Arbitrage

In short: A betting strategy where you place bets on all possible outcomes of an event across different sportsbooks, guaranteeing a profit because the combined implied probabilities are less than 100%.

Arbitrage betting (or arbing) exploits pricing disagreements between sportsbooks. When the combined implied probabilities at different books drop below 100%, you can bet every outcome and lock in a guaranteed profit regardless of result. The math is clean. The execution is where everything actually happens.

How an arbitrage opportunity actually looks

Consider a Lakers vs Celtics game. DraftKings has the Lakers moneyline at +115. FanDuel has the Celtics moneyline at +110. Most of the time these prices would be in line with each other, but tonight FanDuel is slow updating after a Celtics injury report and DraftKings has already moved.

Check the combined implied probability:

• DraftKings Lakers +115 → 46.5% implied
• FanDuel Celtics +110 → 47.6% implied
• Combined: 94.1%

The 5.9% gap below 100% is the arbitrage edge. Bet both sides with sizes that produce equal payouts and you profit no matter who wins.

To work out the stakes for a $1,000 total outlay:

Lakers stake = $1,000 × (46.5% / 94.1%) = $494
Celtics stake = $1,000 × (47.6% / 94.1%) = $506

If Lakers win: $494 × 2.15 = $1,062 returned. Profit: $62
If Celtics win: $506 × 2.10 = $1,063 returned. Profit: $63

Either way, ~$62 profit on $1,000 staked, a 6.2% return locked in before tip-off.

The 5.9% combined-implied gap and the 6.2% realized return don’t quite match because of how the math compounds, but the headline is the same: this is a guaranteed profit, not a probabilistic edge.

Where arbs actually come from

Arbs aren’t random. They exist because of structural disagreements between book types:

• Sharp vs retail mispricing. Pinnacle accepts large sharp action and adjusts prices instantly. DraftKings adjusts more slowly because their pricing accounts for where the public is betting. When sharp money moves Pinnacle but retail hasn’t caught up, the gap is the arb.
• Cross-book reaction lag. When one book moves a line, others usually follow within minutes. The arb window is the lag between the first book moving and the slowest book catching up. Sometimes 30 seconds. Sometimes 20 minutes.
• Promotional pricing. Books occasionally run “boosted odds” promotions that briefly create off-market prices. These produce easy arbs because the boost is intentional rather than an error.
• Exchange vs sportsbook gaps. Novig and Betfair prices come from peer-to-peer order flow without vig. When their prices diverge from sportsbook lines, the gap can be larger and last longer.
• Prediction market vs sportsbook gaps. Kalshi, Polymarket, and other prediction markets price sports events using different mechanisms than sportsbooks. The structural difference produces persistent arbitrage opportunities, especially in markets where the prediction market and sportsbook have different views on the same event.

Typical sportsbook-to-sportsbook arbs run 1-3%. Sportsbook-to-exchange arbs run 2-5%. Prediction market arbs can run 5-15% or more, though they often come with execution and settlement complications.

What arbitrage is not

A few things commonly confused with arbitrage:

Arbitrage vs. +EV betting. Arbs guarantee profit on every outcome by betting both sides at different books. +EV is one-sided and probabilistic. You expect to profit on average but any individual bet can lose. Arbs have effectively zero variance per bet; +EV bets have substantial variance. Most serious bettors do both.

Arbitrage vs. middling. A middle is a two-leg bet that can win both legs if the result lands in a specific window, but loses a small amount otherwise. Arbitrage profits regardless of outcome; middles trade certainty for higher possible upside. Middles are positive-EV bets, not guaranteed profits.

Arbitrage vs. hedging. Hedging means placing a second bet to reduce exposure on an existing position you already have, usually accepting a worse net price to lock in profit. Arbitrage is taking both sides simultaneously at favorable prices from the outset.

“Risk-free” promotions. Books advertise “risk-free bets” and “bet $5, get $200.” These are not arbitrage. They’re promotional credits that often pay out in bonus bets rather than cash, with their own EV math. The marketing language borrows from arbitrage’s appeal but the underlying structure is different.

The hard parts nobody mentions

Most arbitrage explainers stop at the calculation. The practical reality is harder than the formula suggests:

Arbs disappear fast. Once a sharp book moves and slow books start catching up, the window closes. A 3% arb at 6:00 PM might be a 1% arb at 6:03 PM and gone by 6:05. Speed of execution matters more than spotting the opportunity. By the time you’ve calculated stake sizes manually, the price often has moved.

Stake sizing limits real-world arbs. A theoretical 3% arb on $10,000 is $300 of guaranteed profit. A 3% arb on $200 (because that’s the max bet size your limited account allows) is $6. Many arbs that show up in scanners are technically real but practically too small to bother with after factoring in execution time.

Account limiting accelerates with arbing. Books can only see their own activity, but the patterns that signal arbitrage are distinctive: bets placed within seconds of line moves at sharp books (suggesting reaction to a scanner), consistently taking the absolute best available price the moment it appears, stake sizes that look calculated rather than round, concentrated activity on the soft side of markets that just moved, and regular withdrawal cadences. Pure arbers tend to trigger these patterns more reliably than mixed +EV bettors and get limited faster as a result.

False arbs from stale data. A scanner showing a 4% arb on a market where one book hasn’t updated in 30 minutes isn’t a real arb. It’s stale data. When you try to place the bet, the book will either reject it, void it, or have already moved. Filtering out stale-data false positives is a meaningful part of running arbitrage in practice.

One leg gets voided. If you place both sides of an arb and one gets voided (pitcher change in MLB, player scratch in NBA, postponement), you’re left with an unhedged position. Most arbs are robust to this, but it happens.

The scanner-edge versus realized-edge gap is real. Scanners surface arbs at their headline edge percentage. By the time you’ve placed both legs, the realized edge is typically 50-80% of the headline number after accounting for line movement during placement, slight stake sizing errors, and the occasional missed leg.

What “running arbs” looks like as a habit

The serious arbitrage workflow:

1. Run a scanner across many books to identify combined-implied-probability gaps
2. Filter ruthlessly. Discard anything with stale prices, anything below your minimum edge threshold (typically 2%+ after expected friction), anything on a market with known settlement quirks
3. Open both books simultaneously, calculate stake sizes for equal-payout structure
4. Place the leg at the book most likely to move or limit first (usually the soft side)
5. Place the second leg immediately, accepting whatever line is current
6. Log the bet, the prices taken, the final closing lines, and the actual realized profit when settled

Volume matters. Individual arbs are small. A bettor placing 2-3 arbs per day at 1-3% on bankroll-appropriate stakes is running a real operation; one arb a week at 5% is a hobby.

For deeper coverage of how to identify and execute arbitrage opportunities, see Arbitrage Betting Explained.