Register
Free toolNo sign-up

Estimate 99% Confidence Interval Downside Risk

Value at Risk, or VaR, estimates how much a portfolio could lose over a set period at a given level of confidence. A one-day 99% VaR of ₹28,000 means that, on 99 days out of 100, losses should not exceed ₹28,000, if the assumptions behind the estimate hold. Banks and funds use VaR to set risk limits; traders can use it to understand how much their positions could lose on a bad day.

Confidence
₹
%
days
Value at Risk, parametric method
₹27,91299% VaR, 1 day

2.79% of the portfolio

Expected shortfall₹32,008Average loss on the days beyond VaR
One standard deviation₹12,000Over the holding period
Days worse than VaR1 in 100If returns are normal
VaR by holding period
Days95% VaR99% VaR
1₹19,740₹27,912
5₹44,140₹62,413
10₹62,423₹88,265
20₹88,280₹1,24,826

Assumes normally distributed returns, which understates large market moves. Options positions need full revaluation rather than this shortcut.

Limits, sizing and hedges for options positions.Read the risk guide

The calculator above uses the parametric method: you enter the portfolio value, its daily volatility, a confidence level and a holding period, and it returns the VaR, the expected shortfall beyond it, the one standard deviation loss and a table for 1, 5, 10 and 20 days. This page explains the three main ways to calculate VaR, works through the calculator's sample and two other examples, shows why the shortcut fails for option positions, and sets out what VaR can and cannot tell a trader in India. VaR is an estimate of normal bad days, not a limit on losses.

What the VaR calculator does

PartWhat it does
InputsConfidence (95% or 99%), portfolio value in rupees, daily volatility in per cent, holding period in days
VaRPortfolio value × z × daily volatility × √days, with z = 1.645 at 95% and 2.326 at 99%
Expected shortfallThe average loss on the days worse than VaR, under the same normal assumption
One standard deviationThe size of an ordinary bad day over the holding period
Days worse than VaR1 in 20 at 95%, 1 in 100 at 99%
Table95% and 99% VaR for 1, 5, 10 and 20 days

A short history of Value at Risk

VaR grew out of a request at J.P. Morgan in the late 1980s, when its chairman, Dennis Weatherstone, asked for a single page at 4:15 each afternoon showing how much the bank could lose over the next day across all its trading. The answer, one number at a stated confidence, became the "4:15 report".1 In 1994 the bank published its method and data as RiskMetrics, free for anyone to use, and VaR spread across banks and funds within a few years.2

Regulators adopted it next. The Basel Committee's 1996 amendment on market risk let banks use their own VaR models, at 99% over ten days, to set the capital they held against trading positions.3 Criticism followed as well: in 1999 Philippe Artzner and colleagues showed that VaR can fail to reward diversification, since two positions combined can show a larger VaR than the sum of the two, and proposed measures without that flaw, of which expected shortfall became the best known.4 The 2007–2009 crisis, when losses ran far past what VaR models had shown, pushed regulators further: the Basel market risk rules finalised in 2019 replaced 99% VaR with 97.5% expected shortfall for capital.5

Indian markets use the same family of ideas for margins. Exchanges' clearing corporations charge VaR-based margins in the cash market, and futures and options margins are set with SPAN, a scenario method first built by the Chicago Mercantile Exchange in 1988, plus an exposure margin.6

Three ways to calculate VaR

Parametric VaR

Assume returns follow a normal distribution and use the portfolio's standard deviation:

Formula
VaR = portfolio value × z × daily standard deviation × √days

where z is 1.645 for 95% confidence and 2.326 for 99%. For short periods, the average return is usually small enough to ignore. This is the method the calculator uses. It is quick and needs only one number, the volatility, but it assumes losses follow the normal curve.

Historical VaR

Take the portfolio's actual daily returns over a past period, such as the last 500 days, sort them from worst to best, and read off the loss at the chosen percentile. For 99% VaR with 500 days, that is the fifth-worst day. Historical VaR makes no assumption about the shape of returns, so it captures the fat tails that real markets have, but it knows only the period it was calculated on.

