City Building Game
I like the idea of making a game based on your city planning prototype, however I don’t think your game will be a great vehicle for your city-planning ideas.
Modelling The Interesting Stuff
In order to make your city-planning ideas work in a game you would either have to model incentives based on individual in-game agents, and thus give agents complex AI for long-term decisions like when to buy or rent or move or renovate or change jobs, how much money to save or to invest with some risk or to spend, how many children to have and so on. You would have to model trust and civic engagement and social cohesion. Or all these factors would just be variables in an abstract cellular automaton based on a system of differential equations, like the original SimCity. In that case, you would have to make simplifications and judgements that look like begging the question.
The middle path would be putting agents into a grid-based world in which they make some decisions individually, but are influenced by grid-based environmental factors. On every grid update, grid cells are first updated based on the aggregate of agents living in the cell, then grid cells update based on surrounding cells. On every decision, an agent consults the values in the current cell it is in, or a weighted combination of the cells it was in most often over a period of time.
You probably want to model trust, safety, length of commute, crime levels, civic engagement, savings, disposable income, taxes, rent and rent controls, property developers, landlords, homeowners, family…
Communicating With The User
You need a way to make the player see what the agents are thinking, when they are making important decisions, and why.
It is important for two reasons:
- If things happen but you can’t see them, the game feels boring
- If important things happen and you don’t know, the consequences feel unfair
I like the idea of making a game based on your city planning prototype, however I don’t think your game will be a great vehicle for your city-planning ideas.
I agree, but in this case I can use some of those ideas as starting places to give more depth to the simulation, so there can be some simulation of those ideas that isn’t feasible in existing city builders, without the kind of in-depth total simulation we might do if this were a university research project.
Thus, the OTV Game can be differentiated by support for mixed-used buildings where the bottom is commercial and the upper portion is residential, zoning regulations with more potential control, rent bidding, etc.
The middle path would be putting agents into a grid-based world in which they make some decisions individually, but are influenced by grid-based environmental factors. On every grid update, grid cells are first updated based on the aggregate of agents living in the cell, then grid cells update based on surrounding cells. On every decision, an agent consults the values in the current cell it is in, or a weighted combination of the cells it was in most often over a period of time.
This is essentially my plan, along with a goal of 1,000,000 agents and 64km2 of area. Decision trees can be manageable for each agent if they are very small, and various heuristics will be used to make the simulation feasible, including use of grids and hierarchical routing.
Initial simulation will be simpler and focus on the core economic elements, and more complexity will be added over time. For instance, once basic markets are implemented and tested, more industries and specializations can be added and simulation load and difficulty observed.
You probably want to model trust, safety, length of commute, crime levels, civic engagement, savings, disposable income, taxes, rent and rent controls, property developers, landlords, homeowners, family…
Yes, some of that is definitely on the initial slab of what I want to develop, and how to rig up property developers will be one of the interesting questions, since I plan to track firms’ profits and accumulated capital.
However, I think sufficiently complex behavior can be obtained with fairly simple rules - for instance, that firms have a base cost and marginal cost, and scale up production when they make a profit and scale down production when they fail to make a profit, and that when they accumulate enough saved up capital and are profitable, they move to a bigger building to expand.
This pent-up capital accumulated for bigger buildings could then be part of the heuristic used by property developers. (Which probably would skip being physically represented as owning offices in the city, unlike other businesses.)
You need a way to make the player see what the agents are thinking, when they are making important decisions, and why.
It is important for two reasons:
1. If things happen but you can’t see them, the game feels boring
2. If important things happen and you don’t know, the consequences feel unfair
Yes. This will require combing the grid for issues and representing them as visual cues for the users. The form that takes will have to depend on magnitude and kind.
The stylized aesthetic provides plenty of room to provide visual cues in addition to cues such as floating event bubbles above businesses going bankrupt. It also provides room for some interesting overlays.
Ideally, we could also access individual citizens at their home or workplace and get more detailed information about them, but in practical terms this isn’t efficient for a city of 200,000, so there must be other ways to display this data.
Preparations are now in motion. I will be evaluating the difficulty of the development path I want to pursue, level of interest, and so on.






