I like it. My only suggestion is to log your axes before performing the regression. The residuals are not normally distributed, and you'll likely get an even stronger correlation this way. Although from visual inspection I don't think it's going to change the results very much if at all.
Where do you find information on how well a movie does after its theatrical release with streaming and DVD sales? I'm curious to find out how well "Mission: Impossible—Dead Reckoning" and "Indiana Jones and the Dial of Destiny" performed after they left the theaters.
Am I misunderstanding your last graph, or are you forgetting to compare base rates? Female director is the highest appearing tag, but is that because it's a more common tag than something like murder? I couldn't find the statistic on what percentage of movies are directed by female directors, but at least of current directors, 14% are female, so for future movies, you would expect approximately 14% of underappreciated to include female directors.
Maybe the graph is instead trying to say what percentage of movies with those tags are underappreciated? If so, then your point is supported, but it's not clear from the article.
I think another factor to consider in the analysis is who leaves online reviews for movies.
Normally, the crowd watching one to four movies a year, most or all of them highly marketed blockbusters, will not center their whole existence around their appreciation for movies and therefore, will not tend to use letterboxd or IMDb.
On the other hand, most of the underappreciated cult classics are also loved by professional critics or at least somewhat credited for being transgressive. So movie buffs will be more likely to watch such movies and rate them online.
This does not invalidate the analysis, but maybe puts 'over appreciation' into perspective. The biggest box office numbers are often hauled in by blockbusters with a huge budget that try to attract an audience as big as possible (many of them online non-commenters), so by design of this metric they will be overappreciated when extrapolating from their online review counts.
Actually, that would be an interesting analysis: How to infer actual viewership numbers from online ratings of movies. But if letterboxd is any indication, on that platform, correlation between the two seems very weak:
Thanks for sharing! Just a quick correction: "....high-concept and challenging to market, with premises not easily conveyed via a movie poster or tagline" is using high-concept incorrectly. High-concept means the premise is easily conveyed. It doesn't mean highbrow; more commonly, it means the opposite.
I like it. My only suggestion is to log your axes before performing the regression. The residuals are not normally distributed, and you'll likely get an even stronger correlation this way. Although from visual inspection I don't think it's going to change the results very much if at all.
Where do you find information on how well a movie does after its theatrical release with streaming and DVD sales? I'm curious to find out how well "Mission: Impossible—Dead Reckoning" and "Indiana Jones and the Dial of Destiny" performed after they left the theaters.
Am I misunderstanding your last graph, or are you forgetting to compare base rates? Female director is the highest appearing tag, but is that because it's a more common tag than something like murder? I couldn't find the statistic on what percentage of movies are directed by female directors, but at least of current directors, 14% are female, so for future movies, you would expect approximately 14% of underappreciated to include female directors.
Maybe the graph is instead trying to say what percentage of movies with those tags are underappreciated? If so, then your point is supported, but it's not clear from the article.
I think another factor to consider in the analysis is who leaves online reviews for movies.
Normally, the crowd watching one to four movies a year, most or all of them highly marketed blockbusters, will not center their whole existence around their appreciation for movies and therefore, will not tend to use letterboxd or IMDb.
On the other hand, most of the underappreciated cult classics are also loved by professional critics or at least somewhat credited for being transgressive. So movie buffs will be more likely to watch such movies and rate them online.
This does not invalidate the analysis, but maybe puts 'over appreciation' into perspective. The biggest box office numbers are often hauled in by blockbusters with a huge budget that try to attract an audience as big as possible (many of them online non-commenters), so by design of this metric they will be overappreciated when extrapolating from their online review counts.
Actually, that would be an interesting analysis: How to infer actual viewership numbers from online ratings of movies. But if letterboxd is any indication, on that platform, correlation between the two seems very weak:
Fast five, 429k ratings
Everybody wants some, 123k ratings
Citizen Kane probably fits your criteria. It did not recoup its budget during its initial theatrical run, but was praised by critics.
Thanks for sharing! Just a quick correction: "....high-concept and challenging to market, with premises not easily conveyed via a movie poster or tagline" is using high-concept incorrectly. High-concept means the premise is easily conveyed. It doesn't mean highbrow; more commonly, it means the opposite.