About Nyaastats

Nyaastats tracks download statistics across torrents on Nyaa, aggregated across individual torrents and visualized using data from the AniList API.

Why?

Educational and entertainment purposes, as the saying goes. And as far as I know, official streaming sites (Crunchyroll, Netflix, etc) don't publish metrics publicly anywhere, so this seemed to be the closest I can get to actual quantitative metrics.

Nyaa downloaders are also closer to my preferred subculture of anime fandom anyway, vs the ratings of normies on MAL or whatever.

What do the numbers actually represent?

The numbers are the downloaded count from Nyaa'a torrent tracker. Most torrent clients will tell the tracker when they finish downloading a particular torrent, which makes the number go up. This is also the same number the "Completed Downloads" column on Nyaa's website.

The download numbers on this website per show/episode are aggregated across all releases, using some fuzzy matching to compensate for different titles/filename formats.

Does one download equal one viewer?

Not necessarily. If you assume people only bother downloading one release per episode, download totals do approximate unique viewers though.

Which torrents are included?

Only torrents in the "English-translated Anime" category.

What is "Endurance"?

Endurance is a metric I made up that attempts to measure how well a show maintains its audience after the first episode.

It's calculated as: average downloads of episodes 2 and later divided by episode 1 downloads.

100% means later episodes get the same downloads as episode 1
80% means later episodes average 80% of episode 1's downloads (typical drop-off)
Above 100% is rare—it means the show grew in popularity after premiering

We cap at episode 14 to keep comparisons within a single season, even for shows that continue longer.

Presumably shows with unusually high endurance kept viewers engaged. Shows with low endurance had many people try episode 1 but drop it later.

What is "Late Starters"?

Late Starters is another metric I made up to try and identify "sleeper hits"—shows that weren't hyped but people picked up later.

It's calculated as the episode 1 downloads in the first week divided by episode 1 downloads after the first week.

10% means most people downloaded episode 1 right away.
50% means half the people started the show weeks later.

Why group shows by premiere size?

Shows with huge premieres (like sequels to popular series) behave differently than smaller shows. Also it keeps the beeswarm/scatterplots mangeable.

How do you handle different subtitle groups and video qualities?

We sum downloads across all releases of the same episode. If SubGroup A's 1080p release got 10K downloads and SubGroup B's 720p got 5K, we report 15K total for that episode.

This assumes most viewers pick one version to watch. It may slightly overcount if many people download multiple versions, or undercount if we miss some releases.

How do you define a "season"?

However AniList does.

How fresh is the data?

The scraper runs hourly to collect new torrents and update download counts. Weekly statistics are computed once per week. For ongoing seasons, you're seeing data up to the most recent weekly update.

Why is the data for some show missing/wrong/weird?

The matching process is fuzzy, and some torrent releases have weird filenames, and some shows just don't fit nicely into the assumptions made by the parser (off-season, episode 0s/OVAs). If you think you can improve any of the methods, Open an issue on GitHub.