NFT Alchemy: Pricing 101 # 134

Author: Sally, IOSG Ventures

Editor: Olivia, IOSG Ventures

NFT Alchemy: Pricing 101

How to price NFTs has always been an interesting topic. Pricing is an inevitable intermediate operation that contains both computable and non-computable components, which need to be addressed in any NFT finance application scenario. To widely apply NFTs in DeFi and fully activate the liquidity of NFTs, we first need to make a valuation judgment on the asset value of NFTs that is as realistic and widely accepted as possible.

However, due to the following three aspects, NFTs are difficult to evaluate and calculate simply like traditional financial assets:

1. The inherent subjectivity and illiquidity caused by the non-fungible nature of NFTs

2. The relative ambiguity of NFT rarity, and the fact that rarity and price levels are not completely positively correlated

3. The extreme price volatility of NFTs (often experiencing sharp rises and falls due to team and policy issues)

If the pricing mechanism cannot be well resolved, then behaviors such as NFT lending transactions often struggle to gain market trust due to high risks, leading to two problems:

• Lack of sufficient liquidity to support trading pool depth
• Difficulty in constructing diversified financial derivatives in the form of NFTs

To address this issue, an increasing number of NFT pricing platforms and emerging methods are appearing in the market. Here we can simply categorize these pricing solutions into two types:

• Peer pricing: which can be further divided into (a) Subjective assessment pricing and (b) Liquidity pool game pricing
• Oracle pricing: which can be further divided into (a) TWAP pricing and (b) Off-chain computation pricing

Peer Pricing

a. Crowds Subjective Assessment Pricing

Subjective assessment pricing is currently a pricing form with a strong subjective component. Liquidity lending protocols represented by Taker V1 can combine the interests of the DAO with those of the lenders, using subjective assessment and voting decisions to whitelist and price NFTs, thereby reducing the risk exposure faced by lenders to some extent. At the same time, this method has no restrictions on the quality requirements of NFT assets and can be widely applied to price discovery for long-tail and emerging NFT collectibles. However, this approach heavily relies on the judgment ability of curators and cannot provide real-time updates on NFT prices, resulting in overall low efficiency.

Taker V1

Taker is a liquidity protocol for NFT assets that primarily provides liquidity for NFT lending in the form of a DAO, supporting various forms of assets including NFTs, securities, synthetic assets, etc. Holding TKR tokens grants DAO membership, allowing participation in decisions regarding lending rates and fair pricing designs. Additionally, holding TKR can also yield extra income through staking.

Within Taker DAO, there are multiple curator DAOs (sub-DAOs), each of which can manage its own whitelist and the floor price of any NFT on the whitelist to prevent borrower defaults. Furthermore, members of the sub-DAO collectively vote to decide to allocate funds from their treasury to specific types of NFT assets. For example, some sub-DAOs may focus solely on Metaverse land assets, while others may focus only on PFP art assets.

In the Taker peer pricing model, the complete lending process is as follows:

1. Taker community members (lenders) deposit funds into the DAO.

2. The DAO mints DAO tokens (TKR) to represent members' shares.

3. Curators (initiators) subjectively price the NFT collectibles initiated in the DAO.

4. Borrowers use NFTs as collateral to take loans based on the curator's assessed price.

5. Borrowers repay the loan with interest.

6. DAO members receive rewards (based on interest yield).

7. As the DAO grows, the earnings obtained by DAO members accumulate.

b. Spot Liquidity Pool Game

The liquidity pool game pricing mechanism is similar to DeFi, primarily using an optimistic staking voucher mechanism, where liquidity providers stake based on their price expectations, thus binding the valuation of NFTs to the asset prices within the liquidity pool. This is also a pricing method that largely relies on the subjective decisions of LPs, with the advantage of achieving real-time valuation of NFTs, linking transaction value to real value, and releasing greater liquidity. However, it also faces issues such as a complex pricing mechanism that is not suitable for rapid pricing of low-value long-tail NFTs.

Abacus

Abacus is a simple and clear NFT valuation system that primarily utilizes optimistic PoS to create a liquidity pool-based NFT valuation method. The valuation methods of Abacus are divided into two types: one is the peer pricing mentioned above, and the other is the liquidity pricing discussed here. The value within its liquidity pool converted into ETH essentially represents the value of opening the NFT pool. Under this mechanism, it establishes a trading pair of ETH/NFT, allowing for real-time reflection of NFT prices similar to Opensea.

Based on its liquidity pool pricing method, the complete lending process is as follows:

1. Open the pool:

a. NFTs are non-custodial; owners need to sign on the pool as proof of life.

b. Owners will receive an NFT (ERC721) token to represent their property and earn transaction fees.

2. Traders lock ETH in the pool.

a. Decide the amount of ETH to lock: If the fund pool is 2 ETH, but the trader believes the NFT is worth 2.5 ETH, the trader can put 0.5 ETH into the fund pool, making the pool worth 2.5 ETH.

b. Decide the locking time: If the trader believes the price will only stay briefly at the current price, they can lock it for 2 weeks. Conversely, if the trader is confident in the floor price, they can lock it for a longer time to earn more rewards.

