Most sabermetric comparisons of pitchers will utilize metrics such as Fielding Independent Pitching (FIP), x-FIP, and SIERA. These measures have become the "go to" metrics because of the prevalence of DIPS theory as a framework for sabermetric analysis of pitching performance. And these metrics have been adopted for good reason---FIP, x-FIP, and SIERA are better seasonal measures of a pitcher's true performance than ERA or Runs Allowed (RA). The volatility and randomness of BABIP, which drives ERA and RA, can distort the traditional measures of pitching. Thus, sabermetrics has gravitated toward metrics which avoid the influence of BABIP.
The focus on BABIP volatility, though, assumes that all contact is created equal---which obviously is not true. Because some pitchers appear to sustain over performance of their defense independent pitching metrics, we can hypothesize that these "exceptions" are pitchers who consistently suppress hard contact. Mike Fast, in his pre-Astros life, called attention to this issue and suggested that advanced technologies, such as hit f/x, could help discern pitchers who exhibit a skill for weak contact. We know from Fast's occasional comments at TCB that the Astros utilize those types of technologies to analyze pitchers and hitters. However, those results are proprietary and unavailable to those of us who are not employed by major league teams. So, the question is: Can we utilize any available metrics as a proxy for identifying weak contact skills?
Fangraphs' Tony Blengino, formerly an analyst with the Mariners, considers "contact management skill" in his evaluation of pitchers, as discussed here. You will notice that he uses variations of SLG%-against and batting average on specific types of batted balls to develop measures of contact management. Interestingly, Roger Clemens' stint with the Astros has the best contact management score in the time period since 2000.
Brett Talley at Rotographs followed up Blengino's observations with two articles, here and here, which utilize the batted ball splits for sOPS+ at Baseball-Reference.com as a proxy for weak contact measures. Because the sOPS+ is based on batted balls, it excludes walks and therefore is driven to a large extent by slugging percentage. He identified Garrett Richards, Adam Wainwright, and Johnny Cueto as this season's elite contact management pitchers.
Talley examined whether sOPS+ might explain variations in the difference between pitchers' SIERA and ERA. His test found it to be a poor explanation of the difference. However, he found that sOPS+ exhibited a more robust explanatory power for a sample composed of the highest level performers, as defined by the variance from average sOPS+. He concluded that this supports his contention that a sub-set of elite pitchers exhibit a skill for inducing weak contact.
I suspect that sOPS+ was chosen, in part, for convenience. As a measure of hard contact, sOPS+ may overweight the impact of singles, by giving equal weight to OBP and SLG. Therefore, the focal point of my review will be the SLG component.
Sample size is important for putting SLG-against in perspective. Here is a Baseball Prospectus article which shows the sample sizes necessary for various pitching statistics to stabilize. SLG-against (550 AB) stabilizes much more quickly than BABIP (2000 balls in play). So it is a more reliable contributor to pitching outcomes than BABIP. SLG-against generally stablizes faster than ISO, HR rate, and line drive rate. On the other hand, strike out rate and walk rate, which are the principal components of defense independent pitching metrics, stabilize much more quickly than SLG. My advice---ignore SLG-against in small sample sizes.
Correlation analysis provides us information as to whether the metric is descriptive, predictive or some combination. This is usually a key issue in understanding pitching metrics. A descriptive statistic can explain what happened, but it may not tell us if the performance reflects a repeatable skill. I utilized Steve Staude's pitching correlation tool at fangraphs, set to 95% confidence level, for pitchers with 70 IP/year since 2007. The two relevant statistics are correlation coefficient (r) and R-squared (R^2). R-squared is notable because it indicates the percentage of variation explained by the explanatory variables.
Not surprisingly, SLG is highly explanatory of the pitching outcomes which occurred. SLG has a correlation between 0.6 and 0.85, explaining between 40% and 64% of the variation in same year ERA, FIP, SIERA, and RE24. SLG is moderately correlated with same year BABIP, and weakly correlated with ERA minus FIP. SLG is moderately correlated with contact rate and weakly correlated with fastball velocity.
SLG is not very predictive of the next year pitching performance though. SLG has very little or no correlation with next year ERA, Runs Allowed Per 9, and BABIP. It should be noted, though, that these are volatile pitching statistics on a year to year basis, and are not strongly explained on a forward-going basis by any variable. SLG has some correlation (r between .34 and .37) with next year FIP and SIERA.
SLG-against shows some indication of repeatability, though it isn't strong. SLG has a 0.38 correlation with next year SLG. By comparison, SIERA has a 0.6 correlation with next year SIERA. By contrast, BABIP has virtually no year to year correlation (0.18).
