At the time of this game, Krause had a rating of 2234.

<3.g3>

In the 2nd round of the same tournament, in the game Carlsen vs P Johansson, 2002, Carlsen had played 3.e3 here, with the following moves: 3...Nf6 4.d4 g6 5.Bd3 Bd7 6.0-0 0-0 7.Nc3 Bg4 8.h3 Bxf3 9.Qxf3 dxc4 10.Bxc4 Nbd7 11.Rd1 e5! 12.d5! e4! 13.Nxe4 Nxe4 14.Qxe4 Nb6 15.Bb3 cxd5 16.Bxd5 Nxd5 17.Rxd5! Qb6 18.Qd3 ⩲

Carlsen either had analyzed the previous game, which he had won, and found this fianchetto line to be more promising. Or, more probably, he feared some preparation by Rause, which he decided to avoid.

Of course, Magnus could also have picked 3.d4, transposing into well trodden paths, or 3.b3, or even 3.cxd5, all moves played thousands or, in the case of 3.cxd5, at least hundreds of times! But so is 3.g3...

<3...Bg4 4.Bg2 Nd7> An interesting line of the Reti Opening, getting the bishop out before playing Nbd7 and e6. Black intends to trade this 'bad' bishop for a white knight.

<11.Nc3 Bd6 ⩱> Looks quite natural, but better seems 11.f4=/ ⩱

<12.Re1 Ne7! ⩱> 12.Re1 was threatening to capture on d5. Still, maybe better was 12.Bd2 Rc8 ⩱ / = Stockfish 9 - 64 bits POPCNT.

<14...Bxe7▢> If 14...Qxe7? 15.Nxd5! ± Stockfish 9 - 64 bits POPCNT.

<16.Rad1?! Rac8 ∓/ ⩱> 16.Rec1 ⩱ or 16.Rac1 ⩱ seem more logical to me. Also, there is the interesting 16.Ne2!? ⩱ all moves validated with Stockfish 9 - 64 bits POPCNT.

<19.Ne4?! Bg7 20.Nc5 Qc6 ∓> Nice looking move, but 19.Ne2! Qa4 (or 19...h5 20.Rd2 ⩱ / ∓ Stockfish 9 - 64 bits POPCNT) 20.Qe3 ⩱ / ∓ Δ if 20...Qxa2 21.Ra1 Qxb2▢ 22.Reb1 Qc2▢ 23.Rxb7 ⩱ / ∓ Stockfish 9 - 64 bits POPCNT.

After 20 moves, it becomes clear that Carlsen has been outplayed by Rause, Black having a clear advantage with a strong bishop on g7, a better pawn structure and a firm implantation on the c-file, while the white knight appears exposed on c5.

<21...Bxd4? 22.Rxd4▢ Qxc5=> Krause is too eager to collect some material! Patience was required here : 21...h5! 22.Nb3 b6 ∓ with multiple ways to harass White: e.g. Qa4, Re8, Rc8, in a kind of cat and mouse game. Stockfish 9 - 64 bits POPCNT.

<23.Red1? Rxd4 24.Rxd4! h5 ∓> Back into trouble, justifying Black's 21st move... Carlsen should not have been afraid of trading the queens: 23.Rxc4! and now either

A) 23...Qxe3 24.Rxe3! dxc4 25.Rc3! Rc8 26.b3▢! b5 27.Kf3!= Stockfish 9 - 64 bits POPCNT.

<26.b3?! Rc8!-+> White is now on the verge of losing and must be very cautious as there is no margin for errors anymore. 26.Rb4! ∓ Stockfish 9 - 64 bits POPCNT.

<29...Qc5(?!)-+/ ∓> Krause is trying to trade queens, but better is 29...Qa1!-+ with a possible Rc8-c1. Stockfish 9 - 64 bits POPCNT.

<30.Rd4(?!)> Were the players in time trouble here? 30.a4 or 30.Qe5 were better, though Black is possibly winning whatever White plays. Stockfish 9 - 64 bits POPCNT.

