How much does the pre-season data tell us about a rookie’s first season? Last year there were 17 rookies who played at least 100 minutes in the pre-season and at least 500 minutes in the regular season. Yes, 17 is a small sample. Still, it’s all we got so let’s tell a story about Kevin Durant and the 2007-08 rookie class.
The 2006-07 Rookies
Table One: Evaluating the 2006-07 Rookies in the Pre-Season and Regular Season
Our story begins with Table One. This table reports the 2006-07 regular season performance of our 17 rookies, as well as what each did in the 2006 pre-season (pre-season data taken from Doug Steele’s web site). At the top of the list are Paul Millsap and Rajon Rondo. Millsap led all rookies in Estimated WP48 (see table for calculation) while Rondo finished second. In the regular season, Rondo led this sample in WP48 while Millsap finished second. So the identity of the top two rookies was pretty consistent from pre-season to regular season.
And then there is the bottom of the list. Adam Morrison was the least productive rookie in the pre-season last year. And in the regular season he maintained this ranking. In fact, three of the four rookies who were in the negative range in WP48 in the pre-season were also in the negative range in the regular season.
Overall, pre-season Estimated WP48 has a 0.70 correlation with regular season WP48. No, it’s not a perfect correlation. And the sample is too small. But pre-season performance seems to be telling us something about rookie performance in the regular season.
The 2007-08 Rookies
The 2007 pre-season is now history. Again we consult the data from Doug Steele. With this data I have calculated Win Score, PAWSmin, and Estimated WP48 for the 23 rookies who logged at least 100 minutes in the pre-season. The results are reported in Table Two.
Table Two: Evaluating the 2007-08 Rookies in the Pre-Season
The analysis reported in Table Two reveals that this rookie class didn’t play well in the 2007 pre-season. Only four rookies posted an above average WP48 in October. Last year, in our sample of 17 (which leaves out the rookies who played 100 minutes in the pre-season but didn’t play 500 minutes in the regular season) there were seven above average rookies. And last year was not a “good” rookie class. So although it didn’t look like we could see less productive rookies, it appears we might in 2007-08.
Although most rookies played badly in October, there was one big exception. Al Horford posted a 0.302 WP48 in the pre-season. Let’s put that number in perspective. If Horford is an average player (i.e. posts a 0.100 WP48), and he plays about 2,300 minutes this year, he will produce about 4.8 wins. If he posts a 0.200 WP48, his Wins Production rises to 9.6. And if he does what he did in the pre-season he will produce 14.4 wins. In sum, if Horford goes from average to what he did in the pre-season, the Hawks win about 10 more games. And that might be good enough to put Atlanta in the playoffs.
But would that be good enough to win Horford Rookie of the Year? No, that honor will probably still fall to Kevin Durant. Durant played very well at Texas last year. Then he played very badly this summer in two rookie leagues. Then he had a good game with Team USA in a team scrimmage. But in the pre-season, where he posted a 0.006 WP48, he was again quite bad. Nevertheless, he did lead all rookies in the pre-season in points per 48 minutes. And scoring is what attracts attention in the NBA. Consequently, if Durant scores, he can expect to be named to the first All-Rookie team by the coaches and he will likely be named Rookie of the Year by the media.
Unfortunately for Seattle, although Durant will score, he won’t produce many wins if he keeps playing like he did in the pre-season. And in the first game of the season, that kind of play did continue. This is what Durant did against Denver tonight.
Points = 18
Rebounds = 5
Steals = 3
Blocked Shots = 1
Assists = 1
Field Goal Attempts = 22
Free Throw Attempts = 4
Turnovers = 2
Personal Fouls = 3
That’s the data. Let’s look at these numbers via Win Score.
Win Score = PTS + REB + STL + ½*BLK + ½*AST – FGA – ½*FTA – TO – ½*PF
Given this formulation, here is Durant’s Win Score against Denver.
Durant’s Win Score = 18 + 5 + 3 +0.5 +0.5 – 22 – 2 – 2 – 1.5 = -0.5
Now what Win Score means depends on the position played. At least, that’s true when Win Score is a positive number. When you see a negative number there is only one meaning for all positions – you played badly.
Yes, this is just one game. But Durant didn’t play well in the pre-season. And he didn’t play well in summer ball. At some point, all these little samples are going to start telling us something.
For now, it doesn’t look like Durant is going to be the most productive rookie in 2007-08. But if he gets to take 22 shots each game, he certainly is going to get a chance to score. And scoring is going to earn you some honors and get you paid.
This all leads to an obvious question – if Durant plays badly this year but is told by the coaches and the media that he’s “great, will he have any incentive to improve in the future?
– DJ
Our research on the NBA was summarized HERE.
