Being scientific research a process of social interaction, this process can be studied from a game-theoretic perspective. Some conceptual and formal instruments that can help to understand scientific research as a game are introduced, and it is argued that game theoretic epistemology provides a middle ground for ‘rationalist’ and ‘constructivist’ theories of scientific knowledge. In the first part (‘The game theoretic logic of scientific discovery’), a description of the essential elements of game of science is made, using an inferentialist conception of rationality. In the second part (‘Sociology of science and its rational reconstructions’), some ideas for the reconstruction of case studies are introduced, and applied to one example: Latour's analysis of Joliot's attempt to build an atomic bomb. Lastly, in the third part (‘Fact making games’), a formal analysis of the constitution of scientific consensus is offered.