I've been busy during the last fortnight but have come out with the latest update on the Evans Gambit.
Firstly, a recap on 4...Bxb4 5.c3 Be7. In Nigel Short's preferred line, 6.d4 Na5 7.Bd3 d6, if White continues with 8.dxe5 then Black has the independent option 8...Bg4, as pointed out to me by Mark Nieuweboer, so White should in my opinion consider the move-order trick 8.0-0, intending 9.dxe5, and in that case 8...Bg4 is met by 9.Nbd2 reinforcing the knight on f3.
Instead, after 5...Bc5 6.d4 exd4 7.0-0 Bb6 8.cxd4 d6 (or 7.cxd4 Bb6 8.0-0 d6) we reach the so-called Normal Position.
Tim Harding wrote extensively about this in his two-part series about Chigorin and the Evans Gambit. I don't really trust 9.d5 Na5 10.Bd3, but 10.h3, preventing 10...Bg4, might be good enough. Instead 9.Nc3, the preference of Morphy and Chigorin, looks good enough to provide enough compensation for the pawn following 9...Na5 10.Bd3 Ne7 11.Nd5 (or 11.h3) and 10...Bg4 11.Nd5 (sacrificing a second pawn, also good is 11.Be3). Black generally retains the extra pawn, so the "Normal Position" offers both sides good scope to play for a win.
Instead critical is 5...Ba5, whereupon after 6.d4 (I don't trust 6.Qb3 Qe7, followed by 7...Nf6, or 6.0-0 Nf6, both of which give Black chances to safely castle kingside) Black can play 6...d6 whereupon 7.0-0 Bb6 leads into Lasker's Defence.
Traditionally this has led to a double-edged game after a long and fairly forcing line: 8.dxe5 dxe5 9.Qb3 (9.Qxd8+ Nxd8 10.Nxe5 Nf6 is equal) 9...Qf6 10.Bg5 Qg6 11.Bd5 Nge7 12.Bxe7 Kxe7 13.Nxc6 Qxc6 14.Nxe5 Qe6 15.Nc4 Rd8. However, Mihail Marin pointed out the deviation 11...f6 in Beating the Open Games, which appears to be better for Black, and was successful in Kogan-Anand, Venaco 2005. Thus White should avoid this line.
Thus White tends to continue 7.Qb3 Qd7 and grandmasters have largely abandoned the old main line 8.dxe5, striving to open the d-file. I am guessing that this is because of Black's method of returning the gambit pawn with 8...Bb6, rather than the attempt to hold it with 8...dxe5. White typically gets better piece play, but an inferior pawn structure, e.g. 9.Nbd2 Na5 10.Qc2 Nxc4 11.Nxc4 d5 12.exd5 Qxd5 13.Ne3 Bxe3 14.Bxe3 Ne7 (Li-Ni, Manila 2007).
Nigel Short has tried out 8.Nbd2 Bb6 9.a4 which worked well for him in Short-Sargissian, Wijk aan Zee 2008, but 9...Nf6 may be a more critical test than the 9...Nh6 played in the game.
Finally we come to the Compromised Defence with 6...exd4 7.0-0 dxc3, which is reminiscent of some lines of the Danish and Goring Gambits, but with improved chances for White. The most critical line runs 8.Qb3 Qf6 (8...Qe7 9.Nxc3 sets up the dangerous threat of Nc3-d5, and if 9...Bxc3 10.Qxc3 and if 10...f6 11.e5) 9.e5 Qg6 10.Nxc3.
Most reliable is 10...Nge7 11.Ba3 0-0, getting the king to comparative safety on the kingside, but White can generate very strong kingside threats, particularly with the black queen exposed on the kingside. Computers tend to assess these positions as equal but in practice Black's defence is very difficult.
An interesting sideline after 7.0-0 is 7...Nf6!?, also examined by Tim Harding at one of his later articles (http://www.chesscafe.com/text/kibitz119.pdf). I think White's best bet is probably 8.e5 rather than the old book recommendation 8.Ba3 and the resulting positions are somewhat messy, but Black has to exercise considerable care to navigate through the complications without trouble.
Of course my coverage of the Evans isn't finished- there is still 7.0-0 Nge7 and the 7.Qb3 lines to examine, which are probably the three most important lines for the theoretical assessment of the Evans.