The lack of liquidity in these markets is well known, with the term ‘illiquid JPEGs’ being used tongue-in-cheek by some when referring to NFTs. Low liquidity hinders price discovery and burdens market participants with the possibility of never being able to sell their assets.

To quantify the illiquidity of the current market, let's look at some numbers for Bored Ape Yacht Club (BAYC), one of the most popular NFT collections. At the time of writing, there have been a total of around 23,000 sales for a collection that contains 10,000 assets. This means that the collection averaged only 2.3 sales per asset since its release over 8 months ago. To put it in simpler terms, if you were to use the asset as collateral for a loan, the liquidation engine could only liquidate 2.3 times in 8 months

Trying to bring liquidity to a market that is by definition not uniform, is a challenge. Rather than minimizing the spread between buyers and sellers of a single asset, one needs to minimize the spread between highly similar assets, as well as the spread between all other asset groups in any given collection, depending on the relationships that exist between them. Bringing those two groups closer together is where market makers actually start creating value for the industry

Market making is the process of offering simultaneous buy and sell quotations for a given asset. When performed correctly, the market benefits from tighter spreads and deeper liquidity, making it easier for market participants to buy and sell their assets. In this context, the spread refers to the difference between the highest bid (buy) price and the lowest ask (sell) price. Market makers are compensated for their efforts by capitalizing on the bid-ask spread.

We believe that the challenge posed by pricing non-fungible assets is at least partly responsible for the liquidity shortages the NFT market is experiencing. Over the past several months, our team has been working on something that we hope you will find useful and interesting. Special thanks here to Furkan, Cyrill, Karsten, Gill and Ahmad from NUMEUS.

Niftymate

TL;DR: Niftymate is a free-to-use, on-chain price stream offering up-to-date ‘fair value’ pricing estimates for NFT collections. We currently support the top 13 NFT collections on OpenSea by volume and aim to add support for an additional 100 collections over the coming weeks.

What exactly is a ‘Niftymate’?

In short it’s an accurate estimate for the current ‘fair’ price of an NFT asset based on a combination of rarity, similarity, transaction history, on-chain data, sentiment analysis and some other stuff that we’re keeping secret for now! In theory, our pricing engine is able to accurately price any collection, provided there is sufficient available data out there.

We don’t claim to have built an omniscient model - there are some variables that simply cannot be modelled, but what we have found is that these models are able to price individual assets with remarkable accuracy. You will have noticed that we refer to our pricing as a reflection of the ‘fair’ value for a given asset, but what exactly does this mean, particularly for something as subjective as art? We explore this a little further below and will go into more depth in subsequent posts.

We invite you to test out the stream and start playing around with it. We’re excited to see where and how Niftymate will be integrated and look to the community to provide feedback that will help shape future iterations.

Official NUMEUS price issuing address on Arweave

xyBU2hGQqOFjtoFnopyV_PfTilnpb63sqZDNNlsJhow

1. Collection Slug: BoredApeYachtClub

2. ​​Collection Contract Address: 0xbc4ca0eda7647a8ab7c2061c2e118a18a936f13d

3. App-Name: NiftymateByNumeus

4. App-Version: 0.0.1

5. Time Batch: 2021-12-03T13:59:13Z

Contained within the batch files is a Niftymate price (in ETH) for every single asset in a given collection, irrespective of whether it is on sale or not.

Here is a quick guide on how to query the Niftymate price stream: https://github.com/numeusxyz/niftymate-stream

Why we built it

Niftymate is free for everyone to use. In building this tool, we wanted to give something back to the community that has given us so much over the years.

We believe Niftymate benefits both buyers and sellers. Prospective buyers are better informed as to what the ‘fair’ value of a given NFT is, whilst prospective sellers have a ‘fair’ value upon which to base their listing price. It goes without saying that buyers and sellers are free to transact at whatever price they see fit. Our hope is that this will ultimately result in a more efficient and liquid market which would be a win for the industry as a whole.

Furthermore, we encourage the development of competing and complementary products. Other teams in the space are approaching the problem of NFT pricing from a different angle e.g. using peer prediction mechanisms to crowdsource NFT appraisals. We envision a future where price streams from various sources can be combined with different weightings depending on the application or end-user, thereby preventing overreliance on any one source.

How we built it

For the time being we are keeping the secret sauce that drives our pricing engine just that. Instead we would rather give you a brief overview of how the price stream was created and what goes into creating a “Niftymate”.

The first step was to gather and import all the available on-chain and off chain data. This includes things like the metadata and properties of the NFTs themselves, as well as recent transactions, transfer history, and wallet distribution.

Next, we cleaned up the data before feeding it into the models. Whilst we believe the majority of NFT volume is organically driven - that is real trades between unrelated parties - one cannot ignore the fact that there are clear instances of washtrading by a few actors within this space. We will refrain from commenting on the possible motivations behind these actions and naturally need to be very careful with such statements. As such our pricing engine is able to precisely identify and filter out actual washtrades from hype driven, but otherwise genuine transactions.

