Dr. Siggy: In this classical opening, 4...Bb6 followed by 5...Qe7 is a good old defense, reintroduced to tournament practice by Alekhine in 1923. Its objective is to maintain the pawn at e5 and thus to prove that 4.c3 is a non developing move.
By the continuation 6.d5 Nb8/d8 7.d6!?, White offers an important pawn to hinder for some time the development of Black's queen's bishop. It is clear that, after the usual reply 7...Qxd6 8.Qxd6 cxd6, Black's game is seriously cramped. The question is whether this compensates White for the loss of such an important pawn.
I'm inclined to believe it does not, specially if Black proceeds as Alekhine did in this particular game: 6...Nd8! 7.d6!? cxd6(!) - which seems much, much better than 7...Qxd6, since the queen is one of Black's best placed figures on the board and the queen's knight is able to go not only to c6 but also to e6.
According to P. Moran, Alekhine played this game like he did when he was at his apex. Check out 13...Ne6! (threatening 14...Nf4! and obtaining the initiative), 17...b4! and 18...b3! (crippling White's queen's wing completely), 20...Bb7 (getting the queen's bishop out of the prison), 23...Bd8!, 24...f5! and 25...Bg5! (activating the king's bishop powerfully), 31...hxg5 (unabling White to react with f4), and finally 34...Rc5! (sacrificing a pawn to put it an end): simply brilliant!...
This Alekhine's gem repays a very lengthy study, at the very least because it's the only game I know where 7...cxd6!? is used (and quite successfully...).