Basketball is in the midst of a statistical revolution, as has been said ad nauseum by many people who don’t seem to understand what they’re talking about.
Like the “Moneyball” inspired SABERMETRIC mode of thought that has become prime in baseball over the past decade (don’t make outs), basketball’s revolution is built on a simple, almost laughably obvious, idea. Good players do things that increase the likelihood of their team scoring and avoid doing things that decrease the likelihood of their team scoring.
What though, are those things?
Using regression analysis* and lots of caffeine, Dave Berri and a team of sports economists have ferreted out what, exactly, causes a team to succeed. Below is a simplified version of their formula, which they call win score.
*Regression analysis is the method used in economics to separate the impact of one variable on an outcome from that of other variables. For example, in basketball, there is a negative correlation between offensive rebounds and winning. Teams that get more offensive rebounds win less. Regression analysis though shows that this is true only because teams that are getting a lot of offensive rebounds are missing a lot of shots, and missed shots obviously lead to losses. A regression analysis untangles the impact of offensive rebounds from missed shots and shows that offensive rebounds, other things equal, are good news for your team.
Win Score = PTS + REB + STL + ½*BLK + ½*AST
– FGA – ½*FTA – TO – ½*PF
The average win score for a center is 10.8, 10.3 for a power forward, 7.3 for a small forward, 6.1 for a shooting guard, and 6.3 for a point guard.
Win score is, again, a simplified version of a stat called Wins Produced, or its derivative WP48 (wins produced per 48 minutes). Because a team wins on average 0.5 games per 48 minutes, and has 5 players on the court at a time, an average player has a WP48 of 0.1. Any player with a Wp48 above 0.1 is above average, any player below 0.1 is below average. A WP48 of 0.2 is excellent and a wp48 of 0.3 or above denotes unofficial superstar status.
How accurate is this method? It explains 95% of team wins. So, when this model attempts to predict team wins for a given season, based only on that team’s player’s box score statistics for that given season, it does so with a margin of error of only about 1.5 wins. That’s pretty accurate.
For a detailed explanation of wins produced go here. http://www.wagesofwins.com/CalculatingWinsProduced.html
Or just visit Mr. Berri’s website and tool around. www.wagesofwins.com
— Tom Sunnergren