Stops That Belong to the Chart

The stop marks where the idea is wrong

A stop has exactly one honest job: it states the price at which the reason for the trade no longer holds. Below a reclaimed level, below the base low that supported the entry, below the breakout candle. Stated that way, the placement is a fact about the chart rather than a preference about risk — and it is the only placement that can be evaluated after the fact, because a stop that is hit tells you something specific: the market disagreed with the premise. The alternative that most people use is a **percentage stop**: always −7%, or whatever the rule is. It has the virtue of being simple and two faults that matter. On a volatile instrument, a fixed percentage can sit inside the ordinary daily range, which means it is hit by randomness and produces losses that carry no information. On a quiet one, it can sit far beyond any structural level, which means the account risks more per trade than it needs to in order to be wrong. In both cases the stop stops describing the idea and starts describing the trader. The check that fixes the volatility problem is **ATR**: the average true range, which is the typical distance the instrument travels in a bar. A stop closer than about two average ranges is inside the noise, and no amount of analysis protects it — the instrument will wander that far without changing anything about the thesis. Practical form: measure the structural distance in ATR. If it is above two, the structure agrees with the noise, and the trade is sized from it. If it is below two, either the structural stop is genuinely tight (a small base, a compressed range) or you are looking at a trade whose invalidation is closer than its weather. The setup in the workbench — Entry: $84.00 · Structural level: $78.60 · Stop distance: 6.43%, or $5.40 a share · Distance in ATR: 2.57 — outside the noise · Shares at a 1% budget: 111, a $9,333 position ← The order of operations is the point: chart → stop → distance → size. The position is a consequence of the first three, which is what makes it computable before the trade rather than negotiable during it.

What a stop cannot do

A stop is an instruction to sell at the market once the trigger price trades, which means it controls **when** you exit and not **what you get**. Most of the time those coincide, because liquid markets trade continuously and the fill is within a cent or two of the trigger. The exception is the gap: an earnings release, a takeover, a sector-wide shock, anything that changes the price while the market is shut. The stock reports after the close and opens 8% lower, and the stop fills at the open — several points below the trigger, and nowhere near the price you had in mind. This is not a defect in the order type and it cannot be fixed by choosing a different one. A stop-limit protects the price by refusing to fill, which leaves you holding the position through the gap, and that is the trade rather than a solution to it. What can be done is arithmetic: the true worst case for a position is the risk budget **plus** the largest overnight gap the instrument has delivered, which argues for smaller positions in names that can gap and for not carrying full size through scheduled events. A position stress test that assumes the stop is the floor is testing a market where the close is a promise. The two limits together give the honest description of a stop-loss plan. It bounds the loss when the market is continuous, which is most of the time and is worth having. It does not bound the loss when the market jumps, which is when it matters most, and the defence against that is size rather than an order type. Written down that way, a stop is a useful instrument with a stated failure mode — which is a much better thing to own than a guarantee that does not exist. A gap through the stop, from the workbench — Planned loss at the stop: $12,000 — the 1% the sizing was built on · Earnings gap, filled at the open: $18,900 — 1.575% of the account · Overshoot past the trigger: 5.75% of the entry price ← · What would have prevented it: a smaller position, or no position through the event The most common way to lose more than the plan allows is not a missed stop but a gap through one, which is why the sizing rule uses a worst case rather than the stop price.

What stop distance does to the other statistics

Choosing where a stop sits changes more than the size of a loss, and the direction of the change is counter-intuitive enough to be worth stating plainly. Tighten a stop without changing the entry and the **win rate rises** — most adverse excursions that would have become losses are cut off before they turn into one, and more trades end in a small profit. At the same time the **payoff ratio falls**, because the trades that survive a tight stop are frequently the ones that never went anywhere, while the winners you are most interested in are the ones with enough room to need it. Win rate and payoff ratio move in opposite directions, and because expectancy is the product of the two, a tighter stop can raise expectancy, lower it, or leave it unchanged. It is not a lever that improves things by being turned. Whether a given stop distance helps therefore depends on the instrument’s noise, and that is the only reason the ATR check belongs in the procedure. A stop inside the ordinary range of a position’s daily movement will be triggered by noise rather than by being wrong, and each of those exits is a transaction cost plus a re-entry decision made in worse conditions. The structural level — the price at which the thesis is actually disproved — is the anchor; the noise check is whether that level is inside the market’s everyday breathing. When the structural level sits closer than the noise, the honest conclusion is usually that the position is too large or the idea too tight to trade, not that the stop should be widened to give it room. There is a further cost that the statistics hide and the account statement does not: **the re-entry problem**. A stop that fires puts you flat and facing a decision you did not plan — whether to buy back, at what price, and with what size, now that the original premise has been contradicted at least locally. Systems that stop out and re-enter mechanically pay the spread twice and the impact twice and still hold the same exposure. Systems that stop out and never re-enter avoid that cost and accept a different one, which is being absent from the winners that start with a shakeout. Both are legitimate; what is not legitimate is treating the stop as the end of the trade when the plan has no re-entry rule, because that means the decision is being made by the market rather than by you. Tightening the same entry — Wide stop, thesis-based: Lower win rate, higher payoff ratio, fewer exits to re-enter · Tight stop, inside the noise: Higher win rate, lower payoff ratio, more spread paid twice · Tight stop with no re-entry rule: The market decides when you are invested — which is not a plan ← When testing a stop distance, look at expectancy rather than at the win rate, and count the re-entries and their cost. Most of the apparent improvement from a tighter stop disappears once the round trips are priced.

