Quick answer
Quick answer: With only bishops left, one extra pawn is usually a draw, and two or even three can be too. The defender blockades on the squares their bishop controls and you can never attack those squares. Keep other pieces on and everything changes.
That last sentence is the part most players don't know, and it's worth more than the rest.
I read the endgame theory before writing this rather than trusting the usual rule of thumb. The details sit on the opposite colored bishops page, and the geometry behind it is the bishop itself.
Why these endgames draw so often
Your bishop covers one color. Theirs covers the other. Neither can ever attack a square the other defends.
So the defender picks the squares their bishop controls, puts everything there, and sits.
You can bring your king, push your pawns and improve your bishop as much as you like. None of it touches the squares that matter.
That deadlock is the whole story, and it's why material counting fails so badly here.
One extra pawn is almost never enough
This is the result that shocks players the first time.
A clean extra pawn in a normal bishop endgame is often winning. With opposite colored bishops it's usually a draw, and not a difficult one for the defender.
The pawn needs a square to promote on, and the defending bishop simply stands on that diagonal forever.
So before you trade into this structure, ask whether a pawn up is actually going to be worth anything.
Two pawns are often not enough either
Two extra pawns feels overwhelming and frequently isn't.
The defender builds a blockade on their bishop's color, puts their own pawns on that color to protect them, and holds. Three pawns can fail against a good fortress as well.
So the question is never how many pawns. It's whether the position lets you break the blockade.
Our note on holding a fortress covers what the defending side is actually building here.
Separated pawns beat connected ones
This is counterintuitive and it's the most useful thing on this page.
In most endgames you want your passed pawns side by side. Here you want them far apart, because a single bishop cannot cover two distant promotion squares at once.
The rule is sharp. One file between them is a draw. More than one and the attacker generally wins.
So if you get to choose which pawn to keep, choose the one that widens the gap.
When connected pawns do win
They aren't useless, they just need to get further.
Connected passed pawns that reach the sixth rank normally win, because the defender runs out of squares to stand on.
The exception is the wrong rook pawn, where the promotion square is a color your bishop can never touch. That one draws no matter what you do.
Check the promotion square color before you commit your whole plan to a rook pawn.
The defender's method in three steps
If you're the one holding, this is simple enough to do under time pressure.
Put your pawns on the same color as your bishop, so your bishop defends them and their bishop can't touch them. Put your bishop on the diagonal that stops the passed pawn. Bring your king to help.
Then do nothing. Genuinely nothing. Shuffle and let them try.
Most losses from a held position come from the defender getting bored and moving something.
Keep the rooks on if you want to win
Here's the reversal that decides a lot of real games.
With rooks still on the board, the stronger side has far better winning chances, because a rook covers both colors and can sometimes be given up to smash a fortress.
So trading into a pure bishop ending is often the move that throws away a win.
Our guide to rook endgames covers the technique you'll need once you've kept them on.
Opposite bishops favor the attacker in the middlegame
The same structure that draws an endgame is dangerous earlier on.
With queens and rooks still around, your bishop attacks squares the defending bishop can never defend. That effectively gives you an extra piece on one color.
Attacks against a castled king are noticeably stronger with opposite bishops on the board.
So the same feature is a drawing tool in the endgame and a weapon before it. The difference is what else is still there.
How to decide before you trade
Ask three questions before allowing the last piece off.
Do I have two passed pawns with more than one file between them? Can my pawns reach the sixth rank together? Is there any way to break the blockade square?
If the answer to all three is no, keep a piece on and play for something else.
That one habit converts more won positions than any amount of endgame technique does.
Do not let the clock talk you into a trade
Under time pressure, trading pieces feels safe. In this structure it's the most dangerous thing you can do.
A position that was clearly winning becomes a dead draw in one move, and there's no way back from it.
So when you're low on time and a piece trade is offered, that's exactly the moment to spend ten seconds you feel you can't spare.
One question is enough. Can I still win this with only bishops left?
Zugzwang almost never helps you here
In most endgames, running your opponent out of useful moves wins games. Opposite bishops break that idea.
The defending bishop usually has a long diagonal with several squares on it, so it always has somewhere harmless to go.
That means you can't squeeze them the way you would in a king and pawn ending.
So stop looking for the waiting move and start looking for the pawn break, because a break is the only thing that changes the position.
The king matters less than usual
In most endgames, king activity is the first priority. Here it's demoted.
A very active king still can't step onto the squares the enemy bishop controls, and those are the ones the blockade sits on.
So an active king helps only if it can support a pawn on your own bishop's color.
Work out which color your winning attempt runs on, then bring the king to that color rather than just towards the center.
Pawns on the wrong color are the real problem
This is where the position is usually lost or saved, long before the endgame arrives.
If your pawns sit on the color of your own bishop, your bishop is blocked and your pawns are safe. If they sit on the other color, your bishop is free and your pawns are targets.
Which one you want depends entirely on your role in the position. Attackers want a free bishop. Defenders want safe pawns.
Our explainer on good bishops and bad bishops covers that trade off properly.
How to practice this
Set up a position two pawns up with opposite bishops and try to win it against a bot.
You'll fail the first several times, which is the lesson. Then move the pawns further apart and watch it become winnable.
That single experiment teaches the whole subject faster than reading about it.
You can set the position up and play it out on our board as many times as you like.
What to remember at the board
Three things, and they fit on one hand.
A pawn up means nothing here. Passed pawns want distance, not company. And the last piece you trade decides whether the game is a win or a draw.
I compared a set of my own games against these rules and the losses were almost all the same mistake, which was trading into a pure bishop ending while a pawn up.
Our guide on which endgames to study first puts this alongside the ones that come up more often.