The old school way was to set the servo arm pointing straight back (or nearly so) at the closed position. That's assuming the throttle is closed by pulling back; the other type would be set up straight forward. Idle speed would be set with the pushrod length, which was the most precise way to do it given how inaccurate servos were 40 years ago. That setup gave the secondary benefit of throttle exponential, which the method I described above also does. I personally think that spending time getting the mechanical linkage just so on a throttle isn't worth it since servos are so much more precise these days and there is no concern at all for torque or resolution on a throttle. It's different on a control surface, but with throttles let things be easy if they can.