Mechanical trading rules seem to have more potential than previous tests found. Fama and Blume (1966), looking at the Dow 30 of the late 1950s, found no profits for the best (1/2-percent) rule after adjusting for transactions costs. Fifteen of these stocks looked profitable in their sample, however; for the same rule, the surviving fourteen show statistically significant profits for 1970-1982 for transactions costs obtainable by floor traders. The test used here assumes constant risk premia, or more generally, that risk premia are on average approximately the same on days "in" as for the total period.