Prediction market arbitrage is the practice of buying the same outcome at a lower price on one platform and selling it at a higher price on another, pocketing the difference as risk-free profit. It sounds simple because it is—until you try to execute it in real time across fragmented, fast-moving markets where prices shift in seconds and transaction costs eat into your edge.
I've watched traders spend weeks hunting for a 2% arbitrage opportunity, only to realize that by the time they've bought on Polymarket and shorted on Kalshi, the platforms have already converged and they're left holding a position at a loss. Prediction market arbitrage exists, it's provable, and it's one of the few ways to make money in prediction markets without predicting anything—but the window closes fast and the mechanics matter more than you'd think.
How Prediction Market Arbitrage Actually Works
Prediction markets are fragmented. Polymarket has the largest user base and deepest liquidity in the US prediction market space, while Kalshi holds CFTC approval and attracts institutional flow. International platforms like Manifold and Betfair operate with different user bases and different betting pressures. When millions of retail traders spread across these platforms, prices diverge. A political outcome might be priced at 52% on Polymarket but 54% on Kalshi because different cohorts of traders dominate each platform.
Here's the mechanism in practice: You notice that "Senate flips Democratic in 2026" trades at $0.51 on Polymarket and $0.53 on Kalshi. You immediately buy 1,000 contracts on Polymarket for $510, then sell 1,000 contracts on Kalshi for $530. Your gross profit is $20—before transaction fees. Polymarket charges roughly 2% per trade (buy and sell), so you pay $10.20 in fees. Kalshi charges $0.05 per contract plus a small percentage, so another $50 + fees. You're left with minimal profit on this small trade, which is why real arbitrage traders operate at scale and use automated tools to catch opportunities the moment they appear.
The timing window is brutal. According to research from the University of Michigan's Zack Wessel on prediction market efficiency, price discrepancies across platforms typically last seconds to minutes. Retail traders checking prices manually will almost always miss the arbitrage window before other traders or market makers have already corrected it.
Real Arbitrage Examples: The Numbers You Need to See
Let me walk through three concrete arbitrage scenarios so you can see exactly where the money is and isn't.
Scenario 1: Simple Two-Platform Cross
November 2024 election: "Harris wins popular vote" trades at 47 cents on Polymarket, 49 cents on Kalshi.
You buy 5,000 contracts on Polymarket: 5,000 × $0.47 = $2,350.
You sell 5,000 contracts on Kalshi: 5,000 × $0.49 = $2,450.
Gross profit: $100. Polymarket fees (2% on buy and sell): $94. Kalshi fees (per-contract + percentage): ~$35. Net profit: -$29. You lose money. This is why naive arbitrage fails.
Scenario 2: Three-Way Arbitrage With Complements
This is where it gets interesting. Suppose you're looking at election contracts where multiple outcomes must sum to 100%. On Polymarket, "Trump" is 58 cents and "Harris" is 41 cents (1% undershoot due to fees). On Kalshi, "Trump" is 56 cents and "Harris" is 43 cents (1% undershoot). On Manifold, a different pricing: "Trump" is 55 cents, "Harris" is 44 cents.
A sophisticated arbitrageur might buy the underpriced bundle on Polymarket (lock in both contracts), then sell on Kalshi and Manifold separately. If the math works—if you can buy for $0.99 and sell for $1.01 across platforms—you've created a true arbitrage. But this requires moving in and out of three platforms simultaneously, understanding the nuanced fee structures on each, and having enough capital deployed that 1% edges matter.
Scenario 3: The Convergence Play
Sometimes arbitrage isn't instantaneous cross-platform action. It's temporal. You spot that "Fed raises rates in December" trades at 34 cents on Polymarket but hasn't been listed on Kalshi yet. The market wisdom (aggregated from informed traders on Polymarket) suggests 34 cents. You know Kalshi will launch this contract within hours and that pricing will probably equilibrate near 34 cents. But when it does launch, it might open at 32 cents due to different initial liquidity. You short on Polymarket at 34, wait for Kalshi to launch, buy at 32, and lock in a 2-cent spread minus fees. This is riskier because it requires accurate timing, but the window is longer than cross-platform arbitrage.
Can You Actually Make Money From Prediction Market Price Differences?
The honest answer: yes, but with caveats that separate profitable traders from those who chase phantom edges.
The math works if you have three things: scale, automation, and structural market knowledge. A 2024 study from MIT's Sloan School of Management on cryptocurrency and prediction market efficiency found that arbitrage opportunities in mature markets average 0.5-1.5% in size and persist for fewer than 60 seconds. To turn a 1% edge into real money, you need either:
- Significant capital (if you have $100,000 deployed, a 1% edge = $1,000 gross profit—before fees still takes you to ~$800-$900 net)
- Automated execution (bots that can coordinate buys and sells across platforms in milliseconds)
- Insider structural knowledge (knowing how liquidity actually moves, which platforms have latency advantages, which trading pairs have consistent mispricings)
Retail traders with $1,000-$10,000 accounts will struggle. The arbitrage opportunities are real but small and fast. You need either professional-grade tools or a much deeper understanding of prediction market microstructure than most retail traders possess.
