Cricket OKF Conformance Levels

What Is a Conformance Level?

Cricket OKF conformance levels describe how completely a bundle implements the cricket OKF profile on top of Google OKF v0.1. They are:

• Honest — a bundle claims only the level it can demonstrate.
• Graduated — each level includes all requirements of levels below it.
• Implementation-neutral — any cricket data provider can use them, not just CricketStudio.
• Independently verifiable — each level's requirements are explicit enough to check with a validator.

Level 0 — Cataloged

The bundle uses Google OKF file format with cricket-domain type values. No provenance, no metric definitions, no citation rules.

Requirements:

• Every .md file has valid YAML frontmatter
• Every file has a non-empty type field using cricket OKF vocabulary
• Every file has title and description

Limitation: Claims cannot be cited independently. No source traceability.

Level 1 — Citable

Every file in the bundle can be cited with a stable URL and a declared source boundary.

Requirements (all Level 0 +):

• Every data-bearing file has canonical_page or resource (a citable URL)
• Every file declares source_boundary
• Every file declares license
• Every file declares source_system

What this enables: Journalists and agents can link to a specific file and disclose its source. Still no metric definitions or sample-size rules.

Level 2 — Evidence-Backed

Every data-bearing claim in the bundle has a traceable evidence reference, a declared freshness date, and is backed by explicit metric definitions and sample-size rules.

Requirements (all Level 1 +):

• Every data-bearing file has a provenance block with source and confidence
• Every file has last_verified (ISO date)
• Every file has dataset_version where applicable
• The bundle includes metric files for every metric cited in ranked claims
• Every metric file has: formula, sample-size floor, limitations, ranking rule
• The bundle includes methodology files for: sample-size floors, phase definitions, ranking eligibility
• Leaderboard/ranking files declare the floor and scope they apply

What this enables: Agents can safely answer questions with correct scope, sample size, and confidence. Claims are traceable to source and dated.

CricketStudio OKF reference bundle self-certification: ✓ Level 2 (June 2026) — superseded by Level 3 (July 2026)

Level 3 — Agent-Safe ✓ CricketStudio OKF reference bundle

The bundle provides structured, machine-readable claim objects that agents can parse without natural language extraction.

Requirements (all Level 2 +):

• The bundle includes machine-readable claim objects with stable cs_claim_* IDs, metric reference, value, sample size, date window, provenance, and canonical URL — served at https://players.cricketstudio.ai/claims.jsonl
• The bundle includes a published JSON Schema for the claim object — schema/claim-object.schema.json in this repo; canonical at https://okf.cricketstudio.ai/schema/claim-object.schema.json
• The bundle includes evaluation cases: https://players.cricketstudio.ai/evals/cricket-qa-v1.jsonl — 50 seed Q&A rows (target 1,000 via OKF dossier) with mustInclude / mustNotInclude assertion fields for automated LLM accuracy testing
• The bundle includes a machine-readable agent usage specification: https://players.cricketstudio.ai/trust-manifest.json — declares corpus scope, sample floors, provenance chain, metric registry link, and all Trust OS artifact URLs

What this enables: RAG systems, MCP tools, and agent pipelines can ingest structured claim objects without relying on prose parsing. Any LLM provider can run the benchmark to independently verify cricket factual accuracy.

CricketStudio OKF reference bundle self-certification: ✓ Level 3 (July 2026)

Level 4 — Auditable (Roadmap: v1.0)

The bundle is versioned, regression-tested, and independently auditable. External implementations exist.

Requirements (all Level 3 +):

• Versioned schema releases with breaking-change documentation
• Published conformance reports per release
• Automated regression tests on claim objects (not just schema validation)
• Published eval results showing pass/fail per eval case per release
• At least one external implementation or community validation of conformance

What this enables: OKF becomes a credible open standard. Implementors can certify conformance. Researchers and journalists can audit the standard's own compliance.

Conformance Checklist

Use this checklist to assess a bundle:

Level 0 Checklist

• All .md files have valid YAML frontmatter
• All files have a non-empty type value
• All files have title and description

Level 1 Additional Requirements

• All data-bearing files have canonical_page or resource
• All files declare source_boundary
• All files declare license
• All files declare source_system

Level 2 Additional Requirements

• All data-bearing files have provenance.source and provenance.confidence
• All files have last_verified
• Metric files exist for every cited metric
• Each metric file has: formula, floor, limitations, ranking rule
• Methodology files exist for: sample-size floors, phase definitions, ranking eligibility
• Ranking files declare floor and scope

Level 3 Additional Requirements

• Machine-readable claim objects with full schema — /claims.jsonl (players.cricketstudio.ai)
• JSON Schema for claim objects published — schema/claim-object.schema.json
• Eval cases published — /evals/cricket-qa-v1.jsonl (50 seed rows, target 1,000)
• Agent usage specification — /trust-manifest.json (corpus scope, floors, provenance, artifact links)

Level 4 Additional Requirements (Roadmap)

• Versioned releases with change docs
• Published conformance reports
• Automated regression test suite
• Eval results published per release
• External implementation or community validation

Self-Certification

To self-certify your bundle's conformance level:

1. Run the Level N checklist against your bundle.
2. Confirm every item passes (or explicitly acknowledge what is missing and why).
3. Publish your self-certification in your bundle's README or a conformance.md file at the bundle root.
4. Link to this specification page as the authority.

CricketStudio OKF self-certification:

How Another Provider Uses This

Any cricket data provider can implement the cricket OKF profile:

1. Use Google OKF v0.1 as the base format (Markdown + YAML, type field).
2. Follow the cricket OKF type vocabulary (spec/types).
3. Declare provenance per the provenance convention (spec/provenance).
4. Write metric files per the metric standard (spec/metrics).
5. Apply sample-size floors per the doctrine (spec/sample-size).
6. Self-certify at the level they achieve using this checklist.

No license fee, no registration, no approval required.

Related

• Type Vocabulary
• Provenance Convention
• Metric Standard
• Sample-Size Doctrine
• Claim Discipline
• Releases