free NBA basketball picks
Home | NBA PLAYOFFS | Regular Season | Commentary | Contact Us

Which position matters the most?

At 82games we track a number of stats on a "production by position" basis, both for teams (see Pacers positional numbers for example), and for individual players (see Tim Duncan's stats by playing position).

There can be complexities in evaluating which players on court in a game are playing which position at that moment, but by scaling the roster of a team from “most point guard?to “most center?you can organize a group into logical positional placement. For instance when the Kings have Bibby-Christie-Peeler-Stojakovic-Webber on the floor, Webber is acting as the center, Stojakovic is the power forward, whereas in a lineup of Bibby-Christie-Stojakovic-Webber-Miller, Webber is now at the four spot, and Stojakovic is playing the three.

Moreover, we track the stats of the opposing counterpart position, which enables us to build out all kinds of tables with information comparing the net production for each team and player.

One of the many research projects underway and continuing through the off-season will be to assess the correlation between various statistics and the wins accrued for the team in actual reality. While there are many stats we track for position production purposes, by turning to the ratings based on the excellent PER formulas developed by John Hollinger (and as detailed in his Pro Basketball Prospectus books), we get to a summary rating that tackles the heart of the matter.

1) Correlation of Team Positional Production and Wins
We'll first focus on just the team's actual player PER rating (ignoring the counterpart of the opposition), and have three views based on the 2002-03 season, the 2003-04 season, and the two seasons combined.

Position
02-03
Correlation
03-04
Correlation
2Yr
Correlation
95% CI
PG
.53
.42
.49
.27 to .66
SG
.21
-.17
.07
-.19 to .32
SF
.19
.27
.23
-.03 to .46
PF
.42
.38
.40
.16 to .60
C
.28
.40
.33
.08 to .54

First off we are dealing with only 29 observations per year, so not a giant sample to work with. Not surprisingly then, the standard errors are such that using a 95% confidence interval, we cannot say with total conviction that Point Guard production has been more valuable than Shooting Guard play, since the "best case" SG is 0.32 correlation, while PG is 0.27 in the "worst case."

Still the correlation values from year to year are pretty consistent for PG and PF, and overall it comes in as PG-PF-C-SF-SG in order of importance. Now, before all the GM's rush out to trade their high priced Shooting Guards, perhaps what we are seeing is merely a reflection of the NBA currently and not a long term indication of true worth of positions. For instance, when Jordan was in his prime and ostensibly a shooting guard, it seems likely that the SG production would correlate somewhat more substantially with Wins! These days it is the Power Forwards and Point Guards getting the most notice -- Duncan, Garnett, Kidd and co are some of the stars of the immediate pro basketball era. However, if another crop of Jordans comes along who's to say that the SG's won't become the dominant "correlators."

2) Correlation of Opponent Positional Production and Wins
This time we'll look at the production allowed to opponents by position:

Position
02-03
Correlation
03-04
Correlation
2Yr
Correlation
95% CI
PG
-.42
-.48
-.44
-.63 to -.21
SG
-.54
-.65
-.58
-.73 to -.38
SF
-.57
-.59
-.56
-.72 to -.36
PF
-.40
-.38
-.39
-.59 to -.14
C
-.37
-.52
-.44
-.63 to -.21

For the opposing production by position, the order of importance according to the correlation comes in as SG-SF-PG-C-PF, which is interesting to say the least. Whereas having a great shooting guard is downplayed, being able to defend the opposing two guard is significant!

(And no, this study was not funded by the agents of Bruce Bowen, Trent Hassell, et al.)

It's tight though among the positions for this look however, suggesting the ordering could easily change with a different sample of years.

3) Correlation of Net Positional Production and Wins
Finally, we'll run the numbers for the net production, or "PER Difference" to use the term being popularized in certain quarters.

Position
02-03
Correlation
03-04
Correlation
2Yr
Correlation
95% CI
PG
.61
.60
.60
.41 to .74
SG
.39
.36
.38
.13 to .58
SF
.43
.62
.53
.31 to .69
PF
.58
.48
.53
.32 to .70
C
.38
.52
.45
.21 to .63

For the net production by position, it's PG-PF-SF-C-SG, but again with the wide standard deviations, the 95% CI does not allow for any conclusive stance on one position over another.

At the same time, we can infer from the data that for the past two years, the teams with high producing point guards and forwards have been faring better than those with juggernaut two guards or centers.

Perhaps the most relevant mention of this could be seen in the current Spurs-Lakers playoff series (with San Antonio up 2-0 at this time), where the Spurs have been fabulous in the PG battle (Parker running wild against Payton) and the PF spot (Duncan over Malone), while the Lakers haven't extracted wins from their edge at SG and C.

Anyway, the above tables have served as the warmup...we'll be back next week with the real test when we run regressions on the PER team stats for all positions at once!


Rate this Feature
Poor   Fair   Good   Excellent

Enter your comments in the box

Email (optional)

We want your feedback! Tell us your thoughts

Copyright ?2003 by www.nbatop.com, All Rights Reserved