The verification rail
systematic finance
was missing.

The market wasn't waiting.
It was blocked.

This market wasn't waiting on demand. It was blocked. To prove an edge was real you had to reveal the strategy, and once you revealed it, it wasn't yours. To keep it secret you had to ask for trust. Allocators have lived inside that trade-off for thirty years, and no amount of diligence dissolves it, because the constraint was mathematical rather than commercial.

Zero-knowledge proofs dissolve it. That cryptography only became practical, at this cost and this speed, in the last two years. We didn't invent it. We pointed it at systematic finance: seventeen conditions checked on a single credential, a 220KB receipt, seconds to verify.

The constraint that made this market impossible is gone. The standard is unclaimed.

When MIZAN is the standard,
this is what an allocator sees.

One screen. The engine is asset-class agnostic and verifies any market, the trend family verified live today, with more markets as managers submit them. In the built product, strategies are funded through the allocator's own brokerage account. Custody never leaves them. They see their own fills — never the manager's logic; confidentiality covers the position stream itself.

Don't trust us.
Run the verifier.

The credential is a cryptographic receipt. Anyone can re-run it locally and confirm, without MIZAN, the quant, or the data provider in the loop. Click below to run the audit on our first issued credential.

New: you can now re-verify a credential entirely in your browser: client-side, no server, including the STARK seal itself (proof that the computation actually ran, not just the numbers), with no trusted setup. Try it at studio.mizan.market →

And we don't just defend the engine — we attack it: an adversarial red-team ledger with named findings (seed-grinding, rounding-direction, floor-slack), each one fixed and published in full. Read the attack ledger →

Newer still: chained tracks. A credential is no longer frozen at mint: the same sealed strategy is re-proven over a strictly longer window as data accrues, each link cryptographically bound to its parent (same spec commitment, Merkle-prefix window, same locked gate). A backtest that grows into a proven live track. Re-verify the first real chain yourself →

The credential also carries a Deflated Sharpe: Bailey & López de Prado's correction for the trials you ran to find a result. Held to the honest 0.95 significance bar, MIZAN's own flagship credential reads "NOT significant at 95%": and the credential says so, out loud, by design. A verifier you can't fail is theater. As of 2026-07-22, the two further López de Prado tests are live too: minted as real STARKs on GPU and re-verifiable at mizan.market/verify with the source-available verifier: PBO (probability of backtest overfitting, held to ≤0.5, the flagship sample reads 0.0759 at S=16 — all 12,870 balanced partitions, López de Prado’s own published setting, not overfit) and CPCV (purged, embargoed combinatorial cross-validation for leakage, headlined by the 5th-percentile path Sharpe, the sample reads +0.003). All three sins López de Prado named — selection bias, overfitting, leakage — are now cryptographic credentials rather than roadmap — to our knowledge the first time the full framework has been made re-verifiable this way. The honest caveats stand: the trial count N is forward-only (it counts what you ran after the ledger opened, and can't recover experiments from before it), the Deflated Sharpe is computed at an autocorrelation-corrected effective sample size (n_eff) — serial dependence makes the test harder, never easier, and no strategy has yet cleared all three bars, the flagship still fails the 95% Deflated-Sharpe test, in public.

Some strategy classes MIZAN refuses on principle — HFT, permanently. A verification layer is only as valuable as what it refuses to bless.