The explosion of profile picture (PFP) projects and large-scale art drops has popularized NFTs, but the true mechanism behind their perceived value often remains mysterious to newcomers. Most of these popular collections (like CryptoPunks or Bored Apes) are generative. They are not individually drawn; they are programmatically assembled from hundreds of distinct traits.

The Generative Process

1. ​Trait Layering: An artist creates multiple versions of component layers: backgrounds, bodies, clothes, eyes, hats, etc. (e.g., 50 types of hats, 10 types of backgrounds).
2. ​Rarity Weights: The artist or developer assigns a specific weight or probability to each trait. For instance, a "Golden Laser Eyes" trait might be assigned a 0.01% chance of appearing, while a "Blue Background" might have a 10% chance.
3. ​Algorithmic Assembly: A computer script runs through these layers and weights, randomly selecting and stacking traits to generate the final collection (e.g., 10,000 unique images).

The Role of Metadata and Rarity Tools

​The generated image itself is only half the story. The core of the NFT's verifiable uniqueness and rarity lies in its metadata, typically stored in a JSON file off-chain (IPFS) and referenced by the token. 

- Metadata Structure: This file explicitly lists every trait the NFT possesses:

{

  "trait_type": "Background",

  "value": "Cosmic Void"

},

{

  "trait_type": "Eyes",

  "value": "Golden Laser" // Rarest Trait

}

- Rarity Scoring: Third-party rarity tools (like Rarity Tools or icy.tools) analyze the entire collection's metadata. They calculate the percentage frequency of every single trait across the collection.

• ​Calculation: If a "Golden Laser Eyes" trait appears 10 times in a 10,000-piece collection, its rarity is 10/10000 = 0.001, or 0.1%.
• ​Score Aggregation: These tools use various methodologies (like Trait Normalization or Statistical Rarity) to combine the rarity of all traits on a single token into a single Rarity Score. A higher score equals a rarer NFT.

Therefore, an NFT's market value often correlates directly with its objective Rarity Score, which is ultimately derived from the weighted probability tables set by the creators and verifiable through its on-chain metadata.