This paper introduces an improved Tumbleweed Algorithm (FDB-TA) based on a fitness-distance balance guidance mechanism to address the tendency of the traditional Tumbleweed Algorithm to get trapped in a local optimum. By integrating the fitness-distance balance mechanism, the algorithm enhances the equilibrium between global and local search capabilities, demonstrating superior performance in tackling complex optimization problems. Extensive experiments conducted on the CEC2013 benchmark set validate the efficacy of the improved algorithm. Results indicate that FDB-TA achieves outstanding outcomes across various test functions, significantly outperforming the original algorithm.