Karpova: This game was annotated by Adolf Zinkl in the 'Neue Freie Presse' of 12 November 1909.
Zinkl considers <4.Be2> to be better, but 4.Qe2 looks fine. However, 5.fxe5 is an interesting alternative to 5.exd5 which only seems to help Black develop.
Zinkl criticizes 8...Qxc4, suggesting instead <8...Bxf3 9.Qxe6+ fxe6 10.gxf3 Nd4> with an advantage for Black, but after 11.Kd1 Nxf3 12.Bh3 Nxe5 13.Bxe6+ Kb8 White may not be that much worse. The text move looks fine, but 8...Qd7 was another alternative.
11.Bd5 appears to be a mistake, but the move was not annotated. Better looks the other way to cover f3 - 11.Be2 - where the ♗ is cannot easily be attacked. But Black doesn't choose the most forceful continuation, i. e. 11...Ne7 12.Be4 f5.
Black seems to lose his advantage after 15...f6. The move was not annotated, but 15...g6 (16.f6 Nef5 looks fine for Black). But the text move, cutting off his own ♘h4, makes 16...g5 necessary.
I wonder why White didn't opt for 18.fxg6, when the Black f-♙ remains weak.
Geor Marco considers 18...Nc6 to be the decisive mistake as it allows Ne4 with tempo. He suggests <18...c6> instead, which looks fine. Another possibility was 18...Bb4 to trade off the ♘c3 (raising the question if the intermediate check 16...Bc5+ was necessary). I'm not sure if the text move was really a mistake. But undoubtly, Black's position becomes worse, until he finally cracks with 23...Bd8.
Source: 'Wiener Schachzeitung', September 1910, p. 313