A method of measuring potential losses in a portfolio by analyzing actual past market movements. Unlike Analytical VaR, which relies on statistical assumptions, Historical VaR uses real historical data to estimate how much value a portfolio could lose over a set time horizon at a given confidence level. 

To calculate it, the model applies historical price changes to the current portfolio to simulate a range of possible outcomes. The worst losses over a selected look-back period (e.g., 1 year) are ranked, and the Value at Risk is the loss at the chosen percentile, often 95% or 99% confidence. 

For example, if a portfolio shows a 99% Historical VaR of $80,000 over 10 days, it means there is a 1% chance the portfolio could lose more than $80,000 based on past data. 

Key features of Historical VaR: 

• Model-free, relies on actual historical returns, not normal distribution assumptions. 

• Responsive to market events, especially in volatile periods. 

• Can be less predictive if future market behavior differs from the past. 

It is commonly used in risk management, regulatory reporting, and stress testing, and is often compared with Analytical and Monte Carlo VaR to validate risk assessments and capture different scenarios.