The strategy that delivers the best returns, it turns out, is to divide the funding equally among all researchers. And the second- and third-best strategies involve distributing it at random to 10 or 20 percent of scientists.