How predict.pm works
predict.pm works in three core steps, repeated every 3 minutes: it fetches events and markets from each platform's API, an LLM groups the same real-world event across platforms, and the model aligns each outcome so the prices line up in one side-by-side comparison.
- Refreshes every 3 min
- AI-matched
- Free · Beta
- 1
Fetch every market
predict.pm continuously ingests events and markets from each platform's API — Polymarket, Kalshi and Futuur.
- 2
Match events with AI
A large language model reads the differently-worded titles and groups the same real-world event into one cross-platform event.
- 3
Match outcomes
Within each event, the model aligns every market's outcomes so equivalent results line up across platforms.
- 4
Compare side by side
The aligned probabilities are shown in one grid, so you can read every platform's price at a glance and spot the best one.
How fresh is the data?
predict.pm fetches new markets and refreshes prices every 3 minutes. Live prices are read from each platform's native API when you open an event, and probabilities are normalized to a 0–100 scale.
Does predict.pm hold funds?
No. predict.pm is an independent comparison tool — it shows odds and links out to each platform, but never accepts bets or custodies funds. It is currently in beta.
Compare the odds yourself
See the same event priced across Polymarket, Kalshi and Futuur.
Learn more
Frequently asked questions
How does predict.pm work?
predict.pm works in three steps, repeated every 3 minutes: it fetches events and markets from each platform's API, an LLM groups the same real-world event across platforms, and the model aligns each outcome so the prices line up in one side-by-side comparison.
Where does predict.pm get its data?
From the native APIs of each supported platform — Polymarket, Kalshi and Futuur. Live prices are read from each platform at view time.
How often does it update?
predict.pm fetches new markets and refreshes prices every 3 minutes.
Does predict.pm use AI?
Yes. A large language model performs the event and outcome matching across platforms, even when the wording is completely different.