3. NFT owners initiate release.

4. NFT owners send loan requests.

5. Ownership is transferred to the lending platform.

6. The lending platform checks the spot size and locking time.

7. The lending platform issues the loan.

8. If the borrower fails to repay, the NFT will be auctioned immediately.

Oracle Pricing

a. TWAP

A typical way for oracles to price NFTs is based on a simple traditional algorithmic trading strategy, calculating a time-weighted average price (TWAP) by weighting the sales prices and floor prices of NFTs. For example, if we want to calculate the TWAP price over a time interval of 1 hour, we can take the cumulative prices P1 and P2 at the start and end, as well as the time differences T1 and T2, and divide the former by the latter to obtain the TWAP for that hour. The most well-known TWAP oracles include Chainlink and Uniswap V2.

In fact, TWAP pricing is the easiest to implement and the most efficient way; by integrating, scraping, and cleaning price data from NFT trading platforms, selecting multiple prices within a set time series to average can reduce the possibility of malicious manipulation and provide an acceptable, relatively accurate NFT price.

However, TWAP is not a perfect solution, as in extreme market conditions, when prices fluctuate sharply, TWAP oracles can easily be affected and become inaccurate. Therefore, TWAP is considered suitable only for pricing blue-chip NFTs with high market activity, good liquidity, and relatively stable prices.

BendDAO

BendDAO is a lending protocol that addresses NFT liquidity issues, allowing borrowers to borrow ETH by collateralizing NFTs. Currently, BendDAO supports lending for 9 types of blue-chip NFTs, including BAYC, Cryptopunks, Azuki, MAYC, CloneX, World Of Women, Coolcats, CyberKongz, and Doodles.

The NFT pricing method adopted by BendDAO is typical TWAP pricing. By collaborating with Chainlink, it collects the floor prices of the collateralized NFTs from the two trading platforms Opensea and LooksRare through multiple nodes, feeding the floor prices to the chain via contract interfaces to calculate the corresponding TWAP, thus filtering out the impact of price fluctuations on trading platforms. As shown in the figure below, for the Cryptopunk collection, the floor price provided by the oracle is consistent with the transaction average price and TWAP.

Similar to BendDAO, protocols using TWAP oracles for pricing include JPEG'd, DeFrag, DropsDAO, Pine, and others.

b. Off-chain Computation

Off-chain computation based on AI and machine learning is gradually becoming an emerging NFT oracle pricing method. Due to the non-fungibility of NFTs, their main attribute classifications, rare features, historical sales data, and other value information can be decomposed through metadata for use as model indicators. Protocols can then model and process based on this series of indicators and datasets to provide a relatively reliable and accurate pricing or pricing range.

This valuation method has a high technical barrier and is relatively friendly to long-tail NFT assets, making it one of the most promising solutions for large-scale applications. However, the problem lies in the high requirements for computing power and metadata; since the algorithms are not public, we cannot determine whether the training and fitting results are effective. Additionally, if the attributes of NFTs change, the model may become invalid, necessitating continuous iteration.

Banksea

The oracle protocol represented by Banksea primarily uses AI models to train datasets of NFTs, generating accurate and efficient predicted prices for different NFT assets. Its overall structure consists of two modules: the collection layer and the NFT layer.

In the collection layer, Banksea collects on-chain NFT transaction and listing records to calculate three types of prices in real-time: market floor price, AI floor price, and 24-hour average price. The AI floor price represents the lowest price among all AI valuations, serving as a risk control and stabilization measure during extreme market fluctuations or oracle attacks.

In the NFT layer, Banksea extracts multi-dimensional features of NFTs, conducts AI model training based on time series, and periodically generates two results: standard valuation and valuation range. Additionally, it fits and regresses the valuations calculated by the AI model against real-time transaction prices to optimize the final results and narrow the error range.

The off-chain pricing process based on AI models in Banksea is as follows:

1. External API queries: Monitor and scrape comprehensive NFT data from sources including trading platforms, social platforms, public chains, and collateral platforms.

2. Data aggregation: Clean the collected NFT data, extract feature attributes, and input them into AI nodes.

3. AI modeling: AI node clusters train and deploy models based on the input datasets, calculating predicted prices and risk scores, and returning the results to Banksea's smart contract.

4. Data submission: After removing outliers, the on-chain smart contract extracts data within a reasonable range to submit to third-party programs. Besides Banksea, Upshot and NFTBank also provide oracle solutions for precise NFT pricing based on more refined machine learning (ML) methods. Additionally, community tools like Defi Kingdom and Axie Infinity integrate AI off-chain algorithms for pricing.

Conclusion One More Thing

Finally, we can summarize the four specific solution strategies in the two major paradigms of NFT pricing currently on the market through the following dimensions:

It can be seen that regardless of the valuation method, there are certain advantages and disadvantages. We look forward to more emerging NFT pricing methods being explored and refined in the near future. Especially for the off-chain computation oracle pricing method, we believe that with technological advancements and the participation of more quality projects, more AI algorithm technologies such as deep neural networks (DNN) can be投入到拟合评估函数中，使定价决策树得到更准确和快速的修剪。

NFT pricing is like a game of Go, a seemingly simple yet complex game, a problem composed of a series of decisions. We can use peer intuition to judge the range or use oracle algorithms to predict the future.

And if we must ask what constitutes a good pricing paradigm? I believe the key to the question may be, as Wu Qingyuan said, not in how many times or how far one can calculate, but in how broadly, quickly, and accurately one can calculate.