Extant scholarship has generally ignored relational interdependencies when attempting to understand why and whom non-state armed groups fight during civil war. This paper investigates the interconnected web of alliances and rivalries between armed groups in the ongoing Yemen Civil War to address the following question: why do armed groups fight each other? This paper adopts an inferential network approach to investigate determinants of intergroup conflict. In particular, it emphasises the role of group identity, arguing that operating in salient cleavages necessitates that groups align with or distinguish themselves from each other. We hypothesize that shared identity increases the likelihood of conflict in dyads, while moderate similarities do not exert this degree of influence. Additionally, we argue that informal alliances between groups counterintuitively incentivises violence the longer the war continues. These arguments are tested using exponential random graph models to account for endogenous structures arising from repeated interactions over time and explore the causal pathway between identity, alliances and fratricide. Our findings indicate that shared identity is a significant driver of hostilities, moderates the influence of cooperation and amplifies the effects of group attributes. We find no support for the traditional proposition that power asymmetry significantly affects group rivalry. Using a variety of robustness measures and simulation, it is also shown that network models more accurately capture the underlying data-generating process to predict fighting between groups in this case.