Black's most critical test of the gambit is to accept it, and if you know what you're doing, accepting the gambit offers Black good winning chances. However, it also gives White what he/she wants, and gives White the majority of the piece play and attacking chances, and the resulting positions often become complicated very quickly.
With 5.Nxc3 White settles for sacrificing just one pawn. This is less risky than sacrificing a second pawn with 5.Bc4, but it allows Black to get into lines following 5...Bb4 and 5...Bc5 that White can avoid by using the 5.Bc4 move-order (if 5.Bc4 Bb4, 6.0-0 and 6.bxc3 are probably both preferable to 6.Nxc3, and if 5.Bc4 Bc5, White has the tactic 6.Bxf7+ followed by 7.Qd5+ and 8.Qxc5.)
Even if you prefer to sacrifice the second pawn with 5.Bc4, it is worth looking at these 5.Nxc3 lines to familiarise yourself with some of the key ideas, and especially the lines with 5.Nxc3 d6, for after 5.Bc4 d6 White has nothing better than 6.Nxc3 transposing to 5.Nxc3 d6 6.Bc4.
White's main ideas are:
To put pressure on f7 with Bc4, Qb3 and/or Ng5. Ng5 can also be followed by the advance of the f-pawn.
If Black's bishop on f8 ends up cut off from the defence of the kingside (notably in the lines with ...Bc5 followed by ...d6), it can be better to attack on the kingside dark squares with Bg5 and Nd5. (This is not as effective against the 5...Bb4 lines because Black can prevent Nc3-d5 by exchanging off the knight on c3).
If Black plays ...Nf6, to be able to kick the knight by playing e4-e5. This often accelerates White's initiative. The threat of hitting ...Nf6 with e4-e5 makes Black's kingside development difficult, a key part of White's compensation for the pawn.
To catch the black king in the centre.
Black's main ideas are:
If White goes for the Bc4, Qb3 approach, a key resource is to "fork" the queen and bishop with a well-timed ...Nc6-a5, forcing the exchange of the white bishop on c4 for the black knight on c6. This reduces White's attacking potential.
Nf3-g5 from White can often be met by ...Nc6-e5 followed by ...h7-h6, kicking the knight, although in many variations White can then mix things by attacking the knight on e5 with f2-f4. Alternatively in some variations parrying the threat to f7 with ...Ng8-h6 can work, but this does develop the knight to a sub-optimal square.
To try and develop the kingside pieces quickly and get castled to comparative safety.
Other ideas for Black vary depending on what Black chooses to do with the bishop on f8. Black can cover the e5-square with the immediate 5...d6 intending ...Nf6, ...Be7 and ...0-0. Alternatively Black can develop the bishop actively at b4 or c5, with the idea of playing ...d6, ...Nf6 and ...0-0. If the bishop goes to b4 it is often with the aim of chopping off the knight on c3, exchanging a pair of minor pieces, but at the cost of giving White the bishop-pair in an open position. Here is an overview of the key variations.
Illustrative games
Overview of variations
5...Bb4
5...Bb4 has traditionally been regarded as the most critical response. Black pins the knight on c3 to the king and intends to swap it off, eliminating a potentially strong attacking piece but at the cost of giving White the bishop pair in an open position. Black generally intends to play ...Nf6 and ...0-0 with a solid position. The main line is then 6.Bc4, continuing development and preparing to attack Black's weak point on f7. Then Black's main counter is 6...d6, which adds protection to the e5-square and so prepares to play ...Nf6 and ...0-0 without allowing White to kick the knight with e4-e5.
5...Bb4 6.Bc4 d6
After these moves we have:
a) 7.Ng5 is the most reliable way to get compensation for the pawn, attacking Black's weak point on f7 and forcing Black to make concessions: a1) 7...Ne5? 8.Qa4+ and Black loses material, for the bishop on b4 is attacked and if Black defends the bishop with 8...Nc6, f7 is left undefended. a2) 7...Nh6 8.0-0 0-0 (or 8...Bxc3+ 9.bxc3 with f4 to follow) 9.Nd5 Bc5 10.Nf3!, threatening Bxh6, gives White good compensation, as pointed out by Karsten Müller and Martin Voigt in Danish Dynamite. a3) 7...Bxc3+ 8.bxc3 Ne5 (or 8...Nh6 9.0-0 intending 10.f4) 9.Bb3 leads to complications and approximate equality with best play, envisaging 9...h6 10.f4.
