Having reached the training score of 2099 for the first time, I resolved to be extra careful on the next puzzle in order to cross the 2100 bar :) But the puzzle I got was
lichess.org/training/64168I couldn't find anything clear, and when I analyzed with Stockfish, it couldn't either. It's presently at depth 28 and has found an edge of 1.5 with the suggested move d5, and if we play the suggested moves after that (starting with Ne5) there is nothing very clear. The puzzle on the other hand shows Nxd5 (knight sac from the opponent) after d5, which does look good (+3), but is not what Stockfish suggests.
That doesn't look quite right... Thanks.
The idea is to remove the knight on c6 from the queenside to get a nice attack with
1.d5 ne5 2. Qa4
If you don't like the puzzle. Down vote it!
If many people do it will be removed.
Never mind... (edited)
Yes, not very easy to understand this one. The line with the knight sacrifice is just a computer immediately realising that he is worse because it expects only good moves to be played in every position.
However after d5 Ne5 Qa5 a6 white needs to sacrifice his bishop for a kind of an uncertain attack. We are not used to do that at all because we think if we sacrifice it must be over for the oponent which is not true. Still trying to understand this concept.
By the way, defense is often „weak“, at least against us humans. Sometimes you don‘t have to show that you have calculated well, the defender throws in the towel with say jettisoning material right away.
Having d5 Nxd5?? in the PV is likely just a result of a very shallow search.
At very low search depths, d5 Ne5 gets evals around +2.5, and while my SF never actually showed Nxd5 as the best reply, it was the second best reply only .1 or .2 away sometimes (again, at very, very low depths).
Most likely the analysis was just really shallow for some reason.
Also, we don't have to play Qa4 after Ne5 just because SF likes it the most. Bxa7 just takes a pawn for free, and is a pretty stable +1 (although black can make it interesting with Bxa7 Nxd5?! with some compensation for the piece, but it's not enough).
Winning a pawn is a pretty simple way to make progress :)
Having played through a bunch of different lines, I would still prefer to play Qa4, even though the punch line comes much later.
At any rate, it's clear that Nxd5?? is poor response to d5, but d5 is clearly the best move in the start position, and exd5 is clearly the best response to Nxd5??, so other than the fact that our "opponent" in this tactic plays poorly, I don't really see a problem.
Every move we're required to play is the best one. It just looks weird, is all.
Thanks guys. So, I think the puzzle does have a problem, in fact two problems: the gain is smaller or more uncertain than in almost all other puzzles, and Nxd5?? is a wrong following based on a search that for some reason was very shallow. I usually don't upvote or downvote puzzles but I will downvote this one :)
