Ask YC: Creating Models | Hacker News Reader

Ask YC: Creating Models

2 pointsssharp2 comments

I'm starting to plan out a small online game and need to get some advice on building mathematical models. These models are definitely not scientific and don't really need to be exact, just something that is close and works as expected. I don't even think they need to be particularly large. For most of what I'm doing, there will only be 2-5 inputs.

Are there any good online guides to building formulas and how various functions can control the output? I don't have any great background in math but I'd expect that there has to be some sort of examples online that I could dissect and rearrange.

View on HN

If you don't mind tracking down books, I would recommend a few that do a good job introducing modeling.

B.S. Bennett, "Simulation Fundamentals" covers basic simulation from a programmer's perspective. It covers both discrete and continuous systems. It has an "older" feel to it than many resources (for instance, it compares analog vs digital computer implementations), but for me this basic approach made it a useful introduction.

Andrew Ford, "Modeling the Environment: An introduction to system dynamics modeling of environmental systems" is nice because it looks at real (albeit simplified) environmental systems, which gives it a useful grounding. Focus is continuous systems (i.e., characterized by ordinary differential equations).

Arthur Few, "System Behavior and System Modeling" treats simpler systems than Ford's book. I recall this being a very direct, easy-to-read book (I was researching books to teach systems dynamics to high school students). Unfortunately, this one may be hard to find now.

I'm not really clear what you are trying to do with these models, but agent based modeling may be related : http://en.wikipedia.org/wiki/Agent-based_model This may be a useless comment, but I would also recommend you focus on the software engineering aspect - provide a clean, simple interface that describes what you want and do a naive implementation that will allow you to work out the rest of your system, then you can keep working while studying the ideal mathematical model for behavior. I think the general search term for this field is "numerical analysis". One quick and dirty technique that is sometimes useful if you have a set of input vectors and a set of goal states is to do polynomial interpolation with something like a lagrange polynomial.

Reply on news.ycombinator.com