b) 7.0-0 intending Nc3-d5 also leads to many complicated variations, but after 7...Bxc3 8.bxc3 Nf6, preparing kingside castling, it is harder for White to get enough compensation for the pawn. White can then go for positional pressure with 9.Bg5 or sacrifice a second pawn with 9.e5, but both most likely leave White a bit short of full compensation. If 9.Ba3 then White lacks an adequate response to 9...Bg4!. Black can instead try 8...Bg4!? envisaging 9.Qb3 Bxf3+ 10.Bxf7+ Kf8, allowing f7 to drop off in return for attacking chances against White's weakened kingside, although objectively this line is less good - if White defends accurately (admittedly a big "if"!) then White has good chances of emerging from the complications with a slight edge.
c) Similarly 7.Qb3 Bxc3+! 8.bxc3 (8.Qxc3 is met by 8...Qf6) 8...Qe7 (8...Qd7 9.Ng5 Nh6 threatens ...Na5, but then 10.Qc2 with f4 to follow gives White decent compensation) 9.0-0 Nf6, preparing kingside castling, gives White difficulties getting enough compensation. Again White has a choice between seeking positional compensation with 10.Bg5 or sacrificing a second pawn with 10.e5, both of which again probably leave White a bit short of full compensation, and 10.Ng5?! leads nowhere after 10...0-0. d) With Black having played ...d6, 7.Bg5 can't be met by ...Be7, but on the other hand Black can now play ...Nf6 without having to worry about e4-e5. Thus 7...Bxc3+ 8.bxc3 Nf6 is a solid defence, whereupon 9.0-0 is probably best, transposing to 7.0-0 Bxc3 8.bxc3 Nf6 9.Bg5.
6th move alternatives
There are also some important sixth move alternatives for both sides: a) 6.Bg5 intends to pin the g8-knight to the queen, and then build up positional pressure for the pawn, but Black does not have to comply: a1) 6...Nge7 7.Qb3 (White can also consider 7.Qc2 envisaging queenside castling and a kingside pawn storm, but this plan is probably too slow, as was demonstrated in Gufeld-Stein, Kiev 1959) 7...0-0 8.Bc4 gives White a fair amount of positional pressure for the pawn, although if Black goes for development with 8...d6 I am not sure if it is sufficient. a2) 6...Be7!? is a surprisingly robust defence despite the loss of a tempo (...Bb4 and then ...Be7), as White can't make much happen after 7.Bxe7 Ngxe7 8.Bc4 0-0. Black intends ...Nf6 and ...0-0 with a solid position. Perhaps White's best bet is to retreat 7.Bf4 and then try to mix things by castling queenside and envisaging a kingside pawn roller.
After 6.Bc4 Black has various alternatives to 6...d6, but no convincing independent alternatives: a) Unless Black is convinced by the line 6...d6 7.Ng5 Nh6, there is nothing wrong with flicking in 6...Bxc3+ 7.bxc3 and then 7...d6, as with best play it will usually transpose.
b) 6...Nge7 intends kingside castling but after 7.Ng5 Black can't play 7...0-0? because 8.Qh5 attacks f7 and h7, so play tends to continue 7...Ne5 8.Bb3 h6 9.f4 and in this line the retreat 9.Nf3!? is also playable. The slower 7.0-0 also gives enough compensation, e.g. 7...0-0 8.Nd5 Nxd5 9.exd5 Ne7 and Black's king lacks defenders, or 7...Bxc3 8.bxc3 followed by Nd4 and f4.
c) 6...Qe7 threatens ...Bxc3+ followed by ...Qxe4+, but then 7.0-0 is a strong reply, as now 7...Bxc3+ 8.bxc3 Qxe4? is met by 9.Re1 pinning and winning the queen. The greedy 8...Ne5?! leaves Black facing a big attack after 9.Nxe5 Qxe5 10.Qb3. 8...d6 is probably Black's best bet, but then White continues the attack with 9.Re1 or 9.Ba3, threatening e4-e5 in both cases.
d) 6...h6 is often seen at club level, fearing Nf3-g5, but as ...h6 does not contribute to development, both 7.0-0 and 7.Qb3 now give White good compensation for the pawn.
Classical development: 5...Bc5
This is a tough and generally underrated defence. The idea is again ...Nf6 and ...0-0 with a solid position. On c5 the bishop reduces White's control over the d4-square and ensures that if White castles kingside it will be difficult to engineer the advance of the f-pawn. For these reasons, White's immediate attacks on f7 are less strong than against most of Black's other defences. The downsides are that the white knight on c3 is free to come to d5 in this line, and that the bishop on c5 typically gets cut off from defending the black kingside after 6.Bc4 d6. As a result, the plan of putting pressure on the black kingside with Bg5 and Nd5 is quite strong.
