I'd argue that point a bit mate. Simulation software now IS incredibly powerful, some of the stuff i've been using recently is industry standard both at OEM and F1/WRC level. The simulations are incredibly accurate subject to the quality of the data entered into them.

For example, even going back to 2002 when Cosworth were developing the Focus ST170 with Ford, their engine simulation was validated on an engine dyno to within 2% accuracy. Moving forward nearly 10 years, things are even more accurate now.

My dissertation this year is to do with variable valve timing and i'm hoping that my own simulation will validate within a 5% accuracy margin. With the software simulation I can see the affects of changing my cams to assymetrical, increasing/decreasing duration, increasing/decreasing lift, alterations to the timing where peak lift occurs etc. Not only that but I can see full analysis of the valvetrain in terms of cam follower inertia, stress, oil film thickness, wear load, camshaft bending dynamics - even down to the cam belt tension!

Last time I checked, it was pretty hard to check all those variables by trial and error testing?

Admittedly the software license for this is $30,000 a year, (fortunately i've got access to it as part of my degree) but you can literally target any changes/areas of interest with the click of a mouse and evolve a camshaft through to the final stages before getting a prototype made.