The Technical Notes at wagesofwins.com provides more information on the published research behind Wins Produced and Win Score
Wins Produced, Win Score, and PAWSmin are also discussed in the following posts:
Simple Models of Player Performance
What Wins Produced Says and What It Does Not Say
Introducing PAWSmin — and a Defense of Box Score Statistics
Wages of Wins Journal 
Gareth Lewin
November 1, 2007
If I have my math correct:
Iverson = 25 + 5 + 7 + 0 + 7 – 25 – 9 – 7 – 1 = 2
Anthony = 32 + 5 + 1 + 0.5 + 2.5 – 27 -3 – 3 – 1.5 = 6.5
Camby = 4 + 15 + 0 + 2.5 – 5 – 1 – 2 – 1 = 12.5
So Durant was WORSE THAN IVERSON.
And Anthony wasn’t that bad? (Is a 6.5 win score good?)
And Camby (One of my favorite players of all times) was – as usual – very good it seems.
It would be interesting if there was a webpage that got NBA boxscores and let you see winscore, wp48 etc updated automagically.
If you know of some place I could get that data legally and freely, I might setup such a webpage.
Westy
November 1, 2007
You note that so far the ’07-’08 rookies don’t look good. This is pretty surprising given the hype this class received, and considering how bad last year’s class was. In your analysis of the draft prospects you (or Erich Doerr) also noted that it looked like a good class based on their college statistics. Why are we not seeing the production expected?
And specifically, you noted at that time that, indeed, Durant played very well in college. Based on that a good pro career was anticipated. So far, though, we’re not seeing it. Looking back at those calculations based on college, what was incorrect about our expectations? Has there ever been anyone as good as Durant (or at least close) in college who ended up a bad pro?
Owen
November 1, 2007
Gareth – I think those winscores should be
Anthony – 10.5
Iverson – 9.5
I think it would be awesome to have a live updated win score feed. You could actually do it for a fantasy league on CBS Sportline I think…
Re Durant, he looked terrible. He needs to fill out. Not ready for the NBA….
Owen
November 1, 2007
P.S. You subtracted his points rather than his shot attempts, and you also have to divide fts by 2 before you subtract. It works best if you look at the box score and read right to left I find.
Chip Crain
November 1, 2007
What? No Michael Conley stats? I realize he didn’t play that great but I can’t believe the #4 pick in the draft is already being ignored when ranking rookies this year.
Chip Crain
November 1, 2007
For that matter where is Juan Carlos Navarro as well?
dberri
November 1, 2007
Chip,
Steele’s stats say Conley didn’t play 100 minutes in the pre-season. In 96 minute he has a 0.141 Win Score per-minute, which is pretty good for a point guard. And Navarro had a -0.044 Win Score per-minutes in the pre-season, which is quite bad.
About Win Score data… Jason Chandler says his Win Score/Wins Produced website will be up and running in a few weeks. His website was updated last year.
dberri
November 1, 2007
Westy,
I would have to look at the data to be sure, but I don’t think there has been a player who played as well as Durant did in college in the past twelve years who did not at least make to average in the NBA.
Someone did note in the comments last summer that Durant might be like Carmelo Anthony and Jerry Stackhouse. Both Anthony and Stackhouse were good college players who have struggled in the NBA. Durant was better than both in college, but like these two, he is struggling.
The good news for Seattle fans is that our samples so far are very tiny. Durant can certainly play better. And I expect that will happen. But so far, he has played pretty poorly.
Ryan Schwan
November 1, 2007
Excuse me while I go giggle over the Hornets landing Julian Wright 13th in the draft. Then I’ll call up Byron Scott and tell him to play the rookie, since he didn’t play him last night.
Was the PAWSmin based on last years averages or averages in the preseason? I’m hoping you say last years averages.
dberri
November 1, 2007
Ryan,
PAWSmin are for the pre-season.
Gareth Lewin
November 1, 2007
Oh I missed this
“About Win Score data… Jason Chandler says his Win Score/Wins Produced website will be up and running in a few weeks. His website was updated last year.”
Yay, I like it when stuff I want to do has already been done.
Gareth Lewin
November 1, 2007
(Reposting this, since it seems to have not gotten posted)
Owen,
I think I did divide by 2 before I subtracted, and I didn’t add either of their points.
But I did get them both wrong. It was really late hehe. For Iverson I forgot to /2 his FTA and he should end up as 6.5
For Anthony I seem to have taken his 3ptA instead of his FTA and should end up 4.5
Still we came to completely different values. This is strange.
I assume FGA includes 3 point attempts? If not then Iverson would be a 9.5 and Anthony would be a 10.5 which lines up with your values. But I’m pretty sure win score takes 3 point attempts into account?
dberri
November 1, 2007
FGA includes 3 point attempts. I ran the numbers and I get what Owen got — 10.5 for Anthony and 9.5 for Iverson. Camby is at 14. All are above average for their position. In fact, and this is not surprising given that they won going away, almost every Denver player was above average last night (except for Nene).
This could mean a) Denver is great b) Seattle is bad c) its only one game and doesn’t mean anything. All are somewhat true, although I lean to (b) and (c).