Where the stop actually lives, and what it does to the market

A stop is usually described as an order at a price, and that description hides three mechanical facts that decide whether it works. The first is **where the order is held**. Most retail stops are not resting on the exchange; they are monitored by the broker, which watches the market and submits a market order when its trigger condition is met. That matters because the trigger is defined on the broker’s feed, not on the consolidated tape: an order to sell that triggers when the last trade prints at or below the stop can be triggered by a single small print in a thin moment, and it can equally fail to trigger when the quote is at the stop but the last trade is a cent above it. Two brokers can therefore fill you at different moments on the same stock, and a stop that “should have” triggered may not have, because the price that triggered yours never printed on the broker’s data. The second fact is what the stop becomes when it triggers. A stop-market order becomes a **market order**, which means it takes whatever liquidity is available in the book at the moment it arrives — and if several participants have stops at the same level, which is common at round numbers and at obvious technical levels, they trigger together and arrive together. A cluster of stop orders is a cluster of market orders, and market orders consume the book from the top down: the first fills at the touch, the second at the next level, and the last one a few cents or a few percent worse. That is the mechanism behind the moves people describe as stop hunting, and the honest version is that no hunter is required — a visible level at which a crowd has placed the same order is a pool of liquidity for anyone willing to trade into it. The third fact is the **stop-limit offset**, which is the only lever available: a limit a few cents away from the trigger will fill in ordinary conditions and refuse to fill in a fast one, converting an uncertain price into a certain non-exit. The choice between those two bad outcomes is a stated risk preference, which is why it belongs in the plan rather than in the moment. There are two further mechanics that decide what the order does to your risk and they are both about time. A stop that is not marked good-till-cancelled expires at the close, so a position held overnight may be unprotected for the gap — and a stop that is good-till-cancelled persists through the next session, which is what you want if you cannot watch the open. The other is the trigger reference itself: a sell stop monitored on the **last trade** behaves differently from one monitored on the **bid**, and the difference is largest in exactly the conditions where the stop matters, because a bid can collapse several cents before any trade prints. The practical rules that follow are unglamorous and they are the whole content of the read: give the stop an explicit time-in-force, know whether it is held by the broker or resting at the exchange, place it on the side of the book that matches what you are actually exposed to, and prefer a level that the crowd is not also using if you can find one — because the technical level everyone can see is also the level at which the most orders will arrive at once. • Most retail stops are held by the broker and triggered on the broker’s feed, not the exchange’s book. • A stop-market order becomes a market order and consumes the book from the top down. • Stop clusters at visible levels arrive together and are the pool that produces the sharp move. • The stop-limit offset chooses between an uncertain price and a certain non-exit — decide it in the plan. • Set time-in-force explicitly, and match the trigger reference to the side of the book you are exposed to. The link to the market-structure lesson is direct: the stop is an order like any other, so it is subject to queue priority, to the liquidity in the book, and to the difference between the last trade and the quote. Treating it as a property of the position rather than as an instruction submitted to a market is how a plan that looked complete turns out to have a hole in it.

Sources

• Average True Range as a measure of instrument noiseJ. Welles Wilder, "New Concepts in Technical Trading Systems" (1978)
• Structural invalidation versus percentage stopsStandard technical-analysis practice; Elder, "Trading for a Living"
• Gap risk and why stops fill beyond their triggerInvestopedia and exchange documentation on stop and stop-limit orders
• The stop as the input to position sizingVan Tharp, "Trade Your Way to Financial Freedom"