Tricky Ending

FATHER Sebastian Kneipp did not just promote water cures - he advocated abstinence from meat and alcohol in any but small quantities.
He thought clothing should be self-spun and made from linen or hemp, rather than wool, and he believed breathing hot air was a major cause of disease.


Father Kneipp's statue dominates the town square

Kneipp was widely praised in his lifetime, although his bulky body-shape in the statue above suggests he may not have always followed his own tenets.
Then again, I am a great advocate of taking your time over a game of chess, but yesterday in round two of Bad Wörishofen U2000 I was well ahead on the clock as usual.
Nevertheless I was still doing fine when my game reached the following position:

I was offered a draw as I was thinking about my 25th move in Thomas Bergmann (1796) - Spanton (1890)

25....Qxd2
Considering the rating difference, I was not about to take a draw if I could help it, but anyway I felt Black's endgame prospects were slightly preferable.
26.Nxd2 f5 27.f3 Kf7 28.Kf2 Ke6 29.Nb3 Bf8!?
White threatened to win a pawn by 30.Nc5. This could have been stopped with 29...Kd5, as then 30,Nc5?? loses to 30...f4. I played  the text in the hope of inducing …
30.Nc5+ Bxc5 31.Bxc5
The analysis engine Komodo9 agrees with 30.Nc5+, and I presume TB played it because a bishop is usually significantly better than a knight when players have rival pawn majorities. However, I felt Black's better king was a key factor here, and making it an endgame of different minor pieces would increase the chances of the game not ending in a draw.
31...Kd5 32.Bb6 e4 33.Bc7 Ne5 34.Bxe5?
Swopping bishop for knight is not necessarily bad, but doing it this way is wrong, as will be seen.
34...Kxe5 35.fxe4 Kxe4
If White had played 34.fxe4+ Kxe4 35.Bxe5, then after 35...Kxe5 we would get the same position, with the same player to move, except Black's king would be on e5 instead of e4. That seems to make the difference between a draw and a Black win.
36.Ke2 f4?
I thought it was important to fix White's kingside pawns, but Komodo9 and Stockfish10 much prefer 36...g5, which keeps open the option of Black's king attacking White's kingside pawns via f4.
37.b4?
This loses by a tempo, but a tempo in the endgame is often vital. The engines give 37.a4, which Stockfish10 reckons equalises, whereas Knmodo9 at first believes Black has the upper hand but, if given plenty of time, agrees the position is drawn.
37...b5 38.Kd2 g5 39.c4
This was TB's idea, but, as noted above, it fails by a tempo.
39...bxc4 40.Kc3 Kd5 41.a4 h5 42.b5 axb5 43.axb5 g4 44.hxg4 hxg4
44...h4? only draws.
45.b6 Kc6 46.b7 Kxb7 47.Kxc4 f3 48.gxf3 g3 0-1