Ever since this blog launched I have been using certain numbers to show luck and skill to find the best players in fantasy baseball. We have discussed many of them and you can find my favorites at this post. These are the ones I use first to rate a player.
One of the important ones for pitchers is FIP, but as we have discussed there are some holes in the numbers. FIP does not account for HR/FB% or FB% numbers. I was stubborn to switch to any other analysis since I liked the results and I knew to look for these factors.
I have decided to switch to xFIP, which is an adjustment of the original FIP. FIP was created by Tom Tango who was nice enough to talk with us earlier this year. The creation of xFIP was just an adjustment of the HR in the calculation done by Dave Studemen at THT. Let's look at the calculations:
FIP = (HR*13+(BB+HBP-IBB)*3-K*2)/IP
So this leaves a stat(HR) that is effected by luck and variance.
xFIP = ((FB*.11)*13+(BB+HBP-IBB)*3-K*2)/IP
Looking this over the big change is a removal of HR for fly balls allowed times .11, which was the league average HR/FB. How much of a change did this make in a large amount of data?
Colin Wyers also at THT wrote on that today and found the root mean square error for each projection (including tRA). His findings showed that ERA was the worst predictor of future ERA as we would all assume, but xFIP and tRA are slightly better than FIP. He assumes we should stick to larger projection systems, but that is not in the context of fantasy sports. In fantasy baseball we have a different requirement for statistical analysis and can use these much differently.
So from know on I will be using xFIP in my analysis and any differences between xFIP and ERA are more tightly linked to a factor of luck. We will still have the occasional case with low HR% or a Javier Vazquez case, but the analysis should be greatly improved overall.
