Incorrect Puzzle sKutB

lichess.org/training/sKutB

Rf8+ is given as the only solution, but b7 is also a clear win. Let stockfish run to depth 31+ and you will see they are both given scores of +50 (overwhelming win).

:OO

how did you even figure that out

This one is especially tricky because the advantage doesn't come from having a rook up. The difference comes from the position of white's king.

If white's king is anywhere except e1 or f1 after white takes the rook, then black can force a draw through very specific maneuvers.

I suspect that the only reason this puzzle is only rated 1000-ish is because the best move is easy for even low level players to guess. The mindset probably being: "If either option leads to a similar end, why not start with the check?"

I absolutely agree that the level of depth to this puzzle is well beyond how easy it is to guess the correct moves.

(Also, stockfish at depth 40+ shows that b7 is only worth +4 (if you ignore the tablebase which says draw) when Rf8+ is worth +56. Endgames can be weird like that.)

if you go to the game you'll see that b7 was played and from the engine analysis that that is refuted by 45...Re2+ which draws.

If you play out the game using the engine moves material is immediately traded and the LiChess tablebase reports it's a draw.

Yes @danegraphics and @StingerPuzzles you're both right and thanks for the analysis @danegraphics . My own version of stockfish quickly looked all the way to the end of the game and told me the truth.

Something I've found about Lichess puzzles is that as long as there are only two choices, a puzzle that is very hard to understand can have a low rating. If 50% of puzzlers (including those rated 900 or below) can get it right, it will have a low rating. This is a frustration I have with puzzle storm / racer, and why I think it's inferior to regular puzzles where you can take your time. Puzzle storm will often reward you for guessing and you have learned nothing.

I suppose if 45 b7 Re2+!! 46 Kf1 Rf2+!! were played in a game, it would deserve double exclamation points. The patzers who played this in a rapid game, of course went 45 b7? Kxg3? . I wouldn't be shocked to see that continuation in a GM game either.

I studied the position just now and I think to be able to force a draw black needs to be able to
1. move between guarding the two pawns with a single move (they are separated by two files, which is fine).
2. guard a pawn to queen without interference from the white king if white wins the other pawn

To prevent the above and win, white needs to stop black from having its king on e3 or f3 and advancing a pawn to d3 or g3. White can defend this by having the king on e1 or f1 and playing Rb3+ immediately to drive the king back to the 4th rank.

Kudos to any player that sees that over the board (or even in a puzzle setting).

P.S. after doing the above, white needs to attack from behind with the rook to win the pawns--and insert a check at the right moment to force the black king in front of one of the pawns while taking the other one with the king--but I don't think that's nearly as difficult to figure out.

@castall - Well, that is how chess is, after all, which IMO is part of what makes it fun. There's always something more for us to learn about the game.

Having watched GMs analyze endgames before, I suspect that a GM might actually be able to deduce the reasoning for the king positioning. I have my own hunches, but nothing concrete (like being well positioned to juggle stopping both of black's passed pawns without losing your rook).

Regardless, computers will always be able to see things that humans will never be capable of seeing, so sometimes puzzles like this will pop up. Good catch on your part!

@danegraphics I was just analyzing the position and I edited my latest response above. I think there is something to learn from this puzzle.