The trade, in one paragraph
Arbitrage in sports betting is the same idea that drives currency triangulation in foreign-exchange desks and convertible-bond hedging in equity derivatives: two markets price the same underlying at prices that, summed correctly, leave a sliver of guaranteed profit between them. The arber backs every possible outcome at the right book, in the right size, and locks the profit before the event begins. Each individual arb is small — typically 1-4% of total stake — but the trade is mathematically risk-free in a way that almost nothing else in retail betting is.
Arbitrage is also the most surveilled activity in the modern sportsbook. Books detect arbers within weeks through deposit patterns, bet-selection profiles, and the specific signature of an arber's stakes hitting mispriced lines minutes before they move. The arber's edge is therefore not just mathematical; it is operational — building a multi-book account profile, varying stake patterns, mixing arbs with 'mug' bets that look recreational, and rotating between operators before any single book limits the account into oblivion.
The math of the locked profit
The mechanics are arithmetic, not algebra. Convert every American odds price to implied probability. Sum the implied probabilities across all possible outcomes. If the sum is less than 100%, an arb exists. The locked profit margin is (1 / sum) - 1, expressed as a percentage of total stake.
A concrete example. Book A has the Lakers at -130 (56.5% implied). Book B has the Celtics at +145 (40.8% implied). Sum: 97.3%. The arb margin is (1 / 0.973) - 1 = 2.77%. To lock the profit, the bettor sizes each side by inverse probability: 56.5% / 97.3% = 58.1% of total stake on the Lakers, and 41.9% on the Celtics. Place $581 at -130 (potential return: $1,028.85) and $419 at +145 (potential return: $1,026.55). Either side returning roughly $1,027 on $1,000 staked equals a locked $27 profit, regardless of which team wins. The two payouts are not exactly equal because the conversion rounds; a sharp arber accepts a few cents of imbalance to size in whole dollars.
The stake-sizing formula
The general formula for arbitrage stake sizing is straightforward. For each outcome i with decimal odds d_i, the stake on that outcome as a fraction of total bankroll-for-this-arb is: s_i = (1 / d_i) / Σ(1 / d_j), where the sum is across all possible outcomes. The total return on any winning outcome is then s_i × d_i, which by construction is equal across all outcomes when properly sized.
Translating American odds to decimal: a positive American odd a becomes (a/100) + 1, so +145 = 2.45. A negative American odd a becomes (100/|a|) + 1, so -130 = 1.769. Plug into the formula: 1/2.45 = 0.408; 1/1.769 = 0.565; sum = 0.973. Stake fractions: 0.408/0.973 = 41.9% on the +145, 0.565/0.973 = 58.1% on the -130. Profit margin: (1/0.973) - 1 = 2.77%. Every arbitrage calculator in the world implements exactly this math; the spreadsheets are public and the formula has been the same since Kelly's 1956 paper laid out the inverse-probability betting framework that arb sizing inherits.
Finding arbs — tools and tradecraft
The 2024-2025 US legal sports-betting market has 8-15 active operators per state, each with independent pricing engines, model latency, and promotional offers. Cross-book price gaps are routine. Tools like OddsJam, RebelBetting, BetBurger, and Crazy Ninja Odds scrape live prices across 30-60 US and international books and surface arbs in a continuous feed. Subscription tiers range from $50 to $300 per month depending on data refresh rate, included markets, and number of supported books.
Subscriptions cost money, but the bigger constraint is execution. An arber sees an alert for a 3% arb on a niche MLB total. They have approximately 2-8 minutes before one of the two books recalibrates and the arb closes. In those minutes they must: confirm the price still exists on both books, size both sides correctly, place both bets without one side being voided for 'palpable error,' and confirm both wagers settle as accepted before treating the position as locked. A bettor who is slow, distracted, or attempting to size by hand will routinely place the first leg and watch the second leg move against them — leaving directional exposure rather than a locked arb. Sharp arbers use multi-book betslip pre-staging, two monitors, and (in advanced cases) browser automation that breaks operator terms of service.
Account longevity — the arber's hardest problem
The single biggest constraint on arbitrage is not finding arbs but staying alive long enough to keep placing them. Sportsbooks profile bettors aggressively. Within 4-12 weeks of identification, a flagged arb account will see max wagers cut from $5,000-$10,000 down to $20-$100 — making the account effectively useless for further arbing. The detection signals include: large deposits followed by quick withdrawals, exclusive betting on lines that move within 60 seconds of the bet placement, no parlays or props (recreational signal), bet sizes that map to inverse-probability stake fractions across books, and steam-following on cross-book line movement.
The defensive playbook is well-documented in the arber community. Mix mug bets: place a handful of recreational-looking parlays and SGPs for small stakes each week. Vary stake patterns: do not always bet exactly $487.32 — round to $500, sometimes $450, sometimes $525. Use multiple books: spread action across 8-20 operators so no single book sees enough volume to flag the pattern. Withdraw less aggressively: pull money to bank only after building a deposit history. Bet some games to lose: take recreational angles on Sunday-night football, Monday-night basketball, anything that looks like a fan bet. The objective is to look like a casual bettor who occasionally wins, not a quantitative trader.
Three-way arbitrage in 1X2 markets
Soccer's three-way moneyline market (home win / draw / away win) creates a richer arbitrage surface than two-way US markets. An arber needs prices on all three outcomes from up to three different books, summed to less than 100%. Because three-way pricing has more degrees of freedom and softer books often misprice the draw probability, 3-5% three-way arbs are more common in soccer than two-way arbs are in US sports. The math is identical — sum the implied probabilities, size by inverse probability — but the execution requires three simultaneous bets, multiplying the risk that one leg is voided before all three settle. Three-way arbs are the bread and butter of European arbitrage operations; in the US they are a smaller part of the picture because soccer handle is lower.
