My Content Rating System

This is how I rate movies, television shows, books, and music.
Mrs Slocombe and Captain Peacock
Mrs Slocombe and Captain Peacock

A New Rating System

I realized a while ago that when I rate a book or a film, I sometimes conflate the concepts of quality and enjoyment. It’s entirely possible to like a “bad” movie or to dislike a “good” book. In fact, this happens a lot. So I wanted to devise a rating system that accounts for both, is easy to use, and is adaptable to different ratings schemes.

Simplicity

To keep things simple, I decided to limit the possible values for quality and enjoyment to five:

Adaptability

I envisioned a lookup table in which I could look up my quality assessment along one axis and my enjoyment assessment along the other axis to find the appropriate rating. Here are the lookup tables for some common rating schemes:

Lookup Table Possible Outcomes Description
10 1–10 (no half stars) Possibly the most common scheme; I use it to rate movies and TV shows in Trakt. You might be tempted to covert these ratings to 1–5 scale (no half stars) by dividing by 2, but it doesn‘t quite work.
10 ½–5 (with half stars) This is equivalent to the 10-star scheme above. Rotten Tomatoes uses this scheme.
9 0–4 (with half stars) This is similar to the 4-bone scheme used by VideoHound’s Golden Movie Retriever, in which a 0 rating was called “Woof!” and there was no ½-bone rating. Leonard Maltin’s Movie Guide also used this, without any ½ or 1 rating.
6 0–5 (no half stars) Used by The Times
5 1–5 (no half stars) One of the most common schemes; Used by Yelp!, Common Sense Media, and Goodreads
5 0–4 (no half stars) Equivalent to the 5-star scheme above; Used by Gene Siskel (Chicago Tribune) and Roger Ebert (Chicago Sun-Times), among others.
4 1–4 (no half stars) Used by Matt Birchler’s Quick Reviews app
3 1–3 (no half stars) I don’t know of anyone who uses this hypothetical scheme.

The Math

In mathematical terms, a rating is a vector in two-dimensional space.
In mathematical terms, a rating is a vector in two-dimensional space.

Think of a rating as a vector in 2-dimensional space, with quality being one dimension and enjoyment being the other. The vector’s length is the Euclidean distance, rescaled by dividing by √2:

where

As you can see, the values for all three variables depend on the rating scheme in use. The values for “Terrible” and “Hate” will always equal min. Similarly, the values for “Great” and “Love” equal max. The values in between are equally spaced. For example, in a 1—5 scheme, the values are simply 1, 2, 3, 4, and 5. But for a 1-10 scheme, they are 1, 3.25, 5.5, 7.75, and 10. The result s is rounded either to the nearest whole or half, depending on whether half stars are allowed.

References