Introduction

Sharpe ratio (return per unit volatility) and Sortino ratio (return per unit downside volatility) are ubiquitous in finance. Yet both have limitations. Sharpe assumes volatility is the right risk measure; Sortino ignores tail risk. This article explores advanced performance metrics that address these shortcomings, providing richer pictures of strategy risk-adjusted returns.

Limitations of Sharpe and Sortino Ratios

Sharpe Ratio Issues

Sharpe ratio assumes normally distributed returns and treats upside and downside volatility equally. A strategy with frequent small gains and rare large losses could have "good" Sharpe ratio while being undesirable. Sensitive to returns distribution shape: non-normal returns (common in finance) make Sharpe misleading.

Sortino Ratio Limitations

Sortino improves on Sharpe by ignoring upside volatility (treating only downside as risk). But it still ignores tail risk: a strategy with 95% of returns normal and 5% catastrophic losses looks acceptable on Sortino but is undesirable.

Calmar Ratio and Return Over Maximum Drawdown

Calmar ratio = annual return / maximum drawdown. Focuses on largest loss experienced, not volatility. A strategy returning 10% annually with 5% max drawdown has Calmar of 2.0; a strategy returning 15% with 20% max drawdown has Calmar of 0.75.

Advantage: incorporates drawdown severity directly relevant to investors. Disadvantage: single worst day dominates metric; doesn't account for recovery time or frequency of drawdowns.

Information Ratio and Active Management

Information ratio = (strategy return - benchmark return) / tracking error. Measures excess return per unit of deviation from benchmark. Most relevant for active managers: how much alpha do you generate relative to benchmark risk taken?

Requires clearly defined benchmark and meaningful tracking error. If tracking error is low (strategy closely hugs benchmark), even small alpha looks good. Useful for comparing managers within same style but problematic for comparing across different strategies.

Omega Ratio and Probability of Gain

Omega measures probability and magnitude of returns above a minimum acceptable return (MAR) vs below MAR. Omega = [E(max(0, R-MAR))] / E(max(0, MAR-R)). Unlike Sharpe, doesn't assume symmetry.

More intuitive: Omega above 1.0 means excess gains exceed excess losses; Omega of 1.5 means gains are 50% larger than losses. Captures non-normal return distributions better than Sharpe.

Conditional Value-at-Risk (CVaR) and Expected Shortfall

VaR (value-at-risk) says "worst 5% of days lose more than 2%." Doesn't tell you how much worse than 2% can get. CVaR (Conditional VaR) measures average loss on worst 5% of days: "worst 5% days lose average 3%." More informative than VaR.

Use CVaR as risk measure: minimize portfolio CVaR subject to return constraints. Results in portfolios more robust to tail events than Sharpe-optimized portfolios.

Kappa Ratio Family

Generalization of Sharpe and Sortino to higher moments. Kappa3 incorporates skewness (asymmetry of return distribution), Kappa4 incorporates kurtosis (tail thickness). More complex but captures distribution shape better than Sharpe.

Disadvantage: requires estimating higher moments which are noisier in limited data.

Stability and Robustness Metrics

Herfindahl Index of Returns

Measure concentration of returns: do returns cluster in few days or spread evenly? Sum of squared daily returns (normalized). High index = concentrated returns (feast/famine). Low index = consistent returns. Prefer lower index: less dependence on few big days.

Return Autocorrelation

Positive autocorrelation = good days followed by good days (persistent momentum). Negative autocorrelation = reversals (mean-reversion). For trading strategies, persistent returns are preferable to mean-reverting ones (less whipsaw).

Scenario and Stress Testing

Rather than single-number metrics, evaluate strategy across scenarios: how does it perform in 2008-2009? COVID crash? Low-vol environments? Best practices: report returns in multiple market regimes, document performance under stress scenarios, identify conditions where strategy breaks.

Multi-Objective Evaluation

Rarely does one metric capture everything. Best approach: evaluate strategies on multiple dimensions: Sharpe, Sortino, max drawdown, recovery time, tail risk (CVaR), correlation to other strategies, turnover/implementation cost. Visualize tradeoffs (Pareto frontier of strategies).

This prevents over-optimizing on single metric at cost of others.

Practical Implementation

Most backtesting platforms (QuantConnect, Backtrader) compute Sharpe/Sortino by default. Calculate additional metrics: Calmar (internal max drawdown method), Information Ratio (vs specified benchmark), Omega (specify MAR), CVaR (percentile of returns). Build dashboards showing multiple metrics.

Conclusion

Beyond Sharpe and Sortino lie richer metrics capturing important aspects of strategy risk-return profiles. Calmar and CVaR incorporate drawdown severity. Information Ratio measures relative performance. Omega captures non-normal return distributions. Best practitioners evaluate strategies across multiple dimensions, avoiding over-reliance on single metrics and understanding tradeoffs between return, volatility, drawdown, tail risk, and consistency.