I used to size every trade the same. One percent of risk, every time, no matter what the market was doing. It felt disciplined. It felt like the responsible thing to do. And for a long stretch it worked well enough that I never questioned it. Then I went back and looked at where the account actually made and lost money over a couple of years, and the pattern was uncomfortable. The winners were clustered in a handful of clean trending stretches. Everything in between was a slow, grinding drawdown that the big weeks papered over.
That is the quiet problem with flat sizing. It treats a market that is trending cleanly and a market that is chopping sideways as the same opportunity. They are not. A trend gives you follow-through, so a good entry keeps paying you after you are in. Chop gives you a fill and then yanks it back. If you bet the same size in both, you are overbetting the environment that punishes you and underbetting the one that rewards you.
Why flat sizing bleeds across regime changes
Think about what a fixed one percent actually does across a full cycle. In a trend, your edge per trade is higher and your trades tend to run, so a fixed bet leaves money on the table. You had a green light and you drove at the same speed as everyone stuck in traffic. In chop, your edge per trade collapses, whipsaws eat your stops, and that same fixed one percent is now too much. You are paying full freight for trades that mostly break even before costs.
The damage is not one dramatic loss. It is the accumulation. A hundred slightly-too-large bets in a hostile regime, each one costing you the spread and a little slippage and the occasional stop-out, adds up to a real hole. Meanwhile the trending weeks that should have carried the year get sized like ordinary weeks. The average masks it. You look at your win rate and it seems fine, and you never notice that your sizing was working against the shape of your own returns.
Position sizing is the one lever that touches every single trade you take. Get it slightly wrong in a systematic direction and the drag compounds. That is why this matters more than most people give it credit for. You can have a decent entry signal and still hand most of the gains back through sizing that ignores the environment.
A concrete multiplier keyed to regime state
The fix is to keep a base risk unit and then scale it by a multiplier that depends on the regime. Your base stays the same. The regime decides how much of it you actually deploy. Something like this is a reasonable starting frame, and you should treat the numbers as a shape to tune, not gospel:
- Strong trend, low-to-normal volatility: multiplier around 1.3 to 1.5. You have follow-through and the noise is not eating you. Press.
- Trend, but elevated volatility: multiplier around 1.0. The direction is your friend but the swings are wide, so hold at base and let the volatility term keep your dollar risk sane.
- Range or unclear regime: multiplier around 0.5 to 0.7. Reduce. You are trading through friction, so pay for less of it.
- High volatility with no clear direction: multiplier around 0.3, or flat. This is the environment that quietly does the most damage. Size like you expect to be wrong.
The other half of this is volatility scaling underneath the regime multiplier. Even within a single regime, a market breathing twice as wide should get roughly half the position for the same dollar risk. So the full shape is base unit, times regime multiplier, times a volatility adjustment that targets constant dollar risk. Practitioners often anchor the volatility piece to something like average true range or realized volatility over a lookback window. The regime multiplier then presses or pulls back on top of that stable base.
How you classify the regime is its own decision, and you have options. A long moving-average slope and price above or below it gives you a crude trend read. The gap between a short and long ATR tells you whether volatility is expanding or contracting. Some people lean on an ADX-style directional strength reading. The specific detector matters less than being honest about what state you are in and letting that state, not your mood, set the multiplier.
Not overreacting to one noisy print
Here is where most people break their own system. They wire up a regime detector, it flips to trend on a single strong bar, and they immediately size up into what turns out to be a fakeout. Now the regime tool is not protecting them, it is amplifying the exact noise it was supposed to filter.
The whole point of regime awareness is to respond to a change in character, and character does not change on one candle. So you have to build in some stickiness. A few ways to do that, and I would use more than one:
- Require confirmation. The regime has to hold for two or three bars, or a couple of days on a daily system, before the multiplier moves. One print does nothing.
- Smooth the input. Feed your detector a smoothed value rather than the raw reading, so a single spike cannot flip the state on its own.
- Step, do not jump. Move the multiplier toward its new target gradually instead of snapping from 0.5 to 1.5 in one trade. If the regime is real, you will get there. If it was noise, you never overcommitted.
- Add hysteresis. Use a higher threshold to enter a regime than to leave it, so you are not flip-flopping around a single line every time the reading wobbles across it.
The tradeoff is honest and worth stating. Every filter you add to avoid whipsaw also makes you slower to catch a real regime change. You will size up a little late into the best trends and hold reduced size a little too long when chop finally ends. That lag is the price of not getting faked out constantly. In my experience it is a price worth paying, because the fake regime changes vastly outnumber the real ones, and the cost of overreacting is immediate while the cost of a small lag is mild.
One more failure mode to name. Do not let the multiplier and the stop compound in the wrong direction. If you size up in a trend and also widen your stop because volatility rose, you can accidentally double your dollar risk without noticing. Keep the dollar-risk target fixed and let the multiplier express itself through position size against that target, not on top of it.
How to actually run it
If you want to test this without rebuilding your whole stack, do it in the least glamorous way possible. Take a strategy you already trade, keep every entry and exit identical, and change only the sizing. Run it once flat and once regime-aware over the same history, and compare the equity curves and the drawdowns, not just the totals. What you are looking for is whether the regime version gives up less in the bad stretches, because that is where flat sizing does its quiet damage. If you have a backtesting setup that lets you tag each trade with its regime state, the comparison gets a lot easier, which is part of why I built regime tagging into the tooling at Blockcircle in the first place.
A couple of things I would check before trusting the result. Make sure the volatility and regime inputs only use data available at the time of the trade, because it is very easy to leak the future into a regime label and convince yourself the whole thing works. And watch the turnover. If the multiplier is changing on nearly every trade, your stickiness is too loose and you are just adding noise with extra steps.
None of this makes a mediocre strategy good. It will not save a broken edge. What it does is stop a decent edge from handing back its gains to the environment. You still need something that works. Regime-aware sizing just makes sure you are betting more of it when the market is paying and less of it when the market is charging you for the privilege of showing up.