Some of the variations in the analysis of this game are very long. They're definitely worth playing through, though, especially the extraordinary note to White's 26th move. Most of the analysis is the computer's work; I've just selected the most interesting and important variations.
1.e4
e5
2.Nc3
Nf6
3.Bc4
Bc5
[3...Nc6
is the main alternative here.]
4.d3
Nc6
5.Nge2
I can only find two previous games with this move: Mieses-Von Scheve, Leipzig 1888 (1-0) and Marsalek-Blatny, Chocen 1950 (0-1). If White wishes to stay with the Vienna (rather than transposing to the Giuoco Pianissimo with 5.Nf3), 5.f4 is the natural move.
5...d6
6.Na4
[Marsalek chose the horrible 6.Bg5?
against Blatny, who finished him off in some style: 6...Bxf2+
7.Kf1
Bb6
8.Ng3
h6
9.Bd2
Ng4
10.Qf3
0-0
11.Nf5
h5
12.h3
Bxf5
13.exf5
Nd4
14.Qd1
Qh4
15.Ne4
Nxf5
16.hxg4
Qxh1+
17.Ke2
Qxg2+
18.Ke1
Qg1+
19.Ke2
Nd4#
]
6...Bb6
7.Nxb6
axb6
8.Bg5
[Having removed the bishop White's natural choice is 8.f4
, although Black should have few problems.]
8...h6
9.Bh4
Be6
10.Nc3
g5
Both sides having been waiting to see where the other intends to castle. This move signals that Black will go the queenside, but more or less forces White to do the same, since the advanced black pawns would mean a ready-made attack if White should castle kingside.
11.Bg3
Qd7
12.Bxe6
Qxe6
13.Qd2
0-0-0
14.0-0-0
d5
Despite my slightly shaggy pawn structure, I've come out of the opening pretty well. This pawn break gives me the initiative in the centre.
15.exd5
Nxd5
16.a3
This stops any tricks aimed at a2, but weakens b3.
16...Nd4
I was very happy at this point. I have a nice position in the centre and more space, and the kingside pawns are ready to roll.
17.Rde1
My first instinct was that this is a mistake: the other rook gets locked out of the game. But there is one significant benefit, as we shall see.
17...f6
18.f3
Qc6
The original idea here was to threaten to break up White's queenside pawns. The secondary point is the Nb3+ tactic if the knight moves away. Of course, 19.Nxd5 threatens a fork of its own, which means that 19...Nb3+ loses a piece. But the longer I looked at this line, the more I began to suspect that I could get a strong attack by sacrificing the piece. Without analysing very much, I decided to go for it.
19.Nxd5
Nb3+
[Now 19...Rxd5
20.Kb1
Rc5
gives me a very comfortable position, so the knight sacrifice is hardly necessary. So why play Nb3+? It turns out that the defence is very difficult, whereas my moves are easy to find. White's problems stem from the fact that the bishop and the king's rook are effectively out of play, while his king remains vulnerable in the centre.]
20.Kd1
Rxd5
[Only possible because of 17.Rde1, so perhaps that wasn't so bad after all. Originally I had intended to play 20...Qxd5
21.cxb3
Qxb3+
but after 22.Ke2
my rooks are slow getting into the action and the attack can easily fizzle out. So I had to think again, wondering whether or not I'd overdone the optimism. Fortunately, there is a decent alternative.]
21.cxb3
Rhd8
22.Ke2
Qb5
23.Rd1
f5
[Up to here both players' moves have been more or less forced. Now computer analysis suggests that 23...Qxb3
24.Qb4
Rxd3
25.Qxb3
Rxb3
26.Rxd8+
Kxd8
27.Rd1+
Ke7
28.Rd2
is best, when Black doesn't have enough compensation. Having spent a long time analysing the position after 23.Rd1 I agree with the computer, but entering this line would be almost equivalent to resigning. Therefore it makes sense to avoid chasing material, and to try to maintain pressure against the white king.]
24.Qc3?
