Anti-Engine Puzzle

[xr_a_y]

[fen]8/p3r2k/Pp2q3/2pQ4/K1P2p2/1P6/7P/3R4 w - - 56 89[/fen]

How fast can your engine find the brilliant move in this position ?

Is it Rd2 ?
Minic says that at depth 40 after some minutes.

[peter]

[fen]8/p3r2k/Pp2q3/2pQ4/K1P2p2/1P6/7P/3R4 w - - 56 89[/fen]

How fast can your engine find the brilliant move in this position ?

Is it Rd2 ?
Minic says that at depth 40 after some minutes.

You might be right, Rd2 looks best (at least most beautiful) even if it's not the single one game changer (which are less and less easy to be found as for test postions hard enough for modern software- hardware- combis, just concerning Jouni's question of his other one thread).
After some Forward- Backward in depth 34:

8/p3r2k/Pp2q3/2pQ4/K1P2p2/1P6/7P/3R4 w - - 0 1

Analysis by Crystal 251221:

1. +- (4.79): 89.Rd2 Qxd5 90.cxd5 Rf7 91.d6 f3 92.Rf2 Rf6 93.d7 Rd6 94.Rxf3 Rxd7 95.Kb5 Rc7 96.Rf8 Kg6 97.Rb8 c4 98.bxc4 Rc5+ 99.Kb4 Ra5 100.c5 Rxc5 101.Rb7 Rf5 102.Rxa7 Rf4+ 103.Kb3 Rf3+ 104.Kb2 Rf2+ 105.Kb1 Rf1+ 106.Kc2 Ra1 107.Kb3 Kh6 108.Kb4 Ra5 109.Ra8 Kg7 110.a7 Kh7 111.h4 Kg7 112.h5 Kh7 113.h6 Ra2 114.Kb5 Rb2+ 115.Kc6 Rc2+ 116.Kb7

2. +- (4.45): 89.Qh5+ Qh6 90.Qf3 Qf6 91.Qh3+ Qh6 92.Qd3+ Qg6 93.Qd5 Re6 94.Ka3 Rf6 95.Qf3 Rd6 96.Re1 Rf6 97.Re7+ Rf7 98.Qxf4 Rxe7 99.Qh4+ Kg8 100.Qxe7 Qb1 101.Qg5+ Kh7 102.Qh4+ Kg6 103.Qg3+ Kh5 104.Qf3+ Kh4 105.Ka4 Qe1 106.Kb5 Qe6 107.Qg2 Qe8+ 108.Qc6 Qe3 109.Qf6+ Kh3 110.Kc6 Qxb3 111.Qe6+ Kg2 112.Qd5+ Kf1 113.h4 b5 114.cxb5 c4 115.h5 Qc3 116.Qf7+ Ke2 117.b6 Qb4 118.Qe6+ Kf1 119.Qf6+ Kg2 120.bxa7 Qa4+ 121.Kc7 Qa5+ 122.Qb6 Qe5+ 123.Qd6

3. +- (4.39): 89.Qf3 Qf6 90.Qh3+ Qh6 91.Qf5+ Qg6 92.Qd5 Re6 93.Ka3 Re7 94.Qf3 Qf6 95.Ka4 Rf7 96.Kb5 Qe6 97.Qh5+ Kg8 98.Qg5+ Kh7 99.Qh4+ Qh6 100.Qg4 Qg6 101.Qh3+ Kg7 102.Rd8 Rf8 103.Qd7+ Rf7 104.Qe8 Qf6 105.Qh8+ Kg6 106.Rg8+ Rg7 107.Rc8 Qc3 108.Rc6+ Kf5 109.Qh5+ Rg5 110.Qh7+ Kg4 111.Qd7+ Kh5 112.Rd6 Qxb3+ 113.Kc6 Qf3+ 114.Kc7 Kh4 115.Rf6 b5 116.cxb5 c4 117.b6 c3 118.b7 c2 119.Qh7+ Qh5 120.Rxf4+ Rg4 121.Rxg4+ Kxg4 122.Qe4+ Kh3 123.Qe3+ Kg2 124.b8Q Qh7+ 125.Kc6

30 threads of Ryzen CPU, 32G hash, 6men Syzygys regards

[Joerg Oster]

[fen]8/p3r2k/Pp2q3/2pQ4/K1P2p2/1P6/7P/3R4 w - - 56 89[/fen]

How fast can your engine find the brilliant move in this position ?

Is it Rd2 ?
Minic says that at depth 40 after some minutes.

