1. Bitcoin. Cryptocurrencies. So hot right now.
Since the launch of Bitcoin in 2008, hundreds of similar projects based on the blockchain technology have emerged. We call these cryptocurrencies (also coins or cryptos in the Internet slang). Some are extremely valuable nowadays, and others may have the potential to become extremely valuable in the future*. In fact, the 6th of December of 2017 Bitcoin has a market capitalization above $200 billion.
The astonishing increase of Bitcoin market capitalization in 2017
* WARNING: The cryptocurrency market is exceptionally volatile and any money you put in might disappear into thin air. Cryptocurrencies mentioned here might be scams similar to Ponzi Schemes or have many other issues (overvaluation, technical, etc.). Please do not mistake this for investment advice.
That said, let’s get to business. As a first task, we will load the current data from the coinmarketcap API and display it in the output.
# Importing pandas import pandas as pd # Importing matplotlib and setting aesthetics for plotting later. import matplotlib.pyplot as plt %matplotlib inline %config InlineBackend.figure_format = 'svg' plt.style.use('ggplot') # Reading in current data from coinmarketcap.com current = pd.read_json("https://api.coinmarketcap.com/v1/ticker") # Printing out the first few lines current.head()
2. Full dataset, filtering, and reproducibility
The previous API call returns only the first 100 coins, and we want to explore as many coins as possible. Moreover, we can’t produce reproducible analysis with live online data. To solve these problems, we will load a CSV we conveniently saved on the 6th of December of 2017 using the API call
# Reading datasets/coinmarketcap_06122017.csv into pandas dec6 = pd.read_csv("datasets/coinmarketcap_06122017.csv") # Selecting the 'id' and the 'market_cap_usd' columns market_cap_raw = dec6.loc[:,['id', 'market_cap_usd']] # Counting the number of values market_cap_raw.count()
id 1326 market_cap_usd 1031 dtype: int64
3. Discard the cryptocurrencies without a market capitalization
Why do the
market_cap_usd differ above? It is because some cryptocurrencies listed in coinmarketcap.com have no known market capitalization, this is represented by
NaN in the data, and
NaN's are not counted by
count(). These cryptocurrencies are of little interest to us in this analysis, so they are safe to remove.
# Filtering out rows without a market capitalization #cap = market_cap_raw.query('market_cap_usd > 0') # this would be a better way to do this cap = market_cap_raw[market_cap_raw['market_cap_usd'] > 0] # Counting the number of values again cap.count()
id 1031 market_cap_usd 1031 dtype: int64
4. How big is Bitcoin compared with the rest of the cryptocurrencies?
At the time of writing, Bitcoin is under serious competition from other projects, but it is still dominant in market capitalization. Let’s plot the market capitalization for the top 10 coins as a barplot to better visualize this.
#Declaring these now for later use in the plots TOP_CAP_TITLE = 'Top 10 market capitalization' TOP_CAP_YLABEL = '% of total cap' # Selecting the first 10 rows and setting the index cap10 = cap.iloc[0:10,].set_index('id') # Calculating market_cap_perc cap10 = cap10.assign(cap_perc = lambda x: cap10.market_cap_usd / cap.market_cap_usd.sum() * 100) # Plotting the barplot with the title defined above ax = cap10['cap_perc'].plot.bar(title=TOP_CAP_TITLE) # Annotating the y axis with the label defined above ax.set(ylabel=TOP_CAP_YLABEL, xlabel="")
[Text(0,0.5,'% of total cap'), Text(0.5,0,'')]
5. Making the plot easier to read and more informative
While the plot above is informative enough, it can be improved. Bitcoin is too big, and the other coins are hard to distinguish because of this. Instead of the percentage, let’s use a log^10 scale of the “raw” capitalization. Plus, let’s use color to group similar coins and make the plot more informative*. For the colors rationale: bitcoin-cash and bitcoin-gold are forks of the bitcoin blockchain**. Ethereum and Cardano both offer Turing Complete smart contracts. Iota and Ripple are not minable. Dash, Litecoin, and Monero get their own color.
- * This coloring is a simplification. There are more differences and similarities that are not being represented here.
- ** The bitcoin forks are actually very different, but it is out of scope to talk about them here. Please see the warning above and do your own research.
