The Black-Litterman model

Plain mean-variance optimization has a notorious flaw: feed it your return forecasts and it produces wildly concentrated, unstable portfolios that change drastically when a forecast nudges slightly. The Black-Litterman model, built at Goldman Sachs in the early 1990s, tames this.

The problem it solves

An optimizer treats your estimates as certain truth. Since expected returns are nearly impossible to forecast precisely, tiny estimation errors get amplified into absurd allocations — 200 percent in one asset, short another. The output is mathematically optimal and practically useless.

The two-step intuition

Black-Litterman starts somewhere sensible and adjusts gently:

1. Begin with the market's own view. Reverse-engineer the expected returns that would make today's market-capitalization weights optimal. This equilibrium is the neutral starting point — what you would hold if you had no opinions.

2. Blend in your views, weighted by confidence. You state specific views ("I expect European equities to beat Japanese equities by 2 percent") and how confident you are in each. The model combines the equilibrium with your views in proportion to their certainty.

   Final expected returns = a confidence-weighted blend of (the market-implied equilibrium returns) and (your stated views)

Why the output behaves

Because you only nudge the equilibrium where you actually have a view — and only as hard as your confidence justifies — the resulting portfolio stays close to the sensible market weights, tilting modestly toward your opinions. Strong, confident views move it more; weak views barely register. No more 200-percent positions.

A feel for the result

No views at all        -> you hold the market portfolio
One confident view     -> a measured tilt toward it, rest near market
Many uncertain views   -> small tilts, portfolio stays well diversified

The honest caveat

Black-Litterman is more robust than naive optimization, but it does not manufacture insight. Garbage views, confidently held, still produce a tilted-toward-garbage portfolio. It also relies on the market portfolio being a sound anchor and on you quantifying "confidence" — a subjective number dressed as precision. It disciplines the optimizer; it does not replace judgment.