Time, and lots of correlation data. The training course is OK if you want to build templates, and that is one thing they make a big deal about, but in reality a car company might need a new template once in 5 years, the nitty gritty of getting your models to (a) run and (b) give reasonably accurate results is mostly done by talking to other people.
You should probably grab a copy of Damian Harty's book, The Multibody Systems Approach to Vehicle Dynamics, as should I!
I'm pretty bad at the theoretical side of the Solver but when I'm given the time I can make it do what I need. Here's some nice curves showing parking efforts on concrete and on a greased plate. Black is ADAMS, orange and brown is test. The weird spike in the test data is due to bending in the tie rod we think.
https://www.mediafire.com/convkey/2247/ ... 0xu36g.jpg