Stop-Loss and De-Risking Rules

Every trader believes, while a position is losing money, that it is about to turn around — that is precisely what makes losing positions so dangerous to manage in real time. Stop-loss and de-risking rules exist to take that decision away from in-the-moment judgment and replace it with something decided in advance, when nobody's ego or sunk cost is on the line.

The basic mechanics

The simplest version is a single trigger: exit a position once its loss from entry, or from a recent high, exceeds a set amount.

$$\text{trigger: exit when } \frac{P_t - P_{peak}}{P_{peak}} \le -L$$

In words: track the position's value against its own high-water mark, and once it has fallen by more than a threshold percentage $L$, act — not "consider acting," act. A de-risking ladder softens this into steps: cut the position by a third at a 5% loss, by half of what remains at 10%, and flat at 15%, rather than an all-or-nothing switch.

Worked example

A trader holds a $10 million position with a de-risking ladder: cut a third at -5%, cut to half of what remains at -10%, go flat at -15%. The position falls 5%, triggering a cut to $6.67 million. It keeps falling to -10% from entry; the ladder cuts the remaining position to $3.33 million. It falls further to -15%; the trader exits entirely. Total realized loss across the staged exits comes to roughly $0.83 million, versus the $1.5 million a full $10 million position would have lost held flat to -15% — the ladder cost some of the recovery upside if the trade had turned around early, but it capped the downside path.

What this means in practice

Systematic strategies build stop-losses directly into their code so there is no human moment of hesitation; discretionary desks write them into a risk mandate precisely because humans anchor on entry price and rationalize holding through pain. The trade-off is real: a strict stop-loss guarantees you exit some fraction of positions that would have recovered, converting a paper loss into a realized one right before the rebound.