The distribution of long-run compound returns to portfolio strategies is greatly affected by periodic rebalancing. Over time, buy-and-hold portfolios gradually lose diversification as extreme long-run skewness in individual stock returns leads to increasingly concentrated holdings. For long investment horizons, small rebalanced portfolios holding only a fraction of all stocks therefore achieve better diversification than much larger marketwide buy-and-hold portfolios. Consequently, over long horizons, rebalanced portfolios tend to outperform buy-and-hold portfolios, and risk-averse investors prefer the former. Empirical results strongly support the theoretical predictions and add further evidence to the strong empirical performance of (small) equal-weighted portfolios.