Jim, That is good and I see many cases of algorithms, but not much in trying to match to the experimental observations. I don't know what it is ... are they saying the problem is intractable, but here are some algorithms that you can play around with?
That is what is intriguing about doing the modeling, in that if you can start to make some headway then the possibilities open up regarding real analysis. I have yet to see someone make some obvious headway. Dara's approach is still one of the best alternatives .. to consider a model-based prior history and see how well it can predict in the near-term time frame.