Evaluating the 2008 NBA Draft

Conventional wisdom is that it takes about 3 seasons for a draft to shake out, in terms of evaluating talent. The 2008 draftees have now had an opportunity to play 3 full seasons, so let's look at how they stack up. I'm simply going to use ezPM here, but maybe in a future post will re-visit with RAPM.

Before I post a table of the draft picks, I'll make one significant observation of the data (well, I guess I should let you decide whether it is a "significant" observation). I wanted to see whether there was a correlation between playing time (possessions) and draft slot. To do this, I regressed possessions (in 2010-11) on draft slot and the past season ezPM100 ratings (obviously, I didn't want to use WARP, since that is directly correlated with both possessions and ezPM100). The results are somewhat surprising to me (although I recall others — including Berri using WP as the player metric — getting similar results):

Call:
lm(formula = POSS ~ DRAFT + EZPM100)

Residuals:
    Min      1Q  Median      3Q     Max 
-3290.8  -786.3   125.8  1000.6  1894.2 

Coefficients:
            Estimate Std. Error t value Pr(>|t|)    
(Intercept)  4458.18     390.08  11.429 1.07e-13 ***
DRAFT         -64.86      14.32  -4.528 5.99e-05 ***
EZPM100       106.79      72.50   1.473    0.149    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 

Residual standard error: 1313 on 37 degrees of freedom
  (20 observations deleted due to missingness)
Multiple R-squared: 0.3886,	Adjusted R-squared: 0.3555 
F-statistic: 11.76 on 2 and 37 DF,  p-value: 0.0001115

It turns out that there is a strong correlation between playing time and draft slot, but essentially no correlation for ezPM100. You can see these relationships in the plots below:

2008 draft re-visited three seasons later.

You see in the first plot (upper left) that WARP generally declines with draft slot, but this is mostly a function of playing time, which, in turn, is highly correlated with draft position (lower right for possessions or upper right for games started). There is essentially no relation between ezPM100 and draft position, although it should be noted that most of the last 10 draft slots didn't play a single possession this season (you can see those data points lying along the x-axis in the plots for GS and POSS). If you're curious, that's Semih Erden at #60. Overall, 40 of the 60 draft picks played at least a single possession this season.

Here are the players sorted in descending order by WARP:

 2008 Draft Class in 2010-11

RANK NAME TEAM POS DRAFT WARP EZPM100 GS POSS
1 Kevin Love MIN 4.5 5 15.48 6.19 73 5246
2 Derrick Rose CHI 1.0 1 10.18 2.53 81 5612
3 Russell Westbrook OKC 1.0 4 10.12 2.70 82 5421
4 Serge Ibaka OKC 4.5 24 7.16 2.23 44 4158
5 Eric Gordon LAC 2.0 7 6.73 2.02 56 4062
6 DeAndre Jordan LAC 5.0 35 6.73 2.23 66 3910
7 George Hill SAS 1.0 26 6.26 1.60 5 4098
8 Luc Mbah a Moute MIL 4.0 37 5.40 1.21 52 3838
9 Danilo Gallinari NYK 3.5 6 4.62 -0.66 48 3246
10 Ryan Anderson ORL 4.0 21 4.60 2.21 14 2682
11 Omer Asik CHI 5.0 36 3.95 3.68 0 1819
12 JaVale McGee WAS 5.0 18 3.73 -0.33 74 4021
13 Brandon Rush IND 2.0 13 3.71 0.18 21 3416
14 Courtney Lee HOU 2.0 22 3.56 0.15 1 3299
15 Nicolas Batum POR 3.0 25 3.52 -0.88 67 4637
16 Jason Thompson SAC 4.5 12 3.51 -0.02 39 3426
17 D.J. Augustin CHA 1.0 9 2.32 -1.87 82 5144
18 Brook Lopez NJN 5.0 10 2.11 -2.05 82 5317
19 Darrell Arthur MEM 4.0 27 1.98 -1.24 9 3060
20 Chris Douglas-Roberts MIL 2.0 40 1.72 0.02 12 1660
21 Goran Dragic HOU 1.0 45 1.62 4.65 3 725
22 O.J. Mayo MEM 2.0 3 1.57 -1.91 17 3583
23 Roy Hibbert IND 5.0 17 1.48 -2.25 80 4431
24 Bill Walker NYK 3.0 47 1.31 -0.61 1 1555
25 Joey Dorsey TOR 4.5 33 1.03 0.02 9 988
26 Michael Beasley MIN 4.0 2 1.02 -2.63 73 4736
27 D.J. White CHA 4.0 29 1.01 -1.38 0 842
28 Mario Chalmers MIA 1.0 34 0.79 -2.47 28 2984
29 Anthony Randolph MIN 4.5 14 0.75 -0.71 3 924
30 Darnell Jackson SAC 4.5 52 0.61 -1.32 2 986
31 J.J. Hickson CLE 4.0 19 0.48 -2.97 65 4307
32 Donte Greene SAC 3.0 28 0.31 -2.88 21 2213
33 Marreese Speights PHI 4.5 16 0.29 -2.65 1 1405
34 Kosta Koufos DEN 5.0 23 0.18 4.91 1 200
35 Jerryd Bayless TOR 1.5 11 0.08 -3.00 14 2537
36 Kyle Weaver UTA 1.5 38 0.07 -1.67 0 132
37 Semih Erden BOS 5.0 60 -0.19 -3.94 7 993
38 Alexis Ajinca DAL 5.0 20 -0.55 -8.77 2 143
39 Robin Lopez PHX 5.0 15 -0.69 -4.46 56 1968
40 Sonny Weems TOR 2.0 39 -0.80 -4.26 28 2776
41 Joe Alexander 8 0 0
42 J.R. Giddens 30 0 0
43 Nikola Pekovic 31 0 0
44 Walter Sharpe 32 0 0
45 Nathan Jawai 41 0 0
46 Sean Singletary 42 0 0
47 Patrick Ewing Jr. 43 0 0
48 Ante Tomic 44 0 0
49 Trent Plaisted 46 0 0
50 Malik Hairston 48 0 0
51 Richard Hendrix 49 0 0
52 DeVon Hardin 50 0 0
53 Shan Foster 51 0 0
54 Tadija Dragicevic 53 0 0
55 Maarty Leunen 54 0 0
56 Mike Taylor 55 0 0
57 Sasha Kaun 56 0 0
58 James Gist 57 0 0
59 Joe Crawford 58 0 0
60 Deron Washington 59 0 0

