There have been some great articles on how to use regression towards the mean and (roughly equivalently, I’d claim) Bayes Theorem to make projections but in case they left you still wanting more, here are some interactive slides that demonstrate the concept. For a full discussion of the mathematics, Kincaid has you covered. Anyway, here you go:
Using Bayes Theorem to Make a Projection
We started keeping a weekly log of past projections in June. This might give you a sense of just how quickly these things change 
To get a better sense of what Steamer reliability scores look like (and because I’m starting to play around with Tableau) here’s a graph of Projected ROS wOBA v. reliability score. (more…)
We set up a page where you can cook your own Steamer Percentile Projections using a player’s projected mean, his playing time and his reliability score. Let us know what you think.
A possible first step towards building better heat maps:
Next steps include using Run Expectancy and making a web app to update and publish these once a month.