A stop loss is a price level that closes a position automatically once price reaches it, and it exists to answer one question: at what point does this trade idea stop being true? It isn't a safety net bolted on after the fact. On a position that's actually been thought through, the stop is decided at the same moment as the entry, before either number gets acted on.
Mechanically, a stop-loss is just a conditional order sitting on the book: once the market trades at the trigger price, it becomes a market order and fills at the next available price, not necessarily the exact stop price itself. That gap matters more in fast-moving or illiquid conditions than it does on a deep market like EUR/USD, but it's part of why a stop is a plan for an exit, not a guarantee of one.

Why the stop isn't a round number
A stop placed at "50 pips away" or "2% of the account" is a risk decision wearing the clothes of a market decision. It says nothing about the chart.
A real stop sits at the price where the original reason for the trade is actually disproven: past the level that would have to hold for the setup to still make sense. On a EUR/USD position built from a Fibonacci retracement on the last completed swing, that's beyond the next retracement level out, past the point where the swing itself would be invalidated if price kept going. Corporate Finance Institute's rundown of stop-loss mechanics covers the same basic idea in general terms: the stop is supposed to represent an acceptable, pre-decided loss, not a reaction once the position is already underwater.
What happens when it's hit
The position closes at that exact price, the loss is recorded, and nothing about it gets softened or explained away after the fact. A stop doing its job isn't a mistake: it's the mechanism working exactly as designed, at the exact price it was supposed to.
That's also why the loss gets published with the same detail a winning call gets. A stop hit is information, not something to bury.
The only move a stop is allowed to make
Once a position is live, the stop can move in exactly one direction and for exactly one reason: to breakeven, once the trade has moved far enough in its favor to justify locking in a scratch instead of risking the original loss. When that happens, it's logged as a new, dated event on the position with a public explanation attached, not a quiet redraw of the original plan.
What a stop never does here is get moved further away to "give the trade more room." Widening a stop after the fact is the single most common way a disciplined plan turns into hope. If the original level was right, it didn't need to move. If it was wrong, the position should have already closed.
What a stop loss doesn't do
It doesn't cap your loss at a comfortable-sounding percentage, and it doesn't replace deciding how much to actually risk on the trade. Position sizing is a separate question, answered by your own account and your own tolerance for drawdown, not by the stop's distance from entry.
A stop 40 pips away and a stop 15 pips away can carry the exact same dollar risk, depending entirely on how much you size into each one. That's exactly why conviction grades exist alongside every stop the desk publishes: a relative signal for how much confidence sits behind the level, never a lot size or a dollar figure. Reading the stop correctly alongside the rest of a published trade plan is what turns a single number into something actually useful.
The stop does its job and the loss is published with the same write-up a winner gets, explaining whether the read was wrong or the market simply moved first.
A stop that's tight because it's measured off a real swing, not tight because someone wanted a small number, is the difference between a stop that protects a real thesis and one that just gets clipped by ordinary noise before the trade ever had a chance to work.