Most post-season awards in the NBA are determined by the media. It’s the sportswriters we turn to when we want to know who is Most Valuable, Rookie of the Year, All-NBA, Best 6th Man, and Most Improved. Okay, we don’t always learn who these players actually are, but we do learn about the opinions of the sportswriters (which has to be worth something).
Back in May I looked at the top rookies and concluded that in terms of Wins Produced, Rajon Rondo just edged out the media’s choice of Brandon Roy. Still if you really thought Roy was ROY, I wouldn’t argue too much.
Today I want to build upon yesterday’s post – The Playing Time Illusion – and look at the award for Most Improved Player (MIP).
The Media’s Selection
From Patricia Bender’s wonderful website we see that 129 members of the media voted for this award. Each voter selected three players. The player ranked as most improved received five points, with the second and third choice receiving three and one point respectively. When the votes were tallied, the MIP was Monta Ellis.
When we look at the votes we see that 36% of voters thought Ellis was the MIP, while 34% chose Kevin Martin. In all, 29 players received some consideration for this award, but most members of the media seemed to think that the MIP was either Ellis or Martin.
Not Ellis or Martin
Okay, one could argue that Roy was ROY. But Ellis for MIP? Not sure how to defend this choice. In a moment I will get to the numbers, but for now I will say that almost all these numbers disagree with the choice of Ellis.
Kelly Dwyer of Sports Illustrated indicated that Martin was clearly the MIP. Last March I responded to this claim by noting that although Martin was more productive in 2006-07, this increase in production was tied almost entirely to Martin playing more minutes and taking more shots per minute played. Martin’s shooting efficiency was unchanged from 2005-06, and his performance with respect to the non-scoring aspects of the game were also unchanged.
So Who is MIP?
Okay if it’s not Ellis or Martin, who should be the choice? To answer this question requires some kind of criteria. Perhaps the easiest criterion is to simply look at how Wins Produced changed from season to season. But that approach can suffer from the “Playing Time Illusion.” A player could simply see his Wins Production increase because he played more minutes.
An alternative approach is to look at changes in production per-minute. Or one could compare how many wins we saw from a player to how many wins we could have expected from the player had his per-minute production from the previous season remained unchanged. In other words, we can look at changes in Wins Produced per 48 minutes [WP48], or we can compare actual wins to expected wins. One would think these approaches would give the same answer, but it turns out they don’t.
Let me illustrate by looking at the seven players who were named on at least 10% of the ballots for MIP.
Table One: The Top Seven in MIP Voting
When we compare actual wins to expected wins for these seven players, the choice is Deron Williams. When we look at changes in WP48, the choice is Al Jefferson. Before I get to these players, I want to comment again on Ellis.
Monta Ellis averaged 6.8 points per game in 2005-06. Last season he averaged 16.5, for an increase of 9.7 points per game. If this was your only consideration, and these seven players were your only candidates, then Ellis would be MIP. But when we consider playing time, we see that most of the increased scoring we see from Ellis was simply a result of him appearing on the court more frequently. In Table One we compare what Ellis did on a per-game basis in 2006-07 to what we would have seen from Ellis in these 06-07 minutes with his per-minute production from 05-06. When we take this approach we see that the Warriors should have expected Ellis to score 12.9 points per game in 06-07 (given his per-minute scoring mark from 05-06). Yes, Ellis did improve with respect to scoring (and scoring efficiency) but the improvement is quite a bit less when you control for playing time.
When we look at other aspects of his performance, though, we see that Ellis actually declined in 06-07. His rebounds were down and his turnovers were up. Overall his WP48 did improve from 0.027 to 0.043, but his improvement was the smallest leap among the seven players examined.
So the choice of Ellis appears to be motivated entirely by changes in his points production per game. And this change in scoring per game is mostly due to changes in playing time. So not only are sportswriters failing to consider most of a player does on the court, they are making a mistake with the one factor they are considering.
Okay, which of the other players would have made a better choice? As noted, when we compare actual wins to expected wins, the choice is Deron Williams. A few weeks ago I noted that Chris Paul is still a more productive player than D. Williams. When we look at Paul and D. Williams as rookies it’s hard to argue with the contention that Paul was the better player. D. Williams, though, improved enough his sophomore season to make this more of a debate. Had D. Williams maintained his per-minute performance from 2005-06, he would have only 1.2 wins in 06-07. The 9.2 wins he actually produced results in an eight win improvement, or a better mark than any other player listed (Paul produced 13.2 wins last year, so he is still the more productive player).
