Goal:
- perform Exploratory Data Analysis on historical prices of cryptocurrencies
- filter dataset to rows and columns of interest and make insightful visualizations
- identify potential relationships for deeper research and machine learning application
- in particular, how do other coin prices move relative to Bitcoin (BTC)?
Dataset: "Cryptocurrency Historical Prices" dataset from Kaggle: https://www.kaggle.com/sudalairajkumar/cryptocurrencypricehistory
Historic Prices of top 23 Cryptocurrencies by marketcap
- Date
- daily Prices in USD (High, Low, Open, Close) dating back to April, 2013 (for BTC)
- Volume of daily transactions
- Market Capitalization
* additional columns derived: daily price change, percent change, price up/down, standardized closing price * all numerical data except for the categorical variable 'Up' which represents whether the price went up or not
Why are we interested in how other coins track with Bitcoin (BTC)?
- BTC is 46% of the total cryptocurrency market cap, tracks the entire cryptocurrency market well
- very first (oldest) cryptocurrency created in 2009 (most historic data for this currency)
- most trusted, widely-used cryptocurrency
Historic Bitcoin Price:
- 2 major bubbles: 2017 and 2021
- BTC tracks entire crypto market well
- note on scale: Bitcoin(BTC) in the 34,000 USD, Ethereum 2,100 USD, Cardano 1.21 USD , Dogecoin 0.19 USD
- wide range (orders of magnitude)
Cryptocurrencies by Marketcap and Volume
- coins with the largest marketcaps: Bitcoin (46%), Ethereum (17%), Tether (4.8%), Binance Coin (4.0%), Cardano (2.9%)
- largest average daily volume: Tether, Bitcoin, Ethereum, XRP, Litecoin
Correlation Matrix of Cryptocurrency Prices:
- stable coins (Tether, USDCoin) are uncorrelated or negatively correlated with other coins
- coins with the highest correlations: Polkadot, Uniswap, Cosmos, Ethereum all have correlations that are > 0.90
Standardized Movements of 5 most correlated Coins
- Legend:
- Bitcoin (2013): blue
- Ethereum (2015): green
- Cosmos (2019): orange
- Uniswap (2020): red
- Polkadot (2020): purple
Standardized Movements of last 500 days
- can visually confirm that that highly correlated coins track together
Future Studies:
- make the time series data stationary by removing seasonality and trends
- test for changes before and after big events like the onset of Covid-19 pandemic
- fit predictive models with price as outcome
- incorporate other financial data like interest rate, unemployment rate, and S&P 500 data as possible predictors
- look at more obscure coins with small marketcaps