The Spotlight

Which asset class matters most right now?

ProductionVerdict · SANITY-VALIDATEDRecall · 6/6

The Lens home — the wheel, the class summaries, the one-pager rows · anchor metric: × its usual move

What it measures

Every series in the book is measured on its own clock — market prices day to day, monthly releases print to print. Each new reading becomes one number: how big was this change against the same series' typical change over the last two years. A 4× day for crude and a 4× print for payrolls are the same size of story, so slow data competes with fast data on equal terms. Series come grouped into 52 themes (spreads, factor returns, the default cycle…) across 8 asset classes. A theme's loudness is its loudest member scored against that theme's own three-year history — a theme that is loud every week stops being news, and every class has the same chance of reaching the top. Levels never score, only fresh moves do: a spread parked at an extreme goes quiet within days. That is the habituation law, built in so the wheel cannot get stuck on one chronic story.

\text{move}_i(t) = \frac{|\Delta x_{i,t}|}{\operatorname{median}(|\Delta x_i| \text{ over } \sim 2y)}

\text{score}_g = \operatorname{pctile}\big(\max_{i \in g} \text{move}_i \,\big|\, \text{own history, } \sim 3y\big)

\text{LQ}_c = \frac{\text{board share}_c}{\text{fair share}_c}

2.71× on a class node means that class holds nearly three times its fair share of today's board — the spotlight is on it. 1.0× is neutral, 0.0× means none of its themes made the board. On the one-pagers, 2.3× its usual move is the theme's loudest member against its own normal pace, Quiet (0.8×) means below usual, and 0.0× means the latest reading simply didn't change. The wheel recalculates every trading day and describes the recent past — never a forecast.

Limitations

T+1 on official data — FRED series publish a day behind market prices; the wheel publishes the freshest day on which ALL 8 classes score (in practice, yesterday's close).
Release cadence — a monthly series can only make news when it prints; between prints it reads 0.0× by design, not by malfunction.
Burn-in — a series needs about two years of history before its yardstick is trustworthy; young series stay out until then.
The tape validates sanity, not skill — on the hand-built event tape the expected class reached the top 2 in 6 of 6 testable windows, but a random class does so ~79% of the time over a random window. The claim is that the wheel points at the right things with honest receipts — not that it predicts.

In the product

Powers the whole of The Lens: the wheel's node sizes and rank order, the per-class summaries, the ×-kickers and row order on every class one-pager, and the per-theme receipts (“JNK ETF flows went from -0.50 to -0.90”) — all recomputed daily from the same artefact.

Grading & stability

6/6

Event tape
testable daily-clock windows — sanity, not detection skill

79%

Placebo base rate
top-2-any-day under a random class × random window null

5y daily

History
~2y burn-in before the first scored day

Validated on a hand-authored event tape over the daily clock, with the placebo base rate printed inside every report. Verdict SANITY-VALIDATED.

The Spotlight · The Lens methodology← all engines