1. d4 d5 2.c4 c6 3.Nf3 Nf6 4.Nc3 e6 5.e3 Nbd7 6.Bd3 dxc4 7.Bxc4 b5 8.Bd3 Bb7
8...Bb7 (which Steffen Pedersen calls the Wade Variation) is probably the main move, though Anand played the riskier 8...a6 against Kramnik and other top players sometimes prefer it. (8...Bd6 and 8...b4 are also playable though much less often seen.)
Anyway, after 8...Bb7 play habitually proceeds 9.O-O a6 - mark that sequence well, you need to remember this - 10.e4 c5
and now because 11.e5 isn't going to work after 11...cxd4 12.exf6 dxc3 White normally plays 11.d5, known as the Reynolds Variation, in which many grandmaster games have been played and which is currently seen as all right for Black. (Provided that is, a number of pitfalls are avoided, starting by not capturing on d5, which error was committed against me one weekend in the 4NCL.)
All right, but what if we play 9.e4 instead of castling? This is the trick. If Black, knowing that the idea is to move the bishop first and then the a-pawn (and possibly not knowing much more theory than that) plays naively 9...a6 then White has 10.e5!
Now Black is suddenly in huge trouble. If this is not immediately apparent, the point is that after the natural 10...Nd5 and the subsequent exchange on d5 Black either doesn't have the freeing ...c5 any more, because after 11...cxd5 there's no c-pawn, or, after 11....exd5, is going to lose very quickly to something involving e6 with the Black king trapped in the centre. Pedersen calls the position after 11...cxd5 "very good for White" while Vera (Gambit, 2007) calls it "clearly unfavourable for Black". Black gets a dead bishop: White gets a big space advantage and an attack on the king.
So why is 9.e4 more rare than 9.O-O? Largely because if Black knows what they're doing (or thinks about it instead of playing the automatic move) there is 9...b4!
9...b4 is, by the way, playable in the main line with 9.O-O, but not so good because there, the central e4 square is available for the knight: now it's not, and as we know that 10.e5 doesn't work when the pawn's already on b4, White must play 10.Na4 after which the knight can now come to d5 and we have 10...c5 11.e5 Nd5.
This position is no disaster for White - many fine attacking games have been won from here - but we can see that Black has successfully got in both Nd5 and c5, the moves he wanted to play. In general it's not so testing as the line with 9.O-O and for that reason is less popular with grandmasters. (Come to that, the whole 5.e3 line may be less testing than 5.Bg5, though that's a very complicated story.)
However, my experience has been as follows. I've tried the trick 9.e4 four times (I think) all against players rated a little lower than I. Every time, they've followed up with 9...a6?, allowing me to play 10.e5. Now as it happens I've not won all these games - I actually think I may only have won one of them - but nevertheless, a plus-over-minus position is not to be sniffed at after fewer than a dozen moves if somebody's going to hand it to you. The problem with it is - in order to try the trick, you have to play the second-best line. If they know what they're doing, you've deliberately put yourself in an inferior line. But on the other hand, everybody does seem to fall for it.
So what do you do? How far are you prepared to base your openings on what you hope your opponent doesn't know? Or put it another way - how far are you prepared to play for tricks?