Monte Carlo VaR

Simulate thousands of possible outcomes using a model of how the portfolio's prices move, revalue the positions in each outcome, then read the percentile from the simulated results. This method handles options and complex positions better than the parametric method, at the cost of more computing and more model choices.

MethodNeedsHandles optionsFat tailsMain weakness
ParametricVolatilityPoorlyNoAssumes normal returns
HistoricalPast returns of the positionsYes, with repricingYes, as seen in the windowOnly knows the past window
Monte CarloA price modelYesOnly if the model has themDepends on the model

A worked example: the calculator's sample

The calculator opens with a portfolio of ₹10,00,000, daily volatility of 1.2%, 99% confidence and a one-day holding period.

Figure95%99%
One-day VaR₹19,740₹27,912
Expected shortfall₹24,747₹32,008
One standard deviation₹12,000₹12,000
Days worse than VaR1 in 201 in 100

And by holding period:

Days95% VaR99% VaR
1₹19,740₹27,912
5₹44,140₹62,413
10₹62,423₹88,265
20₹88,280₹1,24,826

So on about one day in a hundred, the portfolio would be expected to lose more than about ₹27,900, and on those days the average loss would be about ₹32,000, if returns behaved as the normal distribution assumes. Real returns have more extreme days, so treat both figures as lower bounds for the bad days, not upper bounds.

Scaling VaR to longer periods

Under the same assumptions, VaR scales with the square root of time:

Formula
10-day VaR = 1-day VaR × √10

The 10-day 99% VaR in the example is about ₹27,912 × 3.16 = ₹88,265. This shortcut assumes returns are independent from day to day and that the portfolio does not change. In a falling market, bad days tend to cluster, so the true ten-day figure is often larger. And a trader who cuts positions after a loss changes the portfolio, so the true figure can also be smaller. The square-root rule is a starting point, not a forecast.

Expected shortfall

VaR tells you the threshold, not how bad losses are beyond it. Expected shortfall, also called conditional VaR, is the average loss on the days that are worse than VaR. Under the normal assumption the calculator gives ₹32,008 at 99% for the sample, about 15% more than the VaR. With real market returns the gap is usually wider, because the worst days are worse than the normal curve allows.

Expected shortfall can be measured historically too: in the example further down, the average of the five worst days in 500 gives the 99% figure directly, with no assumption about the shape of returns. Expected shortfall is the better guide to how much capital a bad day could take, which is why bank regulators moved to it. For a trader, it answers the question VaR leaves open: when the bad day comes, how bad is it likely to be?

A worked example: historical VaR by hand

A trader keeps 500 days of daily returns for a ₹10 lakh portfolio. Sorted from worst to best, the five worst days are −4.1%, −3.6%, −3.2%, −2.9% and −2.7%.

FigureHow it is foundResult
99% one-day VaRThe fifth-worst day of 500, 2.7%₹27,000
99% expected shortfallThe average of the five worst days, 3.3%₹33,000

The historical figure is close to the parametric one here, but it carries a different kind of warning: the worst day in the record was 4.1%, and the record contains only the days that happened to occur in those two years. A window that includes a crash will show a higher VaR than one that does not, for the same positions.

A worked example: two positions together

A ₹10 lakh account holds ₹6 lakh in stock A, with daily volatility of 1.5%, and ₹4 lakh in stock B, with daily volatility of 2%. Their daily returns have a correlation of 0.4.

Position99% one-day VaR
Stock A alone₹20,934
Stock B alone₹18,608
Sum of the two₹39,542
Both together, with correlation 0.4₹33,108
Formula
Combined VaR = √(VaR A² + VaR B² + 2 × correlation × VaR A × VaR B)

Holding the two together lowers the 99% figure by about ₹6,400, because the two stocks do not always fall on the same day. The saving depends on the correlation, and correlations rise in a sell-off, when stocks tend to fall together. On those days the combined VaR moves towards the simple sum. Using the sum as a cautious figure is never wrong; using the diversified figure assumes the correlation holds when it matters.

VaR for option positions

Parametric VaR works poorly for options because option values do not change in a straight line with the underlying. A short straddle loses little on quiet days and a great deal on a large move in either direction, a pattern a single standard deviation does not capture.

