This is very much a filler course. There's not a lot of work, and what work there is isn't challenging. But you can pick up credits towards graduation easily.
The course structure is almost entirely built around the final group project (80%). You will have several bites at the apple so to speak. There are something like 4 "interim" presentations, which should generally be early iterations of your final presentation (on which you are graded). These are worth 20% in total. The final presentation is worth 40% and the group work feedback is worth 20%. The last 20% of the course is class discussion stuff. This will involve readings, one of which is about 100+ pages (the rest are shorter). In the 1st half of the course be prepared to read about 30 pages a week and write 500 word commentaries on what you read (2nd half of course these go away). Presentations are literally 10-15 minutes of recorded talking over a powerpoint.
Don't be freaked there is a "soft" 95% cap on all assignments, the last 5% is if you've truly gone above and beyond in some crazy way (and in many cases, I think the reading/commentary grades are just flat 95% if you submitted something of roughly the right length, and a lower grade otherwise. I don't know how 2 TAs are supposed to read 500 words per student per week).
For most OMSCS students, be aware that this is a policy track class for cybersec folks. Therefore, you might be expected to shoulder the bulk (I ended up doing 100%) of the coding for your final project. This coding is in netlogo, which is a weird functional language. Thankfully, it's not hard to learn (being very high level) and besides you don't need to make very complicated simulations anyway. I think we did a couple hundred lines of code for our project, but a lot of it was just variable setting, comments or whitespace for readability.
The structure of "do presentation, get feedback, do another presentation" will really help you iron out the kinks in your presentation. These interim presentations get TA and peer feedback, so they should keep incrementally improving from iteration to iteration. That's a great way to learn how to present. One gripe I have is the window for writing code, which means you basically have to finish coding the final project in 2 weeks (learn new language, architect and implement an application). The deliverable before "present your model" is "present your concept", so you should use that as a chance to basically pseudocode everything out, so you can implement fast once you need to code the model itself. Personally, since I was going to code everything anyway, I tried to be firm with my team about what I was comfortable I could implement quickly. If I hadn't done that, there was a risk of scope creep. Also if you have any tabletop wargaming experience, it's fairly straightforward to think about your simulation rules as a wargame ruleset you are creating and then having to implement.
For final projects, if you choose a battle to simulate that is more "modern", then the impact of luck is increased, since modern combat has a lot of interlocking systems, where a lucky hit somewhere can yield a failure cascade. For "ancient" or "medieval" battles, however, luck plays less of a part. With thousands of people swinging sharp pieces of metal at each other, some lucky hit will be offset elsewhere with a lucky hit for the opposing side. TL;DR: modern battles can be more interesting to simulate due to the effect of luck. It would also be fun to see the causal chain of events from some freak accident.
Finally, my group had good success using regression analysis to present results. It's not taught in the class, but if you have any data science background at all, you can just dip into statsmodels or R (excel if you are hard pressed) and run a regression (logistic or OLS, your choice), pull out the p-values and t-stats, and use them to tell a story. This seemed to impress a lot, even though it's a fairly basic step to take. Since you control the input dataset via your simulation, you can pretty much make sure that you get 1) "correct" signs and 2) significant p-values.