The Math of Arbitrage Betting and Why Your Favorite Book Closes Your Account

Arbitrage betting is the one strategy in sports betting that the math actually supports without any edge estimate or probability judgment on your part. You find two books with different opinions on the same event, size your stakes correctly, and you lock in a guaranteed profit. No handicapping needed.

The catch: the account that lets you execute this strategy has a lifespan measured in bets, not years. Soft books have spent the last decade building risk systems to identify you faster. Massachusetts regulators made news in 2024 by becoming the first US state to require sportsbooks to tell you why they limited you. The reason matters. Understanding what triggers the limit is how you extend your runway, or decide the whole enterprise isn't worth it.

This article covers the mechanics of the arb, the worked math, the detection signals sportsbooks actually use, and what the regulatory record says about where this is going.

What Creates an Arbitrage Opportunity

Every sportsbook builds a margin into its lines. At -110/-110 on a standard point-spread market, the two implied probabilities sum to 104.76%. That 4.76% overround is the vig, the book's guaranteed take if action balances perfectly. In practice, books shade lines based on public betting tendencies, but the overround is always there.

An arb opportunity exists when you pull the best available price on each outcome from different books and the implied probabilities sum to less than 100%. That gap below 100% is your margin.

The formula:

• Convert each American price to decimal. For favorites: decimal = (100 ÷ |American odds|) + 1. For underdogs: decimal = (American odds ÷ 100) + 1.
• Implied probability of each outcome = 1 ÷ decimal odds.
• If IP₁ + IP₂ < 1.00, you have an arb. The gap is your margin.

Worked Example: A Two-Way Arb

Say you find this setup across two books:

Sum is 98.89%, below 100%. Arb margin: 1.11%.

Now size the stakes. With $1,000 total to deploy:

• Stake on Team A (DraftKings): $1,000 × 0.5238 ÷ 0.9889 = $529.68
• Stake on Team B (FanDuel): $1,000 × 0.4651 ÷ 0.9889 = $470.32

Results:

• If Team A wins: $529.68 × 1.909 = $1,011.17 returned. Net profit: $11.17.
• If Team B wins: $470.32 × 2.150 = $1,011.19 returned. Net profit: $11.19.

Guaranteed $11.18 on $1,000 deployed, 1.11%, regardless of which team wins. The math delivers exactly what it promises.

Typical arb margins run 1-3% for pre-match opportunities. Live betting arbs can hit 3-5% briefly but vanish in seconds. Most pre-match windows close in 30 seconds to 5 minutes as books update toward the consensus price.

How Often Do Arbs Appear

The empirical answer comes from a 2013 study in Economica by Franck, Verbeek, and Nüesch, who analyzed inter-market arbitrage across the top five European soccer leagues. Their finding: a combined bet at a bookmaker and a betting exchange yielded a guaranteed positive return in 19.2% of matches. Bookmakers, on average, experienced negative margins from the posts that created those arbs. They were giving money away on those specific lines and they knew it. The book's loss on the arb post is a cost of doing business, recouped from the recreational bettor majority on every other line.

That 19.2% figure captures a broader market (UK, which has accessible exchanges). In the US market, where true exchanges are limited, the number is lower. Arbs between soft books (DraftKings, FanDuel, BetMGM, Caesars) and sharper references do appear regularly, particularly in the minutes after Pinnacle moves a line and before the retail books catch up.

Why Soft Books Create the Gap

The structural reason arbs exist at US retail books comes down to how they set lines. DraftKings and FanDuel do not operate as market makers. They source opening lines from third-party feeds, then shade those prices toward the public, adding extra margin on the popular side. This approach maximizes hold from recreational bettors but creates pricing inconsistencies across books.

Pinnacle, operating with margins around 2% versus 4-6% at retail books, functions as the sharpest reference price in the market. Its closing lines show a correlation of 0.997 (r-squared) with realized outcome probabilities across nearly 400,000 European football matches, as close to a true probability estimate as the betting market produces. When Pinnacle moves, other books lag. That lag is the arb window.

