Originally Posted by
uva3021
i use a combination of xFIP, FIP, ERA/WHIPx, xERA, and a projected ERA based on running a regression model with the lehman database using FIP and xERA, then weight all the following concomitantly with the pitcher's average line, and come up with a line that way if you just use ERA you'll get some lines taht are like -1500 because there may be a 5 run difference, but by season's end, and running some sort of regression, the era tightens up i've been able to get down to a 3-5% RMSE with a .92-.93 correlation coefficient when compared to Pinny's openers running my model, which filters the plays that have value