Coverage criteria

Every candidate is measured against the same set of hard checks, every day, by the same automated pipeline. Nothing is scored, weighted, or judged at this stage — each check is a number against a threshold, and a token either clears it or it doesn’t.

A token that clears every check is a candidate, not coverage. A human decides what gets covered; nothing publishes automatically.

Three public sources, each used for what it is best at:

• Jupiter — the daily candidate feed, holder counts, top-10 concentration, mint and freeze authority status, and deployer history.
• DexScreener — per-pair liquidity, 24h volume, pool creation dates, and project social links.
• RugCheck — per-pool liquidity lock percentages.

When sources disagree on liquidity, we record the lower number. A filter that quietly takes the flattering source is not a filter.

A token’s age is the earliest pool timestamp any source reports, so a token cannot look younger by adding a new pool. In published coverage, every figure is linked to its source at the point it appears, dated on the day it was measured.

Any single failed check disqualifies — there is no compensating a failed check with a strong one elsewhere. The disqualifiers that end the most candidacies:

• A live mint authority (someone can print more supply) or live freeze authority (someone can freeze your tokens). These are the two we will never waive.
• Unlocked liquidity: under 90% of material-pool liquidity locked, or a deepest pool under 95% locked. Withdrawable liquidity is rug risk, and we weigh money, not pool counts.
• A deployer wallet with a history of serial launches or migrations.
• Concentration: a top-10 that can move the market on its own.
• No verifiable public presence.

Unknown counts as failed. If a source cannot tell us a number — holder concentration, age, lock status — the token fails that check rather than getting the benefit of the doubt. Every failure is recorded with the measured value that caused it, and the full failure log is kept for every candidate we have ever measured, not just the ones that pass.

Coverage is assigned a lane at approval, by a human, and the lane is shown on the ledger. The lane states what kind of asset it is, so a reader never has to infer it:

• meme — memecoins. No claim of intrinsic value is implied, and where a project’s own materials say it has none, we repeat that plainly.

Lanes are added the same way criteria change: as a versioned, dated addition — never by quietly reclassifying existing coverage.

Coverage does not end at publication. Every covered token is re-measured on a schedule: market cap and liquidity now, alongside market cap and liquidity on the day of coverage, both shown on the ledger. The comparison is the grade — we do not editorialize it.

Three rules make the record trustworthy:

• Nothing is deleted. Coverage that aged badly stays on the ledger with its numbers.
• Corrections are published, dated, and show the original figure next to the corrected one. The post is amended, never removed. Policy: /research-corrections.
• A covered token is not covered again within 90 days, so the ledger cannot be padded by re-covering winners.

Clearing the checklist means measured and not disqualified. Nothing more.