Lil Nouns is a spin of Nouns DAO, where every 15 minutes a new randomly generated piece of art of certain characteristics is created. Users can bid and after a certain point (settling) the highest bidder wins. Since they are minted every fifteen minutes, do we see much action on the secondary market, specifically within 24 hours? This analysis does a deep dive into the lil nouns secondary marketplace

Additionally python code for visualization, processing, and ML models can be found here: https://github.com/jhackworth42/Crypto-Analytics/blob/main/lil_nouns.ipynb

Introudction

Header

As of writng there has been over 135 sales within 24 hours representing close to selling withing 24 hours being 5% of mints. Also, there is a decent premium between mint and secondary sell with a rolling average of .24 ETH.

Below, we can see an interesting trend emerge. Looking at secondary selling within 24 hours we can see two things off the bat. 1) The price difference over time between minting and secondary selling within 24 hours has gone down but has remained somewhat steady now with some variations. While the rolling average is .24, the average price premium is more like .15 ETH now 2) # of Bids does not appear to be a factor in secondary selling (initial thought would be more bids, the more popular and thus the desire for a higher price on the secondary market) 3) Opensea is close to where all the trading is done for lil nouns (this is also just the overall trend of the market though)

So what factors may go into putting a Lil noun on the market and are there premiums on price? I broke out python to do a quicker analysis to loop through the visualizations (Github link found here: https://github.com/jhackworth42/Crypto-Analytics/blob/main/lil_nouns.ipynb)

Here are the top insights for this piece:

1. Sellers tend to sell in the later hours of the day and earlier, however, the price difference is increased in the early morning and midday. In fact, we even see one hour with a negative price difference!
2. Glasses in terms of # of sales are somewhat evenly distributed but see price difference premium for 0,7, and 10 characteristics. We can see a similar trend for other characteristics where the more sales, the less likely it is going to have as high of a premium
3. Popular days to sell are Wednesday and Friday but fall in the back in terms of the price difference with Sunday having the highest at.3 ETH
4. While having almost the same amount of sales, the background characteristic of 1 carries a slight premium

To see the rest of the outputs check my github!

Characteristics

So why would someone buy vs mint? Are people selling for profit or just because they don't like their lil noun? In the below tables we can gather a few insights for buyers and sellers

For Sellers, just because you spent a lot of ETH on mints does show a relationship with the average amount of eth profited (excluding gas and creator fees). Those that spent the least amount of ETH were the ones who probably thought they could make a quick buck and then came out to negative or slightly above what they made. The highest number of sales one address did is 9 for the timeframe but has came second in overall profit (highest total profit only had 4 sales)

Sellers / Buyers of Seconday

For buyers, we see that there are some buyers spending a decent amount of eth on secondary sales. Most wallets only do 1 buy of a lil noun, suggesting these collectors may not be pouncing on lil Nouns. The distribution of buys vs sells also seems to converge around .3 ETH the further # of buys they do,

Machine Learning Analysis

Could we be able to predict the price difference is for this lil noun dataset? The dataset is extremely limited so there is probably a bit of overfitting of the data. Building a quick Random Forest Regressor, we can predict the price and the price premium (slight difference in variables)in ETH. For the RMSE of the first price prediction, we got .16 ETH (Our prediction will stay within this bound) and for the price premium, we got .21 ETH. Since this is such a limited dataset this won't reveal much. However, we can see what features our model thought were important. Obviously, time was an important factor given the downward trend, but we also see that certain days, the hour of the day, and characteristics, played an effect in our models. Once again the code for the models can be found. I wouldn't use this model for trading, but expanding the dataset might be helpful for all Lil noun trades!

Shap Plot for Random Forest Price Prediction (Feature Importance)

Overall this was a pretty fun experiment. We saw there are users selling within 24 hours, but its quite small (users holder on longer to their lil noun) The price premium has been going down most likely due to the overall market and the increasing supply of lil nouns. Certain users are quite skilled at buying/selling lil nouns. Finally, we did see that the time of day, weekday, and characteristics such as glasses and body time can make a value more desirable. I would hope to continue my analysis on lil nouns by expanding to the overall datset and finding some more insights on a broader dataset

Shap Plot for Price Premium (Feature Importance)

Conclusion