I develop a model of collaboration between tournament participants in which agents collaborate in pairs, and an endogenous structure of collaboration is represented by a weighted network. The agents are forward-looking and capable of coordination; they value collaboration with others and higher tournament rankings. I use von Neumann-Morgenstern stable sets as a solution. I find stable networks in which agents collaborate only within exclusive groups. Both an absence of intergroup collaboration and excessive intragroup collaboration lead to inefficiency. I provide a necessary and sufficient condition for the stability of efficient outcomes in winner-takes-all tournaments. I show that the use of transfers does not repair efficiency.