Ask HN: An alternative to IMDB scores, whats a better way to pick movies?

[pauljeba]

I like IMDB rating, helps find the best movies easily. But, recently when I decide to watch a movie based on the scores alone, I have been greatly disappointed. Eg. The Joker 2019 movie is highly rated, but its not the movie for me. I don't generally like such sad and mind-fuck kind of movies.

Has anyone thought about this? I would like to create a new movie rating system, that will be easy to pick the right movies.

[s1t5]

The best indicator is a recommendation from someone whose opinion on movies is largely aligned with yours. I suppose you need a platform where the user is profiled and matched against reviewers/raters/critics with similar tastes. Has anyone done this?

[bobsoap]

criticker.com does exactly this by normalizing individual user ratings into percentiles and calculating a "Taste Compatibility Index", floating those users whose ratings are closest aligned with yours to the top (and their highest-rated movies). In my case, the resulting predicted score for most movies is close to 80% accurate. Highly recommended site.

https://www.criticker.com/

[Someone]

Every major “movie provider” will have something like that. Netflix certainly has (https://en.wikipedia.org/wiki/Netflix_Prize).

[bobsoap]

Yes, but Netflix's recommendations are limited to Netflix's catalog.

[pauljeba]

Yes, that's how most recommendations are being effected. I think we need also an explicit way of communicating our preferences, and getting scores against them. The problem with recommendations is, they are not transparent enough and can fail.

[mattbrewsbytes]

When it comes to picking things you are spending your free time on, movies, dining out, video games, books, etc. you are more likely to immediately not even consider stuff you strongly dislike. Its really frustrating to see things in lists you wouldn't even consider. You lose any hope in that system being able to help you find something to escape reality for a while.

What is needed is a movie pick service and that starts with asking what genres you hate. Then get into specific movies you hate to further refine the list of things I don't want to see but at the same time clarifying questions to narrow down the like/dislikes because there are movies that combine elements from different genres. For example, The Joker could be considered in the #superhero, #sad, #mind-fuck categories and if you like #superhero movies but strongly dislike #sad and #mind-fuck then those should be weighted higher and the system shouldn't show The Joker on any personal lists.

The service should have an unfiltered search so people can just find stuff too.

[pauljeba]

Exactly, free time is a premium and has to be treasured.

I think a quiz like what you meantioned would be effective. The problem is finding data that could help power such a quiz. Example, a #tag of #sad could be assigned to a movie that's just 20% sad, leading to misuse/misinterpretation. What we may need is a variable-tag metric... But I don't think there is any available data for this yet..

[slightwinder]

It's impossible to press a complex object into singular value. Everyone has different key aspects and you can't see from a single value which key aspects are strong and which are weak.

Read some reviews and critics to get a hint on the vibes and content of a movie and whether it matches your personal taste. IMDB has them both.

[pauljeba]

What about multiple metrics, which each can talk about specific aspects of why someone would watch a movie. I could then weigh these metrics as per my preference and choose the movie of my liking. Do you get it?

[slightwinder]

Sure, you can do that. There are sites doing this, many more did that in the past. Most failed at some point because the amount of work, knowledge and dedication for this to reach a critical mass of uptodate meaningful data is to big to survive longterm.

And in the first place you will be very busy to figure which metrics you want in the first place. You basically need a complete set of them before you even begin for serious, because otherwise you will invalidate all older entries if you add metrics later.

Thinking about, tags are popular today, several media centred-sites are using them to allow a kind of freeform-metric to describe content. Maybe with modern solutions you can do something machine learning-voddoo to extract tags from reviews and generate meaningful metrics for describing impressions of a movie.

[pauljeba]

You are very right, especially with the challenges in making it successful. Crowd sourcing or Machine Learning are the options I guess, each with its pros and cons.

[dodgyb]

Metacritic consistently come out on top when compared to IMDB and Rotten Tomatoes. This article is a good dive into the pros and cons:

https://thenextweb.com/syndication/2018/02/04/imdb-rotten-to...

