I have completed the 2005 SSPS projections for pitchers. Pitcher projections are based off regression estimates of pitcher performances in 150-inning seasons from 1998-2004. I generated two projections, one for ERA corrected by 3-year Lahman pitcher park factors, the other estimated based on the park/team of each pitcher. I projected 2005 estimates for players who pitched on the same team for an entire year, and pitched more than 30 innings.
Like the estimates for batters, I ran several regressions on historical data until I found the variables that maximized the fit. The estimates based on the park of the pitcher (“Straight-Prediction ERA,” as I call it) were much stronger than the “Park-Corrected” estimates, explaining 48% of the variance of ERA. The variables that were important are 100% in line with DIPS/FIP theory. The projections are based on the previous season’s strikeout-rate, walk-rate, home run-rate, team BABIP, age, handedness, and league. Consistent with DIPS, the previous season’s BABIP did not impact the next season’s ERA or BABIP for pitchers.
The variance of the predicted 2005 ERAs (SD = 0.65) is much smaller than the 2004 variance of ERAs (SD = 1.34) This is not surprising, since much of ERA is determined by BABIP, which is largely random. The fact that such a large determinant of ERA is random makes predicting ERAs for any one pitcher very tricky. I guess that was the whole point of the DIPS revolution, right?
Anyway, here you go. As always, I welcome your thoughts and suggestions.