Most of the losing streaks I have looked at closely were not strategy failures. The signal was fine. The entries were fine. What changed was the weather. The market got roughly twice as loud, the stop that used to sit safely below the noise was now sitting inside it, and the trader kept using the same fixed number of points or percent because that is what the rules said. So they got tapped out on every trade, took the small loss five times in a row, and concluded the edge was gone. The edge was not gone. The stop was just too tight for the room the market was now using.
A stop is a bet about how far price can wander against you and still be inside normal behavior. When volatility roughly doubles, normal behavior doubles too, and a stop measured in fixed points quietly becomes a much tighter stop in volatility terms. Nobody moved it. The ground moved under it. So before touching entries or indicators, the first job is to notice which regime you are actually in.
Flagging the regime before you flag the trade
You do not need a fancy classifier. Two cheap measures get you most of the way, and they agree with each other often enough that you can trust the signal when they do.
The first is realized volatility, which is just the standard deviation of recent returns, annualized if you like round numbers. Take a rolling window, something like 20 bars, and compare it against a longer baseline, something like 100 bars. When the short window sits well above the long one, roughly one and a half times or more, you are in an elevated phase. This is slow and smooth, which is what you want for a regime flag. You are not trying to catch the exact turn. You are trying to name the environment.
The second is range expansion, and I like Average True Range for this because it captures the gaps and the wicks that a close-to-close return calculation misses. Compare current ATR to its own moving average. When ATR is running well above its typical level, the bars are physically bigger, and your stop distance has to respect that whether or not the closing prices look dramatic.
A rough workflow I keep coming back to:
- Compute 20-bar realized vol and 100-bar realized vol. Flag elevated when the short is roughly 1.5x the long or more.
- Compute ATR and its own 50-bar average. Flag expansion when ATR is running clearly above that average, say 1.3x or more.
- When both agree, treat it as a high-volatility regime and switch your mechanics. When only one fires, treat it as a warning and tighten your risk without fully committing to the wider settings.
- Require the regime to persist for a few bars before you act on it, so a single violent candle does not flip you back and forth.
That persistence rule matters more than it sounds. Volatility measures are jumpy near the transition, and if you toggle your whole approach on every bar you will get the worst of both. A little hysteresis, where you need a couple of confirming bars to enter the regime and a couple of calm bars to leave it, saves you from thrashing.
Widening stops and targets without widening your risk
Here is the part people get backwards. Widening your stop sounds like taking more risk, and if you widen the stop and keep the same position size, it is. The trick is that stop width and position size move together, in opposite directions, so your dollar risk stays roughly constant while your stop finally sits outside the noise.
The cleanest way to do this is to stop thinking in fixed points and start thinking in volatility units. Set your stop as a multiple of ATR rather than a fixed distance. If your stop is 2x ATR, then when ATR doubles, your stop distance doubles automatically, and you never have to remember to adjust it. The regime handles itself.
Then size the position off that distance. Decide how much you are willing to lose on the trade as a fraction of the account, something small and constant, and divide by the stop distance to get the number of units. When the stop is wider, the position is smaller. When the stop is tight, the position is larger. Same risk either way. The math is boring and that is the point.
- Fixed dollar risk per trade, chosen once and held constant across regimes.
- Stop distance in ATR multiples, so it breathes with volatility on its own.
- Position size derived from those two, never chosen by feel.
Targets need the same treatment, and this is the half people forget. If you widen the stop but keep the old fixed target, you have quietly cut your reward-to-risk ratio, because the stop grew and the target did not. In a loud regime the target should scale with the same ATR the stop uses. The move you are trying to catch is bigger now. Reaching for the same old fixed target means you leave the larger swing on the table and get a worse ratio than you did in calm markets. Scale the target with volatility and the ratio holds.
The failure modes worth naming
The most common one is the death by a thousand tiny losses I opened with. Tight stop, loud tape, tapped out repeatedly, each loss small enough to feel harmless, all of them together larger than the drawdown you were trying to avoid. If you find yourself stopped out and then watching price go your way within a bar or two, over and over, that is the regime telling you the stop is inside the noise.
The opposite failure is worse and sneakier. You widen the stop, correctly, but you forget to shrink the size. Now a single loss in a high-vol regime is two or three times the loss you budgeted for, and one bad trade does the damage of a whole cluster. This is how people blow up trying to fix the first problem. Widening the stop is only safe if the size comes down with it. They are one adjustment, not two.
A quieter failure is being too slow to leave the regime. High volatility does not last forever, and when it fades your wide stops start giving back more than they need to on losers while your smaller size caps the winners. The same persistence rule that kept you from thrashing on the way in should let you step back down to normal mechanics once realized vol and ATR have both settled for a stretch. The regime flag works in both directions or it does not really work.
None of this makes a bad strategy good. A signal with no edge in calm markets will not suddenly find one when things get loud. What regime-aware sizing does is stop a decent strategy from dying during the phase where it should have made the most money. Volatile markets are where the big moves live. The goal is to still be in the game with enough size to catch one, and small enough stops-in-volatility-terms that the noise does not bleed you out first. Measure the weather, let the stop breathe with it, and let the size shrink to keep the risk flat. The rest is just holding those three things constant while everything else moves.