HOW IT WORKS
We read your Steam library through the Steam Web API. Every game you own is classified as either played (any playtime at all) or unplayed (zero hours).
For each anchor game, we pull metadata from the Steam Store—genres, categories, developers, and publishers. These become feature vectors, weighted by square-root of playtime so that no single game dominates.
WEIGHT: √(playtime) — dampens outliers, keeps signal from medium-playtime games
Every unplayed candidate is scored against your taste profile using cosine similarity, combined with a quality signal from Metacritic scores and Steam recommendation counts.
Single-player games get a small boost. Early Access games get a penalty.
Results are split into three recommendation lists:
Click any anchor game to trigger a "If you liked X..." search. This uses Jaccard similarity on genres (78%), categories (17%), and developer (5%), then re-ranks using Steam user-defined tags when available (72% weight).
Your Backlog Hunter level is based on the percentage of your library you've played.
The HP bar shows your played percentage. The MP bar shows your unplayed percentage. Play more of your backlog to level up.
Your Steam profile's "Game details" must be set to Public for us to read your library. We don't store your data—everything is computed on the fly and cached briefly for performance.