Analyzing it quickly with the engine shows that your bishop move fails to 10...f6, where the pawn is pinned, and 17...Rh6, preparing to block the check, but these moves are again difficult for opponents to see c:
A beautiful Halloween Gambit's game analyzed

Analyzing it quickly with the engine shows that your bishop move fails to 10...f6, where the pawn is pinned, and 17...Rh6, preparing to block the check, but these moves are again difficult for opponents to see c:

@Eized, yeah, I find the Nc6 retreat slightly weaker and enjoy facing it as well, and I was glad to be lucky enough to still force a pawn to the 6th rank with 14N.d6 after the Nb6 retreat.
Yeah, when I take the time hand analyse a rather quiet game afterward, my analysis is pretty decent, save for one or two missed tactics of various difficulty. For this game, computer analysis reveals that mine is far from objective, to say the least ^^.
12h5 is also inaccurate as after hxg5 hxg6 fxg6, I have to either agree to some kind of queen trade or lose a tempo, and my initiative vanishes.
As good as Nd6 is, 13Ne4 also is extremely dubious, although I can't think of another move that keeps the pressure high and the position difficult to asses.
I don't know why I said Qf8 was forced, I also saw f6 before playing Bg5 and found it an annoying defense, but now, I can't figure a way to edit the analysis, so I'll left it there.
My answer would have been
During the game, when I analyzed continuations to choose a move to make, I assumed I was already a piece down, and considered ok continuations where, against best defense, I was no more than a piece down at the end. Among ok continuations, I went for the trickiest, and also disregarded slower moves that are probably objectively better. I thought f6 could be between -1 and -3 (my acceptation threshold), turns out it is a bit more in favor of black.
But more importantly, I failed to correctly assess the (objective) value of white initiative, I thought it was only worth between half a pawn and a whole pawns, turns out it is worth even more (1.5), and computer says the position before Bg5 is at -0.5, so I should not have been satisfied with continuations where I end up down a whole piece and where my attack is no longer as strong, thing that I also assessed incorrectly.

Beautiful game, indeed ! That 7. ..., Bb4 line it’s my most feared one when playing the Halloween. My opponents usually trade their Bishop, and, thus, not only they simplify the game (simplify when winning rule) but also eliminates the Knight that could become an important threat.
So, when facing that line, the pawns become my most important weapon to fight with.
Since I’m not good at analyzing games (the language barrier), would you analyze this one I just finished playing ? It’s another approach to the 7. ..., Bb4 line. You’ll be surprised with how it ends...
Hey, just wanted to show this miniature of mine where I played the Halloween's gambit and analyzed it afterward.
I would gladly take any inputs about this game and my analysis of it, the game was very dense and tactical and I might have missed strong moves that me or my opponent could have played.
Also don't mind the possibly bragging tone of the analysis, I made it for a game showcase initially (yep, bragging and advertising on game analysis).
The game :
https://www.chess.com/daily/game/153857756
the analysis :