We then set about building models to price NFTs (Non Fungible Tokens). Perhaps unsurprisingly, it’s very difficult to provide a ‘fair’ price for individual pieces of art (e.g. 1of1’s), as their value is almost entirely subjective. When pricing collections however, the task becomes a little easier as we are able to use reference values and gather significantly more data points and relationship data to feed into our models.

Since most NFTs outside of gaming do not yet possess much utility besides use as a PFP or being displayed in a gallery, we gave a lot of weight to the rarity scores within a collection to make sense of it all. We anticipate that this will change over time as digital experiences are built around them and will adapt the models to reflect this in due course. Market beta becomes a key driver with rarity scores and similarity scores lead to the first grouped pricing.The pricing model also takes group dynamics, subjectivity (to a degree) and social psychology into account. We have seen what an influence pop culture and meme value can have on price and whilst some have lasting effects, most are fleeting moments of unreasonable subjectivity (i.e. hype). As a result, we are working on ways to identify and flag moments of hype to give markets a more accurate indication of driving forces.

Interestingly enough social psychology is actually driven by quantifiable indicators underneath such as last paid price and the speed at which the pricing curve evolves. In social behaviour we often observe that a given audience will look at historic price evolution and use that as a calculation baseline of how much premium they're willing to pay to hopefully take a profit within the same timeframe and the same price evolution down the line. By paying this premium on top, the buyer even increases the angle of the price curve and thus reinforces that cycle into a “hype”.

To bring the data on-chain, we construct a dictionary for each collection which contains collection slug (as found in OpenSea), date and time at which price predictions are generated, list of token ids and list of predicted prices. After that dictionary is constructed for all collections we generate predictions for all assets in a given collection in the format of a single JSON file. We then upload the JSON file to Arweave network using a simple Python script.

We chose Arweave for its:

1. Data availability and data permanence

2. Cost efficiency - given how our price stream is free to use, it wasn’t economically viable for us to fund the cost of an oracle data solution on a different chain

3. Available python libraries - this allows the data to easily be queried by dapps

4. Simplicity - it took us 30 minutes from reading into it, to publishing the data for the first time.

Roadmap

Everyone loves a roadmap, so here’s ours. We’d like you to think of V0.0.1 as a “teaser” version of the Niftymate price stream. We are currently working on some major upgrades to the pricing engine and our integrations, but hope that this first iteration gets the conversation going - after all we wouldn’t want to be the only ones building something like this!

Integrations

The most obvious use cases for Niftymate currently include NFT marketplaces and collateralized NFT lending platforms. We are currently working together with a number of projects who are excited about integrating our price stream.

More collections

We currently support the top 13 NFT collections on OpenSea by volume and aim to add support for an additional 100 collections over the coming weeks, with a view to eventually pricing 1000s of collections in the future.

Price predictions

In V0.0.1 all the necessary data is updated (not published) on a per-block basis. Future versions will source data directly from the mempools, further improving speed and allowing the model to predict future price movements.

Multi-chain

In the near-term, we are working on bringing Niftymate to various L2s and potentially even Ethereum Mainnet. We are also exploring the idea of issuing price stream updates in the form of NFTs - where pricing data is contained within the metadata of the NFT itself. This would allow Smart Contracts and Marketplaces to work with significantly less outside sources, thus reducing the number of attack vectors.

Washtrade filtering

The pricing engine is able to identify and flag suspected washtrades, preventing them from affecting price. We invite the community to provide feedback on whether we should include the option of switching off washtrade filtering.

Speed

Currently, the price stream is pushing updated prices every 4 hours. We are working on updates that will significantly increase the frequency of these updates (think seconds, not hours).

Decentralization

There’s no getting around the fact that in its current form, the price stream is highly centralised. Whilst we are staking our reputation on this, we acknowledge this does not provide sufficient guarantees and aim to open source aspects of our price stream over time.

Additionally, we are working on a supplementary product that we’re referring to as a ‘risk stream’. This is a collection/asset risk monitoring system intended for use by lending platforms to protect lenders and borrowers. The risk stream will flag collections with highly centralized holdings, floor sweeping by large asset holders, a decrease in liquidity, a decrease in participation of wallets, scams/hacks to name a few. We will be completely open sourcing this tool and open up its governance to the community.

We hope the community is as excited about Niftymate as we are and we can’t wait to share more details soon. In future posts, we’ll be dropping some alpha and doing a deeper dive on the underlying tech so be sure to follow along! You can find us at:

Twitter: @numeusxyz

GM.xyz: numeus.eth

ENS: numeus.eth

Website: numeus.xyz

We invite anyone looking to integrate our price stream into their project or provide feedback to reach out to the team @numeusxyz or Cyrill @cybedifferent.