I always saw it as constructing versus running models. Historians dispassionately collect and interpret data. In doing so they can come to surprising conclusions, e.g. around how violence [1] or hegemony [2] propagate. Counterfactualists run these models to simulate decisions, both past and present.
Have there been any serious attempts to generate more rigorous models that can be simulated? I'm sure any such attempt would fall far short of reality, but even a terrible failure would be fascinating.
I basically agree with DanielBMarkham's comment below, but I have nonetheless been thinking about trying to do something like this for awhile now. I've been struck by how many historians my age got interested in history in part because of the Civilization game franchise. When I was a kid I became obsessed with setting up Civ II games to run over night with every civ played by the AI, then seeing which one ended up conquering the New World. At some level, this surely influenced my decision to study the Columbian Exchange in grad school, which is what I work on now. So even if it doesn't yield any usable empirical data, I think it would be an amazing teaching tool. And who knows, maybe in 10 years, some hyper-advanced version of an historical simulation game might end up proving useful in answering certain highly specific historical question (i.e., why did Spanish galleons suffer a lower rate of shipwreck than Portuguese ones, or the like).
At any rate, if anyone knows about projects along these lines, I'd be very interested to hear about them. I think the trouble would be 1) convincing funding agencies that it's not a fool's errand and 2) arranging for an effective collaboration between historians and computer scientists/game designers/whoever else would have relevant expertise. But it does strike me as something which could be fascinating.
[1] http://www.amazon.com/The-Better-Angels-Our-Nature/dp/149151...
[2 http://www.amazon.com/Tragedy-Great-Power-Politics-Updated/d...