On Friday I compared the performance of the top pick in the NBA draft to the performance of the most productive player selected in each draft since 1991. Or as I put it – I compared The Top Pick vs. The Top Pick.
In this post was the following table, which detailed the career performance of each number one pick since 1991.
Table One: The Career Performance of the Number One Picks (1991-2006)
Now back in May I wrote a column examining the value of having the number one pick in the draft. In that post was a table that reported the productivity of each number one choice his rookie season (since 1992).
Table Two: Performance of Number One Picks – 1992 to 2007
Today I want to put these two tables together. Specifically I want to compare what the number one picks did their rookie season to what they did the rest of their career. For this we turn to Table Three.
Table Three: Evaluating the Consistencey of the Number One Picks – 1991-2005
Table Three compares the productivity of each number one pick his rookie season– measured via WP48 (Wins Produced per 48 minutes) – to what that player did after his first year.
Across the past 16 seasons we see that the number one pick posted a WP48 above 0.100, or above average, nine times. Of these nine players, only Joe Smith failed to be above average the rest of his career. Five players posted a WP48 in excess of 0.200 their rookie season and only Chris Webber – whose career numbers have declined since his injury in 2003 – failed to post a career WP48 in excess of 0.200. In sum, virtually all the “good” rookies have been “good”. And the “great” rookies were generally “great.”
At the other end of the table are many of the top picks who did not have stellar careers. Seven players have been below average their rookie season, but Andrea Bargnani has yet to have a career beyond his first season so he is excluded from the analysis. Such analysis suggests, though, that Bargnani might not be worthy of his number one selection. When we look at the career performance of the six below average players we see five players who went on to post below average career numbers. The lone exception, LeBron James, was virtually average his rookie season and went on to be extremely productive.
Ideally we would do more than just stare at the numbers. Unfortunately our sample is only 15 players, which is a bit small for us to do any real analysis statistically. Still I would note that the correlation coefficient between First Year WP48 and After First Year WP48 is 0.79.
These numbers do “suggest” – and I want to emphasize the word “suggest” – that we might be able to tell a great deal about Greg Oden by the end of the 2007-08 season. If Oden fails to deliver an above average performance his rookie season, it would be fighting a trend for him to suddenly deliver later on. And as noted above, the same story can be told about Bargnani.
Again the word is “suggest.” A sample this small cannot provide us “proof” or the “truth” (not that we are in the “truth” business).
Looking at a Larger Sample
What would be interesting to see is the predicitive power of the first year performance for all players. And I wasn’t planning on looking at this today, but looking at the data I have collected I decided this would only take a few minutes.
So after a few minutes of analysis I found a 0.69 correlation between first year WP48 and WP48 after the first season (for all drafted players since 1991 who played at least 800 minutes his rookie season). In sum, if all we know about a player is his rookie year per-minute performance, we know a great deal about what that player will do the rest of his career.
In other words, even if we don’t know about a player’s injury status, his role on his team, his coaching, his attitude, his shoe deal, etc… we still know much about a player’s career performance if we simply know what he offered his first few games as an NBA player.
Let me put this in perspective. The correlation coefficient between a baseball batter’s OPS this season and last season is less than 0.60. In football the correlation between this season and last season’s performance for a quarterback is less than 0.4o. So we are seeing more consistency between what an NBA player does his rookie season and the rest of his career than we see from year-to-year in baseball or football.
The consistencey of NBA performance is one of the more important stories the statistical analysis of the NBA reveals. Relative to baseball and football, what you see is much closer to what you are going to get in the NBA. This consistency may be bad news for Bargnani fans, but it should be good news for decision-makers in professional basketball.
– DJ
Wages of Wins Journal 
Jason
July 29, 2007
It’s curious that there’s still ‘conventional wisdom’ that ‘big men take longer to develop.’ I’m curious if there’s any WP support for this. From what I’ve seen, it looks like guards and forwards may have a learning curve, but it appears that big men don’t take longer to develop so much as they *never* develop if they weren’t any good to begin with.
Owen
July 29, 2007
You wonder too, what difference chronological age makes. I can’t believe Duncan would have been in the .300’s without three or four years of college. More on that later in the summer I guess…
Ap
July 30, 2007
Wow, more tremendous analysis. I’ve been reading your blog for a couple months now, and while your analysis has been nothing but insightful, I do wonder one thing. What are the examples where Wins Produced fails, and fails miserably. As in, has there been a team comprised of a great number of players whose wins produced were high and didn’t win (like your Garnett vs. Duncan example). The stat, is, by its basis, supposed to be a competent measure of the amount of wins produced. So if you look at a teams total wins produced then it should get close to that teams actual wins. Does this happen? And if you’ve done this analysis plenty a time before, could you or one of your other fine readers link me to it perhaps?
Ap
July 30, 2007
NVM…I’ve done my own digging and seen that, of course, your model works to predict real wins. Otherwise it wouldn’t be a model! Silly me, I’m new, but I’m learning.
McCoy
July 30, 2007
I think Owen has a valid point. Controlling for age, in this era of high school and college freshman pickes, would seem to be important (see Lebron James large improvement above).
Ben
July 30, 2007
But is there a difference in the correlation between #1 draft picks depending on how old they were and the level of competition when they entered the league? I would assume the correlation would go up for players who were older before they came into league? For instance, Kobe Bryant or Tracy McGrady’s correlation’s for their first year wouldn’t have been as high as Shaquille O’Neal’s?
As for the big men take longer to develop theory, as well as point guards. There is much qualitative evidence in the careers of Brad Miller, Mehmet Okur, Chauncey Billups, Gary Payton, John Stockton, Steve Nash, etc… Of the notable swingmen i can remember, most were pretty darn good off the bat. The same has been true of all-time great big men. Above average bigmen though, i think it might be true.
Tom Mandel
August 1, 2007
Re: your final point — some rookies have come straight out of high school; others have played 4 years of college. It would be interesting to see whether and how it makes a difference how old you are, and how much organize basketball experience you’ve had, when you come into the league. Now, at least in theory, younger players are able to get in young, because they are more gifted. This could be tested for too.
Douglas
August 5, 2007
I’m afraid I don’t think the statistics here imply what is thought, at least necessarily. For instance, the high correlation (0.69) might more be a factor of where and with whom they play, rather than their own abilities or potential. If a really good player happens to land on a really bad team, then immediately is called upon to play lots of minutes, his “WPA” will suffer greatly in comparison to a relatively bad player who happens to land on a really good team, since in the latter situation the bad player likely will not get many minutes, but the team will garner many wins. Perhaps I’m not completely understanding how the tables and such were derived.
Douglas J. Bender
(Elkhart, IN)
Douglas
August 5, 2007
Make that “WP48”, not “WPA”.
autontart
October 7, 2008
Подскажите шооблончег под WordPress 2.6.2, чтобы был похож на ваш dberri.wordpress.com.
Заранее благодарю)