Arbitrage betting explained: the math, the fees and the real risks

What is arbitrage betting?

Arbitrage betting is placing bets on every outcome of one proposition, at two or more venues, so the combined payout beats the combined stake whichever outcome happens. It works because each venue sets its own prices and adds its own margin, so one venue's price on a side is sometimes better than another venue's price on the opposite side by more than the two margins.

The test uses decimal odds. Convert each price to decimal (the odds converter does it), take 1/d for each outcome and add the results. One sportsbook's own two prices add up to more than 1, and the excess is its margin, called the vig. If the best price on each side, taken from different venues, adds up to less than 1, there is an arbitrage, and the locked return is 1 / sum(1/d) - 1. Use the price after any exchange fee, because an exchange pays out less than its quoted price once its fee is taken.

How does an arbitrage bet work?

You bet both sides in the right proportion, so either result pays the same amount. Take a game total with the over at +106 at Venue A and the under at +104 at Venue B. Those are decimal prices 2.06 and 2.04. The prices are made up to show the math; they are not a live quote.

The two values of 1/d add up to 0.9756, so the locked return is 1 / 0.9756 - 1 = 2.50%. Split $500 as in the table and either side returns $512.49, a profit of $12.49.

How do you calculate the stakes?

Each stake is the total times that outcome's share of the 1/d sum. That is what makes every outcome return the same amount.

1. Convert every price to decimal odds.
2. Compute 1/d for each outcome and add them. If the sum is 1 or more, there is no arbitrage.
3. Stake on each outcome = total x (1/d) / sum. For the over: $500 x 0.4854 / 0.9756 = $248.78.
4. Round to what each venue accepts, then re-check the worst case.

Rounding matters at small margins. Rounded to whole dollars the example becomes $249 and $251: the first returns $512.94, the second $512.04, so the worst case is a $12.04 profit. SlateProof's planner keeps the round-up and round-down combination with the best worst case. Exchange contracts trade in whole units, which adds another rounding step. The arbitrage calculator does this for you.

Is arbitrage betting legal?

It is as legal as the bets themselves: each bet has to be legal where you are when you place it. Arbitrage is ordinary betting at two or more venues, with no special status.

Availability depends on the state you are in, not where you live, because every venue checks your location. The state pages list which of the ten sportsbooks and exchanges SlateProof tracks are available in each state, with the date we checked. The venue's own app has the final say, and this is not legal advice. Sports contracts at exchanges are contested in several states and that can change quickly (see the prediction markets guide). SlateProof never prices offshore books or sweepstakes apps, and uses pick'em apps only for entries whose every pick is hedged at a legal venue. Faking your location, using someone else's account or opening duplicate accounts is not arbitrage: it breaks venue terms and can be fraud.

Will sportsbooks limit or ban you for arbitrage betting?

They can limit you, and many do. A sportsbook may cut your maximum stake or restrict an account it judges to be winning consistently, and prop-market stakes are often capped from the start, so a $500 total may not be accepted on the sportsbook leg. SlateProof's board warns when a leg is a sportsbook prop for that reason.

Exchanges match one trader with another, so their usual limit is depth: you can only trade what other people have resting at that price. Place the leg most likely to move or be limited first. If one leg is rejected after you place the other, you hold an ordinary bet, not an arbitrage.

How long do arbitrage opportunities last?

Usually minutes, often less, so assume you have less time than a board shows. SlateProof's own figures so far come from offline replays, not its live forward record, and the samples are small.

• A replay of 44 captured snapshots (October 5 and 6, 2026), run under an earlier rule that needed DraftKings or FanDuel on one leg, found 12 distinct arbitrages of 0.24% to 2.75% after fees, about $1 to $14 at a $500 total. Two big sportsbooks against each other almost never crossed.
• A speed test on historical prop prices valued the whole lane of arbitrages that involved DraftKings or FanDuel at about $309 a day if each were placed instantly, about $117 a day at five minutes and about $7 at 30 minutes, before tax.
• Arbitrages with every leg at an exchange were still there five minutes later only about 27% of the time in a separate offline test.

These are dated internal tests, not results. SlateProof now freezes each arbitrage when it first appears and logs whether it is still there at every later snapshot, so the track record measures how often one survives long enough to place. Snapshots are minutes apart, so the board shows you where to look, not what will still be there.

