Lesson 3 of 8 · All lessons in this module
- The simulation framework: four steps, three sources of scenarios
- Historical simulation and historical VaR
- Historical simulation in practice: forecasts, data gaps and bonds
- Bootstrap resampling
- Bootstrapping in practice: price paths, confidence intervals and limits
- Monte Carlo simulation: what it is and what it is for
- Monte Carlo valuation with log returns: European and Asian options
- Correlated scenarios: Cholesky and multivariate Monte Carlo
This lesson is part of the 2026 and 2027 curriculum. You are following the 2026 curriculum, where it is not taught in this form. Switch if you are sitting the exam under the 2026 and 2027 curriculum.
Lesson 3 of 8 · 12 min
Historical simulation in practice: forecasts, data gaps and bonds
Historical simulation can stress a forecast with past forecast errors, but it needs clean data, proxies for missing history, yield-based factors for bonds, and an awareness that it only knows the past.
In short
- Past forecast errors (forecast versus actual) can be replayed on a current forecast to give a realistic range of outcomes and curb overconfidence.
- Percentiles of a ranked sample come from linear interpolation between neighbouring observations.
- Missing or unreliable data arise from asynchronous markets (holidays, closures), securities that did not yet exist, and thinly traded instruments. A proxy, often a weighted average of similar instruments, fills the gap.
- For bonds, simulate benchmark yield changes, not the bond's own past price changes: a bond about to mature carries almost no price risk whatever its history shows.
- Strengths: simple, no distributional or correlation modelling, real data, cheap to compute. Weaknesses: backward looking, limited scenario variety, dependent on data quality, sensitive to outliers.
Unlock this lesson free for 7 days
Create a free account to get 7 days free: every lesson, video and flashcard, and the question bank. No card needed.