Volatility Drag: Why Positive Average Returns Can Still Shrink an Account

Average trade return can be positive while account balance falls. The calculation need not be wrong: the method of averaging differs from how capital actually grows. The gap is volatility drag.

+50% Followed by −50% Is Not Breakeven

Start with two trades, +50% and −50%. Add and divide by two and the average is zero, but the account is below its starting point.

Returns multiply: 1.5 × 0.5 = 0.75. Each percentage acts on the then-current amount, so the later loss in this example removes more dollars. This multiplication is another side of compounding; volatility drag describes its erosion effect.

Arithmetic vs. Geometric Average

The geometric average measures actual compound growth. It is the constant per-trade return that would produce the same final result.

The gap is drag: 0% − (−13.40%) = 13.40 percentage points lost to variation. Arithmetic average is always at least the geometric average, and greater return fluctuation widens the gap. With zero variation and identical returns, they coincide.

Estimating Drag: σ²/2

A practical approximation subtracts half the return variance from the arithmetic average. Let σ be return standard deviation.

The square is the key. A small volatility increase produces a faster-growing deduction. Conversely, reducing volatility can improve growth like increasing returns. This helps explain the popularity of the Sharpe ratio, which compares return with volatility.

The Same Win Rate, Different Outcomes

Consider 100 trades with exactly 50% wins and vary only the gain/loss magnitude.

Directional skill is identical. Only amount risked per trade differs, changing results through σ². See position sizing and portfolio heat for setting exposure.

Recovery also differs. A's −39.5% requires +65.3%, calculated as 0.395 ÷ 0.605, to return to starting capital. B's −2.0% needs about +2.0%. This is the recovery asymmetry described in drawdown.

Leverage Increases Drag by Its Square

Leverage L multiplies expected return and volatility by L, but drag scales with variance and therefore L². Beyond a point, more leverage reduces compound growth.

A positive-expectancy strategy can become negative through excessive size. The Kelly criterion addresses the growth-optimal leverage, matching the 2.5× maximum above. In practice, estimation uncertainty often leads to using half that amount or less.

The calculation entirely excludes forced liquidation. A liquidation boundary can end the account before long-run drag plays out, lowering the practical leverage limit further.

Four Ways to Reduce Drag

① Reduce size per trade. Halving σ quarters drag. This is more directly controllable than raising win rate.

② Cut the large-loss tail. Drag comes from dispersion. One large loss increases σ² substantially, so stops and daily loss limits can reduce drag without targeting win rate.

③ Avoid stacking correlated assets. Five highly correlated coins can effectively be one position. Their combined volatility enters σ; what matters is whether they actually move differently, not the count.

④ Keep risk magnitude consistent. Switching between 1R and 4R increases σ even if average size is unchanged. Varying size itself has a cost.

Five Practical Checks

① Judge performance by balance change, not arithmetic average. The simple average of trade returns is not the account's growth rate.

② Record return standard deviation. The mean alone hides drag. Rising σ means more risk even when performance looks good.

③ Judge leverage changes by geometric growth. Doubling arithmetic expectancy can still reduce actual growth.

④ Keep trade risk consistent. Increasing size when confident can erode long-term growth through σ.

⑤ Check simulations. Use your win rate and payoff in Monte Carlo simulation to see how size changes the outcome distribution despite equal expectancy. Pair this with risk of ruin to inspect irrecoverable paths.

Three Key Points

① Returns multiply rather than add. A 0% arithmetic average can mean −13.4% geometric growth, which represents actual compounding.
② Drag is approximately σ² ÷ 2. Double volatility gives fourfold drag; leverage L gives L²-fold drag.
③ Even a winning strategy has a size where growth becomes negative. Smaller exposure and trimming large losses can improve growth without changing win rate.

Notice

Returns, standard deviations, leverage, and balances are hypothetical examples, not measurements from a strategy or exchange. The σ²/2 approximation works for small returns and distributions without excessive skew; larger moves create error. Fees, slippage, funding, and liquidation are excluded, so actual outcomes can be worse. Leveraged trading can lose all principal. Investment decisions and responsibility are yours.