Ask HN: Anyone trading prediction markets programmatically?
Been building automated systems to compare prices between Kalshi and Polymarket on the same events. The spread between venues on sports contracts is surprisingly wide - sometimes 5-8 cents on identical outcomes. Curious if others here are doing similar cross-venue work and what your stack looks like. Im running Python/FastAPI with direct websocket integrations to both platforms. I do. Doing the arbitrage game is tough. I tried (with mixed results) to play the 5m and 15m crypto markets for arbitrage but it's way too competitive and HFT for me. Right now I'm finally done with my blockchain indexer and I'm parsing the data. Already have a live order book snapshotting stored as well but it's a massive amount of data (up to 2gb per 15m market). I still to work on compressing the data there. Once I have the data collected I will finally get to try to mine for some strategies. Did not try kalshi too much yet. Coming from cryto algo trading I've learned there are more attainable strategies that work compared to arbs. Arbs are tough and hyper-competitive. Ping me if you wannt to connect and share notes I’ve given it a go, the main issue I’ve found is historical data is near impossible to get prior to even this year. I’d happily pay for it but it doesn’t even go far back. So you have to run live scrapers which will take some time to build a good historical dataset. I know someone who is trading daily weather temperature predictions on prediction markets and using the best weather predictions he can find. It thin markets but I think there is some edge there.