Crowd suggestion: Application to simulation & video games for ice hockey could be interesting.

Yes, we have systems of events that involve major actions on the ice & also actions of action perception, planning, & categorization by various players that affect the actions which they initiate. One can image, for example, a coach focused game that allowed special slowing isolation of play to observe the state of various players & abilities of the coach to try & improve their conception of how the player should be reacting.

A lot of work has been done on statistical analysis of causal models which are sensitive to temporal sequencing along with correlation. One sort of abstract approach could be based on Bayesian distributions over causal graph relationships between the variables & then conditional timing sequences based on the causal graph. That lots of computing, but it potentially offers a lot of flexibility in modeling with good regularization. If it works well, computational short cuts will be found.