Pragmatic encroachment theories have a problem with evidence. On the one hand, the arguments that knowledge is interest-relative look like they will generalise to show that evidence too is interest-relative. On the other hand, our best story of how interests affect knowledge presupposes an interest-invariant notion of evidence. This paper sketches a theory of evidence that is interest-relative, but which allows that ‘best story’ to go through with minimal changes. The evidence someone has is just what evidence a radical interpreter says they have. And a radical interpreter is playing a kind of game with the person they are interpreting. The cases that pose problems for pragmatic encroachment theorists involve coordination games between the interpreter and the interpretee, and like most coordination games they have multiple equilibria. The best solution to these games involves the notion of risk-dominant equilibria. So the evidence the person has is the evidence the interpreter says they have in the risk-dominant equilibria. This will be interest-relative, but consistent with the ‘best story' about how interests usually affect knowledge.