Top N-precentile

[Paxy]

Can I suggest you implement an N-percentile average feature?
It will be used to get more accurate average reading from not so accurate sensors like PM sensors.
The way to implement is just simple to sort the data range by value and drop some percent of top/bottom values before calculating the average.

Configurable variables should be:

• N percent
• Drop from: top, bottom, both(N/2 each)

[Limych]

Such calculations require a double pass of the algorithm over all source data. And this can very negatively affect the work of HA. I will try to come up with something, but I doubt that it will work out.

[Paxy]

Perhaps it should be available only for a limited array of values to avoid a huge impact on HA performances.
For ex. the option is only available if sample_limit is less than 100.