Prediction Market Arbitrage: Quick Reference for Power Users
Prediction market arbitrage is the practice of exploiting price discrepancies for the same outcome across two or more prediction market platforms to lock in a risk-free (or near risk-free) profit. Done right, a power user can systematically capture 2–8% edges on individual trades before fees, with some cross-platform gaps reaching 15%+ during breaking news events. This guide is your go-to cheat sheet for identifying, executing, and automating arbitrage plays on platforms like Polymarket, Kalshi, Manifold, and beyond.
What Is Prediction Market Arbitrage and Why Does It Matter?
At its core, arbitrage in prediction markets means buying "Yes" on one platform at a low price and "Yes" (or "No") on another platform where the market has mispriced the same event. Because prediction markets trade in probabilities expressed as prices (e.g., 62¢ = 62% chance), any time the same event is priced differently across venues, a gap exists to exploit.
Why do these gaps appear?
- Liquidity fragmentation: Different user bases on each platform create localized price discovery
- Information latency: News hits one platform's order book before another's
- Structural differences: Kalshi is CFTC-regulated with different counterparty pools than Polymarket's crypto-native user base
- Settlement mechanics: Different resolution criteria for "the same" event can create permanently justifiable price gaps
For power users who treat prediction markets as a serious edge — not just entertainment — arbitrage is often the most reliable path to consistent returns. PredictEngine is built specifically for traders who want to capture these opportunities systematically.
The Four Core Arbitrage Structures
Not all arbitrage plays are created equal. Here's a breakdown of the four main structures you'll encounter:
1. Cross-Platform Directional Arbitrage
Buy the same outcome on Platform A (cheap) and sell or hedge on Platform B (expensive). If both contracts resolve identically, you collect the spread.
Example: "Will the Fed cut rates in September?" trades at 58¢ on Polymarket but 64¢ on Kalshi. Buy on Polymarket, sell on Kalshi. Net spread = 6¢ per dollar of exposure.
2. Complementary Arbitrage (Yes + No Hedge)
Buy "Yes" on one platform and "No" on another. If the combined cost is under $1.00, you guarantee profit regardless of outcome.
Formula: Profit = $1.00 − (Price_Yes + Price_No) — only valid when this is positive.
3. Correlated Market Arbitrage
Two markets aren't identical but are logically linked. For instance, a "Democrat wins presidency" market and a "Democrat wins Pennsylvania + Michigan + Wisconsin" market should have correlated prices. Divergences signal an edge.
4. Temporal Arbitrage
The same market on the same platform but with different expiry dates. A "Yes by year-end" contract trades cheap relative to a "Yes by Q3" contract, creating calendar spread opportunities.
Platform Comparison: Key Arbitrage Metrics
Before you execute any trade, understand the structural differences between platforms. This table covers the metrics that matter most for arbitrage:
| Platform | Fees | Settlement Speed | Liquidity Depth | Crypto/Fiat | Key Edge |
|---|---|---|---|---|---|
| Polymarket | 2% taker | 1–3 days post-resolution | High (top markets) | USDC (crypto) | Fastest news pricing |
| Kalshi | 3–7% maker/taker | 1–5 business days | Medium | USD (fiat) | Regulated, institutional flow |
| Manifold | None | Variable | Low | Play money / Mana | Sentiment leading indicator |
| Metaculus | None | Days–weeks | Very low | None (reputation) | Aggregated forecaster signal |
| PredictIt | 10% profits, 5% withdrawal | Days | Low–Medium | USD | Political niche |
Key insight: The Polymarket–Kalshi pair offers the best liquidity-adjusted arbitrage opportunities because both platforms price real-money outcomes on identical political and economic events, yet serve structurally different user bases. For a deep dive on how these platforms diverge in practice, check out common mistakes traders make when switching between Polymarket and Kalshi.
Step-by-Step: How to Execute a Cross-Platform Arb Trade
Here's the standard execution workflow for a directional cross-platform arb:
Scan for price discrepancies — Use a price aggregator or build your own feed pulling Polymarket's CLOB API and Kalshi's REST API simultaneously. Flag any identical event where the price gap exceeds your fee threshold (typically 5%+ to cover both-side fees).
Verify contract equivalence — Read both resolution criteria carefully. "Fed cuts rates in September" might resolve on different dates or use different source data. Mismatched resolution criteria turns an arb into a directional bet.
Calculate net expected value — Subtract both platforms' fees from the gross spread. Account for slippage on larger positions.
Net EV = Spread − Fee_Platform_A − Fee_Platform_B − SlippageSize your position — On thin order books, your own order moves the price. Start with a 10–20% fill at quoted price assumption for markets under $50K liquidity.
Execute simultaneously — Delays between legs introduce timing risk. Manual execution is fine for slow-moving markets; automated bots are essential for news-driven gaps.
Track resolution timelines — Capital is locked until both contracts settle. Model your opportunity cost. A 4% arb over 45 days is only ~32% annualized — good, but factor this into sizing.
Log and audit — Track every trade's gross spread, net profit after fees, and actual vs. expected resolution time. Your data becomes the foundation for improving your scanner.
For traders looking to automate this workflow, the trader playbook on AI agents for prediction markets walks through how power users are deploying automated pipelines to capture these gaps at scale.
Fee Math: When an Arb Is Actually Worth It
The most common mistake new arbitrageurs make is ignoring fees until after they've entered a trade. Here's the reality check:
Polymarket charges a 2% fee on taker orders. Kalshi's fees range from 3% to 7% depending on market and volume tier.
