If you select one value for x, execute doIt for that value, and get a correct answer, how confident are you that doIt works?  What if you run 50 values through and get correct answers? Feel better?  The more completely a domain is covered by test cases, the more likely the result is to be correct.