moonblock

Motivation 1: Assessing the Price of an NFT is really hard
Motivation 2: This is a problem for buyers and sellers
Screenshot of Price Estimate and Similar Asset Screen
Tech Stack: Moralis and Covalent API to ingest data for data analytics

Inspiration

NFT investments are hard, especially for beginners. Finding the right price to buy or to sell an NFT can be confusing and people often invest a lot of time in researching collections and rarities. Often we see behavior like in the real estate industry: What are the prices listed or recent sales prices of the houses within my neighborhood, similar number of rooms or age of the building. This is currently manually and tedious using numerous filters at NFT marketplaces like Opensea.

What it does

The NFT Price Estimator helps users by providing a price estimate based on historical sales of NFTs in a collection and comparables. Price estimates are derived based on similar items in the collection taking into account their last sales price and the premiums of that sale compared the the floor price of that given day.

How we built it

The tech stack is built on Vue.js in the Frontend, FastAPI in the backend, and deployed on EC2 instances on Amazon Web Services. We used the Covalent API to ingest the NFT metadata (i.e. trait data) and Moralis API to ingest NFT trades/sales data our PostgreSQL database. Python pandas is doing the heavy lifting of performing data analytics on the NFT market data. Similar NFTs are calculating be vectorizing trait data of NFTs and calculating the absolute distance of the one-hot encoded feature vectors.

Challenges we ran into

The data-intensive queries using pandas in the Backend have been time- and memory-consuming, which is why we implemented a cache for the most heavy data queries.

Accomplishments that we're proud of

The app can crawl through a collection can be extended for any collection that is ingested into the PostgreSQL database. It is ready to be scaled in the future if necessary.

What's next for NFT Price Estimator

We heard some first voices and feedback that this is helpful for users removing the tedious comparison work using manual filtering. Next, we would like to extend the prototype by ingesting more popular collections and expanding the feature scope including more rarity aspects (e.g. rarity ranking and price visualization across rarity space). We imagine working on this project beyond the hackathon :)