Prediction market arbitrage is the practice of exploiting price discrepancies for the same event across different platforms. Unlike directional trading, pure arbitrage requires no view on the outcome — you simply capture the spread between two mispriced markets.
What is prediction market arbitrage?
Binary prediction markets like Polymarket and Kalshi trade contracts that resolve to $1.00 (YES) or $0.00 (NO). When the same event is listed on both platforms, their prices should theoretically sum to $1.00. When they don't, an arbitrage window opens.
Example: "Will candidate X win?" priced at YES 52¢ on Polymarket and NO 46¢ on Kalshi. Combined cost: 98¢. Guaranteed return: $1.00. Edge: +2.04%.
Single-platform vs cross-platform arbitrage
Single-platform: Exploiting YES/NO mispricing within the same market on one platform. Rare, since market makers correct these quickly.
Cross-platform: The same event priced differently on Polymarket vs Kalshi. More common because the platforms have different liquidity pools, different user bases, and different fee structures causing persistent divergence.
The math behind the edge
Let P = price of YES on Platform A, Q = price of NO on Platform B. An arbitrage exists when:
P + Q + fees < 1.00
The gross edge is (1.00 - P - Q). After accounting for platform fees (typically 0.5–2% per leg), the net edge must still be positive to be worth executing.
Typical gross edges on Polymarket/Kalshi arbitrage: 1.5% to 6%, depending on market activity and how long the window has been open.
Sizing with Kelly criterion
For pure arbitrage the Kelly formula simplifies significantly — since the probability of winning is 100% (assuming both legs fill), the optimal bet size is limited only by:
- Available liquidity at the quoted price
- Platform position limits
- Leg-fill risk (the risk that the second leg doesn't fill before price moves)
In practice, bet sizing is constrained by market depth. Arbitrage Agent uses Kelly sizing with a cap at the available depth at the quoted price.
Risks and how to manage them
Leg risk: The biggest operational risk. If Leg 1 fills (e.g., YES at 52¢) but Leg 2 fails to fill before the price moves, you're holding a directional position. Mitigation: execute both legs near-simultaneously; implement a kill switch that triggers if Leg 2 doesn't fill within a timeout.
Liquidity risk: Thin markets may not have enough depth to fill the desired size at the quoted price. Always use limit orders and check order book depth before committing.
Fee drag: Fees erode the gross edge. On a 2% gross edge, a 0.5% fee per side (1% total) leaves only 1% net. Model fees explicitly; don't trade if net edge is below ~1%.
Execution latency: Arbitrage windows on active markets close in seconds. Manually spotting opportunities misses most of them. Real-time detection with instant alerts captures far more — and automated execution (in final testing) will close the gap entirely.
Tools and automation
Monitoring thousands of markets manually is not feasible. Arbitrage Agent handles event matching (the hard part — the same event has different titles on each platform), price comparison, edge calculation, and dual-leg execution automatically.
Getting started
The best approach: start in dry-run mode. Let the agent scan for opportunities for a week and review the signals. Verify that the event matching is accurate, the edge calculations are correct, and the fills are realistic. Then move to live trading with a small bankroll and scale from there. Join the waitlist for early access.