Reproduce “Directly Learning Stock Trading Strategies Through Profit Guided Loss Functions” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Devroop Kar, Zimeng Lyu, Sheeraja Rajakrishnan, Hao Zhang, Alex Ororbia, Travis Desell, Daniel Krutz
The authors adapt time-series neural networks to learn portfolio positions directly rather than forecast prices for a trading decision. Their final layer outputs bounded allocations and optional cash, trained with profit-guided losses using subsequent returns. On data for 50 S&P 500 companies, Crossformer with StockLoss-L2 returned 51.04% in 2022 versus 2.81% for PPO; results varied and are not live-trading evidence.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.