In this chapter, the authors introduce an approach to address the ordering issue. It is based on the idea of averaging the value of a variable’s attribution over all possible orderings. The authors show that Shapley values could be presented as a unification of a collection of different commonly used techniques for model explanations. Players are not identical, and different players may have different importance. Cooperation is beneficial, because it may bring more benefit than individual actions. The picture for variables fare and class is more complex, as their contributions can even change the sign, depending on the ordering. Shapley values provide a uniform approach to decompose a model’s predictions into contributions that can be attributed additively to different explanatory variables. An important drawback of Shapley values is that they provide additive contributions (attributions) of explanatory variables. An important practical limitation of the general model-agnostic method is that, for large models, the calculation of Shapley values is time-consuming.