interesting insider article from the other day..
The truth about analytical methods is that once in a while you'll get a result that flies in the face of conventional wisdom. When that happens, it means one of two things: 1. The analytics saw something that everybody else couldn't see. Or 2. Everybody else saw something that the analytics couldn't see.
And in the case of two particular players in this year's NBA draft, it will be very interesting to find out the answer.
The draft is Thursday, June 25, and now that we know who's in and who's out, it's time to unveil this year's Draft Rater -- a statistical projection of the top NBA prospects coming out of the college ranks.
<offer></offer>
To review for the uninitiated, the Draft Rater looks at a player's college production in a variety of metrics and a few other salient facts (such as his height, birth date and years of college experience), and from that projects what a player's player efficiency rating will be when he reaches his peak.
<!-- INLINE MODULE -->
Draft Rater History
<!-- END INLINE MODULE -->The basic idea is to use the NBA's past to predict its future. The Draft Rater looks back at prospects from past drafts and then, using regression analysis, identifies which attributes determined pro success and which didn't. My database of college players goes back to 2002, which is still a bit limited, but the rater gets smarter each year because it has more information with which to work -- not only an extra year of drafts but also an extra year of pro seasons from every prospect.
This year, several subtle changes helped reduce the error rate when back-tested on previous drafts. First, I ran a separate regression for each of the three position categories -- point guards, wings and bigs -- something that wasn't really feasible when I started doing this. But now that the pool of prospects is large enough, this method has produced greater accuracy.
Second, instead of tying the projection to a player's third-year PER, I used a more general descriptor of what his peak value was, allowing me to minimize the impact of fluke seasons and better adjust for some players who entered the league young and didn't max out until their fourth or fifth season. (Some of these players will perform much better than projected, but keep in mind that it's all relative. For more on why the projections seem low,
see this explanation.)
Using those changes, I was able to reduce the standard error in the projections from last year's 4.0 to this year's 2.8. This means nothing to 98 percent of you, but the number geeks in the crowd will recognize that this is still quite large -- as you might expect when you're trying to project what a 19-year-old will do when he's 25. Nonetheless, it represents a significant improvement from last year.
The one area where the method still appears to struggle is with one-and-done freshmen, and this speaks to a more general problem: Information is the key to making this thing work, and the more information we have, the better. For players who leave after their first year, the picture is often incomplete, whether we're using a statistical model or traditional scouting.
I bring this up because last year, in particular, was a rough one for the projection system. First, it was an unusual rookie class in general because nearly every player taken in the first round was at least somewhat productive; generally, a draft will have 10 to 12 impactful players and the rest will be filler, but this past season blew that standard away.
Moreover, a number of those players played only one college season, and although the rater had an accurate view of a few (such as
Kevin Love and
Michael Beasley), it missed the boat on some who performed extremely well (including
Derrick Rose to an extent, and
O.J. Mayo,
Anthony Randolph and
Eric Gordon). Gordon is perhaps easier to understand because he was playing hurt at Indiana and his primary skill (shooting) didn't show through statistically, but that doesn't excuse the others.
One important thing to point out is that the Draft Rater is rating pro potential, which is sometimes different from pro performance, depending on the professionalism and work ethic of the player involved. In other words, the fact that
Michael Sweetney and
Shawne Williams rated very highly in previous seasons isn't necessarily a damnation of the system. Rather, their off-the-court habits are the type of thing every general manager has to take into account when evaluating players and something that is usually invisible when looking at their college performance.
That said, before last season, the Draft Rater had performed extremely well.
From 2002 to 2007, 15 players were (a) among the first 10 collegians drafted and (b) excluded from the top 12 by the Draft Rater. In other words, these were the college players the Draft Rater thought were drafted too high. Of those 15, not one has played in an All-Star Game. The only two who have started a significant number of games in the long term have been
Kirk Hinrich (who was 13th in the Draft Rater in 2003) and
Charlie Villanueva.
Who were the other top-10 picks not favored by the Draft Rater?
Spencer Hawes,
Acie Law,
Fred Jones,
Melvin Ely,
Marcus Haislip,
Jarvis Hayes,
Rafael Araujo,
Ike Diogu,
Channing Frye,
Randy Foye,
J.J. Redick and
Patrick O'Bryant.