When we look at changes in WP48, though, we see a different name top the list. Al Jefferson’s WP48 increased by 0.139 from 05-06 to 06-07. Only two sportswriters voted for Jefferson as MIP, but the numbers suggest his candidacy should have received more support.
Moving Past the Top Seven
If we move past the top seven we see a few other players who should have received more support. David Lee saw his WP48 increase from 0.197 to 0.378. His increase of 0.181 exceeded almost everyone who received votes. The exceptions are Andrew Bynum, Amare Stoudemire, and Nene. These three players, though, really didn’t play much in 05-06, so it’s hard to say they actually “improved.”
Three additional players – Emeka Okafor, Zach Randolph, and Matt Barnes — were also strong candidates. Surprisingly none of these players received any consideration at all. Let me conclude with a quick word on each.
The improvement in Okafor was examined last January. Okafor saw his WP48 increase by 0.188. The difference between his actual wins and expected wins was 9.1, a mark that bests D. Williams. Okafor, though, was injured in 2005-06. So he might not have improved as much as he simply recovered from an injury. One should add that Okafor probably had trouble earning votes because he played for the Charlotte Bobcats. The Bobcats did improve by seven wins, but this improvement was muted by the play of Adam Morrison. Had the Bobcats not employed Morrison the team would have likely made more progress in the standings and Okafor would have received more votes for this award.
Another player who should have received some consideration is Zach Randolph. Randolph’s wins production (comparing actual to expected) increased by 7.5. His WP48 also rose by 0.149. These numbers are not quite what Okafor did, but still it’s surprising Randolph received no consideration for this award at all.
Finally there is the play of Barnes, a teammate of the player chosen MIP. The WP48 of Barnes increased by 0.183, which is quite close to the leap we saw for Okafor. His actual wins production also exceeded his expected by 6.9. Unfortunately Barnes failed to average double figures in points in 2006-07, a fact that probably doomed his candidacy.
So who is MIP? I think Okafor might be the best choice, but one could make an argument for D. Williams or Lee. Although the numbers do not give us a clear choice, they do tell us quite clearly that it’s not Ellis. Well, it isn’t if you consider something beyond points scored per game.
One last note on Barnes
Let me close with one last note on Barnes. Barnes is a free agent, and as of last Friday was having trouble landing an offer. According to the Sacramento Bee:
Wednesday was the first day teams could sign free agents, and Barnes said he has not received the offers he anticipated.
“Right now, there’s nothing happening,” he said. “I was thinking the midlevel exception was realistic, but my agent has been telling me teams have questioned whether the success I had was because of (Warriors coach Don) Nelson’s system and whether or not I could have the same success in another system.
“It’s pretty frustrating because I thought I had a good year,” said Barnes, who averaged 9.8 points, 4.6 rebounds and 2.1 assists and shot 43.8 percent from the field, including 36.6 percent from three-point range, in 23.9 minutes per game during the regular season. “It was not a great season, but it was good. And from what other players have received, I thought I’d be right there in that mix. But Golden State has not come around to what I thought they might, so we’ll just have to wait and see what happens.”
One suspects that if Monta Ellis were a free agent he would command more than a mid-level exception. Yet Barnes, who is easly more productive, is not even getting that much. Which returns us to a common story-line in The Wages of Wins and this forum. To maximize the wages of wins you must score. Players like Barnes who try and produce wins without scoring might help their teams, but they do not help their bank account. I think his accountant would appreciate if Barnes could learn this lesson as his career progresses.
– DJ
Wages of Wins Journal 
Sam Cohen
July 16, 2007
I think your commentary on Kevin Martin misses something important, but I’m not sure if there’s a way for statistical measures to capture it well. Namely, as Kevin Martin began becoming a more important part of the King’s offense, opposing teams began to pay more and more defensive attention to him. Therefore, the fact that he could continue to score with the same shooting efficiency IS most likely an indication that he improved substantially. For an offensive player that does not create his own shot (like a Mikki Moore), I completely buy the “more minutes as explanation for supposed improvement,” but I think it doesn’t quite fit when a player moves from being a back-up offensive option to being a primary offensive option. I’d be curious to hear your thoughts on this notion.
Dan
July 16, 2007
Tremendous post.
howlingfantods
July 16, 2007
Zach Randolph couldn’t really receive consideration for the MIP, seeing as he’s already received it once before in 2004. Giving it to him again would basically be like admitting that they screwed up the first time.
