This paper parsimoniously characterizes how past returns affect the cross-section of expected returns. Using Fama-MacBeth regressions, it shows that the momentum and reversals associated with past returns over various horizons are strongly affected by a turn-of-the-year seasonal that differs for winter and losers, depending on both the tax environment and the month of the year, and differs by exchange listing. The analysis also uncovers a consistent winners effect – high fractions of positive return months tend to increase expected returns. Out-of-sample evidence suggests that the documented relation between past returns and expected returns cannot entirely be due to data snooping biases.