Standard Deviation of Returns: The Noise Around Your Edge

Two traders with the identical average profit per trade can have wildly different experiences — one sleeps fine, the other blows a trailing drawdown. The difference isn’t their edge. It’s the noise around it.

Two systems, same expectancy, very different pain

Imagine two strategies. Both average +0.2R per trade — identical expectancy. System A wins 55% of the time at roughly 1:1. System B wins 30% of the time but its winners run 4R while its losers stay at 1R.

On paper they’re twins. In practice they’re nothing alike. System B strings together six, seven, eight losers in a row while it waits for a big winner, and each of those streaks digs a deep equity hole. System A grinds along with shallow, frequent recoveries. Same destination, radically different road — and on a prop account, the road is what kills you, because a trailing drawdown limit doesn’t care about your long-run average. It cares how deep the worst stretch goes.

That “how bumpy is the road” is exactly what standard deviation of returns measures. Expectancy tells you where you’re going; standard deviation tells you how violently you’ll be thrown around getting there.

Computing standard deviation of R per trade

Work in R — each trade expressed as a multiple of the amount you risked — so results are comparable across instruments and position sizes. A trade that made twice your risk is +2R; one that hit your stop is −1R.

The recipe:

Convert every closed trade to its R-multiple.
Compute the mean R across all trades — that’s your expectancy.
For each trade, take the difference from the mean, and square it.
Average those squared differences (that’s the variance).
Take the square root — that’s your standard deviation in R.

A quick read on the output: if your expectancy is +0.2R with a standard deviation of 1.5R, the noise per trade is over seven times the size of the signal per trade. That’s not a broken system — it’s normal, and it’s precisely why single trades and short streaks tell you almost nothing.

How variance drives drawdown depth

This is the connection that matters for staying funded: drawdown depth is driven far more by variance than by expectancy. Two systems with the same edge but different standard deviations will have very different worst-case equity dips, and the higher-variance one will breach a drawdown limit that the lower-variance one clears comfortably.

The intuition: higher standard deviation means longer and deeper losing runs are not just possible but expected. A system that’s dead-on profitable over 500 trades can still carve a drawdown deep enough to end a challenge somewhere in the middle — purely from variance, with no change in the underlying edge.

For prop traders this reframes the whole risk problem. Your enemy isn’t a losing strategy; it’s a winning strategy having an ordinary bad run at the wrong moment. Model how your per-trade variance translates into equity dips with a prop firm drawdown calculator, and read up on how trailing drawdown mechanics can amplify a rough stretch — the trail ratchets up on your highs and then measures your dip from that peak, so a high-variance equity curve gives the trail more to work with against you.

The signal-to-noise view of an edge

Borrow a concept from engineering: signal-to-noise ratio. Your edge (expectancy) is the signal. Your standard deviation is the noise. What you actually care about is how loud the signal is relative to the noise — because that ratio determines how many trades it takes before skill becomes visible through the randomness.

High signal-to-noise (strong edge, low variance) — the equity curve is smooth, the edge shows up fast, drawdowns are shallow. Rare and precious.
Low signal-to-noise (real but small edge, high variance) — the edge is genuine but buried under noise. It takes a large sample to see it, and the ride is punishing.

Most profitable retail and prop strategies live in the second bucket: a real edge, badly obscured by noise. That’s why small samples lie, why a hot streak proves nothing, and why you need error bars on your win rate. The noisier the system, the more trades you need before your results mean anything — and the more a confidence interval on your live edge earns its keep. Shibiki tracks that interval per strategy precisely because a raw average, on a high-variance system, is confidently wrong most of the time.

Reducing variance without killing expectancy

You can lower the noise — but only some methods keep the signal intact. The goal is to raise signal-to-noise, not just flatten the curve.

Methods that usually help:

Consistent position sizing. Wildly varying size is a massive, self-inflicted variance source that adds zero expectancy. Fixing your risk per trade is the single cleanest variance reduction there is — size every entry off a position size calculator rather than gut feel.
Trading more setups (if uncorrelated). More independent trades smooth the curve without touching per-trade edge — the square-root law working for you.
Taking partial profits — genuinely lowers variance, though it also trims expectancy, so measure the trade-off rather than assuming it’s free.

Methods that quietly destroy expectancy:

Cutting winners short to “lock in” gains. Feels like lower variance; actually amputates the fat tail that carries your edge.
Skipping trades after losses. You’ll disproportionately skip the winners that end the streak.

The test for any variance-reduction tweak: did expectancy survive? If smoothing the curve dropped your edge more than it dropped your noise, you made the system worse while making it feel better.

Volatility as an input to position sizing

Finally, standard deviation shouldn’t just be measured — it should feed back into how much you risk. A higher-variance system demands smaller risk per trade to keep the same probability of surviving a drawdown limit. This is the honest version of position sizing: size to the noise, not just the edge.

Practical loop:

Measure your standard deviation of R over a meaningful sample.
Stress-test the worst plausible losing run it implies against your firm’s drawdown room (confirm the exact limits with your firm — they vary and change).
Set risk per trade so that ordinary bad luck can’t breach the limit — the higher the variance, the smaller the risk.
Re-measure periodically. Variance drifts as markets and your behavior change.

There’s a hard-limits angle here too: knowing your variance is one thing, not exceeding your per-trade risk in a heated moment is another. Enforcing max risk at the broker — a cap that holds even when you’re tempted to double down mid-streak — is what keeps a high-variance system from turning an ordinary drawdown into an account-ending one. Measure the noise, then remove your ability to make it worse.