NBA Draft Projection Model and More
Edit: I’ve substantially changed the model since this post. I have also settled on using just height for the time being because any added accuracy by measurements is minimal. For more information, see this post.
The yearly draft is one of the most captivating things about NBA basketball. Part of its lure is its unpredictability. Throw in the stakes – pretty high for many struggling franchises – and we can’t help but watch with wide eyes and watering mouths. And while we’re still a long way from being able to predict players’ performance with great accuracy (30 college games is a small sample by anyone’s measure, and the inconsistent level of competition makes it harder), it’s fun to try. So I’ve been working on developing a model that attempts to project NBA performance of college players. Basically I ran multiple linear regressions on pretty much all the data I could collect (like pace-adjusted box score numbers, team sos, measurements, etc.) for college players from 2002-2008
who have played at least 2,500 minutes in the NBA (Edit: I have now included all drafted players) against their career NBA RAPM. I of course cut out unhelpful factors, and eventually came up with a fairly reasonable predictor. It’s far from perfect though, and I’ll continue to work toward accuracy. It’s better at projecting who will be successful and who won’t than it is at predicting exactly how good a player will be – and it’s really good at predicting who will perform poorly in the league. Edit: I want to point out that the model in its current form is much more accurate – and is probably best at predicting who will be an all-star caliber player in the NBA.
It’s worth noting that the model is more accurate when we add in players’ measurements – particularly standing reach and wingspan – that are available at Draft Express. But these measurements usually aren’t available, at least for many players, until right before the draft. So for now, I’ll rate some players based on a regression without those measurements – which still does a pretty reasonable job. For those interested, I have found (and I think this confirms something John Hollinger has said in the past) that the single most important predictor for NBA success is steals. For a little sneak peak, here are the players from Draft Express’s top 25:
It’s early, but it looks like Nerlens Noel is THE guy this draft class – something I have suspected for some time now – and I wouldn’t be remotely surprised if he went #1. Kentucky’s other freshmen aren’t nearly as intriguing, which is also reflected in my NCAA Estimated Impact. Needless to say, this is the reason the team is struggling to meet its lofty expectations. Elsewhere, this could be a good draft for teams that need point guards – Carter-Williams, Smart, and Burke all project pretty well at this point.
In other, probably less exciting news, I’ve updated NBA Impact by introducing pure RAPM as part of the equation, so this stuff (and check it out, I update the current NBA Impact pretty much daily) is much, much more accurate now. While we’re on the topic, by the way, just as I felt obligated to point readers to xRAPM when J.E. first put it out, I feel obligated to point you to IPV (Individual Player Value ratings). This is very similar to xRAPM – the primary difference is the SPM prior: IPV places much more emphasis on usage (particularly usg%*ast%), which the SPM prior in xRAPM ignores. The creators seem to have certainly done their due diligence, and the result is a very accurate predictor of player performance.
Anyway, stay tuned, I’ll have much more on the draft in upcoming months.