Summary: Some concepts are introduced with a view to dealing with the problem of choosing the best potential alternative(s) in a multi-attribute ordinal data context. We assume that we know the possibilistic weights of attributes and ordinal evaluations of each potential alternative on all attributes. We define what we refer to as positive/negative coalitions of attributes and positiveness/negativeness numerical bags associated with each potential alternative. We also define what will be called positional advantage operators and we describe how all these concepts are used to lay the foundation for the choice method appearing in this paper.