This dissertation consists of three related chapters in the field of empirical asset pricing. Broadly speaking the chapters investigate issues related to active portfolio management, stock-picking, portfolio optimization, and asset pricing model performance. Chapter 1 introduces a systematic portfolio choice solution that advances contemporary models of return predictability by implementing multivariable cross-sectional regressions of key stock characteristics. These models generate tradeable portfolios which significantly outperform common benchmarks. Chapter 2 is the first study to conduct a comprehensive portfolio analysis using individual stock data. Results show that naive diversification consistently outperforms active timing strategies and parametric portfolio choice solutions. These results add to the mounting evidence that practical implementation of portfolio theory often performs poorly out-of-sample. Chapter 3 investigates whether signals from conditional asset pricing models can be used to construct tradeable portfolios and also revisits the characteristics vs. factors debate using a larger set of factors. I conclude that factors do provide reasonable proxies for the observable characteristics. To extend my current body of work, I intend on investigating the appropriate functional form of return predictability regressions and using these results for portfolio construction. Extensions of these studies offer several avenues for future contributions to the field of empirical asset pricing.