And yet I've seen no algorithmic approach for optimizing speedruns for this kind of game. Most have a very large state space, so MILP solvers are not applicable.