Well my idea was that black has to retreat the bishop anyway because of Nxc6. And if black tries Bc3 (with the queen on b2) you play Qb6 which nearly forces Rxb7. And you still get the extra pawn with Nxc6 at some time.

Now on with the retreat of the bishop.

With Queen on b2:

If he goes to c7 I can play Ra1 which frees me the a-file. Good idea I thought.

If he goes d8 I can play Rb8 Rxb8 Qxb8 Re8 Ra1. Which is pretty dangerous also. At least I think so. And I still have Nxc6 in my hand to make pressure on the bishop good foresight for withe.

---

With queen on b3:

I can neither play Ra1 nor Rb8 and black can keep my attack under control more easily. And what moves are then playable for white? Hmm maybe the nearly null-move Re1 with no real use. So only Qb1 remains. The exchange would be good for white but black will probably play Qf6 and you have to follow with Qf5 Qxf5 Nxf5 to mantain iniative. And this move order brings you away from the important d4 -> xc6 field

What a <DISCOVERED ATTACK> exhibition this is by Kamsky (Black).

Position after 27 ... ♕f6-g6:

click for larger view

Here Kamsky has not one but -two- <DISCOVERED ATTACK CHAINS>:

White b7-♖ + Black c7-♗ + Black e7-♖
White c2-♕ + Black e4-♘ + Black g6-♕

So Kamsky is threatening a <DISCOVERED ATTACK> against the <UNDEFENDED> White b7-rook -and- a <DISCOVERED ATTACK> against the <UNDEFENDED> White c2-queen.

Final position 39 ... ♖a1-d1! 0-1:

click for larger view

And in the final position, Kamsky threatens yet another <DISCOVERED ATTACK> which prevents the White d7-pawn from promoting: 40 d7-d8=♕ ♗e7x♕d8 41 ♖c8x♗d8 ♘d5xe3+ <discovered attack> 42 f2xe3 ♖d1x♖d8, leaving Black a rook(!) ahead.

So this final <DISCOVERED ATTACK CHAIN> would be