The irony: soft books benefit from this system. They let arbers find the stale line, then update it. The arber sharpens the book's price for free, then gets limited. The book got a free line correction. The arber got 50-200 profitable bets before the account was capped.

The Detection Playbook

Modern sportsbook risk systems do not wait for you to be obviously profitable over hundreds of bets. They score every account after every wager using signals that are more specific than your win rate alone.

Precise stake amounts. An arb calculator spits out stakes like $473.82. That number, precise to the cent, is a fingerprint. No recreational bettor bets $473.82. Risk engines flag two-decimal precision immediately. When the same account places five bets in a week with odd-cent amounts, the system escalates for human review.

Bet timing vs line movement. The window between Pinnacle moving and DraftKings updating typically runs 3-30 seconds. If your account consistently places bets within seconds of a line move at another book, that pattern is logged. Timing correlation across dozens of bets is statistically identifiable long before you hit any win-rate threshold.

Consistent closing-line value. Sharp bettors get better odds than the closing line more often than chance. The risk engine measures your CLV on every bet. An account that consistently closes above market is profiled as sharp, regardless of realized win rate. CLV predicts long-term profitability better than short-term results.

Market selection. Arbers concentrate in specific markets where gaps appear: full-game spreads and totals at launch, not halftime props. A pattern of betting the same market type across multiple books triggers a flag.

Dynamic risk scoring. Every licensed US sportsbook uses a risk engine that updates each account's score after every bet. The score aggregates historical P&L, bet timing, stake precision, market concentration, and promo usage. When the score crosses a threshold, limits are applied automatically, often before a human reviews the account. By 2026, ML systems were flagging sharp accounts roughly 100 bets into their lifetime, far earlier than the "50-bet limit" narrative from earlier years.

What the Limits Look Like

The account does not get closed immediately. The process runs in stages:

1. Stake factoring. Maximum bet on flagged markets drops, from a $500 limit to $20 or $50 without notice. You find out when you try to place a bet and the book accepts a fraction of your requested amount.
2. Market restrictions. Certain market types get locked while others remain open. A book limits your spread bets but still accepts futures. This extends the account's utility while minimizing the book's risk on your sharpest markets.
3. Account closure. Full closure is the final step, often reserved for accounts that continue probing after stake factoring.

DraftKings' terms reserve the right to "limit or refuse any wager, restrict your betting activity, or close your account at our sole discretion." That language is standard across every US retail book. Terms reference prohibited "syndicate betting," coordinated action, and cite "structured wagering" and "live market latency exploitation" as specific grounds for limiting.

The Regulatory Record

On February 26, 2024, the Massachusetts Gaming Commission voted 5-0 to adopt regulation 205 CMR 238.30. Effective June 1, 2024, every licensed sportsbook in Massachusetts must notify bettors within 48 hours of any account limitation. The notice must explain the reason for the restriction and specify which markets are affected. Accounts limited before June 1 and accounts where limits carry over from other states are also covered.

Massachusetts is the first US state to require this transparency. The regulation followed a rocky public process: the MGC called a roundtable in May 2024 to discuss limiting practices in an open meeting, and no sportsbook sent a representative. A second roundtable in September 2024 brought all licensed Massachusetts books to the table, where they defended the practice.

The practical effect: bettors in Massachusetts now know which markets triggered their limit and can identify the specific signals that caused the restriction. This helps arbers understand what the book caught. It does not stop the limiting itself.

Other states are watching. The MGC's approach has been cited as a model by regulators in at least three other jurisdictions. Whether this becomes national policy depends partly on how much political pressure recreational bettors, the constituency regulators actually respond to, put on legislators. Arbers are a small and specialized group. The policy momentum is real but slow.

