EV Betting vs Arbitrage Betting

Both approaches start from the same observation: prices for one market differ across books, and sometimes differ from a defensible estimate of the truth. They use that disagreement in opposite ways.

EV betting takes one side when the price is better than your fair probability. You accept the outcome risk and rely on many such positions averaging out toward the expected gain.

Arbitrage takes every side of a market at different books when the best prices, converted to implied probabilities, sum to less than 100 percent. Correct stake proportions return the same amount whatever happens, so the outcome stops mattering. No probability estimate is needed at all.

Illustrative prices for a two-way market: book X offers side A at 2.10, book Y offers side B at 2.05, and a sharp reference market, once its margin is removed, puts A at 49.5 percent.

As an EV position. Backing A at book X alone has expected value of 0.495 times 2.10, minus one, or about plus 4 percent per unit. Each unit returns either plus 1.10 or minus 1, so the result of any single bet is dominated by variance, and the 4 percent depends entirely on the 49.5 percent being right.

As an arbitrage. The implied probabilities are about 47.6 and 48.8 percent, summing to about 96.4 percent. Staking in proportion, about 49.4 percent of the total on A and 50.6 percent on B, returns about 1.037 times the total stake either way: plus 3.7 percent, locked, with no view on who wins. Gaps this wide are uncommon and brief; the numbers are for illustration.

Arbitrage does not remove risk; it swaps outcome risk for operational risk.

EV betting's main risk is the estimate. If the fair probability is wrong, every position built on it is wrong in the same direction, and the error does not average out. Results take a long time to reveal this, which is why EV work leans on closing line value and calibration.

Arbitrage's main risks are execution and settlement. The legs are not placed at the same instant, so a price can move or a bet can be rejected after the first leg is accepted, leaving an unbalanced position. Books may void bets on palpable errors under their terms, and settlement rules differ on overtime, retirements, abandoned games and player participation. If one leg voids and the other stands, the position is no longer an arbitrage.

EV betting needs a fair-price source (a devigged sharp market, a model, or both), the best available price, and closing prices for evaluation. It can work from one or a few books.

Arbitrage needs synchronised, fast quotes from as many books as possible, because opportunities appear briefly when one book is slow to update. It also needs funded accounts everywhere at once, since you cannot move money between books in the seconds a gap lasts. In a data pipeline, most apparent arbitrages turn out to be stale quotes or mismatched markets, so quote timestamps and market equivalence checks matter more here than anywhere.

Neither approach is, in itself, a distinct offence: both consist of placing ordinary bets with licensed operators. The constraint is contractual. Book terms typically allow stakes to be limited or accounts closed, and both styles of betting attract that attention.

In Great Britain, the Gambling Commission's chief executive Andrew Rhodes wrote in July 2025 that operators "are entitled to act in their commercial interests and manage liabilities", quoting the Gambling White Paper, and added that being a successful bettor is not a protected characteristic in discrimination law and that operators should be transparent about how and when they restrict accounts. In the United States, sports betting legality has been decided state by state since the Supreme Court struck down PASPA in Murphy v. NCAA in May 2018, so only bet where a licensed operator is permitted. This is not legal advice.

This page compares two approaches as methods and is not betting advice.