Last week I wrote about two stats, Value Hit Percent (VH%) and Poorly Hit Percent (PH%), which both serve to measure the quality of contact on the high and low ends of the spectrum, respectively. Value Hits represent the highest quality of contact, making up a huge number of doubles, triples, and home runs, while Poorly Hit balls represent balls in play that are almost automatic outs. Poorly hit balls register in an earned base, whether by a hit or through an error, about 2.5% of the time. This is a success rate roughly on par with infield fly balls, which results in a base about 2.2% of the time, according to the MLB classification of pop ups over the same two year time span. Granted, differing methods for defining pop ups can decrease the success rates dramatically, down to 1% or lower. No matter how you cut it, though, we’re talking about near automatic outs with both PH and IFFB.
Many have combined IFFB rates with strike out rates when calculating FIP, and the calculations work because both the strike out and the infield fly have roughly equivalent run values during a game, sitting between -.25 and -.28 depending on the season. Considering PH have very similar success rates to IFFB, it seems safe to assume their run value is very similar as well, and PH could be added to the strike out totals in the FIP formula. In order to account for this slight change, we just need to go through and calculate a new constant to go along with the inclusion of PH%, and for the 2016 season it turns out to be 4.817.
I’ll call this stat PHFIP, and I am comparing it to standard FIP. The equations are as follows:
FIP = ( 13 x HR + 3 x BB – 2 x K) / IP + C
PHFIP = ( 13 x HR + 3 x BB – 2 x (K + PH) ) / IP + C
The following chart shows the ten pitchers who benefit the most from using PHFIP.
Eight of these ten pitchers outperformed their FIP, and for each of these eight the PHFIP is closer to the ERA than their FIP. These pitchers combined to out perform their FIP by 25 points, and while their PHFIP is a tad lower than their ERA, it is much closer.
I feel a bit dirty writing these numbers, since David Cameron found very similar results when he compared FIP to IFFIP back in 2013. He too lifted the ten pitchers who gained the most from the IFFIP over FIP, and he also found that eight of those ten had an ERA lower than their. All eight of these pitchers had an IFFIP closer to their ERA than their FIP. These ten pitchers cumulatively outperformed their FIP by 22 points, although his IFFIP happened to be a tad closer to the actual ERA than mine is here.
It is obviously a coincidence that our numbers match up so well, but both methods are offering a small correction to the overall FIP numbers. A very small correction, as it turns out, but a correction nonetheless. Cameron found the correlation between ERA and FIP to be .77 and .78 for IFFIP for the 2012 season. I found the correlation between ERA and FIP to be .71, and .73 for PHFIP this year. By using PH%, I am placing the vast majority of near automatic outs into the equation. Some may slip through the cracks here and there, the system is by no means perfect, but the vast majority are accounted for. Incorporating this much additional weak contact only to see a very small change surprised me a little bit. Infield fly balls alone aren’t especially common, but these PH hits represent about a fifth of the total plate appearances in the game. The PH explain a slice of the variance, although perhaps focusing on the weakly hit balls alone isn’t quite enough to fill in the gap completely.
Moving on from here, it seems necessary to pay closer attention to the medium and hard hit balls. To this point, I have largely ignored those middle range hits, which represent about 45-50% of all contact, in favor of looking at the extreme high and low end contact. The high end has proven to be a good indicator of overall offensive performance, correlating with wOBA (.46). Meanwhile, the extremely weak contact negatively correlates with strike out rate (-.44) and BABIP (-.41). None of these correlations are especially strong, but they are okay by the standards of many other stats, such as year to year BABIP (.20) and HR (.42), whereas year to year PH% is .34.
Ultimately, these statistics are not necessarily predictive, but rather indicate what results are reasonable to expect given a certain set of input. Hopefully, with refinement, the result will narrow down on the most likely set of results given the input, that’s the goal. This isn’t necessarily predictive, though, it is still prescriptive in nature, with currently unknown predictive value. It may come out that this data does bring predictive value, perhaps in the form of more accurate, and potentially more stable, measures of ‘skill’, which could then be input for other forms of analysis.
Getting back to the PHFIP for a moment, adding PH to the calculation seems to be offer a small correction to the overall stat, bridging the gap between FIP and ERA for many pitchers. Like IFFIP, it fails to account for enough of the variance between ERA and FIP to make the stat strong enough to truly differentiate from standard FIP. Below I’ve included a workbook of all the PHFIP and FIP data for ‘qualified pitchers’ which I’ve defined as 126 IP.