On the limits of VaR

WTI has risen 38.94% in one month. (2/6/2026 - 3/6/2026) 

To put that in perspective, I looked at monthly WTI returns from 1980–2026 and calculated: 6

• Average monthly return: 0.63% 

• Standard deviation: 10.04% 

• Z-score for a 38.94% move, 3.82 

Under a normal distribution, a move of 38.94% in one month is a 99.9932th percentile event. That sounds dramatic, and it is. In plain English, if monthly oil returns were perfectly normal, a move like this would be expected only about 1 time in 14,707 months — roughly 1,226 years. 

VaR gives us a disciplined framework for thinking about risk. But it also has limitations: 

 • It assumes returns behave “normally”. 

 • It underestimates the likelihood of extreme moves. 

 • It is built from the past, while markets enter new regimes (new normal) very quickly. 

 • It tells us a lot about what is likely in “regular” times, but nothing about what happens when the world stops behaving normally. 

 So the lesson is not that VaR is useless. The lesson is that models are tools, not truth. When a price move looks like an “1,226-year event,” the correct reaction is not blind confidence in the calculation. The correct reaction is to ask: Are markets really normal? Are extreme events more common than the model assumes? What risks live in the tails that our spreadsheet may not fully capture? 

That is where good financial modeling begins: not with worshiping the output, but with questioning the assumptions. 

I leave you with a dad joke. According to the model, this was a once-in-800-years event. According to markets, it just happened.