Problem 1356, after the win

#1

Spoiler alert, do the puzzle first before looking at this.

I got the congratulation part right away. But I couldn’t finish the win. I figured out all the main pathways (see below) but I couldn’t figure out {3} even though I almost saw the 9-6 sacrifice. {2} took me a long time to get. The computer likes 1-6 right away but I found that easy to counter. So I would have got to the “Congratulations” part but would have ended up drawing.

[FEN “B:W5,10,12,13,K16,27:B1,3,6,20,K26”]

  1. 6x15 27-23 2. 26x19 16x23 3. 1-6 {1}5-1 4. 6-10 1-6 5. 10-14 6-9 6. 14-17 9-6 7. 3-7 12-8 8.
    17-21 8-3 9. 7-11 6-10 10. 15-18 23x14 *

{1} 3. 3-7 12-8 4. 7-10{2} 8-3 5. 10-14 13-9 6. 15-18 23-19 7. 14-17{3}
3-7 8. 1-6 9x2 * {Note the threat of 9-6}

{3}7. 18-22
9-6 8. 1x10 3-7 9. 22-26 5-1 10. 26-30 1-5 11. 30-26 5-9 12. 26-30 9x18 13. 10-14 18-9 *

{2}4. 7-11 8-3 5. 11-16 3-8 6. 15-19 23-26 7. 1-6 5-1
8. 6-10 1-6 9. 10-14 6-10 10. 14-17 10-14 11. 17-21 14-18 12. 21-25 8-12 13. 20-24 26-23
14. 19x26 12x28 *

#2

I love these gems hidden in the end game. I would have drawn this one too.

And thank you for these posts, seeing what comes afterwards gives me some insight into what the engine and problem generator are doing. The puzzles usually end when you win material and I wanted a king to be enough so that some puzzles end when the player forces promotion. I think that was a good choice but you’re right it does lead to situations where you have the king or extra piece but no easy win.

I’m not sure what the solution is or if there is one. On one hand maybe it should force the player to demonstrate the win. But on the other hand I want the puzzles to be ‘snappy’. If it’s 40 moves long its more like an endgame study than a tactics problem. I think I will add some text to the user guide that says the win afterwards may still be tough, your task is only to find the material win.

Maybe a separate section for endgame practice would be fun. Puzzles where you start up material but have to demonstrate the win. In chess for instance practicing rook + pawn vs rook is a useful exercise for amateurs. Maybe I could build an endgame puzzle generator that asks users to follow the ‘moves to conversion’ database for positions that occurred in games. The more moves required over perfect play, the worse your score, unless you end up in a draw where you get the problem wrong. That would require a lot of work including some major infrastructure changes but it does sound like the puzzles would be pretty fun.

Does anybody have any thoughts on what to do for puzzles where the win is far from clear? Any ideas on how to detect it? What to do about it?