Editorial policy

Independence

No asset, protocol, issuer, exchange or fund can pay for coverage, for placement, for a higher position in a table, or for the removal of anything. Coverage is decided mechanically by size. Advertising, where it appears, is served by a third-party network and has no relationship to editorial or data decisions and no ability to influence them.

No advice, ever

Nothing on this site is a recommendation, a suitability assessment, or a price view. We do not publish price targets, ratings, buy or sell language, or the vocabulary that stands in for it — undervalued, overvalued, cheap, expensive, attractive, promising. Where a figure could be read as an implicit ranking, the page states what the figure does not measure.

This extends to the tools. The screener returns a list, never a selection. The rankings order one measurable thing at a time and each carries a note on what it omits. The AI assistant is instructed to refuse questions about what to do.

Sourcing

Every figure names its provider and its reference period. Every news item names its publisher and links to the original; we never reproduce an article body, and press-release wires are blocked because a press release is not journalism. Items written as price calls — “could hit”, “price prediction”, “set to surge”, “best coins to buy” — are rejected regardless of publisher.

Historical claims in educational content are limited to well-established, uncontroversial events. Where a specific date or figure is uncertain, the mechanism is described instead of the statistic.

AI assistance, and how it is labeled

An AI language model is used in four places, all disclosed at the point of use:

• The educational library and reference definitions — drafted with a model against the rules on this page and edited before publication. The requirement that every metric definition include its own failure modes exists partly because it is the section a model is least inclined to write unprompted.
• Research notes — generated from stored figures. Each note lists the exact inputs it was given and is marked as AI-written on the note itself.
• Per-asset summaries — written from the figures on the same page and labeled underneath.
• The Ask the data assistant — reads a JSON extract of this site's own tables, has no internet access, no memory between questions, and is instructed to say so when the data does not contain the answer.

AI is never used to produce a data value. Every number is either read from a named provider or calculated from named providers' numbers by a formula published in the methodology.

Translation

Interface text is machine-translated into every language the site offers, from English source strings. Legal text remains in English deliberately: a translated disclaimer that drifts from its original is worse than one a reader has to read in English. Data labels and provider names are not translated.

Corrections

Errors are corrected at the source rather than patched at the point of display, because a wrong figure is almost always a wrong mapping or a wrong period rather than a wrong cell. When a correction changes a figure that was published, the change flows through every page that used it on the next refresh. Report one here — name the page, the figure and what it should be.

Missing data

A blank on this site is never a zero and never an oversight. Where a figure does not apply to an asset, the site says why. Where it applies but nobody publishes it, the site says that instead. Where we have decided not to publish a class of figure at all, it appears in the list on the methodology page with the reason.