sachistu: Although it may seem a minor point, it's not clear the score given here is correct. The tournament book (edited by Kagan) says the game went 21.Bf7+ Kxf7 22.g6+ hxg6 23.fxg6 Kxg6 reaching the same position as the game cited here.
Chess Review,1945, Jan, p14 and the American Chess Bulletin, 1921, p208 and Rubinstein's Games Vol2 (Donaldson and Minev), p38-39 all give the order currently on the site e.g. 21.g6.
However, the tournament book (mentioned above) page 51-2 with notes by Teichmann, Deutsches Wochenschach, 1922, #7-8, p43-44 and Glatman's book (Rubinstein's Chess Academy) as cited by Donaldson, all give the order mentioned above (e.g. 21.Bxf7+).
Donaldson/Minev claim they used the version with 21.g6 "as it is given in all other sources". Clearly, this is not quite accurate. Oddly, it appears Donaldson and Minev had access to the tournament book, which is cited in other Rubinstein games from The Hague, 1921, thus making their statement strange indeed.
The authors go on to say (in their opinion), the order of 21.Bxf7+ etc "is inaccurate", but do not explain why. I looked at 22.Rf1!? with the idea of Qe2, but this seems unconvincing after ...Nd5.
So I remain puzzled.
I should mention Shakhmatny Bulletin, 1960, p303 skips the moves 29.Qh3+ Kg6 30.Qf5+ Kh6 and gives immediately 29.Qxe5 with rest of the moves as given here, thus their version ends two moves sooner.
By the way, Shakhmatny Bulletin gives the order with 21.g6, so the score is now 4-3 in favor of 21.g6!
<Karpova> The annotators (J.W. Collins, Teichmann, and Kmoch) are universal in their condemnation of 14...dxe4. Teichmann suggests 14...Qa7 with the idea of ...Bb6 and ...Nc5 whereas Kmoch suggests 14...Rad8.
Teichmann and Bogolyubov both suggest 16...Qa7 in place of ...Qc7?. The point being that if the game continued as it did Black would have 20...Nc4! to close off the a2-g8 diagonal (pointed out by Bogolyubov).
Teichmann suggests 24...b4 as does Bogolyubov (citing Mieses) with the idea of controlling f1. However, their line of 25.axb4 Bxe3 26.Qxe3 Nd5! with counterplay does not seem like best play for White. Much stronger seems 25.Rg1+ Kh7 26.Qf7! Qf6 (what else?) 27.Qxf6 gxf7 28.Bxc5 and White is a Rook up although he has to solve some technical problems to free his Queen Rook.
Finally, Bogolyubov thought 31...Nc4 offered better resistance, with not so clear a win for White. I was a bit skeptical, but despite a lot of checks, I have not found a clear win for White. An engine can probably find something.
Teichmann was probably right that 14...dxe4 cost Black the game, but Marco missed some defensive chances to make things more difficult for Rubinstein.
Rubinstein won a 40 guilder prize for the most beautiful game of the tournament.