In this post, we look at correlation of the cryptocurrency market, using Python, to analyze and visualize how these markets have been evolving, and provide a base framework for deeper analysis.
New to crypto? You can find a great overview about block-chain technology, cryptocurrency and how they are related to Satoshi Nakamoto’s innovation.
TLDR: The correlation and trend between various cryptocurrencies are evaluated to understand:
- Which coins can help diversify your portfolio: There is generally a strong correlation between coins, hence it is difficult to diversify risk (Just buying and holding Bitcoin is a safe bet).
- Are stable coins actually stable: Generally they appear to be decoupled from altcoin price fluctuations.
- If it is possible to predict the next bull run: Speculative at best. Nevertheless, end of 2019 looks similar to 2015, 2016 and 2017 … get ready to fill your bags.
Note: All code used for the analysis can he found on my github page and the data comes from coinmarketcap.com, from 2014 - 2019, for the top 100 cryptocurrencies by market cap.
As can be seen from the figure above, in general, the trends for each coin appears to be correlated. However, it is well known that calculating the correlations directly on a non-stationary time series data can give biased correlation values . Hence, the daily percentage change is used to evaluate the correlation. Below are the correlations of the price change for the top 20 coins by market cap.
In general, three main trends stand out for the top 20 cryptocurrencies by market cap:
- All the stable coins (a cryptocurrency that is pegged to fiat or some other commodity e.g. Tether is pegged to the USD) and Ravencoin, has a very low correlation with percent price change of bitcoin and the other altcoins.
- There is a general trend of the correlation increasing over the years as shown in the plot below.
- Only 45% of the coins currently in the top 20 existed 5 years ago.
It is notable that almost all of the cryptocurrencies have become more correlated with each other across the board. In practice, when investing in the stock market, having a diversified portfolio reduces risk. However, having a diversified portfolio of the top 20 coins, does not have the same effect as in trading stocks.
Interesting enough, Ravencoin, which is a fork of Bitcoin with the added feature of asset issuance, and transfer has a very low correlation with Bitcoin and most of the altcoins in the top 20. I am not sure of the reason for this, but it is definitely one of the coins to keep an eye on.
Also, while Bitcoin, Etherum and Litecoin has been in the top 5 for the past 5 years, it would be interesting to further evaluate what are the drivers (technology, marketing, innovation, use case) for other coins moving into the top 20. However, this is beyond the scope of this post.
Stable coins such as Tether, appear to be uncorrelated with the fluctuations in the bitcoin price. Stable coins help eliminate doubt about conversion rates, making cryptocurrencies more practical for buying goods and services. This is so because the underlying asset is not correlated with Bitcoin. However, the main criticisms of stable coins (besides those that are centralized), is proving that the amount of the commodity/fiat used to back the stable coin, reflects the circulating supply of the stable coin.
This analysis is largely speculative, and is based on the box plots above, which is grouped by months, for the past 6 years. In 2015, 2016 and 2017 we see a spike in the Bitcoin price around months 6 and 7, after which we see a bull run at the end of the year.
Likewise, 2014 and 2018 are quite similar, where there is growth in months 5 and 6 but the months following, are sluggish and signals a bearish market.
For 2019, if august looks as strong as July, we could be in for another bull run breaking previous records based on the trends for 2015, 2016 and 2017.
When should we expect to see these highs … the Histogram below shows that the Bitcoin highs, tend to take place in in December and January (Signified by the lager number of outliers in those months).
There is a lot more that can be done to build on this analysis. A few that come to mind are:
- Use API from a popular exchange to have an up to date set.
- Add other metrics such as market cap and volume to enrich the analysis.
- Add some sort of sentiment analysis based on google tends or twitter feed analysis to predict future prices.
- Train a predictive machine learning model to predict future prices.
Thanks for reading, and hopefully this has sparked your interest in the field. Constructive feedback is welcomed and if you find problems with the code, you can open an issue in the Github repo here.
Note from Towards Data Science’s editors: While we allow independent authors to publish articles in accordance with our rules and guidelines, we do not endorse each author’s contribution. You should not rely on an author’s works without seeking professional advice. See our Reader Terms for details.
