What is arbitrage betting?

Take an NFL total of 44.5. One sportsbook has the over at +110. Another has the under at -102. Every price carries an implied probability, which is one divided by the decimal odds. Add the two sides together:

The two prices imply 98.11%, which is less than 100%. That gap is the edge. To lock it in, split the money in proportion to each side's implied probability: with $1,000, that is $485.34 on the over and $514.66 on the under. Either bet then returns $1,019.22, a profit of $19.22 whatever the final score.

Nobody bets $485.34, so here are the same bets rounded to whole dollars:

About $19 either way, a return of roughly 1.9% on the money staked. It is not much on one game. Arbitrage bettors make it work by repeating it many times, and the rest of this guide is about why that is harder than the arithmetic. The arbitrage calculator does this split for two or three legs and shows what rounding costs.

Every sportsbook sets its own prices, and every book builds a margin into them. A market at -110 on both sides implies 52.38% twice, or 104.76% in total. The extra 4.76 points is how the book gets paid: on a balanced market it keeps about 4.5% of the money wagered. That margin is called the hold.

Because each book's own two sides add up to well over 100%, an arb needs two books to disagree by more than their combined margins. That happens when books react to news at different speeds, when one book shades a price toward the side its customers like, and when one book moves on sharp money and another has not caught up yet. Some markets are also corrected more slowly than others: player props, priced across hundreds of players, stay out of line much longer than moneylines, as the measured lifetimes below show.

The book with the out-of-line price is usually the one that moves, and it moves as soon as it notices. Books watch each other and watch where money goes, so a gap between them closes fast.

ArbWins keeps a public track record of every edge it found and how long both prices held. Measured on October 14, 2026, over the first day and a half of data, the median edge stayed open about 11 minutes. Moneylines closed fastest, at a median nearer 1 minute. Player props lasted longest, nearer 25 minutes. That is a short sample, and the live numbers on the track record page will move as data comes in.

Two habits help. Bet the price that is out of line with the rest of the market first, because it is the one most likely to move. And check both prices again right before you stake the second leg. If the second price is gone, you are holding a single ordinary bet, and you have to decide whether to keep it or hedge it at a worse price.

The two books settle the bet differently. An arb only works if both books grade the market the same way. A hockey line that pays on regulation time only, paired with a full-game line that includes overtime, can lose on both sides. Every book ArbWins reads voids a player prop when the player does not take part, but the test differs: dressed, one snap, one minute played. In baseball, listed-pitcher rules vary by book. If one leg is voided and the other loses, you lose a full stake. The settlement rules table shows how each book ArbWins reads grades each market.

Your accounts get limited. Sportsbooks limit accounts that win, and arbitrage bettors win. Books notice calculator-exact stakes, bets placed seconds after a line moves, and accounts that only bet prices that are out of line. Expect it to happen at some of your books eventually. Nothing prevents it. Rounding stakes to amounts other bettors use makes your bets look ordinary, at a small cost in edge.

A bet gets voided. Sportsbooks reserve the right to cancel bets taken at an obvious pricing error, known as a palpable error or palp. The biggest edges are the most likely to be errors. If the error leg is voided after the other leg has lost, there was never an arb. ArbWins marks unusually large edges as likely errors that may be voided.

Rounding and minimums eat a thin edge. Prices of +105 and -103 add up to 99.52%, an edge of 0.48%. On $100, stakes of $49.02 and $50.98 make $0.48 whichever side wins. Round both stakes to $50 and the over pays $102.50, but the under pays only $98.54: a $1.46 loss if the under wins. Maximum bets cause the same problem from the other end: if one book caps a leg, the other leg has to shrink with it.

Yes. Placing bets at licensed sportsbooks in a state where you are allowed to bet is legal, and no law stops you from betting both sides of a game at different books. It is legal, but books do not welcome it: their terms let them limit or close an account at their discretion, and they use that against bettors who win consistently.

Arizona, like most states with legal sports betting, requires you to be 21 or older, and so does ArbWins. ArbWins reads prices the books publish and links you to their betslips. It never places a bet or holds your money.

1. Open accounts at several licensed books in your state. Arbs come from disagreement, and more books means more disagreement.
2. Fund each one. Both legs have to be placed within minutes, so the money has to be sitting at each book before the arb appears. Only use money you can afford to have tied up, and to lose if a leg goes wrong.
3. Do the math before you bet. Run each pair through the arbitrage calculator with the stakes you will actually type, and check the profit after rounding.
4. Check the settlement rules for the sport and market on the rules table before you place the first leg.
5. Keep a record of every bet at every book: stake, price, result. It is the only way to know whether you are actually ahead.

A related approach that bets one side at a time is covered in what +EV betting is, and a bet that wins both sides on one result is explained in what a middle is.