You might be right, Rd2 looks best (at least most beautiful) even if it's not the single one game changer (which are less and less easy to be found as for test postions hard enough for modern software- hardware- combis, just concerning Jouni's question of his other one thread).
After some Forward- Backward in depth 34:

8/p3r2k/Pp2q3/2pQ4/K1P2p2/1P6/7P/3R4 w - - 0 1

Analysis by Crystal 251221:

1. +- (4.79): 89.Rd2 Qxd5 90.cxd5 Rf7 91.d6 f3 92.Rf2 Rf6 93.d7 Rd6 94.Rxf3 Rxd7 95.Kb5 Rc7 96.Rf8 Kg6 97.Rb8 c4 98.bxc4 Rc5+ 99.Kb4 Ra5 100.c5 Rxc5 101.Rb7 Rf5 102.Rxa7 Rf4+ 103.Kb3 Rf3+ 104.Kb2 Rf2+ 105.Kb1 Rf1+ 106.Kc2 Ra1 107.Kb3 Kh6 108.Kb4 Ra5 109.Ra8 Kg7 110.a7 Kh7 111.h4 Kg7 112.h5 Kh7 113.h6 Ra2 114.Kb5 Rb2+ 115.Kc6 Rc2+ 116.Kb7

2. +- (4.45): 89.Qh5+ Qh6 90.Qf3 Qf6 91.Qh3+ Qh6 92.Qd3+ Qg6 93.Qd5 Re6 94.Ka3 Rf6 95.Qf3 Rd6 96.Re1 Rf6 97.Re7+ Rf7 98.Qxf4 Rxe7 99.Qh4+ Kg8 100.Qxe7 Qb1 101.Qg5+ Kh7 102.Qh4+ Kg6 103.Qg3+ Kh5 104.Qf3+ Kh4 105.Ka4 Qe1 106.Kb5 Qe6 107.Qg2 Qe8+ 108.Qc6 Qe3 109.Qf6+ Kh3 110.Kc6 Qxb3 111.Qe6+ Kg2 112.Qd5+ Kf1 113.h4 b5 114.cxb5 c4 115.h5 Qc3 116.Qf7+ Ke2 117.b6 Qb4 118.Qe6+ Kf1 119.Qf6+ Kg2 120.bxa7 Qa4+ 121.Kc7 Qa5+ 122.Qb6 Qe5+ 123.Qd6

3. +- (4.39): 89.Qf3 Qf6 90.Qh3+ Qh6 91.Qf5+ Qg6 92.Qd5 Re6 93.Ka3 Re7 94.Qf3 Qf6 95.Ka4 Rf7 96.Kb5 Qe6 97.Qh5+ Kg8 98.Qg5+ Kh7 99.Qh4+ Qh6 100.Qg4 Qg6 101.Qh3+ Kg7 102.Rd8 Rf8 103.Qd7+ Rf7 104.Qe8 Qf6 105.Qh8+ Kg6 106.Rg8+ Rg7 107.Rc8 Qc3 108.Rc6+ Kf5 109.Qh5+ Rg5 110.Qh7+ Kg4 111.Qd7+ Kh5 112.Rd6 Qxb3+ 113.Kc6 Qf3+ 114.Kc7 Kh4 115.Rf6 b5 116.cxb5 c4 117.b6 c3 118.b7 c2 119.Qh7+ Qh5 120.Rxf4+ Rg4 121.Rxg4+ Kxg4 122.Qe4+ Kh3 123.Qe3+ Kg2 124.b8Q Qh7+ 125.Kc6

30 threads of Ryzen CPU, 32G hash, 6men Syzygys regards

I have some doubts.
I get rather drawish scores for your 3 moves, btw.
Don't forget it's an anti-engine puzzle.

As a human it looks very tempting to try to promote the a-pawn through a capture on b7.

[ImNotStockfish]

[fen]1r2b3/5rbk/n2p2p1/2pPp1Pp/p1P1P2q/R4P2/1P2NR2/KB1N1Q2 w - - 0 163[/fen]

This one is a draw... or is it?

[Paloma]

Not a New,

FIRST: What is the solution of the previous puzzle ?

I don't find a *brilliant move*

[peter]

I have some doubts.
I get rather drawish scores for your 3 moves, btw.
Don't forget it's an anti-engine puzzle.

As a human it looks very tempting to try to promote the a-pawn through a capture on b7.

