The Alchemy of NFT: Pricing 101

How can we price the NFT is prone to be an interesting topic considering pricing is an unavoidable intermediate operation, containing both computable and non-computable parts of the problem. It needs to be solved in almost any NFT Fi application scenario. To universally apply NFT to DeFi and completely activate the liquidity of NFT, we first need to make a valuation judgment about the value of NFT as an asset that is as authentic and widely accepted as possible.

However, NFT is difficult to be valued simply as traditional financial assets due to the three aspects below:

1. The endogenous subjectivity and illiquidity of NFT due to its non-fungibility property.

2. The rarity of NFT is relatively ambiguous. Besides, the rarity and price levels are not in a completely positive correlation.

3. NFT prices are dramatically volatile (often subject to sharp pulls and dumps due to team and policy issues)

And if pricing mechanisms are not well addressed, then actions such as NFT loaning often struggle to win market trust due to high risk, which in turn leads to two problems.

(1 ) Lack of sufficient liquidity to support the depth of trading pool ;

(2) difficulty in structuring diversified financial derivatives in the form of NFTs.

To tackle this problem, more and more NFT pricing platforms and novel approaches are emerging in the market. Here we can simply divide these pricing solutions into two categories:

Peer Appraisal: which can be subdivided into (a) Crowds and (b) Spots

Oracle Appraisal: which can be subdivided into (a) TWAP and (b) off-chain computation

1. Peer

a.Crowds

Crowds appraisal is a form of pricing that is currently more subjective. The liquidity lending protocol represented by Taker V1 can whitelist and price NFT in the form of subjective assessment and voting decision by combining the interests of DAO with those of lenders, thus reducing the risk exposure faced by lenders to a certain extent. At the same time, this approach has no restrictions on the quality requirements of NFT assets and can be widely used for price discovery of long-tail and emerging NFT collections. However, it relies heavily on the curator’s judgment ability, and its inability to provide real-time updated NFT prices makes it less efficient overall as well.

Taker V1

Taker is a liquidity protocol for NFT that provides liquidity for NFT lending primarily through a DAO that supports multiple forms of assets including NFT, securities, synthetic assets and more. Holding TKR tokens gives you access to DAO membership and participation in making decisions about lending rates and fair pricing, among other designs. At the same time, holding TKR also allows you to earn additional income through staking.

Taker DAO has multiple curator DAOs (sub-DAO), each of which can control its own whitelist and the floor price of any NFT on the whitelist to prevent borrowers from defaulting. In addition, the members of a sub-DAO decide, through a collective vote, to invest their own treasury funds in specific types of NFT assets. For example, some sub-DAO’s can focus only on Metaverse land assets and some sub-DAO’s can focus only on pfp art assets.

Mechanism

1. Community members (lenders) deposit capital to DAO
2. DAO mints DAO tokens to represent the shares of members
3. Curator (voted by community members) provides subjective appraisal (duration, APR, amount) on NFT collections
4. Borrowers collateralize NFTs to take out loans based on curator’s appraisal price
5. Borrowers pays back loans with interests
6. DAO members gain from rewards (interest yields)
7. DAO members will gain more as DAO grows

b. Spot

Similar to defi, the pricing mechanism of the liquidity pool works mainly through an optimistic pledge certificate mechanism, i.e., a pledge by the liquidity provider based on its own expectations of the price, thus tying the valuation of NFT to the price of the assets in the liquidity pool. This is also a pricing approach that relies heavily on the subjective decisions of LPs. The advantage is that it enables real-time valuation of NFTs, linking the transactional value to the true value and further releasing more liquidity. However, it also suffers from complex pricing mechanisms and is not suitable for pricing a large number of low-value long-tail NFTs in a time-efficient way.

Abacus

Abacus is a simple and straightforward NFT valuation system that primarily uses optimistic PoS to create a liquidity pool based NFT valuation methodology. abacus’ valuation method is divided into two parts — one is the group pricing as we mentioned above, and the other is the liquidity pricing we discuss here. The value within its liquidity pool converted to ETH is equivalent to the value of the open pool NFT. Under this mechanism, it is equivalent to developing a ETH/NFT trading pair that can reflect NFT prices in real time, akin to opensea.

Based on its liquidity pool pricing methodology, the complete lending process is shown as below:

Mechanism

1. Open the pool
2. NFT non-custody, owner need to sign the pool as proof of life
3. Owners will receive an NFT(ERC721) token to represent their properties and earn trading fees
4. Traders lock ETH in the pool
5. Decide the amount of ETH you want to lock: If the pool is at 2eth, but you think the NFT is worth 2.5eth, you can put 0.5eth in the pool and the pool is now worth 2.5eth.
6. Decide how long you want to lock: If you think the price will only stay there for a short time, you may want to lock for 2 weeks. On the contrary, if you are confident about the floor price, you can lock longer and get more rewards.
7. NFT owners enable emissions
8. Owners request loaning
9. Transferring ownership to a lending platform
10. Lending platform checks spot scale & lock duration
11. Lending platform issues a loan
12. NFTs will be auctioned immediately once repayment is not executed

2. Oracle

a. TWAP

A typical way for an oracle to evaluate the pricing of NFT is based on a simple traditional algorithmic trading strategy in which a weighted average of the NFT sales price and the floor price is calculated to obtain a TWAP (Time Weighted Average Price). For example, if we want to calculate the TWAP price within the time interval of 1 hour, we can take the difference between the cumulative starting and ending prices P1 and P2, and the difference between the starting time T1 and the ending time T2, and divide the two to calculate the TWAP for the 1 hour. The most well-known TWAP oracles include Chainlink and Uniswap V2.