Take one lot of a short Nifty straddle: Nifty at 25,000, the 25,000 call and put sold with five days to expiry and implied volatility of 13%, a lot of 65. The straddle is worth about 304 points. A one-day 99% move in Nifty at that volatility is about ±476 points.

ApproachOne-day 99% loss estimate
Delta only: the straddle's delta is close to zeroAbout ₹1,600
Full revaluation, Nifty down 476 pointsAbout ₹11,800
Full revaluation, Nifty up 476 pointsAbout ₹13,900
Full revaluation, Nifty down 476 points and implied volatility up to 18%About ₹15,300

The delta method says the position is almost riskless because it is balanced at the start; revaluing the options at the moved price shows a loss about nine times larger. Add the rise in implied volatility that usually comes with a fall, and the loss grows again. For options, use historical or Monte Carlo methods that reprice the positions at each level, or look at the full payoff in the options profit calculator, which shows the loss at every index level.

Choosing the volatility input

The parametric figure is only as good as the volatility you give it. Three common sources:

  • Historical volatility of the portfolio. Take the portfolio's daily values for the last month or three months and measure the standard deviation of daily returns. This captures how the actual mix of positions has moved.
  • A weighted figure. Giving recent days more weight, as RiskMetrics did with a decay factor of 0.94, makes the figure react faster when markets turn volatile.2
  • Implied volatility. For a Nifty portfolio, India VIX is the market's expected annual volatility; dividing by √252 turns it into a daily figure. At an India VIX of 14, that is about 0.88% a day.

When the three disagree, the largest is the cautious choice. A portfolio measured over a calm month can show half the volatility it will show in the next volatile one.

VaR for a futures position

One lot of Nifty futures at 25,000, with a lot size of 65, controls ₹16.25 lakh of index exposure, whatever the margin blocked for it. At daily volatility of 0.9%:

ConfidenceOne-day VaR on one lot
95%About ₹24,100
99%About ₹34,000

The VaR is set by the exposure, not by the margin. A trader with ₹3 lakh in the account and two lots open has a one-day 99% VaR of about ₹68,000, close to a quarter of the account on a day that comes once in a hundred. Seen that way, leverage is simply the ratio of the VaR to the capital behind it.

VaR and margin are not the same

Exchange margins and VaR use similar ideas but answer different questions. A futures and options margin is set by the clearing corporation with SPAN, which revalues a position across a set of price and volatility scenarios, plus an exposure margin on top.6 It protects the clearing system if a member's client cannot pay. A trader's own VaR measures the loss to the trader's capital at the confidence the trader chooses.

The two can differ widely. A hedged option spread may need a modest margin and still lose most of it on a large move; a stock held for delivery needs only its price paid and can still fall 20% in a week. Use the margin to know what the broker will block, and VaR, stress tests and the payoff chart to know what you could lose.

Checking a VaR model against reality

A VaR model can be tested by counting the days when the actual loss was larger than the VaR. At 99% over 250 trading days, about 2 or 3 such days are expected. Bank regulators use a traffic-light rule on the same count: up to 4 is green, 5 to 9 is yellow, and 10 or more is red, meaning the model badly understates risk.3

A trader can do the same with a daily record: write down each morning's VaR and each evening's actual profit or loss, and count. If losses exceed the 95% figure much more often than one day in twenty, the volatility input is too low or the positions have changed faster than the figure.

Questions to ask about a VaR figure

  • At what confidence and over what period? A one-day 95% figure and a ten-day 99% figure differ by a factor of about 4.5.
  • From which method? Parametric, historical or Monte Carlo, and over what window?
  • Does it include options at full value, or only their delta?
  • Does it allow for correlations, and what happens if they rise to one?
  • What is the expected shortfall, and how does the worst day in the record compare with both?

Bought options: when VaR overstates the risk

The shortcut fails in the other direction for bought options. A trader who buys a Nifty call for ₹150 a unit, one lot of 65, can lose at most the premium paid, ₹9,750, however far the index falls. A parametric VaR built from the option's own daily volatility, which can be 30% or 40% a day near expiry, may show a 99% figure larger than the whole premium. For a bought option the worst case is known exactly: the premium. For a sold option it is not, which is why sold options need the most care with any risk estimate.

