Help with middlegame pawns/bishop endgame

Hello. I'd like some help looking at this recent game. I did some self-analysis, but didn't use any engines (meeeep helped a lot though). 15|10 Live Standard, I play white in a classical e4 e5. It gets weird, and I have 2 main questions.

1) What should I have done with the pawns in the middlegame, around moves 16-20?

2) Can someone help evaluate this mess of an endgame (starting from about move 30+)? Who wins?

Thanks in advance.

I didn't really look too closely at the endgame.  However at first glance it seemed black was much better.  In same color bishop endgames you want to put all of your pawns on the opposite color square of your bishop.  All your pawns were fixed on light squares while his were fixed on dark, making your pawns a target for black's bishop.  What made it interesting though is you had the a pawn so white may have been winning.  Seems like alot to calculate.

I think less pawn moves in the middlegame would have resulted in a more manageable endgame.

@mattyf9: It is a lot to calculate. As you can see by the sidelines. Which is why I'm asking for some concrete help here.

@VULPES_VULPES: Specifically how? That's what I want to know, since around move 20 I botch the queenside structure a little.

Maybe I should refine what I'm asking. I'm interested in specific moves and variations here, not general principles. (1) Instead of 15. Rhc1, what can I do? (2) Instead of 17. Rc2, what can I do? (3) After 31...Kg7, I think white is either winning or drawing. Can I actually win, since my analysis shows only a draw? Did I miss a black win?

Perhaps you could've tried 17. axb5, after which black will have to take with 17. ... axb5 if he doesn't want to lose his a-pawn. After that, you can do 18. Ra7 Ra8 19. Ra1 Rxa7 20. Rxa7 with an eventual Nb1 or Nf1 to redeploy the knight and c4 - after which if b4 is played, c5! Will fix the black c pawn on c6 - will result in more active pieces and the a file to yourself.

I think my analysis works. I did it all in my head so pardon me if things don't sound right.

A very complicated game. My rough evaluation of the endgame is that black must be better, so can white draw ?

I thought that maybe the incisive 32...Bh3 instead of 32...Rxd2+ might give black the win, as 33.Rxd4 cxd4 34.Kf1 d3 35.Bb7 Bxg2+ crashes through and 34.Bb7 Bxg2 35.c6 Bh1 looks winning too.

So maybe white should refrain from Rd2, but then Black goes Bc8 and slowly improves his position. Tough to defend.

@VULPES_VULPES: It looks good. Thanks for the idea, since I saw 17. axb5 axb5 and hadn't the faintest idea what to do after that. Although an immediate 18. Ra7 b4 19. c4 c5 might be an issue, it looks OK-ish.

@hicetnunc: Thanks! 32...Bh3 does seem to win. The most critical line (I think) goes 32...Bh3 33. Rxd4 cxd4 34. Bb5!? Bxg2 35. c5!? axb5 36. c6 d3+ 37. Kxd3 Bxf3 38. c7 g2 39. c8=Q g1=Q and white should lose.

Which seems to suggest 31...Kg7 was not a total waste of tempo, and perhaps 32. Bb5!? or 32. Rd1 was the right move. I'm still not totally convinced that 31. Rd2 lost, but you've given more food for thought.

32... Bh3 33. Bb5! (for effect, or Bb7 does the same) Bxg2

34. Bxa6 Bh3 (Rxd2 Kxd2)

35. Ra2 g2

36. Ra1.

Find winning plan for black?

@meep : in your line Black has 34...Bxf3+ followed by Rxd2+ I think

Oh that's true; totally missed that. Seems winning.

Sorry for the bump, but there's a lot of unsaid ideas.

Black can win in the final position with no bishops (pick off a and h pawns), thus, he has the plan Bf7-e6-d7. White's only refuge is a6. This gives Black time to prepare an f5 break, activate his bishop, then attack the queenside pawns.

You could have untangled your pieces with Rhd1, Ne1! Ke2! Ne3.

Note that your bishop is white, g3-f4 allows you to create a favorable pawn structure if he tkes, or at least constrict his kingside pawns.

This is off the top of my head, so there's probably some degree of error.