Prediction Market Arbitrage: Why Cross-Venue Gaps Persist
Polymarket charges takers by category and makers nothing; Kalshi fees vary by market. How fees, rules and settlement delays turn price gaps into risk.
Jump to a section (9)
Prediction market arbitrage means holding both sides of the same question on two venues for less than the payout one side is certain to earn. The gaps exist, but most are priced rather than free. Polymarket and Kalshi charge differently, settle differently and resolve on their own rules text, and each difference is a cost or a risk the quoted spread does not show. Before booking a gap as profit, check three things: the fee on each leg, whether both contracts resolve on the same terms, and how long each venue takes to settle.
Why cross-venue price gaps exist
The two venues are not selling the same instrument, even when the headline matches. A gap between them is partly the market pricing that difference.
On Polymarket, winning tokens redeem for $1 each and losing tokens become worthless. What counts as winning is set by the market's rules, not its headline: Polymarket states that the rules define how it resolves, including the source, the end date and edge cases. Kalshi works the same way on paper. The outcome rules, the information used and its source sit in each contract's terms and conditions.
So a "yes" on one venue and a "no" on the other is a hedge only if both rulebooks answer the question identically. Traders who price that risk differently, or who face different fees, will quote different prices for what looks like one event.
Fees: the first cost a spread has to clear
Polymarket charges only takers, with a fee that depends on price and category. Kalshi's fees vary by market and can fall on makers too. A spread that ignores either schedule is overstated.
Polymarket's taker fee formula is fee = C × feeRate × p × (1 - p), where p is the price. Two consequences follow. First, makers are never charged fees; only takers pay. Second, the fee is largest in the middle: it peaks at 50% probability, where Polymarket's own example puts it at $1.75, and shrinks toward both extremes. The rate also depends on the contract category, and geopolitical and world events markets are fee-free. These terms are as of October 2026.
Kalshi says some markets carry different fees from others, so one fee assumption does not hold across its contracts. It also charges maker fees on resting orders in some markets, applied when the resting order executes. These terms are as of April 2026.
| Cost line | Polymarket (as of October 2026) | Kalshi (as of April 2026) |
|---|---|---|
| Who pays | Takers only | Takers, and makers in some markets |
| Fee basis | C × feeRate × p × (1 - p), set by category | Varies by market |
| Resting orders | No maker fee | Maker fee in some markets, charged on execution |
| Zero-fee markets | Geopolitical and world events | Not stated in Kalshi's fee page |
The table turns into a practical rule. A leg taken near even odds on Polymarket pays the most. A leg posted as a resting order avoids Polymarket's fee but may not fill, which leaves the other leg unhedged. On Kalshi, resting does not guarantee a fee-free fill. The net spread is the quoted gap minus both legs' fees, computed per contract, not per venue.
Resolution rules: the risk the price does not show
Fees are a known cost. Resolution mismatch is an unpriced risk, and it is the one that turns a hedge into two losing positions.
Because each venue resolves on its own rules text, two contracts can share a headline and still differ on the reporting source, the cut-off time or how an edge case is treated. If one venue's source reports an outcome that the other venue's source does not, both legs can settle against the holder. No fee schedule captures that. The check is to read both rulebooks side by side before entering, not after.
Settlement timing and the dispute path
The venues also settle on different clocks. Capital sits in both legs until the slower one pays, and a dispute can reopen an outcome the other venue has already closed.
Polymarket resolves through UMA. After an outcome is proposed there is a 2-hour challenge period in which anyone can dispute it. If a second proposal is also disputed, the question escalates to a UMA token holder vote through the DVM. Kalshi determines outcomes from source-agency data, and that can take one hour to more than twelve hours after the market closes.
For a cross-venue position this means two separate exposures. One leg can be paid while the other waits, which ties up capital that the spread was meant to reward. And a disputed Polymarket outcome can land differently from a Kalshi determination already made, which is the resolution mismatch arriving late.
What the best prediction market arbitrage looks like
The best candidates are the dull ones: identical resolution terms, low combined fees and short, predictable settlement on both sides. The size of the gap matters less than how much of it survives those checks.
- Matching rules. Same source, same end date and the same treatment of edge cases on both venues.
- Fees netted per leg. Polymarket's category rate and price-dependent fee on one side, Kalshi's market-specific schedule on the other, including any maker fee on a resting order.
- Category advantage. A Polymarket leg in a fee-free geopolitical or world events market lowers the cost floor.
- Settlement exposure. Allow for Polymarket's dispute path and Kalshi's source-agency delay before counting the capital as free again.
A gap that passes all four is a real spread. A gap that fails one is a position with risk the price does not disclose.
Intel Q&A
How risky is prediction market arbitrage?
Riskier than the word suggests. The price gap is visible; the fee on each leg, a mismatch in resolution wording and a delayed or disputed settlement are not. Any one of them can turn a hedged pair into a loss, and a wording mismatch can make both legs lose.
Is prediction market arbitrage still profitable after fees?
Only where the spread clears both venues' fees. Polymarket's taker fee is largest near even odds and zero in geopolitical and world events markets, while Kalshi's fees vary by market and can include maker fees. Net the spread against both before counting it.
Can AI be used for arbitrage trading?
Software can scan prices and compute fees faster than a person. It cannot settle whether two contracts resolve the same way unless someone has compared the rules text on both venues, and that comparison is where most apparent arbitrage fails.
Who decides the outcome when a Polymarket market is disputed?
A proposed outcome can be challenged during a 2-hour window. If a second proposal is also disputed, the question goes to a vote of UMA token holders through the DVM, which can push settlement well past the other venue's.
Source file
- Polymarket Documentation, Polymarket Resolution, undated, accessed October 2026
- Polymarket Documentation, Polymarket Fees, undated, accessed October 2026
- Kalshi Help Center, Kalshi Fees, April 2026
- Kalshi Help Center, Kalshi Market Outcomes, March 2026
