There's a 5% Chance Your Portfolio Loses More Than This on a Bad Day. That Number Is Called Value at Risk
5 min read
Loading...
5 min read
Your portfolio drops 8% in one bad week. Is that a normal Tuesday for a portfolio like yours, or a warning sign? Most people have no number to check it against, just a bad feeling.
Value at Risk (VaR) is that number. It does not promise you will never lose money. It states, at a chosen confidence level, how much you should not expect to lose more than, over a set period. A one-day 95% VaR of 3% means: on roughly 95 days out of 100, your portfolio should not fall more than 3% in a day. On the other 5, it might fall more, sometimes far more.
VaR needs three inputs, and each one is simpler than it sounds.
A time horizon. One day, one week, one month. Shorter horizons need less data and update faster. This post uses one day.
A confidence level. Usually 95% or 99%. This is not a guarantee, it is a stated frequency. A 95% VaR is expected to be breached about one trading day in twenty. A 99% VaR, about one day in a hundred.
A loss amount, in currency or percent. This is the number VaR actually produces: how much you should not expect to lose more than, at the confidence level and horizon you picked.
Put together: a one-day 95% VaR of Rp1.25 million means that on about 19 days out of 20, your portfolio's one-day loss should stay under Rp1.25 million. It says nothing about the 1 day in 20 where it does not, only that it happens roughly that often.
The simplest way to calculate VaR uses your own portfolio's real history. Called historical simulation, it skips any assumption about how returns are distributed and just reads the answer off what actually happened.
Take a hypothetical Rp50 million portfolio split 60% BBCA, Indonesia's largest bank by market value, and 40% Bitcoin, the kind of blend a diversified NetWort user might hold. Pull its daily returns for the last 100 trading days, then sort them from worst to best. For a 95% VaR, the cutoff sits at the 5th-worst day, the observation right at the 5% mark.
| Rank | Daily return |
|---|---|
| Worst day | −4.8% |
| 2nd worst | −3.6% |
| 3rd worst | −3.0% |
| 4th worst | −2.7% |
| 5th worst (the 95% VaR cutoff) | −2.5% |
This is an illustrative dataset built to show the mechanics, not a real 100-day history. The method is the real part: rank the actual daily returns, then read the loss off the observation that sits at your chosen percentile.
The one-day 95% VaR here is 2.5%, or about Rp1.25 million on a Rp50 million portfolio. On roughly 95 days out of 100, this portfolio's daily loss should stay under that. On the other 5, expect more, and the worst single day in this window lost nearly double that, at −4.8%.
This is where VaR gets misread. A 95% VaR of 2.5% sounds like a safety number. It is really a frequency statement, and it goes silent exactly where you need it most, the tail beyond the cutoff.
Historical simulation at least keeps the real shape of your own past losses in that tail. A different, more common shortcut, parametric VaR, assumes returns follow a normal distribution, and real markets do not, because they crash harder and more often than a normal distribution predicts. Fat tail risk is the name for that gap, and it means any VaR number, historical or parametric, tends to understate how bad the worst days can get. VaR tells you how often you cross a line. It was never built to tell you how far past it you could fall.
This is also why VaR should never stand alone. It pairs naturally with maximum drawdown, which looks at the single worst peak-to-trough loss your portfolio has actually lived through, the number VaR's own tail stays silent about.
Building a full historical simulation by hand is more work than most people want for a rough answer. A faster shortcut exists, called parametric VaR, and it starts from a number you already have.
Your Portfolio Health card on NetWort shows your portfolio's annualized volatility. Divide it by the square root of 12 to get a monthly figure, then multiply by 1.645. That is a rough one-month VaR at the 95% confidence level, in the same percent terms as your volatility number.
For 99% confidence, use 2.326 instead of 1.645. Both multipliers come from the standard normal distribution, and both carry the same warning as any parametric shortcut: they assume your returns are shaped like a bell curve. A portfolio heavy in crypto, or concentrated in a handful of volatile stocks, usually is not, which is exactly the fat-tail gap the section above describes. Treat the parametric number as a fast, rough floor, not a final answer.
VaR is not a metric you check once. It shifts as your holdings, and market volatility, change. A calm quarter understates it. A volatile one overstates the old number if you do not recompute.
Start with your portfolio's actual volatility, visible on your Portfolio Health card next to your Sharpe ratio and max drawdown. Run the quick parametric estimate above to get a working number, then treat it as a floor rather than a ceiling: the tail beyond it is where the real damage happens, and no VaR calculation, however careful, tells you how deep that tail goes.