The best test for any given size $\alpha$ is the test with the highest [[statistical power]], or highest probability of rejecting the null hypothesis when the alternative hypothesis is true. The best test is also referred to as the "most powerful test" for this reason. Recall that any properly constructed test of size $\alpha$ will have the same probability of committing a [[Type I Error]], incorrectly rejecting the null hypothesis when it is in fact true, by definition. Thus, the best test will be the one with the smallest probability of committing a [[Type II Error]].