NFTsRank.com Methodology
Versioned principles for collection identity, peer groups, market metrics, normalization, component scoring, confidence, and responsible presentation.
Scope of this methodology
This static edition documents how an NFT ranking or research product can be designed. It does not claim to operate a current market feed or assign live positions to collections. The framework is intended for educational comparison of NFT collection data across supported chains and categories.
1. Collection identity
Use chain-native identifiers as the canonical key. Names, symbols, marketplace slugs, and image assets are descriptive fields. Multi-contract groupings require a documented relationship and source. Bridged, wrapped, migrated, and derivative assets should remain distinct unless a verified method justifies aggregation.
2. Eligible universe and peer groups
Define supported chains, standards, collection types, minimum age, data coverage, and activity requirements before scoring. Create peer groups for generative collectibles, art and editions, gaming assets, membership or access, digital fashion, and unique works where appropriate. A global view may summarize peer-group results but must preserve category and chain context.
3. Observation windows
Use multiple windows. Seven days can describe current momentum, thirty and ninety days can describe more persistent activity, and lifetime records can provide history. Never present a change percentage without the starting base. Record the data cutoff and source update times.
4. Reference components
| Component | Reference weight | Examples |
|---|---|---|
| Liquidity | 25% | Sales frequency, listing depth, offers, spread, time-to-sale |
| Demand breadth | 20% | Unique buyers and sellers, active days, repeat participation |
| Ownership | 15% | Distribution, concentration, contract-held supply, change |
| Data integrity | 15% | Coverage, freshness, mappings, filters, reproducibility |
| Technical quality | 15% | Contract controls, metadata, storage, supply authority |
| Operating history | 10% | Continuity, rights, incidents, governance, communication |
These weights are illustrative. A liquidity leaderboard should use a different structure from a technical-quality screen. Material weight changes require a new methodology version.
5. Metric normalization
Normalize metrics within relevant peer groups. Percentile ranks are readable, while log transforms can reduce the dominance of extremely large values. Cap or winsorize outliers only under a published rule. Preserve raw values beside normalized scores.
6. Suspicious activity and exclusions
Potential filters include duplicate events, zero-value transfers, bundle allocation problems, self-transfers, rapid round trips, repeated wallet pairs, prices far from nearby observations, and invalid orders. No heuristic establishes intent. Publish gross and adjusted data when possible, and preserve exclusion reasons.
7. Confidence
Confidence is separate from the score. It reflects data coverage, freshness, sample size, mapping certainty, metadata completeness, and unresolved anomalies. Suggested labels are high, moderate, low, and insufficient data. A high score with low confidence should not be displayed as a strong conclusion.
8. Rarity
Rarity is calculated only within a defined collection and metadata snapshot. Publish the formula, eligible supply, missing-trait rule, trait-count treatment, and observation time. Do not include rarity directly as a market-value signal unless the methodology separately establishes a demand relationship.
9. Publication and corrections
Every output should include methodology version, observation time, source coverage, and material limitations. Corrections should preserve a record of the prior output and explain what changed. Historical results should remain associated with the methodology used at the time.
10. Interpretation
A ranking is a research queue, not a recommendation. It cannot establish legal ownership of media, guarantee contract security, predict returns, or determine suitability. Readers should verify current chain and market data and consult qualified professionals for legal, tax, security, or financial decisions.
Sensitivity testing
Ask whether small choices change the story
Recalculate with reasonable alternative weights, time windows, provider coverage, and outlier rules. A fragile rank should be shown as a broad tier with lower confidence.
Reproducibility test
An independent analyst using the same source inputs and rules should be able to produce a close result. Differences should be explainable through timing, coverage, or implementation—not hidden discretion.