In fact, TWAP is one of the easiest and most efficient ways to reduce the possibility of malicious manipulation and provide an acceptable and relatively accurate NFT price, by simply integrating, crawling and cleaning price data from the NFT trading platform and selecting multiple prices to take the average value within a set time series.

However, TWAP is not the perfect solution because in extreme market conditions, when prices fluctuate dramatically, TWAP-based oracle is susceptible to be inaccurate. Therefore, TWAP is considered only suitable for pricing blue chip NFTs with high market activity, good liquidity and relatively stable prices.

BendDAO

BendDAO is a lending protocol to solve the liquidity problem of NFT. Borrowers can borrow out ETH by staking NFT. Currently, BendDAO can support the lending of nine types of blue chip NFT including BAYC, Cryptopunks, Azuki, MAYC, CloneX, World Of Women, Coolcats, CyberKongz and Doodles.

The NFT pricing adopted by BendDAO is a typical TWAP pricing method. In collaboration with ChainLink, it uses multiple nodes to collect the floor price of the NFT on Opensea and LooksRare, uses the contract interface to feed the floor price to the blockchain, and calculates the corresponding TWAP, thus filtering out the impact of price fluctuations on marketplaces. As can be seen from the following figure for Cryptopunk, the floor price provided by the oracle, the average price and TWAP are consistent.

Similar to BendDAO, other protocols adopting TWAP oracle for pricing include JPEG’D, DeFrag, DropsDAO, Pine, etc.

b. Off-chain computation

Off-chain computing based on AI and machine learning has also gradually become a new way to NFT pricing oracle. Due to the non-fungibility of NFT, its main attribute classification, rare features, historical sales data and other valuable information can be used as model indicators through metadata decomposition. The protocol can then model and process based on this series of indicators and data sets to give a relatively reliable and accurate price or price range.

This valuation approach has high technical barriers, is relatively friendly to long-tail NFT and can be considered the solution with the most potential for large-scale application. However, the problem is that this method requires high computational power and metadata, because the algorithm is not public, we can not determine whether the training and fitting results are effective. In addition, once NFT attributes and characteristics change, the model is likely to fail, so it needs to be iterated continuously.

Banksea

The oracle protocol represented by Banksea mainly uses AI models to train NFT datasets, so as to generate accurate prediction prices efficiently for different NFT assets. The whole system consists of two modules: the data acquisition layer and the NFT layer.

At the data acquisition layer, Banksea collects on-chain NFT transactions and listing records to calculate three kinds of prices in real time: market floor price, AI floor price, and 24-hour average price. AI floor price represents the lowest value of all AI valuations and plays a role in risk control and stability maintenance when the market has drastic fluctuations or the oracle is attacked.

At the NFT level, Banksea will extract the multidimensional attributes of NFT, conduct AI model training based on time series, and regularly generate two results: standard valuation and valuation range. In addition, it will fit and do the regression to estimate the outcome calculated by the AI model with the real-time transaction prices to optimize the final results and narrow the error margin.

Banksea’s off-chain pricing process based on AI model can be seen below:

Mechanism

1. External API Query: Monitor and capture integrated NFT data from trading platforms, social platforms, public chains, collateral platforms, etc.
2. Data aggregation: Clean the collected NFT data, extract their features, and input them to AI nodes
3. AI modeling: The AI node clusters conduct model training and deployment based on the input data set, calculate the predicted price and risk score, and return the results to the Banksea smart contract
4. Data submission: Remove outliers from smart contracts, extract data within a reasonable interval and submit it to a third-party program

Except for Banksea, Upshot and NFTBank also provide oracle solutions for accurate pricing of NFTs based on more subdivided ML (machine learning) methods in AI. In addition, community tools such as Defi Kingdom and Axie Infinity integrate AI off-chain algorithms for pricing as well.

One More Thing

In the end, we can summarize the four specific solution strategies of the two major mechanisms currently on the market towards NFT pricing through the following dimensions.

It can be seen that there are certain advantages and disadvantages of either valuation method at present. We look forward to more emerging NFT pricing methods being discovered and improved in the near future. Especially for the oracle pricing method of off-chain computing, we believe that with the advancement of technology and the participation of more high-quality project parties, more AI algorithm technologies such as deep neural network (DNN) can be put into the fitting evaluation function, so that the pricing decision tree can be pruned more accurately and quickly.

It can be seen that there are certain advantages and disadvantages of either valuation method at present. We look forward to more emerging NFT pricing methods being discovered and improved in the near future. Especially for the oracle pricing method of off-chain computing, we believe that with the advancement of technology and the participation of more high-quality project parties, more AI algorithm technologies such as deep neural network (DNN) can be put into the fitting evaluation function, so that the pricing decision tree can be pruned more accurately and quickly.

NFT pricing is analogous to the game of Go, consisting of a series of decisions. It is a complex game that looks simple. We can both use peer intuition to decide the extent, and the oracle algorithm to predict the future.

If you must pursue the question of what is a good pricing paradigm, I think the crux of the problem is perhaps, as Go Seigen said, the greatest Go player in the 20th century, is not how many times and how far to count, but how wide, how fast and how accurate it is.