Mates that appear to be draws

[Dann Corbit]

Hello:

Not sure where this one came from:
[d]qnb5/rp1p4/pPpP4/P1P5/2P4k/5K2/8/5N2 w - -
I can't find a source for it.

Nor can't I, [...]

[...]

I found the almost exact position:

https://pdb.dieschwalbe.de/P1298591
https://yacpdb.org/#284718

[d]qnb5/rp1p4/pPpP4/P1P5/2P5/6K1/8/6Nk w - - 0 1

There are alternative solutions to the original problem according to PDB database. There are tons of variants that can be replayed at YACPDB. The position originally posted in this thread is a variation with an alternative arrangement of the kings and the white knight.

The book is available online at Google Books:

Baltische schachblätter, Issues 1-5

The problem is at Issue 3 of Baltische schachblätter, Problem No. 135 at page 157, by Friedrich Amelung (Original. Dedicated to the Riga Chess Club). Long solutions with variants are given at pages 168 and 169:

No. 135. — Zunächst sei bemerkt, dass die Stellung sich als eine mögliche ergiebt und zwar, indem Schwartz vorher gezogen hatte 1. c6, 2. Sc6—b8, 3. Db8—a8, 4. Ta7—a8, 5. Da7—b8.
Der Gang der Lösung ist nachfolgender, wobei zuerst ein kleines nettes Manöver nöthig wird, um den König aus der Ecke zu vertreiben. Nun sind beiläufig allgemein folgende Sätze für eine derartige Position (wie z. B. hier Kg3, Sg1, Kh1) aufzustellen: 1. Der Springer kann den König aus der Ecke vertreiben, 2. der Springer kann den König von jeder Stellung in der Mitte des Brettes an den Rand, dort aber nicht in ein vorher bestimmtes Eckfeld treiben, — wäre z. B. in unserer Aufgabe die Position hier Kf2, Sh2—Kh1, so kann der König zur Ecke a1 hinflüchten und es ist dann kein Mat möglich.
Die Lösung beginnt mit 1. Kf2, Kh2, 2. Se2, Kh1!, 3. Sf4, Kh2 und darauf macht Weiss den Tempozug 4. Kf3!, wonach Kh1! und 5. Ke2, Kg1!, 6. Ke1!!, Kh1, 7. Kf1, Kh2, 8. Kf2, Kh1, 9. Se2, Kh2, 10. Sg3, Kh3, 11. Sf1. — — Nunmehr gilt es, den König bis in die Ecke h8 zu bringen, — es kann natürlich blos das Verfahren gezeigt werden, ohne alle Varianten auszuführen.
[...] TOO LONG TO COPY BY HAND.

No. 135. — First, it should be noted that the position emerges as a possible one, namely because Black had previously played 1. c6, 2. Nc6—b8, 3. Qb8—a8, 4. Ra7—a8, 5. Qa7—b8.
The solution proceeds as follows, beginning with a neat little Maneuver to drive the King out of the corner. Now, the following general principles can be formulated for such a Position (such as Kg3, Ng1, Kh1 here): 1. The Knight can drive the King out of the corner, 2. the Knight can move the King from any position in the center of the board to the edge, but not into a predetermined corner square. For example, if the position in our problem were Kf2, Nh2—Kh1, the King could flee to the corner a1, and mating would then be impossible.
The solution begins with 1. Kf2, Kh2, 2. Ne2, Kh1!, 3. Nf4, Kh2 and then White makes the tempo move 4. Kf3!, followed by Kh1! and 5. Ke2, Kg1!, 6. Ke1!!, Kh1, 7. Kf1, Kh2, 8. Kf2, Kh1, 9. Ne2, Kh2, 10. Ng3, Kh3, 11. Nf1. — — Now the goal is to get the King to the h8 corner, — of course only the procedure can be shown, without executing all the variations.
[...] TOO LONG TO TRANSLATE.

Regards from Spain.

Ajedrecista.

Without a doubt, that is the source of the problem. I see it also is found in HHDB. I will update the note in my position.

[peter]

Without a doubt, that is the source of the problem. I see it also is found in HHDB. I will update the note in my position.