I'll try to wrap it up with a conclusion. With sufficient sample size, SLG may tell you how well the pitcher managed contact during a particular period. But I wouldn't rely on this statistic, alone, to project the pitcher's future contact management. Perhaps examining the SLG-against in combination with other statistics like SIERA and FIP may be useful in assessing the pitcher's skills. It's possible that a sub-set of extreme weak contact pitchers exhibits the ability to sustain the contact management on a year to year basis, as suggested in the rotographs article.
ASTROS PITCHERS' CONTACT MANAGEMENT
In order to evaluate Astros' pitchers' contact management in 2014, I relied upon the baseball reference.com pitching splits for batted balls. I utilized the pitchers' batted balls in fair territory, constructing a sSLG+ and sOPS+ statistic. The sSLG+ stat compares the pitcher's SLG-against on fair balls to the AL average SLG on fair balls (.517). The sOPS+ is a ratio of the pitcher's sOPS+ on fair balls to the AL average sOPS+ (101) on fair balls. In both cases, a value less than 100 in better than the AL average and a value greater than 100 is worse than the AL average.
The pitchers listed below are confined to players currently on the team and whom are identified as "qualified" for the SLG statistic by baseball-reference. The pitchers are listed in descending order of overall SLG-against (which is different than the SLG on batted balls).
|Batted Balls In Fair Territory|
Based on the table, above, Feldman, Keuchel, McHugh, and Sipp have exhibited above average contact management skills this season. The previously referenced B-Pro article indicates that only Keuchel and Feldman have a sufficient sample size for SLG to stabilize. However, both McHugh and Oberholtzer are very close to the required number of at bats and should have a sufficient sample after their next starts. So, it's probably fair to give some credence to the results for McHugh, Oberholtzer, Keuchel and Feldman. I will comment on the results for the other pitchers, but take those players' results with a grain of salt since it may not be meaningful.
McHugh's results are so good that he probably would make the elite top 15 in contact management (as listed in the rotographs article) if he had qualified in innings pitched. Keuchel and Feldman are well above average in sample sizes which are comfortably above the size for SLG to stabilize.
Peacock is well below average in contact management, but keep in mind that the sample size is too small to draw conclusions. Also, Peacock and Oberholtzer do not have the advantage of being groundball pitchers, which makes it more difficult for them to control the SLG percent. A major ingredient in Keuchel's and Feldman's success on contact is their high groundball rate, which tends to minimize extra base hits.
Pitchers who rely on high K rates and flyballs to succeed are at a disadvantage using a contact-based metric. And this is exhibited by both Fields and Peacock. Because the two metrics in the table are based only on batted balls, the pitchers aren't given credit for at bats which end in a strike out. Fields, for instance, is well below average in overall SLG%, but limiting the analysis to at bats which end in contact moves him above average in SLG-allowed. Blengino's article states that good contact management rates provides the pitcher with a greater "margin of error" on the K-BB side. And I would speculate that the pitching style of Fields and Peacock doesn't leave much margin for a decline in K rates. Peacock's recent struggles were accompanied by a mediocre K rate, rather than his normal above average K rate.
All of this makes McHugh's results more impressive. He appears to have elite contact management skill, but has done so with a high K rate and without the advantage of a groundball pitcher profile. If you look at the rotographs' list of current elite contact management pitchers, none of them really fit McHugh's profile. The closest may be Adam Wainwright, who also is known for a plus curveball.
Consider McHugh's profile in light of this quote from Blengino's fangraphs article:
There are two ways to successfully limit damage on fly balls – to manage the vertical angle and the exit velocity off of the bat. Splitting the fly ball category into upper and lower groups equidistant in size from the popup and line drive borders yields starkly different results. The "higher" fly balls yield an .098 AVG-.234 SLG, while the "lower" ones yield a .380 AVG-.990 SLG.
TCB articles have discussed how McHugh's approach has been tweaked in the direction of pitching high in the strike zone in order to take advantage of "effective velocity" with his curveball.
Sipp has too small a sample, but if we overlook that issue, he has achieved excellent contact management results in much the same manner as McHugh---despite a high K rate and a low groundball rate. Sample size is always a problem for evaluating relievers, which is why they are notoriously volatile from year to year. All we can say is that Sipp's contact results have been intriguing, and he could be quite a find---if he can keep it up over the long haul.
As noted previously, the Astros utilize hit f/x type technologies to evaluate pitchers. We have reads hints that evaluation of contact management was a factor in the signing of Feldman. McHugh and Sipp were available practically for free, and one wonders whether the Astros knew something about their contact management skills.