Shifting responses to the shift

Shifting is, increasingly, not a novelty in the baseball world. You can tell, because TV broadcasts talk about it constantly, and even they don’t seem to think it’s particularly new. Your casual-fan acquaintance or neighbor probably knows shifting is A Thing now. It’s simply a part of modern baseball.

Figuring out what it means for modern baseball, however, has proven more difficult, at least for those of us working with public data. The closest we can get to peeling it apart is using data provided by FanGraphs and Baseball Info Solutions, where each non-home run ball in play is flagged as having been hit into no shift, a traditional (“Ted Williams”) shift, or some other kind of non-traditional shift. Last year, our own Matt Jackson used that data to examine BABIP with and without the shift, and Joe Vasile did a similar thing to look at wBABIP (i.e., BABIP, but with doubles and triples weighted as they are in wOBA).

But analyses based on the BIS/FanGraphs data are necessarily incomplete, as people are quick to point out. The impact of the shift is not limited to the balls hit into it. Batters and pitchers presumably approach plate appearances differently when the shift is on, and those different approaches will lead to changes in the numbers of home runs, strikeouts, and walks, which aren’t reflected in the data we have.

Unfortunately, this is the only shift data that’s available to the public, so we are forced to make do. Nonetheless, I wanted to get at the ways in which batters and pitchers have changed their approach to pitching in front of the shift, and while all the direct ways of doing so are beyond the scope of this data, there are some indirect methods we can use. Namely, I wanted to compare the MLB-wide batted ball tendencies in the face of shifts and not, and see how they’ve changed over time. It won’t tell us whether batters are striking out more often thanks to the shift, but altogether it might tell us something about how they’re attacking these plate appearances.

First, as a precursor, here’s the rate of balls in play that have occurred while the defense was in some form of shift (traditional or not).

Again, it’s important to be clear about what this does and doesn’t communicate. It doesn’t show exactly how often teams shift, since it doesn’t pick up any shifts that happen in plate appearances that end in a walk, strikeout, or home run. But it does show that shifts have gotten much more popular over the last five seasons, and that the trend has shown no signs of slowing down.

So the shift has exploded in popularity. What’s been the response? Over the same time period, here’s the ratio between ground ball rate without the shift, and ground ball rate with the shift. (A figure over one indicates more ground balls against the shift; less than one, more ground balls against no shift.) For this and the subsequent comparisons, I limited the analysis to traditional shifts, since without knowing exactly what a nontraditional shift is, it’s hard to predict how it might impact batter/pitcher behavior.

In 2010 and 2011, among balls put into play against the shift, there was a higher rate of ground balls than among balls put into play against a traditional infield. This makes some sense, but perhaps not for the first reason you’d think. While batters aren’t selected at random for the shift, the most shifted players aren’t picked because they hit a lot of ground balls; indeed, of the ten most shifted batters of 2016, only three had above-average ground ball rates, and even then only slightly. They’re selected because they pull an abnormally large rate of their ground balls, and that often correlates with high fly ball rates, meaning we’d expect this ratio to be below 1.0 if neither the batter nor pitcher were approaching shifted PAs any differently than usual.

As a result, the most plausible explanation for the above chart is that pitchers were changing their approach in some way that made ground balls more likely. Without better data, we unfortunately can’t figure out how they changed it, or even confirm that they changed at all. But it stands to reason that pitchers might be inclined to throw a few more sinkers, or pitch to contact a bit more readily, when they know the infield is shifting behind them. We also can’t say whether it was a good idea — maybe more grounders are a good thing when the shift is on, but maybe these added grounders were hit the other way, or maybe there was an accompanying decrease in strikeouts that made the whole endeavor counterproductive.

However, starting in 2012 and continuing through the present, the pattern has reversed, with the ratio falling to and holding steady at about .95. I see two possible causes for this. On the one hand, this could be batters trying to counteract any pitcher adjustments and hit fewer grounders. We know players are aware of the shift, and we also know that uppercut swings are increasingly in vogue.

But I’m skeptical. I suspect it’s a lot harder for batters to change their swing path than it is for pitchers to change their pitch selection or location, and I suspect that making that change of change from one plate appearance to the next would do more harm than good. Instead, I think the above trend just reflects a reversion back to normal, where pitchers pitch no differently with the shift on than they do when it’s off.

Like I said above, we can’t know for sure whether intentionally trying to get grounders helps or hurts a pitcher — I know I keep talking about the limitations of the data, but they come into play a lot when you’re trying to do this kind of thing — but the decision of whether to shift a batter is made based on their tendencies when faced with normal pitching, and trying to “shift pitch” might throw all kinds of things off. It wouldn’t surprise me at all to learn that, once shifts became more common, and once pitchers and coaching staffs and analytics departments got a little more on the same page, pitchers stopped trying to induce grounders, and just pitched normally.

I will spare you the chart of fly ball rate, since it is (unsurprisingly) the same basic thing in reverse. Instead, here’s the ratio of the rate at which batted balls were pulled against a shift and not against the shift.

Pulling the ball is precisely what makes a batter a candidate for the shift, so it’s no surprise to see that the ratio has always been above 1.0. There’s been a definite downward turn in the ratio, however, falling all the way from 1.19 in 2010 to 1.02 in 2016.

I don’t think it’s a change in approach that’s responsible for this one, though again, all this explaining I’m doing is pure speculation. Instead, what I think is behind this is simple diminishing returns. You saw the first table, where the number of balls hit into the shift skyrocketed; it’s not just the extreme pull hitters — the David Ortizes and Adrian Gonzalezes of the the world — who get shifted anymore. As players with moderate pull tendencies face the shift increasingly often, the pool of shifted players starts to look less and less distinct, and more and more like MLB as a whole.

Jeff Sullivan recently noted that the major-league BABIP has changed very little over the last decade, but it shouldn’t be a surprise that MLB is reaching a sort of equilibrium in relation to the shift. Batters and pitchers face new strategies all the time, and they’re constantly adjusting. The youngest players today came up through the minor leagues while shifts were an established part of the game.

Shifts, like every other strategy — high fastballs, stolen bases, bunting for a hit — are most effective when they’re unanticipated and novel. We’re past the point where the strategy is new, and are firmly into its middle age. There have been adjustments, and counter-adjustments, and counters to the counters, and the result is a baseball that doesn’t look much different than it ever has.