Every risk metric covered so far in this series — Sharpe, Sortino, Calmar, max drawdown — is backward-looking. They describe what already happened to an equity curve. Value at Risk (VaR) is the one number in a trader's toolkit that tries to answer a forward-looking question instead: given how this account has behaved, how much could it plausibly lose over the next day, week, or month?
It's a staple of institutional risk desks for exactly that reason — a bank doesn't just want to know its trading desk lost money last quarter, it wants a number it can act on before the next quarter starts. Individual traders can use the same idea, at a much smaller scale, to size positions and set realistic expectations for a bad stretch before it arrives.
What Is Value at Risk?
Value at Risk estimates the maximum loss a portfolio or account is likely to experience over a given time horizon, at a given confidence level. A "1-day VaR of $2,000 at 95% confidence" means: on 95 out of 100 trading days, this account should not lose more than $2,000. It says nothing about the other 5 days — those are exactly the tail-risk days VaR is not designed to bound.
VaR answers: "What's the most I should expect to lose on a normal bad day?" — not "what's the worst that could ever happen?"
That distinction matters more than any formula. VaR is a statement about the likely range of outcomes, built from historical volatility. It is not a hard ceiling, and treating it as one is the single most common — and most costly — misuse of the metric.
Three Ways to Calculate VaR
- 1Historical simulation — take your actual daily P&L history, rank it worst to best, and read off the loss at the desired percentile (e.g., the 5th percentile for 95% confidence). No assumptions about the shape of your returns; the tradeoff is that it only knows about scenarios that already happened in your sample.
- 2Parametric (variance-covariance) method — assumes returns are normally distributed and calculates VaR from the mean and standard deviation of your P&L: VaR = (Average Return − z-score × Standard Deviation) × Account Value. Fast to compute, but breaks down for strategies with fat tails or skewed return distributions — which describes most active trading strategies.
- 3Monte Carlo simulation — generates thousands of hypothetical future return paths based on the statistical properties of your historical trades, then reads VaR off the simulated distribution. More flexible than the parametric method and not limited to your literal trade history the way historical simulation is, but only as good as the assumptions fed into it.
A Worked Example
Take a trader with 100 days of P&L history and a 1-day 95% VaR calculated via historical simulation. Sorting all 100 daily results from worst to best, the 5th-worst day was a loss of $1,800.
| Confidence Level | Interpretation | 1-Day VaR |
|---|---|---|
| 95% | Losses should exceed this on ~5 of every 100 trading days | $1,800 |
| 99% | Losses should exceed this on ~1 of every 100 trading days | $3,400 |
Both numbers are correct simultaneously — they describe different points on the same distribution of outcomes. A trader sizing positions for a prop firm's daily loss limit cares more about the 99% figure; a trader just trying to understand a typical rough day cares more about the 95% figure.
VaR vs. Max Drawdown: Different Questions Entirely
It's tempting to treat VaR and max drawdown as versions of the same idea. They're not.
- Max drawdown is a single historical fact — the worst peak-to-trough decline that has already occurred in your account
- VaR is a statistical estimate of a plausible near-term loss, derived from the distribution of past returns, not a record of one specific past event
- Max drawdown accumulates over a losing streak spanning many trades or days; VaR is typically scoped to a single time horizon (commonly 1 day or 10 days)
- A strategy can have a small historical max drawdown and still carry a high VaR, if recent volatility has picked up — VaR reacts faster to changing conditions because it's built from a rolling window, not a single all-time low
TradeEdge tracks daily P&L, Sharpe, Sortino, Calmar ratio, and max drawdown automatically from your logged trades — giving you the return history VaR needs without exporting to a spreadsheet first.
How Traders Actually Use VaR
The practical use case isn't predicting the future — it's sizing risk before it happens, in a few concrete ways:
- 1Position sizing — if 1-day 95% VaR on a proposed position exceeds your daily risk budget, the position is too large before you even enter it
- 2Setting realistic daily loss limits — a prop firm's max daily drawdown rule can be checked against your account's VaR to see how much room for error actually exists on a normal bad day
- 3Comparing strategies on equal footing — two strategies with similar average returns can carry very different VaR, exposing which one is quietly taking more risk to get there
- 4Flagging regime changes — a sudden jump in rolling VaR, even before a drawdown shows up, is often the earliest signal that a strategy's risk profile has shifted
Where Traders Get VaR Wrong
The biggest mistake is treating VaR as a worst-case number instead of what it actually is: a threshold that gets breached on purpose, a known percentage of the time. A 95% VaR is expected to be exceeded roughly 1 day in 20 — if it never gets breached, the estimate is too conservative, not impressively safe. The second mistake is the same one that inflates Sharpe and Sortino ratios on thin samples: calculating VaR from too few days of data understates it, because a short history hasn't yet included a genuinely bad day. A third, specific to the parametric method, is assuming normally distributed returns for a strategy that actually has fat tails — options sellers and trend-followers in particular tend to have return distributions that make textbook parametric VaR meaningfully understate real tail risk.
Why This Needs a Journal, Not a Quick Mental Estimate
Calculating VaR by hand means pulling every day's realized P&L into a single sorted series and re-running the calculation as new days close — tedious enough that most traders who try it by spreadsheet do it once and never update it. A trading journal that logs every closed trade already has the daily P&L series VaR is built from, and can recompute the historical-simulation figure the same way it already recalculates Calmar and Sortino ratio as fresh trades close — turning a once-a-quarter spreadsheet exercise into a number that stays current with every session.