Nylon Calculus: 2017 P-AWS NBA Draft projections
For the last couple of years, I have been releasing a draft model called the P-AWS model. The model is trained on a box score metric, Alt Win Score, adjusted by playing time. It’s done pretty well out of sample in the last few years, so I wanted to get it on the record as we head into tonight’s draft.
The list below is ordered by the rank in the model, with the score listed as well. The easiest way to reference the model scores is that a 5.0 is about an average starter projection, an 8.0 is a borderline All-Star, while a sub 3.0 score is not at all good.
In addition to that list view, I created a visualization for statistical performance broken into categories. The statistics have been normalized so that an average score for the draft class is rated as zero, and anything below average is a negative number. They are shown in box and whisker graphs by position which lets you compare how different aspects of each prospects game rank as compared to the average of other players at their position.
In the visualization you can filter by mock draft order, position or name. Scroll over the dots to see the names of the players if not displayed.
Obviously, take all those projections with a grain of salt. While this model has proven to be a fairly accurate projection of talent and potential, there are very few sure things when it comes to the NBA Draft.