Observational Reasoning & Behavior-Integrated Trading
A machine learning system for predicting overnight price movements in equity markets using multi-modal data sources.
ORBIT is a quantitative trading research platform that combines multiple signals to predict overnight gap movements in US equity markets. The system focuses on predicting close-to-open returns for SPY/VOO (S&P 500 ETFs).
Key Features:
- Multi-modal prediction (multi-head)
- Walk-forward validation with strict temporal discipline
- Production-ready pipeline for daily scoring and execution
- Comprehensive backtesting and regime analysis tools
Current Phase: Production deployment (backtest_v1 model validated)
- ✅ Data ingestion pipeline
- ✅ Feature engineering and model training
- ✅ Backtesting framework with regime analysis
- ✅ Out-of-sample validation (Nov 2024 - Nov 2025)
- ✅ Paper trading infrastructure
- 🚧 Live trading deployment (in progress)
backtest_v1 Model (Nov 2024 - Nov 2025 out-of-sample):
- Return: +76.49% (long-only overnight strategy)
- Sharpe Ratio: 6.09
- Max Drawdown: -3.25%
- Win Rate: 99.4% (157/158 trades)
- Regime-tested: Robust across bull/bear/high-vol/low-vol conditions
This software is for research and educational purposes only.
- Not financial advice or investment recommendations
- Past performance does not guarantee future results
- Trading involves substantial risk of loss
- Only trade with capital you can afford to lose
- Consult a licensed financial advisor before trading
The authors make no warranties about the accuracy, completeness, or reliability of this software. Use at your own risk.
See LICENSE for details.
For questions or collaboration: issues
Note: This is a research project. The system is under active development and results should be validated in paper trading before any live deployment.