theodosius: Stockfish 5 analysis.
White maintains opening advantage until 13. Rg1 loses about .6 for white (O-O was better).
16...Bf6? lost .25 (Nec7)
18. Bg5 lost it back (Bxg7)
20...Kh8? lost .75 (Nec7 again)
21. Bd3 gave .5 back (h4)
21...Nf6 lost .8 (to 1.5, Qe7)
23. Rb1 loses .5 (f3) but played it a move later gaining it back
Throughout the 20s after this, SF wants White to play h4, and black to play Bf6/Nc7. The analysis swings widely as these giants ignore these developing moves.
SF then wants white to move Qd7. Adding that to 28...Qe7, it seems that both players wasted a move. White is rated +1.58 with this continuation.
28. h4 h6 29. Kf2 Kh7 30. h5 b6 31. Kg2 Ne7 32. Rh1 g5 33. b4 Kg8 34. Nf5 Nxf5 35. exf5 cxb4
With 32. Qg2, white loses .75 and has descended into draw territory (Qd2, retains about a pawn advantage).
33...Bf6 undevelops his bishop for no reason, when SF wants Nc7. Black moves 34...Nc7 after 20 moves of SF's pleading!
35. Qf1 is a dead draw. At this point, SF wants b4 and Kg2 for white and Bg5 for black for several moves.
39...Bh4 is another major blunder, losing a theoretical pawn (Bxe3). The next move, black undevelops his bishop again to Bf6. At this point White is +2.1 with a powerful kingside continuation:
41. g5 Bg7 42. h4 f6 43. h5 gxh5 44. Nxh5 fxg5 45. Bxg5 Qf7 46. Nxg7 Qxg7 47. Qh2 Qf7 48. Be2 Nc7 49. b4 b6 50. Rh1 Rf8 51. bxc5 bxc5
But why would white want to do that? 41. Rb2?
Throughout the 40s, it's clear that black's bishop should have stayed at h4 instead of waltzing around. 44. Na2 lost a "grim" .7 (Kg2), but black's bishop wastes time. SF would have white develop Kg2 and pawn break f4.
50. Nf5 moves white back to draw territory (Nc3).
Then 51. Bf2? (a clear blunder per SF) gives black his first fleeting chance at a win -.5 (Nc3), which is squandered by ...Bg5 (gxf5).
53...Be7 (again undeveloping his dark bishop) puts white a pawn up (Bf4).
Through the rest of the 50s, SF wants to initially expand on the kingside with h5 and a rook on the h file, but after redirecting a rook to the b file continues its desire to get it's rim knight to c3.
Black continues to waste moves on the dark-square bishop, instead of moving knight to f6 as SF would like.
Seesaw continues through the 60s. SF has been demanding Nc3 nearly ever one of the last 25 moves.
Black, on the other hand, had multiple development options, Ne7, h5, cxb4, which it throws away for futile bishop moves. It appears knights are more valuable to SF in these positions.
73. Nc3. After 30 moves.
78. Ba4 was a -1 point error (Kg1), but ...Bc8 was worse (cxb4) as white's 79. Bc6 could force a white passed d pawn, giving white a 1.5 advantage in SF. White declined 80. Bxb7, returning to business as usual.
90...Na6 is a bad idea for black, losing .7 (Nf6), and again white must make a combative g5 to capitalize, but instead undevelops it's knight to a2.
93...Ne7 blocks the correct Be7, giving a +2 to white. SF would like white to respond with f4, which appears to be one of white's key moves in controlling the board.
<AylerKupp>'s megapost gloriously expounds on the tragedy of white's first capture, and SF concurs with his assessment.
Trading of the rooks began white's descent to a draw, SF attempted to avoid this with Nb3 or Ne2. 99. Bc2 would have robbed white's minor advantage if ...Bxc2.
SF doesn't say 104. Bc6? is a blunder, though what else would you call ignoring a free pawn? The pawn stays in play for white until the draw.
Many here have speculated on possible continuations (<David2009> is right on) after 106...Qa5, so I will not.
In general, both sides ignored attacking chances, white on the kingside and black on the queenside.
I hope you enjoyed reading this as much as I have enjoyed writing and analyzing it.