@Stefan_vom_Bodensee you mean Rc4+? That leaves the king open to d3, d5, and f3. And the stockfish analysis says Nc4, which would take the rook.
an example also for the unexpectable reaction: puzzle 66805: I play 1.Ra1, which attacks the black Queen on a7 and it is the correct solution; and now comes a strange answer by black: 1...Qc5: ??? - an absolutely senseless sacrifice of the queen. That's the way the lichess puzzles/training becomes an absolutely senseless exercise. In this example, i cannot say "Thank you" for the points. --- By the way: it the same thing with the puzzles on chess.com. The only good tactic trainer I know is the TIMLEG app in the google play store. All solutions there a controlled by stockfish 8.
@Stefan_vom_Bodensee if you notice, most puzzles come from blitz games, where it's more prone to blunders. I feel that most of the time you shouldn't really anticipate the answer, and just find the move best "at that point". Also, stockfish thinks Qxc5 is a +5.3 move ¯\_(ツ)_/¯
But, I do think that some are seriously pointless. For example, lichess.org/training/61903 this one where all you need to do is queen the d6 pawn.
Ok, a few things here.
First, one should be careful in picking examples of supposedly "bad" puzzles.
In the case of 66805, Qxc5 makes at least as much sense as any other move.
The queen is trapped, and on b8 would just hang to the bishop.
You can play Qxa1, giving it up for a rook. However, after Qxc5 dxc5, there's Bxc3!, and after queen moves and Bxa1, you've given up the queen for rook and pawn.
Of course it's all losing for black anyway, but Qxc5 is actually objectively the best move there.
Second, I do agree with the concern in general
This is why I still prefer doing tactics at places like chess.emrald.net, which explicitly rejects puzzles with more than 1 solution, with very specific criteria defining what that means.
Their criteria are not ideal, in my opinion, but a move in that general direction would be nice.
It's just much easier to accurately check that a singular best move exists in a position than it is to correctly flag every secondary solution as "Nice move, but..." etc.
An alternative, which may be something the devs are already working on, is to switch to a model more like the "Practice" section of the site, where it just checks the position after your move to see that you're actually holding the correct evaluation (or correct winning chances, based on the whole logarithmic fanciness), instead of just comparing the move played to a "correct" stored move.
Just my $0.02.
First, one should be careful in picking examples of supposedly "bad" puzzles.
In the case of 66805, Qxc5 makes at least as much sense as any other move.
The queen is trapped, and on b8 would just hang to the bishop.
You can play Qxa1, giving it up for a rook. However, after Qxc5 dxc5, there's Bxc3!, and after queen moves and Bxa1, you've given up the queen for rook and pawn.
Of course it's all losing for black anyway, but Qxc5 is actually objectively the best move there.
Second, I do agree with the concern in general
This is why I still prefer doing tactics at places like chess.emrald.net, which explicitly rejects puzzles with more than 1 solution, with very specific criteria defining what that means.
Their criteria are not ideal, in my opinion, but a move in that general direction would be nice.
It's just much easier to accurately check that a singular best move exists in a position than it is to correctly flag every secondary solution as "Nice move, but..." etc.
An alternative, which may be something the devs are already working on, is to switch to a model more like the "Practice" section of the site, where it just checks the position after your move to see that you're actually holding the correct evaluation (or correct winning chances, based on the whole logarithmic fanciness), instead of just comparing the move played to a "correct" stored move.
Just my $0.02.
This topic has been archived and can no longer be replied to.