In a previous post we talked about ratings and that they are often distributed according to a Gaussian distribution. This is related to the wisdom of the crowd that roughly says that if we average all opinions, we usually get a good result which is often better than individual results. In terms of ratings it means that if a movie has an average rating of “6” that this value is supported by the majority of the voters, up to some fluctuations.
However, as usual, there are exceptions. For instance, if we consider science fiction B-movies, especially those that copy themes from popular big-budget movies, bi-modal distributions, which we call “hate-it-or-love-it” distributions, seem to be more often. So often, that one might call it a pattern. For instance, movie X has a mean rating of “3”, but if we analyze the ratings, we can see that almost 10% gave the highest rating while 30% gave the lowest rating. Of course small numbers are not very trustworthy, but the trend that about 10%, or at least a notable fraction of the voters, gave the highest rating is present in many sci-fi B-movies.
So, why there is often such a large group of loyal voters that “liked” the movie so much? Maybe it became a hobby to watch low-budget scifi movies? Or even a trend to watch those movies because they are so entertaining? Is it a trend at all, or did we stare too long at bar graphs? Bottom line, even without highly sophisticated algorithms, we can squeeze a lot of knowledge out of movie ratings and this additional knowledge can be, for instance, incorporated into factor models.