The "official" Lichess solution gives 1. ... Rca7 2. c4 but what if 2. g5? In fact after 2. g5 black seems hardpressed for initiative.
Many Lichess puzzles leave me puzzled because of such ignored lines. Play with Stockfish on those positions and you will know!
The puzzle is correct. After 1...Rca7 2.g5, White is still losing.
Just 2...hxg5 3.hxg5 Bc3 or just 2...Bc3, White has no good way of stopping the straightforward idea of Rxa2 and Ra1 with mate to follow.
White has to settle for Rd2, just giving up the exchange, and black is winning.
Strange. Stockfish Level 8 plays ...g5.
But I have another question: in the main solution, what's wrong with Rxa1 -- why must the other rook be skewered instead?
There's nothing wrong with Rxh1. The solution is actually wrong to make Rxh1 fail.
On the 2.c4 or 2.g5 thing, that doesn't matter so much. The point of puzzles is just to make the trainee find the right moves, not necessarily to make the puzzle play the best moves.
In this case, though, 2.c4 and 2.g5 are equally bad, and one of them has to played, so that doesn't really matter.
Skewing the king to take the rook on the g-file removes the defender of the g-pawn making g5 less of a threat.
Yeah, that's one way to try to make sense of it, but objectively g5 isn't a threat.
Evals are very similar for both Rxh1 and Ra2+. It's actually a bit weird that Rxh1 isn't listed as an alternative, since the evals are about 0.4 apart almost from depth 1.
The reason the engine prefers Ra2+ and Rxg2 by a little bit over Rxh1 is that after 1...Rca7 2.c4 Rxb3+ 3.axb3 Ra1+ 4.Kc2 Rxh1, white can play 5.cxb5.
Of course that is losing, but the point of the other line is that after 1...Rca7 2.c4 Rxb3+ 3.axb3 Ra1+ 4.Kc2 Ra2+ 5.K-moves Rxg2, if white now tries cxb5, then black has Rb2, immediately picking up one of the b-pawns.
It's a little thing, but that's why it prefers Ra2+ and Rxg2. Still, the eval difference is too small for it not to get at least a "You have better".
Probably just a quirk in how that one was parsed.
We'll have to talk about this sometime now that I'm starting to have a little more free time (work's been insane the last couple months), but I do think there are some nice ways to handle at least new puzzles (obviously redoing existing ones is too much) that will make this all a much less frequent occurrence.
At any rate, I've already started to ramble too much.
Cheers!
All this discussion definitely shows that that particular puzzle is puzzling and hence does not so much teach beginners as it confuses them. Let us ananalyse it more before we give a solution to it. Till then, we'd rather not show it among the puzzles.
See another puzzle : http/en.lichess.org/training/31997. After
1. Be5 Rb2 2. Ka1 Qe5 3. c4, the official solution gives 3 .. Qc3 and rejects 3 .. Qf6, but the latter is *clearly* better, allowing black to keep the extra rook.
BTW how is training content, including puzzles, vetted?
#9 By users voting on the content.