

SlowChess Blitz 2.9 avx2 (32768 MB hash, 31 threads)
FEN: 1n2k1n1/ppp1pppp/8/4P3/3p4/8/PPPP1PPP/RNBQKBNR w HA - 0 3

[37] M15 01:09.3 1226.28M Qg4 Nd7 Qxg7 Nh6 Qxh6 c6 Qxh7 Kd8 Qg8+ Kc7 Qxf7 Kd8 e6 Nf6 Qf8+ Ne8 d3 b5 Bf4 c5 Be2 c4 Bh5 Kc8 Qxe8+ Kb7 Qxb5+ Kc8 Qb8#

So at least not more than #17 in all for this one line.

[towforce]

These postions are some of the most difficult ones for engines to count best DTM because of the many near to each other moves along the lines, search- tree explodes from the very start. Specialised mate search programs like Gustav or Chest don't stand any chance neither, the are based on reduction of search trees, which doesn't work for such positions at all.

Go for the universal solution: uncover the immutable rules of chess which will enable you to quickly work out the distance to mate from any position.

[peter]

These postions are some of the most difficult ones for engines to count best DTM because of the many near to each other moves along the lines, search- tree explodes from the very start. Specialised mate search programs like Gustav or Chest don't stand any chance neither, the are based on reduction of search trees, which doesn't work for such positions at all.

Go for the universal solution: uncover the immutable rules of chess which will enable you to quickly work out the distance to mate from any position.

According to the uncovered rules this one's mate in 18