The math — why arbs exist
Different sportsbooks have different probability models. They process the same information (statistics, injuries, public action) but reach slightly different conclusions and price accordingly. Normally, vig keeps the combined two-side market above 100% even with disagreement. But when two books disagree significantly — one bullish on the favorite, another bullish on the dog — the disagreement can exceed vig, creating an arb.
The mechanics: if the implied probabilities of opposing sides sum to less than 100%, you can bet both sides in proportions that guarantee a fixed profit. The catch is that this state is unstable — books detect the discrepancy and adjust quickly. Arbs live in narrow time windows.
# Arb detection DraftKings: Lakers +145 → decimal 2.45 → implied 40.82% FanDuel: Celtics -135 → decimal 1.741 → implied 57.45% sum = 40.82 + 57.45 = 98.27% # < 100%, arb exists arb_margin = 100% − 98.27% = 1.73%
Stake sizing — the inverse-probability rule
To lock equal profit on both outcomes, stake inversely to implied probability.
# Total stake budget total = $10,000 # Inverse-probability allocation stake_LAL = 10000 × (0.4082/0.9827) = $4,154 stake_BOS = 10000 × (0.5745/0.9827) = $5,846 # Payout on either outcome LAL wins: 4154 × 2.45 = $10,177 BOS wins: 5846 × 1.741 = $10,177 guaranteed profit = $177 ≈ 1.77% on stake
Round to whole dollars before placing; rounding error is typically < 0.1% and doesn't materially affect locked profit.
Where arbs come from — the three sources
1. Book-to-book disagreement
Different books, different models, different opinion. Most common during weekday afternoon when liquidity is thin and lines aren't yet calibrated by sharp money. Window: 30 seconds to 5 minutes. Margins: 0.5-2.5%.
2. Exchange vs. book disagreement
Betfair/Smarkets exchange prices (set by individual layers) versus book-set prices. Books and exchanges have different probability views; exchange prices reflect real money put up by counterparties, often sharper than books on niche markets. Margins: 1-4%. Windows can last longer (10-30 minutes) because exchanges don't auto-adjust to books.
3. Bonus-arb (promotional arb)
A book offers risk-free bet or odds boost; you arb against another book. Margins: 5-15% per bonus, but capital intensive and one-shot per account. Volume comes from rolling through many promotions.
Arb tools — the infrastructure
Modern arb operations rely on commercial scanning tools that monitor 50+ books in real time and surface arbs by margin, market, and book pair.
| Tool | Coverage | Monthly cost | Notable features |
|---|---|---|---|
| OddsJam | 40+ US books | $199-$499 | Strong US focus, real-time alerts, +EV mode included |
| RebelBetting | 100+ global books | $199-$349 | European/Asian markets, exchange integration |
| BetBurger | 200+ books | $179-$269 | Highest book coverage, surebets and value bets |
| ArbHunter | 80+ books | $129-$249 | Mid-range coverage, cheaper entry |
| Crazy Ninja Odds | 50+ books | $99-$199 | US emphasis, lower-cost option |
Tools alone don't produce profit. The arber must have: accounts at multiple books (typically 10-30); capital allocated across them; banking infrastructure for fast funds movement; and execution discipline to place both legs in under 30 seconds.
Account longevity — the central problem
Books detect arbing patterns reliably within 50-150 bets. Signals: ① small precise stake sizes (round to nearest dollar in inverse-probability ratios); ② bet sequencing across books within 30 seconds; ③ preference for low-vig markets and certain bet types; ④ withdrawal patterns matching arb profit cycles; ⑤ never betting parlays or single-book systems retail bettors favor.
Once detected, accounts get: ① stake-limited (max bet drops from $5K to $50-$200); ② withdrawal-restricted (slower processing, additional verification); ③ closed outright with no warning. Some books (DraftKings, FanDuel) have publicly disclosed that they limit arbers; others (BetMGM, Caesars) do so quietly.
Operational tradecraft
Career arbers deploy multiple tactics to extend account life:
- Stake variation — round to non-obvious amounts ($87 instead of $85.30); include occasional flat-stake bets to blend.
- Account aging — bet recreationally for 30-90 days before deploying arb action; build a "casual bettor" profile.
- Loss-leader bets — periodically place obvious -EV bets (parlays, longshots) to maintain plausible recreational pattern.
- Multi-name strategy — family members (legal in most jurisdictions) hold separate accounts; arber executes through them.
- Geographic spread — operate accounts across multiple states (US) or countries (EU) to access more book ecosystems.
- Exchange-heavy mix — Betfair/Smarkets don't limit arbers because exchange economics work on commission, not vig. Some arbers do 60-80% of volume on exchanges to extend retail account life.
Risk-free in theory, friction in practice
The arb edge looks clean on paper. In execution, frictions accumulate:
- Line movement — the +145 you saw at DraftKings may move to +135 in the 8 seconds it takes to navigate to FanDuel.
- Max bet caps — DraftKings allows $3K max on your account; you can only place 60% of the desired stake on the LAL side.
