Models take into account current conditions throughout many layers of the atmosphere on both a global and local scale. These conditions change, and when a previous model run's +6 hour prediction doesn't match what happens 6 hours after the run, then the run 6 hours later will be different. What you are seeing is the models continually updating themselves, and thus pushing the storm further west, based on differences between actual conditions and what they had prediction the conditions would be.

If you want to read up on how the various tropical models work, take a look here: http://www.nhc.noaa.gov/aboutmodels.shtml

Frankly, hurricanes are very erratic systems, and very difficult for models to predict. Think of them as a spinning top that doesn't always go where you think it should.

A lot of this comes back to the action principle (also known as the Lagrangian). Any system will take the path of least action. Action includes energy and momentum. The problem is understanding what the conditions that effect the energy and momentum of a hurricane are, and accurately detecting them in the atmosphere. While dynamical models are good on normal systems, hurricanes just cause problems sometimes