Intraday traders and VaR

A one-day VaR measured on closing values says little about the swings inside the day. An intraday trader who is flat at the close carries no overnight risk, but can still meet a large move between 9:15 am and 3:30 pm, or 3:40 pm for F&O. Two adjustments help. Measure volatility on intraday returns, such as hourly closes, and scale it to the time you actually hold positions. And set the daily loss limit, not the VaR, as the line that stops trading for the day, because the limit acts on what happens and the estimate only describes what usually happens.

The limits of VaR

  • Fat tails. Market returns have more extreme days than the normal distribution predicts. Parametric VaR underestimates how often large losses happen.
  • It says nothing beyond the threshold. A 99% VaR does not describe the worst 1% of days. Expected shortfall does, partly.
  • History may not repeat. Historical VaR knows only its window. A calm period produces a low VaR just before a volatile one.
  • Correlations change. In a crisis, assets that normally move independently often fall together, and diversification shrinks.
  • Liquidity. VaR assumes positions can be closed at market prices, which may not hold in a fast market or a thin strike.
  • Gaps. A position held overnight can open past any stop. One-day VaR measured on closes includes gaps; intraday risk limits do not.
  • It can be gamed. Positions that sell rare large risks, such as far out-of-the-money options, show a small VaR until the day they do not.

VaR is a useful summary of normal risk, not a guarantee against extreme loss. Use it beside stress tests, drawdown limits and hard loss limits.

Stress tests beside VaR

A stress test asks a different question: not "how much on a bad day in a hundred?" but "how much if this specific thing happens?". Useful scenarios for an Indian derivatives account include:

  • Nifty falls 5% at the open on an overnight event.
  • Nifty moves 3% in either direction on an expiry day in the last hour.
  • India VIX doubles while the index falls 4%.
  • A stock in the portfolio falls 15% on results.
  • The broker's session ends or the account cannot trade for an hour.

For each, revalue every position and add up the loss. If any single scenario takes more than you can afford, the positions are too large, whatever the VaR says.

Using VaR in trading

  • Set a daily risk budget, such as keeping 99% one-day VaR below 2% of capital.
  • Compare strategies on the same basis, at the same confidence and holding period.
  • Watch changes. A rising VaR shows the portfolio's risk is growing, whether from larger positions or higher volatility.
  • Size positions from the budget. If the budget is ₹20,000 at 99% and a stock's daily volatility is 1.8%, the most you can hold is about ₹20,000 ÷ (2.326 × 1.8%), close to ₹4.8 lakh.
  • Combine with hard limits. VaR estimates; a loss limit acts. An MTM stop loss for the whole account, called Smart Exit in QuantPower, squares off every position and stops the algos when the day's loss reaches your amount.

A VaR budget works best as a rule about size, set before the day starts, rather than a figure checked after a loss. Each morning, work out the VaR of the positions you plan to hold, and if it is above the budget, trade smaller. If volatility rises through the week, the same positions take up more of the budget, so the size comes down automatically. This is the main practical use of VaR for a trader: it turns a change in market conditions into a change in position size before the loss arrives, not after.

A worked example: a daily risk budget for a ₹10 lakh account

A trader sets a 99% one-day VaR budget of 2% of capital, ₹20,000, and plans two positions.

StepCalculationResult
Budget2% of ₹10,00,000₹20,000
Stock with 1.5% daily volatility₹20,000 ÷ (2.326 × 1.5%)Up to about ₹5.7 lakh alone
Stock with 2.5% daily volatility₹20,000 ÷ (2.326 × 2.5%)Up to about ₹3.4 lakh alone
Both, half the budget each₹10,000 eachAbout ₹2.9 lakh and ₹1.7 lakh

Splitting the budget evenly is the cautious choice, because it ignores any benefit from the two stocks moving differently. A trader who counts that benefit can hold a little more, at the cost of relying on a correlation that may not hold on the worst days.

