What a match used to say, and when it stopped saying it
The changes made to a match's own fields, newest first, each with the value **before** and **after**. Everywhere else reports THAT a match moved — `?changed_since=` on the match list and the WebSocket `changed` field — because every other record is updated in place; this is the only surface that retains the prior value, which is what lets you reconcile a stored copy rather than overwrite it. Only the match's own observable fields are tracked: `status`, `score_a`, `score_b`, `winner_team_id`, `scheduled_at`, `started_at`, `ended_at`, `stage`, `format`, `result_type`, `team_a_id`, `team_b_id`, `tournament_id`. An entry whose `changed` is `["data_available"]` and whose `changes` is `{}` is an ARRIVAL, not a change — a sub-resource (maps, stats, rounds, depth, lineups) landed for the first time and the match's own fields did not move. `was_completed` marks whether the match was ALREADY `completed` when it moved, which is the line between a correction to something we published as final and ordinary lifecycle progress. History begins when this endpoint shipped and cannot be backfilled — earlier values were overwritten in place. Entries are retained for 180 days. A match with no recorded changes yields an empty list, as does an unknown id. Not cursor-paginated (`next_cursor` is null); page backwards with `before`. Ids inside `changes` are as recorded at the time and are never rewritten by a later merge, so one may no longer resolve.
The changes made to a match's own fields, newest first, each with the value before and after. Everywhere else reports THAT a match moved — ?changed_since= on the match list and the WebSocket changed field — because every other record is updated in place; this is the only surface that retains the prior value, which is what lets you reconcile a stored copy rather than overwrite it.
Only the match's own observable fields are tracked: status, score_a, score_b, winner_team_id, scheduled_at, started_at, ended_at, stage, format, result_type, team_a_id, team_b_id, tournament_id. An entry whose changed is ["data_available"] and whose changes is {} is an ARRIVAL, not a change — a sub-resource (maps, stats, rounds, depth, lineups) landed for the first time and the match's own fields did not move.
was_completed marks whether the match was ALREADY completed when it moved, which is the line between a correction to something we published as final and ordinary lifecycle progress.
History begins when this endpoint shipped and cannot be backfilled — earlier values were overwritten in place. Entries are retained for 180 days. A match with no recorded changes yields an empty list, as does an unknown id. Not cursor-paginated (next_cursor is null); page backwards with before. Ids inside changes are as recorded at the time and are never rewritten by a later merge, so one may no longer resolve.
Primary auth (live). Authorization: Bearer <api-key> — the header takes precedence over the query fallback. The raw key is hashed (SHA-256, hex) and looked up in api.keys.key_hash; the raw value is never stored.
In: header
Path Parameters
Game slug. cs2 is the only populated title today; an un-onboarded game 404s cleanly.
Resource identifier — either a UUID or the resource's human-readable slug, scoped to {game}. /v1/cs2/teams/natus-vincere and /v1/cs2/teams/019f23d1-fb5c-7987-b522-47c3a5e72111 address the same row, so you can go straight from a slug you already have without first looking up its id.
Slugs work on teams, players and tournaments. Matches have no slug, so a match id must be a UUID. Either way, an identifier that resolves to nothing returns 404 not_found.
Query Parameters
Maximum entries to return. Default 50, maximum 200.
1 <= value <= 20050RFC 3339 timestamp. Returns only entries strictly EARLIER than this instant. Exclusive, so passing the changed_at of the last entry you received returns the next page with no overlap and no gap. A malformed value is a 400.
date-timeResponse Body
application/json
application/json
application/json
application/json
application/json
curl -X GET "https://example.com/v1/cs2/matches/019f2858-e253-7f3d-bdff-bc7738bd1036/history"{ "data": [ { "changed_at": "2026-09-21T19:42:07.318402Z", "changed": [ "data_available" ], "changes": {}, "was_completed": true }, { "changed_at": "2026-09-21T11:08:44.901233Z", "changed": [ "score_b", "winner_team_id" ], "changes": { "score_b": { "from": 1, "to": 0 }, "winner_team_id": { "from": "019f2858-e255-78e4-9e7c-32f307dfac91", "to": "01a0916e-90e4-7469-903a-463c4fc69c4b" } }, "was_completed": true }, { "changed_at": "2026-09-20T21:15:02.447810Z", "changed": [ "status", "score_a", "ended_at" ], "changes": { "status": { "from": "live", "to": "completed" }, "score_a": { "from": 1, "to": 2 }, "ended_at": { "from": null, "to": "2026-09-20T21:14:58Z" } }, "was_completed": false } ], "meta": { "count": 3, "next_cursor": null }}{ "error": { "code": "not_found", "message": "not found", "request_id": "019f3d18-c15f-7319-81a7-343e8a80a578" }}{ "error": { "code": "not_found", "message": "not found", "request_id": "019f3d18-c15f-7319-81a7-343e8a80a578" }}{ "error": { "code": "not_found", "message": "not found", "request_id": "019f3d18-c15f-7319-81a7-343e8a80a578" }}{ "error": { "code": "rate_limited", "message": "rate limit exceeded", "request_id": "019f3d18-c15f-7319-81a7-343e8a80a578" }}A single match GET
One match by id, with the same fields as a list row minus the denormalised names — a detail row carries bare `team_a_id`/`team_b_id`/`tournament_id`, so resolve those separately or read them off the list route instead. The `data_available` flags tell you which sub-resources (`maps`, `stats`, `map_stats`, `vetoes`, `rounds`, `depth`, `lineups`, `odds`) actually exist for this match, so you can skip requests that would return an empty list. Pass `include=` to embed related resources or the sub-resources themselves in this one response.
Pre-match roster snapshots for a match GET
Each team's listed roster as it stood in the 48 hours before the match, oldest snapshot first within each team. **This is a roster snapshot, not a confirmed starting five.** Substitutes are listed too, so `player_ids` usually holds 6 or 7 ids and can hold anywhere from 1 to 11 — never assume a length of 5. The five who actually played are the player rows on `/stats`; compare the two to see who was benched or who stood in. A new row is written only when a team's roster differs from its previous snapshot, so most teams have exactly one row per match, and a second row means the roster changed in the run-up. An old `captured_at` is therefore not staleness. Coverage starts on 2026-07-09 and reaches roughly 6 in 10 matches since then; every earlier match returns an empty list. `data_available.lineups` and `?has=lineups` tell you which. An unknown match id also returns an empty list.