This study aims to derive the epistemic bias in shi-bu-shi questions, a type of A-not-A question in Mandarin Chinese. I propose: (i) the focus marker shi presupposes that its prejacent is a possible complete answer to the current Question Under Discussion (QUD); (ii) accordingly, shi-bu-shi questions are presupposed to be part of the Focus-strategy of inquiry; (iii) the Focus-strategy of inquiry indicates the questioner's intention to close the current QUD as soon as possible, and to achieve this goal, the questioner should check the answer that she considers most likely to be true. By assuming such completeness-to-likelihood reasoning, a novel link between focus in polar questions and question bias is established. The ramifications of this proposal for related phenomena (e.g., bias in embedded questions, evidential bias) are then discussed.