In other words, when the Draft Rater has suggested avoiding a player, that has turned out to be good advice.
The Draft Rater also has spotted some of the biggest steals in recent years:
•
Carlos Boozer was the 26th collegian taken in 2002; Draft Rater had him second.
•
Josh Howard was 17th in 2003; Draft Rater had him fifth.
•
Danny Granger was the 13th collegian in 2005; Draft Rater had him third.
•
Rajon Rondo was the 16th collegian taken in 2006, but Draft Rater had him second.
•
Rodney Stuckey was the 14th collegian chosen in 2007; Draft Rater had him fifth.
• And last year, two players the Draft Rater had rated much higher than others did,
Mario Chalmers and
George Hill, had productive rookie seasons.
So, most of the time, when the Draft Rater puts a player in the top five, there's a good reason.
All of which leads us to 2009 and whom the Draft Rater likes and doesn't like.
This year, the Draft Rater is closer to the general draft consensus than usual, with two glaring exceptions that I referenced above.
Let's get to them:
The pleasant surprise: Ty Lawson
Two players are neck and neck for the top spot in this year's Draft Rater. You could easily guess that one of them is
Blake Griffin, but most folks never would have guessed that the other is Lawson.
Lawson, who is coming off an electric performance in leading North Carolina to the championship, grades out highly for several reasons: Although he's short for a point guard, his shooting numbers (47.1 percent on 3-pointers), strong assist rate and microscopic turnover ratio (9.1, first among point guard prospects) all point to him as an NBA keeper.
The Draft Rater puts Lawson slightly ahead of Griffin for first, but this doesn't mean a team should take Lawson first. The standard error in the projections for point guards is higher than it is for big men, which means random noise could be putting Lawson ahead just as easily as on-the-court performance. If the consensus is that Griffin is the better player, I don't think Lawson's statistical record alone is strong enough evidence to refute it. Additionally, we've heard questions about Lawson's work ethic and injuries.
But the rating is emphatic enough for me to say Lawson should be at the top of the college point guard ladder, ahead of
Jonny Flynn,
Jrue Holiday,
Jeff Teague & Co. (If you're wondering about
Ricky Rubio, I'll have more on him next week.)
The unpleasant surprise: DeMar DeRozan
I'd be hard-pressed to name a potential high lottery pick throughout the years about whom the Draft Rater has been less excited. I rated 90 prospects for this draft, and DeRozan ranked 54th among them. Two of his USC teammates --
Daniel Hackett and
Taj Gibson -- outranked him, as did assorted other nonentities such as Kevin Rogers,
Chinemelu Elonu and
Ben Woodside. I'll wait here while you Google them.
Why? Because there really isn't anything in DeRozan's statistical profile that makes you think "NBA star." He rarely took or made 3-pointers, and he had a strongly negative pure point rating, which are two powerful indicators for a wing player. His numbers in other areas were unimpressive, too. In particular, he was a bad free throw shooter, which indicates that his outside shot might never be a strong suit.
Some scouts I have talked to have compared DeRozan to
Rudy Gay in terms of being an NBA athlete but having a questionable motor. But that comparison falls flat, according to the Draft Rater: Gay was the top-rated player in his draft class, while DeRozan is nowhere close. And although he's supposed to be a great athlete, he didn't show it on the court often enough: His rebound, block and steal totals were all very ordinary.
As I mentioned above, one-and-done players sometimes fool the system -- they're the youngest, least experienced guys in the pool, and thus, a major factor is how much they improve post-draft rather than just how good they are pre-draft.
Nonetheless, I would back away from DeRozan if the 12 relatively safe guys at the top of the Draft Rater are still on the board.
Speaking of which, let's take a look at the collegians for 2009.
Rankings: The top 12
Top 12 rated collegians for 2009
<table> <thead> <tr><th>Player</th> <th>School</th> <th>Draft Rater</th> </tr> </thead> <tbody> <tr class="last"> <td>1. Ty Lawson</td> <td>North Carolina</td> <td>16.34</td> </tr> <tr class="last"> <td>2. Blake Griffin</td> <td>Oklahoma</td> <td>16.21</td> </tr> <tr class="last"> <td>3.
