Risk Metrics · 6 min read

What Is Value at Risk (VaR)? A Plain-English Guide

VaR puts a dollar figure on a bad day: the loss your portfolio is unlikely to exceed at a given confidence level. Here is how it works, how to read it, and why CVaR matters too.

Value at Risk — almost always written VaR — answers a deceptively simple question: how bad could a normal bad day be? More precisely, it estimates the loss your portfolio is unlikely to exceed over a set period, at a chosen confidence level.

A one-day 95% VaR of $4,200 on a $100,000 portfolio means: on 95 out of 100 typical days, you should not lose more than $4,200. On the other 5 days, you might — and VaR alone will not tell you how much.

The three ingredients of any VaR number

Change any one of these and the number changes, so a VaR figure is meaningless without its horizon and confidence level attached. "$4,200" is noise; "one-day 95% VaR of $4,200" is information.

How VaR is actually calculated

There are three mainstream approaches. The historical method replays the portfolio across real past returns and reads off the relevant percentile. The parametric method assumes returns follow a known distribution (often normal) and computes the cutoff from the mean and volatility. The Monte Carlo method simulates thousands of possible futures and measures the loss distribution directly.

Each makes different assumptions, so they rarely agree to the dollar. The historical method captures real-world fat tails but is anchored to whatever happened in the lookback window. The parametric method is fast but understates extreme moves. Monte Carlo is flexible but only as good as the model driving the simulation.

Where VaR quietly misleads you

VaR is a threshold, not a ceiling. It tells you the edge of the cliff but nothing about the drop beyond it. Two portfolios can share an identical 95% VaR while one loses a manageable amount in the worst 5% of days and the other is wiped out. That blind spot is exactly what Conditional VaR measures.

VaR also assumes the future resembles the recent past. During volatility spikes, yesterday’s VaR can badly understate today’s risk — which is why forward-looking GARCH volatility models are often layered on top.

Key takeaways

Want your own portfolio’s VaR computed on live data, in plain English? Run the free diagnostic — no account required.

Run the free diagnostic · Create a free account