I am not exclusively an 1.e4 player, and in a recent online game I tried the Queen's Gambit (1.d4 d5 2.c4). In my Exeter days I also used this occasionally and had some exciting games against a Slav player which went 2...c6 3.Nc3 e6 4.Nf3 dxc4 and then I turned it into a real gambit with 5.e4 b5. But on this occasion, the hunter became the hunted, as I met with the Hennig-Schara Gambit: 2...e6 3.Nc3 c5 4.cxd5 cxd4!?.
I haven't played this with Black (not yet, anyway) as I have been having too much fun with the Albin Counter-Gambit, and have occasionally experimented with the Slav, but I think it's a pretty reasonable gambit. The main line then goes 5.Qa4+ Bd7 6.Qxd4 exd5 7.Qxd5 Nc6, but I allowed my opponent a couple of additional options by playing 5.Qxd4 Nc6 6.Qd1 exd5 7.Qxd5. Here most players transpose with 7...Bd7, but 7...Bd6 and 7...Be6 are both possible.
When I was growing up in the 1990s and early 2000s, usually the recommended approach for Black was to castle queenside and go for a kingside pawn storm, but with a normal edge for White. But a few recent articles (e.g. by Nigel Davies over at Chesspublishing) have been suggesting that Black is doing better by castling short and relying on piece play. I also saw a ChessPub forum post from a couple of years ago by MNb which suggested that Black's queenside castling plan is close to refuted if White plays a3 and then meets ...g7-g5-g4 with Nh4.
The latest version of Stockfish agrees with all of this. The most critical position arises after 7...Bd7 8.Nf3 Nf6 9.Qd1 Bc5 10.e3 Qe7 11.Be2...
Here Stockfish assesses 11...0-0 as about +0.5 for White, which I reckon is pretty decent for a gambit with the black pieces. In the Lumbras GigaBase Black has played this line in a number of grandmaster level games with reasonable results, and in some of them Black still ends up building a kingside attack, while White doesn't have much in the way of attacking chances. Black tends to follow up with ...Rfd8 and ...Ne5.
But the line with 11...0-0-0 12.0-0 g5 13.a3 g4 14.Nh4 is assessed as close to +1. Looking over the games, Black still gets attacking chances here, but Black's attack is slowed down quite a bit and White's attacking chances on the queenside are as good as Black's on the kingside, hence the engine assessment of Black having very little compensation for the pawn. In the database, White has only got this far in seven games out of 528, and Black's results have been quite good overall, so I'd say that the 0-0-0 plan is still viable if the opponent isn't strongly prepared, especially at lower levels and faster time limits. But if White is well prepared, White will be doing very well, and certainly most GM games played since around 2005 have seen Black castle short.
I guess I had mixed feelings when my opponent deviated with 7...Be6 8.Qxd8 Rxd8.
This is an example of where computer evaluations can only get you so far, as Stockfish reckons that this line isn't substantially inferior, assessing it as +0.6 for White, but I reckon that in practice it is substantially inferior. With the queens off, Black has to play more accurately in order to maintain some compensation for the pawn. I continued with 9.e3, preventing ...Nd4, and my opponent was a little hesitant, so I ended up trading down into a winning king and pawn endgame - probably the biggest risk associated with playing gambits, but this time I was the one defending against one.
From a practical/results point of view, this was a good outcome, but I think a far more interesting game would have ensued if my opponent had gone ...Bd7, or the less common ...Bd6, I was happy to take on the challenge of meeting one of Black's more aggressive approaches.