turingMachine

The last few years I have found less and less interesting movies to watch. Maybe I am just getting old ;)

One simple, but effective way I have used in the past to determine if a movie is worth watching is to use its IMDB rating. Except for few comedies (e.g. Monty Python) I have always enjoyed any movie in the IMDB TOP 250.

But what is a good ranking? Let us try to answer that question…

I have data from the IMDB up to May 2013. One of the files that they provide includes, for each movie, a weighted rank (from 0 to 10), and the number of votes (plus some other info that we are not going to use).

There are 210,371 movies. But not all of them have the same importance. The minimum number of votes is 5 and the maximum 971,041 –The Shawshank Redemption (1994), one of the IMDBs all-time favorites.

Here is the distribution of the number of votes in that dataset. 75% of movies have less than 118 votes!

    Votes       
Min.   :     5  
1st Qu.:    11  
Median :    26  
Mean   :  1524  
3rd Qu.:   118  
Max.   :971041  

Let us choose only movies that have a significant number of votes. Let us use 10k as the threshold. This leave us with 4,634 movies (I wonder how many movies Pic A Flic has). This is a Violin plot of the rank of movies in the IMD with at least 10,000 votes. It shows, in a single plot, two ones: a density kernel (the outside) and boxplot (the inside). The kernel tell us that this has an almost normal distribution and its peak is around 6.7. In fact, the median is 6.7 (that means that 50 percent of movies with at least 10k votes are above 6.7) and the average is 6.6. In fact, the 75% is 7.4 (the black area in the middle of the plot ends at the first 25%, and 75% quartiles–i.e. 25% of the movies are below 6.0). So if you go to the Rental store, and see a movie that is relatively popular and has a rank of 7.5, then it is within the top 25% of movies. If you are very picky, use a threshold of 8.0: there are only 430 with higher than 8, or approximately 9% of the movies in the IMDB with at least 10k votes.

This only holds true for Movies. For TV-shows the numbers are very skewed towards 10, and there have significantly less votes.

So next time they ask you: what movie should we rent? Simply answer: any with at least 10,000 votes in the IMDB and at least a rank of 8. Or even better, one of the top 250.

One word of caution: latest movies tend to be overvalued and their rankings lower as time passed by. For example, CM101MMXI Fundamentals (2013) has a current rating of 9.2 with 690,768 votes!!!! A movie I never heard of, but maybe a total success in Turkey, has more votes than all but 8 other movies, and a rank that is the highest except for one movie. Suspicious, if you ask me. Also, the voting data the IMDB provides does not seem to match exactly what is online.

–dmg