BP Daily Podcast: Effectively Wild Episode 802: The Podcast We Put Away Wet

I would guess that Sam is right about the BP projections. I would take the median to be the model's closest approximation of a player's true talent, or better middling performance level. The 90th, etc. percentile seem likely to be estimates of what happens if regular skill variation or luck favors the player over the course of a year. The 10th seems to be what would happen if the player had a bad year, either as a result of bad luck or poor performance (relative to the player's typical performance).