I see an experiment coming up here. ;) Take two identical Rybka engines and have them play against each other, one with and one without book with alternating colors. Let them play at least 100 games and see what happens.

I will post the result here in a couple of days :)

     I think that what I have demonstrated, in both the Ehlvest match and in my various private engine-engine tests, is that the disadvantage of minimal opening knowledge can be kept to close to 50 Elo points by choosing openings that are slightly inferior but relatively safe. Naturally if you try to play a poison pawn Najdorf or any other razor-sharp opening, the value of theory is more than fifty points. But there are many openings and defenses that can be played with little knowledge that are only slightly inferior. As White, you can easily get an equal game with openings like the Colle or London system that require little knowledge, and since White's advantage is less than fifty Elo this means that the difference between complete knowledge of theory and just minimal knowledge is worth less than fifty Elo for White. As Black, the difference is greater, but not much greater. By choosing something a bit passive like the Philidor or Steinitz defense (to the Spanish), your disadvantage will surely not be more than a hundred Elo, and if this is compared to the normal forty or so for White it means that Black might lose sixty Elo playing this way, which requires only minimal knowledge.
     Openings are overwhelmingly important in GM play because GMs cannot concede fifty Elo points, and if they try to play sharp openings they must know every detail.