The statistic I decided to use as a general-purpose comparison is PER*MP (Player Efficiency Rating)*(Minutes Played). This gives a heavy consideration for actually being on the floor, and a heavy consideration for contributing while they are playing. John Hollinger, creater of the complicated PER, has a similar statistic - called Value Added (VA). PER*MP and VA differ in that Value Added corrects for position. Value Added is an attempt to compare a players contributions as it compares to a replacement player. For a variety of reasons VA does not lend itself well to our analysis of the NBA draft, and we will stick with PER*MP.
I calculated the PER*MP performance of every draft pick since 2002 (the last ten drafts) for every season. I then calculated the average performance for every pick, for every year since they were drafted and smoothed the line out in Excel. Basically, I found the PER*MP you should expect from a pick for every year after being drafted. For example, the 8th pick in the draft, in their 4th season since being drafted, is expected to post a PER*MP of 24154. If they average 20 minutes a game and play in 75 games, the PER would have to be 16.1 to reach the "expected" performance. This expected performance accounts for all the busts at the draft position - Rafael Araujo in 2004, for example - as well as the standouts like Rudy Gay in 2006.
The expected performance for each draft position increased until year three, then leveled out until year seven, then dropped down year eight. This behavior was much more evident early in the draft - the late picks peaked at year 3 as well, but with much smaller differences since very little is expected of a late draft pick. It should be noted that few picks reach the expected value late in the second round. Most of them post very low numbers, with a few players that go on to have good careers. The expected value is the smoothed-curve approximation to the average performance. Showing these trends on a plot was pretty hard in two dimensions, so I tried to get a decent 3D chart in excel.
Next, I had to decide how to weigh each year given the PER*MP every season. I took the expected values and figured out a formula to weigh the seasons to get a fair comparison. Obviously, good performance early is more valuable than good performance in three years. For picks with 8 full seasons to compare, the weighing is as follows:
2nd Year: 17.0%
3rd Year: 16.3%
4th Year: 15.3%
5th Year: 13.6%
6th Year: 11.2%
7th Year: 7.5%
8th Year: 1.8%
The picks that have fewer years are weighed in the same pattern, with the first year always the most valuable. For example a second-year player would have his first season weighed at just over 50%.
The 2002-2004 drafts have already had eight seasons to judge by, so these Modified Pick Values (MPVs) won't change going forward. For the more recent drafts, we don't have 8 years of performance to judge the picks on. These expected and actual MPVs will both change. Of course, this means that the expected MPV will be different for every draft after 2004, as you can see in the slideshow below.
To view any image in full resolution, click the gear in the top-right corner and select "view full resolution". Or, you can view the charts at picasa or imgur.
Since ranking things is probably the most-fun thing we can do with statistics, lets see what these numbers tell us. A more appropriate ranking system would probably be the raw value over the expected value - but with only a ten-year window, it would extremely favor the 2002-2004 drafts. So, instead we will rank them by percent of expected value - the best and worst value picks.
This is a very unfair ranking system, its just for fun. No one is suggesting that these are the best picks of the last decade - they are just the highest value picks at their spots. And since we are ranking players who have yet to finish more than a year or two, a lot of the results will be quite a bit ridiculous. So, with those qualifications, on to the rankings!
