Abstract: In order to improve the intelligent of computer games (CGs) system, the efficiency of traditional methods in optimizing training of static evaluation parametric combination and the quality of training, a distributed genetic operator was proposed for clan crossover. The adaptive genetic algorithm was improved in the course of population enlarging and chromosome duplication. Analysis results show that, the ability has been improved with the improved genetic operators, to inherit superior character from parent generation and to explain gene expression. Better training results have been gained in actual optimization of draughts. Besides, simulation experiment has validated the stability and accountability of the algorithm in solving general problems.