What we publish, and what we don't
Every scoring product faces the same tension. Publish nothing and you are asking for trust you have not earned. Publish everything and the people you are scoring optimise against it by Tuesday.
Most tools resolve this by saying nothing and calling it proprietary. We would rather draw the line explicitly and tell you where it is.
What we publish
Every factor. You know all six conviction channels, all nine safety layers, every filter dimension, and what each one looks at. Nothing on a card comes from an input you have not been told about.
Every ceiling and every cutoff. Channel maxima, tier boundaries, penalty magnitudes, round-trip bands, the volume and ratio thresholds behind each trade label. The numbers on conviction, trade labels and the safety stack are the real ones.
Every measured claim, with its denominator. When we say a filter has an edge, you get the base rate it beat, the sample it came from, and whether it survived the test that removes the biggest winners. When it fails that test, we say so on the same page.
What we could not measure. Blank fields stay blank. An unmeasured round-trip earns no credit in the score and shows as absent on the card.
What we got wrong. The reason L7 exists is a real customer losing real money to a check we had not thought to build. That is written on the safety stack page, not buried in a changelog.
What we do not publish
The per-condition point increments inside a channel. You know Codebase is worth up to 25 points and that it reads repository, age, language, docs, paper and audit. You do not get the exact points for "the repository predates the token", because that is a number a launcher can optimise directly — and backdating a repository is cheap.
The distinction we are drawing: we publish what a token is judged on, and what disqualifies it. We do not publish the weights that would let someone assemble the minimum viable legitimate-looking token.
Live wallet identities. Where a labelled wallet's identity is public we may show it; where the label is derived from our own analysis, we show the behaviour, not the address list.
How this compares
Worth knowing what the rest of the market does, because "proprietary" is doing a lot of work out there:
| Approach | Who does it | What you get |
|---|---|---|
| Publish factors, hide weights | Credit scoring, most risk scores | You know what matters, not how much |
| Publish everything | Web performance audits, vulnerability scoring | Fully reproducible, fully gameable |
| Publish nothing | Most crypto smart-money and wallet-label products | A number and a brand |
We sit in the first row, with the ceilings published — a little more open than the convention, and deliberately short of reproducible.
If you think a specific number should be public, say so. This line is a judgement, not a law. The test we apply is simple: does publishing it help you evaluate a token, or help someone build one that games us? Where the answer is genuinely the first, we will publish it.
The thing we will not do
We will not show you a confident number where we have an unmeasured field. Every product in this space faces the temptation to fill a gap with something plausible, because a complete-looking card converts better than an honest one.
A card with a blank field tells you we did not know. A card that quietly substitutes a default tells you nothing at all, and you cannot tell which one you are holding. That is the difference this entire product is built on, so it is the one place we have no flexibility.
➡️ See also: Conviction score · Filters
Anything on this page unclear, or wrong?
Tell us →