By Kurt Bullard
With Brandon Ingram and Ben Simmons slated to go with the first two picks of the 2016 NBA draft, a lot of the pre-draft talk—at least in the Boston area—has centered around what the Boston Celtics should do with the No. 3 pick; namely, about whether to trade it or keep the pick. One of these trades being specifically floated out by the media is a straight-up trade for Jahlil Okafor, since the Celtics in the past have reportedly targeted the former Duke center (though it was never confirmed).
From a qualitative standpoint, Okafor’s season was up-and-down. The rookie missed 29 games due to injury, but still managed to put up 17 points and 7 rebounds per game while he was on the court. While his traditional stats seem great for a rookie, advanced metrics tend to point the other way, as he finished the year with a negative VORP and a -4.1 BPM. His game draws a lot of proponents and critics. While detractors tend to point out his subpar defense, lack of fit in the modern NBA, and that someone has to score for the Sixers, supporters claim that he is destined to become one of the best low-post scorers in the association.
At the end of the day, this potential rumored trade comes down to whether or not the Celtics should trade this year’s No. 3 pick—an unknown commodity—for last year’s No. 3 pick—an asset that has been observable for one year. To answer that question, one has to look at how much information Jahlil’s first year can tell us about how his career will pan out.
For this undertaking, I created two models using player data whose careers started between 1985 and 2007 and were selected with a top 15 pick:
Regressing Win Shares over the first eight years of a player’s career against a player’s pick (Control Model)
Regressing Win Shares over the first eight years of a player’s career against a player’s pick AND the player’s rookie year Win Shares per Minute
Both of the above were general additive models, which allow for smoothing and use cross-validation to determine the amount of smoothing to optimize the predictive power of the model.
Testing whether the inclusion of the Rookie Win Shares per Minutes Played is significant in the model using ANOVA testing yields a significant p-value, meaning that a player’s first-year performance is predictive of a player’s future performance, controlling for his draft pick position.
Below shows how draft pick selection and WS/MP during one’s rookie season influences the prediction of the model.
Using this model and the model that uses only draft pick position, one can predict how Jahlil Okafor would be predicted to do in his career against an unknown No. 3 pick. Okafor’s WS/MP last year was .0007, which is a very low figure.
Jahlil Okafor Career Prediction: 36.5 WS
No. 3 Pick Prediction: 40.7 WS
Jahlil isn’t projected to do better than an average No. 3 pick, so, on average, it would be better for the Celtics to just make the pick than it would be to make this trade straight up.
This model assumes, however, that each draft has an equal amount of talent. This may not be true this year, as is evident with all of the Boston media talk about trading the pick. Okafor might also be able to help sooner than a rookie would if the Celtics want to contend within the next few years, since he is further along in his career than would be Jamal Murray, Buddy Hield, or Dragan Bender.
The following table shows how the top 15 picks from last season, with KAT leading the field by far. Mudiay had negative win shares this year, which my model does not handle well, so I excluded him from the analysis. Win Shares I believe tend to favor offense as well, so Justise is perhaps projected a bit low as well.
So in the end, a player’s rookie year can’t just be thrown out the window when projecting for his future. But, Danny Ainge has a track record of success, and, as a Celtics’ fan, I trust whatever he does.
Model 1 Summary
Model 2 Summary