Selecting managers on the basis of past performance is an intuitive strategy that seems trusted by investors. Indeed, many studies report a positive correlation between past fund returns and investor cash flows. Evidence also suggests that at least at shorter-term horizons investing with the top performing managers may generate future out-performance. This was initially dubbed the “smart money” effect, the idea that investors are capable of identifying manager skill. More recent studies suggest that manager exposure to momentum stocks, on one hand, and the persist-ency of fund flows, on the other hand, are the more likely explanations of the fact that past and future fund alphas seem to be positively correlated at shorter-term horizons in the pooled cross-section of funds. However, at longer-term horizons, e.g. over the standard investment evaluation horizon which is roughly three years for the average institutional investors, there is evidence of a negative correlation between past and future fund alpha. Managers with stronger performances seem more likely to under-perform subsequently over longer-term horizons. If there is mean reversion over the horizon of interest, then the modern hiring/firing practice should lead to worse outcomes than the apparently paradoxical strategy of investing in managers with poor performances and firing the successful ones. While such a contrarian approach to manager selection might seem counter intuitive and unreasonable, there is at least, a justification to it under this “dumb money” effect. A more practical approach consists on using other factors than performance for manager selection. Literature supports the use of factors such as the theoretical soundness of the “investment thesis” driving the fund’s strategy, the link between manager compensation and fund performance or the fund active share.