This study pulls data from thousands of top NFTs to determine how features like headwear, eyewear, and background color influence price.
Updated Jun 22, 2022
Updated Jun 22, 2022
Many companies on MoneyMade advertise with us. Opinions are our own, but compensation and in-depth research determine where and how companies may appear.
NFTs
Art & Culture
Crypto
Research and data analysis by Ashley McKillips.
It's clear by now that non-fungible tokens (NFTs) are here to stay. Not only that, but they've become one of the most accessible investments thanks to their often low initial price and the fact that anyone can buy one in minutes on NFT marketplaces like Rares and OpenSea.
It might be easy and low-cost to invest in NFTs, but they definitely don't all skyrocket in value. For every CryptoPunk that's gone from $100 to $1 million, there are hundreds of NFTs that are worth almost nothing.
Whether you're looking to turn a quick profit or hold onto an NFT that will increase in value over time, the goal is simple: buy low, sell high. In other words, you're looking for NFTs that are undervalued when comparing their current sell price to their predicted future sell price. But how do you predict that?
In this study, we pulled data from thousands of the top NFTs on OpenSea and analyzed key features like headwear, eyewear, background color, smoking, and sale history to determine if it's possible to predict whether or not an NFT will increase in value. Here's what we found.
We gathered data from the four of the most popular NFT collections on OpenSea: Azuki, Bored Ape Yacht Club, CyberBrokers, and DourDarcels. These are all PFP or "profile picture" NFTs, which are collections built around a character. They're a form of generative art in which creators use code to generate hundreds, or even thousands, of iterations of their character, employing varying combinations of different hair colors, skin colors, accessories, backgrounds, and more. Collectors often buy these NFTs as a way to express their digital identity, using them as avatars or profile pictures on social media and within the metaverse, hence the name "PFP."
Using character-based NFT collections made it easy to compare common characteristics such as headwear, eyewear, and background color. Each of the four collections chosen contain around 10,000 randomly generated NFTs. These NFTs sell for anywhere from around $800 on the low end (DourDarcels) to upwards of $500,000 on the high end (Bored Ape Yacht Club).
We looked at four different easily identifiable physical features to determine aesthetic predictors of a successful NFT: headwear, eyewear, background color, and whether or not the character is smoking. We also looked at sales data to determine how the frequency with which an NFT is traded might act as a predictor of its success. Here's what we found for each of these factors.
Source: OpenSea.io, From left to right: DourDarcels, CyberBrokers, Azuki, Bored Ape Yacht Club
Overall, headwear didn't seem to be a significant predictor of NFT success. A high percentage would indicate that headwear could be a predictor of success, while a low percentage would indicate that a lack of headwear could be a predictor of success. However, exactly half of the top NFTs we pulled data from were wearing headwear—defined as anything worn on the head of the character that isn't hair, including hoods and clips—indicating that headwear doesn't necessarily make a character NFT any more or less desirable.
However, when we zoomed in, there was some indication that headwear might have helped the success of certain collections.
Given that Bored Ape Yacht Club is by far the most popular and valuable collection of the four studied (and one of the top three collections of all time on OpenSea), it may be the case that the prevalence of headwear within this collection helped it garner attention. From halos and bunny ears to sailor caps and party hats, the headwear used in this collection is certainly eye-catching and adds a lot to the unique personality of each character, and that personality is exactly what collectors of PFP NFTs are looking for.
An example of how heavy accessories can be used to ascribe a human-like personality to a non-human NFT character
Source: OpenSea.io, Bored Ape Yacht Club
It's also worth noting that Azuki and CyberBrokers, the two collections in which headwear was the least prevalent, are very different in style from Bored Ape Yacht Club and DourDarcels. Rather than featuring fantastical, cartoonish, non-human characters, these two collections feature more realistic depictions of human characters. If it can be said that PFPs are often chosen largely on the basis of self-expression, then accessories like headwear may play a more central role on non-human characters than on human characters. After all, these accessories serve to anthropomorphize non-human creatures for collectors who are looking to find an image that represents their style or personality.
Source: OpenSea.io, From left to right: Azuki, DourDarcels
Eyewear was much less prevalent in these four collections than headwear, suggesting that it might not be a significant predictor of NFT success. Defined as anything worn on the eyes of the character (excluding laser beams shooting out of eyes in the Bored Apes Yacht Club collection), only 18.5% of the top NFTs were sporting eyewear.
When broken down by collection, the numbers stay low across all four.
An example of how the visibility of a character's eyes is often necessary to convey mood and emotion
Source: OpenSea.io, Bored Ape Yacht Club
This may come as a surprise considering the role that accessories in general play in personalizing a PFP NFT. However, eyewear covers a character's eyes, one of the most important features an NFT artist can use to convey mood and emotion. The absence of eyewear on the vast majority of top NFTs, then, might be an indication that animated facial expressions are what is sought-after by collectors. If you browse the Bored Ape Yacht Club collection, (again, one of the top NFT collections of all time), you'll see that exaggerated facial expressions are one of the most noticeable features of its characters.
Examples of a cool, neutral, and warm background color
Source: OpenSea.io, From left to right: Bored Ape Yacht Club, Azuki, CyberBrokers
After breaking down background color into categories of warm, cool, and neutral tones, we found that the most common background color across the top NFTs was warm, followed by cool, and then neutral. However, when broken down by collection, there was no clear pattern.
Warm colors include red, orange, yellow, pink, and brown backgrounds as well as some warmer purple backgrounds, while cool colors include green and blue, and neutral colors include gray and black. Given that the warm tone contained the largest number of colors, this information alone doesn't necessarily indicate that warm-toned NFTs are more successful.
Source: OpenSea.io, CyberBrokers
That said, the CyberBrokers collection in particular has an incredibly low percentage of warm-toned top NFTs and a huge percentage of neutral NFTs. This particular collection features bright, eye-catching colors on the characters themselves (lots of neon green, magenta, oranges, and bright blues and purples). It may be that the NFTs in this collection with neutral-toned backgrounds are more popular because they allow the colors of each character's attire to stand out more.
Source: OpenSea.io, From left to right: Bored Ape Yacht Club, Azuki
Only 11.33% of the top NFTs we looked at featured a character smoking something, suggesting that smoking may be a feature that makes an NFT less popular.
We excluded DourDarcels from this analysis as none of the characters in this collection are smoking. Smoking accessories were largely unpopular across all three of the other collections.
This may be partly due to the importance of facial expression. In the same way that eyewear can obscure expression, smoking may have a similar effect. However, it's worth nothing that many characters in these collections have their mouths obscured by other objects, such as gas masks, surgical masks, knives, and even blades of wheat. Across all collections, variations in the eyes seem to be more integral to conveying mood and emotion in a character than the mouth. It's more likely that smoking is simply a low-demand feature.
Across all four popular NFT collections, we also looked at the number of times these NFTs changed hands. On average, the top NFTs have been sold 2.065 times each. The maximum times any of the top NFTs included have been sold was 6, and the most common figure for times sold (the mode) was 1.
Here's what that "times sold" data looks like when broken down by collection.
CyberBrokers is a bit of an anomaly in this category, having been traded significantly fewer times than the other collections. This might be due to the fact that the CyberBrokers collection has substantially fewer owners than the other collection (3.8k at the time of writing versus 5k+ for the other three collections).
Overall, though, our findings indicate that the top NFTs are traded twice on average and up to 6 times. While these numbers might seem low, it's worth considering that with the exception of Bored Ape Yacht Club, all of these collections launched this year. Even Bored Ape Yacht Club has only been around for about a year as of this study. Even when compared to some of the most liquid investments, such as stocks, these numbers are fairly high. People certainly aren't day-trading the top NFTs, but they aren't always holding onto them for the long-run either.
On top of data suggesting NFTs that have been traded a couple of times are more likely to increase in value, it also makes sense that prior sales numbers can serve as evidence that an NFT is in-demand. That said, price is just as important as times sold. If you're looking for the biggest return on investment, NFTs with a low Earliest Price and a high Latest Price are more likely to have a large Highest Current Bid.
Our data suggests that if you're looking for NFTs with high potential for price growth, you should look for warm or neutral toned NFTs that have headwear but aren't wearing eyewear or smoking. You'll also want to look for NFTs that have been sold at least twice and have a low Earliest Price and high Latest Price.
Source: OpenSea.io, Bored Ape Yacht Club
Of course, style and art aren't a science—even when they're created with code. There are plenty of other factors that can influence the popularity of an NFT that aren't as easily measured. For instance, the absence of eyewear may generally be more popular, but an individual NFT might incorporate eyewear in a unique or particularly aesthetically-pleasing way that causes it to take off. But hopefully by combining your own artistic judgment with our data, you can pick some winners.
Data was gathered on NFTs from 4 popular OpenSea collections: Azuki, Bored Ape Yacht Club, CyberBrokers, and DourDarcels. These collections each contain around 10,000 randomly generated, iterative, and character focused NFTs. We picked these specific four because they were all listed in the top collections on OpenSea as of March 2022. Bored Ape Yacht Club was chosen primarily due to its name recognition while the other three were chosen because they had easily determinable headwear, eyewear, and smoking characteristics that made data collection more clear-cut.
Characteristic and price data for the top 50 highest latest price NFTs in each collection was collected during March 17-24 of 2022. We focused on the top-priced ones to determine if there were commonalities in the successful NFTs even across collections. Since the data was gathered there have been some ranking changes in these collections, meaning that time sold data, highest current bid, and latest price will have changed since the study was conducted.
We used a linear regression model to figure out what the best predictors of Highest Current Bid and Latest Price are. We used stepwise predictor selection to add and remove predictors until we got the best models.
If we were to complete this study again, there are some things we would modify. We'd want access to data that updates in real time so prediction models are as realistic as possible. We'd also like to gather more data, expanding this study to a wider range of NFT collections and to more NFTs in the collections we used. A deeper color analysis could also prove useful, testing how each individual color impacts price. Finally, including more predictors, such as body characteristics and clothing, could also provide valuable insights.