Broad cross-sector value & global-vs-US screens

Claim

Five broad screens run across 2026. This note is about the screen type rather than any field, because the pattern across them is the reusable part.

Broad screens keep re-surfacing names that field-specific screens already covered, because both draw from the same listicle-and-vendor sources. The marginal information per token is lower, and the output overlaps work already done.

The one thing they do well is negative space — the March 19 adjacent-sector screen's most useful output was the portfolio-gap list (zero exposure to consumer defensive, industrials, utilities, materials, defense), not its shortlist. Run broad screens to find what's missing, then run a field screen to fill it.

The global screen's specific failure

screen-global-vs-us-value-2026-07-29 flagged Japan as discounted on EWJ −8.4%. The yen fell 8.41% over the same period and the TOPIX set a record close. The screen was measuring currency. Full correction in [[japan-tech]].

Unhedged country ETFs are not valuation signals. Any international screen using them as a starting filter inherits the FX move as if it were a discount.

What would falsify this

A broad screen surfacing a genuinely novel name that field screens missed. Worth tracking rather than assuming — the claim is about observed yield, and five runs is a thin sample.

History

• 2026-03-27 → 07-29 — five broad screens.
• 2026-07-30 — pattern recorded, including the global screen's FX error.