Seriously, the traditional method seems primarily designed to ensure that the top two seeds meet in the finals. Thus, if the higher rated player always wins, the top seed will have the easiest game in each round, the second seed will have the second easiest, and so on.

Another way to put it is that the system is designed to permit the round of N players to consist of the top N seeds, which does seem obviously correct to me. However, if you assume that the higher seed always wins, then as long as the top N/2 play only the bottom N/2 in each round, it doesn't matter how the opponents are chosen.

There's an interesting article at http://en.wikipedia.org/wiki/Single-elimination_tournament which suggests that I may have led a sheltered life. Another article at http://www.tcnj.edu/~curley2/Tournaments/Tournament.htm

The latter points out that it's not necessary to seed all players according to skill; it suggests seeding about the top quarter of participants by skill and the rest randomly.