Has it really been over a month since I posted Fischer - Larsen, Portoroz 1958? At that time Fischer's death was still fresh in memory, but the memory has now faded.
The first flurry of annotator's symbols occurs after the diagrammed position. White continued 14.Nd5 Bxd5, where both Fischer and Kasparov note that 14...Nxd5 is bad because of 15.Bxg7 Kxg7 16.exd5 Bd7 17.Rde1. Somewhat curiously, Fischer referenced a game ('Suetin - Korchnoi, USSR Championship prelims') that Kasparov says was never played.
Now instead of 15.Bxd5, both World Champions give 15.exd5 Qb5 16.Rhe1 a5 17.Qe2 as better.
Portoroz Interzonal 1958
[FEN "r4rk1/p3ppbp/3pbnp1/q7/1p1BP3/1BN2P2/PPPQ2PP/1K1R3R w - - 0 14"]
Fischer gives Black's 15...Rac8 ('after the game Larsen explained he was playing for a win') a '?'; Kasparov gives '?!'. Better was 15...Nxd5, when Kasparov agrees with Fischer that 16.exd5 Qxd5 17.Qxb4 offers more chances than 16.Bxg7 Nc3+ 17.bxc3 Rab8 18.cxb4 Qxb4+ 19.Qxb4 Rxb4+ 20.Bb2 Rfb8.
The retreat 16.Bb3 merits a '!'. Fischer wrote
Now I felt the game was in the bag if I didn't botch it. I'd won dozens of skittles games in analogous positions and had it down to a science: pry open the h-file, sac, sac, ... mate!
Since 16...Qb5 loses a Pawn to 17.Bxa7, Black tried 16...Rc7. After 17.h4, the idea 17...Qb5 was too slow, while 17...h5 fails to 18.g4. White continued 18.h5, and Fischer won on the 31st move.
Later: Re 'Fischer referenced a game ('Suetin - Korchnoi, USSR Championship prelims') that Kasparov says was never played', this was explained in My 61 Memorable Games: A Mystery by GM Larry Evans: 'the 10th edition (1965) of my Modern Chess Openings page 195 note "d" of the Yugoslav Attack in the Sicilian Defense sourcing "Suetin-Korchnoi." This error probably was carried over from the 9th edition. In those days no massive databases were available to check these references.'