This paper explores the medium-run behaviour of bounded rational players in repeatedly played games when they occasionally experiment or make mistakes. The formal analysis introduces a hierarchical structure of limit sets to characterize the most possible medium-run behaviour over gradually...