How betting markets were made

Betting between two parties on the outcome of a contest is ancient and requires no institution beyond an agreement and, usually, a stakeholder to hold the money. Such a wager has no margin: whatever one side loses the other wins, and the stakeholder's only interest is a fee for the service. The structure is symmetrical and there is no house.

Its limitation is practical. Finding a counterparty who wants the opposite side at an agreed price is slow, and the price two individuals settle on carries no information beyond their own opinions. The commercial history of betting is essentially the history of solving that matching problem, and every solution introduced a party who took a position rather than merely holding the stakes.

The professional layer solved the matching problem by standing ready to take the opposite side of any outcome at posted prices. This is a genuine service and it carries genuine risk, and the layer's compensation is built into the price set. Posting prices whose implied probabilities sum to more than one—the overround described in this board's first mathematics section—creates a margin that is realised when the money taken is balanced across outcomes.

This makes a layer's business fundamentally about balance rather than about prediction. A perfectly balanced book pays the same amount whichever outcome occurs and retains the overround with no exposure at all. Price movement in such a market is therefore driven substantially by where money has already been placed, which is why prices move in response to demand as well as to information, and why reading a price as a pure forecast is a mistake.

For a long period the characteristic venue was a physical ring, with prices called and chalked and adjusted continuously in response to what was being taken. Off-course and telephone betting extended the same relationships beyond the venue, and legalisation in various jurisdictions turned an informal trade into a licensed one with premises, records and duties.

The shift to screens changed the arithmetic more than the structure. Prices could be updated continuously and compared instantly across sources, which compressed margins in liquid markets, while the number of distinct propositions on a single contest multiplied enormously—and the additional propositions generally carry wider margins than the principal ones, for the same reason that exotic bets on a wheel layout do.

The exchange model returns to the original symmetric structure with the matching problem solved by software. Participants post prices on either side and the platform matches them, charging a commission on winnings rather than embedding a margin in the price. There is no house position, and the prices that result are widely treated as a relatively clean read on market opinion precisely because no overround has been added to them.

It is worth being clear about what this does and does not change. It removes the layer's built-in margin and replaces it with an explicit commission, which is smaller but still a cost applied to every settled position. The expected value of participation remains negative for the average participant; the structure of who takes the margin, and how visible it is, has changed rather than the direction of the arithmetic.