I think that the point of the puzzle is missed when the computer doesnt test you.
lichess.org/training/62723White should take with the pawn not the queen
If you take with the pawn : Rg8+, Kh1, Qxh2, Kxh2, Rh4#
I know, but that's the variation the puzzle should have given you in my opinion.
Yeah. I agree gxf3 is much cooler and more important line.
The computer is supposed to play the best solution in puzzles, no ?
Qxf3 is logical if it is the case.
Afaik the computer tries to play the best move. Gxf3 is a mate in 3 so the engine doesn't like that.
So we agree that Qxg3 is a logical move for the computer !
I understand
@HaakonGM because some players would have play gxf3 but as you said
@lovlas the computer plays the best move.
So this puzzle is logical.
Sure. But Qxf3 doesn't really illustrate the point of the exercise. But ofc its the best move.
But I think that gxf3 is a more important line. If you dont see the follow up, you lose.
Lichess puzzles currently have so many issues:
--Not timed, making ratings meaningless.
--Many puzzles are simply opponent's blunders and you just take his pieces.
--You need to guess the best computer moves for 4 or 5 moves, and a small inaccuracy at the end costs you all the points. Partial points should be awarded if the inaccuracy doesn't lose.
--As HaakonGM has pointed out, the computer shows "best" (or least bad) evaluation moves, not most challenging moves.
--Many has duals.
Improvement of the puzzle-generating algorithm and/or some human curation is necessary, I think. At present, I can't recommend these puzzles to newcomers to chess.