Load balancing is crucial to computational grids. Based on the typeof information on which corresponding decisions are made, load balancing approaches can be generally categorized into two classes: average-based and instantaneous approaches. Both these two classeshave merits and demerits. It's therefore advantageous to combineboth of them to work in tandem to form a hybrid approach that makesthe best of the strong points of each. In this paper, we present ahybrid load balancing strategy for sequential tasks in computationalgrids with the purpose to improve the performance by a joint effort.The simulation results show that our algorithm can achieve a betterload balancing performance than its counterparts provided insimulation.