I’m interested in creating a formal model for how to incorporate player behavior into an economic or systems model.
For example, when creating an initial model there are several types of variables- predictors of behavior (20% of players will wan this), aesthetic settings (score should be in 1000s not 10s), balance knobs (enemies should do 50 damage). Then you create your model, and it has several outputs- “Gold earned per second”, “Damage dealt per player” “Average HP for a level 20 enemy”.
When we’ve balanced systems, we tend to pick inputs and outputs we want to set for aesthetic or pacing reasons, and then we want to solve for the remaining inputs and outputs through monte carlo (or other method) of simulation. If our collected metrics fail to match the model, we figure out whether a behavior predictor was wrong or if our model was wrong and adjust accordingly.
We’ve never done this formally (although I am sure several horseshoers have), but it would be really useful to have a good way of setting up this iteration that could be applied to many game design problems.