Methods and sources

How the numbers were made

Every figure in the atlas comes from one snapshot of the Steam catalogue. This page explains each field, how it was cleaned and where it falls short.

Sources

Two files make up the snapshot, taken in late October 2026. The latest release date in it is 25 October 2026. The first is a catalogue export of 144,751 Steam apps with store fields and SteamSpy estimates. The second holds the press quotes shown on 13,568 store pages.

Which games are included

Names and text

All text is written in plain ASCII. Accents and other marks are removed, ligatures are spelled out, curly quotes become straight quotes, dashes become hyphens and symbols such as trademark signs are dropped. Cyrillic and Greek letters inside Latin names are converted to their Latin forms. Store descriptions, content notes and press quotes written mainly in another script are omitted rather than shown in part. Page addresses are built from the cleaned name; when two games share a name, the less reviewed one gets its release year or app ID added.

Dates

Release dates are the dates on the store page. Year and month pages use them as given. Trend charts start in 2006, when the catalogue first passed a few dozen releases a year. 2026 is partial and runs to 25 October 2026.

Reviews and rating bands

Steam reviews are the count of positive plus negative reviews. Positive share is positive reviews divided by that total. Shares and medians of positive share use games with at least ten reviews unless stated. Review counts for the newest games are incomplete in the snapshot: 62% of 2024 releases show at least one review, against 24% of 2025 releases and 1% of 2026 releases, so the share of games with no reviews is inflated by recent years. Rating bands are computed from the counts with Steam's published thresholds; Steam's own label can differ slightly because the store applies its own filters.

Steam rating bands
BandRule
Overwhelmingly Positive95% or more positive, 500 or more reviews
Very Positive80% or more positive, 50 or more reviews
Positive80% or more positive, 10 to 49 reviews
Mostly Positive70% to 79% positive
Mixed40% to 69% positive
Mostly Negative20% to 39% positive
Negativeunder 20% positive, 10 to 49 reviews
Very Negativeunder 20% positive, 50 to 499 reviews
Overwhelmingly Negativeunder 20% positive, 500 or more reviews

Rankings and percentiles

On each game page, a bar shows the share of games that the game outranks: games with fewer reviews, a lower positive share among games with ten or more reviews, a lower price among paid games released the same year, a lower recorded peak among games with one, or a lower owner band among games with an estimate. Ties do not count as outranked. Overall and yearly ranks order games by review count, then by positive reviews.

Owners, players and playtime

Estimated owners come from SteamSpy as bands such as 20,000 to 50,000. Games reported as 0 to 0 are shown as having no estimate. The share of owners held by the top games uses the geometric midpoint of each band, so it is indicative, not exact. Peak concurrent players is SteamSpy's peak for the day before the snapshot and is recorded for 19,420 games. Playtime is shown only where SteamSpy reports it, which is 9,851 games.

Genres, features and tags

Genres and store features are chosen by developers. A handful of genre and feature labels stored in French or Spanish were mapped to their English equivalents. Community tags are applied by players, and Steam keeps the twenty most applied for each game with the number of players who chose each one. Tags only appear once a game has reviews, so tag shares are computed over the 82,043 games with at least one tag. Genre pages exist for genres with at least 20 games and tag pages for tags with at least 10.

Prices

Prices are the current US list price at the snapshot, so older games show what they cost today rather than at launch. A game is free when the store marks it free or its price is zero. Games the store marks as unavailable are shown as not sold and left out of price statistics.

Languages

Interface and full audio languages come from the store's language list. Footnote markers and formatting were removed, and only names used by at least 20 games are kept, which drops a few garbled entries.

Press quotes

Each quote on a store page was split into text, score and outlet. Scores were normalised to 100: fractions such as 8/10 and 4/5, percentages, bare numbers up to 10 multiplied by ten, and letter grades mapped to a fixed scale. Outlet names were merged on a lowercase alphanumeric key, and well known outlets were given their usual spelling. 11,712 of 28,263 quotes carry a usable score. Developers choose which quotes to show, so these are not a sample of all reviews. Outlets with ten or more quotes have their own page.

Similar games

For games with at least three tags, similar games are found by comparing tag lists. Each tag is weighted by how rare it is across tagged games, the ten most applied tags count fully and the rest at about half, and games are compared by the cosine of those weights with a small bonus for games with more reviews. Games from the same developer and games with the same name are left out, because the page lists those separately. Games with fewer tags show the most reviewed releases that share their main genre and year instead.

Studios

Developers and publishers are matched on a lowercase alphanumeric key, so small differences in spacing or punctuation are merged, while different legal names such as a studio and its parent company stay separate. Studios credited on three or more games have their own page. Studio names written only in a non-Latin script are described by their script rather than shown.

Charts

Categorical colors follow a fixed order, magnitudes use one blue ramp, and pairs that run above or below expectation use blue against red. Charts are drawn when the pages are built, need no script to display and have a table view with the exact values.

Corrections

The atlas reflects the store as it was on the snapshot date. If a game's details have changed since, the store page linked from its page is the current record.