In my ancient book on the Scotch (Botterill & Harding, Batsford, 1977), they say "This move [(3...Nxd4)] doesn't look right, but it's not easy to point to any specific line that shows it up. Probably white should simply content himself with [4.Nxd4], transposing to variation A in chapter 10." Variation A in chapter 10 is: 1.e4 e5 2.Nf3 Nc6 3.d4 exd4 4.Nxd4 Nxd4. There, they say "This is a rather naive plan that condemns Black to a passive stance in the centre." and each line in their analysis leads to a "+/-" position for white.

If 3...Nd4 really equalised against the Scotch, I would be tempted to analyse 1.e4 d6 2.d4 e5 to a draw.

But really, this cannot be good. 

If you want to mix things up a bit, why not try Malaniuk's 3...ed4 4.Nd4 Bb4 5.c3 Be7!?. I posted a file on this line on the Forum somewhere, if you look in the list of Forum pgn files you will find it.

I've been thinking about changing my approach against the Scotch, and have come across the move 3...Nxd4.  I'm not about to take it up, but would be interested to know a bit more about it.

From Wikipedia (not, I realise, the most reliable source of chess knowledge):
"3...Nxd4 is possible, though rarely played today by strong players. It was popular in the 19th century, and receives five columns of analysis in Freeborough and Ranken's opening manual Chess Openings Ancient and Modern (3rd ed. 1896 p. 53). It is often described today as a strategic error, since after 4.Nxd4 exd4 5.Qxd4 (5.Bc4 is the Napoleon Gambit) White's queen stands on a central square, and is not developed too early since it cannot be chased away very effectively (5...c5? is a seriously weakening move that blocks Black's king's bishop). Nonetheless, the Encyclopaedia of Chess Openings (volume 3, 3rd edition 1997, p. 251 n.28, referring to p. 252 line 1) concludes that Black equalises with 5...Ne7 6.Bc4 Nc6 7.Qd5 Qf6 8.0-0 Ne5 9.Be2 c6 10.Qb3 Ng6 11.f4 Bc5+ 12.Kh1 d6. Similarly, Harald Keilhack concludes in Knight on the Left: 1.Nc3 (p. 21) that although ...Nxd4 is a "non-line" these days, if Black continues perfectly it is not clear that White gets even a small advantage. Keilhack analyses 5.Qxd4 d6 6.Nc3 Nf6 7.Bc4 Be7 8.0-0 0-0 9.Bg5 c6 10.a4 Qa5 11.Bh4 and now after 11...Qe5 or 11...Be6, "White has at most this indescribable nothingness which is the advantage of the first move." (Id. p. 25) The Encyclopaedia of Chess Openings also concludes that Black equalises after the alternative 4.Nxe5 Ne6 5.Bc4 Nf6 6.Nc3 Be7 7.0-0 0-0 8.Be3 d6 8.Nd3 Nxe4 10.Nxe4 d5 (p. 251 n.28)."

In MegaBase 2011 I have only 242 games with this move, which seems a staggeringly small number if it does indeed equalise without too much difficulty for black.  Furthermore, white scores exceptionally well after both Nxd4 and Nxe5, averaging 74%.  Some of this is no doubt due to the fact that the move is probably mainly played by weak players, but it still seems a surprisingly high percentage for a move which it is claimed is equal.

Can anyone shed any more light on this variation?

Will