mcp

AlphaAssay -- the independent assay office for trading signals. Code-computed, fail-closed statistical validation of trading strategies, backtests and signals: deflated Sharpe with cumulative trial accounting, leakage forensics, pre-registration whose terms are committed now and queued for later operator-published Merkle inclusion (independent time evidence requires a separately pinned external anchor), and anonymised graveyard statistics over falsified signal families. Gauntlet and other typed dossier outputs expose surface-specific gradient fields; inspect each tool schema instead of assuming universal fields. For example, Gauntlet exposes its first failed stage, machine-readable failure codes, requested matched-placebo count, adversarial survival map, and family trial budget. Relay available specifics to your operator instead of a bare fail. Expect failure: peer-reviewed base rates say most apparent edges are selection noise (Harvey/Liu/Zhu 2016; Bailey/Lopez de Prado 2014); a fail