This is a mind-blowing mate in 4. Let's call it "Schrödinger's Mate".

It was a few years ago since I last played chess but let's try it 😅
The pawn on b6 moves forward two times and get a queen, the black king has nowhere to escape, because of the two white pawns on d6 and g6 so that's a mate. But why a Schrödinger's mate then?

@cafecortado If 1. b6, Black can just castle. That's where the rabbit hole starts. ^^

@Fast-Nop What about Qh3, then b7? castling would be prevented and Nf5 would be a threat..

@simulate Qh3 does not prevent Black from kingside castling, and once the king is there, it's safe for the next three moves, e.g:
1. Nf5 O-O
2. Qe2 b4
3. Qe7 Rf7
4. Qxf7+ Kh8
5. Qxg7# 1-0

SPOILER ALERT!

OK, this is really hard. The key here is that it is uncertain what the castling rights are. It can be proven as below that either White or Black can have castling rights, but not both. However, the starting position has both possibilities until one player castles and, Schrödinger style, thereby forces the whole position into the one where the other never had castling rights.

Let's assume White can castle. Then the Bc1 must have been taken on c1, and the Ra1 must have been taken without ever having left the fortress. On the other hand, Black must have made two pawn captures on the queenside - one for the white Bf1, and the other for some "other piece". However, White's pawns on b6/d6 have needed five captures to arrive where they are. Since Black is missing seven pieces in total, that leaves two captures.

The "other piece" must have been a promoted pawn because no further white pawn can have captured towards the queenside with only two available captures. That was the original white h pawn (or g-pawn, and the h-pawn captured on g6). It can only have promoted via either f7, in which case the black king must have moved out of check, or via h8, in which case the rook must have moved - in both cases, Black cannot castle.

If however Black can castle, then the "other piece" cannot have been a promoted pawn. It must instead have been the white Ra1 which came out after the Bc1 had been taken. Or it could have been the white rook from h1, after which the Ra1 moved to h1. Both ways require the white king to have moved, i.e. no white castling anymore.

With that in mind, White's task is to castle before Black can do so, and then promoting the pawn on b6. The queen sacrifice clears f1 and prevents Black from castling right away because the queen checks g8. Then White castles so that Black can never have had castling rights. Total Schrödinger mindfuck.

1. Qc4 bxc4
2. 0-0 Kd8
3. b6 Ke8
4. b8Q#

I can never solve such riddles 😅 but I will come back to it tomorrow

@simulate The remarkable thing is that it's a mate in 4 that computers cannot solve. ^^

Take it to chessrant guys 🍿 cold popcorn ❄️

@Fast-Nop I’m lost, why can’t black castle after white castle?

@jesustricks Because the backwards analysis as above shows that if White castles, this implies a game history where Black must have moved either king or rook before.

I see, so it’s assumed that this board state is the result of a valid chess game. I’m a noob in chess riddles. And chess.

I thought riddles were all just hypothetical placements of pieces without that condition I mentioned.

I curse the day that chess was given internal state beyond what the board shows.

I mean I'm kinda into chess (Hikaru Nakamura and Agadmator on Youtube 😊) but I never would have guessed such an awesome puzzle exists and that you can prove something about castling rights even in a position like this one.

I say let's have the chessbot AI solve this lol

Black king is gone

If ypu are black and don't resign. You are going for a remi and waste time, because this is a loosing position.

@karma Won't work because "AI" is glorified pattern matching while what you need is reasoning. Though neuro net pattern matching can play amazing chess these days as witnessed by Leela Chess Zero.

@simulate Yeah that's one pretty amazing puzzle. :-)

@Fast-Nop I think it would be possible to make a reasoning AI, be it an actual neural net or just a heuristic atop a clever data model, that takes a ruleset and a question as input and proves the question true or false - although this is of course NP hard so the implementation will have some drawbacks like unbounded runtime.

@Lor-inc I don't think that this is possible because AIs don't understand anything. What they do is pattern matching as well as correlation finding.

Their success in regular chess shows is that understanding isn't even required to play chess better than humans can.

@Fast-Nop alpha star does outplay alot pros

@SHA-16384 It's usually white's turn.

@SHA-16384 In mate problems, it's White to move and deliver checkmate.

One rare fun exception is below where the task is that Black COULD mate in 3, but actually can't (the eigth rank is on the low side in the diagram because it's from Black's view):

The mate in 3 itself would be easy enough with:

1. Qa8×e4+ Nc3×e4
2. Rf7-f3+ Nd4×f3
3. Rb2-e2#

But then Black would form the crucifix. Such problems are attributed to the mythical Dr. Faustus who used to annoy Mephistopheles (none else then the devil) with such problems that are unsolvable only for Mephistopheles. That's also why Black is to move, to represent the devil.

Also find realy bad chess in appstore

This is so jacked up haha