However, after, say
28...Qd2
29 Nf6 NxN
30 QxN

it looks pretty hopeless for black to me.....what can black do after QxN?

Course there's 28..Qd8...

[...] So had Carlsen found 28.Nf5-g3!! and the winning ideas associated with it, he would already have an Immortal Game.>

Maybe you're right - it seems to lead to a position where White retains a pawn advantage in the endgame, in a better position than the one resulting from the 32.g6 line analyzed above - 33...Bh6 (33...Rc2? 34.Rg3 Bh6 35.g7+ Kg8 36.Nf4!) 34.Bh3 (34.gxf7 Re7 35.Nf6 Nxf6 36.Rxf6 Rc1) 34...Qg5 and apparently it boils down to 35.Bxe6 Rxe6 36.Qxg5 Bxg5 37.Rxf7+ Kg8 [White's last moves can be played in some different orders, but I don't see that it makes a difference] 38.Rxd7 Rxg6 39.Rxb7:

click for larger view

<28.Ng3!! Bg7 29.Nxh5! gxh5> and now <30.Qxh5!> (rather than 30.Nf6+ as in the Marin-Torrecillas analysis published on chessbase at the time) threatening, among other things, 31.Nf6+ and 31.g6 (e.g. 31...f6 32.Qh7+ Kf8 33.Nxf6! Nxf6 34.Rxf6+ Bxf6 35.Rf3 Qd8 36.g7+ Ke7 37.g8Q+; g7+ also wins after 35...Qe1+ 36.Bf1 Qxe4) <30...Qd2(!)> (again, the defence that puts up the most resistance; 30...Qd8 31.Rhf3 Re7 [a move which isn’t playable with the queen on d2] 32.Nf6+! Kf8 33.Qh7 Re8 34.Bh3!; or 31...Bxd5 32.exd5 f6 [32...Re7 33.d6] 33.gxf6 Nxf6 34.Rxf6! Bxf6 35.Qg6+ Bg7 36.Rf7) <31.Rhf3!> (now 31.Nf6+ fails to 31...Nxf6 32.gxf6 Bxh3 and there’s no 33.Qg5; or 31.g6 f6 32.Qh7+ Kf8 33.Nxf6 Nxf6 34.Rxf6+ [34.Rhf3 Qh6 35.Rxf6+ Ke7] 34...Bxf6 35.Rf3 Qd1+ 36.Rf1 Qxf1+ and Black gets too much material for the queen) <31...Bxd5> (that’s the most Black can do – he can’t hold on to f7 by 31...Rf8 because of 32.Ne7#) <32.exd5 Re7> (note that on d2 the black queen prevents 33.d6 here) <33.Rxf7 Rxf7 34.Qxf7+ Kh8> (34...Kh7 35.Be4+ and mate) <35.Qxd7 Qxg5 36.d6!>

click for larger view

with winning initiative for White (besides picking up the b7 pawn, there are ideas of Qc7, Qe7, or Rg1)

An interesting sideline is 32...f6 (instead of Re7) 33.Rg3! (33.gxf6 Nxf6 34.Rxf6 Bxf6 35.Qg6+ Bg7 36.Rf7 Qh6) 33…f5 (33...Rf8 34.Rd1! Qc2 35.d6! [preparing Bd5+] 35...Qh7 [35...Nb6 36.gxf6 Rxf6 37.d7 Rd8 38.Qe8+ Rf8 39.Qe6+ Rf7 40.Rf1 and Black will be mated shortly] 36.Bd5+ Kh8 37.Qxh7+ Kxh7 38.g6+ Kh8 39.Rh3+ Bh6 40.Rxh6+ Kg7 41.Rh7+ Kxg6 42.Rxd7) 34.Rh3! (now that there’s no Qxg5) 34...Rf8 (34...Nf8 35.Rxf5 Re7 36.Rf1! [preparing Be4] Qd4 37.d6! Rd7 38.Rd1 followed by 39.Bd5+ if the black queen retreats) 35.Qh7+ Kf7 36.Rh6!; or 32...f5 33.Rh3 reaching a move earlier the same position as in the line above.

And after 32...f6 33.Rg3 Rf8 34.Rd1, if Black tries 34...Qb4 so as to meet 35.d6 Nb6 36.gxf6 Rxf6 37.d7 Rd8 38.Qe8+ with 38...Qf8, White has 37.Rg1! (threatening Bd5+) 37...Qd4 (37...Rf7 38.Qxe5 Qb5 39.Qe6 Qf5 40.Rxg7+ Kxg7 41.Be4+) 38.Bf3 Rf7 39.Rh3 and the only way to prevent an immediate mate is 39...Qxg1+.

also, i think that currently there are many elite gms that play this kind of position as well (and imperfect) as carlsen; he might not be worse than the majority of his peers at it, but the super-tactical nature of this position doesn't play into carlsen's (unique) strengths the way other, more strategical ones do.

i'm only speculating now, but experiences like this game might have led carlsen to leave his initial approach of going all out to inflict losses on anand (and others), instead aiming for game plans that exploit his special skills better while involving less risk.

so, in some sense games like this may have helped carlsen to identify his own strengths as a player. the initial failed attempts at "running over" anand the way he had run over lesser opponents in his pre-2700 days might be seen as part of carlsen's elite education - and hence useful. that some antagonists still try to use carlsen's early losses to anand as evidence of the latter's superiority over the former is mostly evidence of a flawed tradition of abusing life-time records as measure of individual strength relationships.

Judgment can be a tricky business indeed, and the flexibility to switch between different types of advantage is an important attribute for any player, which only comes with experience. When one has burnt up much clock time calculating all sorts of sharp lines (as here), to settle for 'only' a superior ending requires discipline which may be outside the realm of even top GMs in the heat of battle. We need only turn to the recent Wang-Carlsen battle for an object example where White failed to adjust to a change in the situation.

<....the super-tactical nature of this position doesn't play into carlsen's (unique) strengths the way other, more strategical ones do....>

On examining a number of Carlsen's games, his opponents generally come under great pressure throughout all phases. It's one thing to calculate one's way through a mess of tactics and something else again to fathom the more amorphous aspects of positional understanding.

<....some antagonists still try to use carlsen's early losses to anand as evidence of the latter's superiority over the former is mostly evidence of a flawed tradition of abusing life-time records as measure of individual strength relationships.>

certainly - but often the sheer time spent to realize that there's no directly deciding continuation in an attack will cripple the chances of finding/choosing the best non-decisive continuation.

anyway, to me it seems that carlsen's choice of later years - whether deliberate/conscious or not - has been to treat many situations & positions that *might* hold something concrete in a very pragmatic way. i.e. instead of searching for the position's perfect "solution", he instead moves relatively quickly, going for something he considers appealing without necessarily being absolutely best. sometimes this leads to "criticism" from people like me who would've liked him to "always" find the knock-out blow where those exist. ;o)

i guess it's hard to tame a strong urge to win. :o)

well, if one can say that about a game including 7. g4 Nxg4 8. Rg1 Qf6 9. Rxg4 Qxf3 ... ;o)