Capablanca's Endgame Sense

The sort of verbal analysis I used in Capablanca's Positional Sense, is helpful in the endgame when precise calculation of all variations for many moves ahead is beyond human capacity. Later in that same game -- 1921 Capablanca - Lasker, Game 10 -- I highlighted a possible divergence between Capablanca and Kasparov in the evaluation of a specific position: '44.Ke2 {CAP: '!?'; KAS: '?'}'.

While I was looking at that I spotted a couple of other points worth noting. First, there is a discrepancy in the moves of the game given by Capablanca and given by Kasparov. The Cuban's book on the match and his Primer both give the move 41.Ne3, while Kasparov gives 41.Nc3. The next few moves are the same in all sources, the lines reconverge with 43.Nd1, and I couldn't see any significant deviations into sidelines, so I don't think the difference is too important.

More important is the evaluation of the position in the first diagram. Kasparov assigns the next two moves a '!', 43.Nd1! (setting a trap) and 43...Rb1! (avoiding the trap), where the trap is 43...Nb4 44.Rd2 Rb1 45.Nb2 Rxb2 46.Rxb2 Nd3+ 47.Ke2 Nxb2 48.Kd2. The trap is also given by Capablanca. The engine disagrees with this, pointing out that by playing 45...Nc6, instead of 45...Rxb2, Blacks wins a Pawn and keeps the superior position. By a significant margin the engine prefers several moves over 43.Nd1, for example 43.h4, where White avoids the immediate loss of a Pawn and carries on with the grim defense. As no one has given alternatives to 43.Nd1, this merits further analysis.

After 43.Nd1 Rb1, we arrive at the position where Lasker played 44.Ke2. Kasparov wrote,

Finally White blunders a Pawn. Of course, more tenacious was 44.Ke1 Na5 45.Kd2 Rxb3 46.Rxb3 Nxb3+ 47.Kc3 "with drawing chances in view of the insignificant amount of material remaining on the board", (Panov) although objectively this Knight endgame a Pawn down is also lost (Knight endgames being like Pawn endgames).

In his notes to the match, Capablanca commented on 44.Ke2, 'Not a mistake, but played deliberately. White had no way to protect his b-Pawn', and in his Primer gave the same variation up to 46...Nxb3+. He concluded that 'there would have resulted a rather difficult Knight ending, which should nevertheless be won for Black.'

The second diagram shows the position after 46...Nxb3+ in the variation given by both Kasparov ('this Knight endgame a Pawn down is also lost') and Capablanca ('a rather difficult Knight ending, which should nevertheless be won for Black'). The difference between the game's 44.Ke2, and the suggested 44.Ke1, is that the Rooks stayed on the board in the game. In 'Endgame Strategy', Shereshevsky wrote,

[White] should have aimed for the exchange of Rooks, since in the Knight ending Black would have encountered certain technical difficulties, in view of the limited number of Pawns. Correct was 44.Ke1, with drawing chances.

Is the second diagram a draw or a win for Black? I spent some time analyzing the position and failed to find a definite path to victory, so this also merits further analysis. The simplest chess positions can sometimes be the most difficult to understand.