Coinbase Algo Trading (based on Elon's thoughts..haha)

Description

• A simple to use repo to automate trading crypto on Coinbase
• It listens to Elon Musk's related accounts (@elonmusk, @tesla, @spacex) in real-time to see if either account is posting anything related to bitcoin
• Each tweet's polarity (sentiment) is analyzed and if the polarity is greater than a specific threshold, the algorithm sends a BUY market order to coinbase

Under Development

• Advanced algorithm to minimize execution slippage
• Additional statistics to improve execution performance
• Portfolio management to analyze open positions
• VWAP execution algorithm
• PAIRS execution algorithm based on statistical arbitrage vs other crypto pairs

Getting Started

• Clone this repository
• Sign up for a Coinbase Account
• Generate API credentials for the Coinbase Sandbox Account (https://public.sandbox.pro.coinbase.com/), so you don't run a real P&L initially!!
• Enter your API credentials in the script coinbase_creds.py like the following:

api_secret = "SOME_STRING"
api_key = "SOME_STRING"
api_pass = "SOME_STRING"

• Sign up for a twitter developer account and generate API credentials to enter into the twitter_credentials.py script like the following:

# Twitter API Credentials
API_KEY = "SOME_STRING"
API_SECRET_KEY = "SOME_STRING"
BEARER_TOKEN = "SOME_STRING"
ACCESS_TOKEN = "SOME_STRING"
ACCESS_TOKEN_SECRET = "SOME_STRING"

• You're now pretty much good to go, just run the following:

python coinbase_algo.py

Under the hood

Just for fun, let's dig into what the coinbase_algo.py script is doing in a bit more detail

• We utilise the cbpro library to open a websocket connection to receive data in real-time from coinbase (please ensure you point to the sandbox before going live!!)

class TextWebsocketClient(cbpro.WebsocketClient):
    def on_open(self):
        self.url           = 'wss://ws-feed-public.sandbox.pro.coinbase.com'
        self.message_count = 0
        self.initial_date = datetime.now()

• We place a market order, worth $1,000, to buy BTC if Elon says something pretty +ve about it

# ------------------------
# Trade based on the below
# ------------------------

# Get df with Elon's related tweets from
df = get_elons_tweets()
# Filter the df to listen out for upcoming tweets
filtered_df = df[(df['date'] > self.initial_date) & (df['crypto_mention'] == True)]

if len(filtered_df) < 1:
    print('Nothing to look for yet')
else:
    # Get average polarity
    sentiment_array = np.array(filtered_df['sentiment'])
    average_polarity = np.mean(sentiment_array)

    # Check if the statements are +ve enough
    if average_polarity > 0.6:
        # Send a buy order to coinbase for $1000 of bitcoin
        # # Authenticated Client
        url='https://api-public.sandbox.pro.coinbase.com'

        client = cbpro.AuthenticatedClient(
            api_key,
            api_secret,
            api_pass,
            api_url=url
        )
        client.place_market_order(product_id='BTC-USD',side='buy',funds=1000)