Statistics from databases can be very misleading.  They do not take into account advances in theory, for one thing.

The Tarrasch 3 Nd2 was extremely popular for many years when White players were shying away from the Winawer's complications.  During that time, top players played it and ran up a great score. 

The Nadjorf Poisoned Pawn is a great example of how stats get skewed.  For many years it was the most  popular opening in master competition, new moves and new ideas were coming out all the time.  A new move is introduced, scores some good wins, gets very popular very fast.  Then, someone finds the answer - it doesn't have to "refute" the line, just change its evaluation to worse than others, suddenly everyone drops it like it's hot.

But the stats will show that sideline with a big positive score because it ran up that score before the answer was discovered, and then only a few played it before it fell out of favor.  But if you go by statistics alone, you will think it is a good line still, even if refuted 30 years ago.

I seem to recall that in another thread, you claimed that a move was good because Karpov played it. Therefore, because Nd2 was played by Karpov, it must be good.

Seems like this argument is more sound than one based off databases 

Oh, I just double checked and it was Kramnik.

New argument:

1. If Karpov plays Nd2 and Karpov is as good or better than Kramnik, then Nd2 is a good move.

2. While I cannot play like Karpov, neither can my opponents

And maybe 2. h3 is a good move!?

because Karpov lost against a6 by miles

I agree with pfren but I think his arguments are out of context for the overall level of the thread.  Yes, the Tarrasch offers white almost nothing against a well prepared black opponent, and Nc3 is the objectively best way to pursue an advantage in the position.  With that being said, players below a certain level are better off in the simpler positions arising from the Tarrasch instead of entering the minefields of the Winawer or the MacCutcheon.  

In other words, yes, the best way to chase advantage is Nc3, but generally speaking if the players involved are of a high enough caliber for it to matter, they are already aware of this fact.  It doesn't make an idiot any smarter to call them an idiot.

Of course the best way to pursue an advantage is Nc3, but that doesn't mean that Nd2 isn't a good move or that Nd2 allows black to equalize.

I agree, the fact that the best way to pursue an advantage is Nc3 doesn't mean that Nd2 allows black to equalize.  The lines and the theory show that Nd2 allows black to equalize rather than the fact that Nc3 is better.

to be honest i dont understand enough about chess but winawer seems to be much more natural and easier to play then the tarrasch. Maybe im just too bad to realize how i misplay it, but seriously the winawer seems to me the most natural to play, if i play the tarrasch and the opponent plays differently....

IM pfren says that c5 and Qxd5 equalizes, but I feel like white has a slight edge after this, although I will agree he doesn't have that much:

Of course, and that's why I only looked at recent games (2012) and why I even emphasized the word currently in my post. I looked at only 2012 games between 2500+ players, and Nd2 scored a trivial amount better for White. If there have been major changes in theory in the last year, you would have a point.

I never argued "in favor of 3.Nd2". That's your misreading of what I wrote. I only asked why Nc3 doesn't score better than Nd2 at the highest levels currently if it's objectively better.

To be more specific, I previously looked at 2012 games between 2500+ players (and found Nc3 59.5% [n=205] versus Nd2 59.7% [n=129]).

You mentioned 2700+ players. There aren't many games, but for 2012 games between only 2700+ players the numbers are again comparable: 16 games for Nc3 (White=62.5%) and 6 games for Nd2 (White=66.7%). I'm sure you're right that Nc3 is objectively stronger, but it doesn't seem to be enough stronger in practice at the highest levels to actually win more games, which would generally be expected if one move was much stronger than the other.

nc6 is a weird move for a french player to make the c-pawn attack is a good resource in the french

I checked chessbase just for games in 2010 and 2011 with both players >2400, and there white scores 57.2% with Nc3 and 55.7% with Nd2. However when I dug a little deeper into the critical lines, by move 8 in the Tarrasch white was scoring 45% in the Be7 lines and only about 50% in the 3...c5 lines. White maintained a solid 60% or so in the main lines of 3.Nc3 through at least move 10. 

nc3 allows the winawer which is dangerous at low levels so many players don't want to go into the complexities of it ecen if nc3 is objectivley better

Using database statistics is a fool's errand.  There are far too many factors involved. 

If they were indeed dispositive of anything, the highest scoring lines in the database would be the lines used in current practice, wouldn't they?  Why not?

It's fine to check stats as a very general guide, but I like the way Chessbase presents them (press "S" after highlighting the move in the opening key):  not just the # of games, wins for White & Black & draws, but also the average rating and performance rating for both sides. 

The most informative number (assuming there are an adequate number of games in the sample) is the difference between the average rating and the performance rating for each side.  If one side consistently outperforms the rating, the line is more likely good for that side.  This is not exact, either - no statistics tell the whole story - but it is a better guide than the won/loss percentage alone.

i think the number of draws compared to the wins and loses could also be an indicator though