[IndexPointer]

Only you can know what you like.

The more movies you watch, the better your intuiton for predicting if you'll like the movie by watching the trailer will get.

I found imdb scores are good to have a general idea. Eg I probably won't like a movie with a 5.9 score, even if the trailer seemed good. Or I'll give a movie with a >8 score a chance if the trailer is okay, but sometimes I know a movie is not for me by the trailer, no matter how high the score is.

For everything in between I decide mostly by the trailer and the past movies of the director.

[pauljeba]

What do you look for in a trailler ? Could you describe the process in few words?

[IndexPointer]

>Could you describe the process in few words?

Not really, I wouldn't know how to describe it. I just have an intuition of when a movie is "promising" according to my taste. An intuition that gets better the more trailer and movies I watch.

[Jugurtha]

I like a few actors, directors, screenwriters. I watch what they did, and watch movies when one of them makes one. I can't watch anything else. It limits the movies I can watch to one or two per couple of years, but it's okay.

I used to be a voracious movie consumer when I was younger, but it's due to the fact there was a backlog. Now there is none.

[avocadoLife]

I've had some good recommendations from https://movielens.org/

Also, if you like a particular movie/genre then it's best to google for similar ones. People tend to recommend 'if you like X then you should also check out Y'. I've found this to be a decent indicator.

[pauljeba]

I like these profile based recommendations, but they tend to be black box ones. They miss out on the how I may like a future movie that I have not come across before. I like how nomadlist.com has done it for places. You can easily figure out which place to go, just using their fitlers.

Would a similar filter be useful for movies?

[dairylee]

In the past I used to use Criticker (https://www.criticker.com/) and found that once I'd rated some movies the recommendations were very good.

But like others have said the easiest way is to find a friend/critic who has similar taste to you and just see what they recommend.

[pauljeba]

I am really looking for an answer to this. I promise to incorporate the best answer arriving out of our discussion in my site flixcatalog.com. I don't think the existing filter on imdb scores makes sense.

[11874442]

Isn't the current iMDB score just an average of everyones scores combined?

How do you propose a scoring system would work where movies are scored according to your preference in movies?

[pauljeba]

If you look at what nomadlist.com has done for places - I can easily figure out if I would like to visit a place or not, by selecting their emoji based filters.. I think a similar emoji based rating, that could say how sad a movie is, how happy a movies is, etc, could be useful. What you think?

[11874442]

Personally I would associate a frowning/sad emoji with something negative; the movie is bad etc.

It sounds like you want to seperate the rating and emotion/vibe of a movie, and score/rate both?

[pauljeba]

Yes, emotions- "how does one feel while watching the movie". This is missing from the traditional genre tagging - "whats the movie about"

[IndySun]

You could, instead of looking at the score only, also use the tags on imdb to at least guide you to films of interest to you.

[pauljeba]

You mean the genre tags, like crime, thriller, etc?

[IndySun]

Yes. Things like that, plus the certificate, the mood you're in, the usual reasons for sitting down to some visual presentation. Also, on imdb the scores are incredibly mean average for very popular movies and can vary wildly on the lesser known. If you take a look at a movie you love, and one you hate, you can find a surprisingly short width of score. What this means is that the meaningful scores are 6ish to 8ish. Most popular Movies sit in that area, hence a 6ish in genre you like may actually be rated great by you and an 8ish in genre you dislike might mean you'd appreciate it though not love it.

[5041us]

letterboxd.com - The score is a good reference, but you will also find really good lists from users with the same taste as yours.

[pauljeba]

I like letterboxd, but unless I read through all the comments, I can't make out if I would like a movie or not. There scoring is similar to imdb. I dont think just a simple rating system would be a solution.

[baspey]

The most efficient solution is to give the user a predicted rating which tells him at first glance how much he'll like a movie. Of course the prediction is based on how much the user has liked other movies (collaborative filtering). Netflix used to do that quite well until their recommendation system went down the drain for marketing reasons. My app Coollector Movie Database works on that principle.