Evans Above

I MESSED up an excellent position against the Evans Gambit when playing for Battersea last night.

James Stevenson (172) - Spanton (163), London League 3 (Battersea 2 v Hammersmith)

1.e4 e5 2.Nf3 Nc6 3.Bc4 Bc5 4.b4 Bxb4 5.c3 Ba5 6.d4 d6!?

Black has played 6...d6!? instead of the much more common 6...exd4

Easily most popular is 6...exd4, but the text has been played by Adolf Anderssen, Anand and Caruana.
7.0-0
7.dxe5 was played in Alan Spice (155) - Spanton (151), Highbury (rapid) 1990. After 7...dxe5 8.Qb3 Qe7 9.Ba3 Qf6 10.0-0 Bb6, White has some compensation for his pawn sac, but Black has no weaknesses and both Stockfish9 and Komodo9 prefer Black. I did eventually win the game in a time scramble.
The main line in ChessBase's 2018 Mega database runs 7.Qb3 Qd7 8.dxe5 Bb6. The most recent game in the database between two 2400+ players continued: 9.Bb5 (correspondence master Tim Harding's recommendation in the revised second edition of Evans Gambit And A System Vs. Two Knights' Defense [punctuation is Chess Digest's]) Nge7 10.0-0 0-0 with an unclear but approximately level position in Bogdan Belyakov (2487) Jure Borisek (2561), World Blitz Championship 2016 (1/2-1/2, 62 moves).
7...Nf6 8.Qb3?!
Qd1-b3 is a common move in the Evans, but here it may be a mistake.
Perhaps White should settle for winning his pawn back by 8.dxe5 dxe5 (8...Nxe5!? is also possible) 9.Qxd8+ Nxd8 10.Nxe5, although Black is comfortable after 10...0-0 or 10...Be6.
It's important to realise that 8.Qa4 is well-met by 8...Nd7.
8...0-0 9.dxe5 Nxe5?
As JS suggested afterwards, I should have played 9...Nxe4, which my analysis engines consider to be nearly winning.
Instead, the game continued:
10.Nxe5 dxe5 11.Ba3
Black is losing the exchange. There is serious compensation, but I mishandled the position and lost without putting up any decent resistance (1-0, 33 moves).