Puzzle - Mate in three - shows wrong position

I do puzzles, mate in three, every day since years.
Recently I encountered positions which were definitely not a "mate in three".
Last example: https://lichess.org/training/GlQac

I just saw, that it is possible to add a new puzzle theme for a given position.
If this is not engine checked, this is not a good idea.
Maybe somebody just added the theme "mate in three" which is not correct.
It is annoying if I want to solve "mate in three", but get positions which do not fit in this category.

I think the mate-in-n tags are automatically assigned, based on the assumption that the engine played the "best" line, i.e. prolongs checkmate as long as possible.

In this puzzle, the engine plays into a checkmate in three instead of a checkmate in 6 or so... which looks pretty unusual.

There should be a pop up to report a faulty solution, if you let the engine run long enough.

But I'm not getting it here, it may be that my mobile is too weak in this situation im not sure. It does also show mate in 7 though maybe not deep enough for the pop up to trigger.

@lonelypeanut said in #3:

There should be a pop up to report a faulty solution, if you let the engine run long enough.

The solution is correct, so there won't be any popup. it's just that the "mate-in-3" tag is wrong. And this is one of the categories which you cannot even change manually.

@nadjarostowa said in #4:

The solution is correct

If the solution assumes that a bad move is played, in this case lower than number five should that not count as wrong?

@lonelypeanut said in #5:

If the solution assumes that a bad move is played, in this case lower than number five should that not count as wrong?

I guess you can argue that way from a theoretical (puzzle-mining) perspective. From the user's side, the puzzle is correct, as the opponent is by no means forced to pick one defense over another.

My guess is that the "popup thing" only checks for multiple solutions for the user, as not playing the optimal move for the other side is only a slight annoyance but doesn't render the puzzle faulty as such.

@nadjarostowa said in #6:

I guess you can argue that way from a theoretical (puzzle-mining) perspective.
From the user's side, the puzzle is correct, as the opponent is by no means forced to pick one defense over another.

But you can't solve puzzles if you can't assume the best move is always played by the opponent. To solve it you need to make assumptions and randomly picking a move from the top ten makes that impossible imho.

At least that's what I thought. I might be wrong.

@lonelypeanut said in #7:

But you can't solve puzzles if you can't assume the best move is always played by the opponent. To solve it you need to make assumptions and randomly picking a move from the top ten makes that impossible imho.

At least that's what I thought. I might be wrong.

No. You can solve the puzzle just fine. Naturally, the other side is completely free to play whatever they like, and you need to have good answers to all of them.

That's one of the differences between computer generated puzzles and those curated by a human. The computer might question you with moves that make no sense, while the human will test you on critical lines. Like a computer would give away all its pieces to not walk into a forced mate in 20 (but is completely lost), but that mate variation is the critical line and would ask for many tough decisions.

@nadjarostowa said in #8:

No. You can solve the puzzle just fine. Naturally, the other side is completely free to play whatever they like, and you need to have good answers to all of them.

Interesting. Maybe that's why I'm sometimes surprised (aside from being stupid and missing stuff) sometimes I feel like the move played is just some random worse move than what I found for the other side.

I'll be paying attention to that in the future.

@lonelypeanut

A random example that comes to my mind:

https://lichess.org/analysis/3rk2r/pp2ppbp/5p2/q2N4/3Q1P2/8/PPP3PP/2KR3R_w_k_-_0_1?color=white#0

The critical line is 1. Qa4+ Qxa4 Nc7+ etc.

If the computer would extract this puzzle from a game, it would probably ask for 1. Qa4 Rd7 2. Qxa5, which is, well, somewhat dull.