The Game Census

Open data & API

The census's own recurring measurements of Steam games, free to download and reuse: four CSV files and a read-only JSON API. Everything below is also what powers the game pages on this site.

Licence and citation

The data is released under the Creative Commons Attribution 4.0 International licence (CC BY 4.0): you may copy, redistribute and build on it, commercially or not, provided you give credit. Please cite it as:

The Game Census (thegamecensus.com)

Every API and download response also carries an X-Census-Licence header stating the licence and citation.

Downloads

UTF-8 CSV with a header row. The first three files list only games that have a public page on this site, so a row there always has a matching /v1/game/{app_id} response. sales-calibration.csv lists the comparisons shown on the sales-estimate calibration page and can include games without one.

Read-only API

JSON over HTTPS, no key required, open to requests from any origin (CORS). Responses are cached at the edge for an hour. An id that is not a number, is unknown, or has no public page returns HTTP 404 with {"error":"not found"}.

GET /v1/game/{app_id}

curl https://thegamecensus.com/v1/game/730

FieldMeaning
app_idSteam application id.
name, slugGame name as listed on Steam, and its page slug (https://thegamecensus.com/game/{slug}).
release_dateSteam release date, YYYY-MM-DD; null when Steam gives none.
is_freetrue when Steam lists the game as free to play.
reviews.positive, reviews.negativeSteam review totals at the last census snapshot.
reviews.snap_dateDate of the most recent daily review snapshot we hold for the game.
health.tier, health.labelCensus health tier (thriving, healthy, cooling, quiet, dormant, unknown) and its display label; health is null when no tier has been computed.
health.computed_atWhen that tier was last computed (ISO 8601, UTC).
ccu.latest, ccu.latest_atMost recent concurrent-player sample we recorded and when (ISO 8601, UTC). It can be older than a day for games sampled less often.
ccu.peak_30dHighest concurrent-player sample in the last 30 days; null when there was no sample in that window.
sales_estimate.low, sales_estimate.highEstimated copies-sold range: total Steam reviews times a release-era multiplier range (the review-multiple method described on the methodology page). null for free-to-play games and for games with fewer than 50 reviews.
urlThe game's page on this site.

GET /v1/ccu/{app_id}

Daily maximum and average concurrent players for one game from our own samples, oldest day first, at most the latest 400 days.

FieldMeaning
app_id, name, slug, urlAs above.
series[].dateUTC calendar day, YYYY-MM-DD.
series[].maxHighest concurrent-player sample that day.
series[].avgMean of that day's samples, to one decimal.
series[].samplesNumber of samples behind that day's max and avg; days with few samples are less reliable.

GET /v1/health-tiers

One JSON array with an entry per game that has a computed tier: {"app_id", "slug", "tier", "current_players"}. current_players is the in-game count at the last tier computation, null when no sample from the preceding two hours existed.

GET /v1/reviews/{app_id}

Daily Steam review totals for one game from our own snapshots, oldest day first, at most the 400 most recent days that have a snapshot.

FieldMeaning
app_id, name, slug, urlAs above.
series[].dateSnapshot day, YYYY-MM-DD.
series[].positive, series[].negativeCumulative Steam review totals at that day's snapshot.
series[].steam_review_scoreSteam's own 0-9 review-score bucket for that snapshot; null when not recorded.

GET /v1/search?q={name}

curl "https://thegamecensus.com/v1/search?q=witcher+3"

Games with a public page whose name matches the query (2 to 80 characters), best match first, at most 10. A query that is missing or too short returns HTTP 400 with {"error":"q must be 2-80 characters"}.

FieldMeaning
qThe normalised query.
results[].app_id, results[].name, results[].slug, results[].urlAs above.
results[].reviewsTotal Steam reviews (positive plus negative).
results[].steam_review_labelSteam's own review label, e.g. Very Positive; null when Steam gives none.

GET /v1/movers?direction={rising|falling}&limit={1-50}

Review-velocity movers: the latest complete Steam review-histogram bucket against the same game's average over the 8 buckets before it. Both directions by default; pass direction for one list and limit (1 to 50, default 25) for its length. A bad value returns HTTP 400.

FieldMeaning
method, computed_atHow the ranking is built, and when the leaderboard was last materialised (ISO 8601, UTC; null when it was ranked live).
rising[], falling[]The two leaderboards (only the requested one when direction is given). Each row has the fields below.
rank, app_id, name, slug, urlPosition, and the game as above.
recent_reviewsSteam reviews in the latest complete histogram bucket.
baseline_reviewsThe game's average reviews per bucket over the 8 buckets before it.
delta_pctPercent change of recent_reviews over baseline_reviews.
recent_period_startStart date of the latest complete bucket, YYYY-MM-DD.

GET /v1/badge/{app_id}.svg

An embeddable SVG badge with the game's current player count, refreshed every 15 minutes at the edge. A game with no public page returns HTTP 404 with a JSON error.

GET /v1/openapi.json

The whole API above as an OpenAPI 3.1 document, for client generators and API explorers.

Feed

/feed.xml is an Atom feed of the monthly State of Steam reports and the weekly review movers.

What is included, and what is not

Included: the census's own observations (concurrent-player samples, review snapshots, health tiers) and facts Steam publishes publicly (name, release date, free-to-play flag, review totals).

Not included: any field that comes from a third-party database, critic or ratings aggregator, console storefront, community wiki or monthly-history service. We show some of those on game pages, but we have not cleared their terms for redistribution as an open dataset, so they are left out of the files and API on purpose rather than risk relicensing something that is not ours to relicense. The one exception is sales-calibration.csv, which carries Wikidata's reported unit sales and their reference links, as already shown row by row on the calibration page. Wikidata publishes its structured data under CC0, so those figures carry no extra terms. Its pc_only column is the page's cohort label (PC-only or all-platform), worked out from IGDB platform data; no IGDB field itself is included.

Update cadence

Downloads are rebuilt from the live database whenever their 24-hour edge cache expires, so a file can be up to a day behind. API responses can be up to an hour behind. Player-count sampling began on 6 August 2026, so there is no earlier player-count history in this dataset. How often an individual game is sampled varies; the samples column of /v1/ccu and the latest_at and computed_at timestamps show what each number is based on.

Fair use

A fair-use rate limit applies to the API and downloads: more than 20 requests in 10 seconds from one IP address gets HTTP 429 with a Retry-After header. Fetch each file once rather than polling it, and use the API for individual lookups. If you need something larger or more frequent, get in touch via the contact page.

For how the tiers and sales estimates are computed, see the methodology.