I recently managed to get in an attractive queen sacrifice in an online "thematic" game in the Steinitz Gambit (1.e4 e5 2.Nc3 Nc6 3.f4 exf4 4.d4). The game was not without mistakes but the end result was very satisfying.
It was vaguely inspired by a similar queen sacrifice that I once pulled off in the French Defence, Winawer Poisoned Pawn variation at my local chess club some years ago, admittedly in a well-known theoretical line:
1.e4 e6 2.d4 d5 3.Nc3 Bb4 4.e5
I like the line 4.a3 intending 4...Bxc3+ 5.bxc3 dxe4 6.Qg4, which leads to dynamically equal play and sometimes leads to Black coming under pressure on f6 after Bc1-g5, but on that occasion decided to take my opponent on in the main line.
4...c5 5.a3 Bxc3+ 6.bxc3 Ne7 7.Qg4 Qc7
I was involved in reading through some analysis of 7...0-0 at the local chess club a couple of months ago. Although objectively Black is probably OK, I much prefer White in that line, for White gets reasonable long-term attacking chances against the black king. My results as White against 7...Qc7 have been mixed. I have also tried it out a couple of times from the black side in online games, with success.
8.Qxg7 Rg8 9.Qxh7 cxd4 10.Ne2 Nbc6 11.f4 Bd7
11...dxc3 12.Qd3 d4 is most popular nowadays according to John Watson.
12.Qd3 dxc3 13.Nxc3
To my knowledge 13.Qxc3 is considered more critical, but the text move sets up a trap.
13...a6 14.Rb1 0-0-0?
Black needed to play a preparatory move like ...Na5 before castling queenside. White wins material with 15.Qxa6! because if 15...bxa6 16.Bxa6+ Black's only legal move is to block with 16...Qb7. I can't remember how the rest of the game went, but it was a fairly comfortable win due to the extra material and Black's exposed king.
Now onto the game in the Steinitz Gambit, which went as follows:
1.e4 e5 2.Nc3 Nc6 3.f4 exf4 4.d4
This was a "thematic" game. I would not normally play 2.Nc3 because of 2...Nf6 intending 3.f4 d5. The only way the Steinitz Gambit can be reached via a 2.f4 move-order is 2.f4 exf4 3.Nc3 Nc6 4.d4, or 3.d4 Nc6 4.Nc3, but in both cases Black does better to play 3...Qh4+ immediately.
4...Qh4+ 5.Ke2 d6
This move isn't too bad but White should be able to get full compensation for the pawn against this. More critical is 5...b6 intending ...Ba6+.
6.Nf3 Bg4 7.Bxf4 O-O-O 8.h3
The move played isn't clearly bad, but 8.Kd2 and 8.Ke3 are probably stronger here.
Steinitz would have been pleased with White's king position, reinforcing the centre. However Black could have caused White a few problems here with 9...Qf6, pinning the bishop on f4 and attacking d4.
9...Nf6 10.Qd2 Qh5+ 11.Ke3
Probably a bit too bold. 11.Kf2 is fine for White.
Not objectively best (it may well be deserving of a question mark), but I sensed that it was probably White's best practical choice, especially for an online game. 12.Kf3 g5 leaves Black with slightly the better of a messy position.
12...Qxh1 13.Nb5 a6 14.d5 Ne5 15.Qa5
Objectively best was 15.Na7+ Kb8 16.Nc6+ Nxc6 (Not 16...bxc6? 17.Qb4+ followed by 18.Bxa6) 17.dxc6 with some attacking chances on the queenside, although Black stands better.
Best was 15...Qg1+ intending 16...Qc5 and Black should be winning with accurate play.
16.Kd2 Rd7 17.Na7+ Kb8?
17...Kd8 was best although White has the upper hand after 18.Qb4.
If 18...bxc6, 19.Qb4+ followed by 20.Bxa6+ wins for White.
19.Qxa6 Qxg2+ 20.Kc1
The computer points out that 20.Kc3 actually forces checkmate because Black cannot exchange off the queen for White's bishop on f1. 20...Qf3+ 21.Kb4 Qxe4+ 22.Ka3 Qf3+ 23.c3 and Black runs out of checks. Steinitz would certainly have been proud of White's king boldly marching onto the fourth rank. However, the move played in the game wins comfortably because Black's only way to stop checkmate is to give up the queen. In the game, Black allowed the checkmate.