The story is familiar by now. While it's great to have a top 3 or top 5 lottery pick, that is certainly not a guarantee of future success. Likewise, a late first round pick can easily turn out to be a quality starter in the league.

So, who made out like bandits in the 2008 draft? Clearly, the winners were Chicago (Derrick Rose and Omer Asik), Oklahoma City (Russell Westbrook and Serge Ibaka), and Los Angeles Clippers (Eric Gordon and DeAndre Jordan). Milwaukee got very good value out of their second round pick in Luc Mbah a Moute (drafted #37). The biggest disappointments thus far appear to be O.J. Mayo (drafted #3 and still with Memphis) and Michael Beasley  who was drafted second and is no longer with the team that drafted him (Miami).

11 thoughts on “Evaluating the 2008 NBA Draft”

  1. About half of the current starters are lagging by this measure. How much better would they look if they came off the bench?

    Might be interesting to sum draftees by starter vs bench and look at when guys converted from one to the other at some point and how EZPM changed from that exact point and see whether the decision was a beneficial one.

  2. Using those who started 39+ games (J Thompson fit better with the top group than the lower one) I find the "starters" averaged a 3.6 on position assignment compared to 3.1 for the 'non-starters". The starters average 14.5 on draft position, non=starters 29.4. The starters average WARP 4.9, non=starters 1.5. The starters average 0.1 on EZPM100, non=starters -0.8. Ah, the importance of barely getting above average performance.

    1. haha

      This particular draft was pretty much a crap shoot. Every year, there are going to be a couple of elite players who go in the top 5, although we're never really sure who they are until a few years down the road. The rest of the draft is pretty much noise.

      It makes me think that the whole notion of "reach" is kind of bogus. In order for a pick to be a "reach", that means that there is a definite correlation of value with draft position. But if there is no correlation, then you can't really be reaching. Any pick is as good as any other.

      Ibaka, Asik, Jordan all would have been called "reaches", if they were taken top 10. Hindsight, yada, yada.

      1. There are cases of "best player available" at certain draft junctures you might take; but, given these numbers of top guys barely above league average performance and the others who played only 1 point behind on EZPM100, on average it may be better to usually go with fit to system and need.

        1. Or maybe that's what teams are doing now, when they should be taken BPA (or best future talent available)? Hard to say.

        2. Teams talk about building a team thru the draft. the Thunder got 4 top 5 picks, kept 3 and used one to get Perkins. That worked.

          I don't think you can build a next gen good team without multiple top picks and / or very good / fortunate picking.

          You can get role players thru the draft but you can get role players anywhere.

        3. Yeah, you'd have to look pick by pick, team by team and for multiple years to make judgments about draft strategy and the performance of that strategy vs the alternatives.

  3. Those with position assignment 1.5 or less (PGs & close) had an average draft rank of 18.6, a WARP of 4.5, an EZPM100 of 0.6 and started 42 games.

    Those with position assignment between 1.5 and 3 (wings) had an average draft rank of 26.2, WARP of 2.2, EZPM100 of -1.0 and started 22 games.

    Those with position assignment between 3.5 and 4 (forwards) had an average draft rank of 20.1, WARP of 2.7, EZPM100 of -0.8 and started 37 games.

    Those with position assignment of 4.5 or higher (big men) had an average draft rank of 24.4, WARP of 2.8, EZPM100 of -.045 and started 34 games.

    PGs delivered more performance value in the most minutes.

    Wings and forwards delivered the least average performance.
    3rd year Bigs played the least.

  4. Probably generally harder to put up big EZPM100 numbers as a starter compared to as a bench player because of strength of opponent and competition for offensive production but it is not simple and clear cut in every case. It depends on the player and the role and whether the role draws out the best of a player and minimizes the weaknesses or is less favorable than that.

    Have you look at as starter vs off the bench splits of EZPM for specific players and aggregrated for the league?

    1. For the past season, average ezpm100 of players with >avg GS was ~0.50. Average of players with less than average GS and greater than 500 possessions was -1.43. So, on average, starters do have higher ezpm ratings.

  5. Ok, thanks.

    EZPM Splits for players with >avg GS and with less than average GS against 4+ opponent starters, or 3, or 2 or less might be interesting as well. Players will vary but general patterns seem worth knowing too especially since individual player patterns are more likely to involve noise.

Leave a Reply