What you will learn
- Parameter instability
- Scenario perturbation
- Estimation error
- Robust alternatives
Parameter instability
Parameter instability occurs when small input changes produce large changes in the selected solution. Measure this directly by perturbing assumptions or resampling under a justified design. A precise output can still be unstable.
Scenario perturbation
Scenario perturbation varies returns, costs, correlations or other inputs within plausible ranges. State why the changes are meaningful. A stress selected only because the candidate survives it is not an informative robustness check.
Estimation error
Estimation error affects expected returns and risk inputs. Optimizers can amplify small noisy differences by assigning extreme weights. Compare against simple baselines that require fewer estimated quantities.
Robust alternatives
Robust alternatives seek acceptable behavior across a set of assumptions rather than the best value under one estimate. The uncertainty set is itself a modeling choice and should be reviewed for relevance and conservatism.
Worked example
Two strategies have nearly equal estimated means, but the optimizer allocates almost everything to the slightly higher one. A small change in the sample flips the allocation to the other strategy.
Try it yourself
Describe a comparison that tests whether this sensitivity materially affects out-of-sample decisions.
Show the worked solution
Compare the optimized policy with capped or simple allocations under predefined input perturbations and later data, including costs. Report both performance and weight turnover; do not assume the most stable policy is automatically best without the decision context.
Apply this to your course project
Compare a simple rule with an optimized alternative under perturbation and held-out scenarios.
Keep the calculation inputs, assumptions and decisions with your work. Practical exercises are self-reviewed; the scored knowledge checks assess the questions shown, not an independent certification of practical competence.