NBA Draft Analysis

I've wanted to look at the NBA draft for a while now - I had lots of questions. I tried to answer a few of them by looking at the last ten NBA drafts (2002-2011) and looking at how their careers turned out relative to their draft positions.

The first thing I had to figure out was how to compare careers. Simple box-score metrics obviously don't work - looking at points per game would be a very poor single indicator of career success in the NBA. I did some cursory investigating into advanced basketball statistics (APBRmetrics) and found a lot of ideas are out there.

I get the majority of my data from the wonderful Basketball-Reference site, and found that they list Player Efficiency Rating (PER) and Win Shares. Other sites, like the NBA Geek, list Wins Produced. Another metric is adjusted plus/minus, and while I'm sure it has its merits, I'm not really interested in assessing NBA drafts in a world that has Eric Bledsoe as the best per-minute player in the NBA last year. Its not just refusing to accept something that goes against my pre-concieved notions of what happened last year, its that public perception and box-score statistics are how players and drafts are evaluated.

I liked Win Shares and Wins Produced - but they both attempt to gauge a players defensive contributions by looking at how the team did. I don't want to give a player credit for playing with good defenders, it doesn't seem fair to me.

The most important metric is minutes played - if you're good enough to get on the floor and stay on the floor, you're a contributing member of the team. No other stat can replace that. I decided the best statistic to use for measuring a players quality of contribution is the Player Efficiency Rating. Despite its flaws, it gives a great picture of a player's ability to contribute. Most importantly, for this exercise, it is relatively consistent with perception. If a GM drafts someone with high PER for his draft position, chances are that pick will be viewed as a "success" when the GM is evaluated.

What I calculated is the PER*Minutes Played - PERMP - for each player drafted since 2002. You can view each photo in full resolution by clicking the gear in the top right corner, or just view the album in full here.



These results were very interesting to me and once I averaged each draft pick's performance per year, it gave relatively nice curve.

The formula for the percent of the first pick each pick is worth: 318307*e^(-0.06167*Pick)/299270. This gives:

1 100%
2 94.0%
3 88.4%
4 83.1%
5 78.1%
10 57.4%
20 31.0%
30 16.7%
45 6.63%
60 2.63%

Of course, the numbers at the top are silly - in the 2012 draft, the top pick was probably worth double any other pick - it is expected that Anthony Davis be a superstar, and everyone else would be a longshot for superstar status (see: 2004 Dwight Howard or maybe 2003 LeBron (one pick being significantly better, despite there being a lot of superstars in 2004)).

But, in the 2007 draft, the top pick was only marginally better than the second pick - you were still getting Oden or Durant (which, at the time, was a toss-up). But once you leave the top 5, the pick values are a lot more useful and consistent from draft to draft.

Again, much thanks to basketball-reference.com for all their data. If you have time, check out an analysis posted at 82games.com - he analyzed the drafts from 1980-2003 with a significantly different process and got very similar results.

Note: Originally posted on my tumblr blog.