I keep coming back to trailing stops because they are the part of a system where good backtests go to die. You test an entry, it looks great, and then you bolt on a trail and half your edge evaporates. Not because the trail is bad, but because every trailing method is a bet on how the market moves after you are in, and if that bet is wrong the trail either strangles your winners or gives back too much before it triggers. The four I actually use are the ATR trail, the chandelier exit, moving-average trails, and swing-structure trails. They are not interchangeable, and picking the wrong one for the entry is one of the quieter ways to lose money.
The tension underneath all of them is the same. A tighter trail locks in more profit per trade and cuts your average loss, but it also stops you out of the rare monster move that pays for a whole quarter of small losers. A looser trail lets the big winner run but gives a chunk of it back on every trade and sits through drawdowns that feel awful in real time. There is no setting that wins on both. You are choosing which failure mode you can live with.
ATR trails and the chandelier are cousins
The ATR trail is the workhorse. You take the Average True Range over some lookback, multiply it by a factor, and trail your stop that distance behind price. Because it scales with volatility, the stop breathes. It widens when the market gets jumpy and tightens when things calm down, which is exactly what you want if you are trying to stay in a move without getting shaken out by normal noise. A common starting point is a 14-period ATR with a multiplier somewhere around 2.5 to 3.5. Below roughly 2 you are trading a lot of intraday wiggle for stop-outs. Above roughly 4 you are basically holding through anything short of a real trend break.
The chandelier exit is an ATR trail with one important twist. Instead of trailing from the current price, it trails from the highest high since you entered, hanging the stop down from that peak like a chandelier from the ceiling. The practical effect is that it ratchets. Once price makes a new high the stop moves up, and it never moves down. This makes it noticeably stickier on the way up than a naive price-based ATR trail, and it is my default for clean trend-following on daily bars. Typical settings are a 22-period ATR with a 3x multiplier, which is close to where the method was originally described. The failure mode is the same as any ATR method though. In a sharp V-shaped reversal, the stop is anchored to a high that price blew past, so you give back the full ATR-times-multiplier distance before you are out.
Moving-average trails trade lag for smoothness
A moving-average trail is dead simple. You exit when price closes back through a moving average, or you trail your stop to the value of the average itself. The appeal is that it filters noise for you. Price can poke below intrabar all it wants, and as long as it closes back above your line you are still in. A fast average, say a 10 or 20 period, keeps you close to the action and gets you out early on a stall. A slow one, 50 or higher, sits through corrections that would trigger any ATR trail, which is great when the trend resumes and infuriating when it does not.
The catch with any moving average is lag. The average is a smoothed version of where price already was, so by the time price closes decisively through it you have often given back more than a volatility-based trail would have cost you. Moving averages also behave badly in chop. When price is oscillating around the line you get a string of whipsaw exits, each one a small loss, and there is no volatility term to widen the stop and save you. If your entries fire in ranging conditions, a moving-average trail will quietly bleed you.
Swing-structure trails follow the chart, not a formula
The last one is the most discretionary and, in trending markets, often the best. You trail your stop just below the most recent higher low in an uptrend, or above the most recent lower high in a downtrend. You are letting the market's own structure define the stop instead of a number. As long as the trend keeps printing higher lows, your stop keeps climbing and you stay in. The moment structure breaks, you are out, usually with more of the move captured than a fixed-distance trail would have given you.
The problem is that swing points are only obvious after they form, and defining a valid swing takes judgment or a fractal rule, which is harder to automate cleanly. In fast, low-pullback moves there may be no fresh higher low for a long time, so your stop sits far below price and you carry a lot of open risk. And in chop there is no clean structure at all, so the method has nothing to grab onto.
Matching the trail to the entry
Here is the rough mapping I use, and it comes down to how long you expect to hold and how the entry was generated.
- Momentum and breakout entries: use a chandelier exit or a tight ATR trail, multiplier around 2.5 to 3. You are trying to catch an impulsive move and get out before it gives everything back, so let the ratchet do its work.
- Position and trend-following entries: use a swing-structure trail or a slow moving average. You want to sit through pullbacks that a tight trail would exit, and you are willing to give back more per trade to stay in the one that runs for months.
- Anything firing in a range: reconsider trailing at all. A fixed take-profit or a time-based exit usually beats a trail in chop, because every trailing method assumes a trend to follow and there is not one.
One habit that saves real money: before you trust any trail, run it two ways in a backtest, once on a clearly trending sample and once on a clearly choppy one, and look at the two equity curves separately. A blended number hides the fact that the same trail can be brilliant in one regime and a slow leak in the other. When I test exit variants inside Blockcircle, the split view is the thing I actually look at, because the average across regimes lies to you.
If you take one thing from this, make it the volatility scaling. A trail measured in raw points or a fixed percent will be too tight on a volatile name and too loose on a quiet one, and you will blame the strategy when the real problem was that the stop never fit the instrument. Start with an ATR-based distance, size the multiplier to how long you intend to hold, and only reach for structure or a moving average once you know what you are giving up to get the smoother behavior.