Liquefaciton has caused serious damages to buildings, roads and humanlife during earthquakes. Gravelly soils have often been thought to resist liquefaction because of their large particles and high peamiability. However, many cases of gravelly soil liquefaction have been observed in past earthquakes. Previous studies have examined how gravel content(GC) affects liquefaction strength and strain behavior. They found that higher GC generally increases liquefaction strength, but the full understanding of this behavior is still unclear. The basic mechanisms behind the liquefaction of gravelly soils have not been completely explained. In this study, numeriacl analyses using the Discrete Element Method (DEM) were carried out to explore the fundamental mechanisms of gravelly soil liquefacton. The DEM results showed that liquefaction resistance changes with GC and follows similar trends to those found in earlier reserch. DEM represents each particle individually, but it faces difficulties when the number of particles becomes very large. This makes it hard to understand the role of individual particles and the structures they form. To overcome these challenges, network theory was applied to the DEM results. Using contact information between particles, a contact network was created where particles are nodes and contacts are edges. Betweenness centrality, an important measure in network theory, was used to identify particles that act as hubs for stress transfer. The results showed stress is more likely to concentrate on nodes and edges with higher centrality, which was also found in experiments in previous research[1]. Furthermore, it was found that when gravel particles, which exhibit extremely high centrality, are present, stress becomes highly concentrated on those particles. These finding indicates that the way stress is transmitted can be predicted using betweenness centrality and gravel plays an important role in soil.