Dre at Nerd Numbers had a post the other day about the MVP race where he mentions that Steve Nash is one of only four players to win the award while his team won less than 55 games (in 2006). That was the year Amar’e blew out his knee after playing 50 minutes of the season, yet the Suns won 54 games and were second in the West (they went on to lose to Dallas in the conference finals the year that Dwayne Wade free throw-ed his way to a ring). Dre also casually mentions that Steve got a lot of credit for Shawn Marion’s play that year. This reminded me of a project I did probably a couple years ago. Despite the typical Wins Produced view that interactions are a small part of basketball, I was convinced that Steve makes his teammates better. After all, he is pretty awesome.
At the time I think all I used was field goal percentage, which I figured out at a painfully slow rate from ESPN. I divided things up into three pointers and two pointers, because I thought Steve could probably pass to open shooters but maybe they wouldn’t make it whereas he could definitely get guys lay-ups and dunks. I don’t remember all the details, but I think I just used players on Phoenix that season who had also played on other teams in the past and a couple of notable departures (Marion and Joe Johnson), and I don’t think it was significant at all, but players did numerically shoot better when playing with Nash. I emailed Prof. Berri to ask if he knew about anything showing that some players make their teammates better, and he was kind enough to reply back. I don’t remember what all he said besides that there was evidence that more assists lead to more points (that sounds obvious, so I assume he meant when controlling for some other things).
But, now that Dre had jogged my memory and I had access to his automated Wins Produced numbers, I thought I would try again. The numbers only go back to 2001, so Steve has already been in the league for four years (2000-2001 is his fifth). There are four seasons in Dallas followed by six in Phoenix (I’m not including the current season since there have been so few games), and in those ten years Steve has played 26,523 minutes and generated 140.7 wins for a career(ish) WP48 of .255 (basketball-reference.com only gives him credit for about 102 in that period, presumably because they hate Canada). In that time frame, he is estimated to have assisted on 43% of his teammates’ baskets, which is a ton. His career mark of 40.3% is behind only Stockton, Chris Paul, Deron Williams, and Magic.
What I did was get all of the Phoenix and Dallas rosters from the automated site for 2001 to 2010. I cut out anyone that played under 500 minutes or was an obvious back-up that wouldn’t have played much with Steve; obviously people who aren’t on the court can’t benefit from his play. I also cut out anyone who had only played with Nash (or never with Nash for some Phoenix players in the early 2000s and Dallas players in the late 2000s) so that I only have players who had both played and not played with Steve. That way each player can serve as his own control. Cool, huh? I also dropped one or two people if they were at the end of their career and limited minutes/huge drops in productivity would potentially skew the results. A full list of the players is at the end of the post if you want to see them; there are 39 in total. If you think I should add or drop anyone for a good reason, let me know.
I got each player’s data from the automated site. For any year they were traded within the season, I just averaged it together (thanks a lot, Tim Thomas) unless part of the season was with Nash and part without. Then I added up their minutes and wins produced separately for with-Nash and without-Nash seasons and calculated WP48. What I wanted to see: if players had a higher WP48 when playing with Nash than without. Here come numbers!
The 39 players had 370,789 minutes played without Nash and generated 814.4 wins for a .0925 WP48. They also had 145,495 minutes played with Nash and generated 399.3 wins for a .1048 WP48. Thus, players were .0122 WP48 better with Nash. A paired t-test says this has a p value of .07, so it isn’t quite significant by typical standards. That doesn’t seem like a big difference, but think of all the minutes they played without Nash. I took each player’s non-Nash minutes and figured out how many wins they would have produced in that time if they played at their with-Nash WP48. Remember that in real life they produced 814.4 wins; if they had played with Nash they project to have produced 920.6 wins, or an extra 13%. That sounds like a fair amount, right? 13% turns a 50 win team into a 56 win team.
Let’s look at some of the players to get a sense of the close/mixed result. The five biggest declines in WP48 belong to Tim Hardaway, Shaq, Laettner, Tony Delk, and Tim Thomas. Shaq put in the most time with Nash at a season and a half (actually 3000 minutes played), and all of these guys were at the end of their careers except for Laettner. The five biggest gainers are Marquis Daniels, Louis Amundson, Amar’e Stoudemire, Greg Buckner, and Shawn Marion. Daniels has not played well since leaving Dallas, and the same is true for Buckner; Amundson was a rookie before playing in Phoenix so improvement might be expected. As for Amar’e and Marion, oh the schadenfreude. The story is that they are both finishers who benefited from Nash getting them the ball but they couldn’t create for themselves, and that seems to show up in their WP48. It’s also true that Amar’e’s only two years without Nash were his first two, but he isn’t exactly lighting up the league this season so far. Marion had three years with Stephon Marbury and one with Jason Kidd before Nash, then Kidd again last year and a Dwayne Wade/Mario Chalmers/Jose Calderson combo in 2009.
I think it’s also fun to look at who would have benefited most (or least) from Nash in terms of total wins produced. The lowest name on the list is Shaq, who produced 99 wins while not playing with Nash since 2000-2001; at his with-Nash rate he would have only produced 67 wins. Of course, the non-Nash time includes his Lakers heyday. The biggest beneficiary would have been Antawn Jamison, who would have produced 96 wins instead of 60, followed by Shawn Marion.
Moving a bit toward subjective decisions, I could note that taking Shaq out of the sample is enough to make the t-test significant. If you counter by taking out Jamison or Daniels, the p value moves back to around .05. I think taking Shaq out is more defensible given the timing of when he played with Nash, but it’s easy to slip into rationalization. So I’ll end by saying that while the WP48 numbers are somewhat unclear, I think Nash makes his teammates better. Given that Wins Produced adds together a variety of measures, it’s a tall order to expect one teammate to improve another’s productivity; there’s no reason Nash should impact rebounding, free throws, steals, or blocks. I would wager that he improves his teammates’ shooting accuracy (maybe someone with more detailed player data on hand could take a look?), and maybe reduces turnovers just by handling the ball so much (or potentially getting them the ball in better positions), but he might also cut into their assists for the same reason. So Nash should only influence maybe two or three pieces that go into WP48, but we’re still seeing a bit of an improvement. At any rate, I would definitely take this trade in a second.
(Did you look at that trade? Sending Ben Gordon to the Suns costs them one win, and giving Nash to the Pistons costs them four wins? Everyone loses and the Pistons come out further behind? Sweet fancy Moses).
Players in the sample: Adrian Griffin, Amar’e Stoudemire, Antawn Jamison, Antione Walker, Boris Diaw, Brian Skinner, Channing Frye, Christian Laettner, Danny Fortson, Dirk Nowitzki, Eduardo Najera, Grant Hill, Greg Buckner, Howard Eisley, Hubert Davis, James Jones, Jared Dudley, Jason Richardson, Jim Jackson, Joe Johnson, Josh Howard, Juwan Howard, Kurt Thomas, Leandro Barbosa, Louis Amundson, Marquis Daniels, Matt Barnes, Michael Finley, Nick Van Exel, Quentin Richardson, Raef LaFrentz, Raja Bell, Shaq, Shawn Marion, Steven Hunter, Tim Hardaway, Tim Thomas, Tony Delk, Walt Williams.