I've taken to using the Scotch again and on those rare occasions I see a Steinitz Variation (4. ... Qh4), I like to employ a gambit line that Karjakin has used to good effect:

8. Nb5 Kd8 traps the Black King in the centre while White is strong there (8 ... Qxg2 is bad.) It looks like a fairly strong move.

I'm curious about whether castling is a better, worse, or roughly equal option to 8. Nb5. After 8. O-O Nxd4 the position seems pretty passable for Black. White has some compensation for the pawn in the bishop pair, space in the centre, and a small development edge, but that exchange seems to mute a lot of White's attacking chances. Is there something I'm missing in this continuation? (8. ... Nf6 looks pretty weak as an alternative for Black.)

Thoughts?

Well, 8 Nb5! seems so strong and forcing I don't know why we need to look at alternatives.  8 ... Kd8  9 0-0 Nf6 (what else?)  10 Re1 with Bd3 to follow, and a sweet game for White.

If 8 0-0 Nxd4 (the first signal of the weakness of Black's position is that if he needs to do this undoubling the pawns, he's in real trouble, but 8 ... Nge7, for example, just allows White another chance for 9 Nb5!), 9 cxd4 and Black is already uncomfortable.  But this edge doesn't compare with the huge initiative 8 Nb5! yields.

But it doesn't look much better to decline the gambit with 6 ... Nf6 instead, due to 7 Nf5! Qxe4  8 Nxg7+ Kf8  9 Bh6 and White seems to be having more fun.  And there really aren't many alternatives to 5 ... Bb4, since 5 ... Bc5 is met with 6 Be3.  That leaves 4 ... Qh4 as the culprit, which means the whole variation may be busted by this gambit!