Deep Learning for Statistical Arbitrage
- Paper Link
- Replicated and applied a statistical arbitrage framework to cryptocurrency markets, leveraging residual portfolios from conditional latent asset pricing factors to construct arbitrage portfolios.
- Implemented optimal trading policies based on these signals, achieving steadily high out-of-sample mean returns of 8.2% and Sharpe ratios exceeding 1.5, and lowered the transaction costs by 17%.