Back

TradFi student thinking about doing some small-scale market making on prediction markets for fun/learning.

Inventory risk and hedging seem manageable, but adverse selection from informed traders feels like the biggest blindside risk. I'm wondering if it makes sense to model informed order arrivals as a Poisson process (or similar) and adjust spreads/size accordingly.

Has anyone looked at this problem before? How would you think about modelling or detecting informed flow in prediction markets?
Jun 16, 2026 · 05:05 PM · 52 views
1 Reposts
1 Likes
2 Replies
@stylish · Jun 17, 2026 · 08:27 AM
Jump owns 30% of Polymarket and they will always have latency advantage over you for market making. Polymarket give maker rewards when they want liquidity for markets that are way too risky for institutionals to provide liquidity in. Together with natural swings in public sentiment and probability of an event, this makes it for a very tough environment to make a market in.

That being said I think toxic flow detection is too hyper specific. I think you should look at market swings as a whole and toxic flow may make part of that. Because both blow out your position and one is a superset of the other
Reply