TwitchTracker - Game & Channel Rankings

Datadory delivers TwitchTracker game and channel rankings data covering thousands of games and millions of Twitch channels — average and peak concurrent viewers, hours watched, minutes broadcast, follower gains and share-of-platform percentages across trailing seven-day leaderboards, thirty-day summaries and multi-year monthly history, delivered daily, weekly, or hourly.

What is the TwitchTracker - Game & Channel Rankings dataset?

Every night, between the raid trains and the ranked grind, the largest live-audience market in gaming re-sorts itself — and TwitchTracker keeps the standings. The service publishes ranked leaderboards of top games and top channels alongside follower and viewership movers, tracking millions of channels and thousands of games while surfacing the top roughly 100 of each on its front shelves. The numbers underneath are industrial-scale: the platform logged 1,530,544,896 hours watched in the most recent complete month, against an all-time monthly peak of 2,307,473,592 hours in May 2021. Datadory turns those standings into a clean, delivered feed — average and peak concurrent viewers, hours watched, minutes broadcast, followers gained and totals, and each game's share of the platform's entire audience — across trailing seven-day leaderboards, trailing thirty-day summaries and monthly history reaching back years. The record scores 7/10 on our quality rubric and sits among the 16 datasets catalogued under Interactive Home Entertainment.

Get a sample of this dataset and we cut rows toward the games, channels and windows you actually model.

What does a sample of the data look like?

One row per entity-window observation — five captured straight from the August 2026 research pass:

# one row per game summary (trailing 30 days)
entity_type      : game
name             : Just Chatting
rank             : 1
avg_viewers      : 283193
avg_channels     : 4458
hours_watched    : 47670747

# one row per channel summary (trailing 30 days)
entity_type      : channel
name             : caedrel
rank             : 5
avg_viewers      : 47514
max_viewers      : 113208
minutes_streamed : 12098
hours_watched    : 9580448
followers        : 46807
followers_total  : 1665103

# one row per weekly game ranking
game             : Just Chatting
avg_viewers_7d   : 283193
twitch_share     : 13.177%
change_7d        : +1.9%
game             : Dota 2
avg_viewers_7d   : 135849

# platform-wide monthly aggregate
scope               : Twitch total
hours_watched_month : 1530544896
peak_month          : 2021-05
peak_hours_watched  : 2307473592

(Sample observations from the August 2026 research pass; fuller cuts ship with your sample.) Three things worth reading off them. First, the grain is exact — one entity, one window, one set of aggregates — so a game-level or channel-level series is a filter away, not a reshaping project. Second, concentration is the story: Just Chatting alone held 13.177% of the platform's whole audience at 283,193 average viewers, nearly double Dota 2's 135,849, with League of Legends further back at 99,514 — a hierarchy any category-sizing model needs as a series, not a screenshot. Third, peaks and averages diverge in ways means alone hide: caedrel averaged 47,514 concurrent viewers across 12,098 broadcast minutes yet spiked to 113,208 at peak, converting 46,807 new followers toward 1,665,103 overall.

Which fields does the field dictionary define?

Twelve fields ride on every record, all defined below with examples taken from captured observations and marked verified against the live card during the August 2026 pass. Two of them — the ranked-list value and the calendar-month key — arrive from the rendered history tables rather than the summary aggregates, so their worked examples get pinned once your sample scopes which tables your pipeline keeps.

How wide does coverage run, and at what grain?

Three chips summarize the footprint:

  • Geography: global — every language community on the platform rolls into the same leaderboards, with language-level breakdowns available as a separate slice.
  • Temporal: trailing seven-day leaderboard windows, trailing thirty-day summary aggregates, and per-entity monthly histories that stretch back years — far enough to contain the platform's all-time peak month of May 2021.
  • Granularity: per game and per channel at day and month steps, plus platform-wide monthly aggregates for sizing the entire market.

Context for the collectors: this record scores 7/10 on Datadory's quality rubric against a catalogue average of 7.81 — a mark shared by 296 of the 1,744 datasets catalogued. Set it against the rest of the shelf in our best interactive home entertainment datasets ranking.

How is the data delivered through Datadory?

API, files, or your warehouse. Daily, weekly, or hourly.

Pick the channel your stack already speaks and set the cadence to match the decision being fed — bulk files for overnight warehouse loads, structured payloads for event-driven pipelines, spreadsheet-friendly extracts for analysts living in a BI tool. Leaderboard-shaped data rewards a faster drumbeat than most feeds because a stale board misleads quietly: hourly pulls suit trend desks watching which titles break out, daily loads fit competitive-intelligence briefs, weekly cuts cover quarterly category reviews. The monthly histories argue for the opposite discipline — delivered through one unchanging schema, each vintage appends cleanly to the per-game and per-channel spines instead of reshaping them, so joins on entity and month survive untouched. Pin a frozen cut whenever you need a point-in-time reconstruction of what the board looked like the day a title launched.

Who builds on this collection?

Four jobs these standings do that a screenshot cannot.

Track which games own the culture, week by week. Average viewers, share of the platform's total audience and week-over-week change turn "everyone is playing it" into a measurable claim — Just Chatting at 13.177% share against Dota 2's 135,849 and League of Legends' 99,514 average viewers is a market-structure statement, not a vibe.

