Thank you, this exactly! Seems like you want to reduce the standard deviation only considering pieces that are larger than the mean, but still relative to the mean. Would be very curious to see the results redone using that approach.