Disaster model

[Cynary]

We need a way to model disasters.

[Cynary]

Implementation idea:
There will be a weather class, instead of a specific disaster class.
The weather class will be a simple weather simulator, based on a Markov Model of the weather.
The model is as follows:
Observations/output are the amount of rainfall for the day.
There is a state that represents a normal day: the amount of rainfall is chosen from some distribution (e.g. exponential if we want our map to be a sunny map, gaussian with some >0 average if we want to live in a rainy place).
There are infinite states that represent a disaster occurring: each state represents a day in the disaster. As the number increases, the higher the probability of going back to normal day. The amount of rain depends on which day we're in, but I'm thinking of having it be a gaussian distribution whose average increases according to a logistic function (increases quickly at first, but then it stabilizes at some threshold).

[Maplenormandy]

I can cover the research and development of a disaster model. Sorry about the notification spam lol.

So I think instead of breaking down into normal day vs. disaster states, it might make more sense to break it down into no storm, storm developing, and storm hitting states:

• During the no storm state, there's little or no rain, i.e. no rain that affects the game state.
• During the storm developing state, the game starts tracking the possible intensity of the storm (either disaster or no disaster) and time to hit.
• During the storm hitting state, the game starts sending rain that affects the game state, i.e. through flooding or other in-game effects

This is similar to the combined prototype model we made in class, except for the model we made in class, we assumed every storm would result in some disaster.

This focuses player attention to make decisions in response to specific events (i.e. do you prepare for this specific storm, rather than just having the player do things day-by-day), and makes it so that you can prepare for a storm but have wastage due to the storm not being a disaster, which is something that I think we haven't captured yet but is important to FbF since it's one of the stumbling blocks for acceptance of FbF.

[Maplenormandy]

Additionally, we could add seasonal variation, i.e. there is an additional underlying variable that determines whether or not a season would be very wet or not. This adds an additional layer of complexity, but does add another possible interesting decision in that you know that the season will be wetter on average so you prepare more, but you don't know which storm exactly will be the one that's a disaster.