Value at Risk (VaR)
Explained

"How much can I lose?" is the first question every serious risk manager asks. For a simple stock position, the answer is relatively straightforward — you multiply the position size by the stock price. For an options book with multiple legs across different underlyings and expirations, it's far more complex. Value at Risk (VaR) is the industry-standard framework for answering this question with mathematical precision.

Used by every major investment bank, hedge fund, and increasingly sophisticated retail traders, VaR translates a complex portfolio's risk into a single, interpretable number. Understanding it — including its significant limitations — is essential for anyone running more than one or two options positions simultaneously.

What VaR Actually Says

Value at Risk answers a specific question: what is the maximum loss my portfolio will suffer over a given time horizon, at a given confidence level?

A VaR statement always has three components: a dollar amount, a time horizon, and a confidence level. For example: "Our 1-day 95% VaR is $2,500." This means: with 95% confidence, the portfolio will not lose more than $2,500 in a single trading day. Equivalently, on roughly 1 in 20 trading days (5% of the time), losses will exceed $2,500.

Three Ways to Calculate VaR

VaR can be computed by three distinct methods, each with different assumptions, strengths, and weaknesses. For options portfolios, the choice of method matters significantly.

1. Parametric VaR (Variance-Covariance Method)

The parametric approach assumes portfolio returns are normally distributed. Given the portfolio's expected return (μ) and standard deviation (σ), VaR at confidence level c is:

VaR(c) = μ − z(c) × σ    where z(c) is the standard normal z-score for confidence level c

For a 95% confidence level, z = 1.645. For 99%, z = 2.326. This method is fast and computationally cheap — once you know the portfolio's mean and variance, VaR is a single calculation. The problem: options portfolios have non-linear payoffs. Their P&L is emphatically not normally distributed — it's skewed, fat-tailed, and changes shape as the underlying moves. Parametric VaR applied to options gives a deceptively clean answer built on a fundamentally wrong distributional assumption.

2. Historical Simulation VaR

Historical simulation takes the actual returns observed over a past period — typically 250 to 500 trading days — applies each historical day's return scenario to the current portfolio, and ranks the resulting simulated P&Ls from worst to best. The 95% VaR is simply the 5th percentile of this empirical distribution (the 13th worst day out of 250).

This approach makes no distributional assumptions — it uses real market history, including all the fat tails, skewness, and volatility clustering that actually occurred. The limitation is look-back dependency: if the historical window doesn't include a stress event similar to what you're about to face, historical simulation will underestimate risk. A VaR model calibrated on 2015–2019 data completely missed the COVID crash of March 2020.

3. Monte Carlo VaR

Monte Carlo simulation generates thousands of hypothetical future scenarios by randomly sampling from a specified statistical model — typically allowing for fat tails, stochastic volatility, and correlation dynamics. Each simulated scenario produces a portfolio P&L, and VaR is the percentile of the resulting distribution.

Monte Carlo is the most flexible and accurate method for options portfolios. It can incorporate the non-linear payoffs of options correctly, model volatility as stochastic (capturing vega risk), and stress-test correlation assumptions. The cost is computational — running 10,000 full scenario simulations across a complex options book takes time and processing power. It's the standard at professional options desks for precisely this reason.

Why VaR Is Especially Tricky for Options

Options have properties that make them systematically difficult for standard VaR frameworks:

Non-Linear P&L Profiles

A stock position's P&L is linear — if the stock moves 2%, your P&L moves proportionally. Options are non-linear — a 2% move in the underlying has a completely different P&L impact depending on whether you're ATM or deep OTM, whether you're long or short, and how far you are from expiration. This non-linearity means simple variance-based VaR badly misrepresents options risk. A short gamma position that looks manageable under parametric VaR can suffer catastrophic losses in a large gap move that the normal distribution assigns near-zero probability.

Volatility Risk (Vega)

Standard VaR models focus on price risk — moves in the underlying. For options, volatility itself is a risk factor. A portfolio that's delta-neutral (no price direction exposure) can still suffer large losses if implied volatility spikes, because vega exposure creates P&L even when the underlying doesn't move. A VaR model that ignores vega risk is missing a critical dimension of options portfolio risk.

Theta Decay and Time Risk

Options lose value as time passes even if the underlying stays flat. A multi-day VaR horizon needs to account for theta decay on long option positions, which erodes value every day regardless of market moves. This time-based erosion is unique to options and is typically absent from stock-portfolio VaR frameworks.

CVaR: The Honest Measure of Tail Risk

Conditional Value at Risk (CVaR) — also called Expected Shortfall (ES) — answers the question VaR refuses to: when losses do exceed the VaR threshold, how bad do they get on average?

CVaR is the expected loss given that you're already in the worst (1 − c)% of outcomes. If your 95% VaR is $2,500, your 95% CVaR tells you the average loss on the worst 5% of days. That number is always larger than VaR — and for fat-tailed, negatively skewed distributions like options P&L, it can be dramatically larger.

This example illustrates why CVaR is particularly important for options traders. Naked options selling can post favorable VaR numbers — many small wins mean the VaR threshold is rarely breached. But when it is breached (in a gap move, a volatility spike, a flash crash), the losses are not modestly above VaR — they can be 5–10× VaR. CVaR captures this; VaR masks it.

Regulatory frameworks have moved toward CVaR: Basel III and its successors require banks to report Expected Shortfall rather than VaR precisely because the 2008 financial crisis revealed how catastrophically VaR had underestimated tail losses in structured products. For options traders running premium-selling strategies, the same lesson applies.

Stress Testing: What VaR Can't Model

Both VaR and CVaR are statistical measures based on historical or simulated data. They estimate risk under "normal" probability distributions. Stress testing complements them by asking: what happens in a specific catastrophic scenario that may be outside the historical distribution?

Useful stress scenarios for an options portfolio:

• VIX spike to 60+: How much does your book lose if implied volatility doubles overnight? (As it did in March 2020 and February 2018)
• Underlying gap move of 15%: What's the P&L on your short delta positions if the underlying gaps past all your strikes before you can adjust?
• Correlation spike: All your "diversified" positions correlate to 0.95 simultaneously. What's your actual book P&L?
• Liquidity freeze: Bid-ask spreads widen 5× and you can't close positions at mid-price. What does it cost to unwind the book in that environment?

Stress testing is not a statistical exercise — it's an imagination exercise. The goal is to identify the specific scenarios that would produce your largest losses and verify you can survive them. If a stress scenario produces a loss larger than your monthly drawdown limit, you need to either reduce position size or add protective hedges before that scenario can occur.

A Practical VaR Framework for Options Traders

Putting it all together in a workflow you can actually use:

1. Define your risk budget. Decide on a maximum acceptable 1-day 95% VaR as a percentage of account (e.g., 3%). This is your overall risk ceiling.
2. Estimate position-level VaR. For defined-risk positions, use max loss × probability of breach. For undefined-risk, use broker margin requirement as a conservative proxy.
3. Aggregate across the book. Sum position VaRs, adjusting for correlations between positions (correlated positions don't diversify; uncorrelated ones do).
4. Check CVaR. For each short options position, estimate the expected loss in the worst 5% of scenarios — this is your true tail exposure. Ensure CVaR is within a survivable range (e.g., < 15% of account).
5. Run stress tests. Apply the VIX doubling and 15% gap scenarios to the current book. If either produces an unacceptable loss, hedge or size down before entering new positions.
6. Review weekly. VaR is not a one-time calculation. Market conditions change, positions drift in delta and gamma as the underlying moves, and what was a 3% VaR position on Monday can be a 7% VaR position by Friday.