The Wilkes-Barre/Traxler is so tricky it can even surprise a GM using the most powerful chess engine. What has been known since Williams, 1978, The Real American Wilkes Barre, is that 6...Rf8 is a blunder leading to a big advantage White. Critical is 6...Qe8 7.d3 d6 8.Nf3!? What I pointed out to the GM is that White has retreated the Knight without being provoked by h6 so this loses a tempo in a sharp position. In the internal harmony of chess, if Black makes nothing but attacking moves here, he should be okay. so 8...Qg6 9.Be3 Bg4! 10.Nbd2 Raf8!! (This is the critical improvement; now Black's position hangs by a thread. It is the ability of Black to walk his King to the Queenside without hemming in his Queen Rook that is critical.
11.Bxc5 dxc5 12.h3 Bxf3 13.Qxf3 Nd5 14.Qg4 Nf4 15.Rh2 Qf6 16.g3 h5 17.Qf3 Nd4 18.Qe3 Nfe6 19.h4 Kd8 20.O-O-O Nxb3 21.axb3 b6.
According to the GM White is still winning by +.6 but I am not so sure. For White to make progress he has to be able to play f4, but at the present time the f-pawn is backwards on an open file. Black can triple on the f-file, then play Qh6. If White tries to build for f4 by playing Ng2, then Black has time to arrange his Queenside with b5/c6/a5/Kb6. When the White Knight invades on f5, Black just plays g6 driving it back. White was never able to muster threats against the e5 pawn.
I see no way for White to make progress.
The residual danger here for White is that this is almost a heavy piece ending where it is recognized that this is very difficult to win a pawn up because of the threat to draw by perpetual check. This is a very difficult theoretical/practical win.
I retained a Grandmaster to analyze the Wilkes-Barre/Traxler. He used a computer he characterized as "just below industrial grade". His final comment, "It is very hard to get greater advantage in any other normal opening."
1. e4 e5 2. Nf3 Nc6 3. Bc4 Nf6 4. Ng5 Bc5 5. Bxf7+ Ke7 {Bc4 and Bd5 are ok
also, but this is the most accurate move} 6. Bb3 Rf8 (6... Qe8 7. d3 d6 8. Nf3 Qg6 9. Be3 $1 Bxe3 10. fxe3 Rf8 11. Qd2 $16) (6... d6 7. d3 {Even the engines like Houdini on powerful hardware prefer accurate and safe way to play to secure the winning edge} Qe8 8. Nf3 Qg6 9. Be3 Bg4 10. Nbd2 {Black has no shadow of compensation for a pawn...white is technically winning}) (6... Nd4 7.c3 Nxb3 8. Qxb3 Qg8 {the best, but also quite bad for black} (8... Qf8 9. Nf3 d6 10. d4 Bb6 11. dxe5 Nxe4 12. O-O $18) 9. Qc2 {In the middlegame, white's advantage is a lot greater} d6 10. Nf3 Bg4 11. d4 Bxf3 12. gxf3 Bb6 13. Rg1 g6 14. Be3 Qe6 15. Nd2 {And again...white is a pawn up with the great position}) 7. d3 d6 8. Nf3 Qe8 9. Be3 $16 Qg6 10. Nbd2 Bxe3 11. fxe3 Qh6 12. Qe2 $16 (What the Grandmaster didn't point out in the final move sequence is 12.Qe2 Ng4 13.Nf1 +/-)