In turn, the endgame was lost only by <72.Kg7?>. White 72.Ke7 or 72.Kf5 would have readily held the game. In order to collect the h7 pawn, Black would have had to surrender both of his own remaining pawns.

I personally have a hard time viewing <66.Ke4> as an error. Some of the possible variations after 66.Kxf6 just transpose back into the essentially the same positions as those played in the game. And even after the game 66.Ke4, the encounter would have been easily drawn were it not for the misguided pursuit (72.Kg7?) of the 'attractive nuisance' at h8.

- <20.Qc3> forks c7 and g7, so Black is losing a pawn('s worth) regardless.

- If Black allows Qxc7, White probably eats a few more pawns (or mates).

- If Black allows Qxg7, it "forces" Raf8. Then Black is stuck playing defense against White's <AWACS in the throat> Nd5.

- <20..c6!> offers R for N+P, which is also about a 1-pawn deficit. But it snuffs all White counterplay. Then White's RR is an awful piece mix to face Black's mobile pawn chain.

Capablanca demonstrates how to handle the N side in a N vs R game. He exploits the holes in his opponents' position where his Knight can't be dislodged by his opponent's Rook.

In such positions, the chess principle

<There are holes in the R side which the N side can exploit favorably with the Knight, on which the N can't be challenged by an opposing B or N (because they have been exchanged off with a R). This is usually can happen because the position retains a semi closed character just after the sac is made, in which the interlocked pawn structures allows for such holes.>