Yes, this seems like a difficult challenge. However, this and even more complicated scenarios are routinely dealt with by quantitative methods. Statisticians solved problems like these long ago, and currently have so many models available at their disposal to deal with these scenarios. Any reasonably modern data analysis system (we don't even have to go to expensive Oracle / SAP / SAS solutions) can model data such as you describe and proceed to give you probabilty distributions.
In other words: What you describe is not a problem for people whose job is to actually analyze data. We've solved it long ago, even before computers, but now that we have powerful computers in place, it's even easier. I don't know how complex and capable Funcom's data analysis tools are, but as long as they aren't stone age, we can safely assume they are capable of modelling and running simulations for the complex data you describe.