[This looks natural (it prevents ...e4), but it gives up most of White's advantage. White's other main choices, 24.Qc2 and 24.Qb4!, should win. Providing a detailed analysis here is next to impossible; some sample variations follow. 24.Qc2
This is good and probably sufficient to win. However, Black has plenty of chances. 24...Rc5
25.Qb1
e4
26.fxe4
fxe4
27.Rhe1
h5
28.b4
(28.a4
is also good for White) 28...exd3+
(28...Rcd5
29.Qc1
(The attempt to simplify to an endgame with two extra pawns by 29.Kf2?
leads only to draw at best. The following almost forced variation shows why: 29...h4
30.Bxc7
Kxc7
31.dxe4
Rxd1
32.Rxd1
Rf8+
33.Kg1
Qe2
34.Re1
Qf2+
35.Kh1
h3
36.Rg1
Qe2
37.Qc1+
Kb8
38.Qxg5
hxg2+
39.Qxg2
Rf2
40.Qg3+
Ka7
41.h3
Qxe4+
42.Rg2
Qb1+
43.Kh2
Rf1
44.Rg1
Qc2+
45.Rg2
(45.Kh1?
Qe4+
46.Qg2
Rf3
47.Kh2
Qf4+
48.Kh1
Qf5
and Black wins) 45...Qb1
with a draw by repetition.) 29...c6
30.Kf1
and White should be able to hold on to the extra material.) 29.Rxd3
Rcd5
30.Rc1!
Qxd3+
(30...c6
31.Rcc3
) 31.Qxd3
Rxd3
32.Rxc7+
Kb8
33.Rc3+
Rxg3
34.Rxg3
with a probable win for White in the endgame; 24.Qb4!
This is the move that refutes Black's sacrifice. I think it's hard to see because it appears to abandon the d-pawn, but White gets counterplay along the a-file and threatens to exchange queens. 24...Qc6
25.Rhe1
Qg6
(25...Rd4
26.Qc3
) 26.Kf1
f4
27.Bf2
Rxd3
28.Rc1
and White's king is safe.]
24...Rc5
25.Qd2
[25.a4
Qa6
26.Qd2
may be good enough to win, but it's very tricky. Now I thought about repeating the position with 25...Rd5, but I felt that I could try for more.]
25...e4
White's excursion with the queen has cost time, and now my attack is very dangerous.
26.Rc1??
[26.fxe4!
leads to some fantastic variations: 26...fxe4
27.Qb4
exd3+
28.Kf2
Rf5+
(28...Rc2+
29.Kg1
Qxb4
30.axb4
Rxb2
31.h4
Rxb3
(31...g4
32.h5
Rxb3
33.Bf4
Rxb4
34.Bxh6
is wild. Either side might win.) 32.hxg5
hxg5
33.Rh7
Rd7
is a probable draw) 29.Kg1
Qxb4
30.axb4
d2
31.h4
g4
32.Kh2
(32.Bf2
leads to something similar, but White might just be able to sneak a win. In some variations White can untangle his kingside pieces only to find that all of his b-pawns have vanished and the black queenside is ready to roll forward.) 32...b5!
Otherwise White can play b5 and extract a rook from its misery via a4. 33.Ra1
Kb8
34.Rhd1
Rd3
and astoundingly Black has a positional draw: neither side can improve the position of their pieces or force a further weakness. After the move played, White is lost.]
26...Rxd3
27.Qb4?
[Allowing me to produce a neat mating finish. Instead, 27.fxe4
forces Black to find the following elegant win: 27...Rxd2+
28.Kxd2
Qd7+
29.Ke2
Qd4
30.Rxc5
Qxe4+
31.Kf2
Qd4+
32.Kf1
Qd3+
33.Kf2
bxc5
34.Re1
(White cannot save the bishop: 34.Be5
Qd2+
35.Kf1
(35.Kf3
g4+
36.Kg3
Qe3+
37.Kh4
Qxe5
with mate to follow) 35...Qd1+
36.Kf2
Qxh1
) 34...f4
]
27...Rd2+!
[27...Rd2+
28.Kxd2
Qd3+
29.Ke1
Rxc1+
30.Kf2
e3#
] 0-1