Note: this is another rules-lawyering question.
I first ask for an official source I am calling into question the phrasing of the rules. Then I consider a few plausibly official sources, and point out a loophole (unsurprisingly not the first time castling has come under fire) - is my reading correct?
Preamble: Is there an official source for the rules on Chess960 castling?
Common knowledge at this point is that castling in Chess960 can be informally described as:
- the king and rook end on the same squares as in regular chess,
- the king and rook should not have moved before,
- the squares the king and rook need to "slide through" must be empty,
- the king should not be in, or pass through, or end in check.
- (And the castling cannot be made with a promoted rook).
How about an official source?
Well, one would start by looking at the FIDE Laws of Chess as usual; the section Guidelines II is on Chess960 Rules. Disappointingly, only articles II.220.127.116.11 (specifying king and rook final squares) and II.18.104.22.168 (specifying passed-through squares must be vacant) seem relevant; no mention is made of the king being in or passing through check. So we assume that regular article 3.8.2 (specifically 22.214.171.124.1) applies to cover those.
Alternate sources would be Wikipedia (which cites for some reason dwheeler.com), and the organisers of the recent World Fischer Random Chess Championships, chess.com. Those read largely the same as the FIDE Laws.
Consider the following position in Chess960. (wKb1, wRb1, bKe8, bRa1).
4k3/8/8/8/8/8/8/rR2K3 w Q - 0 1
Assume also that it's white's turn, and the white king and rook have never moved before. Can white castle left? Applying a little common sense, the answer is no (if we make the move, the white rook goes to d1 and the white king ends in check on c1).
However, according to every ruleset above (FIDE, Wikipedia, chess.com), all the criteria for castling have been met. In particular, in the diagram position, the black rook does not attack c1 nor d1, so the king is not moving through check. The relevant clause of the FIDE Laws is 126.96.36.199.1:
[Castling is prevented temporarily] if the square on which the king stands, or the square which it must cross, or the square which it is to occupy, is attacked by one or more of the opponent's pieces.
Which is not the case in the diagram.
One way out of this would be to invoke article 3.9.2 of the FIDE Laws:
No piece can be moved that will either expose the king of the same colour to check or leave that king in check.
Which by unfortunate phrasing, is ambiguous in the context of castling, which is "A move of the king and either rook of the same colour..." (Article 3.8.2), i.e. a move of two pieces. In the diagram position, moving the king alone, or the rook alone, will not expose the white king to check.
So I call attention to this loophole here. To summarise this question:
- Is there an official source for castling in Chess960?
- Is there an error in my reading of the rules that there is a loophole?