Arbitrage locks in a profit whichever way a market resolves. On Polymarket, and between Kalshi and Polymarket US, the gaps are real but small, and the fees and the rules decide whether one is worth taking. This guide shows where they open and prices them at the fees each venue charges today.
Every prediction market contract pays one dollar to the winning side and nothing to the losing side. If you can own every possible winner for less than a dollar, you've locked in the difference whatever happens. That gap, after fees, is the arbitrage.
The textbook example, Yes at 55¢ and No at 42¢, can't happen on Polymarket. Yes and No trade on one shared order book, so an order to buy No at 48¢ is the same order as one to sell Yes at 52¢. The best No price is always a dollar minus the best Yes bid, which means buying both sides costs a dollar plus the spread.
Kalshi's book works the same way. The gaps worth looking for sit between markets, not inside one.
Each trade below owns every possible winner for less than a dollar. They differ in where the mispricing comes from, how long it tends to last, and what it costs to take.
When an event splits into one market per outcome, exactly one of them resolves Yes. Each outcome has its own order book, so their prices can drift apart. If the Yes asks across the full set add up to less than a dollar, owning one of each pays a dollar for less.
The set has to be complete, including the catch-all outcome, or you're betting on the ones you left out. In this example 100 sets cost $96.00 and pay $100.00. After $2.54 in taker fees at Polymarket's politics rate, $1.46 is locked in whichever outcome wins. The rates for every category are on our Polymarket fees page.
Some markets imply others. A candidate can't win the election without first winning the nomination, so the election can never be worth more than the nomination. When it trades higher anyway, buying Yes on the nomination and No on the election pays at least a dollar in every case.
These gaps are rarer and slower to close, because no order book links the two markets. Someone has to notice the relationship first.
The same question often trades on Kalshi and on Polymarket US at different prices. Buy Yes on one and No on the other, and if the two cost less than a dollar together, one of them pays. This is the gap most traders mean by prediction market arbitrage, and it's the one fees hit hardest, because you pay two venues.
Polymarket's main exchange doesn't open new positions for traders in the United States, so for a US trader the pair is Kalshi and Polymarket US. Both are regulated by the CFTC.
Kalshi and Polymarket US both charge takers a rate times contracts times price times one minus the price. Fees peak at 50¢ and fall toward the extremes, so the gap a pair needs depends on where it trades. Near 50¢ a pair needs about 4.1¢ of gap just to break even. Near 90¢ it needs about 1.5¢.
You pay the taker rate on each venue. Today that's 0.07 on Kalshi and 0.095 on Polymarket US. Polymarket US lowers its rate for traders with more volume, and Kalshi rounds each order's fee up to the next cent.
Polymarket US charges makers nothing, so posting a resting order on that leg and taking only the Kalshi side cuts the fees to 1.8¢ at 50¢, against 4.1¢ taking both. The cost is legging risk. Your resting order might fill after the Kalshi price has moved, or not fill at all.
Kalshi charges makers on 68% of the markets TAPE tracks, at a rate of 0.0175. Resting the Kalshi leg instead saves less, so the Polymarket US side is usually the one to post.
How many cents of gap a Kalshi and Polymarket US pair needs before it makes anything, at today's rates. Fees are symmetric, so a Yes leg at 20¢ costs the same as one at 80¢.
| Yes leg price | Taking both legs | Resting the Polymarket US leg |
|---|---|---|
| 5¢ or 95¢ | 0.8¢ | 0.3¢ |
| 10¢ or 90¢ | 1.5¢ | 0.6¢ |
| 20¢ or 80¢ | 2.6¢ | 1.1¢ |
| 30¢ or 70¢ | 3.5¢ | 1.5¢ |
| 40¢ or 60¢ | 4.0¢ | 1.7¢ |
| 50¢ or 50¢ | 4.1¢ | 1.8¢ |
Two contracts can share a headline and still resolve differently. If they can, you don't hold an arbitrage. You hold two bets that can both lose. Read both rulebooks for these four things before you trade the gap.
The data each venue settles on. One may use an official agency and the other a news outlet or a different feed.
The cutoff date, time, and time zone. A result announced late in the evening can land inside one window and outside the other.
The exact threshold and definitions. Above versus at or above, or a figure as first reported versus as later revised.
Postponements, cancellations, and ties. One venue may void the market while the other resolves it to a side.
A three cent gap isn't three cents of profit. Six things take a bite before you keep anything, and most gaps that look free are one of these in disguise.
Each leg you take pays a taker fee, and a gap smaller than the fees isn't an edge. Price it first with our payout calculator.
The best price is good for a small size. Buy more than sits there and your average price walks up, which can close the gap before you finish.
If one leg fills and the other price moves before you get there, you're left holding a one-sided bet.
The payout comes when the market resolves, which can be months away. Three cents for three months is a very different return from three cents for three days.
A cross-venue pair needs funded accounts on both exchanges, and moving money between them takes time. Cash on the wrong venue can't take the gap.
A market that's canceled or resolved 50/50 breaks the math. Our Settlement Risk Index tracks how often markets are voided.
The clean gaps get taken in seconds by bots watching every market. What's left is spotting the gaps that open when a market is thin or slow, and knowing whether the book is deep enough to take them at size. That's a question about the order book and who's standing in it, which is what TAPE shows: the live spread, the depth at each level, and whether the trader on the other side is a market maker or directional money.
Price a single trade: the payout, the profit after fees, and the implied odds.
The payout is locked once you own every possible winner, but the trade isn't free of risk. Fees and spread can turn a small gap into a loss, the legs can fill at different times, your cash is tied up until the market resolves, and contracts on two venues can resolve differently. A real arbitrage is one where the gap survives all of that.
No. Yes and No share one order book, so the best No price is always a dollar minus the best Yes bid, and buying both sides costs a dollar plus the spread. Real gaps open between markets: across the outcomes of one event, between markets that depend on each other, and between venues.
Yes, when the same question trades at different prices and the gap clears both venues' fees. Near 50¢ that takes about 4.1¢ at today's taker rates. Polymarket's main exchange doesn't open new positions for US traders, so they pair Kalshi with Polymarket US. Check that both contracts settle on the same source, at the same time, and on the same wording.
The obvious ones close in seconds, because bots watch every market for exactly this. The gaps that last are in thinner markets, where there isn't enough depth for a bot to take size, and in markets whose link to each other isn't obvious. That's where reading the order book matters most.
A good bet pays only if you're right about the outcome. An arbitrage pays whichever way the market resolves, because you own every winner. Most trades that feel like free money are really directional bets with a thin cushion.
No. TAPE is the intelligence layer: the live spread, the book depth, and the trader labels that tell you whether a gap is real and takeable at size. You place the trades yourself.
More free tools in resources.
Book a demo and we will show you the spread, the depth, and who is on the other side of your markets.