The relationships between complexity and stability are one of the central themes in ecology. As a model real food web, soil food web had been contributed a lot to the complexity-stability relationships. This paper reviewed the methods, results, deficiencies and perspectives of exploring ecosystem complexity-stability relationships based on soil food web that developed by Moore, de Ruiter and Neutel and other theoretical ecologists. The complexity of soil food web including interaction strengths and patterns, compartments, productivity and energetic organizations was described by incorporating the data of functional groups biomass with the Lotka-Volterra-based and process-oriented soil food web models, and the local stability was analyzed through linking the Lotka-Volterra-based soil food web mode with community matrix theory. After that, the general complexity-stability relationships in ecosystem can be discussed. Moore, de Ruiter and Neutel proved that compartments, the pattern of interaction strengths in real food web enhanced the ecosystem stability compared with random food web, and both the productivity and dynamic stability limited the food chain length. However, the complexity-stability relationships based on soil food web must be explored further because the description of complexity and stability was based on the equilibrium assumption and neglected the temporal and spatial variability in soil food web structure. Therefore, dynamic soil food web mode would be the direction of exploring ecosystem complexity-stability relationships by means of soil food web.