Your mention of taste is interesting as https://www.criticker.com engine basically recommends movies based on other people who have liked and disliked the same movies as you liked other films and it’s really effective!
I admittedly haven’t tried this one (I guess I will), but this was also the promise of Goodreads, and Netflix since the DVD era, and it just never seems to work for me.
I have admittedly discovered a handful of great authors via Goodreads recommendations, but they’re big names I think I would have eventually found anyway (e.g. Brandon Sanderson), and I spent a lot of time adding books I’d read and carefully considering how to rate each one.
I’m actually broadly curious why these systems aren’t more effective. It seems like such a perfect system on paper.
One reason is that Netflix killed theirs after paying a prize of $1M for a 10% improvement to their already good Cinematch recommender. See: https://www.netflixprize.com/
Yes, that's what the site is supposed to do. It is entertainment, not facts, and so I want to be entirely in the bubble of "Hot Fuzz" fans or "Dead Snow" fans and very much without intersection with any superhero stuff.
I've just imported 300+ ratings from IMDb into Criticker, and the recommendations it gave me seemed so far off that I didn't even feel like giving any of the movies a chance. Maybe I'm not watching enough classics, but it feels like the site hasn't really gotten to the long tail of movies.
In comparison, when I go rate movies on IDMb I'm rarely more than one star away from the average rating, so IMDb serves my needs quite well, anyway.
It seems to be? I get a lot of recommendations I’ve never seen before. I think if you rate a lot of non mainstream stuff it’ll pair you with people who do the same and voila - new non mainstream stuff!
I have admittedly discovered a handful of great authors via Goodreads recommendations, but they’re big names I think I would have eventually found anyway (e.g. Brandon Sanderson), and I spent a lot of time adding books I’d read and carefully considering how to rate each one.
I’m actually broadly curious why these systems aren’t more effective. It seems like such a perfect system on paper.