Research
Drawdown vs Volatility: Which Actually Predicts Strategy Survival
Volatility is what statisticians measure; drawdown is what investors live through. The two metrics tell different stories — and the case for drawdown as a primary survival metric is stronger than the textbook implies.
Two strategies. Same expected return. Same annualized volatility. One has a maximum drawdown of 12%. The other has a maximum drawdown of 35%. Are they equivalent risk?
Textbook portfolio theory says yes. Practical experience — and the math of how compounding works — says emphatically no.
What volatility actually measures
Volatility is the standard deviation of period returns. It assumes returns are independent across periods (no autocorrelation) and treats positive and negative deviations symmetrically. Under those assumptions, volatility fully describes the risk distribution.
Both assumptions fail in real strategies.
What drawdown captures that volatility doesn't
Maximum drawdown is a path-dependent metric. It captures the cumulative effect of a string of bad observations clustered in time — the thing investors actually live through, the thing that triggers redemptions, and the thing that compounds against you when leverage is applied.
Two strategies with identical volatility but different return autocorrelations have different drawdown profiles. A strategy with positive autocorrelation in losses (losses cluster) will have deeper drawdowns than its volatility predicts. Trend-following is the canonical example — drawdowns of 25% are common in a strategy with 12% annualized volatility.
The compounding asymmetry
A 50% drawdown requires a 100% gain to recover. A 20% drawdown requires 25%. The recovery math is non-linear and brutal.
This is the strongest case for drawdown as a primary risk metric: it directly measures the amount of theoretical compounding that has been destroyed. Volatility doesn't.
When volatility is the right primary metric
- Mean-reversion strategies with low return autocorrelation. Drawdown closely tracks volatility.
- Symmetric strategies where upside and downside variability are similar in scale.
- Position-sizing decisions where you need a forward-looking estimate. Volatility forecasts are more stable than drawdown forecasts.
- Comparing strategies on the same risk basis. Volatility-targeted comparison is fairer than drawdown-targeted comparison because drawdown is sensitive to a single observation.
When drawdown is the right primary metric
- Asymmetric strategies where loss clusters matter — trend, momentum, options-selling.
- Allocator-facing reporting. Allocators ask about worst-case experience, not theoretical variance.
- Leverage decisions. Leverage amplifies drawdown faster than volatility because of the compounding asymmetry.
- Capital adequacy / kill-switch design. Stop-loss thresholds should reference drawdown, not volatility.
What we recommend
Use both. The single-metric debate is a distraction. Sentivue's internal evaluation rubric requires:
- Annualized volatility for risk budgeting and vol-targeting.
- Maximum drawdown and recovery time for capital adequacy.
- Calmar ratio as a single number for allocator-facing reporting.
- Sortino ratio for asymmetric strategies where upside skew matters.
A strategy that scores well on Sharpe but has unacceptable drawdown or vice versa is rejected at allocation review, regardless of how good the headline number looks.
Practical takeaways
- Volatility predicts what statisticians worry about. Drawdown predicts what investors do.
- The deeper the drawdown, the more non-linear the recovery cost. This is structural and unavoidable.
- Position-size on volatility; report on drawdown; capitalize on worst-case drawdown × 2.