Middling vs. arbitrage — sibling strategies
A close cousin to arbitrage is middling, which exploits the same cross-book price differentials in a different shape. An arb backs both sides at prices that overlap and lock a small profit. A middle backs both sides at non-overlapping spreads or totals, locking a small loss on most outcomes and a large profit if the final result lands inside the 'middle' between the two lines. Take Team A -3 at Book A and Team B +5 at Book B. If the favorite wins by 4, both bets win — a huge payout. If the favorite wins by any other margin, one bet wins and the other loses, losing only the small juice.
The expected value of a middle depends on the probability the result lands in the middle window. A 2-point middle window in NFL is roughly 7-9% probable; the middle pays roughly 19x the per-leg vig loss when it hits, giving the strategy a positive expected value if the window is wide enough. Middles are higher-variance than arbs, lower-frequency, and have the same operational challenges (limit risk, void risk, cross-book execution). Sharp bettors run both strategies in parallel because they exploit the same market inefficiencies through different shapes.
Bonus-driven arbitrage and promotional EV
The single highest-EV form of arbitrage in the modern US market is bonus-driven arbing. New US states going live trigger massive operator promotional spend: $1,000 deposit matches, $200 'bet $5 win $200' offers, risk-free first bets, profit boosts on specific markets. These promos shift the EV math because the bettor's effective price on the boosted leg is higher than the posted odds — meaning an apparent 'losing' two-sided bet can be locked at a positive EV when the bonus is included.
The classic example: a 'bet $5 win $200' first-bet bonus. The bettor places $5 on Team A at -120 at the new book (the boosted leg) and $5 on Team B at +110 at an established book (the natural hedge). If Team A wins, the bettor pockets the $200 bonus payout. If Team B wins, the bettor collects $10.50 from the +110 hedge and loses the $5 on Team A. The net expected value, after weighting both outcomes by implied probability, is positive — often in the $20-80 range per promo. Multiply across 8-15 books going live in a new state and a careful bonus arber can clear $1,000-5,000 in promo EV over the first weeks of state legalisation. The strategy is well-documented and lawful; the books accept it as the cost of customer acquisition.
Exchange arbitrage — lay vs. back
Betting exchanges (Betfair, Smarkets, and increasingly Sporttrade and Prophet in the US) let users back an outcome (bet for it to happen) or lay an outcome (bet against it, taking the other side of someone else's back). This creates a clean arbitrage surface: a sportsbook may price a moneyline at -130 (56.5% implied) while a Betfair exchange may have a lay price at 1.74 decimal (57.5% implied). Backing the underdog at the sportsbook and laying the favorite on the exchange — or equivalently, backing on the exchange and using the sportsbook as the hedge leg — locks a small guaranteed profit.
Exchange arbs have two advantages over sportsbook-to-sportsbook arbs: exchanges do not limit profitable customers (they take commission on net winnings instead of running a sportsbook against the bettor), and exchange liquidity is concentrated, making execution faster than chasing prices across 8-15 retail books. The trade-off is liquidity depth — exchanges in early US markets often have shallow order books, meaning large stakes move the price and reduce the arb margin. The arber's playbook in 2025 is increasingly to use exchanges as the second leg of the trade whenever possible, because exchange action does not trigger limit reviews.
The regulatory and operational risk landscape
Arbitrage is legal in every US state that has legalised sports betting; no state regulator prohibits cross-book hedging. Operators, however, retain the right under their terms of service to limit, restrict, or close any account at their discretion, and 'persistent inverse-stake-sizing across books' is grounds for closure at every major US sportsbook. Bettors should expect their arbitrage accounts to have finite lifespans — typically 6-18 months at any given book — and plan rotation accordingly.
The two operational risks that matter most: void risk (one leg of an arb is cancelled by an operator, leaving directional exposure) and limit risk (the bet is accepted but at a reduced max stake, throwing off the planned sizing). Sharp arbers manage both by pre-checking limit profiles on each book monthly, avoiding arbs on niche or volatile markets, and keeping per-arb stakes below the smallest current limit across both legs.
Practical checklist for arbitrage bettors
- Use a calculator, every time. The inverse-probability sizing formula is unforgiving; a single mis-sized arb erases ten correctly-sized ones.
- Pre-stage betslips across both books before confirming either bet. Execution speed is the difference between locked profit and directional exposure.
- Mix mug bets to look recreational. Limits come faster than most arbers expect.
- Read void rules at every book. A voided leg turns a locked arb into a directional loss.
- Rotate books. Plan on 8-20 active accounts and assume each will be limited within 6-18 months.
- Track every arb. The 1-4% return per ticket only compounds into a livable income if the volume and accuracy are real.
Sources & further reading
- Kelly, John L. "A New Interpretation of Information Rate." Bell System Technical Journal, 1956 — the foundational paper on inverse-probability bet sizing that underlies arbitrage stake calculation.
- Thorp, Edward O. "The Kelly Capital Growth Investment Criterion." Handbook of Asset and Liability Management, 2008 — extension of Kelly to multi-outcome and arbitrage contexts.
- Shin, Hyun Song. "Prices of State Contingent Claims with Insider Traders, and the Favourite-Longshot Bias." Economic Journal, 1992 — the model of bookmaker margin that arbs implicitly exploit.
- Snowberg, Erik & Wolfers, Justin. "Explaining the Favorite-Longshot Bias." Journal of Political Economy, 2010 — cross-book pricing inefficiencies that produce arbitrage opportunities.
- Pinnacle Betting Resources — public documentation on arbitrage calculation, sure-bet sizing, and the structure of three-way and exchange arbs.