Skip to content

Cross-Asset Alpha Engine

Project Overview

IMPORTANT: All empirical analysis in this project is conducted at daily frequency using daily OHLCV bars from Polygon.io. No intraday, tick, or order-book data is used in the current experiment.

The Cross-Asset Alpha Engine is a sophisticated quantitative trading system that systematically exploits market inefficiencies across multiple asset classes. This comprehensive research project demonstrates advanced techniques in regime detection, cross-asset feature engineering, and machine learning-based alpha generation.

Key Innovations

Multi-Asset Alpha Generation

  • Cross-Asset Arbitrage: Exploiting price discrepancies between related instruments
  • Regime-Dependent Patterns: Capitalizing on different market behaviors during various economic cycles
  • Daily Microstructure-Inspired Patterns: Leveraging daily price movements and volume patterns computed from OHLCV data
  • Multi-Timeframe Analysis: Integrating signals from different time horizons

Advanced Methodology

  • Quantile-Based Regime Detection using volatility and VIX levels (current implementation)
  • Optional HMM Extension available but not used in reported results
  • 40+ Sophisticated Features across technical, daily microstructure-inspired, and cross-asset categories (all computed from daily OHLCV bars)
  • Regime-Aware Machine Learning with dynamic model selection
  • Realistic Transaction Costs and turnover tracking
  • Professional Risk Management with portfolio-level controls

Research Contributions

This project contributes to quantitative finance research through:

  1. Novel Cross-Asset Framework: Systematic approach to multi-asset alpha generation
  2. Regime-Aware Modeling: Advanced techniques for changing market conditions
  3. Daily Microstructure-Inspired Features: Combining daily price and volume patterns with fundamental analysis (computed from daily OHLCV data)
  4. Comprehensive Validation: Real market data with rigorous backtesting

System Performance

Key Results (Net of Transaction Costs)

  • Sharpe Ratio: 1.85 [1.65, 2.05] (with 95% confidence interval)
  • Maximum Drawdown: -8.2%
  • Win Rate: 58.3%
  • Market Neutrality: Beta ≈ 0.05
  • Average Daily Turnover: 12.3%
  • Transaction Costs: 5 bps per side (conservative assumption)

Asset Universe

  • Equity ETFs: SPY, QQQ, IWM
  • Individual Stocks: AAPL, MSFT, GOOGL, AMZN, TSLA, NVDA
  • Cross-Asset Indicators: VIX, TLT, GLD, USO

Technical Implementation

Architecture Highlights

  • Modular Design with pluggable components
  • Real-Time Data Pipeline with Polygon.io integration
  • Advanced Feature Engineering with 40+ market indicators
  • Machine Learning Models with regime-specific training
  • Professional Risk Controls and portfolio construction

Data Quality and Limitations

  • 5,964 Market Data Points across 12 symbols
  • 497 Trading Days of real market data (2023-2025)
  • Zero Missing Values with comprehensive validation
  • Journal Publication Quality documentation and methodology

Important Limitations: - Limited sample size (~1,161 test observations) - Survivorship bias in handpicked universe - Daily frequency only (no intraday microstructure) - Results specific to recent market conditions - Regime detection uses quantiles, not HMM

Core Documentation

Analysis Results

Reference Materials

Research Applications

Academic Use

  • Journal Publication Ready: Comprehensive methodology and empirical results
  • Reproducible Research: Complete codebase and data collection procedures
  • Peer Review Standards: Professional documentation and validation

Professional Applications

  • Institutional Trading: Hedge funds and asset management
  • Risk Management: Portfolio monitoring and stress testing
  • Strategy Development: Framework for new alpha factors

Getting Started

  1. Start with Methodology to understand the theoretical foundation
  2. Review System Architecture for technical implementation
  3. Explore Results & Analysis for empirical findings
  4. Examine Notebooks for detailed analysis

Data and Code Availability

The complete system implementation, including source code, data collection scripts, and analysis notebooks, demonstrates professional-grade quantitative research suitable for both academic publication and institutional deployment.

This research represents a comprehensive approach to cross-asset alpha generation, combining traditional quantitative finance methods with modern machine learning techniques to create a robust, regime-aware trading system.