Knight v Bishop

I WAS on top board last night in round two of the open section of the Pimlico Summer congress, which is being held over five consecutive Thursdays.

Black has just played 34...Ke7-d6 and offered a draw in Spanton (171) - Gergely Kiss (207)

*****
*****
*****
*****
*****
*****
*****
Normally a bishop outplays a knight in an ending with rival pawn-majorities. But here White can stop Black activating his queenside majority, after which it is just a question of whether White can win or only draw with his kingside majority. Bearing in mind that Black's bishop is fairly bad - four out of six black pawns are on light squares - Black's practical chances of holding this ending are very low.
35.Kd4 Be2 36.Nc3 Bf3 37.a4 bxa4 38.Nxa4 Bd1 39.Nc5 Bf3 40.Nb7+
40.e5+ is probably a quicker way to win, but there is no harm in probing with the knight as Black cannot prevent White coming back to the e5+ idea.
40...Ke6 41.Nd8+ Kd7 42.Nf7 Ke6 43.Nd8+
I got cold feet about 43.Nh8 because of 43...f5, when 44.Nxg6 seems to be the only way to keep winning chances. White then has to be certain he is winning after both 44...fxe4?! and 44...Bxe4. I was not sure, but I was fairly confident that returning the knight to the queenside would win.
43...Kd7 44.Nb7 Ke6 45.Nc5+ Kd6 46.Nd3 Ke6 47.Ke3 Bg2?
This speeds defeat, as does 47...Bg4? 48.Nf2. Black had to play 47...Bd1 or 47...Bh1, but in each case White wins slowly but comfortably, eg 47...Bd1 48.Nc5+ Kd6 49.Na6 Ba4 50.Kd4 Bc2 51.e5+ fxe5 52.fxe5+ Kd7 (52...Ke6 53.Nxc7 Kf5 54.e6 Kf6 55.Kc5) 53.Ke3 Bf5 54.Kf4 Bc2 55.Nb8+.
48.f5+ gxf5
If 48...Kf7, then 49.Nf4.
49.Nf4+ 1-0