The shuffle — the prelude of our Evaluation Machine
We have talked extensively about how our NFT Evaluation Machine differentiates us from other similar projects but haven’t yet unpacked it, or how is it expected to price NFTs? In the following series, we’ll gradually unpack each of the key areas without revealing the secret sauce that makes it work. At the end of the series, you’ll be able to explain how our Evaluation Machine works to your grandma, and, hopefully, she won’t be able to reverse-engineer our secret sauce!
This series is structured like a good Texas Holdem hand, and the next pieces will come out as indicated:
- The shuffle — The prelude of our Evaluation Machine
- The hand — How salty is the OpenSea? @ 30/01/2022
- The flop — In the eye of the (AI) beholder @ 27/02/2022
- The turn — How much is the fish? @ 27/03/2022
- The river — The price of everything and value of nothing @ 30/04/2022
What is AI-based pricing?
AI, Machine learning, and big data are the most abused keywords in the industry, usually positioned as a solution for all the problems you have in life! In practice, I usually draw the line at Excel — if you can solve it in Excel without substantial extensions/effort, then it’s definitely nothing fancy requiring any of the previously mentioned labels. In our case, we’ll drill deeper where even simple copy/pasting code from StackOverflow won’t suffice! Don’t get me wrong, AI/ML can solve a broad range of problems IF they are framed precisely. A snap from “I, Robot” below summarizes it well:
Let’s focus first on the key aspects of our Evaluation Machine before going deeper into nuts and bolts.
At the moment, most of the NFT pricing is done heuristically, and arbitrary prices are posted as asking bids. As a result, two subsequent problems arise — the majority of posted auctions are never filled, causing a major cost in terms of listing gas fees and opportunity costs. A fraction of sold NFTs are rarely resold on the secondary market — either because of owners “just like the stock” or due to loss aversion. It’s hard for homo economicus (rational individuals) to admit that one has overpaid for an asset and, therefore, will keep on holding it. Two processes influence the final price: the probability of selling it at the selected listing price and the overall value of the intrinsic attributes of the asset (measuring how much they like the stock). We’ll come back to an in-depth look at these two in the ‘’How much is the fish” piece.
One of the most important drivers of any NFT price is its novelty. Everyone wants to own the original, and due to the decentralized and anonymized nature of the asset — it’s the first one that got listed on a reputable chain. A clear analogy to how intellectual property is protected and defended is applicable here. Tackling novelty requires scoping out the whole on-chain market, and we’ll showcase our approach to one example in the “How salty is the OpenSea?” piece.
The next natural question is, what if I just make a copy of the original and repost it as a second copy? Will it be worth a bit less than the original? We can clearly see that a major amount of the value is derived from the overall collection it belongs to. In a similar spirit, I can make a perfect copy (theoretically!) of any of Banksy’s works, but I doubt that I’ll be able to retire for the sale of it. We’ll see this in the final piece of the series.
The next trick would be to change only a piece of an NFT and work on the similarity aspect. Similar how videos with weird framing are posted on youtube to comply with their % of novelty requirement. A human eye can easily spot differences and similarities — how can this be tackled in a systematic fashion for our Evaluation Machine to ‘see’? If we limit our scope for illustration purposes only to images, we can see how this has been well researched in a deep transfer learning field. The human has been replaced with a silicone one with impressive precision. We’ll cover our application in the “In the eye of the (AI) beholder” piece.
We’ll cover in more depth how all these drivers interact together on the final price in the ‘the price of everything and value of nothing” piece. Briefly, we’ll see how the analogy between RBG (Red-Blue-Green) can be applied here to solve pricing questions.
I hope everyone is as excited as I’m and has a clear expectation of what is to come and where we are going with this!