Donn Bryan Julian
Abstract Dynamic exposure rules can appear effective simply because they reduce risky participation, not because they time exposure well. This study evaluates Pogi, a recursive FULL/PARTIAL/NONE controller that separates portfolio composition from total risky exposure and uses a non-executed shadow path to observe recovery during defensive states. Using daily cryptocurrency data from 2014–2026, eight chronological test folds, recursive transaction costs, and CRRA certainty-equivalent welfare, Pogi is compared with static scaling, volatility targeting, CPPI, drawdown throttling, moving-average control, fractional Kelly scaling, and an exact ex-post exposure-matched diagnostic. The analysis also tests initialization and memory sensitivity, timing nulls, search capacity, selection-aware inference, and external validation using U.S. industry portfolios and frozen cross-market transfer. The completed evaluation did not establish robust welfare superiority for Pogi or support a broader methodological contribution under the pre-specified evidence criteria. The results instead show why dynamic exposure rules should be judged against exposure-matched benchmarks, model-search controls, recursive-state diagnostics, and genuinely external validation.