Markowitz framework selecting weights boldsymbolw to minimise boldsymbolw^T boldsymbolΣ boldsymbolw subject to boldsymbolw^T boldsymbolμ = R^*, boldsymbolw^T boldsymbol1 = 1. Closed form boldsymbolw^* propto boldsymbolΣ^-1(boldsymbolμ - r_fboldsymbol1) for the tangency portfolio. The infamous flaw: boldsymbolw^* is extraordinarily sensitive to noise in hatboldsymbolμ (1% input change to corner-solution weights), motivating shrinkage covariance (Ledoit-Wolf), Black-Litterman, risk-parity, and resampled-efficiency alternatives.