RamaY seems to like "NetLogo," so maybe that is something that could work for you. I am quite comfortable with eclipse and use it routinely for all coding work. If you want to go this route, then you have a bit of setup to get started:
Look at: http://www.aass.oru.se/~fpa/teaching/In ... clipse.pdf
Look at: http://www.vogella.de/articles/UML/article.html
as an overview.
Once you get that far, you are ready to start creating diagrams.
But, first, the purpose, at least as I see it, is to develop a framework for geostrategic modeling that a variety of "thinkers" can then use to create scenarios and worlds and understand and control their evolution.
Second, to build the model, we choose uml2 because it allows us to create structure, behavior and interaction
in very general terms and this is useful create geospatial, dynamical, and game elements in our model. There are partial things that are easily available. For example, if we assumed there was no geospatial element and there was no dynamics, but rather just a game, then we have interesting things like excel spreadsheets to model games (hawks/doves etc.). Much can be learned from that. Geospatial representations can be easily had in things like Arcview or other GIS software, but it has been difficult to embed games (though some dynamics is now possible). Similarly, if you just wanted to do dynamics, what happens when n-bodies apply some control laws, then you could simply do it in matlab. Putting all these three together, so that there is an environment, there are events, there are actions there is interaction and there are games is what requires a more generic modeling framework and uml comes very close to doing that. But, as with many things, the increase in flexibility comes with an increase in learning curve. That is ok, I think, because I am not pressed for time, just yet.
Have a look at how to create various kinds of diagrams, e.g. http://www.agilemodeling.com/essays/umlDiagrams.htm
Once we have a reasonable set of tools at hand, both modeling and learning the toolkit goes hand in hand.
How are we to model then and this is where everybody needs to chime in, in terms of a generic systems architecture, which is at a slightly "lower" level than "India takes Hambanthota, Hu Jintao pisses in his pants" level. This is simple things.
Structure: What is the domain, how big must each cell represent. What does each cell contain, who owns this cell etc. These are just creating various classes.
Behavior: How do attributes in a cell change. What is a goal, how is it represented. What kind of actions are available to make a goal. etc.
Interaction: There is a master clock. There is a "script" of events. The clock goes, events happen, agents behave; the environment monitors the behavior, arbitrates, scores and redistributes attributes etc.
At that point we have an easy way to run scenarios.
Much of this work could already be done, and if that's true, we should go there. I haven't though may be what RamaY suggests is a possibility. But you should look at the use of uml2 to model the great game, shown in this thread or somewhere else, earlier. That got my attention.