A game I played against Mac OSX's built-in chess program (which is, I
think, based on GNU Chess). I played the stonewall attack (which hardly
seems fair against a computer sometimes) and reached the following
position):

r3qnk1/rb2bppp/4p3/pp1pP1PP/2pP3N/P1P1P1Q1/1PB5/4BRRK w - -

This isn't quite a white-to-play and win situation. Or maybe it is. You
tell me. White's pieces are close to optimally placed, but how to break
through?

I played:

26.Rf6! and won quickly because black accepted the sacrifice, as I knew
he would. (Computers have such a hard time not grabbing material).

Obviously 26. ... gf gets white his material back after, say, 27.gf+
Ng6 28.fxe6 and 29.hxg6. And this is not even best, eg 29.hxg6 fg
29.Nxg6!! wins a lot of material.

Black played 26...Bxf6 and lost quickly after 27.gf g6 28.Qg5 Nd7 (the
threat of Qh6-g7 drives off a defender of g6) 29. hg fg 30.Bxg6 (Nxg6
also works) hg 31.Nxg6 Qf7 32.Qh6 Qh7 33.f7+ Kxf7 34.Qxh7+ Ke8 35.Qe7
mate.

I was not afraid of black declining the sacrifice with g6. This just
weakens black's king position and although I didn't see an immediate
refutation, it strikes me as nothing but a delaying tactic. If he ever
plays Bxf6 we'll transpose into something along the above lines, or I
simply prepare my heavy pieces for an eventual attack down the h-file,
with or without retreating my rook.

But...

There's one possible defense I didn't account for, and it's quite
possible that it works. Black is getting overrun in the kingside, but
his heavy pieces are poised for action on the queenside. So what if he
plays (from the FEN position)

26.Rf6!? b4!

It seems impossible that black can survive all the pieces swirling
around his king, but...

does anyone see a clear win for white after this move? How do you break
through if black basically ignores the sacrifice?

-Ron