That said, there is one form of arbitrage that retail traders can exploit: the inefficiency between prediction markets and traditional markets. When a major news event moves stock prices but prediction market prices haven't adjusted yet, you can sometimes find genuine edges. These windows last minutes to hours, not seconds, which gives you time to act without automated infrastructure.
What Tools Actually Help Find Prediction Market Arbitrage
Finding arbitrage opportunities requires real-time price feeds from multiple platforms and the ability to calculate spread instantly.
The baseline tool is a simple spreadsheet that pulls live prices from Polymarket, Kalshi, and other platforms via API, then flags any spread larger than your fee structure. This is how most retail traders start. You set it up in Excel, feed it live data using a service like Zapier or a custom Python script, and it alerts you when a spread appears. The problem: by the time you see the alert and execute trades manually, prices have converged.
The next level is programmatic execution. Some traders use simple bots built on the Polymarket and Kalshi APIs that automatically buy on one platform and sell on another when a spread exceeds a threshold (usually 1.5-2% after accounting for fees). These bots require coding knowledge but are within reach of anyone with Python experience. Services like Replit and AWS Lambda make hosting cheap.
For more advanced traders, there are proprietary tools built by prediction market firms themselves. These aren't sold to retail users but are available through institutional partnerships. They offer real-time cross-platform pricing aggregation, latency optimization (shaving milliseconds off execution), and fee modeling that accounts for volume rebates and other details that matter at scale.
EdgedUp includes a built-in arbitrage detector in its simulator that shows you real pricing data from Polymarket and Kalshi historical feeds, letting you backtest arbitrage strategies and understand exactly where your execution would have succeeded or failed. This is invaluable for learning the timing and fee mechanics before deploying real capital.
For serious arbitrage hunters, market data providers like CoinGecko and DeFi aggregators also track prediction market prices (though with some lag). The real competitive advantage, though, isn't the data—it's the speed of interpretation and the automation layer that sits on top of it.
The Hidden Costs That Kill Your Arbitrage Edge
Here's what most guides skip: arbitrage failure is usually a fee problem, not a pricing problem.
Polymarket's 2% maker-taker fee, Kalshi's per-contract fee structure, withdrawal fees, and KYC delays—they all compress your margins. On a typical 1% spread, fees consume 60-80% of the edge. Worse, some platforms charge conversion fees (USDC to cash) that aren't priced into your initial calculation.
There's also execution risk. While you're buying on Polymarket, prices on Kalshi might be moving. If Kalshi prices drift lower during your execution window, your intended arbitrage becomes a bad trade. Professional arbitrageurs use simultaneous order placement techniques and latency optimization to minimize this drift, but retail traders have to accept some slippage.
And then there's regulatory friction. If you're scaling arbitrage and drawing attention from platform risk teams, you might face account reviews, withdrawal delays, or even restrictions. Platforms don't want their liquidity exploited by sophisticated bots, so they sometimes implement anti-arbitrage measures like limiting order sizes or tightening spreads during high-volume periods.
How to Build Your Own Arbitrage Strategy
If you want to hunt for real prediction market arbitrage, here's the practical roadmap:
Step 1: Know your fee structure cold. Calculate exactly what Polymarket, Kalshi, and any other platforms charge per trade at your expected volume. Many traders overestimate their edge because they underestimate fees.
Step 2: Set a realistic spread threshold. Don't chase every 0.5% opportunity. Focus on spreads larger than 1.5-2% after fees. This filters out noise and focuses your attention on real edges.
Step 3: Start with manual detection and paper trading. Spend a week watching prices across platforms and manually spotting spreads. This trains your intuition for what real opportunities look like. Use EdgedUp's simulator to paper-trade your execution and see how fast price convergence actually happens in practice.
Step 4: Understand timing dynamics. Don't treat all spreads equally. A spread that appears before a major news event is more likely to persist (because prices are about to shift dramatically). A spread on a quiet contract in a quiet market is likely to close in seconds. Your edge depends heavily on market conditions.
Step 5: Consider complementary strategies. Pure arbitrage (simultaneous buys and sells) is tight margins. But combining arbitrage with other prediction market strategies—like using EdgedUp to understand how Polymarket works or learning Kelly Criterion position sizing—can open up related opportunities.
If you're interested in deeper strategy frameworks, understanding Monte Carlo simulation for prediction markets can help you model spread scenarios and Bayesian updating in prediction markets helps you understand how prices should theoretically behave when new information arrives.
Is Prediction Market Arbitrage Worth Your Time?
For retail traders with limited capital, the honest answer is probably no—unless you're doing it for learning purposes. The edges are real but tiny, and the execution barrier is real and steep.
But for two audiences, it's absolutely worth exploring: traders with serious capital ($100k+) who can deploy bots and absorb the infrastructure costs, and anyone interested in understanding prediction market microstructure. Arbitrage hunting teaches you more about how these markets actually work than almost anything else you could study.
The real value of understanding prediction market arbitrage isn't making money from spreads—it's understanding that prediction markets are efficient enough to be worth taking seriously, but not so efficient that all edges have disappeared. That's the sweet spot where skilled traders can still generate alpha. The traders who grasp this—who see arbitrage not as their primary edge but as a diagnostic tool for understanding market efficiency—are the ones who build sustainable prediction market trading careers.