Where does this come from?

1.d4 Nf6 2.c4 d5?!

In my experience it isn't as bad as it looks, but still - it's a Grunfeld without the Grunfeld bishop and play against the center. So white really should be pretty happy seing this.

It's pretty common here I've noticed, at least amongst the lesser mortals.

http://www.365chess.com/opening.php?m=5&n=126&ms=d4.Nf6.c4.d5

Ah, so anything but the most natural move seems playable for black. Pity the most natural move (cxd5) has a crushing percentage of white wins.

I finally remembered that it was GM Ron Henley who told and analyzed 3.cxd5 with me. His assessment of the move is White's capture on d5 is premature. It breaks the tension in the center too soon and the settled situation in the center gives Black a clear plan. The first 20 moves of the game analyzed by Houdini 4 pro64 at 25 moves deep are: 1.d4 d5 2.c4 Nf6 3.cxd5 Qxd5 4.Nc3 Qa5 5.Bd2 Qf5 6.Nf3 Nc6 7.h3 h6 8.Qb3 e6 9.g4 Qa5 10.e4 Qb6 11.Qxb6 axb6 12.Bf4 Bb4 13.Bd3 Bxc3 14.bxc3 Kd8 15.Bg3 Bd7 16.a3 Ne7 17.0-0 Bc6 18.Ne5 Rf8 19.f3 Nd7 20.Nxc6+ Bxc6.

The main problem that Black faces initially is the fear of having his Q trapped or losing it in close quarters with many threats against it. From move 1 thru 20 Houdini's assessment of the position changes from (+.076) to (+.046). A final assessment of +.046 after move 20 is less than half a pawn advantage for White. Almost impossible to convert that advantage to a winning position without several suboptimal moves by Black.

I am running Houdini 4 pro 64 on a top of the line Toshiba pc with 2 CPUs. It is a velociraptor. It considered 7.e3 and rejected it in favor of 7.h3.

What are you chess engine are you running , how many CPUs and what type of PC?

@Moses2792796

Have to go to work now. Will post Houdini's analysis for both moves when I return.

IMO the above "mine is bigger than yours" argument is pretty pointless. Why would I want to play a shallowly analyzed computer line that ends with -0.5 when playing black. Computer apparently assess the position as clearly better for white. That's enough for me. Also, unassisted computer analysis of openings are pretty naff as is.

FWIW I ran this on my phone where Droidfish (Stockfish on Android) gives the lines after Qxd5 and Nc3 as + 0.7 while Critter gives it 0.5. Not encouraging either of it really.

It also seems to indicate that "better than it looks" might be a reasonble conclusion, but still much worse than for the example the proper Grun or Slav if you insist on 1. - d5.