Another reason to roll average the readings is that the scheme where you want to just look for 5 readings the same has a bit of flaw in that the readings on any sampled measurement system can be off as much as +/- 1. If you look for 5 counts the same do plan to put a tolerance on the acceptable value values you are comparinng.
even simpler, look for 5 consequtive readings above the set limit, then to reset look for 5 consequtive readings below the set limit. This would even add a bit of hysteresis which, I'm sure, will be needed

Erik