Originally Posted by rowan

Now that I think about it more, perhaps a better metric would be takedowns versus unique objects. For example, a site with 100 DMCA takedowns and 500 vids (high percentage of piracy); versus a site with 100 DMCA takedowns and 500,000 vids (mostly genuine vids, or at least not yet complained about)

Google may not be able to reliably determine "number of potentially copyrightable objects" on a site, but they could use something simpler like a count of pages.

100 DMCAs with 500 pages = severe penalty
100 DMCAs with 500,000 pages = slight or zero penalty