Value at Risk (VaR): What’s the Worst That Could Happen?

The answer, dear reader, is everything.

I’ve been staring at price charts for twenty-three years. I’ve watched algorithmic trading destroy fortunes in milliseconds. I’ve seen traders with PhDs in mathematics get absolutely demolished by a Tuesday afternoon in the GBP/USD. And you know what? Most of them never even heard of Value at Risk, or if they did, they thought it was something that happened to other people.

It isn’t.

Let’s Start With the Depressing Truth

Value at Risk—VaR, as the cool kids call it (and by cool kids, I mean the ones who haven’t yet blown up their accounts)—is basically a fancy statistical way of asking: “How badly can this realistically go tits up?”

More formally: VaR estimates the maximum loss your position could suffer within a specific confidence level over a given time horizon. Say you have a 95% VaR of $5,000 over one day. That means on any given trading day, there’s a 95% probability you won’t lose more than $5,000. Which also means there’s a 5% probability you will lose more. Sometimes significantly more.

You know what a 5% probability feels like to a retail trader? Like it’ll never happen to them. Until it does. Twice in a row.

I watched a bloke in 2015 wake up to find the Swiss franc had unpegged from the euro. His VaR models? Absolutely useless. He’d calculated a 99% confidence level assuming the peg would hold forever—because apparently, some people think central bank policy is written in God’s personal diary. His 10:1 leverage position? Gone. His house? Also gone. His marriage? Double gone.

That’s the thing about VaR that nobody tells you at trading seminars: it assumes history repeats in normal, predictable ways. Markets don’t work that way.

The Three Flavours of VaR (Or: Pick Your Poison)

There are three main approaches to calculating VaR, and choosing between them is like choosing between different types of financial misery.

1. Historical VaR

Take your last, say, 250 days of trading data and simulate what would happen if historical price movements occurred again. Simple. Elegant. Completely useless when the market does something unprecedented.

The 2008 crisis? Black swan. The March 2020 COVID crash? Another black swan. The time Turkey’s central bank governor got fired via Twitter and the Lira plummeted 15% in an afternoon? Black swan. History isn’t just non-repeating; sometimes it’s batshit crazy.

2. Parametric VaR

Assumes price movements follow a normal distribution (a bell curve). This is what most banks use because it’s mathematically clean and makes spreadsheets look professional in front of the board of directors.

It’s also catastrophically wrong during volatile markets. Forex moves don’t follow nice Gaussian distributions. They follow what nerds call a “fat-tailed distribution”—meaning extreme moves happen way more often than the math pretends they should.

I knew a risk manager at a major tier-1 bank who got absolutely filleted when the GBP/JPY crashed 300 pips in what should have been a 0.003% probability event. His parametric model said it couldn’t happen. The market had other ideas. He’s now managing portfolios at a pension fund in Milton Keynes, which is basically exile for London traders.

3. Monte Carlo VaR

Run thousands of simulated price paths using random scenarios. It’s the “throw everything at the wall” approach. More flexible, but computationally expensive and still fundamentally reliant on your assumptions about volatility and correlation.

Monte Carlo is what you use when the other two methods have already failed you at 3 AM on a Friday.

Why Your Broker’s VaR Calculator Is Lying to You

Most retail brokers offer “VaR calculators” that are about as useful as a chocolate teapot. Here’s why:

They assume you’ll actually stick to your risk management. You won’t. I know you think you will, but you won’t. Humans are wonderful creatures of contradiction. We promise ourselves we’ll stop loss at X, then when the position moves against us, we suddenly become philosophers wondering if we should “give it more time.”

They don’t account for liquidity disappearing. During a flash crash or geopolitical event, the bid-ask spread can widen faster than you can say “margin call.” Your calculated 5% risk becomes 12% in actual practice. I’ve seen it happen in major currency pairs during bank hours.

They ignore correlation breakdown. In normal times, EUR/USD and GBP/USD move together. In crisis mode? They decouple violently. VaR models built in calm markets assume correlations remain stable. They don’t.

The Honest Application of VaR

Despite all this pessimism, VaR does have a purpose if you’re not completely dense about it.

VaR should be used as a starting point for thinking about risk, not as a destination. It’s a conversation starter, not an oracle.

Here’s how I actually use it:

1. Portfolio Positioning: I calculate VaR across my entire portfolio to see where my concentration risk lives. If 60% of my VaR is tied up in GBP pairs during a major political event, that’s useful information.

2. Position Sizing: VaR helps me scale position size proportionally to volatility. When VaR spikes, I reduce leverage. When it’s calm, I might incrementally increase it. This isn’t rocket science.

3. Stress Testing: I use VaR as a baseline, then deliberately break the assumptions. What if volatility quadruples? What if my three favourite pairs all move against me simultaneously? What if the unthinkable happens? This is where actual preparation happens.

4. Calibration: I compare my predicted VaR to actual losses. When my model consistently underestimates losses, I adjust. When it overestimates, I adjust differently. VaR should evolve with your actual trading experience.

The Uncomfortable Conclusion

Value at Risk is a tool for recognising that the worst can happen. It won’t prevent it. It won’t even predict it accurately. What it does do, if you’re not a complete muppet, is force you to confront the reality that there are scenarios where you lose more than you planned.

The traders who survive and prosper aren’t those with the most sophisticated VaR models. They’re the ones who understand that VaR is merely a lower bound on possible misery. They trade smaller. They diversify. They respect volatility. They assume they’re wrong and prepare accordingly.

That Swiss franc story I mentioned? His VaR said he was fine. He wasn’t. That’s the whole lesson right there.

Use VaR. Calculate it. Understand it. But for God’s sake, don’t let it make you complacent.

Because the worst that could happen? It usually does. Eventually.

---

Now stop reading blogs and go check your position sizing.