The complexity and state space of the Poker is not large. Therefore it is possible to just exhaustively search all the combinations. In fact you can even calculate the probability of getting each cards with some arithmetic. I would recommend you to read Poker Theory and Analytics by Kevin Desmond on MIT Open Courseware to understand more.
Your idea is exactly what decision trees or random forests are. They are machine learning techniques that can build human-understandable rules with a tree of if-s.
On the other hand, there are also non-human-understandable machine learning. Neural network is a kind of that. The models are non-intuitive to human. They are mainly used for the problems that are hard to formulate or with extremely large state space, for example, playing Go (just like AlphaGo), identifying all the objects in an image, or how to win Starcraft.
solved Teaching the computer to play texas Hold’em poker [closed]