Second moment (direction vs volatility)

Every return carries two separate pieces of information, and traders routinely confuse them. The first moment is the mean — direction, whether price went up or down and by how much on average. The second moment is the variance — how big the swings were, and how they moved together with other assets. One tells you where. The other tells you how hard.

Condition a return on a market state — a volatility regime, a session, a calendar phase, a dealer-gamma regime — and something asymmetric happens. The second moment moves. Volatility clusters, correlation shifts, tails widen or shrink, all in ways that hold up out-of-sample. The first moment doesn't. Direction stays close to a coin flip across every state we tested, and none of them nudge the sign of the next return.

The pattern isn't unique to our data. Markets price in information about magnitude far more efficiently than they leak information about sign. A dealer sitting short gamma has to chase price in whichever direction it already moves, which amplifies whatever's happening — but it doesn't tell you which direction that will be. A quiet Tuesday session predicts a quiet Tuesday, not an up Tuesday.

That distinction is the whole reason intermarket analysis earns its keep as a filter and fails as a signal generator. A regime read on one market can tell you the next few hours will be calm or violent, correlated or decoupled. It cannot tell you whether the next candle closes green or red. Sell it as direction and you're selling a coin flip with a research budget attached.

We test for this by splitting return series into first-moment and second-moment questions separately, before conditioning on any state variable. Does the regime predict sign? Does it predict magnitude? Across every regime we've checked the split holds: the volatility question answers, the direction question shrugs. The same discipline sits underneath volatility risk premium — a genuine structural edge, but one built entirely on the second moment, never the first.

Get the two moments mixed up and a real finding turns into a fake one. A backtester who conditions entries on a volatility regime and reports a directional edge has usually just relabeled variance as alpha. The honest version of the finding is narrower and less exciting: this state tells you how big the move might be, not which way it goes. Size the trade around that. Don't pretend it points anywhere.