As I see it, the initial Vegas line is an assessment of the difference between two teams, taking sundry factors into account. And it's generally a pretty good take.
I mean, to me it seems more complicated than the bookies saying, OK, where will we set the line to attract equal action between two teams. Rather, it starts with an assessment of a reasonable point differential between the two teams, based on all factors, so that bettors see it that way and make their choices accordingly.
If the initial line is reasonable and plausible, the rest will take care of itself for the bookies. The more accurate the initial assessment is, the more likely it will attract roughly equal action on both teams from the get-go.
Generally, the line doesn't move significantly. Maybe a point or two. There are exceptions of course, and they happen every week. All in all, however, the general stability of the initial line tells you not how clever the bookies are in enticing equal money on each team but rather how accurately they've assessed the contest to start so that money will flow in equally on both teams. Accuracy first. Cleverness second.
The initial line is usually pretty straight-on. Penn State, for example, opened at 4.5...moved to 5.5...and is now back in the range of 4.5-5, depending on the book. This suggests the bookies got it right to begin with, though that doesn't mean that either Penn State or USC couldn't end up coasting to an easy beat of that line. Happens all the time.
As I said above, over time the FPI winner predictions are roughly on par with those of Vegas. But the FPI doesn't predict victory margins...only winners. That's easier to do than nailing margins of victory.
I'm generally not a big fan of computer wizardry when it comes to predicting football games...and definitely not when it comes to predicting margins of victory. That said, I do think ESPN's FPI algorithms are comparatively good for the former purpose...comparatively as in compared to other computer programs.
Oh well, this is just my amateur blather on the issue...for what it's worth.