Moment Time-Series & Skew Flips

Until now you have treated skewness and kurtosis as numbers you measure once and report. That is the wrong mental model for trading them.

A moment is not a constant of the asset. It is a quantity that breathes. The skew you measure this week is a different animal from the skew you measured last month, and the path it took to get from there to here is often the signal — more so than the level it happens to be resting at today. So in this chapter you stop asking "what is the skew?" and start asking "what is the skew doing?" You turn each higher moment into a time-series, and you learn to read its motion.

Let us build the intuition first. Picture a window of returns — say the last sixty trading days. Inside that window you can compute the four moments you already know: the mean, the variance, the skewness, and the kurtosis. Now slide the window forward by one day. The oldest return drops off the back, a new return joins the front, and you recompute. Do that every day and you no longer have four numbers — you have four curves marching through time. The skewness curve might sit gently negative for months, the ordinary state of an equity index, and then, over a couple of weeks, climb toward zero and cross into positive territory. That crossing is what we call a skew flip, and it is one of the most informative events in this entire course.

Now the formalism, stated carefully. For a window of returns, you first standardize: subtract the window mean and divide by the window standard deviation, so every observation becomes a z-score. The skewness is the average of those z-scores raised to the third power. Because you cubed them, the sign survives — a fat left tail, where the big surprises are losses, produces a negative average, and a fat right tail produces a positive one. The kurtosis is the average of those same z-scores raised to the fourth power. The fourth power destroys the sign but enormously magnifies anything far from the mean, so kurtosis measures how much of your variance is being manufactured by rare, violent observations rather than ordinary chop. A Gaussian has a kurtosis of three; the excess kurtosis, which is what most platforms plot, is that number minus three, so zero means "as tail-heavy as a normal" and anything positive means "heavier."

The crucial move — the one that separates this tier from the descriptive statistics you learned in Tier 200 — is to compute these on a rolling basis and watch the series. Three distinct events live in those series, and you must not conflate them.

The first is a kurtosis spike. The fourth moment lurches upward. Tail mass is arriving; the distribution is fattening. This typically means the market is starting to produce outsized moves in both directions — a regime where the calm is breaking and the size of the next surprise is growing, regardless of its sign. A kurtosis spike is a volatility-of-volatility warning. It tells you the magnitude of coming moves is rising, but it is deliberately silent on direction.

The second is a skew swing. The third moment makes a large move while staying on the same side of zero — for instance, skew that was mildly negative plunging to deeply negative. The asymmetry is intensifying. Left-tail risk is being priced in faster than right-tail risk; the crowd is paying up for downside protection or dumping it. A swing is a change in degree.

The third, and the headline of this chapter, is a skew flip: the third moment crosses zero, changing sign. This is a change in kind. An asset whose surprises were biased to the downside now has surprises biased to the upside, or the reverse. Flips often mark the exhaustion of one regime and the birth of another — capitulation bottoms frequently show skew flipping sharply positive as the relentless down-gaps stop and violent relief rallies begin.

Here is where the institute's throughline returns. A moment-series alert is only evidence if the series earns it. A single catastrophic day can drag a sixty-day skewness deeply negative all by itself, and then, sixty days later, drop out of the window and produce a "flip" that is pure bookkeeping — the artifact of one outlier aging out, not a genuine change in the market's character. So before you trade a flip, you interrogate it: is the move driven by many observations shifting, or by one observation arriving or departing? Does it persist when you lengthen the window? Genuine evidence is robust; an artifact dissolves the moment you poke it.