The paper deals with the analysis and modification of the expectation submodel of a game theory-based vehicle-pedestrian interaction model. The model modification concerns inclusion of additional explanatory variables, particularly the pedestrian position, in order to extend the model applicability and improve the overall model accuracy. The expectation model parameters are identified by using a real-world drone-recorded dataset, which includes several hours of vehicle-pedestrian interactions at an unsignalized crosswalk. The identified expectation model and the overall pedestrian model are validated against the experimental data.