I want to show you the line I've been working at. I seriously think it may be our best chance to win and certainly it is the line with highest engine evals.
I will not say that it is winning, but in all variations the eval should be better than -0.6 (Houdini 3). Of course it has to be said that even with this eval some lines may be drawn, for example I noticed that some lines convert to rather drawish 2R vs 2R endgame with Black a pawn up.
I would like to point out that Black can obtain -0.6 advantage regardless of what White does. Also, in my line Rh7 is avoided although in one subvariation it is actually played later, but with a good compensation.
I would be nice if you looked through my lines to see if I maybe missed a better move for White, but remember that I probably checked all options up to 30 ply depth.
Have fun browsing through the lines :)
If he plays 15.Qf3, this is the line I assumed.
<15. Qf3 Nb4 16. 0-0-0 Nxa2+ 17. Kb1 Nb4 18. Qf5 d3 19. c4 Qxf5 20. gxf5 f6 21. Rdg1 Kf7 22. Rg6 h5 23. Rhg1 Rag8>
[-0.41] d=29 24.Bc3 Nc6 25.f3 Bf8 26.R6g2 Bd6 27.Ka2 a5
[-0.49] d=29 24.R6g3 a5 25.h4 Nc6 26.Bc3 Rh7 27.f3 Ra8
[-0.50] d=29 24.Rf1 Rb8 25.Nc3 b6 26.f4 exf4 27.Rxf4 Bd6
[-0.51] d=29 24.f3 h4 25.R6g4 Rh5 26.f4 Rxf5 27.Rxh4 exf4
At first I stopped my analysis here because the first line did not look promising. Fortunately, <YouRang> followed this line deeper and posted in the Tree some analysis which made me interested in this line again.
Let's try to identify potential breakthroughs for Black.
<1> playing h5-h4-Rh5-Rxf5
<2> playing b5
<3> playing a5-a4
<4> playing g6/g5
Along the lines that I will mention, the first two are most likely.
What are White's threats? Mostly maintaining the clamp on g7, in some lines also f3-Nf2-Nxd3.
Let's proceed to the concrete lines now. First I will analyze the line in which White allows Black to play h4.
<A> After <24.f3 h4> we have this:
[-0.47] d=32 25.R6g4 Rh5 26.f4 Rxf5 27.Rxh4 exf4 28.exf4 Nc6 29.Ng3
[-0.56] d=32 25.Bc3 Nc6 26.R6g4 Rh5 27.f4 Rxf5 28.Rxh4 exf4 29.exf4 b5
<A1> <25.R6g4> and now comes a sequence forced for White:
<25...Rh5 26.f4 Rxf5 27.Rxh4 exf4 28.exf4 >
Here the eval rises significantly, I have 39-ply Houdini here:
[-0.61] d=39 28...Nc2 29.Ng3 Ra5 30.Ne4 b5 31.c5 b4 32.Rg3 Na3
[-0.54] d=38 28...Nc6 29.Ng3 Ra5 30.Ne4 b6 31.Rg3 Re8 32.Nf2
So we have -0.6 eval. Let's go back to <A> and try the second line.
<A2> <25.Bc3 Nc6 26.R6g4 Rh5 27.f4 Rxf5 28.Rxh4 exf4 29.exf4>
It is basically the same line as line <A1>, but with Bc3 Nc6 inserted. It is also forced for White. I discovered it should be even better for Black because we have a b5 break (breakthrough <2>) later in this variation (see Houdini's line). Note that in this line Black may be sacrificing an exchange.
We may then conclude that White will probably play h4 to stop Black from advancing h4. This stops forever the breakthrough <1>.
Followed in part 2...
<B> <24.Bc3 Nc6 25.h4 a5>
Another crucial position. The eval is now in -0.5 range, but still not -0.6.
[-0.52] d=36 26.R6g2 Ba3 27.Ka2 Bb4 28.Rf1 Ra8
[-0.53] d=36 26.R6g3 Rh7 27.R3g2 Ra8 28.f3 Bf8
[-0.63] d=36 26.Kb2 Bb4 27.R6g3 Ne7 28.Rf3 Ra8
I have tried both first lines here. My conclusion is that R6g2 is better for White than R6g3 because the latter runs into breakthrough <2> with Rh7+Rb8+b5.
So: <26.R6g2 Ba3> (threatening Ne7)
[-0.57] d=32 27.Ka2 Bb4 28.Rf1 Ra8 29.Kb2 Be7
[-0.62] d=32 27.Rf1 Ra8 28.Ka2 Be7 29.f3 b6
[-0.63] d=32 27.Rg3 Ne7 28.Rf3 Ra8 29.Ka2 Bb4
[-0.70] d=31 27.f3 Ne7 28.Rf2 b5 29.Ka2 Bb4
Still the first line is below -0.6...
<27.Ka2 Bb4 28.Rf1>
Another crucial position. Black has a lot of options:
[-0.54] d=32 28...Ra8 29.Kb2 Be7 30.Rfg1 Bf8
[-0.51] d=32 28...Re8 29.Bxb4 axb4 30.Nc5 Rhg8
[-0.49] d=32 28...b6 29.Kb2 Ra8 30.f4 exf4
[-0.49] d=32 28...Rd8 29.Bxb4 axb4 30.Kb2 Ne7
Black has 28...Ra8 -0.59/34 and 28...b6 -0.51/38. But I prefer the second line:
<28...Re8> (anticipating f4)
Now engines prefer Bxb4 but it is no good for White.
<B1> <29.Bxb4 axb4 30.Nc5 Rhg8> (forced after Bxb4)
[-0.52] d=36 31.Kb2 e4 32.f4 Nd8 33.Ra1 Re7
[-0.75] d=35 31.f4 Ra8 32.Kb2 Ra5 33.Nxb7 Ra3
Engines still do no see it, let's move forward the first line.
<31.Kb2 e4> Suddenly the eval jumps to -0.7 because 32.f4 drops to -0.88/34.
[-0.74] d=34 32.f3 Re5 33.Nxe4 Rxf5 34.f4 Ra5 35.Ra1 Rxa1
[-0.87] d=34 32.Ne6 Ne7 33.Nxc7 Rc8 34.Nb5 Nxf5 35.Rh2 Rc6
So we may conclude that 29.Bxb4 is rather inferior. Let's roll back to move 29.
<B2> Other White options are:
[-0.61] d=33 29.Kb2 Ne7 30.Ng3 Reg8 31.Ra1 Ra8 32.Bxb4 axb4 [-0.65] d=33 29.f4 exf4 30.Rxf4 b6 31.Kb2 Be7 32.Nf2 Bd6
<29.Kb2 Ne7 30.Ng3 b6>
[-0.67] d=33 31.Bxb4 axb4 32.f4 exf4 33.Rxf4 Nc6
Again a very high score.
after <15. Qf3 Nb4 16. 0-0-0 Nxa2+ 17. Kb1 Nb4 18. Qf5 d3 19. c4 Qxf5 20. gxf5 f6 21. Rdg1 Kf7 22. Rg6 h5 23. Rhg1 Rag8>
the eval is -0.6 in Black's favor in all lines. We avoid putting rook on h7 and use threats <1>, <2>, <3> and <4> to gradually improve our position. Black should be able to break through with advantage, although it also must be said some of the resulting rook endgames may be drawn in the end.