There are several issues I see with this.
One is the fact that at the top I am quite sure the opening theory extends past 5 moves on each side in most games.
Second is that if one person plays very badly such as hanging pieces, walking into obvious tactical traps, or has to deal with a very long forced variation, the computer will agree with most of his moves. The point I make with this, is that computer agreement with moves alone in one game is not enough, you must look at the moves themselves, and see if they are obvious moves, or something else.
Not every game needs to be watched, but this could be streamlined into if you wish to submit a questionable game, then this could help filter out which ones the game moderators should look at.
Also, how closely does Rybka, Crafty, and Fritz agree on most moves?