## Does IFFB% Correlate with HR/FB Rates?

Over the last week or so, various reputable baseball analysis sites have been digging into the relationship between infield fly ball rates (IFFB%) and home run per fly ball rates (HR/FB). The discussion was prompted by a blog post by Rory Paap at Paapfly.com called “Matt Cain ignores xFIP, again and again,” which generated a response from Dave Cameron here at Fangraphs.

Paap suggested FIP and xFIP do Cain a disservice because they don’t give him his due credit for possessing the “unique skill” of inducing harmless fly ball contact, a theory that David Pinto at Baseball Musings attempted to quantify last October. Cameron’s response included some interesting analysis that looked at the best pitchers from 2002-2007 in terms of HR/FB rate and compared their IFFB% over that span to what they posted the next three seasons. His conclusion?

Is there some skill to allowing long fly outs? Maybe. But if you can identify which pitchers are likely to keep their home run rates low while giving up a lot of fly balls before they actually do it, then you could make a lot of money in player forecasting.

Simply out of curiosity, I decided to throw my hat into the ring and see if I could find a trend between IFFB% and HR/FB rate. My theory was that if IFFB% and HR/FB rate showed some sort of correlation, then plotting HR/FB rate as a function of IFFB% would show a clear inverse trend (meaning that a higher IFFB% would more likely generate a lower HR/FB rate, and vice versa).

To do this, I looked at all pitchers from 2008 to 2010 who threw at least 162 innings and plotted their IFFB% and HR/FB rate as described above. This three-year range generated 257 such data points, and these were the results:

*Note: IFFB% is on the x-axis and HR/FB rate is on the y-axis.*

Just by looking at the chart, it’s tough to visually decipher any sort of trend. If there actually is an inverse relationship between IFFB% and HR/FB rates, we would expect the data points to slope from the top-left (low IFFB%, high HR/FB rate) to the bottom-right (high IFFB%, low HR/FB rate).

By adding a best-fit trend line to the data set, we see that there is a very slight slope in the direction we anticipated, but to say it shows any sort of useful relationship is a stretch. The data has an R-Squared value of just 0.0126, which tells us there was very little correlation between IFFB% and HR/FB rate. If you don’t know what R-squared is, it’s simply a representation of one variable’s ability to forecast another. R-Squared values range from 0 to 1, and the closer they are to 1 the more of a correlation there is between the two sets of data. An R-Squared value of 0.0126 between IFFB% and HR/FB rate shows very little correlation.

What conclusions can we draw from this? Perhaps it is possible to tell if a pitcher like Cain is more prone to lower HR/FB rates by virtue of his ability to induce weaker contact, but IFFB% alone is not enough to draw any conclusions. More sophisticated analysis, like that provided by Pinto’s article at Baseball Musings, might unveil some usable relationships, but we cannot simply look at Clayton Kershaw’s 4.1 percent HR/FB rate in 2009 and say his 13.5 percent IFFB% explains it. For now, I’m still skeptical about pitchers like Cain, but there’s no doubt his performances these last few seasons have given us plenty to think about.

Print This Post

Baseball Professor is one of the most personable fantasy baseball site's on the web! Visit us at www.baseballprof.com and check out our daily articles, podcasts, and fantasy rankings.

I don’t understand why everyone has problems with the analysis; because there is always problems with an analysis of this magnitude. We’re talking about human individuals who act not according to some mathematical formula, but perhaps merely based on what they had to eat before a game or what their personal life was like at such time. As much as statistics are helping to base our expectations of player performance, let us not forget that these people are persons like us; I feel that that flies out the window all too often on a site like this when everything must fall into some quantified field with variables including only a team or the environment and not necessarily the psyche or determination of a particular individual player.

I don’t think anyone will argue with what you just said; grit and determination are major motivating factors for competitive individuals and cannot be wholly quantified as predictive measures of success. I do disagree, however, that when analyzing past performance it is impossible or pointless to conduct research such as this.

Maybe Cain is an extremely competitive individual who really digs in when he’s on the mound. Maybe that explains why he has become such a successful pitcher. However, that is completely separate from this here. Looking at HR/FB rates and IFFB%s to try and find a correlation speaks only to the relatedness of the two stats to see if one can be used as a predictor of the other. Somewhere down the road we’ll probably invent some new stat or find a new way of measuring Cain’s performance (average trajectory of fly balls in play? that sounds interesting) that may allow us to see why or how pitchers like Cain continue to spit in the face of our modern day understanding of baseball statistics.

Just because we are yet to find an explanation doesn’t mean one does not exist.

If you expand your model to include other controls, and estimate it for 2008-2010, you find that IFFB is a good predictor of HR/9:

Dependent variable: HR/9

iffb -0.0298***

(0.00735)

k9 -0.0536***

(0.0164)

fb 0.0266***

(0.00353)

y2009 -0.0505

(0.0806)

y2010 -0.145*

(0.0772)

Constant 0.741***

(0.164)

Observations 480

R-squared 0.140

Can you clear something up for me when looking at the stats glossary it says GB% and FB% are those events ‘in play’ but I don’t think this is the same definition of ‘in play’ that is used in BABIP correct? It seems that here’in play’ means not fould, whereas BABIP, balls in play mean they are in the field of play. I ask this because it seems we are using the same FB% here in the two stats, which I believe that since IFFB% is a portion of FB%, and HR/FB is also using the same denominator, so by default you are going to find a positive correlation between HR/FB and IFFB%.

I think what you are really looking for is if a pitcher who induces IFFB is also inducing other ‘less harmful’ FB. so it seems you should take out the IFFB from the rest of the FB and look at HR/(FB-IFFB), and lthen look at correlations – or compare to other pitchers.

Maybe I don’t understand the stats being used, if not, please let me know..

Seems to me you’d want to look at the correlation for the subset of pitchers with high IFFP%. You kind of already know that xFIP generally “works”- i.e. flyball pitchers as a whole don’t particularly out-perform their xFIP. The question you’re really trying to answer is “do some guys have a knack for inducing weak flyballs”. You’d get at this by looking at the guys with the highest IFFB% and seeing if they also had low HR/FB%. i.e. do extreme flyball pitchers outperform their xFIP? To me, it makes perfect sense that they might. And that extreme groundball pitchers might outperform their FIP and xFIP.

Another way to get at this would be to strip out everybody’s defensive support and park effects and see if the FIP and xFIP v ERA outliers seem to have some common characteristics. I’m not really hooked into the SABR community; for all I know Tom Tango looked at this 10 years ago. But it makes perfect sense to me that extreme GB and FB might beat their FIPs and xFIPs. In fact, I think I’d expect this.