We examine the economic gains produced by equity (and bond) conditional strategies when the out-of-sample predictive power, R2, is close to zero. To address this issue, we design a model which expresses the performance of the conditional strategy as a function of any given level of R2. The relation between R2 and performance is studied in different settings, including one and multiple assets, as well as short-selling constraints. In all cases, a level of R2 as small as 0.5% is sufficient to greatly outperform the unconditional (passive) strategy. For a given R2, the profits are much higher if the source of predictability comes from asset-selectivity rather than factor-timing. In addition, we find that under asset-selectivity, a single-asset strategy requires a very high R2 in order to perform as well as a multi-asset strategy.