Methodology

Methodology

How the daily snapshot is collected, how each metric is defined, and where the limits are.

Pipeline

1. A collection script pulls public feeds once a day: Steam's charts and store APIs, Hugging Face's model API, benchmark snapshots published by third parties, Liquipedia's tournament page and RSS news feeds.
2. Raw responses are stored as a dated snapshot and written to a local database.
3. The site is generated from that snapshot: leaderboard pages, one page per tracked entity, a daily brief and open data exports.
4. Numbers in the generated pages come from the snapshot; the build fails rather than inventing a missing value.

Definitions

• Peak players — the highest concurrent-player count Steam reports for a game in the latest daily window. It is not a live count and it is not total owners or daily active users.
• Weekly rank change — current rank minus last week's rank on the same chart. Positive means the game moved up.
• Hugging Face likes / downloads — lifetime counters reported by the model page at snapshot time. Downloads are cumulative and can include evaluation and CI traffic.
• Benchmark score — a figure published by the benchmark owner on a specific date, reproduced with a link. Scores from different boards are not comparable.
• Series and game record — counted only from completed matches with both teams known, using the winner field from the bracket.
• Timestamps — snapshots are stored in UTC. Match times are shown in UTC and Beijing time (UTC+8) because the source schedule is reported in Beijing time.

Update cadence

The snapshot is collected daily; this build used the November 7, 2026 snapshot. Benchmark boards update less often, so each board shows its own "as of" date. Entity history starts with the first stored snapshot — there is no backfill, and the site says so on pages without history.

Known limitations

• Steam's public chart exposes a limited top list, not every game.
• Benchmark figures come from published pages, not a first-party API; the note on the AI page records this.
• Bracket data can lag official announcements and does not include odds or in-game statistics.
• News headlines are shown for context only; the linked publisher owns the article.

AI crawlers and reuse

The site is server-rendered HTML. Crawlers for search and AI answers — including GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Bingbot — are allowed in robots.txt. Google-Extended is also allowed so the content can be used for Gemini grounding; AI Overviews use regular Googlebot and cannot be opted out of separately.

The JSON and CSV exports are free to reuse with attribution to the page and the original source. See open data.

Corrections

If a figure is wrong, send the page URL, the correct value and a source link via the contact page. Corrections are applied to the next daily build.