For the previous four seasons, my model correctly ranked 85% of top-10 players outside of the bottom two tiers (>= Tier 5).

An equivalent number of players for the given seasons was chosen according to their ADP. These correctly ranked top-10 players only 73% of the time.