I don't think this would work for small-ish subs, would it? Most upvotes one gets might actually come from the same group of people...
Yeah. This seems as likely to detect cohorts of similarly interested groups of people as it would a voting ring.
This might be useful though, in that things which trend towards honest are actually trending toward general interest.
Maybe you could compare the user's ratio to the median ratio for all users who have posted in that particular sub, or something along those lines.
Ya. I would be curious at the false positive rate for this. For a website with millions of members and few voting rings you would need the false positive rate to be really really low for this to work