Traders argue about where stops belong as though it were a matter of judgement, and part of it is not. Three numbers describe any strategy, two of them are mechanically linked, and moving one moves the other in a direction that cancels most of the benefit. Knowing which part is arithmetic narrows the argument considerably.
Any strategy can be described by how often it wins, how much it makes when it wins, and how much it loses when it loses. Those three determine everything else, and the useful fact about them is that the first and the third are joined at the hip: the loss size is set by where you put the stop, and where you put the stop is what determines how often price reaches it.
This is why comparisons of win rate between strategies are close to meaningless without the other two numbers. A system that wins nine times out of ten and a system that wins three times out of ten can produce identical results, and either can be the better one, depending entirely on the sizes. Quoting a win rate alone conveys almost nothing, and it is the number most often quoted alone.
The arithmetic that ties them together is short. Expectancy per trade is the win rate multiplied by the average win, minus the loss rate multiplied by the average loss. Every argument about stop placement is an argument about how a change to one term moves the others, and most of those arguments are conducted without writing the expression down.
Win rate on its own describes nothing. It is one of three numbers, and the other two are what decide whether the first is good news.
Move a stop further away and something reliable happens: fewer trades are stopped out, so the win rate rises. It rises for a mechanical reason rather than because the strategy improved, since the price now has further to travel before the exit triggers. That feels like progress and the equity curve frequently looks better for a while.
What also happens is that each loss is now larger, by exactly the amount the stop moved. If the underlying edge did not change, the two effects offset: you lose less often and more each time, and the expectancy is where it was. The comfort improved and the arithmetic did not, which is the specific reason this adjustment is so popular and so rarely productive.
The case where widening genuinely helps is narrow and worth naming, because it does exist. If the original stop sat inside normal price movement, it was being triggered by noise rather than by the strategy being wrong, and those exits were losses the strategy never needed to take. Widening past that threshold converts noise-triggered losses into trades that get a chance to work, and it improves expectancy for a real reason. Beyond that threshold, the offsetting begins and the improvement stops.
A stop is not a tool for making money, and treating it as one produces most of the confusion around it. It is a decision made in advance, at a moment when you are calm, about an action you would otherwise have to take at a moment when you are not. Its entire value is that it removes a choice from a situation in which choices are made badly.
That reframing settles several arguments. The right stop is not the one that maximises a backtest, because a stop optimised on past data is fitted to noise that will not repeat. It is the one that reliably caps the loss at an amount you can absorb repeatedly, placed where the reason for the trade would be demonstrably wrong rather than where the arithmetic happens to look best.
It also explains why removing a stop mid-trade is so consistently expensive. The decision to widen or cancel is taken at the exact moment the stop was designed to protect against, by the version of you least equipped to take it, with information that feels new and usually is not. The mechanism was working; the override is the failure.
Stops are not placed at random. They gather at round numbers, just beyond recent highs and lows, and at the levels that widely used indicators produce, because most participants are looking at the same charts and applying the same rules. This clustering is observable and it has consequences for anybody whose stop is in the crowd.
The consequence is not that somebody is targeting you. It is that a price reaching a cluster triggers many stops at once, all of which become market orders in the same direction simultaneously, into a book that is thinner than usual because the move that got there consumed some of it. The result is a fast extension beyond the level, followed frequently by a partial recovery once the forced flow is exhausted.
That pattern is what people describe as their stop being hunted, and the description gets the mechanism backwards. The move happened because the stops were there, not because someone knew yours specifically. The practical adjustment is to place stops where the trade thesis fails rather than at the obvious level, accepting a slightly larger loss in exchange for not being part of the cascade.
The strong version of the claim, that a venue or a large participant can see your individual stop and moves the price to trigger it, is almost never what happens and is not necessary to explain the observation. Clusters of stops are visible from the shape of the market without anybody seeing any individual order, because the levels where they gather are predictable from public information.
The weaker version is true and worth taking seriously: participants who can see a concentration of resting stop orders have an incentive to push toward it, because triggering the cluster produces a burst of one-directional flow they can trade against. That is not a conspiracy, it is a rational response to visible liquidity, and it happens in every market that has stops in it.
What follows is unglamorous. Placing a stop where everybody else places theirs makes you part of a predictable pool of forced orders, which is a bad place to be regardless of anybody's intent. Placing it where your reason for the trade would be wrong puts you outside the cluster, and the small extra distance is the price of not being in it.