- Connection issues — slow Wi-Fi mid-bet, app crash, two-factor delay.
- Bet rejection — book flags the bet for review, returns "stake adjustment required" error after 90 seconds (often too late to arb the other side).
- Banking delays — capital tied up in book #1 can't fund book #2 immediately. Most arbers maintain 1.5-2x notional capital across all books to avoid this.
- Currency conversion (international arbing) — FX fees of 1-3% can eat smaller arb margins entirely.
Net effect: a 1.5% theoretical arb margin often becomes 0.7-1.2% realized after friction. Larger arbs (3%+) are robust to friction; smaller arbs (<0.8%) are usually not worth executing.
Bonus arbs — the highest-margin sub-strategy
Sportsbook promotions create predictable, high-margin arb opportunities for first-time deposits. Standard offers:
- Risk-free first bet — bet up to $1,000; if it loses, get your stake back in bonus credit. Arb: place the bet on a heavy favorite at competitor book; if you lose the favorite, you have $1,000 in bonus credit to redeploy. Expected value: $400-$600 net depending on bonus conversion rate.
- Deposit match — book matches deposit up to $500-$2,500 with rollover requirement (5-15x). Arb: optimize through low-vig markets to clear rollover with minimal expected loss. Realized EV: 40-70% of bonus amount.
- Odds boost — book offers Lakers +200 instead of standard +150. Arb the boosted side against the unboosted Celtics at competitor. Realized EV: $20-$50 per boost.
A bonus arber rolling through 8-15 books in their state captures $3,000-$8,000 in promotional EV over 2-4 months before exhausting first-time bonuses. Some pros maintain spouse/family identities to extend.
Middle — the high-variance cousin
A middle is two bets on the same side at different point spreads, hoping the result lands between them. Not risk-free, but very high upside when middled.
# NFL spread middle DraftKings: Lakers +7 (-110) stake $1,100 to win $1,000 FanDuel: Lakers -3 (-110) stake $1,100 to win $1,000 # Outcome: Lakers lose by 1-6 → both bets win → profit = $2,000 # Outcome: Lakers lose by 7+ → bet 1 pushes/loses → -$110 # Outcome: Lakers win → bet 2 wins → -$110 # Expected hit rate on a 10-point middle ~ 5-7% # EV positive if hit rate > 5.5% (covers vig on losing weeks)
Sharp middlers know which markets have historical landing-zone clustering — NFL key numbers (3, 7, 10) and NBA totals near 220-225 are favorites. Middle strategy is +EV but high-variance; not for the faint of heart or thin bankrolls.
The arb ecosystem — who participates
Three player types:
- Mass-market arbers — 5,000-50,000 globally, capital $5K-$50K, using scanning tools to grab 1-2% arbs. High volume, low margin. Account longevity 2-6 months.
- Pro-tier arbers — 500-2,000 globally, capital $100K-$1M, sophisticated multi-account ops, blend arbs with +EV and bonus farming. Long-term operations.
- Market-making arbers — institutional firms (often quant fund spinouts) that arbitrage between exchanges, market-make on Betfair, and capture commission spreads. Capital $5M-$50M+, employees count.
Sources & further reading
- Hausch, Donald B., Lo, Victor S.Y., Ziemba, William T. (eds.). Efficiency of Racetrack Betting Markets. World Scientific, 2008 — academic foundation for arbitrage in betting markets.
- Buchdahl, Joseph. "How profitable is sports arbitrage?" Football Data Blog, 2018.
- Franck, Egon, Verbeek, Erwin, Nüesch, Stephan. "Inter-market arbitrage in sports betting." Journal of Money, Credit and Banking, 2013.
- OddsJam Documentation — "How sports arbitrage works" (commercial publication, practical detail).
- RebelBetting — "What is sure betting? A complete guide" (operator-side methodology).
FAQ
How do I know an arb exists?
1/decimal_A + 1/decimal_B < 1.0. Example: DraftKings offers Lakers +145 (decimal 2.45, implied 40.82%); FanDuel offers Celtics -135 (decimal 1.741, implied 57.45%). Sum = 98.27% < 100% → 1.73% arb margin. Practical tools: OddsJam, RebelBetting, BetBurger, ArbHunter scan thousands of markets per second and surface arbs in real time. The challenge is execution speed — arbs typically last 30 seconds to 5 minutes before books adjust.How do I size each bet to lock profit?
stake_A = S × (1/decimal_A) / (1/decimal_A + 1/decimal_B); stake_B = S × (1/decimal_B) / (1/decimal_A + 1/decimal_B). Example with $1,000 total stake on the Lakers +145 / Celtics -135 arb: stake_LAL = $1000 × (0.4082/0.9827) = $415.41; stake_BOS = $1000 × (0.5745/0.9827) = $584.59. Both outcomes return $1,017.30 — guaranteed profit of $17.30 (1.73% on stake). Most arb tools auto-compute these sizes. Practical wrinkle: round to whole dollars at each book to avoid suspicious stake patterns.