NVDA, second year of trading. I bought 40 shares at $118.40 on a breakout above a level I'd marked three days earlier. My stop was $115.90. Simple plan, written down before I clicked buy: if it closes below $115.90, I'm out, no discussion.
It closed at $115.70 that day. I didn't sell.
I told myself the close was still inside the wick from earlier in the session, that the level hadn't really broken, that I'd give it until the next candle. Three days later I sold at $109.80. That's $340 on 40 shares, on a trade where my own plan had capped the loss at $100. I didn't lose $100. I lost $340, and the extra $240 came entirely from a decision I made after I already knew the rule I was breaking.
I could tell you what a stop-loss is in one sentence, and had been able to since about week two of trading. That sentence did nothing for me on the day that mattered.
What a stop-loss actually is
A stop-loss is a standing order sitting on your broker's books, waiting. You tell it: if the price touches this number, sell, automatically, without asking me again. It's not a note in your head. It's not a level you "plan to sell at." It's an instruction the exchange executes whether you're watching the screen or asleep.
There are two common forms. A stop-market order triggers at your price and then sells at whatever the next available price is, which might be a few cents worse in a fast market. A stop-limit order triggers at your price and then only sells at that price or better, which sounds safer until a stock gaps through your level and your limit order just sits there unfilled while the price keeps falling. Most beginners should use a stop-market on liquid names like NVDA or SPY. The few extra cents of slippage matter far less than the risk of an unfilled order in a stock that's dropping fast.
That's the entire mechanism. Set a price, the order lives on the books, it fires without your permission when the price gets there. Anyone can learn this in the time it takes to read this paragraph.
The part the definition doesn't cover
Here's what nobody tells you in year one: knowing what a stop-loss is has almost nothing to do with keeping one. Those are two different skills, and the gap between them is where most beginner accounts actually bleed out.
A stop-loss only protects you if it sits on the exchange's books, untouched, before the trade goes red. The moment you can still reach it — the moment it's a mental level instead of a live order, or a live order you have the ability to cancel — it stops being a stop-loss and becomes a suggestion you're free to argue with. And you will argue with it, because the argument only happens once real money is already losing, and that's exactly the moment your judgment is worst.
On the NVDA trade, my stop wasn't even a phantom mental level. It was a real order, sitting there, ready to fire. I canceled it manually about ninety seconds after the close I didn't like. The rule existed. The tool existed. I removed both, by hand, in under two minutes, because a red number on a screen is a different experience than a red number in a backtest.
Why the moment beats the rule
The plan gets made when you're calm, looking at a chart with no position on, no money at risk, thinking clearly about where the level breaks. The decision to abandon the plan gets made when you're not calm, staring at a number that's costing you real dollars every second you look at it. Those are two different people making two different decisions, and the second one has access to a cancel button the first one never accounted for.
That's the mechanism behind "I'll give it until the next candle." It sounds like patience. It's actually your brain negotiating with a loss that already happened, using hope as the currency. Nobody moves a stop when the trade is green. It only happens once the price has already proven the original plan wrong, which means the version of you doing the negotiating is the version with the worst information and the most emotional stake in being right.
The math on the NVDA trade
Plan: buy 40 shares at $118.40, hard stop at $115.90. Max loss if the plan holds: $2.50 a share times 40 shares, $100, plus whatever slippage a stop-market order eats on a bad tick.
What happened instead: I canceled the stop after the $115.70 close, held through two more red days, and sold at $109.80 on the third day when I couldn't stomach watching it anymore. Loss on 40 shares: $8.60 a share, $344. The plan capped my downside at $100. My decision to override the plan cost me $244 more than the plan ever exposed me to, on a trade where the market didn't do anything unusual — it just kept going the direction it was already going when my stop first got hit.
I didn't lose that $244 because I read NVDA wrong. My entry logic and my stop level were both fine; the level held for two more weeks after I finally sold, which tells you the original plan wasn't even bad. I lost it because I had the physical ability to cancel an order that was working exactly as designed.
Why beginners ignore a rule they can recite
Ask any beginner what a stop-loss is and they'll get it right. Ask them three months later why their last five losing trades all ran bigger than planned and you'll hear some version of what I told myself on NVDA — the level hadn't really broken, it was just a wick, one more candle would tell the real story. None of that is a knowledge problem. It's what happens when the person setting the rule and the person enforcing it in the moment turn out to be two different versions of you, and only one of them had access to a cancel button.
The fix people usually reach for is more discipline. Journal the trade, write the rule down twice, promise yourself next time will be different. I tried all of it for about a year and a half. It worked until the next time a real position went red at 2pm on a Tuesday and canceling an order took one click and about four seconds.
What actually changed it for me
What worked wasn't a better rule. It was removing my own access to the cancel button. I use Alertsify now, which copies the entries and exits of a trader I follow directly into my account. The stop gets placed the way it was planned, and I'm not the one holding the trigger on whether it stays there once the trade turns red. I still mark my own levels, still have opinions on where NVDA or SPY should hold, still think a trade through before I'd ever consider taking it. What I don't have anymore is the two-minute window where a losing position and a cancel button sit in the same hand.
That's the actual gap this closes. Not the definition — I had that in week two. The moment where knowing the rule and needing to act on it under pressure turn out to require two different skills, and only one of them can be taught from a textbook.
The honest limits here
A stop-loss doesn't guarantee your exit price. In a fast-moving stock, a stop-market order can fill worse than your trigger level, sometimes meaningfully worse. Copy-execution doesn't fix a bad entry, a bad level, or a stock that gaps overnight past every stop that existed. Nothing here removes risk. It removes one specific failure point — the human hand reaching for the cancel button once a position is already red — and that's a narrower claim than it might sound like. Trade size that actually matches what you can afford to lose on a single bad print, regardless of who or what is placing your orders.
Where that leaves me
I still think about that NVDA trade sometimes, mostly because the numbers were so clean. A hundred dollars planned, three hundred forty-four taken, with nothing in between except a decision I made with my thumb over a mouse button. The level held two weeks later. I'd already been wrong about my own ability to leave a working order alone, which turned out to be a more expensive thing to be wrong about than the trade itself.
These days my account copies a trader I follow through Alertsify instead of me managing my own stops in real time — it didn't teach me anything new about what a stop-loss is, it just took away the moment where knowing wasn't enough. If you want to see what that actually looks like:
Disclosure: that's an affiliate link — I may earn a commission if you sign up for a paid plan, at no extra cost to you. There's a free trial if you want to look around first.