Modeling Flow state in an Action-Roguelike
Using machine learning to read player challenge from gameplay.
For this research project, we tested whether a small action roguelike could estimate player flow from gameplay data. We built a Vampire Survivors-inspired prototype, logged player behavior, and asked participants to report how challenging the game felt.
We trained a model on those signals to explore dynamic difficulty adjustment. The result gave us a practical starting point for linking challenge, emotional response, and player modeling inside a live game loop.