How do exchange and prediction-market fees change the math?

They shrink it, sometimes to nothing. An exchange charges a fee that depends on the contract price, so convert each exchange price to an effective decimal price before you test for arbitrage.

Fees checked October 6, 2026. Polymarket US: its fee page (banker's rounding, effective October 1, 2026). Novig: its fee page. Kalshi: the 7% rate is the one SlateProof models, from third-party summaries; Kalshi's own fee schedule is the authority and some series carry multipliers. ProphetX: the 2% rate is from third-party summaries.

Here is the effect. A sportsbook offers one team at +110 (decimal 2.10) and a Kalshi contract on the other side costs 50 cents (decimal 2.00 before the fee). Before the fee the gap is 2.44%. Kalshi's fee at 50 cents is $1.75 per 100 contracts, so the effective price is 1.9324 and the gap falls to 0.63%. With a $500 budget the plan is $240 at the sportsbook and $261 on Kalshi ($501 in all), with a worst-case profit of $3.00. SlateProof asks for at least 0.2% after fees, or 0.5% when a leg is at an exchange, because whole contracts, fee rounding and price rounding eat the rest.

What can go wrong with an arbitrage bet?

Most failures come from one leg not behaving like the other. A locked profit is a promise only if every leg is accepted, kept and settled the same way.

• A leg is rejected or the price moves. If only one leg fills you hold a plain bet. Place the leg that is most likely to move first.
• A stale or erroneous line is voided. A price far better than other books' prices on the same outcome is more often a mistake than a gift, and a venue can cancel a bet at an obvious error while the other leg stands. SlateProof flags a leg roughly 4 to 5 points better than the market (at even money) and blocks one roughly 15 points better.
• A player does not play. Sportsbooks void a player prop when the player does not play. Exchanges do not refund: Kalshi settles at the last fair price and others may settle differently, so an exchange-only prop arbitrage can lose on both legs.
• Settlement rules differ. A postponed game that is not replayed in time settles at a last fair price on an exchange, not as a refund. A period or 90-minute scope can differ between venues. An NFL tie is a push at a sportsbook, while Polymarket US settles a no-ties winner contract at 50 cents if the game ties. A whole-number line is a push at sportsbooks and may not be refunded at an exchange. SlateProof blocks the combinations it cannot reconcile and discloses the rest.
• Exchange depth is thin. Resting size caps the total stake. A quote whose two sides add up to 100% is a midpoint, not a price you can buy, and SlateProof drops it.

Does tax change whether arbitrage pays?

Often, yes. Gambling winnings are taxable income, losses are deductible only if you itemize, and from the 2026 tax year only 90% of wagering losses count (IRS Topic 419; 26 U.S.C. 165(d)). An arbitrage always has a winner and a loser, so you can owe tax on profit you never made.

Here is the $500 example at a 24% federal rate, with whole-dollar stakes of $249 and $251. Before tax the profit is $12.94 if the first bet wins and $12.04 if the second wins.

Add up a year of arbitrages and the same pattern holds. In an offline test at a 24% bracket, SlateProof found that under the 2026 rule an itemizer loses money after tax on arbitrages below about 1.5%. This is an illustration, not tax advice: state tax, your bracket and your records differ, so ask a tax professional.

Arbitrage, +EV or promo hedging: which fits?

Promo hedging has shown the most measurable value in SlateProof's tests, arbitrage margins are small, and +EV betting is a long-run edge rather than a locked one.

Arbitrage margins are small (SlateProof has seen a median of about 1%), so $500 per arbitrage earns a few dollars each, and books limit accounts that arbitrage consistently.

Where can I see live arbitrage?

The live dashboard lists current arbitrages across the legal US venues, with exact stakes, the accounts each one needs and a warning for every risk above. It never places a bet for you, and a locked profit holds only if every leg is accepted at the shown price.

Keep going

Put your own prices into the arbitrage calculator or the hedge calculator. Read how fair value is built in the expected value guide, how exchange contracts differ in the prediction markets guide, and compare the exchanges and states. The methodology page explains how SlateProof vets an arbitrage.