YouTube Milestones

YouTube Milestones

Around April 2018, I had an idea for a "YouTube Top 250" list of videos, modeled after the IMDb Top 250 movies.

However, finding an unbiased way to rank videos turned out to be difficult. If you just take the like/dislike ratio, you get a list heavily skewed towards new videos and videos with few views. I tried adding weights to the formula to counterbalance that, but never could get a list I was happy with.

Finally, I hit on the idea of using a Pareto frontier. Say that video A "dominates" video B if A is older, has more views, and has a higher like/dislike ratio than B. Now take the set of all videos on YouTube, and remove every video which is dominated by any other. You are left with the Pareto frontier, or set of undominated videos. There were about 270 of these (out of 1,000,000 videos in my database). Listing them by date gives a kind of "milestones" of YouTube.

The milestones are a nice mix of videos familiar because they have so many views, videos notable because of their age (e.g. "Me at the Zoo", the first video on YouTube), and videos qualifying due to their high like ratio (e.g. videos of YouTube celebrities). There are also a large number of videos of Korean boy bands, but this seems to be unavoidable!