return-bench-mars

What problem does it solve?

Benchmarking MarS return prediction pipelines to quantify their ability to predict price direction using order_index derived slots; it provides metrics to evaluate predictive performance and robustness.

Core Features & Use Cases

• Measures IC, Ranked IC, and Direction Accuracy for MarS-generated order sequences against real price movement.
• Applies to evaluating model-driven trading signals across multiple horizons and tickers.
• Use Case: You want to validate whether your MarS return benchmark pipeline reliably forecasts price direction on GOOG or other assets.

Quick Start

Run the return_bench.py script with your generated and real indices to compute IC, rank IC, and direction accuracy for the specified horizons.