Here are the 25 top value picks since 2002 by percent of expected value:
- 2011 Isaiah Thomas 60 3844.80%
- 2007 Ramon Sessions 56 986.01%
- 2006 Paul Millsap 47 811.82%
- 2007 Marc Gasol 48 798.47%
- 2003 Mo Williams 47 664.00%
- 2005 Ryan Gomes 50 651.34%
- 2003 Kyle Korver 51 638.68%
- 2009 Marcus Thornton 43 607.14%
- 2005 Monta Ellis 40 572.20%
- 2002 Carlos Boozer 34 532.45%
- 2010 Landry Fields 39 525.20%
- 2011 Chandler Parsons 38 520.83%
- 2005 Andray Blatche 49 477.58%
- 2005 Amir Johnson 56 470.19%
- 2005 David Lee 30 457.89%
- 2005 Louis Williams 45 454.08%
- 2002 Rasual Butler 52 446.74%
- 2009 Chase Budinger 44 446.35%
- 2005 Marcin Gortat 57 440.77%
- 2009 DeJuan Blair 37 401.45%
- 2008 Goran Dragic 45 379.82%
- 2004 Trevor Ariza 43 86166 370.50%
- 2003 Josh Howard 29 148134 370.09%
- 2003 Zaza Pachulia 42 86890 361.49%
- 2011 Lavoy Allen 50 5598 345.41%
All of these top value picks are in the later part of the draft due to the lower expected values. LeBron would have had to be drafted 6th to take the 25th spot overall! Here are the 25 best value picks in the top 10:
- 2004 Andre Iguodala 9 237.22%
- 2003 Dwyane Wade 5 229.23%
- 2002 Amare Stoudemire 9 221.43%
- 2005 Chris Paul 4 221.12%
- 2008 Brook Lopez 10 203.72%
- 2002 Caron Butler 10 201.58%
- 2010 Greg Monroe 7 199.31%
- 2003 LeBron James 1 189.96%
- 2009 Brandon Jennings 10 189.94%
- 2003 Chris Bosh 4 188.13%
- 2006 Rudy Gay 8 187.23%
- 2007 Kevin Durant 2 174.80%
- 2008 Russell Westbrook 4 174.44%
- 2008 Kevin Love 5 170.17%
- 2003 Carmelo Anthony 3 164.88%
- 2006 Brandon Roy 6 159.71%
- 2004 Luol Deng 7 159.57%
- 2011 Kemba Walker 9 159.47%
- 2009 Stephen Curry 7 156.05%
- 2003 Kirk Hinrich 7 154.93%
- 2010 DeMarcus Cousins 5 153.99%
- 2005 Deron Williams 3 151.39%
- 2007 Joakim Noah 9 144.75%
- 2004 Dwight Howard 1 143.88%
- 2009 DeMar DeRozan 9 140.63%
This ranking shows the importance of not getting injured to be considered a valuable draft pick, with players like Blake Griffin, who would be over 150% of his expected value if he could be judged for only two seasons (but has to be judged for three since he sat a season after being drafted by the Clippers). And finally, the worst top 10 picks of all time:
- 2011 Jonas Valanciunas 5 0.00%
- 2006 Mouhamed Sene 10 3.21%
- 2006 Patrick O'Bryant 9 4.89%
- 2004 Luke Jackson 10 5.36%
- 2002 Nikoloz Tskitishvili 5 6.24%
- 2004 Rafael Araujo 8 7.51%
- 2008 Joe Alexander 8 8.39%
- 2002 Jay Williams 2 11.28%
- 2009 Hasheem Thabeet 2 13.65%
- 2006 Adam Morrison 3 13.85%
- 2002 Dajuan Wagner 6 14.65%
- 2007 Greg Oden 1 16.84%
- 2009 Ricky Rubio 5 24.10%
- 2003 Darko Milicic 2 31.04%
- 2005 Ike Diogu 9 33.04%
- 2007 Brandan Wright 8 36.07%
- 2004 Shaun Livingston 4 36.33%
- 2006 Shelden Williams 5 41.80%
- 2003 Mike Sweetney 9 42.35%
- 2009 Jordan Hill 8 48.57%
- 2011 Enes Kanter 3 48.74%
- 2007 Yi Jianlian 6 54.47%
- 2009 Jonny Flynn 6 55.80%
- 2010 Ekpe Udoh 6 56.76%
- 2005 Martell Webster 6 59.74%
Of course, #1 in this ranking, Jonas Valanciunas, will actually be playing next year, so it isn't very fair to call him the worst pick ever. But generally top 10 picks are drafted to play right away. He will still easily be able to exceed his Expected Value with a nice career though.
A huge thanks to the fine folks at basketball-reference.com for their wonderful stats database, I got all of the raw data from their site. Also, Grantland.com's Bill Barnwell probably inspired this series (and maybe this website) with his NFL draft analysis. And finally, 82games.com did something similar to this a few years ago but did it the easy way, which didn't do the job justice.
This was Part I of the StatDance.com NBA draft analysis.
No comments:
Post a Comment