Games have become a very successful setting for modelling logics and programming languages. If we view a game as a tree then the full history of play is encoded in every position. If we view them it as a graph then different plays may lead to the same position, and once we have reached a position we no longer know how we got there. Strategies then compose via relational composition, and that makes reasoning about strategies and proving equations much easier. In this talk we present the main ideas of this approach.