Standard deviation is a poor measure because you care more about avoiding big pieces than small ones. Penalizing for having a few tiny pieces doesn’t make sense.

You probably don’t even care about the “standard” deviation at all. You care about the deviation from some desired size. Probably the more accurate problem is “what is the fewest number of straight cuts I can make such that all pieces are below some target size”.

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.

A few tiny bits will just melt away or brown faster, no big deal. But one big undercooked chunk in a bite of otherwise soft onions? That's noticeable