Are Groundball Pitchers Overrated? by Matthew Murphy February 20, 2015 Justin Masterson is one of the game’s best groundball pitchers. (via Erik Drost) Getting a ground ball is a good outcome for a pitcher. While I don’t think that anyone reading this article would dispute this fact, here’s a table from Dan Farnsworth’s 2014 THT piece on batted ball types to illustrate the point: Batted Balls Outcomes for 2013 MLB regular season Batted Ball Type OBP SLG OPS ISO wOBA Grounders .232 .250 .483 .018 .213 Liners .685 .883 1.568 .193 .681 Flies .213 .621 .834 .403 .346 Ground balls turn into hits only about 23 percent of the time, and almost never go for extra bases, for an average wOBA of .213. To put that number into perspective, the lowest wOBA by a qualified position player in the past decade is .247 (by Alcides Escobar in 2013, if you must know). So, if ground balls are good, then groundball pitchers must be good, right? This is one assumption that makes ERA estimators such as xFIP and SIERA so appealing — they place a value on the pitching skill of inducing ground balls. Also appealing is that inducing ground balls is a repeatable skill with a very high year-to-year correlation (R = 0.85). This stability, along with the absence of home runs (which are weighed very heavily by FIP and fluctuate more drastically) makes these estimators more stable. In fact, Bill Petti found that SIERA and xFIP- were the most stable of the commonly used ERA predictors, and that they were significantly better at predicting future ERA than ERA itself. However, a closer look at these correlations reveals a surprising piece of information. While xFIP is much more stable year-to-year than FIP, FIP is actually slightly better at predicting future ERA. Year One to Two Correlation Comparisons Year Two Y1 GB% Y1 ERA- Y1 FIP- Y1 xFIP- Y2 ERA- -0.05 0.36 0.45 0.42 Y2 FIP- -0.14 0.41 0.58 0.58 Y2 xFIP- -0.25 0.42 0.60 0.70 Why might this be the case? Once again, let’s look at the correlations. While ground-ball percentage (GB%) has an inverse correlation to future xFIP (R = -0.25), there is practically no correlation to future ERA (R = -0.05). These trends hold true even when looking at the same year. Among qualified starters from 2010-2014 (443 player-seasons), there was no correlation between GB% and ERA- (R = -0.09). So, if ground balls are good for pitchers, and inducing them is a repeatable skill, then why aren’t pitchers who induce ground balls getting better results? Fly Ball Contact Management The key here lies in two other pitching outcomes that appear to be intertwined with ground balls: pop-ups and home runs. Pop-ups are nearly as good as strikeouts, in that they are almost always an automatic out. While inducing pop-ups appears to be a repeatable skill (IFFB% has a year-to-year correlation of 0.37), it is not nearly as stable as strikeout or walk rates. (Note: IFFB% measures the percentage of fly balls that are infield flies, not total infield fly percentage.) As it turns out, groundball pitchers not only have lower total pop-up rates because of fewer fly balls, but groundball rate actually has a strong inverse correlation to IFFB% (R = -0.45), meaning that fly balls hit against groundball pitchers are less likely to be pop-ups than if they came against an average or flyball pitcher. Meanwhile, while groundball pitchers give up fewer total home runs, they give up more home runs than expected given their low flyball rate. That is, there is a slight positive correlation between groundball percentage and home runs per fly ball (R = 0.13 for the same year, R = 0.17 for the following year). This might not be surprising to those of you who are intimately familiar with the factors involved in calculating SIERA. Matt Swartz noticed this relationship and included it in the formula, stating that “pitchers who have higher flyball rates allow fewer home runs per fly ball” and mentioning Matt Cain (circa 2011 – not the new version) as a relevant example. In addition to these correlations, ground balls also have a very slight negative correlation with strikeouts and a slight positive correlation with walks. This makes some sense, since pitches up in the zone are both more likely to be fly balls and more likely to be whiffs. So what if groundball rate doesn’t correlate with run prevention because of this association? What if ground balls are good, but the pitchers who induce them are simply worse at striking out batters and avoiding walks? To see if this is the case, we can look at how ground balls correlate to a pitcher’s