Methodology v0.1

Every published number is calculated from stored test results under the declared weights below. Commercial factors, portfolio membership and editorial preference are not inputs to any score, to eligibility, to retest priority or to a correction outcome.

Category weights

Overall
Input	Weight
Relationship	25%
Roleplay	20%
Memory	20%
Product and local language	20%
Value	15%

Relationship
Input	Weight
Conversation	40%
Continuity	25%
Emotional response	20%
Boundaries recovery	15%

Roleplay
Input	Weight
Character fidelity	35%
Narrative continuity	25%
Initiative	20%
User control recovery	20%

Memory
Input	Weight
Delayed recall	40%
Contradiction resistance	25%
Relationship continuity	25%
Memory controls	10%

Value
Input	Weight
Usable free experience	20%
Paid experience per cost	45%
Paywall friction	20%
Local price stability	15%

Evidence minimums

An app is rank-eligible only with a complete minimum evidence bundle: all 18 declared criteria observed under methodology v0.1, a current price observation for the edition, and a verified official destination. An incomplete bundle means the product stays outside the published roster — there is no placeholder position.

Tie-breakers

Equal calculated values display the same score while positions stay consecutive, resolved in this declared order:

evidence-completeness
recency
alphabetic

Stale evidence and retests

Prices and policy-sensitive facts are rechecked monthly; full scores are retested on a schedule and after a material release. Evidence that ages past the published threshold is marked stale, the app is shown as retest due, and it leaves the ranking until the retest completes. A stale price keeps its exact observation date visible rather than being quietly refreshed.

Version history

Version	Status	Note
v0.1	Active	First published calibration. Changing any weight creates a new version and recalculates from stored results; historical scores keep the version they were published under.