Risk controls in QuantPower

  • Before an order. The order sheet shows the margin the order needs before it is sent, and the strategy builder shows the payoff of a multi-leg position at every index level, with its margin.
  • On the option chain. Implied volatility and the Greeks for every strike, and the expected move from the at-the-money straddle.
  • During the day. Positions show live profit and loss across every connected broker. The daily loss limit refuses new orders once the day's loss reaches it, Smart Exit squares off and stops the algos at your profit target, stop or trailing level, and the kill switch stops a segment until its next session.
  • Before going live. Backtests show the worst day and the maximum drawdown of a strategy, and Forward test runs it on live prices with test money first.

Charts, the option chain and the strategy builder are free; a backtest costs 2 tokens after 10 free each week, and a live algo 200 tokens per 28 days (1 token = ₹1, GST extra).

How to use the calculator, step by step

  1. Enter the portfolio value in rupees: the market value of the positions, not the margin blocked.
  2. Enter the daily volatility. Use the volatility calculator on the portfolio's daily values, or on a single stock's closes for one position.
  3. Choose the confidence: 95% for an everyday figure, 99% for a cautious one.
  4. Set the holding period: 1 day for intraday and short-term traders, 5 or 10 days for positions held longer.
  5. Read the VaR and the expected shortfall, and compare both with the loss you could accept.
  6. For option positions, use the result only as a rough lower bound and check the full payoff instead.

Common mistakes

  • Using the margin as the portfolio value. VaR is about the value at risk, not the money blocked.
  • Using annual volatility. The calculator needs daily volatility; divide annual volatility by √252.
  • Reading VaR as the maximum loss. Losses beyond it happen, by definition, on one day in twenty or one in a hundred.
  • Applying it to a short straddle. The delta shortcut can understate an option position's risk many times over.
  • Ignoring the window. A volatility measured in a calm month gives a VaR that a volatile month will break.
  • Adding diversified VaRs of strategies that lose together. Strategies on the same index before the same expiry share their bad days.

Where risk measurement is going

The direction is from one number to several. Banks have moved from VaR to expected shortfall plus stress scenarios; brokers and exchanges recompute margins several times a day; and trading platforms now show each position's payoff at every price level and each strategy's drawdown as it happens. For a trader, the practical change is that the tools to ask "what if?" are on the same screen as the order, so a VaR figure can be read beside the loss at a specific move, not instead of it.

Start with 500 free tokens

Connect your broker, build a strategy and test it on paper before a single live order goes out.

Frequently asked questions

What does a 99% VaR mean?

That losses over the chosen period should exceed the VaR figure on only about 1% of occasions, if the model's assumptions hold.

What is the difference between VaR and expected shortfall?

VaR is the threshold that bad days should cross only rarely. Expected shortfall is the average loss on the days that do cross it, so it says how bad the bad days are.

Can losses exceed VaR?

Yes. VaR is a threshold, not a maximum. On the worst days, losses can be far larger.

Why does the calculator use daily volatility?

The formula scales a one-day move by the square root of the holding period. If you have annual volatility, divide it by √252, about 15.87, to get the daily figure.

Should I use 95% or 99%?

Use 95% for an everyday sense of a bad day and 99% for limits and capital. A 99% figure is about 41% larger than the 95% figure for the same portfolio under the normal assumption.

Does VaR work for an iron condor or a short straddle?

Only as a rough lower bound. Option positions lose much more than their delta suggests on large moves; check the loss at specific index levels on the payoff chart instead.

References

  1. Philippe Jorion, *Value at Risk: The New Benchmark for Managing Financial Risk*, 3rd edition, McGraw-Hill (2006), on the origin of the 4:15 report
  2. J.P. Morgan and Reuters, *RiskMetrics: Technical Document* (1994 and 1996 editions)
  3. Basel Committee on Banking Supervision, *Amendment to the Capital Accord to Incorporate Market Risks* (January 1996) (bis.org)
  4. Philippe Artzner, Freddy Delbaen, Jean-Marc Eber and David Heath, "Coherent Measures of Risk", *Mathematical Finance*, 9(3), 203–228 (1999)
  5. Basel Committee on Banking Supervision, *Minimum capital requirements for market risk* (January 2019) (bis.org)
  6. NSE Clearing, margins in the capital market and derivatives segments: VaR margin, SPAN and exposure margin (nseindia.com)
The examples on this page are hypothetical. Risk estimates have inherent limits and do not guarantee maximum losses. This tool and guide are for education and are not investment advice.