# Colors for the bar plot COLORS = ['orange', 'green', 'orange', 'cyan', 'cyan', 'blue', 'silver', 'orange', 'red', 'green'] # Plotting market_cap_usd as before but adding the colors and scaling the y-axis ax = cap10["market_cap_usd"].plot.bar(color=COLORS, logy=True) # Annotating the y axis with 'USD' ax.set(ylabel="USD") # Final touch! Removing the xlabel as it is not very informative ax.set(xlabel="") # Is there a better way to remove a label ? plt.show()
6. What is going on?! Volatility in cryptocurrencies
The cryptocurrencies market has been spectacularly volatile since the first exchange opened. This notebook didn’t start with a big, bold warning for nothing. Let’s explore this volatility a bit more! We will begin by selecting and plotting the 24 hours and 7 days percentage change, which we already have available.
# Selecting the id, percent_change_24h and percent_change_7d columns volatility = dec6.loc[:,["id", "percent_change_24h", "percent_change_7d"]] # Setting the index to 'id' and dropping all NaN rows volatility = volatility.dropna().set_index("id") # Sorting the DataFrame by percent_change_24h in ascending order volatility = volatility.sort_values("percent_change_24h", ascending=True) # Checking the first few rows volatility.head()
7. Well, we can already see that things are a bit crazy
It seems you can lose a lot of money quickly on cryptocurrencies. Let’s plot the top 10 biggest gainers and top 10 losers in market capitalization.
#Defining a function with 2 parameters, the series to plot and the title def top10_subplot(volatility_series, title): # Making the subplot and the figure for two side by side plots fig, axes = plt.subplots(nrows=1, ncols=2, figsize=(10, 6)) # Plotting with pandas the barchart for the top 10 losers ax = volatility_series[:10].plot.bar(color="darkred", ax=axes) # Setting the figure's main title to the text passed as parameter fig.suptitle(title) # Setting the ylabel to '% change' axes.set(ylabel="% change", xlabel="") # Same as above, but for the top 10 winners ax = volatility_series[-10:].plot.bar(color="darkblue", ax=axes) axes.set(xlabel="") # Returning this for good practice, might use later return fig, ax DTITLE = "24 hours top losers and winners" # Calling the function above with the 24 hours period series and title DTITLE fig, ax = top10_subplot(volatility["percent_change_24h"], DTITLE)
8. Ok, those are… interesting. Let’s check the weekly Series too.
800% daily increase?! Why are we doing this tutorial and not buying random coins?* After calming down, let’s reuse the function defined above to see what is going weekly instead of daily.
- Please take a moment to understand the implications of the red plots on how much value some cryptocurrencies lose in such short periods of time
# Sorting in ascending order volatility7d = volatility.sort_values("percent_change_7d", ascending=True) WTITLE = "Weekly top losers and winners" # Calling the top10_subplot function fig, ax = top10_subplot(volatility7d["percent_change_7d"], WTITLE)
9. How small is small?
The names of the cryptocurrencies above are quite unknown, and there is a considerable fluctuation between the 1 and 7 days percentage changes. As with stocks, and many other financial products, the smaller the capitalization, the bigger the risk and reward. Smaller cryptocurrencies are less stable projects in general, and therefore even riskier investments than the bigger ones*. Let’s classify our dataset based on Investopedia’s capitalization definitions for company stocks.
- Cryptocurrencies are a new asset class, so they are not directly comparable to stocks. Furthermore, there are no limits set in stone for what a “small” or “large” stock is. Finally, some investors argue that bitcoin is similar to gold, this would make them more comparable to a commodity instead.
# Selecting everything bigger than 10 billion largecaps = cap.query("market_cap_usd > 10e9") # Printing out largecaps largecaps.head()
10. Most coins are tiny
Note that many coins are not comparable to large companies in market cap, so let’s divert from the original Investopedia definition by merging categories. This is all for now. Thanks for reading!
# Making a nice function for counting different marketcaps from the # "cap" DataFrame. Returns an int. # INSTRUCTORS NOTE: Since you made it to the end, consider it a gift :D def capcount(query_string): return cap.query(query_string).count().id # Labels for the plot LABELS = ["biggish", "micro", "nano"] # Using capcount count the biggish cryptos (Greater than 300 000 000) biggish = capcount("market_cap_usd >= 3e8") # Same as above for micro ... (Greater than 50 000 000 Less than 300 000 000) micro = capcount("market_cap_usd >= 50e6 and market_cap_usd < 3e8") # ... and for nano (Greater than 50 000 000) nano = capcount("market_cap_usd <= 50e6") # Making a list with the 3 counts values = [biggish, micro, nano] # Plotting them with matplotlib plt.bar(LABELS, values)
<BarContainer object of 3 artists>