See how these little flank moves (h3 and ...a6) become very good moves when there is a direct purpose behind them.

Exactly archduke_piccolo, when there is real justification behind the moves and you aren't just pushing them to simply keep away a knight or bishop early, they become real daggers later on.

It may be a mistake - it puts my queen on a8 where it has influence down the a file and along the long diagonal.

Is this a mistake or is white just playing a slightly worse non-computer line. I wonder what the difference between this and Nd2 really is, for example.

Well the other lines leave c4 protected.

It is possible White is simply gambiting the c-pawn for the sake of attack. But Black's defences, though held back, look pretty solid. Another alternative seems to have been 16.b3, followed by 17.Re1.

Nxc5! not a move I would have seen in the position. You are also forgiven for missing it. This is why computers are so good.

A possible alternative to 17.exf6 e.p. might have been 17.Nc3 bxc4 18.Re1, though how one proceeds from there is hard to say for sure.

It seems to me that white is helping black solve his problems with this move. That knight on e8 isn't good but on d6 is controling important squares. I have heard that trades are either forced, fantastic, or failures (the idea of course is that even trades are basically failures) and this looks like a failure of a trade to me.

I agree with Charles: White should build up the tension with Re1.

Yes Re1 needs to come first.

I think the pawn capture is preferable, less to improve the structure than to cover the e5-square ...

I also prefer the pawn capture. Disconnecting the pawns usually does not bold well in the endgame.

19.Ne5 seems to offer some interesting possibilities. Sample line: 19.Ne5 Ba4!? 20.Qe2 Bb4!? 21.Nxg6! hxg6 22.Qe7! with a strong attack.

White is certainly in a gifting mood this game.

White's attacking prospects have taken a big hit, it seems to me.

This trade (or taking with the bishop) on the other hand, looks forced. There is not real advantage to be gained but of course you have no option to leave the knight on that lovely square. Compare to white taking on move 19 where there were many other options and so not forced but also not fantastic. This is a good learning idea for new players to determine when trades are good or not. Perhaps white thought he was winning and so it was fantastic but I don't see what he was seeing.

I would have taken with the knight, myself, but there is in my view almost nothing in it. After 20...Nxe6 21.dxe6 Bc6, Black has considerable pressure down the long diagonal. But after 20...Bxe6 21.dxe6 d5 22...Re8 or 22...Bxb2 (depending on White's 22nd) Black would have a fine game.