RLXBT is local research infrastructure for AI feature discovery, event-driven backtesting, robustness analysis, strategy evaluation, and trading reinforcement learning.
Website · Research Atlas · Start free · Headless server
A backtest can be wrong while still producing a plausible number. RLXBT is built around making the assumptions visible and testing the result beyond a single equity curve: execution timing, costs, risk controls, out-of-sample evidence, sensitivity, and robustness.
The public Quant Research Log documents that process with bundled data and reproduction commands. It includes negative results and defects found in RLXBT itself, because evidence is more useful than a polished claim.
- Explore the Research Atlas to see the questions, methods, and evidence.
- Reproduce a study from the Quant Research Log.
- Start free on macOS or deploy the headless server.
- Use GitHub Discussions for integration questions and research collaboration.
- macOS application for local research;
- headless Ubuntu/Docker server at
ghcr.io/sergio12s/rlxbt-server; - HTTP API for data loading, backtests, and research workflows;
- public studies with exact parameters and reproduction scripts.
This is the public product, documentation, and discussion entry point for RLXBT. It intentionally does not contain the private source repository or private trading research. Public evidence and reproducible studies live in quant-research-log; product downloads and current documentation live on rlxbt.com.
- Questions and collaboration: Discussions
- Reproducible product problems: Issues
- Security reports: see SECURITY.md