This thesis investigates the role of military strategy in the geospatial variation of conflict severity in civil wars. Arguing on the basis of rationality of the conflict parties, the strategic behavior is determined by a cost-benefit assessment of each area according to their attractiveness to fight. Thus, the differences in characteristics of these areas can be utilized as predictors for where civil war is most severe. The effects of these structural variables change over time and are accompanied by the endogenous spatial dependency of conflict severity. To test this theory a standard OLS multiple regression and a spatial autoregressive model are applied to the case of the Taliban insurgency in Afghanistan. The analysis showed supporting evidence for the theory. Spatial dependency is constantly significant in explaining conflict severity but its importance is fading over time. Also, local population support makes areas more attractive to insurgents. This variable also becomes less important in the course of the conflict. The findings indicate that fighting moves from rural to urban areas. Thus, the analysis of the conflict confirms the developed theory. Further research could strengthen the confidence in the findings and expand the theory.