The polling website fivethirtyeight.com has an excellent article about why it's so difficult to make a good COVID-19 model. If you are interested in some of the challenges you face when you attempt to make anything other than a fairly trivial model, like the ones above, this article explains some of them.

Sorry it's taken me so long to reply.

Thanks for your comments. 

WRT models that produce unexpected results: one of the most important features of complex systems is that what are often called higher-level phenomena "emerge" from lower-level phenomena. Two examples we've seen are (a) gliders and other patterns in the Game of Life and (b) flocks in the flocking model. In both cases, the underlying model says nothing about generating these phenomena; yet they appear. 

That was one of the features of complex systems that generated the most initial excitement. In fact, systems that produce higher-level phenomena based on lower-level activities that aren't explicitly structured to create those phenomena are all around us. The best and still most puzzling example is life: how can non-living chemicals produce living organisms? We don't have a good answer to that. One reason we don't know how to answer that question is that we don't have a good way to define what it means for something to be considered alive. (Viruses are a great example in that they fall on the boundary between living and non-living.)

Another example that you're all familiar with is software. Software consists of statements in a programming language. (1) The statements themselves don't do anything. At a minimum (and at a greatly simplified view) software needs a computer to execute it. But (2) a computer does nothing but very simple operations. Think of what you learned about assembly language and machine language. Each operation is very trivial: load, store, add, etc. But put together you get a program that does often amazing and surprising things. They are particularly amazing and surprising when you see the results without knowing how the software works. But even when you know how some software works, e.g., the greatest-common-divisor algorithm, the individual steps are not independently connected to the final result. It's only all the steps together when executed by a computer that you get the final result. (A good example of how disconnected software can be from the results it produces is how difficult it often is to predict what software will do if you haven't seen it before. This week's assignment -- explaining how the NetLogo model works -- illustrates this!)

Think of any of the models we've built. Run one to someone who is not knowledgeable about software and ask that person what they think they will find if they look inside the computer while it is running. For example, in the NetLogo Baess paradox model, a software-naive person might imagine that if they looked inside the computer while it was running they would see something like individual agents (cars and trucks) being moved around by some sort of director, e.g., cars and trucks being manipulated by a choreographer. (When you look at the individual frames in an animated movie, you do see something like that.) Yet when you actually look inside the computer while it's running the NetLogo Baess paradox model you don't see anything like cars and trucks. All you see is lines of code. So, how did those lines of code become cars and trucks on a screen? (A virus is something like software. It consists of an RNA, or sometimes DNA, sequence that by itself does nothing. It requires the processor of an external entity, i.e., our cells, to make it come alive.) That's the excitement of complex systems.