Evidence Engine:
source processing, not an oracle.
The Evidence Engine is GambleLABS Network's source-processing and verification system. It reviews documents already registered as sources, extracts the entities they name, maps proposed relationships, records contradictions and preserves provenance for every supported claim.
The stages, in order: source processing, entity extraction, relationship mapping, provenance, verification and publication review. Nothing skips a stage and nothing bypasses the last one.
Then it stops. Everything it produces is a proposal addressed to an editor, held in an internal queue, invisible to readers. It has no knowledge of the industry that does not arrive as a supplied document: it may not state a fact its sources do not contain, and recognising a company's name is not treated as evidence about that company. Every claim it makes carries the source it came from, or it is discarded.
- Not a prediction engine
- It does not forecast. It will not tell you whether an operator is about to fail, whether a licence is likely to be revoked, or what a regulator will do next. It reports what named sources already state, in the past tense, and nothing else.
- Not a system for winning at gambling
- It produces no tips, odds, predictions, strategies, staking plans or recommendations about where or how to bet. Nothing it does can improve a gambling outcome, and nothing on GambleLABS NETWORK is designed to help anyone gamble more successfully.
- Not a chatbot
- There is no conversational surface anywhere on this platform. It answers no reader questions and holds no dialogue. Its only input is a fixed set of registered source documents supplied with a written instruction by an authorised account.
- Not permitted to publish an accusation without human review
- No finding reaches a reader until a named editor has approved the run that produced it. There is no confidence score high enough to bypass that, no automatic approval after a delay, and no threshold above which the gate opens by itself.
- Not a source of facts of its own
- It may not state anything the supplied sources do not contain, and it is instructed not to draw on prior knowledge of any company, brand, domain, licence or person. A run that cites a source it was never given is discarded in full rather than partly kept.
01The pipeline
Eight stages, one of them human
A document travels through these in order. With no provider connected there is no run to count, so the diagram documents the route a finding must take to reach publication and states nothing about volume.
No analysis provider is connected to this installation. Nothing is generated in the meantime, and no analysis on this platform is ever synthesised to fill a gap.
Once a provider is connected, this page publishes completed runs that a named editor has approved: each finding with the registered source documents it cites, its evidence status, its score and the editor who approved it. The diagram below is the route such a finding must take. It is a description of the method, not a running system.
Read the Evidence Engine methodology- 01
Source intake
Every document is registered as a source record carrying its publisher, retrieval date and archive link before any analysis starts; a reviewer can open the full list of sources admitted to the run.
- 02
Verification
Each source is checked back to the body that issued it and graded for reliability; a reviewer can see which sources were graded as primary official records and which were downgraded.
- 03
Entity extraction
Companies, licence holders, providers and authorities named in the verified sources are extracted as candidate entities; a reviewer can read every extracted name beside the source it was taken from.
- 04
Relationship mapping
Extracted entities are linked into ownership, supply, hosting and payment relationships; a reviewer can trace the path of each proposed edge and the sources offered in support of it.
- 05
Context
Proposed findings are set against the records the platform already holds and the regime of the jurisdiction concerned; a reviewer can inspect each contradiction the engine raises against the existing record.
- 06
Confidence
Every finding is scored from the reliability and number of the sources behind it, never from the engine’s own certainty; a reviewer can see the score attached to each individual claim.
- 07
Human review
An editor approves, revises or rejects the run in full; nothing proceeds without a recorded approval, and the reviewing account and timestamp are stored against the analysis.
Mandatory human gate - 08
Published intelligence
Approved findings are attached to the record they describe and carry their sources with them; a reader can open exactly the same source list the reviewer worked from.
Reader-visible
02Scope
What the engine analyses
A definition of the material the engine is permitted to work over. It is a statement of scope, not a claim about volume: the section indexes report how much of each kind the platform actually holds.
- 01Player reports
- Reports submitted privately by players. They are analysed only inside the editorial console, to help a reviewer see whether several unconnected reports describe the same operator or the same payment route. No analysis of a report is ever reader-visible, and a report becomes public only as a redacted case file a reviewer writes deliberately.
- 02Companies and operators
- Entity dossiers: the corporate body, its registered and trading names, the brand a player believes they are dealing with, and the licensee actually named on a regulator’s register.
- 03Domains
- Hostnames attached to an operation — their registrar and registration records, their status, and whether an authority has named them on an unauthorised-operator list.
- 04Providers and infrastructure
- Platforms, white-label providers, aggregators, game studios and hosting providers: the layers a consumer-facing brand is assembled from and rarely discloses.
- 05Payments
- Payment providers and the routes deposits and withdrawals travel along, including the intermediaries that sit between a player and the operator they believe they are paying.
- 06Licences
- Licence records as they appear on a regulator’s own register: the number, the type, the holder’s name, the status and the date that status changed.
- 07Jurisdictions
- Country and territory profiles — the regime in force, the authority responsible, and the register and warning list that authority publishes.
- 08Regulatory records
- Enforcement notices, suspensions, revocations, penalties, public warnings, licence conditions and court judgments, each taken from the issuing body’s own notice.
- 09Public sources
- Every document registered as a source: official registers, regulatory notices, court records, corporate filings, company statements, press releases, reporting, academic and technical material — each with its publisher, retrieval date and, where one exists, an archive link.
- 10Investigation documents
- The evidence attached to an investigation or a case file — register extracts, correspondence, transaction records, screenshots and technical artefacts — in the state a reviewer collected and graded them.
03Explainability
Every finding shows its working
A finding a reader cannot check is not evidence, it is an assertion. Every published Evidence Engine-supported finding carries all seven of the following, on the record it describes — the investigation, research note, company dossier, case file or regulatory action — without the reader having to ask. A cross-record run describes no single record, so its findings are published in the output list further down this page.
- Source references
- Every source the finding cites, resolved to the record it came from. A finding that cites no source is not stored and cannot be displayed; a run that cites a source it was not given is discarded in full.
- Confidence level
- A score from 0 to 1 derived from the reliability and number of the supporting sources — never from the engine’s own certainty.
- Evidence status
- The claim’s standing in the platform’s own evidence vocabulary: corroborated, single-source, unverified or disputed. It is derived from the record — how many sources carry the claim, and whether the run itself reports it as contested — and it never reads “verified”, because verifying a document against the body that issued it is a human act the engine cannot perform. Thin evidence is labelled thin.
- Contradictory information
- Where sources disagree, the disagreement is published beside the finding with both sources named. It is not resolved, averaged, or quietly dropped.
- Human-review status
- Who approved the run and when. An unapproved run is not readable on any public page, at any confidence level.
- Last-processed date
- When the run completed. A relationship recorded in March is not evidence about April, and the date says so.
- Relationship path
- For a relationship finding, the ordered chain of entities — brand, operator, licensee, parent — so a reader can follow the same route the engine proposed.
What a reader can always do
- Open every source the finding cites, in full.
- See the score, and read what the score does and does not mean.
- Read the contradiction, where the sources disagree.
- See which editor approved the run, and when.
- Ask for a correction — and see it published permanently.