Why Bill James May Have Missed Early Success

Addendum: Further investigation has revealed that data from the strike season included in the following study is responsible for the results. When corrected, batting average is no longer a better predictor of offense than RC. Thanks to Phil Birnbaum.

Economists are baffled at Billy Beane’s discovery of Bill James. James publishes his first book in 1977 but few in MLB take him seriously until the mid-to-late 1990s in Oakland. As I discuss in a previous post, maybe James’s ideas got pushed aside due to some bad luck. Mainly, maybe some of his good ideas looked bad due to anomalies in baseball that occurred at the time he came on the scene. So I decided to look at one of James’s first discoveries, Runs Created (RC), which was published in 1979.

RC = [(H + BB)*TB]/(AB + BB)

In the 3 seasons that followed RC’s introduction (1980-1982) the R-squared of the linear regression estimate of RC on Runs per game (RG) was .44. For the effect of BA on RG the R-squared is .64. If I’m a GM, who do I listen to? The guy from the pork and beans factory or my baseball scouts? In hindsight, it is easy to see what a truly great man James is. But such an observation may not have been so easy 20 years ago.

P.S: James tested the RC model on the 1975 season (according to Moneyball), for which the model predicts with an R-squared of .85. Today RC has been modified to where it is an excellent predictor of offense.

Comments are closed