In this paper, some characteristics of GA(genetic algorithm) deceptive problems are discussed, the performance of the mutation operators is analyzed theoretically, and a new method for solving GA deceptive problems is proposed. The proposed method can change the direction and the possibility of the mutation operators in the process of the genetic searching, making mutation operators adjustable. In this way, GAs can eliminate effectively their deceptive conditions, keep the diversity of population, and thereby converge conveniently on global solutions.