NBA True Winning Percentage

[mickeyg13]

I know it's been studied in baseball that looking at run differential can be a more accurate predictor of future winning percentage than even current winning percentage (some form of Pythagorean wins). Has anyone studied anything similar in the NBA? This might have an obvious answer, but I'm not as much of an NBA fan as I am MLB, but the NBA seems to be the only other major sport where sample size issues wouldn't trivialize findings.

[Dudd]

[ QUOTE ]
Pythagorean winning percentage - A Bill James invention in baseball, Pythagorean records are based on the knowledge that team winning percentages are generally closely related to points scored and points allowed (and, in cases where they differ, the reason is usually temporary luck). This relationship can be approximated by PF^x/((PF^x) + (PA^x), where x depends on the total points scored. In baseball, over the course of a season, x is close enough to be approximated by two. In basketball, x is a little more difficult to calculate. When Oliver did it a decade and a half ago, it was about 16.5. Now, because point totals are lower, the exponent is believed to be closer to 13 or 14. It is also possible to calculate Pythagorean percentages in different ways, like a multiplier (2.7) by the team's point differential (for an 82-game season). This is slightly less accurate, but much easier. A third, more complicated method is employed by Oliver, which takes into account the variability in a team's points scored or points allowed and is thus more accurate.
Statistically: PF^13.5/((PF^13.5)+(PA^13.5))
Expected wins = 2.7*(PF/G-PA/G) + 41

[/ QUOTE ]

From here

[mogwai316]

You can see the updated values of this stat every day in the expected W-L and WP columns here:
http://sports.espn.go.com/nba/stats/rpi

It does use the 16.5 exponent value. You could easily compute the numbers using a different exponent if you want, though.