by: r. Liou (firstname.lastname@example.org)
Summary: In the past year, over 83% of the price variation of NEO can be traced to changes in the price of Bitcoin, negative investor sentiment on overseas-listed Chinese assets, and investor preference toward inflationary safehavens.
In May of 2021, I wrote an article analyzing which commodities/ cryptos most closely influence the price of NEO. Using 4 years worth of financial data obtained from Yahoo Finance, I utilized linear and multivariable regressions within the R programming language to find the factors with the highest R-squared (correlation) score. We were able to tease out a basket of cryptocurrencies and commodities that explained over 85% of the variation in the price of NEO.
In the year since that article has been written, the price of NEO has gone on a rollercoaster ride, skyrocketing past $120 before falling back below $20 (as of May 2022). How great would it be if we could predict which factors most drive NEO’s price, and use them to help prepare us for the next bull or bear run?
This article will utilize the Pandas library in Python as well as price data from Yahoo Finance for various crypto-currencies and commodity indices to determine what factors are most responsible for changes in the price of NEO in the past year (May 2021- May 2022).
To review- a linear regression measures the size/strength of a relationship between variables, and helps us to predict one variable if we know the value of the other. This measurement of relationship is known as the R-squared score. Given an R-squared score of 80% for example, it indicates that 80% of the variation in the output variable is explained by the input variable. Regressions can help us to sift through large amounts of data and make predictions on things from the prices of houses, to used cars, to insurance premiums.
The conclusion from last year’s paper was that using a basket of specific commodities and cryptos, we could create a model which could reliably forecast the price of NEO. However, the basket grew to be quite large, and many of the variables had overlapping features, making it hard to use it for predictive purposes.
This ‘kitchen sink’ approach included a number of cryptocurrencies as well as all the prices of a handful of commodities including lumber, crude oil, and gold. It returned an R-squared score of 83%.
From 2021-2022, this same basket of commodities returned an R-squared of 73.1%, indicating that the model still has some predictive power. However, as it contains a lot of variables, many of which have overlapping traits, it becomes hard to use. Would it be possible to reduce the number of variables we have in our basket, thus simplifying the model? And with the price of NEO diverging from other chains, can we pinpoint what has caused this divergence in the past year? To answer these questions, let’s use Pandas to perform regressions on each of the two asset classes that we examined last year: other cryptocurrencies and inflation hedges.
From the period 09/17- 04/21, we had the following R-squared results with NEO:
- Tron (44.4%)
- Bitcoin (36.6%)
- ETH (36.4%)
- Binance (7.4%)
Back then, the price of NEO was much more strongly correlated with the price of Tron, ETH, and Bitcoin, than it was with Binance. But did this still hold true from the period 05/2021-05/2022?
Interestingly, in the past year, the correlation with Binance has dropped to just .00037%. Meanwhile, for ETH, the correlation dropped to just .009%.
In other words, the price trends of Binance and ETH have diverged strongly from the price of NEO.
Tron’s regression, meanwhile, declined to just 22% in the past year.
Surprisingly, from 2021-2022, NEO’s correlation with BTC has risen from 36.6% all the way to 57.4%.
We suspect one reason may be because Ethereum and Binance gained popularity in the past year, at the expense of Bitcoin.
Meanwhile, another crypto that had a high correlation with NEO (but in a negative way) was Terraluna, at 53%:
This indicates that as the price of NEO was falling, the prices of Terraluna tended to rise. However, it should be noted that with the recent free fall in price of Terra Luna the past week, the R-squared regression with NEO has dropped to 43%. Therefore, this may portend a reversal in trend with NEO going forward.
From the period 05/21- 05/22, we had the following R-squared results with NEO:
- Bitcoin (57.2%)
- TerraLuna (53%)
- Tron (22%)
- ETH (0.9%)
- Binance (.01%)
From an inflation-hedge (commodities) perspective, we wanted to see if investors viewed NEO as a hedge against rising inflation. Therefore, we picked some common commodities indices and did a linear regression against NEO to find the R-squared score.
From 2017-2021, the commodities with the strongest correlation to NEO were:
- Lumber (13.99%)
- Crude oil- Brent (10.4%)
- Corn (4.3%)
- Gold (3%)
- Stocks (DJIA) (.1%)
However, in the past year, the correlation to commodities has become much stronger, in fact more than doubling!
NEO/Corn has risen to 42.3% from 2021-22
NEO/Lumber rose to 32.3% from 2021-2022
NEO/Gold rose to 30.2%. from 2021-2022
NEO/Dow Jones rose to 10% from 2021-2022
However, by looking at the graphs, we can see that the correlation against commodities is negative. This tells us that in times of high inflation, people began to stock up on commodities such as lumber or corn, but at the expense of selling NEO.
Going back to our model, if we stopped here and did a multivariable regression between the two best performing variables in each class, say BTC and corn, with NEO, our total R-squared score is just 4 2.9%. Not very high or convincing just yet.
Therefore, let’s ask ourselves-
What other factors may have affected the price of NEO that are testable? Has NEO fallen victim to anti- China sentiment, similar to US – listed Chinese stocks? In this case, the MCHI ETF, an index of 50 popular overseas-listed Chinese stocks such as Alibaba, etc. , may serve as a useful proxy for us.
Let’s run the regression for MCHI and NEO to see what we get? A correlation of 68%- That demonstrates a pretty strong correlation.
This tells us that as investor appetite for US-listed stocks such as Alibaba and Didi waned in 2021, the same happened to NEO.
So it appears that as we have identified some key variables that have changed in the previous year, perhaps we can edit and reduce our basket of goods to a smaller and more focused set of variables. Let’s pick one variable from each class and see what we get.
From cryptocurrencies, NEO’s price is much more in line with BTC, so we will select that as a variable for our model.
What about inflation hedges? NEO’s price is more in line with traditional inflation hedges than before, though from a negative relationship perspective. The correlations for both corn and lumber have increased significantly from 2021-22, so we can pick one of those two.
Lastly, since we got a good result with MCHI, we’ll add that to our basket.
Our final, simplified basket is:
And the result is:
An r-squared of 83.4%
BTC-lumber-MCHI is even higher, at 84.2%.
Wow- Quite impressive! This high R-squared indicates a strong probability that our model has good predictive capability for the price of NEO.
Let’s backtest it over the period 2016-2021. We get: 68.9%
If we run the regression over the entire period 2016-2022, however, the cumulative R2 is 80.5%, which is still a very good score. Great!
Let’s summarize what these conclusions are telling us. There are 3 correlations that increased significantly from 2021-22: NEO and BTC, NEO and commodity staples such as corn and lumber, and NEO and overseas-listed Chinese assets.
As far as the recent decline in price of NEO is concerned, it is the combined result of the recent plunge in the price of Bitcoin (but not of ETH or BNB), investor sentiment in overseas-listed Chinese assets, and investor bias toward inflation-hedge commodity safehavens like corn and lumber. Forecasting the direction of these 3 trends would strongly assist a NEO investor to determine what the future NEO price will be.
Furthermore, NEO’s price divergence away from ETH/BNB/Tron has only exacerbated these trends.
To mitigate these trends, one strategy NEO could attempt is during tough economic times such as now, to adopt more real-world usage so that investors will feel comfortable holding it as they would an inflation-safe commodity such as corn or lumber.
Future articles may focus on how to do additional forecasting and analysis within these areas.
Thanks for reading, and please don’t hesitate to share any comments or feedback.