Trying Too Hard As White

I HAVE  a database of my games - 3,448 in all.
There are some rapid and some correspondence games, but the vast majority are at normal time limits.
As Black I score 47 percent - slightly above the average for black of 45-46 percent in large databases. But as White I score 51 percent - significantly below the average for white of 54-55 percent.
Why is this? I have come to the conclusion the main problem is trying for too much as White.
Take today's game as an example.
Spanton (1923) - Mike Waddington (2096), Guernsey Round Two
Leningrad Dutch
1.c4 f5 2.Nc3 Nf6 3.g3 g6 4.Bg2 Bg7 5.d3 0-0 6.e4 e5
Much more popular are 6...fxe4 and 6...d6, but the text has been tried by some strong players.
7.exf5!?
This scores just 28 percent in ChessBase's 2018 Mega database, but it is the choice of my main analysis engines, Stockfish9 and Komodo9.
The idea is to capture on f5 when Black is forced to take back with a pawn rather than a piece.
7...gxf5 8.Nge2 Nc6 9.d4?!
This extravagant move has been played by a 2460, but it is asking too much of the position.
Natural was 9.0-0, and if the normal 9...d6, then 10.f4!? fixes Black's f pawn, while the engines also like 10.d4!? - better now there is no check on the e file.
9...e4?
Black has no difficulties after the simple 9...Nxd4 10.Nxd4 exd4 11.Qxd4 Re8+.
The problem with the text is it gives White a juicy outpost at f4.
10.0-0?
"Castle because you must or because you want to, not because you can."
It was important to play 10.Bg5, setting up a pin that prevents Black grabbing his fair share of the centre with 10...d5.
10...d6?
Failing to challenge White's central control.
After 10...d5 11.cxd5 Nxd5 12.Nxd5 Qxd5 13.Be3, White may well be better, eg 13...Rd8 is met by 14.Nf4 and 15.d5, but Black is in no danger of being overrun.
11.Bg5 Qe8

Try to guess White's boneheaded 12th move

****
****
****
****
****
****
12.Bxf6?
Continuing development with 12.Qd2, or immediately challenging Black's e pawn with 12.f3, are superior ideas.
Stockfish9 very much likes 12.Rc1?!, which I confess I do not understand.
12...Rxf6 13.Nd5 Rf7 14.b4?
Ridiculously trying for too much.
Again, attacking Black's e pawn with 14.f3 was good.
14...Be6 15.Ne3?
Retreating in a position like this is unlikely to be the answer.
Stockfish9 gives 15.Qd2 Bxd5 16.cxd5 Ne7 17.Nc3, with a small edge for Black.
Komodo9 prefers 15.f3 Bxd5 16.cxd5 Nxb4 17.fxe4 fxe4 18.Qb3, with some compensation for the pawn.
15...Nxb4 16.d5?
I rejected 16.Rb1 because of 16...c5, but the text is weaker.
16...Bc8 17.Rb1 Nd3 18.f3 Bh6 19.f4 Bg7 20.Nc1 Bd4 21.Qxd2 Nxc1
I now picked up my queen, intending to play 22.Qxd4??, saw the reply 22...Ne2+, and resigned. I should have played on with 22.Qxc1, but perhaps early resignation was for the best.