ability (or lack thereof) to pitch up to (or outperform) their ERA estimators. If ground balls are good, than the pitchers who inducing them should prevent runs as well or better than their FIP and xFIP would predict. Groundball and Infield Fly Ball Percentage Correlations Correlation to: FIP- minus ERA- xFIP- minus ERA- xFIP- minus FIP- GB% -0.07 -0.14 -0.13 IFFB% 0.17 0.25 0.18 Because of the nature of adjusted ERA and ERA estimators, a positive number means that the pitcher “outperformed” his peripherals. In terms of this graph, it means those pitchers’ ERAs were lower than their FIP or xFIP, or that their FIP was lower than their xFIP. With groundball rate, the correlation is very small, but negative across the board. At the very least, groundball pitchers do not outperform their peripherals (which might be expected, given that the ERA estimators do not take into account the fact that ground balls are less likely to turn into doubles or triples). Meanwhile, pitchers with higher IFFB% – which, as I mentioned earlier, has a significant inverse correlation to groundball percentage – did outperform their peripherals, particularly xFIP. Comparing High- and Low-Groundball Pitchers However, these correlations are fairly small. What happens if we look at groundball pitchers? To do this, I looked at the 80 pitchers who threw at least 600 innings between 2010 and 2014. These pitchers had an average groundball rate of 45.3 percent, with a standard deviation of 5.4 percent. So, I looked at pitchers who were at least one standard deviation above or below the mean groundball rate, and defined them as “High GB%” and “Low GB%” pitchers. The High GB% group includes 10 pitchers, including notable worm-burners such as Rick Porcello, Felix Hernandez, Justin Masterson and Tim Hudson. Meanwhile, the Low GB% group (13 pitchers) includes the previously mentioned Cain, along with flyball machines Jered Weaver and Max Scherzer. Let’s look at the average statistics from these two groups: Statistical Comparison, High GB% vs. Low GB% Pitchers Group n GB% FB% HR/FB IFFB% BABIP K% BB% ERA- FIP- xFIP- High GB% 10 54.4% 26.8% 11.0% 7.6% 0.293 17.2% 7.9% 102.2 100.5 96.6 Low GB% 13 36.8% 42.7% 9.4% 11.2% 0.285 19.9% 7.3% 99.7 100.6 103.0 Average 80 45.3% 34.7% 10.2% 9.4% 0.291 19.5% 7.1% 96.6 96.8 96.6 Starting near the left, we see that the High GB% pitchers have above-average HR/FB rates and below average IFFB% (i.e. more home runs and fewer pop-ups per fly ball). Meanwhile the fly balls generated by the Low GB% group are more likely to turn into pop-ups and less likely to leave the park. While these differences may not seem huge, they are both statistically significant (P < .001). Low GB% pitchers also have a slightly below-average BABIP due to the increase in pop-ups (not significant), and have a slightly higher strikeout rate than their groundball brethren (P < .05). When we combine all of these factors, we see that the High GB% pitchers allow more damage on their fly balls, leading them to underperform both their xFIP and FIP by a rather large margin. Meanwhile, the ability of the Low GB% group to manage flyball contact allowed them to beat their xFIP, while their FIP (which accounted for the reduced HR/FB%) estimated their ERA fairly accurately. And ultimately, while the High GB% pitchers have a strong edge in xFIP, it is in fact the Low GB% group that posted the better ERA. One item of note is that despite similar K% and BB% rates, the Low GB% group was actually inferior to the average pool. This could mean that the ideal pitcher doesn’t rely too heavily on inducing ground balls or weak flyball contact, but has an average balance of the two and can, perhaps, use either strategy depending on the situation. In this sample, a dozen pitchers were within one percentage point of the average GB% and FB%. This group includes studs such as Clayton Kershaw, Madison Bumgarner, David Price and Cole Hamels, and has an average ERA- of 90.5. Employing a similar strategy as outlined above, we can see that pitchers who significantly under-perform their peripherals (ERA- at least nine percent worse than their xFIP-) tend to induce more ground balls (48.5 percent) than their over-performing counterparts (43.1 percent groundball rate). Also, high-IFFB% pitchers tend to out-perform their FIP and xFIP, while those with low IFFB% rates tend to underperform. (Notably, only three high-IFFB% pitchers managed to maintain an average or better GB% and HR/FB% – Clayton Kershaw, Roy Halladay and Jon Lester.) It is clear that High