ALLOCATOR-V1: the gate class for books.
Everything MIZAN issues today certifies one strategy. An institutional allocator does not buy strategies; they buy books: tens of models, hundreds or thousands of instruments, and a research process that renews itself.
This page specifies the credential for that, shows the specimen it issues, and states honestly which fields exist today, which need a new engine era, and which no proof system will ever reach.
A complete credential, as it will read when the class ships.
It is a specimen: the holder is fictional and several fields are not yet issuable. No credential on the registry carries unproven fields, and none ever will; a specimen lives here, on a specification page, nowhere near a credential URL.
Fictional holder “Meridian Research LLP”. Figures illustrate the format, not a real record. The legend printed at the foot of the specimen marks every field as live today, pending an engine era, or gated on the world.
What is real, what needs a new judge, and what never comes.
Under era law an engine is superseded, never revoked: credentials verify forever against the era that judged them, and a changed judge is a new judge by definition. The live engine is era v12 (659bdb67). Several fields below therefore cannot arrive quietly; they require a published engine era with its own diff and review.
Trial count N · exposures · regimes · costs · out-of-sample · sealed path · the chainin the engine now
The trial count is the leaf count of a committed ledger, and the credential states plainly when a given credential does not bind one. Exposure and regime panels are derived in-circuit from the same position series the P&L compounds.
Holder record: the denominatorshipped 2026-08-29
Every credential a holder has ever minted, passes and refusals together, computed from the append-only registry. Opt-in and permanent once chosen; a holder who has not elected shows exactly that.
Breadth · velocity floor · book tier · extended regressors · regime floors · short-vol payoff · scalea governed engine release after v12
All require changes inside the guest circuit: a new image, re-proving, and a published review. They will ship as one governed era release rather than as quiet edits.
Adherence: did they trade what they provedneeds a counterparty attestation
Requires a counterparty willing to sign a position file: a prop firm, a pod platform or a prime broker. No amount of cryptography substitutes for someone attesting to what was actually traded. Until one exists, the field reads not attested, never inferred.
Percentiles: comparability across booksneeds many books on one gate
Meaningful only once enough books share one hash-pinned gate, identical cost treatment and identical drawdown marking. The comparison set must stay public and independently recomputable, or the percentile is void.
Correlation to the allocator’s own book, in zero knowledgeresearch-grade engineering
The allocator commits their return series locally; the holder proves correlation below a threshold the allocator chooses. Neither counterparty, nor MIZAN, learns the other’s data. A research-grade engineering project, and the second hard pair in this industry.
Capacity · crowding · book management · custody · counterparty competenceoutside every proof system
These depend on future liquidity, on people, and on everyone else in the trade. No proof system reaches them and no MIZAN credential will imply it does. They are printed on the artifact as refusals so that everything above them can be believed.
Why publish an unfinished specification. Because a standard other people are meant to build on cannot be a surprise, and the fastest way to get it right is to be corrected early. If your own screen carries a criterion missing from this page, we would rather hear it than guess: that is how the list was written.
Tell us what we left out →