What Is Expected shortfall?

Expected Shortfall (ES), also known as Conditional Value at Risk (CVaR), Average Value at Risk (AVaR), or Expected Tail Loss (ETL), is a risk-adjusted return measure that quantifies the average loss in the worst *q*% of scenarios. While Value at Risk (VaR) answers “what is the most I can lose at a given confidence level?”, Expected Shortfall goes further and asks “if losses exceed that VaR threshold, how bad are they on average?” This makes it a true tail risk measure — it looks beyond the cliff edge rather than merely identifying where it is.

Expected Shortfall is a coherent risk measure, meaning it satisfies the mathematical properties of sub-additivity, monotonicity, positive homogeneity, and translation invariance. VaR, by contrast, is not sub-additive: two portfolios can each have acceptable VaR individually, yet their combination can have a worse VaR than the sum of the parts. This theoretical superiority is one reason the Basel Committee’s Fundamental Review of the Trading Book (FRTB) framework shifted bank capital requirements from VaR at 99% to Expected Shortfall at the 97.5% confidence level.

For algorithmic trading strategy evaluation, a useful derived metric is the Return-to-CVaR ratio — the annualised return divided by the Expected Shortfall at a chosen quantile (typically 95% or 99%). This ratio directly measures how much return a strategy earns per unit of tail risk assumed, and it is far more informative than raw Sharpe for strategies with skewed or fat-tailed return distributions. Strategies that look attractive on a Sharpe basis can reveal hidden fragility when measured against CVaR, because occasional catastrophic losses drag the Expected Shortfall up even if day-to-day volatility appears benign.

Formula

ES_α = -(1 / (1-α)) × ∫ from α to 1 of VaR_γ dγ

In discrete terms: sort returns from worst to best, take the worst (1−α)% of them, and compute their average. For example, ES at the 95% level is the average of the worst 5% of returns.

Pros

  • Captures the severity of tail losses, not just their threshold

  • Coherent risk measure (sub-additive), unlike VaR

  • Adopted by Basel FRTB for bank capital requirements

  • Return-to-CVaR ratio is more informative than Sharpe for fat-tailed strategies

Cons

  • Requires more data than VaR to estimate accurately, especially at extreme quantiles

  • Harder to backtest reliably because tail events are rare by definition

  • Estimation can be sensitive to distributional assumptions (historical vs. parametric vs. Monte Carlo)

  • Less intuitive to communicate to non-technical stakeholders than simple drawdown metrics

Expected Shortfall on Wikipedia.

Man Group — Covering Your Tail: The Case for Expected Shortfall.

QuantInsti — CVaR calculation in Python and Excel.

See also