The best move here is 9....Be6! to challenge the Bishop at c4 while blocking the e-file. If Black instead plans on developing his Bishop to f5 he does best to play 9....Bf5 10.Rhe1 O-O, transposing to the game continuation. The problem with 9....O-O?! is that it allows White an immediate attack by 10.Bd3! h6 (10....g6 11.Rhe1 Be6 12.Ne4 Nxe4 13.Rxe4 Bxg5+ 14.Nxg5 h5 15.g4 [15.Rxe6!?] 15....Kg7? 16.Rxe6 1-0 Kokholm--Christensen, Denmark 1993) 11.Bxh6 (11.Rhe1 is also good) 11....gxh6 (11....Ng4 12.Bg5!) 12.Qxh6 Ne5 (12....Nb4?! 13.Ng5 Nxd3+ 14.Rxd3 Bf5 15.Rg3 Bg6 16.Ne6! 1-0 Nejstadt--Amateur, Exhibition 1950) 13.Nxe5 dxe5 14.Qg5+ Kh8 15.Bf5 Qe8 16.Rd3 +- Nejstadt.
As indicated above, White could gain an immediate advantage by 10.Bd3!
The Bishop moves to defend the Black kingside from 11.Bd3. The Bishop is rather loosely placed here, though, and offers White a target for his attack.
A fascinating sacrifice with more potential than Tartakower may have realized during the game. The more standard move is Keidanski's 11.Qf4! Bg6 12.g4, gaining time by attacking the loose Bishop at f5. Less successful is 11.Nd4?! Nxd4 12.Rxd4 h6! 13.Bxh6 gxh6 14.Qxh6 Nh7! and White's attack fails.
Black must capture with the Knight since 11...Qxe7? loses to 12.Re1! (not 12.Nd5? Qe4! =+) 12....Qd7 13.Bxf6 gxf6 14. Nd5 or 12....h6 13.Bxh6! and White is winning.
This move appears at first glance to be a loss of time, and closer analysis reinforces that assumption. Perhaps Tartakower thought he needed to first deprive the Black Bishop of its natural retreat to e6 before attacking it. More direct, though, was 13.g4! with excellent compensation for the exchange.
1) 13....Ng6 14.Qh6! Bxg4 (14....Be6 15.Nd4! Bxc4? 16.Nf5 +-) 15.Rg1 Be6 (15....b5 16.h3 bxc4 17.hxg4) 16.Bxe6 fxe6 17.h4 +=.
2) 13....Be6 (Did Tartakower fear this retreat?) 14.Ne4!! Ng6 (14....Bxc4 15.Nxf6+ Kg7 16.Nh5+ Kg8 17.Qg5+ Ng6 18.Qh6 mates) 15.Nxf6+ Kg7 16.Qxh7+ Kxf6 17.g5+ Ke7 18.Bxe6 Kxe6 19.Nd4+ Kd7 (19....Kd5 20.Qh3! Qxg5+ 21.Kb1 Rh8 22.Qb3+ +-; 19....Ke7 20.Re1+ Ne5 21.f4 c5 22.Nf3 Rh8 23.Qf5; 19....Ke5 20.f4+ Kxf4 21.Qh3 Kxg5 22.Rf1 +-) 20.Qh3+ Ke8 21.Re1+ Ne5 22.f4 +=
3) 13....Bg6 14.Qxf6 Nc8 15.Qf4 Nb6 16.Be2 Nd7 17.h4 Qf6 18.Qg3!? (18.Qxf6 +=) 18....Nc5 19.h5 Be4 20.g5 Qe7 21.Nd4 +=.
Some of these lines would have been difficult to find over the board. But Tartakower must have recognized that he would only gain a draw out of the line he chose, so he should have searched for an alternative.
The Knight is off limits since 16....Bxf3? allows a mating attack by 17.Rxg6+ hxg6 18.Qxg6+ Kh8 19.Qh6+ Kg8 20.Bd3! f5 21.Bxf5.
Black gets nothing from 17....Qd7?! 18. Nh4! with a strong attack.
It is not clear why Black risks his King for more than the forced draw that follows 19....hxg6! 20.Qxg6+ Kh8 21.Qh6+ with perpetual check. After all, White can easily force a draw after 19....Kf7 as well, so it is not clear what Black gains from his risky wanderings. Perhaps Shoosmith was concerned about his standing in the tournament, but he should have played the position on the board rather than his position in the standings.
White seems to do even better after 21....Kd6 22.Qf4+ Ke6 23.Qxf3 Rg8 24.Rxb7 Rg1+ 25.Kd2.
There is no time for 22.f3? when Black gains the advantage by 22....Rf7 or 22....Qb6. But White's attack is actually quite sufficient after this exchange.
Clearly White can force the draw at any time. But Black's play has encouraged him to pursue a win!
There does not seem to be a saving move at this point. White now controls all of the critical squares around the Black King.
The King is now forced out from behind its pawn cover and mate is inevitable.
Black resigns since nothing can be done to prevent mate. Black can only delay the inevitable by surrendering all of his material: 38....Qd6 39.Rxd6 Rad8 40.Rxc6 Rd1+ 41.Kxd1 Rd8+ 42.Kc2 Rd2+ 43.Kxd2 h6 44.Ra6#