Two Perspectives on the Top 10 across the Past Decade

Posted on September 24, 2009 by

31


This past week, Kelly Dwyer of Yahoo! Sports constructed a list of The top 10 individual statistical seasons of the last decade.  Dwyer introduced the list with the following statement: “….let’s go cold. Best statistical season. Pure production. I don’t care if a player’s wife gave birth to triplets in the same season he was able to guest star on stage in a David Mamet production the year his team finally won a title while he averaged a career-high in rebounds per game. Don’t care about the sweet story, only care about the sweet stats.”

In looking at the list it was not quite clear what “production” Dwyer was seeking to measure.  He mentioned points, blocks, steals, rebounds, Player Efficiency Rating (PER), Win Shares, etc….   What he never mentioned is how he decided Kevin Garnett’s performance in 2003-04 (ranked 3rd) was better than what Chris Paul did this past season (ranked 7th).  Numbers can be used to differentiate each performance, but you have to have some sort of methodology that allows such differentiation.  And that’s missing from this ranking.

In the comments on my last post, Simon noted the following about Dwyer’s rankings:

“…there’s nothing surprising about the list. It’s basically the list of players’ with highest PER number, with one exception. Dwight Howard’s 08-09 season was actually behind Iverson’s 05-06(25.9) and Elton Brand’s 05-06 season(26.5) in PER but the Superman curiously got the nod as the sole exception.”

The Wins Produced Rankings

So Simon argues that Dwyer – despite mentioning a number of stats – essentially relied on PERs.  Other people – via comments at the WoW Journal and e-mail – have asked that I re-construct the list via Wins Produced.  And without further comment, here is that list:

Rank

Player

Wins Produced

1

Kevin Garnett (2002-03)

30.7

2

Kevin Garnett (2003-04)

29.6

3

Kevin Garnett (2004-05)

29.4

4

Chris Paul (2008-09)

28.2

5

LeBron James (2008-09)

27.8

6

Shaquille O’Neal (1999-00)

27.1

7

Ben Wallace (2002-03)

27.1

8

Kevin Garnett (2005-06)

26.2

9

Ben Wallace (2001-02)

25.6

10

Tim Duncan (2001-02)

25.0

As one can see, Kevin Garnett dominated the NBA from 2002-03 to 2005-06.  Unfortunately, his teammates were not very productive.  So KG had to wait until 2007-08 to finally play for a team that could dominate the NBA as he did individually.

Standard Deviations Above Average (SDAA)

One issue in looking at this list is that it’s dominated by big men.  Wins Produced compares a player to the average at his position.  As noted in the Wages of Wins, though, there’s a “short supply of big men.”  As a consequence, teams are forced to employ big men who are not very productive.  In contrast, as players get smaller the supply increases.  And that means, there are many more productive little guys. All of this means that it’s easier for the very best big men to distance themselves from the pack at their position.

To adjust for this issue – again, as noted in The Wages of Wins — we can examine how many standard deviations a player is above (or below) the mean at his position. For example, in 2002-03 Kevin Garnett posted a 0.443 WP48. The standard deviation of WP48 for a power forward is 0.110.  Consequently, Garnett was 3.11 standard deviations above average (average WP48 is 0.100). 

To put this in perspective, Michael Jordan – in 1990-91 – posted a 0.437 WP48.  If we compare WP48 numbers, KG’s 2002-03 season looks slightly better.  However, the standard deviation of WP48 for a shooting guard is only 0.092.  So Jordan’s performance that season was 3.65 standard deviations above average.  One should note that this wasn’t even Jordan’s best season. In 1988-89 MJ was 4.18 standard deviations above the average shooting guard.  Since 1977-78, only Magic Johnson in 1982-83 managed to perform four standard deviations above the average at his position (Magic’s mark that year was 4.02).  In fact, if we look at the past thirty years, only Magic and Jordan managed to post marks that were 3.5 standard deviations above average (Magic did this six times, Jordan did this three times).

Well, at least that was true until last year.  As the following table reveals, this past year both Chris Paul and LeBron James surpassed the 3.5 mark.

Rank

Player

Standard Deviations Above Average

1

Chris Paul (2008-09)

3.68

2

LeBron James (2008-09)

3.50

3

Kevin Garnett (2004-05)

3.18

4

Shawn Marion (2000-01)

3.16

5

Jason Kidd (2006-07)

3.12

6

Kevin Garnett (2002-03)

3.11

7

Kevin Garnett (2003-04)

3.08

8

Ben Wallace (2002-03)

3.06

9

Chris Paul (2007-08)

3.02

10

Kevin Garnett (2005-06)

2.94

So if we consider how far a player statistically surpasses the average productivity at his position, both CP3 and King James this past season offered the best performance across the past decade.  In fact, one has to go back to the 1990-91 season to find a player who performed at the level of Paul and James in 2008-09.

Let me close by briefly commenting on Kobe Bryant. Dwyer argues that what Kobe did in 2005-06 ranks in the top 10; and if we focus on scoring (or PERs) that might be true.  Kobe’s WP48 in 2005-06, though, was only 0.203; or only 1.12 standard deviations above average.  And that wasn’t even Kobe’s best season. In 2002-03, Kobe posted a 0.260 WP48, a mark that was only 1.73 standard deviations above average.

Kobe is often compared to Jordan.  However, when we compare Kobe to MJ – via Wins Produced, WP48, or SDAA – it’s clear that Kobe is not like Mike.  Kobe is more like MJ Lite.  Or perhaps, MJ Very Lite.

– DJ

The WoW Journal Comments Policy

Our research on the NBA was summarized HERE.

The Technical Notes at wagesofwins.com provides substantially 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

Wins Produced vs. Win Score

What Wins Produced Says and What It Does Not Say

Introducing PAWSmin — and a Defense of Box Score Statistics

Finally, A Guide to Evaluating Models contains useful hints on how to interpret and evaluate statistical models.