Here 6.Bc4 d6 is again the usual continuation. The developing move Bc4 prepares kingside castling and hits Black's weak point on f7, and threatens Bxf7+ followed by Qd5+ and Qxc5, taking advantage of the position of the undefended bishop on c5. Black does best to play 6...d6, defending the bishop and preparing ...Nf6 and ...0-0.
5...Bc5 6.Bc4 d6
Here play branches out into the following lines: a) 7.Bg5 is the most usual continuation, attacking the black queen, and as Black has played 6...d6, Black cannot retreat the bishop on c5 back to e7. All of Black's ways of addressing the attack on the queen involve making a concession: a1) 7...Qd7 blocks in the c8-bishop. Here White has generally done well with 8.Qd2 envisaging queenside castling and a kingside pawn storm. a2) 7...f6 weakens the kingside light squares, but it is playable. 8.Bh4 appeals to me the most, pinning the f6-pawn against the queen, and if 8...Bg4, 9.h3, since the tactic 9...Bxf3 10.Qxf3 Ne5, hitting the queen on f3 and the bishop on c4, is thwarted by 11.Qh5+, taking advantage of the weaknesses created by 7...f6. a3) 7...Nf6 develops the knight to what is normally its best square, but this line is one of the exceptions. 8.Nd5?! is premature due to the tactic 8...Bxf2+, so 8.0-0 is probably best, intending Nd5 on the next move. Then play often continues 8...Bg4 9.Nd5 Ne5 10.Be2 or 8...h6 9.Bxf6 Qxf6 10.Nd5 Qd8 11.b4 with a4 to follow, and in both cases White has good compensation. 8.Qc2!?, suggested by Karsten Müller and Martin Voigt in Danish Dynamite, envisages queenside castling and also looks fully playable, though it has not yet been tested as of 2018. a4) 7...Nge7 is the most critical response, aiming to eventually break the pin, although Black has to be careful as in some variations Nd5 followed by b2-b4-b5 can end up costing Black a piece. Here again 8.Nd5 is probably too hasty because 8...h6! causes difficulties, and 8.Qc2?! and queenside castling is not as effective here because Black can castle kingside and kick the bishop on g5 back to g3 with ...h6 and ...Ng6, whereupon the bishop on g3 blocks the advance of the g-pawn. So White should reply 8.0-0, and then follow up with 9.Nd5 in most cases.
b) 7.0-0 is a reasonable alternative, intending 8.Bg5. It does allow Black the independent possibility 7...Bg4 intending 8.Bg5 Qd7!?, although 9.Nd5 then gives White decent compensation. Otherwise play tends to transpose back into 7.Bg5 lines, viz. 7...Nf6 8.Bg5 or 7...Nge7 8.Bg5. An advantage of playing 7.0-0 first is that it tends to tempt Black into the more active-looking 7...Nf6 when 7...Nge7 is probably the more critical line, and it also avoids the lines 7.Bg5 Qd7 and 7.Bg5 f6, but it also denies White the possibility 7.Bg5 Nf6 8.Qc2!?.
c) 7.Qb3 attacking f7 is dubious in this line because of 7...Qd7! threatening ...Na5 which would force the exchange of the Bc4 for the Nc6, and after 8.Ng5 Nh6, if White pushes the f-pawn it makes White's castling problematic as the bishop on c5 then covers the g1-square. So White often tries a trap with 8.Nd5 intending 8...Na5? 9.Qc3, but 8...Nge7 is then a robust defence.
d) 7.Ng5 attacking f7 also doesn't seem to work in this line. One example is the line 7...Ne5 8.Bb3 h6 9.f4 Bg4, when White can't play 10.Qd4 due to the bishop on c5, and so Black is better.
6th move alternatives
After 5...Bc5 neither side has any good reason to deviate from the continuation 6.Bc4 d6.
a) If White tries 6.Bg5 play will often transpose to the line 6.Bc4 d6 7.Bg5, but Black has the possibly superior deviation 6...Be7!?, as per 5...Bb4 6.Bg5 Be7. And after 6.Bc4, Black doesn't have a good alternative to 6...d6:
a) 6...Nge7 7.Ng5! is problematic for Black, as 7...0-0? loses to 8.Qh5, and 7...Ne5 can be met by 8.Nxf7 Nxf7 9.Bxf7+ Kxf7 10.Qh5+ and Qxc5.
b) 6...Nf6?! 7.e5! is even stronger for White than in the analogous line 5...Bb4 6.Bc4 Nf6 7.e5. If 7...d5 8.exf6 dxc4 then 9.fxg7! Qxd1+ 10.Kxd1 with the deadly threat of Nd5 gives White a very strong attack. And if 7...Ng4 then 8.Bxf7+ intending 8...Kxf7 9.Qd5+ and Qxc5 or 9.Ng5+ and Qxg4 regains the pawn with a continued attack.
c) 6...Qe7? 7.0-0 followed by Nd5 is pretty disastrous for Black.
d) And 6...h6?! is met by 7.Bxf7+ with the idea 7...Kxf7 8.Qd5+ and Qxc5.