GB% and Low GB% pitchers achieve significantly different batted ball outcomes, but we haven’t yet looked at how they might be doing this. Fastball Comparison, High GB% vs. Low GB% Pitchers (PITCHf/x stats) Group FA% vFA FA-Z High GB% 21.9% 91.1 6.2 Low GB% 43.4% 90.3 9.4 Average 32.8% 91.2 8.3 One of the keys appears to be in fastball usage. Groundball pitchers are less likely to rely on their four-seam fastball, often mixing in a heavier dose of sinkers/two-seam fastballs. Interestingly, while the Low GB% pitchers throw more four-seamers, it’s not because they have better velocity. In fact, this group had below-average velocity. One thing they did have, however, was extra rise on their fastball (P < .001). The far right column indicates the average vertical movement on these pitches. The higher the number, the less a fastball drops on its way to the plate. All but one of the 13 Low GB% pitchers had above-average vertical movement on their fastballs, with an average of 9.4 inches. While the 1.1 inch extra rise relative to the average population might not seem like a big deal, keep in mind that the maximum allowed bat radius in the majors is 1.3 inches. One inch is the difference between putting the barrel on the ball and a can of corn. Four-seam movement also has a strong correlation to batted ball outcomes in the total pitcher pool, such as GB% (R = -0.63), IFFB% (R = 0.37) and HR/FB% (R = 0.29). Pitchers with more rise on their fastball throw it more often, and get more weak fly ball contact at the expense of grounders. While this analysis was all performed with pitchers who tossed at least 600 innings, reducing the requirement to 400 innings (and expanding the player pool) yields similar results. Ground ball pitchers still have less rise on their fastball, and get fewer pop-ups and more home runs on the fly balls they do allow, all with a high degree of statistical significance. Closing Thoughts So, are ground balls good? Of course they are! However, this fact has led to a strong bias toward groundball pitchers (and against their flyball counterparts) in the baseball analytics community. Just because a ground ball in itself is a good outcome for a pitcher, it doesn’t necessarily mean that we should place a significant value on pitchers who are skilled at inducing grounders. In fact, pitchers who get a lot of ground balls often struggle in other facets of their game, particularly in managing fly ball contact. And while there are groundball pitchers who can avoid the long ball and maintain a solid pop-up rate (I’m looking at you, Felix), these pitchers are the exception rather than the rule. As the strike zone has expanded downwards, pitching down in the zone has become an increasingly popular strategy. While this has allowed some pitchers to rack up grounders without sacrificing strikeouts, this analysis has shown that there can be a tradeoff associated with chasing ground balls. Many of today’s successful groundballers, such as Sonny Gray, Alex Cobb and Tyson Ross (all of whom ranked in the top-5 in GB% among qualified pitchers in 2014) actually have above-average “rise” on their four-seam fastball, and throw it a decent amount of the time. This has allowed them to maintain a high strikeout and groundball rates, and their contact management skills aren’t quite as poor as many other High-GB% pitchers. In addition, pitch location certainly matters. A fastball low in the zone can still turn into a pop-up if the batter thinks that it is going to drop more than it actually does. Future research must be done to clarify this relationship, in addition to looking at how these trends might play out with specific pitch types. We use ERA estimators like xFIP because they are straightforward. xFIP relies on only three factors, all of which the pitcher has a great deal of control over and are very stable from year to year. We like it because it removes the sometimes huge fluctuations seen in home run rates that can wreak havoc on a pitcher’s ERA and FIP. And as a tool for doing that, it is excellent. However, while xFIP is a useful tool, when we rely on it too much we fall into a trap: the trap of forgetting that not all fly balls are created equal, and that while neither inducing pop-ups nor avoiding home runs is not as repeatable as inducing ground balls, they nonetheless remain valuable skills. So next time you’re getting ready to write off a young pitcher who’s succeeding despite giving up “too many” fly balls, just remember: Ground balls may be good, but groundball pitching is also overrated.