For a complementary arb (Yes on Polymarket + No on Kalshi):
- Buy Yes at 55¢ on Polymarket → effective cost after 2% fee: 56.1¢
- Buy No at 42¢ on Kalshi → effective cost after 5% fee: 44.1¢
- Total cost: 100.2¢
That's a losing trade despite an apparent 3-cent gross spread. You need the raw spread to exceed combined fees by at least 1–2¢ to have any margin after slippage.
Rule of thumb for power users: Only pursue arbs where the gross spread is ≥ 7–8% on Polymarket–Kalshi pairs. On lower-fee venue pairs, 4–5% can still be viable.
This fee sensitivity is especially critical for high-frequency plays like earnings surprise markets, where multiple trades per event compound fee drag quickly.
Automation and Tooling for Serious Arbitrageurs
Manual scanning doesn't scale. Power users running serious arbitrage operations use at least two of these tooling layers:
Data Layer
- Polymarket CLOB API — Real-time order book data, REST and WebSocket
- Kalshi Market API — RESTful endpoints for market prices and positions
- Custom aggregators — Python scripts pulling both feeds into a unified price matrix, flagged on spread threshold
Execution Layer
- Smart order routing — Automatically routes buys to the cheapest platform and sells to the most expensive
- Latency optimization — Co-locate execution scripts on cloud instances close to API endpoints
- Position limits — Hard-coded guards to prevent oversizing on illiquid markets
Monitoring Layer
- PnL dashboards — Track realized vs. unrealized across all open arb positions
- Resolution trackers — Alert when a contract approaches its resolution date to flag capital release timing
- Anomaly detection — Flag unusual gaps that may signal resolution disputes rather than genuine price inefficiency
PredictEngine provides a unified interface for power users who want to monitor cross-platform market data without building the full data pipeline from scratch.
For a more technical look at how automated systems analyze order books at scale, the AI-powered order book analysis piece covers backtested results on a $10K portfolio using algorithmic approaches.
Common Arbitrage Traps and How to Avoid Them
Even experienced traders fall into these:
1. Resolution criterion mismatch — Two platforms resolve "the same" event on different triggers. Always read the fine print on both sides before entering.
2. Correlated liquidity crunches — During high-volatility news events, both platforms widen spreads simultaneously. Your arb entry may look clean at 10am, then unprofitable by the time your second leg fills.
3. Counterparty/platform risk — Crypto-native platforms carry smart contract risk. Diversify platform exposure; don't put 100% of arb capital on one platform.
4. Capital lock-up underestimation — A 5% arb that takes 60 days to resolve is only ~30% annualized. That sounds good until you miss a 15% arb because capital was locked.
5. The "obvious" arb that isn't — If a gap looks huge and obvious, assume information asymmetry first. Other traders may know something about resolution that you don't. The gap in house race prediction markets around election night 2024 looked like free money — until precinct reporting delays triggered non-standard resolution on one platform.
Frequently Asked Questions
What is the minimum capital needed to start prediction market arbitrage?
You can technically start with as little as $200–$500, but transaction fees and slippage will eat most of your edge at that level. Most serious arbitrageurs find that $5,000–$10,000 per position is where fee drag drops to a manageable percentage of gross spread. Below $2,000, arbitrage is primarily a learning exercise, not a profit center.
How do I find prediction market arbitrage opportunities in real time?
The most reliable method is building or subscribing to a price aggregator that polls multiple platform APIs simultaneously and flags markets where the same event is priced more than 5–8% apart. Platforms like PredictEngine offer monitoring tools so you don't have to build the full pipeline yourself. Manually checking platforms once per hour can catch slower-moving gaps but will miss news-driven spikes.
Are prediction market arbitrage profits taxable?
In the United States, profits from prediction market trading are generally treated as ordinary income or capital gains depending on your jurisdiction and holding period. CFTC-regulated platforms like Kalshi issue 1099s for qualifying users. Crypto-native platforms like Polymarket require self-reporting. Always consult a tax professional familiar with derivatives and digital assets — this is an evolving area.
Can I automate prediction market arbitrage completely?
Yes, and many power users do. You need API access to both platforms (Polymarket provides a CLOB API; Kalshi has a REST API), a matching engine that identifies and sizes opportunities, and a smart order router that executes both legs with minimal latency. The AI agents for prediction markets playbook covers the full technical architecture. The biggest challenge isn't automation — it's maintaining the resolution-criteria verification step programmatically.
What's the difference between arbitrage and market making on prediction markets?
Arbitrage exploits price gaps between platforms for the same event; profit comes from the spread, not from taking directional risk. Market making involves posting limit orders on both sides of a single market to capture the bid-ask spread; profit comes from providing liquidity. The two strategies can be combined — many algorithmic market makers also run cross-platform arb scanners to pick off mispricings when they appear.
How quickly do prediction market arbitrage gaps close?
On high-liquidity political markets, obvious gaps (>5%) often close within minutes to hours as other arbitrageurs spot and act on them. On niche markets or those with fewer active traders, gaps can persist for days or even weeks. The biggest windows typically open immediately after major news breaks — the first 10–30 minutes often have the widest spreads before liquidity catches up across platforms.
Final Thoughts and Next Steps
Prediction market arbitrage isn't a get-rich-quick strategy — it's a systematic, process-driven edge for traders willing to do the fee math, read the fine print, and either build or access the right tooling. The most consistent performers in this space treat every trade as a data point, refine their scanners continuously, and scale carefully as they validate their execution pipeline.
Whether you're running your first manual arb or looking to automate a multi-platform scanning system, PredictEngine gives power users the infrastructure to monitor markets, track positions, and identify cross-platform price gaps without building everything from scratch. Explore the platform, check the pricing page to find the tier that fits your trading volume, and start capturing the edges that casual traders leave on the table.
Run your own quant desk
Felix lets you research, backtest and run automated trading strategies, with risk controls built in.
Try Felix