1.) Mapping O and X states
Again, there may too many states to cover, so we will do is get the difference of height percentage of the bird and the two pipes:
2.) Adding states scenarios as inputs (even if its incomplete that's ok, as long as we achieve a goal)
Syntax: state action reward next-state
Results: (Note: Press compute policy button multiple times until it gets the optimal policy)
Per Action:
Now let's try to beat this AI, press the compute policy couple of times first to train the AI.
Click on the panels to start
Conclusion: Now we know that our algorithm can create a tough AI that's impossible to beat, although training for a simple game requires a lot of states.
Applying a shortcut algorithm using approximation for tictactoe