Nylon Calculus: Joel Embiid, Jamal Murray and hope for the 2016 draft Class
Most draft projections kinda stink. That applies to my model, most other analytic models, informal scouting systems, most team’s internal processes, and certainly the casual fans’ strongly held, but lightly researched, Twitter opinions. Or at least the draft projections stink relative to our ability to appraise talent and predict future performance after a player has actually played in the NBA, which is saying something.
In part because of that relative malodorous nature of draft projections, I created a few models that add information from the player’s rookie campaigns to their pre-draft projections to get an updated forecast on their future performance. My general take away is that these models are significantly more accurate than draft projections, increasing R^2 from just over .40 to about .60, but also that the pre-NBA performance is still a significant contributor.
I have two models that I’ve used in the past, one that focuses on efficiency as measured by box score production, and another that mimics total production something like total Win Shares, though based on the AWS measure. Both were trained on the best performance of the players’ third and fourth year in the league.
For both models the key factors are rookie efficiency, rookie minutes played, pre-NBA efficiency, age and 3-point attempt rate. Perimeter players tend to improve somewhat more in box score based measures. Below is a scatter plot of this year’s sophomores and their projected improvement by age as of their rookie year.
The big difference between the two models is the relative weight of minutes played, which is higher in the production model. For player’s losing significant time to injury, like say Joel Embiid, the efficiency number represents a slightly more optimistic take on their future health.
Below is the list for rookies centered into a plus-minus format for ease of reference. The table is sorted by best projection in the Efficiency projection:
The leader on the efficiency list is Embiid, though even here his relative lack of minutes played lowers his estimate. If he had been able to play say 2,000 minutes at the same level, not a guaranteed proposition, his projection would climb to +3.6. It also should be said that, like any box score based measure, this projection is weaker on defensive measurements.
After Embiid the model indicates that the Knicks should be putting more development into Willy Hernangomez, his efficiency last year was surprising enough that the model projects him to remain at about his current level. Still, his current level is good enough to warrant more playing time. The next best player Efficiency projection is for Jamal Murray, a rookie that went a little under the radar in his rookie year.
In fact on the Production Model list below, Murray has the highest projected AWS Shares measure. This model also indicates some hope for those on the Brandon Ingram Atoll, as he projects to be second in the Production Model.
Next: The Washington Wizards are to be taken seriously
Interpreting a model like this for an injured player like Embiid is difficult. For most rookies their playing abilities and the trust of their coach are limiting factors for getting minutes on the court, but for Embiid the issues were obviously health related. The data I am working here with isn’t fine tuned enough to separate the whys of a rookie’s playing time totals. In general, it can be inferred that this model is the more pessimistic take on Embiid’s health.
Even those these models are better than the draft models they are still fairly uncertain. Intuitively they show that for guys coming off one year in the NBA we shouldn’t totally dismiss their pre-NBA performance. Guys like Ingram still have some so called second draft potential, just maybe not as much as they had in the first draft.