Created listed of all artists I have in my spotify playlists. Then printed all the "similar artists" for all of them, ignoring the ones I already have in playlist. Then sorted and printed the ones with biggest count = the artists that are most similar to the ones I am having in my playlists.
The result was very interesting. Especially comparing when I asked the same (similar artists) from the last.fm.
Unfortunately the code initiated extremely huge number of API calls, so definitely not ready for public usage.
I would be very interested if anyone tried to dig out something from the WEB api data in this manner.