Since the nineties, I’ve been a big fan of movie recommendation engines; my video store had a system called “Clair V.” It was a touchscreen interface; I think you logged in with your phone number. They got the initial data set from customers; I remember filling out worksheets with movie ratings.
Lots of semi-colons in that paragraph.
Since the early aughts, I’ve been using Movielens, a non-commercial engine run out of the University of Minnesota. Despite my lava-hot takes, it’s always been very solid for me. Unfortunately, my dataset isn’t the movie log I kept from 1998 to 2004, which is lost to time. I emailed it to someone once, but it’s not in the Gmail sent folder, so I’m out of luck.
Movielens has an “unusual likes” and “unusual dislikes” section. My unusual dislikes include some Christopher Nolan, District 9 (no shit), and X2; amongst others, not worth listing. However, my unusual likes list contains all fantastic films. On these hills, I stand: Critical Care, Bright Victory, The Funhouse, One Night Stand, August, The Faculty, Purple Violets, and Interrupted Melody.
But Movielens isn’t great for discovery. My top picks haven’t changed much for years. I just haven’t gotten around to watching The 400 Blows again or Ran. I wish I could, but the problem with watch and read lists is they just keep growing.
Starting last year, I’ve also been using Criticker. It’s a little bit different, assigning you cohorts based on your ratings, so while Movielens says I’ll give Licorice Pizza ★½ because Movielens knows The Master sucks, Criticker says I’ll give it ★★★★. Because I’m a cishet white guy who loves Magnolia so I’m in that cohort.
A long time ago, I had a friend who thought the machine learning movie recommendation engines were a fundamentally flawed concept because tastes are too individualized. The last twenty years have proven that sentiment wholly inaccurate, of course, thanks to the Internet and various communities building up around x, y, or z. We’re all basic in one group or another; it’s just a matter of finding that group.
Though there are still times I’ll throw Movielens for a spin. Not as often as Criticker, which didn’t think I’d like The Rider.
One of my very back burner projects is developing a methodology for watch lists based on initial success, critical responses, critical re-evaluation, and legacy. I’m not sure why; I’ll obviously never have time to watch everything, but it seems fun as a project. Though, even in the earliest planning stages, it’d be absurdly United States-centric, just because of available data.
I suppose this post could be an immediate precursor to a Licorice Pizza post.
It isn’t. The next items on the watch list are a Pacino and Pfeiffer duet of Scarface and Frankie and Johnny. Movielens and Criticker are sure I’ll like one of them.