Quantifying the unquantifiable

The problem: loud, unfalsifiable, unresolvable

Pick any consequential subject and look at what is written about it. There is no shortage of material — there is a surplus. Analysts, wire copy, think-tank briefs, threads, television panels. What almost none of it contains is a quantity.

Instead it contains adjectives: tensions are escalating, the debt path is unsustainable, the aviation network is strained, the race is tightening. Adjectives have three properties that make them useless for actually knowing anything:

• They do not compare. "Elevated tension" this week and "elevated tension" last March are the same phrase attached to two different worlds. You cannot subtract one from the other.
• They cannot be wrong. A claim with no number attached can never be checked against what happened next, so the people making them are never scored, and the confident ones win by volume.
• They hide the reasoning. When a verdict is a paragraph, you cannot see which input drove it, or what would have had to change to reverse it.

That is the noise. It is not that these subjects are genuinely unmeasurable — it is that almost nobody commits to a measurement, because committing is the part that can make you look wrong.

What one number actually buys you

An index is a promise: here is the rule I will use to turn this mess into a figure, and I am publishing the rule before I see the answer. Everything useful follows from the promise, not from the figure.

It makes the topic comparable across time. A reading of 128 this week against 112 in April is a statement with content. The number carries its own history, so trajectory — usually the thing that actually matters — becomes visible instead of anecdotal.

It makes the reasoning inspectable. A composite is a sum of parts, and every index here shows its parts. When the score moves you can see which input moved it, and you can then disagree with a weight rather than with a mood.

It makes the author accountable. A published series cannot be quietly revised after the event. The rule that produced today's reading is the rule that produced the readings which aged badly, and both stay on the page.

It moves the argument somewhere productive. The interesting disagreement stops being "is this serious" and becomes "is that the right weight for that input". That is an argument that can be settled.

What every index here has in common

The subjects have nothing to do with each other. The discipline is identical, and it is the reason a number here should be worth more than an opinion elsewhere.

A fixed, absolute definition of normal

Each index is scored against a concrete calm state, chosen once and never recalculated. The convenient alternative — scoring each input against its own rolling average — quietly rewrites history every time a new observation lands, and can only ever say "worse than usual", never "bad".

The rule is published before the answer

Inputs, weights and thresholds are decided in advance and written down. Changing one restates the entire series, so it is treated as a documented version change rather than a quiet correction.

The number always shows its parts

Nothing is published as a bare figure. The breakdown sits next to the headline, so a reader can rebuild the composite — or reject the composite and keep the components.

Degradation is visible, not smoothed over

When an input goes stale or a source fails, the affected component is flagged and the reading is marked degraded. A model that silently carries forward last week's value while presenting itself as current is worse than no model at all.

Every figure is free to reuse

Index scores and their history are published under CC BY 4.0. Cite them, replot them, argue with them.

What the inputs actually are, what each weight is, where each threshold was set and why — those are specific to one index, so they are documented on that index's own dashboard, beside the number they produce, rather than restated here.

What a single number cannot do

• The weights are judgement, not physics. They are argued on each dashboard, and they remain opinions — held constant, published, and open to being wrong.
• A composite can hide its own composition. Two very different weeks can score the same. That is exactly why the breakdown sits beside the headline figure on every page.
• The stress indices are one-sided. They measure stress against a calm baseline and have a hard floor at it, so they cannot express conditions genuinely calmer than normal.
• Some inputs are keyed by hand. Where no free machine-readable source exists, the value is entered manually and refreshes on a slower cadence. Every component states its own source and last-updated date.
• None of this tells you what to do about it. These are measurements, not advice — see the terms and disclaimers.

The indices, today

Each live series is one noisy subject collapsed into one figure. Every dashboard carries its own methodology — the full construction of that index, with the reasoning behind every choice in it.