We propose the following genetic algorithms (GA) to effectively solve job-shop scheduling problems by introducing the following judgments of the present status of the genes in the population and characteristics of the task sequence: (1) Extracting excellent genes, i.e., the effective part in the string from individuals with good fitness; (2) Exploiting local search around an elite by checking all probable mutation within one Hamming distance; (3) Introducing an overrule to examine only the cases which seem to give good results. These algorithms are tested by applying them to the Muth-Thompson Job-Shop Scheduling Problems (1963), and we prove that the application of GA with combined use of these algorithms can generate practically good schedules even for job-shop scheduling problems of relatively large scale.