# 2026-07-10 — Interpretation addendum: AZ boundary tie-break, [85.0, 85.1)

Pre-registered **before** the 2026-07-21 Arizona result exists, and anchored independently (see
`calibration/ANCHORS.md`). This addendum does NOT modify, reinterpret, or supersede any anchored
text. `calibration/hypothesis_log.md` and `calibration/shadow_adjustments.json` remain byte-identical
to their OpenTimestamps-anchored digests; nothing here touches the lock, `locks.log`,
`certified_results`, or any client-facing number.

## The ambiguity

The anchored AZ interpretation policy (`hypothesis_log.md`, 2026-07-09) fixes the raw lock at Biggs
**91.9** and the pre-registered shadow at Biggs **78.3**. It states:

> **REJECTED — overshrink** if `actual >= 85` (above the raw/shadow midpoint) … (`shadow_mae > raw_mae`)

Two clauses that are meant to coincide do not, on a narrow interval. The exact raw/shadow midpoint is
`(91.9 + 78.3) / 2 = 85.1`, not `85.0`. Therefore, for a certified two-way Biggs share in
**`[85.0, 85.1)`**:

- the **bright line** (`actual >= 85`) says REJECTED — overshrink; while
- the **parenthetical MAE gloss** (`shadow_mae > raw_mae`) is FALSE — the shadow is marginally
  closer. At `actual = 85.0`: `raw_mae = 6.90`, `shadow_mae = 6.69`, shadow ahead by `0.21`.

Outside `[85.0, 85.1)` the two clauses agree, and this addendum has no effect.

## The tie-break, fixed now

For the AZ pair (`us_arizona` / `us_az_gov_r_primary_2026`, lock `72bda729`), if the certified
two-way **Andy Biggs** share lands in **`[85.0, 85.1)`**:

1. **The anchored bright line governs.** The `actual >= 85` threshold is the operative test.
2. The outcome reads **REJECTED — overshrink**.
3. The shadow earns **no promotion credit** in that band. The pair does not count toward the three
   prospective in-regime pairs the anchored promotion rule requires, and the transform stays
   `unvalidated_shadow`.

This holds **notwithstanding the parenthetical MAE gloss**, and notwithstanding that
`sonar score-cycle` will still print the true `raw_mae` / `shadow_mae` comparison on the day (it
reports what it measures; the reading of that measurement is what this addendum fixes).

## Rationale

Conflicts resolve against our own hypothesis.

## Scope

Applies to the AZ 2026-07-21 pair only, on the interval `[85.0, 85.1)` only. It creates no new
threshold, relaxes none, and grants the shadow nothing it did not already have. A future contest
needing a midpoint rule must pre-register one before its result exists, as this one was.