Indeed, here's the .pgn given in HHdbVII:
[pgn][Event "no ty"]
[Site "; Baltische Schachblatter#135"]
[Date "1891.??.??"]
[Round "?"]
[White "Amelung=F"]
[Black "(+3334.54g3h1) U2"]
[Result "1-0"]
[Annotator "Martan,Peter"]
[WhiteFideId "-1"]
[BlackFideId "-1"]
[SetUp "1"]
[FEN "qnb5/rp1p4/pPpP4/P1P5/2P5/6K1/8/6Nk w - - 0 1"]
[PlyCount "109"]
[GameId "2282429966123790"]
[EventDate "1891.??.??"]

{stipulation: mate in 55 source: Baltische Schachblatter#135 no.3 1891; U2: Wiereyn=P HHdbV 13-2-2011} 1. Kf2 Kh2 2. Ne2 Kh1 3. Nf4 Kh2 4. Kf3 $1 Kh1 5. Ke2 Kg1 6. Ke1 Kh1 7. Kf1 Kh2 8. Kf2 Kh1 9. Ne2 Kh2 10. Ng3 Kh3 11. Nf1 Kg4 12. Ke3 Kf5 $1 (12... Kh4 13. Kf4) 13. Nh2 Ke5 14. Ng4+ Kf5 15. Kf3 Kg5 16. Ne3 Kh4 17. Kg2 Kh5 $1 18. Kh3 Kg5 19. Kg3 Kh5 (19... Kg6 20. Kf4 Kh5 21. Nf5 Kg6 22. Ng3 Kf6 23. Nh5+ Kg6 24. Kg4 Kh6 25. Nf4 Kg7 26. Kf5 Kf7 27. Nh5 Kg8 28. Kg6) 20. Nf5 Kg5 21. Nh4 Kh5 22. Nf3 Kg6 23. Kf4 Kf6 $1 (23... Kh5 24. Kf5 Kh6 25. Ne5 Kh5 26. Ng6 Kh6 27. Nf4 Kg7 28. Kg5 Kf7 29. Nh5 Ke6 30. Kf4 Kf7 31. Kf5 Kg8 32. Kg6) 24. Ne5 Ke6 25. Ng4 Kf7 26. Kf5 Kg7 27. Ne5 Kh6 28. Kg4 (28. Kf6 $1 {<cook PW>} Kh5 29. Ng6 Kg4 30. Ne7 Kf3 31. Nxc8 Ke3 32. Ke7 Kd4 33. Kd8 Kxc5 34. Kc7 Kxc4 35. bxa7 Qxa7 36. Nxa7 c5 37. Kxb8 Kd3 38. Kc7 c4 39. Kxd7 c3 40. Ke7 c2 41. d7 c1=Q 42. d8=Q+ Ke4 43. Qb6) 28... Kh7 $1 29. Kh5 $1 Kg7 30. Kg5 Kh7 31. Nf7 (31. Kf6 $1 {<cook PW>} Kh6 32. Ng6 Kh5 33. Ne7 Kg4 34. Nxc8) 31... Kg7 32. Nh6 Kh7 33. Nf5 Kg8 34. Kg6 (34. Kf6 $1 {<cook PW>}) 34... Kh8 (34... Kf8 35. Nh6) 35. Kf6 $1 (35. Ne3 $1 {<cook PW>} Kg8 36. Ng4 Kh8 37. Ne5 Kg8 38. Nf7 Kf8 39. Nh6 Ke8 40. Kg7 Kd8 41. Nf5 Ke8 42. Ne7 Kd8 43. Kg8 Ke8 44. Nxc8 Kd8 45. Nxa7 Ke8 46. Kg7 Kd8 47. Kf7 Qxa7 48. bxa7) 35... Kh7 36. Kg5 (36. Kf7 $1 {<cook PW>}) 36... Kh8 37. Kg6 Kg8 38. Nh6+ Kh8 39. Kf6 Kh7 40. Nf7 Kg8 41. Kg6 Kf8 42. Nh6 Ke8 43. Kg7 Kd8 44. Nf5 (44. Kg8 $1 {<cook PW>}) 44... Ke8 45. Ne7 Kd8 46. Kg8 Ke8 47. Nxc8 Kd8 48. Nxa7 Ke8 $1 49. Kg7 Kd8 50. Kf7 $1 Qxa7 51. bxa7 Kc8 {<eg>} 52. a8=Q b5 53. axb6 a5 54. b7+ Kd8 55. Qxb8# 1-0
[/pgn]
But there are several side lines commented as <cook> with different DTM than main line has, SlowChess after backwarding quickly says:

