Skip to content
Anime26

Methodology & data quality

Anime26 treats methodology as a feature. Every enriched value is either backed by evidence or left as an honest Unknown. This page documents what is measured, how, and where the limits are.

mode culturalupdated May 26, 202660 titleshash 26333736521631c1Audit passed

What Anime26 measures

The dataset separates title-level and character-level metrics and never conflates them. It can answer, distinctly: what share of action anime are female-led (a title-level metric), and what share of MAIN characters in action anime are female (a character-level metric). These have different denominators and are reported separately throughout the site.

AniList and Jikan/MAL source roles

AniList is the structured spine: ids, titles, season/year, format, genres, tags with rank, studios (with isAnimationStudio), staff roles, and characters with a structured gender and age. Jikan/MAL provides continuity and popularity: score, members, favorites, genres, themes and demographics. Public web pages are used only where structured sources are insufficient, always cited.

Identity resolution

AniList and MAL records are resolved to one internal id per title via structured id mappings and fuzzy title matching. AniList and MAL popularity are kept on their native, non-equivalent scales and are never combined into a single synthetic number.

Protagonist classification

Protagonist structure (Female-led, Male-led, Mixed-led, Ensemble) is computed from AniList's structured MAIN-character gender field, corroborated by Male/Female Protagonist tags. Character gender is taken from the structured field — never inferred from a name.

Adult-cast classification

Cast age (Adult Cast, Mostly Adult Cast, Mixed Age Cast, Youth Cast) comes from AniList cast tags and structured character ages — never inferred from demographic alone. A seinen title can star teenagers; a shoujo title can star adults, and the data reflects that.

Settings, themes and studios

Settings and themes are mapped from AniList tags (weighted by rank) and MAL genres/themes through a conservative controlled ontology; ambiguous tags stay Unknown. Studio aggregates count animation studios only — producers and licensors are excluded.

Confidence bands

Each classification carries a confidence score. Values below a field's acceptance threshold are shown as Unknown rather than as low-confidence conclusions.

UI labelDataset range
Verified0.95–1.00
High confidence0.85–0.94
Supported0.70–0.84
Needs review0.50–0.69

Data coverage

Known vs. Unknown by field. Unknown is surfaced as a count and excluded from percentage denominators — never hidden.

FieldKnownUnknownTotalKnown %
Protagonist class5646093.3%
Adult-cast class36246060.0%
Primary setting50106083.3%
Primary themes60060100.0%
MAIN-character gender133613995.7%
Evidence coverage
100.0%
92/92 classifications
Verified
15
classifications
High confidence
73
classifications

Audit summary

Audit passed

Checked Jul 24, 2026 · 0 errors · 0 warnings · publication is blocked unless the audit passes.

Duplicate AniList ids
0
must be 0
Duplicate MAL ids
0
must be 0
Contradictions
0
cross-source
Manual-review queue
42
flagged rows

Checks

  • Duplicate identifier check
    0 duplicate AniList ids, 0 duplicate MAL ids
    pass
  • Evidence-backed classifications
    0 non-Unknown classifications lack evidence metadata
    pass

Known limitations

  • The shipped snapshot is materialized in cultural mode (top-N per season) for tractability.
  • Structured character gender/age thins for older or niche titles, raising Unknown rates.
  • Setting/theme ontologies are conservative — ambiguous tags stay Unknown rather than guessed.
  • Sequel detection is a title-pattern heuristic; franchise graphs are not fully materialized.
  • AniList and MAL popularity are not equivalent and are kept on their native scales.