Modern Portfolio Theory outlines a two-step wealth allocation process, yet accurately predicting asset returns and utilizing those predictions effectively remains a challenge. This difficulty stems from a weak link between forecast accuracy and economic value, a topic of debate among researchers. While some researchers question whether returns are predictable at all, others aim to enhance prediction methods by inter alia borrowing techniques from the realm of machine learning. Our study addresses the gap in how to systematically use predictability for better economic outcomes, exploring various investment strategies and assets. By linking return predictability directly to economic performance, we challenge the focus on statistical significance over economic relevance. Our findings suggest that even slight predictability, when strategically applied, can yield substantial economic benefits.