Disclaimer and evidence policy

Version 2026-11-03. The short form appears on every report (nocap/disclaimer.py).

What we quote, and how we prove it

• Only what was published. Every product name, trading term, claim, rule or figure we attribute to a vendor, signal channel, EA seller or firm is taken from material they published publicly (their website, marketplace listing, channel posts, public statements), or from a file supplied to us by the person asking for the audit or by the vendor. We do not paraphrase a claim into something stronger than its words.
• A dated, unaltered copy. When we capture a source we keep the exact bytes the server sent, the address we asked for and the address actually served, the HTTP status, the server's own Date, Last-Modified and ETag headers when given, and our capture time in UTC.
• A fingerprint. Each copy is stored under its SHA-256 hash. Any later change to the copy changes the hash.
• A trail that can't be quietly edited. Every capture, upload and quote is an event in an append-only, hash-chained log: each event contains the hash of the one before it, so changing or deleting any past event breaks every link after it. The chain's latest hash is anchored daily outside our servers (OpenTimestamps on the Bitcoin blockchain), so we cannot rewrite the trail without it showing.
• Quotes are checked by machine. A quote is only accepted if its words are found in the stored copy. A report cannot show a vendor claim that has no captured source; it shows "not shown" instead.
• Footnotes. Each quoted item carries a footnote: where it was published, when we captured it, the start of its SHA-256 fingerprint, and its evidence number. On request we provide the full copy and fingerprint so anyone can check it.
• Robots and privacy. We capture public pages only, honour robots.txt, never log in to a vendor's site to capture it, and never capture private messages. Where a site blocks automated capture we ask for a PDF or screenshot and record it as an upload.

What a grade is, and isn't

• A grade (A to D, or "Not enough evidence") is our analysis of the captured evidence using the published method in METHOD.md. The same evidence always gives the same grade, and payment never changes a grade.
• It is not financial advice, not a recommendation to buy, sell or hold anything, and not a promise about future results. Trading leveraged products can lose more than the money put in.
• Wording stays factual: "not supported by the evidence", never "scam" or "fraud".

Right of reply and corrections

• Anyone whose product is graded on a public page can reply ([email protected]); the reply is shown next to the verdict.
• If they show us a genuine mistake, we correct it as a new, dated entry in the trail. The original stays visible in the history, so corrections are open, not silent.
• Sources that are taken down stay in our trail: the copy proves what was published at the time.

Legal review (UK first) of this wording, the verdict language and the right-of-reply process is still recommended

before public launch; the plan budgets about £1–2k for it.