|Credit: Tim Krabbe (see below for details).
In the position above Adams (Black) played 24... Nd7? which was answered by 25. 0-0-0! winning the exchange (and later the game). As Krabbe notes, missing the possibility of castling happened in the past many times.
It is not surprising Krabbe points out this particular game on his web site. After all he is -- if such a thing exists -- the world's expert on castling. Wikipedia notes that in the early 70s he composed a chess problem exploiting the fact that the FIDE definition of castling in force then allows White to castle "extra long". Krabbe gives such castling the natural designation 0-0-0-0 in his 1986 book Chess Curiosities; for some reason, Wikipedia prefers 0-0-0-0-0-0.
What is "extra long" castling? With a promoted rook on e8 and a king on e1, Krabbe noted (following, if Wikipedia is correct, Max Pam's original discovery) that castling was legal, so long as the newly-created rook had not moved, and all other conditions (the king is not in check, etc.) are fulfilled. It is done -- again, taking the FIDE rules in force at the time literally -- by moving the rook to e2 and the king over it, to e3, in one move. Of course, the reverse -- with a promoted rook on e1 -- is possible for Black.
Later, FIDE amended the rules, explicitly requiring both king and rook to be on their own first rank -- much like, in order to avoid similar difficulties, the rules now explicity require that a pawn be promoted to a piece of its own color.
|Credit: Wikipedia (see link above)