Largest peak-to-trough decline of an equity curve: MDD = max_t ≤ Tleft[ dfracmax_s ≤ t W_s - W_tmax_s ≤ t W_s right]. Path-dependent and grows roughly with √(T) under iid Gaussian returns (Magdon-Ismail-Atiya scaling) — longer histories almost always contain deeper drawdowns. MDD drives institutional allocator decisions far more than volatility because redemption clauses and risk budgets trigger off drawdowns, not realised vol. Calmar = annual return / |MDD|.