No doubt Black had played 36..Rb7 to prepare ..b4, but it allowed White after 37 gxf5 gxf5 38 Nxf5 to win a pawn. White was then threatening a fork at d6 (winning the exchange) and Black sidestepped it by 38..Qc6 – which also protected the Rook on e8, meaning that White’s Knight is now threatened.

Before analyzing the possible moves it might make, first it should be noted that the Knight MUST move: if White delays moving it by 39 Qe2? Qf6, and yet again by 40 Qh5? (40 Qg4? Bxf5 wins) ..Rf8, then follows 41 Ng3 Qxf4 (diagram) when Black has recovered his pawn and threatens, among other things, the crushing ..Qf2+. White’s only conceivable reply of 42 Rgf1 (42 Ref1? Qd2+ forces mate) is answered by 42..Bf3! (Black’s other threat) when Black wins the exchange and even forces the Queens off with an elementary win.

click for larger view

(position after 41..Qxf4 in above analysis where White moved the Knight WAY too late)

So back to the position after 38..Qc6.

click for larger view

At this point White has an extra pawn (the f-pawn), but it’s isolated. But more importantly, this is still a middlegame position, one where White’s King is a bit exposed just like Black’s. As long as the Queen and Rooks are still on the board, White must stay at least somewhat concerned about the Black Bishop’s control of light squares in the vicinity of White’s King. Obviously, simplification is desirable for White. If White could swap off all the major pieces, he’d have an easily winning Knight-vs-Bishop ending. On the other hand if he swapped off just the minor pieces, I feel (and this is where I’m *far* less sure) he might still have to work hard to win the resulting Queen-&-Rooks ending. With all this in view, White could’ve considered any of three Knight moves; I’m not sure which one is best:

<1]> 39 Nd4 puts the Knight back on a fine central outpost (with tempo), when White might plan to follow with Rg4 and Reg1. Black can consider 39..Qf6 (when if White then plays 40 Nf3? – relying on the pin to transfer his Knight to e5 – Black plays 40..Rf8, breaking the pin and picking up White’s f-pawn next move); Black can also consider 39..Qd6 (diagram), aiming to achieve the ..b4-break as soon as possible (there’s much to analyze here, way more than I’m able to – for ex., should White himself play 40 b4, to freeze Black’s Q-side pawns on light squares, at the expense of saddling himself with a backward pawn?).

click for larger view

(position after 39..Qd6 in above analysis)

<2]> 39 Ng3 might have either of two intentions: either to simply trade off on e4, or to play Nh5 hoping to create combinational motifs on the dark squares around Black’s King. And if Black tries to meet both of these possibilities by responding 39..Bg6 (diagram) – I don’t know if this is best; Black surely has other options (again, too many to analyze) – White can consider two replies:

click for larger view

(position after 39..Bg6 in above analysis)

<2A]> 40 Re5 (hitting the d-pawn and preparing to double Rooks) and if ..Rxe5 41 fxe5 (diagram); I had thought White’s advantage might be growing ever so slightly, but after 41..b4 now I’m not so sure.

click for larger view

(position after 41 fxe5 in above analysis)

<2B]> 40 f5 Bf7 and then maybe 41 Qd4 (or Qf4); or, to keep Black from replying 41..Rbe7 and thus keeping control of the e-file, perhaps White should swap Rooks first, 41 Rxe8 Qxe8 42 Re1 (diagram) and if ..Re7 (or should Black play ..Qb8 or ..Qf8 to aim for the ..b4-break?), then 43 Rxe7 Qxe7 and now maybe 44 Qd4 (or Qf4).

click for larger view

(position after 42 Re1 in above analysis)

<3]> A third option is 39 Ne3 (with the idea of going via g4 to e5). Now, it appears 39..Rf7? could be met by 40 Nxd5 (I can’t find how Black refutes it or gets comp for it); and it looks like 39..b4 would also drop a pawn in similar fashion after 40 axb4 axb4 41 Nxd5. Thus Black’s best move seems to be 39..Rf8. Then White plays 40 Ng4 (second diagram below). Now Black can’t play 40..Rbf7? (41 Ne5), nor 40..Qd6? (41 Rxe4). A third move to consider, 40..Qc7 (same idea as ..Qd6, to pressure the f-pawn, but without the flaw of dropping a piece), just “looks” too slow – firstly because it keeps the Rook on b7 from joining the action on the Kingside; and secondly because, after 41 Ne5, Black seems not quite ready to initiate Queenside counterplay by 41..b4, due to 42 axb4 axb4 43 c4! (diagram; and if ..Qc5 44 b3).

click for larger view

(position after 43 c4! in above analysis)

So after 40 Ng4 (diagram) best seems either:

click for larger view

(position after 40 Ng4 in above analysis)

<3A]> immediate Queenside counterplay by 40..b4; this might continue either 41 cxb4 axb4 42 Ne5 Qf6, or 41 axb4 axb4 42 Rg3 Rf5 43 Ne5 Qf6; or,

<3B]> 40..Rg7; and this might continue 41 Ne5 Qf6.