Two real fixes, eighteen years apart, for the same problem this whole education hub keeps coming back to.
Classical Markowitz optimization has a known bad habit: hand it a slightly optimistic return estimate for one asset, and it will happily bet everything on that one asset β the "error-maximizer" problem you've already met on this hub. Black-Litterman was built to fix exactly this, by starting from a sensible, market-implied baseline and only nudging it where you actually have a real opinion, instead of trusting raw historical averages outright.
Entropy Pooling asks the same question in a more general way: instead of just nudging an expected return, what's the least distorted way to reshape your entire set of assumptions to match a belief you actually hold β without overcorrecting or introducing bias you never intended?
Rather than starting from raw historical average returns (the thing that makes classical MVO so fragile), Black-Litterman starts from an equilibrium prior β the returns implied by the market already being roughly correctly priced β and then blends in your own specific views, weighted by how confident you say you are in each one. A view you're barely sure about moves the result a little. A view you're highly confident in moves it a lot. Zero views, and it simply returns the market-implied baseline, unchanged.
The posterior (post-view) expected return blends the equilibrium prior Ξ with your views P, Q β weighted by how much you trust the prior (ΟΞ£) against how much you trust each view (Ξ©). No solver guesswork: it's a closed-form Bayesian update, the same shape as combining two independent measurements weighted by their own precision.
Black-Litterman only lets you tilt expected returns. Entropy Pooling, developed nearly two decades later, generalizes the same underlying idea to reshape an entire probability distribution β views on volatility, correlation, or even which market regime is more likely, not just the average. It finds the new set of scenario probabilities closest to your original data (measured in relative entropy β the same concept behind "information") that still makes your stated view exactly true. The smallest possible change that honors what you actually believe, nothing more.
Both run as real, working tabs in MFL-PORTFOLIO-SCIENCE today. Black-Litterman's equilibrium prior is reverse-optimized from an equal-weight starting portfolio (real market-cap weights aren't available in this app yet β an honestly-labeled stand-in, not silently presented as genuine market data). Everything downstream β the posterior blend itself β is real math on your real loaded data, verified against a hand-derived closed form and a second, independent implementation before shipping.
Entropy Pooling's dual optimization was checked against a direct constrained solve (SciPy's SLSQP, with no shortcuts baked in) β posterior probabilities agreed to about one part in a million.
Fischer Black β yes, the same Black behind the Black-Scholes options formula β and Robert Litterman built this at Goldman Sachs in 1990, as an internal fix for a real, practical headache their own portfolio managers kept running into. It reached the wider world in 1992 ("Global Portfolio Optimization," Financial Analysts Journal), and became far more widely understood after Guangliang He and Litterman published a clearer walkthrough of the underlying intuition in 1999 β the version most practitioners, including this app, actually build from.
Attilio Meucci, a working quant and professor, published the entropy-pooling generalization in 2008 ("Fully Flexible Views: Theory and Practice") β nearly two decades after Black-Litterman, solving the same fundamental problem their model first identified, just with a far more flexible mathematical toolkit behind it.
Both techniques are only as good as the views you feed them. They protect you from letting raw historical noise dominate your portfolio β they cannot protect you from a confidently-held view that's simply wrong. Garbage in, carefully-blended garbage out.
General education only. Nothing here is personal financial advice or takes into account your objectives, financial situation or needs. Past performance is not a guarantee of future results.