Tyreke Evans</td> <td>Memphis</td> <td>15.02</td> </tr> <tr class="last"> <td>4.
Austin Daye</td> <td>Gonzaga</td> <td>14.24</td> </tr> <tr class="last"> <td>5.
Stephen Curry</td> <td>Davidson</td> <td>14.18</td> </tr> <tr class="last"> <td>6.
Nick Calathes</td> <td>Florida</td> <td>13.66</td> </tr> <tr class="last"> <td>7.
DeJuan Blair</td> <td>Pittsburgh</td> <td>13.56</td> </tr> <tr class="last"> <td>8.
Danny Green</td> <td>North Carolina</td> <td>13.28</td> </tr> <tr class="last"> <td>9. Jonny Flynn</td> <td>Syracuse</td> <td>12.99</td> </tr> <tr class="last"> <td>10.
James Harden</td> <td>Arizona State</td> <td>12.97</td> </tr> <tr class="last"> <td>11.
Hasheem Thabeet</td> <td>Connecticut</td> <td>12.90</td> </tr> <tr class="last"> <td>12.
Earl Clark</td> <td>Louisville</td> <td>12.88</td> </tr> </tbody> </table>
For starters, let's talk about two of the players who play multiple positions -- this matters now that we're rating players in part based on position.
Stephen Curry graded out at 14.18 as a wing but only 12.86 as a point guard. Either way, it puts him in the top dozen players, but by this rating, he's a much better prospect if he's able to defend against wings.
The difference for Earl Clark was less dramatic, but he rated slightly better as a wing than as a big man (12.14), which would have dropped him from 12th to 15th.
A couple of other names on here are likely to raise eyebrows:
Austin Daye might not have had a great season, but the Draft Rater looks favorably upon a 6-foot-11 small forward who can shoot (assuming he can play the 3 in the NBA). His numbers were strongest in the categories that project best to the pros, including 42.9 percent shooting percentage on 3s and 2.1 blocks per game, which is why he moves all the way up to No. 4 on this list.
Nick Calathes is under contract in Greece but still will be draft-eligible, and he rates higher than the hot point guards most teams are discussing in the top 15. Although he has been knocked for his athleticism, he had high rates of rebounds and steals and a strong 2-point shooting percentage. Teams in luxury tax trouble should look particularly hard at him because he can be stashed in Europe for a year or so.
Danny Green is the other surprise on this list. He's rated highly every year I've rated him, so seeing his name doesn't shock me anymore, but he has received little attention nationally. Still, he's a great shooter who can defend, and he rates as the third-best wing after Daye and Tyreke Evans.
Rankings: 13 to 25
Collegians: No. 13 through 25
<table> <thead> <tr><th>Player</th> <th>School</th> <th>Draft Rater</th> </tr> </thead> <tbody> <tr class="last"> <td>13. Jrue Holiday</td> <td>UCLA</td> <td>12.73</td> </tr> <tr class="last"> <td>14. Jeff Teague</td> <td>Wake Forest</td> <td>12.50</td> </tr> <tr class="last"> <td>15.
Gerald Henderson</td> <td>Duke</td> <td>12.17</td> </tr> <tr class="last"> <td>16.
Paul Delaney</td> <td>UAB</td> <td>11.85</td> </tr> <tr class="last"> <td>17.
Aaron Jackson</td> <td>Duquesne</td> <td>11.83</td> </tr> <tr class="last"> <td>18.
Darren Collison</td> <td>UCLA</td> <td>11.80</td> </tr> <tr class="last"> <td>19.
Terrence Williams</td> <td>Louisville</td> <td>11.80</td> </tr> <tr class="last"> <td>20.
Leo Lyons</td> <td>Missouri</td> <td>11.53</td> </tr> <tr class="last"> <td>21.
Eric Maynor</td> <td>VCU</td> <td>11.35</td> </tr> <tr class="last"> <td>22.
John Bryant </td> <td>Santa Clara</td> <td>11.30</td> </tr> <tr class="last"> <td>23.
DeMarre Carroll</td> <td>Missouri</td> <td>11.18</td> </tr> <tr class="last"> <td>24.
