KEG: Trading down to a slightly inferior endgame against Lasker was always a recipe for disaster. And so it was for Cohn here.
Through move 9, Cohn had the better game. But fearing his famous opponent, he traded off two sets of minor pieces on moves 10-12. Even then, he was fine, but his 13. Qg4 check and 14. QxQ check reduced to an endgame in which Lasker had a small edge. It was also an endgame with lots of play, Lasker having 2 Rooks, a Bishop, and 7 pawns against Cohn's 2 Rooks, Knight, and 7 pawns. According to Sem, Vidmar thinks the exchanges allowed White to "breathe freely." Similarly, the Tournament Book opines that had Cohn played "quietly for a draw [after the exchange of Queens], he need not have been apprenhensive of losing the game." A review of what happened in the game itself, however, shows otherwise. Once in the endgame with his slight edge, Lasker attacked on both sides of the board and posed repeated tactical problems for Cohn. Was the game theoretically drawable for Cohn? The answer is almost certainly "Yes," and I would not have been surprised if endgame geniuses like Capablance or Rubinstein (or more recently Smyslov, Karpov, or Korchnoi)were able to hold the position. But like Bobby Fischer, Lasker was able to torture opponents trying to hold slightly inferior positions for hours and make them sweat if they wanted to draw. More often than not, these opponents erred at some point, and got killed. This is exactly what happened to Cohn here, who should never have abandoned his better middle game to try to survive in this endgame thicket against a killer endgame opponent like Lasker.
A review of this endgame was, at least for me, both thrilling and educational. Lasker attacked on both wings with 16...a5, 17...g5, and then 18...a4. Seeing Lasker gobble up space and squares all over the board, Cohn tried to get counterplay with 24. g4, which he no doubt thought would only be a temporary pawn sacrifice. But after this ill-judged advance, Lasker never gave Cohn a chance. And his winning play was both efficient and elegant.
Try an experiment. Feed Lasker's moves into an engine like Fritz. I did, and found that, while Fritz occasionally differed from Lasker and sometimes found faster ways to dispose of Cohn, Lasker's play always kept the win secure and often was far more aesthetically pleasing than Fritz' efforts.
Note, for example, Lasker's 33...d5. At first blush, it might appear that Cohn could have won a pawn with 34.exd5. But Lasker would have crushed this move with 34...Rf8 check. (Indeed, Fritz shows that this would have been a forced mate for Lasker).
As a second example, after Lasker's 35...Rf8, it appears that Cohn could have won a pawn with 36. Nxb7. But had he done that, Lasker was ready with the murderous 36...Rf4 check.
Most pretty of all, after Cohn's 39.Nb6, Lasker sacrificed the exchange with 39...Kf5!! (instead of the prosaic 39...Rd3). It took Fritz a fair bit of time to realize that Lasker's line is best.
Once Cohn had accepted the exchange sacrifice, Lasker's pawns steamrolled White's two Rooks. After Lasker's 44...e3, it appears at first that Cohn could grab a pawn with 45. Rxe3, but Lasker had a lovely Bishop skewer combination planned had Cohn gone pawn hunting. After 45. Rxe3, Lasker would have played 45...RxR check, and if then 46. RxR 46...Bc5 cleans the house.
As Cohn learned to his chagrin, Lasker's tactically wizardry was just as dangerous in the endgame as in the middlegame.
Although as a theoretical matter 24. g4 was the move that caused Cohn to lose the game, in my book he lost the game back on moves 13 and 14 when he let Lasker trade off the Queens.