A common problem with computer resolved tactical puzzles.

The puzzle here ( en.lichess.org/training/27989 ) has a problem I have not run into so much on lichess but it is common on a lot of tactical training sites. The perfect solution is very complex with a 9 move variation to finally win the queen with a million different viable options including numerous inferior pawn, piece, and exchange wins before that. However, on move two of the combination the computer sees the 9 move queen win variation follows quickly by mate as well and so gives away its queen on move 2 completely ruining the value of the problem.

Perhaps there should be a way to report problems that should have a human override the computer solution?

As another little issue, I don't think it's very appropriate for this problem to have been selected as an "easy" problem for me. I understand it's rating relative but it's inappropriate going from mate in 2 to 'that' in the same problem set.

In theory the voting system is supposed to help filter out bad problems. In practice players upvote problems that they solved correctly and downvote problems that they failed.

Do we have numbers to back this hypothesis up, @Toadofsky?

When solving a puzzle it is important to classify it - fork, skewer, mating net etc. That seems to be intrinsic to the benefit gained in training. I vote that you switch the numeric voting system to a tag based one that might also contain tags such as 'non unique solution' that might help weed out rogue ones. Further to this I'd like to be able to study one tag type - to reinforce a concept.

I started coding to track the numbers but the puzzle schema changed. What we have now are three things:

1. High-rated mate puzzles with hundreds of karma having only one solution despite having multiple mates or ways to win heavy material

2. High-rated material gain puzzles with negative karma which are "simple" (solutions 1-2 moves deep) but the obvious check or queen sacrifice doesn't mate

3. Anecdotal evidence from watching people solve puzzles versus my own experience: it's rare for players to upvote a problem after failing it, or to downvote a problem after passing it, and the opposites are common

The problem with the voting is that after solving/failing you only vote based on your own experience. "My alternative solution wasn't accepted" -> downvote. "I found the checkmate in 3" -> upvote. But you can't see the puzzle as a whole.

Would it be possible to display the whole tree of variations, which are accepted? That might help to "explore" the puzzle better. Because often I wonder, if this or that move would have been okay also.

@mrburns123 in theory there are tools to help you with that at the bottom of the problem: the "from game xxx", where you can click and see the analysis, and play the variations with the local engine loaded; or the analysis button, for local analysis; or even the "play from position", where you can play the position against stockfish.

So in principle you have plenty of ways to test your ideas (I do it when I do not see a refutation to my solution)

Yeah sure, I do that also. But I don't know what the puzzle would have accepted or not. And that's why it is hard to up- or downvote it. I don't want to upvote a puzzle, just because it accepted "my" solution.

I don't think it's a bad problem, just a bad solution. If the solution was the most critical line it'd be a reasonable problem.

A typical downvote from me:
en.lichess.org/training/22411
I saw that Re1+ leads to mate soon: Re1+ Kd6, b4, with the threat of Qc5#. Black can throw in b6 and Rxg2, but that's it. A simple solution, mate in 5.

But apparently there is a mate in 4, so my solution didn't count.
I just don't care for finding a mate in 4 when I have found a mate in 5.
Unless the puzzle would specifically state "mate in 4". Because otherwise I will waste 5+ minutes on finding a shorter mate, when there might not even be one.