An Unusual Line Against The Petrov

MY round-six game from the Fareham open.
Spanton (170) - Derek McGovern (154)
Petrov Defence
1.e4 e5 2.Nf3 Nf6 3.Nxe5 d6 4.Nf3 Nxe4 5.c4!?

Position after 5.c4!?

The main line runs 5.d4 d5 6.Bd3 Nc6 7.0-0 Be7 (Black's last two moves are often played in reverse order) 8.c4, when the argument revolves around whether the black knight on e4 represents a gain of tempo, or will it be forced to retreat, or exchange itself, for a loss of time?
The text is recommended in The Chess Advantage In Black And White by Raymond Kaufman (McKay Chess Library 2004). Kaufman says 5.c4!?, like 5.d4, gives White "a tiny edge but has much less theory associated with it." He adds that its "practical results are good," and this has stood up since the book's publication - White scores 60% with 5.d4 in ChessBase's 2020 Mega database, but 64% with 5.c4!? (both numbers are well above White's usual percentage across large databases of about 55%).
David Hooper in A Complete Defence To 1 P-K4 (Pergamon Press 1967) stated: "White hopes for strong central pressure if Black plays ...P-Q4. However, by not further advancing his QP Black gets a very satisfactory game. White will merely have made a useless and in some cases a weakening pawn move."
This opinion is more or less echoed by Lasha Janjgava in The Petroff (Gambit 2001): "The idea behind this move is to obstruct ...d5, so that when White finally does attack the e4 knight (by d4 followed by Bd3) it will be forced to retreat to f6. The problem is that it is rather slow, and additionally Black can retreat the knight to g5 rather than f6. After the exchange of knights, Black will be less worried by White's space advantage and shouldn't have many problems equalising."
However, as I have pointed out, the move scores well, and has been played by strong GMs including Karjakin, Leko and Wei.
5...Be7
Easily Black's most-popular reply.
6.d4 Bg4
The main line runs 6...0-0 7.Bd3 when 7...d5 is very narrowly the most-popular move in ChessBase's 2020 Mega database, but White scores 72%.
7.h3 Bh5 8.Bd3!?
Tal played 8.Qb3 against Smyslov in a 1974 22-move draw, but Stockfish10 and Komodo10 prefer the text.
8...Nf6
If 8...Ng5!? the engines reckon White is better after 9.Bxg5 Bxg5 10.g4 Bg6 11.Qe2+ Be7 12.Nc3.
9.0-0 0-0 10.Nc3 Nc6
10...c6 11.Re1 was better for White in Yi Wei (2728) - Yankai Li (2451), Chinese team championship (Beijing) 2017 (1-0, 59 moves).
11.Be3 d5
Georg Marco - William Napier, Cambridge Springs (Pennsylvania) 1904, saw 11...Qd7 12.Be2 Rae8 13.d5 when White was better but the game was drawn.
12.c5!?
Pushing past with c4-c5, or ...c5-c4 for Black, is always controversial. I have not seen this put forward as a rule-of-thumb, but it may well be that the push-past is usually only good if it comes with tempo, typically by hitting an enemy bishop on d6 (or d3).
12...a6
Black often meets c5 with ...b6, but here 12...b6? runs into 13.Qa4, when the engines reckon best is 13...Bxf3 14.Qxc6 Be4 15.Bxe4 dxe4, after which White can win a pawn by 16.Nxe4.
The engines suggest 12...Re8 13.a3 Bf8 14.Re1 Ne4 with near-equality.
13.a3 Re8 14.b4 Bg6?!
Voluntarily giving up the pin seems strange. The engines give 14...Qd7 eg 15.Re1 Bf8 16.Be2 with a slight edge for White.
15.Bxg6 hxg6 16.Qb3!?
This proves quite difficult to meet. The d5 pawn is awkward to defend, and the white queen has latent pressure against f7.
16...Bf8?!
The engines suggest 16...Qd7 but 17.Bf4 is good, eg 17...Rad8 18.Rfe1 Bf8 19.Rxe8 Rxe8 20.Be5 Ne7 21.Re1 with continuing pressure.
17.Bg5 Ne7 18.g4!?
Denying the f5 square to the black pieces, as well as generally gaining space on the kingside.
18...c6 19.Rae1 Qc7 20.Ne2 Ne4
The engines give 20...b6, but after, for example, 21.Bf4 Qb7 22.Nc3, prefer White.
21.Bf4 Qd8 22.Ng3 Nc8
The engines suggest 22...a5, which at least gives the prospect of counterplay even though White remains better.
23.Ne5 Qf6 24.Bc1 Ng5
Komodo10 gives 24...Qe6 25.f4 f5 26.Kh2 a5 27.bxa5 Rxa5 28.Qxb7 Ne7 29.Qb3 Rb5 30.Qa2 Nc3 31.Qc2, when White is a clear pawn up.
25.Bxg5 Qxg5 26.f4 Qd8 27.Qf3 Re6?!
This cannot be good, but the engines reckon Black is lost anyway.
28.Re2 Qh4 29.f5 gxf5
Best, according to the engines, is 29...Rf6 30.Nd7 Ne7 31.Nxf6+ Qxf6 32.Qf4 g5, but Black has no compensation for the exchange.
30.Nxf5 Qh7 31.g5
Even stronger is 31.Nxf7.
31...Ne7 32.Nd6 f6 33.Qg4 f5 34.Nxf5 Nxf5 35.Rxf5 Rae8 36.Ref2 Be7 37.Rf7 Bf8? 38.Rxf8+ 1-0