The cascade happens because the stops were there. Nobody had to see yours; the level where you put it was public information.
A trailing stop moves in one direction as the position works, locking in progress and never retreating. It solves a real problem, which is the tendency to hold a winning position until it stops being one, and it does so without requiring a decision at the moment the decision is hardest.
The cost is that it converts a position with an open-ended outcome into one with a ratchet, and ratchets exit early in markets that move in bursts with pullbacks between them. A trailing stop tight enough to protect most of a gain will exit on an ordinary retracement in the middle of a move that continues, and one loose enough to survive those retracements gives back a large part of what it was meant to protect.
There is no setting that resolves the tension because the tension is the instrument. What can be done is to make the choice deliberately: a tight trail is a decision to take frequent modest results, a loose one is a decision to accept larger give-backs in exchange for staying in longer moves. Both are defensible and neither is the setting that avoids the trade-off.
A price stop asks whether the trade has gone wrong. A time stop asks a different and often better question: whether the trade has gone anywhere at all. A position that was entered on an expectation of movement, and has not moved after the period in which the movement was supposed to happen, has been invalidated just as surely as one that hit a level.
The argument for adding one is that capital sitting in a position that is doing nothing has a cost that never appears as a loss. It is unavailable for anything else, it continues to carry risk, and on a perpetual contract it is paying or receiving funding the whole time. None of that shows up in a profit and loss column until something happens, and frequently nothing does.
The implementation is trivial and the discipline is not: decide, when you enter, how long the thesis needs to work, and close the position when that period passes regardless of where the price is. Traders who add this usually find that the trades it removes were not the ones they would have chosen to remove, which is the point of writing the rule in advance.
Two measurements answer the question and neither requires a backtest. The first is how many of your stopped-out trades subsequently reached your intended target: if the number is high, the stop is inside normal movement and is manufacturing losses the strategy did not require. The second is how far your stops actually fill from their trigger price, which tells you what the mechanism costs on top of the placement.
The first of those is the one that changes behaviour. Most traders have never counted it, and the count is available in any trade record with an entry, an exit and a target. If a substantial share of stopped trades would have worked, the placement is wrong in a specific and fixable way; if almost none would have, the placement is fine and the losses are the strategy rather than the stop.
The distinction matters because the two problems have opposite remedies, and a trader who has not measured which one they have is as likely to make it worse as better. Widening a stop that was already outside normal movement increases every loss and improves nothing, and that is the more common of the two mistakes.
Only if the original stop was inside normal price movement, in which case it was being triggered by noise rather than by the trade being wrong. Beyond that threshold, widening raises the win rate and enlarges every loss by exactly the same amount, and the two offset. The comfort improves and the arithmetic does not.
Because stops cluster at round numbers and just beyond obvious highs and lows, and triggering a cluster produces a burst of forced orders in one direction into a book that is already thin. The extension beyond the level and the partial recovery afterwards are what that flow looks like, and nobody needed to see your order specifically.
The version where somebody sees your individual order is almost never what happens and is not needed to explain what you observed. The version that is real is that concentrations of resting stops are predictable from public information, and participants have an incentive to push toward them. The remedy is the same either way: do not place stops where everybody places theirs.
There is no such number in isolation. Win rate is one of three quantities, and a system winning three times in ten can outperform one winning nine times in ten depending on the sizes. Any figure quoted without the average win and the average loss beside it conveys almost nothing.
It solves the problem of holding a winner until it stops being one, and it costs you exits during ordinary retracements in moves that continue. There is no setting that avoids that trade-off, because the trade-off is the instrument. Choose a tight or loose trail deliberately rather than searching for the one that does both.
An exit based on elapsed time rather than price. It asks whether the trade has gone anywhere, which is a different question from whether it has gone wrong. Capital in a position that is doing nothing is unavailable, still carrying risk, and on a perpetual contract still paying or receiving funding, none of which appears as a loss.
Count how many of your stopped-out trades later reached the target you had intended. If a substantial share would have worked, the stop sits inside normal movement and is manufacturing losses the strategy did not need. If almost none would have, the placement is fine and the losses belong to the strategy.
Moving it closer to lock in progress is a different action from moving it further away, and only the second is the problem. Widening under pressure is taken at the exact moment the stop existed to protect against, by the version of you least equipped to decide, on information that feels new and usually is not.