Gareth Lewin
November 1, 2007
Ok, I don’t get how I am getting the numbers wrong.
Here is a dumb question, according to http://www.nba.com/games/20071031/SEADEN/boxscore.html
Antony had fgm-a 10-21 and 3pm-a 3-6
I assume that meant 21 2 point attempts and 6 3 point attempts. Maybe I am reading box scores wrong and the FGM-A includes 3 pointers?
If so then that is why my math was off.
Jason
November 1, 2007
10-21 includes both 2 and 3 point baskets.
Anthony was 7 for 15 on two pointers.
Pete
November 1, 2007
I believe I’m remembering correctly that Win Score correlates well with Wins Produced, but does not adjust for certain factors. What major factors are missing? Position played? Pace of game? Other?
That should mean that one can probably compare a players performance to previous performances, and perhaps compare two players playing the same position to each other if they played in the same game(s). I’ve been playing with Win Score / 48 a bit in this way…
Is there a simple linear adjustment that one can make to compare performance between positions? What about pace (if that’s indeed taken into account in Wins Produced)? Can one get a rough estimate of pace by looking at total number of shots taken in a game?
Phil B
November 1, 2007
Hi, so I’m kinda new to this, but I’m looking at your formula and you subtract for personal fouls. Why is that? A personal foul is not always a bad thing. There are lots of times when you want to foul an opposing player, lots of times when your coach will tell you to foul. Why do you subtract this from the score? That being said, I do agree that as of now, I don’t see Durant as being all that good for his team. Give him time though and we’ll see what he can do. Of course, I don’t need numbers to tell me that, I just watch the games.
Pete
November 2, 2007
“A personal foul is not always a bad thing.”
Actually, a personal foul is always a bad thing, because it always gives the opposing team a free attempt to score. It’s just that sometimes the alternative is worse (ie the clock runs down, an easy basket is made, etc).
Jason
November 2, 2007
Fouls are substracted because empirically, more fouls means a lower probability of victory.
The values of the stats are determined from regression in a large sample based on the rather unsurprising (yet still sometimes ignored) notion that being more efficient results in more wins. In large sample, it works. In any particular game, a certain statistical datum may not factor in, but on average, these garbage time stats aren’t where most stats are accumulated and across time, they seem to balance pretty well. That late foul or rebound when the game is decided has no bearing on the final outcome.
It may be that a foul is strategically the proper thing to do to prevent something more certain but this doesn’t really dispute the fact that fouls are going to lower probability of victory. If you are fouling to prevent the clock from running out, you’re in a situation where the probability of victory is pretty low, so this correlation holds on logical principles. If you’re fouling to prevent the easy bucket, it’s probably a valid correlation as well as the foul is a marker of a situation (probably an easy bucket–a worse outcome on average) that is bad and doesn’t favor winning either.
Phil B
November 2, 2007
Personal fouls do not always lower the probability of victory. Let’s use today’s post subject, Mr. Shane Battier (who I hate cause he’s a Duke player, but that’s beyond the point, he’s obviously a good player). Let’s look at the game vs. the Lakers, you could say that his foul on Kobe at the end of the game won the game, because Kobe at the end of games is quite deadly and may have hit the three. He didn’t, because Battier fouled him. That foul (along with his three pointer) won the game and the foul was subtracted from his win score? No thanks, don’t sign me up for this cult.
Rashad
November 2, 2007
Phil B, any statistical model can only look at things on average. Something like 95% of fouls throughout a game are going to hurt your team by giving the other team possession if your team had possession (offensive foul) or by giving your opponent free throws, or using up fouls leading to later fouls causing free throws.
While the occasional foul helps your team, you have to calculate for the average foul, and the average foul definitely hurts your team.
Rashad
November 2, 2007
Phil B, any statistical model can only look at things on average. Something like 95% of fouls throughout a game are going to hurt your team by giving the other team possession if your team had possession (offensive foul) or by giving your opponent free throws, or using up fouls leading to later fouls causing free throws.
While the occasional foul helps your team, you have to calculate for the average foul, and the average foul definitely hurts your team.
Phil B
November 8, 2007
It’s a broken formula. End of story.
Tom Mandel
November 13, 2007
Yeah it’s broken — that’s why it correlates so well with reality, because it’s broken. We all know that the right team doesn’t really win. ;>
A couple of surprises this early season: AI is off to what looks like a great start, and the Wizards aren’t playing Dominic McGuire (made a few rookie errors early). We’re 1 and 5, another confirmation that the right team doesn’t win. It’s the world that’s broken, not the formula.
Where’s this website w/ a winscore feed going to live — I *must have* the URL; I must!
Tom Mandel
November 13, 2007
Oleksiy Pecherov was chosen in the ’06 draft but is a rookie this year and played in this preseason. I’d like to see his numbers, where he stands in this analysis.