As all the various lines become a little bewildering already, I have decided to make a summary. It includes most lines from above and also a few comments of mine.
1.e4 e5 2.f4 exf4 3.Nf3 d6 4.d4 g5 5.Nc3 Bg7 6.g3 g4 7.Nh4 f3
7...Nc6 Furhoff-Sandor, Budapest 1994, see above, must also be taken seriously.
8.Bc1-e3,
It is not clear yet, when White must insert the moves h3 and h5. Frendo's analysis runs 8.h3 h5 9.Be3 Bf6 (Nc6 10.Qf2!?) 10.hxg4 hxg4 11.Qd2 Bxh4 12.Rxh4 Rxh4 13.gxh4 Qxh4+ 14.Qf2 Qxf2+ 15.Kxf2 but Ne7 16.Bc4 Kf8 17.Bh6+ (17.Rh1 Kg7) Kg8 leaves Black with an advantage. White has 10.Nf5 Bxf5 11.exf5 eg d5 12.hxg4 hxg4 13.Rxh8 Bxh8 14.Qd2 Nc6 15.o-o-o Qd7 16.Bg5 f6 17.Bf4 o-o-o 18.Nb5 with decent compensation.
8.....,Bg7-f6;
8...Nc6 9.Qd2 Nce7 10.h3 h5 11.o-o-o c6 Kennaugh-Bellin, ENG 1999, 12.Bf4 idea e4-e5 unclear.
8...Nge7 9.Qd2 (9.h3 h5 10.Qd2 is possible)
a)9...Nd7 10.o-o-o Nb6 11.Bd3 Be6 12.h3 h5 13.Rde1
1)13...c6 14.Nf5 (14.hxg4 hxg4 15.Nf5 Bxf5 16.Rxh8+ Bxh8 17.exf5 Kd7 18.Bg5 is possible)
Bxf5 15.exf5 d5? (Qc7 or Qd7 16.Bf2 with full compensation or Kd7 16.Ne4!?) 16.Bg5 f6
17.Bf4 Qd7 18.hxg4 hxg4 19.Rxh8+ Bxh8 20.Qh2 o-o-o 21.Qh7 +-.
2)13...Nc4 14.Bxc4 Bxc4 15.Bf4 Qd7 16.Nf5 Rg8 17.hxg4 hxg4 18.b3 +=Shulman-Kamberi,
Stillwater 2001.
b)9...Ng6 10.Nf5 Bxf5 11.exf5 Ne7 12.Bg5! f6 13.Bf4 Nxf5 14.o-o-o with compensation (Frendo).
9.Qd1-d2,
Quite dubious, but not really clear is 9.Nxf3 gxf3 10.Qxf3 Be6 11.e5.
9.Nf5 Bxf5 10.exf5 Bg5! 11.Bxg5 Qxg5 12.Qd3 Nc6 is probably good for Black.
9.....,Nb8-c6;
9...Bxh4 10.gxf4 Qxh4+ 11.Bf2 Qh6 12.Qxh6 Nxh6 13.Nd5 with compensation in Furhoff-Eriksson, Helsingborg 1991.
10.o-o-o,
Another idea is 10.Nd5 Bxh4 11.gxh4 Be6 12.o-o-o Bxd5 (Qd7 13.Nb4!?) 13.exd5 Nce7 14.Bb5+.
10.....,Ng8-e7;
10...Bxh4 11.gxh4 Bd7 12.Qf2 13.h3 leaves Black with the question, how to finish his development.
11.Nc3-d5,
Or 11.Qf2 Bd7 12.h3 h5 13.Re1.
11.....,Ne7xd5;12.e4xd5,Nc6-e7;13.Be3-h6,
Dragonslayer also suggests 13.Bc4.
13.....,Ne7xd5;14.h2-h3,
He, who knows how to evaluate all this stuff, please stand up.
It is the remarkable games of the Swede (these fellows are more dangerous in the KG than on the footballfield, it seems) Furhoff indeed, who inspired me to investigate this variation. Still one should not consider me an expert, as my only experience runs
MNb - Koller,G [C37]
em ICCF EM/H/135, 2004
1.e4 e5 2.f4 exf4 3.Nf3 g5 4.d4 d6 5.Nc3 g4 6.Bxf4 gxf3 7.Qxf3 Bg7 8.0-0-0 Qe7 9.e5 dxe5 10.dxe5 Nc6 11.Nd5 Qd8 12.e6 1-0
which was much too easy, to be relevant.
The starting point of this thread was my claim 1.e4 e5 2.f4 exf4 3.Nf3 d6 equal. Overambitious defenders must also find something in
Rechel,B (2420) - Law,A (2310) [C37]
4NCB-Qiv1 0304 (2.6), 19.10.2003
1.e4 e5 2.f4 exf4 3.Nf3 g5 (or d6 4.d4 g5 5.Nc3 Bg7 6.h4) 4.d4 Bg7 5.Nc3 d6 6.h4 h6 7.hxg5 hxg5 8.Rxh8 Bxh8 9.g3 g4 10.Bxf4 gxf3 11.Qxf3 Bxd4 12.0-0-0 Nc6 13.Bb5 Bxc3 (retreats will be punished with 14.e5!) 14.Qxc3 Bd7 15.Bxc6 Bxc6 16.Qh8 Kd7 17.Qh3+ Ke8 18.Qh8 Kd7 19.Qh3+ Ke8 ½-½
There is also 3...d6 4.Qe2. See TalJechin's website:
http://hem.passagen.se/tjmisha/ Click chess!
Click King's Gambit Stuff
Click No Bust.
Only for KG enthusiasts!
My conclusion is, that 3...d6 offers Black less winning chances than 3...g5.