This article considers how statistical reasoning changes conceptions of evidence and proof. Beginning with three Court of Appeal judgments in which proof is quantified, it traces the implications of statistical ways of thinking about proof through the law of criminal evidence. This leads to the bizarre conclusion that proof is, by and large, impossible. The argument then takes a more constructive turn. The way in which the presumption of innocence is conceptualised in statistical argument is criticised and it is suggested that proof depends on a precondition of trust in the way suspects are selected by the police. For that trust to be deserved, police suspects must be chosen in a legitimate manner.