Ledgerproof — Real-World Validation

Public reproducibility artefacts for Ledgerproof's first real-world validation run. This repository contains only evidence — the recorded output of running the frozen Ledgerproof risk-decision engine against a real, publicly published gambling dataset. It contains no engine or product source code.

Result, in one line

On a real, payments-only public dataset (138,928 transactions), the frozen engine (casino-risk-api@0.1.0, unchanged) returned a Compatibility Report with zero findings and no runtime error — the correct, honest result. Payments data (deposits and withdrawals) carries no bets, decisions, or risk signals, so no risk finding can be evidenced. The engine produced none: it declines to fabricate what the data cannot support.

What this validates — and what it does not

Shows

• The engine runs on a real public dataset without modification.
• Evidence discipline holds: no required data ⇒ no finding, no fabricated number.
• It fails safe — a Compatibility Report, not a crash and not a false positive.
• The run is reproducible: a fixed input (SHA-256) → a recorded summary and log.

Does not show

• This is not a certification and not a passed audit.
• Payments-only data cannot demonstrate fraud, capital, or player-protection detection quality. That is measured on bet-level operator data during a pilot.
• No commercial, regulatory, or ROI claim is made from this dataset.

Dataset & attribution (CC BY 4.0)

Ghaharian, K., Puranik, P., Abarbanel, B., Taghva, K., Kraus, S.W., Singh, A., Feldman, A., Bernhard, B. (2023). Raw payments transaction data from online casino players and online sports bettors. Mendeley Data, V1. doi:10.17632/9j5gcygnwg.1 · Companion article: Data in Brief v48, 109214. Licensed CC BY 4.0.

The raw CSV is not redistributed here — download it from the Mendeley landing page and verify it against CHECKSUMS.sha256.

Files

Path	What it is
SOURCE.md	Provenance, licence, and attribution for the dataset.
EXPECTED_OUTCOME.md	The predicted outcome (a schema analysis) and the observed authoritative result.
DEVIATIONS.md	The two documented data-vs-paper metadata differences.
CHECKSUMS.sha256	SHA-256 of the exact source file that was run.
mapping.json	The source-column → canonical-field mapping used (no engine logic).
results/real-data/validation-summary.json	Machine-readable run summary.
results/real-data/run-log.txt	The run log (counts, coverage, outcome).
results/real-data/report.html	The rendered Compatibility Report.

Reproduce

1. Download Online_casino_DIB.csv from the Mendeley landing page (https://data.mendeley.com/datasets/9j5gcygnwg/1), respecting the CC BY 4.0 licence and the site's terms.
2. Verify the file: sha256sum Online_casino_DIB.csv must equal the hash in CHECKSUMS.sha256. If it differs, you have a different file — stop and note it.
3. Review mapping.json and EXPECTED_OUTCOME.md: the file is payments-only (ReqTimeUTC, TransactionType, TransactionAmount, Status, AccountIdentifier), so every risk-analysis module is UNAVAILABLE and the correct outcome is a Compatibility Report with zero findings.
4. Confirm against the recorded artefacts in results/real-data/: row count 138,928; deposits 68,599 + withdrawals 70,329 = 138,928; outcome COMPATIBILITY_REPORT; findingsFired: []; runtimeError: null.

The Ledgerproof engine (casino-risk-api@0.1.0) is a pinned private dependency; these artefacts let anyone independently verify which bytes were analysed (by checksum) and what the engine produced (the recorded summary, log, and report). The correct result on payments-only data — no findings — is verifiable from the data's shape alone.

Amounts

Monetary figures in the artefacts are in the source file's own units; the file carries no currency field, so no currency is claimed. Deposit and withdrawal totals are descriptive context, not a Ledgerproof output.