RLER

Rust · reinforcement learning

A Rust reinforcement-learning paper-trading workbench for NSE equities on 15-minute candles.

View source on GitHub

Stack: Rust, reinforcement learning, Q-learning, NSE, paper trading, walk-forward, Yew, MIT

What it is

RLER is a workbench for teaching an agent to trade on 15-minute candles — and, more honestly, for teaching me how reinforcement learning behaves on noisy financial data. It is paper-only. It does not place broker orders and it does not give investment advice.

The current Rust implementation does tabular Q-learning, CSV and Yahoo candle loading, backtests, paper replay, walk-forward validation with Monte Carlo robustness, SQLite run history, and a Rust/Yew dashboard. It ships with sample data so it works fully offline. DQN is intentionally deferred — getting the tabular case honest comes first.

What each command does

• train — fit one tabular Q-learning model
• backtest — train on earlier rows, evaluate on later rows
• paper — replay paper trading with a saved model
• validate — rolling walk-forward validation, the test that actually matters

At a glance

Project boundaries, stated up front: paper trading only, no real broker execution, no profit guarantee, deep RL is a future phase. Honesty about scope is part of the design.

All projects