NFTs are illiquid digital assets that change hands infrequently. Hence, the open market cannot be relied on as the sole price oracle or NFTs, which limits their application.
Appraisals offer one such mechanism to efficiently price NFTs. There are other mechanisms that are commonly used today, though they each fall short of appraisals.
To date, floor price has been the most used metric for valuing NFTs and NFT collections. A floor price is simply the price of the lowest-listed NFT in a collection. This works for some use-cases but is not without issue.
First, a floor price only reveals information about what the least-valuable NFTs in a collection are worth (referred to as “floor NFTs”). NFTs are, by definition, non-fungible, meaning the NFTs in a collection represent a potentially large spectrum of differing values.
In light of this, one way to improve the informative efficiency of floor price is to consider “Trait Floor Price” instead. This means taking the lowest ask of one (or multiple) of the trait(s) that an NFT possesses. For example, if the lowest ask for Gold Apes is 250 ETH, the trait floor for the Gold Fur trait would be 250 ETH. However this too fails to overcome two other shortcomings with relying on floor prices.
Floor prices may not exist; collections and especially small baskets of assets that share common traits may not admit a single listed asset. In fact, the assets bearing the rarest of traits or rarest combinations thereof — and therefore those typically of greatest interest — tend to be listed for completely nonsensical prices if they are even listed at all.
Further, a floor price represents only one side of the order book: the ask-side; it does not give us any information about the bid-side of the order book. A floor price only tells us what people are willing to sell their NFTs for, not the price they’re willing to spend to buy NFTs. In other words, floor price ignores demand.
Another mechanism people use to value NFTs (albeit less commonly) is by referencing the last sale price.
The benefit of this approach is that it yields a more informative sense of what someone is willing to actually pay for an NFT. By definition, someone did in fact pay that price!
A glaring issue with this approach however is that NFTs often change hands infrequently so the last sale price quickly becomes stale, rendering it an insufficient measure of value over time. The failure of this method to communicate any information is especially true for those assets that have yet to sell since their mint, of which there are many.
One can chip away at this illiquidity issue by considering the last sale price of sets of assets, including baskets of assets of common traits, but there is a clear tradeoff between grasping at more data (by increasing the size of such a set) and the applicability of any sale to any constituent NFT.
A newer, more robust approach to pricing NFTs is appraising them. An appraisal is an in-the-moment estimation of sale price that would be yielded if the asset were to be sold now. Appraisals are a powerful tool for valuing NFTs and creating more efficient NFT markets, as they offer:
- More granular pricing: Instead of assigning a value at the collection-level or trait-level, appraisals assign a value at the NFT-level — every NFT gets its own price estimate. This is important because each NFT is unique, meaning each may warrant a unique value.
- Frequent updates: Appraisals enable assets to be re-priced without them needing to change hands — something that happens infrequently with NFTs.
- More accurate: A much larger set of inputs are considered in generating appraisals.
Appraisals may take both floor and last sale price (among other factors) into consideration, making them strictly more informative than either alone.
Appraisals can be produced in a number of ways, including crowdsourcing insights from experts to statistical analyses, machine learning models, and combinations thereof.
After exploring several of these strategies for appraising NFTs, we have found machine learning to be the most effective in achieving high coverage, high accuracy, and frequent price updates.
At Upshot, we developed our machine learning models to emulate, enhance and automate the manual process an expert collector might use when valuing NFTs. These processes may include reviews of:
- the net present value of a collection’s overall desirability and utility
- the desirability of the asset itself, as a function of rarity and desirable traits
- recent achieved sales of similar assets (”comparables” or “comps”)
Using ML, we are able to produce accurate, near-real-time appraisals for a wide range of NFTs. To better understand the underlying our underlying process, visit Appraisal Methodology.
Updated 10 months ago