T-002 FREEZE RECORD

Document title: T-002 Preregistration v1.0
Document version: Preregistration v1.0
Package revision: Final supplied implementation clarifications integrated
Package creation timestamp (UTC): 2026-09-20T19:30:00Z
Phase: Phase 0 only
Status: DATA-FREE VERDICT EXECUTABILITY AUDIT PASS; immutable registration pending.

SHA-256 HASHES (exact final file bytes)

File: T-002_Preregistration_v1.0.md
SHA-256: fa1303478361ff6fe2aaf465fd5af15ac756489e7204101b4d71209a4de8f982
Size (bytes): 54147

File: T-002_Preregistration_v1.0.pdf
SHA-256: c119271f936dd6854fc4005337bd13c550e74b2497c6d029f03eebde47028848
Size (bytes): 98211

INTENDED DATASET VERSIONS
COW National Material Capabilities v7.0
COW State System Membership v2024
V-Dem v16
Polity5 v2018

These are intended execution versions, not downloaded or inspected datasets.
Source dataset-file hashes must be recorded at the registered execution stage.

CHANGE CONTROL
The six sections in Pasted markdown(20260920-192637).md were inserted verbatim:
- Leave-One-IDP-Component-Out Construction: Section 22.3
- Model Estimability and Non-Convergence: Section 10.1
- Usable Temporal Forecast Origin: Section 14.1
- Bootstrap Failure Rule: Section 16.1
- Feasibility of Verdict-Relevant Robustness Analyses: Section 22.4
- Exhaustive Meaning of Measurement Failure: Section 9.1

Two superseded sentences in Section 16 were reconciled with the supplied rule:
2,000 attempted resamples, and percentile intervals from successful replicates
subject to the newly specified minimum of 1,900 successful estimates.
Only Sections 9, 10, 14, 16 and 22 changed in this revision.
All other numbered sections and the Freeze Checklist remain unchanged.
The preceding approved verdict rules remain integrated in Sections 14, 21-23.
No additional methodological rule was selected.
These hashes supersede those of the preceding pre-registration package.

SUBSTANTIVE EQUIVALENCE
PASS. The PDF was generated from the exact final Markdown. Extracted PDF body
text matched rendered Markdown text after removal of Markdown styling, PDF page
footers and whitespace differences. All substantive text, formulae, thresholds
and checklist entries matched. All 22 PDF pages were visually reviewed.

REPEATED DATA-FREE EXECUTABILITY AUDIT
Result: PASS for the requested verdict-executability audit.

The consolidated classification rules are execution-deterministic within this
audit scope: the registered validity/availability conditions, statistics and
ordered rules determine exactly one evidence category without a further
qualitative verdict judgment.

Findings and resolving sections:
- Major robustness contradiction and its category consequence: Section 22.
- Unique, exhaustive evidence-category precedence and equality boundaries:
  Sections 21 and 23.
- Qualitative verdict terms replaced by explicit numerical conditions:
  Sections 14.2, 21 and 23.
- Pooled temporal decision rules, with no temporal interval or per-origin veto:
  Section 14.2.
- O* robust-opposition veto, distinct from corresponding-signal and
  good-governance exclusions: Sections 21.3-21.5.
- Leave-one-component-out subscale weights and fixed sample: Section 22.3.
- Estimability, unavailable temporal origins and their consequences:
  Sections 10.1 and 14.1.
- Failed bootstrap attempts and minimum successful estimates: Section 16.1.
- Robustness feasibility, fixed/subset samples and unavailable analyses:
  Section 22.4.
- Exhaustive measurement-failure conditions: Section 9.1.
- 100-year sensitivity cannot affect the final verdict: Sections 19 and 22.

Verification included all 16 Boolean combinations of the four ordered decision
predicates and 16 hypothetical numerical boundary cases. Every case received
exactly one category. These are logic checks, not fitted models, simulated
historical datasets or empirical results.

No further unresolved decision was identified within the repeated verdict-level
audit. This audit does not establish actual dataset-schema compatibility,
empirical feasibility, successful model fitting or completion of execution code.
The registered stop-on-implementation-ambiguity rule remains in force.

OUTCOME-ACCESS DECLARATION
No registered breakdown outcomes were accessed during creation or revision of
this freeze package. No datasets, historical case identities, breakdown counts
or outcome values were accessed, downloaded, inspected or summarised.
Only supplied design documents, user instructions and data-free logical checks
were used. No predictor panel was constructed and no models were fitted.
Outcome execution and Phase 1 have not begun.

REGISTRATION / TIMESTAMP STATUS
Registration URL/reference: PENDING; none supplied or created.
Immutable registration timestamp: PENDING.
The creation timestamp above records file generation, not independent proof of
immutable registration. The package and hashes alone do not satisfy all
outcome-access prerequisites.

PRESERVATION AND NEXT GATE
The frozen execution brief remains in Section 25. The Freeze Checklist is
included as requirements to confirm, not a certification that registration or
execution prerequisites have already occurred.
Preserve these exact file bytes for registration. Record the registration
reference and timestamp separately before outcome exposure. Source hashes,
outcome-blind predictor implementation and the remaining registered prerequisites
must also be satisfied before any registered outcome is accessed.