<30...Qd6?! ∓> The Black queen should remain inside White's position, making sure the c-file control is not challenged by Carlsen and always threatening to enter the black rook inside White's position as well. Better 30...Qc2 31.Rd2-+ repeating the position after 28.Rd2, or 30...Qc1 31.a4 a5!-+ Stockfish 9 - 64 bits POPCNT.

<31.Rd2?!-+> Right square, but wrong piece! 31.Qd2! ∓ Stockfish 9 - 64 bits POPCNT.

<31...b5?!-+> Probably still winning, but Krause appears to be unsure as to how to make some progress, an might as well have been short on time to devise a clear plan here. 31...Rc1!-+ Stockfish 9 - 64 bits POPCNT.

<34...Qc2> 34...Qe1! 35.Qe5▢ Qxe5! 36.fxe5 a5!-+ with a strong passed pawn in the center and nothing to compensate for White... Stockfish 9 - 64 bits POPCNT.

<35.Qe5?!> Better is 35.a4 Qxb3 36.f5 gxf5▢ 37.axb5 axb5▢ 38.Qg5+ Kf8▢ 39.Qh6+ Ke7! 40.Qg5+ Kd6! 41.Qxh5-+ With this line full of forced black moves, if Krause was short on time there was a possibility she would miss one of them while trying to reach the time control... Stockfish 9 - 64 bits POPCNT.

<35...Qc7?!> Krause is still trying to trade queens, while on the previous move she could have forced it advantageously... Now, here why not 35...Qxa2! 36.f5 gxf5! 37.Qe3 Qb2 (or 37...Qa1 38.Qg5+ Kf8▢-+ Stockfish 9 - 64 bits POPCNT) 38.Qg5+ Kf8▢-+ Stockfish 9 - 64 bits POPCNT.

<36.Qf6> Better: 36.a4 Qc2!-+ Stockfish 9 - 64 bits POPCNT.

<36...Qd8?! 37.Qe5-+> The queen trade is becoming an obsession for Black! 36...Qc2! repeating the position after 34...Qc2. Stockfish 9 - 64 bits POPCNT.

<37...Rc2 38.a4!-+> 37...Qc7! repeating the position after 35...Qc7 and getting back on the path to victory!

<38...bxa4?> 38...Qc7! 39.axb5 axb5 40.Rb4 Qc5 41.Rd4 Qc7 42.Rb4 (repeating the position) Qb7-+ passing the time control! Stockfish 9 - 64 bits POPCNT.

<39.bxa4?> Did Carlsen consider 39.Rxa4!-+ Stockfish 9 - 64 bits POPCNT.

<39...Qc7?!-+> 39...a5!-+ Stockfish 9 - 64 bits POPCNT.

<40.Qe3?! Rc4!-+> 40.Qxc7 Rxc7 41.Rb4! Rc6! 42.Rb8+ Kg7▢ 43.Ra8 Kf6-+ Stockfish 9 - 64 bits POPCNT; also 40.Qf6-+ Stockfish 9 - 64 bits POPCNT.

<41...Qc5 42.Rxc4 Qxc4▢-+> Why not 41...41...Rxd4! 42.Qxd4 Qc4!-+ followed with d5-d4. Stockfish 9 - 64 bits POPCNT.

<43.Qb6?! d4! 44.Qb8+ Kh7> Better is 43.a5 d4! 44.Qe4 Qa2 45.Qxd4 Qxa5-+ Stockfish 9 - 64 bits POPCNT.

<45.Qf8 Qc7▢ 46.Qb4-+> Better is 45.Qb7 Kg7▢ 46.Qe4-+ Stockfish 9 - 64 bits POPCNT.

<46...Qd8> 46...d3!-+ Stockfish 9 - 64 bits POPCNT.

<48...Qd5> 48...Kh7! Stockfish 9 - 64 bits POPCNT.

<52...Qb4 53.Qc8+ Kg7 54.Qxa6 Qe1+! 55.Kg2 Qe4+! 56.Kh2 Qe1 57.Kg2> repeating the position after 55.Kg2. The simple 52...d3! 53.Qd8+ Kg7▢ 54.Qxd3 Qa1+! 55.Kg2 a5!-+ was doing the job, controlling the long dark diagonal, while White has nothing to stop Black's a-pawn. Stockfish 9 - 64 bits POPCNT.