shr0pshire: There is really one move in this game that concerns me and it is 15. ...Nb8!.
This isn't the first time this move has been played, even in the chessgames database it has been played a few times before. Although now that it has been played a few times in games between grandmasters it's results are pretty staggering.
First, I would like to note that this moves strength surprises me not because it seems to force white to trade down material into a slightly worse endgame, but because of the passivity of this move. It is returning the knight to the same square it occupied before black's turn on move 2.
To contrast this with other moves that have been played on move 15. When faced with this same position in 2004, Topalov chose a more active route of castling. Vallejo-Pons vs Topalov, 2004
In this same position Wang Yue against Almasi also chose to castle on move 15. Z Almasi vs Wang Yue, 2006
There are other games with Karjakin and other grandmasters who have played this same position, and until recently the theory was to choose something more active like castling. However, white won slightly more games after the castle, even on the grandmaster level.
Later today, I want to look at what makes 15. ...Nb8 so much better than the more active options.
With the three games with 15. ...Nb8 in this database the record is 0-1-2. Which is not too shabby for black, especially since all the players involved were rated higher than 2500 at the time.