What Statcast Reveals About Contact Management as a Pitcher Skill by Craig Edwards August 17, 2017 While there are certain events (like strikeouts, walks, and home runs) over which a pitcher exerts more or less direct control, it seems pretty clear at this point that there are some pitchers who are better at managing contact than others. It’s also also seems clear that, if a pitcher can’t manage contact at all, he’s unlikely to reach or stay in the big leagues for any length of time. Consider: since the conclusion of World War II, about 750 pitchers have recorded at least 1,000 innings; of those 750 or so, all but nine of them have conceded a batting average on balls in play (BABIP) of .310 or less. Even that group of nine is pretty concentrated, the middle two-thirds separated by .029 BABIP. The difference between the guy ranked 125 out of 751 and the guy ranked 625 out of 751 is just three hits out of 100 balls in play. Those three hits can add up over a long period of time, of course, but it still represents a rather small difference even between players with lengthy careers. For that reason, attempting to discern batted-ball skills among pitchers with just a few seasons of data is difficult. Thanks to the emergence of Statcast, however, we have some better tools than just plain BABIP to evaluate a pitcher’s ability to manage contact. Let’s take a look at what the more granular batted-ball data reveals. Statcast has provided some new information that allows us to compare a few seasons’ worth of contact quality against pitchers. For the purposes of looking at weak contact, let’s focus on two pursuits. We’ll begin with xWOBA, a metric recently added to Baseball Savant that includes strikeouts and walks to determine a wOBA-like stat based on launch angle and exit velocity. We can use their search to drill down to just those balls which were put in play. Here are the leaders and laggards by xwOBA on contact so far this season for all pitchers with at least 1250 pitches: Pitcher xwOBA Leaderboard on Contact Rank Pitcher xwoba ERA FIP 1 Brandon McCarthy .308 3.84 3.33 2 James Paxton .312 2.78 2.49 3 Dallas Keuchel .315 2.77 3.71 4 Alex Wood .316 2.30 2.70 5 Brad Peacock .316 3.30 2.75 6 Kyle Freeland .317 3.74 4.75 7 Joe Biagini .319 5.11 3.88 8 Aaron Nola .321 3.02 3.14 9 Andrew Cashner .321 3.32 4.41 10 Chase Anderson .321 2.89 3.43 … 127 Matt Moore .407 5.71 4.67 128 Vince Velasquez .407 5.13 5.50 129 Derek Holland .409 5.68 6.22 130 Jesse Chavez .410 5.29 5.38 131 Josh Tomlin .410 5.38 4.25 132 Danny Salazar .411 3.92 3.44 133 Kevin Gausman .411 5.08 4.56 134 Johnny Cueto .419 4.59 4.66 135 Chris Tillman .420 7.94 6.22 136 Ricky Nolasco .427 5.24 5.36 SOURCE: Statcast So what does that tell us? The first part is kind of obvious: pitchers who get hit hard tend to give up runs; those who don’t get hard, prevent them. Only Danny Salazar, who strikes out one-third of the batters he faces, gives up a bunch of hard contact and has managed a decent season. This general observation would tell you that xwOBA is doing something right. We know that strikeouts and walks play a significant role in a pitcher’s efficacy, which is why r-squared for xwOBA on contact with ERA (.32) and FIP (.23) show some relationship, but not one that is incredibly strong. When you include strikeouts and walks into xWOBA, r-squared is much stronger for both ERA (.58) and FIP (.69). That might put us closer to an ERA estimator or predictor, but that doesn’t really get us to the contact question. So let’s look at xWOBA on contact and compare that to wOBA on contact. This graph shows the 137 players this season with at least 1250 pitches. So again, it looks like xwOBA is doing something right, as there’s a relationship between xwOBA on contact and wOBA on contact. I ran these same figures for pitchers in 2016 with at least 2,000 pitches and got pretty similar results, although the relationship was stronger in 2017 than in 2016, perhaps a result of more accurate Statcast data this season. When I looked at xwOBA for hitters earlier in the season, I found a pretty good relationship between xwOBA and itself in the future, finding it a more accurate predictor than past wOBA. We can do something similar for pitchers. While looking to see if xwOBA against for pitchers might do a better job at estimating and predicting ERA than FIP might be a worthwhile study, we already know FIP does a pretty good job. Trying to see if xwOBA can tell us anything about contact skill might be a bit more interesting and potentially more worthwhile. There are 81 pitchers who recorded at least 2,000 pitches in 2016 and also 1,250 pitches so far this season. For a comparison, here’s a graph showing wOBA on contact against in 2016 and wOBA on contact against in 2017. So we don’t see a great relationship from year to year, and keep in mind these numbers include home runs. There really isn’t much in there to indicate we can see a skill in the numbers year over year. Perhaps xwOBA can do a little better, like it did on the hitter side. The graph below shows the xwOBA on contact for the same sample of pitchers over the last two seasons. That’s not really great. Pitchers differed quite a bit from year to year. Some guys were terrible in 2016 and then great in 2017 (Andrew Cashner and Chase Anderson) while other guys went the wrong way (Johnny Cueto, Kyle Hendricks, and Masahiro Tanaka). All of which is to say, it’s hard to look at how a pitcher did against contact in one year and project for the next. We could be dealing with a change in talent level and we could be dealing with some luck or the individual sample sizes just might not be big enough. As part of my research, I also compared xwOBA on contact in 2016 to wOBA on contact in 2017 and got a similar result (r=.18) to wOBA on contact between the two years. These figures are similar to BABIP between the two years as well (r=.20). Before calling it a day, I used one more Statcast tool: the type of batted ball. Over at Baseball Savant, they separate the quality of contact into six categories: Barrel, Solid Contact, Flare/Burner, Poorly/Under, Poorly/Topped, and Poorly/Weak. For the same groups of pitchers in 2016 and 2017, I looked at the percentage of batted balls that were classified as one of the last three categories. My thought: perhaps there’s something that a look at the entire data set misses, that inducing weak contact itself needs to be separated from the rest of the batted balls. It didn’t work. There wasn’t a strong relationship here (r=.23), but keep in mind almost all of the pitchers are grouped together between 58% and 68%, so it might be difficult to find a relationship when the distribution is so narrow. I’m not ready to give up trying to solve pitcher contact, and there’s probably a lot more that could be done even with just the data above, but the data probably does help show something we’ve known for quite a while: the hitter has a lot more control of what happens to the ball once it makes contact.