I get that, I'm just not sure there's a good reason for it. Extending discrete models to their probabilistic versions is often very useful. Probabilistic knowledge representation is much less brittle than a logic-based system. Stochastic grammars and probabilistic automata have been essential to both language modeling and bioinformatics. Game graphs are nice for perfectly observable systems, but stochastic games are much more common in real life (and have to be modeled using some probabilistic mechanism).

This page has some good fundamentals, it's just weird to me that probabilistic extensions weren't also presented.