I'd love studying sages to highlight white's mistakes in this game!

Don't knock the end game! But I agree: the Latvian Gambit has an undeserved notoriety.

Probably White's best. For a long time I thought 3.Bc4 was the way to go, and I have played this line and annotated it on Gameknot. The one game extant in my personal database for Black, I had this knight move played against me: game

Yes, 3. Nxe5 is regarded as white's best continuation. I have excellent results against 3. exf5. If white is booked up on this opening, it can be pretty frightening. You can suddenly find yourself playing "hope chess", which is never good.

...and this invites white to play what he played, thinking it's a good and surprising idea.

3...Nf6 was my response in the above-mentioned game. The received wisdom used to be that 3...Qf6 was the right move. I've never trusted that advice.

Yeah me neither, I used to play 3...Qf6 as well. I had some good times with it...but also some really bad times. Bad, meaning embarrassing. I've since found that 3...Nf6 gives me much more flexibility and stable chances (if such a term can be used in this opening) to equalize and set traps.

This is a thematic threat in the London opening I often play. But like this, in its most pure form, has its flaws, as the game seems to illustrate. White needs to further prepare the attack in order to make it successful. A very good lesson!

Qh4 seems good, but it isn't, since after Rg8 the N can't retire on f4 (Rg4!), and even after NxB, Rg4!, followed by Rxe4+ wins.

Yes, you are correct. Good point.

6.Qh4 does appear to be a playable line, after 6...Rg8 7.Nxf8 Rg4 8.Qh3 ... (8.Qh6 Rxe4ch etc, just the same as what follows) Re4ch 9.Be2 Qe7 10.Nc3 Rxe2ch 11.Nxe2 Nd4 12.Qd3 holds everything. That final queen move is not available after 8.Qh6, and that seems to make all the difference.

I don't think White has any choice in the matter...

archduke is absolutely on target, nothing else makes sense

7...Qe7 seems to be worth more than a glance.

also, did you consider 7..Kxf8

I think this is The big white's mistake. White should be prepared to play g4 and support his f5 pawn, but it'd better than exposing himself like that. And... pray!  

Maybe 8...Nc3 was the way to go?

how about 8.exf5?!

This advance looks very strong.

8 ... d5 !! did change the tempo, it was double attack queen and pawn, good move.

10. Be2, looking for (say black plays the aggressive Nd4), 11. Bh5+, NxB, 12. QxN+, Rf7, 13. Qe5+, a nice clamp on black, since if Kf8, Qh8+, and if Re8, Qh+ forces the Q's exchange, and with it white solves the worst part of his problems with Na3 or even Kd1.

Qd4 was very tempting, thinking in Ng4. White can only 1) castle (c3 is bad after Qb6). interposing at any time 2) BxN+. Hard to assess! Say 10... Qd4, 11. o-o, Ng4. f2 can't be easily defended, but white has 12. BxN+, bxB, 13. Qh5! and either black retreats one of his attacking pieces on f2, or pins it. ...Kd8, Qh4+ just gives white more defence on f2. True white is very badly placed, but if white survives the storm, he will be winning.

I was looking at 10...Qd4, too. But it does seem, as damalfi indicates, that it is too blunt an instrument.

12... Ng4 comes to mind, with the same result: strong attack on f2 white can survive with checks on the BK while trying to find a move.

I like the nonce word 'aggress'!

I think you understimate Ng4's potential, and white could have castled, yes, but this is not meaningless. White also had 13. Qb5, virtually forcing the Q's exchange, after which the attack on f2 can be dealt with.

After 13. Qb5 Bd7 14. Qxc6 Bxd6 15. h3 Rg8 white can't reasonably castle without the pressure that follows: 16. 0-0 0-0-0 17. b3 Rg6 et cetera. But you make excellent observations. Truth be told, I overlooked Qb5