3 ms·
For all developers creating city-builder games, it seems there's an inherent dilemma between designing your game at a macroeconomic level and a microeconomic le
by lasagnaphil 6y ago
For all developers creating city-builder games, it seems there's an inherent dilemma between designing your game at a macroeconomic level and a microeconomic level.
Games with macroeconomic simulations (SimCity 4 for example) use sampling-based methods to get statistics about the city and then plug in some simplified equations to simulate things like demand and growth. Your simulations generally behave in the way you want to, with all your assumptions and beliefs about how economics should run. The problem with this is that it fails to capture the nuance of local economic effects, which leads to your game becoming too simplistic (exemplified by the Magnasanti project which is referenced in the article)
Games with microeconomic simulations (ex. Cities Skylines) treats every citizen as an independent agent, with their own jobs, families, and movement for work/shopping/travel. The problem with this is that you have less control of how the game will actually play like, and that it's incredibly hard to balance the game (Cities Skylines has a reputation for their systems being "too easy" except for transportation, but I imagine that they made it that way because any further fundamental economic tweaks that make the game harder might actually break the whole game). There's also the technical issue of simulating too many agents.
NewCity, one of the more recent city-builder games on early access, is trying to achieve some balance between the two, and it's interesting to read their design decisions (https://lonepine.io/2021/01/15/macroeconomics.html https://lonepine.io/2021/01/15/macroeconomics.html). It seems like a really hard gamedev design problem though.
- BlueTemplar 6y agoWait, so SimCity 4 actually is NOT microeconomic, or did Maxis lie about it ?
- hypertele-Xii 6y agoSimCity 4 doesn't simulate individual agents. You can blatantly see vehicles phasing in and out of existence to give the look of a bustling city, without actually simulating one. SimCity (2013) is the first SimCity game to simulate individual agents. And it takes some pretty ginormous liberties there, too. Citizens, for example, return to any random available house after their workday. That is, they switch homes every day. It's less human city and more... anthill.
- bobthepanda 6y agoSC:2013 had a lot worse problems as well with how it chose to do agents. - In C:S agents are mostly used to allocate work/home slots, and to depict traffic. SC used agents for everything, including simulating how individual poops would reach a sewer outlet, or how power or water reach homes. So they used computationally expensive algorithms to simulate things that no one cared about actually seeing like that. - Because of the very heavy computation, to make the load workable they limited city size to 2km by 2 km, or 4 sq km. To put this in perspective, the land area of Manhattan is 59 sq km. SC:2013's scale would be as if the city you built would only get to be built as big as Central Park (3.5 sq km). This is fine if you want to simulate a medieval town, maybe, but probably not a modern city with trains and highways and stadiums and things that take up a lot of space.
- fakedang 6y agoAnd I think the Cities Skylines model is the way to go here, although consuming computationally. I read somewhere that they were being used for city planning exercises at MIT and in Sweden.
- bobthepanda 6y agoThey're of limited utility though, because any models, particularly for a computer game meant for consumer hardware and designed for "fun", are going to be very simplified.
- fakedang 6y agoIn city planning, some utility is better than no utility at all, which is how most cities were designed.