In the last article we analyzed the NFT portfolios of some of our users for their yieldedness. In this one, we will take a closer look at our NFT scoring method based on statistical methods. The rarity of an NFT determines how rare an NFT is and, in turn, how valuable it is. Rare NFTs are most sought after by collectors, which pushes their price up. This is why in this article we will provide our users with a better understanding of the traits of different NFT collections that influence their rarity and, consequently, give a solution of better and more successful budget allocation.
Trait shares
Let’s consider the traits of Ninja squad official #6104 an example. We can see that CLOTHING, HAT, FACE and WEAPON traits have very low shares and this should make this NFT rare.
In fact, if we compare traits of that NFT with another from this collection, we will see that the CLOTHING trait has 83 unique values, about 1–2% share each. It means that most of the values are very close to the average, but they are not unique. We can consider only traits which share less than 0.01% as unique. The same is true for the HAT, FACE and WEAPON traits in this collection. From this you can make a conclusion that NFT with a small share of a trait is not always rare.
For example, the average CLOTHING share for the collection mentioned above is 1.48%. It means that CLOTHING trait of our NFT is more popular than the average of the collection.
Deviation scoring method
Main idea of our scoring method is to calculate the deviation of a trait’s share from the average share for the collection. This sum of the deviations by each trait is an unnormalized NFT score. Then we should scale it to the [0, 10000] range that more score more rarity. Further, we will call this the Rarity Score.
NFT comparison
Now, we can compare NFTs in a collection with filled traits by our scoring method. Also we can sort NFTs by rarity score and build the order in the collection. For example, Ninja squad official collection has 8888 items in total. The most rare item will give order 1 and the most common will give order 8888.
We split all collections by rarity order for the convenience of users:
- top 2% of items will give PLATINUM label
- 2–10% — GOLD
- 10–25% — SILVER
- 25–50% — BRONZE
- 50–100% — IRON
Let’s consider three NFTs from this collection: #6104, #5518, #7758. We calculated rarity scores using our method in Table 1.
First two NFTs are very similar in Rarity score, although they differ in the order. The last NFT is the rarest one. Let’s look at the trait shares (Fig. 2). We can see that Male GENDER of #6104 and #5518 NFTs is very common. Meanwhile, the other traits are almost equally rare. The main reason that #7758 NFT is the rarest, is the GENDER trait. In this collection, GENDER trait has the Male value in 90% and Female in 10% of cases.
It still remains difficult to compare NFTs based on trait shares, because it’s not informative. Therefore we built a non-linear transformation that represents a trait’s importance. The transformation enhances really rare traits and smoothes common ones (Fig. 3). It can be much more informative which trait is important.
Blockchain has rapidly grown over the years, with NFTs being an exciting innovation gaining more and more adoption. To give our users the possibility to buy inexpensive and unique NFTs and understand their rarity traits, Amoss developers provide you with the best solutions for analyzing NFT collections by applying different approaches and exceptional algorithms for your proper budget management.
Join Amoss here:
