BGAN Punks V2 Grails & Floors: Applying the Fat Middle Theory
First a TL;DR for the Fat Middle theory: in essence every NFT collection has good and bad outliers — grails and floors — and it’s the substance in the middle that really defines a collection. Crypto Orca has a great article on this here: The key to a great long-form generative art collection: the fat middle (It’s really not that long. You should read it).
Long-form, in the context of generative art, describes a collection of artworks that is generally too large for the artist to curate and hand-pick the best ones. Typically we see this with collections of a couple hundred pieces or more. Tyler Hobbs writes about this here: The Rise of Long-Form Generative Art.
The observation Muiy makes here is really spot-on for our beloved BGAN Punks. They don’t trade like other PFP (profile picture) collections, where the floor pieces create a bulk of the trading volume. In these other collections the pricing between the floor, the middle, and grails is generally pretty tight. With BGAN Punks listings there is the floor, a very thin middle and an even thinner top tier. A large portion of the collection has never been listed or sold. There’s also a lot of speculation baked into the upper tiers, and probably once the floor:grail price asymmetry lessens we’ll begin to see more grails listed and sold. Until then we’ll likely continue trading like separate independent collections from the same artist.
There’s a lot of speculation involved with trying to find the fair price of any NFT collection. We can see this play out by looking at XCOPY’s works. People see his earlier creations sell for a boat load of ETH, so they expect his new works to at least capture some of this success. He has a small collection of 1 of 1 originals and small bundles, Grifters a collection of 666, and Max Pain with 7,469. So here we have a lot of the same variables within the separate collections, but the biggest difference is the collection size. If all of XCOPY’s creations where part of the same long-form collection the price asymmetry would be very similar to what we see with BGAN Punks.
When applying the Fat Middle theory to XCOPY the results are unfavorable. This theory basically says that the quality of a long-form collection is only as good as it’s middle and the middle must be fat. And since 90% of the collection in our example is Max Pain, it fails the Fat Middle test. Thankfully XCOPY released his creations as separate collections.
BGAN Punks and Long-Form Art
In Tyler Hobbs’ The Rise of Long-Form Generative Art the first thing he discusses is the history of generative art — it’s been around since the 60s — but he then goes into the amount of work which is required vs releasing a short-form collection. Which generally starts as a long-form collection, but then the artist trashes the weak outputs giving the illusion of a strong algorithm with strong outputs. A great long-form collection requires a strong algorithm to produce a visually appealing variety while also providing cohesion throughout the large number of outputs in the collection.
There’s also a section on how Art Blocks operates: the collector mints, the algorithm produces an output, everyone sees it for the first time (including the artist). In most other generative collections the artist has had a chance to curate the outputs before collectors mint them as NFTs. With these collections the artist has a general idea of how the final output looks.
With Bastard GAN Punks V2 there was no curation by Berk. What we have is the direct result of the GAN mind. Typically with long-form PFP collections the artist creates all the traits, and then randomly assigns them to generate the final output. For CryptoPunks these are things like hats, hair, skin color, sex, zombie, etc. BGAN Punks were created by feeding these same traits — plus some Berk originals —to a home-brewed AI known as a Generative Adversarial Network (GAN). On the surface some BGAN Punks may resemble their CryptoPunks parents; but, a deeper dive shows the GAN ignored gender, and colors were assigned randomly, and many have indistinguishable traits where the output was generated somewhere between multiple traits. This is how we get front facing, and X Bastards.
We can see the GAN mind at work by observing HYPED AFs within the collection. Berk also added original generative glitch-art backgrounds to some static outputs, which adds to the overall visual aesthetic and helps to define tiers within the collection. When asked about how the glitchy backgrounds were assigned, Berk responded, “randomly.” But marrying AI and GLIC generated images wasn’t enough, Berk also assigned each Bastard with their own generated punk lyrics pooling from 34 languages — one of which is fictional. These Bastard songs often act as a tie breaker when choosing from two pieces in the same tier.
When Berk was generating the V2 collection he ran the code through five different GANs. This is evident through the outputs generated by the five the types of HYPED AFs. This method is likely responsible for the insane variety seen across the 11,305 outputs in the collection. And because they’re GAN made, there’re no two BGANs who look alike sans an attribute or two; which, is an issue seen in nearly all other generative PFP collections.
BGAN Punks and the Fat Middle
Objectively defining the Fat Middle tiers of BGAN Punks is a difficult task. With similar collections like CryptoPunks collectors can sort by rarity, and then divide the collection into tiers. While there are some traits which are more aesthetically pleasing than other rare traits, buyers still trend towards rarity. With BGANs this is just straight up not the case.
But thankfully the BGAN community has some tools to make this task easier. Pop Rank is probably the most accurate tool we have when it comes to defining the top tier — the grails. For those of you unfamiliar with Pop Rank, it’s basically a game. Two images from the same collection pop on screen, and you click the one you like the most.
Judging from the top 50, top tier traits are HOODIE, TASSLE HAT, GLITCHY AF, double bad habits, HYPED AF, FRONT FACING, and X BASTARD. These traits make up about 32% of the collection. But keep in mind a large percent of FRONT FACING, GLITCHY AF, and double bad habits fall into the middle tiers, and so really all this tells us is these are not bottom tier BGANs.
On the low end the NFTX vault holds 371 BGANs. This is an instant liquidity vault, and selling into it does not guarantee you can buy the same BGAN back in the future. This is where we go when we’re forced to sell, but the vault also generally reflects the current floor price across other marketplaces and traders will sometimes buy/sell into the vault for profit.
So generally people are only selling BGANs into the vault which they either want to trade up for another piece in the vault, or if they don’t mind trading sideways. The NFTX vault is a very good sample of bottom tier traits.
An overwhelming percentage of these NFTX BGANs have NUTTIN’ traits, MEDICAL MASK, and EDIBLES???. Other than MEDICAL MASK these are all basically blank traits and make up a majority of the collection. But the MEDICAL MASK trait is widely regarded as the least aesthetically pleasing trait in the entire collection.
When we use other resources, like Gem.xyz Analytics, to gauge the desirability of these tiered traits it gives a clearer picture of how collectors actually value their BGANs. Below we see MEDICAL MASK in every tier, and FRONT FACING in all except the bottom.
In finding the Fat Middle for BGAN Punks V2, we ignored the outliers on each end. So HOODIE BGANs, POLICE HATS, and BLACK BACKGROUND traits are not counted from the top 50; for, there is a social consensus that each one is a grail. And their rarity has granted them grail status. The same is true for HYPE AFs as they’re in a league of their own, since they don’t share traits with static BGANs. NUTTIN’ EAR and NOSE are excluded because there aren’t enough other traits in those categories and about 85% of the collection shares them. Based on our samples, the Fat Middle for BGAN Punks V2 is 55%, the top tier is 12%, and the bottom is 33%.