Benchmark creators and esports organisations. Peak versus average viewers, minutes broadcast and follower velocity separate event-driven audiences from sustainable ones — the caedrel row above doubles its mean at peak, which reads very differently in a talent valuation than a flat high-average channel.

Backtest engagement against release and tournament calendars. Monthly histories reaching back years put the May 2021 all-time peak of 2,307,473,592 hours watched on the same axis as today's 1,530,544,896-hour months, so launch effects and esports seasons can be measured against the long baseline.

Size the live-streaming market itself. Platform-wide monthly watch-time aggregates give investors and analysts a single defensible denominator for share-of-attention math across the whole industry shelf.

Which personas get the most value?

Market Researchers & Consultants size the live-audience market and benchmark individual titles and talent with share-of-platform figures instead of anecdote. See market researchers.

Competitive Intelligence & Product Teams watch which games and channels are compounding week over week — follower gain against minutes broadcast is the cheapest early-warning signal on the shelf. See competitive intelligence & product teams.

Journalists, Academics & Students cite audience numbers with a verifiable trail — ranks, windows and aggregates that can be re-checked line by line rather than eyeballed off a chart. See journalists & academics.

Investors & Quants treat attention as the leading indicator for the games business, pairing platform watch-time aggregates with sales and concurrency series to model where spend follows eyeballs. See investors & quants.

Which notes pair with this dataset?

  • Steam Charts - Concurrent Players — storefront-side concurrency for the same titles; the two make the natural PC-market pair (head-to-head).
  • RAWG Video Games Database API — metadata, ratings and store availability for the games climbing these boards, so viewership joins to catalog context on game identity (record).
  • VGChartz - Video Game Sales Charts — the purchase side of the ledger, useful when testing whether watch time leads or lags unit sales (record).
  • itch.io Indie Games Browse — the long tail of independently released titles whose breakout moments first show up as audience moves here (record).
  • Twitch Developers — the platform's own first-party statistics surface, a useful cross-check on the aggregates this feed repackages (profile).
  • Interactive Home Entertainment hub — every dataset in the industry, this one included (hub).

Field dictionary

Every field below is documented against real records. The full dictionary ships with the sample.

Field dictionary — the twelve fields verified against captured records during the August 2026 research pass (examples from observed rows)
fieldtypedefinitionexample
rankintegerPosition within the window — trailing-thirty-day standing on summary records, weekly position on game leaderboards.1
avg_viewersintegerAverage concurrent viewers over the window: trailing thirty days on summary aggregates, last seven days on weekly game rankings.283193
avg_channelsintegerAverage number of concurrently live channels broadcasting the game.4458
max_viewersintegerPeak concurrent viewers reached within the window.113208
hours_watchedintegerTotal viewer-hours accumulated across the window.47670747
minutes_streamedintegerTotal minutes the channel broadcast during the window.12098
followersintegerFollowers gained by the channel within the window.46807
followers_totalintegerCumulative follower count for the channel.1665103
Twitch SharestringPercentage of the platform's total audience represented by the game's viewers.13.177%
7-day ChangestringPercentage movement in average viewers versus the prior week.+1.9%
ri-valueintegerRanked-list value carried on rendered history tables — average viewers for games, concurrent viewers for live channels.-
MonthstringCalendar-month row key in per-game and per-channel history tables.-

Questions buyers ask

What does one record of TwitchTracker game and channel rankings data contain?

One row per entity-window observation: the game or channel, its rank, average concurrent viewers, average live channels or peak viewers depending on entity type, hours watched, minutes streamed on channel rows, followers gained and total followers, plus share-of-platform and week-over-week percentage change on game rows. Twelve fields carry the full payload.

How far back does the monthly history reach?

Per-game and per-channel monthly histories extend back years — deep enough to contain the platform's all-time peak month of May 2021, when watch time hit 2,307,473,592 hours against 1,530,544,896 in the most recent complete month. Scope the span you need and the sample confirms exactly how far a given entity's series runs.

Can the data separate a breakout spike from sustained viewership?

Yes — peak and average sit side by side on every channel record. Caedrel averaged 47,514 concurrent viewers across 12,098 broadcast minutes in the observed window but peaked at 113,208, more than double the mean, which is the signature of event-driven audiences rather than a flat daily schedule.

What can I combine this dataset with?

Storefront concurrency from Steam Charts is the natural counterpart — we maintain a head-to-head comparison of the two — and sales charts, indie storefront catalogs and speedrun records extend the picture from who watched to who bought, listed and finished. All sit on the same interactive home entertainment shelf.

Can I get a scoped sample before committing?

Yes. Name the games, channels and window lengths you model and we cut the sample to match, including the monthly history depth behind each entity. Samples ship in the same twelve-field schema as the full feed, so nothing gets reshaped between evaluation and production.

How current are the numbers on this page?

They are photographs from the August 2026 research pass: the Just Chatting, caedrel and platform-wide figures quoted here were verified then. Delivered rows carry their own window stamps rather than inheriting this page's date, so your models always know how old a figure is.

See the rows before you pay anything.

Name this dataset and we send real records from it — scoped to the fields you asked for.

See pricing