Glossary · the vocabulary of verifiable track records

The terms, defined plainly.

Sixteen definitions. Each one is the meaning used across this estate, the VTR-1 standard and the registry. Where a term has an academic source, the source is named.

Verifiable track record
A trading record whose returns, costs, out-of-sample split and trial count have been recomputed inside a cryptographic proof, so that anyone can check the record is real without seeing the strategy that produced it. The term is defined by the VTR-1 standard.
Trial count (N)
The number of strategies, parameter sets or variants tried before the reported one was chosen. Every correction for backtest overfitting divides by it. For forty-six years it was self-reported by the person being judged; in a MIZAN credential it is the leaf count of a committed Merkle tree and cannot be understated.
Deflated Sharpe Ratio (DSR)
Bailey and López de Prado, 2014. The probability that a backtest Sharpe ratio is significant after adjusting for the number of trials, the sample length, and the skew and kurtosis of returns. A high Sharpe with a large trial count deflates to noise.
Probability of Backtest Overfitting (PBO)
Bailey, Borwein, López de Prado and Zhu, 2014. The probability that the strategy selected in-sample underperforms the median out-of-sample, estimated by combinatorial cross-validation. A second, independent school of backtest honesty.
Register-before-run
The rule that a strategy's hash and its full trial set are committed before evaluation begins. It is the condition under which the proven trial count binds. Without it, N is only what the submitter chose to commit.
Zero-knowledge proof (ZK proof)
A cryptographic proof that a computation was performed correctly on given inputs, without revealing the inputs. MIZAN uses a STARK generated inside the RISC Zero zkVM: the strategy stays on the quant's machine, the verdict is public.
STARK receipt
The output of the prover. About 225 KB, it verifies offline in roughly 81 milliseconds and depends on no server, no trusted party, and no MIZAN infrastructure.
Credential
One record judged by the gate: dataset root, trial-count commitment, disclosed metrics, verdict, engine era, proof hash and anchor, published on the append-only registry. A credential can be a pass or a refusal; both are published.
Refusal
A credential whose record did not clear the gate. MIZAN publishes refusals with the same weight as passes; 54 of the first 77 credentials were refusals, including Apple buy-and-hold and the company's own flagship. A gate that passes everything certifies nothing.
The gate
The locked set of conditions a record must clear: after-cost Sharpe, drawdown, worst bar, walk-forward stability, deflated significance, independence from market beta. The verdict is decided in-circuit; no hand reaches inside.
Era law
Every engine version is a sealed era with its own verifier image. A credential verifies forever against the era that minted it. Eras are never relabelled and old credentials are never re-judged silently.
Live-chained track record
A strategy anchored to Bitcoin and RFC-3161 before the data it will be judged on exists, then re-proven on the growing window at published deadlines, each link bound to its parent. It answers the question 'was this built in hindsight' with a millisecond check.
Anchor
A hash of the credential committed to the Bitcoin blockchain (via OpenTimestamps) and to an RFC-3161 timestamp authority, proving the credential existed at that time.
VTR-1
The Verifiable Track Record standard: forty-one normative requirements, three conformance levels, conformance vectors and public errata, published on SSRN under CC BY. Anyone may implement it; nobody may overclaim it.
The trust layer for machine finance
MIZAN's position: the neutral verification rail every marketplace, vault and allocator screen can run on, and the credential that proves a strategy's record, human or AI, without revealing it. Built on proof, not disclosure.
The biggest solvable problem in quantitative finance
The deadlock in which a track record could not be checked without revealing the strategy, and a revealed strategy stops being an edge. Non-stationarity, capacity and crowding cannot be repealed; this deadlock could be, and was, with zero-knowledge proofs and a committed trial ledger.

Go deeper: the problem · how it works · the standard · the deflated Sharpe calculator · every credential.