Many endgames are tough but pawn endgames can be very deceptive. Here is an OTB game that I played a few weeks ago where I had many winning chances. This endgame was reached in the last round, late at night, after 10-12 hours of chess play. For many reasons, I had a mental block causing me not to see what should have been a straightforward win.
When we reached this position, I stopped and spent at least 5 minutes trying to work out how to triangulate in order to advance the c pawn to a queen.
Of course, the simplest win is to just put black into stalemate and force him to advance his a pawn, where white then advances his! I only seriously looked at the line where white plays bxa5, which of course ends the game in stalemate.
I had lost a won endgame a few rounds earlier in the tournament and was a bit gun shy. The thought of black being allowed to freely advance his own a pawn down the file unopposed made me stop even considering that line, even though it is obvious that I checkmate first with my own pawn advance! (By the way, this win was pointed out to us immediately by the TD Sevan Muridian, after the draw was agreed.)
Ok, but there is another win! In fact, my intution was correct and there is a triangulation win. Try to find it:
As you'll see in the annotations, black's king is prevented from moving to c7 (or else the white king gets to b6). Because of this restriction and the fact the black king is on the last rank and cannot step backwards, triangulation works to "lose a move" and promote the c pawn. (This win was shown by FM Kevin Bachler.)