Prediction market arbitrage sounds straightforward: find the same event priced differently on Polymarket and Kalshi, buy both sides, collect the guaranteed spread. In practice, doing this manually is nearly impossible at any meaningful scale. Here's why automation is essential — and what a good Polymarket arbitrage agent actually does.
Manual arbitrage: the limitations
Manually monitoring arbitrage opportunities across thousands of prediction market contracts requires:
- Checking both Polymarket and Kalshi simultaneously for price divergences
- Matching events correctly across platforms (same event, different titles)
- Calculating edges after fees fast enough to act before the window closes
- Executing both legs near-simultaneously to avoid leg risk
Arbitrage windows on active markets last 30–200 seconds. A human checking manually every few minutes will miss most opportunities. Even monitoring continuously is ineffective — you can't watch thousands of markets at once.
What an arbitrage agent does
A proper Polymarket/Kalshi arbitrage agent does the following in a continuous loop:
- Event matching: Maps the same real-world event across both platforms despite different titles, contract structures, and resolution criteria.
- Price monitoring: Subscribes to real-time WebSocket price feeds on both platforms to detect divergences the moment they appear.
- Edge calculation: Computes the net edge after both platforms' fees to verify the opportunity is actually profitable.
- Dual-leg execution: Fires both orders near-simultaneously to minimise leg risk.
- Position management: Tracks open positions, monitors for fill confirmation, and triggers kill-switch logic if a leg fails to fill.
Arbitrage Agent: how it works
Arbitrage Agent uses AI-powered event matching to identify the same event across Polymarket and Kalshi — the hardest part of cross-platform arbitrage. It monitors 2,400+ active markets in real time, calculates net edges explicitly (including fees), and executes both legs in under 200ms when an opportunity meets the threshold.
Key features: dry-run mode for safe testing, Kelly criterion sizing, configurable edge threshold, Telegram alerts, and a real-time dashboard showing all open positions and resolved P&L.
What to look for in an arbitrage agent
- Event matching accuracy: False matches (different events mistakenly identified as the same) will cause losses. The matching algorithm is the most important component.
- Execution speed: Sub-200ms execution is the target for capturing most opportunities.
- Fee modelling: The agent must calculate net edge after fees, not gross edge.
- Leg risk management: Kill switch and retry logic for failed second-leg fills.
- Dry-run mode: Essential for verifying the agent's behaviour before committing real capital.
Join the Arbitrage Agent waitlist to get early access to the only agent built specifically for Polymarket/Kalshi cross-platform arbitrage.