Exponential moving averages have some advantages over using simple windowed averages: no large changes as the window edges advance through outliers, and it's easy to tweak how sensitive they are to change through the decay parameter. Would there be an interest in iterating on this?