If you've traded on both Polymarket and Kalshi, you've probably noticed the same event listed on both platforms at noticeably different prices. This isn't a glitch — it's a structural feature of how prediction markets work, and it creates a persistent arbitrage opportunity.
How Polymarket works
Polymarket is a decentralised prediction market built on the Polygon blockchain. It uses an automated market maker (AMM) model with a CLOB (central limit order book) layer. Liquidity is provided by market makers who set YES/NO prices. Trading is open globally (with restrictions in some jurisdictions including the US). Settlement is handled by UMA's optimistic oracle.
How Kalshi works
Kalshi is a CFTC-regulated event contracts exchange based in the US. It operates as a traditional order book exchange. It's legally available to US residents — a key difference from Polymarket. Settlement is handled by Kalshi's own resolution process, aligned with regulatory standards.
Why prices differ
Several structural factors cause persistent price divergence:
- Different user bases: Polymarket attracts a more international, crypto-native audience. Kalshi has a larger proportion of US-based traders. These groups often have different information and different biases.
- Different liquidity: Each platform has its own independent liquidity pool. A large trade on one doesn't move the other.
- Different fee structures: Fee differences shift the effective price for traders on each platform.
- Different resolution sources: Subtle differences in contract wording and resolution criteria can cause legitimate price divergence on close calls.
- Latency: News breaks and prices update — but not always at the same speed on both platforms.
How to profit from the difference
When the YES price on Polymarket plus the NO price on Kalshi (for the same event) adds up to less than $1.00, you have a risk-free arbitrage. Buy both sides simultaneously and you're guaranteed $1.00 on resolution regardless of outcome.
The challenge is execution speed. These windows typically last 30–200 seconds before arbitrageurs close the gap. Manual trading catches very few of them. An automated agent monitoring all events in real time catches far more.
The event matching problem
The hardest part of cross-platform arbitrage isn't the execution — it's identifying that the same event is listed on both platforms. "Will the Fed cut rates in June 2026?" might be titled differently, cover slightly different date ranges, or have subtly different resolution criteria on each platform. Arbitrage Agent uses AI-based event matching to solve this automatically. Join the waitlist to get access.