I think the current criterion for deciding whether a track has been skipped is whether more than the first 5 seconds are played, the same as for deciding which track Previous jumps to.

I think a better criterion would be whether ¼ of the song has been played - this is a better metric for songs whose intros are so quiet it takes more than 5 seconds to even realise a new song has started, never mind that it's one you're not in the mood for.

Similarly, ¾ would be a good criterion for updating the "played" count for the song.