This paper proposes a new formulation of the Maximum Diversification indexation strategy based on Rao’s Quadratic Entropy (RQE). It clarifies the investment problem underlying the Most Diversified Portfolio (MDP) formed with this strategy, identifies the source of the MDP’s out-of-sample performance, and suggests dimensions along which this performance can be improved. We show that these potential improvements are quantitatively important and are robust to portfolio turnover, portfolio risk, estimation window, and covariance matrix estimation.