Arbitrage betting, explained

Two venues price the same bet. One of them is willing to sell you the yes side for 46¢ and the other is willing to sell you the no side for 52¢. Together that is 98¢ for a pair of contracts that is certain to be worth exactly $1 when the game is over, because one of them must settle at a dollar and the other must settle at nothing. Buy both and you have paid 98¢ for a dollar. That is the entire idea, and everything else on this page is a complication of it.

The 2¢ does not care who wins. You are not forecasting anything — you have covered both outcomes — so the usual question of whether a bet was a good idea does not apply. What replaces it is a set of much more boring questions: were both prices real, did both bets actually get accepted, and was the 2¢ still 2¢ after each venue took its cut. Those are the questions that decide whether arbitrage betting works for a person, and they are almost the only things worth thinking about.

Two footnotes on the arithmetic. It is the same arithmetic in American odds, just harder to see: −110 and +120 on opposite sides of the same line come to 98.03% together, the same 2¢ wearing a different unit. And the stake is not split in half — each side takes a share in proportion to what it costs, which is what makes the two returns come out equal.

Yes, everywhere sports betting itself is legal. There is no law anywhere in the United States that treats two ordinary bets at two ordinary licensed venues differently because the person placing them worked out that the prices disagreed. You are using two regulated products exactly as they are advertised, at the prices they published.

The confusion comes from a real thing wearing the wrong word. What arbitrage can breach is a sportsbook's terms of service, which is a contract between you and a company rather than a law — and even there, most books do not prohibit it outright. What their terms reserve is the right to limit your stake size, decline a bet, or close the account, at their discretion and without explaining why. That is a commercial decision, not a legal one, and the worst realistic outcome is a $4 maximum stake rather than anything that follows you out of the building.

On the exchange side the question barely arises. Kalshi is regulated by the CFTC as a derivatives exchange, and holding offsetting positions in related contracts is ordinary trading there rather than a strategy anyone objects to. An exchange makes its money on volume rather than on your losses, so a profitable account is a good customer.

One thing that is genuinely your problem: winnings are taxable income in the United States whether or not a venue issues you a form, and the two legs of a position are two separate events for that purpose — the winning one is income and the losing one is a deduction only if you itemise. Talk to someone who does this for a living before the volume gets large.

Start with what the book is actually looking at, because most advice on this gets it backwards. A sportsbook's risk team is not comparing your bets against some other venue's board. It is watching how your money behaves around its own line moves — and the pattern that stands out is small, precise, fast: an odd stake to the dollar, placed at the exact posted number, minutes after that number moved, always on the side the move went against. Do that thirty times and you are legible whether or not anyone knows why.

So the things that help are the ones that make the money look ordinary. Round the stake. Do not always take the maximum. Leave a gap between the line moving and your bet landing where the position lets you. Have some ordinary bets in the account — parlays, the team you actually watch — because an account with nothing in it but sharp sides at closing prices is an account with a shape. Spread the volume across more than one book instead of taking everything at the one with the best prices.

The structural version of the same advice: take the exchange leg at the sportsbook's expense rather than the other way round. Kalshi, Polymarket and Novig have no reason to restrict a profitable account, so a habit of putting the side that has to be big on an exchange keeps the book side small and unremarkable. It also means the leg that can vanish is the one you place first.

None of this is a method, and nobody should sell it as one. Books limit accounts for winning, and an account that keeps winning will eventually be limited regardless of how it is dressed. What the habits above buy is time, and time is the whole of it.

Capital, in more than one place. A position needs both legs funded at the same moment, which means money sitting in at least two accounts doing nothing between opportunities. A 2% margin on a $500 round trip is $10, so the interesting question is not the percentage, it is how many times a week you can turn the same capital over — and the answer, on public venues, is not many.

Speed you probably do not have. The people doing this at scale are automated. A position you find by refreshing a page has usually been visible for long enough that its best size is gone. This is the single biggest gap between what the arithmetic promises and what a person collects.

Tolerance for the leg that does not fill. It will happen. A bet gets declined, a contract disappears mid-order, a book voids a price it says was posted in error. You are then holding a one-sided bet you never wanted, at a price you only accepted as half of a pair. Deciding in advance what you do in that moment — take the loss, or hedge it back at a worse number — is more useful than any amount of screening for bigger margins.

The live board applies the freshness and identical-line rules that remove most of the fake positions before anyone sees them, and the calculator applies the venue fees the rest of this category leaves out. Neither removes the three things above.