That was actually a great read. Learned a lot from the article.

For example:

Here’s how this works: YouTube URLs look like this: https://www.youtube.com/ watch?v=vXPJVwwEmiM

That bit after “watch?v=” is an 11 digit string. The first ten digits can be a-z,A-Z,0-9 and _-. The last digit is special, and can only be one of 16 values. Turns out there are 2^64 possible YouTube addresses, an enormous number: 18.4 quintillion.

And:

YouTube likes recommending videos with at least ten thousand views, while the median YouTube video has 39 views

Most interesting part for me:

Perhaps the most important thing we did with our set of random videos is to demonstrate a vastly better way of studying YouTube than drunk dialing. We know our method is random because it iterates through the entire possible address space. By comparing our results to other ways of generating lists of YouTube videos, we can declare them “plausibly random” if they generate similar results. Fortunately, one method does – it was discovered by Jia Zhou et. al. in 2011, and it’s far more efficient than our naïve method. (You generate a five character string where one character is a dash – YouTube will autocomplete those URLs and spit out a matching video if one exists.) Kevin now polls YouTube using the “dash method” and uses the results to maintain our dashboard at Tubestats.

And, of course, the discovery of Tubestats:

https://tubestats.org/