1.d4 Nf6 2.c4 g6 3.Nc3 d5 4.Nf3 Bg7 5.Qb3 dxc4 6.Qxc4 0–0 7.e4 Nc6 8.Be2 e5 <Quite a rare continuation. 8...Bg4 is main line.> 9.d5 Nd4 10.Nxd4 exd4 11.Qxd4 c6 <So the knight can take on d5 if white plays 12 e5.> 12.d6 <Novelty I think, 12 Qd1 or 12 Qc4 have been played before.> 12...Nd5 13.Qd3 <13 e5 Qxd6> 13...Nxc3 14.bxc3 Qf6< Some punching and counter-punching here. Black obviously threatens 15...Qxc3 winning >15.Bb2 <Now white threatens 16 d7 winning.> 15...Rd8 16.Rd1 Qe6 17.f4< Black will get his sacrificed pawn back, but Carlsen decides it's not going to be the d6 pawn. [17.a4 Be5]> 17...Qxa2 18.Rd2 Qa5 <This doesn't seem to achieve anything, so could be considered some what passive if the idea is to retreat to d8. The computer seems to like ...a5 and ....b5. Whites center does seem intimidating though, and it is probably best to keep an eye on that, rather than hitting out with a wing attack, because Dominguez isn't a computer [18...a5 19.e5 b5 20.0–0 a4 and black is at least creating some counterplay.]> 19.Qe3 Bd7 20.Kf2 Re8 21.Ra1 Qd8 22.c4 Bxb2 23.Rxb2 b6 24.Bf3 Qh4+ 25.Kg1 Qf6 26.Qd2 g5 27.g3 gxf4 28.gxf4 Kh8 29.Kh1 Rg8? <Carlsen now gets a free move to start his central push. [29...Qh6 30.e5 f6 and black can attack whites pawn chain.] >30.e5 Qh4 31.Qd4< An innacuracy by Magnus. If Dominguez now plays the most accurate defence 31...Rae8 then if 32 e6+ f6 white can't play 33 exd7 because of 33...Re1+ winning. With the white queen on c3, the e1 square is protected. So 31 Qc3! was much better. Magnus still has a strong position though.> 31...Rg7? <[31...Rae8 is better - see above]> 32.Rg2 <[32.Rxb6!] >32...Rag8?< [32...c5 33.Qf2 Qxf2 34.Rxf2 Rag8 35.Bd5 Puts up a better fight.]> 33.e6 Bxe6 34.Rxg7 Rxg7 35.Bxc6 f6 36.d7 Bxd7 37.Bxd7 Re7 38.Be6< Go on then, you can have it!> 38...Rxe6 39.Qd8+ Kg7 40.Rg1+ Kf7 41.Qg8+ Ke7 42.Rg7+ Kd6 43.Qf8+

Nice game by Carlsen! Dominguez never solved the problem of the d6 pawn, and consequently created no play of his own and got slowly strangled.

Yeah, apparently it's a novelty; both Christian Bauer on chessdom and Dennis Monokroussos on his blog have suggested <12...Re8> in response as an improvement for Black - apparently White would have to part from one of his two central pawns, instead of the a-pawn in the game. Monokroussos gives the following analysis:

<31.Qd4 An innacuracy by Magnus. If Dominguez now plays the most accurate defence 31...Rae8 then if 32 e6+ f6 white can't play 33 exd7 because of 33...Re1+ winning. >

I don't think Qd4 was an inaccuracy, at least objectively (nor does Rae8 appear to be better, subjectively speaking). White doesn't play exd7 if 32. ... f6. Instead white plays e7!

After 31. Qd4 if black responds with (sorted from best to worst for black):

Be6: then +1.88 for white
Rg7: then +2.72 for white
Rae8: then +3.73 for white

31. Qc3 which you suggested is significantly weaker than Qd4 (black is able to put up a tougher defense):

31. ... Rg7 32.e6 Bxe6 33.Bxc6 Rag8 34.Qd4 Qd8 35.Rg2 f6 36.Re1 Bd7 37.Be4 Rxg2 38.Bxg2 Rg7 39.h3 Qc8 40.Kh2 Qf8 41.Be4 Rg8

I did see the e7 move was strong, but didn't let the computer run for very long to give the higher evaluation for white. I knew that no defence was that good.

Yes, I shouldn't be calling Carlsen's 31 Qd4 an innacuracy as I didn't study the other lines thoroughly enough. My bad

The analysis is at the bottom of page one of my blog, and I used the program Chessviewer Deluxe to post this analysis on my blog. I highly recommend this program to those of you interested in posting chess games to a blog or website. Go to http://chesstuff.blogspot.com/ for more information about this free product.