今早寫的：Why Your Risk Models Are Lying to You

?

Why Your Risk Models Are Lying to You: Fat Tails the Illusion of Diversification
Experienced investors know market returns don’t follow a neat bell curve. Negative skewness dominates: extreme negative returns occur far more frequently than normal distributions predict, especially in bear markets.

Predicting expected returns is like forecasting the future. As Alan Turing established with "undecidability," no matter how smart you are, certain algorithmic systems cannot be pre-determined.

Here are 5 counter-intuitive truths every asset allocator must confront:

1?? The Paradox of the Signal

Sébastien Page noted that a signal with strong negative predictability is actually a goldmine—you can just trade the exact opposite. The truly dangerous signal is one with zero predictability. If you found the world’s worst investor—consistently wrong and losing money—and inverted their strategy, you’d be wildly successful.

2?? The Death of the Normal Distribution

Extreme returns represent the "fat-tail effect." Market tails are growing larger and heavier. Negative skewness is far more common than positive skewness; a 10% sudden loss happens much more frequently than a 10% gain. Traditional normal distribution is no longer a reliable tool for measuring risk.

3?? The Diversification Trap

Most investors allocate by style and size to diversify. Yet during a financial crisis, diversification among risk assets almost universally fails. In extreme environments, even equity/bond correlations break down. When markets crash, correlations spike exactly when you need safety. The timing of these drops remains entirely unpredictable.

4?? The Value Investor’s Long Game

Value investors aim to buy low, waiting for turbulence to clear. But valuation signals are virtually useless over a 1-year horizon. They require a 5-to-10-year time frame to achieve optimal effect. Over the long run, price converges with fundamentals, and mean reversion rewards the patient.

5?? The Myth of the "25-Sigma" Event

Nassim Taleb argues that the fat-tail effect is routinely ignored, even by brilliant minds. In standard models, a 7 or 25-sigma event is mathematically supposed to happen once every 3.1 billion years. Yet, these "Black Swan" events occur repeatedly. The industry consistently underestimates regime-shifting risks.

Bottom Line:

Stop managing risk based on smooth curves. Build portfolios that survive the unpredictable, embrace the fat tails, and remember that when markets drop, the rules change entirely.