Karpova: This game was played in the German Chess Club Vienna. Grünfeld annotated the game (I only give some excerpts).
On 1...d5, Grünfeld notes that recently <1...Nf6> became more popular. Both, Bogoljubov's <2.Nf3 e6 3.c4 Bb4+> and Grünfeld's <2...g6>, are sufficient for Black to equalize. He refers to his article "Das Pistyaner Turnier in seiner Bedeutung für die Eröffnungslehre" (Arbeiter-Schachzeitung, number 9/10).
3...Nf6: Tarrasch's <3...c5> has vanished from tournament play completely, since White has several good ways of handling it. The best one is the variation introduced by Schlechter and elaborated (<ausgearbeitet>) by Rubinstein - <4.cxd5 exd5 5.Nf3 Nc6 6.g3! Nf6 7.Bg2 Be6 8.0-0 Be7 9.dxc5 Bxc5 10.Na4! Be7 11.Be3! b6 12.Nd4> (Reti vs Tarrasch, 1922).
Grünfeld considers 5.Nf3 to be slightly better than <5.e3> (<5.Nf3 dxc4 6.e4!>).
According to Grünfeld, 6...h6 was only played to avoid theory. It's usually not advisable to drive the ♗ to h4 in the Orthodox Defense, since White can avoid the trade against the ♗e7 later (<...dxc4 Bxc4 ...Nd5 Bg3>).
9...Be6: <9...Bb7> is in accordance with the position, but White plays <.Ne5> and <.f4> and gets an attacking position, against which no antidote has been found yet.
With 14...Bh4, Tartakower was already planning the exchange sacrifice, since he felt he had been outplayed. The sacrifice is not sound, though. Grünfeld gives <14...Bb4! 15.Bb7 Bxc3+ 16.bxc3 Qe7 17.0-0 Rad8 18.Ba6 Nf6> instead.
16.Bxa8<!> (<16.hxg3 Rb8 17.Bxd5 Bxd5 18.Nxd5 Re8+ 19.Ne3 Nf6 20.0-0 Re4> and the d-♙ becomes too weak, according to Grünfeld).
Grünfeld gives <22.Qd3! Bf5 (or 22...f5 or 22...Nxc3 23.Qxg6) 23.Rhf1 Qd6+ 24.Kg1> as much clearer for Wite.
Grünfeld suggests <27.h4>.
According to Grünfeld, weaker than 27...f4 was <27...Qxe5+ 28.dxe5 e3 29.Rfd1> and the White ♖s invade the Black position (and White has a passed ♙).
30.Rc6<?> Grünfeld gives <30.Rc7> (sample line <30.Rc7 Rf7 31.Rc6 Re7 32.d5 Bxd5 33.Rxh6 Kg7 34.Rd6 Bb7 35.a4>. Now the Black ♔ can advance and support the passed ♙s.
36.Rd7 threatens to win the ♗ (<37.Re7+ Kf5 38.Re5+>), but I wonder if 36...Rf7 was not a better way to meet the threat, than 36...Re8.
On 37.Rg7, Grünfeld notes that also <37.Kg1> would be insufficient to save the game (<37.Kg1 Be6! 38.Rg7 e2 39.Re1 Ke3 40.Rg5 Bxh3! 41.gxh3 f3>).
Source: 'Österreichische Schachrundschau', July 1922, issue 5, pp. 34-36