I am not happy with the algorithm at all. We should replace it with some more sophisticated subplugin, using ICC or some similar statistics.
Also fixes small bugs in workshop api constructor and in using "scales" instead of "scale" for the core table name.