The British Guiana 1c magenta was sold for USD 9.5M back in 2014. Why? Because it is the only specimen — it is a rare stamp. A stamp’s rarity is one of the major drivers for philatelists keen to possess them. And the rarity drives the price.
The guy who pasted the British Guiana stamp to an envelope 165 years ago would be shocked by the fact that this piece of paper is valued that much in the brave new world of blockchain, crypto-currencies, and NFTs. The digital universe of the Non-Fungible Tokens is driven by the same incentives that have been driving collectors for many centuries. Fortunately, the new transparent ways to measure items’ rarity have come among other fancy tools in the NFT space.
TL;DR
Artacle automatically indexes all Art Blocks collections and calculates three rarity metrics, they consider the distribution of the tokens’ traits within a project, but in a different manner. Each metric is represented by a number — rank. Smaller rarity ranks get assigned to more rare tokens. So NFTs can be easily compared using these ranks.
Also, Artacle provides REST API to read the rarity ranks. The API documentation is available here: https://artacle.github.io/api-docs/
In the example below we can observe that Fidenza #938 has rarity rank #1, and mapping of the trait’s values to the rarity score explains why. For example, it has the Collision Check trait set to ‘Relaxed’, and there are only 17 tokens in this collection that have the same value for this trait — it is a large share of the final score.
Rarity metrics used by Artacle
Rarity rank
This is the metric that should be in focus when analyzing the rarity and how it impacts prices. It can well explain what is the fraction that each specific trait contributes to the rarity. The ranking is fairly simple: calculate each trait’s value frequency across all tokens in a collection; then calculate the individual trait’s score by dividing 1 by that frequency. The token’s score is a sum of each individual trait’s scores for this token. The token’s rarity rank is the token’s position in a list ordered by the score. The traits that a token does not have, which are presented in some other tokens in this collection, are also considered for score calculation.
Single trait rarity
This measure is strongly driven by the rarest traits, some not rare traits may be not considered in the final score. We calculate the rarity of the token’s traits (including the missing ones) across a collection; sort traits by these rarities (the trait having the lowest frequency goes first, etc); sort tokens as per each trait’s rarity: if it’s a draw — use the next trait.
Statistical rarity
Statistical rarity accounts for all traits, providing a more comprehensive although less intuitive measure. We calculate the frequency of each token’s trait (including the missing ones) across a collection; treat these frequencies as a probability to meet a certain trait’s value across all tokens; multiply these probabilities assuming the traits have independent distributions; then sort the tokens as per their joint traits probability — the token’s position in this sorted list would be its statistical rarity score.
Note: this approach provides exactly the same result (with respect to a token’s position in the ranked list) as the Information Content-based method, suggested by the OpenRarity initiative.
What’s so special about Artacle’s rarity scores?
There is a number of tools and resources that calculate rarity for NFTs, e.g. rarity.guide. Artacle leverages the best practices from different approaches, but pays attention to details:
- The main difference — we leverage information not only about the traits that a token possesses but also about the traits that some other tokens have, which are missing in this specific token. For example: if 999 out of 1000 tokens in a collection have a specific trait (the value does not really matter), while 1 token does not have it — this is what can make the token very special.
- The tokens having exactly the same traits and their values will have exactly the same rarity rank in Artacle. Some other tools just put always increasing rank (e.g. for the single trait rarity) leading to ‘identical’ NFTs having different ranks.
- The ‘spread’ traits — those traits that have a wide range of possible values. For example, a Rotation may be any value in the 0–360 degree range. Some tools just ignore such traits (sometimes without any transparent rules which traits are treated as ‘spread’). Artacle always takes all traits into consideration.
- Traits normalization may be applied to avoid the spread traits heavily impacting rarity. Low-variance traits will have a higher rarity score, while high-variance traits (i.e. spread traits) will be assigned a penalty thus lowering their rarity score. The traits normalization can be turned on/off on Artacle. Example (the Vortex Layers is the spread feature):
- The number of traits may be considered or excluded when rarity is calculated. If enabled, the Traits Count is treated as a separate trait. I.e. a token having a special number of traits compared to other tokens may have a higher rarity rank.
- Traits Weighting can be applied if this was implied by a creator. The author may indicate that specific traits were supposed to mean something special, and in this case, Artacle multiplies their score as specified by the author. Here is an example for House Of Legends — we can see that the Special Vibes and Special Edition traits were scored x10:
How a token’s rarity impacts the price
The rarity impact on the tokens’ price is not as straightforward as one may expect. There are many other external factors that impact the price:
- The artist defines what to put to the traits. Some of the visual appearance factors are not encoded in the form of traits while being important for those who collect NFTs.
- The community ‘mood’ and the things being discussed in Discord right now may impact the prices significantly. Some NFTs may get hype although their traits are not rare.
- The market conditions, ETH exchange rates, etc — that impacts the prices very much. E.g. an uprising trend of ETH to USD rates can drive high demand, so buyers just want to buy tokens as a middle-term investment, disregarding their rarity.
Even having said that, we can still observe a general (although not that strong) correlation between rarity and the price: the tokens with a nice rarity score usually get traded at higher prices. This correlation can also be very much different for each collection, that’s why Artacle provides different visualizations to capture important insights and find possibly under-estimated tokens. For example, here is the Rarity vs Last Sale Price chart for the Seadragons project:
Other features
Artacle also provides more features, both in the ‘fun with art’ space (such as automated similar images search for each token) and the trading insights to analyze price dynamics for each collection and token. See more at artacle.io
Artacle API documentation is available here: https://artacle.github.io/api-docs/
If you have any feedback or new feature request feel free to drop a mail at feedback@artacle.io
