Core Path · L6.2.04

Volatility-weighted and filtered historical simulation

14 minFiltered historical simulationVolatility weightingAge weightingEWMAGARCHVaR backtest

Know these first

By the end you will be able to

  1. 1Weight a historical simulation by age and say what it fixes and what it breaks.
  2. 2Rescale historical moves to today's volatility with EWMA (Hull and White) and GARCH (filtered HS), and revalue in full.
  3. 3Compare the models' one-day VaR on the sleeve from 2007 to 2025, through 2008, March 2020 and the 2022 gilt crisis, by their exceptions.
  4. 4Read the forecasts for 2 January 2026 and the caveats behind them, including the look-ahead in fitted parameters.

Why it matters

The same 250 days, the same sleeve, the same engine. For 2 January 2026 plain historical simulation says the 99% one-day VaR is £1,811,177. Rescale each day's moves to today's volatility and it is £1,181,614. Run both every day since 8 Jan 2007 and count the days the loss beat the forecast: HS 63, the rescaled model 46, where a correct model expects 46.9. Which number would you put in front of the risk committee?