FAQ

🗄️ Database

Which one? We can probably add it. Our data comes from The Movie Database, which holds a staggering 1.3 million titles. Since indexing every single one would be a performance nightmare, we only support the ~15% most popular, which is normally plenty. If you'd like to see something that you're sure already exists on TMDB, just drop us the link and we can easily add it. 👀

No, sorry. Mature movies are welcome, R-rated and even NC-17, but explicit adult content is a whole different thing. Putting it on an open EU-based site would drag us under the strict rules for keeping that stuff away from minors, which means locking everyone behind age checks. Not happening.

Unfortunately we can't, as none of the big streaming platforms let outside apps read your viewing history and most shut down their public APIs years ago. If any of those services ever opens an export function, we'll happily revisit.

🤖 Discovery Queue

Not exactly the kind you're familiar with. We use a standard matrix factorization model from Microsoft's ML.NET library that looks at how you reviewed films compared to everyone else, then generates numeric vectors (your "taste profile") and uses what could be described as smart algebra (the "algorithm") to calculate compatibility scores. All these cool things existed decades before ChatGPT and image generators.

We gave the model a running start! To skip the awkward "empty room phase" during launch, we've integrated the MovieLens dataset, a public research export of tens of millions of real, anonymous film ratings gathered by the University of Minnesota's GroupLens lab. As more of you review films here, the model leans harder on our own community and lighter on that borrowed data. 😉

Ratings are kind of the whole point. Your reviews, specifically how much they differ from our predictions, are what shape your taste profile which is the foundation of the entire sorting and recommendation system. If you only care about tracking movies in generic lists, you can technically just remove them... though there are probably better sites for that.

Doesn't really matter! Machine learning vectors adapt to any scale you gravitate toward, so you're free to assign ratings however you like - as long as you're consistent over time. Everyone starts from their favorites, but for better results you should make sure that you're also including what truly disappointed you.

Not Sure is more flexible. Both options send the movie to your /skipped page, where you can revisit titles sorted by ML score. Since Not Sure entries can be reset separately, you can refresh your discovery queue without reintroducing films you've more permanently dismissed with Not Interested. It's a handy way to keep tabs on titles you're curious about, without committing them to your watchlist.

Nope! It's just a convenient social feature that lets you view their reviews and send or receive direct recommendations. So don't worry, adding that one friend who only watches mumblecore won't mess up your results.

❌ Missing Features

It's mostly about keeping things simple. The interface is already packed with features, so adding dropdowns and buttons to filter your watchlist by streaming service would make things even more cluttered for most users. If you need more control, try using custom sections which support batch editing.

Out of scope, on purpose. ReelsGraph is built around backlog management and discovery, not chronological journaling. If keeping a dated log of every viewing is important to you, Letterboxd is genuinely great at that and you should keep using it. We'll happily import your CSV.

No reliable source. Streaming providers don't expose a clean machine-readable list of dub and subtitle tracks per title, and the data that is available varies by region and over time. Adding the filter without accurate underlying data would just produce false negatives. If a clean source ever shows up, we'll revisit.

The data from TMDB is spotty. Most languages are missing entirely, so showing them would be wildly inconsistent from one title to the next. On top of that, the definitive launch trailer is usually English-only, while secondary languages tend to get just a short teaser.

🤔 Miscellaneous

The alternatives invite lazy ratings. On a five-star or 1-to-10 scale, almost everyone drifts to the top and an 8/10 quietly becomes the new "just okay", while the bottom half goes basically unused except for protest reviews. Words like Mediocre and Excellent break that reflex, and reserving Amazing or Loved for what you truly treasure hands the recommender an unusually strong signal.

Through external means. To keep things safe and spam-free, we don't currently support in-site messaging. That said, if someone has filled out Discord or Reddit info on their profile, you're welcome to reach out there.

No worries, your IP isn't stored. To cut down on spam - especially for users with public profiles - we track a scrambled fingerprint called a hash. There's absolutely no way to turn that back into your real info. It just means you can't recommend the same movie multiple times, even with tricks like incognito mode.

Just one person - hey, that's me👋. ReelsGraph is the movie-shaped sibling of GamesGraph, a project I started in 2019 as a way to learn web development with ASP.NET Core. The same backlog, taste-profile, and discovery ideas turned out to fit films beautifully, and a lot of friends and power users keep helping me sharpen the rough edges. The "we" form includes them, too.