Gutsy endgame plan pays off.

I was browsing GM games a while ago when I came across this.  I had been meaning to try out the Grunfeld for a while, but had never studied or practiced it.   Still, I felt inspired to give it a try the next time someone opened up with d4.   Here's the resulting game.

I'm not convinced black had a sure draw there.  If white defended the pawn with the bishop, there would be no way for black to take it out, as moving the king to the other side and trading the bishop would result in a kingside pawn being promoted.  Annotating quickly I came up with 35. Bb6 Ke6 36. a4 Kd5 37. a5 Kc6 38. h3 Kb5 39. Kg2 Bc3 40. Kf3 Bxa5 41. Bxa5 Kxa5 42. Kf4 Kb4 43. Kg5 Kc3 44. Kh6 Kd2 45. Kxh7 46. Ke2 Kxg6 47. Kxf2 h4 1-0, the king is outside the square of the h pawn.

Edit: Forgive me if trading off bishops is pointless, I'm uneducated in the ways of the bishop endgame, but I know a won pawn endgame when I see one.

On a related note engines are much worse then humans at playing and analysing many endgames, this is one of them.

Oh there's no question I should have played 34... Bc7 before bringing my king over to the a pawn.  I recognized my mistake shortly after I submitted the move.  Luckily my opponent didn't, and I was able to correct my error on move 35.  I meant to include that in my annotation, but I seem to have forgotten. :p.