Hello James, firstly, great work! The thought process is very well elaborated on and crystal clear. I’m probably incorrect here, but it seems like the tools used to arrive at protocol seems to ignore conventional established statistical tools. For example, trimmed/weighted mean has shown to outperform medians in many peer reviewed articles parametric and non-parametric (less assumptions) sample sets. Why was the decision made to use median? The buckets you mentioned seems like a classical histogram, why not use density distribution to give a more smooth approximation rather than a staggered groups, in addition, how were the bucket thresholds arrived at? Have you conducted any hypothesis testing for performance of the algorithm, I didn’t see and p/q values. Keep up the good work!