To discuss that on a somewhat more concrete base, we'd need two things, Jörg: your drawish lines you get drawish scores with and another one line to show a win with a capture on b7 regards

[Paloma]

Btw,
i have here with SF11 :

1.Rd2 b5+ 2.cxb5 Qxd5 3.Rxd5 Kg6 4.Rd6+ Kf5 5.b6 axb6 6.Kb5 f3 7.Rd1 Kf4 8.Rf1 Rf7 9.Kxb6 Rf6+ 10.Kb5 Rf7 11.Kxc5 Rf5+ ....
+- (4.64) Tiefe: 64/132 05:36:20 152267MN, tb=81077459

but is that a *brilliant move* ?
I think there is no brilliant here. ImNotStockfish?

[Joerg Oster]

I have some doubts.
I get rather drawish scores for your 3 moves, btw.
Don't forget it's an anti-engine puzzle.

As a human it looks very tempting to try to promote the a-pawn through a capture on b7.

To discuss that on a somewhat more concrete base, we'd need two things, Jörg: your drawish lines you get drawish scores with and another one line to show a win with a capture on b7 regards

Dear Peter,

to get drawish scores is most likely a matter of using the correct fen.
Please note the 50-move and full move counters in the original fen!

Here a quick MultiPV=3 search up to depth 24 with Crystal 4, 1 GB Hash, 3 Threads, Syzygy6 (8/p3r2k/Pp2q3/2pQ4/K1P2p2/1P6/7P/3R4 w - - 56 89):

Code: Select all

info depth 24 seldepth 56 multipv 1 score cp 77 nodes 54991979 nps 2096529 hashfull 353 tbhits 18879 time 26230 pv d5h5 e6h6 h5f3 h6f6 d1f1 e7f7 f3h5 h7g8 f1g1 f7g7 h5d5 g8h7 g1e1 f6f7 d5e4 g7g6 e1f1 f7d7 a4a3 d7d2 h2h4 h7g7 e4f4 d2a5 a3b2 g6g2 f1f2 g2f2 f4f2 a5a6 f2g3 g7h6 g3g5 h6h7 g5h5 h7g7 h5e5 g7h7 e5e4 h7h8 e4e8 h8h7 h4h5 a6b7 e8g6 h7h8 h5h6
info depth 24 seldepth 60 multipv 2 score cp 70 nodes 54991979 nps 2096529 hashfull 353 tbhits 18879 time 26230 pv d5f3 e6f6 d1f1 e7f7 f3h5 h7g8 f1g1 f7g7 h5d5 g8h7 g1e1 g7g5 d5d7 g5g7 d7d3 g7g6 a4a3 f6d6 d3h3 h7g7 e1e8 g6h6 h3f5 h6f6 f5g5 f6g6 g5e7 d6e7 e8e7 g7f6 e7a7
info depth 24 seldepth 50 multipv 3 score cp 70 nodes 54991979 nps 2096529 hashfull 353 tbhits 18879 time 26230 pv d5d3 e6g6 d3f3 g6f6 d1f1 e7f7 f3h5 h7g8 f1g1 f7g7 h5d5 g8h7 g1e1 g7g5 d5d7 g5g7 d7d3 g7g6 a4a3 h7h6 e1f1 g6g5 d3d7 f6c3 d7d6 h6g7 d6f4 c3a5 a3b2 g5g2 f1f2 g2f2 f4f2 a5a6 f2g3 g7f6 g3d6 f6g5 d6e7 g5g6 e7e6 g6g5 h2h3 a6a5 e6g8 g5h5 g8g4 h5h6
bestmove d5h5 ponder e6h6

Regarding the pawn promotion, that was just an idea from a human point of view, and from a lousy player.

[peter]

to get drawish scores is most likely a matter of using the correct fen.
Please note the 50-move and full move counters in the original fen!

Here a quick MultiPV=3 search up to depth 24 with Crystal 4, 1 GB Hash, 3 Threads, Syzygy6 (8/p3r2k/Pp2q3/2pQ4/K1P2p2/1P6/7P/3R4 w - - 56 89):

