I think you're confused about what sociologists do. They are not concerned with the outcome of wars, but more with the impact of social policy or phenomena. People as individuals may not obey immutable laws, but en mass they can be modeled quite effectively, just like gasses of particles can be modeled without knowing the motion of any individual particle. It is impossible to model the individual particles accurately due to their number, but given extensive properties (Temperature, Density, Volume etc) I can tell you how the gas as a whole will behave. Based on the Copenhagen interpretation of quantum mechanics, I can never tell you exactly how any one particle will behave, only give you probabilities of certain behaviors, just like people. They aren't "rational".
You absolutely can say within social science that certain things will happen at certain rates within tolerances - this is exactly what my partner did in her PhD. She performed statistical tests to show (at a 99% level of confidence) differing care-giving levels (measured by hours worked for family members) based on variables such as fertility of the individuals concerned.
She can certainly tell you, with 99% accuracy, based on the number of children a woman in Togo has and her desired levels of fertility, what the probability of her getting AIDS in the next year is. She can compare this with Benin, which has a different social support structure, and show that, for instance, the more localized family networks reduce this fertility desire and in turn reduce instances of people developing AIDS. And she can absolutely tell you, ahead of time, what the impact of building a new road to a remote village will be in terms of fertility desires, migration and infection prevalence. With measurable, repeatable numbers, statistical significance, etc. If you changed "fertility preference" for "quark mass" and connectivity of a village for "phi^4 coupling" what she and I do end up looking almost identical, so I can't claim to be doing science if she isn't.