Mutation is one of the important operators in genetic algorithm. In traditional genetic algorithm, mutation is activated stochastically. In this way it is unknown and cannot be controlled for which individuals to be mutated. Therefore, it is unavoidable that some good individuals are destroyed by mutation and then the evolutionary efficiency of the genetic algorithm is dampened. Owing to this kind of destructivity of mutation, the operator of mutation has to be limited within a very small probability, and the potentiality of mutation is consequently limited. In this paper, we present an evolutionary chain-based mutation and a control strategy of reasonable competition, in which the heuristic information provided by the evaluation function is well utilized. This method avoids the blindness of stochastic mutation. The performance improved in this method is shown by two examples, a fuzzy modeling for the identification of a nonlinear function and a typical combinatorial optimization problem-the traveling salesman problem.