This is because the process of training the algorithms has historically been different for the two types ... design a technique that can work for both perfect and imperfect information games ...