Bridgeburner: <paladin at large>
The queen exchange was definitely the losing move, at least in the proximate sense (White was probably lost in any case), as it made the <Pa3> indefensible. Lasker played the rest of the game very nicely but it was essentially an easy, somewhat routine endgame win for him.
Rubinstein was very much off his game here...he played the opening very conservatively, never getting around to <e4>, which would have been much more competitive than allowing the lame isolani he ended up with, which in turn allowed <d5> to be occupied as a staging point for Black's pieces.
Rubinstein's middle game was even worse, with lackluster placement of pieces allowing Lasker free rein for his.
I thought Lasker's <26...Nd5>, apart from being an excellent move in its own right, was also a psychological move aimed at provoking the weakening <27.a3>, the fate of which decided the game.
<Ulhumbrus: This suggests that exchanging the Rooks on the c file is a mistake on White's part. Instead of 21 Rxc8, 21 Rd1 avoids the first exchange of Rooks, while 22 Rd1 or 22 Re1 instead of 22 Rc1 avoids the second exchange of Rooks.>
I'm not sure.If <21.Rcd1>, then <21...Nf6> and the White isolani is fully blockaded. Moreover, <22.Re1> wouldn't be a good move as it becomes a target for Black's DSB if it reaches <b4> (there's the echo of that fateful <a3> again), eg: <21.Rcd1 Nf6 22.Rfe1 Nxe4 23.Bxe4 Bb4 24.Bxb7 Qxb7 25.Qe4 Qxe4 26.Rxe4 Rc2>:
Such a position wouldn't be defensible against Lasker.