Inspired play by Psakhis, I stopped looking when I saw I was "letting black off the hook" by trading his queen.

White wants to open the g file, this might be worthy a piece, so I try with: 23 Nxf5 gxf5 24 Bh5 and now black has two possible moves: 24 ... Qg8 (but it looks wrong) and 24 ... Qe6. I'll start with the second one.

After 24 .. Qe6 I have three possibilities:

1) sac the exchange on g7 and then play Qg1 with check, but know I'm a rook down and don't see good ways to continue my attack.

2) 25 Rg6 gaining a tempo with the idea of Qg1 but black plays 25 .. Qxg6 26 Bxg6+ Kxg6 and I'm down in material and don't have an attack any more..

3) 25 Rg3 with the idea of doubling (maybe tripling) heavy pieces on the g-file, but this plan is very slow.

In all this variations the N on d3 is completely useless, so maybe I should sac that knight, but it doesn't seem to work...

Anyway, after 25...Kh8, I figured 26 Qh5, below.

click for larger view

That's as far as I got.

Here's the analysis (what I like to think the continuation would have been if I had been playing OTB):

A) 25...Qe8 26. Rxg7+ Kxg7 27. Bxe8 Raxe8 28. Qh5, with 29. Rg1 to follow. A clear win.

B) 25...Qf7 26. Rxc6 Qg8 27. Rxc7. White is up two s -- also a win.

C) 25...Qg8 26. Rxc6 Rfc8/Rac8 27. Nxf5. The threat of Ne7 or Nxg7 followed by a push of the center s looks overwhelming.

D) 25...Qxg6 26. Bxg6+ Kxg6 27. Nxf5. If 27...Rxf5, then 28. Qg4+ Kf7 29. Qxf5+ Ke8 29. e6, and white again has a positional and material win. If 27...Kxf5, then 28. Qh5+ Ke6 29. Qg6+ Bf6 30. exf6 Nxf6 31. Re1+ Kd7 32. Bxf6, with a similar outcome.

That leaves (ta-da!) 27...Ne8 28. Qg4+ Kh7 29. Rg1 Rg8 30. Qg6+ Kh8 31. Nxh6.

Black has a lot of options. It takes a great deal of time to analyze them completely and ascertain black's best move. That's where I got stuck. I don't know what black's best play is.

One option is a bishop sacrifice, 26...Be2?!

click for larger view

Material: even. White has K-side pressure based on the g-file controlled by Rg1. The Bh4, Pe5, and Pf4 control dark squares near the Black Kh7; Bf3, Ne3, and Ng3, light squares. In particular, the 2Ns are focused on Pf5, and the capture gxf5 makes the Black Qf7 vulnerable to the move Bh5 from the battery, Qd1 and Bf3. White has maximal development: only Ra1 remains inactive, but can move to the K-side in a single move under the right conditions. In addition, the usual conditions for a K-side strike exist, namely, the Black Ba3, Nc7, and Nd7 are essentially out of defensive play on the Q-side, giving White the local superiority a sacrificial attack requires.

Candidates (23.): Ngxf5

23.Ngxf5,

increasing the scope of Rg1 and threatening 24.Nxg7 25.f5 or 24.Bg4.

23…gxf5 (or lose a P under unfavorable conditions) 24.Bh5

Black has 2 feasible responses.

(1) 24…Qg8 25.Bg6+ Kh8 26.Bxf5 (threatening 27.Bxd7)

Black can sacrifice an exchange with 26…Rad1, 26…Rfd1, or 26…Rxf5, leaving White with R+2Ps vs. B+N and a winning attack fueled by 2 central passers. There are 2 other feasible responses, a dismal retreat 26…Nb8, bottling the Q-side up even more tightly than at present, and 26…Rf7.

(1.1) 26…Nb8 27.Bf6 Rf7 28.Qh5,

threatening 29.Rxg7 Rxg7 30.Qxh6+ and possibly 31.Qh7#.

There is no adequate defense.

(1.2) 26…Rf7 27.Bxd7 Rxd7 28.f5,

threatening 29.f6 30.Qg4.

There is no defense leaving Black less than a P down.

(2) 24…Qe6 25.Bg6+ (threatening 26.Qh5 27.Bg5 28.Bxh6)

25…Qxg6 26.Rxg6 Kxg6 27.Qf3,

threatening 28.Rg1+ 29.Qh5 then 30.Qg6 or 30.Nxf6.

There is no way to defend both Pf5 and Bg7, and once one of them falls, the White pieces pour in.

Time to peek. (No crash and burn <today>.) Time to check the kibitzing. I should fail so handsomely as you, <dzechiel>: initially, I also quailed at throwing more pieces onto the fire with my line (2) 25...Qxg6. The Bh4 made a good anchor for the attack, however, so positionally, it was nearly certain the Q would find a stray piece in Black's Q-side tangle. The calculations <were> reassuring, however.

24.Rxg6 Qxg6 25.Bh5

netting the Black Q.

After acceptance of 23.Ngxf5, Toga gives

[ply 15/53, time 01:06, value +2.21]

23...gxf5 24.Bh5 Qe6 <25.Rg6> Nf6 26.exf6 Bxf6 27.Qf3 Ne8 28.Rag1 Qe4 29.Qxe4 fxe4 30.Nf5 Bg5 31.fxg5 Rxf5 32.Rxc6 Rf4 33.Rxh6+ Kg7 34.Rxb6 Rxh4 35.Bxe8 Rxe8 36.Rxa6

Dynamic evaluation: Knight taking f5 would expose the King if g6xf5. The g3 knight is preferable as this allows the rook to take part in the action. Another route could be Bg5, if not taken by h6 then Nxf5. Yet another move could be Bf3-h5. Let's try that one.

23.Bf3-h5 ...g6xh5 24.Qxh5 ...QxQ 25.Nxh5

PM:=