Today's roasting lends further weight to the notion that my algorithm works better. Each batch ended up within 0.1-1.0 of expected vs. the old algorithm showing error in the range of 2-6.