This is the capstone. Six lessons took you from the bare optimization problem to a portfolio that survives a real trading desk. You learned that a portfolio is a weight vector summing to one; that the two inputs are expected returns μ and the covariance matrix Σ; that Σ must be symmetric and positive semidefinite, explodes in size with the number of assets, and becomes dangerous when ill-conditioned; that the minimum-variance and maximum-Sharpe portfolios both invoke Σ⁻¹; that feeding noisy estimates into that inversion makes the optimizer an “error maximizer” whose weights are so unstable that 1/N often beats it; and that the cure is robustness — Ledoit–Wolf shrinkage, shrinking the means, Black–Litterman, risk parity, constraints, and turnover penalties. No formula sheet, no hints, no take-backs: every answer locks the instant you submit, the wrong options are the exact traps that fool real desks, and your score stays hidden until the end.
Big picture
Portfolio Optimization — the whole ladder
- Portfolio Optimization
- From frontier to weights
- A portfolio = weight vector summing to 1
- Two inputs: μ (reward), Σ (risk)
- Return = wᵀμ; Variance = wᵀΣw
- The covariance matrix
- Symmetric, positive semidefinite
- N(N+1)/2 parameters — explodes with N
- Conditioning κ; near-singular if correlated
- Mean-variance mechanics
- Min-variance: Σ⁻¹𝟙, needs only Σ
- Tangency: Σ⁻¹μᵉ, needs μ and Σ
- Two-fund separation; all use Σ⁻¹
- Estimation error
- Optimizer = error maximizer
- μ errors ~10× worse; Σ⁻¹ amplifies
- Unstable weights; 1/N often wins
- Shrinkage & robust estimators
- Bias–variance trade-off
- Ledoit–Wolf shrinks Σ to a target
- Shrink μ; Black–Litterman blends views
- Risk parity & constraints
- Equalize risk contributions, not dollars
- Long-only/box constraints ≈ shrinkage
- Turnover penalty; optimize NET of costs
- From frontier to weights
How this exam works
This is a graded exam. Questions arrive one at a time. Once you submit an answer it is final — there is no going back, no second try, and a wrong answer simply fails that question. Your score stays hidden until the very end, where you need 70% to pass. Read every option before you commit.
What concrete object does a portfolio-optimization solver ultimately produce?
Select an answer to continue.
Passed? Here's what you now own
You can take a universe of assets all the way to a tradeable portfolio: state the optimization problem in weights, build and sanity-check a covariance matrix, compute the minimum-variance and tangency portfolios, recognize when an optimizer is maximizing error rather than return, and deploy the full robustness toolkit — Ledoit–Wolf shrinkage, shrunk or Black–Litterman means, risk parity, long-only and box constraints, and turnover penalties — to produce a portfolio that survives both estimation noise and the broker’s bill.
That’s portfolio optimization, end to end — the bridge from the elegant efficient frontier to the messy, error-aware discipline of actually computing weights you’d put real money behind. You now own both the clean math and the hard-won judgment about when to trust it.