Value at risk and CVaR
Value at risk quantifies how bad an ordinary bad day looks. With a VaR 95 % of 2 %, the daily loss stays below 2 % on 95 out of 100 trading days. What happens on the other 5 % it does not say. CVaR answers that.
- Calm profile
- Typical for portfolios with a large bond or money market allocation. On 10,000 € that is under 100 € on a bad day.
- Balanced portfolio
- The usual range for portfolios of equities and bonds. On 10,000 € that is 100 to 200 € on a bad day.
- Equity-typical
- Where broadly diversified equity portfolios sit. On roughly one trading day in twenty, the loss runs higher than this.
- Markedly elevated
- Common for single stocks, crypto or leveraged products. Check whether a few positions drive the number, and what the matching CVaR looks like.
How it is calculated
VaR 95 % = 5th percentile of daily returns (historical)
A portfolio has 500 trading days of history. Sorting every daily return by size, 25 days fall below the 5th percentile. The value at that boundary is the VaR.
- Trading days
- 500
- Worst 5 %
- 25 days
- Value at boundary
- −1.9 %
- VaR 95 %
- 1.9 %
What the number does not tell you
- It describes precisely the days that hurt least well. VaR is a threshold, not an expected value. Whether a 3 % or a 25 % loss hides behind the worst 5 % stays open. CVaR closes that gap by averaging those very days, which is why it always sits above VaR.
- The parametric version flatters. Derived from a normal distribution instead of real trading days, the figure comes out too low in calm phases. A wide gap between historical and parametric VaR means your history held more extreme days than theory expects.
- One day is not one month. A daily VaR cannot simply be scaled up. Losses cluster in crises rather than spreading evenly. Longer horizons need a separately calculated monthly VaR.
- It only knows the past. A window without a crisis delivers a harmless VaR. That is a statement about the window, not about the portfolio.
Related metrics
How Evergrova calculates it
Performance & Risk shows VaR 95 % and CVaR 95 % in the metric band, with VaR 99 %, CVaR 99 % and the parametric variant in the detail section. The stress test adds a monthly VaR.
Data basis: historical quantile of daily returns over the selected window, horizon 1 trading day.
View the demo portfolioCommon questions
What is the difference between VaR 95 % and VaR 99 %?
Both answer the same question at different strictness. VaR 95 % describes the worst day in twenty, VaR 99 % the worst in a hundred. The 99 figure therefore always sits higher and statistically hits two to three days a year.
Why is CVaR always larger than VaR?
Because it averages exactly those days that lie beyond the VaR threshold. An average of values that all exceed a boundary must itself exceed that boundary.
Can I use VaR to estimate my maximum loss?
No. VaR is a daily figure with a probability attached, not a ceiling. For how deep a portfolio fell in a real crisis, maximum drawdown is the right number.
Last reviewed: 2026-08-08