Risk management

Arithmetic and arrangement rather than prediction. The components are consistent across the literature even where the conventions differ: the size of a position, the distance to the level at which it would be closed, how much of the account those two imply is at stake, how many positions are open in instruments that move together, and how much of the account is committed as margin at once. None of it forecasts anything. It bounds what a forecast being wrong would cost.

The conventions traders quote are bounded conventions, not findings. A fixed fraction of account equity treated as the amount at stake on any one position is the most cited; a ceiling on total exposure and a ceiling on aggregate portfolio heat extend the same idea to a whole account. The broker's own arrangements sit alongside these and are not a substitute for them: a margin call and a close-out are constraints the firm applies to protect itself, and they act after most of the account's own limits would already have been reached.

Three limits are less firm than they look, and each is a common misreading. A protective level is a level at which an order becomes executable, not a guaranteed exit price, so the realised loss can exceed the planned distance in a gapping market. A fixed percentage rule applied position by position does not bound the loss across positions that move together, since two strongly correlated positions behave closer to one position of double the size. And on a margined contract the arithmetic runs on the full contract value, so losses are calculated on the whole position and are not limited to the amount deposited. Practitioners disagree about much in this area, but the published work on sizing is unusually consistent in finding that how much is committed matters more to the distribution of outcomes than where a position was entered.