VaR measures the worst monthly loss by a strategy, at 95% confidence. Technically, VaR is the 95% percentile of the distribution of % returns on equity, projected over a 1-month horizon.
A strategy trading with stable risk will experience losses larger than VaR one month every 20 months (1-95% = 5% or 1/20).
VaR is a risk measure widely applied in finance. In a trading context, many mistake VaR for historical drawdown even though both are conceptually different.
VaR tracks risk, describing what could happen. Drawdown is backward looking: it captures what did happen. Contrary to misguided conventional wisdom, it is entirely possible for trading strategies with extremely high risk (VaR!) to muster low drawdown, for a while.
Unsure about what we mean? Read on!
Why VaR is NOT Drawdown?
Play this (thought!) experiment: introduce a single bullet in an empty 6 load revolver barrel. Randomly rotate drum. Aim revolver at your head. Pull the trigger.
- You might survive 50 runs of Russian Roulette… this can´t be ruled out statistically,
- This doesn't change a fact: on average, you're dead after the worst of 6 runs
The 95% percentile is overkill for this experiment, 83.3% percentile (1 - one sixth) captures the range of all possible outcomes :-(
Surviving 50 rounds of Russian roulette isn't low risk. Mistaking Drawdown for risk amounts to thinking that what did happen is the same as what could happen. Historians care about Drawdown, but smart investors focus on risk.
VaR and Drawdown would converge, given enough repetitions of a controlled experiment in which conditions stay constant. Unfortunately, markets don´t stay constant and can´t be controlled. Don't learn this lesson the hard way!