I get your argument, but I think Deron is a poor candidate — if you gave the MIP to top 5 lottery picks who improve between their first year to their second year, the award would be pretty meaningless — all young players have a pretty big jump their second and third years, and # 3 picks are supposed to be good.
Okafor is coming back from injury, which this award is not supposed to be about. Matt Barnes, like Ime Udoka (another guy worth mentioning) basically got their first shot at significant playing time.
I would’ve been tempted to vote Lee for almost everything (sixth man, MIP) except it’s always a bit of a bummer to vote for someone in street clothes for most of the second half of a season.
I guess I’d go for Big Al just sort of by eliminating all others. He improved a lot in terms of efficiency stats plus got the all important ppg increase that the scribbler classes love to list in their news releases.
Jed
July 16, 2007
Hi DJ,
A minor question regarding your playing time hypothesis. I’d be curious to see if there’s any evidence that we can treat production per unit time as a linear function. That is, if I double player X’s playing time from 15 min/game to 30 min/game, should I expect to see all of his stats (scoring, rebounding, etc.) about double?
It seems to me that we can’t make this assumption, at least for relatively large increases (in absolute terms) . Although I’m no expert, it seems like it’s quite a different matter to sustain a given level of play for 40 minutes versus 10 or 20, especially night in and night out. Players get tired or injured, defenses lock down, etc. and surely this must have some kind of effect on production.
If this is true, then perhaps it might really be worth it to pay a premium for the guy who can produce at level T for 40 minutes. Of course, this wouldn’t explain why a guy like Mikki Moore gets the MLE, since according to your metric he’s below average for all playing times, but maybe it makes front offices look slightly less irrational.
Anyhow, looking forward to your reply, and love the site.
Jed
July 16, 2007
I just noticed that I inadvertently echoed a lot of what Sam said in his post–apologies for that.
Jason
July 17, 2007
From the numbers I’ve played with, some stats scale well, others not as well. Rebound rates don’t seem to change as much as scoring rates with changes in playing time. A 10pt, 5 rebound player who sees his time doubles is likely to grab about 10 rebounds, but his scoring doesn’t appear so easily predicted based on points per minute.
Dave probably has better data on this.
One anecdotal comment: it does seem appropriate to credit Biedrins as being greatly improved even if most of his increased productivity is a matter of playing time. His playing time in previous seasons was limited largely by his inability to stay out of foul trouble. Being able to stay on the court was something he needed to improve and did. It doesn’t show up in wp48 as much as it does in total wins but in a real game, the penalty for the 6th foul is greater than the penalty for any of the previous 5. WP does not make this distinction. For most players it may not be an issue. For Andris, it certainly was.
MT
July 17, 2007
Agree with Dan – tremendous post. Excellent comments as well.
dberri
July 17, 2007
A few quick comments…
First, I hate how the comments look on our page now. I can’t read these very easily as I write. Not sure what to do about that, though.
Okay, some responses.
Sam: We have not been able to find a link between shot attempts and shooting efficiency. So I remain unconvinced that a player generally shoots worse as he shoots more. But even if that was true, how much more effective does this make Martin? Players like Lee, Okafor, DWilliams, and Barnes improved tremendously. Not sure we can look at Martin in such a way to close that gap.
About the minutes and production link that Jed also noted… I have found that minutes and production are positively linked. As a player plays more, his per-minute numbers go up. There is a causation issue here. It could be the player gets to play more when he is perceived to be better by the coach. So the direction of the relationship is unclear to me. I would add that the impact is not very large. It is not the case that changing minutes has a huge impact on over-all performance.
I will note that Mikki Moore has played semi-regular minutes before in his career, and his past numbers do not suggest he played much better for the Nets last season. In fact, back in 2000-01 he played 1,154 minutes in 81 games for the Pistons and posted a 0.087 WP48. So relative to his 2000-01 campaign, he was not as good last year. I should add that in 2004-05 he played 1,178 minutes for the Clippers and posted a 0.001 WP48. So relative to that season he was better for the Nets. Regardless, I still would argue that relative to his career, he was not much better for the Nets last season and it seems hard to argue that a player who was only worth a 10-day contract three years ago is now worth $18 million.
howlingfantods: Neat comment. I did not remember about Zach Randolph winning the award in the past.
Jason: Great comment as always. Not sure how to incorporate a propensity to foul out in the analysis. In a sense it is there, since if you foul out your productivity goes to zero.