Which of 2005's division winners won't repeat in 2006?
We start with the assumption that the very best teams of 2005 won't be as good in 2006.
Why? The technical term is "regression to the mean." In nonstatistician terms, the reasoning is simply this: Teams that win a significantly impressive number of games were probably a little bit lucky (among other things), and can't count on their luck to hold.