Prediction markets like Polymarket, Kalshi and PredictIt often show different prices for what looks like the same election, because each has its own traders, money, fees, limits and contract rules. Arbitrage means buying the cheap side on one platform and the opposite side on another so that one of the two positions pays out whatever happens. Traders doing that push the prices back together, which is why big gaps rarely last. It sounds risk-free, but fees, locked-up money and differences in how each contract resolves usually eat most of the edge.
If you follow markets as a signal rather than trade them, the useful takeaway is that a two- or three-cent difference between platforms means very little. This is an explainer, not trading advice.
How are prediction market prices set? #
Each contract pays $1 if an event happens and nothing if it doesn’t. A “Yes” price of 55 cents is read as a 55% chance, and “No” on the same outcome trades near 45 cents. Prices are set by traders buying and selling against each other, so they reflect the views and money of whoever happens to be on that platform. Our guide to reading prediction market odds covers the conversion in more detail.
Why do prices differ between platforms? #
Several things keep the platforms from agreeing perfectly:
| Reason | How it creates a gap |
|---|---|
| Different traders | Each site draws different users with different views |
| Different money | Polymarket’s international site settles in crypto; Kalshi and PredictIt use dollars and bank transfers; moving money between them takes time |
| Fees | PredictIt charges 10% of profits and 5% on withdrawals; other platforms charge per-trade fees, so the break-even price differs |
| Position limits | PredictIt caps each trader at $3,500 per contract, which limits how much any one person can correct a mispricing |
| Contract wording | “Which party controls the Senate” and “which party wins more seats” aren’t the same bet when the vice president breaks ties |
| Liquidity | Thin markets have wide gaps between the best buy and sell prices |
For a fuller comparison of the three platforms, see PredictIt vs. Polymarket vs. Kalshi.
How does cross-platform arbitrage work? #
Say two exchanges list the same question: will Party A control the House after the 2026 election?
- Exchange 1: Yes costs 47 cents.
- Exchange 2: No costs 49 cents.
Buy 1,000 Yes contracts on Exchange 1 ($470) and 1,000 No contracts on Exchange 2 ($490). You’ve spent $960. Whatever happens, one set pays $1,000 and the other pays nothing, so you collect $1,000. That’s a $40 gross profit, about 4.2%.
Now the costs:
- Fees. If the winning side sits on a platform that takes 10% of profits, and your profit on that leg is $530 (you paid $470 for contracts that pay $1,000), the fee is $53, more than your entire $40 edge. On platforms with per-trade fees, you pay on both legs whether you win or lose.
- Time. Your $960 is tied up until the market settles, which for a November election bought in September is at least two months, longer if a race goes to a recount. A 4% gross return over several months has to beat what that money would earn in a savings account or Treasury bills.
- Moving money. Funding two platforms, especially one in crypto and one in dollars, costs time and sometimes transfer fees.
That’s why most visible gaps sit there untouched: after costs, they aren’t profitable.
What is “resolution risk”? #
This is the trap that turns a sure thing into a loss. Two contracts that look identical can be written differently:
- One market might settle on which party’s senators hold a majority, another on which party organizes the chamber. In a 50-50 Senate, the vice president’s party controls it.
- One might resolve when the Associated Press calls the race, another only when the result is certified, and the deadlines might differ if there’s a recount or legal challenge.
- A market on “who wins the election” may resolve differently from one on “who is inaugurated” if something unusual happens in between.
If the two contracts resolve differently, you can lose both legs. Always read each contract’s rules and resolution source before assuming two prices describe the same event. Our explainer on how prediction markets pay out covers settlement rules.
What about arbitrage within one platform? #
In a market with many candidates, such as a party’s presidential nomination, the Yes prices for all candidates should add up to about $1, since only one can win. They often add up to more, because traders like buying Yes on favorites and long shots and fewer bother selling. If the Yes prices add up to $1.10, buying No on every candidate can lock in a small gain, since all but one of those No contracts pays out.
Some platforms offer a feature, sometimes called “negative risk,” that recognizes those No positions offset each other and asks for less collateral. Fees and the cost of trading many thin markets still cut into it.
Is correlated-market trading arbitrage? #
Not really. Traders sometimes compare state-by-state markets with a national market and argue the prices are inconsistent. That’s a bet on a model, not a locked-in payout: the states aren’t independent, and the national result can surprise everyone. It’s a judgment call dressed up in arithmetic.
What should you take from price gaps as a follower? #
- Small gaps are noise. Two or three cents between platforms usually reflects fees and liquidity, not different information.
- Big gaps deserve a closer look. They usually come from different contract wording or a thin market, or occasionally from one large trader, which we cover in can election prediction markets be manipulated?
- Compare markets with polls. A market that disagrees with the polling average is making a claim about something the polls might be missing, which may or may not be right.
Election Tracker shows current Polymarket prices, with 24-hour moves, for the races and control markets it’s configured to follow, alongside its own polling averages for the same races. It doesn’t show Kalshi or PredictIt prices and it doesn’t trade: the Market Sentiment tab is read-only, needs no account and labels prices as trader sentiment, not forecasts. It’s free on iPhone.
Frequently asked questions #
Is prediction market arbitrage legal? #
Trading the same event on two regulated platforms is legal, as long as you follow each platform’s rules and eligibility requirements. Using a VPN to reach a platform that doesn’t serve your location breaks its terms.
Why don’t Polymarket and Kalshi prices match? #
They have different users, fees, funding methods and contract rules, and moving money between them takes time. Traders narrow the gaps, but small differences persist because closing them isn’t worth the cost.
Is arbitrage risk-free? #
Rarely. Fees, the time your money is locked up and differences in how each contract resolves can each wipe out a small edge. Resolution risk can turn a supposedly hedged position into a loss on both sides.
What is negative risk in prediction markets? #
It’s a feature on some platforms for multi-outcome markets. Because only one outcome can win, a set of No positions across all outcomes can’t all lose, so the platform requires less collateral to hold them.
How big are price differences between platforms? #
On heavily traded markets, usually a few cents or less. Larger gaps tend to appear in thinly traded markets or when contracts are worded differently.