How to work with prediction market data: testing a strategy on Kalshi's history, reading Kalshi and Polymarket files side by side, the terms you will meet, and where else to get the data.
How to backtest Kalshi markets
A Kalshi backtest in DuckDB, step by step: outcomes, entry prices you could have traded, fees, gaps, and the mistakes that make backtests look better than trading.
Read the guideKalshi vs Polymarket data: what differs
Identifiers, order book shape, update rules, units, fees and resolution on the two venues, and what each difference means for code that reads both.
Read the guidePrediction market data glossary
Order book, market structure, file and backtesting terms, each defined in a sentence or two.
Read the guidePrediction market data providers compared
Kalshi Eye, Probalytics, TickFoundry, Oddpool, Tardis.dev, Dune, the pmxt archive and the venues' own APIs: coverage, depth, formats and prices, as each states them.
Read the guideReference
- Data dictionary: every dataset's columns, with free sample files.
- Data by venue: what the archive holds for Kalshi, Polymarket, Limitless and Chainlink, and since when.
- Bulk downloads: getting the daily Parquet files with the Python client, a script, or S3 tools.
- Free one-day sample on Hugging Face: October 13, 2026 for seven Kalshi series, every dataset, no account needed.
- Examples on GitHub: runnable scripts that read that sample in place.
- MCP server: search the archive from Claude, Cursor and other AI assistants.