Derived by the Research Lead from Bruno P. C. Levy and Hedibert F. Lopes: adapt the paper's dynamic logistic classifier and momentum-model selection to a long-short basket of LTC, BCH and ADA perpetuals. Preserve calendar-month forecasts, lookbacks of 1, 2, 4, 6, 8, 10 and 12 months, monthly holding periods, cutoff 0.5, coefficient-discount grid {0.98, 0.99, 1}, model-forgetting grid {0.99, 1}, and 36 labeled training months. Use daily bars to construct completed UTC calendar-month closes. This is an adaptation of the supplied full paper, not reproduction of its 56-contract results. Added implementation choices are zero-mean Gaussian coefficient priors with identity covariance, deterministic tie-breaking, and conservative exposure caps. Source volatility scaling is retained with a 40% annualized target before caps; retrospective scaling to 10% realized volatility is excluded from executable sizing. Monthly decisions execute at the completed signal bar's close, the platform's supported equivalent of next-bar-open execution on a continuous market. No exchange-hosted protection or selectable same-bar policy is required.