RL Trading Web Tool

Purpose and scope are to enable reproducible evaluation of reinforcement learning trading policies within a controlled backtesting environment. It accepts market data, a policy interface, and backtest settings to produce performance and risk metrics for comparisons. The tool is intended for quantitative researchers, portfolio managers, and algo developers who require auditable experiment logs and standardized outputs.
The core logic simulates price evolution, applies agent actions, and accounts for costs and execution constraints. Outputs include per-asset exposure, equity curves, and summary statistics suitable for research notebooks and ML pipelines.
Users can perform parameter sweeps, compare policy variants, and assess robustness across seeds and market regimes. The framework emphasizes deterministic reproducibility when seeds and data are fixed, while allowing stochastic evaluation for robustness analyses.
Unique value lies in an RL-centric evaluation pipeline with structured, exportable results and explicit reward shaping controls that align with common research workflows. It supports multiple backtest configurations, data quality checks, and standardized reporting for publication-ready results.