Arbitrage

Arbitrage betting is the attempt to cover every mutually exclusive outcome of the same event at prices that create a positive return regardless of which outcome wins. The opportunity exists only when the best available prices, taken together, imply less than 100% probability. The arithmetic can be riskless in theory while the real execution is not.

The quickest arbitrage test

Convert each decimal price into implied probability:

Implied probability = 1 / decimal odds

Then add the implied probabilities for a complete outcome set.

• total above 1.00: the combined prices contain an overround;
• total equal to 1.00: the combined prices are break-even before practical costs;
• total below 1.00: a mathematical arbitrage exists if every leg can actually be placed and settled on the same event definition.

For two outcomes priced at 2.10 and 2.05:

• 1 / 2.10 = 0.476190;
• 1 / 2.05 = 0.487805;
• total = 0.963995, or about 96.40%.

Because the total is below 100%, the quoted prices leave about 3.60 percentage points of room before stake allocation and execution effects.

This is the mirror image of Overround, where one bookmaker’s complete market normally sums above 100%.

How to split the stakes

If you want the same gross return from every outcome, allocate the bankroll in proportion to each outcome’s reciprocal price.

For total bankroll B, price Oᵢ, and reciprocal sum S = Σ(1/Oᵢ):

Stakeᵢ = B × (1/Oᵢ) / S

Using $1,000 across the 2.10 and 2.05 example:

• reciprocal sum S = 0.963995;
• stake at 2.10 ≈ $493.97;
• stake at 2.05 ≈ $506.03.

Both legs return roughly the same amount:

• $493.97 × 2.10 ≈ $1,037.34;
• $506.03 × 2.05 ≈ $1,037.36.

So the theoretical profit is about $37.35 on $1,000, or roughly 3.74% of the money staked.

Notice the percentage is not simply 100% minus 96.40%. The equalized return is:

Arbitrage return on bankroll = (1 / S) - 1

That distinction matters when comparing opportunities.

Why the same-event requirement is strict

Two attractive prices are not an arbitrage if they settle different propositions.

Common mismatches include:

• one market includes overtime and the other does not;
• one tennis bet is void on retirement while another grades a completed set as action;
• one soccer market is “90 minutes” while another includes extra time;
• one side of a player prop requires the player to start while the other has different participation rules;
• one price is for a regulation result and another is for qualification or advancement.

The labels can look similar while the settlement rules leave a gap in which both wagers lose or one is voided. Before calculating stakes, define the event as precisely as a contract.

Execution risk is the part spreadsheets cannot eliminate

An arbitrage calculation assumes every quoted price remains available while the bettor places every leg. Real markets move.

Suppose the first $500 wager is accepted at 2.10. Before the second wager is placed, the other price drops from 2.05 to 1.85. The original equalized-profit structure is gone. The bettor now has an exposed position and must choose between accepting directional risk or completing the hedge at worse terms.

Other execution risks include:

• stake limits lower than expected;
• partial acceptance;
• odds changing during submission;
• market suspension;
• palpable-error or obvious-price rules;
• different void rules;
• account restrictions;
• currency-conversion cost;
• commission on exchanges;
• withdrawal or settlement delays.

This is why “risk-free” should refer only to the payoff table after every valid leg has been accepted under compatible rules. It should not describe the entire process.

Arbitrage is not the same as ordinary hedging

A hedge reduces or reshapes an existing exposure. It may lock in a profit, reduce a loss, or simply lower variance. The combined position can still have negative expected value.

Arbitrage begins with prices that, together, permit full outcome coverage with positive equalized value.

Example: you hold a long-shot ticket that has increased in value and place an opposing wager to guarantee some cash. That is Hedging. It becomes arbitrage only if the combined accepted prices produce a positive result across every settlement state.

The language matters because “I bet both sides” is not enough. Betting both sides at poor prices can guarantee a loss.

Arbitrage and casino table games

True arbitrage is uncommon in ordinary fixed-rules casino table games because the same casino controls the offered paytable and normally builds an advantage into the complete set of wagers. Placing Banker and Player in baccarat, red and black in roulette, or Pass and Don’t Pass in craps does not automatically create an arbitrage. Ties, zeros, commissions, pushes, and payout rules prevent the simple “cover both sides” idea from becoming free profit.

A casino promotion can occasionally create unusual combinations, but those must be evaluated under the actual promotional terms. A rebate, free bet, coupon, progressive, or cross-property offer changes the payoff structure; it does not remove the need to enumerate every outcome.

The core test remains Expected Value: what is the net result in each possible state, multiplied by its probability?

A three-outcome example

Suppose the best available decimal odds on a three-way event are:

Because the total is 0.96825, the theoretical equalized bankroll multiplier is 1 / 0.96825 ≈ 1.03279. A $1,000 bankroll can therefore be split to target about $1,032.79 gross return whichever of the three outcomes wins, before rounding and costs.

Approximate stakes are:

• A: $368.84;
• Draw: $286.89;
• B: $344.27.

Rounding to the allowed bet increment will slightly disturb equality. On very thin arbitrages, rounding alone can remove part of the profit.

Why headline percentages can mislead

An “arb percentage” can be quoted in several ways. One site may report the reciprocal sum, another may report 100% minus that sum, and another may report profit divided by bankroll. Those are related but not identical measures.

Always ask what the percentage means.

For the two-outcome example above:

• reciprocal sum: about 96.40%;
• gap below 100%: about 3.60 percentage points;
• equalized profit on bankroll: about 3.74%.

The last figure is the one directly connected to the return after correct proportional allocation.

When an apparent arbitrage disappears after costs

Small pricing gaps deserve special caution because real costs can be larger than the mathematical edge. Suppose an equalized position appears to earn 0.8% on a $1,000 bankroll. If an exchange charges commission on the winning leg, one account uses a costly currency conversion, or the allowed stake increments force unfavorable rounding, the final locked result can fall to zero or become negative.

The same applies to taxes or fees that are assessed asymmetrically. A calculation performed on headline odds assumes the stated payoff reaches the bettor exactly as modeled. If one venue quotes net odds and another applies a deduction after settlement, the legs are not economically equivalent even if the screen prices look compatible.

For thin opportunities, calculate the net cash settlement of every outcome after all known costs. The opportunity is not an arbitrage merely because the pre-fee reciprocal sum is below 1.00.

The practical checklist before calling a position an arbitrage

First, confirm that the outcomes are exhaustive: one and only one must settle as the winner under the shared rules. Second, confirm the settlement definitions match. Third, verify the available limits and minimum stakes. Fourth, calculate the allocation with the actual accepted prices, not screenshots taken minutes earlier. Fifth, include commissions, taxes, currency costs, and rounding if they apply. Finally, keep evidence of accepted bets and terms because a grading dispute can destroy the equalized result.

For related concepts, read Payout Odds and House Edge. Arbitrage is a pricing relationship, not a prediction method: the bettor does not need to know which outcome will occur, but does need every leg to survive execution and settlement exactly as modeled.