Kalshi Arbitrage: Settlement, Dutch-Book & Cross-Market Edges

What Arbitrage Means in Prediction Markets

Arbitrage is locking in a profit that does not depend on how an event resolves. In prediction markets it leans on one hard fact: every binary contract settles at exactly $1 if the event happens or $0 if it does not. Kalshi arbitrage means finding contracts whose prices are internally inconsistent, then buying the cheap side so the settlement math returns more than you paid.

It sounds like free money. It rarely is. True riskless arbitrage on Kalshi is rare and fleeting, and fees, order limits, and settlement timing usually erase the gap before you can close it. Most of what traders label "arb" is really a statistical edge that still carries a substantial risk of loss.

Prediction market arbitrage comes in a few recognizable shapes. This guide walks through each, then covers why the clean version is so hard to actually bank.

YES + NO Mispricing on a Single Market

On one binary market, YES and NO are opposite bets on the same question. Exactly one of them settles at $1. In an efficient market their prices sum to $1. When they sum to less than $1, buying both sides guarantees a $1 payout for less than $1 of cost.

Example: a contract shows YES at $0.42 and NO at $0.55. The pair costs $0.97. Whatever happens, one side pays $1.00.

The gross edge is $0.03 per pair. That is the textbook YES + NO mispricing. In practice these gaps are tiny, appear for seconds, and shrink once fees and the bid-ask spread are counted. You also buy at the ask on both legs, not the mid, so the real cost is usually higher than the quoted sum suggests.

Mutually-Exclusive Sets That Don't Sum to 100%

Many Kalshi events split into several mutually exclusive contracts — "which candidate wins," or "what range will CPI land in." Across a complete, mutually-exclusive-and-exhaustive set, exactly one outcome resolves YES. If you buy YES on every outcome and the combined cost is below $1, one of them pays $1 and you keep the difference.

When the outcomes are all trading and their prices sum to more than $1, the set is overpriced; when they sum to less than $1, it is underpriced. A set that sums to $1.06 or $0.94 instead of $1.00 is a "Dutch book" — an internally inconsistent set of odds.

The catch: exhaustiveness has to be real. If the market has an "other" or "none of the above" bucket you missed, your set is not complete and the trade is not riskless.

Settlement and Near-Certain-Outcome Arbitrage

Sometimes an outcome is effectively decided but the contract has not officially settled. A race is called, yet YES still trades at $0.96 for a few minutes. Buying at $0.96 to collect $1 looks like a locked 4-cent gain. This is settlement arbitrage — trading the repricing window around resolution.

It is not free. The result can be disputed, the official settlement source can lag or differ from the news, and your capital is tied up until the exchange pays out. Kalshi settles to a specific written rule, not to what a headline says, so read the resolution criteria before assuming an outcome is certain. See the settlement arbitrage glossary entry for the mechanics and the failure modes.

Kalshi vs Polymarket: Cross-Venue Arbitrage and Why It's Hard

The same real-world question is often listed on both Kalshi and Polymarket. When Kalshi prices NO at $0.52 and Polymarket prices the matching YES at $0.45, the two venues disagree, and the gap looks arbitrageable. This is Kalshi vs Polymarket arbitrage, and it is much harder to execute than a single-market gap.

Reasons it is hard for US traders:

• Different settlement currency. Kalshi settles in US dollars; Polymarket settles in USDC on-chain. Moving and converting capital adds cost, time, and friction.
• Legal and state-level uncertainty. Polymarket returned to US users in late 2025 under a CFTC-regulated Designated Contract Market license, but several states are challenging event contracts. Availability is not uniform across the country. Check the current rules for your state before assuming access.
• Resolution wording differs. The two venues can word "the same" market with different sources, dates, or edge-case rules. A gap can be genuine disagreement about resolution, not free money.
• Capital on two platforms. You need funded accounts on both, which kills the speed a fleeting gap demands.

A side-by-side of the two venues lives at Kalshi vs Polymarket.

How Fees and Slippage Eat Thin Edges

This is where most prediction market arbitrage dies. Kalshi charges trading fees that scale with contract price and quantity, and you pay them on entry — often on both legs of a two-sided trade. A 3-cent gross edge can turn negative once fees hit both sides.

Slippage compounds it. Order books on many event contracts are thin. If a $0.42 YES only has 200 contracts available, your 1,000-contract order fills the rest at worse prices, and the average cost climbs above the level that made the trade look profitable. The larger you size to make a thin edge worthwhile, the more you move the price against yourself.

The honest summary: a gap has to clear fees on every leg, survive the spread, and fit inside available depth before it is an edge at all. Screenshots of "guaranteed" arbitrage rarely account for all three.

Automating Arbitrage Detection

Because these gaps are small and short-lived, they reward scanning breadth and speed over deep analysis. A detector continuously pulls prices across every open market, computes the implied probabilities, and checks the consistency rules: does YES + NO sum below $1, does a mutually-exclusive set sum away from $1, does one venue disagree with another. Crucially, it must subtract fees and check available order-book depth before flagging anything — an edge that ignores those is not an edge.

Even fully automated, this is competitive. Other bots watch the same books, and the cleanest gaps close in seconds. Realistic automated "arbitrage" is usually a blend of true consistency trades and statistical edges, run with strict risk controls. None of it guarantees a profit, and prediction markets carry a substantial risk of loss.

Tradewink Predictions is an autonomous agent that scans Kalshi markets continuously and can watch far more contracts at once than a person can — the breadth that catching fleeting mispricings requires. It sizes positions with fractional Kelly, enforces per-bet and daily-loss limits, and defaults to paper mode so you can watch its behavior before risking capital. It is a research and execution tool, not investment advice, and it makes no performance promises. Read the Kalshi trading strategies it uses, or see the Predictions overview.