Methodology
UmaTierList stores each entity once and connects related records by stable identity. Trainees, Support Cards and Skills do not gain strength labels from rarity, stars, Potential, level or uncaps alone. When we publish an evaluation, it must name its context and may not be inherited silently from another Scenario.
We distinguish:
Canonical fact: a normalized value supported by identified evidence.
Localization or alias: another label for the same concept, kept separate from the structured key.
Editorial recommendation: a human assessment with a stated context and review date.
Hypothesis: an interpretation that remains unverified and is not presented as fact.
Conflict: sources disagree or a record cannot be reconciled without further evidence.
Missing, locked, unavailable, unknown and zero are different states. Pending proposals are isolated from the public read model until a human review approves them.
Review and corrections
Records retain source and evidence references where available. We prefer leaving a field incomplete to filling it by inference. When a source changes or a conflict is resolved, the record should receive a new review date and preserve the reason for the correction.
Material corrections should identify the affected entity, field, previous state, evidence used and date of change. Editorial rankings should identify their owner, method, context and revision date.
Sources
Our source hierarchy is:
- Official Global/English game interfaces, announcements and publisher material.
- Official material from another region, clearly labelled and never used to imply Global availability by itself.
- Established public English-language references used for verification and contrast.
- Other public sources used only when their limitations are documented.
Every important factual claim should have identifiable evidence. Conflicts remain visible rather than being resolved silently. A source link is evidence for a statement; it is not an endorsement of the external site.