How Much Could We Lose Tomorrow?" Value at Risk (VaR) Explained in Plain English

16 Nov 2026 6 min read 17 views
How Much Could We Lose Tomorrow?" Value at Risk (VaR) Explained in Plain English
16 Nov 2026 · 6 min read

Every evening, in trading floors and risk rooms across the world, someone has to answer one deceptively simple question for senior management: "If tomorrow is a bad day, how much could we lose?"

The number they report is called Value at Risk (VaR). It is the most widely used measure in market risk, it appears in bank annual reports and regulatory capital calculations, and it shows up in almost every market risk interview. Here's what it means and how it's calculated, with no heavy maths.

VaR in one sentence

VaR is the loss that should not be exceeded over a given period, at a given confidence level, under normal market conditions.

So a statement like "1-day 99% VaR is ₹3.5 lakh" means:

  • On 99 days out of 100, the portfolio should not lose more than ₹3.5 lakh in a day.
  • On roughly 1 day in 100, the loss could be larger.

Notice what VaR does not tell you: how much larger the loss could be on that bad day. We'll come back to that.

Three ways to calculate VaR

Let's use one simple portfolio throughout: ₹1 crore invested in a stock portfolio.

1. Parametric (variance-covariance) VaR

Assume daily returns follow a normal distribution. If the portfolio's daily volatility is 1.5%, then at 99% confidence (z ≈ 2.33):

VaR = 2.33 × 1.5% × ₹1,00,00,000 ≈ ₹3.5 lakh

  • ✅ Fast, easy, great for intuition
  • ❌ Real markets have "fat tails" — extreme days happen more often than the normal curve predicts

2. Historical simulation VaR

Take the last 500 days of actual returns, apply each one to today's portfolio, and sort the resulting profits and losses from worst to best. At 99% confidence, 1% of 500 is 5 — so VaR is roughly the 5th-worst loss.

  • ✅ No assumption about distributions; uses real market behaviour
  • ❌ Assumes the past 500 days represent the future; a calm period gives a low VaR right before a storm

3. Monte Carlo VaR

Build a statistical model of how prices move (including volatilities and correlations), simulate thousands of possible tomorrows, compute the portfolio's profit or loss in each, and read off the 1st percentile loss.

  • ✅ Flexible; handles options and complex products with non-linear payoffs
  • ❌ Computationally heavy and only as good as the model behind it
MethodSpeedHandles fat tails?Handles options well?
ParametricFastPoorlyPoorly
HistoricalMediumYes (if in the data)Yes
Monte CarloSlowDepends on modelYes

Making volatility smarter: EWMA and GARCH

Markets aren't equally calm every day — volatility clusters. A quiet month can be followed by a turbulent week. Two popular techniques capture this:

  • EWMA (Exponentially Weighted Moving Average): gives more weight to recent days, so VaR reacts faster to changing conditions.
  • GARCH: models volatility as something that rises after shocks and gradually reverts to a long-run average.

For the true extremes, risk managers also use Extreme Value Theory (EVT) to model the tail directly, and copulas to capture how assets move together in a crisis.

The problem with VaR — and Expected Shortfall

Remember the question VaR doesn't answer? "On the worst 1% of days, how bad does it get?"

Two portfolios could have the same VaR of ₹3.5 lakh, but on their worst days one loses ₹4 lakh and the other loses ₹40 lakh. VaR can't tell them apart.

Expected Shortfall (ES) fixes this: it is the average loss on the days beyond the VaR threshold. It looks into the tail instead of stopping at its edge.

This is a major reason global regulators moved bank market risk capital from VaR towards Expected Shortfall under the Fundamental Review of the Trading Book (FRTB), which also introduced revised Standardised and Internal Model Approaches.

Is the model any good? Backtesting

A 99% VaR should be breached on about 1% of days — around 2 to 3 times in a 250-day year. Backtesting simply counts the actual breaches:

  • Far too many breaches → the model underestimates risk
  • Almost none → the model may be too conservative, tying up excess capital

Backtesting and model validation are core parts of any market risk team's work.

Beyond VaR: counterparty credit risk

When a bank trades derivatives, it also faces the risk that the other side defaults before paying. Measuring that exposure brings in frameworks such as SA-CCR, IMM and CVA (Credit Valuation Adjustment) — natural next steps once you understand VaR and simulation.

Why learn this hands-on

Reading about VaR takes 10 minutes. Building it is what employers test. In a market risk interview you may be asked to:

  • Compute historical VaR in Excel from a price series
  • Code a Monte Carlo simulation in Python
  • Explain why your VaR jumped yesterday
  • Compare VaR and ES for a portfolio with options

Learn market risk with Peaks2Tails

Peaks2Tails — whose very name refers to the peaks and tails of a loss distribution — was founded by risk practitioners Satyapriya Ojha and Karan Aggarwal (both FRM® and CQF®).

Market Risk & Counterparty Credit Risk – 180+ hours, Python + Excel

  • VaR and Expected Shortfall: parametric, historical and Monte Carlo
  • GARCH, EWMA, EVT and copulas
  • FRTB Standardised and Internal Model Approaches
  • Derivative valuation, sensitivities and hedging
  • Model validation and backtesting
  • Counterparty credit risk: SA-CCR, IMM, BA-CVA and SA-CVA
  • Capstone projects: build a Monte Carlo VaR engine, an FRTB capital charge simulator and an SA-CCR/IMM exposure engine

Starting from zero? CPRF covers markets, derivatives, statistics, Python and market risk modelling across its four semesters, with 100% live weekend classes in Hinglish.

For banks: Peaks2Tails runs corporate training on market risk, valuations and FRTB through on-site, hybrid and mentoring formats.

Key takeaways

  • VaR = the loss you shouldn't exceed on most days, at a chosen confidence level.
  • Three methods: parametric (fast), historical (data-driven), Monte Carlo (flexible).
  • EWMA/GARCH make volatility realistic; EVT and copulas handle extremes.
  • Expected Shortfall shows how bad the tail really is, and underpins FRTB.
  • Backtesting keeps the model honest.

Next steps

  • 📈 Explore the Market Risk & CCR program on the Peaks2Tails short courses page
  • 💬 Book a free 1-to-1 counselling call on WhatsApp
  • ▶️ Watch free sessions on the Peaks2Tails YouTube channel

Peaks2Tails — turning risk into opportunity.

FAQs

What does 99% VaR mean? That on 99 out of 100 days, losses should not exceed the VaR figure; on about 1 day in 100, they may.

What is the difference between VaR and Expected Shortfall? VaR is the threshold loss; Expected Shortfall is the average loss beyond that threshold.

Which VaR method is best? It depends on the portfolio. Historical and Monte Carlo methods handle real-world behaviour and options better; parametric is fastest.

Can I calculate VaR in Excel? Yes. Parametric and historical VaR are easy to build in Excel; Monte Carlo is usually done in Python for speed.

Article enquiry

Need Help? Contact Us

Fill out the form and our team will contact you shortly.

Continue reading

Related articles

WhatsApp Us Call Now