Sometimes I take a look at the db levels of the music just prior to the crowd coming in. Find the db value, then hard limit to a point just before the music level db (highlight that area and get the statistics).
Sometimes I do this on both channels separately, sometimes I do 1 real hard, and another less so. When it's less so, it's because I would like to keep a bit of the clapping (not the annoying loud stuff) just for ambience and a natural feel.
Dave