Ranking Methodology

Genre and tag pages on GeekyAlien (like /genres/{genre}) present games as a "best of" list. That word is a promise, and this page explains exactly how we intend to keep it — including where we currently fall short.

Where we are right now

GeekyAlien's catalog is still in early development. At the time of this writing it holds a small set of fixture games used for building and testing the site, not a real import of Steam's catalog or review data. Genre pages currently sort games alphabetically, which is a placeholder, not a ranking. The algorithm below is the plan for when real review data lands — we're publishing it now, before it's live, so the rule is fixed in public before any game benefits from it.

Why not just sort by "% positive"

Steam shows each game a positive-review percentage. It looks like a ranking signal, but sorting on it directly breaks the moment small sample sizes get involved. A brand-new game with 10 reviews, all positive, would show 100% — and rank above a genre-defining game with 50,000 reviews and a "merely" excellent 97% positive rate. The first game isn't proven better; it just hasn't been reviewed enough times for a bad review to show up yet. Ten reviews are not enough evidence to trust an average, but a naive percentage treats them as if they were.

The planned approach: Wilson score lower bound

Instead of the raw percentage, we plan to rank games by the lower bound of the Wilson score confidence interval for the share of positive reviews. In plain terms: rather than asking "what percentage of reviews are positive?", it asks "given how many reviews this game has, how confident can we actually be that its true positive rate is at least this high?"

A game with 10 reviews and 100% positive gets a low score under this method, because 10 data points don't rule out the true rate being much lower — the interval is wide, and we rank by its pessimistic end. A game with 50,000 reviews at 97% positive gets a high score, because that large a sample pins the true rate down tightly, close to 97%. The effect is exactly what a "best of" list needs: a handful of enthusiastic reviews from friends and family can't outrank a game the community has actually tested at scale, but a well-reviewed game with a modest audience still isn't buried under one with a bigger marketing budget and a similar quality bar.

We expect to combine that score with recency (so an old, stale review base doesn't dominate a list forever) and relevance to the specific page (a game only counts toward a genre or tag page's ranking if it genuinely belongs there). The exact weighting will be documented here once it's implemented, and the computed score will be stored and recalculated on a schedule rather than on every page load, so the ranking is stable between visits.

What never affects ranking

Affiliate partnerships, advertising deals, and any other commercial relationship never influence where a game appears in a "best of" list. Ranking is computed from review data only. Where a page includes sponsored or affiliate content — a deal callout, a promoted listing — it is kept visually separate from the ranked list itself, and labelled as such, so it's never mistaken for part of the ranking.

Questions about a specific ranking

If a list looks wrong to you, or you think a game is mis-ranked, use the report link on that game's page or see our contact page. This page will be updated whenever the algorithm changes, with a new "last updated" date above.