Tyler Hansbrough</td> <td>North Carolina</td> <td>11.11</td> </tr> <tr class="last"> <td>25.
Wayne Ellington</td> <td>North Carolina</td> <td>11.04</td> </tr> </tbody> </table>
This part of the list is an interesting mishmash of potential sleepers and potential busts. In general, players in this range have some kind of NBA career but always can count on getting some quality time with the family during All-Star Weekend.
We're awash in point guards in this draft, and six of the top nine names in this section play the position. The lesson is this: If you're in the market for a point guard, one will fall to you, and they're more or less the same after the first couple.
Down at No. 13, Holiday is a bit of a surprise -- given that he's projected to go higher -- but he has the two characteristics that produce the greatest error rate in the Draft Rater: He's a point guard and has played only one year. In other words, his real value might be much higher or much lower, and because the consensus is much higher, it wouldn't bother me to use a top-eight pick on him.
Delaney and Jackson are second-round sleepers at the point, but because projections for point guards are a bit more volatile, perhaps they shouldn't really be this high. The other "who's he?" on the list, Bryant, is a 6-11, 275-pound center from Santa Clara who could have a fine 10-year career as a third center in the
Greg Kite/
Aaron Gray mold.
Rankings: Potential disappointments
Collegians: Other notables
<table> <thead> <tr><th>Player</th> <th>School</th> <th>Draft Rater</th> </tr> </thead> <tbody> <tr class="last"> <td>26.
Jordan Hill</td> <td>Arizona</td> <td>10.97</td> </tr> <tr class="last"> <td>28. B.J. Mullens</td> <td>Ohio State</td> <td>10.81</td> </tr> <tr class="last"> <td>30.
James Johnson</td> <td>Wake Forest</td> <td>10.63</td> </tr> <tr class="last"> <td>31.
Chase Budinger</td> <td>Arizona</td> <td>10.51</td> </tr> <tr class="last"> <td>45.
Derrick Brown</td> <td>Xavier</td> <td>9.55</td> </tr> <tr class="last"> <td>48.
DaJuan Summers </td> <td>Georgetown</td> <td>9.38</td> </tr> <tr class="last"> <td>51.
Jodie Meeks</td> <td>Kentucky</td> <td>9.35</td> </tr> <tr class="last"> <td>52.
Sam Young</td> <td>Pitt</td> <td>9.34</td> </tr> <tr class="last"> <td>54. DeMar DeRozan</td> <td>USC</td> <td>9.26</td> </tr> <tr class="last"> <td>62.
Toney Douglas </td> <td>Florida State</td> <td>8.56</td> </tr> <tr class="last"> <td>68.
Patrick Mills</td> <td>Saint Mary's</td> <td>8.36</td> </tr> <tr class="last"> <td>83.
Jack McClinton</td> <td>Miami</td> <td>6.64</td> </tr> </tbody> </table>
And here's where we get to the players the Draft Rater is down on.
Several potential first-round picks don't pass muster here, with short, shoot-first combo guards in particular bearing the brunt of the Draft Rater's wrath -- Jack McClinton, Patrick Mills and Toney Douglas were the three lowest-rated "name" prospects, and Jodie Meeks didn't fare a whole lot better.
The other big surprise down here is Jordan Hill, who could go as high as No. 4 but rates 26th in the Draft Rater. Hill had solid rebounding and scoring numbers, but his percentages weren't off the charts, and his poor assist and turnover numbers were a red flag. Although one might think that ballhandling categories wouldn't matter for a power forward, apparently they do -- pure point rating (a measure of how a player passes and handles the ball) is a pretty strong success indicator for frontcourt players, and only four prospects rated worse than Hill.
One of those players was Mullens, who was the absolute worst at -2.85. Everyone concedes he's a project, so perhaps it's not such a big surprise to see him down this low. But the Draft Rater is saying that maybe even the middle of the first round is too high to take the risk on him.
Pitt's Sam Young also graded out extremely poorly. He had the worst pure point rating of any wing player, and the other thing that hurt him is that he's one of the oldest prospects in the pool. How old? He's 19 days older than six-year vet
Darko Milicic and a full half-decade older than Holiday.
John Hollinger writes for ESPN Insider. To e-mail him, click here.