2 minute read

Hover over the interactive plots, made with Plotly

It turns out the film I value more than most on Letterboxd is 22 Jump Street and my most disappointing was The Terminator. Now, comedies are subjective, so I feel no need to defend one of my favourites, but I might get some flack for my Terminator review. In my defence, this film is so important for sci-fi that I knew the whole plot and infamous moments before watching. Maybe its my own fault, but I wasn’t expecting it to be quite so low budget, particularly the mirror scene…

Superior Streaming Service?

Interestingly, I watched more than twice as many films on streaming services than in the cinema, but which is superior?

From the below figure, the box plots show that I found films on Netflix the highest quality overall (excluding iplayer which only had 3 films). While the cinema had the largest spread in ratings, it did provide the most 5 star films: Barbie, Oppenheimer, Mission: Impossible Dead Reckoning and 2001: A Space Odyssey

Can you guess when I finished my PhD thesis?

Getting average ratings from Letterboxd

While Letterboxd provides some neatly packaged .csv files for users to explore their own data, the API for gathering information about each film is private. Therefore, to compare my ratings with those of the general public, I needed to first find the average rating of each film.

Originally I turned the Name column into a hyphen separated URL, such as https://letterboxd.com/film/no-time-to-die/. However, this referenced a little known film about a hearse driver, rather than the James Bond one I had seen, with the actual URL of https://letterboxd.com/film/no-time-to-die-2021/.

Within the .csv files, Letterboxd provide the url of each of personal film review, such as https://boxd.it/41lX6F. Thankfully, this redirected to the full address like https://letterboxd.com/rlaker/film/no-time-to-die-2021/, which just needed my username removing with the following function:

import requests
import pandas as pd

watched = pd.read_csv("watched.csv")

def get_final_url(redirect_url):
    try:
        response = requests.get(redirect_url)
        if response.history:
            # Extract final URL after redirection
            final_url = response.url
            # Remove '/rlaker' part from the URL
            final_url = final_url.replace('/rlaker', '')
            return final_url
        else:
            return "No redirection occurred"
    except requests.RequestException as e:
        return str(e)

watched['Letterboxd_URL'] = watched['Letterboxd URI'].apply(get_final_url)

After originally trying to find the right <span> with BeautifulSoup, I realised the data was actually stored in a dictionary, which was then served by Javascript later (as explained in this Reddit answer).

It turned out regular expressions were enough for this task:

import requests
import re  # Regular expression module

def get_average_rating(url):
    try:
        headers = {
            'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/58.0.3029.110 Safari/537.3'
        }
        response = requests.get(url, headers=headers)
        if response.status_code == 200:
            # Use regular expression to find "ratingValue"
            match = re.search(r'"ratingValue":(\d+\.\d+)', response.text)
            if match:
                return float(match.group(1)) # The first group captures the rating value
            else:
                return None
        else:
            return None
    except Exception as e:
        return str(e)

# Applying the function to each URL
watched['Average_Rating'] = watched['Letterboxd_URL'].apply(get_average_rating)

You May Also Enjoy