Guides
Short, code-first guides for working with per-day Parquet market data. Each one runs against the real files.
Understanding the data
- OHLCV data: how each bar is built: auction prints in the daily bar, bar-open timestamps, missing minutes, resampling and futures settlement, measured on real files.
Backtesting engines
- Backtrader with Parquet market data: one-symbol extraction, daily and 5-minute PandasData feeds, session filtering, a moving-average example.
- vectorbt with Parquet market data: build the symbols-by-dates close matrix and run a portfolio, up to the whole market.
Loading and querying
- DuckDB stock data Parquet guide: schema per timeframe, multi-file globs, VWAP and resampling in SQL.
- Polars market data Parquet guide: lazy scans, symbol filters, and resampling with Polars.
- Build a local market data lake: the by_date layout, a downloader, and querying across years.
- Stock market data Parquet datasets: which sources ship Parquet, the free datasets that exist, and seven checks before trusting one.
Getting the data
- Free historical stock data: what the free year includes and how to pull it in a minute.
- Download all historical US stock data: sizes per timeframe and the one-command loop over the whole archive.
- Free NQ and MNQ futures historical data: what each free source gives up, and the roll checks to run first.
Choosing a source
- yfinance limitations for backtesting: where the free wrapper breaks down once a notebook becomes a backtest.
- Historical intraday data providers compared: 1-minute history depth per plan, sessions, venues and delivery for 15 sources.
- Polygon.io is now Massive: new hosts, packages and environment variables, and what still works.
- Compare providers: MarketParquet against sixteen market-data vendors, one page each.