fix(statements): republish the 2026-07-28 weekly, and render the denominator

[tkowalczyk]

Closes the last open item on auditmos/ogsfrompoly#236 — 2026-07-28-weekly.md has
been off the live site at draft: true since #58.

Republishing

resolved_count landed in #59, so an unresolved window is finally expressible.
Adding resolved_count: 0 clears findHitRateInconsistencies, which is the guard
that held this file as a draft:

before:  alert_count=263 with hit_rate=0 but hypothetical_pnl_usd=199.11 > 0
         (0 resolved alerts collapsing 0/0 → 0?)
after:   clean

No number changed. The frontmatter is exactly what #58 landed — regenerated by
poly-track report public weekly at auditmos/ogsfrompoly@554f537 over the identical
window. Its summary already read "0 of 263 resolved", and that sentence is the
count, summed over every category; the LP-mapped subset it bounds can only be 0 too.
So the new field states what the prose already claimed, rather than asserting
anything new.

Why this PR also touches the render

Flipping draft: false alone would have put Hit rate 0% back on the homepage,
the statements index, the statement page and the social card — beside this stat's own
caption:

Share of resolved alerts that hit the predicted side. 0.50 ≈ a coin flip; above 0.50 is signal.

That is the precise misreading ogsfrompoly#236 exists to stop, sitting on the most-skimmed
surface, while the correction lives only in the summary paragraph. Republishing without
this would have re-shipped the bug in a quieter place.

	value	note
resolved_count: 0	pending	"No outcomes have settled yet — 0 of 263 alerts have resolved, so there is no rate to report. Macro markets typically resolve months after the alert fires."
resolved_count: 240	47%	standing note + "Measured over the 240 of 263 alerts that have resolved."
absent	47%	standing note, unchanged

Three properties the tests pin:

• Keyed off the denominator, never the rate — a genuine 0-in-favour-of-240 week
  still reads as a real 0%, not "pending".
• Absent means unknown, never zero — the back catalogue renders byte-identically.
  Verified across all twelve statements, not just the changed one.
• The standing caption is replaced, not appended to — "above 0.50 is signal" reads
  as a verdict on the number beside it, so it must not survive next to a vacuous rate.

A known non-zero denominator is now shown as well. Publishing a rate without one is
what let the original 0.00 pass unnoticed for a week.

Verification

Full Verify gate run locally:

• pnpm test — 416 passed (408 before; 7 new + 1 content case the now-published statement adds)
• pnpm lint — clean, 115 files
• pnpm knip — clean
• pnpm types — 0 errors, 0 warnings (1 pre-existing hint in DisplayHeading.astro, untouched)

Rendered value for every statement in the collection:

2026-05-31-weekly    LIVE   51%      | Share of resolved alerts that hit the predicted side…
2026-06-09-weekly    LIVE   52%      | Share of resolved alerts that hit the predicted side…
…
2026-07-21-weekly    LIVE   50%      | Share of resolved alerts that hit the predicted side…
2026-07-28-weekly    LIVE   pending  | No outcomes have settled yet — 0 of 263 alerts have resolved…

Producer half is auditmos/ogsfrompoly@6c6328f; no address leak (all wallet references
remain truncated ids).

Refs auditmos/ogsfrompoly#236