Age and FIP-ERA differential

On Wednesday's episode of Effectively Wild, Sam Miller's Play Index segment was about whether older pitcher out-perform their FIP more often than younger pitchers, due to increasing "craftiness", etc., as they age. The Play Index can divide players into four age bins: 25 & under, 26-30, 31-35, and 36+. Sam looked at the last five years and found that the youngest two groups tend to very slightly under-perform their FIP, with an ERA 0.02 runs higher than FIP for each, while the older groups tend to out-perform their FIP, by 0.02 runs for pitchers 36+ and by 0.09 runs, by far the largest margin, for the 31-35 group. I took a further look at the data to try to tease out any additional trends, and in doing so found a possible indicator of whether a pitcher will pitch into their 30s.

First, I attempted to re-create Sam's results. I found that I couldn't get literally identical results, actually, though what I got was close enough to call them the same. I went further, too, going as far back as 2000, and found FIP-ERA on a yearly basis. I then took the average over the whole sample weighted by IP for each data point -- this ended up being about 700,000 total innings pitched, split roughly 25:45:20:10 among the age bins. Over the entire period, I found that the two young groups under-performed their FIP by 0.02 runs (same as in Sam's results), while the two older groups each over-performed their FIP by 0.05 runs. When I narrowed my data set to what Sam used, I found an over-performance of 0.08 runs for the 31-35 group instead of 0.09, but otherwise got the same results.

What I was most curious about relating to this data set was whether this was an older-pitcher phenomenon, as speculated by Sam and Ben, or whether there was a different factor at play. Specifically, I wondered whether these FIP over-performers developed the skill as they aged or if they had it all along.

To look at this, I first identified all the pitchers in the age 31-35 cohort between 2011-2015. I then used Fangraphs to determine their IP, FIP, and ERA at every age available, and binned that data using the same age ranges as the Play Index. Sam hypothesized that the reason for the FIP-ERA split was an increase in craftiness and/or wisdom and/or some other experiential factor; for the sake of having a competing idea, I guessed that FIP over-performace was always a characteristic of these pitchers, and said characteristic was selected for over the course of their careers (hence them still being around at ages 31-35 while the under-performers no longer are). If Sam's right, I should see an age pattern that mimics that of the entire sample; if I'm right, I'll see a consistent FIP-ERA across all the age bins. The answer is...

... somewhere in the middle. At each age range this cohort of pitchers out-performs their FIP; however, in no other age range do they do so as dramatically as they do during their age 31-35 seasons. I also don't trust the number for the 36+ group, which is 0.004 runs better than FIP, due to an extremely small sample size.

These results may indicate that a preference exists in MLB, whether conscious or subconscious, for pitchers who have a better ERA than FIP. I don't find this to be particularly surprising -- to me, it essentially says that ERA is still trusted above FIP when it comes to personnel decisions. These findings also suggest a potential indicator for whether a young pitcher will last in the majors into his 30s; however, I haven't done the work to see whether there's any predictive power here whatsoever. I'd also love to see this repeated, either by me or by someone else, using the same methodology but on a different cohort in order to see whether the effect can be observed across time or is unique to the most recent five seasons.

. . .

All data courtesy of FanGraphs.

John Choiniere is a researcher and occasional contributor at Beyond the Box Score. You can follow him on Twitter at @johnchoiniere.