This project integrates advanced techniques in U.S. equities trading, combining pairs trading, Hidden Markov Models (HMM) for market regime classification, sentiment analysis, and robust risk management.
- Pairs Trading Strategy: Stock pair selection through unsupervised machine learning.
- Regime Classification with HMM: Identifies market regimes and allocates assets.
- Sentiment Analysis: NLP techniques to analyze news streams.
- Risk Management: Includes Triple Barrier method and Equal Contribution to Risk approach.
- In Sample (Jan 1, 2017, to June 1, 2023): 17.28% profit, Sharpe Ratio of 1.248, max drawdown 6.4%.
- Out of Sample (Jan 1, 2022, to Jan 1, 2023): 31.99% profit, Sharpe Ratio of 1.476, max drawdown 9%.
- Stress Test - March 2020: 1.64% profit, Sharpe Ratio of .937, max drawdown 4.5%.
- Blind OOS (Jan 1, 2023, to Apr 1, 2023): 11.47% profit, Sharpe Ratio of 1.529, max drawdown 8.2%.
- Live Paper Trading Result: 0.05% profit.
For detailed results please check the Algo_585_Group_Project pdf report.
Python and QuantConnect Platform - see the QuantConnect for details.
This project was developed as part of an academic final project for the Algorithmic Trading Course. Contributions were made solely by Dev Patel, Andres Caicedo, Cindy Chiu, and Nathan Luksik.
This project is licensed under the MIT License - see the MIT License for details.
Gratitude to David Ye and the course staff for their invaluable guidance.