Volcach: The computer recommends 10 h4 here instead of Bxf6. It's a common idea mirroring the h5 for Black in several lines of the Marshall Gambit.
Instructive moments for me:
<17 Qc4!> Not a move I'd even look at in the position. It pins the f7 pawn but seems to hang the e5 pawn. For me however, I'd have lost the Bishop as Re1 pins the Bishop and my plan of f6 to support the Bishop doesn't work! It's pinned by the aptly placed c4 Queen. Brilliant foresight to see that hanging the pawn, doesn't hang the pawn.
<20 c3!> The computer has this as an inaccuracy. But the unveiled plan of activating the king surprised me. Capablanca shows that his king can take an active role in the position while Black's is doomed to passivity, shuffling between f8 and g8
<35 Ke2!> Somehow knowing thatthe king is better placed on g3 than d3. I'm not sure I would've considered this idea, even not knowing now why it's better than staying in the middle.
<41 Qf5!> Beautiful square for the Queen, supporting Rd7 in the near future.
<43 Kg4!!> The only move that instantly wins. If Nxe5 Qxe5 Qxe5 Rxe5, the Endgame is probably still winning for white but it's not a trivial affiar. Now after Kg4 however both the Queen and Bishop are hanging, after Qf6-g7+, Black seems to be hoping for a Khx "blunder" alloweing Rxe5 now supported by the Queen. But Rd8 instantly wins in either case
Great game that I'll definitely commit to memory. Although the King and Queen Shufflings will be a tad annoying.