Running the application:

• create a virtual environment: python3 -m venv venv
• activate virtual environment: source venv/bin/activate
• installing packages: pip install -r requirements.txt
• create .env file with API keys from Alpaca Trading API:

API_KEY = 
API_SECRET = 

• run application: streamlit run app.py

Process:

• building strategy structure
• position sizing: cash at risk metric = how much of our cash balance do we want to risk at every trade
• incorporating informative news data
• develop ML algorithm for sentiment analysis
• incorporate ML predictions into trading strategy
  • if positive sentiment (expected increase in asset prices): close shorts, place buy
  • if negative sentiment (expected decrease in asset prices): close long, place sell
• interactive interfaces

technical challenge

1. leveraging the api: trial and error & debugging
2. continuously improvising trading strategy with the integration of new informative data:

• intially only relying on prices: limited -> cash management strategy stop loss and take profit
• dynamically incorporating new information (news: sentiment of the market) & data processing -> sentiment analysis: only want to trade on strong positive or negative sentiment (need proabability)
  1. NLTK that counts the occurence of positive & negative words --> doesn't consider context
  2. RNN that has vanishing gradient problem only considers short window of context
  3. finbert: BERT language model finetuned in finance domain for financial sentiment classification (https://huggingface.co/ProsusAI/finbert)

personal note:

• paper trading doen't include commission and fees
• short position: selling assets you don't own with the expctation of buying them back at a lower price (taking loans)
• long position: buying assets with the expectation of selling them at a higher price later
• lumibot: easy algo trading framework