Stop losses: what they do, and what they do not

The stop did exactly what the paperwork said it would, which is why Jane spent a morning believing her brokerage had cheated her. She owned a stock at $52, placed a stop loss at $46, and considered the downside handled: whatever happened, she was out at $46. Then came an earnings report, released after the close, and a market that digested it overnight. The stock’s last trade before the news was $48.30. Its first trade the next morning was around $41, and that is roughly where Jane’s stop sold her shares, five dollars beneath the number she had chosen, without the order malfunctioning in any way. Her morning happens after every earnings season, and the gap between what she believed she had bought and what she had actually bought is this article’s entire subject.

What she believed she had bought was insurance: a guarantee that her loss ends at $46. What a stop loss actually is, is an instruction: sell once the market touches $46, at whatever price the market pays after that. A stop is an instruction, not insurance. The difference between those two products is invisible on calm days, which is precisely how the confusion survives, and it becomes the whole story on the one morning the protection was for.

What the order is, mechanically

The definitional part is short, and worth having exactly right. The plain version is a stop order, which FINRA’s investor guidance defines in a single line: “A stop order is an order to buy or sell a stock once the stock reaches a specified price.” Until that price trades, the order rests, invisible to the market. The moment it trades, your stop converts into a market order, an instruction to sell immediately at the best available price, and from that moment forward the stop price you selected has no further role. It was a trigger, and it has been pulled.

There is a second version, built to fix that, which trades one uncertainty for another. A stop-limit order converts, when triggered, into a limit order at a price you set, so, again per FINRA, “your shares will only be bought or sold once the stop price is reached if your brokerage firm can obtain the specified limit price or better.” The catch sits in the if. A market falling fast through your limit leaves the order unfilled and you still holding, all the way down.

A third variant moves instead of resting. A trailing stop follows the price up at a distance you choose, a dollar amount or a percentage, and stops following the moment the price turns down, so the exit level ratchets higher behind a winning position and locks in more of the gain the further it runs. It is a genuinely useful automation for letting winners breathe without surrendering them, and it changes nothing else described here: when touched, it converts to the same market order, with the same unguaranteed fill, and a trail set tighter than the stock’s ordinary wobble simply hands the position to the first ordinary wobble. Every stop variant is the same machine with a different trigger. The machine is what the next section is about.

Where the protection ends

Three conditions defeat the insurance reading of a stop, and none of them is rare.

The first is Jane’s: the gap. Stocks trade in sessions, news does not, and a stock that closes at $48.30 can open at $41 with no trade in between. A stop at $46 cannot fill at $46 if $46 never trades again; the order triggers on the way past and fills near the open. Here is the shape of her morning:

The gap has an intraday relative that wears an official uniform: the halt. Exchanges pause trading in a single stock after sufficiently violent moves, and news that matters, a merger, an investigation, a withdrawn forecast, often arrives wrapped in one. While the halt lasts, nothing trades, so no stop can act; when trading resumes, it resumes at whatever price the reopening auction finds, which can sit far below your level. The stop then triggers on the first prints of the new reality, exactly like Jane’s morning compressed into twenty minutes.

The second failure mode is the fast market, the gap’s everyday cousin. FINRA’s guidance runs the example plainly: “If the market’s moving fast, you could receive less—and potentially significantly less—than $50 per share by the time your order is executed.” Speed is the enemy because a market order joins a queue; in a rout, the queue is long and the price is a moving target.

The third defeat runs the opposite direction, and it costs more traders more money than either crash scenario: the fluke trigger. “Rapid price movement during a short period of time could trigger a stop order,” FINRA notes. “Due to market volatility, the stock might later rebound and resume trading at its prior price level.” A stop placed inside a stock’s ordinary daily wobble is not protection; it is a standing offer to sell your position to the first noisy dip that comes past, which is why stops set at the round number just under the purchase price, where thousands of other people also set them, have a way of being collected minutes before the market turns.

