Expectancy says what a branch earns. The σ ratio says how much to believe it.

Two readings, one branch

Magnitude and stability are independent, and only one of them is obvious.

A branch's expectancy and its volatility are separate properties, and they can move independently in all four combinations. Positive expectancy with contained volatility is the condition the system is built to find. Positive expectancy with expanding volatility is the condition it is built to warn about, because the expectancy figure will continue to read well for some time after the branch has stopped being dependable. The reason this needs a dedicated measurement rather than intuition is that expanding dispersion around a stable mean produces no signal at all in the mean — the average is exactly as favourable as it was, computed over observations that are further apart. Every summary statistic that operators naturally reach for is a measure of centre, and the property that is changing is a property of spread. Without the volatility columns the change is genuinely invisible, not merely easy to miss.

The ratio

A short window against a long one is an expansion detector.

The useful construction is not either standard deviation alone but their relationship. σ12 is the slow reading and functions as the branch's established dispersion; σ6 is the fast reading and responds to recent behaviour. Their ratio therefore answers a question neither answers separately: is this branch currently more scattered than it has been? A ratio near one says recent dispersion matches the established level and nothing has changed. A ratio meaningfully above one says the recent period is wider than the branch's own history, which is the earliest available indication that the conditions producing the edge may be shifting. The longer window is the denominator throughout the system — the established behaviour is the reference against which the recent period is measured, not the other way around — and keeping that convention fixed is what allows the ratio to be read the same way everywhere it appears.

Why it discounts

The same expectancy over a wider distribution is a weaker claim.

An expectancy figure is an average, and the informativeness of an average depends on the spread of what it averages. Computed over tightly clustered outcomes it is a strong summary of what the branch does. Computed over widely scattered outcomes it is a weak summary of the same thing, and the uncertainty around it grows with the dispersion. This is why volatility functions as a discount rather than as an independent warning: it does not contradict the expectancy reading, it widens the range of realities compatible with it. Practically, an expanding ratio means the branch's next stretch has a materially wider set of plausible outcomes than its headline suggests, and committing capital on the headline is committing to the centre of a distribution that has quietly become less informative about its own tails. The discount is not pessimism; it is the correct handling of a number whose error bars have grown.

The branch dimension

Some branches are volatile by design, and the ratio handles that correctly.

The branches are not expected to share a dispersion profile. Trend No-Partial is the true fat-tail accelerator and accepts high variance as the price of geometric growth; Overflow is the stability branch and earns its role by being narrow. Comparing raw σ across branches would therefore penalise the accelerator for behaving as designed and flatter the stability branch for the same reason. The ratio construction avoids this entirely, because each branch is measured against its own longer-window history rather than against the others. An expanding ratio on the accelerator means it has become more volatile than the accelerator normally is, which is a genuine finding. A high absolute σ on the accelerator means only that it is the accelerator. Reading absolute dispersion across branches is one of the more common ways to arrive at a confident and completely inverted view of which branch is misbehaving.

What it changes

The response is to size differently, not to conclude differently.

An expanding ratio is not a verdict that the branch has stopped working, and treating it as one produces its own errors — branches get abandoned during ordinary widening and re-entered after the informative period has passed. The correct response operates on confidence rather than on belief. The branch's expectancy remains the estimate; what changes is how much weight that estimate carries when it enters a deployment decision, and therefore how much capital is committed on it. A branch reading strong expectancy with an expanding ratio is a candidate for continued participation at reduced conviction, held while the ratio resolves in one direction or the other. This keeps the operator in the branch's real outcomes, which is the only way the question gets answered, while limiting what a wrong answer costs.

• Ratio near parity: expectancy carries its full weight in the decision.
• Ratio expanding: same expectancy, reduced conviction, smaller commitment.
• Compare each branch to its own history — never to another branch's σ.

The key idea

Confidence is a measured quantity here rather than a felt one.

Discretionary allocation usually carries an unstated confidence term — a sense of how much a number is to be trusted, formed from how recent the evidence feels and how the last few trades went. That term is real, it is decisive, and it is unmeasured, which makes it the largest unaudited input in most trading decisions. The volatility block replaces it with something computed from the branch's own dispersion history, visible in the same view as the expectancy it modifies, and identical every time it is read. Making a confidence term explicit does not make it correct, but it makes it inspectable, and an inspectable input is one that can be argued with when it turns out to be wrong.