Code: Select all

info depth 24 seldepth 56 multipv 1 score cp 77 nodes 54991979 nps 2096529 hashfull 353 tbhits 18879 time 26230 pv d5h5 e6h6 h5f3 h6f6 d1f1 e7f7 f3h5 h7g8 f1g1 f7g7 h5d5 g8h7 g1e1 f6f7 d5e4 g7g6 e1f1 f7d7 a4a3 d7d2 h2h4 h7g7 e4f4 d2a5 a3b2 g6g2 f1f2 g2f2 f4f2 a5a6 f2g3 g7h6 g3g5 h6h7 g5h5 h7g7 h5e5 g7h7 e5e4 h7h8 e4e8 h8h7 h4h5 a6b7 e8g6 h7h8 h5h6
info depth 24 seldepth 60 multipv 2 score cp 70 nodes 54991979 nps 2096529 hashfull 353 tbhits 18879 time 26230 pv d5f3 e6f6 d1f1 e7f7 f3h5 h7g8 f1g1 f7g7 h5d5 g8h7 g1e1 g7g5 d5d7 g5g7 d7d3 g7g6 a4a3 f6d6 d3h3 h7g7 e1e8 g6h6 h3f5 h6f6 f5g5 f6g6 g5e7 d6e7 e8e7 g7f6 e7a7
info depth 24 seldepth 50 multipv 3 score cp 70 nodes 54991979 nps 2096529 hashfull 353 tbhits 18879 time 26230 pv d5d3 e6g6 d3f3 g6f6 d1f1 e7f7 f3h5 h7g8 f1g1 f7g7 h5d5 g8h7 g1e1 g7g5 d5d7 g5g7 d7d3 g7g6 a4a3 h7h6 e1f1 g6g5 d3d7 f6c3 d7d6 h6g7 d6f4 c3a5 a3b2 g5g2 f1f2 g2f2 f4f2 a5a6 f2g3 g7f6 g3d6 f6g5 d6e7 g5g6 e7e6 g6g5 h2h3 a6a5 e6g8 g5h5 g8g4 h5h6
bestmove d5h5 ponder e6h6

I used the fen from the original posting as you can see from my Crystal- output, the fen cb- GUI gives is just a translation, move counter's correct.
And then you get drawish scores also of course with drawish lines, which will not be the ones with best moves here, as you can see on your own by comparing your output of Crystal's and mine for more then the first few plies only.

You simply cannot expect Crystal or any other engine here to find best moves along the output- lines with a short- time MultiPV- search only.
As I wrote, at my output Forward- Backward was done before stand- alone- pondering, try a Backward of the first one line (Rd2) only of at least 5 moves with enough hardware- time and you'*ll see, what happens.
Then you can try the same with my second one MultiPV- output- line, If you still get drawish lines and drawish evals only, your hardware- time still wasn't enough.

Edit: E.g. try first deviation for White in your Qh5- line with 91.Qh3+ instead of Rf1 and then give in the following moves also:

(89.Dh5+ Dh6 90.Df3 Df6) 91.Dh3+ Dh6 92.Df3 Df6 93.Dh3+ Dh6 94.Df5+ Dg6 95.Dd5 Tf7

That should be enough to see evals climb at single primary line of Crystal's with 3 good cores, 4G hash and maybe 3 minutes. Then go backward at to 91.Qh3+ again, if you want to see, that the winning eval is kept in hash with enough hardware- time from ply to ply agin, and you should get output- lines with eval of clear +- like this one here then too:

8/p3r2k/Pp3q2/2p5/K1P2p2/1P5Q/7P/3R4 b - - 0 1

Analysis by Crystal 251221:

91...Qh6 92.Qd3+ Qg6 93.Qd5 Qe8+ 94.Ka3 Qg6 95.Qf3 Qf6 96.Ka4 Rf7 97.Kb5 Qe6 98.Qh5+ Kg8 99.Qg5+ Kh7 100.Qh4+ Qh6 101.Qg4 Qg6 102.Qh3+ Kg7 103.Qf3 Qf5 104.Ka4 Kh6 105.Rd6+ Rf6 106.Rd8 Rf8 107.Rd5 Qg6 108.Rd2 Qf5 109.Rg2 b5+ 110.cxb5 Re8 111.Rd2 Re4+ 112.Ka3 c4 113.Rd6+ Kg7 114.Qf2 Re3 115.Qg2+ Kf7 116.Qb7+ Re7 117.Qd5+ Qxd5 118.Rxd5 cxb3 119.Rd4 f3 120.Rf4+ Ke6 121.Rxf3 Rh7 122.Kxb3 Kd6 123.Rc3 Rxh2
+- (2.38) Depth: 33/94 00:01:14 3033MN, tb=2525949

Happy New Year regards

[xr_a_y]

[fen]1r2b3/5rbk/n2p2p1/2pPp1Pp/p1P1P2q/R4P2/1P2NR2/KB1N1Q2 w - - 0 163[/fen]

This one is a draw... or is it?

Looks like Minic is trying to trap the black queen, starting with Qg2. This will allow the f pawn to be pushed several time and finally break trough forcing a bishop sac. Not sure this is "the" way...