How seriously should you take these failure modes? Consider what the New York Stock Exchange itself concluded. In a 2015 client notice, the exchange announced that it “will no longer accept new Stop Orders and Good Till Cancelled (‘GTC’) Orders beginning February 26, 2016,” cancelling every one resting on its book. Brokerages still offer stop orders by simulating them on their own systems, so nothing changed for the ordinary customer’s app. But the country’s oldest exchange looking at this order type and deciding to stop taking it is a fact worth exactly the pause it just gave you.

When stop rules help, measured

None of this means stops are a scam, and the careful research on them says something more interesting than either the marketing or the horror stories. Kathryn Kaminski and Andrew Lo, in a Journal of Financial Markets study titled When Do Stop-Loss Rules Stop Losses?, analyzed stop rules as policies, predetermined triggers that cut a portfolio’s exposure after a threshold of cumulative loss, and asked when the policy adds value. The answer turns on a single property of the market being traded. If prices move like a random walk, where the next move owes nothing to the last, their conclusion is blunt: stop-loss rules never stop losses; the rule just sells you out of positions whose future was, on average, unchanged. But where returns have momentum, where losses tend to be followed by further losses, the arithmetic reverses, and the benefit of stopping is, in their phrase, directly proportional to the magnitude of return persistence.

Read that as a user’s manual. A stop is not a general-purpose amulet you tape onto anything you own; whether it protects or merely bleeds you depends on whether the thing you trade trends. Trend-following strategies and momentum markets are where stop rules earn their keep, and the rule is close to pointless wherever returns are genuinely unforecastable, which is the case their random-walk result describes. One finding in the same paper deserves to be reported alongside that, because it cuts against the folklore in the other direction: when the authors tested stop rules on a plain buy-and-hold position in U.S. index futures from January 1993 to November 2011, moving into long-term bond futures whenever the stop triggered, they found that at longer sampling frequencies certain stop-loss policies added value over buy-and-hold while substantially reducing volatility. So the honest summary is narrower than either slogan: the stop belongs to the strategy and the holding period, not to the instrument, and a trader who cannot say what property of their market makes their stop sensible is holding Jane’s insurance policy.

The job a stop actually does

Placement is where the fluke-trigger risk gets managed, and the principle fits in a sentence even though the craft does not: a stop belongs at the price where your reason for the trade is established as wrong, not at the loss that feels tolerable. The two numbers rarely coincide. The tolerable-loss stop lands wherever your comfort ran out, which is usually inside the market’s routine noise, in the crowd of everyone else’s comfort, at the obvious round number the next dip passes through on its way to rebounding. The invalidation stop lands where the market structure that justified the entry has actually failed, which might be closer than comfort wanted or further than comfort allows, and if it is further than your size can afford, the honest conclusion is a smaller position, not a nearer stop. Reading where invalidation truly sits is level-reading craft, and it is a skill, not a formula.

So what is the honest product? A stop loss is the enforcement arm of a decision you made while calm. Before entry, with no money at stake, you decide the price at which your idea is established as wrong, the concession point this section’s sizing post derives everything from. The stop’s job is to execute that decision at three in the afternoon while you are in a meeting, or at three in the morning while you are asleep, or at the precise moment you would otherwise be renegotiating with yourself. It automates your discipline, not your outcome. That is a genuinely valuable product, arguably the most valuable one a beginner can buy for free, and it is a different product from a floor under your money. Reading where that concession price actually sits, where the chart says your idea failed rather than where your nerve runs out, is chart craft, and Chapter 16 of The Complete Trader, on support and resistance, is where the book teaches it.

Which is why the last word belongs to size. A stop caps the ordinary loss, the slow drift against you on a liquid day. The gap, the halt, the fast market pay no attention to it, and the only variable in your control that caps those is how large the position was when the news hit.

Rerun Jane’s morning under that rule and the story changes shape without a single order changing type. Her stop still fills near $41; the gap was never negotiable. But a position sized to survive a five-dollar surprise turns the same morning from an ending into an expense, and her stop, understood at last as an instruction, goes back to doing the one job it was always doing: making sure that when wrong arrived, she left, promptly, without being asked twice.