New Season Underway

MODERN chess is a year-round activity, but league chess generally keeps to a traditional autumn-winter-spring timetable.
So it was that my second season for Battersea began today - yesterday, technically, as it has gone midnight as I type this - when I turned out for Battersea2 in division one of the Central London League against … Battersea1.
Aldo Camilleri (196) - Spanton (167)
Göring Gambit
1.e4 e5 2.Nf3 Nc6 3.d4 exd4 4.c3 d5
The most popular move in ChessBase's 2018 Mega database is accepting the gambit by 4...dxc3, but the various ways of declining - chiefly 4...d5, 4...d3, 4...Nf6 and 4...d6 - combine to outnumber the one method of acceptance.
5.exd5 Qxd5 6.cxd4 Bg4 7.Be2 Bb4+ 8.Nc3 Bxf3
I had previously preferred 8...Nf6, but singularly without success.
9.Bxf3 Qc4
Capablanca's patent, first played against Marshall at Lake Hopatcong (New Jersey) 1926.
10.Bxc6
This quickly became the main line, but Marshall chose 10.Be3. That game continued 10...Bxc3+ 11.bxc3 Qxc3+ 12.Kf1 Qc4+ 13.Kg1 Nge7 14.Rc1 Qxa2 15.Ra1 Qc4 16.Rc1 Draw agreed.
10... bxc6
Counter-intuitive, at least to me, but the 'natural' 10...Qxc6 leaves White with something of an initiative after 11.0-0. However, Stockfish9 reckons the position is dead equal, and 10...Qxc6 has been played by some strong players, including Svidler.
11.Qe2
Getting the queens off is not forced, but it seems sensible since Black is preventing White from quickly castling, and Black has more pawn weaknesses, which could prove an endgame headache.
11...Qxe2+ 12.Kxe2 0-0-0 13.Be3 Ne7

Black's position looks iffy, but is it really?

George Botterill in Open Gambits (Batsford, 1986) wrote: "There is no reason to think that Black stands worse. His doubled c pawns are quite useful, giving a firm grip on important squares."
14.Rhd1 Rhe8 15.a3 Ba5 16.Rac1?!
My main analysis engines, Stockfish9 and Komodo9, prefer 16.Rd3 and 16.Kd3.
16...Nf5
Exposing the d pawn as the weakest pawn on the board.
17.Na4

Can you find a good move for Black here?

17...Nxd4+
Winning a pawn, but Black's extra pawn is doubled and isolated, which is why the engines think the position is equal. They give 17...Bb6!, when White has the unpleasant choice of losing the d pawn in less favourable circumstances, or unisolating Black's queenside pawns by playing Nxb6.
18.Kb1 Nb3 19.Rxd8+ Rxd8 20.Rc2
I expected 20.Rxc6, which I intended meeting with 20...Kb7, when White has at best a very minimal edge thanks to his better pawn-structure.
20...Kb7 21.Ke2 Bb6
Better than the obvious 21...Nd4+, which runs into 22.Bxd4 Rxd4 23.Nc5+, eg 23...Kc8 (not 23...Kb6?? 24.b4) 24.Nb3 Rd5 25.Nxa5 Rxa5 26.Rxc6 with a drawn rook-and-pawn ending.
22.Nc5+ Nxc5 23.Bxc5 Rd5 24.Be3 Re5 25.Rc4 Bxe3 26.fxe3

Black to play and win?

So we have a rook-and-pawn ending which must be better for Black, but it is hard not to think that White has very good drawing chances.
26...Rh5 27.h3 Rb5 28.Rc2 Kb6 29.g4 h5 30.Kf3 hxg4+ 31.hxg4 Rb3 32.g5 c5 33.Kf4 c4 34.e4 Kc5 35.g6 f6
Maybe 35...fxg6 was better. It was certainly the less-likely-to-lose option.
36.Rd2 Rd3 37.Re2 Kd4 38.Kf5 Rg3 39.Rd2+ Ke3 40.Rc2 Rg5+ 41.Ke6 Rxg6 42.Rxc4 Rg2?
Better was 42...Rg4, when 43.Rxc7 Rxe4+ is very promising for Black as the f pawn is quick. Stockfish9 gives 43.Rc3+! Kxe4 44.Rc4+ Kf3 45.Rc3+ Kf2, and now 46.Rxc7, but rates Black as winning (Komodo9 reckons Black is 'merely' much better).
43.b4 Ra2 44.Rxc7 Rxa3 45.Rxg7 Kxe4 46.Kxf6 Kd5 47.Rc7 a6 48.Rc8
White offered a draw.
48...Ra4 49.Rc5+ Kd6 50.Ra5 ½-½
Black can win White's final pawn, but cannot prevent the White king from reaching a1.