For most of financial history, arbitrage has been a game for institutions. You needed co-located servers, direct exchange access, prime brokerage relationships, and enough capital to make fractional-cent price differences worth capturing. Retail traders could read about it in textbooks. They couldn't do it.
Prediction markets just changed that. Platforms like Polymarket and Kalshi have created a new class of markets where the barriers to entry are a crypto wallet and some Python scripts. And the inefficiencies are real.
Arbitrage in 30 Seconds
Arbitrage is buying something in one place and simultaneously selling it in another for a higher price. The classic example: if gold trades at $2,000 on the New York exchange and $2,005 on the London exchange, you buy New York and sell London. Pocket the $5. No directional risk—you're not betting on whether gold goes up or down. You're exploiting the fact that two markets disagree on the same thing.
On Wall Street, these gaps close in milliseconds. High-frequency trading firms spend hundreds of millions on infrastructure specifically to capture these micro-discrepancies before anyone else can. The opportunities exist, but they're invisible to anyone without nanosecond-level execution.
Prediction markets are different. The gaps are wider, they persist longer, and the infrastructure required to exploit them is orders of magnitude simpler.
Why Prediction Markets Have Gaps
Polymarket runs on the Polygon blockchain. Kalshi is a CFTC-regulated exchange operating on traditional rails. They both offer binary outcome contracts—"Will X happen? Yes or No"—but they operate in completely different ecosystems with different user bases, different liquidity profiles, and different settlement mechanisms.
This means the same event can trade at meaningfully different prices on each platform. If Polymarket prices "Will the Fed cut rates in June?" at 62 cents (YES) and Kalshi prices the same outcome at 57 cents, there's a 5-cent spread. Buy YES on Kalshi at 57, sell YES on Polymarket at 62, and you've locked in 5 cents of profit per contract regardless of what the Fed actually does.
These spreads show up because the platforms have fragmented liquidity. Polymarket skews crypto-native and international. Kalshi skews US-based and more traditional. News hits them at different speeds. Sentiment shifts propagate unevenly. And neither platform has the kind of institutional market-making infrastructure that keeps traditional exchanges tightly aligned.
How It's Different from Wall Street
On traditional exchanges, arbitrage is a speed game. The information is the same everywhere; the only question is who can act on it fastest. The edge is infrastructure: fiber optic cables, FPGA chips, co-location contracts.
Prediction market arbitrage is more of an information and access game. The inefficiencies aren't microsecond price lags—they're structural. Different platforms, different user demographics, different regulatory regimes, different deposit and withdrawal mechanics. The gaps can persist for minutes or hours, not milliseconds.
This means you don't need a $50 million infrastructure stack. You need a way to monitor prices on both platforms, a funded account on each, and a script that can execute when the spread exceeds your threshold. A laptop, some API calls, and basic automation. It's genuinely accessible to anyone who can write a bot.
What a Setup Actually Looks Like
The realistic version: you write a price monitor that polls both Polymarket's API and Kalshi's API for overlapping markets. When the implied probability diverges beyond a threshold that accounts for fees, slippage, and settlement risk, you execute offsetting positions on both sides. You need funded accounts on both platforms—USDC on Polymarket (Polygon), USD on Kalshi. You need to handle the settlement asymmetry: Polymarket settles on-chain, Kalshi settles in dollars.
The edge cases are where it gets interesting. Liquidity isn't always deep enough to fill both sides at the prices you saw. Order books move. Gas fees on Polygon are cheap but not zero. Kalshi has its own fee structure. You need to model all of this into your threshold calculation or you'll find yourself locking in spreads that are negative after costs.
More sophisticated setups add a third data source—polling aggregates, news sentiment, or their own probability models—to identify not just cross-platform discrepancies but cases where both platforms are mispriced relative to an independent estimate. This starts to blur the line between arbitrage and informed speculation, but it's where the bigger edges probably live.
The Lines You Should Know About
This is where the excitement needs some cold water.
Polymarket's terms of service prohibit US-based users from trading on the platform. The site is built on crypto rails partly to operate outside US regulatory jurisdiction. If you're in the US and accessing Polymarket through a VPN to mask your location, you're violating their terms of service at minimum—and potentially running afoul of CFTC regulations around unregistered derivatives trading. The enforcement reality has been inconsistent, but "nobody has stopped me yet" is not a legal strategy.
Kalshi is CFTC-regulated, which means it's legal for US residents but comes with restrictions. Not all event types are available. Political event contracts were in regulatory limbo for years before Kalshi won its legal battle in 2024. But regulated doesn't mean anything-goes—there are position limits, and automated trading has its own compliance considerations.
Using external data sources to inform your trades—polling data, private models, scraping news feeds—is generally fine. But if your strategy involves non-public information (insider knowledge about event outcomes), you're in manipulation territory. Prediction markets are still markets, and market manipulation rules still apply. The CFTC has been explicit about this.
The crypto layer adds its own risk. Polymarket runs on Polygon, which means your capital is exposed to smart contract risk, bridge risk if you're moving funds from Ethereum, and the general volatility of holding USDC in a DeFi-adjacent environment. These aren't theoretical risks—bridge exploits have cost users hundreds of millions across the broader crypto ecosystem.
Is It Worth It?
The honest answer: probably, for the right person with the right expectations. If you're a developer who can write bots, understand market microstructure, and are comfortable with the regulatory ambiguity, there's a genuine window here. Prediction markets are young, fragmented, and inefficient. That's the textbook environment for arbitrage.
But the window is exactly that—a window. As these markets mature, the spreads will tighten. More sophisticated participants will enter. Market makers will bridge the platforms. The structural inefficiencies that make this possible today are symptoms of an immature market, and markets grow up.
On Wall Street, arbitrage was once available to anyone with a phone and a seat on two exchanges. Then electronic trading compressed the gaps. Then HFT compressed them further. Now you need a nine-figure budget to compete. Prediction markets might follow the same arc on a compressed timeline.
If you're going to play this game, play it with eyes open. Understand the regulatory landscape before you fund an account. Model your costs honestly. And don't assume the opportunity persists forever—the whole point of arbitrage is that it eliminates itself.