SF 17 needs depth 25 to solve a two-mover.

[Ajedrecista]

Hello:

I composed a two-mover based on other interesting-to-me 5+0 game that I lost on time, where I missed some checkmate in N moves in an interesting king hunt. I did not save the game, by the way.

I rearranged the white knight (on c3 in the original game) and added other black pieces to increase the difficulty for engines. The surprise came when SF 17 needed depth 25 (!) to find the correct solution (there are two solutions indeed). First of all, here is the position:

[d]2rr4/3R3n/b4p1p/8/2p1R1P1/pp3kNP/5P1n/6K1 w - - 0 1

I friendly challenge you to solve the problem on your own before running any engine.

.
.
.

After all of you have solved it, this is the time to say that the conditions of SF 17 search were single core and Multi-PV disabled, to get deterministic results:

Code: Select all

FEN: 2rr4/3R3n/b4p1p/8/2p1R1P1/pp3kNP/5P1n/6K1 w - - 0 1

Stockfish-17-windows-x86-64-sse41-popcnt:
Found 145 WDL and 145 DTZ tablebase files (up to 5-man).
Available processors: 0-3
Using 1 thread
NNUE evaluation using nn-1111cefa1111.nnue (133MiB, (22528, 3072, 15, 32, 1))
NNUE evaluation using nn-37f18f62d772.nnue (6MiB, (22528, 128, 15, 32, 1))
  1/4	00:00	          44	22.000	-3,08	Rd7xd8 Rc8xd8
[...]
  5/4	00:00	         191	95.500	-3,08	Rd7xd8 Rc8xd8
  6/8	00:00	         274	137.000	-2,89	Re4e3+ Kf3f4 Re3e4+ Kf4g5
  7/8	00:00	         338	169.000	-2,93	Re4e3+ Kf3f4 Re3e4+ Kf4g5
  8/11	00:00	         621	207.000	-1,75	Re4e3+ Kf3f4 Ng3e2+ Kf4g5 Rd7g7+ Kg5h4 Kg1xh2
  9/10	00:00	         696	232.000	-1,88	Re4e3+ Kf3f4 Ng3e2+ Kf4g5 Rd7g7+ Kg5h4 Kg1xh2 c4c3 Ne2xc3
 10/12	00:00	       2.372	338.857	-0,94	Re4e3+ Kf3f4 Re3e4+ Kf4g5 Rd7g7+ Kg5h4
 11/10	00:00	       2.872	359.000	 0,00	Re4e3+ Kf3f4 Re3e4+ Kf4f3 Re4e3+
 12/5	00:00	       2.961	370.125	 0,00	Re4e3+ Kf3f4 Re3e4+ Kf4f3
[...]
 23/5	00:00	       6.392	456.571	 0,00	Re4e3+ Kf3f4 Re3e4+ Kf4f3
 24/14	00:00	      10.383	432.625	+2,68	Re4e3+ Kf3f4 Re3e4+ Kf4f3 Re4e5 Rc8c5 Re5xc5 c4c3 Rc5xc3+ Kf3f4 Rd7xd8
 25/20	00:00	      15.762	404.153	+M3	Re4e5 Rc8c5 Re5xc5 c4c3 Rc5f5+
 26/6	00:00	      16.432	410.800	+M3	Re4e5 Rc8c5 Re5xc5 c4c3 Rc5f5+
 27/6	00:00	      17.075	426.875	+M3	Re4e5 Rc8c5 Re5xc5 c4c3 Rc5f5+
[...]
 244/6	00:01	     277.733	263.004	+M3	Re4e5 Rc8c5 Re5xc5 c4c3 Rc5f5+
 245/6	00:01	     278.907	264.116	+M3	Re4e5 Rc8c5 Re5xc5 c4c3 Rc5f5+

[pgn][Event ""]
[Site "Madrid (Spain)"]
[Round ""]
[Date "2025.10.15"]
[SetUp "1"]
[FEN "2rr4/3R3n/b4p1p/8/2p1R1P1/pp3kNP/5P1n/6K1 w - - 0 1"]
[PlyCount "5"]
[Result "1-0"]

1. Re5 fxe5 (1. ... Rc5 (1. ... Rxd7 2. Rf5#) 2. Rxc5 Rxd7 {(any)} 3. Rf5#) 2. Rf7+ Nf6 3. Rxf6# 1-0[/pgn]

OTOH, here is AnMon 5.60 from year 2006:

Code: Select all

FEN: 2rr4/3R3n/b4p1p/8/2p1R1P1/pp3kNP/5P1n/6K1 w - - 0 1

AnMon 5.60:
   1+	00:00	           3	0	-7,54	Rd7xh7
   1	00:00	           4	0	-7,54	Rd7xh7
   1+	00:00	          18	0	-7,53	Re4e3+ Kf3f4 Ng3h5+ Kf4g5 Rd7g7+ Kg5h4 Rg7xh7
   1	00:00	          84	0	-7,43	Re4e3+ Kf3f4 Rd7xh7
   2	00:00	         391	0	-7,35	Re4e3+ Kf3f4 Ng3h5+ Kf4g5 f2f4+ Kg5h4 Rd7xh7
   2+	00:00	       1.520	0	-7,24	Re4e5 Ba6b7 Rd7xb7 Rd8d1+ Kg1xh2 f6xe5 Rb7xh7 Kf3xf2 Rh7xh6
   2	00:00	       1.810	0	-4,90	Re4e5 Rc8c5 Re5xc5
   3+	00:00	       2.565	0	+M3	Re4e5 Rc8c5 Re5xc5 Ba6c8 Rc5f5+
   3	00:00	       2.603	0	+M3	Re4e5 Rc8c5 Re5xc5 Ba6c8 Rc5f5+

I guess other old engines solve the position with low serch depths.

------------

I had lost two games against the same player before and I felt he/she was stronger than me. I sacrificed a piece for a pawn in the third game, attempting a threefold repetition that was enough for me. There was not a threefold repetition and a funny king hunt began. Once again, a change of plans led to interesting motifs, like here. Digging deep in my memory, I had reconstructed the game completely after hours of playing it, quite a feat for me:

[pgn][Event ""]
[Site ""]
[Round ""]
[Date "2025.10.15"]
[PlyCount "80"]
[Result "0-1"]

1. d4 d5 2. c4 e6 3. Nc3 Nf6 4. Bg5 c6 5. cxd5 exd5 6. Nf3 Bd6 7. e3 O-O 8. Bd3 Nbd7 9. Rc1 b6 10. Na4 Bb7 11. Qe2 a5 12. O-O Re8 13. a3 Nf8 14. Bxf6 gxf6 15. Nh4 Ng6 16. Qg4 Bc8 17. Qh5 Bd7 18. Bxg6 fxg6 19. Nxg6 hxg6 20. Qxg6+ Kf8 21. Qh6+ Kf7 22. Qh7+ Ke6 23. e4 Rh8 24. exd5+ Kxd5 25. Nc3+ Kc4 26. Qf7+ Kxd4 27. Rcd1+ Ke5 28. Rfe1+ Kf4 29. Rxd6 Rf8 30. Qxd7 Qxd7 31. Rxd7 Rac8 32. h3 Kf5 33. g4+ Kf4 34. Rg7 Kf3 35. Rd7 Rcd8 (35. ... Rfd8 36. Re5) 36. Rxd8 Rxd8 37. Re3+ Kf4 38. Kg2 Rd2 39. Ne2+ Kg5 40. Kg3 Rxb2 {(white lost on time)} 0-1[/pgn]

I played Rd7 (from g7), threatening Rd3# and black instantly played ..., Rcd8 instead of ..., Rfd8 (a losing move that gave me the idea to compose the problem). It is needless to say that I would have never found the key move after ..., Rfd8.

Regards from Spain.

Ajedrecista.

[RubiChess]

Calling mate-in-three a two-mover took me 15 minutes of my life.

[Ajedrecista]

Hello Andreas:

Calling mate-in-three a two-mover took me 15 minutes of my life.

Yes, my bad. I hope you have found the key move and enjoyed the problem despite my typo.

Still quite surprising for me that SF 17 needs to search up to depth 25 to find a checkmate in three moves. Too much pruning on this particular case, which usually helps in real gameplay.

Regards from Spain.

Ajedrecista.

[flok]

