Definition
Prediction market arbitrage exploits price discrepancies for the same binary event listed on two different platforms. Binary contracts resolve to exactly $1.00 (YES wins) or $0.00 (NO wins). When YES on Platform A and NO on Platform B for the same event cost a combined total below $1.00, buying both guarantees a profit at resolution regardless of the outcome.
How it works
Example: "Will the Fed cut rates in June?" is listed on both Polymarket and Kalshi.
- Polymarket: YES at $0.47
- Kalshi: NO at $0.51
- Combined cost: $0.98
- Guaranteed payout: $1.00
- Gross edge: +2.04%
After platform fees (~1% per side), the net edge might be ~0.04% — or with a better spread, meaningfully positive. The key is finding spreads wide enough to cover fees.
Why the opportunity exists
Polymarket and Kalshi have different user bases, different liquidity pools, and different fee structures. A large trade on one doesn't move the other. News breaks and prices update at different speeds on each platform. These structural differences create persistent price divergences that arbitrageurs can exploit.
The matching problem
The hardest part of prediction market arbitrage isn't execution — it's correctly identifying that the same event is listed on both platforms. "Fed cuts rates at May FOMC" and "Federal Reserve rate cut — May meeting" are the same event but share no keywords. AI semantic matching solves this at scale.
Automation advantage
Arbitrage windows typically last 30–200 seconds before market makers close the gap. Manual execution catches very few opportunities. Automated agents monitoring all markets in real time via WebSocket and executing both legs in parallel under 400ms capture far more.