Best option is to use Statistical API. Probably this example is the simplest, just extend the timeRange to cover all four years, and add all relevant bands.
You will have to define the polygon (source data are in raster format so a concept of “point” is not well defined). To be as accurate as possible we recommend:
- work in UTM CRS (in the appropriate zone - i.e. 33N for Austria), to avoid any reprojection
- construct a polygon so that it aligns with the Sentinel-2 grid perfectly - i.e. if you have a point in Wien with E,N: 602003, 5340511, you can construct a BBOX by simply taking coordinated divisible by 10 (
"bbox": x602000, 5340510, 602010, 5340520]
if I did not do an error)
To get cloud-free points, do output also CLM and/or SCL band in the avalscript, so that you can filter these out later on.
As a side note - if you really have lots of points and you want to explore a bit more, you might want to give a try to our latest addition - Batch Statistical API. You will need AWS account (free one will probably be sufficient) and Enterprise-level Sentinel Hub account (we are happy to “promote” you for a month, so that you can try it out).