The objections above — variance is symmetric, returns are not Gaussian, correlations are not stable — are usually filed as footnotes to a theory that still works “well enough”. They are not footnotes. They describe the conditions under which the theory fails completely, and those are the conditions you most need a portfolio to survive.
Variance measures the width of a distribution; deep uncertainty is not knowing the distribution at all. Real return series have fat tails: the rare, violent move is far more likely than a normal curve predicts, and it is that move — not the daily wiggle variance captures — that decides whether you stay solvent. A model that treats an extraordinary year and a ruinous one as equally “risky” has mislabeled the only risk that matters. You are never wiped out by an upside surprise.
Diversification, MPT’s central promise, is measured in calm markets and spent in violent ones. In a liquidity crash, correlations across the risk-asset bucket — equities across countries and sectors, credit, most “alternatives” — converge toward : everything you bought to zig zags together, exactly when you needed the offset. The classic counterweight, high-quality government bonds, has usually held up, but “usually” is not “always” — in 2022 stocks and bonds fell in step. Underwrite your portfolio as though its risk assets are a single position when it counts, because in the moment that counts, they are. Concentration is the same trap worn the other way: a capitalization-weighted index can quietly become a leveraged bet on a handful of names, and the efficient frontier, recomputed on the new data, will cheerfully bless it. Owning “the whole market” is not the same as being diversified.
The fix is structural, not statistical — you do not patch a fragile model by feeding it better estimates. Build for survival first. Size positions so that a run of bad outcomes cannot end the game; this is the discipline behind Edward Thorp’s decades of compounding, formalized in the Kelly criterion and developed in full in section “Beyond the Questionnaire: Kelly Sizing and the Barbell”. Hold genuine dry powder, or an explicit tail hedge that pays off precisely when everything else does not, so a crash becomes an opportunity to deploy rather than a forced sale. Aim past mere robustness for what Taleb calls antifragility: a portfolio positioned so that disorder, when it arrives, works in your favor. The efficient frontier is a fair sketch of the risk–return trade-off in ordinary weather. It is not a survival plan, and the costliest error in this chapter is mistaking the first for the second.