notatiger54: A few concrete analytical notes:
14...Bd7!? This was my original contribution to the line that I dubbed the "Phoenix Variation" in my published articles, in recognition that the whole 8...h6 variation had been left for dead after a series of spectacular miniatures between GMs in the early '70s (starting with Grefe-Browne, as mentioned above). Instead of leaving his king in the center by going for ...Rb8 and ...b5, I decided it would be wise for Black to quickly finish queenside development, preparing to castle long.
A key line of my analysis justifying 14...Bd7 goes: 15.Nxe5 dxe5 16.Rhf1 hxg5 17.Bxg5 Nh7 18.h4 0-0-0 19.Rxf7 Bxg5+ 20.hxg5 Nxg5 21.Rff1 Nxe4=
19.Kb1?! A few months before this game, I had suggested this move in my Chess Horizons article, as a possibly promising alternative to 19.Nxg5. I didn't intend to mislead anyone; I simply discovered the strong Black response that follows in this game after my article came out. (Better is 19.Nxg5 Rxg5 20.g4, with advantage.)
23.Rdd1 At first glance, 23.Rd2 looks strong, with the idea 23...Rxg2 24.Ne5 threatening Bh5+ with discovered attack on the g2-rook. But then the tactical justification behind 19...f5! and 20...Nc6! appears: after 23.Rd2 Rxg2 24.Ne5 Nxe5 25.Bh5+ Rg6 26.Rg1 Kd8! 27.Bxg6 Nf3, and Black escapes with better chances (2Bs + united passed pawns).
30...Nd4!? I was quite proud of this positional pawn sac. But Stockfish is unimpressed: it says the sac actually diminished my winning chances.