Karpova: This game is from the Winter tournament 1908/09 of the Vienna Chess Club. It was a double round robin, with the first cycle ending at the end of January. The second cycle ended on February 20th. The game was played in the first cycle, on January 20th.
8.Bd3 is the more common reply.
With 9...0-0, the game deviates from the database. 9...a5 was played in Tartakower vs Prins, 1949.
Reti considers <11.O-O-O> best, as White doesn't need to fear <11...Bxe3+ 12.Qxe3 Qb4>.
The annotator awards 12...Rb4 a <?>, suggesting <12...Re8> instead, which would win the ♙-e4 in the future.
The annotator awards 14.Bxc5 a <!>, calling <14.axb4 exd3 15.Bxc5> to get 2 ♖s for the ♕ mistaken due to <15...Qd7! 16.Bxf8? Qd4+ 17.Ke1 Qe3+ 18.Kd1 dxc2+ 19.Kxc2 Bf5+>, but White appears to be fine after 16.Bxd3.
The annotator calls 14...Ng4+ a mistake, which should have lost. Better was <14...Qxc5+ 15.Qe3 Qxe3+ 16.Kxe3> with a slightly better endgame for White.
16.Kg1 is called bad, and <16.Qf3
exf3 17.axb4 fxg2+ 18.Kxg2 Bxg4 19.Be2> suggested instead. I'm not sure if that's the case.
After 18.Rc1, the annotator calls White's situation pretty bad (<recht übel>), but White appears to have a clear advantage.
It may have been better to keep the strong ♗ with 21.Bxb4.
<22.Nb1> is given as a possibility instead of 22.Na2. The text move appears to be a mistake.
The move is not annotated, but 22...c5 may not have been too precise. 22...Qc5 looks very interesting, covering the ♙b4, pinning the ♖f2 and taking the c-file under control.
23.h3 is called best and only the mistaken <23.Bc4> discussed, but 23.Nc1 looks interesting.
Again not annotated, but worth a look is 23...Be6. It seems inferior, driving the ♘a2 where it wants to go anyway. Why not the the more restricting 23...Qe3 or 23...Qe1! (when ...Be6 is really threatening)?
On 26.Na2, the annotator claims that <26.Ne2 Bxg4!> was hopeless for White, but I don't see an advantage for Black after 26.Ne2 Bxg4 27.Nf4 c3 28.bxc3 bxc3 29.Bd3.
After 26...Qb3, White seems to panic without a reason. He could have played 27.Nc1 again.
Source: 'Wiener Schachzeitung', June 1909, pp. 172-175