Expected Value of an NBA Draft Pick

Using data from Stathead.com, I built a dataset of 1,322 draftees across 22 NBA draft classes (1989–2011), measuring each player's value through "Prime Win Shares" — the sum of their five best seasons by Win Shares, a metric adapted from Brocato Prime Wins. Players who never appeared in the NBA are included with zero production, since a pick that goes nowhere is still a pick spent.

The analysis progressively refines the model: starting with a simple OLS regression, testing its assumptions, applying a log transformation to address skewness and non-linearity, grouping picks into draft tiers (Top Pick / Lottery / First Round / Second Round), and using bootstrap aggregation (bagging) to stabilize coefficient estimates.

The final models achieve an R² of ~0.326–0.328 — meaningful, but deliberately modest. The results highlight something the numbers make clear: draft position sets an expectation, not a guarantee. The variance in outcomes is the story.