Puzzles End Too Early

In about half of the puzzles I do, the solution line ends in a position that still looks very unclear. After opening the analysis board and seeing the continuation, I see why the solution is what it is, but the full line should really be in the tactic itself so we don't have to open the analysis board for every problem.

I understand why this might be difficult if there are multiple winning lines, since the puzzle trainer only allows for one line. But in most cases there is exactly one continuation that is much stronger than the others, so there's no reason not to include it in the solution.

Please downvote these puzzles. Downvote all puzzles that you don't love.

"But in most cases there is exactly one continuation that is much stronger than the others, so there's no reason not to include it in the solution."

Please post puzzle URLs as examples of this.

I often find this too. I play a winning move, that is right, but then get the follow up slightly wrong and get the puzzle wrong. I don't remember to downvote it often enough. i will try to downvote the bad puzzles in the future

So I just tried a few more puzzles to find examples. Here's one: lichess.org/training/uemtV

So after fxe6, the puzzle ends. Material is equal, and black's rook appears to be "hanging" on d2. But of course, the rook isn't actually hanging - because the whole idea of the sequence is ...Qxd2 Qh1+ Kf2 Ne4+ winning the queen. Both Qh1+ and Ne4+ are by far the best moves in the position (Qh1+ is -62 for black, 2nd best move is +7 for white. Ne4+ is -63 for black, 2nd best move is -7).

So if black is looking to solve the tactic, he might see that Rd2 distracts the queen from the defense of the king. But he doesn't have to see the concrete win to get the tactic right, and will have to enter the analysis board to see the full sequence if he isn't able to find it himself.

I probably should have proposed a solution to this in my original post, since looking back it just looks like I'm just complaining. Which was not my intention at all! I love Lichess and everything you guys do here. I do downvote tactics I don't think are great, but the thing is, it's not a bad tactic at all. Sacrificing the rook to distract the queen, forcing the king to a square where you can play a royal fork is actually a pretty great tactic.

I think one of the issues might be that instead of Qxd2, the engine shows that Nd4 is technically more accurate, since it's only -8 for black instead of -52. But if Nd4 is played the continuation is trivial, you take white's queen and are up tons of material. Qxd2, although it is objectively worse than Nd4, is the critical line in this tactic, since the whole reason that Nd4 is more accurate, is because Qxd2 doesn't work tactically! But making this differentiation between critical line vs. the technically accurate line probably would have to be done manually by human input, which is what I assume is the main issue here.

Maybe in addition to letting users upvote/downvote, and select which tags are appropriate, there could be a way for users to suggest additional moves to be added to the tactic so it's complete? I don't know how feasible that could be, but it's really all I can think of.

Anyways, thanks for all you do Thibault. Have a great day!

i also played the move correctly but next move is not correct

"Maybe in addition to letting users upvote/downvote, and select which tags are appropriate, there could be a way for users to suggest additional moves to be added to the tactic so it's complete?"

Way too often the human moves will be wrong, in the sense that they will have an alternate move that is not bad enough that it should be refused. And by too often I mean all the time. Since that's exactly how the puzzle generator works: it stops when the only-move sequence stops.

Instead, let's downvote puzzles that humans don't enjoy. We have more than a million puzzles. Let's only keep the great ones.

If you plan to keep only the great ones, then people will have to select them as there favorite.
Star the puzzle to bookmark it in their profile and at the same time those puzzle will get a book mark number as being one of the great ones. The users will never see the number of times it got bookmarked.

A Lichess database of bookmarked great puzzles could then be created for study purposes or to help raise funds for lichess.

If a player down votes a puzzle, they will not bookmark it into their favorites.
If they up vote and bookmark it in their favorites, those are the great ones.

Even if they don't remain bookmarked for ever, they will have been bookmarked at least once.
The amount of players that book mark a puzzle and the type of rated player that bookmarks a puzzle, makes a puzzle great for that rating level.

So for starters, a star is needed to bookmark the puzzles, similar to bookmarking a game, but you don't want that star hidden. You might want a star that is transparent and shows a color one if we have already bookmarked it into our public or personal profile.

If an admin goes through these 1 million puzzles, working one minute on every puzzle, eight hours a day, five days a week, he will need around 8 years.

Let's better down vote those puzzles :-)