← Leaderboard

Methodology & API

How RealGitRank turns public GitHub events into an engineering impact index — and where to read the data yourself.

The score

Rank Score = (Delivery × 0.35 + Code Health × 0.25 + Adoption × 0.25 + Consistency × 0.15) × (1 − BotPenalty)

Every pillar is computed from GitHub public events (pushes, merged pull requests, issues, stars, forks) over the ingest window, then combined into a 0–1000 index. Commit count is never a pillar: it measures typing speed, not impact. Percentile is reported against the whole ranked population, and the final score is rounded to one decimal.

The four pillars

PillarWeightWhat it measuresWhy it is hard to game
Delivery 35% Merged pull requests in owned and external repositories. External PRs merged by other maintainers count more; self-merged PRs are discounted. Opening PRs is free, merging is not. Only reviewed, merged code counts.
Code health 25% Test-file ratio, multi-file commit ratio, language diversity. One-line churn and empty markdown edits score close to zero.
Adoption 25% Stars, forks and downstream dependents after fraud filtering. Star-cartel rings are filtered out before scoring.
Consistency 15% Active shipping weeks and the variance of push hours. Bots that push around the clock are penalised instead of rewarded.

Pillar sub-scores (s_pr, s_quality, s_impact, s_consistency) appear in the API only when the pipeline produced them for that day; carried-over users keep their blended score without fresh pillar values.

The bot penalty

Each account gets a penalty between 0 and 1, built from behavioural signals in its push stream:

The penalty multiplies the final score. Accounts at 0.70 or higher are classified as bots; accounts below 0.05 are marked Verified Human Engineer. Human engineers dominate the main board because automated commit farms lose most of their score.

Tiers

TierPercentile
Elite 99th percentile and above
Diamond 95th percentile
Platinum 90th percentile
Gold 75th percentile
Silver 50th percentile
Bronze everyone else

A tier is a relative band of the current ranked population, so it moves as the population grows. Tier and percentile are recomputed on every daily run.

Daily snapshots, movement and decay

Data sources and freshness

A scheduled GitHub Actions workflow (daily, ~02:00 UTC) ingests the previous day of public GH Archive events into DuckDB, scores every account, runs data-quality gates, then rebuilds and deploys this static site on GitHub Pages. Enrichment (avatar, country, company, language) comes from the GitHub API when a token is configured. Everything is free, static and versioned in git — there is no database and no runtime server.

Known limitation: code-health and adoption signals depend on the depth of the ingest window, so they can be missing for low-activity accounts. Missing metrics lower a score; they never invent one.

Static JSON API

All endpoints are pre-built files on GitHub Pages, so they are cacheable, CORS-free for normal fetches, and cost nothing to serve.

Example record

{
  "username": "sirine2907",
  "rank": 1,
  "rank_score": 128.9,
  "tier": "Elite",
  "percentile": 100.0,
  "country": "United States",
  "primary_language": "Dart",
  "merged_prs": 17,
  "is_human_verified": true,
  "is_bot_classified": false,
  "bot_penalty": 0.0,
  "profile_url": "https://realgitrank.bharatdn.com/profile/sirine2907",
  "badge_url": "https://realgitrank.bharatdn.com/badges/sirine2907.svg",
  "api_url": "https://realgitrank.bharatdn.com/api/user/sirine2907.json"
}

Field names follow the pipeline schema; optional fields are omitted rather than set to null. Please link back to realgitrank.bharatdn.com when you publish a rank or score.

Related