Metrics and Interpretation

DevHunt's formulas, inputs, sample-size behavior, and why these descriptive heuristics have limits.

Language weights

Two implementations exist.

Primary-language distribution skips repositories that have no primary language. Each remaining repository gets a weight of log(1 + stars) when that value is positive, falling back to 1 when stars are zero. Repository weights are then summed by primary language and the language totals are normalized.

Byte-weighted V2 multiplies each language's byte count by log(1 + stars). It skips a repository when log(1 + stars) is zero, when language data is absent, or when the repository's total language bytes are zero. The remaining weighted bytes are normalized.

Both are repository summaries, not proficiency scores.

Focus Ratio

Focus Ratio aggregates raw language bytes across all repositories, then divides the largest language total by the sum of all observed language bytes.

sample_size is the total counted bytes.
When there are no language bytes, the value is null and the sample size is zero.

Activity proxies

The hours histogram bins pull-request creation timestamps after applying a timezone offset.

Night Ratio excludes merge commits and invalid author timestamps, applies the selected timezone, then computes:

night included commits / all included commits

for local hours 22, 23, 0, 1, 2, 3, and 4. Its sample_size is all included commits. When there are none, the value is null and the sample size is zero.

These timestamps are activity proxies. They are not work schedules or health evidence.

Collaboration ratios

UOI classifies a pull request as self-authored when the login and the repository owner match case-insensitively. It then divides the external pull-request count by external plus self pull requests. When that denominator is zero, it returns null. The generated profile records uoi_sample_size as all observed pull requests.

External acceptance divides merged external pull requests by all external pull requests. When there are no external pull requests, it returns null. external_pr_sample_size is the external pull-request count.

Grit Factor

This project-defined ratio considers only non-fork repositories owned by the requested login.

Lifecycle days are calculated as floor(max(pushedAt - createdAt, 0) / 1 day). The expression pushedAt ?? createdAt falls back to createdAt only when pushedAt is null or missing. Invalid or negative timestamp differences clamp to zero.

Three categories:

long-term — lifecycle of 90 days or more
gem — shorter-lived, but at least 5 stars
churn — everything else

The value is (long-term + gem) / included repositories. When there are no included repositories, the value is null and sample_size is zero.

Interpretation boundary

Always present the formula, inputs, sample size, and unknown-value behavior together. None of these heuristics validates skill, personality, intelligence, employability, reliability, or objective ability, and this documentation publishes no individual result values.