How a signal earns a capital allocation

Why we don't headline live returns. A short window of live performance says far more about market regime than about signal quality. Our research models are presently beta-driven and mid-restructure, and validated alternative-strategy books (for example, options income) remain in research validation and are not yet capital-allocated. Advertising a recent return or alpha figure would misrepresent what has actually been proven.

Pre-registration. Every signal is registered with a hypothesis, explicit null conditions, a required sample size, and a decision threshold — all frozen before observations are scored. This removes the degrees of freedom that let in-sample fitting masquerade as edge.

Forward validation. Candidate signals accrue genuine, prospective, out-of-sample observations from a live research book. No backtested or hypothetical returns enter the validation record. Over 65,000 prospective observations are tracked across the registry.

Promotion bar. A signal is promoted to live position sizing only after it clears the Harvey (2016) t>3 threshold with multiple-testing correction on a frozen sample. Signals that fail are retired and logged in the research record — not quietly re-tuned until they pass.

Audit trail. Hypothesis registrations, frozen thresholds, and outcome logs are retained for independent verification upon request. Nothing here is investment advice or a performance representation.