One major challenge facing policy-makers is to design education and workplace training programs that are appropriately challenging. We review previous research that suggests that difficult training is better than easy training. However, surveys we conducted of students and of expert sport coaches showed that many prescribed easy rather than difficult training for those they coached. We analyzed the performance of National Collegiate Athletic Association (NCAA) basketball teams in postseason tournaments to see whether the existing research, largely on individuals in short-term situations, would generalize to teams in the long run. Indeed, playing difficult nonconference (training) games modestly improved performance for NCAA teams in the postseason. Difficult training particularly benefitted teams that lost many nonconference games, and the effect of difficulty was positive within the range of difficulty NCAA teams actually encounter, making it clear that difficult training is superior. We suggest that our results can be generalized beyond sports, although with careful consideration of differences between NCAA basketball teams and other teams that may limit generalizability. These results suggest that policy-makers might consider amplifying the difficulty of team training exercises under certain conditions.