How can crime be disrupted effectively without increasing resources? To answer this question, we develop a spatial network model of crime diffusion, using London as a case study. Moving beyond traditional hot spot policing, we identify key player neighborhoods-highly connected areas in the network. Counterfactual analysis shows that targeting top 10% of key players reduces crime by 10.7% (5.8 percentage points) more than targeting top 10% of hot spots, resulting in potential annual savings exceeding 130 million pounds. Examining the underlying mechanisms, we find that while hot spots attract crime locally, key players facilitate its propagation across areas.