Sole author · 2026
RCBot
A small-cap momentum trading bot that runs the same code against historical replay and Interactive Brokers.
- Python
- ib_async
- Parquet
- pytest
Problem
I wanted replay and live trading to feed the bot exactly the same events, so that a test on recorded data shows how the bot would behave live.
What I built
- One event format (book, tape, status) for both Parquet replay and IB Gateway, with incremental features, micro-pullback and new-high detectors, a risk governor, and position management.
- A conservative simulated broker that models sweeps, queue position, and latency, plus an Interactive Brokers adapter for paper trading. Paper, shadow, and detect-only modes share the same decision code.
- Tools for threshold sweeps, labeled-candidate analysis, and execution stress tests, plus a run ledger that saves the full configuration and results of every run.
Results
- event format for replay and live
- 1
- paper trading
- IBKR
Limits
The replay data is for measuring how often setups trigger and for finding bugs. It does not show that the strategy is profitable. Choosing which stocks to trade is still done by hand.