Puzzle doesn't accept solution for a rare reason

While trying the very addictive and nice Puzzle Streak, I came across a puzzle, lichess.org/training/eqHqG, from this position: lichess.org/Yqcx7F0d/white#62
I must admit I missed the correct mate in two because I saw a forced mate in three and thought that was good enough. However, my solution (Rf8+ Kh7 Qh6+ Kxh6 Rh8#) was denied because apparently after Rf8 black can repeat the position for the third time. But obviously I cannot know these details :-) Normally, all puzzle departments get around multiple solutions very nicely.

Thanks in advance for looking into it and keep up the good work!
Bert

Generally speaking, shorter solutions are more accurate and therefore better.

@KeithDenning You're right, but this isn't chess.com where a mate in 3 is rejected :-) If I understand it right, the puzzle would have been rejected if there are multiple winning first moves that aren't a mate in one. However, because of the repetition the script will have thought the winning line was unique while in fact there is another one avaliable just one move longer.

@TacticalBert It isn't about "rejecting" a mate in 3. It is about the shortest solution being most accurate and therefore correct.

I mean, take a position that can be solved with a mate in one. I'm sure there are lines where the same puzzle can be "solved" with a mate in five or twenty-eight (just like in a real game), but nobody would argue that the longer mates are of the same quality.

I'll agree that if a puzzle has two solutions of equal length it should be removed. Lichess is pretty good at doing this, even on those occasions where the script gets it wrong. But when there's on solution that is shorter than all others, that solution is best.

@KeithDenning But that is not the policy here, as far as I know. If there is e.g. a mate in 5, but also something as little as a +3 solution, it won't be made into a puzzle. However, the script got tricked here and that is why I made the thread :-)

Thank you @TacticalBert, you are correct. I removed the puzzle.

I remember the old "Openings" puzzles (or maybe it was just the last edition of normal puzzles) had "alternate" solutions that, while not accepted as the "correct" answer, didn't make you lose the puzzle. They returned a message that read "Good Move, but you can do better". Would it be possible to re-implement this response for cases such as this? Seems a waste of a good puzzle to remove it entirely.

@thibault

Could you explain in what way @TacticalBert was correct? The goal is to find the best move and a mate in 3 is just not as good as a mate in 2, I have to agree with @KeithDenning here although I admit that this solution sure has it's beauty, too.
I strongly doubt that it was rejected because of a repetition.

Lately I have been pleasantly surprised by the added improvements like Puzzle Streak and Racer.

The only thing that bothered me in that area lately is the way Stockfish defends in a desperate position sometimes. The worst case I saw once was a five move combination ending with checkmate. White's best defense whould have given Black only one move to mate, a doublecheck with Rc4. Instead he offered him the opportunity to choose between 6 mates in 1, almost any move of the rook on the c-file. Doesn't seem very logical to me as you should always expect the strongest defense.

You can have good arguments in favor or against accepting longer forced mates. The issue here was that lichess accepts longer mates in all puzzles except the one posted.

Hi everyone.

I had not initially understood that the puzzle rejected one line because it would have led to a threefold repetition in the game from which the puzzle was derived. Obviously nobody could have known this.

I think I misunderstood the situation. Though I will stick to my position that shorter forced mates are by definition more correct; that's not really the situation here.

@Cellini: I appreciate the support, but now that I (hopefully) understand the situation, I agree with @TacticalBert that the puzzle merited removal. And it has been removed, so I consider it case closed, and appropriately so.