How is CVaR different from VaR?

VaR provides the boundary cutoff for losses at a given confidence level (for example, "losses will not exceed $1,000 on 95% of days"). CVaR tells you the average severity of losses when that cutoff is breached (for example, "on the 5% of days when losses exceed $1,000, the average loss is $2,400").

What is the formula for calculating CVaR?

In historical simulation, CVaR is calculated as the arithmetic mean of all returns that fall below the (1 - confidence level) percentile cutoff. Formally, CVaR_alpha = E[Loss | Loss >= VaR_alpha].

Why is CVaR considered a coherent risk measure?

CVaR satisfies all four mathematical axioms of coherent risk measures: translation invariance, sub-additivity (diversification reduces or maintains risk), positive homogeneity, and monotonicity. Standard VaR fails the sub-additivity requirement under non-normal return distributions.

What is a good CVaR level for a trading strategy?

A good CVaR depends on account sizing and risk tolerance, but generally CVaR should remain close to the VaR threshold. When CVaR is several multiples higher than VaR, it indicates severe fat-tail risk and large outlier losses in the strategy.

Backtesting guide

Conditional Value at Risk (CVaR) & Expected Shortfall

Q: What is Conditional Value at Risk (CVaR)?

Conditional Value at Risk (CVaR), or Expected Shortfall, is the average of all losses that exceed the Value at Risk threshold at a specified confidence level, such as 95% or 99%.

Conditional Value at Risk (CVaR), also referred to as Expected Shortfall (ES) or Average Value at Risk (AVaR), is an advanced risk assessment metric that quantifies the expected average loss of an investment or trading portfolio strictly within the worst-case percentile tail beyond the Value at Risk threshold.

Read the practical guide

Key Takeaways

CVaR calculates the average loss on the days or trades when losses exceed the Value at Risk (VaR) cutoff.
Unlike standard VaR, CVaR satisfies the mathematical condition of sub-additivity, classifying it as a coherent risk measure.
At the 95% confidence level, VaR identifies the loss threshold exceeded 5% of the time, while CVaR calculates the average loss across that worst 5% of outcomes.
Basel III banking regulations and quantitative portfolio managers prioritize CVaR over VaR for tail risk modeling.

What is Conditional Value at Risk (CVaR) and its meaning

Conditional Value at Risk addresses the primary vulnerability of Value at Risk: blindness to the magnitude of tail losses. If a trading strategy has a 95% daily VaR of $2,000, VaR tells you that on 5% of trading days, losses will exceed $2,000. However, it does not reveal whether those tail losses average $2,100 or $15,000. CVaR calculates the conditional expectation of loss given that the loss is greater than or equal to the VaR threshold.

Expected Shortfall: Measures the expected value of losses conditional on exceeding the quantile cutoff.
Tail Severity: Captures extreme downside events, liquidity dry-ups, and market gap risks that standard VaR ignores.
Regulatory Standard: Adopted by the Basel Committee for banking book and trading book market risk calculations.

Conditional Value at Risk formula and calculation steps

To calculate historical CVaR from an empirical distribution of returns, identify all return observations that fall below the (1 - alpha) percentile, and compute their arithmetic mean. For continuous parametric distributions, CVaR integrates the loss distribution density across the entire tail region beyond the VaR quantile.

Mathematical Definition: CVaR_alpha = E[Loss | Loss >= VaR_alpha].
Historical Calculation Step 1: Rank all trade or daily returns from worst to best.
Historical Calculation Step 2: Determine the VaR cutoff index corresponding to the chosen confidence level (such as the bottom 5% of observations).
Historical Calculation Step 3: Average all losses that equal or exceed the VaR cutoff value.

VaR vs CVaR: Key differences and comparative advantages

Comparing VaR and CVaR highlights why quantitative risk management has shifted toward Expected Shortfall. VaR is a quantile (a single point on the distribution curve), whereas CVaR is an expectation (the average of all points in the tail). CVaR is convex and sub-additive, meaning the CVaR of a combined portfolio is always less than or equal to the sum of the individual CVaRs of its components, incentivizing true diversification.

Metric Type: VaR is a threshold cutoff; CVaR is an average conditional loss.
Sub-additivity: VaR can penalize diversification in non-normal distributions; CVaR is always sub-additive.
Sensitivity to Outliers: CVaR directly responds to extreme outlier changes in the tail; VaR remains unchanged if an extreme loss becomes worse without crossing the threshold count.

Using CVaR in algorithmic strategy backtesting

In strategy backtesting, strategies that rely on tight stop-losses with high win rates often hide catastrophic left-tail risk. For instance, short option selling or martingale-style position scaling can show consistent gains alongside a mild 95% VaR, but an extreme CVaR. Reviewing CVaR alongside maximum drawdown ensures that the strategy can survive black swan market dislocations.

Audit whether strategy edge is dependent on unhedged left-tail exposure.
Size positions such that account capital can withstand multiple consecutive average CVaR loss events.
Evaluate whether parameter changes reduce both peak drawdown and Expected Shortfall.

Direct Pineify workflow

Evaluate Expected Shortfall and tail loss distribution

Pineify Backtest Deep Report calculates Conditional Value at Risk (CVaR), return skewness, and trade loss distribution from your TradingView trade export CSV.

Backtest Deep Report

Upload your TradingView CSV export to audit Expected Shortfall, tail risk, and trade distribution metrics.

Analyze strategy CVaR

Boundary: Backtest Deep Report analyzes historical trade logs; past tail metrics do not eliminate market execution or liquidity risk.

Primary sources

R. Tyrrell Rockafellar and Stanislav Uryasev, Optimization of Conditional Value-at-RiskVerified 2026-09-02
SEC Investor.gov, Performance ClaimsVerified 2026-08-16

Related Resources

Value at Risk (VaR) Guide Tail Risk in Trading Downside Risk & Semi-Deviation Risk-Adjusted Return Metrics Backtest Deep Report

This page is educational and does not provide investment advice. Backtests are hypothetical, depend on their data and assumptions, and do not guarantee future results. Trading can result in substantial loss.