White is pretty much forced to exchange here. Moving the R to d4 is met by ... NxNf3+, whereupon White loses the exchange on d4. White's pieces are stumbling over each other here, so an exchange on e5 might seem to improve his position. However, I have a plan ...
16. Nf3xe5 Bg7xe5
Black's plan: exert pressure on both flanks at once. 1) The c-file has now been cleared for Black's heavy pieces, so White must keep a watch for possible exchanges on c-3 which might lead to breakup of the White Q-side. 2) The Black B has also got h2 in his sights, with Rg8 and Qc7 and advance of Black's h-pawn as possible moves to put the White K in danger. White will have difficulty defending both flanks at once because his pieces are clumped together and blocking each other.
17. Ra1d1
Black must be wary of a possible breakthrough sacrifice on d-5, leaving his K exposed in the center.
17... Bd7c6
Immediately: the pressure on the d-pawn was getting uncomfortable, so I voluntarily blocked the c-file (temporarily) to shore up the center. For later: another long-range piece is bearing down on the White K-side (potentially). Black's position is like a coiled spring: full of active potential. White's position is like a nautilus shell: coiled, but cramped.
18. Be2h5
Not what I expected (which was Bf3 or b4 or f4). White wants to get some more space for his pieces to maneuver in; this B move allows the White N to migrate via e2 to f4.
18... Rh8g8
According to plan ... slowly building up potential threats on both flanks.
19. Nc3e2 Be5xb2
I thought longer and harder about this move than any other in the game. I half-suspected a trap: I was not certain I could extract the B if White replied c3, Bxa3; Ra1, b4 etc. The alternative I considered here was Rg5, pretty much forcing the White B to retreat to f3; this seemed sound, but slow. I opted for the pawn grab after all.
20. Ne2f4
Threatening Nxe6.
20... Ke8e7 21. Qd2b4
Rather than trying to trap the Black B with c3, White opts for reducing the pressure by exchanging queens.