SlowChess Blitz 2.9 avx2 (32768 MB hash, 31 threads)
FEN: qnb5/rp1p4/pPpP4/P1P5/2P5/6K1/8/6Nk w - - 0 1

[78] M53 00:20.2 970.73M Kf2 Kh2 Nf3+ Kh3 Ne5 Kh2 Ng4+ Kh1 Kf3 Kg1 Ne3 Kh2 Kf2 Kh1 Nf5 Kh2 Ng3 Kh3 Nf1 Kh4 Ke3 Kg4 Ke4 Kh4 Kf4 Kh3 Kf3 Kh4 Ne3 Kh5 Kf4 Kh4 Ng4 Kh3 Kf3 Kh4 Nf2 Kh5 Ke4 Kg5 Ke5 Kh5 Kf5 Kh4 Kf4 Kh5 Nh3 Kh6 Kg4 Kg6 Nf4+ Kf6 Nd3 Kg6 Ne5+ Kf6 Kf4 Ke6 Ng4 Kf7 Kf5 Kg7 Ne5 Kh7 Kg5 Kg7 Nd3 Kf7 Nf4 Kg7 Nh5+ Kf7 Kf5 Kg8 Kg6 Kh8 Nf4 Kg8 Nd3 Kf8 Ne5 Ke8 Kg7 Kd8 Ng6 Ke8 Ne7 Kd8 Kg6 Ke8 Nxc8 Kf8 Nxa7 Kg8 Nc8 Kf8 Ne7 Ke8 Kg7 Kd8 Kf8 Qa7 bxa7 b5 axb8=Q#

In CHESTUCI.epd of version 5.2 it's stored as #51, pity without best parameters.

So it seems to be #51with this one position, and there aren't any duals at first move, even thus the single best move is to be found easily now, and the correct solution again should be counted "found" only togehter with exactly right DTM, regards

[peter]

SlowChess Blitz 2.9 avx2 (32768 MB hash, 31 threads)
FEN: qnb5/rp1p4/pPpP4/P1P5/2P5/6K1/8/6Nk w - - 0 1

[86] M51 01:20.0 3036.27M Kf2 Kh2 Ne2 Kh1 Ng3+ Kh2 Nf1+ Kh1 Kf3 Kg1 Ng3 Kh2 Kf2 Kh3 Nf1 Kh4 Ke3 Kg4 Ke4 Kh4 Kf4 Kh3 Kf3 Kh4 Nh2 Kh5 Kf4 Kh4 Ng4 Kh3 Kf3 Kh4 Nf2 Kh5 Ke4 Kg5 Ke5 Kh6 Nh3 Kg6 Ke4 Kh6 Kf5 Kh5 Ng5 Kh4 Kf4 Kh5 Nf3 Kh6 Nh4 Kh5 Nf5 Kg6 Ng3 Kf6 Nh5+ Kg6 Kg4 Kh6 Nf4 Kg7 Kg5 Kf7 Nh5 Ke6 Kf4 Kf7 Kf5 Kg8 Kg6 Kh8 Ng3 Kg8 Ne4 Kh8 Ng5 Kg8 Nf7 Kf8 Nh6 Ke8 Ng8 Kd8 Ne7 Ke8 Nxc8 Kd8 Nxa7 Ke8 Kg7 Kd8 Kf8 Qxa7 bxa7 Kc8 a8=Q b6 Qa7 bxc5 Qc7#

[jp]

...
SF has a particularly hard time with them, solving only one in a reasonable time.
Often, the key move is exactly right, but the plan is exactly wrong.

So what are the conclusions now?

Is there any position in the set that Stockfish actually gets wrong, i.e. misevaluates as a draw when it's a win, or gives an incorrect winning line for (meaning a move for the attacking side anywhere in the variation can be refuted by the defending side, leading to a draw)?

Otherwise, it's not really SF's problem...