Covering every side of a bet

A bookmaker builds a margin into every line. Quote both sides of a coin flip at 1.95 and the implied probabilities add to 102.6%. That 2.6% is the house edge, and it is why backing both sides at one book always loses.

Now take the best price on each side from two DIFFERENT books. One book is slow to move on the over, the other is slow on the under, and the two best prices can add to less than 100%. That is a price gap.

Every leg sits at a bookmaker that posts its own price and honours it up to its own limit, so the only real question is whether the margin is worth two logins. They are rare, they appear when one book lags a line move, and they close quickly.

At an exchange, a peer-to-peer venue or a prediction market, the price is not the house's opinion. It is somebody else's open order. That changes two things and neither is in the number on the card.

The first is the fee. An exchange takes a commission on winnings rather than baking a margin into the price, so a headline 2.02 does not pay 2.02. Every margin on our board is already net of that: the card shows both the quoted price and what it actually pays.

The second is depth, and it is the one that costs people money. Behind a bookmaker's 2.02 is that book's limit. Behind an exchange's 2.02 is one person's order for one amount, and that amount is not in any odds feed, including ours. Push more than it through and you eat that order and average into the next one at a worse price. A 1% margin does not survive much of that.

So on an exchange leg, check the available size at the venue before you place anything else.

Two books can drift onto different lines on the same market. Take Over 2.5 at one and Under 3.5 at the other and you own everything between them.

• Exactly 3 goals: both legs win.
• 2 or fewer, or 4 or more: one leg wins and nearly covers the other.

So the downside is not a loss of the whole stake, it is the small difference between two prices, and you know it exactly before you place either leg. That number is the one on the card, printed as “worst case” and deliberately not in the profit colour. A middle's upside is real and conditional, and printing it in green beside a locked-in gap would invite a comparison that is not true.

Lines have to drift apart first, so middles are rarer than price gaps.

Sometimes covering every side of a line costs almost nothing, 2% or less after fees. You will not make money on it. What you get is turnover at a known, small, fixed cost, without taking any position on the game.

This is the section that clears a deposit bonus. A bonus that demands you turn over your deposit some number of times is asking you to place bets, not to win them. Doing that on ordinary lines costs you the bookmaker's margin on every rollover, call it 4 to 6% a turn. Doing it on a low hold costs 2% or less. Over a rollover requirement of several times your deposit, that difference is most of what the bonus was worth.

Two things to hold onto. Turnover requirements only exist at bookmakers, so a row with an exchange or prediction-market leg is cheap cover but will not clear anything. And read the bonus terms for a minimum-odds rule: many exclude bets under about 1.50, and half of a low hold is usually below that.

A hold this thin is also one price move away from being an outright gap, which is the other reason to watch them.

A covered position is only covered once every leg is on. Between the first bet and the last you are holding a one-sided position, and if the second price moves or vanishes you are stuck with it at a price you did not choose.

1. Open both venues before placing anything, and check the quote ages on the card. A nine-minute-old quote on a gap that lives two minutes is already gone.
2. Place the least reliable leg first: the exchange order, the smaller book, the one most likely to move. If it fails you have lost nothing.
3. Then the leg you are confident will still be there.
4. If the second leg has moved, you are not obliged to complete a worse position. Taking a small loss to close is usually cheaper than holding a bet you never wanted.

The obvious question about a position that wins whatever happens is why not put your whole balance on it. Four answers, and every one of them scales with the amount.

Execution. The second leg fails and the position is naked, for the full amount. This is the most common way arbitrage actually loses money.

Depth and limits. A book has a maximum it will accept, and an exchange price has one person behind it. Beyond that you average into worse prices and the margin goes.

Account life. Bookmakers limit and close accounts that arb, and a large, round, perfectly hedged stake is the most recognisable pattern there is. A position that nets you fifty and costs you the account has lost you far more than it made.

Settlement. Venues can settle the same event differently: void at one, action at the other, different rules on an abandoned match. Then one leg stands alone for its full stake.

This is why the stake calculator on each card suggests a fraction of your bankroll rather than everything you have, and why the fraction differs by section: 5% when every leg is at a book that honours its posted price, and 2% when one leg is an open order at an unknown size. Both are starting points you can type over. The real ceiling is what the venues will take and how much attention your accounts can afford.

Kelly does not apply here, which is why the calculator does not use it. Kelly sizes a bet by how much variance you are taking on, and a covered position has almost none, so Kelly would say “bet everything” and be wrong for all four reasons above.

• Every margin is net of what the venue actually takes, not the headline price.
• No two legs at the same trading desk. Several brands share one operator, and a “gap” between two of them is the same book disagreeing with itself, which it will void.
• Quotes we cannot confirm are current are dropped, and the page says how many.
• Prices too far from the fair line to believe are dropped, because those are errors, not opportunities.
• Stakes are rounded to what a book will actually accept and the margin is recomputed from the rounded numbers. On a small outlay that rounding can eat the whole gap, and when it does the card says so rather than quoting you the ideal.