There is a document on how epoc is calculated at https://assets.firstbeat.com/firstbeat/uploads/2015/10/white_paper_epoc.pdf
There is an Ambit app that uses stryd power reading to calculate your energy efficiency similar to the running performance on ambit which was far more accurate than relying on gps data.
http://www.movescount.com/apps/app10645993-Average_energy_efficiency
Notice the level of filtering on Ambits Running Performance graph relying on gps vs the energy efficiency app relying on power data.
By the time you notice changes in Running performance graph to figure out whether you have proper hydration or enough carbs; energy efficiency app would react much sooner as it gets its data from accelerometers vs. gps and barometer on watch used by running performance. Your body makes subtle changes to adapt to current situation, it does not think about finishing line, thats what brain is for (or not). The sooner you notice the sooner you can react because physiological response to recover gets Progressively slower the father you fall in anaerobic energy debt.
Constantly changing wind speeds would render ascent descent baro sensor data invalid even on a flat surface no matter ambit or s9.
Epoc modeling has a delay, see white paper, basically loose ohr or belt ruins it.
Real time running economy would react sooner than any other method. Your model is based on power, sensor will not become loose or fall off (less likely than belt or ohr)
Or better yet, Having a small “sandbox“ for custom equations and graphs sent to SA from s9 like we did on Ambit To MC would stop avalanche of nagging user requests to add features. It would also let us innovate without consuming other peoples time on development. Few 32 bit floating data as temporary registers to implement equations is not much of a footprint these days. Add a few kb of data to support at 10s logging interval. It would just be much more fun even if it fed you meaningless numbers.
<nudge> <nudge>