Analyzing My Shows and Movie Habits

Today, I decided to readd a watches page, but this time it isn’t built from hundreds of posts, but from the data that I get directly from Trakt’s API. I built a small tool called trakt-collector used to collect your history and save it in JSON format.

The Trakt API gives you so much information about every episode and every movie: from the title, to the rating, description, channel where it aired, when it aired first, the countries where it aired, etc, etc. I don’t actually need all that information, but it’s never too much to store.

So, and about the watches page: I decided to build a simple thing. It’s an exhaustive list, sorted by watch date with the most recent on top. I decided to put the series before the movies because it’s what gets updated more.

I have a script on my server that runs every day and updates the page on this website with the new information provided by the API. It generates a simple JSON (as you can see bellow) that is then fed to the page and the HTML gets generated.

{
  "movies": [
    {
      "title": "Airplane Mode",
      "watched": "2020-01-27T07:57:30.000Z",
      "url": "https://trakt.tv/movies/airplane-mode-2020"
    },
    {
      "title": "Murder Mystery",
      "watched": "2020-01-24T22:19:02.000Z",
      "url": "https://trakt.tv/movies/murder-mystery-2019"
    },
    // ...
  ],
  "series": [
    {
      "title": "Young Sheldon",
      "watched": "2020-02-22T14:39:07.000Z",
      "url": "https://trakt.tv/shows/young-sheldon"
    },
    {
      "title": "#CasaDoCais",
      "watched": "2020-02-20T23:54:53.000Z",
      "url": "https://trakt.tv/shows/casadocais"
    },
    // ...
  ]
}

Meanwhile, I thought: “what about doing some pseudo-analysis to the data I’ve been collecting?”. After all, Trakt’s “Year in a Review” feature is for premium users, which I’m not. So I’ve never been able to get yearly statistics.

I decided to use a library I discovered recently on Hacker News called vega-lite which allows me to produce beautiful SVGs representing the data. All the code needed to produce the figures you’ll see below can be found on the repository.

First of all, I wanted to know which genres I’ve been watching more. I separated shows (collections of episodes) from movies to get an overview and this is what I got:

The unique shows are marked as and the movies marked as . It’s clear that most of the content I watch is drama-related or comedy. Also, please note that a movie or show can be categorized as more than one genre.

After that, I decided it would be interesting to see how’s my monthly activity. In this case, I did not reduce the show’s episode. Here, I’m counting every episode as one and every movie as one.

It’s clear that most of what I watch are TV shows, but that’s normal due to their intrinsic nature. It’s also interesting to see that November and December were the months where I watched most TV shows. However, I think this might’ve been an exception that happened last year. In addition, it may also be related to the schedule of the shows I watch. Most of them may be released on that time of the year.

Regarding to which day of the week I usually watch more TV shows - colored with -, it seems it doesn’t matter much. Please note that this chart takes the sum of all episodes and movies I watched on each weekday through 2019. In opposition to TV shows, it seems that I clearly watch more movies on Saturdays and Sundays.

And… I decided to leave the most exciting graph to the end. Here’s a year map where you can visualize in which days of the year I watched more media.

The closer a square is to , the more episodes or movies I watched on that day. By looking to the graph, I’d say that I watched something on about 50% of the days of the year, if not more.

This was actually a nice experiment and I will probably repeat this kind of analysis with other data.