The same live ATR is calm against one session baseline and extreme against another.

Why four references exist

An instrument does not have one volatility profile; it has one per session it trades in.

The sessions differ structurally rather than incidentally. Participation, liquidity, which side of the world is originating flow and whether scheduled events cluster in the window all vary, and they vary enough that the same instrument produces materially different typical ranges depending on the hours observed. Comparing a live reading against an all-sessions baseline therefore compares it against an average of conditions that never occur simultaneously. The four-reference design exists so the operator can ask the question that actually matters: is this reading unusual for the hours in which the trade will live.

Which reference governs

The session the trade will be managed in, not the one the reading was taken in.

The selection rule is simple and frequently got wrong. The relevant baseline is the one covering the hours during which the position will actually be open and managed, because that is the environment the stop has to survive. A reading taken at the end of one session and used to size a trade that will be held into the next has been compared against the wrong reference, and the error is directional — it will typically understate the width required if the position is moving into a more active window, and overstate it moving the other way. When a trade spans sessions, the honest reference is the more active of the two.

Reading the disagreement

A wide spread across the four verdicts is itself a classification.

When the four references broadly agree, the instrument is behaving uniformly and the choice of baseline barely matters. When they diverge sharply, something specific is happening: the instrument's activity has concentrated into particular hours, which usually means a scheduled event, a regional catalyst or a liquidity condition. That state has management consequences beyond stop distance — it argues for participating in the concentrated window rather than around it, and it argues against holding through the quiet hours on a trade whose thesis depends on continuation. The spread is available on the same screen as the readings and is easier to interpret than any of them alone.

• Agreement across references means the baseline choice is not load-bearing today.
• Sharp divergence indicates activity concentrated in specific hours — usually for a reason.
• The spread argues about participation timing, not only about stop width.

The all-sessions trap

The most convenient reference is the one that is wrong most often.

The all-sessions baseline is the default in most workflows because it is the one that always has the deepest sample and therefore the most mature status. That combination — most convenient, best-supported, and structurally averaged across conditions that never co-occur — makes it the most quietly misleading field on the panel. It is genuinely the right reference for an instrument traded across all hours without preference, and almost nobody trades that way. An operator with a session preference should be reading their session's baseline and treating the all-sessions figure as background.

The sample trade-off, again

Session baselines are thinner than the all-sessions baseline by construction.

The four references do not have equal standing, and it connects directly to baseline maturity. Slicing by session divides the observations, so a session baseline is always less mature than the all-sessions one for the same instrument, and a session-plus-timeframe slice is thinner still. The correct reading is therefore two-stage: check the sample status of the session baseline you intend to use, and if it is not yet decision-grade, fall back to the broader reference while knowing it is averaged. The fallback is a compromise rather than an equivalent, and it should be recorded as one.

The key idea

A ratio is only as meaningful as the denominator you chose, and the denominator is a decision.

The matrix presents four ratios because there are four defensible denominators, and refusing to pick one for the operator is the honest design. Every volatility judgement is implicitly a comparison against something, and most operators make that comparison against an unexamined mental reference — usually the last few days of the chart in front of them. Making the reference explicit, offering the four that matter, and letting them disagree in public is what converts a vague sense that the market feels busy into a statement with a denominator attached.