Princeton Senior Sauhard Sahi (under the supervision of Princeton Professor Ed Felton) conducted a study of 1021 randomly selected files pulled from the trackerless variant of BitTorrent.  10 of the 1021 files were found to be likely non-infringing, which would mean that more than 99% of the files sampled were copyrighted content.

Figure 1 – Types of BitTorrent files sampled by Sauhard Sahi

While 99% of these files sampled were pirated, it doesn’t directly translate to the percentage of BitTorrent network traffic.  The network traffic number could simply be computed by weighting the files by the number of active non-seeding peers (the number of people actively downloading).  This method already factors in the file size because the larger the file, the longer it takes people to download something which means more active non-seeding peers.  Even if a file gets downloaded by the same number of people, small BitTorrent files will have very few active peers because people will quickly finish the download and either begin seeding or leave the torrent.

Because the pirated files are always the most popular; and we know this because every top 100 list on popular BitTorrent sites are pirated files; we can surmise that the percentage of BitTorrent traffic flowing through a network will be much higher than the 99% file make up.  Of course, I hope Mr. Sahi can verify this from his sample and I’ll be contacting him to request this and update this post when the results are in.