Know these first
- L5.1.07: Monte Carlo pricing
Math Atoms
By the end you will be able to
- 1Simulate the market model by log-Euler and predictor-corrector steps, under the spot and the terminal measure, with 3 normals a step for 40 rates.
- 2Check a simulation against what has a closed form (caplets, the 5y × 5y swaption by Rebonato's formula, a deflated bond), always with the standard error, the paths and the seed.
- 3Separate discretisation bias from sampling error, using common random numbers.
- 4Cut the standard error with antithetic pairs and a control variate, and read what each buys (1.32× and 2.02× here).
Why it matters
L5.3.11: Bermudan swaptions priced the 10-non-call-1 Bermudan on a Hull-White tree, because one factor recombines. The market model has 40 forward rates whose drifts depend on each other: no tree holds them, and the next lesson (L5.5.08: Longstaff-Schwartz for Bermudans) prices the Bermudan on simulated paths instead. Before trusting a Bermudan, which has no closed form, the simulation has to pass on everything that does. This lesson runs those tests, and learns to tell an error in the scheme from an unlucky seed.