Skip to content
All library documents

Modeling NFT Collection Floor Price Distributions

Article Quant Q&A · Author: Kalev Maricq
Read the original

Summary

The document frames the problem of forecasting a collection’s lowest NFT ask from current bid and ask information. It contrasts modeling every NFT as a correlated geometric Brownian motion with modeling the floor itself as a geometric Brownian motion. The first approach is computationally demanding and depends on estimating a very large covariance matrix; the second is more tractable but may miss how the floor forms.

The author highlights discrete jumps and asymmetric price moves: bullish demand can quickly lift the floor as buyers take existing listings, while a lower floor requires an owner to submit a cheaper ask. The document offers no tested model, data, or empirical results; it is a request for literature and modeling ideas. Its discussion is specific to NFT collection floors, and the proposed dynamics remain open questions rather than validated methods.

Key ideas

  • The target is the future distribution of the minimum ask across an NFT collection.
  • Modeling each token separately creates a difficult covariance estimation problem.
  • Treating the floor as geometric Brownian motion is simpler but may misrepresent how listings change.
  • Floor prices may react asymmetrically to buying pressure and newly submitted lower asks.
  • The document proposes no empirical evidence or established solution.

Tags

CryptoStatisticsMarket microstructureOn-chain data
Full text
# NFT Floor Price


# NFT Floor Price












I'm interested in modeling NFT Floor Price. Specifically, I'm trying to answer the question:

> Given current bid-ask info on an NFT collection, what is the probability distribution of the lowest ask on any NFT in the collection at some future time t?

So far, I've tried:

- Searching for existing literature on modeling NFTs. I've found a lot about how to try to speculate on the price of a specific NFT, but very little on modeling them in the abstract.

- Searching for literature on modeling low-liquidity assets, including real-estate. I've found a lot about valuation, but very little about distributions.

- Modeling NFTs in a collection as 10,000 highly correlated assets with geometric brownian motion and then finding the minimum. Not only is this computationally intensive, but the massive covariance matrix is virtually impossible to estimate properly.

- Treating the floor price itself as a geometric brownian motion variable. While this is much more tractable, I'm not sure if it's theoretically sound. It also doesn't account for discrete jumps in the floor price and assymetry between upward and downward movement (the floor price can react very quickly to bullish news as speculators buy existing NFTs, but in order to drop, one of the 10k individuals in the world who own one must submit an ask lower than the current floor).

Is there any existing literature on modeling the future price distributions of illiquid assets or portfolio order statistics? Any ideas on how to modify a standard geometric brownian motion/lognormal model to incorporate bid/ask info, allow for discrete jumps, or have assymetric movement?

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.