Solid defence: 5...d6
This prepares solid development with ...Nf6, ...Be7 and ...0-0. This line also commonly arises after 5.Bc4, as 5...d6 is probably the most reliable way of declining the second pawn, and then White has nothing better than 6.Nxc3. Best play then continues 6.Bc4 Nf6, which leads to another complicated mess: a) 7.Qb3 is the most usual continuation, attacking f7. a1) 7...Qe7?! defends f7 but then the queen ends up vulnerable to being hit by Nc3-d5 after either 8.0-0 or 8.Bg5. a2) 7...Qd7! is the right move. While it blocks in the c8-bishop, the queen blocks the a4-e8 diagonal so that Qa4 is not check, and so if 8.0-0?!, 8...Na5 forces the exchange of the Bc4 for the Nc6. Play then typically continues 8.Ng5 Ne5 (8...Nd8 is rather passive, and White continues with f4 with the idea of pushing the central pawns) 9.Bb5 (9.Be2, as favoured by Martin Voigt, is more passive) 9...c6 10.f4, with board-wide chaos. This is a heavily tested and analysed position and so I feature it in a few of the illustrative games.
b) 7.Ng5 is equally good, also attacking f7, for having played ...Nf6, Black cannot parry the threat with ...Nh6. Then Black's only good response is 7...Ne5! and now: b1) 8.Bb5+ c6 9.f4 is not so effective because of 9...Ng6 intending 10.Bc4 d5 striking out in the centre. b2) 8.Bb3 h6 (8...Bg4 is also playable, but then 9.f3 Bh5 10.0-0 gives White good compensation) 9.f4 leads to considerable complications again: b2a) 9...Bg4 10.Qd4 gives White reasonable compensation (the retreat 10.Nf3 also appears to be playable, but it has scored badly for White in practice) b2b) 9...hxg5 10.fxe5 Bg4 11.Ba4+ c6 12.Qb3 is another plausible continuation, again with decent compensation for White.
c) 7.0-0?! cannot be recommended as it does not put Black under enough pressure. Black executes the plan with 7...Be7! intending kingside castling, but not 7...Bg4?! 8.Qb3!. c1) 8.Nd5 is the trickiest answer, offering the pawn on e4, but after 8...0-0! (not 8...Nxe4?! 9.Re1 with a dangerous attack down the e-file) Black threatens 9...Nxd5 and then 10...Ne5. The e5-square is an important outpost for Black's knights in this 7.0-0 Be7 line, and it is mostly this resource that denies White full compensation for the pawn. c2) 8.h3 prevents the idea of ...Nf6-g4-e5, but then 8...0-0 is again a good reply, and if, say, 9.Bf4, then 9...Nd7!? prepares to bring the knight to e5 anyway.
6th move alternatives
In this line 6.Bc4 is obligatory, as White needs to attack f7 to disrupt Black's development plan of ...Nf6, ...Be7 and ...0-0, and 6.Bg5?! is well met by 6...Be7, leaving Black a tempo up on the already strong line 5...Bb4 6.Bg5 Be7. So here I will focus on Black's sixth move alternatives.
a) 6...Be7?! 7.Qb3! leaves Black unable to defend f7. Thus Black tends to try to grab the bishop pair as compensation for losing the extra pawn and having an exposed king, viz. 7...Na5 8.Bxf7+ Kf8, whereupon White can't hold onto the light squared bishop (if 8.Qd5?! c6! and White has to continue accurately not to lose a piece) so 9.Qa4 is the usual continuation, rounding up the knight on a5, and 9...Kxf7 10.Qxa5 typically follows. White has some advantage, mainly due to Black's problems with king safety. Black can also consider 9...c6 whereupon White again can't preserve the bishop on f7, as 10.Bh5? Nf6! is embarrassing. 10.Bxg8 followed by 11.0-0 is the usual continuation.
