Not intending to be rude, but this is a textbook example of reasoning that I call the liberal-arts fallacy. It goes something like this: If I can describe X in a very vague way, then it must be very simple to build Y. It's just details.
For example: There are very tall buildings in many cities in modern times. In the old days, buildings were only a couple of stories tall. Modern engineering has allowed us to make them much taller. Adding more height is just putting one story on top of another, right? So I expect to see million-story buildings tomorrow. (And if I don't it's because people are too lazy. -- You didn't say this, but this is often included in the argument.)
Also, your hockey comparison is a bad example, because the collision calculations in that example would be relatively simple compared to what (I think) would be required for satisfactory combat. Combat would require a system that involves angles, surfaces, many different types of micro-movements, sway, balance, and probably many more dimensions. And the calculations would also have to factor in many different elements related to particular weapons.
I'm not sure where you would draw the line at approximating all of the above, and I'm sure that different people would have different opinions on what would be good enough. But you wouldn't have to go very deep to end up with a system that would be more complicated and compute-intensive than the physics engines we're seeing in games right now.