@ViciousPecan said:

Currently, when you click Explore, you can choose three sorting methods:

• Rating

If you choose "Rating Descending" you will get many trash results, like "Satan Moves to the Suburbs (2016)", which only turns up because one twerp rated 10 So it's mostly useless for the purpose of finding good movies, but still turns up one or another interesting movie.

I believe that IMDb awaits at least five ratings before showing average rating. Maybe doing something like that here would be advisable. Also, it definitely would help to show number of ratings for a movie, so folks know the number or ratings that the average is based upon; but I'm not so sure that using number of ratings as a sorting criterion would make much sense as a sort of surrogate for average rating. E.g., Plan 9 from Outer Space is generally regarded as one of the worst movies ever made. At IMDb it has an average rating of 4.0/10, but that terrible average is based on a huge 31,540 user ratings. (The 4.4 average it gets here has no count of number of user ratings, but I'm guessing it's minuscule relative to the number over at the other site.)

Actual average rating would work well as long as they await a minimum number of ratings before showing average. And I think it's obvious that showing number of ratings next to the average would be really helpful, so that users can judge for themselves whether the sample is large enough to be representative.

I believe that IMDb awaits at least five ratings before showing average rating. Maybe doing something like that here would be advisable. Also, it definitely would help to show number of ratings for a movie, so folks know the number or ratings that the average is based upon; but I'm not so sure that using number of ratings as a sorting criterion would make much sense as a sort of surrogate for average rating. E.g., Plan 9 from Outer Space is generally regarded as one of the worst movies ever made. At IMDb it has an average rating of 4.0/10, but that terrible average is based on a huge 31,540 user ratings. (The 4.4 average it gets here has no count of number of user ratings, but I'm guessing it's minuscule relative to the number over at the other site.)

Actual average rating would work well as long as they await a minimum number of ratings before showing average. And I think it's obvious that showing number of ratings next to the average would be really helpful, so that users can judge for themselves whether the sample is large enough to be representative.

I remembered reading this about averages the other day in "Site Reliability Engineering" p. 42:

Most metrics are better thought of as distributions rather than averages. For example, for a latency SLI [Service Level Indicator], some requests will be served quickly, while others will invariably take longer–sometimes much longer. A simple average can obscure these tail latencies, as well as changes in them. Figure 4-1 [omitted] provides an example: although a typical request is served in about 50ms, 5% of requests are 20 times slower! [...]

Translated into TMDB terms:

Most metrics are better thought of as distributions rather than averages. For example, for a rating metric, some ratings will be 9, while others wil invariably be 4–sometimes even 1. A simple average can obscure these tail ratings, as well as changes in them. Figure 4-1 [omitted] provides an example: although a typical rating is 7, 5% of ratings is 20 times worse.

A well-known example where ratings are broken up into segments is Amazon. You can see at a glance the distribution of ratings. It might be worthwile to do this as well on TMDB.

@ViciousPecan said:

A well-known example where ratings are broken up into segments is Amazon. You can see at a glance the distribution of ratings. It might be worthwile to do this as well on TMDB.

Yes, IMDb's system that allows us to see distributions is preferable, but a measure of central tendency is also helpful as long as it is based on a reasonable minimum number of ratings and if the number of ratings is indicated. And for a measure of central tendency, the median is, in this case, preferable to the mean, the latter of which can be more easily influenced by a few extreme ratings.