b) 6...Be6 7.Bxe6! (Nd5 is often a good answer to ...Be6, but here 7.Nd5 Be7 is a good reply) 7...fxe6 8.Qb3! attacks e6 and b7: b1) 8...Qc8 9.Ng5 attacks e6 again and Black is obliged to play 9...Nd8, leaving the black pieces clustered on the back rank. White can then start the slow advance of the central pawns with 10.f4, aiming to catch the black king in the centre. White has very good compensation for a pawn. b2) 8...Qd7 is trickier because if 9.Ng5 Nd8 the black queen is more actively placed on d7, where it helps defend the second rank, and it is harder for White to demonstrate enough compensation for a pawn. Thus regaining the pawn with 9.Qxb7 is the more reliable continuation, which gives White a small advantage following 9...Rb8 10.Qa6 Be7 11.0-0.
c) 6...Bg4 7.Qb3! attacks b7 and f7: c1) 7...Bxf3?? loses to 8.Bxf7+ Ke7 9.Qe6#. c2) 7...Qd7 is the best reply: c2a) White the tactic 8.Bxf7+ Qxf7 9.Qxb7, attacking the rook on a8 and the knight on c6 and so regaining the pawn. Black's only way to generate counterplay here is to sacrifice the exchange with 9...Kd7! 10.Qxa8 Bxf3 11.gxf3 Qxf3. White has scored heavily from this position but objectively Black's counterplay may be sufficient. c2b) 8.Ng5 is a good alternative, keeping up the pressure: if 8...Nh6 threatening ...Na5, 9.Bd5 preserves the light squared bishop.
d) 6...h6 is a common reaction at club level, fearing Nf3-g5, and it is not as bad as it looks. d1) 7.Qb3 Qd7! Black again threatens ...Na5, forcing the exchange of the Bc4 for the Nc6, although with Black well behind in development, White can afford to allow this and still get enough compensation, viz. 8.0-0 Na5 9.Qb5 Nxc4 10.Qxc4 Be7 11.Rd1 with the threat of e4-e5. If White does not want to go in for this line, White can back out with 9.Qc2 with f2-f4 to follow. d2) A good alternative is 7.0-0, a waiting move that aims to play Qb3 under better circumstances, as Black cannot defend f7 after 7...Be7?! 8.Qb3. Black's best is 7...Nf6 intending 8.Qb3 Qd7 and then White can consider either 9.Rd1 Na5 10.Qb5 Nxc4 11.Qxc4 with the threat of e4-e5, or 9.Bb5!?. White has scored heavily in these lines.
5th move alternatives for Black
There are various 5th move alternatives to consider, but they are not really a serious test of White's gambit.
a) The natural 5...Nf6 is not played as often as 5...d6, even though it often transposes after 6.Bc4 d6 (other moves tend to be met by 7.e5, hitting the knight on f6. White does have the alternative 6.e5!? which looks good. Black can try and round up f2 with 6...Ng4 when 7.Qe2 Bc5 8.h3!? leaves Black in a tangle, and Black's best is probably to sacrifice a piece for two pawns and some attacking chances. For this reason Black is probably right to prefer 5...d6 to 5...Nf6.
b) 5...Be7 cannot be recommended because of 6.Bc4, and then if 6...Nf6 7.e5, or if 6...d6 7.Qb3 (transposing to 5...d6 6.Bc4 Be7 7.Qb3) and Black cannot defend f7.
c) 5...Bd6 was brought to my attention by Stefan Bücker, citing Siegfried Kalkofen. The idea is to gain a foothold on the e5-square. After 6.Bc4 Nf6 we get a position that more commonly arises from 5...Nf6 6.Bc4 (note 6.e5!?, as discussed above) 6...Bd6. White's best is probably 7.0-0 (or 7.Bg5, which tends to transpose) 7...0-0 8.Bg5, with the idea of meeting the retreat 8...Be7 with 9.e5.
d) 5...g6 intends a kingside fianchetto, with the idea of 6.Bc4 d6 (not 6...Bg7?! 7.Qb3!) and now 7.Bg5?! is too hasty as it is well met by the surprisingly strong 7...Be7!. It is not so easy to exploit Black's dark square weaknesses once the dark squared bishops are exchanged. However, White has two very good alternatives that give dangerous attacking chances: d1) 7.0-0 was suggested by Mark Nieuweboer, essentially a waiting move intending 7...Nf6?! 8.Ng5 and 7...Bg7 8.Bg5. d2) 7.Qb3 is also good. Then 7...Qd7 threatens ...Na5, but 8.Nd5 keeps up the pressure, for 8...Na5? is well met by 9.Qc3 attacking the rook on h8, and if 8...Bg7 9.Bd2!?, preventing ...Na5.
Conclusions
In summary, the 5.Nxc3 line appears to be reasonably sound as well as dangerous, but objectively Black's resources are enough to keep the chances level. More dangerous but less definitely sound is to offer the second pawn with 5.Bc4.