Where the Math Still Works

Once you're limited at soft books, three places remain where the arb model survives without account attrition:

Pinnacle. The book explicitly welcomes arbitrage and has since 1998. Its "Winners Welcome" policy is a business decision. Sharp action sharpens Pinnacle's own lines, reducing the book's exposure on stale prices. You will never get limited at Pinnacle for winning. What you give up: margins of about 2% versus the retail book's 4-6%, which narrows arb windows but doesn't eliminate them. Pinnacle operates outside most US states.

Betting exchanges. Betfair charges a standard 5% commission on net profit from winning bets. Smarkets charges 2%. Matchbook charges 1-2%. On an exchange, you bet peer-to-peer. There is no bookmaker to limit you. The exchange takes its commission regardless of outcome. The model survives indefinitely because the exchange profits from volume, not from betting against you. Betfair is not licensed in most US states. Smarkets operates in limited US markets.

US peer-to-peer platforms. Novig operates as a peer-to-peer sports betting exchange in the US, charging a 1-2% commission on winning bets. No vig baked into lines. No account limitations for winning. Kalshi and Polymarket offer prediction market contracts on sports outcomes, also with no bookmaker to restrict you. Liquidity is thinner than traditional books, which limits stake size on any single bet.

The trade-off at all three: lower margins mean fewer arb windows versus soft books. An arber using Pinnacle or an exchange as the "lay" side of a two-way arb against a US retail book still faces the same account attrition at the retail book. The exchange leg is unlimited. The retail book leg has a clock.

The Honest Assessment

Arbitrage betting is a real edge, not a system sold by someone who profits from your tuition. The math is verifiable, the profits are guaranteed, and the academic literature confirms the opportunities exist. Franck et al. found arbs in nearly one in five soccer matches across the top European leagues.

What the edge does not promise: longevity at any specific soft book. Your account at DraftKings or FanDuel is a depleting asset. Each arb bet extracts a small guaranteed return and moves you closer to the limit. Risk scoring catches this faster now than it did five years ago. The 2026 reality is that a disciplined arber with careful execution might sustain 100-200 profitable bets at a soft book before hard limits drop.

The business of full-time arbitrage requires constant new account openings, careful stake sizing, variation in markets, and timing, all to extend the productive life of each account. That operational friction is the real cost the math does not capture.

For a bettor with multiple accounts at different books and access to a reliable odds-comparison tool, selective arbing, taking the opportunities without the fingerprints of a full-time arber, extends account life. Avoid precise cent-level stakes. Vary your timing. Do not hit the same market type repeatedly. No single tactic guarantees indefinitely unlimited action, but each extends the clock.

The structural edge is real. The runway is finite. Knowing both is how you size the investment of time and capital correctly.

Sources

• Franck, E., Verbeek, E., & Nüesch, S. (2013). Inter-market Arbitrage in Betting. Economica, 80(318), 300-325. doi:10.1111/ecca.12009
• Franck, Verbeek, Nüesch. SSRN preprint. SSRN 1503375
• Massachusetts Gaming Commission. Regulation 205 CMR 238.30 (Feb 26, 2024). Via Gaming America
• Massachusetts sportsbook limit notice regulation, effective June 1. Via Sports Betting Dime
• BettingUSA. Vig rates: FanDuel 13.68%, DraftKings 13.99% (June 2026). bettingusa.com
• Rebel Betting. Pinnacle margins ~2%, arbitrage welcome since 1998. rebelbetting.com
• Shark Betting. Exchange commission rates: Betfair 5%, Smarkets 2%, Matchbook 1-2%. sharkbetting.com
• BettingUSA. Novig 1-2% commission, peer-to-peer exchange model. bettingusa.com/novig
• DeucesCracked. ML risk scoring catches winners earlier. Dynamic score updates after every bet. deucescracked.com
• SportsBettingDime. Precise stakes flag arbers. Timing patterns tracked. sportsbettingdime.com