More thorough analysis has led me to believe that Wells has a point  . The knight on h3 is no improvement for black at all, it will go to f4 to prevent black's d5 and white is better. Wells considers 12...d5 instead and though I thought this was nonsense because black doens't try to take advantage of the weakness on e3 it looks like we get more or less the same positions as after 12...Nd5 13.e4 fxe3 or 12...Bc5 13.Nh3 Nd5 14.e4 fxe3, but with black's pieces better coordinated. Especially the variation 13.e3 fxe3 14.Ke2 Bc5! (14...Bd6 has been played more often, but seems illogical, as Wells points out) 15.Kxe3 Bd7! (Black immediately supports the b5-b4 advance) seems playable enough to me.

I wonder what Davies says about this line... Since he considers 8...Qxd4 I suspect he doesn't like it or doesn't consider it at all?

Joris:

    Davies considers 8...Qxd4 (avoiding the doubled b-pawns) 9.cxd4 e6 to be quite playable and this line seems likely to lead to a complicated French-like endgame after subsequent moves like ...Nc6, ...d5 and ...Bd7 for Black and e2-e4-e5, Be3 and Rc1 for White.  If you don't mind playing interesting endgames (really, more of a middlegame without Qs) I think this line provides chances for both players to outplay an opponent.  I am not sure who has the advantage in the resulting positions (White has some space, Black's position is more compact and he can try to target White's advanced central pawn structure) but I would guess that they are close to = although the position is quite a ways from being a dead draw.

       - Geof

I think that the variation after 6.Nd2 cxd4 7.Nb3 Qb6 (7...Qd8 is also very interesting, but maybe slightly better for White also) 8.Qxd4 Nc6 9.Qxb6 axb6 10.Nd4! as suggested by Wells and also covered by Davies is somewhat better for White if he knows what he is doing.  In particular, I seem to win a lot of blitz games in the 10...e5!? 11.Nxc6 exf4 12.Nd4 lines.  Basically, White must be careful for the next 5-10 moves and then he gets a favorable endgame where he can torture his opponent forever.  If you like grinding your opponents down in a += endgame (especially opponents that are hoping for more double-edged positions, hence 2...Ne4 and 3...c5) this line is not a bad choice at all.   

I think it's pretty clear now that the 6.d5 Qb6 7.Bc1 (7.b3 is interesting but I am not convinced it is any good) e6 8.c4 exd5 9.cxd5 c4! 10.e3 Bc5 lines are very problematic for White, so I wouldn't recommend it.  Although White has some tries at earlier deviations, none are convincing.

I don't think there is much hope for a White advantage in the 4.d5 e6 5.f3 Bd6!? lines, either.

So if you play the Tromp with White and you don't like the 6.Nd2 endgame, you should consider 3.Bh4!?, which leads to a number of sharp possibilities for both players.  GM Davies makes a pretty good case that this move has been underestimated, IMO.

      - Geof

Perhaps I was a bit fast in my judgement  . After 8.Qxd4 Nc6 9.Qxb6 axb6 10.Be3 e5!? 11.Bxb6 d5 seems to offer black enough compensation. Very interesting stuff!!!

1) I have tried 10.e4 Bc5 11.Nh3 d6 12.Bxc4 Bxh3 13.gxh3 in an email game. Black postponed castling with Nbd7 14.Nc3 Rc8. As soon as the game is finished, I will post it. All I can say now is, that I have good drawing chances.
12.Nd2 was played in Lerner,D-Rajlich,V/Stratton Mountain 1999. I feared c3! 13.bxc3 Bxh3 14.Nc4 Bf2+ 15.Kd2 Qc7 16.gxh3 o-o.
2) I am not sure if White gets anything after 12.Nd2 b5 13.a3 Qa5 14.b4 Bxb4 15.Rb1 bxc4 16.Rxb4 Qxd5.