patzer2: Alternatives (Opening & Otherwise):
3...d6 Fischer's Defense or Fischer's Bust may be a good alternative for Black to 3...g5, which appears to be almost like tossing a coin to determine the winner among players of equal strength, as both sides have about even chances in the double-edged complications that follow. Of course the move 3...d6 is no where near a refutation or bust of the King's Gambit, as Fischer once claimed. See Hector vs Jonkman, 2003 for an example of the difficulties Black faces in equalizing, let alone winning in this line. Yet, it is a sound equalizing Black strategy that may be better than the mainline.
5. Ne5 (most popular move) leads to a clear White advantage according to Kasparov and Keene in BCO, while 5. Ng5 (The Allgaier Gambit) should result in a clear advantage for Black according to theory. Interestingly, the Opening explorer indicates better practical results for White with 5. Ng5 (54% wins to 36% losses) than with 5. Ne5 (43% wins to 42% losses), though modern theoretical improvements make those statistics suspect. However, results such as Black's victory against 5. Ne5 in A Fedorov vs Shirov, 2000 , against a top GM King's Gambit specialist, makes me wonder if 5. Ng5 might not be just as good a try for a White win, especially for those brave souls who like wild speculative sacrifices early in the opening.
Perhaps Black should play 7...f3!? as in Gunsberg vs Schlechter, 1903 or Gunsberg vs Bird, 1889 As long as Black does not get too greedy trying to grab material, as in Taubenhaus vs W Pollock, 1886 he should be OK. Black won 6 of 7 games in the ChessGames.com database with this move. However, the draw result in Fleissig/Marco vs Schlechter/Faendrich, 1903 seems to give a good line for White, though Black managed a win against this plan in Guinefolleau vs Pouplard, 1979
Perhaps 8...dxe5 is better as in B Wall vs R Illorde, 1988 where Black missed a winning opportunity on his 12th move.
10...h5? is too passive, and an opening blunder. Black is winning instead after 10...Nxe4!
Fritz 8 analyzes the win as 10...Nxe4! Nxe4 11. Nxe4 dxe4 12. Bc4+ Kg7 13. 0-0 Nc6 14. c3 <14. d5? Bc5+ > 14...Rf8 15. Qe2 Bxh4 16. Qxe4 Bf5 17. Bxh6+ Kxh6 18. Rxf5 Rxf5 19. Qxf5 Qg5 20. Qe6+ Qd7 Qg6 21. Qxg6+ <21. Qd7 Rf8 > 21...Kxg6 (-1.81 @ 15/58 depth & 709kN/s)