FIP and the Ball-Strike Count II

In my previous post about FIP, I discussed how pitchers give up harder contact in a hitters count resulting in more extra base hits and weaker contact in a pitchers count resulting in more singles. This means that while BABIP doesn’t change much based on count, the effect of the hits based on the count does. I attempted to find metrics that might help predict which players could benefit from this but was unable to do so.

In this post, I decided to look at pitchers that gave up weaker than average contact and see how their ERA and FIPs compare to each other. From 2000-2013, I looked at all pitchers that gave up fewer than average doubles, triples and home runs while giving up a larger than average percentage of singles on balls in play. Presumably, pitchers that fit this profile give up weaker contact than pitchers that don’t and therefore should have a lower ERA than FIP. This test should help determine the impact of giving up weaker contact.

There were 117 pitchers that fit these criteria and many of them were ones that one might expect. For example, star closers such as Mariano Rivera, Craig Kimbrel, Joakim Soria, Ryan Cook, Jim Johnson (not including his terrible 2014), B.J Ryan and David Robertson were all on this list. So were guys like Rick Porcello, Brett Anderson, Brandon Webb, Chien-Ming Wang, Doug Fister and Derek Lowe. These pitchers are well known for giving up a lot of ground balls and allowing only weak contact. A list of the entire 117 pitchers can be found here.

It should come as no surprise that the pitchers on this list record more saves than their average usage would suggest. These pitchers threw only 9% of total innings while recording 20.8% of total saves. It makes sense that pitchers that can avoid giving up hard contact end up being used as closers.
However, there are some surprising results when we compare their ERAs to their FIPs. Only 58 of the 117 pitchers on this list actually have an ERA lower than their FIP. The mean ERA is 3.73 while the mean FIP is 3.78 or about a difference of .05. This is minimal and unimportant. This suggests that even pitchers that end up allowing weaker than average contact still have an ERA that’s similar to their FIP.  It appears that despite the fact that FIP doesn’t differentiate between a single, double or triple the stat still accurately describes performance.

The pitchers that primarily are able to outperform their FIP are those that are able to avoid giving up any type of hits whether they’re singles or for extra bases. There have been 26 pitchers that give up fewer 1B, 2B, 3B and HR than the 35th percentile. They have an average ERA of 3.03 and an average FIP of 3.55. But then again, it’s questionable whether that should be attributed to pitcher skill or to defense.

This doesn’t mean that FIP is necessarily 100% accurate. Some have proposed that Chris Tillman has extra value not measured via traditional statistics because he is able to keep opposing batters close to the bag and therefore prevents steals, prevents runners from advancing multiple bases on a hit and creates more double plays. But it is interesting to see that FIP “works” even in a situation where we’d expect it to fail.

This analysis indicates that pitchers do have some effect on how hard their pitches are hit and therefore whether impact the likelihood of a batter getting an extra base hit. Some pitchers appear to be better than average at preventing hard contact than other pitchers while most pitchers give up weaker contact in favorable pitch counts. I have been unable to determine an impartial way of determining which pitchers should be expected to give up weaker contact than their peers but it appears that someone using the eye test or knowledge about pitchers can predict this with reasonable accuracy. However, FIP appears to be accurate even when looking at pitchers that allow weaker contact than their peers. I am not quite sure how this can be the case but facts are facts. This possibly indicates that FIP is a better metric than its detractors suspect and suggests that people shouldn’t necessarily dismiss it simply because they find things that it doesn’t measure.