Optimal line for Muzio gambit (Sarratt defense)

This is the famous Muzio gambit sacrificing the king's knight :

There are the following possibilities:

• 6...Qf6 Sarratt defense (ev=-0.9, d=23)
• 6...Qe7 From defense (ev=-0.4)
• 6...Nc6 Holloway defense (ev=+0.0)

and others. The best option for white seems to be 7. d3 (ev=-1.0, d=33). The first historical game: Wood-Lewis (1815) http://www.chessgames.com/perl/chessgame?gid=1075443

It gives the following optimal line (by SF9+ evaluations):

Now white has many options evaluated at about -1. 

Ummmmmm.....

Would suggest otherwise.

BigBadHadron: Thanks for your analysis! So, normally, black wins except for your 21. Qxe6+ variation.  

Ahoy!

Not my analysis. There used to be a massive website run by a Mr. Thomas Stock of Germany, part of which was devoted to the Muzio-Polerio Gambit (another sizable chunk was devoted to the Belgrade Gambit).  I managed to get most of the relevant material (databases) off the site before it disappeared (although, some of it still remains on the wayback machine website).

I have always had a soft spot for the MP gambit because my late chess coach for some reason poured a lot a time into analyzing the resulting unbalanced positions (this was before the time of relaying on chess engines).

His favorite saying was (said in a thick Welsh accent) "I would always play 4.Bc4 if 4...g4 was always a given". Fortunately for me, it took me a few thrashings to work out that 4...g4 was not a given, 4...d6 was clearly better move intending Bg7 & h6 and he only ever used against me in blitz after that.

He died 15 years ago but his influence on my chess was imminence 

BigBadHadron: Thanks for telling your interesting personal story!  I see that in the following KGA position:

SF9+ (d=24) suggests Blachly gambit 4...Nc6 evaluated at -0.7 rather than canonical 4...g4 (ev=-0.5). Here 4...d6 seems to be a mistake (ev=+0.3). 

4...d6 seems to be a mistake? Awesome! There goes 181 years worth of chess theory & practice, disappeared in a puff of binary driven logic. 

Well, bro, that's the modern life with robots and engines.  However, it turns out that many classical chess lines are confirmed by modern computer analysis.