Let me start this column by repeating something I said just a few days ago:
Last summer the sportswriters named Monta Ellis as the Most Improved Player of 2006-07. In response to this selection, I wrote the following column:
The title pretty much summarizes the argument. If you look at the numbers – and I mean at more than points scored per game – it would be hard to argue that Ellis was the most improved player last year. Of the seven candidates I examined last summer, Ellis was the least improved.
As I noted less than two weeks ago, Ellis improved tremendously in 2007-08. In fact, he improved so much that if Wins Produced was your metric of choice (and of course, it should be), Ellis would be the Most Improved Player in 2007-08. In other words, the sportswriters were right last summer, they were just 12 months too early.
Finding the Most Improved
The Monta Ellis conclusion is based on the following analysis. First I looked at players who played at least 1,000 minutes in both 2006-07 and 2007-08. Then I compared the Wins Produced a player posted in 2007-08 to what his Wins Produced would be this season if he maintained his 2006-07 WP48 [Wins Produced per 48 minutes]. Following these steps reveals that Monta Ellis – with a WP48 that increased from 0.043 to 0.177 – posted the largest increase in Wins Produced. In all, Ellis produced 8.6 more wins than his WP48 of 2006-07 would suggest.
As Table One reveals, Monta Ellis just edged out Chris Paul for this award. Hedo Turkoglu – the player the media chose – ranks fourth, just behind Chris Kaman (who I would have guessed was the winner before I did my analysis). Rudy Gay – who finished second in the media’s voting – ranks 20th. Yes, Gay did improve, just not as much as a number of other players (including Kaman).
The media’s vote indicated that LaMarcus Aldridge was the third most improved player. But if we look at the 25 most improved, Aldridge’s name is nowhere to be found.
We do find Aldridge’s name, though, in Table Two.
Table Two reports the players who changed the least from 2006-07 to 2007-08. The 20th player on this list is Aldridge. Yes, a player who members of the media think is the most improved really hasn’t changed much at all.
At least, not in the aggregate. There were some changes in Aldridge’s performance if we look at the individual stats.
As Table Three indicates, Aldridge did manage to take more shots this season. And despite a lower level of shooting efficiency, he was able to score more points. Of course, if you score less efficiently, you are not helping as much as you could.
In addition to scoring, Aldridge also posted an increase in assists and personal fouls. But he rebounded slightly less and committed a few more turnovers. The net effect was a very small change in Win Score per minute, and as noted, WP48.
Of course the media doesn’t look at Win Score or Wins Produced. In fact, they don’t appear to look at shooting efficiency or per-minute stats. What they did see is that Aldridge increased his per-game scoring from 9.0 to 17.8. And apparently, that was enough to get Aldridge some votes.
The Least Improved
At least, Aldridge did improve a bit. A few players didn’t quite fare this well. Table Four reports the least improved players – or the players who posted the biggest declines – in 2007-08.
Topping the list is Rashard Lewis, who was primarily done in by the switch from small to power forward. The next two names on the list – Quentin Richardson and Dwyane Wade – were done in by injuries.
Assuming Wade can recover from injury, he’s the early favorite for Most Improved in 2008-09. At least, the early favorite if Wins Produced is the metric of choice. Given the media’s fixation on per-game scoring, Wade is going to have a problem winning this award. Wade averaged 24.6 points per game in 2007-08. Unless he can average well over 30 this next season, some members of the media might not see any improvement next season (regardless of what Wins Produced might say).
Our research on the NBA was summarized HERE.
Wins Produced, Win Score, and PAWSmin are also discussed in the following posts:
Finally, A Guide to Evaluating Models contains useful hints on how to interpret and evaluate statistical models.