Library Update: xFIP
Last week, we updated our Library entry for Fielding Independent Pitching (FIP). This week, FIP’s sister-stat Expected Fielding Independent Pitching (xFIP) got the same treatment. If you’re looking for a defense independent pitching statistic that also tries to strip out some of the inherent fluctuation in home run rate, xFIP is for you. Head over to the full entry to learn more about it.
If you missed it last week, there’s also a blog post in the Library that breaks down the basics of why using metrics other than ERA will help you develop a better understanding of how well a pitcher is performing.
As always, please feel free to pose questions or comments below or find me on Twitter @NeilWeinberg44 if that’s more convenient. Also remember to stop by our weekly FanGraphs Q&A chat (Wednesdays at 3pm eastern) to learn more about our stats and our site.
Neil Weinberg is the Site Educator at FanGraphs and can be found writing enthusiastically about the Detroit Tigers at New English D. Follow and interact with him on Twitter @NeilWeinberg44.
In mention of statistics that are meant primarily for their predictive qualities, such as xFIP and SIERA, I feel the idea of sample size needs to be emphasized. If the primary goal of a predictive stat is to strip out random variation, then the other primary way to strip out random variation should be acknowledged, and that is simply to get a lot of data.
Yes, single-season xFIP is a better predictor on next season’s ERA than single-season ERA. The main reason this is true is that even over a full season, a pitcher doesn’t give up that many home runs, so HR/FB is still working on a relatively small sample, and is thus subject to a lot of random variation. However, we do know that pitchers exert some control on their HR/FB rate, and this effect can be observed with some confidence once we have a lot of data on the pitcher.
For example, if a pitcher has a HR/FB rate of 7% over 5 full seasons, and you want to predict his next season ERA, you’re not just going to assume he’ll have a league-average HR/FB rate next year, which xFIP does. 5 full seasons is enough that we should give him some credit, and project a next-season HR/FB rate that is better than average.
On the other hand, if we have a pitcher who used to be mediocre, and now he’s 3 months into an excellent season, then his xFIP over those three months will be far more instructive than his ERA over those months.
I suppose this may be more of a general statistical literacy issue which you address elsewhere. If you do not, either a page explaining some of this stuff (like small sample size, random variation, etc.) might be helpful to some, or just a link to another resource.
We have a sample size page, which could use some updating. But also, I have a lot of this info in the xFIP entry. I talk about sample size and that some pitcher can beat lgHR/FB% and if they demonstrate that xFIP underrates them, etc.