EVIDENCE STANDARD

How This Guide Decides What It Can Claim

Every useful claim needs a source boundary. This page explains the labels used across the tracker, analyzer, and guides.

Evidence levels

Verified

Confirmed against a primary source or a complete, reviewable run.

Observed

Seen directly in the live interface on the stated date.

Community

Reported by a player but not yet independently reproduced.

Unknown

Not established by the available evidence.

Run review threshold

A public run needs its role, mode, complete pick chain, visible grades, final result, observation date, and editor review. A partial screenshot is not a complete run.

Analyzer heuristics

The analyzer counts complete player-and-trait pairs, lists missing slots, compares only recognized letter grades, and shows whether a final OVR was recorded. It does not estimate hidden weights.

Dates and corrections

“Observed” dates refer to the underlying game evidence. Publication dates refer to this site. Corrections are recorded in the change ledger.

Machine-readable evidence

The current Guard Classic baseline is published as versioned JSON with its scope, source, observation date, slot labels, and unresolved claims. A future correction must change the version or review date and appear in the ledger.

Download baseline JSON

What we will not infer

  • A universal best build from one successful run.
  • An official formula from correlation alone.
  • Role parity from a different role's board.
  • Current rules from an undated screenshot.