Black–Litterman

Black–Litterman attacks mean-variance instability at the input stage. Start from a neutral equilibrium: the market-capitalization weights imply returns consistent with CAPM. Then blend in the investor’s views, expressed as expected returns with stated confidence, and Bayes-style updating shifts equilibrium returns toward the views in proportion to their precision. Running the ordinary optimizer on the blended inputs yields portfolios that move smoothly as views change instead of lurching between corner solutions. Views may be absolute or relative, and assets without views keep their implied equilibrium returns. It is the standard answer when a raw optimizer demands an implausibly concentrated book, leaning on the CAPM equilibrium of the previous node.