Battery percentage from cubic equation?

[mrmabs]

I find the battery percentage to be still not a good predictor of capacity, and I had a thought that it might be possible to get a slightly better battery percentage from a cubic equation. I have had a dig through the source and see that a lookup table is currently used in power.h.

This site gives a live example and explanation of how to solve for x/y pairs: https://www.omnicalculator.com/statistics/cubic-regression

I worked with the input of (x,y) 100,4.2; 80,3.8; 20,3.6; 0,3.2. It should be possible to do this with other chemistries as well, as most follow a similar curve. There could also be a config file option for the user to specify points, four seem to make sense in most cases.

I think the only concern would be the cost of cpu cycles, having to find the best fit formula on boot, and then calculating a cubic equation with floating point numbers to get the battery percent.