Social signals behind the reputation of NFTs - An invitation to create open NFT rankings with Karma3 Labs

NFT Collection rankings based on social consensus, using on-chain data and verifiable compute

The abundance of NFTs came with an abundance of noise

The Defi summer in 2020 led to an exponential growth of on-chain transactions, enabling a transparent lens for crypto and web3 community to understand trends and behaviors, providing useful insights for decision-making.

The explosion of NFTs later in the summer opened doors for millions of new users to experience on-chain ownership and access. Today there are 111M NFTs on Open Sea alone. According to NFTScan’s latest data, there are around 5000 NFT smart contracts deployed and 2 million NFTs minted every day.

NFT search and discovery are Painful

The abundance of NFTs comes with an information overload for users and adds chaos in decision-making, amidst the fakes, frauds and copy cats.

Some NFT marketplaces have started to run a basic identity check for NFT collectors and collections. They try to aggregate off-chain social data (Twitter, discord, website) along with on-chain data (wallet, ENS, etc). OpenSea introduced verified collection blue check marks to provide safety assurance to users through hand-picking of relatively trustworthy NFTs. This has been a good first step to enabling trust and safety for onboarding users. But it is hard to scale and may not be sufficient to enable a resilient identity and reputation system for NFT transactions.

The context of crypto supports pseudonymity and permission-less creators and communities, so it is hard to create a system for trust, safety and fraud prevention. A broken search and discovery experience for NFTs also discourages the next wave of millions of users to participate in this on-chain economy.

Lack of trust and reputation signals for NFTs

Most NFT marketplaces currently rely on volume data as an approximation for ranking or displaying an NFT collection. Search and discovery tabs display NFTs that are trending based on volume data, or are hand-picked by marketplaces for a variety of reasons, which may not always be public and transparent. How a collection makes it to the Popular or Trending lists may remain a subjective black-box decision.

The perceived value of an NFT is driven by Social Capital?

Price discovery of digital collectibles and NFTs has been challenging so far. A collection’s floor price has been used to approximate the value of an NFT. NFT prices can be extremely volatile, yet and the perceived blue chip NFTs experience relatively less of price fluctuations.

The communal nature and social discovery of NFTs of value

As we examine the behaviors of people when it comes to purchasing NFTs, more often than not, they trust the NFTs being collected by social circles. The friends, the communities of projects or interest groups, the recommendations from trusted peers, and the collectors of fame who they follow — are the most important references for a reputable NFT collection.

  • A key insight — NFT marketplaces realize that today users do not necessarily come for the discovery of new NFTs, instead, users discover NFTs from their social circles in Discord or Twitter, therefore when they come to marketplaces to look for a collection, they already come with intent. Such social circle discovery and recommendation signals are hard to capture for marketplaces. If they did, it would help make their platforms the go-to place for first-time search and discovery of NFTs.

Our Thesis: compute the reputation of NFTs based on peer-to-peer trust signals

We believe that the value of a digital collectible such as an NFT is a function of social consent among the collectors or holders. People collect an NFT because of many reasons: they like the NFT from a design point of view based on personal tastes; they like the creators/artists that they have been a fan; they see value in the associated utility of the NFT; they trust the social media trends as indication of the popularity of the NFTs; they assess that the NFT can be a good investment based on trading metrics.

The reputation of NFTs is the result of the social consent of your trusted circles of people or communities. Such examples range from the initial group of owners of Cryptopunks and BAYC, to the introduction of FWB social camps and NFTs, to the surging popularity of LOOT as it exploded on our social media feeds over a few days only. NFTs are not alone reflections of social consent, we also see the same heuristic in the collection of art pieces, swags of popular IPs, and unique artifacts.

Launching the Verifiable Compute tool for NFT Ranking

We are announcing the launch of, a people’s choice ranking of NFT collections on Ethereum. It’s a tool for NFT marketplaces, NFT social projects, data wizards, and enthusiasts to use for creating their own ranking systems for better NFT discovery and recommendation. And we have created a Dune dashboard as an example of analysis as well.

Our demo ranking - NFT reputation based on "smart and reputable buyers" behind blue chips

We analyzed NFT transaction data on Ethereum and applied the EigenTrust algorithm to compute a trust or reputation score of NFT collections based on social consensus. The scores generated from EigenTrust compute have been used to display a ranking of NFTs.

We started with a small set of 20 blue chip NFTs - popular and well-acknowledged collections - (BoredApeYachtClub, BoredApeKennelClub, MutantApeYachtClub, Otherdeed, Moonbirds, Azuki, Art Blocks, Doodles, PudgyPenguins, Meebits, Wrapped Cryptopunks, Cool Cats, Cryptoadz, CloneX, Loot, Sorare, World of Women, Nouns, VeeFriends, mfer), and traverse the NFT ownership graph to measure social consensus on all NFT collections on Ethereum.

NFTs that are ranked high on the list are among the most reputable NFT collections as evidenced by social consensus among NFT holders. The meaning of social consensus is outlined in the section below.

Some principles for Karma3’s NFT collection reputation score

  • NFT sales are attestations of true economic activity.

  • The reputation signal between any two collections is a derivative of the cohort of owners’ transitive purchases of other collections.

  • Lindy Effect matters for reputation.

Based on the above principles, we extracted and filtered on-chain data using a variety of sources like Dune, NFTScan, SimpleHash, OpenSea, and NFTPort:

  • 24 months of NFT transactions on Ethereum Mainnet between April 1, 2021, and Mar 23, 2023

  • Filter out all wash trades (as defined by the team at Dune)

  • Filter out all NFT transactions with zero ETH value (Airdrops and Free mints)

  • Filter out all transfers to 0-addresses i.e., filter out NFT burns (price pumps)

Once we extracted and filtered on-chain data, we computed three different EigenTrust-based rankings with different assumptions — any developer can make their own ranking system.

Next step - Personalized recommendation of NFTs

Evidently, the choice of our 20 Blue Chip NFTs and their associated ownership is just one of the many choices to seed the compute. The choice can be customized to any particular community or private individual based on their trusted collections, collectors or traders. The power of our infrastructure is eventually able to power personalization of NFT discovery for everyone based on their preferences, tastes, interests, and trusts of their respective social circles.

For learning more about how the EigenTrust computation runs, feel free to check out our docs.

An open reputation layer - Contribute, propose, and use rankings

This ranking is an initial version with assumptions and principles on what data to use and how to choose the weights in the EigenTrust algorithm. We did the prototype to demonstrate how easy is it to modify these assumptions and enable a sort of marketplace for different ranking and reputation algorithms, suited for a community or developer’s needs.

We invite the NFT community to explore the rankings and propose any alternate strategies to rank NFTs. Our goal is to have an open-source, distributed, and decentralized compute model that different communities can tune to best serve their specific needs. For example, traders may weigh blue-chip statistics higher when computing social consensus whereas collectors may weigh holding periods higher when computing social consensus. The implementation will allow for parameterizing in a transparent and verifiable manner that is simple to interpret and easy to explain to end-users.

Our infrastructure offers a simple way to do the verifiable computation for ranking any on-chain wallet or asset.

Subscribe to OpenRank
Receive the latest updates directly to your inbox.
Mint this entry as an NFT to add it to your collection.
This entry has been permanently stored onchain and signed by its creator.