No, I saw those. I just chose to ignore them :) ... In particular, I did some additional tests using randomization with the traditional bin packing algorithms and found that the less-packed bins were typically filled to only 760 or so. Additionally, since sum(@a)/31 = 851.7, strictly enforcing the 840-900 limits is difficult/unlikely. Thus, I took the 840-900 limits to really mean "as balanced as possible", hence my solution.