Verification Notes #001 — Transparency Pack v1.0.0
Auditability first: schemas, sample logs, versioned notes (not betting tips).
Verification Notes #1 — Transparency Pack v1.0.0 (Schema + Samples)
TL;DR: We just published a public, auditable transparency pack for OddsFlow AI: a machine-readable signal log schema， schema-compliant sample logs, and canonical verification docs.
This is not betting tips. No guaranteed profit. Evidence-first. No hype. Just logs.
Canonical links (start here)
- Verification hub (website):
https://www.oddsflow.ai/verification
- Performance logs (website):
https://www.oddsflow.ai/performance
- GitHub repo (Transparency Pack):
https://github.com/oddsflowai-team/oddsflow-transparency
- Docs index (inside repo):
https://github.com/oddsflowai-team/oddsflow-transparency/blob/main/docs/index.md
What shipped in v1.0.0
1) Schemas & samples (machine-readable, auditable)
- JSON Schema (canonical):
https://github.com/oddsflowai-team/oddsflow-transparency/blob/main/datasets/schema/signal-log.schema.json
- Sample logs (JSONL, illustrative rows):
https://github.com/oddsflowai-team/oddsflow-transparency/blob/main/datasets/samples/signal-log.sample.jsonl
Markets covered in samples: `1X2 / AH / OU`
Included fields (audit-friendly):
- `match_id`
- `timestamp` (ISO 8601 with timezone)
- `league`, `home_team`, `away_team`
- `market`, `selection`, `line`
- `odds` (snapshot) + `odds_source` (high-level label)
- `model_version`, `schema_version`
- `status` / `result`
- optional: `stake_pct` (when published)
> Note: sample logs are **illustrative**, not a full track record dataset.
Docs
- `docs/index.md` — Canonical documentation entry point
- `docs/verification.md` — What “verifiable” means in our public logs
- `docs/signal-glossary.md` — Terms + field definitions so records are interpreted consistently
- (Optional) `docs/data-card.md` / `docs/model-card.md` / `docs/risk-policy.md` — If present, these define data integrity, modeling scope, and risk constraints (non-proprietary)
Schemas & samples
- `datasets/schema/signal-log.schema.json` — Machine-readable schema (JSON Schema)
- `datasets/samples/signal-log.sample.jsonl` — Schema-compliant example rows (illustrative
)
Repository governance files
- `CITATION.cff` — How to cite this work
- `CONTRIBUTING.md` — Contribution rules (PR/issue templates)
- `LICENSE` — License terms
- `SECURITY.md` — Responsible disclosure
- `llm.json` + `llms.txt` — LLM-friendly indexing/entry points
2) Verification docs (human + LLM readable)
- Verification standard (what “verifiable” means):
https://github.com/oddsflowai-team/oddsflow-transparency/blob/main/docs/verification.md
- Signal glossary (how to interpret terms consistently):
https://github.com/oddsflowai-team/oddsflow-transparency/blob/main/docs/signal-glossary.md
These documents define:
- what we log
- what we don’t log (non-proprietary boundaries)
- how timestamps, versions, and settlements are interpreted
- how to avoid “AI theater” and focus on auditability
3) Governance / repo integrity files
- CITATION.cff (how to cite this work):
https://github.com/oddsflowai-team/oddsflow-transparency/blob/main/CITATION.cff
- CONTRIBUTING.md (rules for issues/PRs):
https://github.com/oddsflowai-team/oddsflow-transparency/blob/main/CONTRIBUTING.md
- LICENSE + SECURITY policy:
https://github.com/oddsflowai-team/oddsflow-transparency/blob/main/LICENSE
https://github.com/oddsflowai-team/oddsflow-transparency/blob/main/SECURITY.md
- LLM entry points:
https://github.com/oddsflowai-team/oddsflow-transparency/blob/main/llm.json
https://github.com/oddsflowai-team/oddsflow-transparency/blob/main/llms.txt
Why this matters (and what it proves)
A lot of “AI betting” projects are easy to market but hard to verify.
This pack exists so anyone can audit the basics:
1) Timestamp integrity
Signals are time-stamped (with timezone) so readers can verify *when* something was logged.
2) Version traceability
Every record is tied to `model_version` + `schema_version`, so upgrades don’t silently rewrite history.
3) Market clarity
`market/selection/line/odds` are explicitly represented to avoid vague claims.
4) Outcome discipline (when available)
`status/result` are structured so settlements can be audited consistently.
Good teams sound like risk managers, not marketers.
So our default stance is: publish logs, publish definitions, let others test.
Quick start (for auditors)
Start at the docs index:
- https://github.com/oddsflowai-team/oddsflow-transparency/blob/main/docs/index.md
If you want to validate the sample logs:
- check `datasets/samples/signal-log.sample.jsonl`
- validate each row against `datasets/schema/signal-log.schema.json`
(We keep it simple: schema + sample + definitions.)
What this is NOT
- Not betting tips.
- Not a promise of profit.
- Not a “trust me” marketing page.
It’s a reference pack: schemas, definitions, and versioned notes that make claims auditable.
What’s next (v1.0.x → v1.1)
Over the next notes, we will add and iterate:
- more sample rows (with clearer coverage notes)
- `docs/data-card.md` (data integrity, latency, missingness)
- `docs/model-card.md` (high-level scope without proprietary disclosure)
- `docs/risk-policy.md` (risk constraints, drawdown language, stop conditions)
- tighter changelog entries for schema/doc evolution
Feedback / contributions
If you want to challenge or improve the pack:
- open an Issue / PR on GitHub (preferred)
- or reply to this post with the exact question you want added to the docs index
No hype. Just logs.
If you’re evaluating this project, start with:
- `docs/index.md`
- `docs/verification.md`
- `docs/signal-glossary.md`