You were obviously better before the sacrifice, so i'm not sure why you felt compelled to take such a risk. It looks like your opponent had to blunder twice to make it work for you, first by moving the knight, and allowing your queen to come to d6 was pretty suicidal, but even then black has play until he tries Qb6. I think if black castles instead of taking your knight, and you trade off for blacks bad bishop as you planned, then black has a lot of play. In game, under time pressure, I think the sac is probably sound, particularly in blitz, but if you put it in an engine i think it'll find the move throws away at least some of your development and space advantage.
Thanks, Mike...that's about the answer I expected. It wasn't that I felt compelled to take a risk, it was more of a whim. I had a vague idea that if black declined to take my knight and I did take the bishop, black's kingside would be weakened, but now that I look at it I think it would have been actually strengthened.
So out of curiousity, what would you have played instead of 20. Nc5? Perhaps 20. Qc3? I realized I have the advantage at this point, but it was unclear to me how to capitalize on it.
I would have looked to infiltrate black's camp via the weakness the c file, by doubling my rooks on that file (there's no Qc3 without the knght moving first, which seems like a waste of a tempo, since it only has the very passive Nd2) and then hitting c6 or c7, depending on black really. The game would have been a bit longer, but i think its only a matter of time before your queenside pressure wins either the knight or the b6 pawn, particularly if black remains uncastled.
Looking at the game, the sacrifice is quite the shocker and a move that I couldn't believe when I first saw it, but as the game continued, it started to come together quite prettilly. With best play, though, I think your pawn roller could have failed. Black lost the thread on his 26th move, bringing the knight over from b6, and giving you the chance to win back your material with an advantage then and there with a second sacrifice that immediately jumped out at me, screaming to be played. Had the knight remained where it was, black could have set up a blockade of your passed pawns, and I think it would have been very difficult for you to make any progress. I submit this missed opportunity, on move 27, for your consideration; personally, I think it wins by force but I might have overlooked something. Regardless of the inaccuracies, your knight sacrifice was a beautiful, and astonishing conception. Let me know what you think of my analysis.
Thanks, GambitKnight! That sacrifice did cross my mind as well, but again, I think black could have countered....for example, by threatening checkmate (see below).
Ultimately, I think my ploy not only would have failed against a better player but was unnecessarily risky. But hey, it made for an interesting game. 
(boneheaded alternate scenario deleted)
6 Nxe6 Qe4? 7. Qxc8#
ScottSchafer: your line is fundamentally flawed. ...Qe4?? allows a mate in one. After the exchange sacrifice, with optimal play, white seems to get an excellent endgame, and with poor play, a crushing attack. To be honest, I don't see any decisive counterattacks that do not stem from a blunder on white's part. Of course, the endgame itself could be tricky and difficult to convert, but the winning chances are still there, and barring something catastrophic, I reckon that black would have to fight for dear life just to hold the draw.
ScottSchafer: your line is fundamentally flawed. ...Qe4?? allows a mate in one. After the exchange sacrifice, with optimal play, white seems to get an excellent endgame, and with poor play, a crushing attack. To be honest, I don't see any decisive counterattacks that do not stem from a blunder on white's part. Of course, the endgame itself could be tricky and difficult to convert, but the winning chances are still there, and barring something catastrophic, I reckon that black would have to fight for dear life just to hold the draw.
You're right, my move was fundamentally flawed. I'm glad I didn't play so badly in the actual game. Black is better off playing fxe6 (pawn takes knight) instead of Rxc5, I think, but white takes the a-file pawn and I think the result is the same.
Thanks for taking the time to look at this and present your idea. After looking at it more closely, I like it a lot.
The great thing about sacrifices is they don't have to be entirely sound. They just have to be good enough that your opponent doesn't defend accurately. For example, here's a game I played a while back OTB blitz. It took some serious thought for me and other chess.com people to annotate it and find the accurate reactions that would leave my opponent better. The time controls were 5/0 and accurate defense often takes more time.
