We previously discussed the usability of an NFT AMM (Automated Market Maker). In this post, we explain how the SeaCows NFT AMM protocol handles NFT pricing.

The following assumed a good understanding of AMMs and liquidity pools.

The status quo

To automate NFT trades two things are required. Essential liquidity and automated pricing. To provide the former existing NFT AMMs have adopted the now common notion of liquidity pools. In regards to the latter, other protocols such as sudoswap have suggested bonding curves to provide immediate price quotes.

Bonding Curves

The concept of bonding curves is the following. All NFTs within one pool have the same price. Whenever a buyer or seller interacts with the liquidity pool and conducts a trade, the price for all NFTs inside the pool changes. Every purchase results in an increase, while every sale results in a decrease in the price. The starting price and price change variable are set by the liquidity provider(LP). Generally, the LP can decide between a constant, linear and exponential bonding curve.

Benefits & Drawbacks

• Benefits:

  1. It’s fully on-chain which makes it easier to integrate with other protocols.

  2. It provides the LP with a high degree of control over the prices within the pool (i.e. starting price, price range, delta).

  3. Prices behave predictably.

  4. Immediate quotes optimally allow traders to avoid unexpected slippage.

• Drawbacks:

  1. Since it’s fully on-chain, computation is limited and does not allow for a more sophisticated pricing strategy. “If all NFTs within one pool have the same price, how am I supposed to account for a different level of rarity within one collection?”

  2. Tied to the point above, it assumes homogeneity between NFTs. All NFTs within one pool are treated the same, going against the non-fungibility of NFTs.

A hybrid model

This entire year we have been researching and developing a hybrid NFT AMM protocol that provides essential liquidity via on-chain liquidity pools and automates pricing via a combination of on-chain bonding curves and off-chain machine learning models.

With such a new hybrid model, we are now able to offer the best of both worlds. Highly liquid NFT markets that are able to adopt a flexible pricing strategy.

How does it work?

1. Data Collection: Upon creation of the liquidity pool, all related transaction history and metadata of the selected collection are collected and analysed.

2. Classification: The off-chain model classifies the entire collection into up to 5 tiers based on their rarity and utility.

3. Pricing: A different set of pricing parameters (i.e. starting price, price range, delta) for all 5 tiers is automatically suggested, and can be adjusted by the LP.

4. Pool Creation: Once set, a pool with different tier levels is created.

AI-driven pricing

In order to make tier-based pricing work, we have built a model that is able to generalise across different NFT gaming collections. Based on the in-game stats and previous transaction prices, the model is capable to observe a pattern. This can be used to evaluate all NFTs within the collection, and increase price transparency, as it can highlight what features of the NFT make it valuable in-game.

Conclusion

We believe that in order to properly enable the adoption of NFT AMMs, a hybrid model utilising off-chain pricing computations to account for the non-fungibility of NFTs is needed. At current we are collaborating with many GameFi projects to test its viability and usefulness. At the same time, we also recognise that our approach has its own drawbacks such as the additional waiting time that is required when creating the liquidity pool as we collect data and classify the NFT collection. Our main objective is to learn quickly and build a protocol that can provide substantial value to all players involved.