Code: Select all

info depth 1 score cp -2804 nodes 72 time 0 nps 72000 tbhits 0 hashfull 0 pv e4e3
info depth 2 score cp -2450 nodes 210 ebf 1.312335 time 0 nps 210000 tbhits 0 hashfull 0 pv e4e3 f3f4 g3e2
info depth 3 score cp -2386 nodes 459 ebf 4.217311 time 0 nps 459000 tbhits 0 hashfull 0 pv e4e3 f3f4 g3e2 f4g5 d7g7
info depth 4 score cp -2256 nodes 1185 ebf 1.611953 time 1 nps 1185000 tbhits 0 hashfull 0 pv e4e3 f3f4 g3e2 f4g5 d7h7 a3a2
info depth 5 score cp -2219 nodes 2227 ebf 3.631787 time 2 nps 1113500 tbhits 0 hashfull 0 pv e4e3 f3f4 g3e2 f4g5 d7h7 a3a2 h7g7
info depth 6 score cp -2132 nodes 3822 ebf 1.047802 time 3 nps 1274000 tbhits 0 hashfull 0 pv e4e3 f3f4 g3e2 f4g5 d7g7 g5h4 g7h7 a3a2 h7h6
info depth 7 score cp 0 nodes 6742 ebf 0.553122 time 6 nps 1123666 tbhits 0 hashfull 0 pv e4e3 f3f4 e3e4 f4f3
info depth 8 score cp 0 nodes 6987 ebf 0.308757 time 6 nps 1164500 tbhits 0 hashfull 0 pv e4e3 f3f4 e3e4 f4f3
info depth 9 score cp 0 nodes 8866 ebf 2.317018 time 8 nps 1108250 tbhits 0 hashfull 0 pv e4e3 f3f4 e3e4 f4f3
info depth 10 score cp 0 nodes 12080 ebf 3.621929 time 10 nps 1208000 tbhits 0 hashfull 0 pv e4e3 f3f4 e3e4 f4f3
info depth 11 score cp 0 nodes 16785 ebf 1.582401 time 14 nps 1198928 tbhits 0 hashfull 0 pv e4e3 f3f4 e3e4 f4f3
info depth 12 score cp 0 nodes 24239 ebf 1.522901 time 18 nps 1346611 tbhits 0 hashfull 0 pv e4e3 f3f4 e3e4 f4f3
info depth 13 score cp 0 nodes 44529 ebf 2.076640 time 26 nps 1712653 tbhits 0 hashfull 2 pv e4e3 f3f4 e3e4 f4f3
info depth 14 score cp 0 nodes 58710 ebf 1.379300 time 30 nps 1957000 tbhits 0 hashfull 2 pv e4e3 f3f4 e3e4 f4f3
info depth 15 score cp 0 nodes 105844 ebf 1.524144 time 46 nps 2300956 tbhits 0 hashfull 9 pv e4e3 f3f4 e3e4 f4f3
info depth 16 score mate 3 nodes 447047 ebf 15.839652 time 155 nps 2884174 tbhits 0 hashfull 39 pv e4e5 c8c5 e5c5 d8d7 c5f5
info depth 17 score mate 3 nodes 509616 ebf 3.847366 time 175 nps 2912091 tbhits 0 hashfull 44 pv e4e5 c8c5 e5c5 d8d7 c5f5
info depth 18 score mate 3 nodes 563608 ebf 0.448045 time 192 nps 2935458 tbhits 0 hashfull 45 pv e4e5 c8c5 e5c5 d8d7 c5f5
info depth 19 score mate 3 nodes 690711 ebf 1.425274 time 233 nps 2964424 tbhits 0 hashfull 48 pv e4e5 c8c5 e5c5 d8d7 c5f5
info depth 20 score mate 3 nodes 902949 ebf 1.982654 time 301 nps 2999830 tbhits 0 hashfull 52 pv e4e5 c8c5 e5c5 d8d7 c5f5
info depth 21 score mate 3 nodes 1129884 ebf 1.336204 time 373 nps 3029179 tbhits 0 hashfull 63 pv e4e5 c8c5 e5c5 d8d7 c5f5
info depth 22 score mate 3 nodes 1634903 ebf 1.542561 time 534 nps 3061616 tbhits 0 hashfull 85 pv e4e5 c8c5 e5c5 d8d7 c5f5
info depth 23 score mate 3 nodes 2514071 ebf 1.968272 time 815 nps 3084749 tbhits 0 hashfull 133 pv e4e5 c8c5 e5c5 d8d7 c5f5
info depth 24 score mate 3 nodes 3504464 ebf 1.400393 time 1135 nps 3087633 tbhits 0 hashfull 173 pv e4e5 c8c5 e5c5 d8d7 c5f5
info depth 25 score mate 3 nodes 5651070 ebf 1.562573 time 1815 nps 3113537 tbhits 0 hashfull 264 pv e4e5 c8c5 e5c5 d8d7 c5f5
info depth 26 score mate 3 nodes 11299122 ebf 2.388062 time 3611 nps 3129083 tbhits 0 hashfull 454 pv e4e5 c8c5 e5c5 d8d7 c5f5
info depth 27 score mate 4 nodes 87422749 ebf 5.955025 time 27440 nps 3185960 tbhits 0 hashfull 998 pv e4e5 c8c5 e5c5 d8d7 c5f5

This is THE DOG (4.10.1).