Here's a chess puzzle that requires you to have a very liberal interpretation of the rules. Observe the position below. White to move and mate in 2:
By getting the pawn to the other side you can turn it into a rook or a queen to cut off the black king's escape on the e file which leaves it with only one place to go: g2.
But what can you do in this situation? It looks like you only have 2 ways to check the king now. Moving your new rook from e8 to e2+ doesn't work because the king can just take your rook on h1 (or run away to f3) and any other move your choose either doesn't deliver a check or allows Black to just take a piece.
Instead, let's look at how the rules of chess defined castling back in 1970: “The king is transferred from its original square, two squares toward the rook; then that rook toward which the king has moved is transferred over the king to the square immediately adjacent to the king.” Other criteria are that the rook hasn't previously moved and that the king spends no part of its move in check.
So by the letter of the law, we could castle towards our newly created rook (which hasn't moved) by moving the king 2 spaces upwards and placing the rook on the other side of it. This delivers mate as seen below:
Notice that the movement of the king allows the back rooks to become connected (so that Black can't take one) and it covers Black's escape to f3.
Normally castling Kingside is denoted as O-O and Queenside as O-O-O. These show how many spaces the rook passes through. Therefore O-O-O-O-O-O makes sense for the notation for this bizarre move.
This was published in a series of variation chess puzzles in 1970 and it caused lots of arguments about the technicalities of the rules of chess. FIDE has since updated the rules to prohibit this move by requiring the rook in question to be on the same rank as the king.