Is this a drawn or a won game?

It resulted in a draw. I guess I should have kept the rooks or moved the pawns another way. But how? Whatever I do stockfish tells me it's still winning for black. Even in the end. I could not see any breakthrough. I am confused. Is this a drawn position or a winning position for black after the rook exchange? What would you have played? Can this be won? Advice from better players is much appreciated!

Doesn’t black king just march up to f2 or e1 or even better to e3 and eat everything?

...b4 ...Bb5 will add pressure on d3 which will fall (black king on e3) and that’s it.

Black can always tempo with Bishops if he needs to penetrate or make white move to lose defense of some pawn.

I just had a brief look, but from the position you gave, I think that the key mistake, at least on a conceptual level is the move 33...g5. I think that it stops Black's most obvious attempt at a winning plan which is to walk the King in through f6-g5-f4/h4 etc. This simple plan looks winning to me, as it will stretch White's forces quite thin, and he will probably eventually be forced into a zugzwang or any other form of concession and Black should convert easily.
Also, I think that in this particular case it's fine objectively, but again, on a conceptual level it doesn't feel obvious to immediately push the queenside pawns all the way until they are fixed. Usually the pressing side (side playing for the win) wants to keep the pawns flexible in order to allow for more potential pawn breaks. This can also be useful on a human level, since it forces your opponent to consider all of those pawn breaks in every single position. And that is a problem that fixing the pawns solves for White immediately...
And, to answer your other question, I think that trading the rooks is fine, I probably would have done the same, rather than leave White with a rook on the 7th, and yes, I do believe quite strongly that the endgame after the rook trade is winning for Black. If he applies the correct approach that is.
I hope I helped. Cheers!

@JARANDujo thank you that helped me a lot! I completely blocked myself. I see that g5 was a huge mistake. Do you also have any advice on how to recognize such a situation and be aware of that it's very important now to not block the way for the king? In many cases it's irrelevant. Usually I don't need a plan in the endgame, but I need one in such positions. I should have noticed somewhere at move 25 that this position is very closed and it might end in a draw. I know that bishops of opposite colors can lead to a draw, but as I had both I never thought of a draw. I was really just pushing wood there, not realising that I eliminated the chance of a win.

Also I don't really understand why stockfish calculates -4 on move 48. Isn't that a draw? At this point my king cannot get on the other side, even if I try to sacrifice a bishop. I really don't get it. Thanks for your answer, it already helped me.

Ba8 Ba6 would have been interesting to try, with the idea of xc4 and then making a pawn break.

It may not be winnable anymore after you play g5. I only made a quick glance at it, but your king needed a way to penetrate, and you cut him out of the game.

if white dont take in a4 how do you progress? (e.g. Bd1-c2 ad eternium)

anyway, g5 is a big mistake... You need plan a path to victory when you reach that kind of endgame. The firs thing you think is about theorical draws like 'wrong rook pawn' in bishop endgame, slatemates, etc.. You need look for main routes and the main of all: TIME.


Stockfish's evaluation at move 48 simply is not correct. Classical engines still fall victim to the horizon effect: You are one bishop up, and you will still be at the end of all search paths (of useless bishop moves by both sides), so that material advantage will stick in Stockfish's mind (and +1 due to general positional advantage that you keep). It cannot see beyond its search depth and realize that nothing substantially will happen no matter how long the game goes on (and there are too many bishop moves to run into threefold repetition in all paths). At depth 50, evaluation should eventually flip to 0.0 due to the 50-moves-rule.

How to recognize such a situation?
After the rooks are traded it's more or less impossible to win by a direct mate with your remaining pieces (your light squared bishop is too restricted). Thus the main win condition (for both sides) is to promote a pawn and mate with the new queen. Your opponent still has 7 pawns, so to promote, you need to clear the way, either by arranging a breakthrough that results in a passed pawn for you or by penetrating with your king and eating his pawns from behind.
Once you have realized this, you need to look for weak spots where one of these two options could work. Since White has absolutely no control over the dark squares at all (and you do with your dark squared bishop), penetration via g5->h4/f4 should naturally be at the top of your list. And subsequently, you see that it makes no sense to block off your own way by moving a pawn there.