Choosing Historical Data Sources for LumiBot Backtests
Summary
This documentation overview maps backtesting data sources to asset classes, bar intervals, and setup requirements in LumiBot. It suggests Yahoo for uncomplicated daily stock and ETF tests, ThetaData for intraday stock or options history, Polygon for several asset classes, and Databento for futures and market-data schemas. It also lists Pandas input for locally held data, Interactive Brokers history, Alpaca options data, and Polymarket price history. The practical lesson is to choose a source based on the instrument, granularity, and access credentials or files available.
The page points to supporting guides on running backtests and inspecting performance, trades, indicators, logs, and tearsheets. It also describes an agent-driven workflow in which strategies can use AI agents and point-in-time data, with cached replay of identical agent runs. The document is an orientation page rather than a strategy evaluation: it supplies no comparative accuracy, costs, or performance evidence. Provider availability, access requirements, and historical coverage vary, so researchers need to check the detailed source-specific guidance before relying on a dataset.
Key ideas
- Select a historical data source according to asset class, required bar interval, and access setup.
- Yahoo is presented as a simple option for daily stock and ETF data, while intraday work may require a specialized provider.
- LumiBot can accept locally prepared data through Pandas as well as data from multiple external providers.
- The documentation lists generated artifacts that help inspect backtest behavior and results.
- Agent-driven strategies can use historical point-in-time data and replay cached runs.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.