Most final appeal courts around the world do not sit *en banc*, but rather sit in smaller panels. In some courts, those panels are decided randomly, and where random panel assignment exists, it can help identify the effects of particular (types of) judges on particular types of outcomes. Where panel assignment is not random, it must instead be modelled. However, there are few existing models for this "pick-m-from-n" choice situation. In this paper, I develop such a model, and apply it to panel selection on the UK Supreme Court between 2009 and 2017. I test for the importance of specialisation in the relevant area of law, the importance of workload, and the importance of past agreement with the President and Deputy President of the court (who decide on the panel assignments). In this way I make a methodological and substantive contribution to the study of judicial decision-making.
