Karpova: This game was played on board 2 of the team match between <Arbeiter-Schachklub I> (Fischer) and <Deutscher Schachverein I> (Becker).
Albert Becker annotated it.
7.Bxf6: <7.Bh4 dxc4 7.e3 Nb6 8.Ne5 c5 9.Nxc4 cxd4 10.exd4 Nxc4 11.Bxc4 Be7>.
Becker suggests <7...c5!> with an equal game.
Becker played 8...a6 to prepare a ♙-storm, if White plays 0-0-0.
10.Ne5: White avoids moving the ♗f1 to not lose a tempo after <10...dxc4>. Becker suggests <10.c5> (<10...Nd7 11.Nd2 c6 12.f4> preventing <...e5>).
If <11.dxc5>, then <11...Qc7>.
<11...cxd4> costs a ♙ after <12.dxe6>.
13...b5 threatens <14...cxd4> followed by <...b4>.
Becker considers <14.a3 cxd4 15.exd4 Ne4> worse for White, due to <16.Bd3 Nxc3>, but 16.Rc1 may be ok for White.
Becker considered <14...c4 15.f4 b4 16.Nb1 Ne4 17.Nd2 f5 (17...Bf5) 18.Bf3>, but thought that White got good counterplay via <.g4> and Black didn't want to close the c-file. To me, 16...Ne4 (17.Bxc4) and 17...f5 (18.Ndxc4) don't look good.
<16...Bc8 17.Qb1 g6 18.g4!>, but 17...g6 doesn't look convincing (18.Bxb5), while Black's ♙-storm after 18.g4 (18...b4 19.Nd1 c4) looks scary.
<17.Bd3> (threatening <.Ng4!>) <17...Rfe8!>. However, the text move (17.g4, threatening <.g5>) looks like an inaccuracy, allowing Black's next move.
Becker devotes a long annotation to the <sneaky> 18.Ne4, which appears to me to be the decisive mistake. Even if Black falls for it and follows the line <18...Qxc1 19.Nxf6+ gxf6 20.Bd3 Qxe3+ 21.Kh1>, Black can enter the interesting endgame after 21...Qe4+ 22.Bxe4 dxe4 23.Nd7 e3+ 24.Kg1 Rfd8.
20.Nxf7 is forced, because <...Bc8> threatens.
Source: 'Österreichische Schachrundschau', February 1922, issue 2, pp. 13-14