After 1. d4 d5 2. c4 c6 3. Nc3 e5?! the Opening Explorer says:

Move :: white wins / draws / black wins
4. dxe5 :: 46.4% / 32.1% / 21.4%
4. e3 :: 40% / 32% / 28%
4. cxd5 :: 27.8% / 27.8% / 44.4%

Oddly, I think that the last move 4.cxd5 is probably White's best continuation although it got a bad rap due to the statistics.

4.e3 is what I consider a patzer move, even though it's not bad. I consider 4.e3 it as White's way of saying "I don't want to play into your weird book lines, let me just play timid chess even if I get slightly cramped, and let's get into a reasonable middlegame and play chess."

Also note that the 4.cxd5 variation is something that comes around in the otherwise drawish Slav-exchange: 1.d4 d5 2.c4 c6 3.cxd5 cxd5 4.Nc3 e5!?

Some important games with this line:

Krasenkow vs Shabalov, 1991 - Shabalov, one of the lines big proponents in the early 90's, scores an impressive victory that no doubt made some GM's sit up and take notice of this opening's revival.

Kasparov vs P Nikolic, 1992 - Here Kasparov takes the bull by the horns, snatching the pawn, and goes on to win a brilliant game.

<I don't think 4. cxd5 is a good move - the line 1. d4 d5 2. c4 c6 3. cxd5 cxd5 4. Nc3?! e5! is pretty good for Black.> I don't think you can say that 4.Nc3 is a bad move. Kasparov and other monstrously strong players employ the b-knight before the g-knight in positions like that.

<This is of course the reason why stronger players usually play 3. Nf3 preventing any ideas with ...e5> I think the preference toward moving the KN before the QN (and that's all it is, a preference), stems from the fact that systems that involve an early f3 are not very popular these days. When Kasparov used to play 1.d4 nearly exclusively, he would often play into variations like 1.d4 d5 2.c4 c6 3.Nc3 all the time--because in some variations he would be inclined to move the King's knight to e2 instead of f3, the idea being that now he's better placed to try to make a central break with f3 and e4.

In typical 1.d4 d5 openings, Black's position is very solid, and he develops safely and soundly, however there is ONE tiny wrinkle in Black's development plan... that awkward queen's bishop.

Efforts to develop this bishop prematurely often get slapped down, e.g. 1.d4 d5 2.c4 c6 3.Nc3 Bf5? 4.Qb3! and now Black is practically down a pawn. And that's not the only line where Qb3 hurts Black for just trying to get his bishop into the game.

That move, Qb3, is often the move which punishes Black for developing his bishop early. So, with that in mind, Black decides that he will "surrender the center" and capture on c4. So now we have 1.d4 d5 2.c4 c6 3.Nc3 Nf6 4.Nf3 dxc4! The idea of this pawn grab is twofold--first, it takes away that otherwise killer move Qb3 from White, just long enough to actually get the queen's bishop into the game. Second, unlike most lines in the Queen's Gambit, Black is actually threatening here to win that gambit pawn for good with ...b5. Because of the threat to keep the pawn, many players opt to play a4, but this move too has its drawbacks, as this makes b4 available for Black's purposes.

One of the greatest compliments about the Slav defense was by Hans Kmoch, in "Pawn Power in Chess", when he wrote, <In the Slav Defense, White has no good pawn breaks, and Black doesn't need any.>

Now azaris, <I don't think that 4...dxc4 is worth an exclam since it's the move that gives White the most iniative out of the opening> Well I'm not going to argue exclams, but my thinking is that it's a move that looks bad, but in fact entails a very well thought-out opening plan. I play ...e6 in that position these days, another move that "looks bad" but lays the path for a very solid plan of development. I don't think one move is better than the other, but the Semi-Slav I think gives both players more ways to win, so it makes for exciting chess.

<You can usually let the pawn go and still get a good game going. Certainly I'd like to face this rather than a Semi-Slav or something.> Not me, at least not in a tournament. Semi-Slav games are very double edged, and somebody is bound to win, but the Slav is one tough nut to crack when a good player handles the Black pieces. White's advantage is so tiny that it's not easy to get anything working. Of course no plan is foolproof--see Kasparov vs Timman, 1988 for a neat example of how Garry manages gets a nagging advantage out of the opening and converts it into a point.

<Recent GM practice seems to favor not even trying to hold the pawn> Old GM practice favors that as well. Pawn grabbing in the opening is rarely recommended and not for the feint of heart; but unless something is going on that I'm not aware of, White still plays a4 against the Slav, no? That's Black's cue to abandon any idea of clinging to the gambit pawn, but like I said above, Black can then enjoy the hole created on b4.