Kalshi Prediction Markets

Prediction-market pricing moves faster than a person can watch, and the interesting moments are the ones nobody was looking at.

Standard market interfaces show a snapshot, not a history — there was no way to go back and see how a book built up before it moved, or which markets had gone unusually quiet or unusually loud.

Orderbook monitors on Kalshi and a Polymarket gamma tracker write every observation to an archive, with liquidity and weather-market conditions recorded alongside price.

Large-order detection flags block-sized bids and asks as they land, activity checkers rank markets by how much is happening in them, and a dedicated weather-markets analyzer accounts for the different liquidity profile those contracts carry versus politics or sports markets. Every collector runs on its own polling interval behind rate-limit throttling, so watching more markets doesn't mean tripping the exchange's limits.

A continuous record instead of screenshots: what the book looked like, when it moved, and how thin it was when it did.

That archive is what makes the rest useful — big-block alerts fire in real time, weather-market liquidity gets tracked on its own timeline, and every market has a history to pull up rather than reconstruct from memory. The pipeline runs across two exchanges at 99.9% uptime, generating 500K+ data points a day.

Python with asyncio and websockets carries the real-time collection; Postgres holds the orderbook and market archives; and the Kalshi API and Polymarket Gamma API are the two data sources the whole suite is built around.