NFT Rarity Scoring Tool – Turnkey Development

You launched a collection, tokens were sold out, but holders doubt the fairness of the rarity ranking. The reason is an inappropriate rarity calculation algorithm. We've helped dozens of projects avoid this issue by building tools that account for each collection's specifics. Our approach combines multiple methodologies with flexible weight tuning and aggressive caching. Our team has 5+ years of experience in Web3 and has delivered over 15 NFT analysis tools—a guarantee of quality. Starting from $5,000, our tool saves up to $20,000 compared to in-house development.

Why a Single Algorithm Doesn't Fit All Collections

Trait Rarity Score is the basic method: sum of inverse probabilities of traits. It's simple but overestimates tokens with many attributes. Statistical Rarity (product of probabilities) is fairer but suppresses singular unique traits. Information Content (−log2(p)) yields a balanced distribution but requires normalization.

We use a combination of Trait Normalization (OpenRarity) and Information Content—this reduces score spread by 2.5× compared to the basic method. In one project with a 10,000-token generative avatar collection, our combined algorithm reduced score variance by 40% relative to Trait Rarity Score, and the entire metadata indexing completed in under 30 seconds. For each collection, we select the optimal algorithm based on its statistics.

OpenRarity (official documentation) recommends combining normalization and information content methods for the most balanced ranking.

How We Collect and Index Metadata

NFT metadata is stored on IPFS or a centralized server via tokenURI. We read contractURI and tokenURI(0..n) through Multicall3—up to 500 RPC calls in a single batch. For IPFS, we use Cloudflare gateways and resolve CIDs. We handle edge cases: null traits (converted to "None" attribute), numeric values (binning), and updates after reveal.

We build the index in PostgreSQL: table token_traits(collection_id, token_id, trait_type, trait_value) plus aggregates per trait. After metadata updates, we recalculate incrementally.

How We Assess Rarity in 5 Steps

1. Analyze collection structure—determine trait types, distributions, identify anomalies.
2. Select algorithm—choose a combination of Trait Normalization and Information Content (or another if needed).
3. Fetch metadata—via Multicall3 batches of 500 tokens, IPFS gateways.
4. Compute scores—apply selected algorithms, calculate rank and percentile for each token.
5. Deploy API and frontend—integrate with marketplace, publish documentation.

How Long Does Development Take?

A typical project takes 2 to 4 weeks. The timeline depends on collection size, number of supported blockchains, and required integrations. The first calculation for a 10k-token collection takes seconds; subsequent ones are milliseconds.

Comparison of Rarity Algorithms

Common Mistakes in Self-Implementation

• Using only one algorithm—leads to biased ranking.
• Ignoring null traits—distorts statistics, tokens without attributes get incorrect rank.
• No caching—each IPFS or RPC request takes up to 500 ms; without cache, a 10k collection takes minutes.
• Not accounting for reveal—if metadata updates after sale, the index must be recalculated incrementally.
• Self-implementation can waste up to 50% of budget on fixes. Our tool cuts development costs by 3× using ready-made modules.

What's Included in the Work

• API with endpoints /rarity/{contract}/{tokenId} and algorithm selection.
• Web interface with rarity table and token percentile.
• Integration with your marketplace via OpenRarity npm package or custom solution.
• Documentation and source code access.
• Team training on the tool.
• Deployment on your infrastructure or cloud.
• One month of post-launch support.

Example Display